Tsymbal, Zutic - SpintronicsHandbook (2019)

Spintronics Handbook:
Spin Transport and

Magnetism,
Second Edition
Semiconductor Spintronics—Volume Two
Spintronics Handbook:
Spin Transport and
Magnetism,
Second Edition
Semiconductor Spintronics—Volume Two
Edited by
Evgeny Y. Tsymbal and Igor Ž utić
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2019 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Printed on acid-free paper
International Standard Book Number-13: 978-1-4987-6960-0 (Hardback)
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors
and publishers have attempted to trace the copyright holders of all material reproduced in this
publication and apologize to copyright holders if permission to publish in this form has not been
obtained. If any copyright material has not been acknowledged please write and let us know so we
may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known
or hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC),
222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that
provides licenses and registration for a variety of users. For organizations that have been granted a
photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and
are used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis web site at

http://www.taylorandfrancis.com
and the CRC Press web site at

http://www.crcpress.com
Cover illustration courtesy of Markus Lindemann, Nils C. Gerhardt, and Carsten Brenner.
The e-book of this title contains full colour figures and can be purchased here: http://www.crcpress.
com/9780429434235. The figures can also be found under the ‘Additional Resources’ tab.
Contents
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
About the Editors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
List of Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Section IV—Spin Transport and

Dynamics in Semiconductors
Chapter 1. Spin Relaxation and Spin Dynamics in Semiconductors and

Graphene. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Jaroslav Fabian and M. W. Wu
Chapter 2. Electrical Spin Injection and Transport in Semiconductors . . . . . . . . . 59
Berend T. Jonker
Chapter 3. Spin Transport in Si and Ge: Hot Electron Injection and
Detection Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Ian Appelbaum
Chapter 4. Tunneling Magnetoresistance, Spin-Transfer and
Spinorbitronics with (Ga,Mn)As. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
J.-M. George, D. Quang To, T. Huong Dang, E. Erina,
T. L. Hoai Nguyen, H.-J. Drouhin, and H. Jaffrès
Chapter 5. Spin Transport in Organic Semiconductors . . . . . . . . . . . . . . . . . . . . . . 247
Valentin Dediu, Luis E. Hueso, and Ilaria Bergenti
v
vi Contents
Chapter 6. Spin Transport in Ferromagnet/III–V Semiconductor

Heterostructures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Paul A. Crowell and Scott A. Crooker
Chapter 7. Spin Polarization by Current. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
Sergey D. Ganichev, Maxim Trushin, and John Schliemann
Chapter 8. Anomalous and Spin-Injection Hall Effects . . . . . . . . . . . . . . . . . . . . . . 339
Jairo Sinova, Jörg Wunderlich, and Tomás Jungwirth
Section V—Magnetic Semiconductors,

Oxides and Topological Insulators
Chapter 9. Magnetic Semiconductors: III–V Semiconductors. . . . . . . . . . . . . . . . 371

Carsten Timm
Chapter 10. Magnetism of Dilute Oxides. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
J.M.D. Coey
Chapter 11. Magnetism of Complex Oxide Interfaces . . . . . . . . . . . . . . . . . . . . . . . . 457
Satoshi Okamoto, Shuai Dong, and Elbio Dagotto
Chapter 12. LaAlO3/SrTiO3: A Tale of Two Magnetisms. . . . . . . . . . . . . . . . . . . . . . . . 489
Yun-Yi Pai, Anthony Tylan-Tyler, Patrick Irvin, and Jeremy Levy
Chapter 13. Electric-Field Controlled Magnetism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519
Fumihiro Matsukura and Hideo Ohno
Chapter 14. Topological Insulators: From Fundamentals to Applications. . . . . . 543
Matthew J. Gilbert and Ewelina M. Hankiewicz
Chapter 15. Quantum Anomalous Hall Effect in Topological Insulators . . . . . . . 573
Abhinav Kandala, Anthony Richardella, and Nitin Samarth
Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
Foreword
S
pintronics is a field of research in which novel properties of materials,
especially atomically engineered magnetic multilayers, are the result of
the manipulation of currents of spin-polarized electrons. Spintronics, in
its most recent incarnation, is a field of research that is almost 30 years old.
To date, its most significant technological impact has been in the develop-
ment of a new generation of ultra-sensitive magnetic recording read heads
that have powered magnetic disk drives since late 1997. These magnetoresis-
tive read heads, which use spin-valves based on spin-dependent scattering at
magnetic/non-magnetic interfaces and, since 2007, magnetic tunnel junctions
(MTJs) based on spin-dependent tunneling across ultra-thin insulating layers,
have a common thin film structure. These structures involve “spin engineer-
ing” to eliminate the influence of long-range magneto-dipole fields via the use
of synthetic or artificial antiferromagnets, which are formed from thin mag-
netic layers coupled antiferromagnetically via the use of atomically thin layers
of ruthenium. These structures involve the discoveries of spin-dependent tun-
neling in 1975, giant magnetoresistance at low temperatures in Fe/Cr in 1988,
oscillatory interlayer coupling in 1989, the synthetic antiferromagnet in 1990,
giant magnetoresistance at room temperature in Co/Cu and related multilayers
in 1991, and the origin of giant magnetoresistance as being a result of predomi-
nant interface scattering in 1991–1993. Together, these discoveries led to the
spin-valve recording read head that was introduced by IBM in 1997 and led,
within a few years, to a 1,000-fold increase in the storage capacity of magnetic
disk drives. This rapid pace of improvement has stalled over the past years
as the difficulty of stabilizing tiny magnetic bits against thermal fluctuations
whilst at the same time being able to generate large enough magnetic fields to
write them, has proved intractable. The possibility of creating novel spintronic
magnetic memory-storage devices to rival magnetic disk drives in capacity and
to vastly exceed them in performance has emerged in the form of Racetrack
Memory. This concept and the physics underlying it are discussed in this book,
vii
viii Foreword
together with a more conventional spintronic memory, magnetic random

access memory (MRAM). MRAM is based on MTJ magnetic memory bits,
each one accessed in a two-dimensional cross point array via a transistor.
The fundamental concept of MRAM was proposed in 1995 using local fields
to write the MTJ elements. This basic concept was proven in 1999 with
the subsequent demonstration of large-scale, fully integrated 64 Mbit mem-
ory chips in the following decade. Writing these same elements using spin
angular momentum from sufficiently large spin-polarized currents passed
through the tunnel junction elements emerged in the 1990s and is now key
to the development of massive-scale MRAM chips. The second edition of
this book discusses these emerging spintronic technologies as well as other
breakthroughs and key advances, both fundamental and applied, in the field
of spintronics.
Beyond MRAM and Racetrack Memory, this book elucidates other
nascent opportunities in spintronics that do not rely directly on magneto-
resistive effects, such as fault-tolerant quantum computing, non-Boolean
spin-wave logic, and lasers that are enhanced by spin-polarized carriers. It
is interesting that spintronics is a field of research that continues to surprise
even though the fundamental property of spin was realized nearly a century
ago, and the basic concept of spin-dependent scattering in magnetic materi-
als was introduced by Neville Mott just shortly after the notion of “spin” was
conceived.
Since the first edition of this book, spintronics has so much evolved
that a new name of “spin-orbitronics” has been coined to describe these new
discoveries and developments. In the first edition of this book, spin-orbit
coupling was regarded rather negatively as a property that leads to mix-
ing between spin-channels and the loss of spin angular momentum from
spin currents to the lattice, thereby limiting the persistence of these same
spin currents, both temporally and spatially. In this edition, several physi-
cal phenomena derived from spin-orbit coupling are shown to be key to the
development of several new technologies, such as, in particular, the current-
induced motion of a series of magnetic domain walls that underlies Racetrack
Memory. This relies especially on the generation of pure spin currents via
the spin Hall effect (SHE). The magnitude of the SHE was thought for some
time to be very small in conventional metals, but over the past few years,
this has rather been shown to be incorrect. Significant and useful SHEs have
been discovered in a number of heavy materials where spin-orbit coupling is
large. These spin currents can be used to help move domain walls or to help
switch the magnetization direction of nanoscale magnets. Whether they
can be usefully used for MRAM, however, is still a matter of debate.
Another very interesting development since the first edition of this book
is the explosive increase in our understanding and knowledge of topological
insulators and their cousins including, most recently, Weyl semi-metals. The
number of such materials has increased astronomically and, indeed, it is now
understood that a significant fraction of all extant materials are “topologi-
cal”. What this means, in some cases, is that the spin of the carriers is locked
to their momentum leading, for example, to the Quantum Spin Hall Effect.
Foreword ix
The very concept of these materials is derived from band inversion, which is
often due to strong spin-orbit coupling. From a spintronics perspective, the
novel properties of these materials can lead to intrinsic spin currents and
spin accumulations that are topologically “protected” to a greater or lesser
degree. The concept of topological protection is itself evolving.
Distinct from electronic topological effects are topological spin textures
such as skyrmions and anti-skyrmions. The latter were only experimentally
found 2 years ago. These spin textures are nano-sized magnetic objects that
are related to magnetic bubbles, which are also found in magnetic materi-
als with perpendicular magnetic anisotropy but which have boundaries or
walls that are innately chiral. The chirality is determined by a vector mag-
netic exchange – a Dzyaloshinskii–Moriya interaction (DMI) – that is often
derived from spin-orbit coupling. The DMI favors orthogonal alignment of
neighboring magnetic moments in contrast to conventional ferromagnetic
or antiferromagnetic exchange interactions that favor collinear magnetic
arrangements. Skyrmion and anti-skyrmion spin textures have very interest-
ing properties that could also be useful for Racetrack Memories. Typically,
skyrmions and anti-skyrmions evolve from helical or conical spin textures.
The magnetic phase of such systems can have complex dependences on tem-
perature, magnetic field, and strain. Some chiral antiferromagnetic spin tex-
tures have interesting properties such as an anomalous Hall effect (AHE),
which is derived from their topological chiral spin texture in the absence of
any net magnetization. In practice, however, a small unbalanced moment is
needed to set the material in a magnetic state with domains of the same chi-
rality in order to evidence the AHE. On the other hand, these same chiral
textures can display an intrinsic spin Hall effect whose sign is independent
of the chirality of the spin texture.
The DMI interaction can also result from interfaces particularly
between heavy metals and magnetic layers. Such interfacial DMIs can give
rise to chiral domain walls as well as magnetic bubbles with chiral domain
walls – somewhat akin to skyrmions. The tunability of the interfacial DMI
via materials engineering makes it of special interest.
Thus, since the first edition of this book, chiral spin phenomena, namely
chiral spin textures and domain walls, and the spin Hall effect itself, which is
innately chiral, have emerged as some of the most interesting developments
in spintronics. The impact of these effects was largely unanticipated. It is
not too strong to say that we are now in the age of “chiraltronics”!
Another topic that has considerably advanced since the first edition of
this book is the field of what is often now termed spin caloritronics, namely
the use of temperature gradients to create spin currents and the use of ther-
mal excitations of magnetic systems, i.e. magnons, for magnonic devices.
Indeed, magnons carry spin angular momentum and can propagate over
long distances. Perhaps here it is worth mentioning the extraordinarily long
propagation distances of spin currents via magnons in antiferromagnetic
systems that have recently been realized.
Recently discovered atomically thin ferromagnets reveal how the pres-
ence of spin-orbit coupling overcomes the exclusion of two-dimensional
x Foreword
ferromagnetism expected from the Mermin-Wagner theorem. These two-

dimensional materials, which are similar to graphene in that they can readily
be exfoliated from bulk samples, provide a rich platform to study magnetic
proximity effects and transform a rapidly growing class of van der Waals
materials. Through studies of magnetic materials, it is possible to reveal
their peculiar quantum manifestations. Topological insulators can become
magnetic by doping with 3d transition metals; the quantum anomalous Hall
effect has been discovered in such materials, and heterostructures that con-
sist of magnetic and non-magnetic topological insulators have been used to
demonstrate current-induced control of magnetism.
Spintronics remains a vibrant research field that spans many disciplines
ranging from materials science and chemistry to physics and engineering.
Based on the rich developments and discoveries over the past thirty years,
one can anticipate a bountiful future.
Stuart Parkin
Director at the Max Planck Institute of Microstructure Physics
Halle (Saale), Germany
and
Alexander von Humboldt Professor, Martin-Luther-Universität
Halle-Wittenberg, Germany
Max Planck Institute of Microstructure Physics
Halle (Saale), Germany
Preface
T
he second edition of this book continues the path from the foundations
of spin transport and magnetism to their potential device applications,
usually referred to as spintronics. Spintronics has already left its mark
on several emerging technologies, e.g., in magnetic random access memories
(MRAMs), where the fundamental properties of magnetic tunnel junctions are
key for device performance. Further, many intricate fundamental phenomena
featured in the first edition have since evolved from an academic curiosity into
the potential basis for future spintronic devices. Often, as in the case of spin
Hall effects, spin-orbit torques, and electrically-controlled magnetism, the
research has migrated from the initial low-temperature discovery in semicon-
ductors to technologically more suitable room temperature manifestations in
metallic systems. This path from exotic behavior to possible application contin-
ues to the present day and is reflected in the modified title of the book, which
now explicitly highlights “spintronics,” as its overarching scope. Exotic topics
of today, for example, pertaining to topological properties, such as skyrmions,
topological insulators, or even elusive Majorana fermions, may become suitable
platforms for the spintronics of tomorrow. Impressive progress has been seen in
the last decade in the field of spin caloritronics, which has evolved from a curi-
ous prediction 30 years ago to a vibrant field of research.
Since the first edition, there has been a significant evolution in material sys-
tems displaying spin-dependent phenomena, making it difficult to cover even
the key developments in a single volume. The initially featured chapter on gra-
phene spintronics is now complemented by a chapter on the spin-dependent
properties of a broad range of two-dimensional materials that can form a myriad
of heterostructures coupled by weak van der Waals forces and support super-
conductivity or ferromagnetism even in a single atomic layer. Exciting develop-
ments have also been seen in the field of complex oxide heterostuctures, where
the non-trivial properties are driven by the interplay between the electronic,
spin, and structural degrees of freedom. A particular example is the magnetism
xi
xii Preface
emerging in two-dimensional electron gases at oxide interfaces composed of

otherwise nonmagnetic constituents. The updated structure of a significantly
expanded book reflects various materials developments and it is now the-
matically divided into three volumes, each based on broadly defined metallic
and semiconductor systems or their nanoscale and applied aspects.
Spintronics becomes more and more attractive as a viable platform for
propelling semiconducting technology beyond its current limits. Various
schemes have been proposed to enhance the functionalities of the exist-
ing technologies based on the spin degree of freedom. Among them is the
voltage control of magnetism, exploiting the nonvolatile performance of
ferromagnet-based devices in conjunction with their low-power operation.
Another approach is utilizing spin currents carried by magnons to trans-
port and process information. Magnon spintronics involves interesting fun-
damental physics and offers novel spin wave-based computing technologies
and logic circuits. Optical control of magnetism is another approach, which
has attracted a lot of attention due to the recent discovery of the all-optical
switching of magnetization and its realization at the nanoscale. Chapters on
these subjects are included in the new edition of the book.
Nearly nine decades after the discovery of superconducting proximity
effects by Ragnar Holm and Walther Meissner, several new chapters now
explore how a given material can be transformed through proximity effects
whereby it acquires the properties of its neighbors, for example, becoming
superconducting, magnetic, topologically non-trivial, or with an enhanced
spin-orbit coupling. Such proximity effects not only complement the con-
ventional methods of designing materials by doping or functionalization but
can also overcome their various limitations and enable yet more unexplored
spintronic applications.
We are grateful both to the authors who set aside their many priorities
and contributed new chapters, which have significantly expanded the scope
of this book, as well as to those who patiently provided valuable updates to
their original chapters and kept this edition even more timely. The comple-
tion of the second edition was again greatly facilitated by Verona Skomski,
who tirelessly collected authors’ contributions and assisted their prepara-
tion for the submission to the publisher. We acknowledge the support of
NSF-DMR, NSF-MRSEC, NSF-ECCS, SRC, DOE-BES, US ONR which,
through the support of our research and involvement in spintronics, has also
enabled our editorial work. We are thankful to our families for their sup-
port, patience, and understanding during extended periods of time when we
remained focused on the completion of this edition.
Evgeny Y. Tsymbal
Department of Physics and Astronomy,
Nebraska Center for Materials and Nanoscience, University of Nebraska,
Lincoln, Nebraska 68588, USA
Igor Ž utić
Department of Physics, University at Buffalo,
State University of New York, Buffalo, New York 14260, USA
About the Editors
Evgeny Y. Tsymbal is a George Holmes University Distinguished Professor

at the Department of Physics and Astronomy of the University of Nebraska-
Lincoln (UNL), and Director of the UNL’s Materials Research Science and
Engineering Center (MRSEC). He joined UNL in 2002 as an Associate
Professor, was promoted to a Full Professor with Tenure in 2005 and named
a Charles Bessey Professor of Physics in 2009 and George Holmes University
Distinguished Professor in 2013. Prior to his appointment at UNL, he was a
research scientist at University of Oxford, United Kingdom, a research fellow
of the Alexander von Humboldt Foundation at the Research Center-Jülich,
Germany, and a research scientist at the Russian Research Center “Kurchatov
Institute,” Moscow. Evgeny Y. Tsymbal’s research is focused on computational
materials science aiming at the understanding of fundamental properties of
advanced ferromagnetic and ferroelectric nanostructures and materials relevant
to nanoelectronics and spintronics. He has published over 230 papers, review
articles, and book chapters and presented over 180 invited presentations in the
areas of spin transport, magnetoresistive phenomena, nanoscale magnetism,
complex oxide heterostructures, interface magnetoelectric phenomena, and
ferroelectric tunnel junctions. Evgeny Y. Tsymbal is a fellow of the American
Physical Society, a fellow of the Institute of Physics, UK, and a recipient of the
UNL’s College of Arts & Sciences Outstanding Research and Creativity Award
(ORCA). His research has been supported by the National Science Foundation,
Semiconductor Research Corporation, Office of Naval Research, Department of
Energy, Seagate Technology, and the W. M. Keck Foundation.
Igor Ž utić received his Ph.D. in theoretical physics at the University of

Minnesota, after undergraduate studies at the University of Zagreb, Croatia.
He was a postdoc at the University of Maryland and the Naval Research Lab.
In 2005 he joined the State University of New York at Buffalo as an Assistant
xiii
xiv About the Editors
Professor of Physics and got promoted to an Associate Professor in 2009

and to a Full Professor in 2013. He proposed and chaired Spintronics 2001:
International Conference on Novel Aspects of Spin-Polarized Transport and
Spin Dynamics, at Washington DC. Work with his collaborators spans a
range of topics from high-temperature superconductors, Majorana fermions,
proximity effects, van der Waals materials, and unconventional magnetism,
to the prediction and experimental realization of spin-based devices that are
not limited to magnetoresistance. He has published over 100 refereed articles
and given over 150 invited presentations on spin transport, magnetism, spin-
tronics, and superconductivity. Igor Žutić is a recipient of the 2006 National
Science Foundation CAREER Award, the 2019 State University of New York
Chancellor’s Award for Excellence in Scholarship and Creative Activities,
the 2005 National Research Council/American Society for Engineering
Education Postdoctoral Research Award, and the National Research Council
Fellowship (2003–2005). His research is supported by the National Science
Foundation, the Office of Naval Research, the Department of Energy, Office
of Basic Energy Sciences, the Defense Advanced Research Project Agency,
and the Airforce Office of Scientific Research. He is a fellow of the American
Physical Society.
Contributors
Ian Appelbaum Elbio Dagotto

Department of Physics Department of Physics and
University of Maryland Astronomy
College Park, Maryland The University of Tennessee
Knoxville, Tennessee
and
Ilaria Bergenti
Materials Science and Technology
Institute for Nanostructured
Division
Materials Studies
Oak Ridge National Laboratory
Bologna, Italy
Oak Ridge, Tennessee
J. M. D. Coey
T. Huong Dang
Centre for Research on Adaptive
Unité Mixte de Physique
Nanostructures and Nanodevices
Centre National de la Recherche
Trinity College
Scientifique-Thales
Dublin, Ireland
Université Paris-Saclay
and
Scott A. Crooker Laboratoire des Solides Irradiés,
National High Magnetic Field Ecole Polytechnique, CNRS and
Laboratory CEA/DRF/IRAMIS
Los Alamos National Laboratory Institut Polytechnique de Paris
Los Alamos, New Mexico Palaiseau, France
Paul A. Crowell Valentin Dediu

School of Physics and Astronomy Institute for Nanostructured
University of Minnesota Materials Studies
Minneapolis, Minnesota Bologna, Italy
xv
xvi Contributors
Shuai Dong Ewelina M. Hankiewicz

Department of Physics Institute for Theoretical Physics
Southeast University and Astrophysics
Nanjing, China Würzburg University
Würzburg, Germany
H.-J. Drouhin
Laboratoire des Solides Irradiés, Luis E. Hueso
Ecole Polytechnique CIC nanoGUNE Consolider
CNRS and CEA-DRF-IRAMIS Basque Foundation for Science
Institut Polytechnique de Paris Bilbao, Spain
Palaiseau, France
Patrick Irvin
E. Erina Department of Physics and
Laboratoire des Solides Irradiés, Astronomy
Ecole Polytechnique University of Pittsburgh
CNRS and CEA-DRF-IRAMIS and
Institut Polytechnique de Paris Pittsburgh Quantum Institute
Palaiseau, France Pittsburgh, Pennsylvania
Jaroslav Fabian Henri Jaffrès

Department of Physics Unité Mixte de Physique
University of Regensburg Centre National de la Recherche
Regensburg, Germany Scientifique-Thales
Sergey D. Ganichev Palaiseau, France
Department of Physics
University of Regensburg Berend T. Jonker
Regensburg, Germany Magnetoelectronic Materials and
Devices Branch
Jean Marie George Materials Science and Technology
Unité Mixte de Physique Division
Centre National de la Recherche Naval Research Laboratory
Scientifique-Thales Washington, DC
Palaiseau, France Tomás Jungwirth
Institute of Physics
Matthew J. Gilbert Academy of Sciences of the Czech
Department of Electrical and Republic
Computer Engineering Praha, Czech Republic
University of Illinois at and
Urbana-Champaign School of Physics and Astronomy
Urbana, Illinois University of Nottingham
and Nottingham, United Kingdom
Department of Electrical
Engineering Stanford Abhinav Kandala
University IBM T.J. Watson Research Center
Stanford, California Yorktown Heights, New York
Contributors xvii
Jeremy Levy John Schliemann

Department of Physics and Department of Physics
Astronomy University of Regensburg
University of Pittsburgh Regensburg, Germany
and
Pittsburgh Quantum Institute Jairo Sinova
Pittsburgh, Pennsylvania Johannes Gutenberg Universität
Mainz
Fumihiro Matsukura Institute of Physics
Tohoku University Mainz, Germany
Sendai, Japan
Carsten Timm
T. L. Hoai Nguyen Institute of Theoretical Physics
Institute of Physics, VAST, 10 Technische Universität Dresden
Daotan, Dresden, Germany
Badinh, Hanoi, Vietnam
Duy-Quang To
Laboratoire des Solides Irradiés,
Hideo Ohno
Ecole Polytechnique
Tohoku University
CNRS and CEA-DRF-IRAMIS
Sendai, Japan
Institut Polytechnique de Paris
Palaiseau, France
Satoshi Okamoto
Materials Science and Technology
Maxim Trushin
Division
Centre for Advanced 2D Materials
Oak Ridge National Laboratory
National University of Singapore
Oak Ridge, Tennessee
Singapore
Yun-Yi Pai Anthony Tylan-Tyler

Department of Physics and Department of Physics and
Astronomy Astronomy
University of Pittsburgh University of Pittsburgh
and and
Pittsburgh Quantum Institute Pittsburgh Quantum Institute
Pittsburgh, Pennsylvania Pittsburgh, Pennsylvania
Anthony Richardella M. W. Wu
Department of Physics and Hefei National Laboratory for
Materials Research Institute Physical Sciences at Microscale
The Pennsylvania State University and Department of Physics
University Park, Pennsylvania University of Science and
Technology of China
Nitin Samarth Hefei, Anhui, China
Department of Physics and
Materials Research Institute Jörg Wunderlich
The Pennsylvania State University Hitachi Cambridge Laboratory
University Park, Pennsylvania Cambridge, United Kingdom
Section IV
Spin Transport
and Dynamics in
Semiconductors
Chapter 1 Spin Relaxation and Spin Dynamics in
Semiconductors and Graphene 3
Chapter 2 Electrical Spin Injection and Transport in
Semiconductors59

Berend T. Jonker
Chapter 3 Spin Transport in Si and Ge: Hot Electron
Injection and Detection Experiments 149

Ian Appelbaum
1
2 Section IV. Spin Transport and Dynamics in Semiconductors
Chapter 4 Tunneling Magnetoresistance, Spin-Transfer

and Spinorbitronics with (Ga,Mn)As 175
Chapter 5 Spin Transport in Organic Semiconductors 247
Chapter 6 Spin Transport in Ferromagnet/III–V
Semiconductor Heterostructures 269
Chapter 7 Spin Polarization by Current 317
Chapter 8 Anomalous and Spin-Injection Hall Effects 339
1
Spin Relaxation and
Spin Dynamics in
Semiconductors
and Graphene
1.1 Toy Model: Electron Spin in a Fluctuating

Magnetic Field 5
1.1.1 Motional Narrowing 12
1.1.2 Reversible Dephasing, Spin Ensemble,
Random Walk in Inhomogeneous Fields 12
1.1.2.1 Reversible Spin Dephasing: Spin
Ensemble in Spatially Random
Magnetic Field 12
1.1.2.2 Spin Echo 14
3
4 Chapter 1. Spin Relaxation and Spin Dynamics in Semiconductors
1.1.2.3 Spin Random Walk in

Inhomogeneous Magnetic Field 14
1.1.3 Quantum Mechanical Description 16
1.1.4 Spin Relaxation of Conduction Electrons 16
1.2 D’yakonov–Perel’ Mechanism 18
1.2.1 Spin–Orbit Field 18
1.2.2 Kinetic Equation for the Spin 19
1.2.3 Persistent Spin Helix 21
1.3 Elliott–Yafet Mechanism 24
1.3.1 Electron-Impurity Scattering 25
1.3.1.1 Spin–Orbit Coupling by the Impurity 25
1.3.1.2 Spin–Orbit Coupling by the Host
Lattice 26
1.3.2 Electron–Phonon Scattering 27
1.3.2.1 Yafet Relation 29
1.4 Results Based on Kinetic-Spin-Bloch-Equation
Approach 31
1.4.1 Kinetic Spin Bloch Equations 31
1.4.2 Spin Relaxation/Dephasing 32
1.4.3 Spin Diffusion/Transport 41
1.5 Spin Relaxation in Graphene 47
Acknowledgments 51
References 51
T
he spin of conduction electrons decays due to the combined effect of
spin–orbit coupling and momentum scattering. The spin–orbit cou-
pling couples the spin to the electron momentum that is randomized
by momentum scattering off of impurities and phonons. Seen from the per-
spective of the electron spin, the spin–orbit coupling gives a spin precession,
while momentum scattering makes this precession randomly fluctuating,
both in magnitude and orientation.
The specific mechanisms for the spin relaxation of conduction elec-
trons were proposed by Elliott [1] and Yafet [2], for conductors with a cen-
ter of inversion symmetry, and by D’yakonov and Perel’ [3], for conductors
without an inversion center. In p-doped semiconductors, there is in play
another spin relaxation mechanism, due to Bir et al. [4]. As this has a rather
limited validity we do not describe it here. More details can be found in
reviews [5–9].
1.1 Toy Model: Electron Spin in a Fluctuating Magnetic Field 5
Before we discuss the two main mechanisms, we introduce a toy model

that captures the relevant physics of spin relaxation without resorting
explicitly to quantum mechanics: the electron spin in a randomly fluctuating
magnetic field. We will find certain universal qualitative features of the spin
relaxation and dephasing in physically important situations.
The next part of this review covers the experimental as well as compu-
tational status of the field, discussing the spin relaxation in semiconductors
under varying conditions such as temperature and doping density.
1.1 TOY MODEL: ELECTRON SPIN IN A

FLUCTUATING MAGNETIC FIELD
Consider an electron spin S (or the corresponding magnetic moment) in the
presence of an external time-independent magnetic field B0 = B0z, giving rise
to the Larmor precession frequency ω0 = ω 0z, and a fluctuating time-depen-
dent field B(t) giving the Larmor frequency ω(t) (see Figure 1.1). We assume
that the field fluctuates about zero and is correlated on the -timescale of τc:
w(t ) = 0, wa (t )wb (t ¢) = dab w2ae -|t -t ¢|/ tc . (1.1)
Here α and β denote the Cartesian coordinates and the overline denotes
averaging over different random realizations B(t). We will see later that such
fluctuating fields arise quite naturally in the context of the electron spins in
solids.
The following description applies equally to the classical magnetic
moment described by the vector S as well as to the quantum mechanical
spin whose expectation value is S. Writing out the torque equation, S = w ´ S ,
we get the following equations of motion:
S x = −ω 0 S y + ω y (t )Sz − ω z (t )S y , (1.2)
S y = ω 0 Sx − ω x (t )Sz + ω z (t )Sx , (1.3)
B0, ω0
B(t), ω(t)
FIGURE 1.1 Electron spin precesses about the static B0 field along z. The randomly
fluctuating magnetic field B(t) causes spin relaxation and spin dephasing.
Sz = ω x (t )S y − ω y (t )Sx . (1.4)
These equations are valid for one specific realization of ω(t). Our goal is to
find instead effective equations for the time evolution of the average spin,
S(t ) , given the ensemble of Larmor frequencies ω(t).
It is convenient to introduce the complex “rotating” spins S± and Larmor
frequencies ω± in the (x,y) plane:
S+ = Sx + iS y , S− = Sx − iS y , (1.5)
ω + = ω x + iω y , ω − = ω x − iω y . (1.6)
The inverse relations are
1 1
Sx = ( S+ + S− ) , S y = ( S+ − S− ) , (1.7)
2 2i
1 1
ωx = ( ω + + ω − ) , ω y = ( ω + − ω − ) . (1.8)
2 2i
The equations of motion for the spin set (S+, S−, Sz) are
S+ = iω 0 S+ + iω z S+ − iω + Sz , (1.9)
S− = −iω 0 S− − iω z S− + iω − Sz , (1.10)
1 
Sz = −  i ( ω + S− − ω − S+ ) . (1.11)
2 
In the absence of the fluctuating fields, the spin S+ rotates in the complex
plane anticlockwise (for ω 0 > 0) while S− clockwise.
The precession about B0 can be factored out by applying the ansatz*:
S± = s± (t )e ± iω0t . (1.12)
Indeed, it is straightforward to find the time evolution of the set (s+, s−,
sz ≡ Sz ):
s+ = iω z s+ − iω + sz e − iω0t , (1.13)
s− = −iω z s− + iω − sz e iω0t , (1.14)
1 
sz = −  i  ω + s− e − iω0t − ω − s+ e iω0t  . (1.15)
2 
* This is analogous to going to the interaction picture when dealing with a quantum mechani-
cal problem of that type.
The penalty for transforming into this “rotating frame” is the appearance of
the phase factors exp(±iω 0t).
The solutions of Equations 1.13 through 1.15 can be written in terms of
the integral equations
t t
∫ ∫
s+ (t ) = s+ (0) + i dt ′ω z (t ′)s+ (t ′) − i dt ′ω + (t ′)sz (t ′)e − iω0t ′ , (1.16)
0 0
t t
∫ ∫
s− (t ) = s− (0) − i dt ′ω z (t ′)s− (t ′) + i dt ′ω − (t ′)sz (t ′)eiω0t ′ , (1.17)
0 0
t
1
sz (t ) = sz (0) −
2i ∫
dt ′  ω + (t ′)s− (t ′)e − iω0t ′ − ω − (t ′)s+ (t ′)e iω0t ′  . (1.18)
0
We should now substitute the above solutions back into Equations 1.13
through 1.15. The corresponding expressions become rather lengthy, so we
demonstrate the procedure on the s+ component only. We get
∫
s+(t ) = iω z (t )s+ (0) − ω z (t ) dt ′ω z (t ′)s+ (t ′)
0
t

∫
+ ω z (t ) dt ′ω + (t ′)sz (t ′)e − iω0t ′ − iω + (t )e − iω0t sz (0)
0
(1.19)
t
1
2 ∫
+ e − iω0t ω + (t ) dt ′  ω + (t ′)s− (t ′)e − iω0t ′ − ω − (t ′)s+ (t ′)e iω0t ′  .
0
The reader is encouraged to write the analogous equations for sz (that for s−
is easy to write since s− = s+* ).
We now make two approximations. First, we assume that the fluctuat-
ing field is rather weak and stay in the second order in ω.* This allows us to
factorize the averaging over the statistical realizations of the field
ω(t )ω(t ′)s(t ′) ≈ ω(t )ω(t ′) s(t ′) (1.20)
as the spin changes only weakly over the timescale, τc, of the changes of
the fluctuating fields. This approximation is called the Born approximation,
alluding to the analogy with the second-order time-dependent perturba-
tion theory in quantum mechanics. Going beyond the Born approximation
one would need to execute complicated averaging schemes of the product in
Equation 1.20, since s(t) in general depends on ω(t′ ≤ t).
* More precisely, we assume that |ω(t)|τc ≪ 1, so that the spin does not fully precess about the
fluctuating field before the field makes a random change.
As the second assumption, we consider a “coarse-grained” time evolu-

tion, meaning that we are effectively averaging s(t) over the timescale of the
correlation time τc; we are interested in times t much greater than τc. That
allows us to approximate
t τc t τc

∫0
dt ′ ω(t )ω(t ′) s(t ′) ≈
∫ dt ′ω(t)ω(t ′) s(t), (1.21)
0
since the correlation function ω(t )ω(t′) is significant in the time interval of
|t − t′| ≈ τc only. The above approximation makes clear that the spin s(t) is the
representative coarse-grained (running-averaged) spin of the time interval
(t − τc, t). Equation 1.21 is a realization of the Markov approximation. The
physical meaning is that the spin s varies only slowly on the timescale of τc
over which the correlation of the fluctuating fields is significant. We then
need to restrict ourselves to the timescales t larger than the correlation time
τc. In effect, we will see that in this approximation the rate of change of the
spin at a given time depends on the spin at that time, not on the previous his-
tory of the spin.
Applying the Born–Markov approximation to Equation 1.19, we obtain
for the average spin s+ the following time evolution equation*:
∫
s+ = i ω z (t )s+ (0) − dt ′ ω z (t )ω z (t ′) s+ (t )
0
t
1

∫
+ dt ′ ω z (t )ω + (t ′)e − iω0t ′ sz (t ) − i ω + (t )e − iω0t sz (0) + e − iω0t (1.22)
0
2
t
∫
× dt ′ ω + (t )ω + (t ′)e − iω0t ′ s− (t ) − ω + (t )ω − (t ′)e iω0t ′ s+ (t ) .
0
Using the rules of Equation 1.1, Equation 1.22 simplifies to
t
1
∫
s+ = −ω 2z dt ′e −(t − t ′ )/ τc s+ (t ) + e − iω0t
0
2
(1.23)
t
∫
× dt ′ (ω 2x − ω 2y )e − iω0t ′ s− (t ) − (ω 2x + ω 2y )e iω0t ′ s+ (t ) e −(t − t ′ )/ τc .
0
 
Since we consider the times t ≫ τc, we can approximate
t t

∫0
∫
dt ′e −(t − t ′ )/ τc ≈ dt ′e −(t − t ′ )/ τc = τ c . (1.24)
−∞
* The initial values of the spin, s(0), are fixed and not affected by averaging.
Similarly,
t t
1 ± iω 0 τ c

∫0
∫
dt ′e −(t − t ′ )/ τc e − iω0 (t ± t ′ ) ≈ dt ′e −(t − t ′ )/ τc e − iω0 (t ± t ′ ) = τ c
−∞
1 + ω 20 τ 2c
. (1.25)
The imaginary parts induce the precession of s±, which is equivalent to shift-
ing (renormalizing) the Larmor frequency ω 0. The relative change of the
frequency is (ωτc)2, which is assumed much smaller than one by our Born
approximation. We thus keep the real parts only and obtain
1 τc  
s+ = −ω 2z τ c s+ + 2 2
(ω x − ω y )s− e
−2 iω 0 t
− (ω 2x + ω 2y )s+  . (1.26)
2 1 + ω 20 τ 2c  
Using the same procedure (or simply using s− = s*+ ), we would arrive for the
analogous equation for s−:
1 τc  
s+ = −ω 2z τ c s− + 2 2
(ω x − ω y )s+ e
2 iω 0 t
− (ω 2x + ω 2y )s−  . (1.27)
2 1 + ω 20 τ 2c  
Similarly,
τc
sz = −(ω 2x + ω 2y ) sz . (1.28)
1 + ω 20 τ 2c
For the rest of the section we omit the overline on the symbols for the spins,
so that S will mean the average spin. Returning back to our rest frame of the
spins rotating with frequency ω 0, we get
1 τc  
S+ = iω 0 S+ − ω 2z τ c 2 2 2 2
(ω x − ω y )S− − (ω x + ω y )S+  , (1.29)
2 1 + ω 20 τ 2c  
1 τc  
S− = iω 0 S− − ω 2z − 2 2 2 2
(ω x − ω y )S+ − (ω x + ω y )S−  , (1.30)
2 1 + ω 20 τ 2c  
τc
Sz = −(ω 2x + ω 2y ) Sz . (1.31)
1 + ω 20 τ 2c
Finally, going back to S x and Sy
τc
Sx = −ω 0 S y − ω 2z τ c Sx − ω 2y Sx , (1.32)
1 + ω 20 τ 2c
τc
S y = ω 0 Sx − ω 2z τ c S y − (ω 2x )S y , (1.33)
1 + ω 20 τ 2c
τc
Sz = −(ω 2x + ω 2y ) Sz . (1.34)
1 + ω 20 τ 2c
We can give the above equation a more conventional form by introducing

two types of the spin decay times. First, we define the spin relaxation time
T1 by
1  2  τc
=  ω x + ω 2y  , (1.35)
T1  1 + ω 20 τ 2c
and the spin dephasing times T2 by
1 ω 2y τ c
= ω 2z τ c + , (1.36)
T2 x 1 + ω 20 τ 2c
1 ω 2x τ c
= ω 2z τ c + . (1.37)
T2 y 1 + ω 20 τ c2
We then write
S
Sx = − ω 0 S y − x , (1.38)
T2 x
Sy
S y = ω 0 Sx − , (1.39)
T2 y
S
Sz = − z . (1.40)
T1
Our fluctuating field is effectively at infinite temperature, at which the
average value for the spin in a magnetic field is zero. A more general spin
dynamics is
S
Sx = −ω 0 S y − x , (1.41)
T2 x
Sy
S y = ω 0 Sx − , (1.42)
T2 y
S − S0 z
Sz = − z . (1.43)
T1
where S0z is the equilibrium value of the spin in the presence of the static
magnetic field of the Larmor frequency ω0 at the temperature at which the
environmental fields giving rise to ω(t) are in equilibrium. The above equa-
tions are called the Bloch equations.
The spin components S x and Sy, which are perpendicular to the applied
static field B0, decay exponentially on the timescales of T2x and T2y, respec-
tively. These times are termed spin dephasing times, as they describe the
loss of the phase of the spin components perpendicular to the static field B0.
They are also often called transverse times, for that reason. The time T1 is
termed the spin relaxation time, as it describes the (thermal) relaxation of
the spin to the equilibrium. During the spin relaxation in a static magnetic
field, the energy is exchanged with the environment. In the language of sta-
tistical physics, the relaxation process establishes the Boltzmann probabil-
ity distribution for the system. Similarly, dephasing establishes the “random
phases” postulate that says that there is no correlation (coherence) among
the degenerate states, such as the two transverse spin orientations; in ther-
mal equilibrium such states add incoherently.
For the sake of discussion consider an isotropic system in which
ω 2x = ω 2y = ω 2z = ω 2 . (1.44)
If the static magnetic field is weak, ω 0 τc ≪ 1, the three times are equal:
1
T1 = T2 x = T2 y = . (1.45)
ω 2τc
There is no difference between the spin relaxation and spin dephasing. It
is at first sight surprising that the spin relaxation time is inversely propor-
tional to the correlation time. The more random the external field appears,
the less the spin decays. We will explain this fact below by the phenomenon
of motional narrowing.
In the opposite limit of large Larmor frequency, ω 0 τc ≫ 1, the spin relax-
ation rate vanishes
1 ω2 1
≈ 2 2 → 0, (1.46)
T1 ω0 τc
while the spin dephasing time is given by what is called secular broadening:
1
≈ ω 2z τ c . (1.47)
T2
If secular broadening is absent, the leading term in the dephasing time will
be, as in the relaxation
1 ω2 1
≈ . (1.48)
T2 ω 20 τ c
In this limit the spin dephasing rate is proportional to the correlation rate,
not to the correlation time.
In many physical situations the fluctuating fields are anisotropic. Perhaps
the most case is such that ωz = 0 while
ω 2x = ω 2y = ω 2 . (1.49)
Then, from Equations 1.35 through 1.37, it follows that
T2 = 2T1 . (1.50)
This anisotropic spin relaxation is due to the fact that while the trans-
verse spin can be flipped by two modes of the in-plane fluctuating field,
the in-plane spins can be flipped by one mode only (the one perpendicu-
lar to the spin).
In the cases in which there is no clear distinction between T1 and T2, we
often use the symbol
τ s = T1 = T2 , (1.51)
to describe the generic spin relaxation.
1.1.1 Motional Narrowing

The surprising fact that at low magnetic fields the spin relaxation rate is pro-
portional to the correlation time of the fluctuating field (as opposed to its
inverse) is explained by motional narrowing. Consider the spin transverse to
an applied magnetic field and assume that the field has a single magnitude,
but can randomly switch directions, between up to down, leading to a ran-
dom precession of the spin clock- and anticlockwise. In effect, the spin phase
executes a random walk. A single step takes the time τc, the correlation time
of the fluctuating field. After n steps, that is, after the time t = nτ, the stan-
dard deviation of the phase will be δφ = (ωτ c ) n , the well-known result for
a random walk. We call the spin dephasing time the time it takes for δϕ ≈ 1.
This happens after the time τs = τc/(ωτc)2, or τs = 1/(ω2τc), which is the result
we obtained earlier from the Born–Markov approximation, Equation 1.45.
eversible Dephasing, Spin Ensemble,

1.1.2 R
Random Walk in Inhomogeneous Fields
Our previous calculation was carried out for a single spin in a fluctuating
magnetic field. After the decay of the spin components, the information
about the original spin is irreversibly lost as we have no information on the
actual history of the fluctuating field. Such an irreversible loss of spin is often
termed spin decoherence. We will see that spin dephasing can be reversible.
We will also see that a simple exponential decay, of the type exp(−t/τs), is not
a rule.
1.1.2.1 Reversible Spin Dephasing: Spin Ensemble

in Spatially Random Magnetic Field
There are physically relevant cases in which the decay of spin is reversible.
A typical example is an ensemble of localized spins, each precessing about a
static local magnetic field B0 + B1, with varying static components B1 (of zero
average) giving rise to random precession frequencies ω1 (see Figure 1.2).
Another important example is that of the conduction electrons in noncen-
trosymmetric crystals, such as GaAs, in which the spin–orbit coupling acts
as a momentum-dependent magnetic field. The spins of the electrons of dif-
ferent momenta precess with different frequencies. We are interested in the
total spin as the sum of the individual spins.
S(t)
FIGURE 1.2 Electron spin precesses about the static B0 field along z perpendicu-
lar to this page. The spatially fluctuating magnetic field B1z causes reversible spin
dephasing.
Consider the external field along z direction, and take the fluctuating
frequencies from the Gaussian distribution
1 2 2
P (ω 1 ) = e − ω1 / 2 δω , (1.52)
2
2πδω
2
with zero mean and δω variance. Denote the in-plane spin of the electron a*
1
by Sxa and S ya . This spin precesses with the frequency ω 0 + ω1a , leading to the
time evolution for the rotating spins:
a
S±a (t ) = Sxa (t ) ± iS ya (t ) = S±a (0)e ± iω0t e ± iω1 t . (1.53)
Suppose at t = 0 all the spins are lined up, that is, S±a (0) = S± (0). The total spin
S±(t) is the sum
S± (t ) = ∑S (t) = S (0)e ∫ dω P(ω )e

s
a
± ±
± iω 0 t
−∞
1 1
± i ω1 t
. (1.54)
Evaluating the Gaussian integral we get
2 2
S± (t ) = S± (0)e ± iω0t e − δω1 t / 2 . (1.55)
The in-plane component vanishes after the time of about 1/δω1, but this
dephasing of the spin is reversible, since each individual spin preserves the
memory of the initial state. The disappearance of the spin is purely due to
the statistical averaging over an ensemble in which the individual spins have,
after certain time, random phases. This spin decay is not a simple exponen-
tial, but rather Gaussian.
* This discussion applies to nuclear spins as well.

1.1.2.2 Spin Echo
We have seen that there are irreversible and reversible effects both present in
spin dephasing. It turns out that the reversible effects can be separated out
by the phenomenon of the spin echo. Figure 1.3 explains this mechanism in
detail. Suppose all the localized spins in our ensemble point in one direction
at time t = 0. At a later time, t = Tπ, the spins dephase due to the inhomoge-
neities of the precession frequencies and the total spin is small. We apply a
short pulse of an external magnetic field, the so-called π pulse, that rotates
the spins along the axis parallel to the original spin direction, mapping the
spins as (S x, Sy) → (−S x, Sy). At that moment the spins will still be dephased,
but the one that is the fastest is now the last, and the one that is the slowest
appears as the first. At the time t = 2Tπ all the spins catch on, producing a
large spin signal along the original spin direction. In reality this signal will
be weaker than that at t = 0 due to the presence of irreversible processes, by
exp(−2Tπ/T2). Important, reversible processes are not counted in T2.
1.1.2.3 Spin Random Walk in Inhomogeneous Magnetic Field

Another interesting situation appears when we consider the possibility
that the spin diffuses through a region of an inhomogeneous precession
frequency.* We can model this situation on the system of spins in one dimen-
sion, each spin jumping in a time τ left or right. Suppose the precession fre-
quencies vary in the x direction (see Figure 1.4) as
ω( x) = ω 0 + ω′x. (1.56)
32
1 3
2
1
Before π pulse
3
2
3 1
2
1
Time
After π pulse
FIGURE 1.3 Initially all the spins point up. Due to random precession frequencies
the spins soon point in different directions and the average spin vanishes.
Applying a π-pulse rotating the spins along the vertical axis makes the fastest spin
the slowest, and vice versa. After the fastest ones catch up again with the slowest,
the original value of the average spin is restored. Any reduction from the original
value is due to irreversible processes signal.
* The inhomogeneity could be due to the magnetic field or to the g-factor.

ω 0 + ω´x
x S(t)
FIGURE 1.4 The electron performs a random walk. Its spin precesses by the inhomo-
geneous magnetic field along the x-axis. The average spin dephases to zero with time.
Here ω′ is the gradient of ω. As the electron diffuses, it precesses with the

precession frequency ω(t) = ω[x(t)], given by the position of the electron x(t)
at time t. The time evolution for an individual spin is

∫
S+ = ω(t )S+ (t ) = ω(t )S+ (0) + ω(t ) dt ′ω(t ′)S(t ′). (1.57)
0
We are interested in the averaged quantities over many realizations of the

random walk. We also consider the timescales much greater than the indi-
vidual random walks steps τ.
The accumulated phase after N steps or t = Nτ time is the sum
φ N = x1 + x2 + + x N = N δ1 + ( N − 1)δ1 + + δ N . (1.58)
Here δi = ±1 is a random variable representing the random step left or right.
The variance of ϕN is
N −1
∑( N − i) ≈ 3 N . (1.59)
1
σ 2N = 2 3
According to the central limit theorem, ϕN is distributed normally with the

above variance:
1 2 2
P (φ N ) = e − φ N / 2 σ N . (1.60)
2πσ 2N
We can now transform the dynamical equation into the ensemble averaging
∫
2 3
S+ (t ) = S+ (0)e iφ(t ) = dφ N e iφ(t ) P(φ N ) = e − ω1 Dt /3
, (1.61)
−∞
where we denoted the diffusivity by D = τ/2, for the unit length step.
1.1.3 Quantum Mechanical Description

The Born–Markov approximation to the dynamics of a system coupled to an
external environment can be cast in the quantum mechanical language. We
refer the reader to Ref. [8] for the derivation. The physics of this derivation
is the same as what we did in Section 1.1. Only the formalism is different.
We consider the system described by the Hamiltonian
H (t ) = H 0 + V (t ), (1.62)
in which H0 is our system per se and V(t) is the time-dependent random

fluctuating field of zero average and correlation time τc:
V (t ) = V (t ) = 0, V (t )V (t ′) ~ e −|t − t ′|/ τc . (1.63)
The system is fully described by the density matrix ρ.

The transformation to the rotating frame is equivalent to going to the
interaction picture in quantum mechanics:
ρ I (t ) = e iH 0t / ρe − iH 0t / , (1.64)
V I (t ) = e iH 0t / V (t )e − iH 0t / . (1.65)
Performing the operations, as outlined in Section 1.1, for the classical model,
we arrive at the effective time evolution for the density of state operator:
2 t τc
d ρ I (t )  1 

dt
= 
 i  ∫ dt ′ V (t),[V (t ′),ρ (t)] . (1.66)
0
I I I
The above equation is called the Master equation and is the starting equa-
tion in many important problems in which a quantum system is in contact
with a reservoir (Figure 1.4).
pin Relaxation of Conduction Electrons

1.1.4 S
We will consider nonmagnetic conductors in zero or weak magnetic fields so
that the spin dephasing and spin relaxation times are equal, T2 = T1 = τs. The
formula for the spin relaxation time
1
= ω 2 τ c , (1.67)
τs
that was derived above for the spin in a fluctuating magnetic field applies in
a semiquantitative sense (i.e., it gives an order of magnitude estimates and
useful trends) to the conduction electron spins as well. We analyze below the
D’yakonov–Perel’ [3] and the Elliott–Yafet [1, 2] mechanisms.
The D’yakonov–Perel’ mechanism is at play in solids lacking a cen-
ter of spatial inversion symmetry. The most prominent example is the
semiconductor GaAs. In such solids the spin–orbit coupling is manifested

as some effective magnetic field—the spin–orbit field—that depends on the
electron momentum. Electrons in different momentum states feel different
spin–orbit fields, so that the spin precesses with a given Larmor frequency,
until the electron is scattered into another momentum state (see Figure 1.5).
As the electron momentum changes on the timescale τ of the momentum
relaxation time, the net effect of the momentum scattering on the spin is to
produce random fluctuations of the Larmor frequencies. We have motional
narrowing. Since these frequencies are correlated by τc = τ, we arrive at
1
= ω 2soτ, (1.68)
τs
for the spin relaxation time. The magnitude of ωso is the measure of the
strength of the spin–orbit coupling. The spin relaxation rate is directly pro-
portional to the momentum relaxation time—the more the electron scatters,
the less its spin dephases.
The Elliott–Yafet mechanism works in systems with and without a cen-
ter of inversion. It relies on spin-flip momentum scattering. The spin-flip
amplitudes are due to spin–orbit coupling, while the momentum scattering
is due to the presence of impurities (that also contribute to the spin–orbit
coupling), phonons, rough boundaries, or whatever is capable of random-
izing the electron momentum. During the scattering events, the spin is pre-
served (see Figure 1.6). How do we account for such a scenario with our toy
model? Consider an electron scattering off of an impurity with a spin flip.
This spin flip can be viewed as a precession that occurs during the time of
the interaction of the electron with the impurity. Let us say that the scat-
tering takes the time of λF/υF , where υF is the electron velocity and λF is the
electron wavelength at the Fermi level (considering that it is greater or at
most equal to the size a of the impurity—otherwise we could equally put
a/υF ). Then, the precession angle φ = ωso(λF/υF ), with ωso denoting the spin–
orbit coupling-induced precession frequency. Let us compare this with the
Impurity
Phonon
FIGURE 1.5 D’yakonov–Perel’ mechanism. The electron starts with the spin up. As
it moves, its spin precesses about the axis corresponding to the electron velocity.
Phonons and impurities change the velocity, making the spin to precess about a
different axis (and with different speed). During the scattering event the spin is
preserved.
Impurity
Phonon
FIGURE 1.6 Elliott–Yafet mechanism. The electron starts with the spin up. As it
scatters off of impurities and phonons, the spin can also flip due to spin–orbit cou-
pling. Between the scattering the spin is preserved. After, say, a million scattering
events, the spin will be down.
angle of precession, ωτ, in the motional narrowing model, Equation 1.67. We

see that the spin-flip can be described by the effective precession frequency
ω = ωso(λF/υFτ). We then obtain for the spin relaxation rate
2
1 λ2  ε  1
= ω 2so 2F ≈  so  . (1.69)
τs vF τ  ε F  τ
Here εso = ћωso and ε = (ћkF )vF/2 is the Fermi energy. For the Elliott–Yafet
mechanism holds that the more the electron scatters, the more the spin
dephases.
Below we discuss the two mechanisms in more detail, providing the
formalisms for their investigation.
1.2 D’YAKONOV–PEREL’ MECHANISM

D’yakonov and Perel’ [3] considered solids without a center of inversion
symmetry, such as GaAs or InAs. The mechanism also works for elec-
trons at surfaces and interfaces. In such solids, the presence of spin–orbit
coupling induces the spin–orbit fields, which give rise to the spin pre-
cession. Momentum relaxation then causes the random walk of the spin
phases.
pin–Orbit Field
1.2.1 S
In solids without a center of inversion symmetry, spin–orbit coupling splits
the electron energies:
ε k ,↑ ≠ ε k ,↓ . (1.70)
Only the Kramers degeneracy is left, due to time reversal symmetry:
ε k ,↑ = ε − k ,↓ . (1.71)
This energy splitting at a given momentum k is conveniently described by

the spin–orbit field Ω, giving a Zeeman-like (but momentum-dependent)
energy contribution to the electronic states, described by the additional
Hamiltonian (to the usual band structure):

H1 = W k × s. (1.72)
2
The time reversal symmetry requires that the spin–orbit field is an odd func-
tion of the momentum:
Wk = − W − k. (1.73)
The representative example of a spin–orbit field is the Bychkov–Rashba [10]

(αBR) and Dresselhaus [11, 12] (γD) couplings in 2d electron gases formed at
the zinc-blend heterostructures grown along [001] [8]:
H 1 = (α BR + γ D )σ x k y − (α BR − γ D )σ y k x . (1.74)
The axes are x = [110] and y = [110]. The spin–orbit field is
W k = 2 éëê(a BR + g D )k y ,-(a BR - g D )k x ùûú . (1.75)
This field has the C2v symmetry, reflecting the structural symmetry of the
zinc-blend interfaces (such as GaAs/GaAlAs) with the principal axes along
[110] and [110].
1.2.2 Kinetic Equation for the Spin

Let sk be the electron spin in the momentum state k. The time evolution of
the spin is then described by
¶s k

¶t
- W k ´ sk = - åW
k¢
kk ¢ ( s k - s¢k ) . (1.76)
The left-hand side gives the full time derivative dsk/dt, which is due to the
explicit change of the spin, its direction, and the change of the spin due to the
change of the momentum k. The right-hand side is the change of the spin at
k due to the spin-preserving scattering from and to that state. The scattering
rate between the two momentum states k and k′ is Wkk¢.
There are two timescales in the problem. First, momentum scattering,
which occurs on the timescale of the momentum relaxation time τ, makes the
spins in different momentum states equal. Second, this uniform spin decays
on the timescale of the spin relaxation time τs, which we need to find. We
assume that τ ≪ τs; this assumption is well satisfied in real systems. In prin-
ciple we could go directly to our model of the electron spin in a fluctuating
magnetic field, with the role of the random Larmor precession playing by Ω,
identifying τc = τ. However, it is instructive to see how this two timescale

problem is solved directly.
We separate the fast and slow components of the spins as follows:
s k = s + x k, x k = 0. (1.77)
The symbol denotes averaging over different momenta. Our goal is to

find the effective equation for the time evolution of s, which is the actual spin
averaged over k. The fast component, ξk, decays on the timescale of τ to the
value given by the instantaneous value of s. Our goal is to find the effective
equation for the time evolution of s, which is the actual spin averaged over k.
The fast component, ξk, decays on the timescale of τ to the quasistatic value
given by the instantaneous value of s.
Upon substituting Equation 1.77 into the kinetic Equation 1.76 and aver-
aging over k, we obtain the equation of motion for the averaged spin
s = W k ´ x k , (1.78)
using that Wk = 0 . We need to find ξk. Since s is hardly changing on the

timescales relevant to ξk, we write
x k − Wk × s − Wk × x k = − ∑ Wkk ′ ( x k − x k ′ ).
(1.79)
k′
We make the following assumption:
Ωτ 1. (1.80)
That is, we assume that the precession is slow on the timescale of the
momentum relaxation time (see our note on the Born approximation in
the toy model in Section 1.1). For simplicity, we make the relaxation time
approximation to model the time evolution of the fast component ξk:
xk
xk - W k ´ s - W k ´ xk = - . (1.81)
t
From the condition of the quasistatic behavior, ∂ξk/∂t = 0, we get up to the
first order in Ωτ the following solution for the quasistatic ξk:
x k = t ( W k ´ s ) . (1.82)
The above is a realization of coarse graining, in which we effectively average

the spin evolution over the timescales of τ.
Substituting the quasistatic value of ξk from Equation 1.82 to the time
evolution equation for s, Equation 1.78, we find
s = t W k ´ ( W k ´ s ) . (1.83)
Using the vector product identities, we finally obtain for the individual spin
components α:
sa = WkaW kb tsb - W 2kt sa . (1.84)
These equations describe the effective time evolution of the electron spin in
the presence of the momentum-dependent Larmor precession Ωk.
For the specific model of the zinc-blend heterostructure with the C2v
spin–orbit field, Equation 1.75, we obtain the spin dephasing dynamics
− sx − sy − sy
s x = , s y = , sz = , (1.85)
τs τs τz
where:

1
=
( α BR ± γ D )2
1
,
1 2
= , (1.86)
τ x, y α 2BR + γ 2D τ s τ z τ s
and
1 4m  2
= ε k  α BR + γ 2D  . (1.87)
τ s 4 
The up (down) sign is for the sx (sy). The spin relaxation is anisotropic. The
maximum anisotropy is for the case of equal magnitudes of the Bychkov–
Rashba and Dresselhaus interactions, αBR = ±γD. In this case one of the spin
components does not decay.* The sz component of the spin relaxes roughly
twice faster than the in-plane components.
ersistent Spin Helix

1.2.3 P
Let us consider the case of αBR = γD = λ/2. According to Equation 1.86, the
spin component sx does not decay, while the decay rates of sy and sz are the
same:
−2 s y −2 sz
s y = , sz = . (1.88)
τs τs
It turns out that a particular nonuniform superposition of sy and sz can
exhibit no decay as well. This superposition has been termed persistent spin
helix [13].
Let us assume that the spin is no longer uniform, so that the kinetic
equation contains the spin gradient as well, due to the quasiclassical change
of the electronic positions:
∂s k ∂s
− Wk × s k + k ⋅ v k = − ∑ Wkk ′ ( s k − s′k ) . (1.89)
∂t ∂r k′
* The decay of that component would be due to higher-order (such as cubic) terms in the spin–
orbit fields.
A running spin wave
s k (r) = s k e iq⋅r ; s k ≡ s k (q), (1.90)
then evolves according to
∂s k
− Wk × s k + i ( q ⋅ v k ) s k = − ∑ Wkk ′ ( s k − s′k ) . (1.91)
∂t k′
We again separate the fast and slow spins as follows:
sk = s + x k , x k = 0. (1.92)
For the dynamics of the slow part we get
s k = Wk × x k − i ( q ⋅ v k ) x k . (1.93)
Proceeding as in the previous section, assuming that s is stationary on the

timescales relevant to ξ, we can write
x k − W k × s − W k × x k + i ( q ⋅ v k ) s + i ( q ⋅ v k ) x k
(1.94)
= − ∑ Wkk ′ ( x k − x k ′ ).
k′
We now work with the following assumptions
Ωτ 1, q 1, (1.95)
where ℓ = gτ is the mean free path. We thus assume that the precession is
slow on the timescale of the momentum relaxation time, as well as (this is
new here) the electronic motion is diffusive on the scale of the wavelength of
the spin wave. In the momentum relaxation approximation, also considering
the leading terms according to the conditions in Equation 1.95, we get
x
x k − W k × s + i ( q ⋅ v k ) s = − . (1.96)
τ
In the steady state, corresponding to a given s(t), the solution is
x k = τ ( Wk × s ) − iτ ( q ⋅ v k ) s. (1.97)
Substituting to the time evolution equation for s, Equation 1.95, we find
s = − 2iτ ( q ⋅ v k ) (Wk × s ) + τ Wk × (Wk × s ) − τ ( q ⋅ v k )2 s. (1.98)
Using the vector product identities and introducing the diffusivity
D = v 2kα τ, (1.99)
where α denote the Cartesian coordinates (we assume an isotropic system),

we finally obtain
sα = −2iτε αβγ qδ vkδ Ω kβ sγ + Ω kα Ω kβ τsβ − W2kτ sα − Dq 2 sα . (1.100)
For our specific case of
Ω k = (λk y , 0, 0), λ = α BR + β D , (1.101)
we find that sx decays only via diffusion:
sx = − Dq 2 sx . (1.102)
The spin dephasing is ineffective. This is an expected result.

More interesting behavior is found for the two remaining spin compo-
nents, sy and sz. These two spin components, transverse to the spin–orbit
field, are coupled
mλ
sy = +2i q y Dsz −  τΩ2 + Dq 2  s y , (1.103)

mλ
sz = −2i q y Ds y −  τΩ2 + Dq 2  sz , (1.104)

and we denoted Ω2 ≡ W2k . Interestingly, this set of coupled differential
equations has only decaying solutions. This is best seen by looking at the
rotating spins
s+ = s y + isz , s− = s y − isz , (1.105)
whose time evolutions are uncoupled
 2mλ 
s+ = −  Ω2 τ + Dq 2 − q y D s+ , (1.106)
 
 2mλ 
s− = −  Ω2 τ + Dq 2 + q y D s− . (1.107)
 
Considering that
2 2
 m  m
Ω2 τ = λ 2 k y2 τ = λ 2   v 2y τ = λ 2   D, (1.108)
   
we find that the decay of s+ vanishes for the wave-vector
mλ m
q PSH
y = = ( α BR + β D ) . (1.109)

The abbreviation PSH stands for the persistent spin helix, which is the spin
wave described by s+ at this particular wave-vector (see Figure 1.7). While
FIGURE 1.7 The persistent spin helix is a wave of circularly polarized spin. The
sense of polarization, clock or counterclockwise, depends on the relative sign of
the Dresselhaus and the Bychkov–Rashba spin–orbit coupling.
individually both sy and sz decay at a generic wave-vector (and also in the

uniform case, q = 0), the spin helix they form does not decay in the approxi-
mation of the linear spin–orbit field. The spin wave rotating in the opposite
sense, s−, on the other hand, decays. And vice versa for q PSH
y = −mλ / . The
persistent spin helix was observed in the spin grating experiment [14].
1.3 ELLIOTT–YAFET MECHANISM
Elliott [1] was first to recognize the role of the intrinsic spin–orbit coupling—
that coming from the host ions—on spin relaxation. Yafet [2] significantly
extended this theory to properly treat electron–phonon spin-flip scattering.
The Elliott–Yafet mechanism dominates the spin relaxation of conduction
electrons in elemental metals and semiconductors, the systems with space
inversion symmetry. In systems lacking this symmetry, the mechanism
competes with the D’yakonov–Perel’ one; the dominance of one over the
other depends on the material in question and specific conditions, such as
temperature and doping.
Suppose a nonequilibrium spin accumulates in a nonmagnetic degen-
erate conductor. The spin accumulation is given as the difference between
the chemical potentials for the spin-up and the spin-down electrons. Let us
denote the corresponding potentials by μ↑ and μ↓. The nonequilibrium elec-
tron occupation function for the spin λ is
1  ∂f 0 
f λk ≈ β( ε k − µ λ )
≈ f k0 +  − k  (µ λ − ε F ). (1.110)
e +1  ∂ε k 
Here
1
f k0 = f 0 (ε k ) = , (1.111)
e β( ε k − ε F ) + 1
describes the equilibrium degenerate electronic system of the Fermi energy
εF . The electron density for the spin λ is
n  ∂f 0 ( ε ) 

∫
nλ = d εg s (ε) f λ0 (ε) ≈
2 ∫
+ d εg (ε)  −
 ∂ε 
(µ λ − ε F ). (1.112)
Here gs(ε) is the electronic density of states, defined per unit volume and per
spin, at the Fermi level:
 ∂f k0 
gs = ∑
k
 − ∂ε  . (1.113)
 k
The total electron density n is

∫
n = 2 d εg (ε) f 0 (ε). (1.114)
We assume that the spin accumulation does not charge the system, that is,
the charge neutrality is preserved n↑ + n↓ = n. This condition is well satisfied
in metals and degenerate semiconductors that we consider. We then get
µ ↑ + µ ↓ = 2ε F . (1.115)
The spin density is
s = n↑ − n↓ = g s (µ ↑ − µ ↓ ) = g sµ s , (1.116)
where μs is the spin quasichemical potential, μs(μ↑ − μ↓).

The spin relaxation time T1 is defined by the decay law:
ds dn↑ dn↓ s µ
= − = W↑↓ − W↓↑ = − = − g s s .
dt dt dt T1 T1 (1.117)
Here W↑↓ is the net number of transitions per unit time from the spin ↓
to ↑. Similarly, W↓↑ expresses the rate of spin flips from ↑ to ↓. In the degen-
erate conductors, the spin decay is directly proportional to the decay of the
spin accumulation μs:
dµ s µ
= − s . (1.118)
dt T1
lectron-Impurity Scattering
1.3.1 E
We need to distinguish the cases of the impurity or host-induced spin–orbit
coupling. Although the formulas for the calculation of T1 look similar in the
two cases, they are nevertheless conceptually different.
1.3.1.1 Spin–Orbit Coupling by the Impurity

If the spin–orbit coupling comes from the impurity potential (in this case
the coupling is often termed extrinsic), the spin-flip scattering is due to that
potential.
The number of transitions per unit time from the spin-up to the spin-
down states is
W↑↓ = ∑∑W
kn k ′n ′
k ′n ′↑, kn ↓ − Wkn↓,k ′n ′↑ . (1.119)
The rate is given by the spin-flip events from down to up minus the ones
from up to down. We use the Fermi golden rule to write out the individual
scattering rates:
2π
f k′n (1 − f k′′n ) | U k n↑, k ′n ′↓ |2 δ  ε k ′n ′ − ε k n  , (1.120)
 
Wk ′n ′↑,k n↓ =

2π    
Wk n↑,k ′n ′↓ = f k ′n ′  1 − f k n  | U k ′n ′↓,k n↑ |2 δ  ε k ′n ′ − ε k n  . (1.121)

Substituting for the occupation numbers the linearized expression, Equation
1.110, and using the definition of the spin relaxation of Equation 1.117, we
obtain for the spin relaxation rate the expression
1 2π 1 2  ∂f 0 ( ε ) 

T1
=
gs ∑kk ′
U k n↑,k ′n ′↓  −
 ∂ ε k
k

 
 δ  ε k ′n ′ − ε k n  . (1.122)
If we define the spin relaxation for the individual momentum state k as
2π  ∂f 0 ( ε k )  
∑U
1 2

= k n ↑, k ′n ′↓  − ∂ε  δ  ε k ′n ′ − ε k n  , (1.123)
T1k k′  k 
which has a straightforward interpretation as the spin-flip rate by the elastic
impurity scattering to all the possible states k′, we get
1 1
= , (1.124)
T1 T1k εk = ε F
as the average of the individual scattering rates over the electronic Fermi
surface.
1.3.1.2 Spin–Orbit Coupling by the Host Lattice

If the spin–orbit coupling comes from the host lattice (in this case the cou-
pling is often termed intrinsic), the spin-flip scattering is due to the admix-
ture of the Pauli spin-up and spin-down states in the Bloch eigenstates.
How do the Bloch states actually look like in the presence of spin–orbit
coupling? Elliott showed that the Bloch states corresponding to a generic lat-
tice wave-vector k and band n can be written as [1]
Ψk ,n⇑ (r) =  akn (r) |↑ + bkn (r) |↓  e ik⋅r , (1.125)
Ψk ,n⇓ (r) =  a−∗ kn (r) |↓ − b−∗ kn (r) |↓  e ik⋅r . (1.126)
The states |↑ and |↓ are the usual Pauli spinors. We can select the two
states such that |akn| ≈ 1 while |bkn| ≪ 1, due to the weak spin–orbit coupling;
this justifies calling the two above states “spin up” (⇑) and “spin down” (⇓).
In fact, to “prepare” the states for the calculation of the spin relaxation, they
need to satisfy
k , nλ | σ z | k, nλ ′ = λδ λλ ′ , (1.127)
with λ = ⇑, ⇓. That is, the two states should diagonalize the spin matrix Sz (or
whatever spin direction one is interested in).
The Bloch states of Equations 1.125 and 1.126 allow for a spin flip even
if the impurity does not induce a spin–orbit coupling. Indeed, the matrix
element
k , n ⇑| U | k , n ⇓ ~ ab, (1.128)
is in general non-zero due to the spin admixture. The spin-flip probability is

proportional to |b|2, the spin admixture probability. This quantity is crucial
in estimating the spin relaxation in the Elliott–Yafet mechanism. The spin
relaxation time T1 in this case can be calculated using the formula Equation
1.122, with
U k n↑,k ′n ′↓ = U k n⇑,k ′n ′⇓ , (1.129)
given by Equation 1.128. A useful rule of thumb for estimating the spin relax-
ation time in this case is
1 b2
≈ kn , (1.130)
T1 τp
where the averaging of the spin admixture probabilities b2 is performed over
the Fermi surface (or the relevant energy scales of the problem); τp is the
spin-conserving momentum relaxation time. We stress that b is obtained
from the states prepared according to Equation 1.127.
1.3.2 Electron–Phonon Scattering

The spin flip due to the scattering of electrons off phonons involves the
intrinsic spin–orbit potential. The electron Bloch states are the ones given
by Equations 1.125 and 1.126, prepared according to Equation 1.127.
The net number of transitions per unit time from the spin-up to the
spin-down states is
W↑↓ = ∑∑∑W
kn k ′n ′ qν
k n ↑,qν; k ′n ′↓ + Wk n↑;k ′n ′↓,qν
(1.131)
−Wk ′n ′↓,qν;k n↑ − Wk ′n ′↓;k n↑,qν .
We introduced the rates of the spin-flip transitions accompanied by the pho-

non absorption and emissions as follows. The net transition rate from the
single electron state |k′n′↓ to the electron state |kn↑ , while the phonon of
momentum q and polarization v is emitted, is
2π 2    
Wk n↑,qν;k ′n ′↓ = Mk n↑,qν;k ′n ′↓ f k ′n ′↓  1 − f k n↑  δ  ε k n − ε k ′n ′ + ω qν  . (1.132)

Similarly, the net transition rate from the single electron state |k′n′↓ to the
electron state |kn↑ , while the phonon of momentum q and polarization v
is absorbed, is
2π 2    
Wk n↑;k ′n ′↓,qν = Mk n↑;k ′n ′↓ qν f k ′n ′↓  1 − f k n↑  δ  ε k n − ε k ′n ′ − ω qν  . (1.133)

The same way are defined the remaining two rates, Wk ′n′↓ ,qv ;k n↑ for the spin
flip from kn↑ to k′n′↓ with the phonon emission, and Wk ′n′↓ ;k n↑ qv , for the
phonon absorption:
2π 2
Wk ′n ′↓,qν; k n↑ = M k ′n ′↓,qν; k n↑ f kn↑ (1 − f k ′n ′↓ ) δ  ε k ′n ′ − ε k n + ω qν  , (1.134)

2π 2
Mk ′n ′↓,qν;k n↑,qν f k n↑ (1 − f k ′n ′↓ ) δ  ε k ′n ′ − ε k n − ω qν  . (1.135)
 
Wk ′n ′↓;k n↑ qν =

The calculation of the spin relaxation due to the electron–phonon scatter-
ing is rather involved and we cite here only the final result for degenerate
conductors:
1 4π 1  ∂f k0n  N2

T1
=
gs ∑∑∑
kn k ′n ′ ν
−
 ε  2 NM ω
kn k ′ − k ,ν
(1.136)
2
× e k − k ′, ν ⋅ kn ⇑ ÑV k ′n′ ⇓
{
× nqν − f k0′n ′ + 1 δ ( ε kn − ε k ′n ′ − ω qν )
(1.137)
}
+ nqν + f k0′n ′  δ ( ε kn − ε k ′n ′ + ω qν ) .
The electronic bands are described by the energies ε kn of the state with
momentum k and band index n. Phonon frequencies are ωqv , for the phonon
of momentum q and polarization v; similarly for the phonon polarization
vector εqv . We further denoted by M the atomic mass, by N the number
of atoms in the lattice, and by ∇V the gradient of the electron-lattice ion
potential. The equilibrium phonon occupation numbers are denoted as nqv,
given as
1
nqν = n(ω qv ) = βω qν . (1.138)
e −1
The electronic states |knσ are normalized to the whole space.
103 Spin resonance exp

Spin injection exp
Elliott–Yafet theory
102
T1 (ns)
101
100
50 100 150 200 250 300 350 400
T (K)
FIGURE 1.8 Spin relaxation in silicon. Phonon-induced spin relaxation in silicon

results in an approximate T 3 power law [15]. Shown are the experimental data from
spin resonance [16] and spin injection [17] experiments, and a calculation based on
the Elliott–Yafet mechanism [15].
Two types of processes contribute to the phonon-induced spin flips.

First, what we call the Elliott processes are the Elliott-type of spin flips in
which the Bloch states given by Equations 1.125 and 1.126 scatter by the
scalar part of the gradient of the electron–ion potential V. Second, what we
call the Yafet processes are the spin flips due to the gradient of the spin–
orbit part of the electron–ion potential. These two processes are typically
of similar order of magnitude and have to be added coherently in order to
obtain T1.
Figure 1.8 shows the experimental data of the spin relaxation in
intrinsic (non-degenerate) silicon, obtained by the spin resonance [8, 16]
and the spin injection [17] experiments. The calculation based on the
Elliott–Yafet mechanism of the phonon-induced spin flips reproduces the
experiments [15].
1.3.2.1 Yafet Relation
The expected temperature dependence of the phonon-induced spin flips in
degenerate conductors, is, according to the Elliott–Yafet mechanism, 1/T1 ~
T at high temperatures (starting roughly at a fraction of the Debye tempera-
ture TD) and 1/T1 ~ T 3 at low temperatures, in analogy with the conventional
spin-conserving electron–phonon scattering. The high-temperature depen-
dence originates from the linear increase of the phonon occupation numbers
n with increasing temperature: nqv ~ T, as kBT ≫ ωqv . At T > TD are all the
phonons excited. The low-temperature dependence of the spin-conserving
scattering follows from setting the relevant phonon energy scale to kBT. The
matrix element
kn ⇑ ÑV k ′n′ ⇑ ~ q , (1.139)
which gives the 1/T1 ~ T 3 dependence. However, Yafet showed [2] that for
the spin-flip matrix element the space inversion symmetry modifies the
momentum dependence to
kn ⇑ ÑV k ′n′ ⇓ ~ q 2 , (1.140)
so that
1
~ T 5 , (1.141)
T1
instead of the expected T 3. Since the same temperature dependence holds
for the phonon-induced electrical resistance ρ(T), we can write
1
~ ρ(T ), (1.142)
T1
known as the Yafet relation.
The relation Equation 1.140 holds if both the Elliott and Yafet pro-
cesses are taken into account. Individually, they would lead to a linear
dependence on q. The quantum mechanical interference between these
two processes thus significantly reduces the spin-flip probability at low
momenta q. An example is shown in Figure 1.9. The Elliott and Yafet
processes would individually give much stronger spin relaxation than
is observed. Their destructive interference can modify T1 by orders of
magnitude.
104
Total
103 Elliott
Yafet
102
T1 (ns)
101
100
50 100 150 200 250 300 350 400

T (K)
FIGURE 1.9 Interference between the Elliott and Yafet processes. Individually
the Elliott and Yafet phonon-induced spin relaxation processes give spin relaxation
orders of magnitude stronger than the total one. This example is from the calcula-
tion of T1 in silicon [15].
1.4 Results Based on Kinetic-Spin-Bloch-Equation Approach 31
1.4 RESULTS BASED ON KINETIC-SPIN-

BLOCH-EQUATION APPROACH
It was shown by Wu et al. from a full microscopic kinetic-spin-Bloch-equa-
tion approach [9] that the single-particle approach is inadequate in account-
ing for the spin relaxation/dephasing both in the time [18–22] and in the
space [23–25] domains. The momentum dependence of the effective mag-
netic field (the D’yakonov–Perel’ term) and the momentum dependence of
the spin diffusion rate along the spacial gradient [23] or even the random
spin–orbit coupling [26] all serve as inhomogeneous broadenings [19, 20].
It was pointed out that in the presence of inhomogeneous broadening, any
scattering, including the carrier–carrier Coulomb scattering, can cause
an irreversible spin relaxation/dephasing [19]. Moreover, besides the spin
relaxation/dephasing channel the scattering provides, it also gives rise to
the counter effect to the inhomogeneous broadening. The scattering tends
to drive carriers to a more homogeneous state and therefore suppresses the
inhomogeneous broadening. Finally, this approach is valid in both strong
and weak scattering limits and also can be used to study systems far away
from the equilibrium, thanks to the inclusion of the Coulomb scattering.
In the following subsection, we present the main results based on
kinetic-spin-Bloch-equation approach. We first briefly introduce the kinetic
spin Bloch equations. Then we review the results of the spin relaxation/
dephasing in the time and space domains, respectively.
1.4.1 Kinetic Spin Bloch Equations

By using the nonequilibrium Green function method with gradient expres-
sion as well as the generalized Kadanoff–Baym ansatz [27], we construct the
kinetic spin Bloch equations as follows:
ρ k(r, t ) = ρ k(r, t ) dr + ρ k(r, t ) dif + ρ k(r, t ) coh + ρ k(r, t ) scat. (1.143)
Here ρk(r,t) are the density matrices of electrons with momentum k at posi-
tion r and time t. The off-diagonal elements of ρk represent the correlations
between the conduction and valence bands, different subbands (in confined
structures), and different spin states. ρ k (r , t ) dr are the driving terms from
the external electric field. The coherent terms, in Equation 1.143, ρ k coh are
composed of the energy spectrum, magnetic field, and effective magnetic
field from the D’yakonov–Perel’ term and the Coulomb Hartree–Fock terms.
The diffusion terms ρ k (r , t ) dif come from the spacial gradient. The scattering
terms ρ k (r , t ) scat include the spin-flip and spin-conserving electron–electron,
electron–phonon, and electron-impurity scatterings. The spin-flip terms cor-
respond to the Elliot–Yafet and/or Bir–Aronov–Pikus mechanisms. Detailed
expressions of these terms in the kinetic spin Bloch equations depend on
the band structures, doping situations, and dimensionalities [9] and can be
found in the literature for different cases, such as intrinsic quantum wells
[18], n-type quantum wells with [21, 28–31] and without [22, 32–34] electric
field, p-type quantum wells [35–37], quantum wires [38, 39], quantum dots
[40], and bulk materials [41] in the spacial uniform case and quantum wells
in spacial nonuniform case [23, 24, 42, 43]. By numerically solving the kinetic
spin Bloch equations with all the scattering explicitly included, one is able to
obtain the time evolution and/or spacial distribution of the density matrices,
and hence all the measurable quantities, such as mobility, diffusion constant,
optical relaxation/dephasing time, spin relaxation/dephasing time, spin dif-
fusion length, as well as hot-electron temperature, can be determined from
the theory without any fitting parameters.
pin Relaxation/Dephasing
1.4.2 S
In this subsection we present the main understandings of the spin relax-
ation/dephasing added to the literature from the kinetic-spin-Bloch-
equation approach. We focus on three related issues: (i) the importance of
the Coulomb interaction to the spin relaxation/dephasing; (ii) spin dynamics
far away from the equilibrium; and (iii) qualitatively different behaviors from
those wildly adopted in the literature.
First we address the effect of the Coulomb interaction. Based on the
single-particle approach, it has been long believed that the Coulomb scat-
tering is irrelevant to the spin relaxation/dephasing [44]. It was first pointed
out by Wu and Ning [19] that in the presence of inhomogeneous broadening
in spin precession, that is, the spin precession frequencies are k-dependent,
any scattering, including the spin-conserving scattering, can cause irrevers-
ible spin dephasing. This inhomogeneous broadening can come from the
energy-dependent g-factor [19], the D’yakonov–Perel’ term [20], the random
spin–orbit coupling [26], and even the momentum dependence of the spin
diffusion rate along the spacial gradient [23]. Wu and Ning first showed that
with the energy-dependent g-factor as an inhomogeneous broadening, the
Coulomb scattering can lead to irreversible spin dephasing [19]. In [001]-
grown n-doped quantum wells, the importance of the Coulomb scattering
for spin relaxation/dephasing was proved by Glazov and Ivchenko [45] by
using perturbation theory and by Weng and Wu [21] from the kinetic-spin-
Bloch-equation approach. In a temperature-dependent study of the spin
dephasing in [001]-oriented n-doped quantum wells, Leyland et al. experi-
mentally verified the effects of the electron–electron Coulomb scattering by
closely measuring the momentum scattering rate from the mobility [46]. By
showing the momentum relaxation rate obtained from the mobility cannot
give the correct spin relaxation rate, they showed the difference comes from
the Coulomb scattering. Later Zhou et al. even predicted a peak from the
Coulomb scattering in the temperature dependence of the spin relaxation
time in a high-mobility low-density n-doped (001) quantum well [29]. This
was later demonstrated by Ruan et al. experimentally [47].
Figure 1.10 shows the temperature dependence of the spin relaxation
time of a 7.5 nm GaAs/Al0.4Ga0.6As quantum well at different electron
and impurity densities [29]. For this small well width, only the lowest sub-
band is needed in the calculation. It is shown in the figure that when the
n = 4 × 1010/cm2
τ (ps)
102
101
0 50 100 150 200 250 300
(a) T (K)
n = 1 × 1111/cm2
102
τ (ps)
101
0 50 100 150 200 250 300
(b) T (K)
102
τ (ps)
n = 2 × 1011 cm2
101
0 50 100 150 200 250 300

(c) T (K)
FIGURE 1.10 Spin relaxation time τ versus the temperature T with well width
a = 7.5 nm and electron density n being (a) 4 × 1010 cm−2, (b) 1 × 1011 cm−2, and (c)
2 × 1011 cm−2, respectively. Solid curves with triangles: impurity density ni = n; solid
curves with dots: ni = 0.1n; solid curves with circles: ni = 0; dashed curves with dots:
ni = 0.1n and no Coulomb scattering. (After Zhou, J. et al., Phys. Rev. B 75, 045305,
2007. With permission.)
electron-impurity scattering is dominant, the spin relaxation time decreases

with increasing temperature monotonically. This is in good agreement with
the experimental findings [48] and a nice agreement of the theory and the
experimental data from 20 to 300 K is given in Ref. [29]. However, it is shown
that for sample with high mobility, that is, low-impurity density, when the
electron density is low enough, there is a peak at low temperature. This peak,
located around the Fermi temperature of electrons TFe = E F / k B , is identified
to be solely due to the Coulomb scattering [29, 49]. It disappears when the
Coulomb scattering is switched off, as shown by the dashed curves in the
figure. This peak also disappears at high impurity densities. It is also noted in
Figure 1.10c that for electrons of high density so that TFe is high enough and
the contribution from the electron–longitudinal optical-phonon scattering
becomes marked, the peak disappears even for sample with no impurity
and the spin relaxation time increases with temperature monotonically. The
physics leading to the peak is due to the crossover of the Coulomb scattering
from the degenerate to the non-degenerate limit. At T < TFe , electrons are in
the degenerate limit and the electron–electron scattering rate 1/τee ∝ T 2. At
T > TFe , 1/τee ∝ T−1 [45, 50]. Therefore, at low electron density so that TFe is low
enough and the electron-acoustic phonon scattering is very weak compared
with the electron–electron Coulomb scattering, the Coulomb scattering is
dominant for high mobility sample. Hence, the different temperature depen-
dence of the Coulomb scattering leads to the peak. It is noted that the peak
is just a feature of the crossover from the degenerate to the non-degenerate
limit. The location of the peak also depends on the strength of the inho-
mogeneous broadening. When the inhomogeneous broadening depends on
momentum linearly, the peak tends to appear at the Fermi temperature. A
similar peak was predicted in the electron spin relaxation in p-type GaAs
quantum well and the hole spin relaxation in (001) strained asymmetric Si/
SiGe quantum well, where the electron and hole spin relaxation times both
show a peak at the hole Fermi temperature TFh [36, 37]. When the inhomo-
geneous broadening depends on momentum cubically, the peak tends to
shift to a lower temperature. It was predicted that a peak in the temperature
dependence of the electron spin relaxation time appears at a temperature in
the range of (TFe / 4, TFe / 2) in the intrinsic bulk GaAs [41] and a peak in the
temperature dependence of the hole spin relaxation time at TFh / 2 in p-type
Ge/SiGe quantum well [37]. Ruan et al. demonstrated the peak experimen-
tally in a high-mobility low-density GaAs/Al0.35Ga0.65As heterostructure and
showed a peak appears at TFe / 2 in the spin relaxation time versus tempera-
ture curve [47].
For larger well width, the situation may become different in the non-
degenerate limit. Weng and Wu calculated the spin relaxation/dephasing for
(001) GaAs quantum wells with larger well width and high mobility by includ-
ing the multi-subband effect [28]. It is shown that for small/large well width
so that the linear/cubic Dresselhaus term is dominant, the spin relaxation/
dephasing time increases/decreases with the temperature. This is because
with the increase of temperature, both the inhomogeneous broadening and
the scattering get enhanced. The relative importance of these two compet-
ing effects is different when the linear/cubic term is dominant [28]. Jiang
and Wu further introduced strain to change the relative importance of the
linear and cubic D’yakonov–Perel’ terms and showed the different tempera-
ture dependences of the spin relaxation time [51]. This prediction has been
realized experimentally by Holleitner et al. where they showed that in n-type
two-dimensional InGaAs channels, when the linear D’yakonov–Perel’ term
is suppressed, the spin relaxation time decreases with temperature mono-
tonically [52]. Another interesting prediction related to the multi-subband
effect is related to the effect of the inter-subband Coulomb scattering. From
the calculation Weng and Wu found out that although the inhomogeneous
broadening from the higher subband of the (001) quantum well is much
larger, due to the strong inter-subband Coulomb scattering, the spin relax-
ation times of the lowest two subbands are identical [28]. This prediction
has later been verified experimentally by Zhang et al., who studied the spin
dynamics in a single-barrier heterostructure by time-resolved Kerr rotation
[53]. By applying a gate voltage, they effectively manipulated the confine-
ment of the second subband, and the measured spin relaxation times of the
first and second subbands are almost identical at large gate voltage. Lü et al.
showed that due to the Coulomb scattering, T2 = T2∗ in (001) GaAs quantum
wells for a wide temperature and density regime [54]. It was also pointed
out by Lü et al. that in the strong (weak) scattering limit, introducing the
Coulomb scattering will always lead to a faster (slower) spin relaxation/
dephasing [35].
Another important effect from the Coulomb interaction to the spin
relaxation/dephasing comes from the Coulomb Hartree–Fock contribution
in the coherent terms of the kinetic spin Bloch equations. Weng and Wu
[21] first pointed out that at a high spin polarization, the Hartree–Fock term
serves as an effective magnetic field along the z-axis, which blocks the spin
precession. As a result, the spin relaxation/dephasing time increases dramat-
ically with the spin polarization. They further pointed out that the spin relax-
ation/dephasing time decreases with temperature at high spin polarization
in quantum well with small well width, which is in contrast to the situation
with small spin polarizations. These predictions have been verified experi-
mentally by Stich et al. in an n-type (001) GaAs quantum well with high
mobility [55, 56]. By changing the intensity of the circularly polarized lasers,
Stich et al. measured the spin dephasing time in a high mobility n-type GaAs
quantum well as a function of initial spin polarization. Indeed they observed
an increase of the spin dephasing time with the increased spin polarization,
and the theoretical calculation based on the kinetic spin Bloch equations
nicely reproduced the experimental findings when the Hartree–Fock term
was included [55]. It was also shown that when the Hartree–Fock term is
removed, one does not see any increase of the spin dephasing time. Later,
they further improved the experiment by replacing the circular-polarized
laser pumping with the elliptic polarized laser pumping. By doing so, they
were able to vary the spin polarization without changing the carrier density.
Figure 1.11 shows the measured spin dephasing times as function of initial
130
Experiment 340 Experiment
Full calculation Full calculation
Calculation 320 Calculation
Spin dephasing time (ps)

120 without HF term without HF term
300
280
110
260
240
100
220
200
90
180
0 5 10 15 0 5 10 15 20 25
(a) Initial spin polarization (%) (b) Initial spin polarization (%)
FIGURE 1.11 (a) The spin dephasing times as a function of initial spin polarization for
constant, low excitation density and variable polarization degree of the pump beam.
The measured spin dephasing times are compared to calculations with and without
the Hartree–Fock (HF) term, showing its importance. (b) The spin dephasing times
measured and calculated for constant, high excitation density and variable polarization
degree. (After Stich, J. et al., Phys. Rev. B 76, 205301, 2007. With permission.)
spin polarization under two fixed pumping intensities, together with the
theoretical calculations with and without the Coulomb Hartree–Fock term.
Again the spin dephasing time increases with the initial spin polarization as
predicted and the theoretical calculations with the Hartree–Fock term are
in good agreement with the experimental data [56]. Moreover, Stich et al.
also confirmed the prediction of the temperature dependences of the spin
dephasing time at low and high spin polarizations [56]. Figure 1.12a shows
350
50
300
Electron temperature (K)
250 40
200
30
150
P = 0.7% P = 0.7%
100 P = 2.9% 20
P = 2.9%
P = 11.9% P = 11.9%
50 P = 27%
10 P = 27%
0
0 5 10 15 20 25 30 35 40 45 50 55 0 10 20 30 40 50
(a) Sample temperature T (K) (b) Nominal sample temperature (K)
FIGURE 1.12 (a) Spin dephasing time as a function of sample temperature, for different initial spin polariza-
tions. The measured data points are represented by solid points, while the calculated data are represented
by lines of the same grayscale. (b) Electron temperature determined from intensity-dependent photoillumi-
nance measurements as a function of the nominal sample temperature, for different pump beam fluence and
initial spin polarization, under experimental conditions corresponding to the measurements shown in (a). The
measured data points are represented by solid points, while the curves serve as guide to the eye. (After Stich, J.
et al., Phys. Rev. B 76, 205301, 2007. With permission.)
the measured temperature dependences of the spin dephasing time at differ-

ent initial spin polarizations. As predicted, the spin dephasing time increases
with increasing temperature at small spin polarization but decreases at large
spin polarization. The theoretical calculations also nicely reproduced the
experimental data. The hot-electron temperatures in the calculation were
taken from the experiment (Figure 1.12b). The effective magnetic field from
the Hartree–Fock term has been measured by Zhang et al. from the sign
switch of the Kerr signal and the phase reversal of Larmor precessions with
a bias voltage in a GaAs heterostructure [57]. Korn et al. [58] also estimated
the average effect by applying an external magnetic field in the Faraday con-
figuration, as shown in Figure 1.13a, for the same sample reported earlier
[55, 56]. They compared the spin dephasing times of both large and small
spin polarizations as function of external magnetic field. Due to the effec-
tive magnetic field from the Hartree–Fock term, the spin relaxation times
are different under small external magnetic field but become identical when
the magnetic field becomes large enough. From the merging point, they esti-
mated that the mean value of the effective magnetic field is below 0.4 T. They
further showed that this effective magnetic field from the Hartree–Fock
term cannot be compensated by the external magnetic field, because it does
not break the time-reversal symmetry and is therefore not a genuine mag-
netic field, as said earlier. This can be seen from Figure 1.13b that the spin
relaxation time at large spin polarization shows identical external magnetic
field dependences when the magnetic field is parallel or antiparallel to the
growth direction.
We now turn to discuss the spin relaxation/dephasing far away from the
equilibrium. In fact, the spin relaxation/dephasing of high spin polarization
addressed earlier is one of the cases far away from the equilibrium. Another
case is the spin dynamics in the presence of a high in-plane electric field. The
spin dynamics in the presence of a high in-plane electric field was first stud-
ied by Weng et al. [22] in GaAs quantum well with only the lowest subband
by solving the kinetic spin Bloch equations. To avoid the “runaway” effect
350
300
Spin relaxation time (ps)
Spin relaxation time (ps)
300 Small initial P

250 Large initial P
250
200
150 200
100 Anti-
150 parallel Parallel
50
0.0 0.1 0.2 0.3 0.4 0.5 –0.5 0.0 0.5
(a) Magnetic field Bext (T) (b) Magnetic field Bext (T)
FIGURE 1.13 (a) Spin dephasing times as a function of an external magnetic field perpendicular to the quantum
well plane for small and large initial spin polarization. (b) Same as (a) for large initial spin polarization and both
polarities of the external magnetic field. (After Korn, T. et al., Adv. Solid State Phys. 48, 143, 2009. With permission.)
[59], the electric field was calculated up to 1 kV/cm. Then, Weng and Wu fur-
ther introduced the second subband into the model and the in-plane electric
field was increased up to 3 kV/cm [28]. Zhang et al. included L valley and the
electric field was further increased up to 7 kV/cm [32]. The effect of in-plane
electric field to the spin relaxation in system with strain was investigated by
Jiang and Wu [51]. Zhou et al. also investigated the electric-field effect at low
lattice temperatures [29].
The in-plane electric field leads to two effects: (a) It shifts the center
of mass of electrons to kd = m*vd = m*μE with μ representing the mobility,
which further induces an effective magnetic field via the D’yakonov–Perel’
term [22]; (b) the in-plane electric field also leads to the hot-electron effect
[60]. The first effect induces a spin precession even in the absence of any
external magnetic field and the spin precession frequency changes with the
direction of the electric field in the presence of an external magnetic field
[22, 32]. The second effect enhances both the inhomogeneous broaden-
ing and the scattering, two competing effects leading to rich electric-field
dependence of the spin relaxation/dephasing and thus spin manipulation
[22, 28, 29, 32, 33, 41, 51].
Finally we address some issues of which the kinetic-spin-Bloch-equation
approach gives qualitatively different predictions from those widely used in
the literature. These issues include the Bir–Aronov–Pikus mechanism, the
Elliot–Yafet mechanism, and some density/temperature dependences of the
spin relaxation/dephasing time.
It has long been believed in the literature that for electron relaxation/
dephasing, the Bir–Aronov–Pikus mechanism is dominant at low tempera-
ture in p-type samples and has important contribution to intrinsic sample
with high photo-excitation [61–67]. This conclusion was made based on the
single-particle Fermi golden rule. Zhou and Wu reexamined the problem
using the kinetic-spin-Bloch-equation approach [30]. They pointed out that
the Pauli blocking was overlooked in the Fermi golden rule approach. When
electrons are in the non-degenerate limit, the results calculated from the
Fermi golden rule approach are valid. However, at low temperature, elec-
trons can be degenerate and the Pauli blocking becomes very important.
As a result, the previous approaches always overestimated the importance
of the Bir–Aronov–Pikus mechanism at low temperature. Moreover, the
previous single-particle theories underestimated the contribution of the
D’yakonov–Perel’ mechanism by neglecting the Coulomb scattering. Both
made the Bir–Aronov–Pikus mechanism dominate the spin relaxation/
dephasing at low temperature. Later, Zhou et al. performed a thorough
investigation of electron spin relaxation in p-type (001) GaAs quantum
wells by varying impurity, hole and photoexcited electron densities over a
wide range of values [36], under the idea that very high impurity density
and very low photoexcited electron density may effectively suppress the
importance of the D’yakonov–Perel’ mechanism and the Pauli blocking.
Then the relative importance of the Bir–Aronov–Pikus and D’yakonov–
Perel’ mechanisms may be reversed. This indeed happens as shown in the
phase-diagram-like picture in Figure 1.14 where the relative importance
τBAP/τDP
300 102
100
101
10
(a) (b)
T (K)
300 100
100
10–1
10
(c) (d)
1 10 1 10 10–2
Nh (1011 cm–2)
FIGURE 1.14 Ratio of the spin relaxation time due to the Bir–Aronov–Pikus
mechanism to that due to the D’yakonov–Perel’ mechanism, τBAP/τDP, as function of
temperature and hole density with (a) NI = 0, Nex = 1011 cm−2; (b) NI = 0, Nex = 109 cm−2;
(c) NI = Nh, Nex = 1011 cm−2; (d) NI = Nh, Nex = 109 cm−2. The black dashed curves indicate
the cases satisfying τBAP/τDP = 1. Note the smaller the ratio τBAP/τDP is, the more
important the Bir–Aronov–Pikus mechanism becomes. The yellow solid curves
indicate the cases satisfying ∂ µh [NLH(1) + NHH( 2 ) ]/∂ µh Nh = 0.1. In the regime above the
yellow curve the multi-hole-subband effect becomes significant. (After Zhou, Y.
et al., New J. Phys. 11, 113039, 2009. With permission.)
of the Bir–Aronov–Pikus and D’yakonov–Perel’ mechanisms is plotted as

function of hole density and temperature at low and high impurity densi-
ties and photo-excitation densities. For the situation of high hole density,
they even included multi-hole subbands as well as the light hole band. It
is interesting to see from the figures that at relatively high photo-excita-
tions, the Bir–Aronov–Pikus mechanism becomes more important than
the D’yakonov–Perel’ mechanism only at high hole densities and high tem-
peratures (around hole Fermi temperature) when the impurity is very low
(zero in Figure 1.14a). Impurities can suppress the D’yakonov–Perel’ mecha-
nism and hence enhance the relative importance of the Bir–Aronov–Pikus
mechanism. As a result, the temperature regime is extended, ranging from
the hole Fermi temperature to the electron Fermi temperature for high hole
density. When the photo-excitation is weak so that the Pauli blocking is less
important, the temperature regime where the Bir–Aronov–Pikus mecha-
nism is important becomes wider compared to the high excitation case. In
particular, if the impurity density is high enough and the photo-excitation
is so low that the electron Fermi temperature is below the lowest tempera-
ture of the investigation, the Bir–Aronov–Pikus mechanism can domi-
nate the whole temperature regime of the investigation at sufficiently high
hole density, as shown in Figure 1.14d. The corresponding spin relaxation
times of each mechanism under high or low-impurity and photo-excitation

densities are demonstrated in Figure 1.14. They also discussed the density
dependences of spin relaxation with some intriguing properties related to
the high hole subbands [36]. The predicted Pauli-blocking effect in the Bir–
Aronov–Pikus mechanism has been partially demonstrated experimentally
by Yang et al. [68]. They showed by increasing the pumping density that the
temperature dependence of the spin dephasing time deviates from the one
from the Bir–Aronov–Pikus mechanism and the peaks at high excitations
agree well with those predicted by Zhou and Wu [30].
Another widely accepted but incorrect conclusion is related to the
Elliot–Yafet mechanism. It is widely accepted in the literature that the
Elliot–Yafet mechanism dominates spin relaxation in n-type bulk III–V
semiconductor at low temperature, while the D’yakonov–Perel’ mechanism
is important at high temperature [5, 7, 8, 69, 70]. Jiang and Wu pointed out
that the previous understanding is based on the formula that can only be
used in the non-degenerate limit. Moreover, the momentum relaxation rates
are calculated via the approximated formula for mobility [69]. By performing
an accurate calculation via the kinetic-spin-Bloch-equation approach, they
showed that the Elliot–Yafet mechanism is not important in III–V semi-
conductors, including even the narrow-band InAs and InSb [41]. Therefore,
the D’yakonov–Perel’ mechanism is the only spin relaxation mechanism for
n-type III–V semiconductors in metallic regime.
Jiang and Wu have further predicted a peak in the density dependence
of the spin relaxation/dephasing time in n-type III–V semiconductors
where the spin relaxation/dephasing is limited by the D’yakonov–Perel’
mechanism [41]. Previously, the non-monotonic density dependence of
spin lifetime was observed in low-temperature (T ≲ 5 K) experiments,
where the localized electrons play a crucial role and the electron system
is in the insulating regime or around the metal-insulator transition point
[71]. Jiang and Wu found, for the first time, that the spin lifetime in metal-
lic regime is also non-monotonic. Moreover, they pointed out that it is a
universal behavior for all bulk III–V semiconductors at all temperatures
where the peak is located at TF ~ T with TF being the electron Fermi temper-
ature. The underlying physics for the non-monotonic density dependence
in metallic regime can be understood as following: In the non-degenerate
regime, as the distribution is the Boltzmann one, the density dependence
of the inhomogeneous broadening is marginal. However, the scattering
increases with the density. Consequently, the spin relaxation/dephasing
time increases with the density. However, in the degenerate regime, due
to the Fermi distribution, the inhomogeneous broadening increases with
the density much faster than the scattering does. As a result, the spin
relaxation/dephasing time decreases with the density. Similar behavior
was also found in two-dimensional system [31, 37] where the underlying
physics is similar. The predicted peak was later observed by Krauß et al.
[72], as shown in Figure 1.15 where theoretical calculation based on the
kinetic spin Bloch equations nicely reproduced the experimental data by
Shen [73].
140
Expt
Te =Teff 120
103 Te =Tlat
× 0.2 100
Teff (K)
τ (ps)
80
60
40
102 Nex = 1.4 × 1016 cm–3 20
1014 1015 1016 1017 1018

Doping density (cm−3)
FIGURE 1.15 Electron spin relaxation times from the calculation via the kinetic-spin-
Bloch-equation approach (red solid curve) and from the experiment [72] (blue •) as
function of the doping density. The green dashed curve shows the results without
hot-electron effect. The hot-electron temperature used in the computation is plotted
as the blue dotted curve. (Note the scale is on the right-hand side of the frame.) The
chain curve is the calculated spin relaxation time with a fixed hot-electron tempera-
ture 80 K, and Nex = 6 × 1015 cm−3. (After Shen, K., Chin. Phys. Lett. 26, 067201, 2009. With
permission.)
1.4.3 Spin Diffusion/Transport

By solving the kinetic spin Bloch equations together with the Poisson equa-
tion self-consistently, one is able to obtain all the transport properties such
as the mobility, charge diffusion length, and spin diffusion/injection length
without any fitting parameter. It was first pointed out by Weng and Wu [23]
that the drift-diffusion equation approach is inadequate in accounting for
the spin diffusion/transport. It is important to include the off-diagonal term
between opposite spin bands ρk↑↓ in studying the spin diffusion/transport.
With this term, electron spin precesses along the diffusion and therefore
k·∇rρk(r,t) in the diffusion terms ρ k (r , t ) dif provides an additional inhomo-
geneous broadening. With this additional inhomogeneous broadening, any
scattering, including the Coulomb scattering, can cause an irreversible spin
relaxation/dephasing [23]. Unlike the spin precession in the time domain
where the inhomogeneous broadening is determined by the effective mag-
netic field from the D’yakonov–Perel’ term, h(k), in spin diffusion and trans-
port it is determined by
| g µ BB + h(k ) |
Wk = , (1.144)
kx
provided the diffusion is along the x-axis [42]. Here, the magnetic field is in
the Voigt configuration. Therefore, even in the absence of the D’yakonov–
Perel’ term h(k), the magnetic field alone can provide an inhomogeneous
broadening and leads to the spin relaxation/dephasing in spin diffusion and
transport. This was first pointed out by Weng and Wu back to 2002 [23] and
has been realized experimentally by Appelbaum et al. in bulk silicon [74, 75],
where there is no D’yakonov–Perel’ spin–orbit coupling due to the center
inversion symmetry. Zhang and Wu further investigated the spin diffusion
and transport in symmetric Si/SiGe quantum wells [43].
When B = 0 but the D’yakonov–Perel’ term is present, then the inho-
mogeneous broadening for spin diffusion and transport is determined by
Ωk = h(k)/kx. In (001) GaAs quantum well where the D’yakonov–Perel’
term is determined by the Dresselhaus term [76], the average of Ωk reads
( )
Wk = C k y − k z2 , 0, 0 with C being a constant. For electrons in quan-
2
tum well, this value is not zero. Therefore, the spacial spin oscillation due
to the Dresselhaus effective magnetic field survives even at high tempera-
ture when the scattering is strong. This effect was first predicted by Weng
and Wu by showing that a spin pulse can oscillate along the diffusion in
the absence of the magnetic field at very high temperature [24]. Detailed
studies were carried out later on this effect [25, 42, 77]. The spin oscillation
without any applied magnetic field in the transient spin transport was later
observed experimentally by Crooker and Smith in strained bulk system [78].
Differing from the two-dimensional case, in bulk, the average of Ωk from
( )
the Dresselhaus term is zero, since Wk = C k y2 − k z2 , 0, 0 = 0 due to the
symmetry in the y- and z-directions. This is consistent with the experimental
result that there is no spin oscillation for the system without stress. However,
when the stress is applied, an additional spin–orbit coupling, namely, the
coupling of electron spins to the strain tensor, appears, which is linear in
momentum [5]. This additional spin–orbit coupling also acts as an effective
magnetic field. Therefore, once the stress is applied, one can observe spacial
spin oscillation even when there is no applied magnetic field [78].
Cheng and Wu further developed a new numerical scheme to calculate
the spin diffusion/transport in GaAs quantum wells with very high accuracy
and speed [42]. It was discovered that due to the scattering, especially the
Coulomb scattering, T2 = T2* is valid even in the spacial domain. This predic-
tion remains yet to be verified experimentally. Moreover, as the inhomoge-
neous broadening in spin diffusion is determined by |h(k)|/kx in the absence of
magnetic field, the period of the spin oscillations along the x-axis is indepen-
dent of the electric field perpendicular to the growth direction of the quantum
well [42], which is different from the spin precession rate in the time domain
[22]. This is consistent with the experimental findings by Beck et al. [79].
Cheng et al. applied the kinetic-spin-Bloch-equation approach to study
the spin transport in the presence of competing Dresselhaus and Rashba
fields [80]. When the Dresselhaus and Bychkov–Rashba [10] terms are both
important in semiconductor quantum well, the total effective magnetic field
can be highly anisotropic and spin dynamics is also highly anisotropic in
regard to the spin polarization [81]. For some special polarization direc-
tion, the spin relaxation time is extremely large [81–84]. For example, if
the coefficients of the linear Dresselhaus and Bychkov–Rashba terms are
equal to each other in (001) quantum well of small well width and the cubic
Dresselhaus term is not important, the effective magnetic field is along the
[110] direction for all electrons. For the spin components perpendicular to
the [110] direction, this effective magnetic field flips the spin and leads to a
finite spin relaxation/dephasing time. For spin along the [110] direction, this
effective magnetic field cannot flip it. Therefore, when the spin polarization
is along the [110] direction, the Dresselhaus and Bychkov–Rashba terms can-
not cause any spin relaxation/dephasing. When the cubic Dresselhaus term
is taken into account, the spin dephasing time for spin polarization along the
[110] direction is finite but still much larger than other directions [85]. The
anisotropy in the spin direction is also expected in spin diffusion and trans-
port. When the Dresselhaus and Bychkov–Rashba terms are comparable,
the spin injection length Ld for the spin polarization perpendicular to [110]
direction is usually much shorter than that for the spin polarization along
[110] direction. In the ideal case when there are only the linear Dresselhaus
and Bychkov–Rashba terms with identical strengths, spin injection length
for spin polarization parallel to the [110] direction becomes infinity [82,
83]. This effect has promoted Schliemann et al. to propose the nonballistic
spin-field-effect transistor [83]. In such a transistor, a gate voltage is used to
tune the strength of the Bychkov–Rashba term and therefore control the
spin injection length. However, Cheng et al. pointed out that spin diffusion
and transport actually involve both the spin polarization and spin trans-
port directions [80]. The latter has long been overlooked in the literature.
In the kinetic-spin-Bloch-equation approach, this direction corresponds to
the spacial gradient in the diffusion term [ρ k (r , t ) dif ] and the electric field in
the drifting term [ρ k (r , t ) dr ]. The importance of the spin transport direction
has not been realized until Cheng et al. pointed out that the spin transport
is highly anisotropic not only in the sense of the spin polarization direction
but also in the spin transport direction when the Dresselhaus and Bychkov–
Rashba effective magnetic fields are comparable [80]. They even predicted
that in (001) GaAs quantum well with identical linear Dresselhaus and
Bychkov–Rashba coupling strengths, the spin injection along [110] or [110]*
can be infinite regardless of the direction of the spin polarization. This can be
easily seen from the inhomogeneous broadening Equation 1.144, which well
defines the spin diffusion/transport properties. For the spin diffusion/trans-
port in a (001) GaAs quantum well with identical Dresselhaus and Bychkov–
Rashba strengths (the schematic is shown in Figure 1.16 with the transport
direction chosen along the x-axis), the inhomogeneous broadening is given
by [80]
   π  π  ky 
W k = 2β  sin  θ −  + cos  θ −   n∧ 0
   4   4  kx 
(1.145)
 k x2 − k y2   ky  
+γ  sin 2θ + k x k y cos 2θ  , − 1, 0  ,
 2   kx  
* [10] or [110] depends on the relative signs of the Dresselhaus and Rashba coupling strengths.
(001)
– z (010)
(110)
(110)
(100)
θ
x
Injection direction
FIGURE 1.16 Schematic of the different directions considered for the spin polar-
izations [(110), (110), and (001)-axes] and spin -diffusion/injection (x-axis). (After
Cheng, J.L. et al., Phys. Rev. B 75, 205328, 2007. With permission.)
with θ being the angle between the spin transport direction (x-axis) and [001]
crystal direction. It can be split into two parts: the zeroth-order term (on k),
which is always along the same direction of n̂ 0, and the second-order term,
which comes from the cubic Dresselhaus term. If the cubic Dresselhaus term
is omitted, the effective magnetic fields for all k states align along n̂ 0 (crystal
[110]) direction. Therefore, if the spin polarization is along n̂ 0, there is no
spin relaxation even in the presence of scattering since there is no spin pre-
cession. Nevertheless, it is interesting to see from Equation 1.145 that when
θ = 3π/4, that is, the spin transport is along the [110] direction, Wk = 2m*βn∧ 0
is independent of k if the cubic Dresselhaus term is neglected. Therefore, in
this special spin transport direction, there is no inhomogeneous broadening
in the spin transport for any spin polarization. The spin injection length is
therefore infinite regardless of the direction of spin polarization. This result
is highly counterintuitive, considering that the spin relaxation times for the
spin components perpendicular to the effective magnetic field are finite in
the spacial uniform system. The surprisingly contradictory results, that is,
the finite spin relaxation/dephasing time versus the infinite spin injection
length, are due to the difference in the inhomogeneous broadening in spacial
uniform and nonuniform systems. For genuine situation, due to the pres-
ence of the cubic term, the spin injection length is still finite and the maxi-
mum spin injection length does not happen at the identical Dresselhaus and
Bychkov–Rashba coupling strengths, but shifted by a small amount due to
the cubic term [80]. However, there is strong anisotropy in regard to the
spin polarization and spin injection direction, as shown in Figure 1.17. This
predication has not yet been realized experimentally. However, very recent
experimental findings on spin helix [13, 14] have provided strong evidence to
support this predication [86].
Now we turn to the problem of spin grating. Transient spin grating,
whose spin polarization varies periodically in real space, is excited optically
by two non-collinear coherent light beams with orthogonal linear polariza-
tion [14, 87–89]. Transient spin grating technique can be used to study the
spin transport since it can directly probe the decay rate of nonuniform spin
2.5
2.5
2.0
n̂0
ẑ 2
n̂1
L0−1 (µm−1)
1.5
Ld (µm)
1.5
1.0
1
0.5 0.5
0.0 0
0 0.2 0.4 0.6 0.8 1
θ (π)
FIGURE 1.17 Spin diffusion length Ld (solid curves) and the inverse of the spin
oscillation period L−01 (dashed curves) for identical Dresselhaus and Rashba coupling
strengths as functions of the injection direction for different spin polarization
directions n̂ 0, ẑ and n̂1 (nˆ 1 = zˆ × nˆ 0 , i.e., crystal direction [110]) at T = 200 K. It is
noted that the scale of the spin oscillation period is on the right-hand side of the
frame. (After Cheng, J.L. et al., Phys. Rev. B 75, 205328, 2007. With permission.)
distributions. Spin diffusion coefficient Ds can be obtained from the transient

spin grating experiments [87–89]. In the literature, the drift-diffusion model
was employed to extract Ds from the experimental data. With the drift-dif-
fusion model, the transient spin grating was predicted to decay exponentially
with time with a decay rate of Γq = Dsq2 + 1/τs, where q is the wave-vector of
the spin grating and τs is the spin relaxation time [87, 88]. However, this
result is not accurate since it neglects the spin precession, which plays an
important role in spin transport, as first pointed out by Weng and Wu [22].
Indeed, experimental results show that the decay of transient spin grating
takes a double-exponential form instead of single exponential one [14, 87,
89]. Also the relation that relates the spin injection length with the spin dif-
fusion coefficient Ds and the spin relaxation time τs , i.e., Ls = 2 Ds τ s from
the drift-diffusion model should be checked. In fact, if this relation is correct,
the above prediction of infinite spin injection length at certain spin injection
direction for any spin polarization in the presence of identical Dresselhaus
and Bychkov–Rashba coupling strengths cannot be correct.
Weng et al. studied this problem from the kinetic-spin-Bloch-equation
approach [90]. By first solving the kinetic spin Bloch equations analytically
by including only the elastic scattering, that is, the electron-impurity scat-
tering, they showed that the transient spin grating should decay double
exponentially with two decay rates Γ±. In fact, none of the rates is quadratic
in q. However, the average of them reads [90]
(Γ + + Γ − ) Dq 2 + 1
Γ= = (1.146)
2 τ ′s
with 1/τ′s = (1/τs+1/τs1)/2, which differs from the current widely used formula
by replacing the spin decay rate by the average of the in- and out-of-plane
relaxation rates. The difference of these two decay rates is a linear function
of the wave-vector q when q is relatively large:
∆Γ = cq + d , (1.147)
with c and d being two constants. The steady-state spin injection length Ls
and spin precession period L0 are then [90]
2 Ds
Ls = , (1.148)
| c − 4 Ds (1 / τ ′s − d ) |
2
2 Ds
L0 = . (1.149)
c
They further showed that the above relation Equations 1.146 and 1.147 are
valid even including the inelastic electron–electron and electron-phonon
scatterings by solving the full kinetic spin Bloch equations, as shown in
Figure 1.18. A good agreement with the experimental data [89] of double
exponential decays τ ± = Γ ±−1 is shown in Figure 1.19. Finally it was shown
that the infinite spin injection length predicted by Cheng et al. in the pres-
ence of identical Dresselhaus and Bychkov–Rashba coupling strengths [80]
addressed earlier can exactly be obtained from Equation 1.148 as in that spe-
cial case τ′s = τ s , d = 0, and c = 2 Ds / τ s [90]. However, Ds τ s always remains
0.4 0.075
0.3
0.05
∆Γ (ps−1)
Γ (ps−1)
0.2
0.025
0.1
0 0
0 1 2 3
q (10 cm–1)
4
FIGURE 1.18 Γ = (Γ+ + Γ−)/2 and ΔΓ = (Γ+ + Γ−)/2 versus q at T = 295 K. Open

boxes/triangles are the relaxation rates (Γ+/− calculated from the full kinetic spin
Bloch equations. Filled/open circles represent Γ and ΔΓ respectively. Noted that
the scale for ΔΓ is on the right-hand side of the frame. The solid curves are the
fitting to Γ and ΔΓ, respectively. The dashed curves are guide to eyes. (After Weng,
M.Q. et al., J. Appl. Phys. 103, 063714, 2008. With permission.)
τ−
103
102 τ−
τ (ps)
τ (ps)
τ+
102
101
τ+
100 150 200 250 100 150 200 250 300
(a) T(K) (b) T(K)
FIGURE 1.19 Spin relaxation times τ± versus temperature for (a) high-mobility
sample with q = 0.58 × 104 cm−1 and (b) low-mobility sample with q = 0.69 × 104 cm−1.
The dots are the experiment data from Ref. [90]. (After Weng, M.Q. et al., J. Appl.
Phys. 103, 063714, 2008. With permission.)
finite unless τs = ∞. Therefore, Equations 1.146 through 1.149 give the correct
way to extract the spin injection length from the spin grating measurement.
1.5 SPIN RELAXATION IN GRAPHENE

Spin relaxation in graphene has a convoluted history with many ends still
open [91]. Based on the mechanisms of spin relaxation described earlier
in this chapter, our first guess would be that electrons in graphene lose
their spins at scattering events with impurities and phonons. The spin-flip
would be facilitated by spin-orbit coupling, native in graphene. This is the
Elliott–Yafet mechanism. The other potential mechanism, Dyakonov–Perel,
is nominally absent here, since graphene alone has space inversion sym-
metry and thus possesses no spin-orbit fields. Consequently, spin preces-
sion—which is needed for the Dyakonov–Perel mechanism to work—is
absent. Since graphene is made of light carbon atoms, spin-orbit coupling
in graphene is rather weak, about 10 µeV [92]. This intrinsic mechanism is
expected to give spin relaxation times in the range of 10 ns–1 µs, depend-
ing on the specifics of electron-phonon coupling. The presence of phonons
in the spin relaxation mechanism would also lead to a distinct tempera-
ture dependence of the spin relaxation rate. As is usual with the Elliott–
Yafet mechanism, the spin relaxation rate would be also proportional to the
momentum relaxation rate.
What about graphene on a substrate? Wouldn’t the substrate break space
inversion symmetry to induce large enough spin-orbit fields to relax elec-
tron spins efficiently according to the Dyakonov–Perel mechanism? In most
experiments on spin relaxation, graphene is on substrates, typically SiO2.
Such a substrate could directly induce spin-orbit fields in graphene due to
weak van der Waals hybridization, but also due to unintended charges on the
substrate; the charges generate electric fields which couple with spin-orbit
coupling to induce Rashba fields. The former effect has not been investigated
in detail yet, but it is expected that spin-orbit fields due to hybridization of
the substrate orbitals with graphene pz orbitals are no greater than 10 s of µeV.
The effect of charges on a substrate was investigated in [93]. The outcome of
this theoretical work is that spin relaxation due to substrates is relatively

weak, giving the spin lifetimes of some microseconds. Also, since the sub-
strate causes spin-orbit fields that lie in the plane (graphene sheet), the spin
relaxation rate would be strongly anisotropic: out-of-plane spin would decay
twice as fast as in-plane spins. This is a very distinct feature which can be
probed experimentally. There have been many theoretical modelings of spin
relaxation in graphene based on spin-orbit coupling [94–105], most giving
spin relaxation times for realistic conditions in the 10 ns–µs range.
The outcomes of spin-injection experiments came as a surprise. In the
first spin-injection experiment [106], the spin relaxation time in graphene
was extracted to be about 100 ps [107], see also Chapter 4, Volume 3. The
momentum relaxation time was also some 100 fs. Subsequent experiments
[107–111] have confirmed this order of magnitude for τ s . The current sta-
tus is that the spin relaxation times in graphene on common substrates and
without special treatments, are 0.1–1 ns, essentially independent of tempera-
ture. This is orders of magnitude less than expected theoretically, based on
spin-orbit coupling mechanisms, as discussed above. Recent experiments
have also shown that the spin relaxation time is highly isotropic [112]. There
is then little doubt that neither phonons (absent temperature dependence)
nor spin-orbit fields due to substrates (absent anisotropy) cause spin relax-
ation of the experimental samples. This excludes both Elliott–Yafet and
Dyakonov–Perel mechanisms. What, then, causes the observed ultrafast
spin relaxation of electrons in graphene?
Thus far, the only theory that consistently explains experimental data
on spin relaxation in graphene is resonant scattering of electrons off mag-
netic moments. Magnetic moments in graphene appear due to defects, such
as vacancies or covalently bonded adatoms and admolecules [113]. There
is no need for transition metal ions. Thus, it is expected that, say, polymer
treatment of graphene devices will result in residual bonded organic mol-
ecules which will give a magnetic moment. It has been shown theoretically
[114] that less than 1 ppm of hydrogen adatoms, for example, would lead
to spin relaxation time of the order of 100 ps. The energy dependence of
the spin relaxation rate agrees very well with experiment. The same theory
also explains the rather puzzling behavior of the spin relaxation in bilayer
graphene [115]. The comparison between the theory and experiments
for single and bilayer graphene is shown in Figure 1.20. There have been
additional theoretical works confirming that resonant scattering off mag-
netic moments is sufficient to give experimentally observed spin lifetimes
[116–119].
However, there are graphene samples in which spin relaxation is greatly
inhibited. Namely, graphene on hBN, or graphene encapsulated by hBN,
exhibit spin relaxation times on the order of 1–10 ns [122–124]. There is not
yet much investigation of these samples but it appears that the increase of
spin relaxation times is not accompanied by an increase in the momentum
relaxation time. This indicates that the scatterers causing the momentum
relaxation and spin relaxation are decoupled, which is the hallmark of the
resonant scattering by magnetic moments. But the spin lifetimes of 10 ns
FIGURE 1.20 Spin relaxation rate as calculated using the mechanism of resonant
scattering off local magnetic moments. Left: spin relaxation rate as a function of
the Fermi energy for single-layer graphene. The solid line is the calculation for
0.36 ppm of magnetic moments, at 300 K, and with electron-hole puddles smear-
ing of 110 meV. The experimental data are filled circles [120]. The figure adapted
from [114]. Right: the same but for bilayer graphene. Theoretical calculations are
performed for 0.16 ppm of hydrogen adatoms. The experimental data are from
[110] (empty circles) and from [121] (filled squares), with indicated temperature.
The Fermi-level smearing is 23 meV (the smearing is lower in bilayer graphene due
to the greater electronic density of states). The disagreement between the two
experimental results for bilayer graphene underlines the extrinsic character of spin
relaxation in graphene. (After Kochan, D. et al., Phys. Rev. Lett. 115, 196601, 2015.
With permission.)
can already be marginally large enough to be limited by other mechanisms,

possibly Dyakonov-Perel and Elliott-Yafet. Further experimental investiga-
tions, especially of the temperature dependence and anisotropy, would help
to settle the open questions.
While graphene is an exciting spintronics material, mainly because the
spin diffusion length is typically greater than a micron (despite the ultra-
short spin relaxation time), there has been considerable excitement recently
about graphene on two-dimensional transition-metal dichalcogenides
(TMDC). Namely, it is predicted by the ab initio theory that spin-orbit cou-
pling of graphene on TMDCs such as MoS2 or WSe2, is about 1–10 meV [125,
126], several orders of magnitude more than in pristine graphene. This opens
great possibilities for spintronics and some of the opportunities in TMDCs
and other 2D materials are discussed in Volume 3, Chapter 5. There have
already been experiments measuring the spin relaxation time of electrons in
graphene on TMDCs, resulting in some controversy. On the one hand, weak
localization measurements seem to yield spin relaxation times on the order
of picoseconds [127, 128]. On the other hand, conventional Hanle effect
measurements in spin-transport circuits give spin relaxation times of tens

of picoseconds [129, 130].
One peculiar feature of the induced spin-orbit coupling is the emergence
of the spin-valley “locking”, inherited by graphene from TMDCs. Namely,
the electron spin in K and K’ valleys is opposite. One can imagine a fictitious
magnetic field, out-of-plane, producing Zeeman splitting, which is opposite
at the two K valleys. In addition, there is the usual Rashba field due to the
TMDC substrate, again in meV range. The presence of the out-of-plane val-
ley fields substantially alters the spin relaxation anisotropy. Indeed, if there is
only the Rashba interaction, which makes in-plane spin-orbit fields, the out-
of-plane spins relax faster. But in the presence of valley Zeeman fields, which
are out-of-plane, it is the in-plane spins that decay faster: see Figure 1.21. for
an illustration. It was theoretically predicted [131] that the spin anisotropy,
expressed as the ratio of the out-of-plane to in-plane spin relaxation times,
ts ,^
, is 10–100 in graphene on TMDCs. This prediction has been swiftly
ts ,
confirmed experimentally [118, 132].
What about other 2D materials? At the moment, we have perhaps the
best understanding of the spin relaxation of conduction electrons in phos-
phorene. Phosphorene, a single (or multiple) layer of black phosphorous is a
direct band gap semiconductor which has space inversion symmetry. Like
FIGURE 1.21 Why is spin relaxation anisotropic in graphene on TMDCs. There

are two valleys, K and K’, where conduction electrons in graphene live. The Rashba
spin-orbit fields in combination with somewhat greater out-of-plane valley
Zeeman fields give a specific spin texture for electron spins at the two valleys, as
shown. Intravalley scattering, with the characteristic momentum relaxation time t p
mainly relaxes out-of-plane spins, since the effective fluctuating spin-orbit field lies
in the plane (the out-of-plane component at a given valley is fixed, up or down).
The Dyakonov–Perel mechanism of spin relaxation then gives the spin relaxation
time, t s,⊥ , proportional to the momentum relaxation time, t p , as indicated. On
the other hand, intervalley scattering is governed by the relecation time t iv. The
in-plane spins will be mainly relaxed by the fluctuating Zeeman valley fields in
intervalley scattering, and the Dyakonov–Perel mechanism yields t s, ,
proportional to t iv . (After Cummings, A.W. et al., Phys. Rev. Lett. 119, 206601, 2017.
With permission.)
References 51
graphene, it is expected that the spin relaxation is due to spin-orbit coupling

and momentum scattering by phonons and impurities. In other words, the
mechanism of spin relaxation should be Elliott–Yafet. This is indeed what
is observed experimentally in the first electrical spin injection experiment
into an ultrathin black phosphorous [134]. The experiment finds a distinct
increase of the spin relaxation rate with increasing temperature, which is a
signature of phonons-induced spin relaxation. The observed spin relaxation
times are nanoseconds, which is rather long, making phosphorene a promis-
ing candidate for spintronics applications. The observed spin relaxation time
agrees also very well with microscopic theory [135]. The theory further pre-
dicts a large spin relaxation anisotropy, due to the anisotropic crystalline
structure of phosphorene. This is yet to be observed in experiment.
ACKNOWLEDGMENTS
This work was supported by DFG SFB 689 and SPP1286, Natural Science
Foundation of China under Grant No. 10725417, the National Basic Research
Program of China under Grant No. 2006CB922005, and the Knowledge
Innovation Project of Chinese Academy of Sciences.
REFERENCES
1. R. J. Elliott, Theory of the effect of spin–orbit coupling on magnetic resonance
in some semiconductors, Phys. Rev. 96, 266 (1954).
2. Y. Yafet, F. Seitz, and D. Turnbull (Eds.), G-factors and spin-lattice relaxation
of conduction electronics, Solid State Physics, vol. 14, Academic, New York,
p. 2, 1963.
3. M. I. D’yakonov and V. I. Perel’, Spin relaxation of conduction electrons in
noncentrosymmetric semiconductors, Sov. Phys. Solid State 13, 3023 (1971).
4. G. L. Bir, A. G. Aronov, and G. E. Pikus, Spin relaxation of electrons due to
scattering by holes, Sov. Phys. JETP 42, 705 (1976).
5. F. Meier and B. P. Zakharchenya (Eds.), Optical Orientation, North-Holland,
New York, 1984.
6. J. Fabian and S. Das Sarma, Spin relaxation of conduction electrons, J. Vac. Sci.
Technol. B 17, 1708 (1999).
7. I. Žutić, J. Fabian, and S. Das Sarma, Spintronics: Fundamentals and applica-
tions, Rev. Mod. Phys. 76, 323 (2004).
8. J. Fabian, A. Matos-Abiague, C. Ertler, P. Stano, and I. Žutić, Semiconductor
spintronics, Acta Phys. Slov. 57, 565 (2007).
9. M. W. Wu, J. H. Jiang, and M. Q. Weng, Spin dynamics in semiconductors,
Phys. Rep. 493, 61 (2010).
10. Y. A. Bychkov and E. I. Rashba, Properties of a 2D electron gas with lifted spec-
tral degeneracy, JETP Lett. 39, 78 (1984).
11. G. Dresselhaus, Spin–orbit coupling effects in zinc blende structures, Phys.
Rev. 100, 580 (1955).
12. M. I. Dyakonov and V. Y. Kachorovskii, Spin relaxation of two-dimensional
electrons in noncentrosymmetric semiconductors, Sov. Phys. Semicond. 20,
110 (1986).
13. B. A. Bernevig, J. Orenstein, and S. C. Zhang, Exact SU(2) symmetry and per-
sistent spin helix in a spin–orbit coupled system, Phys. Rev. Lett. 97, 236601
(2007).
14. J. D. Koralek, C. P. Weber, J. Orenstein et al., Emergence of the persistent spin
helix in semiconductor quantum wells, Nature 458, 610 (2009).
15. J. L. Cheng, M. W. Wu, and J. Fabian, The spin relaxation of conduction elec-
trons in silicon, Phys. Rev. Lett. 104, 016601 (2010).
16. D. J. Lepine, Spin resonance of localized and delocalized electrons in phospho-
rus-doped silicon between 20 and 30 K, Phys. Rev. B 2, 2429 (1970).
17. B. Huang and I. Appelbaum, Coherent spin transport through a 350 micron
thick silicon wafer, Phys. Rev. Lett. 99, 177209 (2007).
18. M. W. Wu and H. Metiu, Kinetics of spin coherence of electrons in an undoped
semiconductor quantum well, Phys. Rev. B 61, 2945 (2000).
19. M. W. Wu and C. Z. Ning, A novel mechanism for spin dephasing due to spin-
conserving scatterings, Eur. Phys. J. B 18, 373 (2000).
20. M. W. Wu, Spin dephasing induced by inhomogeneous broadening in
D’yakonov–Perel’ effect in a n-doped GaAs quantum well, J. Phys. Soc. Jpn. 70,
2195 (2001).
21. M. Q. Weng and M. W. Wu, Spin dephasing in n-type GaAs quantum wells,
Phys. Rev. B 68, 075312 (2003).
22. M. Q. Weng, M. W. Wu, and L. Jiang, Hot-electron effect in spin dephasing in
n-type GaAs quantum wells, Phys. Rev. B 69, 245320 (2004).
23. M. Q. Weng and M. W. Wu, Longitudinal spin decoherence in spin diffusion
in semiconductors, Phys. Rev. B 66, 235109 (2002).
24. M. Q. Weng and M. W. Wu, Kinetic theory of spin transport in n-type semi-
conductor quantum wells, J. Appl. Phys. 93, 410 (2003).
25. M. Q. Weng, M. W. Wu, and Q. W. Shi, Spin oscillations in transient diffusion
of a spin pulse in n-type semiconductor quantum wells, Phys. Rev. B 69, 125310
(2004).
26. E. Ya. Sherman, Random spin–orbit coupling and spin relaxation in symmetric
quantum wells, Appl. Phys. Lett. 82, 209 (2003).
27. H. Haug and A. P. Jauho, Quantum Kinetics in Transport and Optics of
Semiconductors, Springer, Berlin, 1996.
28. M. Q. Weng and M. W. Wu, Multisubband effect in spin dephasing in semi-
conductor quantum wells, Phys. Rev. B 70, 195318 (2004).
29. J. Zhou, J. L. Cheng, and M. W. Wu, Spin relaxation in n-type GaAs quan-
tum wells from a fully microscopic approach, Phys. Rev. B 75, 045305
(2007).
30. J. Zhou and M. W. Wu, Spin relaxation due to the Bir–Aronov–Pikus mecha-
nism in intrinsic and p-type GaAs quantum wells from a fully microscopic
approach, Phys. Rev. B 77, 075318 (2008).
31. J. H. Jiang, Y. Zhou, T. Korn, C. Schüller, and M. W. Wu, Electron spin relax-
ation in paramagnetic Ga(Mn)As quantum wells, Phys. Rev. B 79, 155201
(2009).
32. P. Zhang, J. Zhou, and M. W. Wu, Multivalley spin relaxation in the presence
of high in-plane electric fields in n-type GaAs quantum wells, Phys. Rev. B 77,
235323 (2008).
33. J. H. Jiang, M. W. Wu, and Y. Zhou, Kinetics of spin coherence of electrons in
n-type InAs quantum wells under intense terahertz laser fields, Phys. Rev. B
78, 125309 (2008).
34. P. Zhang and M. W. Wu, Effect of nonequilibrium phonons on hot-electron
spin relaxation in n-type GaAs quantum wells, Europhys. Lett. 92, 47009
(2010).
35. C. Lü, J. L. Cheng, and M. W. Wu, Hole spin dephasing in P-type semiconduc-
tor quantum wells, Phys. Rev. B 73, 125314 (2006).
36. Y. Zhou, J. H. Jiang, and M. W. Wu, Electron spin relaxation in P-type GaAs
quantum wells, New J. Phys. 11, 113039 (2009).
37. P. Zhang and M. W. Wu, Hole spin relaxation in [001] strained asymmetric Si/
SiGe and Ge/SiGe quantum wells, Phys. Rev. B 80, 155311 (2009).
38. C. Lü, U. Zülicke, and M. W. Wu, Hole spin relaxation in P-type GaAs quan-
tum wires investigated by numerically solving fully microscopic kinetic spin
Bloch equations, Phys. Rev. B 78, 165321 (2008).
References 53
39. C. Lü, H. C. Schneider, and M. W. Wu, Electron spin relaxation in n-type InAs
quantum wires, J. Appl. Phys. 106, 073703 (2009).
40. J. H. Jiang, Y. Y. Wang, and M. W. Wu, Reexamination of spin decoherence
in semiconductor quantum dots from the equation-of-motion approach, Phys.
Rev. B 77, 035323 (2008).
41. J. H. Jiang and M. W. Wu, Electron-spin relaxation in bulk III–V semiconduc-
tors from a fully microscopic kinetic spin Bloch equation approach, Phys. Rev.
B 79, 125206 (2009).
42. J. L. Cheng and M. W. Wu, Spin diffusion/transport in n-type GaAs quantum
wells, J. Appl. Phys. 101, 073702 (2007).
43. P. Zhang and M. W. Wu, Spin diffusion in Si/SiGe quantum wells: Spin relax-
ation in the absence of D’yakonov–Perel’ relaxation mechanism, Phys. Rev. B
79, 075303 (2009).
44. M. E. Flatté, J. M. Bayers, and W. H. Lau, Spin dynamics in semiconductors, in
Spin Dynamics in Semiconductors, D. D. Awsehalom, D. Loss, and N. Samarth
(Eds.), Springer, Berlin, 2002.
45. M. M. Glazov and E. L. Ivchenko, Precession spin relaxation mechanism
caused by frequent electron–electron collisions, JETP Lett. 75, 403 (2002).
46. M. A. Brand, A. Malinowski, O. Z. Karimov et al., Precession and motional
slowing of spin evolution in a high mobility two-dimensional electron gas, Phys.
Rev. Lett. 89, 236601 (2002); W. J. H. Leyland, R. T. Harley, M. Henini, A. J.
Shields, I. Farrer, and D. A. Ritchie, Energy-dependent electron–electron scat-
tering and spin dynamics in a two-dimensional electron gas, Phys. Rev. B 77,
205321 (2008).
47. X. Z. Ruan, H. H. Luo, Y. Ji, Z. Y. Xu, and V. Umansky, Effect of electron–elec-
tron scattering on spin dephasing in a high-mobility low-density two-dimen-
sional electron gas, Phys. Rev. B 77, 193307 (2008).
48. Y. Ohno, R. Terauchi, T. Adachi, F. Matsukura, and H. Ohno, Electron spin
relaxation beyond D’yakonov–Perel’ interaction in GaAs/AlGaAs quantum
wells, Physica E 6, 817 (2000).
49. F. X. Bronold, A. Saxena, and D. L. Smith, Semiclassical kinetic theory of elec-
tron spin relaxation in semiconductors, Phys. Rev. B 70, 245210 (2004).
50. G. F. Giulianni and G. Vignale, Quantum Theory of the Electron Liquid,
Cambridge University Press, Cambridge, UK, 2005.
51. L. Jiang and M. W. Wu, Control of spin coherence in n-type GaAs quantum
wells using strain, Phys. Rev. B 72, 033311 (2005).
52. A. W. Holleitner, V. Sih, R. C. Myers, A. C. Gossard, and D. D. Awschalom,
Dimensionally constrained D’yakonov–Perel’ spin relaxation in n-InGaAs
channels: Transition from 2D to 1D, New J. Phys. 9, 342 (2007).
53. F. Zhang, H. Z. Zheng, Y. Ji, J. Liu, and G. R. Li, Spin dynamics in the sec-
ond subband of a quasi–two-dimensional system studied in a single-barrier
heterostructure by time-resolved Kerr rotation, Europhys. Lett. 83, 47007
(2008).
54. C. Lü, J. L. Cheng, M. W. Wu, and I. C. da Cunha Lima, Spin relaxation time,
spin dephasing time and ensemble spin dephasing time in n-type GaAs quan-
tum wells, Phys. Lett. A 365, 501 (2007).
55. D. Stich, J. Zhou, T. Korn et al., Effect of initial spin polarization on spin
dephasing and the electron g factor in a high-mobility two-dimensional elec-
tron system, Phys. Rev. Lett. 98, 176401 (2007).
56. D. Stich, J. Zhou, T. Korn et al., Dependence of spin dephasing on initial spin
polarization in a high-mobility two-dimensional electron system, Phys. Rev. B
76, 205301 (2007).
57. F. Zhang, H. Z. Zheng, Y. Ji, J. Liu, and G. R. Li, Electrical control of dynamic
spin splitting induced by exchange interaction as revealed by time-resolved
Kerr rotation in a degenerate spin-polarized electron gas, Europhys. Lett. 83,
47006 (2008).
58. T. Korn, D. Stich, R. Schulz, D. Schuh, W. Wegscheider, and C. Schüller, Spin

dynamics in high-mobility two-dimensional electron system, Adv. Solid State
Phys. 48, 143 (2009).
59. A. P. Dmitriev, V. Y. Kachorovskii, and M. S. Shur, High-field transport in a
dense two-dimensional electron gas in elementary semiconductors, J. Appl.
Phys. 89, 3793 (2001).
60. E. M. Conwell, High Field Transport in Semiconductors, Pergamon, Oxford,
1972.
61. T. C. Damen, L. Vina, J. E. Cunningham, J. Shah, and L. J. Sham, Subpicosecond
spin relaxation dynamics of excitons and free carriers in GaAs quantum wells,
Phys. Rev. Lett. 67, 3432 (1991).
62. J. Wagner, H. Schneider, D. Richards, A. Fischer, and K. Ploog, Observation
of extremely long electron-spin-relaxation times in p-type δ-doped GaAs/
AlxGa1−xAs double -heterostructures, Phys. Rev. B 47, 4786 (1993).
63. H. Gotoh, H. Ando, T. Sogawa, H. Kamada, T. Kagawa, and H. Iwamura, Effect
of electron–hole interaction on electron spin relaxation in GaAs/AlGaAs
quantum wells at room temperature, J. Appl. Phys. 87, 3394 (2000).
64. T. F. Boggess, J. T. Olesberg, C. Yu, M. E. Flatté, and W. H. Lau, Room-
temperature electron spin relaxation in bulk InAs, Appl. Phys. Lett. 77, 1333
(2000).
65. S. Hallstein, J. D. Berger, M. Hilpert et al., Manifestation of coherent spin
precession in stimulated semiconductor emission dynamics, Phys. Rev. B 56,
R7076 (1997).
66. P. Nemec, Y. Kerachian, H. M. van Driel, and A. L. Smirl, Spin-dependent elec-
tron many-body effects in GaAs, Phys. Rev. B 72, 245202 (2005).
67. H. C. Schneider, J.-P. Wüstenberg, O. Andreyev et al., Energy-resolved elec-
tron spin dynamics at surfaces of p-doped GaAs, Phys. Rev. B 73, 081302
(2006).
68. C. Yang, X. Cui, S.-Q. Shen, Z. Xu, and W. Ge, Spin relaxation in submono-
layer and monolayer InAs structures grown in a GaAs matrix, Phys. Rev. B 80,
035313 (2009).
69. P. H. Song and K. W. Kim, Spin relaxation of conduction electrons in bulk
III–V semiconductors, Phys. Rev. B 66, 035207 (2002).
70. D. D. Awschalom, D. Loss, and N. Samarth (Eds.), Semiconductor Spintronics
and Quantum Computation, Springer-Verlag, Berlin, 2002; M. I. D’yakonov
(Ed.), Spin Physics in Semiconductors, Springer, Berlin, 2008, and references
therein.
71. R. I. Dzhioev, K. V. Kavokin, V. L. Korenev et al., Low-temperature spin relax-
ation in n-type GaAs, Phys. Rev. B 66, 245204 (2002).
72. M. Krauß, R. Bratschitsch, Z. Chen, S. T. Cundiff, and H. C. Schneider,
Ultrafast spin dynamics in optically excited bulk GaAs at low temperatures,
Phys. Rev. B 81, 035213 (2010).
73. K. Shen, A peak in density dependence of electron spin relaxation time
in n-type bulk GaAs in the metallic regime, Chin. Phys. Lett. 26, 067201
(2009).
74. I. Appelbaum, B. Huang, and D. J. Monsma, Electronic measurement and con-
trol of spin transport in silicon, Nature 447, 295 (2007).
75. B. Huang, L. Zhao, D. J. Monsma, and I. Appelbaum, 35% magnetocurrent with
spin transport through Si, Appl. Phys. Lett. 91, 052501 (2007).
76. G. Dresselhaus, Spin–orbit coupling effects in zinc blende structures, Phys.
Rev. 100, 580 (1955).
77. L. Jiang, M. Q. Weng, M. W. Wu, and J. L. Cheng, Diffusion and transport
of spin pulses in an n-type semiconductor quantum well, J. Appl. Phys. 98,
113702 (2005).
78. S. A. Crooker and D. L. Smith, Imaging spin flows in semiconductors subject to
electric, magnetic, and strain fields, Phys. Rev. Lett. 94, 236601 (2005).
References 55
79. M. Beck, C. Metzner, S. Malzer, and G. H. Döhler, Spin lifetimes and strain-
controlled spin precession of drifting electrons in GaAs, Europhys. Lett. 75,
597 (2006).
80. J. L. Cheng, M. W. Wu, and I. C. da Cunha Lima, Anisotropic spin transport
in GaAs quantum wells in the presence of competing Dresselhaus and Rashba
spin–orbit coupling, Phys. Rev. B 75, 205328 (2007).
81. N. S. Averkiev and L. E. Golub, Giant spin relaxation anisotropy in zinc-blende
heterostructures, Phys. Rev. B 60, 15582 (1999).
82. N. S. Averkiev, L. E. Golub, and M. Willander, Spin relaxation anisotropy in
two-dimensional semiconductor systems, J. Phys. Condens. Matter 14, R271
(2002).
83. J. Schliemann, J. C. Egues, and D. Loss, Nonballistic spin-field-effect transistor,
Phys. Rev. Lett. 90, 146801 (2003).
84. R. Winkler, Spin orientation and spin precession in inversion-asymmetric
quasi-two-dimensional electron systems, Phys. Rev. B 69, 045317 (2004).
85. J. L. Cheng and M. W. Wu, Spin relaxation under identical Dresselhaus and
Rashba coupling strengths in GaAs quantum wells, J. Appl. Phys. 99, 083704
(2006).
86. K. Shen and M. W. Wu, Infinite spin diffusion length of any spin polarization
along direction perpendicular to effective magnetic field from Dresselhaus and
Rashba spin–orbit couplings with identical strengths in (001) GaAs quantum
wells, J. Supercond. Nov. Magn. 22, 715 (2009).
87. C. P. Weber, N. Gedik, J. E. Moore, J. Orenstein, J. Stephens, and D. D.
Awschalom, Observation of spin Coulomb drag in a two-dimensional electron
gas, Nature 437, 1330 (2005).
88. A. R. Cameron, P. Rickel, and A. Miller, Spin gratings and the measurement of
electron drift mobility in multiple quantum well semiconductors, Phys. Rev.
Lett. 76, 4793 (1996).
89. C. P. Weber, J. Orenstein, B. A. Bernevig, S.-C. Zhang, J. Stephens, and D. D.
Awschalom, Nondiffusive spin dynamics in a two-dimensional electron gas,
Phys. Rev. Lett. 98, 076604 (2007).
90. M. Q. Weng, M. W. Wu, and H. L. Cui, Spin relaxation in n-type GaAs quan-
tum wells with transient spin grating, J. Appl. Phys. 103, 063714 (2008).
91. W. Han, R. K. Kawakami, M. Gmitra, and J. Fabian, Graphene spintronics, Nat.
Nanotechnol. 9, 794 (2014).
92. M. Gmitra, S. Konschuh, and C. Ertler, C. Ambrosch-Draxl, and J. Fabian,
Band-structure topologies of graphene: Spin-orbit coupling effects from first
principles, Phys. Rev. B 80, 235431 (2009).
93. C. Ertler, S. Konschuh, M. Gmitra, and J. Fabian, Electron spin relaxation in
graphene: The role of the substrate, Phys. Rev. B 80, 041405 (2009).
94. D. Van Tuan, F. Ortmann, D. Soriano, S. O. Valenzuela, and S. Roche,
Pseudospin-driven spin relaxation mechanism in graphene, Nat. Phys. 10, 857
(2014).
95. P. Zhang and M. W. Wu, Electron spin diffusion and transport in graphene,
Phys. Rev. B 84, 045304 (2011).
96. S. Fratini, D. Gosálbez-Martínez, P. Merodio Cámara, and J. Fernández-
Rossier, Anisotropic intrinsic spin relaxation in graphene due to flexural dis-
tortions, Phys. Rev. B 88, 115426 (2013).
97. D. Huertas-Hernando, F. Guinea, and A. Brataas, Spin-orbit-mediated spin
relaxation in graphene, Phys. Rev. Lett. 103, 146801 (2009).
98. P. R. Struck and G. Burkard, Effective time-reversal symmetry breaking in the
spin relaxation in a graphene quantum dot, Phys. Rev. B 82, 125401 (2010).
99. H. Ochoa, A. H. Castro Neto, and F. Guinea, Elliot-Yafet mechanism in gra-
phene, Phys. Rev. Lett. 108, 206808 (2012).
100. V. K. Dugaev, E. Sherman, and J. Barnas, Spin dephasing and pumping in gra-
phene due to random spin-orbit interaction, Phys. Rev. B 83, 085306 (2011).
101. S. Jo, D.-K. Ki, D. Jeong, H.-J. Lee, and S. Kettemann, Spin relaxation properties
in graphene due to its linear dispersion, Phys. Rev. B 84, 075453 (2011).
102. H. Dery, P. Dalal, L. Cywinski, and L. J. Sham, Spin-based logic in semiconduc-
tors for reconfigurable large-scale circuits, Nature 447, 573 (2007).
103. B. Dóra, F. Murányi, and F. Simon, Electron spin dynamics and electron spin
resonance in graphene, Eur. Phys. Lett. 92, 17002 (2010).
104. L. Wang and M. W. Wu, Electron spin relaxation in bilayer graphene, Phys.
Rev. B 87, 205416 (2013).
105. D. V. Fedorov M. Gradhand, S. Ostanin et al., Impact of electron-impurity
scattering on the spin relaxation time in graphene: A first-principles study,
Phys. Rev. Lett. 110, 156602 (2013).
106. N. Tombros, C. Jozsa, M. Popinciuc, H. T. Jonkman, and B. J. van Wees,
Electronic spin transport and spin precession in single graphene layers at room
temperature, Nature 448, 571 (2007).
107. A. Avsar, T.-Y. Yang, S. Bae et al., Toward wafer scale fabrication of graphene
based spin valve devices, Nano Lett. 11, 2363 (2011).
108. N. Tombros, S. Tanabe, A. Veligura, et al., Anisotropic spin relaxation in gra-
109. K. Pi, W. Han, K. M. McCreary, A. G. Swartz, Y. Li, and R. K. Kawakami,
Manipulation of spin transport in graphene by surface chemical doping, Phys.
Rev. Lett. 104, 187201 (2010).
110. T.-Y. Yang, J. Balakrishnan, F. Volmer et al., Observation of long spin-relax-
ation times in bilayer graphene at room temperature, Phys. Rev. Lett. 107,
047206 (2011).
111. W. Han and R. K. Kawakami, Spin relaxation in single-layer and bilayer gra-
112. B. Raes, J. E. Scheerder, M. V. Costache et al., Determination of the spin-life-
time anisotropy in graphene using oblique spin precession, Nat. Commun. 7,
11444 (2016).
113. O. V. Yazyev, Emergence of magnetism in graphene materials and nanostruc-
tures, Rep. Prog. Phys. 73, 056501 (2010).
114. D. Kochan, M. Gmitra, and J. Fabian, Spin relaxation mechanism in graphene:
Resonant scattering by magnetic impurities, Phys. Rev. Lett. 112, 116602
(2014).
115. D. Kochan, S. Irmer, M. Gmitra, and J. Fabian, Resonant scattering by mag-
netic impurities as a model for spin relaxation in bilayer graphene, Phys. Rev.
Lett. 115, 196601 (2015).
116. J. Bundesmann, D. Kochan, F. Tkatschenko, J. Fabian, and K. Richter, Theory of
spin-orbit-induced spin relaxation in functionalized graphene, Phys. Rev. B 92,
081403 (2015).
117. J. Wilhelm, M. Walz, and F. Evers, Ab initio pin-flip conductance of hydroge-
nated graphene nanoribbons: Spin-orbit interaction and scattering with local
impurity spins, Phys. Rev. B 92, 014405 (2015).
118. V. G. Miranda, E. R. Mucciolo, and C. H. Lewenkopf, Spin relaxation in disor-
dered graphene: Interplay between puddles and defect-induced magnetism, J.
Phys. Chem. Solids (2017). doi:https://doi.org/10.1016/j.jpcs.2017.10.022
119. D. Soriano, D. V. Tuan, S. M.-M. Dubois et al., Spin transport in hydrogenated
graphene, 2D Mater. 2, 022002 (2015).
120. M. Wojtaszek, I. J. Vera-Marun, T. Maassen, and B. J. van Wees, Enhancement of
spin relaxation time in hydrogenated graphene spin-valve devices, Phys. Rev. B
87, 081402 (2013).
121. W. Han and R. K. Kawakami, Spin relaxation in single-layer and bilayer gra-
122. M. Drögeler, F. Volmer, M. Wolter et al., Nanosecond spin lifetimes in single-
and few-layer graphene–hBN heterostructures at room temperature, Nano
Lett. 14, 6050 (2014).
References 57
123. M. H. D. Guimarães, P. J. Zomer, J. Ingla-Aynés, J. C. Brant, N. Tombros, and

B. J. van Wees, Controlling spin relaxation in hexagonal BN-encapsulated gra-
phene with a transverse electric field, Phys. Rev. Lett. 113, 086602 (2014).
124. M. Drögeler, C. Franzen, F. Volmer et al., Spin lifetimes exceeding 12 ns in
graphene nonlocal spin valve devices, Nano Lett. 16, 3533 (2016).
125. M. Gmitra and J. Fabian, Graphene on transition-metal dichalcogenides: A
platform for proximity spin-orbit physics and optospintronics, Phys. Rev. B 92,
155403 (2015).
126. M. Gmitra, D. Kochan, P. Högl, and J. Fabian, Trivial and inverted Dirac bands
and the emergence of quantum spin Hall states in graphene on transition-
metal dichalcogenides, Phys. Rev. B 93, 155104 (2016).
127. Z. Wang, D.-K. Ki, H. Chen, H. Berger, A. H. MacDonald, and A. F. Morpurgo,
Strong interface-induced spin-orbit interaction in graphene on WS2, Nat.
Commun. 6, 8339 (2015).
128. B. Yang, M.-F. Tu, J. Kim et al., Tunable spin–orbit coupling and symmetry-
protected edge states in graphene/WS2, 2D Mater. 3, 031012 (2016).
129. S. Omar and B. J. van Wees, Graphene-WS2 heterostructures for tunable spin
injection and spin transport, Phys. Rev. B 95, 081404 (2017).
130. A. Dankert and S. P. Dash, Electrical gate control of spin current in van der
Waals heterostructures at room temperature, Nat. Commun. 8, 16093 (2017).
131. A. W. Cummings, J. H. Garcia, J. Fabian, and S. Roche, Giant spin lifetime
anisotropy in graphene induced by proximity effects, Phys. Rev. Lett. 119,
206601 (2017).
132. T. S. Ghiasi, J. Ingla-Aynés, A. A. Kaverzin, and B. J. van Wees, Large proxim-
ity-induced spin lifetime anisotropy in transition-metal dichalcogenide/gra-
phene heterostructures, Nano Lett. 17, 7528 (2017).
133. A. Avsar, J. Y. Tan, M. Kurpas et al., Gate-tunable black phosphorus spin valve
with nanosecond spin lifetimes, Nat. Phys. 13, 888 (2017).
134. M. Kurpas, M. Gmitra, and J. Fabian, Spin-orbit coupling and spin relaxation
in phosphorene: Intrinsic versus extrinsic effects, Phys. Rev. B 94, 1155423
(2016).
2
Electrical Spin
Injection and
Transport in
Semiconductors
Berend T. Jonker
2.1 Introduction 61
2.2 Spin Polarization by Optical Pumping 62
2.3 Spin Polarization by Electrical Injection: General
Considerations 63
2.4 Semiconductor/Semiconductor Electrical Spin Injection 65
2.4.1 Magnetic Semiconductors: Material Properties 65
2.4.2 Magnetic Semiconductors as Spin-Injecting
Contacts: Initial Transport Studies 69
2.4.3 Spin-Polarized Light-Emitting Diode: Spin-LED 69
Contacts: Spin-LED Studies 75
59
60 Chapter 2. Electrical Spin Injection and Transport in Semiconductors
2.4.4.1 Zn1−xMnxSe 75
2.4.4.2 Ga1−xMnxAs 78
2.4.4.3 Spin Scattering by Interface Defects 79
2.5 Ferromagnetic Metal/Semiconductor Electrical Spin
Injection: Optical Detection with Spin-LED Structures 80
2.5.1 Band Symmetries and Spin Transmission 81
2.5.2 Conductivity Mismatch: Description of the
Problem 83
2.5.3 Conductivity Mismatch: Practical Solution 86
2.5.3.1 Tailored Schottky Barrier as a Tunnel
Barrier 86
2.5.3.2 Spin Injection via a Discrete Layer as
a Tunnel Barrier 96
2.5.4 Hot Electron Spin Injection 108
2.6 Ferromagnetic Metal/Semiconductor Electrical Spin
Injection: Electrical Detection 110
2.6.1 General Considerations and Non-Local
Detection 110
2.6.2 Non-Local Spin Injection and Detection in
Silicon: Pure Spin Currents 112
2.6.3 Three Terminal Spin Injection and Detection
in Si at Room Temperature: Spin-Polarized
Charge Current 117
2.6.4 Three Terminal Spin Injection and Detection
in GaAs: Interpretation and the Role of
Localized States 121
2.7 Graphene as a Tunnel Barrier for Spin Injection/
Detection in Silicon 122
2.7.1 Three-Terminal Device Geometry 123
2.7.2 Four-Terminal Non-Local Spin Valve Geometry 127
2.7.3 Spin Transport in Silicon Nanowires 131
2.8 Outlook and Critical Research Issues 135
Acknowledgments 136
References 137
2.1 Introduction 61
2.1 INTRODUCTION
The field of semiconductor electronics has been based exclusively on the
manipulation of charge. The remarkable advances in performance have been
due in large part to size scaling, i.e. systematic reduction in device dimen-
sions, enabling significant increases in circuit density. This trend, known as
Moore’s Law [1], is likely to be curtailed in the near future by practical and
fundamental limits. Consequently, there is keen interest in exploring new
avenues and paradigms for future technologies. Spintronics, or the use of
carrier spin as a new degree of freedom in an electronic device, represents
one of the most promising candidates for this paradigm shift [2, 3]. Like
charge, spin is an intrinsic fundamental property of an electron, and is one
of the alternative state variables under consideration on the International
Technology Roadmap for Semiconductors (ITRS) for processing informa-
tion in the new ways that will be required beyond the ultimate scaling limits
of the existing silicon-based complementary metal-oxide-semiconductor
(CMOS) technology [4].
Semiconductor-based spintronics offers many new avenues and oppor-
tunities which are inaccessible to metal-based spintronic structures. This is
due to the characteristics for which semiconductors are so well-known: the
existence of a band gap that can often be tuned over a significant range in
ternary compounds; the accompanying optical properties on which a vast
optoelectronic industry is based; and the ability to readily control carrier
concentrations and transport characteristics via doping, gate voltages, and
band offsets. Coupling the new degree of freedom of carrier spin with the
traditional band gap engineering of modern electronics offers opportunities
for new functionality and performance for semiconductor devices (examples
can be found in Chapters 16 and 17, Volume 3).
There are four essential requirements for implementing a semiconduc-
tor-based spintronics technology:
1. Efficient electrical injection of spin-polarized carriers from an appro-

priate contact into the semiconductor heterostructure.
2. Adequate spin diffusion lengths and spin lifetimes within the semi-
conductor host medium.
3. Effective control and manipulation of the spin system to provide the
desired functionality.
4. Efficient detection of the spin system to provide the output.
Although the focus of this chapter is nominally on the first factor, each
of these factors will necessarily be addressed to some extent, as each plays a
role in any experimental research effort.
The generation of spin-polarized carrier populations in semiconduc-
tors has historically been accomplished by optical excitation, an approach
referred to as “optical pumping” or “optical orientation” [5]. This has been
the basis for decades of research exploring spin-dependent phenomena in
semiconductor hosts, beginning with the pioneering work of Lampel [6], and
enabling the development of new spectroscopic techniques which provide

extraordinary insight into spin properties [7, 8]. Injection of spin-polarized
electrons from a scanning tunneling microscope tip in ultra-high vacuum
was first reported in 1992 [9], and provides a highly localized probe, suitable
for basic research. However, a broad technology based on spin manipulation
in a semiconductor host requires an efficient and practical means of spin
injection and detection. This is best accomplished via a discrete electrical
contact. Such a contact defines the extent of the spin source (or detector),
provides an interface between the spin system and the canonical electronic
input/output parameters of voltage and current, and offers a very simple
and direct means of implementing spin injection/detection compatible with
existing device design and fabrication technology [10–12].
In this chapter, we will discuss the factors relevant to spin injection and
transport in semiconductors, review the necessary concepts, and illustrate
these by way of examples, to provide both a practical guide and an over-
view of the current state of the art. We will conclude with a brief section
(Section 2.6) on electrical injection and detection of spin accumulation using
multiterminal lateral transport geometries in silicon, where an operation
well above room temperature has recently been demonstrated.
2.2 SPIN POLARIZATION BY OPTICAL PUMPING

The earliest method of generating a spin-polarized population of carriers
within a semiconductor was by optical excitation using circularly polarized
light, often referred to as “optical pumping” [5]. A photon carries a unit of
angular momentum either parallel or antiparallel to its propagation direc-
tion, and the corresponding polarization is referred to as positive (σ+) or
negative (σ−) helicity, respectively. Absorption of a photon whose energy is
equal to or greater than the band gap promotes an electron from the valence
band (VB) to the conduction band (CB). If the light is circularly polarized,
the electron and hole must share the angular momentum transferred in a
manner that satisfies conservation laws [13].
In a direct gap zinc-blende semiconductor such as GaAs, the CB is s-like
in character and twofold spin degenerate, and the electron can occupy states
with values of spin angular momentum mj = ±1/2. The VB is p-like in charac-
ter and fourfold spin degenerate in bulk material, so that the hole can occupy
states with values of angular momentum mj = ±1/2, ±3/2, corresponding to
“light hole” (LH) and “heavy hole” (HH) states, respectively (Figure 2.1a). For
photons incident along the surface normal, interband transitions at the zone
center (k//= 0) satisfy the selection rule Δmj = ±1, reflecting absorption of the
photon’s original angular momentum. The probability of a transition involv-
ing a LH or HH state is weighted by the square of the corresponding matrix
element connecting it to the appropriate electron state, so that HH transitions
are three times more likely than LH transitions. Thus, absorption of photons
with angular momentum +1 produces three “spin-down” (mj = −1/2) electrons
for every one “spin-up” (mj = +1/2) electron, resulting in an electron popula-
tion with a spin polarization of 50% in a bulk material, where the HH and LH
2.3 Spin Polarization by Electrical Injection: General Considerations 63
n n
n n
–1/2 mj +1/2
–1/2 mj +1/2 CB
CB 3 3
3 1 3 1 σ+ σ–
σ+ σ– σ+ σ–
VB
VB –3/2 +3/2
–3/2 +1/2 –1/2 +3/2 +1/2 –1/2
(a) (b)
FIGURE 2.1 Optical transitions allowed due to absorption (or emission—reverse

the arrows) of circularly polarized light for (a) bulk material and (b) a structure in
which the light/heavy hole band degeneracy is lifted, as in a quantum well or
strained material. The HH transitions are three times more probable than the LH.
states are degenerate. In a quantum well (QW), either quantum confinement or

strain can lift this degeneracy, as illustrated in Figure 2.1b, where the HH states
are at lower energy (nearer the center of the band gap). In this case, circularly
polarized light whose energy is just sufficient to excite these states (and not
the LH states) can in principle produce an electron population, which is 100%
spin-polarized. The electron spin lifetime is much longer than that of the hole
in most materials, and the holes are generally assumed to depolarize instan-
taneously. Spin relaxation and lifetimes are discussed in Chapter 1, Volume 2.
Note that the resulting electron spin is oriented parallel or antiparallel
to the propagation direction of the incident photon (typically the surface
normal), consistent with the angular momentum of the absorbed photon. In
addition, the orientation of carrier spin is typically locked along the surface
normal in a quantum well or can be modified by strain. These are important
factors to bear in mind when performing and interpreting an experiment.
2.3 SPIN POLARIZATION BY ELECTRICAL

INJECTION: GENERAL CONSIDERATIONS
A discrete electrical contact to create, manipulate, and detect spin currents
and populations in a semiconductor host enables the development of a much
broader range of technologies than that allowed by optical excitation alone. A
discrete contact clearly defines the source/detector dimensions and is readily
scalable, an indispensable attribute, in contrast with optical approaches which
are diffraction limited. Existing electronics is based upon a voltage or current
for input and output. A spin contact must, therefore, provide a suitable inter-
face between these conventional parameters and the spin system to be utilized
for information processing or sensing in the semiconductor. A magnetic mate-
rial has a spin-polarized band structure and density of states (DOS), and is
well suited for such a contact—the electron population at the Fermi energy, EF ,
naturally exhibits a net spin polarization, which serves as a source for electri-
cal injection under appropriate bias into an unpolarized host.
The idea that a spin current could in principle accompany a charge cur-
rent was first hypothesized by Aronov and Pikus in 1976 [14], when they
noted that the flow of electric current Ie from a ferromagnetic (FM) metal
should exhibit a net spin polarization (magnetization) due to its spin-polar-

ized DOS. If this charge current flows into a nonmagnetic material, it simul-
taneously corresponds to a flow of magnetization, since each electron also
bears the fundamental unit of magnetization, µB, the Bohr magneton. This
magnetization current Im is simply written as
æI ö
I m = hm B ç e ÷ , (2.1)
è e ø
where:
e is the electron’s charge
0 ≤ η ≤ 1 is a phenomenological factor to account for the fact that the
Fermi surface of the ferromagnet may contain both spin bands
(or to account for possible spin scattering at the interface)
This remarkably simple and intuitive relation provided the seed from
which the burgeoning field of spin injection and transport has sprung.
A contact for spin injection/detection should exhibit some specific char-
acteristics. In general, one desires high spin injection efficiency at low bias
power extending well above room temperature for practical device applica-
tions. While it is common to think of these properties as attributes of the
contact material alone, it is essential to note that the interface between the
contact and the semiconductor is equally important and is often more chal-
lenging to understand. Thus, the following should be considered as desirable
attributes for the combination of spin contact material and the correspond-
ing interface:
ªª A Curie temperature Tc > 400 K, although certain specialized applica-

tions (e.g. cooled infrared detectors) may permit Tc’s as low as 100 K.
ªª High remanent magnetization, i.e. the magnetization retained at
zero field. The nonvolatile behavior essential for many applications
(memory, optical isolators, reprogrammable logic) is predicated
upon a remanence that is a substantial fraction of the saturation
magnetization.
ªª The magnetization should be readily switched at modest power
expense, important for applications which require fast and routine
switching (memory, logic). Historically, this has meant a low coercive
field (Hc < 200 Oe), but more recent interest is focused on reducing
the critical current required to switch the free layer in a spin torque
transfer element (see Chapters 7–9 and 15 in Volume 1 for a discus-
sion of spin torque effects), or utilizing an electric field to assist in
magnetization reversal [15].
ªª A highly polarized DOS at EF. This is often viewed to first order as
the source term when discussing spin injection from any contact, and
higher polarizations produce larger signal levels. Half metals such
as select Heusler alloys ideally exhibit 100% spin polarization at EF,
and are of keen interest, although this polarization is rarely, if ever,
achieved in practice.
ªª The contact/interface should have a reasonably low specific resistance

to allow the usable current densities. A 100% spin-polarized current
of 10−3 A cm−2 may not be as useful as a 30% spin-polarized current
of 10 A cm−2, particularly in generating spin accumulation in semi-
conductor structures (with the exception of single-electron devices).
High resistances also degrade signal-to-noise ratios, compromising
device characteristics such as operating frequency. The contact resis-
tance will be a compromise involving consideration of many factors.
ªª The contact/interface should offer spin injection which is robust
against defects. Defects at complex heterointerfaces are to be
expected—their role in spin scattering is poorly understood and
poses an immediate challenge for the development of spin transport
devices.
ªª The contact/interface should be thermodynamically stable, and ther-
mally stable at temperatures typically encountered for growth, pro-
cessing, and device operation.
The above list is neither exhaustive nor intended to be an absolute yard-

stick by which to gauge the merits of a particular contact/interface—per-
formance and intended application define the latter. It is useful, however, to
keep these items in mind as we continue to develop the spin contacts, corre-
sponding interface, and the new functionality that semiconductor spintronic
devices offer.
Two classes of magnetic materials have been utilized as contacts for
electrical spin injection into semiconductors: semiconductors and metals
[10, 12]. FM metals are familiar to all and offer some significant advantages,
but also present some challenges for efficient utilization as spin contacts—
they will be discussed in later paragraphs. Magnetic semiconductors are less
familiar, but have, in fact, been studied for decades. They have enjoyed a
resurgence of research interest since 1990, and were successfully used before
FM metals to electrically inject spin-polarized carriers into another semi-
conductor. An overview of recent developments in III–V and oxide magnetic
semiconductors appears in Chapters 9 and 12, Volume 2.
2.4 SEMICONDUCTOR/SEMICONDUCTOR
ELECTRICAL SPIN INJECTION
2.4.1 Magnetic Semiconductors: Material Properties
Magnetic semiconductors simultaneously exhibit semiconducting proper-
ties and typically either paramagnetic or FM order. The coexistence of these
properties in a single material provides fertile ground for fundamental stud-
ies, and offers exciting possibilities for a broad range of applications. The
use of a magnetic semiconductor as a spin contact enables the design of a
spin-injecting semiconductor/semiconductor interface guided by known
principles of bandgap engineering (CB and VB offsets, doping, and car-
rier transport) and epitaxial growth (lattice match, interface structure, and
materials compatibility). Therefore, they seem ideal candidates for incorpo-

ration in semiconductor spintronic devices. However, their magnetic prop-
erties to date fall well short of those required for practical devices—e.g. the
Curie temperature is typically well below room temperature, despite much
effort to increase it.
Ferromagnetic semiconductors (FMS) were intensively studied from 1950
to 1970—classic examples include the europium chalcogenides (e.g. EuO, S,
and Se) and the chalcogenide spinels, in which the magnetic element forms
a significant fraction of the atomic constituents (see Figure 2.2a) [16, 17].
Simple transport experiments have demonstrated that the exchange split-
ting of the band edges below the Curie temperature could be used as a spin-
dependent potential barrier, which selectively passes one spin component
while blocking the other, leading to current with a net spin polarization. This
“spin filter effect” was demonstrated in EuS- and EuSe-based single barrier
heterostructures in 1967 by Esaki et al. [18], and has been recently revisited
[19, 20] (spin filtering is reviewed in Chapter 14, Volume 1). However, device
applications languished due to low Curie temperatures and the inability to
incorporate these materials in thin-film form with mainstream semiconduc-
tor device materials.
Recent work successfully incorporated epilayers of n-type CdCr2Se4,
another classic magnetic semiconductor, with GaAs-based heterostruc-
tures by molecular beam epitaxy (MBE) [21] and demonstrated electrical
spin injection [22]. CdCr2Se4 is a direct gap (Eg = 1.3 eV) chalcogenide spinel,
which is reasonably lattice matched to technologically important materi-
als such as Si and GaP (∼1.7% tensile mismatch), and to GaAs (5.2% tensile
mismatch), assuming one uses an effective lattice constant of ao/2 for the
CdCr2Se4. Single crystal epilayers grown on both GaP(001) and GaAs(001)
substrates are n-type (n ∼ 1018 cm−3), FM with the easy magnetization axis
in plane (along the GaAs(110)), and have a Curie temperature of 132 K [21].
Subsequent work successfully demonstrated electrical injection of spin-
polarized electrons from CdCr2Se4 contacts into GaAs [22] using the spin-
polarized light-emitting diode (spin-LED) approach, described in detail
below.
Oxide MBE has been used to successfully grow epitaxial films of EuO
on Si(001) [23] and GaN(0001) substrates [24]. EuO has a highly spin-polar-
ized band structure and a large specific Faraday rotation. Andreev reflection
(a) (b)
FIGURE 2.2 Lattice models of a classic magnetic semiconductor EuO (a), and a
diluted magnetic semiconductor Ga1−xMnxAs (b).
measurements show that the electron spin polarization in these films exceeds
90%. However, no spin injection from EuO into either substrate has been
reported to date. It should be noted that EuO has a Curie temperature, Tc, of
only 69 K, a common shortcoming of most FMS compounds, which essen-
tially precludes their use in practical device applications.
Pronounced magnetic behavior can be introduced into some semicon-
ductors by incorporating a small amount (∼1%–10%) of certain magnetic
impurities into the lattice. These materials are referred to as “diluted mag-
netic semiconductors” (DMS), because the material is formed by diluting a
host semiconductor lattice with a substitutional magnetic impurity, most
typically Mn, to form an alloy (see Figure 2.2b). Strong exchange interac-
tions between the magnetic impurity and the host carriers lead to magnetic
behavior. Some of these materials are truly FM, while others are paramag-
netic yet still exhibit very useful and highly flexible magnetic properties.
The paramagnetic compounds are commonly based on the II–VI or
IV–VI semiconductors, where there is a high degree of solubility for the
magnetic element. Classic examples include Zn1−xMnxSe, Cd1−xMnxTe, and
Pb1−xMnxSe. Comprehensive reviews may be found in Refs. [25, 26]. These
materials are often called “semimagnetic semiconductors” (SMS) because
their paramagnetic behavior and Zeeman splitting of the CB and VB are
dramatically enhanced over what might be expected from the magnetic ions
alone. This enhancement originates from very large exchange interactions
between the s- and p-like carriers of the CB and VB of the host, and the d
electrons of the substitutional magnetic impurity. This sp-d exchange leads
to a tremendous amplification of the Zeeman splitting of the band edges in
an applied magnetic field, and produces amplified magneto-optical proper-
ties such as giant Faraday rotation, as well as other field dependent effects.
If this Zeeman splitting is parametrized by ΔE = g*uBH, then the effective
value of the host g-factor is increased from 2, typical of most hosts to a value
of g* ∼ 100 or greater [25]. For modest fields, the spin splitting significantly
exceeds kBT at low temperature. For example, the splitting of the mj = +1/2
and −1/2 electron states in Zn0.94Mn0.06Se is ∼10 meV at 3 T and 4.2 K, so that
the CB effectively forms a completely polarized source of mj = −1/2 electrons,
and this spin polarization can be tuned by the external magnetic field. These
properties are certainly attractive for a spin-injecting contact.
Oestreich et al. initially proposed the use of an SMS as a spin-injecting
contact [27]. They used time-resolved photoluminescence (PL) to demon-
strate that optically excited carriers became spin aligned in a Cd1−xMnxTe
layer on a picosecond time scale, and that the spin-polarized electrons were
transferred into an adjacent CdTe layer with little loss in spin polarization.
However, this pronounced behavior is limited to low temperatures (T < 25 K)
and relatively highly applied magnetic fields (H > 0.5 T), making these para-
magnetic semiconductors less attractive for device fabrication. Nevertheless,
bulk crystals have found application as magnetically tunable optical polar-
izers operating at room temperature [28], where the long optical path length
compensates for the severe reduction in g-factor and magnetization at higher
temperatures.
A concerted effort was made to introduce magnetic order into semi-

conductor compounds already recognized for device applications by alloy-
ing small amounts of Mn with GaAs and InAs to form new DMS materials.
Nonequilibrium growth at relatively low substrate temperatures by MBE
permits incorporation of Mn at levels well above the solubility limit of the
host lattice. Pioneering work led to the discovery of spontaneous FM order
in DMS alloys such as In1−xMnxAs in 1989 [29] and Ga1−xMnxAs in 1996
[30, 31]. Since Mn acts as both the magnetic element and an acceptor, they are
p-type. After much research [32–34], these new FMS materials now exhibit
Curie temperatures up to 82 K [35] and 200 K [36], respectively. Although
their optical and electronic properties are not nearly as clean and controllable
as their nonmagnetic hosts, these materials have been vigorously studied for
their potential in future spin-dependent semiconductor device technologies.
For example, Ga1−xMnxAs has been used as a source of spin-polarized holes
in resonant tunneling diodes (RTD) [37, 38], LED [39, 40], and for current-
induced magnetization switching [41]. Electric field control of FM order has
been demonstrated in In1−xMnxAs [42], MnxGe1−x [43], and Ga1−xMnxAs [44]
heterostructures, demonstrating one of the highly unusual multiferroic prop-
erties of these materials and portending a host of new applications.
The theory based on a mean field Zener model has predicted that FM
order should be stabilized in a wide variety of semiconductor hosts when
diluted or alloyed with Mn at concentrations of ~5% and sufficiently high
hole densities [45, 46]. Subsequent work has indicated that other magnetic
atoms should produce similar effects. This has stimulated a groundswell of
research activity to synthesize new diluted FMS compounds, with the goal
of achieving technologically attractive materials [47]. Unfortunately, prog-
ress has been hampered on both the theoretical and experimental fronts.
Although the mean field model works well for Ga1−xMnxAs for which it was
developed, it is clear that it is not appropriate for many other semiconductor
hosts—the relevant mechanisms leading to potential magnetic order and the
proper choice of theoretical framework depend strongly upon the position of
the electronic states introduced by the magnetic atom relative to the band
edges of the host semiconductor. New theoretical approaches are now being
developed and appear promising [48].
The origins of ferromagnetic order in even the canonical ferromagnetic
semiconductor, Ga1-xMnxAs, are still debated and remain a topic of current
research effort [49]. Further evidence supporting the mean field Zener model
in which the Fermi energy lies within the valence band and magnetic order
is mediated by extended hole states [50] is countered by evidence that the
Fermi energy is instead located within a separate Mn-derived impurity band
of localized states which stabilize ferromagnetic order [51, 52]. The picture is
complicated by the structural disorder introduced by the presence of Mn and
the low growth temperatures required to incorporate Mn into the lattice, and
the impact of such disorder is very difficult to model. The basic physics and
potential device applications are covered in several recent reviews [53, 54].
The principal obstacle in experimental efforts to synthesize new diluted
FMS materials has been the formation of unwanted phases due to the
relatively low solubility of the magnetic atoms (Mn, Cr, Fe, Co) in most of the
hosts considered (e.g. the group III-As, III-Sb, and III-N families). In many
cases, nanoscale and microscale precipitates of known FM bulk phases form,
which are exceedingly difficult to detect by the usual structural probes of
x-ray diffraction and electron microscopy, but are all too readily detected
with standard magnetometry measurements. A major challenge facing the
experimentalist is to utilize characterization techniques which discriminate
against the potential presence of such precipitates, clearly distinguish the
contribution of the FMS material from that of the precipitates, and directly
probe the key characteristics expected for the FMS itself. Electrical spin
injection from the FMS provides one such litmus test, since it is enabled by
the net carrier spin polarization intrinsic to a true FMS.

Contacts: Initial Transport Studies
As noted earlier, the use of a magnetic semiconductor as a contact enables
design of a spin injection semiconductor device guided by known principles
of bandgap engineering and epitaxial growth. This has greatly facilitated
demonstration of electrical spin injection from such contacts.
The first report of electrical injection of spin-polarized carriers from
a diluted FMS, p-type GaMnAs, was obtained for RTD structures [37, 38].
A p-Ga0.965Mn0.035As emitter layer was grown by MBE on a GaAs/AlAs/
GaAs-QW/AlAs/p-GaAs(001) RTD structure—the corresponding VB dia-
gram is shown as the inset of Figure 2.3. The structure was biased to inject
holes from the GaMnAs emitter, and resonant tunneling occurs at bias volt-
ages where the emitter VB edge is aligned with one of the confined LH or
HH states in the GaAs QW (labeled LH1, HH2, etc.). As the temperature is
lowered below the GaMnAs Curie temperature (∼70 K), the resonant peak
labeled HH2 splits, and the temperature dependence of the splitting mirrors
that of the GaMnAs magnetization. The authors attribute this to the spon-
taneous exchange splitting of the GaMnAs VB edge below Tc and resonant
tunneling from the spin-polarized hole states, which accompanies the onset
of FM order in the emitter. Thus, by selecting the bias voltage, one can use
the RTD to preferentially transmit one hole spin state or the other. The fact
that other resonant peaks did not split was tentatively attributed to some
selection rule process, and is likely due to the orbital character/spin orienta-
tion of the GaAs QW hole states relative to the in-plane magnetization of
the GaMnAs emitter contact. Although the RTD provides some qualitative
indication of electrical spin injection, it is difficult to obtain more specific
and very basic information such as the net spin polarization achieved.
2.4.3 Spin-Polarized Light-Emitting Diode: Spin-LED

In order to study and develop electrical spin injection from a discrete con-
tact, one must first have a reliable means of both detecting the presence of
spin-polarized carriers and of quantifying the spin polarization achieved.
FIGURE 2.3 Derivative of current (dI/dV) vs. voltage (V) of a resonant tunneling
diode with FM (Ga,Mn)As emitter. The labeling indicates the relevant resonant
state in the GaAs well. When holes are injected from the (Ga,Mn)As side (positive
bias), a spontaneous splitting of resonant peaks HH2 is observed below 80 K. The
transition temperature of (Ga,Mn)As is expected to be 70 K. The splitting is attrib-
uted to spin splitting of (Ga,Mn)As VB states. (After Ohno, H. Science 281, 951, 1998.
With permission.)
A simple LED structure takes advantage of one of the distinguishing char-

acteristics of semiconductors—radiative recombination of carriers—and
provides a powerful platform for this purpose. In a normal LED, electrons
and holes recombine in the vicinity of a p–n junction or QW to produce
light when a forward bias current flows. This light is unpolarized, because
all carrier spin states are equally populated. However, if electrical injection
produces a spin-polarized carrier population within the semiconductor, the
same selection rules discussed above for optical pumping also describe the
radiative recombination pathways allowed. Inspection of Figure 2.1 reveals
that if injected carriers retain their spin polarization, radiative recombina-
tion results in the emission of circularly polarized light. A simple analysis
based on these selection rules provides a quantitative and model indepen-
dent measure of the spin polarization of the carriers participating.
A schematic of such a spin-LED [55] is shown in Figure 2.4 for surface

and edge-emitting geometries. As an example, we consider injection of spin-
polarized electrons from a magnetic contact layer, which recombine with
holes supplied from the substrate. The net circular polarization Pcirc of the
light emitted is readily determined from the measured intensities of the pos-
itive and negative helicity components of the electroluminescence (EL), I(σ+)
and I(σ−), respectively Equation 2.2. These in turn are directly related to the
occupation of the carrier states. Assuming a spin-polarized electron popu-
lation and an unpolarized hole population in a bulk-like sample—i.e. all of
the hole states are at the same energy and thus have the same probability of
being occupied—a general expression for the degree of circular polarization
in the Faraday geometry (Figure 2.4a) follows directly from Figure 2.1a. Pcirc
can be written in terms of the relative populations of the electron spin states
n↑ (mj = +1/2) and n↓ (mj = −1/2), where 0 ≤ n ≤ 1, and n↑ + n↓ = 1.
Pcirc =
[ I (s+) - I (s-)]
[ I(s+) + I (s-)]
(n¯ - n- )
= 0.5 (2.2)
(n¯ + n- )
= 0.5 Pspin .
The optical polarization is directly related to the electron spin polariza-
tion Pspin = (n↓ − n↑)/(n↓ + n↑) at the moment of radiative recombination, and
has a maximum value of 0.5 due to the bulk degeneracy of the HH and LH
bands.
In a QW, the HH and LH bands are separated in energy by quantum
confinement, which modifies Equation 2.2 and significantly impacts the
analysis. The HH/LH band splitting is typically several meV even in shal-
low QWs, and is much larger than the thermal energy at low temperature
(∼0.36 meV at 4.2 K), so that the LH states are at higher energy and are not
occupied (Figure 2.1b). For typical AlxGa1−xAs/GaAs QW structures with
Al concentration 0.03 ≤ x ≤ 0.3 and a 15 nm QW, a simple calculation yields
a value of 3–10 meV for the HH/LH splitting [56]. Thus, only the HH levels
hv
hv
QW
(a) (b)
FIGURE 2.4 Schematic of spin-LEDs showing relative orientation of electron

spins and light propagation direction appropriate for deducing the spin polariza-
tion of the electrons participating in the radiative recombination process for (a)
surface-emitting and (b) edge-emitting geometries. The holes are assumed to be
unpolarized in these examples.
participate in the radiative recombination process at low temperature, as

shown in Figure 2.1b, and Pcirc is calculated as before.
(n¯ - n- )
Pcirc = = Pspin . (2.3)
(n¯ + n- )
In this case, Pcirc is equal to the electron spin polarization in the well, and can
be as high as 1.
The use of a QW offers several distinct advantages over a p–n junction
in this approach. The QW provides a specific spatial location within the
structure where the spin polarization is measured, and hence depth resolu-
tion. Varying the distance of the QW from the injecting interface may then
provide a measure of spin transport lengths. This feature was utilized by
Hagele et al. to obtain a lower bound of 4 µm for spin diffusion lengths in
optically pumped GaAs at 10 K [57]. In addition, the light emitted from the
QW has an energy characteristic of the QW structure, and may therefore be
easily distinguished from spectroscopic features arising from other areas of
the structure or impurity-related emission.
Note that the carrier spin polarization, Pspin, determined by this proce-
dure is the spin polarization at the instant and location at which radiative
recombination occurs in the structure (Figure 2.5). After injection at the
interface, the spin-polarized carriers (a) must first transport some distance
to the point of radiative recombination (e.g. to the QW), and then (b) wait
some time τr characteristic of the structure before radiative recombination
occurs, where τr is the radiative lifetime. Spin relaxation is likely to occur
during both of these processes—scattering events during transport lead
to loss of polarization, and the spin polarization decays exponentially with
time as exp(−t/τs), where τs is the spin lifetime (see Chapter 1, Volume 1
for a detailed discussion of spin relaxation). It is difficult to correct for (a)
to obtain the spin polarization at the point of injection, Pinj, without very
detailed knowledge of the interface structure and relevant scattering mecha-
nisms. However, a straightforward procedure can be used to correct for (b).
Psource Pinj Pcirc

Po
CB
τr
τs
VB
FIGURE 2.5 Band diagram illustrating lifetimes and polarizations in a spin-LED.

The electron spin polarization in the electrical contact, Psource, results in some net
injected spin polarization Pinj just inside the semiconductor. Through drift and
diffusion, this electron spin reaches the QW and produces a spin polarization Po.
After radiative recombination, this is manifested as a circular polarization in the EL,
Pcirc. If the VB degeneracy is lifted, Pcirc = Pspin. The initial QW spin polarization can be
obtained from a rate equation analysis, Po = Pspin (1 + τr/τs).
Essentially, one takes a snapshot of the spin system with τr as the shutter
speed. If τr was much shorter than τs, this would provide an accurate mea-
sure of Pspin in much the same way that a short exposure time or shutter
speed is used to capture action photographs with a camera. This is rarely
the case, however, and the spin system decays over the time τr . Therefore,
the result measured represents a lower bound for the carrier spin polariza-
tion achieved by electrical injection. A simple rate equation analysis may be
applied to obtain a more accurate measure of the initial carrier spin polariza-
tion, Po, that exists when the electrically injected carriers enter the region of
radiative recombination (e.g. the QW). Po is given by [58]
æ t ö
Po = Pspin ç 1 + r ÷ , (2.4)
è ts ø
where Pspin is the value determined experimentally as described above.
This provides a first-order correction for what is effectively the instrument
response function of the LED to serve as a spin detector and the efficiency of
the radiative recombination process in the particular structure and material
utilized.
The value of τr/τs can be determined by a simple PL measurement using
near band-edge circularly polarized excitation (optical pumping) [5, 58].
Assuming that both LH and HH states are excited, the initial spin polariza-
tion produced by optical pumping is Po = 0.5 (see Figure 2.1a). One measures
the circular polarization of the PL, PPL, and uses Equation 2.4 to solve for
τr/τs: 0.5 = PPL (1 + τr/τs). Typical values are 1 ≤ τr/τs ≤ 10, so that this correc-
tion can be significant.
Thus, the spin-LED serves as a polarization transducer, effectively con-
verting carrier spin polarization, which is difficult to measure by any other
method, to an optical polarization, which can be easily and accurately mea-
sured using standard optical spectroscopic techniques. The existence of
circularly polarized EL demonstrates successful electrical spin injection
(subject to appropriate control experiments), and an analysis of the circu-
lar polarization using these fundamental selection rules provides a quanti-
tative assessment of carrier spin polarization in the QW without resorting
to a specific model. Note that this approach measures the spin polarization
of the carrier population in the semiconductor (electron or hole) achieved
by electrical injection, and not the spin polarization of the injected current.
Detailed knowledge of the transport mechanism and spin scattering is nec-
essary to connect the two.
A number of conditions facilitate application of the spin-LED approach:
1. The analysis of the measurements is considerably more reliable if

the experimental geometry is appropriate for the optical selection
rules [5, 13]. The hole spin, electron spin, and the optical emission/
analysis axes (or projections thereof) must be colinear to extract
information on carrier spin polarization from the circular-polar-
ized EL. The hole states in a QW have preferred orientations due to
quantum confinement and reduced symmetry [59, 60]: the k = 0 HH

orbital angular momentum is oriented entirely along the growth
direction (z-axis), whereas the k = 0 LH orbital angular momentum
has non-zero projections in all three directions. Analysis of the
HH QW exciton requires that the injected electron spin must be
along the z-axis and the optical measurement must also be per-
formed along the same axis, meaning that surface-emitted rather
than edge-emitted light should be analyzed (Figure 2.4a) [61]. Thus,
the Faraday rather than the Voigt geometry must be used. If the
injected spins are oriented in-plane, then an edge-emission geom-
etry (Figure 2.4b) may be utilized to analyze the EL if the radiative
recombination region is bulk-like, or if one measures the LH exci-
ton in a QW. An example of spin analysis for edge emission from a
spin-LED is provided in Ref. [62].
2. The radiative recombination region should not be highly strained.
Strain modifies the selection rules, compromising the quantita-
tive relationship between Pcirc and Pspin. Strain may produce optical
polarization in the absence of carrier spin polarization, or reduce Pcirc
below that expected from the corresponding Pspin. While no system is
perfect due to lattice mismatch, thermal expansion coefficients, and
contact issues, the AlxGa1−xAs/GaAs QW system comes very close
due to the very small variation of lattice constant with Al concentra-
tion. In contrast, the lattice constant varies rapidly with In concen-
tration, and strain is almost unavoidable in the GaAs/InxGa1−xAs QW
system. Other factors complicate the use of InGaAs in spin-LEDs—
the introduction of In, with its stronger spin–orbit interaction, may
reduce spin lifetimes [63], and the relatively large g-factor introduces
significant magnetic field-dependent effects, which complicate inter-
pretation of the EL polarization [64].
3. The origin of the EL must be correctly identified to determine its
spatial origin within the structure, and to confirm that it derives
from recombination processes for which the selection rules are valid
[65, 66], i.e. spin-conserving processes such as free exciton or free
electron recombination. In practice, this means that the EL must be
spectroscopically resolved and standard analyses applied to assist
in the identification of the emission peaks. Note that bound exciton
and impurity-related emission typically involve non-spin-conserving
processes, and therefore cannot be used.
4. The VB degeneracy must be correctly identified, as described above.
In a QW, the LH/HH splitting can be readily calculated for many
materials [56], and should be compared to the measurement tem-
perature to determine their relative contributions.
To briefly summarize, a quantitative and model independent determi-

nation of the initial spin polarization Po can be obtained from a spin-LED
experiment if a few straightforward conditions are met: one employs the
proper experimental geometry, identifies the radiative transition producing
the EL, and determines a value for τr/τs for the samples under consideration.
This is all relatively easy to do in the Al xGa1−xAs/GaAs QW system, and, with
some additional care, in the GaAs/InxGa1−xAs QW system.

Contacts: Spin-LED Studies
2.4.4.1 Zn1−xMnxSe
As an example illustrating the concepts and procedures above, we consider
electrical spin injection from a diluted magnetic semiconductor contact,
n-type Zn0.94Mn0.06Se, into an AlGaAs/GaAs QW/AlGaAs LED structure,
as first described by Fiederling et al. [67] and Jonker et al. [68]. Zn1−xMnxSe
was chosen as the contact material for a number of reasons. It can read-
ily be doped n-type, allowing one to focus on electron transport, and the
giant Zeeman splitting described above provides an essentially 100% spin-
polarized electron population at modest applied magnetic fields, albeit at low
temperature. Zn1−xMnxSe forms high-quality epitaxial films on GaAs due to
a close lattice match, and the CB offset can be tailored to facilitate electron
flow from the Zn1−xMnxSe into the AlyGa1−yAs/GaAs structure by suitable
choice of the Mn and Al concentrations [68].
A flat band diagram of the structure is shown in Figure 2.6 illustrat-
ing the band alignments and the spin splitting of the CB and VB edges of
the Zn1−xMnxSe with applied magnetic field. The ZnMnSe spin polariza-
tion may be varied simply by varying the applied field, a very useful handle
for experimental studies. At sufficiently high fields, the CB spin splitting
(ΔE = g*uBH ∼ 10 meV) is much larger than the measurement temperature
(∼0.4 meV), so that the Zn1−xMnxSe contact is essentially 100% spin-polar-
ized. The samples were grown by MBE and processed into surface-emitting
LED mesas 200–400 µm in diameter using standard photolithographic tech-
niques. Figure 2.7 shows a schematic cross-section and photograph of the
final devices. The top metallization to the ZnMnSe consists of concentric Au
rings to help insure uniform current distribution, leaving most of the mesa
ZnMnSe GaAs
+1/2 QW
–1/2 CB
AlGaAs AlGaAs
n– p+
VB
–3/2
+3/2
FIGURE 2.6 ZnMnSe/AlGaAs spin-LED band diagram, illustrating the giant

Zeeman splitting of the ZnMnSe band edges and electrical injection of spin-polar-
ized electrons into the GaAs QW.
100 µm —
Base
contact
Top
contact
Metal contact
LED mesa
ZnMnSe
Insulator AlGaAs
GaAs QW
p-GaAs
p-GaAs
p-GaAs
GaAs(001)
(a) (b)
FIGURE 2.7 (a) Schematic cross-section of spin-LED illustrating the structure

and processing steps. (b) Photograph of completed surface-emitting devices. The
active mesa areas (dark circular regions) are 200, 300, and 400 µm in diameter.
(After Ohno, H. Science 281, 951, 1998. With permission.)
surface optically transparent. Details of the MBE growth and device fabrica-
tion may be found in reference [68].
The EL was measured by electrically biasing the LEDs to inject
electrons from the n-ZnMnSe into the GaAs QW at current densities
of ∼0.01–1.0 A cm−2 . Representative EL spectra from such a structure—
(in this case with a multiple QW LED (3 × (20 nm Al0.08Ga0.92As/10 nm
GaAs))—are shown in Figure 2.8 for selected values of the applied field.
The light emitted along the surface normal and magnetic field direction
(Faraday geometry) was analyzed for σ+ and σ− circular polarization and
spectroscopically resolved. The energy of the emission confirms that the
radiative recombination occurs in the GaAs QW via the HH ground state
exciton (note that other tests are applied to confirm this identification—
see Ref. [65]). As noted in the previous section (condition (1)), quantum
confinement locks the HH spins along the z-axis (surface normal), and
therefore, both the injected electron spins and the axis for optical analysis
must be oriented along the z-axis as well to interpret any circular polar-
ization observed in terms of spin polarization. Edge emission will give a
spurious or null result [61].
At zero field, the σ+ and σ− components are identical, as expected,
since no spin polarization yet exists in the ZnMnSe. As the magnetic field
increases, the ZnMnSe bands split into spin states due to the giant Zeeman
effect, and spin-polarized electrons are injected into the AlGaAs/GaAs LED
structure. The corresponding spectra exhibit a large difference in intensity
between the σ+ and σ− components, demonstrating that the spin-polar-
ized electrons successfully reach the GaAs QW. The circular polarization,
Pcirc = [I(σ+) − I(σ−)]/[I(σ+) + I(σ−)], increases with applied field as the ZnMnSe
T = 4.5 K
σ+
σ–
EL intensity (a.u.)
4T
2T
0.5 T
0T
1.520 1.530 1.540 1.550 1.560 1.570

Photon energy (eV)
FIGURE 2.8 EL spectra from a surface-emitting spin-LED with a Zn0.94Mn0.06Se

contact for selected values of applied magnetic field, analyzed for positive (σ+) and
negative (σ−) circular polarization. The magnetic field is applied along the surface
normal (Faraday geometry). The spectra are dominated by the HH exciton. Typical
operating parameters are 100 µA and 2.5 V.
polarization increases, and saturates at a value of ∼80% at 4 T. From the dis-
cussion leading to Equation 2.3, the spin polarization of the electron popula-
tion in the GaAs QW, Pspin, is therefore 80%.
This value is the QW spin polarization at the time of radiative recom-
bination, i.e. after a time ∼τr has elapsed. A value for the initial QW spin
polarization, Po, can be obtained using Equation 2.4 to correct for the effect
that the relative values of the radiative and spin lifetimes have on I(σ+) and
I(σ−). Independent PL measurements using circularly polarized excitation
provide a value of τr/τs, as described previously. The PL is analyzed for σ+ and
σ− helicity, and exhibits a polarization PPL = 40%. Applying 0.5 = PPL (1 + τr/τs)
gives τr/τs = 0.25 ± 0.05. Equation 2.4 then gives Po ∼ 100%, demonstrating
that a well-ordered ZnMnSe/AlGaAs interface is essentially transparent to
spin transport. Thus, an all-electrical process using a discrete contact can
produce a carrier spin polarization within the semiconductor that equals or
exceeds that achieved via optical pumping.
Several control experiments rule out contributions to the circular
polarization measured from spurious sources, and are an essential part of
any such experiment. LEDs with nonmagnetic n-ZnSe contact layers show
no circular polarization with magnetic field, as expected. The circular
dichroism resulting from transmission through the ZnMnSe is negligible

because the GaAs QW emission wavelength is very far from that corre-
sponding to the band gap of Zn0.94Mn0.06Se. PL data from the GaAs QW
excited with linearly polarized light from the same LED mesa structures
used for the EL studies show little polarization, providing a very effec-
tive built-in reference for each mesa LED. Such dichroism effects could be
much larger for emission energies very near the ZnMnSe band gap, where
strong absorption occurs, and must be considered when designing the LED
structure.
2.4.4.2 Ga1−xMnxAs
Spin-LED structures have also been fabricated with p-type GaMnAs as
the spin contact. The first report [39] utilized a 300 nm Ga0.955Mn0.045As/
GaAs/10 nm In0.13Ga0.87As QW/n-GaAs(001) heterostructure in which
spin-polarized holes injected from the GaMnAs radiatively recombined in
the strained InGaAs QW with unpolarized electrons from the substrate.
Because the magnetization of the GaMnAs (and nominal spin orientation
of the injected holes) was in-plane, they used an edge-emission geometry
with the optical axis parallel to the magnetization. They measured the EL
at the QW ground state transition, and observed a 1% polarization, which
exhibited the same temperature dependence as the GaMnAs magnetiza-
tion. A reliable interpretation in terms of hole spin injection and polariza-
tion, however, is compromised by the use of a strained QW as the radiative
recombination region—as discussed previously, (1) the orbital angular
momentum of the QW HH ground state is oriented along the growth direc-
tion (z-axis), orthogonal to the nominal in-plane orientation of the spins
injected from the GaMnAs, and (2) strain in the QW leads to admixture
of states and modifies the selection rules. Inducing an out-of-plane mag-
netization in the GaMnAs (e.g. by applying a magnetic field) or utilizing
a structure with a bulk-like recombination region (Figure 2.4b) may have
alleviated these issues.
More definitive measurements were performed on identical sample
structures in 2002 using the Faraday geometry (surface emission) to address
the first issue noted above [40, 69]. The EL spectra at T = 5 K and the field
dependence of the circular polarization for selected temperatures are shown
in Figure 2.9. The field dependence tracks the out-of-plane (hard axis) magne-
tization of the GaMnAs (Figure 2.9b), and the saturation value depends upon
the thickness, d, of the undoped GaAs spacer layer between the GaMnAs
injector and the InGaAs QW. The circular polarization is 4% for d = 70 nm,
and increases to 7% for d = 20 nm, attributed to the hole spin diffusion length
effects in the GaAs spacer. These data indicate that the spin polarization
of the QW hole population is at least 7% at the measurement temperature.
The polarization decreases with increasing temperature, and disappears by
62 K—the Curie temperature of the GaMnAs injector. In Ref. [69], the use of
small mesas to force the easy axis of the GaMnAs contact out of plane per-
mitted the demonstration of hole spin injection without an applied magnetic
field in the surface emission geometry.
4
I = 8.5 mA
d = 70 nm
T=5 K
2 T = 16 K
T = 28 K
∆ Polarization (%)
T = 50 K
100 T = 62 K
T = 5 K, d = 70 nm H (Ga,Mn)As (p) 0
EL intensity (a.u.)
GaAs 3
I = 8.5 mA spacer d (i)
M|| (µB/Mn)
I = 4.5 mA (In,Ga)As QW 2 Tc = 62 K
10 –2
I = 2.5 mA GaAs (i)
1
GaAs (n)
GaAs substrate (n) 0
0 20 40 60 80
–4 T (K)
1
1.25 1.30 1.35 1.40 1.45 1.50 1.55 –5.0 –2.5 0.0 2.5 5.0
(a) Energy (eV) (b) H (kOe)
FIGURE 2.9 (a) Spectrally resolved EL intensity along the growth direction for
several bias currents, I. Inset shows device schematic and EL collection geometries.
(b) Temperature dependence of the relative changes in the energy-integrated
[gray shaded area in (a)] polarization ΔP as a function of out-of-plane magnetic
field. When T < 62 K, polarization saturates at H⊥ ∼ 2.5 kOe. Inset shows M(T), indi-
cating that the polarization is proportional to the magnetic moment. (After Young,
D.K. et al., Appl. Phys. Lett. 80, 1598, 2002. With permission.)
2.4.4.3 Spin Scattering by Interface Defects

Spin scattering by interface defects is an important issue in heteroepitaxial
systems, and can rapidly suppress the spin polarization achieved by elec-
trical injection. A comprehensive understanding of this process remains
a major challenge for future research. Initial work addressed the effect of
interface nanostructure on the spin injection efficiency at the ZnMnSe/
AlGaAs interface [70]. High resolution transmission electron microscopy
(TEM) was used to assess the interface morphology in spin-LED structures,
and found that the most prevalent defects were stacking faults (SF) in <111>
directions nucleating at or near the ZnMnSe/AlGaAs interface, where they
formed line defects (Figure 2.10a). The electron spin polarization in the
GaAs QW determined from the circular polarization Pcirc of the EL exhib-
ited an approximately linear dependence on the density of interface defects,
as shown in Figure 2.10b. This correlation was explained by a model that
incorporated spin–orbit (Elliot–Yafet) scattering—the model showed that
the asymmetric potential of the interface defect results in strong spin-flip
scattering in the forward direction. A simple expression, Pcirc ∼ 1 − 4ron, gave
excellent agreement with the experimental data with no adjustable param-
eters, where ro ∼ 10 nm is the Thomas–Fermi screening length and n is the
measured defect density. These results provided the first experimental dem-
onstration that interface defect structure limits spin injection efficiency
in the diffusive transport regime. It is interesting to note that strong spin
injection persists even when the misfit dislocation density greatly exceeds
values which would be fatal for conventional devices such as III–V-based
diode lasers and field effect transistors. This is especially reassuring, since
100
90
QW spin polarization (%)

80
70
60
50
40
30
20
10
0
0 50 100 150 200
(a) (b) Defect density (103 cm–1)
FIGURE 2.10 (a) Diagram illustrating the linear interface defects resulting from
the intercept of (111)-type SF planes and the interface plane. Only one of the four
possible (111)-type SF planes is shown for clarity. (b) Correlation of GaAs QW spin
polarization with stacking fault density. Two data points nearly overlap at 85%
polarization. The error bars are comparable to the symbol size. The dashed line is
the calculated result with no adjustable parameters.
interface misfit dislocations are a generic defect routinely encountered in

heteroepitaxial device structures.
2.5 FERROMAGNETIC METAL/SEMICONDUCTOR
ELECTRICAL SPIN INJECTION: OPTICAL
DETECTION WITH SPIN-LED STRUCTURES
FM metals offer most of the properties desired for a practical spin-injecting
contact material: reasonable spin polarizations at EF (∼40%–50%), high Curie
temperatures, low coercive fields, fast switching times, and a well-developed
material technology due to decades of research and development driven in
large part by the magnetic storage industry. An FM contact introduces non-
volatile and reprogrammable operation in a very natural way, due to both
the material’s intrinsic magnetic anisotropy and the ability to tailor the mag-
netic characteristics in numerous ways. Spin-torque transfer switching is
rapidly emerging as the mechanism of choice for manipulating the contact
magnetization, and offers many advantages for spintronic applications, such
as embedded memory or field programmable gate arrays (see Chapters 7, 8,
Volume 1, and Chapter 4, Volume 2 for an overview). This emerging tech-
nology can readily be incorporated in semiconductor spintronic devices. In
addition, metallization is a standard process in any semiconductor device
fabrication line, so that an FM metallization could easily be incorporated into
existing processing schedules.
Initial efforts to inject spin-polarized carriers from an FM metal contact
into a semiconductor were not very encouraging. Several groups reported a
change in voltage or resistance on the order of 0.1%–1%, which they attrib-
uted to spin accumulation or transport in the semiconductor [71–73]. Such
small effects, however, make it difficult to either unambiguously confirm spin
2.5 Ferromagnetic Metal/Semiconductor Electrical Spin Injection 81
injection or successfully implement new device concepts. In addition, some

have argued that these measurements were compromised by contributions
from anisotropic magnetoresistance, or a local Hall effect, which can easily
contribute a signal of 1%–2% [74–76]. Thus, great care must be taken in the
experimental design to recognize and eliminate such spurious contributions.
Zhu et al. [77] utilized optical rather than electrical detection in a spin-
LED structure (the spin-LED is described in Section 2.4.3) consisting of an
Fe Schottky contact to a GaAs/In0.2Ga0.8As QW LED detector. They were
unable to observe any clear difference in intensity when they analyzed the
EL for positive (σ+) or negative (σ−) helicity polarization, in contrast with
earlier results for a ZnMnSe magnetic semiconductor contact, as described
in Section 2.4.4.1. However, by examining the high and low energy tails of
the EL peak (which they attributed to LH and HH contributions, respec-
tively) using pulsed current injection and lock-in detection techniques, with
some background subtraction, they identified a signal which they attributed
to electrical spin injection from the Fe contact. They concluded that a spin
polarization of ∼2% had been achieved in the GaAs, and found that this
signal was independent of temperature from 25 to 300 K. The lack of tem-
perature dependence raises some question as to their interpretation of this
signal, since the spin lifetime is known to decrease rapidly with temperature
for GaAs(001) structures [78].
Two fundamental issues are key to understanding, successfully demon-
strating, and optimizing electrical spin injection from an FM metal into a
semiconductor—the role of band symmetries in facilitating spin transmis-
sion across a heterointerface, and the impact of the large difference in con-
ductivity between the metal and semiconductor.
2.5.1 Band Symmetries and Spin Transmission

The first issue can be stated rather simply: the symmetries and orbital com-
position of the bands participating in the transport process in the metal must
be compatible with those of the semiconductor for optimum spin trans-
mission to occur. Theoretical work has explored the electronic structure
at the FM metal/semiconductor interface in an effort to elucidate the role
of band structure in spin injection [79–82]. This work has emphasized the
importance of matching the symmetries as well as the energies of the bands
between metal and semiconductor to optimize spin injection efficiency.
Some initial insight can be gained by examining the bulk band struc-
tures. Figure 2.11 shows selected portions of the band structure along (001)
for Fe, GaAs [82, 83], and Si [84]. For typical semiconductors of interest,
including GaAs, InAs, GaP, ZnSe, and Si, the CB for the (001) surface exhib-
its Δ1 symmetry near the zone center (dashed curves), with significant s and
pz-orbital contributions. The Fe majority spin band (dashed curve), which
crosses EF midway across the zone, is also of Δ1 symmetry, with significant
s-, pz-, and dz2 -orbital contributions. The s and pz components have large spa-
tial extent, and the pz and dz2 orbitals point directly into the semiconduc-
tor, leading to strong overlap with the corresponding states comprising the
Fe GaAs Si
13
12 Γ2΄
1 ∆΄2
5 Γ15
11 2 2΄
Energy (eV)
X1
∆1 ∆1
10
1
9 5 Γ΄25
2 ∆5
8 2΄
X4
7
Γ ∆ H Γ ∆ X Γ X
FIGURE 2.11 Band structures along (001) for Fe, GaAs, and Si. The majority spin
band of Fe crossing the Fermi level is of Δ1 symmetry (labeled “1” in the plot,
dashed curve), and the CB edge of both GaAs and Si (dashed curve) is also of Δ1
symmetry. The Fe minority spin bands are of different symmetry (Δ2, Δ5, labeled
“2” and “5”). This is expected to enhance transmission of majority spin electrons
from the Fe, and suppress minority spin injection, leading to highly polarized
spin currents in the semiconductor. The plots for Fe and GaAs are taken from Ref.
[76]—the underestimate of the GaAs band gap is characteristic of the computa-
tional approach used. The plot for Si is derived from Ref. [78], and the energy scale
is expanded relative to that used for Fe and GaAs. (After Wunnicke, O. et al., Phys.
Rev. B 65, 241306(R), 2002; Chelikowsky, J.R. et al., Phys. Rev. B 14, 556, 1976. With
permission.)
semiconductor CB. The corresponding Fe Δ1 minority spin band is 1.3 eV

above EF and therefore will not contribute to transport. In contrast, the Fe
minority spin bands which cross EF, exhibit quite different symmetries (Δ2, Δ5),
with orbital components that do not couple strongly to the semiconductor.
Thus, the CB of the semiconductor is well-matched in orbital composition/
symmetry and energy to the majority spin bands of the Fe, facilitating propa-
gation of majority spin electrons, while a poor match exists to the Fe minor-
ity spin bands.
Theoretical treatments have addressed several FM metal/semiconductor
systems, and generally assume a well-ordered interface so that the electron
momentum parallel to the interface plane, k//, is conserved to simplify the cal-
culation. For the case of spin injection from Fe into GaAs(001), calculations
show that these band symmetries play a critical role, leading to a significant
enhancement of transmission from the Δ1 Fe majority spin band at EF, and
a suppression of transmission from the Fe minority spin bands (Δ2, ∆ 2′ , Δ5)
[82, 85]. Consequently, majority spin electrons are preferentially transmit-
ted from Fe into the GaAs, while minority spins are blocked, producing a
significant enhancement of spin injection efficiency. In such cases, the metal/
semiconductor interface essentially serves as a band structure spin filter due
to basic issues of band symmetry and orbital composition.
Similar arguments can be made for the case of Fe and Si(001). As seen
in Figure 2.11, the bottom of the Si(001) CB is also of Δ1 symmetry, enhanc-
ing transmission of Fe Δ1-band majority spin electrons, and suppressing
minority spin current. Calculations assuming an ideal Fe/Si(001) interface

indicated that the current injected from the Fe contact should be strongly
spin-polarized [86]. Similar conclusions were obtained for Fe/InAs(001) [87],
and earlier for epitaxial Fe/MgO tunnel barriers by Butler et al. [88].
The band bending which occurs at the metal/semiconductor interface
complicates the picture, but can be included in the calculation to some extent.
Wunnicke et al. [82] and Mavropoulos [86] simulated the effect of a Schottky
barrier by introducing a potential step between the Fe and the semiconduc-
tor, and found that the current remained strongly polarized.
Thus, the knowledge of the interface band structure is important for
understanding and optimizing spin injection efficiency. The physical struc-
ture of the interface is a key ingredient here. The perfectly abrupt interface
typically assumed for calculation is unlikely to exist in practice. Intermixing
leading to compound formation or disorder will alter the band symmetries
and strongly impact the arguments above. Symmetry breaking due to dis-
order at the Fe/InAs(001) interface in the form of Fe atoms occupying In or
As interface sites was shown to rapidly suppress the spin filtering effect and
resultant high spin polarization in the InAs by opening more channels for
minority spin transport [81].
It should be noted that compound formation per se is not necessar-
ily detrimental, provided that bands with appropriate symmetries and/or
orbital composition are preserved or created. It is indeed possible that the
spin injection efficiency at certain interfaces with poorly matched band sym-
metries will improve with compound formation which produces bands of
more favorable orbital composition. However, disorder due to random inter-
mixing, poorly ordered compound formation, or defects will, in general,
lead to mixing of states, compromising any state-specific transmission, and
reducing the spin injection efficiency.
2.5.2 Conductivity Mismatch: Description of the Problem

The second fundamental issue presenting an unanticipated challenge to uti-
lizing an FM metal as a spin-injecting contact on a semiconductor is the
very large difference in conductivity between the two materials. Simply
stated, the ability of the semiconductor to accept carriers is independent of
spin, and much smaller (lower conductivity) than that of the metal to deliver
them. Consequently, equal numbers of spin-up and spin-down electrons are
injected, regardless of the FM metal’s initial polarization, resulting in essen-
tially zero spin polarization in the semiconductor.
This issue, commonly referred to in the literature as the problem of
“conductivity mismatch,” can readily be understood in the context of the
canonical two-channel model typically used to describe spin-polarized cur-
rent flow in FM metals [89]. In this model, spin-up or “majority spin” elec-
trons flow in one channel, while spin-down or “minority spin” electrons flow
in the other.* Using the analogy of water flow through a hose for electrical
* For a discussion of this terminology, see Ref. [90].

(a)
LF/σ 1/Σ 2Ls/σs

I
I Contact I
FM S
interface
I
(b) LF/σ 1/Σ 2Ls/σs
FIGURE 2.12 (a) Schematic of the two-channel model of spin transport illustrat-
ing the conductivity mismatch issue, which must be considered for spin injection
from an FM metal into a semiconductor. A spin-selective interface resistance pro-
vides a solution, as shown in the lower panel. (b) Equivalent resistor model (due to
A.G. Petukhov). (After Jonker, B.T. et al., MRS Bull. 28, 740, 2003. With permission.)
resistance, the relative conductivities of these channels in each material can

be represented by the diameter of the hose, as shown in Figure 2.12a. In the
metal, the diameter of both spin-up and spin-down hoses is very large (high
conductivity), while in the semiconductor, the diameter is very small (low
conductivity). The problem of conductivity mismatch and its impact on spin
injection is thus reduced to the intuitive physical picture of attempting to
effectively transfer water from a sewer pipe to a drinking straw.
In the FM metal, the carriers are partially spin-polarized (by definition),
so that one of the pipes is nearly full while the other may be nearly empty. In
the nonpolarized, low carrier density semiconductor, the two spin channels
are of equal diameter and partially full. If one now considers the flow of water
(spin-polarized current) from the sewer pipes (metal) to the drinking straws
(semiconductor), it is immediately apparent that the comparatively small con-
ductivity of the semiconductor limits current flow. While one drop of water
may be transferred from the majority spin sewer pipe of the metal to the cor-
responding majority spin drinking straw of the semiconductor, an identical
amount will flow in the minority spin channel, even though the metal pipe
may be nearly empty, due to the limited conductivity of the semiconductor.
This results in zero spin polarization in the semiconductor. It is apparent that
this will be the case regardless of how nearly empty the metal minority spin
band may be, due to the very limited ability of the semiconductor bands to
accept current flow. This is the essence of the conductivity mismatch issue.
This simple picture provides rather surprising insight into efforts to
transfer spin-polarized carriers between these two materials. In the diffusive
transport regime, significant spin injection can occur for only two conditions:
(1) the FM metal must be 100% spin-polarized (minority channel completely
empty), or (2) the conductivity of the FM metal and semiconductor must
be closely matched. No FM metal meets either of these criteria. While half-
metallic materials offer 100% spin polarization in principle [91, 92], defects
such as antisites or interface structure rapidly suppress this value [93].
Model calculations by several groups [76, 94–98] have addressed this
issue and provide a quantitative treatment. These authors extended the work
of van Son et al. [99] and Valet and Fert [89] to calculate the spin polariza-
tion achieved in a semiconductor due to transport of spin-polarized carriers
from the ferromagnet. An equivalent resistor circuit explicitly incorporates
the conductivities of the FM metal, interface, and semiconductor [11, 94], as
shown in Figure 2.12b, and permits discussion of spin injection efficiency in
terms of these physical parameters in the classical diffusive transport regime.
The resistances representing the FM metal and semiconductor are given by
LF/σF and Ls/σs, where L and σ are the spin diffusion lengths and conduc-
tivities, respectively, with σF = σ↑ + σ↓ summing the two FM spin channels.
The interface conductivity Σ = Σ↑ + Σ↓ is assumed to be spin dependent. With
some algebra and effort, the spin injection coefficient γ can be shown to be
æ Ds DS ö
ç rF + rc
æ I -I ö sF S ÷ø
g =ç - ¯ ÷= è , (2.5a)
è I ø ( rF + rS + rc )
where:
sF
rF = LF , (2.5b)
4 s- s¯
S
rc = , (2.5c)
4 S- S¯
Ls
rs = , (2.5d)
ss
are the effective resistances of the ferromagnet, contact interface, and semi-
conductor, respectively, I↑ and I↓ are the majority and minority spin current,
Δσ = σ↑ − σ↓ and ΔΣ = Σ↑ − Σ↓.
For an FM metal, rF/(rF + rs + rc) ≪ 1, and the contribution of the first
term in Equation 2.5a is negligible. The second term is significant only if
ΔΣ ≠ 0 and rc ≫ rF, rs. Thus, two criteria must be satisfied for significant spin
injection to occur across the interface between a typical FM metal and a
semiconductor: the interface resistance rc must (1) be spin selective, and (2)
dominate the series resistance in the near-interface region.
2.5.3 Conductivity Mismatch: Practical Solution

A tunnel barrier between the FM metal and semiconductor satisfies both
these criteria, and was suggested as a potential solution to the conductiv-
ity mismatch problem by several groups [76, 94–97]. The spin selectivity
(ΔΣ ≠ 0) comes naturally from the spin-polarized DOS of the FM metal at
EF, which serves as the source term in a mathematical treatment of the tun-
neling process. The resistance of the tunnel barrier is readily controlled by
its thickness, and can easily be the largest in the series resistance such that
rc ≫ rF, rs. A detailed calculation was provided by Rashba [95], and a more
comprehensive overview was provided recently by Fert et al. [100].
A tunnel barrier can be introduced at a metal/semiconductor interface
in at least two ways: tailoring the band bending in the semiconductor, which
typically leads to Schottky barrier formation, or physically inserting a dis-
crete insulating layer such as an oxide. Examples of both are presented below
for GaAs and Si.
2.5.3.1 Tailored Schottky Barrier as a Tunnel Barrier

One avenue is to take advantage of the band bending which occurs at the
metal/semiconductor interface. This approach exploits a natural character-
istic of the interface, and avoids the use of a discrete barrier layer and the
accompanying problems with pinholes and thickness uniformity. Schottky
contacts are also routine ingredients in semiconductor technology. In the
case of an n-type semiconductor, electrons are transferred into the metal,
depleting the semiconductor interfacial region and causing the CB to bend
upward, forming a pseudo-triangular-shaped barrier with a quadratic falloff
with distance into the semiconductor [101]. The depletion width associated
with the Schottky barrier depends upon the doping level of the semiconduc-
tor, and is generally far too large to allow tunneling to occur. For example,
in n-GaAs, the depletion width is on the order of 100 nm for n ∼ 1017 cm−3,
and 40 nm for n ∼ 1018 cm−3 [102]. However, this width can be tailored by the
doping profile used at the semiconductor surface [103]. Heavily doping the
surface region can reduce the depletion width to a few nanometers, so that
electron tunneling from the metal to the semiconductor becomes a highly
probable process under reverse bias.
This approach was first utilized to achieve large electrical spin injec-
tion from Fe epilayers into AlGaAs/GaAs QW LED structures [10, 104, 105].
The n-type doping profile of the surface AlGaAs was designed by solving
Poisson’s equation with several criteria in mind: (a) minimize the Schottky
barrier depletion width to facilitate tunneling; (b) use a minimum amount
of dopant and heavily doped regions to accomplish this, since high n-dop-
ing is associated with stronger spin scattering and short spin lifetimes [106,
107]; and (c) avoid formation of an electron “puddle” or accumulation region
(i.e. pushing EF above the CB edge), which may either dilute the polariza-
tion of the electrons injected from the Fe contact or contribute to spin scat-
tering. A schematic of the doping profile and resultant barrier is shown in
Figure 2.13a, and the band diagram resulting from the Poisson equation
solution for the full LED structure is illustrated in Figure 2.13b. The doping
Fe AlGaAs
100 Å 150 Å 150 Å 550 Å

~1019 Transition 1016 – 1017
(a)
2
CB
1
E – EF (eV)
0 VB
–1
–2
0 200 400 600 800 1000 1200 1400 1600 1800
(b) Depth (Å)
FIGURE 2.13 Design of an Fe Schottky barrier and spin-LED using a doping

profile to facilitate tunneling of spin-polarized electrons from the Fe through the
Schottky tunnel barrier. (a) Doping profile and resultant reduction of depletion
width at the Fe/AlGaAs interface, and (b) Poisson equation solution corresponding
to doping profile and AlGaAs/GaAs QW/AlGaAs structure described in the text.
of the top 150 Å of n-type Al0.1Ga0.9As was chosen to be n = 1 × 1019 cm−3 to

minimize the depletion width, followed by a 150 Å transition region, while
the rest was n = 1 × 1016 cm−3 with a 100 Å dopant setback at the QW. The
LED structure consisted of 850 Å n-Al0.1Ga0.9As/100 Å undoped GaAs/500
Å p-Al0.3Ga0.7As/p-GaAs buffer layer on a p-GaAs(001) substrate. The width
of the GaAs QW was chosen to be 100 Å to ensure separation of the LH
and HH levels and corresponding excitonic spectral features, an important
consideration for quantitative interpretation of the data, as discussed earlier.
Details of the growth may be found elsewhere [104, 105].
2.5.3.1.1 Confirmation of Tunneling: Analysis of the Transport Mechanism

To demonstrate that the tailored doping profile reduces the depletion width
sufficiently to produce a tunnel barrier, it is necessary to analyze the cur-
rent–voltage (I–V) characteristics of the Fe/AlGaAs Schottky contact and
apply well-known criteria to identify the dominant transport process. Such
criteria should also be applied when a discrete oxide barrier is used. The
mere addition of a layer intended to serve as a nominal tunnel barrier does
not ensure that transport occurs by tunneling—indeed, pinholes are a
chronic problem, and extensive work in the metal/insulator/metal tunnel
5
225 K
175 K
4.5
300
125 K
G (Ω/cm2)
4 75 K
200
3.5
R(T)/R(250 K)
100
3
G(V) (a.u.)
–0.1 0 0.1 LO (AlAs-like)
Voltage (V)
2.5
LO (GaAs-like)
2
1
2
1.5 d Zero-bias anomaly
1
50 100 150 200 250 –60 –40 –20 0 20 40 60
(a) Temperature (K) (b) Energy (meV)
FIGURE 2.14 (a) Inset: series of conductance curves taken at different temperatures. The dotted lines are
representative fits to the data. Parameters for the fitting are defined in the schematic of the tunnel barrier.
Normalized ZBR as a function of temperature for an Fe/Al0.1Ga0.9As Schottky barrier contact, showing minimal
temperature dependence consistent with tunneling. (b) Conductance vs. applied voltage at 2.7 K. A zero bias
anomaly, as well as phonon peaks attributed to GaAs-like and AlAs-like LO phonons, is clearly visible.
junction community has shown that their presence cannot be ruled out eas-
ily [108–110]. Pinholes form low-resistance areas, which essentially short
out the high resistance tunnel barrier layer. Application of the “Rowell cri-
teria” for tunneling [108, 111], and observation of phonon signatures and
a zero bias anomaly in the low temperature conductance spectra provide
clear, unambiguous tests to determine whether tunneling is the dominant
transport mechanism.
There are three Rowell criteria. The first—the conductance (G = dI/dV)
should have an exponential dependence on the thickness of the barrier—
cannot be readily applied in this case due to the nonrectangular shape of the
barrier and variations of the barrier width with bias. The second criterion
states that the conductance should have a parabolic dependence on the volt-
age and can be fit with known models, e.g. a Simmons (symmetric barrier)
[112] or Brinkman, Dynes, and Rowell (BDR) model (asymmetric barrier)
[113]. The inset to Figure 2.14a shows G–V data at a variety of temperatures,
and a representative fit (dashed line) using the BDR model. Parameters of
this asymmetric barrier model are defined in the diagram. Fits to the data
between ±100 meV at several different temperatures yield an average barrier
thickness of d = 29 Å, and barrier heights of ϕ1 = 0.46 eV and ϕ2 = 0.06 eV.* The
large potential difference between the two sides of the barrier is physically
* Notice that the voltage range here is much smaller than the 1~2 V applied to the LED in
the spin injection experiment. In the transport measurement, the voltage drop is primarily
across the tunnel barrier interface, while in the LED, there are a variety of series resistances
and contact resistances, as well as the band gap of GaAs that need to be considered.
consistent with a triangular Schottky barrier tunnel junction. Although ϕ1 is

lower than might be expected for an Fe/AlGaAs interface, image force low-
ering of the barrier due to the highly degenerate nature of the AlGaAs can
lead to reduction of the barrier by >0.3 eV [103]. The “goodness” of the fits,
energy range considered, and deviation of the fit parameters from “known”
physical characteristics of the barrier are typical of similar treatments in
the literature [108, 114]. Therefore, one may conclude that the second Rowell
criterion is satisfied.
While the first two criteria are routinely invoked as proof of tunneling, it
has been argued that neither can reliably distinguish tunneling from contri-
butions due to spurious effects such as pinholes [108–110]. It has been shown
that the G–V data can be fit with reasonable parameters even when tun-
neling was not the dominant transport path [108]. Jönsson-Åkerman et al.
have presented convincing evidence that the third Rowell criterion is indeed
a definitive confirmation of tunneling [108]. This criterion states that the
zero bias resistance (ZBR), i.e. the slope of the I–V curve at zero bias, should
exhibit a weak, insulating-like temperature dependence. ZBR data are shown
in Figure 2.14a as a function of temperature for the tailored Fe/AlGaAs
Schottky contact, and clearly exhibit such a weak dependence over a wide
range of temperature. Thus, the third Rowell criterion is also satisfied, con-
firming that single-step tunneling is the dominant conduction mechanism.
Further evidence for tunneling may be provided by the observation of
phonon signatures and a pronounced zero bias anomaly in the conductance
spectra.* For transport across an ohmic contact, electron energies can never
reach a value higher than a few kT above EF , regardless of the applied bias,
and phonon modes are not observed because the injected electron energy
is too low to excite them. A tunnel barrier enables injection of electrons at
higher energies that are sufficient to excite phonons, thereby enabling a cor-
responding spectroscopy [115]—the observation of such features then pro-
vides further proof for tunneling.
The conductance spectrum for an Fe/Al0.1Ga0.9As sample at 2.7 K is
shown in Figure 2.14b, and exhibits two distinct features between 30 and
50 meV. In AlGaAs, two sets of longitudinal-optical (LO) phonon modes
are present, one GaAs-like and the other AlAs-like, with energies of 36 and
45 meV, respectively. The observed features agree very well with these nomi-
nal values, and are labeled accordingly. In addition, the relative intensities of
the GaAs and AlAs phonon interactions are positively correlated with the
relative Ga:Al content [116]—the observed features exhibit an intensity ratio
of ∼10:1, further confirming their identity.
At the lowest temperatures, the G–V data have a pronounced feature
at zero bias. Zero bias anomalies are generally observed in semiconductor
tunneling devices [115]. Although poorly understood, they have been attrib-
uted to inelastic scattering effects arising from acoustic phonons and barrier
defects, and are closely associated with tunneling. The observation of such a
feature in these data provides further evidence for tunneling.
* See, e.g., Ref. [115].

2.5.3.1.2 Examples: Spin Injection via Tailored Schottky Tunnel Barrier

The Fe/AlGaAs/GaAs/AlGaAs LED structures described above (Figure 2.13)
provide conclusive evidence for efficient spin injection from an FM metal
into a semiconductor, using the tunnel barrier to circumvent the conduc-
tivity mismatch [104, 105]. These samples were processed to form surface
emitting LEDs (Figures 2.4a and 2.7b), annealed at 200°C to improve the
interface structure, and biased to inject spin-polarized electrons from the
Fe through the tailored Schottky tunnel barrier and into the semiconductor,
where they radiatively recombine with unpolarized holes. The EL spectra
are shown in Figure 2.15a, where the light emitted along the surface nor-
mal (Faraday geometry) is analyzed for σ+ and σ− circular polarization for
selected values of applied magnetic field. The spectra are dominated by the
GaAs QW HH exciton, with a linewidth of 5 meV, clearly identifying the
location where radiative recombination occurs.
With no applied magnetic field, the Fe magnetization (easy axis) and
corresponding electron spin orientation are entirely in the plane of the thin
film. Although spin injection may indeed occur, it cannot be detected via
surface emission because the average electron spin along the surface nor-
mal (z-axis) is zero, and the σ+ and σ− components are nearly coincident,
as expected. The magnetic field is applied to rotate the Fe magnetization
(and electron spin orientation in the Fe) out of plane, so that any net elec-
tron spin polarization can be manifested as circular polarization in the EL
via the quantum selection rules. As the magnetic field increases, the com-
ponent of Fe magnetization and electron spin polarization along the z-axis
4πM (Fe)
40
σ+ 1 mA, 2 V
σ– 4.5 K
30
Fe/AlGaAs/GaAs
σ+
EL circular polarization (%)
20 spin-LED
5 meV
Spin-LED EL intensity
10
3T
0
Dichroism
–10
2
–20 σ–
1 T = 4.5 K
–30
0
–40
1.53 1.54 1.55 1.56 –6 –4 –2 0 2 4 6
(a) Photon energy (eV) (b) Magnetic field (T)
FIGURE 2.15 EL data from an Fe Schottky tunnel barrier spin-LED. (a) EL spectra for selected values of
applied magnetic field, analyzed for positive and negative helicity circular polarization. (b) Magnetic field
dependence of Pcirc = Pspin. The dashed line shows the out-of-plane Fe magnetization obtained with SQUID
magnetometry and scaled to fit the EL data. The triangles indicate the measured background contribution,
including dichroism, using PL from an undoped reference sample.
continuously increase, and the corresponding spectra exhibit a substantial

difference in intensity of the σ+ and σ− components—this difference rapidly
increases with field, signaling successful electrical spin injection.
The field dependence of the circular polarization Pcirc = [I(σ+) − I(σ−)]/
[I(σ+) + I(σ−)], where I(σ+) and I(σ−) are the EL component peak intensities
when analyzed as σ+ and σ−, respectively, is summarized in Figure 2.15b.
Pcirc rapidly increases with field and directly tracks the out-of-plane mag-
netization of the Fe film obtained by independent magnetometry measure-
ments (dashed line). Pcirc saturates at a value of 32% at a magnetic field value
characteristic of the Fe contact, B = 2.2, T = 4πMFe, where the Fe magneti-
zation is saturated out-of-plane. Thus, the electron spin orientation is pre-
served during injection from the Fe contact, with a net spin polarization
Pspin = Pcirc = 32% at the moment of radiative recombination in the GaAs QW.
Significant polarization is observed to nearly room temperature.
Preliminary analysis of the temperature dependence of Pcirc shows that it is
dominated by that of the QW spin lifetimes, indicating that the injection
process itself is independent of temperature [104], as expected for tunneling.
Previous work has shown that the electron spin relaxation in a GaAs QW
generally occurs more rapidly with increasing temperature [5, 78], suppress-
ing the measured circular polarization—the GaAs(001) QW is simply an
imperfect spin detector, with a strong temperature dependence of its own.
As noted earlier, the optical polarization measured depends upon the values
of the spin and radiative carrier lifetimes. Both have been extensively stud-
ied, vary with temperature, and depend upon the physical parameters of the
structure and the “quality” of the sample material.
One can determine the initial spin polarization Po that exists at the
moment the electrically injected electrons enter the GaAs QW by indepen-
dently determining the value of the ratio τr/τs by PL measurements and apply-
ing Equation 2.4, as described previously for the case of spin injection from
ZnMnSe films. These measurements yield τr/τs = 0.78 ± 0.05, resulting in a
value Po = Pspin (1 + τr/τs) = 57%. It is interesting to note that this value exceeds
the nominal spin polarization of bulk Fe (∼45%), clearly demonstrating that
the bulk spin polarization of FM metals does not represent a limit to the spin
polarization that can be achieved via electrical injection. Effects such as spin
filtering at the Fe/AlGaAs interface and spin accumulation in the GaAs QW
enable the generation of highly polarized carrier populations, and can be
exploited in the design and operation of semiconductor spintronic devices.
A number of control experiments must be performed to rule out spu-
rious effects. For example, LED structures fabricated with a nonmagnetic
metal contact showed little circular polarization and very weak field depen-
dence, eliminating contributions from Zeeman splitting in the semiconduc-
tor itself. Possible contributions to the measured Pcirc arising from magnetic
dichroism as the light emitted from the QW passes through the Fe film
may be determined both analytically and directly measured. This contribu-
tion was calculated to be 0.9% using well-established models at the appro-
priate wavelength for the thickness of the Fe film [117]. This contribution
was also directly measured by independent PL measurements on undoped
Fe/AlGaAs/GaAs QW test structures. Linearly polarized laser excitation

creates unpolarized electrons and holes in the AlGaAs/GaAs, which emit
unpolarized light when they recombine in the GaAs QW. This light passes
through the Fe film, and any circular polarization measured is therefore due
to a combination of dichroism in the Fe, field-induced splittings in the QW,
or other background effects. The solid triangles in Figure 2.15b summarize
these measurements, and show that the sum of any such contributions is
<1%. Note that the effect measured due to electron spin injection is over 30
times larger.
The sign of the polarization demonstrates that mj = −1/2 electrons
injected from the Fe Schottky tunnel contact dominate the radiative recom-
bination process in the QW when the magnetic field is applied along the
surface normal (Figure 2.1b). This is confirmed by the results from the
ZnMnSe-based spin-LEDs, where the CB spin-splitting is known and the σ+
component also dominates for a similar field orientation. In the nomencla-
ture of the magnetic metals community [90, 118], such electrons are referred
to as “majority spin” or “spin-up,” even though the spin is antiparallel to the
net magnetization. In this community, the term “spin-up” is used to describe
an electron whose moment (rather than spin) is parallel to the magnetiza-
tion. The corresponding state is at lower energy and more populated, and is
therefore synonymously referred to as the “majority spin” state. Such carriers
are designated here as “nm↑” where the subscript “m” is used to unambigu-
ously indicate the convention used in this community [90]. Since the elec-
tron’s spin and moment are antiparallel [119, 120], the spin of a “spin-up”
electron is actually antiparallel to the magnetization. Similarly, the terms
“minority spin” and “spin-down” (designated nm↓) are used synonymously
to refer to electrons whose moment is antiparallel to the magnetization, and
therefore have a higher energy than the “majority spin” electrons.
The experimental observation that the polarization of the current
injected from the Fe contact through the Schottky tunnel barrier cor-
responds to majority spin in Fe is consistent with the pioneering work of
Meservey and Tedrow [118], and with the model proposed by Stearns [121].
While one might reasonably expect majority spin injection to dominate,
recent work has shown that a number of factors are likely to contribute to
the spin polarization of the tunneling current from an FM contact, includ-
ing barrier thickness, band bending, and details of bonding at the interface,
which affect the interface electronic structure [122–124].
The interface atomic structure and its correlation with spin injection
efficiency was addressed for the Fe/AlGaAs/GaAs system by combining
the spin-LED results described above with high resolution TEM analy-
sis of spin-LED samples and density functional theory (DFT) calculations
of the Fe/AlGaAs interface [125]. The TEM analysis showed that the as-
grown Fe/AlGaAs interface exhibited some degree of disorder, as shown
in Figure 2.16a. The chemical order and coherence of the interface could
be significantly improved by a low temperature anneal (200°C for 10 min),
resulting in a highly ordered interface (Figure 2.16b) and a 44% increase
in spin injection efficiency. The annealing temperature was well below the
(110)
(001)
Fe AlGaAs Fe AlGaAs
(a) (b)
FIGURE 2.16 HRTEM images ([−110] cross section) of an Fe/AlGaAs spin-LED

sample. (a) As-grown, exhibiting 18% spin polarization, and (b) following a mild
post growth anneal, exhibiting a 26% spin polarization. Image simulations are
inset (white brackets). The rectangles on the bottom of the images indicate Fe,
AlGaAs, and the interfacial region (gray box with cross). Scale bars equal 1.0 nm.
The contrast variation across the Fe regions is a result of changes in thickness.
∼600°C growth temperature of the AlGaAs/GaAs QW structure, and there-

fore did not affect the characteristics of the semiconductor away from the Fe
interface.
Phase and Z-contrast images were compared with those simulated from
four interface models determined by DFT to be likely low energy candidates.
Three of these models, shown in Figure 2.17 (from left to right: abrupt, par-
tially intermixed, and fully intermixed), were previously proposed and stud-
ied theoretically [126]. The fourth model contained several monolayers of
Fe3GaAs (a known stable alloy in the Fe-Ga-As phase diagram) sandwiched
as an interlayer between the GaAs and Fe. For each model, Z-contrast
images were simulated with software [127] using the relaxed atomic coor-
dinates determined by the DFT calculations. Based upon a mathematical
comparison of line scans of the atomic intensity profiles, both parallel and
perpendicular to the interface between the experimental and simulated
images (Figure 2.18), the authors concluded that the interface of the annealed
sample forms by intermixing of the Fe and AlGaAs occurring on a single
atomic plane, resulting in an interface with alternating Fe and As atoms (see
Figure 2.18b, and middle model structure in Figure 2.17).
The 44% increase in spin injection efficiency observed experimentally
was attributed to the greater tunneling efficiency for spin-polarized elec-
trons across the chemically and structurally coherent annealed interface.
As discussed previously (Section 2.5.1), band symmetries play an important
role in spin injection from a metal to a semiconductor. Strong spin filter-
ing is expected to occur at the ideal Fe/GaAs(001) interface (abrupt with no
intermixing), because the Δ1 symmetry of the bulk Fe majority-spin state
near EF matches that of the bulk GaAs CB states, enhancing transmission
Relaxed structures
(110)
FIGURE 2.17 Low energy structures calculated by DFT after layer relaxation
for the As-terminated Fe/GaAs(001) interface. Left—abrupt interface; middle—
partially intermixed, with one plane consisting of alternating Fe and As atoms;
and right—fully intermixed. The gray and light gray spheres represent Ga and
As atoms, and the dark gray spheres represent Fe atoms. Highly strained bonds
∼15%–20% longer than ideal are shown as dotted lines. (After Erwin, S.C. et al.,
Phys. Rev. B 65, 205422, 2002. With permission.)
of majority spin electrons [82]. The symmetries of the Fe minority-spin

bands do not match the GaAs, so that these states decay very quickly, sup-
pressing minority spin transmission. The same analysis was applied to the
partially intermixed and fully intermixed interface models reported in Ref.
[126] and considered in the TEM analysis above. The results show no sig-
nificant change in the Δ1 decay rate between the abrupt and partially inter-
mixed models [125], indicating that both should enable highly polarized spin
injection. However, a significantly faster decay of the majority spin Δ1 state
into the GaAs was found for the fully intermixed model, indicating lower
spin polarization of the injected carriers, and consistent with the measured
polarizations from the spin-LED samples.
This picture of the Fe/AlGaAs interface structure is consistent with ear-
lier detailed studies of the initial nucleation, interface formation, and mag-
netic properties of Fe films grown on the As-dimer terminated GaAs(001)
2 × 4 and c(4 × 4) surfaces reported by Kneedler et al. [128–130] and Thibado
et al. [131]. These authors concluded that the interface that was ultimately
formed after a few monolayers of Fe deposition was planar with little inter-
mixing, and characterized by Fe–As bonding. Excess As diffuses through
the Fe film and segregates to the surface. Complementary studies further
indicated that the electrical character of the Fe/GaAs interface was not dom-
inated by As antisite defects [132, 133], consistent with surface segregation
of As. An overview of the growth and magnetic properties may be found in
Ref. [134].
This tailored doping profile and Schottky tunnel barrier has been widely
used in subsequent studies to elucidate spin injection, transport, and detec-
tion in a variety of FM metal/AlGaAs/GaAs heterostructures [135–140].
Generation of minority spin electrons has been reported for a limited range
6 5 4 3 2 1 0 –1 –2 –3 –4 –5
β α
β΄
α΄
(a)
(110)
(001)
Fe
Ga
As
(b)
FIGURE 2.18 Z-contrast TEM images of the Fe/AlGaAs(001) [−110] interface from
spin-LED samples. (a) Experimental image from a sample with 26% spin polariza-
tion. The arrowheads, α–α′ and β–β′, indicate the location and direction of the line
profiles parallel and perpendicular to the interface, respectively, as used in the
analysis. (b) Simulated image of a partially intermixed interface and inset ball-and-
stick model from which it was calculated (Fe atoms appear in yellow, As in blue,
and Ga in red). The scale bar equals 0.5 nm.
of bias conditions in Fe/AlGaAs/GaAs structures with interface doping pro-

files and structures nominally identical to those described above [135, 137].
Two possible theoretical explanations have been offered. The first invokes
formation of interface states which mediate minority spin transmission
when the structure is biased so that electrons flow from the GaAs into the Fe
(spin extraction) [124]. Note that this calculation was based on the perfectly
abrupt Fe/GaAs interface model (Figure 2.17 left image), rather than the
partially intermixed model (Figure 2.17 middle image), as concluded from
the TEM analysis above. The specific structure of the interface will have a
significant impact on the formation and character of interface states, and
such calculations can provide further insight by addressing other interface
structures. The second explanation invokes formation of bound states which
form in the CB QW produced by the heavy doping near the semiconductor
interface [123]. The relative contribution of these bound states to the spin
transport process can also lead to the generation of minority spin accumula-
tion in the GaAs. These results underscore the fact that several factors need
to be considered in developing a more comprehensive understanding of spin
transmission across a heterointerface. The tailored Schottky tunnel barrier
approach has also been used very recently for spin injection into ZnSe [141]
and Si [142].
2.5.3.2 Spin Injection via a Discrete Layer as a Tunnel Barrier

Discrete insulating layers may also be used to form a tunnel barrier between
the FM metal and the semiconductor. These provide the more familiar and
well-studied rectangular potential barrier, but also introduce an additional
interface into the structure and raise issues of pinholes and uniformity of
layer thickness. FM metal/oxide/metal structures have been extensively
studied, since they form the basis for a tunneling spectroscopy used to deter-
mine the metal spin polarization [143], and for tunnel magnetoresistance
devices being developed for nonvolatile memory [144–147]. Reviews of this
area may be found in Chapters 10 and 11, Volume 1. Various oxides have
been used as tunnel barriers for spin injection into a semiconductor, includ-
ing Al2O3, SiO2, MgO, and Ga2O3. Examples of each are presented below.
2.5.3.2.1 Al2O3/AlGaAs-GaAs
Spin-polarized tunneling has been well-studied in metal/Al2O3/metal het-
erostructures. Careful measurements using a superconductor such as Al for
one contact have provided quantitative information on the spin polarization
of the tunneling current from many FM metals [143, 148, 149].
Several groups have utilized Al2O3 barriers in FM metal/Al2O3/AlGaAs-
GaAs spin-LED structures with great success [150–153]. Motsnyi et al. [150]
used a CoFe contact and the oblique Hanle effect to measure and analyze
the EL, and reported values for the GaAs electron spin polarization Pspin of
14%–21% at 80 K, and 6% at 300 K. In addition, they were able to experimen-
tally determine the radiative and spin lifetimes of the GaAs itself. When they
corrected their measured values for this “instrument response function” of
the GaAs detector as discussed previously, they obtained a value Po = 16%
at 300 K (this quantity is denoted by “Π” in their notation). Manago and
Akinaga obtained spin polarizations of ∼1% using either Fe, Co, or NiFe con-
tacts [152]. van’t Erve et al. obtained a value of Pspin = 40% at 5 K for samples,
which utilized a lightly n-doped (1016 cm−3) Al0.1Ga0.9As layer adjacent to the
Al2O3 [153]. However, they reported lower operating efficiency (higher bias
voltages and currents) than for the Schottky barrier-based spin-LEDs of Refs.
[104, 105]. This was subsequently remedied by including the same semicon-
ductor doping profile shown in Figure 2.13 before the growth of the Al2O3
layer, so that high Pspin = 40% was achieved at T = 5 K and bias conditions
comparable to those utilized for the Schottky tunnel barrier devices. They
determined the efficiency of their GaAs spin detector by optical pumping to
obtain the value τr/τs = 0.75, resulting in a value Po = Pspin (1 + τr/τs) = 70%—a
value that again significantly exceeds the spin polarization of bulk Fe.
2.5.3.2.2 Al2O3/Si
One of the first demonstrations of spin injection into Si utilized an Al2O3
tunnel barrier in an Fe/Al2O3/Si(001) n–i–p spin-LED heterostructure [154].
While Si is clearly an attractive candidate for spintronic devices, its indirect
band gap makes it much more difficult to probe optically, and led many to
believe that the optical spectroscopic techniques, which had proven so pro-
ductive when applied to the III–Vs, would be unable to provide much insight
into spin-dependent behavior in Si. Although the indirect gap complicates
application of polarized optical techniques used routinely in GaAs, it is
worth noting that the first successful optical pumping experiments to induce
a net electron/nuclear spin polarization were performed in Si rather than in
a direct gap material by Lampel [6].
Several fundamental properties of Si make it an ideal host for spin-based
functionality. Spin–orbit effects producing spin relaxation are much smaller
in Si than in GaAs due to the lower atomic mass and the inversion symmetry
of the crystal structure itself. The dominant naturally occurring isotope, Si
[28], has no nuclear spin, suppressing hyperfine interactions. Consequently,
spin lifetimes are relatively long in Si, as demonstrated by an extensive lit-
erature on both donor-bound [155] and free electrons [156, 157]. In addition,
silicon’s mature technology base and overwhelming dominance of the semi-
conductor industry make it an obvious choice for implementing spin-based
functionality. Several spin-based Si devices have indeed been proposed,
including transistor structures [158, 159] and elements for application in
quantum computation/information technology [160].
Jonker et al. electrically injected spin-polarized electrons from a thin
FM Fe film through an Al2O3 tunnel barrier into an Si(001) n–i–p doped
heterostructure, and observed circular polarization of the EL [154]. This
signals that the electrons retain a net spin polarization at the time of radi-
ative recombination, which they estimated to be ∼30%, based on simple
arguments of momentum conservation. This interpretation was confirmed
by similar measurements on Fe/Al2O3/Si/AlGaAs/GaAs QW structures in
which the spin-polarized electrons injected from the Fe drift under applied
field from the Si across an air-exposed interface into the AlGaAs/GaAs
structure and recombined in the GaAs QW. In this case, the polarized EL
can be quantitatively analyzed using the standard selection rules, yielding
an electron spin polarization of 10% in the GaAs. More recently, the theory
provided a quantitative link between the circular polarization of the Si EL
and the electron spin polarization. These results are summarized in the fol-
lowing paragraphs.
The Si n–i–p LED structures were grown by MBE, transferred in air,
and introduced to a second chamber for deposition of the Fe/Al2O3 tunnel
contact. An n-doping level ∼2 × 1018 cm−3 at 300 K was chosen to prevent car-
rier freeze-out at lower temperatures. This is well below the metal–insulator
transition of 5.6 × 1018 cm−3. Details of the growth and processing may be
found in Ref. [154]. A schematic of the sample structure, corresponding band
diagram of the spin-injecting interface, and a photograph of the processed
surface-emitting LED devices are shown in Figure 2.19a. A high resolution
cross-sectional TEM image of the Fe/Al2O3/Si contact interface region is
shown in Figure 2.19b, and reveals a uniform and continuous oxide tunnel
barrier with relatively smooth interfaces. The Fe film is polycrystalline on
the nominally amorphous Al2O3 layer.
Fe 10 nm Fe
Al2O3 70 nm n-Si
70 nm i-Si
n-Si
150 nm p-Si Fe
Al2O3
p-Si(001) Substrate
Si
100 µm
2 nm
(a) (b)
FIGURE 2.19 Fe/Al2O3/Si spin-LED structures. (a) Band diagram of the spin-injecting interface and schematic
of the sample structure, with an optical photograph of the processed spin-LED devices. The light circular
mesas are the active LED regions. (b) Cross-sectional TEM image of the spin-injecting interface region.
Typical EL spectra from a surface-emitting Fe/Al2O3/Si n–i–p spin-LED

structure are shown in Figure 2.20a and b for T = 5−80 K, analyzed for σ+
and σ− circular polarization. At 5 K, the spectra are dominated by features
arising from electron-hole recombination accompanied by transverse acous-
tic (TA) or transverse optical (TO) phonon emission. Subsequent analysis to
determine the origin of the emission features has resulted in a revised iden-
tification of the EL peaks observed. This analysis revealed that the TO/TA
peaks occur in correlated pairs, with the peaks at 1105 meV (TAs) and 1065–
1070 meV (TOs) due to recombination in the p-Si substrate (p ∼ 1019 cm−3),
and the peaks at 1090 meV (TA1) and 1050 meV (TO1) likely arising from
recombination in the interface region. At 80 K, the TOs feature dominates.
At zero field, no circular polarization is observed because the Fe magnetiza-
tion and corresponding electron spin orientation lie in-plane and orthogonal
to the light propagation direction. As discussed previously, although spin
injection may occur, it cannot be detected with this orthogonal alignment.
Therefore, a magnetic field is applied to rotate the Fe spin orientation out-of-
plane, and the main spectral features each exhibit circular polarization, as
shown by the difference between the red (σ+) and blue (σ−) curves in the 3 T
spectra at T = 5, 50, and 80 K.
The magnetic field dependence of Pcirc for each feature is summarized
in Figure 2.20c. As the Fe magnetization (and majority electron spin ori-
entation) rotates out of plane with increasing field, Pcirc increases and satu-
rates above ∼2.5 T with average values of 3.7%, 3.5%, and 1.9% for the TA s,
TA1, and TO1 features, respectively, at 5 K, and 2.1% and 2% for the TO s
feature at 50 and 80 K. Note that, for each feature, Pcirc tracks the magneti-
zation of the Fe contact, shown as a solid line scaled to the data, indicating
that the spin orientation of the electrons that radiatively recombine in the
Si directly reflects that of the electron spin orientation in the Fe contact.
The field at which the Fe magnetization saturates out of plane is charac-
teristic of Fe (Hsat = 4πMFe = 2.2 T), and is unaffected by lateral patterning,
which might alter the in-plane coercive field. This also argues against pos-
sible compound formation (e.g. FeSi, etc.) at the interface due to pinholes,
which would lead to a different field dependence. The monotonic decrease
in Pcirc from the higher energy (TA s) to the lowest energy feature (TO1) as
seen in Figure 2.20c is consistent with spin relaxation expected to accom-
pany electron energy relaxation. The sign of Pcirc indicates that Fe major-
ity spin electrons dominate, as was the case for the tailored Schottky Fe/
AlGaAs and Fe/Al 2O3/AlGaAs contacts. Data from the requisite reference
samples used to determine background effects such as dichroism show
only a weak paramagnetic field dependence ∼0.1%/T, ruling out spurious
contributions. These data together unambiguously demonstrate that spin-
polarized electrons are electrically injected from the Fe contact into the Si
heterostructure.
An analysis of the EL to extract the electron spin polarization Pspin in the
Si is complicated by the indirect gap character and the participation of pho-
nons in the radiative recombination process. Since the electron spin angu-
lar momentum must be shared by the photons and any phonons involved
in radiative recombination, the photons will carry away only some frac-
tion of the spin angular momentum of the initial spin-polarized electron
population. Thus, our experimentally measured value of Pcirc will result in
a significant underestimate of the corresponding Si electron polarization.
A second significant factor affecting Pcirc is the very long radiative lifetime
typical of indirect gap materials—as noted previously, Pspin decays exponen-
tially with a characteristic time τs before radiative recombination occurs.
Despite these issues, the broader rule of conservation of momentum can be
used to estimate a lower bound for the initial electron spin polarization, and
the same rate equation analysis leading to Equation 2.4 applies to correct
for these lifetime effects: Po = Pspin (1 + τr/τs) > Pcirc (1 + τr/τs). Typical values
for τr in doped Si at low temperature are 0.1–1 ms [161, 162]. The radiative
lifetime of emission features associated specifically with acoustic phonon
emission was measured to be 480 µs for 1 K < T < 5 K [161]. The spin lifetime
for free electrons well above the metal–insulator transition has been deter-
mined to be ∼1 µs in both bulk [156] and modulation doped samples [157],
and is expected to increase at lower electron densities [156]. Therefore, using
Pcirc ∼ 0.03 (Figure 2.20c), τr = 100 µs and τs = 10 µs, a conservative estimate
is given by Po > Pcirc(1 + τr/τs) ∼ 0.3 or 30%, comparable to that achieved in
GaAs.
This value is supported by similar experiments on Fe/Al2O3/80 nm
n-Si/80 nm n-Al0.1Ga0.9As/10 nm GaAs QW/200 nm p-Al0.3Ga0.7As hetero-
structures [154], in which electrons injected from the Fe into the Si drift
across the Si/Al0.1Ga0.9As interface with applied bias and radiatively recom-
bine in the GaAs QW. In this case, the standard quantum selection rules
can be rigorously applied to quantify the electron spin polarization, which
FIGURE 2.20 See EL spectra from surface-emitting Fe/Al2O3/Si n–i–p spin-LED structures, analyzed for
positive (σ+) and negative (σ−) helicity circular polarization at temperatures of (a) 5 K, and (b) 50 and 80 K.
The spectra are dominated by features arising from electron-hole recombination accompanied by transverse
acoustic (TA) and transverse optical (TO) phonon emission. (c) The magnetic field dependence of the circular
polarization for each feature in the Si spin-LED spectrum at 5 K. (d) Inset shows the EL spectrum from Fe/
Al2O3/80 nm n-Si/n-Al0.1Ga0.9As/GaAs QW/p-Al0.3Ga0.7As spin-LEDs at T = 20 K and H = 3 T. The dominant feature
is the GaAs QW free exciton, which exhibits strong polarization. The magnetic field dependence of this circu-
lar polarization is plotted for T = 20 and 125 K. The out-of-plane Fe magnetization curve appears in panel (c)
and (d) as a solid line for reference.
eventually reaches the GaAs [163]. The EL spectrum at 20 K and 3 T is shown
as insert in Figure 2.20d, and is dominated by the free exciton emission at
1.54 eV from the GaAs QW. The field dependence of the polarization of this
feature, Pcirc(GaAs), is shown in Figure 2.20d. Pcirc(GaAs) again tracks the
magnetization of the Fe contact, and saturates at values of 5.6% at 20 K and
2.8% at 125 K. Separate optical pumping measurements on the GaAs QW
provide a direct measure of τr/τs (GaAs QW) = 0.8. Thus, the spin polariza-
tion of the electrons in the GaAs is PGaAs = Pcirc(1 + τr/τs) = 0.1 or 10% at 20 K.
This is consistent with our estimate above of PSi ∼ 30%, and establishes a firm
lower bound for PSi > 10%. It is indeed remarkable that the electrons injected
from the Fe contact drift through the Si, cross the Si/AlGaAs interface and
still retain a significant spin polarization, given (1) the relatively poor crys-
talline quality of Si epilayers on GaAs (lattice mismatch 3.9%); (2) the het-
erovalent interface structure; (3) the 0.3 eV CB offset (Si band lower than
Al0.1Ga0.9As) [164]; and (4) the fact that the sample surface was exposed to
air before growth of the Si, and then again before growth of the Fe/Al2O3
contact.
Recent theory provides a fundamental and more rigorous interpretation
of the circular polarization of the TO and TA features in the Si spin-LED
EL spectra, and thereby an independent determination of the electron spin
polarization achieved in the Si by electrical injection. Li and Dery [165] have
shown that the circular polarization of these features is directly related to
the electron spin polarization for a given doping level in the region where
radiative recombination occurs. For p = 1019 cm−3, they conclude that the TA
feature should have a maximum circular polarization of 13% for a 100% spin-
polarized electron population. Thus, the measured value Pcirc ∼ 3.5% for the
TAs feature (Figure 2.20c) indicates an electron spin polarization of 27%. The
estimates of the electron spin polarization summarized here are in remark-
ably good agreement.
The EL intensity decreases rapidly with increasing temperature. It
is too low to reliably analyze at temperatures higher than those shown in
Figure 2.20 for emission from either the Si (80 K) or GaAs (125 K) even
though clear polarization remains—a further limitation of the LED as a spin
detector. Subsequent work using electrical rather than optical spin detec-
tion confirms that spin injection in the FM metal/Al2O3/Si system persists to
room temperature, as described in the next section.
The Al2O3 layer serves as both a tunnel barrier and a diffusion barrier,
preventing interdiffusion between the Fe and Si which is likely to occur. Spin
injection is also observed in Fe/Si n–i–p samples when the Fe is deposited
directly on the Si, where the Schottky barrier serves as a tunnel barrier [166].
However, the net spin polarization achieved is lower due to the inhomoge-
neous character of the interface that is formed.
Grenet et al. have also used the spin-LED approach to demonstrate spin
injection into Si [167]. They employed a (Co/Pt)/Al2O3 tunnel barrier contact,
where the Co/Pt exhibits strong perpendicular magnetic anisotropy with the
remanent magnetization along the surface normal. This obviates the need for an
applied magnetic field to saturate the out-of-plane magnetization, as required
for the Fe/Al2O3 contacts described above. The entire structure consisted of
(3 nm Pt/1.8 nm Co)/2 nm Al2O3/100 nm n-Si/10 nm Si0.7Ge0.3 QW/550 nm
p-Si/p-Si(001), with 50 nm dopant setbacks at the QW. The strained Si0.7Ge0.3
QW served as the radiative recombination region, which produces a Type II
band alignment with strong hole localization in the Si0.7Ge0.3. They measure a
maximum (typical) circular polarization Pcirc = 3% (1.2%) in the EL signal from
the QW, which is nearly constant at 200 K. The magnetic field dependence of
Pcirc tracks the Co/Pt magnetization, confirming that the injected electrons
retain their spin orientation from the Co/Pt contact.
2.5.3.2.3 SiO2/Si
SiO2 has been the gate dielectric of choice for generations of Si metal-oxide-
semiconductor devices, because it is easy to form and provides the low inter-
face state density necessary for device operation. It has been the cornerstone
of the vast Si electronics industry, because no other semiconductor has a
robust native oxide which forms such an electrically stable self-interface.
As such, it is an obvious choice to use as the tunnel barrier in a Si-based
spin transport device. However, there are few reports of its use as a spin tun-
nel barrier in any structure. Smith et al. reported a smaller-than-expected
magnetoresistance of 4% at 300 K in Co/SiO2/CoFe tunnel junctions, which
they attributed to overoxidation at the metal interfaces [168]. A composite
NiFe/SiO2/Al2O3 tunnel barrier was used as the emitter in a magnetic tunnel
transistor by Park et al. [169]. They reported a tunnel spin polarization of 27%
at 100 K, which was lower than the value of 34% obtained for an NiFe/Al2O3
emitter in the same structure. No data were reported for a tunnel barrier
consisting of SiO2 alone.
Li et al. successfully used SiO2 as a spin tunnel barrier on Si, and recently
reported spin injection from Fe through SiO2 into a Si n–i–p heterostructure,
producing an electron spin polarization in the Si, PSi > 30% at 5 K [170]. The
samples were similar to those described above and consist of 10 nm Fe/2 nm
SiO2/70 nm n-Si/70 nm undoped Si/150 nm p-Si/p-Si(001) substrate, form-
ing a spin-LED structure identical to that described above. The SiO2 layer
was formed by natural oxidation of the n-Si surface assisted by illumination
from an ultraviolet lamp. A TEM image of the Fe/SiO2/Si interface region
appears in Figure 2.21a, and shows a reasonably uniform SiO2 layer with
well-defined interfaces. The SiO2 appears amorphous, as expected, while the
Fe is polycrystalline.
Transport measurements based on the Rowell criteria confirmed that
conduction from the Fe through the SiO2 occurred by tunneling. The con-
ductance exhibits a parabolic dependence upon the bias voltage, as shown
in Figure 2.21b for several temperatures. Good fits (solid lines) to the con-
ductance data are obtained within an energy range of ±100 meV using the
BDR model [113], which yield an effective barrier height of ∼1.7 eV and a
barrier thickness of ∼21 Å at 10 K. The temperature dependence of the ZBR,
expressed as R0(T)/R0(300 K), where R0 is the slope of the I–V curve at zero
bias, is summarized in Figure 2.21c. The ZBR exhibits a weak temperature
dependence indicative of single-step tunneling rather than the exponential
FIGURE 2.21 (a) Cross-sectional TEM image of the Fe/SiO2/Si interface region. (b)
Conductance curves for transport through an Fe/SiO2/n -Si(001) contact exhibiting
parabolic behavior, and fits using the BDR model. (c) Zero bias resistance of the
contact as a function of temperature. The weak temperature dependence indicates
single-step tunneling through a pinhole-free SiO2 tunnel barrier.
dependence of thermionic emission. A weak temperature dependence of the

ZBR has been shown to be a reliable indicator for a pinhole-free tunnel bar-
rier [108], as discussed earlier in this chapter.
EL spectra at T = 5 K and B = 0 and 3 T are shown in Figure 2.22a, ana-
lyzed for σ+ and σ− circular polarization. The spectra are very similar to
those of Figure 2.20a (Fe/Al2O3/Si n–i–p spin-LEDs), and are dominated by
TA and TO phonon-mediated recombination, with specific peak identifica-
tions as discussed in Section 2.5.3.2.2. Although emission is initially detected
at a bias of 1.6 V, higher biases are typically used to reduce data acquisition
time, as the polarization measured is relatively independent of the bias. A
magnetic field along the surface normal rotates the Fe magnetization out of
plane, enabling the injected electron spin polarization to be manifested as
circular polarization in the EL. At B = 3 T, where the Fe magnetization is sat-
urated out of plane, the spectra exhibit polarization for each of the spectral
features, with Pcirc ∼ 3%. The magnetic field dependence of Pcirc for the three
main peaks in the spectra is summarized in Figure 2.22b, and the behavior
and discussion parallel that of Figure 2.20c—the electrons that tunnel from
the Fe through the SiO2 to eventually radiatively recombine in the Si clearly
retain a net spin polarization from the Fe. Applying the same rate equation
analysis, discussed above, yields a value of 30% as a lower bound for the elec-
tron spin polarization achieved in the silicon for temperatures up to 50 K. A
similar value is obtained by applying the phonon-mediated recombination
theory of Li and Dery [165] to the TAs feature.
2.5.3.2.4 MgO/AlGaAs-GaAs
MgO has been widely used in metal/oxide/metal tunnel structures, and
more recently in metal spin torque transfer devices (see Chapters 7, 8, and 15,
FIGURE 2.22 (a) EL spectra at 5 K from a Si n–i–p structure with an Fe/SiO2 contact at zero field and 3 T,
analyzed for σ+ and σ− circular polarization. (b) Magnetic field dependence of Pcirc for the TAs, TA1, and TO1
features at 5 K. Pcirc consistently tracks the out-of-plane magnetization of the Fe (dashed line).
Volume 1 for a comprehensive overview). Its rock-salt structure is closely

lattice-matched to bcc-Fe, and high-quality epitaxial growth has been dem-
onstrated. The band symmetry arguments originally applied to the Fe/MgO/
Fe system by MacLaren, Butler et al. [80, 81, 88] show that the Fe Δ1 majority
spin band couples well to the propagating state in the MgO barrier, while the
minority spin bands do not, resulting in highly spin-polarized current. Thus,
Fe/MgO is an attractive candidate as a spin-injecting contact for a semicon-
ductor such as GaAs or Si (see Figure 2.11).
Jiang et al. demonstrated robust spin injection up to room temperature
with CoFe/MgO tunnel contacts on AlGaAs/GaAs QW/AlGaAs(001) spin-
LED structures [171]. A cross-sectional TEM image of the tunnel contact
region is shown in Figure 2.23a. The experimental procedure and analysis are
identical to that described previously in Section 2.5.3.1.2 (Figure 2.15) for Fe
Schottky tunnel/AlGaAs/GaAs QW/AlGaAs spin-LED structures [104, 105].
The EL spectra analyzed for σ+ and σ− polarization for temperatures of 100
and 290 K are shown in Figure 2.23b and c, and exhibit significant circu-
lar polarization of 52% and 32%, respectively, after background subtraction
when the Fe magnetization is saturated along the surface normal.
The authors argue that these high values for Pcirc are due to improved
spin injection efficiency at higher temperatures than either the Fe/Al2O3 or
Fe Schottky tunnel contacts provide. Because spin-dependent tunneling is
expected to have a weak temperature dependence, a more likely explana-
tion lies in the temperature dependence of the spin and radiative lifetimes
of the GaAs QW—the detector efficiency rather than the spin injection
efficiency was much higher due to the particular characteristics of the
GaAs (which can vary significantly from one MBE machine to another).
This was explicitly investigated by Salis et al., who used time-resolved opti-
cal techniques to measure the lifetimes for these same Fe/MgO spin-LED
samples and found that they were strongly temperature dependent, as
shown in Figure 2.24 [172]. Using the values at 10 K of Pcirc (PEL in Figure
2.24b) = 0.47, τr = 300 ps, and τs = 560 ps (Figure 2.24a), from Equation 2.4,
we find Po = Pcirc(1 + τr/τs) = 0.72. For the second sample they studied (see
Figure 3 of Ref. [172]), at 10 K, the corresponding values are Pcirc (PEL) = 0.32,
τr = 500 ps, τs = 700 ps, and Po = 0.55. These values for the initial spin polar-
ization achieved in the GaAs QW are very similar to those obtained using
either the Fe/GaAs Schottky tunnel barrier (Po = 0.57) or Fe/Al2O3 spin-
injecting contacts (Po = 0.70), described in Sections 2.5.3.1.2 and 2.5.3.2.1,
indicating that all three tunnel barriers are of comparable efficiency for spin
injection. These data underscore the importance of quantifying the detec-
tor response function. Nevertheless, these results conclusively show that
efficient electrical spin injection persists to room temperature and presents
no obstacle for future development of semiconductor spintronic devices.
Lu et al. investigated the effect of MgO barrier thickness and crystallin-
ity on the spin injection efficiency in CoFeB/MgO/AlGaAs/GaAs spin-LED
structures [173]. They found that the injection efficiency as determined from
Pcirc increased as the MgO thickness was increased from 1.4 to 4.3 nm. They
attributed this to a suppression of band bending due to hole accumulation in
2 nm
Ta
CoFe
MgO
(a)
Sample I Sample II
T = 100 K T = 290 K
σ+
VT = 1.8 V VT = 2.0 V
σ–
H=5T H=5T
EL intensity (a.u.)
H=0T
H=0T
H = –5 T H = –5 T
800 805 810 815 820 825 820 830 840 850 860 870 880
(b) Wavelength (nm) (c) Wavelength (nm)
FIGURE 2.23 (a) TEM image of the CoFe/MgO/AlGaAs/GaAs spin-LED structure.

Electroluminescence spectra from the spin-LEDs are shown at temperatures of
(b) 100 K and (c) 290 K. (After Jiang, X. et al., Phys. Rev. Lett. 94, 056601, 2005. With
permission.)
the AlGaAs at the MgO interface for the thicker barriers, which prevented
the holes from tunneling into the FM injector. They also found that the spin
injection efficiency was higher for MgO grown at 300°C where it exhibited
some degree of crystalline texture vs. amorphous films grown at room tem-
perature, which they attributed to contributions from the band symmetry
matching arguments discussed earlier in this chapter. Truong et al. com-
bined pulsed electrical input and time-resolved optical spectroscopy to ana-
lyze the temporal response of the spin polarization, following injection with
800
τ
600 τs
Decay time (ps)

Ts
400
200
(a)
0.5
0.4 0.6
0.3 0.4
PEL
ρ
0.2
ρ
0.2
0.1 PEL
0.0 0.0
0 50 100 150 200 250 300
(b) T (K)
FIGURE 2.24 (a) Measured radiative (τ), spin (τs) and combined lifetimes as a func-
tion of temperature for the spin-LEDs of Figure 2.23. (b) Measured circular polariza-
tion PEL and spin detector efficiency ρ = Ts/τr as a function of temperature for B = 0
(open symbols) and 0.8 T (closed symbols). (After Salis, G. et al., Appl. Phys. Lett. 87,
262503, 2005. With permission.)
these same spin-LED structures [174]. They concluded that the buildup time
of the electron spin polarization in the GaAs QW was much faster than the
rise time of the EL signal, which was limited to ∼700 ps due to parasitics in
the device circuit.
2.5.3.2.5 GaOx /AlGaAs-GaAs
Unlike Si, the native oxide of GaAs is difficult to grow with proper stoichi-
ometry and is less stable than SiO2. Hence a metal-oxide-semiconductor
technology, which allows inversion mode operation in a GaAs device, has
never blossomed. Previous efforts reported a very low interface state density
(1010−1011 cm−2 eV−1) at the GaOx/GaAs interface [175], comparable to that
of SiO2/Si, making GaOx an attractive candidate as a gate insulator. Saito
et al. utilized GaOx as the spin tunnel barrier in Fe/GaOx/AlGaAs/GaAs QW
spin-LED structures [176]. They fabricated surface emitting devices and mea-
sured the circular polarization of the EL as a function of applied magnetic
field (Faraday geometry) and temperature. The experimental procedure and
analysis are again identical to that described previously in Section 2.5.3.1.2
(Figure 2.15) for Fe Schottky tunnel/AlGaAs/GaAs QW structures. They
find that Pcirc tracks the out-of-plane magnetization of the Fe contact with a
maximum value Pcirc = Pspin = 20% at T = 2 K. Using a typical value for τr/τs ∼ 1
from the literature, they calculate the initial spin polarization in the GaAs:
Po = Pspin (1 + τr/τs) = 40%. Pcirc decreased rapidly with temperature to a value
of 2% at 300 K, as found previously for GaAs QW based spin-LEDs with both
Fe Schottky and Fe/Al2O3 tunnel barrier contacts, and this is likely due to the
temperature dependence of τr and τs [10]. Thus, the temperature dependence

observed is that of the GaAs QW detector rather than that of the spin injec-
tion process through GaOx.
In summary, the use of a tunnel barrier allows one to circumvent the
conductivity mismatch issue, enabling the use of FM metals as spin-injecting
contacts into a semiconductor. FM metals introduce many desirable attri-
butes to a device technology, including intrinsic nonvolatile and reprogram-
mable operating characteristics. They exhibit reasonable spin polarizations
at EF (∼40%–50%), high Curie temperatures, low coercive fields, fast switch-
ing times, and an advanced materials technology, which permits one to tai-
lor the magnetic characteristics for a particular application. An FM metal/
tunnel barrier contact injects spin-polarized electrons near the semiconduc-
tor CB edge with high efficiency, where the electron energy is determined by
the bias applied across the tunnel barrier (∼10–100 meV). The realization of
efficient electrical injection and significant spin polarization using a simple
tunnel barrier compatible with “back-end” semiconductor processing should
greatly facilitate progress in the development of semiconductor spintronic
devices.
ot Electron Spin Injection

2.5.4 H
We conclude this section by describing a completely different avenue toward
electrical injection of spin-polarized carriers from an FM metal into a semi-
conductor, which also circumvents the conductivity mismatch issue. This is
the process of hot electron injection, which builds on the development of the
spin-valve transistor and magnetic tunnel transistor [175–179]. This topic is
covered in detail in Chapter 3, Volume 2, and we include a brief discussion
here for the sake of providing a more complete overview on electrical spin
injection.
Electrical spin injection into a semiconductor by hot electron injection
was first demonstrated by Jiang et al. using a GaAs/In0.2Ga0.8As multiple QW
structure as the spin detector, incorporating the spin-LED approach [180]. A
schematic of the heterostructure is shown in Figure 2.25, and consists of a
metal emitter/Al2O3 tunnel barrier/FM metal base deposited on GaAs, which
forms the top layer of the GaAs/In0.2Ga0.8As structure on a p-GaAs(001)
substrate. The FM metal base layer forms a Schottky barrier contact with
the GaAs. Electrons from the metal emitter (which need not be FM) tunnel
through the Al2O3 tunnel barrier and enter the FM metal base at an energy
well above EF, determined by the emitter-base bias voltage. Electrons that
are minority spin relative to the base magnetization are rapidly scattered
due to their short mean free path and lose energy in the FM base, relax-
ing to EF, where they are blocked from entering the semiconductor by the
Schottky barrier. Although majority spin electrons have a longer mean free
path, most are also scattered in the base, relax to EF, and are blocked by the
Schottky barrier. But a small fraction of the majority spin electrons do not
scatter, and retain sufficient energy to surmount the Schottky barrier and
enter the semiconductor, forming a spin-polarized current. These electrons
FIGURE 2.25 Schematic energy band diagram of a magnetic tunnel transistor

merged with a QW LED collector. The emitter/base bias VEB controls the energy of
the injected hot electrons, while the collector/base bias VCB can be used to adjust
the band bending of the LED. The emitter consists of 5 nm Co84Fe16. The 2.2 nm
thick Al2O3 tunnel barrier is formed by reactive sputtering of Al in the presence of
oxygen. The base layer consists of 3.5 nm Ni81Fe19 and 1.5 nm Co84Fe16 with the NiFe
layer adjacent to the GaAs. (After Jiang, X. et al., Phys. Rev. Lett. 90, 256603, 2003.
With permission.)
thermalize to the CB and recombine with unpolarized holes in the InGaAs

QWs. The circular polarization of the resultant EL provides an assessment
of the spin injection.
Using the procedures described above, Jiang et al. measure a value
Pcirc = 10% at T = 1.4 K when the FM base magnetization is rotated out of plane
by an applied field (Faraday geometry), signaling successful spin injection
using this hot electron approach. The value of Pcirc is somewhat lower than
measured in other spin-LED experiments, and the authors suggest two likely
reasons: (1) the injected hot electrons lose a significant amount of polar-
ization during the process of thermalization to the bottom of the CB, and
(2) after the hot electrons enter the QW region, further spin relaxation can
occur in the QWs before recombination. The measured result, therefore, sets
a lower bound for the injected electron spin polarization.
This hot electron injection approach was later adopted by Appelbaum
et al. to demonstrate electrical spin injection into Si [181]. The structure is
similar to that of Figure 2.25 (substituting Si for GaAs/AlGaAs), but they
replace the InGaAs QW detector section with a second FM metal layer,
which serves as a spin analyzer, functioning in the same way as described
above for the FM emitter. This enables them to detect spin currents by mea-
suring changes in the current through the device, as one changes the ori-
entation of the magnetization of emitter and analyzer layers from parallel
to antiparallel. Rather than measuring an optical polarization or a voltage
produced by spin accumulation [136], this approach measures the change in
current. A detailed treatment may be found in Chapter 3, Volume 2.
Thus, the combination of hot electron injection, spin filtering in the

FM base or analyzer layer, and energy filtering by a Schottky barrier results
in preferential transmission of majority spin current and a blocking of the
minority spin current, enabling spin injection and analysis. A practical
concern with this approach is that it is very inefficient—most of the current
flows in the emitter-base circuit (or FM analyzer circuit), while a very small
fraction ∼10 −4 enters the semiconductor as spin-polarized current. This is a
recognized problem with the magnetic tunnel transistor [179].
2.6 FERROMAGNETIC METAL/
SEMICONDUCTOR ELECTRICAL SPIN
INJECTION: ELECTRICAL DETECTION
2.6.1 General Considerations and Non-Local Detection
The preceding sections summarize rapid progress in developing a practi-
cal and technologically attractive methodology for electrical spin injection,
enabling one to produce electron populations in a semiconductor with spin
polarizations exceeding those of typical FM metals. It was repeatedly noted
that the temperature dependence observed experimentally was dominated
by the spin detector, e.g. the temperature dependence of the radiative and
spin lifetimes in the spin-LED structures. Indeed, spin injection by tunnel-
ing is expected to depend only weakly upon temperature. Optical detec-
tion of spin polarization in a semiconductor has proven to be invaluable as
a research tool, and will enable a subset of applications such as spin lasers
(see Chapter 16, Volume 3). However, it is not practical for the broader range
of electronic applications, where a voltage input/output with minimal tem-
perature dependence is required to interface to the rest of the circuit. Thus,
utilizing an FM metal as a spin detector has great technological appeal.
The issue of conductivity mismatch has significant implications for elec-
trical spin detection, and places constraints on the range of contact resis-
tances at the FM metal/semiconductor interface, which permit electrical
detection of spin accumulation. This is reviewed in detail in Refs. [97, 100],
and in Chapter 1, Volume 1 by Fert. The specific cases of FM metal contacts
as spin detectors on GaAs and on Si were discussed by Dery et al. [182] and
Min et al. [183], respectively. The criteria for the conventional two-terminal
magnetoresistance device are particularly demanding, and have yet to be
experimentally realized in a conventional semiconductor such as Si or GaAs.
The concept of non-local spin detection was introduced by Johnson
and Silsbee in 1985 in their study of spin accumulation in metals [184] (see
Chapter 5, Volume 1), and developed extensively by others for spin detec-
tion in metals [185, 186], GaAs [136, 137], Si [187, 188], and graphene [189].
It is based upon the fact that spin diffuses away from the point of generation
independently of the direction of charge flow, i.e. a pure spin current flows
outside of the charge current path. This process is similar to that of heat
diffusing away from a hot contact on a surface produced, e.g. by the tip of a
soldering gun. This is illustrated in Figure 2.26, where an FM metal tunnel
2.6 Ferromagnetic Metal/Semiconductor Electrical Spin Injection: Electrical Detection 111
FIGURE 2.26 Schematic layout of four terminal non-local spin valve (NLSV)
device and depiction of the spin-dependent electrochemical potential. A spin-
polarized current is injected at contact 3 and changes in the electrochemical
potential are measured non-locally at contact 2, placed 1 µm from contact 3. The
difference in shape (100 × 24 µm) and (100 × 6 µm) for contact 2 and 3, respectively,
allows for independent control of the magnetizations of the two contacts. The
large reference contacts (100 × 150 µm) are spaced 150 µm (several spin flip lengths)
from the injector and detector contact. The inset shows an image of the actual
device. (After Jansen, R., J. Phys. D 36, R289, 2003. With permission.)
barrier contact (labeled 3 in Figure 2.26) is used to inject a spin-polarized

charge current into the semiconductor. A second FM contact (labeled 2)
is placed outside of the spin-polarized charge current path defined by the
bias applied between contacts 3 and 4. Contact 2 senses the spin-splitting
of the chemical potential, Δµ = µup − µdown, produced by the spin accumula-
tion resulting from spin diffusing away from the FM injector, contact 3. The
spin accumulation is the greatest directly underneath the injector contact,
and decays with increasing distance from the contact with a certain spin
diffusion length, λ sf . The voltage measured at the detector contact 2 is pro-
portional to Δµ and the projection of the semiconductor spin polarization
onto the contact magnetization. Therefore, one can use some mechanism to
manipulate the semiconductor spin polarization (magnitude or orientation)
to encode or process information with the pure spin system, and directly
read out the result as a voltage at this second contact. There are two key con-
cepts to be recognized: (1) the non-local geometry enables one to separate
and distinguish a pure spin current from a spin-polarized charge current,
and (2) the spin accumulation produces a real voltage, which can be sensed
by an FM contact. The following section illustrates these processes, and
summarizes recent work on spin injection and non-local electrical detection
in Si. Non-local spin injection/detection in GaAs is covered in Chapter 6,
Volume 2, and in graphene in Chapters 4 and 5, Volume 3.
Spin accumulation can also be generated and detected using a single
FM metal tunnel barrier contact, which serves as both the spin injector and
detector in a simpler three terminal geometry with two nonmagnetic refer-
ence contacts [136, 190]. This approach is illustrated in Sections 2.6.3 and
2.6.4, using FM metal/oxide tunnel barrier contacts on both Si and GaAs.
2.6.2 Non-Local Spin Injection and Detection

in Silicon: Pure Spin Currents
The non-local spin valve (NLSV) geometry of Figure 2.26 has been used to
generate and detect pure spin current in silicon in a lateral transport device,
and determine the corresponding spin lifetimes via the Hanle effect [187, 188].
The silicon samples were 250 nm thick epilayers grown by MBE on insulating
Si(001) substrates at 500°C. The films were n-doped to achieve an electron
concentration of 2–3 × 1018 cm−3 at room temperature, below the metal–insu-
lator transition (nMIT ∼ 5 × 1018 cm−3), but high enough to avoid carrier freeze-
out at low temperature. The samples were exposed to atmosphere upon
removal from the MBE system. After a 10% HF acid etch and deionized water
rinse to produce a hydrogen passivated surface, the samples were loaded into
a second MBE system, heated to desorb the hydrogen, and a 1.5 nm Al2O3
tunnel barrier and 10 nm polycrystalline Fe film were deposited to form the
spin-injecting tunnel barrier contacts, as described in Ref. [154].
Conventional photolithography and wet chemical etching were used to
define multiple NLSV structures, each consisting of four Fe/Al2O3 contacts
spaced as shown in Figure 2.26. Contacts 2 (24 × 100 µm) and 3 (6 × 100 µm)
are separated by 1 µm and serve as the spin detector and injector, respectively.
The different aspect ratios provide different coercive fields with the easy axis
along the long axis so that the contact magnetizations can be aligned paral-
lel or antiparallel. Contacts 1 and 4 are reference contacts (100 × 150 µm)
located several spin diffusion lengths away from the transport channel to
ensure that the spin polarization is zero at these spin grounds.
When a negative bias is applied to contact 3, a spin-polarized electron
current is injected from the Fe into the Si (Figure 2.27a) to produce (a) a
spin-polarized charge current, which flows to contact 4 due to the applied
bias, and (b) a pure spin current, which diffuses isotropically. The electron
spin is majority spin oriented in the plane of the surface due to the in-plane
magnetization of the Fe injector contact (and opposite the magnetization
of the injector by convention). These spin currents produce spin accumula-
tion in the Si described by the splitting of the spin-dependent electrochemi-
cal potential, Δµ = µup − µdown, which decreases with distance from the point
of injection, as discussed above and shown explicitly in Figure 2.28a. The
FIGURE 2.27 Schematic illustration of (a) injection and (b) extraction of spins
from silicon by means of spin-dependent tunneling.
FIGURE 2.28 (a) Schematic of the spin splitting of the electrochemical potential Δµ corresponding to spin
accumulation in the silicon produced by spin injection from contact 3 in the four terminal non-local spin valve
geometry (reference contact 4 is not shown). Depending upon the direction of its magnetization, the non-
local voltage VNL measured at contact 2 measures either the spin-up or spin-down branch of the chemical
potential relative to the reference contact. (b) Non-local voltage vs. in-plane magnetic field, for an injection
current of −100 µA at 10 K in an Fe/Al2O3/n-Si non-local spin valve structure. Two levels corresponding to the
parallel and antiparallel remanent states are clearly visible. The triangles (circles) correspond to increasing
(decreasing) the in-plane magnetic field. The arrows indicate the relative orientation of the detector (2) and
injector (3) contact magnetizations. (After Jiang, X. et al., Phys. Rev. Lett. 90, 256603, 2003. With permission.)
pure spin diffusion current is analyzed at contact 2, which is outside of the

charge current path (“non-local”), where Δµ is manifested as a voltage. No
charge current flows in the detection circuit defined by reference contact
1 and magnetic detector contact 2, thus excluding spurious contributions
from anisotropic magnetoresistance and local Hall effects.
The non-local voltage VNL measured at the detector contact 2 is sensitive
to the relative orientation of the contact magnetization and that of the spin
polarization in the Si beneath it. In Figure 2.28b, VNL is plotted as a function
of an applied in-plane magnetic field By used to switch the magnetization of
the injector and detector. At large negative field, the magnetizations are par-
allel, and the majority spins injected into the Si are parallel to the electron
spin orientation of the detector, resulting in a positive voltage at contact 2
corresponding to the difference between the point indicated on the “spin-
up” curve under contact 2 in Figure 2.28a and the spin ground. As the field
is increased (triangles), contact 2 with the lower coercive field switches so
that the injector/detector magnetizations are antiparallel. The orientation
of the spin current diffusing to contact 2 is now antiparallel to that of the
detector, producing an abrupt change at By ∼ 50 Oe and a plateau in the non-
local voltage, with the value corresponding to the difference between the
point indicated on the “spin-down” curve under contact 2 (Figure 2.28a) and
the spin ground. As By is increased further, contact 3 also switches so that
injector and detector contact magnetizations are again parallel, and the non-
local voltage abruptly changes to its original value. The process is repeated
as the field is swept in the opposite direction (circles, Figure 2.28b), produc-
ing an output characteristic similar to that seen in metal pseudo spin-valve
structures [191], and confirming remanent (nonvolatile) parallel and antipar-

allel contact orientations.
When a positive bias is applied at contact 3, electrons flow from the Si
into the Fe contact (Figure 2.27b). The higher conductivity of the Fe majority
spin channel results in efficient majority spin current, leaving an accumula-
tion of minority spin electrons in the Si and a reversal of the spin-splitting of
the chemical potential. This bias configuration is often referred to as “spin
extraction [192]”, since majority spins are extracted from the semiconductor.
A minority rather than majority pure spin current now diffuses from the
injector to the detector contact 2, resulting in an inversion of the non-local
voltage peaks for antiparallel alignment of injector and detector contact
magnetizations, as seen in Figure 2.29. The magnitude of the non-local volt-
age is roughly linear with the magnitude of the bias current.
Thus, the orientation of the spin in the Si, and the corresponding split-
ting of the chemical potential, can be reversed either by switching the con-
tact magnetization or by changing the polarity of the bias voltage on the
injecting contact.
FIGURE 2.29 Non-local voltage vs. in-plane magnetic field at 10 K for several values
of the injection current. Graphs are offset for clarity. For negative bias, electrons are
injected from Fe into the Si channel, and the change in non-local voltage is consistent
with majority spin injection. For positive bias, electrons are extracted from the silicon
into the Fe contact. The majority spins are more readily extracted, resulting in the
accumulation of minority spins in the silicon. A change in sign is seen for non-local
voltage peaks for the antiparallel state, consistent with minority spin accumulation.
(After Jiang, X. et al., Phys. Rev. Lett. 90, 256603, 2003. With permission.)
Further confirmation of spin transport and another avenue enabling

manipulation of the spin in the silicon is provided by the Hanle effect [193],
in which a magnetic field perpendicular to the surface (electron spin) induces
precession and dephasing of the spins in the silicon, resulting in a modulation
and eventual suppression of the detected NL MR signal. The Hanle effect is
widely regarded as the gold standard of spin injection and transport, provid-
ing definitive proof of the existence of a spin-polarized electron population
[136, 137, 181, 186, 187]. The Hanle effect curve at T = 10 K and a bias current
of −100 µA (spin injection) is shown in Figure 2.30. The magnetizations of
the injector and detector contacts are placed in the parallel remanent state,
and a magnetic field (Hanle field, Bz) is applied perpendicular to the surface
of the device. As the Hanle field increases from zero, the injected spins in the
silicon precess around Bz during transit to the detector contact, resulting in
both dephasing and an increasing degree of antiparallel alignment relative
to the detector spin orientation, producing a corresponding decrease in the
magnitude of the non-local signal. Variations in transit times in the trans-
port channel due to the width of the injector and detector contacts truncate
FIGURE 2.30 The Hanle effect curve at T = 10 K and a bias current of −100 µA
(spin injection) obtained for the four terminal non-local spin valve contact
geometry with Fe/Al2O3/Si contacts. The spin in the n-Si transport channel
(n ∼ 2 − 3 × 1018 cm−3) precesses during transit between injector and detector con-
tacts due to an applied perpendicular magnetic field. The magnetizations of the
injector and detector contacts are in the parallel remanent state. The solid line is
a fit to the data using the model described in the text, and yields a spin lifetime of
1 ns and diffusion constant of 10 cm2/s.
the full Hanle curve and suppress further precessional oscillations in the
non-local voltage. If the injector bias is reversed to operate in the spin extrac-
tion mode, or the injector and detector contacts are placed in the antiparallel
remanent state, the sign of the Hanle curve reverses [187, 188]. Hanle curves
are visible to measurement temperatures of ∼150 K for the NLSV geometry
and doping levels n ∼ 2 × 1018 cm−3, used in these samples.
The Hanle lineshape can be fit analytically to obtain the spin lifetime
in this pure spin current. A fit to the data using an approach similar to that
of Ref. [137] (and discussed in Chapter 6, Volume 2) is shown by the smooth
curve in Figure 2.30, and yields a spin lifetime of ∼1 ns at T = 10 K. The life-
times for majority and minority spins (referenced to the injector contact 3)
in the Si channel are comparable, as expected for a nonmagnetic semicon-
ductor. This value is shorter than in the bulk at a comparable electron den-
sity and temperature. The relatively short spin lifetimes are attributed to the
fact that this lateral transport geometry probes spin diffusion near the Si/
Al2O3 interface, where surface scattering and interface states are likely to
produce more rapid spin relaxation than in the bulk.
NLSV measurements using the geometry of Figure 2.26 were also
reported recently in degenerately doped n-Si (doped well above the metal–
insulator transition) by Ando et al. using Fe3Si/Si Schottky tunnel barrier
contacts [194], and by Sasaki et al. using Fe/MgO tunnel barrier contacts
[195]. Both groups observed the non-local magnetoresistance behavior
(Figure 2.28b) due to reversal of the contact magnetization. However, neither
reported a Hanle signal, and thus provide no information on spin lifetimes.
By measuring the dependence of the NLSV signal on injector/detector con-
tact spacing, Sasaki et al. obtained an estimate for the spin diffusion length
of 2.25 µm at T = 8 K and a carrier density n ∼ 1 × 1020 cm−3. They also report
that a measurable NL MR signal persists to T = 120 K.
Thus, an FM metal/Al2O3 tunnel barrier contact can be effectively used
as both a spin injector and a spin detector for Si, providing a direct voltage
readout of the Si spin polarization in a lateral transport device. With the non-
local geometry, one can generate and detect a pure spin current that flows
outside of the charge current path. The spin orientation of this pure spin cur-
rent is controlled in one of three ways: (a) by switching the magnetization of
the FM metal injector contact; (b) by changing the polarity of the bias on the
“injector” contact, which enables the generation of either majority or minor-
ity spin populations in the Si, and provides a way to electrically manipulate
the injected spin orientation without changing the magnetization of the con-
tact itself; and (c) by inducing spin precession through application of a small
perpendicular magnetic field. The contact bias and magnetization, together
with spin precession, allow full control over the orientation of the spin in
the silicon channel and subsequent detection as a voltage, demonstrating
that information can be transmitted and processed with pure spin currents
in silicon. The FM contacts provide nonvolatile functionality as well as the
potential for reprogrammability. This is accomplished in a lateral transport
geometry using lithographic techniques compatible with existing device
geometries and fabrication methods.
FIGURE 2.31 (a) Three terminal measurement geometry for injection and detec-
tion of spin accumulation under left contact. (b) Energy band diagram of the junc-
tion, depicting the ferromagnet, tunnel barrier, and n-type Si conduction and VBs.
(After Dash, S.P. et al., Nature 462, 491, 2009. With permission.)
2.6.3 Three Terminal Spin Injection and Detection in Si at

Room Temperature: Spin-Polarized Charge Current
A simpler three terminal geometry can also be used to generate and mea-
sure spin accumulation in a semiconductor [136, 190], as illustrated in Figure
2.31 for Si. In this case a single FM tunnel barrier contact serves as both the
spin injector and detector, and the spin accumulation Δµ measured is that
which develops directly under the contact in the presence of a spin-polarized
charge current. As discussed in the next section, this spin accumulation does
not necessarily develop in the semiconductor itself—it may, in fact, develop in
the interface states or trap states within the oxide tunnel barrier. Thus, spin
injection into the semiconductor is not required to generate spin accumula-
tion and corresponding voltage signal in this measurement geometry. The
density of such states depends upon the procedures used to form the inter-
face and oxide layer, and may vary widely. Spin accumulation in such states
may indeed prove to be just as interesting and technologically relevant as
in the semiconductor, but may complicate and compromise spin injection/
transport/detection in the semiconductor channel.
This three terminal approach was used by Dash et al. [196] to demon-
strate spin accumulation and electrical detection in heavily n- and p-doped
Si at room temperature. They used FM metal/Al2O3/Si tunnel barrier con-
tacts for which spin injection into Si had been successfully demonstrated
[154]. In contrast with previous work, their Si was degenerately doped, with
n-(p-) doping of 1.8 × 1019 cm−3 (4.8 × 1018 cm−3), well above the metal–insula-
tor transition. The Hanle effect was used to observe the spin accumulation
and determine the corresponding spin lifetimes directly under the contact
as a function of bias and temperature.
A cross-sectional diagram of the contact interface region illustrating
the spin injection, accumulation, and subsequent precession induced by the
perpendicular magnetic field Bz is shown in Figure 2.32a. At Bz = 0, the spin
accumulation Δµ builds and reaches a static equilibrium value. For Bz ≠ 0,
the spins precess at the Larmor frequency ωL = gµBBz/ħ, resulting in a reduc-
tion of the net spin accumulation due to precessional dephasing (the Hanle
effect). Here, g is the Lande g-factor, µB the Bohr magneton, and ħ is the
reduced Planck’s constant. Thus, the voltage at the detector contact decreases
as the magnitude of Bz increases with an approximately Lorentzian lineshape
given by Δµ(B) = Δµ(0)/[1 + (ωLτs)2)], and the spin lifetime τs is obtained from
fits to this lineshape.
The Hanle data showing spin accumulation at the Ni80Fe20/Al 2O3/Si
contact are shown in Figure 2.32b and c as a function of temperature. A
clear signal due to spin accumulation is observed at room temperature,
and increases at lower temperatures. The magnitude of the spin accumu-
lation Δµ at the tunnel interface is obtained from the magnitude of the
Hanle signal ΔV = TSP × Δµ/2e, where TSP = 0.3 is the known tunneling
spin polarization [183] for Al 2O3/Ni80Fe20 at 300 K. Thus, Δµ = 1.2 meV
at 300 K, and increases to a value of 3 meV at 5 K. Dash et al. [196] note
that their measured voltage and these Δµ values are about two orders of
magnitude larger than that expected from the model of Fert and Jaffrès
for spin injection and accumulation in the semiconductor itself [97, 100].
They suggest that lateral inhomogeneity of the tunnel current leads to
higher localized spin injection current densities, but an alternate explana-
tion involving accumulation in localized interface states [190] must also
be considered, as discussed below. Al 2O3/Si interfaces formed by sputter-
or electron-beam deposition of the oxide are known to exhibit large inter-
face state densities [197].
The data are well fit by a simple Lorentzian of the form described above,
as shown by the solid line in Figure 2.32b. The full width at half maximum
corresponds to ωL = 1/τs, yielding a spin lifetime τs = 142 ps at 300 K for the
heavily doped n-Si with a measured electron density of 1.8 × 1019 cm−3. The
lifetime does not change significantly with temperature to 5 K. Similar
results were obtained for p-doped Si, with the surprising result that the hole
spin lifetimes deduced were a factor of two longer (∼270 ps at 300 K) than
the electron spin lifetimes, in marked contrast to what is typically reported
for III–V semiconductors such as GaAs. This is inconsistent with the known
physics of spin relaxation in simple cubic semiconductors like Si and GaAs.
FIGURE 2.32 (a) Illustration of the Hanle effect, in which spin precession produced by a perpendicular
magnetic field leads to a decay of the spin accumulation as measured by the voltage at the injection/detector
terminal in the three terminal geometry. A plot of the voltage vs. magnetic field gives a pseudo-Lorentzian
lineshape from which the spin lifetime can be determined. (b) Hanle curves shown for measurement tempera-
tures of 5–300 K. A spin lifetime of 142 ps is obtained from the fit (solid line), and is independent of tempera-
ture. (After Dash, S.P. et al., Nature 462, 491, 2009. With permission.)
The degeneracy of the light hole and heavy hole states in the valence band
results in very rapid hole spin depolarization on the sub-picosecond time
scale, so that hole spin lifetimes are much shorter than those of the electron.
The long hole spin lifetimes reported, and the fact that they are longer than
the nominal electron spin lifetimes measured using similar samples and
methodology, raise concern as to whether the spin signal measured derives
from carriers in the silicon or from localized states at the oxide interface. The
latter will be discussed in Section 2.6.4.
More recent work [198], has utilized Co0.9Fe0.1/SiO2 and Ni0.8 Fe0.2/SiO2
tunnel barrier contacts on Si(001) [170] in a similar three terminal geometry,
and observed spin accumulation and the Hanle effect to 500 K. This is a sig-
nificant temperature milestone, since commercial and military electronic
systems must operate at temperatures well above room temperature. The
temperature 500 K easily exceeds the maximum temperature specification
required to meet commercial (360 K), industrial (375 K), and even military
(400 K) ratings. SiO2 is highly desirable as a tunnel barrier for at least two
reasons: (1) the SiO2/Si interface is well-developed and known to have a low
level of interface trap states, and (2) the existing semiconductor processing
infrastructure is very familiar and comfortable with this system, facilitating
future transition.
The measurements were performed on silicon-on-insulator (SOI) sam-
ples with a doping levels of n(As) = 3 × 1018 cm−3 to 6 × 1019 cm−3, above and
below the metal–insulator transition. The SiO2 tunnel barrier was formed in
situ by plasma oxidation, followed immediately by sputter-beam deposition
of a 10 nm Co0.9Fe0.1 film, which was subsequently patterned to define the
magnetic contacts. Nonmagnetic reference contacts were deposited by lift-
off. The Hanle data obtained are shown in Figure 2.33a. The signal amplitude
is nearly 0.2 mV at low temperature, and signal is clearly visible at 500 K. The
signal amplitude is in good agreement with that predicted by theory—the
measured spin resistance-area product AΔV3 T(Bz = 0)/I is within a factor of
2–3 of the simplest model, which assumes that the observed spin accumula-
tion occurs in the bulk [190] (A is the contact area, I is the bias current, and
ΔV3 T(Bz = 0) is the measured Hanle amplitude).
The Hanle curves exhibit the classic pseudo-Lorentzian lineshape
expected, and can be fit very well with the Lorentzian model, described
above. The spin lifetimes thus obtained are ∼100 ps and essentially
FIGURE 2.33 (a) Hanle curves obtained at various temperatures to 500 K for Co0.9Fe0.1/SiO2 tunnel barrier
contacts on Si(001). The amplitude decreases with increasing temperature. The linewidth and correspond-
ing spin lifetime show little variation with temperature. (b) Temperature dependence of the amplitude of the
Hanle signal. The inset shows the three terminal measurement geometry.
independent of temperature. This is similar to spin lifetimes measured in

bulk Si at similar carrier densities of n = 3.4 × 1019 cm−3 (As), and attributed
to the metallic character of the sample, where impurity dominated spin–
orbit interaction dominates spin relaxation for a wide temperature range
[199]. The amplitude of the signal ΔV3 T(B z = 0) decreases monotonically
with temperature, as shown in Figure 2.33b. This behavior is inconsistent
with that of the spin lifetime, indicating that other factors, such as spin
injection/detection efficiency, may be responsible. Thus, spin precessional
effects in this technologically dominant material are readily observed at
the elevated temperatures necessary for utilization in commercial elec-
tronic systems.
2.6.4 Three Terminal Spin Injection and

Detection in GaAs: Interpretation and
the Role of Localized States
A clear interpretation of such three terminal Hanle data is still being devel-
oped. In contrast with the NLSV (four terminal) geometry, where the spin
current must flow through the semiconductor before producing the spin
accumulation enabling detection at a second contact, in the three terminal
geometry, it is not always clear that spin transport/accumulation occurs in
the semiconductor itself.
Significant insight was provided by recent experiments of Tran et al.
[190] who used the three terminal geometry and Hanle measurements with
Co/Al2O3/n-GaAs(001) samples. The GaAs 50 nm thick channel was heav-
ily n-doped (Si) at 5 × 1018 cm−3 (well above the metal–insulator transition of
2 × 1017 cm−3) on a semi-insulating substrate, with a 15 nm thick surface layer
at 2 × 1019 cm−3 to minimize the depletion region and facilitate spin injection.
They observed a Hanle signal ΔV ∼ 1.2 meV at T = 10 K (Figure 2.34a), corre-
sponding to a spin accumulation Δµ = 2e (ΔV/γ) ∼ 6 meV (where γ = 0.4 is the
tunnel spin polarization for Co/Al2O3), much larger than that expected from
the predictions of the standard theory of spin injection in the diffusive regime
for a single ferromagnet/semiconductor interface [97, 100]. They concluded
that their measured signal did not originate from spin accumulation and pre-
cession in the GaAs channel. Rather, states localized near the Al2O3/GaAs
interface served as a spin accumulation layer with long spin lifetimes leading
to exceptionally large signal levels, as illustrated in Figure 2.34b. These local-
ized states participate in a sequential tunneling process, and act as a confin-
ing layer which enhances the spin accumulation. The thickness of the layer
containing these states was estimated to be ∼1 nm. The authors speculate
that the origin of these states may be either ionized Si donors within the
depletion layer or interface states at the Al2O3/GaAs interface. The oxide/
GaAs interface is well-known to have a high density of interface states, which
has thwarted the development of a metal-oxide-semiconductor transistor
technology [200]. Thus, the voltage measured in such a three-terminal geom-
etry will include contributions from spin accumulation in localized states
associated with the tunnel barrier and corresponding interfaces, and may
FIGURE 2.34 (a) Hanle signal obtained in a three terminal geometry for a 15 × 196 µm2 Co/Al2O3 tunnel bar-
rier contact to GaAs. A small quadratic background has been subtracted. The solid line is a fit to the data using
the model described assuming spin accumulation in interface states. The left inset shows the bias depen-
dence of the Hanle signal, and the right inset illustrates the spin splitting of the electrochemical potential in
the interfacial region. (After Tran, M. et al., Phys. Rev. Lett. 102, 036601, 2009. With permission.) (b) Schematic
of the measurement geometry and cross-section of the region under the magnetic tunnel barrier contact,
illustrating the location of interface states which the authors concluded produced the spin accumulation and
strong Hanle signal measured.
not demonstrate spin injection or accurately reflect spin accumulation in the

semiconductor itself. Four terminal (NLSV) measurements from the same
samples, in which the spin current must flow through the semiconductor to
a second detector contact, should provide valuable insight and clarification.
2.7 GRAPHENE AS A TUNNEL BARRIER FOR

SPIN INJECTION/DETECTION IN SILICON
As discussed in the preceding sections, a tunnel barrier between a ferro-
magnetic metal and the semiconductor circumvents the large conductiv-
ity mismatch and enables electrical spin injection and detection. Extensive
effort has been directed towards developing appropriate tunnel barriers for
spin contacts, with most work focusing on a reverse-biased ferromagnetic
Schottky barrier or an insulating oxide layer such as Al2O3 or MgO. However,
metal Schottky barriers and oxide layers are susceptible to interdiffusion,
interface defects and trapped charge, which have been shown to compro-
mise spin injection, transport and detection. Ferromagnetic metals readily
form silicides even at room temperature, and diffusion of the ferromagnetic
species into the silicon creates magnetic scattering sites, limiting spin dif-
fusion lengths and spin lifetimes in the silicon. Even a well-developed and
widely utilized oxide such as SiO2 is known to have defects and trapped or
mobile charge, which limit both charge and spin-based performance. Such
approaches also result in contacts with high resistance–area (RA) products,
exacerbating practical issues of local heating and power consumption. More
2.7 Graphene as a Tunnel Barrier for Spin Injection/Detection in Silicon 123
importantly, previous work has shown that smaller RA products within a

select window of values are essential for efficient spin injection and detec-
tion [97,100].
An ideal tunnel barrier should exhibit several key material characteristics:
ªª a uniform and planar habit with well-controlled thickness,

ªª minimal defect and trapped charge density,
ªª a low RA product for minimal power consumption,
ªª chemical stability,
ªª compatibility with both the ferromagnetic metal and the semicon-
ductor of choice,
ªª thermal stability to ensure minimal diffusion to/from the surround-
ing materials at the temperatures required for device processing.
Graphene, an atomically thin honeycomb lattice of carbon, offers a

compelling option. Although it is very conductive in plane, it exhibits low
conductivity perpendicular to the plane [201]. Its sp2 bonding results in a
highly uniform, defect-free layer that is chemically inert, thermally robust,
and essentially impervious to diffusion [202]. Recent work has in fact dem-
onstrated that single-layer graphene can be used as a spin tunnel barrier
between two metals in a magnetic tunnel junction [203, 204].
In this section, we show that a ferromagnetic metal/monolayer graphene
contact serves as a spin-polarized tunnel barrier contact that provides effi-
cient electrical spin injection and detection in planar silicon [205, 206] and
silicon nanowires [207] using both the three terminal and four terminal non-
local spin valve transport geometries described previously. The graphene
barrier also enables one to achieve contact RA products that fall within the
critical window of values required for practical devices. We demonstrate
electrical injection and detection of spin accumulation in silicon above
room temperature, and show that the corresponding spin lifetimes correlate
with the silicon donor concentration, confirming that the spin accumula-
tion measured occurs in the silicon and not in the graphene or interface
trap states. The RA products are three orders of magnitude lower than those
achieved with oxide tunnel barrier contacts on silicon substrates with identi-
cal doping levels.
2.7.1 Three-Terminal Device Geometry

Graphene was grown by low-pressure chemical vapor deposition within cop-
per foil ‘enclosures’ [208], and transferred onto hydrogen-passivated n-type
silicon (001) substrates with electron density n = 1 × 1019, 3 × 1019 and 6 × 1019
cm−3. Ni80Fe20 was deposited on top of the graphene via standard liftoff tech-
niques to form the device structures illustrated in Figure 2.35. Details of the
fabrication process may be found in Ref. [205].
Current-voltage (I-V) curves for several devices at room temperature
are shown in Figure 2.36a, including reference contacts consisting of NiFe
deposited directly onto the hydrogen-passivated Si substrate, and onto Al2O3
(a) (b) E
V
2
1 3
NiFe Si
N (E) N (E) N (E) N (E)
reference
NiFe
Graphene
Silicon
FIGURE 2.35 Schematic of the samples. (a) Monolayer graphene serves as a

tunnel barrier between the FM metal contact and the Si substrate. Contacts 1
and 3 are ohmic Ti/Au contacts. (b) Schematic illustrating spin injection and spin
accumulation.
FIGURE 2.36 Electrical characteristics of the NiFe/tunnel barrier/Si contacts. (a) The current-voltage curves
for NiFe/Si (red), NiFe/graphene/Si (black), NiFe/ 2 nm SiO2/Si (green), and NiFe/Al2 O3 /Si (blue) contacts at RT.
The Si doping concentration is 1 × 1019 cm− 3 (triangles), 3 × 1019 cm− 3 (diamonds), and 6 × 1019 cm− 3 (circles).
(b) The normalized zero bias resistance (ZBR) shows a weak insulator-like temperature dependence for the
NiFe/graphene contacts, confirming tunnel transport. Each solid color line is from a different contact, illustrat-
ing the reproducibility of the data. The ohmic NiFe/Si contacts exhibit metallic behavior; results for three
contacts are shown by triangles, red-dashed and green-dashed lines.
and SiO2 tunnel barriers. The highest resistance is obtained for the NiFe/
SiO2/Si samples and the lowest for the NiFe/Si reference samples. The cor-
responding RA products on the Si (6 × 1019 cm−3) substrate are 0.4 kΩ um2
for direct contact (which is ohmic), 6 kΩ um2 for the graphene barrier, and
15 MΩ um2 for the SiO2 barrier, demonstrating that the graphene barrier
produces an RA product orders of magnitude smaller than the oxides for
thicknesses necessary to produce a pinhole free tunnel barrier.
Figure 2.36b shows the temperature dependence of the zero bias resis-
tance (ZBR) for NiFe/monolayer graphene/Si (n = 6 × 1019 cm−3) contacts. The
weak temperature dependence confirms that transport occurs by tunneling
through a pin-hole-free tunnel barrier, as discussed in Section 2.5.3.3.1. The
Al2O3 and SiO2 contacts also show a weak, insulating-like dependence on
temperature, confirming that they form pinhole free tunnel barriers. In con-
trast, the direct NiFe/Si contact exhibits metallic behavior.
Spin accumulation and precession directly under the magnetic tunnel
barrier contact were measured using the Hanle effect in the three termi-
nal geometry as described in Section 2.6.3, and the results are shown in
Figure 2.37. The top three curves in Figure 2.37a are the Hanle measure-
ments of the graphene, SiO2 and Al 2O3 tunnel barrier samples at room
temperature. All three samples show a clear Hanle effect, evidenced by
a Lorentzian line shape caused by field-induced precession of the spin
population, confirming successful spin accumulation. Note that the gra-
phene barrier produces the highest spin signal, yet has a significantly lower
RA product than the oxide barrier contacts. The bottom two curves in
Figure 2.37a show results for additional reference samples consisting of
the NiFe/Si and a non-magnetic/graphene/Si sample, measured at 10 K.
No Hanle effect is observed even at these low temperatures, as expected.
The temperature dependence of the Hanle signal for the NiFe/graphene/Si
(1 × 1019 cm−3) samples is shown in Figure 2.37b. Much larger signals are
observed at low temperature, and the amplitude decreases monotonically
with temperature.
No evidence for a Hanle-like signal is observed when the magnetic field is
applied in-plane along the long axis of the magnetic contact. The appearance
8
a) RT b)
6
20
Spin Resistance (mW)
3T Voltage @ -1mA (m V)
4
NiFe/Gr/Si
2 0.0
NiFe/SiO2/Si
0
NiFe/Al2O3/Si
-20
4K
-2 50 K
Nonmagnetic/Gr/Si @10K 100 K
-40 150 K
-4 NiFe/Si @ 10K 200 K
300 K
-6 -60
-4000 -2000 0 2000 4000 -5000 0 5000
Magnetic Field (Oe) Magnetic field (Oe)
FIGURE 2.37 Hanle spin precession measurements. (a) Room temperature Hanle data for spin injection
NiFe/Al2 O3 /Si (3 × 1019 ) (red), NiFe/SiO2 /Si (3 × 1019 ) (blue), and NiFe/Graphene/Si (1 × 1019 ) (green). Also shown
are the control samples; nonmagnetic/ graphene/Si (1 × 1019 ) and NiFe/Si (1 × 1019 ) measured at 10 K . (b)
Temperature dependent Hanle data for NiFe/Graphene/Si (1 × 1019 ).
of such an “inverted” Hanle effect has been attributed to precessional dephas-

ing produced by random fringe fields arising from surface/interface rough-
ness of the magnetic contact [209], although other factors may contribute
[210]. The absence of an inverted Hanle signal implies that the NiFe/graphene
tunnel barrier produces a more uniform magnetic film and interface which
suppress such effects relative to that provided by the NiFe/oxide interface.
Values for the spin lifetime are obtained from fits to the Hanle curves
[205] using the Lorentzian described above. The spin lifetime depends
strongly on contact bias and silicon donor density, and weakly on temperature
due to the metallic character of the silicon [199]. The spin lifetime decreases
with increasing donor density, as expected from electron spin resonance
(ESR) measurements on bulk silicon [211, 212]. This is shown explicitly in
Figure 2.38, where we plot the spin lifetime obtained from three-terminal
Hanle data as a function of electron density for four different tunnel barrier
materials (graphene, Al2O3, MgO and SiO2) and three different magnetic
metal contacts (Fe, CoFe and NiFe). At T = 4 K, fits to the Hanle data yield
spin lifetimes of 140 ps and 105 ps for the NiFe/graphene/Si(1 × 1019) and
Si(6 × 1019) samples, respectively. These values agree well with those reported
for NiFe/SiO2 tunnel barrier contacts [198], and show a clear correlation with
the character of the silicon.
The spin lifetime measured with the three-terminal Hanle geometry
shows a clear dependence only on electron density, and the dependence is
consistent with literature ESR data on bulk silicon [211, 212]. The spin life-
time is completely independent of the tunnel barrier material or magnetic
400
Graphene
350 SiO2
Al2O3
MgO
Spin lifetime (ps)
300
250
200
150
100
50
18 19 20
10 10 10
-3
Electron concentration (cm )
FIGURE 2.38 Spin lifetimes obtained from three-terminal Hanle measurements

at 10 K as a function of the Si electron density for the tunnel barrier materials
indicated and different ferromagnetic metal contacts (Fe, CoFe, NiFe). The symbol
shape distinguishes the tunnel barrier material: triangles – SiO2 , circles – Al2 O3 ,
and stars – graphene. Solid symbols correspond to devices with Ni0.8 Fe0.2 contacts,
half-solid symbols to Fe contacts, and open symbols to Co0.9 Fe0.1 contacts. The spin
lifetimes show a pronounced dependence on the Si doping level, and little depen-
dence on the choice of tunnel barrier or magnetic metal.
metal used for the contact. The values for the graphene tunnel barriers fall
directly on the curve. These data confirm that the spin accumulation occurs
in the silicon, and not in the graphene or possible interface trap states. The
measured spin lifetimes are shorter than those in bulk silicon because they
reflect the environment directly beneath the contact, where the reduced
symmetry and increased scattering from the interface are likely to produce
additional spin scattering, as discussed in Ref. [198].
The magnetic contact’s conventional RA product is an important param-
eter in determining the practical application of a spin-based semiconductor
device (note that this is the standard resistance–area product and not the
spin resistance–area product). Calculations have shown that significant local
(two-terminal) magnetoresistance (MR) can be achieved only if the contact
RA product falls within a specific range which depends upon the Si channel
conductivity, the spin lifetime, and the contact spacing and width [97, 100,
213]. The RA products of tunnel barrier contacts to date have been much
larger than required, making such devices unattainable. Previous work to
lower the RA product utilized a low-work function metal such as gadolinium
at the tunnel barrier interface, but no spin accumulation in the semiconduc-
tor was demonstrated [214]. In contrast, the low RA products provided by
the graphene tunnel barriers fall well within this window, and enable realiza-
tion of advanced spintronic devices.
We calculate the range of optimum RA products and the corresponding
local MR as a function of the Si electron density using the methodology of
Ref. [97] and the contact geometry shown as the inset to Figure 2.39. Further
details can be found in Ref. [205]. The geometric parameters are chosen to be
consistent with the node anticipated for Si device technology within the next
5 years. The color code in Figure 2.39 identifies the range of useful MR and
the corresponding window of contact RA products required. Tunnel bar-
rier contacts of FM/Al2O3 and FM/SiO2 fabricated in our lab on identical
substrates have been shown to produce significant spin accumulation in Si,
but have RA products that are too high to generate usable local MR for oxide
thicknesses sufficient to produce a pinhole free barrier. In contrast, using
monolayer graphene as the tunnel barrier lowers the RA product by orders
of magnitude, and values for the NiFe/graphene contacts on bulk wafers
fall well within the range required to generate high local MR. Reducing the
RA product also has a positive effect on the electrical properties of the spin
device, as lowering the resistance reduces noise and increases the speed of
an electrical circuit [215].
2.7.2 Four-Terminal Non-Local Spin Valve Geometry

Magnetic metal/graphene tunnel barrier contacts have also been employed
in the four-terminal non-local spin valve (NLSV) geometry (described in
Sections 2.6.1 and 2.6.2) to generate pure spin currents in silicon and deter-
mine spin lifetimes from the corresponding Hanle effect. The spin lifetimes
are the same as those obtained from the three-terminal geometry devices
described above, providing further confirmation that the spin accumulation
FIGURE 2.39 Resistance– area product window for local magnetoresistance.

Calculation of the local (two-terminal) magnetoresistance (MR) as a function of the
conventional resistance– area product of the contact and the Si electron density
for the device geometry shown in the inset, using the theory of Ref. [11]. Data
points are the resistance– area products measured for our ferromagnetic metal/
tunnel barrier/Si contacts using 2 nm SiO2 (triangles), 1.5 nm Al2 O3 (diamond) and
monolayer graphene (circles) tunnel barriers prepared from identical Si wafers in
our laboratory. The ferromagnetic metal/graphene resistance– area products fall
within the window of useful magnetoresistance values. W = w = 11 nm.
indeed occurs in the silicon channel rather than in trap states localized at
the interface or in some oxide. The existence of such localized trap states
and their potential contribution to spin accumulation has been reported for
other spin tunnel barrier structures [190], particularly those purposefully
fabricated to maximize the density of such states [216, 217].
Non-local spin valve devices were fabricated using Ni80Fe20/monolayer
graphene tunnel barriers on the same silicon substrates described above
and in Refs. [205] and [206]. Monolayer graphene grown by chemical vapor
deposition was transferred onto hydrogen passivated Si(001) wafers, and
NiFe injector and detector contacts defined by electron-beam lithography
and liftoff techniques. The contacts were either 200 nm and 1 μm wide, or
200 nm and 200 nm wide, with edge-to-edge separations of 200 nm or 400
nm. These contacts were then used as a hard mask to etch away the exposed
graphene by a light oxygen plasma, confirmed by a resistance check. A sche-
matic of the device and a scanning electron microscope (SEM) image of the
injector/detector area are shown in Figure 2.40.
Spin-polarized charge current injected at the NiFe/graphene contact 2
generates a spin accumulation in the silicon beneath the contact. Spin diffu-
sion causes a pure spin current to flow in the silicon channel away from the
point of injection, and a corresponding spin splitting of the electrochemical
potential is detected as a voltage on the NiFe/graphene detector contact 3.
The results of several devices on a silicon wafer with n = 1 × 1019 cm−3 are
shown in Figure 2.41 at T = 10 K. The non-local resistance is defined to be
FIGURE 2.40 (a) Schematic of non-local spin valve device. Contacts 1 and 4 are
ohmic contacts applied directly to the Si substrate, and contacts 2 (1 μ m wide) and
3 (200 nm wide) are Au capped Ni80 Fe20 /graphene tunnel barriers for electrical
spin injection and detection. (b) SEM image of the active area of the NLSV device,
dashed line indicates 20 μ m × 30 μ m window, contacts are 200 nm and 1 μ m wide
separated edge-to-edge by 200 nm.
the voltage measured at the non-local detector contact 3 divided by the bias
current applied between contacts 1 and 2. An in plane magnetic field is used
to switch the magnetizations of the two electrodes independently. When the
magnetizations of injector and detector contacts are parallel, a low non-local
resistance is measured, and when the magnetizations are antiparallel a high
non-local resistance is measured – this is the classic signature of a NLSV
device. A net change in the non-local resistance of 1–15 milli-ohm was mea-
sured at 10K, with the larger values exhibited by devices in which both injec-
tor and detector contacts were 200 nm.
Application of an out-of-plane magnetic field causes the spins to precess
and dephase as they diffuse towards the detector contact 3, and the non-local
voltage decreases with increasing magnetic field due to this Hanle effect. The
non-local Hanle data are shown in Figure 2.42 after background subtraction,
and can be modeled using the integral spin diffusion approach described in
Section 2.6.2. The model fits the experimental data well, as shown by the red
FIGURE 2.41 (a) NLSV measurement at 10 K where both injector and detector
are 200 nm wide with an edge-to-edge separation of 200 nm. (b) The same for
devices with a 1 μ m wide injector and a 200 nm wide detector, with an edge-to-
edge separation of 200 nm. The silicon electron concentration is 1 × 1019 cm− 3 .
FIGURE 2.42 Non-local Hanle data after background subtraction at T = 10 K . The
solid line is a fit to the data and yields a spin lifetime of 140 ps. The silicon electron
density is n = 1 × 1019 cm− 3 .
curve in Figure 2.42, and yields values for the diffusion constant D = 2.9 cm2/
sec and the silicon spin lifetime of 140 ± 10 ps. This spin lifetime is the same
as that obtained from the three-terminal data described above for the same
electron density, providing further confirmation that the three-terminal
data are not corrupted by parasitic effects such as localized trap states.
These three-terminal and NLSV results demonstrate that a ferromag-

netic metal/monolayer graphene contact serves as a spin-polarized tun-
nel barrier contact with a low RA product, a crucial requirement enabling
future semiconductor spintronic devices. Utilizing multilayer rather than
monolayer graphene in such structures may provide much higher values of
the tunnel spin polarization due to the band structure derived spin filtering
effects that have been predicted for selected ferromagnetic metal/multilayer
graphene structures [218, 219]. Such an increase will improve the perfor-
mance of semiconductor spintronic devices by providing higher signal-to-
noise ratios and corresponding operating speeds.
2.7.3 Spin Transport in Silicon Nanowires

Semiconductor nanowires (NWs) provide an avenue to further reduce the
ever-shrinking dimensions of transistors. Realization of spin-based Si NW
devices requires efficient electrical spin injection and detection, which
depend critically on the interface resistance between a ferromagnetic metal
contact and the NW. This is especially problematic with semiconducting
NWs because of the exceedingly small contact area, which can be of order
100 nm2. Research has shown that standard oxide tunnel barriers can pro-
vide good spin injection into planar Si structures, but such contacts grown
on NWs are often far too resistive to yield reliable and consistent results.
The graphene tunnel barriers described above offer a solution—they form
low RA product tunnel contacts on bulk Si, while providing spin-injecting
properties superior to those of oxide tunnel barriers.
In this section we describe fabrication of nanoscale spintronic devices
comprised of a Si nanowire channel with low-resistance graphene tunnel
barriers for electrical spin injection and detection of pure spin currents.
Both four-terminal NLSV and three-terminal (3T) geometries are used to
probe spin accumulation and transport. In contrast with previous work on
semiconductor NWs which reported no Hanle effect [220–224], Hanle spin
precession is observed here in both measurements [207]. Because spurious
effects can produce spin-valve-like behavior even in the non-local geometry
[225], these Hanle data provide confirmation of spin accumulation and pure
spin transport, enabling direct measurements of spin lifetimes and diffusion
lengths in these challenging NW structures.
Si nanowires were grown using gold seeded vapor-liquid-solid tech-
niques with a <111> growth axis and phosphorous doped to an electron
concentration n ~ 3 × 1019 cm−3 [226]. They were removed from the growth
substrate by sonication and dispersed on a Si3N4 /Si substrate. Appropriate
nanowires ~ 100–150 nm in diameter were chosen optically, and two
outer ohmic contacts were made using electron-beam lithography. NiFe/
graphene tunnel barriers were fabricated as described above to form the
devices illustrated in Figure 2.43 on which both 3T and NLSV measure-
ments could be made. These magnetic contacts, labeled 2 and 3, are 1 μm
and 200 nm wide respectively, separated edge-to-edge by 200 nm. These
contacts are then used as a hard mask to remove the excess graphene by
FIGURE 2.43 (a) False color atomic force microscopy image of a silicon nanowire with the four contacts
used in the spin measurements. The ferromagnetic metal/graphene tunnel barrier contacts used to inject
and detect spin appear as blue, the gold ohmic reference contacts appear as yellow, and the light green line
is the silicon nanowire transport channel. The bright dot on the end of the nanowire is the gold nanoparticle
used to seed the nanowire growth. (b) Schematic of the four terminal nanowire device in the non-local spin
valve geometry. A spin-polarized charge current is injected at the NiFe/graphene contact 2, generating a pure
spin current that flows to the right within the silicon nanowire. This spin current generates a voltage that is
detected on NiFe/graphene contact 3.
a light O2 plasma etch. Measurements of the I-V curves and temperature

dependence of the zero bias resistance were used to confirm the tunneling
character of the NiFe/graphene contacts. Further fabrication details may
be found in Ref. [207].
The spin injecting properties and spin-RA products of the NiFe/gra-
phene tunnel barrier contacts were initially assessed using the 3T geom-
etry in which a single FM contact is used to both generate and detect spin
accumulation in the area of the NW immediately beneath the contact. A
current is applied between contacts 1 and 3 (see Figure 2.44a), and injection
of spin-polarized electrons from the NiFe/graphene contact 3 produces a net
spin accumulation in the Si nanowire. This results in a spin splitting of the
electrochemical potential which is detected as a voltage measured across
contacts 3 and 4. When a magnetic field Bz is applied along the surface nor-
mal and perpendicular to the electron spin direction, the spins precess and
dephase, reducing the net spin accumulation and the corresponding volt-
age at contact 3, as illustrated in Figure 2.44b. A plot of the detector volt-
age versus the applied field reveals a clear Hanle signal with the expected
Lorentzian lineshape, as shown in Figure 2.44c. Fits to the data as described
in section 2.6.3 yield a spin lifetime of 260 ± 10 ps, a factor of 2 longer than
that obtained in planar Si devices of similar doping level with either oxide
[198] or graphene tunnel barrier contacts [205]. No evidence for a Hanle-like
signal is observed when the magnetic field is applied in-plane.
These spin lifetimes are an order of magnitude shorter than the 2.5 ns
lifetime obtained from electron spin resonance measurements on bulk sam-
ples of similar carrier density [199]. We attribute this to the proximity of
the FM contact interface where magnetic fringe fields due to contact edges
or closure domains may lead to strong local spin precession and dephasing
[227, 228]. In addition, spins can scatter at the FM contact interface, or read-
ily diffuse between the Si and FM metal contact through the relatively trans-
parent tunnel barrier.
FIGURE 2.44 Three terminal Hanle measurements. (a) Schematic of the 3T measurement; (b) Simplified
diagram illustrating spin injection from the NiFe across the graphene tunnel barrier producing spin accumula-
tion in the Si NW, and the Hanle spin precession measurement. N↑ (E) and N↓ (E) are the majority and minority
spin density of states, respectively. (c) 3T Hanle measurements for a 150 nm diameter NW at 10 K and − 200 μ A
bias current (spin injection) after a quadratic background subtraction and plotted as resistance. The solid line
is a fit of the data.
One expects a much larger enhancement of the spin lifetime in a true

one-dimensional (1D) structure relative to bulk material due to the discon-
tinuous density of states that reduces available spin relaxation channels
[229]. However, Si nanowires are expected to exhibit true 1D behavior only
for diameters less than ~ 10 nm [230, 231]. The diameter of the nanowires
used here and in measurements by other groups [220–224] are all much
larger due to limitations in fabrication, and therefore too wide to exhibit an
ideal 1D density of states and the corresponding full enhancement in spin
lifetime. The fact that the Hanle-derived spin lifetime is not reduced rela-
tive to values obtained from similar measurements on bulk planar samples
indicates that surface scattering due to the much larger surface-to-volume
ratio for the NW does not play a more significant role at these relatively high
doping levels, or that another relaxation mechanism dominates.
The same samples are used in the NLSV geometry of Figure 2.43b to
demonstrate the generation and detection of pure spin current in the NW
channel produced by the NiFe/graphene contacts. As described in Sections
2.6.1 and 2.6.2, the pure spin current flowing from the injector to detector
FIGURE 2.45 Non-local spin-valve and Hanle measurements. (a) NLSV data on
150 nm wide nanowire, contacts are 200 nm and 1 μ m wide with a 200 nm edge-
to-edge spacing. The arrows indicate the relative orientation of the magnetiza-
tions of the injector and detector contacts, and the blue and red traces indicate
increasing and decreasing magnetic field sweeps. (b) NLSV Hanle data for the Si
NW for parallel alignment of contacts 2 and 3 after background subtraction. All
data are for a bias current of –200 μ A at T = 10 K
contact creates a spin accumulation and spin-splitting of the electrochemical

potential which appears as a voltage on the detector contact. The results are
shown in Figure 2.45a. The non-local resistance is defined to be the voltage
measured at the non-local detector contact divided by the injector contact
bias current. An in plane magnetic field is used to switch the magnetizations
of the two electrodes independently due to their different coercive fields,
producing parallel and antiparallel alignments. When the magnetizations
of injector and detector contacts are parallel, a low non-local resistance is
measured, and when the magnetizations are antiparallel a high non-local
resistance is measured—this is the classic signature of a NLSV device. A net
change in the non-local resistance of ~ 0.7 mΩ is measured at 10K.
2.8 Outlook and Critical Research Issues 135
The magnetizations of the two contacts reverse cleanly at well-defined

coercive fields consistent with shape anisotropy, so that both parallel and
antiparallel alignment of the injector/detector magnetizations are realized.
This is attributed to the smooth bridging of the NW by the graphene layer,
producing a more gradual variation in the contact topography to which
the NiFe film conforms. In contrast, magnetic metal/oxide tunnel contacts
applied directly to the NW exhibit poorly defined switching characteris-
tics attributed to local domain patterns arising from the more complex
morphology created as the contact attempts to conform to the NW side-
walls [221].
Application of an out-of-plane magnetic field causes the spins to precess
and dephase as they diffuse towards the detector contact, and the non-local
voltage decreases with increasing magnetic field due to this Hanle effect. The
NLSV Hanle data are shown in Figure 2.45b, and can be modeled using the
integral spin diffusion approach described in Section 2.6.2. The model fits
the experimental data well, as shown by the red curve in Figure 2.45b, and
yields values for the diffusion constant D = 0.00047 m2/sec and the NLSV
spin lifetime of 190 ± 10 ps. This effective spin lifetime is shorter than the
value of 260 ± 10 ps obtained in the 3T geometry, although the reasons are
unclear at present. The spin diffusion length is calculated as LSD = [D τs]0.5 ~
300 nm, which is significantly shorter than results reported previously [220,
221]. However, these were based only on the amplitude of the NLSV data
rather than direct determination of the spin lifetime provided by the Hanle
measurements.
Achieving good contacts remains one of the largest technological bar-
riers to exploiting geometrical nanowires for both charge and spin-based
applications. The single layer graphene barriers described here provide a high
quality tunnel barrier with a low RA product free of pinholes that smoothly
bridges the nanowire. This results in clean magnetic switching characteris-
tics for the magnetic contacts, and enables the first observation of Hanle spin
precession of both spin accumulation in the nanowire in the 3T geometry,
and of the pure spin current generated in the nanowire transport channel
using the NLSV geometry. The Hanle data provide direct measurements of
the spin lifetime and spin diffusion lengths in these nanoscale spintronic
devices.
2.8 OUTLOOK AND CRITICAL RESEARCH ISSUES

A great deal of progress has been made in spin injection and transport in
semiconductors in the last 15 years. The contributions from many groups
summarized here and elsewhere have rapidly advanced the state-of-the-art
in electrical injection, detection, and spin manipulation in semiconductors
of immediate technological interest such as GaAs and Si. There are several
important areas that require further in-depth research.
Interfaces continue to play a critical role in at least two ways. First,
they impact spin injection efficiency from a given contact, which has been
the primary topic of this chapter. Second, they impact spin lifetimes in lat-
eral transport structures or in any thin film, nanoscale device. While spin
lifetimes have been reasonably well-studied in bulk Si, there is very little
information available on these lifetimes near interfaces. In Si, the reduced
symmetry will enable spin relaxation via mechanisms which are symmetry-
forbidden in the bulk, such as D’yakonov–Perel’. For all semiconductors, the
specific interface structure, defects and impurities, and band bending are
expected to play a role, but the relative significance of these contributions is
yet to be determined.
Contact resistance plays a major role in the relative efficiency of a spin-
injecting contact, as illustrated by several model calculations, and described
in Sections 2.5.2 and 2.5.3. While a large value seems, at first glance, to be
desirable for spin injection, this will severely limit the spin-polarized cur-
rent flow through the contact into the semiconductor, and thereby limit the
spin accumulation achieved. Therefore, a compromise must be reached. If
one considers electrical detection, further constraints apply as discussed by
Min et al. [183]—for the case of moderately doped Si (n ∼ 1016 cm−3), contact
resistances of order 10−2–10−5 Ω cm2 will be required. Similar considerations
apply to other materials such as GaAs, although the specific values vary.
These values are achievable even with tunnel barriers, but will be strongly
case-specific and application-dependent.
Utilizing intrinsic 2D layers such as graphene or hexagonal boron nitride
as tunnel contacts offers many advantages over conventional materials
deposited by vapor deposition (e.g. Al2O3 or MgO), enabling a path to highly
scaled electronic and spintronic devices. The use of multilayer rather than
single layer graphene in such structures may provide much higher values of
the tunnel spin polarization because of band structure derived spin filter-
ing effects predicted for selected ferromagnetic metal/multilayer graphene
structures [218, 219]. This increase would further improve the performance
of nanowire spintronic devices by providing higher signal to noise ratios and
corresponding operating speeds, advancing the techological applications of
nanowire devices.
ACKNOWLEDGMENTS
It is a pleasure to acknowledge the many colleagues and collaborators who
have made significant contributions through either experimental work,
theory, or discussion: Brian Bennett, Steve Erwin, Aubrey Hanbicki, George
Kioseoglou, Jim Krebs, Connie Li, Athos Petrou, Andre Petukhov, Philip
Thompson, Olaf van’t Erve, and many others. I wish to especially acknowl-
edge Gary Prinz, who provided me with the opportunity to enter and
contribute to the field of magnetoelectronics. Various aspects of the work
presented here were supported by the Office of Naval Research, DARPA
and core program at the Naval Research Laboratory. This effort would not
have been possible without the long-term basic research support of both the
Office of Naval Research and NRL.
References 137
REFERENCES
1. G. Moore, Cramming more components onto integrated circuits, Electronics
38 (April 19, 1965).
2. G. A. Prinz, Magnetoelectronics, Science 282, 1660 (1998).
3. S. A. Wolf, D. D. Awschalom, R. A. Buhrman et al., Spintronics: A spin-based
electronics vision for the future, Science 294, 1448 (2001).
4. International Technology Roadmap for Semiconductors, Executive Summary,
Emerging Research Devices, and Emerging Research Materials, 2009. http://
www.itrs.net.
5. F. Meier and B. P. Zakharchenya (Eds.), Optical Orientation, vol. 8, North-
Holland, New York, 1984.
6. G. Lampel, Nuclear dynamic polarization by optical electronic saturation and
optical pumping in semiconductors, Phys. Rev. Lett. 20, 491 (1968).
7. S. A. Crooker, J. J. Baumberg, F. Flack, N. Samarth, and D. D. Awschalom,
Terahertz spin precession and coherent transfer of angular momenta in mag-
netic quantum wells, Phys. Rev. Lett. 77, 2814 (1996).
8. M. J. Stevens, A. L. Smirl, R. D. R. Bhat, A. Najmaie, J. E. Sipe, and H. M. van
Driel, Quantum interference control of ballistic pure spin currents in semi-
conductors, Phys. Rev. Lett. 90, 136603 (2003).
9. S. F. Alvarado and P. Renaud, Observation of spin-polarized-electron tunnel-
ing from a ferromagnet into GaAs, Phys. Rev. Lett. 68, 1387 (1992).
10. B. T. Jonker, Progress toward electrical injection of spin-polarized electrons
into semiconductors, Proc. IEEE 91, 727 (2003).
11. B. T. Jonker, S. C. Erwin, A. Petrou, and A. G. Petukhov, Electrical spin injec-
tion and transport in semiconductor spintronic devices, MRS Bull. 28, 740
(2003).
12. B. T. Jonker and M. E. Flatte, Electrical spin injection and transport in semi-
conductors, in Nanomagnetism, Volume 1: Ultrathin Films, Multilayers and
Nanostructures, D. L. Mills and J. A. C. Bland (Eds.); in Contemporary Concepts
of Condensed Matter Science, E. Burstein, M. L. Cohen, D. L. Mills, and P. J.
Stiles (Eds.), Elsevier, Amsterdam, the Netherlands, p. 227, 2006.
13. C. Weisbuch and B. Vinter, Quantum Semiconductor Structures, Academic
Press, New York, p. 65, 1991.
14. A. G. Aronov and G. E. Pikus, Spin injection into semiconductors, Fiz. Tekh.
Poluprovodn. 10, 1177 (1976) [Sov. Phys. Semicond. 10, 698 (1976)]; A. G.
Aronov, Pis’ma Zh. Eksp. Teor. Fiz. 24, 37 (1976) [JETP Lett. 24, 32 (1976)].
15. D. Chiba, M. Sawicki, Y. Nishitani, Y. Nakatani, F. Matsukura, and H. Ohno,
Magnetization vector manipulation by electric fields, Nature 455, 515 (2008).
16. C. Haas, Magnetic semiconductors, CRC Crit. Rev. Solid State Sci. 1, 47 (1970).
17. E. L. Nagaev, Photoinduced magnetism and conduction electrons in magnetic
semiconductors, Phys. Stat. Sol. B 145, 11 (1988).
18. L. Esaki, P. J. Stiles, and S. von Molnar, Magnetointernal field emission in junc-
tions of magnetic insulators, Phys. Rev. Lett. 19, 852 (1967).
19. J. S. Moodera, X. Hao, G. A. Gibson, and R. Meservey, Electron–spin polariza-
tion in tunnel junctions in zero applied field with ferromagnetic EuS barriers,
Phys. Rev. Lett. 61, 637 (1988).
20. J. S. Moodera, R. Meservey, and X. Hao, Variation of the electron spin polariza-
tion in EuSe tunnel junctions from zero to near 100% in a magnetic field, Phys.
Rev. Lett. 70, 853 (1993).
21. Y. D. Park, A. T. Hanbicki, J. E. Mattson, and B. T. Jonker, Epitaxial growth
of an n-type ferromagnetic semiconductor CdCr2Se4 on GaAs (001) and GaP
(001), Appl. Phys. Lett. 81, 1471 (2002).
22. G. Kioseoglou, A. T. Hanbicki, J. M. Sullivan et al., Electrical spin injection
from an n-type ferromagnetic semiconductor into a III–V device heterostruc-
ture, Nat. Mater. 3, 799 (2004).
23. J. Lettieri, V. Vaithyanathan, S. K. Eah et al., Epitaxial growth and magnetic

properties of EuO on (001)Si by molecular beam epitaxy, Appl. Phys. Lett. 83,
975 (2003).
24. A. Schmehl, V. Vaithyanathan, A. Herrnberger et al., Epitaxial integration of
the highly spin-polarized ferromagnetic semiconductor EuO with silicon and
GaN, Nat. Mater. 6, 882 (2007).
25. J. K. Furdyna and J. Kossut, Diluted magnetic semiconductors, in
Semiconductors and Semimetals, vol. 25, R. K. Willardson and A. C. Beer
(series Eds.), Academic Press, New York, 1988.
26. N. Samarth and J. K. Furdyna, Diluted magnetic semiconductors, Proc. IEEE
78, 990 (1990).
27. M. Oestreich, J. Hubner, D. Hagele et al., Spin injection into semiconductors,
Appl. Phys. Lett. 74, 1251 (1999).
28. K. Onodera, T. Masumoto, and M. Kimura, 980 nm compact optical isolators
using Cd1−x−yMn xHg yTe single crystals for high power pumping laser diodes,
Electron. Lett. 30, 1954 (1994).
29. H. Munekata, H. Ohno, S. von Molnar, A. Segmüller, L. L. Chang, and L. Esaki,
Diluted magnetic III–V semiconductors, Phys. Rev. Lett. 63, 1849 (1989).
30. J. De Boeck, R. Oesterholt, A. Van Esch et al., Nanometer-scale magnetic
MnAs particles in GaAs grown by molecular beam epitaxy, Appl. Phys. Lett.
68, 2744 (1996).
31. H. Ohno, A. Shen, F. Matsukura et al., (Ga, Mn)As: A new diluted magnetic
semiconductor based on GaAs, Appl. Phys. Lett. 69, 363 (1996).
32. S. J. Potashnik, K. C. Ku, S. H. Chun, J. J. Berry, N. Samarth, and P. Schiffer,
Effects of annealing time on defect-controlled ferromagnetism in Ga1−xMn x As,
Appl. Phys. Lett. 79, 1495 (2001).
33. K. W. Edmonds, K. Y. Wang, R. P. Campion et al., High-Curie-temperature
Ga1−xMn x As obtained by resistance-monitored annealing, Appl. Phys. Lett. 81,
4991 (2002).
34. T. Hayashi, Y. Hashimoto, S. Katsumota, and Y. Iye, Effect of low-temperature
annealing on transport and magnetism of diluted magnetic semiconductor
(Ga, Mn)As, Appl. Phys. Lett. 78, 1691 (2001).
35. Y. Yuan, Y. Wang, K. Gao et al., High Curie temperature and perpendicular
magnetic anisotropy in homoepitaxial InMnAs films, J. Phys. D Appl. Phys. 48,
235002 (2015).
36. L. Chen, X. Yang, F. Yang et al., Enhancing the Curie temperature of ferro-
magnetic semiconductor (Ga,Mn)As to 200 K via nanostructure engineering,
Nano Lett. 11, 2548 (2011).
37. H. Ohno, Making nonmagnetic semiconductors ferromagnetic, Science 281,
951 (1998).
38. H. Ohno, N. Akiba, F. Matsukura, A. Shen, K. Ohtani, and Y. Ohno,
Spontaneous splitting of ferromagnetic (Ga,Mn)As valence band observed by
resonant tunneling spectroscopy, Appl. Phys. Lett. 73, 363 (1998).
39. Y. Ohno, D. K. Young, B. Beschoten, F. Matsukura, H. Ohno, and D. D.
Awschalom, Electrical spin injection in a ferromagnetic semiconductor het-
erostructure, Nature 402, 790 (1999).
40. D. K. Young, E. Johnston-Halperin, D. D. Awschalom, Y. Ohno, and H. Ohno,
Anisotropic electrical spin injection in ferromagnetic semiconductor hetero-
structures, Appl. Phys. Lett. 80, 1598 (2002).
41. M. Yamanouchi, D. Chiba, F. Matsukura, and H. Ohno, Current-induced
domain-wall switching in a ferromagnetic semiconductor structure, Nature
428, 539 (2004).
42. H. Ohno, D. Chiba, F. Matsukura et al., Electric field control of ferromagne-
tism, Nature 408, 944 (2000).
43. Y. D. Park, A. T. Hanbicki, S. C. Erwin et al., A group-IV ferromagnetic semi-
conductor: Mn xGe1−x, Science 295, 651 (2002).
References 139
44. A. M. Nazmul, S. Kobayashi, S. Sugahara, and M. Tanaka, Control of ferro-

magnetism in Mn delta doped GaAs based semiconductor heterostructures,
Physica E 21, 937–943 (2004).
45. T. Dietl, H. Ohno, F. Matsukura, J. Cibert, and D. Ferrand, Zener model descrip-
tion of ferromagnetism in zinc-blende magnetic semiconductors, Science 287,
1019 (2000).
46. T. Dietl, H. Ohno, and F. Matsukura, Hole-mediated ferromagnetism in tetra-
hedrally coordinated semiconductors, Phys. Rev. B 63, 195205 (2001).
47. A. H. MacDonald, P. Schiffer, and N. Samarth, Ferromagnetic semiconductors:
Moving beyond (Ga,Mn)As, Nat. Mater. 4, 195 (2005).
48. H. Raebiger, S. Lany, and Z. Zunger, Control of ferromagnetism via electron
doping in In2O3:Cr, Phys. Rev. Lett. 101, 027203 (2008).
49. N. Samarth, Ferromagnetic semiconductors: Battle of the bands, Nat. Mater.
11, 360 (2012).
50. S. Souma, L. Chen, R. Oszwałdowski et al., Fermi level position, Coulomb gap,
and Dresselhaus splitting in (Ga,Mn)As. Sci. Rep. 6, 27266 (2016).
51. M. Dobrowolska K. Tivakornsasithorn, X. Liu et al., Controlling the Curie
temperature in (Ga,Mn)As through location of the Fermi level within the
impurity band, Nat. Mater. 11, 444 (2012).
52. M. Kobayashi I. Muneta, Y. Takeda et al., Unveiling the impurity band induced
ferromagnetism in the magnetic semiconductor (Ga,Mn)As, Phys. Rev. B 89,
205204 (2014).
53. T. Jungwirth J. Wunderlich, V. Novák et al., Spin-dependent phenomena and
device concepts explored in (Ga,Mn)As, Rev. Mod. Phys. 86, 855 (2014).
54. T. Dietl and H. Ohno, Dilute ferromagnetic semiconductors: Physics and spin-
tronic structures, Rev. Mod. Phys. 86, 187 (2014).
55. B. T. Jonker, Polarized optical emission due to decay or recombination of spin-
polarized injected carriers, U.S. patent 5,874,749 (filed June 23, 1993, awarded
February 23, 1999 to U.S. Navy).
56. G. Bastard, Wave Mechanics Applied to Semiconductor Heterostructures,
Halsted, Halsted, New York, p. 109, 1988.
57. D. Hagele, M. Oestreich, W. W. Ruhle, N. Nestle, and K. Eberl, Spin transport
in GaAs, Appl. Phys. Lett. 73, 1580 (1998).
58. M. I. D’yakonov and V. I. Perel, Chapter 2: Theory of optical spin orientation of
electrons and nuclei in semiconductors, in Optical Orientation, F. Meier and B. P.
Zakharchenya (Eds.), North-Holland, Amsterdam, the Netherlands, p. 274, 1984.
59. R. W. Martin, R. J. Nicholas, G. J. Rees, S. K. Haywood, N. J. Mason, and P. J.
Walker, Two-dimensional spin confinement in strained-layer quantum wells,
Phys. Rev. B 42, 9237 (1990).
60. S. A. Crooker, D. D. Awschalom, J. J. Baumberg, F. Flack, and N. Samarth,
Optical spin resonance and transverse spin relaxation in magnetic semicon-
ductor quantum wells, Phys. Rev. B 56, 7574 (1997).
61. R. Fiederling, P. Grabs, W. Ossau, G. Schmidt, and L. W. Molenkamp, Detection
of electrical spin injection by light-emitting diodes in top- and side-emission
configurations, Appl. Phys. Lett. 82, 2160 (2003).
62. O. M. J. van’t Erve, G. Kioseoglou, A. T. Hanbicki, C. H. Li, and B. T. Jonker,
Remanent electrical spin injection from Fe into AlGaAs/GaAs light emitting
diodes, Appl. Phys. Lett. 89, 072505 (2006).
63. A. Tackeuchi, O. Wada, and Y. Nishikawa, Electron spin relaxation in InGaAs/
InP multiple quantum wells, Appl. Phys. Lett. 70, 1131 (1997).
64. A. F. Isakovic, D. M. Carr, J. Strand, B. D. Schultz, C. J. Palmstrom, and P. A.
Crowell, Optical pumping in ferromagnet-semiconductor heterostructures:
Magneto-optics and spin transport, Phys. Rev. B 64, 161304(R) (2001).
65. B. T. Jonker, A. T. Hanbicki, Y. D. Park et al., Quantifying electrical spin injec-
tion: Component-resolved electroluminescence from spin-polarized light-
emitting diodes, Appl. Phys. Lett. 79, 3098 (2001).
66. G. Kioseoglou, A. T. Hanbicki, B. T. Jonker, and A. Petrou, Bias-controlled hole

degeneracy and implications for quantifying spin polarization, Appl. Phys.
Lett. 87, 122503 (2005).
67. R. Fiederling, M. Kelm, G. Reuscher et al., Injection and detection of a spin-
polarized current in a light-emitting diode, Nature 402, 787 (1999).
68. B. T. Jonker, Y. D. Park, B. R. Bennett, H. D. Cheong, G. Kioseoglou, and
A. Petrou, Robust electrical spin injection into a semiconductor heterostruc-
ture, Phys. Rev. B 62, 8180 (2000).
69. D. K. Young, J. A. Gupta, E. Johnston-Halperin, R. Epstein, Y. Kato, and D. D.
Awschalom, Optical, electrical and magnetic manipulation of spins in semi-
conductors, Semicond. Sci. Technol. 17, 275 (2002).
70. R. M. Stroud, A. T. Hanbicki, Y. D. Park et al., Reduction of spin injection
efficiency by interface defect spin scattering in ZnMnSe/AlGaAs-GaAs spin-
polarized light emitting diodes, Phys. Rev. Lett. 89, 166602 (2002).
71. P. R. Hammar, B. R. Bennett, M. J. Yang, and M. Johnson, Observation of spin
injection at a ferromagnet–semiconductor interface, Phys. Rev. Lett. 83, 203
(1999).
72. C.-M. Hu, J. Nitta, A. Jensen, J. B. Hansen, and H. Takayanagi, Spin-polarized
transport in a two dimensional electron gas with interdigital-ferromagnetic
contacts, Phys. Rev. B 63, 125333 (2001).
73. S. Gardelis, C. G. Smith, C. H. W. Barnes, E. H. Linfield, and D. A. Ritchie,
Spin-valve effects in a semiconductor field-effect transistor: A spintronic
device, Phys. Rev. B 60, 7764 (1999).
74. F. G. Monzon, H. X. Tang, and M. L. Roukes, Magnetoelectronic phenomena
at a ferromagnet–semiconductor interface, Phys. Rev. Lett. 84, 5022 (2000).
75. B. J. van Wees, Comment on observation of spin injection at a ferromagnet–
semiconductor interface, Phys. Rev. Lett. 84, 5023 (2000).
76. A. T. Filip, B. H. Hoving, F. J. Jedema, B. J. van Wees, B. Dutta, and S. Borghs,
Experimental search for the electrical spin injection in a semiconductor, Phys.
Rev. B 62, 9996 (2000).
77. H. J. Zhu, M. Ramsteiner, H. Kostial, M. Wassermeier, H.-P. Schönherr, and K.
H. Ploog, Room temperature spin injection from Fe into GaAs, Phys. Rev. Lett.
87, 016601 (2001).
78. R. C. Miller, D. A. Kleinman, W. A. Nordland Jr., and A. C. Gossard,
Luminescence studies of optically pumped quantum wells in GaAs-AlxGa1−x
As multilayer structures, Phys. Rev. B 22, 863 (1980).
79. W. H. Butler, X.-G. Zhang, X. Wang, J. van Ek, and J. M. MacLaren, Electronic
structure of FM/semiconductor/FM spin tunneling structures, J. Appl. Phys.
81, 5518 (1997).
80. J. M. MacLaren, W. H. Butler, and X.-G. Zhang, Spin-dependent tunneling in
epitaxial systems: Band dependence of conductance, J. Appl. Phys. 83, 6521
(1998).
81. J. M. MacLaren, X.-G. Zhang, W. H. Butler, and X. Wang, Layer KKR approach
to Bloch-wave transmission and reflection: Application to spin-dependent
tunneling, Phys. Rev. B 59, 5470 (1999).
82. O. Wunnicke, Ph. Mavropoulos, R. Zeller, P. H. Dederichs, and D. Grundler,
Ballistic spin injection from Fe(001) into ZnSe and GaAs, Phys. Rev. B 65,
241306(R) (2002).
83. P. Mavropoulos, O. Wunnicke, and P. H. Dederichs, Ballistic spin injection and
detection in Fe/semiconductor/Fe junctions, Phys. Rev. B 66, 024416 (2002).
84. J. R. Chelikowsky and M. L. Cohen, Nonlocal pseudopotential calculations for
the electronic structure of eleven diamond and zinc-blende semiconductors,
Phys. Rev. B 14, 556 (1976).
85. S. Vutukuri, M. Chshiev, and W. H. Butler, Spin-dependent tunneling in FM/
semiconductor/FM structures, J. Appl. Phys. 99, 08K302 (2006).
86. P. Mavropoulos, Spin injection from Fe into Si(001): Ab initio calculations and
role of the Si complex band structure, Phys. Rev. B 78, 054446 (2008).
References 141
87. M. Zwierzycki, K. Xia, P. J. Kelly, G. E. W. Bauer, and I. Turek, Spin injection

through an Fe/InAs interface, Phys. Rev. B 67, 092401 (2003).
88. W. H. Butler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren, Spin-
dependent tunneling conductance of Fe/MgO/Fe sandwiches, Phys. Rev. B 63,
054416 (2001).
89. T. Valet and A. Fert, Theory of the perpendicular magnetoresistance in mag-
netic multilayers, Phys. Rev. B 48, 7099 (1993).
90. B. T. Jonker, A. T. Hanbicki, D. T. Pierce, and M. D. Stiles, Spin nomencla-
ture for semiconductors and magnetic metals, J. Magn. Magn. Mater. 277, 24
(2004).
91. R. A. de Groot, New class of materials: Half-metallic ferromagnets, Phys. Rev.
Lett. 50, 2024 (1983).
92. W. E. Pickett and J. S. Moodera, Half metallic magnets, Phys. Today 54(5), 39
(May 2001).
93. D. Orgassa, H. Fujiwara, T. C. Schulthess, and W. H. Butler, First-principles
calculation of the effect of atomic disorder on the electronic structure of the
half-metallic ferromagnet NiMnSb, Phys. Rev. B 60, 13237 (1999).
94. G. Schmidt, D. Ferrand, L. W. Molenkamp, A. T. Filip, and B. J. van Wees,
Fundamental obstacle for electrical spin injection from a ferromagnetic metal
into a diffusive semiconductor, Phys. Rev. B 62, R4790 (2000).
95. E. I. Rashba, Theory of electrical spin injection: Tunnel contacts as a solution
of the conductivity mismatch problem, Phys. Rev. B 62, R16267 (2000).
96. D. L. Smith and R. N. Silver, Electrical spin injection into semiconductors,
Phys. Rev. B 64, 045323 (2001).
97. A. Fert and H. Jaffrès, Conditions for efficient spin injection from a ferromag-
netic metal into a semiconductor, Phys. Rev. B 64, 184420 (2001).
98. Z. G. Yu and M. Flatte, Electric-field dependent spin diffusion and spin injec-
tion into semiconductors, Phys. Rev. B 66, 201202(R) (2002).
99. P. C. van Son, H. van Kempen, and P. Wyder, Boundary resistance of the ferro-
magnetic-nonferromagnetic metal interface, Phys. Rev. Lett. 58, 2271 (1987).
100. A. Fert, J.-M. George, H. Jaffrès, and R. Mattana, Semiconductors between
spin-polarized source and drain, IEEE Trans. Electron. Dev. 54, 921 (2007).
101. E. H. Rhoderick and R. H. Williams, Metal-Semiconductor Contacts, 2nd edn.,
Clarendon. Oxford, UK, 1988.
102. M. Ilegems, Properties of III–V layers, in The Technology and Physics of
Molecular Beam Epitaxy, E. H. C. Parker (Ed.), Plenum, New York, p. 119, 1985.
103. S. M. Sze, Physics of Semiconductor Devices, 2nd edn., Wiley, New York, p. 294,
1981.
104. A. T. Hanbicki, B. T. Jonker, G. Itskos, G. Kioseoglou, and A. Petrou, Efficient
electrical spin injection from a magnetic metal/tunnel barrier contact into a
semiconductor, Appl. Phys. Lett. 80, 1240 (2002).
105. A. T. Hanbicki, O. M. J. van’t Erve, R. Magno, G. Kioseoglou, C. H. Li, and B. T.
Jonker, Analysis of the transport process providing spin injection through an
Fe/AlGaAs Schottky barrier, Appl. Phys. Lett. 82, 4092 (2003).
106. J. M. Kikkawa and D. D. Awschalom, Resonant spin-amplification in n-type
GaAs, Phys. Rev. Lett. 80, 4313 (1998).
108. B. J. Jönsson-Åkerman, R. Escudero, C. Leighton, S. Kim, I. K. Schuller, and
D. A. Rabson, Reliability of normal-state current-voltage characteristics as
an indicator of tunnel-junction barrier quality, Appl. Phys. Lett. 77, 1870
(2000).
109. D. A. Rabson, B. J. Jönsson-Åkerman, A. H. Romero et al., Pinholes may mimic
tunneling, J. Appl. Phys. 89, 2786 (2001).
110. U. Rüdiger, R. Calarco, U. May et al., Temperature dependent resistance of
magnetic tunnel junctions as a quality proof of the barrier, J. Appl. Phys. 89,
7573 (2001).
111. E. Burstein and S. Lundqvist (Eds.), Tunneling Phenomena in Solids, Plenum,

New York 1969.
112. J. G. Simmons, Generalized formula for the electric tunnel effect between simi-
lar electrodes separated by a thin insulating film, J. Appl. Phys. 34, 1793 (1963).
113. W. F. Brinkman, R. C. Dynes, and J. M. Rowell, Tunneling conductance of
asymmetrical barriers, J. Appl. Phys. 41, 1915 (1970).
114. S. H. Chun, S. J. Potashnik, K. C. Ku, P. Schiffer, and N. Samarth, Spin-polarized
tunneling in hybrid metal-semiconductor magnetic tunnel junctions, Phys.
Rev. B 66, 100408(R) (2002).
115. E. L. Wolf, Principles of Electron Tunneling Spectroscopy, Oxford University
Press, New York, 1989.
116. A. M. Andrews, H. W. Korb, N. Holonyak, C. B. Duke, and G. G. Kleiman, Phys.
Rev. B 5, 2273 (1972).
117. J. J. Krebs and G. A. Prinz, NRL Memo Rep. 3870, Naval Research Lab,
Washington, DC, 1978.
118. R. Meservey and P. M. Tedrow, Spin-polarized electron tunneling, Phys. Rep.
238, 173–234 (1999). See 200–202.
119. C. Kittel, Introduction to Solid State Physics, 5th edn., Wiley, New York, pp.
438–439, 1976.
120. N. W. Ashcroft and N. D. Mermin, Solid State Physics, Holt, Rinehart, Winston,
New York, pp. 654, 661, 1976.
121. M. B. Stearns, Simple explanation of tunneling spin polarization of Fe, Co, Ni
and its alloys, J. Magn. Magn. Mater. 5, 167 (1977).
122. J. M. De Teresa, A. Barthelemy, A. Fert, J. P. Contour, F. Montaigne, and
P. Seneor, Role of metal-oxide interface in determining the spin polarization of
magnetic tunnel junctions, Science 286, 507 (1999).
123. H. Dery and L. J. Sham, Spin extraction theory and its -relevance to spintron-
ics, Phys. Rev. Lett. 98, 046602 (2007).
124. A. N. Chantis, K. D. Belashchenko, D. L. Smith, E. Y. Tsymbal, M. van
Schilfgaarde, and R. C. Albers, Reversal of spin polarization in Fe/GaAs(001)
driven by resonant surface states: First-principles calculations, Phys. Rev. Lett.
99, 196603 (2007).
125. T. J. Zega, A. T. Hanbicki, S. C. Erwin et al., Determination of interface atomic
structure and its impact on spin transport using Z-contrast microscopy and
density-functional theory, Phys. Rev. Lett. 96, 196101 (2006).
126. S. C. Erwin, S.-H. Lee, and M. Scheffler, First-principles study of nucleation,
growth, and interface structure of Fe/GaAs, Phys. Rev. B 65, 205422 (2002).
127. E. J. Kirkland, Advanced Computing in Electron Microscopy, Plenum, New
York, 1998.
128. E. Kneedler, P. M. Thibado, B. T. Jonker et al., Epitaxial growth, structure, and
composition of Fe films on GaAs(001)-2 × 4, J. Vac. Sci. Technol. B 14, 3193
(1996).
129. E. M. Kneedler and B. T. Jonker, Kerr effect study of the onset of magnetization
in Fe films on GaAs(001)-2 × 4, J. Appl. Phys. 81, 4463 (1997).
130. E. M. Kneedler, B. T. Jonker, P. M. Thibado, R. J. Wagner, B. V. Shanabrook, and
L. J. Whitman, Influence of substrate surface reconstruction on the growth
and magnetic properties of Fe on GaAs(001), Phys. Rev. B 56, 8163 (1997).
131. P. M. Thibado, E. Kneedler, B. T. Jonker, B. R. Bennett, B. V. Shanabrook, and L.
J. Whitman, Nucleation and growth of Fe on GaAs(001)-2 × 4 studied by scan-
ning tunneling microscopy, Phys. Rev. B 53, 10481 (1996).
132. B. T. Jonker, O. J. Glembocki, R. T. Holm, and R. J. Wagner, Enhanced carrier
lifetimes and suppression of midgap states in GaAs at a magnetic metal inter-
face, Phys. Rev. Lett. 79, 4886 (1997).
133. C. S. Gworek, P. Phatak, B. T. Jonker, E. R. Weber, and N. Newman, Pressure
dependence of Cu, Ag, and Fe n-GaAs Schottky barrier heights, Phys. Rev. B
64, 045322 (2001).
References 143
134. B. T. Jonker, Electrical spin injection into semiconductors, in Ultrathin

Magnetic Structures IV: Applications of Nanomagnetism, B. Heinrich and J. A.
C. Bland (Eds.), Springer, Berlin, Germany, 2005.
135. S. A. Crooker, M. Furis, X. Lou et al., Imaging spin transport in lateral ferro-
magnet/semiconductor structures, Science 309, 2191 (2005).
136. X. Lou, C. Adelmann, M. Furis, S. A. Crooker, C. J. Palmstrøm, and P. A.
Crowell, Electrical detection of spin accumulation at a ferromagnet-semicon-
ductor interface, Phys. Rev. Lett. 96, 176603 (2006).
137. X. Lou, C. Adelmann, S. A. Crooker et al., Electrical detection of spin trans-
port in lateral ferromagnet-semiconductor devices, Nat. Phys. 3, 197 (2007).
138. M. C. Hickey, C. D. Damsgaard, S. N. Holmes et al., Spin injection from
Co2MnGa into an InGaAs quantum well, Appl. Phys. Lett. 92, 232101 (2008).
139. M. Holub, J. Shin, D. Saha, and P. Battacharya, Electrical spin injection and
threshold reduction in a semiconductor laser, Phys. Rev. Lett. 98, 146603
(2007).
140. P. Kotissek, M. Bailleul, M. Sperl et al., Cross-sectional imaging of spin injec-
tion into a semiconductor, Nat. Phys. 3, 872 (2007).
141. A. T. Hanbicki, G. Kioseoglou, M. A. Holub, O. M. J. van’t Erve, and B. T.
Jonker, Electrical spin injection from Fe into ZnSe(001), Appl. Phys. Lett. 94,
082507 (2009).
142. G. Kioseoglou, A. T. Hanbicki, R. Goswami et al., Electrical spin injection into
Si: A comparison between Fe/Si Schottky and Fe/Al2O3 tunnel contacts, Appl.
Phys. Lett. 94, 122106 (2009).
143. R. Meservey and P. M. Tedrow, Spin polarized electron tunneling, Phys. Rep.
238, 173–234 (1994).
144. J. S. Moodera, L. R. Kinder, T. M. Wong, and R. Meservey, Large magnetoresis-
tance at room temperature in ferromagnetic thin film tunnel junctions, Phys.
Rev. Lett. 74, 3273 (1995).
145. J. Daughton, Magnetic tunneling applied to memory, J. Appl. Phys. 81, 3758
(1997).
146. S. S. P. Parkin, C. Kaiser, A. Panchula et al., Giant tunneling magnetoresis-
tance at room temperature with MgO(100) tunnel barriers, Nat. Mater. 3, 862
(2004).
147. S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Giant room
temperature magnetoresistance in single crystal Fe/MgO/Fe magnetic tunnel
junctions, Nat. Mater. 3, 868 (2004).
148. J. S. Moodera, J. Nowak, L. R. Kinder et al., Quantum well states in spin-depen-
dent tunnel structures, Phys. Rev. Lett. 83, 3029 (1999).
149. D. J. Monsma and S. S. P. Parkin, Spin polarization of tunneling current from
ferromagnet/Al2O3 interfaces using copper-doped aluminum superconduct-
ing films, Appl. Phys. Lett. 77, 720 (2000).
150. V. F. Motsnyi, J. De Boeck, J. Das et al., Electrical spin injection in a ferromag-
net/tunnel barrier/semiconductor heterostructure, Appl. Phys. Lett. 81, 265
(2002).
151. V. F. Motsnyi, P. Van Dorpe, W. Van Roy et al., Optical investigation of electri-
cal spin injection into semiconductors, Phys. Rev. B 68, 245319 (2003).
152. T. Manago and H. Akinaga, Spin-polarized light-emitting diode using metal/
insulator/semiconductor structures, Appl. Phys. Lett. 81, 694 (2002).
153. O. M. J. van’t Erve, G. Kioseoglou, A. T. Hanbicki et al., Comparison of Fe/
Schottky and Fe/Al2O3 tunnel barrier contacts for electrical spin injection into
GaAs, Appl. Phys. Lett. 84, 4334 (2004).
154. B. T. Jonker, G. Kioseoglou, A. T. Hanbicki, C. H. Li, and P. E. Thompson,
Electrical spin injection into silicon from a ferromagnetic metal/tunnel bar-
rier contact, Nat. Phys. 3, 542 (2007).
155. G. Feher and E. A. Gere, Electron spin resonance experiments on donors
in silicon: II. Electron spin relaxation effects, Phys. Rev. 114, 1245 (1959).
156. V. Zarifis and T. G. Castner, ESR linewidth behavior for barely metallic n-type
silicon, Phys. Rev. B 36, 6198 (1987).
157. A. M. Tyryshkin, S. A. Lyon, W. Jantsch, and F. A. Schaffler, Spin manipulation
of free two-dimensional electrons in Si/SiGe quantum wells, Phys. Rev. Lett.
94, 126802 (2005).
158. S. Sugahara and M. Tanaka, A spin metal-oxide-semiconductor field effect
transistor using half-metallic contacts for the source and drain, Appl. Phys.
Lett. 84, 2307 (2004).
159. C. L. Dennis, C. V. Tiusan, J. F. Gregg, G. J. Ensell, and S. M. Thompson, Silicon
spin diffusion transistor: Materials, physics and device characteristics, IEEE
Proc. Circuits Dev. Syst. 152, 340 (2005).
160. E. Yablonovitch, H. W. Jiang, H. Kosaka, H. D. Robinson, D. S. Rao, and
T. Szkopek, Quantum repeaters based on Si/SiGe heterostructures, Proc. IEEE
91, 761 (2003).
161. M. L. W. Thewalt, S. P. Watkins, U. O. Ziemelis, E. C. Lightowlers, and M. O.
Henry, Photoluminescence lifetime, absorption and excitation spectroscopy
measurements on isoelectronic bound excitons in beryllium-doped silicon,
Solid State Commun. 44, 573 (1982).
162. P. Yu and M. Cardona, Fundamentals of Semiconductors, Springer-Verlag,
Berlin, Germany, 1996.
163. I. Žutić, J. Fabian, and S. C. Erwin, Spin injection and detection in silicon, Phys.
Rev. Lett. 97, 026602 (2006).
164. J. C. Costa, F. Williamson, T. J. Miller et al., Barrier height variation in Al/
GaAs Schottky diodes with a thin silicon interfacial layer, Appl. Phys. Lett. 58,
382 (1991).
165. P. Li and H. Dery, Theory of spin-dependent phonon-assisted optical transi-
tions in silicon, Phys. Rev. Lett. 105, 037204 (2010).
166. G. Kioseoglou, A. T. Hanbicki, R. Goswami et al., Electrical spin injection into
Si: A comparison between Fe/Si Schottky and Fe/Al2O3 tunnel contacts, Appl.
Phys. Lett. 94, 122106 (2009).
167. L. Grenet, M. Jamet, P. Noé et al., Spin injection in silicon at zero magnetic
field, Appl. Phys. Lett. 94, 032502 (2009).
168. D. J. Smith, M. R. McCartney, C. L. Platt, and A. E. Berkowitz, Structural char-
acterization of thin film ferromagnetic tunnel junctions, J. Appl. Phys. 83, 5154
(1998).
169. B. G. Park, T. Banerjee, B. C. Min, J. C. Lodder, and R. Jansen, Tunnel spin
polarization of Ni80Fe20/SiO2 probed with a magnetic tunnel transistor, Phys.
Rev. B 73, 172402 (2006).
170. C. Li, G. Kioseoglou, O. M. J. van’t Erve, P. E. Thompson, and B. T. Jonker,
Electrical spin injection into Si(001) through an SiO2 tunnel barrier, Appl.
Phys. Lett. 95, 172102 (2009).
171. X. Jiang, R. Wang, R. M. Shelby et al., Highly spin polarized room temperature
tunnel injector for semiconductor spintronics using MgO(100), Phys. Rev. Lett.
94, 056601 (2005).
172. G. Salis, R. Wang, X. Jiang et al., Temperature independence of the spin-
injection efficiency of a MgO-based tunnel spin injector, Appl. Phys. Lett. 87,
262503 (2005).
173. Y. Lu, V. G. Truong, P. Renucci et al., MgO thickness dependence of spin
injection efficiency in spin-light emitting diodes, Appl. Phys. Lett. 93, 152102
(2008).
174. V. G. Truong, P.-H. Binh, P. Renucci et al., High speed pulsed electrical spin
injection in spin-light emitting diode, Appl. Phys. Lett. 94, 141109 (2009).
175. M. Passlack, M. Hong, J. P. Mannaerts, J. R. Kwo, and L. W. Tu, Recombination
velocity at oxide–GaAs interfaces fabricated by in situ molecular beam epi-
taxy, Appl. Phys. Lett. 68, 3605 (1996).
References 145
176. H. Saito, J. C. Le Breton, V. Zayets, Y. Mineno, S. Yuasa, and K. Ando, Efficient

spin injection into semiconductor from an Fe/GaOx tunnel injector, Appl.
Phys. Lett. 96, 012501 (2010).
177. D. J. Monsma, J. C. Lodder, Th. J. A. Popma, and B. Dieny, Perpendicular hot
electron spin-valve effect in a new magnetic field sensor: The spin-valve tran-
sistor, Phys. Rev. Lett. 74, 5260 (1995).
178. K. Mizushima, T. Kinno, T. Yamauchi, and K. Tanaka, Energy-dependent hot
electron transport across a spin-valve, IEEE Trans. Magn. 33, 3500 (1997).
179. R. Jansen, The spin-valve transistor: A review and outlook, J. Phys. D 36, R289
(2003).
180. X. Jiang, R. Wang, S. van Dijken et al., Optical detection of hot-electron spin
injection into GaAs from a magnetic tunnel transistor source, Phys. Rev. Lett.
90, 256603 (2003).
182. H. Dery, Ł. Cywiński, and L. J. Sham, Lateral diffusive spin transport in layered
structures, Phys. Rev. B 73, 041306(R) (2006).
183. B. C. Min, K. Motohashi, J. C. Lodder, and R. Jansen, Tunable spin-tunnel
contacts to silicon using low work function ferromagnets, Nat. Mater. 5, 817
(2006).
184. M. Johnson and R. H. Silsbee, Interfacial charge-spin coupling: Injection and
detection of spin magnetization in metals, Phys. Rev. Lett. 55, 1790 (1985).
185. F. J. Jedema, A. T. Filip, and B. J. van Wees, Electrical spin injection and accu-
mulation at room temperature in an all-metal mesoscopic spin valve, Nature
410, 345 (2001).
186. F. J. Jedema, H. B. Heersche, A. T. Filip, J. J. A. Baselmans, and B. J. vanWees,
Electrical detection of spin precession in a metallic mesoscopic spin valve,
Nature 416, 713 (2002).
187. O. M. J. van’t Erve, A. T. Hanbicki, M. Holub et al., Electrical injection and
detection of spin-polarized carriers in silicon in a lateral transport geometry,
Appl. Phys. Lett. 91, 212109 (2007).
188. O. M. J. van’t Erve, C. Awo-Affouda, A. T. Hanbicki, C. H. Li, P. E. Thompson,
and B. T. Jonker, Information processing with pure spin currents in silicon:
Spin injection, extraction, modulation and detection, IEEE Trans. Electron.
Dev. 56, 2343 (2009).
temperature, Nature 448, 571 (2007).
190. M. Tran, H. Jaffrès, C. Deranlot et al., Enhancement of the spin accumulation at
the interface between a spin-polarized tunnel junction and a semiconductor,
Phys. Rev. Lett. 102, 036601 (2009).
191. A. V. Pohm, B. A. Everitt, R. S. Beech, and J. M. Daughton, Bias field and end
effects on the switching thresholds of “pseudo spin valve” memory cells, IEEE
Trans. Magn. 33, 3280 (1997).
192. A. M. Bratkovsky and V. V. Osipov, Efficient spin extraction from nonmagnetic
semiconductors near forward-biased ferromagneticsemiconductor modified
junctions at low spin polarization of current, J. Appl. Phys. 96, 4525 (2004).
193. M. I. D’yakonov and V. I. Perel, Chapter 2: Theory of optical spin orientation of
electrons and nuclei, in Optical Orientation, Modern Problems in Condensed
Matter Science, vol. 8, F. Meier and B. P. Zakharchenya (Eds.), North-Holland,
Amsterdam, the Netherlands, p. 39, 1984.
194. Y. Ando, K. Hamaya, K. Kasahara et al., Electrical injection and detection of
spin-polarized electrons in silicon through an Fe3Si/Si Schottky tunnel barrier,
Appl. Phys. Lett. 94, 182105 (2009).
195. T. Sasaki, T. Oikawa, T. Suzuki, M. Shiraishi, Y. Suzuki, and K. Tagami,

Electrical spin injection into silicon using MgO tunnel barrier, Appl. Phys.
Express 2, 053003 (2009).
196. S. P. Dash, S. Sharma, R. S. Patel, M. P. de Jong, and R. Jansen, Electrical cre-
ation of spin polarization in silicon at room temperature, Nature 462, 491
(2009).
197. L. Manchanda, M. D. Morris, M. L. Green et al., Multi-component high-K
gate dielectrics for the silicon industry, Microelectron. Eng. 59, 351 (2001), and
references therein.
198. C. Li, O. M. J. van’t Erve, and B. T. Jonker, Electrical injection and detection
of spin accumulation in silicon at 500K with magnetic metal/silicon dioxide
contacts, Nat. Commun. 2, 245 (2011), doi:10.1038/ncomms1256.
199. Y. Ochiai and E. Matsuura, ESR in heavily doped n-type silicon near a metal-
nonmetal transition, Phys. Stat. Sol. (a) 38, 243 (1976).
200. M. Passlack, R. Droopad, Z. Yu et al., Screening of oxide/GaAs interfaces for
MOSFET applications, IEEE Trans. Electron. Dev. Lett. 29, 1181 (2008), and
references therein.
201. K. S. Krishnan and N. Ganguli, Large anisotropy of the electrical conductivity
of graphite, Nature 144, 667 (1939).
202. S. Chen L. Brown, M. Levendorf et al., Oxidation resistance of graphene-
coated Cu and Cu/Ni alloy, ACS Nano 5, 1321 (2011).
203. E. Cobas, A. L. Friedman, O. M. J. van’t Erve, J. T. Robinson, and B. T. Jonker,
Graphene as a tunnel barrier: Graphene-based magnetic tunnel junctions,
Nano Lett. 12, 3000 (2012).
204. E. D. Cobas O. M. van’t Erve, S. F. Cheng et al., Room temperature spin filter-
ing in metallic ferromagnet/multilayer graphene/ferromagnet junctions, ACS
Nano 10, 10357 (2016).
205. O. M. J. van’t Erve, A. L. Friedman, E. Cobas, C. H. Li, J. T. Robinson, and B. T.
Jonker, Low-resistance spin injection into silicon using graphene tunnel barri-
ers, Nat. Nanotechnol. 7, 737 (2012).
206. O. M. J. van’t Erve A. L. Friedman, E. Cobas et al., A graphene solution to
conductivity mismatch: Spin injection from ferromagnetic metal / graphene
tunnel contacts into silicon, J. Appl. Phys. 113, 17C502 (2013).
207. O. M. J. van’t Erve A. L. Friedman, C. H. Li et al. Spin transport and Hanle
effect in silicon nanowire using graphene tunnel barriers, Nat. Commun. 6,
7541 (2015).
208. X. Li C. W. Magnuson, A. Venugopal et al., Large-area graphene single crystals
grown by low-pressure chemical vapor deposition of methane on copper, J.
Am. Chem. Soc. 133, 2816 (2011).
209. S. P. Dash S. Sharma, J. C. Le Breton et al., Spin precession and inverted Hanle
effect in a semiconductor near a finite-roughness ferromagnetic interface,
Phys. Rev. B 84, 054410 (2011).
210. Y. Aoki, M. Kameno, Y. Ando et al., Investigation of the inverted Hanle effect
in highly doped Si, Phys. Rev. B 86, 081201(R) (2012).
211. J. H. Pifer, Microwave conductivity and conduction-electron spin-resonance
linewidth of heavily doped Si:P and Si:As, Phys. Rev. B 12, 4391 (1975).
212. V. Zarifis and T. G. Castner, ESR linewidth behavior for barely metallic n-type
213. H. Dery, L. Cywinski, and L. J. Sham, Lateral diffusive spin transport in layered
structures, Phys. Rev. B 73, 041306(R) (2006).
214. B.-C. Min, K. Motohashi, C. Lodder, and R. Jansen, Tunable spin-tunnel con-
tacts to silicon using low-work-function ferromagnets, Nat. Mater. 5, 817
(2006).
215. L. Cywinski, H. Dery, and L. J. Sham, Electric readout of magnetization dynam-
ics in a ferromagnet–semiconductor system, Appl. Phys. Lett. 89, 042105
(2006).
References 147
216. O. Txoperena, Y. Song, L. Qing et al., Impurity-assisted tunneling magnetore-

sistance under a weak magnetic field, Phys. Rev. Lett. 113 (2014).
217. O. Txoperena, M. Gobbi, A. Bedoya-Pinto et al., How reliable are Hanle mea-
surements in metals in a three-terminal geometry? Appl. Phys. Lett. 102,
192406 (2013).
218. V. M. Karpan G. Giovannetti, P. A. Khomyakov et al., Graphite and graphene
as perfect spin filters, Phys. Rev. Lett. 99, 176602 (2007); ibid., Theoretical pre-
diction of perfect spin filtering at interfaces between close-packed surfaces of
Ni or Co and graphite or grapheme, Phys. Rev. B 78, 195419 (2008).
219. O. V. Yazyev and A. Pasquarello, Magnetoresistive junctions based on
epitaxial graphene and hexagonal boron nitride, Phys. Rev. B 80, 035408
(2009).
220. Y.-C. Lin, Y. Chen, A. Shailos and Y. Huang, Detection of spin polarized carrier
in silicon nanowire with single crystal MnSi as magnetic contacts, Nano Lett.
10, 2281 (2010).
221. S. Zhang, S. A. Dayeh, Y. Li, S. A. Crooker, D. L. Smith, and S. T. Picraux,
Electrical spin injection and detection in Silicon nanowires through oxide
tunnel barriers, Nano Lett. 13, 430 (2013).
222. J. Tarun S. Huang, Y. Fukuma et al., Demonstration of spin valve effects in
silicon nanowires, J. App. Phys. 109, 07C508 (2011).
223. J. Tarun S. Huang, Y. Fukuma et al., Temperature evolution of spin-polarized
electron tunneling in silicon nanowire–permalloy lateral spin valve system,
Appl. Phys. Express 5, 045001 (2012).
224. E.-S. Liu, J. Nah, K. M. Varahramyan, and E. Tutuc, Lateral spin injection in
germanium nanowires, Nano Lett. 10, 3297 (2010).
225. R. Nakane, S. Sato, S. Kokutani, and M. Tanaka, Appearance of anisotropic
magnetoresistance and electric potential distribution in Si-base multi-termi-
nal devices with Fe electrodes, IEEE Magn. Lett. 3, 3000404 (2012).
226. Y. Cui, L. J. Lauhon, M. S. Gudiksen, J. F. Wang, and C. M. Lieber, Diameter-
controlled synthesis of single-crystal silicon nanowires, Appl. Phys. Lett. 78,
2214 (2001).
227. C. Awo-Affouda O. M. van’t Erve, G. Kioseoglou et al., Contributions to Hanle
lineshapes in Fe/GaAs non-local spin valve transport, Appl. Phys. Lett. 94,
102511 (2009).
228. Y. Ando S. Yamada, K. Kasahara et al., Effect of the magnetic domain structure
in the ferromagnetic contact on spin accumulation in silicon, Appl. Phys. Lett.
101, 232404 (2012).
229. H. Song and H. Dery, Analysis of phonon-induced spin relaxation processes in
230. A. Tilke, F. Simmel, H. Lorenz, R. Blick, and J. Kotthaus, Quantum interference
in a one-dimensional silicon nanowire, Phys. Rev. B 68, 075311 (2003).
231. A. Lherbier, M. P. Persson, Y.-M. Niquet, F. Triozon, and S. Roche, Quantum
transport length scales in silicon-based semiconducting nanowires: Surface
roughness effects, Phys. Rev. B 77, 085301 (2008).
3
Spin Transport
in Si and Ge
Hot Electron Injection and
Detection Experiments
Ian Appelbaum
3.1 Introduction 150

3.2 Hot Electron Generation and Collection 150
3.3 Spin-Polarized Hot Electron Transport 151
3.4 Spins in Silicon 153
3.5 Ballistic Hot Electron Injection and Detection Devices 156
3.5.1 Experiments on Silicon 158
3.6 Spins in Germanium 161
3.7 Outlook 164
Acknowledgments 166
References 166
149
150 Chapter 3. Spin Transport in Si and Ge
3.1 INTRODUCTION
The spin injection and detection limitations posed by conductivity and life-
time mismatch between semiconductors and ferromagnetic (FM) metals
[1–4] has been variously solved by lifting the constraints of ohmic transport
with the insertion of a tunnel barrier [5, 6] or the use of ferromagnetic semi-
conductors [7, 8]. However, for many semiconductors, there are additional
materials-dependent obstacles to the observation of spin transport. The
materials properties of elemental group-IV semiconductors silicon (Si) and
germanium (Ge), pose problems such as alloy formation at metallic contacts
and difficulties with the use of optical methods for spin detection due to indi-
rect bandgap [9, 10].
In this chapter, we discuss how all these issues can be circumvented by
using a mechanism for spin injection and detection that differs fundamentally
from ohmic transport in that (i) the inelastic mean-free-path (mfp) is not the
smallest length scale in the device, and (ii) electrochemical potentials can-
not be uniquely defined in thermal equilibrium. With these techniques, the
physical length scale of the metallic electron injection contacts is shorter than
the mfp, and conduction occurs through states far above the Fermi level (as
compared to the thermal energy k BT ), far out of thermal equilibrium. Because
this transport mode utilizes electrons with high kinetic energy that do not
suffer appreciable inelastic scattering, it is known as “Ballistic Hot Electron
Transport.” Both spin injection and detection are performed all-electronically,
and interfacial structure [11] plays only a minor role, enabling observation and
study of spin transport in materials such as Si and Ge which had been previ-
ously excluded from the field. Related background material on spin injection
and transport can be found in Chapter 5, Volume 1, Chapters 2 and 6, Volume
2, and on spin relaxation in Chapter 1, Volume 2, in this book.
3.2 HOT ELECTRON GENERATION

AND COLLECTION
There are several techniques for generating hot electrons for injection from
a metal into a semiconductor. Since metals have a very high density of elec-
trons at and below the Fermi energy E F , all must rely on an electrically recti-
fying barrier to eliminate transport of these thermalized electrons across the
metal-semiconductor interface, which would dilute the injected hot electron
current. This barrier is ideally created by the difference in work functions of
the metal and the electron affinity of the semiconductor [12–15], but in reality
its energetic height qφ, where q is the elementary charge, is determined more
by the details of surface states which lie deep in the bandgap of the semicon-
ductor that pin the Fermi level [16]. There is, of course, always a leakage current
due to thermionic emission over this “Schottky” barrier at non-zero tempera-
ture T given by the Richardson–Dushman expression ∝ T 2 exp ( −qφ/ k BT ) ;
because typical barrier heights are in the range 0.6–0.8 eV [17] for Si and Ge,
hot electron collection with Schottky barriers is often performed at tempera-
tures below ambient conditions to reduce current leakage to negligible levels.
3.3 Spin-Polarized Hot Electron Transport 151
Some of the earliest hot electron work in metal-semiconductor devices

involved the use of subbandgap photon absorption in the metal, which scat-
tered electrons from a state with energy E < E F to another state with energy
E + ω > E F , where ω is the photon energy. If the Schottky barrier height is
qφ, then electrons with initial energy E > E F + qφ − ω can cross the inter-
face, assuming they are injected or generated within a mfp (typically in the
range 1–100 nm) of it. Because this is similar to the photoelectric effect
(except that the electrons cross an internal interface) the method is known as
internal photoemission (IPE). By varying the photon energy, Schottky barrier
heights [18] (and even buried heterojunction band offsets [19]) can be deter-
mined in a model-independent way by measuring the turn-on threshold of
the collected hot electron current without biasing the device, and the hot
electron current can be independently tuned by changing the illumination
intensity.
In a later experiment, hot electrons were generated electrically using
a forward-biased Schottky diode, and again collected by another (lower
height) Schottky interface in a semiconductor-metal-semiconductor (SMS)
device. Because this provides a narrow distribution of energies with width
≈ k BT , the hot electron mean free path of metal thin films could reliably be
measured [20]. However, in contrast to Internal Photoemission, the energy
itself cannot be changed since this is fixed by the emitter Schottky barrier
height [21, 22].
The use of tunnel junctions (TJs, which consist of two metallic conduc-
tors separated by a thin insulator) [23–28] remedies this problem. The elec-
trostatic potential energy qV provided by voltage bias V tunes the energy
of hot electrons emitted from the cathode, and the exponential energy
dependence of quantum-mechanical tunneling assures a narrow distribu-
tion. Hot electrons thus created can be collected by the Schottky barrier
if qV > qφ. Because of its robust insulating native oxide, aluminum (Al) is
often employed during tunnel junction fabrication; although in principle
any insulator can be used, it has been empirically found that the best are
Al xO, MgO, and AlN [29, 30]. The first proposal for a tunnel-junction hot
electron injector was by Mead, who suggested another tunnel junction
or the vacuum level as a collector [31–33]. Very soon thereafter, Spratt,
Schwartz, and Kane showed that a semiconductor collector could be used
to realize this device with a Au/Al xO/Al tunnel junction [34]. Heiblum’s
THETA device is an all-semiconductor version of this design [35, 36].
All three techniques (IPE, SMS, and TJ), along with illustrations of their
corresponding hot electron energetic distributions, are shown in Figure 3.1.
3.3 SPIN-POLARIZED HOT ELECTRON

TRANSPORT
The inelastic mfp of hot electrons in ferromagnetic metal films is spin-
dependent: majority (“spin up”) electrons have a longer mfp than minority
(“spin down”) electrons. Therefore, an initially unpolarized hot electron cur-
rent will become spin-polarized by spin-selective scattering during ballistic
a) Internal Photoemission b) SMS injection
insulator
photon ћω <EC−E V qφ
ENERGY
EC
semiconductor
semiconductor
collector
emitter semiconductor qV
metal collector
base metal
base semiconductor
collector
EV metal
base
anode
FIGURE 3.1 Illustration of band diagrams for three different hot electron generation techniques: internal
photoemission (IPE), semiconductor-metal-semiconductor (SMS) metal-base transistor, and tunnel junction
(TJ) emission. The energetic distributions of the hot electrons generated (shown in red) are roughly homo-
geneous up to ħω above the collector Fermi energy for IPE, peaked at the fixed Schottky barrier height of the
emitter (qϕ) for SMS, and peaked near the emitter cathode Fermi energy qV (where V is the applied voltage
and q is the elementary charge) for TJ.
transport through a ferromagnetic metal film, where the polarization is

given by
l l
− −
λmaj λmaj
e −e
P= l l , (3.1)
− −
λmaj λmaj
e +e
where:
l is the FM film thickness
λ maj is the majority spin mfp
λ min is the minority spin mfp
This “ballistic spin-filtering” effect can be used not only for spin polar-
ization at injection, but also for spin analysis of a hot electron current at
detection, much as an optical polarizing filter can be used both for electric
field polarization and analysis of photons by changing the relative orienta-
tion of the optical axis.
A device making direct use of this effect is called the spin-valve tran-
sistor (SVT) [37–40]. It is a SMS metal-base transistor device [20] with a
multilayer base consisting of at least two ferromagnetic layers (spin polarizer
and analyzer) whose magnetizations can be set in a parallel or antiparallel
configuration; switching between these two relative orientations induces a
large change in the collector current of several hundred percent because of
the ballistic spin-filtering effect.
Other non-SMS SVT-related designs have been demonstrated as well.

To generate hot electrons, Monsma et al. also suggested the use of a mag-
netic tunnel junction together with a Schottky hot electron collector in their
initial work introducing the SVT [37]; this was then later demonstrated using
solid-state devices [41, 42], and with metal-vacuum-metal tunnel junctions
using STM/BEEM [43]. In 2002, Van Dijken et al. suggested that moving
the ferromagnetic polarizer to the emitter cathode of the tunnel junction
comprises a new device, which they named the Magnetic Tunnel Transistor
(MTT) [44]. (This configuration was, however, first proposed by Monsma
in 1998, without describing it as a new device distinct from the SVT) [45].
Optical methods such as IPE with subbandgap illumination [46, 47] and pho-
togeneration via interband transition with super-bandgap illumination [48]
have also been used to demonstrate hot electron spin-valve effects.
Although we can speculate that the hot electrons presumably main-
tain some residual spin polarization after collection by the Schottky barrier
[49], no spin detection after transport through the semiconductor was ever
demonstrated in the above works, so spin injection success is unknown. The
first successful use of spin-polarized ballistic hot electron injection was not
reported until 2003, by Jiang et al. [50], where a MTT was used for injec-
tion into GaAs, and circular polarization analysis of band edge light emitted
from InGaAs quantum wells was used for detection (a “spinLED”). However,
because of the optical selection rules demanding a polar “Faraday” geometry,
a large perpendicular magnetic field was required to induce a perpendicular
magnetization of the magnetic layers in the MTT injector, so spin precession
measurements (necessary to unambiguously confirm spin transport [51, 52])
could not be done.
After injection across the interface, the hot electrons must relax to the
conduction band edge before transport and recombination. Because GaAs
is a non-centrosymmetric zinc blende lattice that does not preserve inver-
sion symmetry, Dresselhaus spin-orbit terms in the Hamiltonian break the

spin degeneracy of the conduction band away from k = 0 and the Dyakonov–
Perel effect [53] can induce strong spin depolarization [54]. For this reason,
injection near the conduction band edge in states close to the Brillouin zone
center is strongly preferred, or application of hot electron spin injection to a
material without this scattering mechanism is desired.
The group-IV elemental semiconductors silicon and germanium form
in the diamond lattice, which does preserve inversion symmetry, and hence
eliminates the deleterious Dyakonov–Perel mechanism. All subsequent work
presented in this chapter will therefore focus on our experience developing
hot electron techniques for spin injection, transport, and detection in these
two materials, and will highlight the resulting knowledge gained pertaining
to their spin-dependent electronic structure.
3.4 SPINS IN SILICON
Spin-polarized electrons in Si were first studied decades ago using resonance
between Zeeman-split levels in a large magnetic field. Of interest in this
field of electron spin resonance (ESR) was not only the conduction electrons
but also electrons bound to donors [55–60]. The microwave absorption line
positions (or electrically detected magnetic resonance (EDMR) [61, 62] con-
ductance features) give the gyromagnetic ratio γ = g µ B / , where g is the
g-factor and µ B is the Bohr magneton, and their linewidths give a measure of
the inverse of the spin lifetime [59, 63].
The gyromagnetic ratio in Si corresponds to a g-factor very close to that
of a bare electron in vacuum ( g ≈ 2) [64], and observed lifetimes were rela-
tively long (several ns even at ambient), with very narrow linewidths. This
apparent superiority of Si over other semiconductors is due to several for-
tuitous intrinsic properties which make it a relatively simple spin system
[65–67].
It is often pointed out that since SOC scales as Z 4 in atomic systems,
silicon’s low atomic number (ZSi = 14) leads to a reduced spin-orbit coupling.
This is only partly true. The geometric symmetry of the diamond lattice and
the proximity of the CBM to the X-point is arguably more important. For
example, the small SOC correction to g-factor can be explained by the fact
that double orbital degeneracy imposed by non-symmorphic elements of the
symmetry group (those containing partial translations) at the X-point is pre-
served even after inclusion of SOI perturbation. Diamagnetic corrections to
g-factor, which can be calculated via evaluation of operator matrix elements
with remote bands [68]
ψ n πˆ x ψ n ′ ψ n ′ πˆ y ψ n ′′
∑
4i
n g orbit n′′ = , (3.2)
m n ′≠ n
En − En ′
are therefore zero for every band at the X-point, because the energy denomi-
nators are the same, whereas the numerators change sign between orbitally
degenerate states. This is why a large valence band split-off is required for
conduction band g-factor renormalization in zinc blende.
At the Si conduction band valley minimum only 15% along the Δ-axis
away from X toward Γ, the double orbital degeneracy of ∆ 5 valence band
states (which transform like axial vector components {zx, zy } [69]) ≈ 4eV
below the conduction band is indeed broken by SOI. However, because this
state is so close to X where it vanishes, the splitting is only ≈ 2 meV . Thus,
even though the numerators are non-zero for momentum πˆ x, y transverse to
the valleys (conduction band is ∆1 that transforms like z), the total contribu-
tion from Equation 3.2 is tiny and only several parts in 103 .
The long spin lifetime in Si (especially for electrons bound in neutral
donors) was exploited to achieve spontaneous emission of tunable micro-
waves in the late 1950s [70], and suggested much later as a fundamental
ingredient in solid-state quantum computation [71]. In fact, Si was the first
material with which optical orientation (generation of spin-polarized elec-
trons in the conduction band through interband transitions induced by
circularly polarized super-bandgap photon illumination [72]) was demon-
strated [73, 74], yet the importance of achieving true spin transport in Si was
often overlooked in reviews of the field [75]. This milestone was accomplished
convincingly for the first time only in 2007, using the ballistic hot electron
techniques described in Section 3.5 [76].
Due to its apparent advantages over other semiconductors, many groups
tried to demonstrate phenomena attributed to spin transport in Si [77–85],
but this was typically done with ohmic FM-Si contacts and two-terminal
magnetoresistance measurements or in transistor-type devices [86, 87]
which are bound to fail due to the “fundamental obstacle” for ohmic spin
injection mentioned in Section 3.1 [1, 3, 4, 6]. Although weak spin-valve
effects can be found in the literature, no evidence of spin precession is avail-
able, so the signals measured are ambiguous at best [51, 52]. Indeed, although
magnetic exchange coupling across ultrathin tunneling layers of Si was seen,
no spin-valve magnetoresistance was observed [88].
These failures were addressed by pointing out that only in a narrow win-
dow of FM-Si Resistance-Area (RA) product was a large spin polarization
and hence large magnetoresistance expected in all-electrical two-terminal
devices [89]. Subsequently, several efforts to tune the FM-Si interface resis-
tance were made [90–94]. However, despite the ability to tune the RA prod-
uct by over 8 orders of magnitude and even into the anticipated high-MR
window (for instance, by using a low-work-function Gd layer), no evidence
that this approach has been fruitful for Si can be found, and the theory has
been confirmed only for the case of low-temperature-grown 5 nm-thick
GaAs [95, 96].
Subsequent to the first demonstration of spin transport in Si [76], optical
detection (circular electroluminescence analysis [97]) was shown to indicate
spin injection from a FM and transport through only several tens of nm of
Si, using first an Al xO tunnel barrier [98] and then tunneling through the
Schottky barrier [99], despite the indirect bandgap. These methods required
a large perpendicular magnetic field to overcome the large in-plane shape
anisotropy of the FM contact, but later a perpendicular anisotropic mag-
netic multilayer was shown to allow spin injection into Si at zero external
magnetic field [100]. While control samples with nonmagnetic injectors do
show negligible spin polarization, again no evidence of spin precession was
presented.
More recently, four-terminal non-local measurements on Si devices
have been made, for instance with Al xO [101] or MgO [102, 103] tunnel bar-
riers, or Schottky contacts using ferromagnetic silicide injector and detector
[104]. Since these initial demonstrations, substantial progress has been made
by Shiraishi’s Osaka/Kyoto group [105–107] using this method.
Demonstrations of electrically injected spin accumulation in nonmag-
netic materials are considered reliable when measured in a non-local four-
terminal geometry. When performed correctly, the experiment electrically
measures the decay of spin polarization in the nonmagnetic bulk material
between the injection and detection regions. Therefore, it is advantageous to
have a submicron separation between the ferromagnetic injector and detec-
tor electrodes if the spin diffusion length of the nonmagnetic bulk material
is not much longer than one micron. To mitigate this requirement, some
researchers have recently resorted to a local three-terminal measurement
wherein one ferromagnetic electrode is used for both injection and detec-
tion of the spin signal in elemental semiconductors and complex oxides
[108–114]. The results of these experiments, however, are often inconsistent
with our understanding of spin transport phenomena in these materials
[115–119]. And indeed, recent mounting evidence show that Pauli block-
ing during transport through localized defect/impurity states successfully
accounts for the observed phenomena, without any need to invoke actual
spin accumulation in the semiconductor [120–124]. It should be noted, how-
ever, that with clean MBE-grown interfaces, signals consistent with actual
spin injection (much smaller in magnitude and with linewidths reflecting the
expected bulk spin lifetime in heavily-doped Si) have been observed [125], as
in the original work applying a local three-terminal method to GaAs [126].
As mentioned in Section 3.1, another way to overcome the conductiv-
ity/lifetime mismatch for spin injection is to use a carrier-mediated ferro-
magnetic semiconductor heterointerface. The interfacial quality may have
a strong effect on the injection efficiency, so epitaxial growth will be neces-
sary. Materials such as dilute magnetic semiconductors Mn-doped Si [127],
Mn-doped chalcopyrites [10, 128–131], or the “pure” ferromagnetic semi-
conductor EuO [132] have all been suggested, but none as yet have been dem-
onstrated as spin injectors for Si. It should be noted that while their intrinsic
compounds are indeed semiconductors, due to the carrier-mediated nature
of the ferromagnetism, it is seen only in highly (i.e. degenerately) doped and
essentially metallic samples in all instances.
Another alternative injection method, spin pumping via ferromagnetic
resonance with microwave absorption, was reported not only for electrons
in n-Si [133], but also for holes in p-Si [134, 135]. We note that from a theoret-
ical perspective, holes in the valence band of any bulk cubic semiconductor
have spin lifetimes on the order of the momentum lifetime (ps or less), since
spin-orbit interaction spin-mixes the light hole states with up/down prob-
abilities of 1/3 and 2/3 so that Elliott–Yafet scattering among these states and
the degenerate heavy holes causes highly efficient spin flips.
Several experimental [136, 137] and theoretical [138] works also
addressed spin-polarized electrons and holes [139, 140] in Si/SiGe 2DEGs,
but this is outside the scope of the present introduction; we deal here only
with spin transport in 3D bulk Si, but do note that SiGe alloys could very
well become useful as a way of tailoring spin transport devices to make use
of local spin-orbit effects.
3.5 BALLISTIC HOT ELECTRON INJECTION

AND DETECTION DEVICES
Two types of tunnel junctions have been employed for injection of spin-
polarized electrons into non-degenerate semiconductors. The first used bal-
listic spin filtering of initially unpolarized electrons from a nonmagnetic Al
cathode by a ferromagnetic anode base layer in direct contact to undoped,
10-micron-thick single-crystal Si(100) [76]. Despite SVT measurements sug-
gesting the possibility of 90% spin polarization in the metal [40, 141], only 1%
polarization was found after injection into and transport through the Si. It
was discovered later that a nonmagnetic Cu interlayer spacer could be used
to increase the polarization to approximately 37% [142], likely due to Si’s ten-
dency to readily form spin-scattering “magnetically-dead” alloys (silicides) at
interfaces with ferromagnetic metals.
Despite the possibilities for high spin polarization with these ballistic
spin filtering injector designs, the short hot electron mfps in FM thin film
anodes causes a very small injected charge current on the order of 100nA,
with an emitter electrostatic potential energy approximately 1 eV above the
Schottky barrier and a contact area approximately 100 × 100 µm 2. Because
the injected spin density and spin current are dependent on the product
of spin polarization and charge current, this technique is not ideal for
transport measurements. Therefore, an alternative injector utilizing a FM
tunnel-junction cathode and nonmagnetic anode (which has a larger mfp)
was used for approximately ten times greater charge injection and hence
larger spin signals, despite somewhat smaller potential spin polarization of
approximately 15% [143, 144]. These injectors can be thought of as one-half
of a magnetic tunnel junction [145], with a spin polarization proportional to
the Fermi-level density of states spin asymmetry, rather than exponentially
dependent on the spin-asymmetric mfp as is the case with ballistic spin fil-
tering described above.
Although the injection is due to ballistic transport in the metallic
contact, the conduction band mfp is typically only on the order of 10 nm
[146], so the vast majority of the subsequent transport to the detector over a
length scale of tens [76, 143, 142], hundreds [144, 147], or thousands [148]
of microns occurs at the conduction band edge following momentum relax-
ation [149]. Typically, relatively large accelerating voltages are used so that
the dominant transport mode is carrier drift; the presence of rectifying
Schottky barriers on either side of the transport region assures that the
resulting electric field does nothing other than determine the drift veloc-
ity of spin polarized electrons, hence the transit time [150, 151]—there are
no spurious (unpolarized) currents induced to flow. Furthermore, undoped
transit layers are primarily used; otherwise band bending would create a
confining potential and increase the transit time, potentially leading to
excessive depolarization [152].
The ballistic hot electron spin detector is comprised of a semiconductor-
FM-semiconductor structure (both Schottky interfaces), fabricated using
UHV metal-film wafer bonding (a spontaneous cohesion of ultra-clean
metal film surfaces which occurs at room-temperature and nominal force in
ultra-high vacuum) [38, 153]. After injection and transport through the semi-
conductor, spin-polarized electrons are ejected from the conduction band
over the Schottky barrier and into hot electron states far above the Fermi
energy. Again, because the mfp in FMs is larger for majority-spin (i.e. parallel
to magnetization) hot electrons, the number of electrons coupling with con-
duction band states in a n-Si collector on the other side (which has a smaller
Schottky barrier height due to contact with Cu) is dependent on the final
spin polarization and the angle between spin and detector magnetization.
Quantitatively, we expect a contribution to our signal from each electron

having spin orientation θ with respect to the detector magnetization
l l
q - lmaj q -
µ cos2 e + sin 2 e lmin
2 2
(3.3)
1 éæ - ö æ - öù
l l l l
- -
= êç e lmaj
- e lmin ÷ cosq + ç e lmaj
+e lmin ÷ú .
2 êç ÷ ç ÷ú
ëè ø è øû
Because the exponential terms are constants, this has the simple form
∝ cosθ + const.; in the following, we disregard the constant term, as it is spin-
independent. The spin transport signal is thus the (reverse) current flowing
across the n-Si collector Schottky interface. In essence, this device (whose
band diagram is schematically illustrated in Figure 3.2) can be thought of
as a split-base tunnel-emitter SVT with several hundred to thousands of
microns of Si between the FM layers.
3.5.1 Experiments on Silicon
Two types of measurements are typically made: “spin valve” in a magnetic
field parallel to the plane of magnetization, and spin precession in a mag-
netic field perpendicular to the plane of magnetization. The former allows
the measurement of the difference in signals between parallel (P) and anti-
parallel (AP) injector/detector magnetization and hence is a straightforward
way of determining the conduction electron spin polarization,
I P − I AP
P= . (3.4)
I P + I AP
injector
detector
Energy
Al2O3
CoFe
Al Single-crystal
undoped Si
350 µm n-Si
Al
Cu
NiFe
VE VC1 IC1
Cu
IC2
10 nm
8 nm
Distance
FIGURE 3.2 Schematic band diagram of a four-terminal (two for TJ injection and
two for FM SMS detection) ballistic hot electron injection and detection device
with a 350 μm-thick Si transport layer.
Typical spin-valve measurement data, indicating ≈ 8% spin polarization after

transport through 350 μm undoped Si, is shown in Figure 3.3.
Measurements in perpendicular magnetic fields reveal the average spin
orientation after transit time t through the Si, due to precession at frequency
ω = g µ B B / , where B is magnetic field. If the transit time is determined
L
only by drift, i.e. t = , we expect our spin transport signal to behave
µE
∝ cosg µ B Bt / . However, due to transit time uncertainty ∆t caused by ran-
dom diffusion, there is likewise an uncertainty in spin precession angle
∆θ = ω∆t which increases as the magnetic field (and hence ω) increases.
When this uncertainty approaches 2π rad, the spin signal is fully suppressed
by a cancellation of contributions from antiparallel spins, a phenomenon
called spin “dephasing”, or the Hanle effect [154].
On a historical note, our device is essentially a solid-state analog of
experiments performed in the 1950s that were used to determine the g-fac-
tor of the free electron in vacuum using Mott scattering as spin polarizer
and analyzer and spin precession during time-of-flight in a solenoid [64]. In
our case, we already know the g-factor (from e.g. ESR lines), so our experi-
ments in strong drift electric fields where spin dephasing is weak can be used
to measure transit time with t = h / g µ B B2 π , where B2π is the magnetic field
period of the observed precession oscillations, despite the fact that we make
DC measurements, not time-of-flight [150, 151]. Typical spin-precession
data, indicating transit time of approximately 12ns to cross 350 μm undoped
Si in an electric field of ≈ 580 V/cm, is shown in Figure 3.4a.
One important application of this transit time information from spin
precession is to correlate it to the spin polarization determined from spin-
valve measurements and Eq. 3.4 to extract spin lifetime. By varying the
internal electric field, we change the drift velocity and hence average tran-
sit time. A reduction of polarization is seen with an increase in average
transit time (as in Figure 3.5a) that we can fit well to first order using an
exponential-decay model P ∝ e −t /τ , and extract the timescale τ [144]. In
this way, we have observed spin lifetimes of approximately 1 μs at 60 K
Spin Transport Signal (IC2) [pA]
120
115
110
sweep
105 direction
T=150K
100
-200 -100 0 100 200
In-Plane Magnetic Field [Oe ]
FIGURE 3.3 In-plane magnetic field measurements show the “spin-valve” effect
and can be used to calculate the spin polarization after transport in Si.
a) 120
experiment T=150K
model
Spin Signal [pA]

115
110
105
sweep direction
100
-200 -100 0 100 200
Perpendicular Magnetic Field [Oe]
b) 1.0
0.8
FFT-normalized
0.6
0.4
0.2 ∆t
0.0
5 10 15
Time [ns]
FIGURE 3.4 (a) A typical spin precession measurement shows the coherent
oscillations due to drift and the suppression of signal amplitude (“dephasing”) as
the precession frequency rises. Our model simulates this behavior well. (b) The real
part of the Fourier transform of the precession data in (a) reveals the spin current
arrival distribution.
in 350 μm-thick transport devices [155].* In Figure 3.5b, the temperature

dependence of spin lifetime is compared to the T −5/2 power law predicted
by Yafet [60]. This power law arose from considering that the intravalley
spin-flip matrix element is quadratic in phonon momentum q; Fermi’s
golden rule gives the lowest-order spin-flip transition rate as this matrix
element squared (q 4 ∝ E 2 ∝ T 2 ) times the familiar E ∝ T density of
states in 3-dimensions.
Yafet’s approximation was improved by the numerical calculations of
Cheng et al. which were found to be closer to T−3 [156]. As Li and Dery [157]
and Song and Dery [158] point out, however, the spin lifetime dependence
on temperature is fundamentally not well-captured by a simple power law
because it is primarily due to intervalley scattering with large-momentum
f-process [159] phonons at high temperature and intervalley scattering with
small-momentum acoustic phonons at low temperature.
* In Ref. [144], a more conservative estimate of the spin lifetime (e.g. 520 ns at 60 K) was
obtained by fitting to the transit time dependence of an alternative quantity expected to be
proportional to the spin polarization, rather than using Eq. 3.4 directly.
60K
85K
15 100K
Spin Polarization [%]

125K
150K
10
0 50 100 150 200 250

Transit time [ns]
Huang & Appelbaum 2008

1000 Yafet 1963
Spin Lifetime [ns]
Cheng, Wu & Fabian 2009
500
0
60 90 120 150
Temperature [K]
FIGURE 3.5 (a) Fitting the normalized spin signal from in-plane spin-valve
measurements to an exponential decay model using transit times derived from
spin precession measurements at variable internal electric field yields measure-
ment of spin lifetimes in undoped bulk Si. (b) The experimental spin lifetime values
obtained as a function of temperature are compared to Yafet’s T−5/2 power law for
indirect-bandgap semiconductors [60] and Cheng et al.’s T−3 derived from a full
band structure theory [156].
3.6 SPINS IN GERMANIUM
Germanium shares the same diamond lattice with silicon and is also an indi-
rect-gap multivalley semiconductor, but spin-orbit coupling in the conduc-
tion band of this material is very different. While it is true that the atomic
number is much greater (32 as opposed to 14 for Si), the primary reason
for stronger spin-orbit coupling is that the conduction band minimum lies
at the L-point, rather than close to the X-point where the symmetry group
retains non-symmorphic lattice symmetry elements that preserve orbital
degeneracy.
Once again, we can illustrate the effects of spin-orbit-coupling by
analyzing the dominant diamagnetic contributions to the Landé g-factor.
Germanium’s L1 conduction band is primarily affected by the SOI-induced
≈ 0.2 eV splitting of the L3′ valence band, ≈ 2.2 eV below it [160, 161]. The
energy denominators in Equation 3.2 are thus substantially different, so
that a quite incomplete cancellation occurs in the sum. Furthermore, the
wave function symmetries ( L3′ states are odd with respect to transverse
directions, whereas L1 is even, and both are even with respect to the val-
ley axis) lead to non-zero numerators only for magnetic moments pointed
along the valley axis. The transverse diamagnetic correction therefore van-
ishes to lowest order and g ⊥ ≈ 2, but the longitudinal contribution is sub-
stantial so that g ~ 1 [162].
In a system like this with conduction band valley degeneracy and aniso-
tropic Landé g-factor, an unusual effect can occur [163]: for an electron in a
valley whose axis is oriented along ẑ at an angle θ with an external magnetic

field B, we can choose x̂ in the plane of ẑ and B, such that the Zeeman
Hamiltonian governing spin state evolution is
 = µ B ( g Bcosθσ z + g ⊥ Bsinθσ x ) , (3.5)
where µ B is the Bohr magneton, and σ x , z are the 2 × 2 Pauli spin-1/2 matrices.
This seemingly trivial Hamiltonian can be algebraically transformed into an
equivalent picture for a free electron with g-factor g 0 ≈ 2:
B
 = µB g0
g0
( ( g − g ⊥ ) cosθσz + g ⊥sinθσx + g ⊥cosθσz )
(3.6)
 g − g⊥ g 
= g 0µ B  B θσ z + ⊥ B ⋅ σ  .
cosθ
 g0 g0 
Note that this transformed Hamiltonian indicates that the electron spin
g
acts as if it were a free electron in a renormalized magnetic field ⊥ B ,
g0
plus another magnetic field, oriented along the valley axis, with magnitude
g − g⊥
B cosθ. This additional field is randomized during the fast interval-
g0
ley scattering process; time-dependent perturbation theory shows that it
opens a new channel of spin relaxation, even if the external magnetic field is
perfectly aligned with the initial spin orientation.
This extraordinary mechanism is reminiscent of the Dyakonov–Perel
spin relaxation process which dominates in non-centrosymmetric semicon-
ductor crystal lattices [53, 72], for example the III-V compound semicon-
ductors like GaAs. In that case, broken spatial inversion symmetry allows
spin-orbit interaction to cause a momentum-dependent spin splitting
(Dresselhaus spin-orbit coupling [164]); intravalley scattering during spin
precession about this random effective magnetic field leads to depolariza-
tion. In the anisotropic g-factor mechanism described above, the origin of
the additional random field is rooted instead in the broken time reversal
symmetry induced by the real external magnetic field, and intervalley scat-
tering allows g-factor anisotropy to drive its fluctuation between four differ-
ent orientations. This subtle phenomenon can be experimentally verified in
appropriate spin transport devices by the suppression of spin polarization

with longitudinal magnetic field in the spin-valve effect.
The long-distance germanium spin transport devices we fabricated to
observe this phenomenon [165] are nominally identical in operation to their
silicon counterparts that use the same ballistic hot electron transport meth-
ods [144, 166]. However, substantial changes to the fabrication procedure
were required due to the details of mechanical, chemical, and metallurgical
properties of Ge. The transport layer of >40 Ωcm 325 ± 25 μm-thick Ge(001)
was bonded to a CoFe(2 nm)/NiFe(5 nm)/Cu(3 nm)/n-Si spin detector struc-
ture, whose layer sequence was chosen to maximize the Schottky barrier
height on the Ge side [167], and minimize barrier height on the n-Si side to
facilitate the ballistic transport of electrons from Ge, through the metal, and
into the Si conduction band.
Figure 3.6a shows experimental results from these devices in an in-plane
magnetic field quasi-statically swept through both injector and detector
magnetic thin film coercive fields at a temperature of 41 K. A linear back-
ground has been subtracted for clarity to show only the spin-dependent cur-
rent ∆IC 2 for magnetic field orientation along the in-plane <110> and <100>
directions. Both data show prominent magnetic field-dependent spin depo-
larization with a profile very different from the ordinary spin-valve effect, for
example in Figure 3.3.
FIGURE 3.6 Experimental and simulated spin signal IC2 vs applied magnetic field
B, in an accelerating electric field caused by a voltage of VC1 = 0.6 V over the 325 μm
transport distance in undoped Ge at a temperature of 41 K. Panel (a) shows the
spin-valve effect for the B field along both 110 (in blue) and 100 (in red) due
to switching injector or detector magnetizations in an in-plane B field. The round
(cross) markers are experimental results when B is swept in the positive (negative)
direction, with the solid (dashed) curve corresponding to theoretical simulation.
Panel (b) shows coherent spin precession in an out-of-plane B field. The diamond
markers are experiment results. The solid (dashed) curve is the theoretical simula-
tion with (without) the g-factor anisotropy-induced depolarization. (After Li, P. et
al., Phys. Rev. Lett. 111, 257204, 2013. With permission.)
With a model for this B-dependent total spin relaxation rate, we are able
to simulate the spin-valve experiment data, using a drift-diffusion model
[168, 169] that takes into account the transit time uncertainty, and hence
spin orientation distribution at the detector. The Elliott–Yafet spin lifetime
τ s , ph is a free fitting parameter in this theory, allowing us to determine relax-
ation rates from a single spin-valve measurement—this is not possible with
ordinary spin-valve measurements in materials where this mechanism is
absent. As can be seen by the direct comparison in Figure 3.6a, the theory
matches the experimental result for both <110> and <100> in-plane mag-
netic field orientations very well when τ s , ph = 258 ns. This long spin lifetime
at 41 K is the consequence of vanishing intravalley spin flips due to time
reversal and spatial inversion symmetries at the L-point up to cubic order in
phonon wave-vector [170].
Figure 3.6b shows data and the corresponding drift-diffusion simula-
tion results for a measurement in out-of-plane magnetic field, perpendicular
to the magnetic and spin axis, causing coherent precession and an oscillat-
ing spin detector signal. In this geometry, we must include both spin-lattice
(depolarization) and spin-spin (dephasing) relaxation. Clearly, the theoreti-
cal simulation matches experimental data very well. For comparison, we also
show the simulated result excluding the g-factor anisotropy-induced contri-
bution to the spin relaxation. Its discrepancy with experimental data is not
apparent at low precession angles in small fields, but becomes prominent at
subsequent extrema corresponding to 2π, 3π and 4π rad rotations when B
increases.
Because the electron temperature is easily decoupled from the lattice
temperature in transport conditions at finite electric field in this material
[166], the spin lifetimes extracted from fitting the spin-valve depolarization
features are typically lower than those obtained by correlating zero-magnetic-
field polarization with mean transit time from spin precession data [144,
166] except at the lowest accelerating voltages and temperatures. Figure 3.7
compares the temperature dependence of spin lifetimes at several internal
electric fields to the theoretical Elliott–Yafet prediction [170], which applies
to the germanium electron-phonon system at thermal equilibrium. The spin-
valve-obtained lifetimes systematically drop with increasing electric field,
and are noticeably temperature-independent in high electric fields, unlike
the “Larmor-clock”-derived values [171] from fitting precession and minor-
loop spin polarization data at VC 1 < 0.6 V . The origin of low-temperature spin
lifetime suppression seen in these data is likely due to extrinsic effects, as
has been observed in electron spin resonance studies of Si [59], and our own
observation of inelastic exchange scattering with neutral donors [172].
3.7 OUTLOOK
There has been significant progress in using ballistic hot electron spin injec-
tion and detection techniques for spin transport studies in Si and Ge. However,
there are limitations of these methods. For example, the small ballistic trans-
port transfer ratio is typically no better than 10 -3 - 10 -2 ; the low injection
3.7 Outlook 165
FIGURE 3.7 Temperature dependence of spin lifetime in undoped Ge. Low-

field measurements at VC1 = 0.2 V yield a monotonically increasing spin lifetime
with decreasing temperature, similar to the Elliott–Yafet theoretical prediction
[170]. Electric-field-induced intervalley scattering causes enhanced suppression
of lifetimes. Lifetimes obtained by correlating transit time (Larmor clock from spin
precession measurements) and zero-magnetic-field polarization (from minor-loop
spin-valve measurements) show low-temperature depolarization. (After Li, P. et al.,
Phys. Rev. Lett. 111, 257204, 2013. With permission.)
currents and detection signals obtained will result in sub-unity gain and limit
direct applications of these devices. In addition, our reliance on the ability of
Schottky barriers to serve as hot electron filters presently limits device opera-
tion temperatures to approximately 200 K—although materials with higher
Schottky barrier heights could extend this closer to room temperature—
and also limits application to only non-degenerately doped semiconductors.
Carrier freeze-out in the n-Si spin detection collector at approximately 20 K
introduces a fundamental low-temperature limit as well [169].
Despite these shortcomings, there are also unique capabilities afforded
by this method, such as independent control over internal electric field and
injection current, and spectroscopic control over the injection energy level
[173]. Unlike, for instance, optical techniques, other semiconductor materi-
als should be equally well suited to study with these methods. The purpose is
to use these devices as tools to understand spin transport properties for the
design of spintronic devices, just as the Haynes–Shockley experiment [174]
enabled the design of electronic minority-carrier devices such as the bipolar
junction transistor.
There is still much physics to be done with ballistic hot electron spin
injection and detection. For instance, spin control via time-dependent reso-
nant fields is apparently feasible [175], and application of strain along the
valley axes can suppress intervalley scattering to greatly enhance the spin
lifetime [176, 177]. Similar fabrication techniques as used for vertical devices
can be used to assemble lateral spin transport devices, where in particu-
lar very long transit lengths [148] and the effects of an electrostatic gate to
control the proximity to a Si/SiO2 interface can be investigated [178–180].
The massive spin lifetime suppression induced by the interface that was
observed in this work suggests that such devices can be used to explore the
effects of broken inversion symmetry [53] and paramagnetic [62] or charged
[181, 182] defects on conduction electron spins at the semiconductor/insula-
tor interface. In addition, this opportunity for sensitive characterization may
be of importance to the electronics community as they continue to push
MOSFET scaling toward its ultimate limits.
Hopefully, more experimental groups will develop the technology nec-
essary to compete in this wide-open field. Theorists, too, are eagerly invited
to address topics such as whether this injection technique fully circumvents
the “fundamental obstacle” because of a remaining Sharvin-like effective
resistance [183], or whether it introduces anomalous spin dephasing [169].
ACKNOWLEDGMENTS
I.A. would especially like to thank Douwe Monsma for introducing him to
the field, and Chagaan Baatar and Henryk Temkin for crucial early encour-
agement and continued support.
Since beginning work on this subject in 2006, many students and post-
docs have made essential contributions. Biqin Huang, Hyuk-Jae Jang, and
Jing Li deserve special thanks for fabrication and measurement efforts which
have made much of this work possible. Pengke Li’s combined experimental
and theoretical strengths have been invaluable in developing a deeper under-
standing of the physics. Fruitful collaboration with Prof. Hanan Dery has
similarly been an essential ingredient in the success of this research.
This work has been financially supported by the Office of Naval Research,
DARPA/MTO, Defense Threat Reduction Agency, and the National Science
Foundation.
REFERENCES
1. M. Johnson and R. Silsbee, Thermodynamic analysis of interfacial transport
and of the thermomagnetoelectric system, Phys. Rev. B 35, 4959 (1987).
2. M. Johnson and R. Silsbee, Spin-injection experiment, Phys. Rev. B 37, 5326
(1988).
3. G. Schmidt, D. Ferrand, L. Molenkamp, A. Filip, and B. van Wees, Fundamental
obstacle for electrical spin injection from a ferromagnetic metal into a diffu-
sive semiconductor, Phys. Rev. B 62, R4790 (2000).
4. G. Schmidt, Concepts for spin injection into semiconductors–A review, J.
Phys. D 38, R107 (2005).
5. J. C. Slonczewski, Conductance and exchange coupling of two ferromagnets
separated by a tunneling barrier, Phys. Rev. B 39, 6995 (1989).
6. E. Rashba, Theory of electrical spin injection: Tunnel contacts as a solution of
the conductivity mismatch problem, Phys. Rev. B 62, R16267 (2000).
7. R. Fiederling, M. Keim, G. Reuscher et al., Injection and detection of a spin
polarized current in a n-i-p light emitting diode, Nature 402, 787 (1999).
8. Y. Ohno, D. K. Young, B. Beschoten et al., Electrical spin injection in a ferro-
magnetic semiconductor heterostructure, Nature 402, 790 (1999).
tions. Rev. Mod. Phys. 76, 323 (2004).
References 167
Rev. Lett. 97, 026602 (2006).
density-functional theory., Phys. Rev. Lett. 96, 196101 (2006).
12. W. Schottky, Semiconductor theory of the barrier film, Naturwissenschaften
26 (1938).
13. W. Schottky, Deviations from Ohm’s law in semiconductors, Phys. Z 41
(1940).
14. N. F. Mott, Note on the contact between a metal and an insulator or semi-
conductor, Proc. Cambridge Philos. Soc. 34, 568 (1938).
15. N. F. Mott, The theory of crystal rectifiers, Proc. R. Soc. Lond. A 171, 27 (1939).
16. J. Tersoff, Schottky barrier heights and the continuum of gap states, Phys. Rev.
Lett. 52, 465 (1984).
17. S. Sze, Physics of Semiconductor Devices, 2nd edn., Wiley-Interscience, New
York, 1981.
18. C. Crowell, L. Howarth, W. Spitzer, and E. Labate, Attenuation length mea-
surements of hot electrons in metal films, Phys. Rev. 127, 2006 (1962).
19. M. Heiblum, M. I. Nathan, and M. Eizenberg, Energy band discontinuities in
heterojunctions measured by internal photoemission, Appl. Phys. Lett. 47, 503
(1985).
20. C. Crowell and S. Sze, Hot electron transport and electron tunneling in thin
film structures , in Physics of Thin Films, vol. 4, G. Hass and R. Thun (Eds.),
Academic press, New York, 1967.
21. C. Crowell and S. Sze, Quantum-mechanical reflection of electrons at metal-
semiconductor barriers: Electron transport in semiconductor-metal-semicon-
ductor structures, J. Appl. Phys. 37, 2683 (1966).
22. S. Sze and H. Gummel, Appraisal of semiconductor-metal-semiconductor
transistor, Solid-State Electron. 9, 751 (1966).
23. A. Sommerfeld and H. Bethe, Electronentheorie der metalle, in Handbuch der
Physik, vol. 24, S. Flugge (Ed.), Springer, Berlin, 1933.
24. C. Duke, Tunneling in Solids, Academic, New York, 1969.
25. J. Fisher and I. Giaever, Tunneling through thin insulating layers, J. Appl. Phys.
32, 172 (1961).
26. I. Giaever, Energy gap in superconductors measured by electron tunneling,
Phys. Rev. Lett. 5, 147 (1960).
27. J. G. Simmons, Generalized thermal j-v characteristic for electric tunnel effect,
J. Appl. Phys. 35, 2655 (1964).
28. W. F. Brinkman, R. C. Dynes, and J. M. Rowell, Tunneling conductance of
asymmetrical barriers, J. Appl. Phys. 41, 1915 (1970).
29. J. S. Moodera, J. Nassar, and G. Mathon, Spin-tunneling in ferromagnetic junc-
tions, Annu. Rev. Mater. Sci. 29, 381 (1999).
30. Y. Yang, P. F. Ladwig, Y. A. Chang et al., Thermodynamic evaluation of the
interface stability between selected metal oxides and Co, J. Mater. Res. 19, 1181
(2004).
31. C. Mead, The tunnel-emission amplifier, Proc. IRE 48, 359 (1960).
32. C. Mead, Operation of tunnel-emission devices, J. Appl. Phys. 32, 646 (1961).
33. C. A. Mead, Transport of hot electrons in thin gold films, Phys. Rev. Lett. 8, 56
(1962).
34. J. Spratt, R. Schwartz, and W. Kane, Hot electrons in metal films: Injection and
collection, Phys. Rev. Lett. 6, 341 (1961).
35. M. Heiblum, Tunneling hot-electron transfer amplifier: A hot-electron GaAs
device with current gain, Appl. Phys. Lett. 47, 1105 (1985).
36. M. Heiblum, Direct observation of ballistic transport in GaAs, Phys. Rev. Lett.
55, 2200 (1985).
37. D. Monsma, J. Lodder, T. Popma, and B. Dieny, Perpendicular hot electron

spin-valve effect in a new magnetic field sensor: The spin-valve transistor,
Phys. Rev. Lett. 74, 5260 (1995).
38. D. Monsma, R. Vlutters, and J. Lodder, Room temperature-operating spin-
valve transistors formed by vacuum bonding, Science 281, 407 (1998).
39. I. Appelbaum, K. Russell, D. Monsma et al., Luminescent spin-valve transistor,
Appl. Phys. Lett. 83, 4571 (2003).
40. R. Jansen, The spin-valve transistor: A review and outlook, J. Phys. D 36, R289
(2003).
41. K. Mizushima, T. Kinno, T. Yamauchi, and K. Tanak, Energy-dependent hot
electron transport across a spin-valve, IEEE Trans. Magn. 33, 3500 (1997).
42. G. Tae, J. Eom, J. Song, and K. Kim, Spin-polarized hot electron injection into
two-dimensional electron gas by magnetic tunnel transistor, Jpn. J. Appl. Phys.
46, 7717 (2007).
43. W. Rippard and R. Buhrman, Spin-dependent hot electron transport in Co/Cu
thin films, Phys. Rev. Lett. 84, 971 (2000).
44. S. van Dijken, X. Jiang, and S. S. P. Parkin, Room temperature operation of
a high output current magnetic tunnel transistor, Appl. Phys. Lett. 80, 3364
(2002).
45. D. Monsma, The spin-valve transistor, PhD thesis, U. Twente, 1998.
(ISBN:903651049X).
46. I. Appelbaum, D. Monsma, K. Russell, V. Narayanamurti, and C. Marcus, Spin-
valve photo-diode, Appl. Phys. Lett. 83, 3737 (2003).
47. B. Huang and I. Appelbaum, Perpendicular hot-electron transport in the spin-
valve photodiode, J. Appl. Phys. 100, 034501 (2006).
48. B. Huang, I. Altfeder, and I. Appelbaum, Spin-valve phototransistor, Appl.
Phys. Lett. 90, 052503 (2007).
49. S. Wolf, D. Awschalom, R. Buhrman et al., Spintronics: A spin-based electron-
ics vision for the future, Science 294, 1488 (2001).
injection into GaAs from a magnetic tunnel transistor, Phys. Rev. Lett. 90,
256603 (2003).
51. F. Monzon, H. Tang, and M. Roukes, Magnetoelectronic phenomena at a fer-
romagnet-semiconductor interface, Phys. Rev. Lett. 84, 5022 (2000).
52. M. Johnson and R. Silsbee, Interfacial charge-spin coupling: Injection and
detection of spin magnetization in metals, Phys. Rev. Lett. 55, 1790 (1985).
53. M. Dyakonov and V. Perel, Spin relaxation of conduction electrons in noncen-
trosymmetric semiconductors, Sov. Phys. Solid State 13, 3023 (1972).
injection into GaAs from a magnetic tunnel transistor, IBM Res. Dev. 50, 111
(2006).
55. A. M. Portis, A. F. Kip, C. Kittel, and W. H. Brattain, Electron spin resonance
in a silicon semiconductor, Phys. Rev. 90, 988 (1953).
56. G. Feher, Electron spin resonance experiments on donors in silicon 1.
Electronic structure of donors by the electron nuclear double resonance tech-
nique, Phys. Rev. 114, 1219 (1959).
57. G. Feher and E. Gere, Electron spin resonance experiments on donors in sili-
con 2. electron spin relaxation effects, Phys. Rev. 114, 1245 (1959).
58. D. Wilson and G. Feher, Electron spin resonance experiments on donors in
silicon 3. investigation of excited states by application of uniaxial stress and
their importance in relaxation processes, Phys. Rev. 124, 1068 (1959).
59. D. Lepine, Spin resonance of localized and delocalized electrons in phos-
phorus-doped silicon between 20 and 300 degrees K, Phys. Rev. B 2, 2429
(1970).
60. Y. Yafet, G-factors and spin-lattice relaxation of conduction electrons in Solid
State Physics-advances in Research and Applications, vol. 14, F. Seitz and D.
Turnbull (Eds.), Academic Press, New York, 1963.
References 169
61. R. Ghosh and R. Silsbee, Spin-spin scattering in a silicon two-dimensional

electron gas, Phys. Rev. B 46, 12508 (1992).
62. M. Xiao, I. Martin, E. Yablonovitch, and H. Jiang, Electrical detection of the
spin resonance of a single electron in a silicon field-effect transistor, Nature
430, 435 (2004).
spintronics, Acta Phys. Slovaca 57, 565 (2007).
64. W. Louisell, R. Pidd, and H. Crane, An experimental measurement of the gyro-
magnetic ratio of the free electron, Phys. Rev. 94, 7 (1954).
65. V. Sverdlov and S. Selberherr, Silicon spintronics: Progress and challenges,
Phys. Rep. 585, 1 (2015). ISSN 0370-1573. Silicon spintronics: Progress and
challenges.
66. R. Jansen, Silicon spintronics, Nat. Mater. 11, 400 (2012).
67. I. Žutić and J. Fabian, Spintronics: Silicon twists, Nature 447, 269 (2007).
68. P. Li and I. Appelbaum, Symmetry, distorted band structure, and spin-orbit
coupling of group-III metal-monochalcogenide monolayers, Phys. Rev. B 92,
195129 (2015).
69. M. Dresselhaus, G. Dresselhaus, and A. Jorio, Group Theory: Application to the
Physics of Condensed Matter, Springer, Berlin, 2008.
70. G. Feher, J. P. Gordon, E. Buehler, E. A. Gere, and C. D. Thurmond, Spontaneous
emission of radiation from an electron spin system, Phys. Rev. 109, 221 (1958).
71. B. Kane, A silicon-based nuclear spin quantum computer, Nature 393, 1331
(1998).
72. M. Dyakonov and V. Perel, Spin orientation of electrons associated with the
interband absorption of light in semiconductors, Sov. Phys. JETP 33, 1053
(1971).
73. G. Lampel, Nuclear dynamic polarization by optical electronic saturation and
optical pumping in semiconductors, Phys. Rev. Lett. 20, 491 (1968).
74. N. Bagraev, L. Vlasenko, and R. Zhitnikov, Optical orientation of Si29 nuclei in
n-type silicon and its dependence on the pumping light intensity, Sov. Phys.
JETP 44, 500 (1976).
75. D. Awschalom and M. E. Flatté, Challenges for semiconductor spintronics,
Nat. Phys. 3, 153 (2007).
77. Y. Q. Jia, R. C. Shi, and S. Y. Chou, Spin-valve effects in nickel/silicon/nickel
junctions, IEEE Trans. Magn. 32, 4707 (1996).
78. K. I. Lee, H. J. Lee, J. Y. Chang et al., Spin-valve effect in an FM/Si/FM junction,
J. Mater. Sci. 16, 131 (2005).
79. W. J. Hwang, H. J. Lee, K. I. Lee et al., Spin transport in a lateral spin-injection
device with an FM/Si/FM junction, J. Magn. Magn. Mater. 272–276, 1915 (2004).
80. S. Hacia, T. Last, S. F. Fischer, and U. Kunze, Study of spin-valve operation in
permalloy-SiO2-Silicon nanostructures, J. Supercond. 16, 187 (2003).
81. S. Hacia, T. Last, S. F. Fischer, and U. Kunze, Magnetotransport study of
nanoscale permalloy-si tunnelling structures in lateral spin-valve geometry,
J. Phys. D 37, 1310 (2004).
82. C. Dennis, C. Sirisathitkul, G. Ensell, J. Gregg, and S. M. Thompson, High cur-
rent gain silicon-based spin transistor, J. Phys. D 36, 81 (2003).
83. C. Dennis, J. Gregg, G. Ensell, and S. Thompson, Evidence for electrical spin
tunnel injection into silicon, J. Appl. Phys. 100, 043717 (2006).
84. C. Dennis, C. Tiusan, R. Ferreira, et al., Tunnel barrier fabrication on Si and its
impact on a spin transistor, J. Magn. Magn. Mater. 290, 1383 (2005).
85. J. Gregg, I. Petej, E. Jouguelet, and C. Dennis, Spin electronics–a review, J. Phys.
D 35, R121 (2002).
86. S. Sugahara and M. Tanaka, A spin metal-oxide-semiconductor field-effect
transistor using half-metallic-ferromagnet contacts for the source and drain,
Appl. Phys. Lett. 84, 2307 (2004).
87. M. Cahay and S. Bandyopadhyay, Room temperature silicon spin-based tran-

sistors, in Device Applications of Silicon Nanocrystals and Nanostructures, N.
Koshida (Ed.), Springer, New York, 2009.
88. R. Gareev, L. Pohlmann, S. Stein, et al., Tunneling in epitaxial Fe/Si/Fe struc-
tures with strong antiferromagnetic interlayer coupling, J. Appl. Phys. 93, 8038
(2003).
90. B. Min, J. Lodder, R. Jansen, and K. Motohashi, Cobalt-Al2O3-silicon tunnel
contacts for electrical spin injection into silicon, J. Appl. Phys. 99, 08S701
(2006).
91. B. Min, K. Motohashi, C. Lodder, and R. Jansen, Tunable spin-tunnel contacts
to silicon using low-work-function ferromagnets, Nat. Mater. 5, 817 (2006).
92. K. Wang, J. Stehlik, and J.-Q. Wang, Tunneling characteristics across nanoscale
metal ferric junction lines into doped Si, Appl. Phys. Lett. 92, 152118 (2008).
93. T. Uhrmann, T. Dimopoulos, H. Brückl et al., Characterization of embedded
MgO/ferromagnet contacts for spin injection in silicon, J. Appl. Phys. 103,
063709 (2008).
94. T. Dimopoulos, D. Schwarz, T. Uhrmann et al., Magnetic properties of embed-
ded ferromagnetic contacts to silicon for spin injection, J. Phys. D 42, 085004
(2009).
95. R. Mattana, J.-M. George, H. Jaffrès, et al., Electrical detection of spin accu-
mulation in a p-type GaAs quantum well, Phys. Rev. Lett. 90, 166601 (2003).
96. A. Fert, J.-M. George, H. Jaffres, and R. Mattana, Semiconductors between
spin-polarized sources and drains, IEEE Trans. Elec. Dev. 54, 921 (2007).
97. P. Li and H. Dery, Theory of spin-dependent phonon-assisted optical transi-
tions in silicon, Phys. Rev. Lett. 105, 037204 (2010).
98. B. Jonker, G. Kioseoglou, A. Hanbicki, C. Li, and P. Thompson, Electrical spin-
injection into silicon from a ferromagnetic metal/tunnel barrier contact. Nat.
Phys. 3, 542 (2007).
99. G. Kioseoglou, A. T. Hanbicki, R. Goswami, et al., Electrical spin injection into
Si: A comparison between Fe/Si Schottky and Fe/Al2O3tunnel contacts, Appl.
Phys. Lett. 94, 122106 (2009).
100. L. Grenet, M. Jamet, P. Noé et al., Spin injection in silicon at zero magnetic
field, Appl. Phys. Lett. 94, 032502 (2009).
101. O. van’t Erve, A. Hanbicki, M. Holub, C. Li, C. Awo-Affouda et al. Electrical
injection and detection of spin-polarized carriers in silicon in a lateral trans-
port geometry, Appl. Phys. Lett. 91, 212109 (2007).
102. T. Sasaki, T. Oikawa, T. Suzuki et al., Electrical spin injection into silicon using
MgO tunnel barrier, Appl. Phys. Express 2, 53003 (2009).
103. T. Sasaki, T. Oikawa, T. Suzuki, et al. Temperature dependence of spin dif-
fusion length in silicon by Hanle-type spin precession Appl. Phys. Lett. 96,
122101 (2010).
104. Y. Ando, K. Hamaya, K. Kasahara et al., Electrical injection and detection of
spin-polarized electrons in silicon through an Fe3Si/Si Schottky tunnel barrier,
Appl. Phys. Lett. 94, 182105 (2009).
105. M. Shiraishi, Y. Honda, E. Shikoh et al., Spin transport properties in silicon in
a nonlocal geometry, Phys. Rev. B 83, 241204 (2011).
106. M. Kameno, Y. Ando, T. Shinjo et al., Spin drift in highly doped n-type Si,
Appl. Phys. Lett. 104, 092409 (2014).
107. T. Sasaki, Y. Ando, M. Kameno et al., Spin transport in nondegenerate Si with
a spin MOSFET structure at room temperature, Phys. Rev. Appl 2, 034005
(2014).
108. S. Dash, S. Sharma, R. Patel, M. de Jong, and R. Jansen, Electrical creation of
spin polarization in silicon at room temperature, Nature 462, 491 (2009).
109. A. Dankert, R. S. Dulal, and S. P. Dash, Efficient spin injection into silicon and
the role of the Schottky barrier, Sci. Rep. 3, 3196 (2013).
References 171
110. C. Li, O. van’t Erve, and B. Jonker, Electrical injection and detection of spin
accumulation in silicon at 500K with magnetic metal/silicon dioxide contacts,
Nat. Commun. 2, 245 (2011).
111. N. W. Gray and A. Tiwari, Room temperature electrical injection and detec-
tion of spin polarized carriers in silicon using MgO tunnel barrier, Appl. Phys.
Lett. 98, 102112 (2011).
112. A. Jain, J.-C. Rojas-Sanchez, M. Cubukcu et al., Crossover from spin accumula-
tion into interface states to spin injection in the germanium conduction band,
Phys. Rev. Lett. 109, 106603 (2012).
113. W. Han, X. Jiang, A. Kajdos et al., Spin injection and detection in lanthanum-
and niobium-doped SrTiO3 using the Hanle technique, Nat. Commun. 4, 2134
(2013).
114. K.-R. Jeon, B.-C. Min, Y.-H. Park, S.-Y. Park, and S.-C. Shin, Electrical inves-
tigation of the oblique Hanle effect in ferromagnet/oxide/semiconductor con-
tacts, Phys. Rev. B 87, 195311 (2013).
115. O. Txoperena, M. Gobbi, A. Bedoya-Pinto et al., How reliable are Hanle mea-
surements in metals in a three-terminal geometry? Appl. Phys. Lett. 102,
192406 (2013).
116. O. Txoperena and F. Casanova, Spin injection and local magnetoresistance
effects in three-terminal devices, J. Phys. D 49, 133001 (2016).
117. H. N. Tinkey, P. Li, and I. Appelbaum, Inelastic electron tunneling spec-
troscopy of local “spin accumulation” devices, Appl. Phys. Lett. 104, 232410
(2014).
118. I. Appelbaum, H. N. Tinkey, and P. Li, Self-consistent model of spin accumula-
tion magnetoresistance in ferromagnet/insulator/semiconductor tunnel junc-
tions, Phys. Rev. B 90, 220402 (2014).
119. H. N. Tinkey, H. Dery, and I. Appelbaum, Defect passivation by proton irra-
diation in ferromagnet-oxide-silicon junctions, Appl. Phys. Lett. 109, 142407
(2016).
120. Y. Song and H. Dery, Magnetic-field-modulated resonant tunneling in ferro-
magnetic-insulator-nonmagnetic junctions, Phys. Rev. Lett. 113, 6 (2014).
121. O. Txoperena, Y. Song, L. Qing et al., Impurity-assisted tunneling magnetore-
sistance under a weak magnetic field, Phys. Rev. Lett. 113, 146601 (2014).
122. H. Inoue, A. Swartz, N. Harmon et al., Origin of the magnetoresistance in
oxide tunnel junctions determined through electric polarization control of the
interface, Phys. Rev. X 5, 7 (2015).
123. Z. Yue, M. C. Prestgard, A. Tiwari, and M. E. Raikh, Resonant magnetotun-
neling between normal and ferromagnetic electrodes in relation to the three-
terminal spin transport, Phys. Rev. B 91, 195316 (2015).
124. M. Tran, H. Jaffrès, C. Deranlot et al., Enhancement of the spin accumulation
at the Interface between a spin-polarized tunnel junction and a semiconduc-
tor, Phys. Rev. Lett. 102, 036601 (2009).
125. Y. Ando, Y. Maeda, K. Kasahara et al., Electric-field control of spin accumula-
tion signals in silicon at room temperature, Appl. Phys. Lett. 99, 132511 (2011).
126. X. Lou, C. Adelmann, M. Furis et al., Electrical detection of spin accumulation
at a ferromagnet-semiconductor interface, Phys. Rev. Lett. 96, 176603 (2006).
127. M. Bolduc, C. Awo-Affouda, A. Stollenwerk et al., Above room temperature
ferromagnetism in Mn-ion implanted Si, Phys. Rev. B 71, 033302 (2005).
128. S. C. Erwin and I. Žutić, Tailoring ferromagnetic chalcopyrites, Nat. Mater. 3,
410 (2004).
129. S. Cho, S. Choi, G. Cha et al., Synthesis of new pure ferromagnetic semicon-
ductors: MnGeP2 and MnGeAs2, Solid State Commun. 129, 609 (2004).
130. S. Cho, S. Choi, G.-B. Cha et al., Room-temperature ferromagnetism in Zn1-
x Mn xGeP2 semiconductors, Phys. Rev. Lett. 88, 257203 (2002).
131. Y. Ishida, D. D. Sarma, K. Okazaki et al., In situ photoemission study of
the room temperature ferromagnet ZnGeP2:Mn, Phys. Rev. Lett. 91, 107202
(2003).
132. A. Schmehl, V. Vaithyanathan, A. Herrnberger et al., Epitaxial integration of

the highly spin-polarized ferromagnetic semiconductor EuO with silicon and
GaN, Nat. Mater. 6, 882 (2007).
133. Y. Pu, P. M. Odenthal, R. Adur, et al., Ferromagnetic resonance spin pump-
ing and electrical spin injection in silicon-based metal-oxide-semiconduc-
tor heterostructures, Phys. Rev. Lett. 115, 246602 (2015).
134. E. Shikoh, K. Ando, K. Kubo et al., Spin-pump-induced spin transport in
p-type Si at room temperature, Phys. Rev. Lett. 110, 127201 (2013).
135. K. Ando and E. Saitoh, Observation of the inverse spin Hall effect in silicon,
Nat. Commun. 3, 629 (2012).
136. S. Ganichev, S. Danilov, V. Bel’kov et al., Pure spin currents induced by spin-
dependent scattering processes in SiGe quantum well structures, Phys. Rev. B
75, 155317 (2007).
137. M. Friesen, P. Rugheimer, D. Savage et al., Practical design and simulation of
silicon-based quantum-dot qubits, Phys. Rev. B 67, 121301 (2003).
138. P. Zhang and M. Wu, Spin diffusion in Si/SiGe quantum wells: Spin relax-
ation in the absence of Dyakonov-Perel relaxation mechanism, Phys. Rev. B 79,
075303 (2009).
139. N. Bagraev, N. Galkin, W. Gehlhoff et al., Spin interference in silicon three-
terminal one-dimensional rings, J. Phys. D 18, L567 (2006).
140. P. Zhang and M. Wu, Hole spin relaxation in strained asymmetric Si/SiGe and
Ge/SiGe quantum wells, Phys. Rev. B 80, 155311 (2009).
141. S. van Dijken, X. Jiang, and S. S. P. Parkin, Giant magnetocurrent exceeding
3400% in magnetic tunnel transistors with spin-valve base layers, Appl. Phys.
Lett. 83, 951 (2003).
142. B. Huang, D. J. Monsma, and I. Appelbaum, Experimental realization of a sili-
con spin field-effect transistor, Appl. Phys. Lett. 91, 072501 (2007).
143. B. Huang, L. Zhao, D. J. Monsma, and I. Appelbaum, 35% magnetocurrent with
spin transport through Si, Appl. Phys. Lett. 91, 052501 (2007).
144. B. Huang, D. J. Monsma, and I. Appelbaum, Coherent spin transport through
a 350 micron thick silicon wafer, Phys. Rev. Lett. 99, 177209 (2007).
Rev. Lett. 74, 3273 (1995).
146. L. D. Bell, S. J. Manion, M. H. Hecht et al., Characterizing hot-carrier transport
in silicon heterostructures with the use of ballistic-electron-emission micros-
copy, Phys. Rev. B 48, 5712 (1993).
147. J. Li, B. Huang, and I. Appelbaum, Oblique Hanle effect in semiconductor spin
transport devices, Appl. Phys. Lett. 92, 142507 (2008).
148. B. Huang, H.-J. Jang, and I. Appelbaum, Geometric dephasing-limited Hanle
effect in long-distance lateral silicon spin transport devices, Appl. Phys. Lett.
93, 162508 (2008).
149. J. Li and I. Appelbaum, Inelastic spin depolarization spectroscopy in silicon,
J. Appl. Phys. 114, 033705 (2013).
150. C. Canali, C. Jacoboni, F. Nava, G. Ottaviani, and A. Quaranta, Electron drift
velocity in silicon, Phys. Rev. B 12, 2265 (1975).
151. C. Jacoboni, C. Canali, G. Ottaviani, and A. Quaranta, A review of some charge
transport properties of silicon, Solid State Electron. 20, 77 (1977).
152. H.-J. Jang, J. Xu, J. Li, B. Huang, and I. Appelbaum, Non-ohmic spin transport
in n-type doped silicon, Phys. Rev. B 78, 165329 (2008).
153. I. Altfeder, B. Huang, I. Appelbaum, and B. C. Walker, Self-assembly of epitax-
ial monolayers for vacuum wafer bonding, Appl. Phys. Lett. 89, 223127 (2006).
154. W. Hanle, The magnetic influence on the polarization of resonance fluores-
cence, Z. Physik 30, 93 (1924).
155. B. Huang, Vertical transport silicon spintronic devices, PhD thesis, U.
Delaware, 2008.
References 173
156. J. L. Cheng, M. W. Wu, and J. Fabian, Theory of the spin relaxation of conduc-
tion electrons in silicon, Phys. Rev. Lett. 104, 016601 (2010).
157. P. Li and H. Dery, Spin-orbit symmetries of conduction electrons in silicon,
Phys. Rev. Lett. 107, 107203 (2011).
158. Y. Song and H. Dery, Analysis of phonon-induced spin relaxation processes in
159. F. J. Morin, T. H. Geballe, and C. Herring, Temperature dependence of the
piezoresistance of high-purity silicon and germanium, Phys. Rev. 105, 525
(1957).
160. J. Tauc and E. Antončík, Optical observation of spin-orbit interaction in ger-
manium, Phys. Rev. Lett. 5, 253 (1960).
161. M. Cardona and H. S. Sommers, Effect of temperature and doping on the
reflectivity of germanium in the fundamental absorption region, Phys. Rev.
122, 1382 (1961).
162. L. M. Roth and B. Lax, g factor of electrons in germanium, Phys. Rev. Lett. 3,
217 (1959).
163. J.-N. Chazalviel, Spin relaxation of conduction electrons in highly-doped
n-type germanium at low temperature, J. Phys. Chem. Solids 36, 387 (1975).
ISSN 0022-3697.
164. G. Dresselhaus, Spin-orbit coupling effects in zinc blende structures, Phys. Rev.
100, 580 (1955).
165. P. Li, J. Li, L. Qing, H. Dery, and I. Appelbaum, Anisotropy-driven spin relax-
ation in germanium, Phys. Rev. Lett. 111, 257204 (2013).
166. J. Li, L. Qing, H. Dery, and I. Appelbaum, Field-induced negative differential
spin lifetime in silicon, Phys. Rev. Lett. 108, 157201 (2012).
167. Y. Zhou, M. Ogawa, X. Han, and K. L. Wang, Alleviation of Fermi-level pin-
ning effect on metal/germanium interface by insertion of an ultrathin alumi-
num oxide, Appl. Phys. Lett. 93, 202105 (2008).
168. S. Crooker, M. Furis, X. Lou et al., Imaging spin transport in lateral ferromag-
net/semiconductor structures, Science 309, 2191 (2005).
169. B. Huang and I. Appelbaum, Spin dephasing in drift-dominated semiconduc-
tor spintronics devices, Phys. Rev. B 77, 165331 (2008).
170. P. Li, Y. Song, and H. Dery, Intrinsic spin lifetime of conduction electrons in
germanium, Phys. Rev. B 86, 085202 (2012).
171. B. Huang and I. Appelbaum, Time-of-flight spectroscopy via spin precession:
The Larmor clock and anomalous spin dephasing in silicon, Phys. Rev. B 82,
241202 (2010).
172. L. Qing, J. Li, I. Appelbaum, and H. Dery, Spin relaxation via exchange with
donor impurity-bound electrons, Phys. Rev. B 91, 241405 (2015).
173. B. Huang, D. J. Monsma, and I. Appelbaum, Spin lifetime in silicon in the pres-
ence of parasitic electronic effects, J. Appl. Phys. 102, 013901 (2007).
174. J. Haynes and W. Shockley, The mobility and life of injected holes and electrons
in germanium, Phys. Rev. 81, 835 (1951).
175. C. C. Lo, J. Li, I. Appelbaum, and J. J. L. Morton, Microwave manipulation
of electrically injected spin-polarized electrons in silicon, Phys. Rev. Appl. 1,
014006 (2014).
176. H. Dery, Y. S. Song, P. Li and I. Zutic, Silicon spin communication., Appl. Phys.
Lett. 99, 082502 (2011).
177. J.-M. Tang, B. T. Collins, and M. E. Flatté, Electron spin-phonon interaction
symmetries and tunable spin relaxation in silicon and germanium, Phys. Rev. B
85, 045202 (2012).
178. H.-J. Jang and I. Appelbaum, Spin polarized electron transport near the Si/SiO2
Interface, Phys. Rev. Lett. 103, 117202 (2009).
179. J. Li and I. Appelbaum, Modeling spin transport in electrostatically-gated lat-
eral-channel silicon devices: Role of interfacial spin relaxation, Phys. Rev. B 84,
165318 (2011).
180. J. Li and I. Appelbaum, Lateral spin transport through bulk silicon, Appl. Phys.
Lett. 100, 162408 (2012).
181. E. Y. Sherman, Random spin–orbit coupling and spin relaxation in symmetric
quantum wells, Appl. Phys. Lett. 82, 209 (2003).
182. J. Shim, K. Raman, Y. Park et al., Large spin diffusion length in an amorphous
organic semiconductor, Phys. Rev. Lett. 100, 226603 (2008).
183. E. Rashba, Inelastic scattering approach to the theory of a magnetic tunnel
transistor source, Phys. Rev. B 68, 241310 (2003).
4
Tunneling
Magnetoresistance,
Spin-Transfer and
Spinorbitronics
with (Ga,Mn)As
4.1 Tunneling Magnetoresistance in the Valence-Band of

p-Type Semiconductors 180
4.1.1 Introduction 180
4.1.2 Overview of TMR Phenomena Including
(Ga,Mn)As 181
4.1.3 Material and Chemical Trends 185
4.1.3.1 Atomic Picture of Mn-Doped GaAs
Host (See Appendix A) 185
175
176 Chapter 4. Tunneling Magnetoresistance, Spin-Transfer and Spinorbitronics
4.1.3.2 Electronic and Ferromagnetic

Properties 186
4.1.3.3 Impact on Tunneling Transport
Properties 188
4.1.4 Experiments vs. 6-Band k · p Theory of
Tunneling Spin-Transport 190
4.2 Tunneling Anisotropic Magnetoresistance in the
Valence Band of p-Type Semiconductors 192
4.2.1 Overview of TAMR Phenomena 192
4.2.2 TAMR with Holes 194
4.2.3 Tunneling Spectroscopy in the Valence Bands
of (Ga,Mn)As: Resonant TAMR in Double
Tunnel Junctions 196
4.3 Skew-Tunneling and Anomalous Tunnel Hall Effect–
Unidirectional Spin-Hall Magnetoresistance with
(Ga,Mn)As 202
4.3.1 Skew Transmission and Skew Tunneling in the
VB from Atomic SOI: A Simplified Perturbative
Approach 205
4.3.2 Transmission and Chirality within the 6-Band
k · p Kane Model 207
4.4 Spin-Transfer in the Valence Band of p-Type
Semiconductors 211
4.4.1 Short Overview of Spin Transfer Phenomena
with Out-of-Plane and In-Plane Current
Injection 211
4.4.2 Spin Transfer with Holes in (Ga,Mn)As Tunnel
Devices 212
4.4.3 Phase Diagram for CIMS with (Ga,Mn)As 213
4.4.4 Discussion 216
4.5 Summary 217
Acknowledgments 218
Appendix A: Exchange Interactions in (Ga,Mn)As: From
the Mn Isolated Atom to the Metallic Phase 219
Appendix B: Description of the 6-Band k · p Theory 220
k · p Approach for Describing Metallic (Ga,Mn)As 221
The 6-Band k · p Kane Model (Issued from the
14-Band) 223
Tunneling Magnetoresistance, Spin-Transfer and Spinorbitronics 177
The 6-Band k · p Kane Model: Eigenvectors and

Effective Mass in III–V Heterostructures 224
Appendix C: The 30-Band Full-BZ k · p Treatment of the
Skew Tunneling 226
Appendix D: Theory of Spin-Transport and Spin-Transfer
within the k · p Framework 230
General Arguments on Spin Transfer Phenomena 232
Spin Transfer Phenomena with SOI 233
References 235
T
his chapter reviews some fundamental properties of tunneling spin-cur-
rent, spin-transport of holes and spin-transfer of angular momenta in the
valence band of magnetic III–V tunneling heterojunctions involving the
ferromagnetic semiconductor (Ga,Mn)As. This material is further discussed
in Chapter 9, Volume 2 and has recently been the focus of some extended
reports where readers could find additional details of the electronic struc-
ture [1, 2]. Discussion complementing this chapter can be found throughout
this book, for example, on electric-field control of magnetism in Chapter 13,
Volume 2, spin torque and spin-orbit torque in Chapters 7–9, Volume 2, Hall
effects in Chapter 8, Volume 2, topological insulators in Chapters 14 and 15,
Volume 2, and magnetic quantum dots in Chapter 6, Volume 3.
The present review includes some more up-to-date developments since
2012 with (Ga,Mn)As involving principles of current-in-plane spin-orbit
torques [3], anatomy of spin transport at interface, as well as spin-Hall mag-
netoresistance (SMR) and its unidirectional character(U-SMR) [4]. These
phenomena together with their experimental evidence are associated with
new physical properties, and these are made possible via the recent strong
interest in the novel spinorbitronics area. Moreover, (Ga,Mn)As like (Ge,Mn)
for group IV semiconductors, remains a unique prototype group III–V
semiconductor which demonstrates non-zero exchange interactions and
carrier-mediated ferromagnetism. This makes possible the development of
spinorbitronics devices with their particular interest and properties appear-
ing as soon as both exchange strengths and spin-orbit interactions (SOI)
come into play. However, the full metallic character of (Ga,Mn)As described
by exchange-split holes in the valence bands (VB) of the III–V host is still
under debate, owing to some latest results and experiments. The picture
of exchange-split delocalized host of holes interacting with localized Mn
via the Zener-like picture of ferromagnetism is strongly debated, favoring
an impurity band description which seems to be demonstrated by recent
spectroscopic tunneling experiments. Although this issue is still unsolved,
(Ga,Mn)As remains an important prototype material to evidence new types
of spinorbitronics phenomena supported by III–V hosts, at least up to its
Curie temperature, TC of about 200 K [1, 2, 5–8]. Very recently, it has been
reported that the (In,Fe)As compound could support ferromagnetism in its
conduction band (CB) which may be very attractive [9] for spintronic appli-
cations in the CB of semiconducting heterostructures.
Among the strong interest in the field of spintronics and today’s spinor-
bitronics with semiconductors [10, 11], this current research has its roots in
the work of Tanaka et al. from the early 2000s [12]. This research is generally
motivated by (1) the wealth of the physics involving transport of spin-orbit
coupled states carrying quantized angular momenta in epitaxial hetero-
structures; (2) its application to spin-transfer torque phenomena and its deep
fundamental understanding; (3) the need to use materials and tunneling
devices providing an efficient spin injection into group III–V semiconduc-
tors [13–16] and able to circumvent the conductivity mismatch issue [17];
and possibly (4) to use the high coherence of hole spin states in a semicon-
ductor quantum dot [18, 19].
Generating a current of angular momenta (both spin and orbital) in a
semiconductor and converting the latter information into an electrical signal
via tunnel injection [20] or tunnel filtering is still challenging. This appears
as two important prerequisites before any spin manipulation. Despite a rela-
tive low TC , currently reaching 200 K nowadays, the p-type ferromagnetic
semiconductor (Ga,Mn)As, largely described in Chapter 9, Volume 2 [21] and
already well documented [1, 2, 21–27], offers the possibility of a good inte-
gration with conventional III–V species and related heterostructures. The
(Ga,Mn)As ferromagnetism properties may, in the simple case, be semi-quan-
titatively understood by the so-called p-d Zener approach [28]. This describes
a mean-field approach for the average exchange interactions between delocal-
ized p-type carriers and 3d Mn local magnetization in the host III–V semi-
conducting host [29, 30]; although a recent alternative approach focuses on an
impurity-band mostly uncoupled with the host [31–34]. In addition, (Ga,Mn)
As exhibits a specific anisotropic character of its different valence bands,
location of exchange interactions, SOI, and of an anisotropic strain-field [29]
which makes it very attractive for further functionalities. In that sense, III–V
heterostructures including (Ga,Mn)As as a ferromagnetic reservoir often
represent model systems to study some new properties of spin-injection.
On the other hand, spin-currents, SOI, and their interplay are the fun-
damentals of intriguing physical phenomena like the spin-Hall effect (SHE)
and the inverse spin-Hall effect (ISHE) [35–41], and inverse Edelstein effects
(IEE) [42–44] as recently observed with Rashba-states or topological insula-
tors (TI). These phenomena manifest themselves by a left/right asymmetry in
the scattering process of spin-polarized carriers along the transverse direc-
tion of their flow, giving rise to spin-to-charge conversion and vice versa.
These are currently the basis of new functionalities like spin-orbit torques
(SOT) at magnetic/SOI interfaces [3, 45–47] in mind of magnetic switching
using in-plane currents (CIP). In that context, investigations of both SOI and
exchange strengths in solids and at interfaces [48, 49] is of a prime impor-
tance e.g. for the determination of the generalized spin-mixing conductances
for spin-transfer. Concomitantly with the numerous literature devoted to
spin-Hall effects in metals and conductors, a mechanism of tunnel-Hall
effects in magnetic tunnel junctions (MTJs) [50, 51] and a mechanism of tun-
neling planar Hall effect (TPHE) emerging at ferromagnet FM/TI junctions
[52] have recently been proposed. Those qualitatively differ from the SHE
Tunneling Magnetoresistance, Spin-Transfer and Spinorbitronics 179
in terms of the relevant geometry, the forward/backward scattering in the

present case, and/or the magnetization configuration. In particular, the latter
effect is maximized for a planar magnetization parallel to the applied bias,
in the longitudinal geometry, where these other Hall effects vanish. This is
also the case for the U-SMR effects arising at heavy metal (SOI)/ferromag-
netic interfaces [53], semiconductor/ferromagnetic semiconductors [4] and
topological insulators/ferromagnetic interfaces [54]. In the latter case, the
non-linear resistance contribution vs. the direction of the transverse magne-
tization or in-plane current direction, as evidenced experimentally (U-SMR),
have been ascribed to carrier asymmetry-scattering (of the same symmetry
property) and to a second order conductivity component derived from it.
In that spirit, we have recently investigated theoretically the particular
properties of skew-tunneling of carriers at semiconductor interfaces or mag-
netic tunnel junction involving exchange and SOI in the host semiconductor
material [50, 51] like possibly played by (Ga,Mn)As or (Ge,Mn) ferromagnetic
injectors [55]. The prototype system is a simple interface or a tunnel junctions
made of the same ferromagnetic layers in an antiparallel (AP) magnetic con-
figuration. It may admit a difference in the carrier transmission (or diffusion)
upon their incidence, that is upon their parallel wave-vector ±k, vs. the reflec-
tion plane defined by the magnetization and the surface normal via SOI. This
new symmetry term in the electronic transport has an impact, even small,
on the conductivity once the Boltzmann diffusive equation is developed at
the second order of the electric field, and as recently experimentally observed
[54]. Moreover, such transport asymmetry may also be associated with a pure
interfacial (tunnel) topological transverse Hall current after summation of
the electron channels over the Fermi surface [50, 51]. This properties of car-
rier transport occurs in the CB involving SOI-Dresselhaus interactions or in
the VB getting benefit of the atomic SOI. In all cases, calculations go in favor
of a robust forward tunnel scattering asymmetry linked to a certain trans-
port chirality in the tunneling region. This specificity of spin-transport with
spin-orbit split electronic states may be of a significant importance to explain
the specific non-linear U-SMR effects, as revealed recently at (Ga,Mn)As/
GaAs:Mn interfaces with current in-plane geometry [4].
This chapter is divided into five sections: (1) In the first section, we will
present results of tunneling magnetoresistance (TMR) obtained on (Ga,Mn)
As/III–V/(Ga,Mn)As junctions [12, 56–59] studied hereafter for spin injec-
tion problem. TMR results on hybrid (Ga,Mn)As/Insulator(I)/transition
metal junctions will be also be briefly discussed [60–62]. In this section,
quantum size coherent tunneling phenomena and tunneling spectroscopy
features observed on resonant (Ga,Mn)As [63] and with III–V (In,Ga)As
quantum wells (QWs) resonant structures [64] will be addressed, to ques-
tion the coherent nature, or not, of transport in heterojunctions and related
systems. This is actually an important topic focused on finding the adequate
resonant position levels of (Ga,Mn)As QWs in order to determine the correct
band structure [31, 65]. We then tackle the role of different material proper-
ties, mainly hole filling and exchange energy, on the TMR properties and will
establish phase diagrams that correlate these parameters to the TMR. (2) The
second section will be devoted to tunneling anisotropic magnetoresistance

(TAMR) effects [66–71] including at least one (Ga,Mn)As electrode. TAMR
generally manifests itself by a significant variation of the tunneling current
with respect to the magnetization direction of the ferromagnetic electrode
(for a review on TAMR, the readers can refer to Matos-Abiague and Fabian
[72, 73]). In the present case, this physical phenomena originates from the
strong anisotropy of the hole fermi surfaces of the different (Ga,Mn)As sub-
bands involved in the tunneling transport of holes. (3) The third section will
be devoted to the description of the skew tunneling effect and related anom-
alous tunnel hall effect involving ferromagnetic semiconductors [50, 51, 55]
and investigated theoretically using analytical developments together with
advanced 30-band k ⋅ p tunneling methods. (4) In the fourth part, we will
describe spin transfer torque experiments [74–77] which have been dem-
onstrated in these systems [3, 59, 79–82] accounting for the conservation of
the angular momenta of holes carried by the current for perpendicular-to-
plane and in-plane current injection. (5) In the last part, we will describe the
skew tunneling and derive in detail the anomalous tunneling Hall effect. The
investigations will be first made via the relevant perturbation calculations of
the scattering process based on the Green’s function techniques developed
in the first order of the SOI parameter. We also demonstrate results of trans-
port asymmetry implemented with advanced 30-band k ⋅ p methods adapted
to the tunneling transport and using relevant wave function and wave cur-
rent matching conditions that one compares to more standard k ⋅ p codes
with excellent agreement.
In the appendix, we will develop the modeling via advanced k ⋅ p kinetic
exchange approach, the angular momentum transport in the valence band
of semiconductors heterostructures. Concerning (Ga,Mn)As this includes,
in a mean-field approach [29, 83], the average p-d interaction between spin
up and spin down holes in the VB of the semiconductor parametrized by
the spin-exchange parameter and added to the Kinetic or Kohn–Luttinger
Hamiltonian [84]. Within the envelope function approximation, the tunnel-
ing transport of holes in the VB of semiconductors will be described in detail
according to the procedure given by Petukhov et al. [85, 86] for matching hole
wave functions and derivatives at each interface. This approach constitutes
the generalization to magnetic systems of the boundary conditions for hole
envelope wave functions and tunneling current examined about 20 years ago
by Wessel and Altarelli for semiconducting heterojunctions [87, 88].
4.1 TUNNELING MAGNETORESISTANCE IN THE

VALENCE-BAND OF P-TYPE SEMICONDUCTORS
4.1.1 Introduction
The quantum mechanical tunnel effect is one of the oldest quantum phenom-
ena and still continues to enrich our understanding of many fields in phys-
ics. The generality of the tunneling phenomenon is such that every dedicated
textbook on quantum mechanics discusses tunneling through a potential
barrier, and the possibility for an electron to tunnel through such a barrier
4.1 Tunneling Magnetoresistance in the Valence-Band of p-Type Semiconductors 181
with a characteristic imaginary velocity and imaginary wave-vector. Despite

this long history and many experimental investigations, the tunneling process
of electrons (or holes) continues to amaze physicists in the solid-state com-
munity. For spin-dependent tunneling between two ferromagnetic materials,
first proposed and reported [89] in 1975, and subsequently observed, it has
taken two decades to be reliably demonstrated at room temperature [90]. The
large TMR effects possible in ferromagnet tunnel junctions and MTJs led to
the development of nonvolatile magnetic random access memories (MRAMs)
and a new generation of magnetic read-heads in hard disk-drives (HDD), for
example, see Chapters 12–14, Volume 3. For sufficiently thin barriers, typi-
cally a few nanometers of insulating layer, carriers possess a finite probability
to be detected on the other side of the potential barrier, due to the coupling
between the evanescent wave functions on each side of the barrier. With the
condition that the density of states (DOS) of the electrodes is spin polarized
in the case of ferromagnets, and in a simple picture of a tunneling current
proportional to the DOS within each spin channel, the transmission prob-
ability is spin dependent, that is, depends on the relative orientation of the two
magnetizations. This is the so-called TMR effect. Additionally, when dealing
with tunneling of holes in the valence band of semiconductors, one can expect
an influence of the large spin-orbit coupling on the transmission probability
through the barrier as well as a possible deflection of the characteristic car-
rier wave-vector leading to the so-called TAMR effects. Today, and as largely
discussed in this chapter, one of the implementation deals with the properties
of tunneling and transmission of carriers involving spin-orbit coupled states
beyond only the spin degree of freedom.
4.1.2 Overview of TMR Phenomena Including (Ga,Mn)As

The first generation of MTJs has integrated amorphous oxide as tunnel bar-
rier, and spin-dependent tunneling effects were mainly described in terms
of the phenomenological Jullière model [89]. This has begun to change with
the synthesis of high-quality epitaxial Fe/MgO/Fe MTJs [91, 92] and related
junctions showing up huge TMR ratio explained thanks to spin-symmetry
filtering of the electron wave function during their tunneling transfer [93,
94], see Chapters 11–14, Volume 2. From a fundamental point of view, the
ferromagnetic semiconductor (Ga,Mn)As provided, at about the same time,
new opportunities to study spin-polarized transport phenomena in semi-
conductors heterojunctions [12]. One of the interests in (Ga,Mn)As lies in
the wealth and varieties of its electronic valence band structure, location of
exchange interactions, and strong spin-orbit coupling. In this vein, advan-
tages of semiconductor tunnel junctions are fourfold: (1) III–V heterostruc-
tures can be epitaxially grown in a wide variety of tunnel devices with abrupt
interfaces and with atomically controlled layer thicknesses; (2) junctions can
be easily integrated with other III–V structures and devices; (3) many struc-
tural and band parameters are controllable, like barrier thickness and bar-
rier height, allowing the engineering of any band profile; and (4) one can
introduce quantum heterostructures much more easily than in any other
material system.
The first large spin-valve TMR ratio (>70% low temperature) observed
on III–V-based (Ga,Mn)As/AlAs/(Ga,Mn)As MTJs [12] were obtained by
Tanaka and Higo in 2001 (Figure 4.1). One of the first interesting proper-
ties of these epitaxially grown MTJs was the evidence of a rapid drop of
TMR with increasing the AlAs barrier thickness demonstrating a change
of the effective tunneling spin-polarization of holes with barrier thickness.
This was ascribed, in the frame of spin-orbit nearest neighbor sp 3 s * tight-
binding model of tunneling, to the conservation of the in-plane wave-vector
parallel to the interface during the tunneling transfer process due to transla-
tion invariance. Some theoretical investigations performed within the 6×6
multiband k ⋅ p theory emphasize the specific role of the spin-orbit coupling
within the barrier to possibly explain such a loss of TMR with the III–V bar-
rier thickness [95, 96]. However, other calculations performed by Krstajic
and Peeters gave opposite conclusions [97].
Beyond single tunnel junctions, TMR obtained on (Ga,Mn)As-based III–V
double barrier structures, (Ga,Mn)As/AlAs/GaAs/AlAs/(Ga,Mn)As [98]
FIGURE 4.1 (a) Schematic illustration of a wedge-type ferromagnetic semicon-

ductor trilayer heterostructure sample grown by LT-MBE. (b) Magnetization of a
Ga0.96Mn0.04As/AlAs 3-nm/Ga0.967Mn0.033As trilayer measured by SQUID at 8 K. (c)
TMR curves at 8 K of the trilayer 200 μm in diameter. Bold solid and dashed curves
were obtained by sweeping the magnetic field from positive to negative, and from
negative to positive, respectively. A minor loop is shown by a thin solid curve. In
both (b) and (c), the magnetic field was applied along the [100] axis in the plane.
(After Tanaka, M. et al., Phys. Rev. Lett. 87, 026602, 2001. With permission.)
and (Ga,Mn)As/AlAs/InGaAs/AlAs/(Ga,Mn)As [64] were integrating

(Ga,Mn)As/AlAs junctions as spin injector/detector devices. These devices
allowed to demonstrate spin injection phenomena into III–V materials in
the current perpendicular to plane (CPP) configuration. However, such
spin-valve effects observed on double barrier structures with in-plane mag-
netization still raise questions about the symmetry-state of the hole wave
function injected in the quantum QW whose natural quantization axis is
the growth direction. In the case of experiments of Mattana et al. [98], the
formation of hydrogenic impurity states inside the QW, Ga vacancies, for
instance, of quasi-spherical or cubic symmetry was suggested. These sym-
metry hole states of extended envelope wave function have been described
theoretically in the k ⋅ p framework by Baldereschi and Lipari in the early
seventies [99–101] and could give rise to resonant tunneling effects in the
quantum well [102]. Such resonant tunneling effects was already observed
in our group through As antisites deep levels in MnAs/GaAs/MnAs tunnel
junctions synthesized at low temperature [103, 104]. Nevertheless, the oscil-
latory character of the TMR highlighted by Ohya et al. [64] on structures of
higher epitaxial quality is more impressive with regard to the hole symme-
try-state. A part of the answer lies in the possible existence of two orthogo-
nal mixed-spin hole states on resonant levels in the QW. This conclusion
should be supported by future theoretical investigations. Several kinds of
III–V (Ga,Mn)As-based tunnel junctions were developed hereafter, includ-
ing thin GaAs layer [56, 57, 81, 82] or (In,Ga)As barriers [58, 80] specifically
grown for their low resistance area product with a view to proceeding to
spin transfer experiments. More surprising are the results of Xiang et al.
[105] about the observation of small in-plane magnetoresistance effects on
(Ga,Mn)As/highly Be-doped GaAs/(Ga,Mn)As trilayers [105] where GaAs
should play the role of barrier for holes.
In parallel to tunneling transport experiments, (Ga,Mn)As/GaAs Esaki
diodes were used to demonstrate spin injection by electrical means on lat-
eral structures with a non-local detection [106], and by optical techniques
[107, 108] giving more than 80% for the degree of helicity of the light emitted
from these spin-light emitting diodes (spin-LEDs). Calculations performed
in the frame of a spin-dependent tight-binding Hamiltonian, including the
exchange interactions within (Ga,Mn)As, were then employed to under-
stand such high spin injection efficiency [109]. On the other hand, Chun
et al. [60] demonstrated spin-polarized tunneling effects on hybrid MnAs/
AlAs/(Ga,Mn)As structures, as well as Saito et al. with hybrid Fe/ZnSe/
(Ga,Mn)As and Fe/GaO/(Ga,Mn)As junctions [61, 62]. This shows the pos-
sibility to observe large TMR ratio with bipolar n/i/p structures and con-
sequently to convert spin-polarized electrons into spin-polarized holes and
vice versa.
From a more fundamental point of view, could tunneling of holes
through a thin (Ga,Mn)As layer and through the whole III–V heterojunction
be realistically described by coherent processes over quite long layer thick-
ness, as was recently observed with tunneling spectroscopic methods [31,
63, 65]? Specifically, clear resonant tunneling effect were observed through
a relatively thin AlAs/(Ga,Mn)As/(Al,Ga)As resonant tunneling diode with

varying QW thickness ranging from 3.8 to 20 nm (Figure 4.2). These experi-
ments seems to support the picture of the coherent nature of spin-polar-
ized holes tunneling processes through relatively thin (Ga,Mn)As QWs,
despite the low-temperature growth procedure (250°C) often described to
be detrimental for the observation of such phenomena. Moreover, these
FIGURE 4.2 (a) Schematic device structure of the Ga0.95Mn0.05As 20-nm/GaAs

1-nm/Al0.5Ga0.5As 4-nm/GaAs 1-nm/Ga0.95Mn0.05As d-nm/GaAs 1-nm/AlAs 4-nm/
GaAs:Be 100-nm resonant tunneling diode (RTD) junction. (b) and (c) Schematic
band diagrams of the resonant tunneling diode (RTD) junction when the bias
polarity is negative and positive, respectively. (d) d2I/dV2-V characteristics of these
RTD junctions with various QW thicknesses d in parallel magnetization at 2.6 K.
Numbers in the parentheses express the magnification ratio for the vertical axis. (e)
dI/dV-V characteristics of the junction with d = 12 nm at 2.6 K in parallel (blue curve)
and antiparallel (red curve) magnetization. (After Tanaka, M. et al., Phys. Rev. Lett.
87, 026602, 2001. With permission.).
experiments are well explained by 4×4 valence-band k ⋅ p [84] model includ-

ing p-d exchange interactions, thus demonstrating the validity of the coher-
ent approach to describe tunneling transport of holes in a wide family of
(Ga,Mn)As/III–V heterostructures.
4.1.3 Material and Chemical Trends

4.1.3.1 Atomic Picture of Mn-Doped GaAs Host (See Appendix A)
According to optical studies, Mn in GaAs is known to form a shallow accep-
tor center characterized by a moderate binding energy [111] of 110 meV,
and a small magnitude of the energy difference of the order of 8 (±3) meV
between the states corresponding to the spin of the bound hole parallel and
antiparallel to the Mn spin [111]. This relative small atomic p-d exchange
energy originates from the long extend of the bound hole wave function
where the Mn d states are mostly localized. A direct consequence is that
the average exchange integral ∆ exc is expected to be enhanced with increas-
ing the Mn content x. The average exchange interaction in (Ga,Mn)As
5
reads ∆ exc = − xN 0β eV , where N 0 is the concentration of cations and
2
16  1 1 1 2 3 
N 0β = −  +  ×  ( pdσ) − ( pd π)2  < 0 is the exchange inte-

S −∆ + U ∆  3 9 
gral found by treating the p-d hybridization as a perturbation in the configu-
ration interaction picture [112], giving rise to antiferromagnetic interactions
between p and d shells (Figure 4.3). Here, S is the localized d spin, U is the
characteristic Coulomb interaction in the d shell, Δ is the characteristic
energy difference between p and majority spin d orbitals, whereas ( pdσ)
and ( pdπ) are characteristic Slater-Koster hopping integrals. The value of
N 0β = −1.2 eV (β = −54 meV nm 3 ) is generally accepted from core level pho-
toemission measurements for (Ga,Mn)As with a TC close to 60 K [113].
FIGURE 4.3 Schematic illustration of the long-range Coulomb and the two
short-range potentials each contributing 30 meV to the binding energy of the
(Ga,Mn) acceptor. (After Larsson, B.E. et al., Phys. Rev. B, 37, 4137, 1988. With
permission.)
4.1.3.2 Electronic and Ferromagnetic Properties

(Ga,Mn)As that we focus on in this chapter is the most extensively studied
III–V dilute magnetic semiconductor (DMS). Ferromagnetism in (Ga,Mn)
As requires Mn content of x  1% [114–116], whereas the metal-to-insula-
tor transition occurs at slightly higher content (x > 1%) [115–117]. Efforts
to achieve the growth of Mn-doped GaAs at high concentration (x > 1%)
by means of nonequilibrium molecular beam epitaxy (MBE) at low sub-
strate temperatures has led to the development of metallicity and fer-
romagnetism on (Ga,Mn)As with respective TC of 7.5 K and 60 K. More
recently, the TC have been pushed up towards 190–200 K [5–8] by improve-
ment in the MBE technique deposition, allowing Mn concentrations up to
nominally x  16% per cation site [1, 5–8, 118]. Using delta-doped doping
technique with Mn, i.e. mainly containing two-dimensional Mn-rich layer,
TC  250 K has been reached [119].
Owing to a relatively large distance between magnetic ions in DMS,
no direct exchange coupling between d-like orbitals localized on Mn ions
is expected. Thus, rather indirect coupling involving band states accounts
for spin-spin interactions. In this section we describe the effects of short-
range superexchange that dominates in the absence of carriers and com-
petes with long-range carrier-mediated interactions if the concentration
of band carriers is sufficiently high. The double-exchange mechanism con-
tributes when relevant magnetic ions are in two different charge states [28,
120–122], and if the system of transition metal electrons is on the metallic
side or in the vicinity of the Anderson–Mott transition. Double exchange
may indeed be considered as a strong coupling limit of the p-d Zener model,
but does not necessarily lead to the formation of an impurity band [123].
Within this model, TC attains a maximum if the impurity band is half-filled,
so that (in the case relevant here) the concentrations of Mn-2p and Mn-3p
ions are approximately equal. Ferromagnetic coupling mediated by bound
holes in the strongly localized regime was also predicted within a tight-bind-
ing approximation for a neutral or singly ionized pair of substitutional Mn
acceptors in GaAs, neglecting on-site Coulomb repulsion U and intrinsic
antiferromagnetic interaction between Mn-2p spins [124].
Zener noted the role of band carriers in promoting ferromagnetic order-
ing in the 1950s. This ordering can be viewed as driven by the lowering of
the carriers’ energy associated with their redistribution between spin sub-
bands, split by the sp-d exchange coupling to the localized spins in DMS. The
sign of this interaction between localized spins oscillates with the spin-spin
distance like the Ruderman–Kittel–Kasuya–Yosida (RKKY) interaction.
However, the Zener and RKKY models were found to be equivalent within
the mean-field approximation. These approximations are valid as long as
the period of RKKY oscillations R = π / k F is large compared to an average
distance between localized spins. Hence, the technically simpler mean-
field Zener approach is meaningful in the regime usually relevant to DMS
for a lower hole concentration. Owing to higher DOS and larger exchange
coupling to Mn spins, holes are considerably more efficient in mediating
spin-dependent interactions between localized spins in DMS. This hole-

mediated Zener–RKKY ferromagnetism is enhanced by exchange interac-
tions within the carrier liquid. Such interactions account for ferromagnetism
of metals (the Stoner mechanism) and contribute to the magnitude of the TC
in DMS. From the aforementioned arguments and mean-field theory, the
Curie Temperature of DMS, TC = TF − TAF may be written as:
TC = xeff N 0 S(S + 1)F χ cβ2 / (3k B ) − TAF (4.1)
where:
T
AF is the antiferromagnetic temperature due to superexchange
interactions
xeff is the effective Mn impurity concentration
N 0 in Equation 4.1 as defined before is the cation concentration
S is the spin quantum number of the local magnetization
F is an enhancement factor
β is the average exchange integral
χ c = mDOS
*
k F / (π 2 2 ) is the DOS for intraband spin excitations at the
Fermi level
For a detailed review, refer to Ref. [1]. Figure 4.4 displays the Curie tem-
perature TC vs hole concentration p normalized by the effective Mn density
(xeff and hole concentration p = N 0 xeff , respectively) for annealed (Ga,Mn)
As thin films, where squares and stars represent samples with hole density p
determined from a high field Hall effect and ion channeling measurements
[125], respectively. The solid curve shows the tight-binding theory [126].
FIGURE 4.4 Curie temperature TC vs hole concentration p normalized by the

effective Mn density (TC = x eff and hole concentration p = N0 x eff , respectively) for
annealed (Ga,Mn)As thin films, where squares and stars represent samples with
hole density p determined from a high field Hall effect and ion channeling mea-
surements [125], respectively. The solid curve shows the tight-binding theory [127].
(After Wang, M. et al., Phys. Rev. B 87, 121301, 2013. With permission.)
4.1.3.3 Impact on Tunneling Transport Properties

From a point of view of material properties, questions remain on the gen-
5
eral trends of TMR vs. exchange interactions, ∆ exc = − xN 0β, where xN 0 is
2
the concentration of Mn atoms as well as hole band filling within (Ga,Mn)
As. Also, what are the possible effects of the low-temperature (L-T) growth
procedure used for the synthesis of these MTJs on the band lineup, VB
offset and related barrier heights of heterostructures integrating (Ga,Mn)
As? Quite astonishingly and as shown by non-linear I–V characteristics
recorded on junctions, both (Ga,As) and (In,Ga)As materials play the role
of a tunnel barrier for holes injected from/into (Ga,Mn)As [56–58, 128]. The
same qualitative features have been demonstrated through optical measure-
ment of the hole chemical potential in ferromagnetic (Ga,Mn)As/GaAs het-
erostructures by photoexcited resonant tunneling [129]. These results go in
favor of a band-edge discontinuity due to the smaller band gap of (Ga,Mn)
As compared to GaAs [130] and indicates a pinning of the Fermi level deep
inside the band gap of the (Ga,As) host. A part of the answer lies in the
incorporation of n-type double-donor As antisites during the low-tempera-
ture growth procedure [27] that governs, by part, the pinning of the Fermi
level at a higher energy position than expected neighboring the midgap of
GaAs. The second reason is due to the positive coulombic-exchange poten-
tial experienced by holes and introduced by Mn species playing the role of
hydrogenoid centers for holes orbiting around it [26]. This is at the origin of
an impurity-band formation at smaller or intermediate doping level in the
host bandgap [131], and there is currently an intense debate about the posi-
tion of the Fermi level relatively to the impurity band [132]. While infrared
measurements [133], dichroism (MCD) [134], reflectivity [33], and tunneling
spectroscopy [31, 34, 65] experimental and theoretical [135] investigations
seem to support the scenario of a detached impurity band, recent low-tem-
perature conductivity [136] and angle-resolved photoemission spectroscopy
(ARPES) measurements [30], as well as k ⋅ p and tight-binding calculations
[112] validate the approach of a valence-band picture for (Ga,Mn)As more
compatible with a k ⋅ p treatment of its electronic properties. In the follow-
ing, we will then emphasize general phase diagrams established from our
k ⋅ p calculations and giving the TMR vs. the hole concentration or (Ga,Mn)
As fermi level the exchange energy ∆ exc (or equivalently, the spin splitting
parameter BG = ∆ exc / 6 introduced by Dietl et al. [29]) and also will compare
some experimental data to our model.
We are going to focus on our own experimental results. Figure 4.5a,
and b display TMR results acquired at low-bias (1 mV) and low-temperature
(3 K) on Ga0.93Mn0.07As(80 nm)/In0.25Ga0.75As (6 nm)/Ga0.93Mn0.07As (15 nm)
tunnel junctions. These are patterned in micronic [58] (Figure 4.5b) (stan-
dard UV lithography) and submicronic pillars [80] (Figure 4.5a) fabricated by
e-beam lithography for spin transfer experiments. Concerning the e-beam
lithography process, the fabrication procedure is discussed in section III,
which is devoted to spin transfer experiments. These samples exhibit 120%
(micronic) and 150% (submicronic) TMR ratios respectively after annealing,
FIGURE 4.5 (a) Tunneling magnetoresistance (TMR) measured on 700 nm × 200

nm nanopillar Ga0.93Mn0.07As(80 nm)/In0.25Ga0.75As (6 nm)/Ga0.93Mn0.07As (15 nm) at
1 mV and at 3 K. (b) TMR obtained on 128 μm2 junction patterned by standard UV
lithography on as-grown and annealed samples. (c) Current induced magnetiza-
tion reversal (CIMS) experiments on submicronic tunnel junction (a) at 3 K. (d)
Tunneling anisotropic magnetoresistance (TAMR) experiments measured at 3 K on
annealed sample (b). The reference for angle 0° is the direction along the growth
axis (m̂ [001]).
which are among the highest values found at this temperature with (Ga,Mn)
As-based MTJs [12, 56, 57]. The TMR is assumed to be larger for submicronic
junctions owing to a better homogeneity of the magnetization configuration.
One can notice that the TMR ratio is only 30% for as-grown micronic junc-
tions before any annealing treatment. This highlights a particular change of
the (Ga,Mn)As properties. Although the magnetic properties derive from
the whole magnetic layers (volume effect), whereas the tunneling process
is more sensitive to interfaces, it seems possible, however, to draw some
qualitative conclusions. A comparative study of the transport and magnetic
properties of these stacks have clearly demonstrated that the rise of TMR
observed after annealing conjugated to the decrease of the specific resis-
tance × area (RA) product from 5 × 10−2 to 3 × 10−3 Ω cm2 was correlated to
the change of the top (Ga,Mn)As layer properties [58]. In addition, the drop
of the RA product conjugated to a reduction of the coercive field (softening
of the magnetic film) observed on the top (Ga,Mn)As layer [58, 137–139]
unambiguously stem from an increase of the hole concentration in (Ga,Mn)
As due to the annealing. The consequences are: (1) a change of the Fermi
energy within (Ga,Mn)As; (2) a reduction of the (In,Ga)As barrier height;
and (3) an increase of the exchange energy ∆ exc within the top (Ga,Mn)As
5
layer. With a typical exchange integral of − xN 0β = −1.2 eV [113], the aver-
2
age spin-splitting in (Ga,Mn)As can be estimated in the mean field approach
A βMS
from the saturation magnetization according to ∆ exc = 6 BG = F [29]
gµB
where AF is the hole Fermi liquid parameter. Considering that only the top
(Ga,Mn)As layer is affected by thermal treatment, it is then possible to evalu-

ate the increase of the spin splitting BG parameter from 17 meV to 24 meV
[58] and then ∆ exc from 100 meV up to 140 meV for a characteristic TC of 60
K [58]. We will see in the following section that such range of values for ∆ exc
is also compatible with TAMR experiments performed on (Ga,Mn)As-based
resonant tunneling devices.
4.1.4 Experiments vs. 6-Band k · p Theory

of Tunneling Spin-Transport
We present in (Figure 4.6) our results of k ⋅ p calculations (details are given in
Appendix B) concerning specific (RA) products of Ga,Mn)As/In0.25Ga0.75As(6
nm)/(Ga,Mn)As junction vs. the valence band offset (VBO), d B, between
(Ga,Mn)As and GaAs or (In,Ga)As. The VBO determines the effective bar-
rier height φ according to d B = − ε F + φ (left inset of (Figure 4.6). The top of
paramagnetic (Ga,Mn)As valence band (∆ exc = 0) is taken here as the refer-
ence for energy, as well as for the Fermi energy. While tunneling experiments
estimate the barrier height between metallic (Ga,Mn)As and high-temper-
ature (H-T) grown GaAs in the range between 100 meV [71] and 90 meV
[56], photoemission spectra gave a barrier height φ of 0.45 eV for L-T grown
GaAs [140]. This is in agreement with our 6-band k ⋅ p model giving a RA
product  10 −3 Ω cm2 for annealed sampled for d B = −0.55 eV eV and with
experimental results of Chiba et al. [56]. Concerning (In,Ga)As, the band
offset between In0.25Ga0.75As and GaAs grown at high temperature is known
to be less than 50 meV for samples grown at optimized temperatures [141].
Nevertheless, this VBO has to be taken larger of the order of d B = −0.7 eV
FIGURE 4.6 Calculated Resistance × Area (RA) product of trilayer structures

vs. the band offset dB between the ferromagnetic semiconductor and a 6 nm
barrier of GaAs or (In,Ga)As. The physical parameters used for (Ga,Mn)As are
∆ exc = 6BG = 120 meV and n = 1 .5 × 1020 cm−3. Insets: (Bottom) RA product as a
function of the (In,Ga)As barrier width d. This gives a quasi-exponential varia-
tion of the RA product vs. the barrier thickness. (Top) Valence band profile of the
heterojunctions.
to match with our calculations. This determines the effective barrier height
at about 0.6 eV. The reason is that the real value of VBO (and barrier height)
is known to depend on (1) the growth temperature that affects the nature
and density of the dangling bonds at interfaces between two semiconduc-
tors [142]; and (2) the local density of ionized defects in the barrier (such
as As antisites). These generally lead to VB bending susceptible to modify
the average barrier height. The bottom inset of (Figure 4.6) gives the expo-
nential dependence of the (RA) product of (Ga,Mn)As/(In,Ga)As/(Ga,Mn)
As junctions ∝ exp(2κd ) vs. the (In,Ga)As barrier width corresponding to an
effective imaginary wave-vector κ InGaAs  1.15 nm −1 . Experimental results on
AlAs barriers with larger effective hole mass (section IV) give an attenuation
transmission factor of κ AlAs  2.9 nm −1 (κ is the evanescent wave in the bar-
rier corresponding to light holes (LH) in the barrier) [12].
Figure 4.7a displays in color code the calculated TMR of Ga,Mn)As/
In0.25Ga0.75As(6 nm)/(Ga,Mn)As junctions vs. the spin splitting parameter BG
and Fermi energy ε F of (Ga,Mn)As. BG represents the exchange splitting
between two consecutive hole bands at the Γ 8 symmetry point of the VB.
Also plotted on these phase diagrams are the energy of the four first bands
at the Γ 8 point, as well as three different iso hole concentration lines corre-
sponding to 1 × 1020 cm−3, 3.5 × 1020 cm−3 and 5 × 1020 cm−3 as a guide for the
readers. High TMR values, up to several hundred percents, can be expected
either for spin splitting parameter BG larger than several tens of meV or for
low carrier concentration, that is, when only the first subband is involved
in the tunneling transport. This corresponds to a quasi half-metallic char-
acter for (Ga,Mn)As. Starting from the first subband (n = 1) and increasing
the carrier concentration to fill the consecutive lower subbands (n = 2,3,4 ),
up and down spin populations start to mix up, leading to a significant drop of
TMR. For high carrier concentration (n = 4), small TMR is expected which
may anticipate difficulties to realize higher TC and higher TMR. We specify
that for low values of spin splitting and Fermi energy, ferromagnetic phase
induced by carrier delocalization may not exist (top right corner of the dia-
gram) which is, of course, not taken into account in our 6-band k ⋅ p model
using a basis of propagative envelope wave function. Comparing with our
experiments, our estimated value of BG gives a good qualitative agreement
for TMR as illustrated by the trajectory represented in Figure 4.7a between
point 1 (as-grown sample) and point 2 (after annealing). TMR ratio obtained
in as-grown sample (Figure 4.5b) are well reproduced for a hole concen-
tration in Ga0.93Mn0.07As approaching 1020 cm−3, in rather good agreement
with the one measured for single (Ga,Mn)As layer and already reported
[143]. From the literature, annealing procedures are known to remove Mn
interstitial atoms and increase hole concentration [27]. This should lead to a
reduction of the effective barrier height, even if the position of the topmost
(Ga,Mn)As valence bands is expected to move upper in the GaAs band gap
due to an increase of the exchange term ∆ exc . The strong reduction of the
RA product in parallel to a rise of TMR is consistent with such assump-
tion. Nevertheless, the hole concentration extracted here after annealing,
~1,7 × 1020 cm−3, appears to be slightly smaller compared to the value derived
FIGURE 4.7 Tunnel magnetoresistance TMR (a) and TAMR (b) vs. Fermi energy
and spin splitting for a 6 nm (In,Ga)As barrier and a band offset of dB = 700 meV.
White lines represent the four bands at the Γ 8 point. Gray lines indicate the Fermi
energy for different hole concentrations.
from Hall effect measurements. The existence of a possible concentration

gradient can be at the origin of such discrepancy [144–145]. A reduction of
the hole concentration at the interfaces with the barrier due to a significant
charge transfer between p-type (Ga,Mn)As and n-type (In,Ga)As (excess of
As antisites) can also be invoked [146, 147].
4.2 TUNNELING ANISOTROPIC
MAGNETORESISTANCE IN THE VALENCE
BAND OF P-TYPE SEMICONDUCTORS
4.2.1 Overview of TAMR Phenomena
In the absence of spin-orbit coupling, spin and orbital quantum numbers
are fully uncoupled. Consequently, no change of the tunnel conductance is
expected when the magnetization of the ferromagnet is rotated in space. This
4.2 Tunneling Anisotropic Magnetoresistance in the Valence Band of p-Type Semiconductors 193
assertion starts to change when the spin-orbit coupling is introduced. The

combination of spin-orbit coupling and exchange interactions in magnetic
tunnel junctions leads to TAMR effects. This can be understood as a variation
of the tunneling current vs. the magnetization direction of the ferromagnetic
electrode(s). TAMR phenomena can be divided into two different classes
according to whether (i) the magnetization always remains perpendicular to
the tunneling current leading to in-plane TAMR effects [66, 67]; or (ii) the
magnetization rotates from in-plane to out-of-plane directions e.g. parallel
to the tunneling current (out-of-plane TAMR) [58, 68]. The physical origin is
similar to that of anisotropic agnetoresistance (AMR) effects observed on a
single ferromagnetic layer except that in magnetic junctions, the transport of
spin-polarized carriers has a tunneling character. In that sense, TAMR can
be viewed as a natural extension of AMR in the tunneling regime of trans-
port. In principle, the latter geometry is more favorable to the observation of a
larger intrinsic TAMR signal, due to a high selectivity in k-space by tunneling
filtering. First discovered on (Ga,Mn)As-based III–V MTJs, TAMR has since
been demonstrated in transition metal-based nanodevices [148–151], mag-
netic tunnel contacts [152–153], and MTJs [154–156]. TAMR is now the basis
of theoretical proposals on resonant tunneling device [157]. For a review on
TAMR phenomena, readers can refer to Matos Abiague and Fabian [72, 73].
The strong spin-orbit coupling λ so acting in the valence-band of III–V
compounds is very favorable to the observation of TAMR phenomena.
Significant TAMR effects have been observed on (Ga,Mn)As-based tunnel
junctions in the respective in-plane [66, 158] and out-of-plane geometry [58,
68]. This also includes huge TAMR effects obtained in GaAs samples doped
with low Mn content [67] at the vicinity of the metal-to-insulator (MIT)
transition (Figure 4.8). This effect originates from the occurence of an Efros–
Shklovskii gap in (Ga,Mn)As at low bias and low temperature [30] controlled
by the extent of the hole wave function on the Mn acceptor atoms within
a certain depletion region [159]. Using k ⋅ p calculations, it has been shown
that this extent could depend crucially on the direction of magnetization in
(Ga,Mn)As. This was explained by the strong anisotropy of hole envelope
function around a single Mn in (Ga,As) in a low Mn doping regime where
anisotropic exchange interactions mediate a significant difference in the size
of hole wave function between r  m̂ and r ⊥ m̂. This argument holds also
in the reciprocal space ( k̂ ) as emphasized in the following section. In the
present case, this leads to a magnetization-dependent hole integral transfer
tuning the system from an insulating phase into a metallic state by changing
the magnetization direction. This anisotropy of the hole envelope wave func-
tion around a single Mn center calculated by an exchange-including tight-
binding method leading to huge TAMR effects near the MIT transition is
sketched in Figure 4.9 [160, 161].
An unconventional TAMR bias dependence, with sign reversal, has
been evidenced on (Ga,Mn)As/GaAs(n+) Esaki diodes [69]. This exhibits a
crossover between positive and negative TAMR signals. This particular fea-
ture was also observed in our group (Figure 4.9a) on (Ga,Mn)As/AlAs junc-
tions with opposite convention for the definition of TAMR giving rise to an
FIGURE 4.8 Amplification of the TAMR effect at low bias voltage and lox tempera-
tures. (a) TAMR along φ = 65° at 4.2 K for various bias voltages. (b) Very large TAMR
along φ = 95° at 1.7 K and 1 mV bias. (c), (c) ϕ at various bias at 1.7 K. (Top) Layer stack
under study. (After Ruster, C. et al., Phys. Rev. Lett. 94, 027203, 2005. With permission.)
opposite sign. The TAMR changes from positive values at positive bias to
negative values for negative bias, that is, when holes are injected from the
GaAs:Be electrode towards lowest hole energy subbands corresponding to
n = 4 (inset of (Figure 4.10)).
4.2.2 TAMR with Holes

Physical phenomena explaining TAMR with holes can be understood with
the following arguments:
1. The thick tunnel barriers (AlAs, GaAs, InGaAs) transmit easily any
LH-projection along the zone axis (or growth direction labeled by ẑ)
and switch off any HH-components because of their well larger effec-
tive mass.
2. As emphasized below, the exchange interactions mediate a strong
anisotropic LH↔ HH mixing. Since holes are mainly spin-polarized
FIGURE 4.9 Atomic structure close to a single Mn atom in substitional site in

GaAs host. As atoms are indicated by S1, S2, S3 and S4. Local spin density surfaces
for Mn magnetization along [001] (b), [110] (c) and [111] (d) showing the anisot-
ropy of the hole envelope wave function. (After Tang, J.M. et al., Phys. Rev. Lett. 92,
FIGURE 4.10 (a) Out-of-plane TAMR vs. bias acquired on a (Ga,Mn)As/AlAs(3

nm)/GaAs:Be tunnel junction. Positive biases correspond to hole injected from
(Ga,Mn)As towards GaAs:Be. (b) Angular variation of the resistance acquired at a
positive bias of +1 V (star) displaying a positive TAMR. The angle 0° corresponds to
m̂ [001] (growth axis). (c) Calculation of the corresponding TAMR with the same
parameters for (Ga,Mn)As as the one used for Figure 4.6. The injection in the nega-
tive bias regime implies an of holes towards at high energy towards the lowest
heavy hole band (triangle) thus giving rise to a negative TAMR.
along their propagation direction (at least in the limit of a large spin-
orbit coupling λ so  ε F ) the LH ↔ HH conversion is efficient for
wave-vectors k perpendicular to the magnetization m̂. This derives
from the introduction of non-diagonal exchange terms in the overall
Hamiltonian H h [29].
3. Such arguments explain the large positive TAMR associated to the
LH subbands near the Г point (k  zˆ ): LH-to-HH conversion along ẑ
is forbidden for out-of-plane magnetization (m ˆ  zˆ ) while it is possi-
ˆ  xˆ ). Inversely, the same argument
ble for in-plane magnetization (m
yields a large negative TAMR for subbands of HH symmetry near the
Г point.
Our calculations performed on (Ga,Mn)As/AlAs(3-nm)/(Ga,Mn)As

tunnel junctions seem to corroborate such conclusions. In Figure 4.10c, one
can observe a change of sign for TAMR on crossing the third subband, that
is, when the LH subbands become dominant in the tunneling transport (star
in Figure 4.10c). The first subband clearly gives a negative contribution to
TAMR, as does the fourth one, which have the effect of reversing the over-
all TAMR sign at higher energy (triangle in Figure 4.10c). This originates
from the predominant heavy hole character of these subbands, an in-plane
magnetization allowing, through off diagonal components, a possible heavy
to light hole conversion, and then a larger transmission through the bar-
rier [71]. This argument is reversed for the second and third subbands, with
the result that TAMR becomes positive when the n = 2 and n = 3 subbands
are dominant in the tunneling transport. This was theoretically established
through tight-binding treatment of TAMR [162]. Reducing the hole concen-
tration through hydrogenation technics should give the possibility of prob-
ing this possible crossover from positive to negative TAMR [163]. Coming
back to experiments on (Ga,Mn)As/(In,Ga)As/(Ga,Mn)As junctions, taking
into account conjugate TMR and TAMR values obtained before and after
annealing, one can roughly evaluate the projection of the corresponding sig-
nal trajectories in the [ε F , BG ] plane followed during annealing (Figure 4.7b).
A good qualitative agreement can be found even though symmetrical junc-
tions were simulated in order to restrict the number of parameters.
4.2.3 Tunneling Spectroscopy in the Valence

Bands of (Ga,Mn)As: Resonant TAMR
in Double Tunnel Junctions
In order to go a little further in understanding TAMR phenomena with
holes, we will now discuss results of resonant TAMR effects observed on
AlAs/GaAs/AlAs RTDs spin-filtered by a (Ga,Mn)As electrode [71]. In this
work, large TAMR oscillations correlated to the on/out-of resonance of
QW’s quantized states have been observed up to 60 K (inset of Figure 4.11a)
corresponding to the TC of (Ga,Mn)As-based RTDs consisting of 50-nm
(Ga0.94, Mn0.06) As/with a 5-nm undoped GaAs spacer/5-nm AlAs barrier/6-
nm GaAs QW/5-nm AlAs barrier/5-nm undoped GaAs spacer/50-nm Be
FIGURE 4.11 (a) TAMR–V signal (dash-dot red line) and dI/dV–V (dashed line)
acquired on the 6-nm GaAs QW. The left axis displays the measured TAMR; the right
one takes into account the up-renormalization (see text). Inset of (a) TAMR-T acquired
at 2.75 V (LH2 peak) and M-T. (b) TAMR–V calculated (straight red line) and transmis-
sion coefficient (unit non shown) calculated at an average energy ε h = −160 meV and
for respective small k (dotted blue line) and intermediate k (dashed black line). (c)
energy & band-resolved TAMR–V calculated for ε h = −160 meV. (After Elsen, M. et al.,
Phys. Rev. Lett. 99, 127203, 2007. With permission.)
doped (p = 1 × 1017 cm−3) GaAs/50-nm Be doped (p = 1 × 1017 cm−3→2 × 1019

cm−3) GaAs/p+ GaAs substrates, from top to bottom. The whole structure
was grown at an optimal temperature of 600°C, except the top (Ga,Mn)
As layer which was synthesized at 250°C to avoid the formation of MnAs
precipitates. Figure 4.10a displays resonant tunnel conductance (dI/dV–V
curves in logarithmic scale) and TAMR vs. bias acquired on the 6-nm QW
16 μm2 mesa under a saturation field of 6 kOe. Positive bias still corresponds
to holes injected from the (Ga,Mn)As layer into GaAs:Be. Note, however
that, (i) the differential conductivity is broadened near the resonances due
to the large hole density within the (Ga,Mn)As injector; and that (ii) the two
first HH1 and LH1 resonances are hardly visible. Comparing with previ-
ous results of Ohno et al. [164], a series of negative differential conductance
peaks is clearly evidenced with successive contributions of heavy (HHn) and
light hole (LHn) hole states. TAMR data have been obtained by plotting the
relative difference of the tunneling current, jT , measured respectively for
out-of plane ( jT[001]), and in-plane (Ga,Mn)As magnetization ( jT[100]) according
to TAMR =  jT[001] − jT[100]  / jT[100] ( ẑ = [001] is the growth direction).
For positive bias V > 0, we very interestingly observe positive TAMR
oscillations characterized by successive peaks and dips whose positions
are strongly correlated to the successive resonances. Note, however, that
the resonance appears at bias much larger than twice the quantized energy
of the discrete levels in the quantum wells. This discrepancy originates
from the introduction of a resistance RDZ in series with the double barrier
structure in the GaAs:Be electrode because of a gradient in the doping [71].

While the measured TAMR amplitude appears rather moderate (10% or
less), the overall intrinsic TAMR signal is much larger, by a factor of ≈ 5,
(right scale of Figure 4.11a) when one takes into account the presence of
such nonmagnetic serial resistance RDZ . Inset of Figure 4.11a displays the
temperature dependence of the TAMR peak acquired at 2.75 V, correspond-
ing to resonant tunneling into the LH2 state. The signal amounts to 7% at 4
K and decreases gradually with temperature to vanish at about 50 K close to
the (Ga,Mn)As Curie temperature (M-T curve). This demonstrates that the
oscillatory behavior of TAMR is induced by the exchange interaction within
(Ga,Mn)As layer. One can thus discard (1) any magneto-tunneling effects
deriving from the quantization of Landau levels on QW states, as well as (2)
deflection of the hole parallel wave-vector by Lorentz force in the plane of
the junction [165]. Moreover, both mechanisms would hardly appear at such
intermediate field (6 kOe). Our conclusion is that such an oscillatory TAMR
behavior reflects the spin-dependent transmission coefficients of the polar-
ized holes injected from (Ga,Mn)As, thus emphasizing the VB anisotropy of
the (Ga,Mn)As. Note, however, that this can include an extension of mecha-
nism (2) i.e a deviation of the hole wave-vector along the in-plane ŷ = [010]
direction normal to the in-plane x̂ = [100] magnetization driven by spin-orbit
interactions. We will come back to this very important point at the end of
this section.
On the basis on the aforementioned arguments on differential transmis-
sion of HH and LH-components, one can suggest that the oscillation charac-
ter of TAMR on/out-of resonance may result from the 4th heavy hole (Ga,Mn)
As subband contribution. To this end, we have plotted in Figure 4.11b the
transmission amplitude of holes of average energy ε h = −100 meV (−160 meV
from the top of the valence band) corresponding to p = 2 × 1020 cm−3 for two
different parallel wave-vectors k, respectively small k x = 0.1 nm−1 (dot curve)
and intermediate k x = 0.35 nm−1 (straight curve) across the double barrier
structure. ε h = −160 meV is then assumed to be the elastic energy transmit-
ted during the tunneling process. The resulting TAMR shape, displayed in
Figure 4.11c, is in excellent agreement with the experimental data exhibit-
ing the successive peaks and dips correlated with the successive positions
of resonances. In addition, the magnitude of TAMR is also well-reproduced
after renormalization by a factor of 5 (right scale of Figure 4.11a). We observe
that most of the TAMR peaks correspond to a specific resonance at inter-
mediate k (what we call out-of-resonance) whereas the TAMR dips match
generally with a specific resonance at small k (what we call on-resonance).
We then considered the respective contributions of the different holes sub-
bands to the TAMR. The four band resolved TAMR signals extracted for
ε h = −160 meV (still normalized by the total tunneling current for an in-
plane magnetization) are plotted in Figure 4.10c. The band indexes n = 1, 4
resp. n = 2, 3 stand for the HH resp. LH symmetries near the Г point. Quite
remarkably, the main contribution to the TAMR oscillations derives from
the first LH subband (n = 2) exhibiting the same trends as the summed signal
(Figure 4.11b) as well as the experimental data (Figure 4.11a). Let us focus on
the specific LH2 resonance (Figure 4.11c) where we hereafter designate V as
the effective bias applied on the double barrier structure. The second sub-
band contribution to TAMR is small, about 10%, on-resonance (V = 0.59 eV:
point A in Figure 4.11c, increases up to about 60% out-of-resonance for inter-
mediate k (V = 0.68 eV : point B) before vanishing at even larger bias and
larger k (V = 0.71 eV : point C).
The following demonstration is based on the calculations of the band-
selected tunnel transmissions mapped onto the different (Ga,Mn)As Fermi
surfaces as sketched in Figures 4.12 and 4.13. These figures display the dif-
ferent band-selected transmission for respective in-plane (Figure 4.12: m ˆ  xˆ )
and out-of-plane magnetization (Figure 4.13: m ˆ  zˆ ). The transmission coef-
åT
ˆ
k,m
ficients n¢ (eh ) mapped on the two-dimensional k Fermi surfaces for
n¢
the two magnetization configurations are plotted in Figures 4.12 and 4.13a–l
for the four subbands. In both figures, the indices a–d correspond to the
mapping of transmission calculated for a bias V = 0.59 eV, indices e–h for a
bias V = 0.68 eV and indices i–l for a bias V = 0.71 eV.
FIGURE 4.12 Band-selected ( n = 1− 4 ) tunnel transmissions (in color logarithmic scale) mapped onto the
different (Ga,Mn)As Fermi surfaces calculated for in-plane magnetization ( m ˆ  xˆ ). The different columns
corresponds to a specific selected band. Calculations have been performed for a bias applied to the double
barrier structure equal to V = 0.59 eV (a-d: first line), V = 0.68 eV ((e)–(h): second line) and V = 0.71 eV ((i)–(l): third
line) around the LH2 resonance.
FIGURE 4.13 Band-selected ( n = 1− 4 ) tunnel transmissions (in color logarithmic scale) mapped onto the
different (Ga,Mn)As Fermi surfaces calculated for in-plane magnetization ( m ˆ  xˆ ). The different columns
corresponds to a specific selected band. Calculations have been performed for a bias applied to the double
barrier structure equal to V = 0.59 eV ((a)–(d): first line), V = 0.68 eV ((e-h): second line) and V = 0.71 eV ((i-l): third
line) around the LH2 resonance.
1. On LH2 resonance (V = 0.59 eV ): point A) On Figures 4.11 and 4.12a–d,

the bright spot with a relative small radius at the center indicates
that the tunnel transmission is mainly driven by small k. The band-
resolved TAMRs obtained by k-integration approaches +125% (resp.
−38%) for n = 2 (resp. n = 4), whereas it gives −38% (resp. +3%) for n = 1,
(resp. n = 3). Correspondingly, the weighted tunneling currents cal-
culated for an in-plane magnetization are respectively 18% and 22%
for n = 2 and n = 4 (30% for both n = 1 and n = 3). The overall TAMR is
less than 10%, since the negative HH contribution (n = 1, 4) is partially
compensated by the positive LH one (n = 2).
2. Out-of-resonance at V = 0.68 eV : point B) The tunneling transmis-
sion in the reciprocal space is now mainly confined within a ring of
an intermediate radius from the zone axis e.g. driven by intermediate
k ≈ 0.35 nm −1 (Figures 4.12 and 4.13e–h. Due to the finite extension
of the minority spin HH hole Fermi surface (n = 4), the resonance
condition can be only partially achieved for this particular band: the
tunnel conductance falls off. Each weighted tunnel conductivity is
then 33%, 30%, 33% and 4% for n = 1−4 whereas the corresponding
TAMRs equals now −30%, +180%, 20% and −90%. The consequence
is an overall increase of the TAMR up to 60%, due to the deficit of

negative TAMR from the minority spin HH hole subband. The over-
all results are gathered in Table 4.1
3. At higher bias V = 0.71 eV : point C) The tunnel transmission is
driven by holes of even larger k (≈ 0.5 nm−1 ) away from the zone
axis ẑ (Figures 4.11 and 4.12i–l). It results in a strong LH↔HH
mixing along ẑ which, in turn, strongly reduces selected TAMR
down to −22%, +40%, 9% and −55% for n = 1 − 4 (Figure 4.11c). The
weighted tunnel conductivities are respectively 60%, 30%, 7% and
3%. As a result, the overall TAMR is low due to the strong reduction
of the second subband LH contribution (from +180% to +40%), as
shown in Figures 4.11j and 4.12j. The conclusion is that the oscilla-
tory behavior of TAMR originates from the two merging mecha-
nisms established at the beginning of the section and based on the
HH and LH character of each subbands. One can then argue that
acquiring a large k out-of-resonance naturally mixes the LH and
HH characters resulting in a significant TAMR drop ( from 180%
to 40%). Conversely, the argument (i) yields a large negative TAMR
for subbands of HH symmetry at intermediate or small k before
decreasing at larger k . The ensemble of band-selected intrinsic
TAMR and their corresponding contribution to the resonant tun-
neling current are reported in Table 4.2 for respective on and off-
resonant transmission on LH2 peak. The overall results are gathered
in Table 4.2.
In addition, the high-resolution of resonance peaks obtained at negative

bias V < 0 (not shown), that is, when holes are injected from GaAs:Be into
(Ga,Mn)As, allows the investigation of other superimposed effects related
TABLE 4.1
Band-Restricted TAMR and Contribution to the LH2 on-Resonance
Tunneling Current (V = 0.59 eV). The Kinetic Energy of Holes in
(Ga,Mn)As is Equal to 160 meV. (HH) and (LH) Indicates the Heavy or
Light Character for Holes Near the Г8 Point
n = 1 (HH) n = 2 (LH) n = 3 (LH) n = 4 (HH)
TAMR −38% +125% +3% −38%
Current 30% 18% 30% 22%
TABLE 4.2
Band-Restricted TAMR and Contribution to the LH2 off-Resonance
Tunneling Current (V = 0.68 eV). The Kinetic Energy of Holes in
(Ga,Mn)As is Equal to 160 meV. (HH) and (LH) Indicates the Heavy or
Light Character for Holes Near the Г8 Point
n = 1 (HH) n = 2 (LH) n = 3 (LH) n = 4 (HH)
TAMR −30% +180% +20% −90%
Current 33% 30% 33% 4%
to new TAMR phenomena. A significant resonant-TAMR signal has been

highlighted in asymmetric (Ga,Mn)As-based double barrier structure at
negative bias [166]. Such experiments exhibiting a magnetization-dependent
shift in bias of the consecutive HHn and LHn resonances seem to origi-
nate from a significant modulation of the Fermi level of (Ga,Mn)As imposed
by the anisotropic strain field. In that sense, this could be closely related
to Coulomb-blockade magnetoresistive (CB-TAMR) phenomena [167] in
the in-plane geometry, although in the latter case, a full understanding of
CB-TAMR is missing.
4.3 SKEW-TUNNELING AND ANOMALOUS

TUNNEL HALL EFFECT–UNIDIRECTIONAL
SPIN-HALL MAGNETORESISTANCE
WITH (GA,Mn)AS
In order to introduce the notion of skew-tunneling phenomena, note that
(Figure 4.12a–l) display in general a clear non-cubic asymmetry of the tun-
neling coefficients for in-plane magnetization (m ˆ  xˆ ). This results from the
action of the spin-orbit coupling on hole motion during their tunneling trans-
fer. This can be viewed as a generalization of the Lorentz force to magnetic sys-
tems by which a deviation of the hole wave-vector along the in-plane ŷ = [010]
direction normal to the x̂ = [100] magnetization vector is expected [64]. From

the general expression of the spin-orbit coupling so = − λ so (∇V (r ) × p) ⋅ S ,

it results in a potential discontinuity due to band offset (∇V (r ) ∝ z ) which

introduces a supplementary energy term so ∝ − λ so ( z × p) ⋅ S = − λ so (S × z ) ⋅ p

breaking the in-plane symmetry as soon as S admits a non-vanishing in-
plane component. This term is at the origin of the transmission asymmetry
between ±k y in-plane wave-vectors. In the envelope function approxima-
tion, this effect is implicitly included in the boundary conditions for the wave
functions and their derivative (current operator). Such an asymmetry in the
transmission coefficient is at the origin of what should be designated as tun-
neling (Spin) Hall effect (Figure 4.14).
On the hand, recent spin-Hall magnetoresistance measurements with
an unidirectional character (U-SMR) have been evidenced in two types
of bilayer systems either metallic bilayers Co/Pt involving Pt [53] and the
(Ga,Mn)As in (Ga,Mn)As/GaAs:Be systems [4] (see Figure 4.15). The same
kind of effect, i.e. a current-direction dependent contribution to the resis-
tance, was very recently reported on magnetic or nonmagnetic topological
insulator (TI) heterostructures, Crx ( Bi1− y Sby )2 − x Te3 / ( Bi1− y Sby )2Te3, effect
of several orders of magnitude larger than in other systems (Ref. [54] and
Figure 4.15). From the magnetic field and temperature dependence, the
UMR is identified to originate from the asymmetric scattering of electrons
by magnons [15(a)]. In particular, the large magnitude of UMR is an outcome
of spin-momentum locking and a small Fermi wave number at the surface
of TI. In particular, this unidirectional character of the magnetoresistance
manifests itself in a change of the resistance by switching the magnetization
4.3 Skew-Tunneling and Anomalous Tunnel Hall 203
FIGURE 4.14 (a) Schematic of the U-SMR phenomenon. Thin arrows represent
the SHE-induced spin polarization by non Ferromagnetic Mn-doped GaAs; thick
arrows represent the easy-axis (EA) magnetization of the (Ga,Mn)As ferromagnet.
(b) Schematic of the device and measurement geometry. (c) Longitudinal resis-
tance measurements at 130 K and different amplitudes and signs of the applied
current as a function of the external magnetic field. Steps correspond to the 180°
magnetization reversal. (d) Difference between non-linear resistance states vs.
current by switching to opposite transverse magnetizations, set by sweeping the
magnetic field from negative or positive values to the zero field, as a function of
the applied current. (After Olejnik, K. et al., Phys. Rev. B 91, 180402(R), 2015. With
permission.)
transverse to the current flow from one direction to the other. Equivalently,
this magnetoresistance contribution changes linearly with the injected
in-plane current and its sign follows the sign of current. It clearly reveals a
novel symmetry in the field of magnetoresistance.
This is explained by some chirality arguments incorporated in the equa-
tion of transport (Boltzmann). If there exists a non-symmetric probability
of magnon-diffusion for electrons or holes on the Fermi surface of the TI
depending on their incident wave-vectors (±k) along the current direction
perpendicular to the (transverse) magnetization, the resistance acquires a
supplementary non-linear contribution proportional to the current.
df (k , t )
Indeed, the Boltzmann equation, = 0, for an in-plane current
dt
(CIP geometry) derived at the second order of the electric field writes for
the respective fermi distribution f and its out-of-equilibrium part g (Fermi
surface displacement) as:
¶d(e - e F ) g (q)
d(e - e F )ev F E cos q + t(q)(ev F E )2 cos2 q = (4.2)
2¶e t(q)
with v F , the fermi velocity, and E x the longitudinal electric field (along
the x direction). This leads to an additional E 2 -dependence contribu-
tion to the current-distribution g once the momentum relaxation time τ
depends on the scattering angle along the respective forward and back-
ward directions (θ angle parameter). This angle dependence on θ is
FIGURE 4.15 (a) Schematic diagram of spin-momentum locking of the sur-

face Dirac state in TI and principle of magnon-asymmetry scattering process.
(b), (c) Schematic illustration of the concept for UMR in TI heterostructures
Crx (Bi1− y Sby )2− x Te3 / (Bi1− y Sby )2 Te3 (CBST = BST) on InP substrate under +J (b) and -J
(c) dc current. Here, magnetic field, magnetization, and dc current are along the
in-plane direction, where dc current is applied perpendicular to the magnetization
direction. (d) Schematic illustration of a “normal” CBST = BST heterostructure. (e)
Magnetic field dependence of resistance R xx for the sample depicted in (d), mea-
sured under J = 1 μA (red) and J=−1 μA (blue) at 2 K. (f) Difference of the resistance
∆R xx of plus and minus current shown in (e), (g)–(i). The same as (d)–(f) for the
“inverted” BST/CBST heterostructure.
responsible for a supplementary longitudinal conductivity term on the form

3 (1) µEδτ an.
∆σ(2) (2)
xx = 1 / ρ xx = σ where δτ an. is the anisotropic part of the scat-
4 vF τ
tering time between forward and backward electronic scattering and where
σ(1) is the linear conductivity. We will give some particular insights on the
properties of the anisotropic part of the scattering time δτ an further on in
the next chapter on the focus of the (Ga,Mn)As compound.
Such a carrier diffusion asymmetry on the Fermi surface may then lead
to U-SMR effects when one considers that SOI are present in one of the
bilayers, at least [50, 51]. We have recently investigated, by theory, this kind
of asymmetry phenomena in tunneling transport in the CB of III–V semi-
conductors contacted by a ferromagnetic (Ga,Mn)As layer [51]. The result is
quite spectacular, and in particular, in some selected points of the Brillouin
zone. In the case of the VB, one expects even larger asymmetry of transmis-
sion/reflection/diffusion in tunneling owing to the larger strength of the SOI.
This effect should also manifest itself by a surface anomalous tunneling Hall
effect when summed over the Fermi surface as evidenced in ferromagnetic
GeMn nanocolumn embedded in Ge [55]. The next part is devoted to the
description of tunneling scattering asymmetry (Figure 4.16) with holes in
the VB of (Ga,Mn)As-based III–V heterostructures constituting a new type
of spinorbitronics phenomenon. We will consider the most favorable case of
bilayer or tunnel junctions, that is, a given interface between the two mate-
rials, with same ferromagnetic contacts in the AP magnetic configuration.
We work in the current perpendicular-to-plane geometry for the transport.
4.3.1 Skew Transmission and Skew Tunneling

in the VB from Atomic SOI: A
Simplified Perturbative Approach
We will now prove, by analytical and numerical methods, that the asymme-
try of tunneling transmission and related anomalous tunnel Hall effect (or
skew tunneling) also occurs in the VB of semiconductor junctions involv-

ing core atomic SOI, H SO = (∇U × p) ⋅ σ . Quite astonishingly, the tun-
4m02c 2
neling scattering asymmetry may then observed in p-type orbitals without
invoking the assist of any odd-potential spin-orbit like played a Rashba or
Dresselhaus field. We demonstrate that such a chiral property in the trans-
port using (Ga,Mn)As [23, 29, 83] may be understood via a perturbative scat-
tering approach. This type of chirality phenomena in transport may explain
the giant Hall-effect observed in GeMn with a nanocolumn growth [55], and
FIGURE 4.16 Scheme of scattering (or transmission) asymmetry process at an

exchange-SOI interface of a bilayer with AP magnetizations M and − M along x.
The propagation direction of carriers (straight arrow) is along z with propagative
wave-vector k1 whereas the in-plane incident component k = + xi (heavy line) or
k = −ξ (dashed line) is along y; xyz forms a direct frame. The dash-dotted curve
denotes the evanescent waves, either reflected or transmitted. Carriers with a
k = +ξ in-plane wave-vector component are more easily transmitted than those
carrying −ξ . Top right inset: Energy profile of the exchange step; E is the longitudi-
nal kinetic energy along z and 2w is the exchange splitting in the magnetic materi-
als. (After Dang, T.H. et al., Phys. Rev., B 92, 060403(R), 2015. With permission.)
the recent U-SMR observed with (Ga,Mn)As [4] or with topological insula-
tors [54] via magnon-assisted asymmetric diffusion process. We define the
asymmetry of tunneling transmission, , for different in-plane component
of the incident wave-vector as:
Tk − T− k
= (4.3)
Tk + T− k
In the case of holes:
1.  appears without specific odd-symmetry potentials but combined

effects of pure atomic spin-orbit and exchange interactions through
the matching rules.
2. The VB brings a supplementary complexity in terms of LH-to-HH
band-mixing, each of them being doubly degenerated due to the spin
degree of freedom. LH and HH-subbands generally mix together at
oblique incidence upon reflection/transmission at interfaces.
3. Under some assumptions, a simple perturbative method may be
implemented in the VB along the guideline developed for the
CB. We consider a 6 × 6 Kane {X , Y , Z } ⊗ ↑, ↓ basis func- { }
tions. By rotation, this basis can be made equivalent to the 6 × 6
k ⋅ p Luttinger basis [29]. Here, one considers atomic SOI only in
the barrier (perturbation) and neglects SOI in the contacts. The
complete investigation including SOI in electrodes will reveal the
perfect agreement between the effect and size to expect with the
analytical approach.
The expression for the correction to the scattering (or transmission

coefficient) and its asymmetry after perturbation by the SOI VSO in a multi-
band picture (case of VB) is:
σ ′σ
∑∫ ,nm ( z ′)
d
VSO
Ψ(inm)0σ ( z ′) d ( k(n)z ′ ) ,
*
δt(σnm
′σ
)=  Ψ(out
n )0 σ ′
( z ′) (4.4)
l
0 2E
where (m), and (n) are the subscript corresponding to the multiband struc-
ture of the respective ingoing and outgoing waves and where  Ψ(out n )0 σ ′
( z ′)
generally admits a complex form through LH-HH band-mixing. However,
for relatively thick barriers, on can disregard the transmission of any HH
character. We will restrict our calculations on the LH-bands only. E is the
kinetic incoming kinetic energy.
Once the atomic SOI is introduced, VSO σ ′σ
= λ SO L ⋅ S = λ SO ( Lˆ z .Sˆz +
ˆL+ Sˆ− + Lˆ − Sˆ+ ) where L̂ and Ŝ are the respective orbital and spin operator, the
above expression becomes:
λ SO
∑∫
d
( z ′) [ Lˆ+ ]lm Ψ(inm)0↑ ( z ′) d ( k(n)z ′ ) ,
*
δt(↓↑
nm) =
 Ψ(out
n )0 ↓
(4.5)
2E 0

l
for an ingoing spin ↑ carrier (Ψ(inm)0↑ ). The result is that the tunneling trans-
mission is connected to the coupling between the (in) and (out) orbital
moments (L+ operator) in the tunnel barrier (chirality).
4.3.2 Transmission and Chirality within

the 6-Band k · p K ane Model
From Equation 4.5, we have calculated the tunneling transitions from the
ingoing LH state into the outgoing Z state. The result is [168]:
2
8 K LH  ∆ SO 
 exp ( −2λK LH d )
↓↑
δTLH →Z = 
λ 2
HH K HH (1 + λ LH )(1 + λ HH )  3 E 
2 2
(4.6)
2
 Mk 
× 1 +
 LK LH 
δTZ↓↑→ LH = δTLH
↓↑
→Z (4.7)
d is the barrier thickness, K LH,HH are the propagative wave-vectors in the

contacts for LH and HH and λK LH , HH are the corresponding evanescent
wave-vectors in the barrier for the incoming waves (in unit of K LH , HH ).
λ LH , HH are then unitless. The transmission is zero at the limit of an infi-
nite HH-mass (large K HH ) [168]. The limit of the zero-transmission has been
observed numerically in the case of 3 nm thick barrier in Figure 4.3 of Ref.
[51] of our previous paper in the limit of a small kinetic energy (and then
small HH-to-LH mixing). Nonetheless, the asymmetry , derived from Eqs
(4.7) still exists and is given by:
2λ LH K LH MLk
AZ → LH = (4.8)
(Mk ) + (Lλ K LH )
2 2
 LH
SOI-assisted tunneling transitions involve the coupling of the orbital

moment, as described by Equation 4.8, with a different probability and effi-
ciency for respective left and right incidence. (Figure 4.17) displays the asym-
metry  Equation 4.8 vs. the hole energy ε calculated analytically for two
different k wave-vectors, ξ = k = 10 −2 , 5 × 10 −3 nm −1 (black straight line). It
is compared with the results of numerical calculations (brown circle sym-
bols) implemented in a 6 × 6 k ⋅ p model in the limit of infinite HH-mass
(M − L = 2). Those are in perfect agreement which proves again the power
of the perturbative scattering method used here. From a vanishing value at
zero kinetic energy (K LH = 0),  gradually increases vs. energy when K LH
increases. The maximum is reached at a certain energy, just below the bar-
M
rier step, corresponding to λ LH K LH = k and then  decreases down to
L
FIGURE 4.17 Asymmetry of transmission  for holes vs. hole energy obtained
by perturbative scattering methods (black straight line) and numerical computa-
tions (brown circle symbols) with a 6-band tunneling code for the case of a p-type
magnetic ⇑ GaAs/3 nm GaAs barrier/p-type magnetic ⇓ GaAs. The barrier height is
equal to the exchange step of 0.3 eV. The calculation is performed for two different
hole incident ξ = k = 10 −2 ,5 × 10 −3 nm−1 in the limit of infinite HH-mass (M − L = 2).
Results corresponding to real Luttinger parameters for GaAs are also represented
(purple star symbols).
zero when λK LH becomes small (no evanescent states). The comparison with
the situation of real 6-band Luttinger parameters for GaAs is also displayed
in (Figure 4.17) (purple star symbol) with a good agreement as far as the
kinetic energy for holes is sufficiently small to sustain an evanescent wave in
the barrier. This also demonstrates that numerical computations are man-
datory to correctly describe the amplitude of  in some real physical situa-
tions, like e.g. the case of (Ga,Mn)As/GaAs/(Ga,Mn)As MTJs like described
throughout this review.
4.3.3 Transmission Asymmetry in the VB from

a 30-Band k · p Point of View
Figure 4.18 display the results of the two-dimensional (2-D) map calcula-
tions, in the (k = k x , k y ) reciprocal space of the 1st Brillouin zone, of the hole
transmission coefficient calculated for a p-type (Ga,Mn)As⇑/GaAs 3 nm/

(Ga,Mn)As⇓ magnetic tunnel junction of barrier height Φ = 0.3 eV. The hole
total energy is fixed at ε = −0.25 eV counted from the top of the VB. These cal-
culations are based on the multiband scattering-matrix technique developed
in Ref. [168], known to be more robust than the transfer-matrix method devel-
oped for the 6-band model [58, 85]. The results of the 30-band calculations on
the skew-tunneling effects are displayed in Figure 4.18d. We have checked that
the 14-band model provides similar data with P ′ = 0 and ∆ ′ = 0 with Figure
4.18c or without Figure 4.18b ghost-band treatment (see Appendix B) as well
as the 6-band Luttinger approach Figure 4.18a. The map of the transmission
coefficients are then very comparable in each case showing an equivalent
transmission coefficient mapping and corresponding transport asymmetry.
Note that the maximum scale of the hole transmission for the 14 band, 14
ghost-band, and 30 ghost-band (see Appendix B) are rather similar (15 × 10 −3 )
whereas the 6-band Luttinger model gives are slightly different (45 × 10 −3 )
because of the limit of validity of the effective Hamiltonian model.
Figure 4.19 displays the hole transmission asymmetry  vs. hole energy
−1
 in the case of a 3-nm-thick tunnel barrier for k = 0.05 nm . The agree-
ment is almost perfect between the different multiband k ⋅ p codes proving
again the relevance of our 30-band k ⋅ p tunneling treatment. The energy
range covers the valence spin subbands, namely, starting from the highest
FIGURE 4.18 Two-dimensional maps of the transmission coefficient T calculated

in the VB for a p-type magnetic tunnel junction in the AP state by respective 6 × 6,
14 × 14, 18 × 18 (14-ghost-band) and 30 × 30 ghost-band k × p methods. The param-
eters are: exchange energy 6w = 0 .3 eV, total kinetic energy  = 0 .25 eV counted
from the top of the VB; band parameters of the 14-band k × p model taken from
Ref. [169], and the one for the 30-band from Ref. [170]. The barrier thickness is 3 nm.
(After Dang, T.H. et al., J. Magn. Magn. Mater. 459, 37, 2018. With permission.)
FIGURE 4.19 (Bottom): Transmission asymmetry  vs. total energy  calcu-

lated in the VB for a p-type MTJ in the AP state by respective 6 × 6, 14 × 14, 18 × 18
(14-ghost-band) and 30 × 30 ghost-band k ⋅ p methods. The parameters are:
6w = 0 .3 eV, parallel wave-vector ξ = k = 0 .05 nm−1, barrier thickness d = 3 nm, and
barrier height 0 .3 eV. The energy zero corresponds to the nonmagnetic upper-
valence-band maximum.
energy, the up-spin heavy-hole band (HH ↑), the up-spin light-hole band
(LH ↑), the down-spin light-hole band (LH ↓), the down-spin heavy-hole
band (HH ↓). The up-spin split-off band (SO ↑) and the down-spin split-
off band (SO ↓) was considered in a previous contribution [51]. We refer to
points (1) to (4) marked by vertical arrows for discussing the contribution
from holes emitted from the different spin subbands in Region I to the cur-
rent injected in Region II. For instance, with these parameters, the energy of
the HH ↑ [HH ↓] maximum, corresponds to 0.15 eV [−0.15 eV], the energy
origin being taken at the top of the valence band of the nonmagnetic mate-
rial, and is indicated by point (1) [(4)].
Correspondingly, one observes an almost fully negative transmission
asymmetry in this energy range for predominant majority spin-up injection,
that is, as far as HH ↓ does not contribute to the current. At more negative
energy [ < − 0.15 eV: point (4)], a sign change of  occurs at the onset of
HH ↓ (in the upper left inset, see the step in the transmission coefficient,
which reaches almost +50%). Moreover, calculations give that the asymme-
try  remains positive after crossing SO ↑ before turning negative again
once crossing SO ↓ [51].  then changes sign twice at characteristic energy
points corresponding to a sign change of the injected particle spin. We
have performed the same kind of calculation for a simple contact [51]. It is
remarkable that , although smaller, keeps the same trends as for the 3-nm
tunnel junction, except for a change of sign, showing a subtle dependence of
the exchange coupling on the barrier thickness. Without tunnel junction, 
4.4 Spin-Transfer in the Valence Band of p-Type Semiconductors 211
abruptly disappears as soon as SO ↓ contributes to tunneling [circle region]

i.e., when evanescent states disappear. In the case of tunnel junction, ,
although small, persists in this energy range and this should be related to the
evanescent character of the tunneling wave function in the barrier.
4.4 SPIN-TRANSFER IN THE VALENCE BAND

OF P-TYPE SEMICONDUCTORS
4.4.1 Short Overview of Spin Transfer Phenomena with
Out-of-Plane and In-Plane Current Injection
The magnetic moment of a ferromagnetic body can be reversed or re-ori-
ented by transfer of the spin angular momentum carried by a spin-polarized
current. This concept of spin transfer has been introduced by Slonczewski
[74] and Berger [75], before being confirmed by extensive experiments on pil-
lar-shaped magnetic trilayers; for a review, the reader can refer to Chapters 7
and 8, Volume 1 [77]. Most current-induced magnetization switching (CIMS)
experiments have been performed on purely metallic trilayers [171–176],
such as Co/Cu/Co, in lithographically patterned nanopillars with detection
of the magnetic switching by giant magnetoresistance effects. Typical criti-
cal current density J C required for magnetization reversal in these systems
has been of the order of 107 A.cm−2 or higher. There have been several reports
on CIMS in transition metal MTJs with low junction resistance at critical
current density below 106 A.cm−2 [177–179]. CIMS experiments on tun-
nel junctions bring new physical problems [180], and are also of particular
interest for their promising application to the switching of the MTJ-MRAM
(Magnetic Random Access Memory). For a review of spin transfer in MgO-
based MTJs, the reader can refer to the paper of Katine and Fullerton [181]
and references therein. (Ga,Mn)As is known to have small magnetization of
less than 0.05 T and a high spin-polarization of holes which must result in
the reduction of critical current according to the Slonczewski spin-transfer
torque model. The results show that a low current density, of the order of 105
A.cm−2, is needed for CIMS with (Ga,Mn)As [79, 80] showing the interest of
magnetic semiconductors for spin transfer. As emphasized by Chiba et al.
[182], the specific VB structure and SOI has to be taken into account and
should have the effect of mixing the spin states of carriers. In the follow-
ing, we present results of CIMS experiments obtained on our own (Ga,Mn)
As MTJs whose structure and TMR properties are described in section II,
comparable to the one obtained in Ohno’s group [79]. We also discuss open
questions related to the VB electronic structure compared to the basis appli-
cable to metallic CPP-GMR structures.
On the other hand, considering the spin-orbit torques (SOT), the com-
bination of exchange and SOI in (Ga,Mn)As makes this materials very
important. Indeed, more generally, magnetization switching at the inter-
face between ferromagnets and SHE nonmagnetic materials controlled by
a current and related current-induced torques are of a particular interest. In
that sense, the size and symmetry of the SOI at the interface with relevant
materials with surface broken symmetry, like using (Ga,Mn)As (III,V),

deserves some clear investigations for potential future applications. With
that in mind, SOI current induced torques have been already demonstrated
with (Ga,Mn)As [3, 183] and more recently, the specific role of the antidamp-
ing Slonczewski-like torque in the ferromagnetic resonance regime of spin-
transfer. This particularly emphasizes the role of the two types, Rashba and
Dresselhaus, symmetry-like terms originating from the unidirectional char-
acter of the interface, together with the symmetry breaking from Td to C2v
symmetry group.
4.4.2 Spin Transfer with Holes in (Ga,Mn)As

Tunnel Devices
Concerning the fabrication of our nanoscale MTJs, a circular resist mask
(height = 200 nm, diameter = 700 nm) is first defined by e-beam lithography
and oxygen plasma etching. A pillar is then etched down to the conduct-
ing GaAs buffer layer by ion beam etching and a Si3N4 layer is sputtered to
cover the bottom electrode. The next step is the planarization of the surface
by spin-coating. The nitride layer on top of the pillar is then removed by
ion etching. Finally, the top of the pillar is cleaned with an oxygen plasma,
and top electrodes are fabricated by a liftoff process. For CIMS experi-
ments, the application of a significant bias of about 800 mV in our devices
is needed to observe spin transfer phenomena, so this switching cannot be
detected directly by a clear change of resistance. Instead we used the fol-
lowing procedure. Starting, for example, from a parallel (P) configuration
at V = 0, the voltage is increased step by step, and, after each step, brought
back to 20 mV to compare with that found at 20 mV before the step. We can
check in this way whether the magnetic configuration has been irrevers-
ibly switched. Of course, only irreversible switchings can be detected and
the reversible changes of the “steady precession regime” [77, 184] cannot be
detected. Examples of results obtained with this procedure are shown in
Figure 4.4c. By comparing with the MR curve at 20 mV of Figure 4.5a, one
sees that the magnetic configuration is switched irreversibly from an almost
parallel (P) to an almost antiparallel (AP) configuration by a positive cur-
rent density (current flowing from the thin magnetic layer to the thick one),
jc+ = 1.23 × 105 A cm − 2 (Vc+ = 810 mV) at 3 K and jc+ = 0.939 × 10 −5 A cm −2
(Vc+ = 680 mV) at 30 K. The configuration is then switched back to paral-
lel by a negative current above a threshold current density of the same
order ( jc − = −1.37 × 105 A cm −2 at 3 K and jc − = −0.986 × 105 A cm −2 at 30 K).
Opposite current directions for the P to AP and AP to P transitions is the
characteristic behavior of switching by spin transfer [77, 171–179]. Reversing
the initial orientation of the magnetizations does not reverse the sign of
the switching current, which confirms that Oersted field effects can be
ruled out. By more complex cycling procedures that would take too long to
describe, we have checked that, as in standard CIMS experiments, the tran-
sition is due to the magnetic switching of the thin layer. The fact that the
bottom thin layer is the free layer is also consistent with the spin-transfer
torque model as the total magnetic moment of the bottom layer is less than
that of the thicker top layer in (Ga,Mn)As as already reported. Further
experiments [80] have demonstrated that CIMS on (Ga,Mn)As-based MTJs
could be obtained only in a small field window of 2–3 Oe around 13 Oe
at 3 K, which may be explained by the combined effects of the Joule heat-
ing and temperature dependence of the magnetic properties. Indeed, from
SQUID and TMR measurements, we know that the anisotropy and coercive
field of the thin layer drops to a couple of Oe above 30 K, or, equivalently,
by heating the sample with a bias voltage value of about 700 mV. This study
has established a particular phase diagram (Figure 4.20) for CIMS vs. the
amplitude of the external magnetic field applied. This study has shown that
the heating of the sample and the manifestation of a reversible switching
regime [77, 172, 174, 176] (which cannot be detected by the experimental
protocol used) can prevent any switching of (Ga,Mn)As out-of a certain
narrow field window.
4.4.3 Phase Diagram for CIMS with (Ga,Mn)As

A significant difference with respect to the behavior with magnetic metals
is that CIMS curves similar to those of Figure 4.4c can be obtained only
in a small field window of 2–3 Oe around 13 Oe at 3 K. Let us first say
that, from the analysis of the shifts of minor MR cycles, we found that the
dipolar field generated by the thick (Ga,Mn)As layer and acting on the thin
layer is close to 13 Oe at 3 K, so that Happ = 13 Oe corresponds to approxi-
mately an effective field H eff = 0 Oe on the thin layer when the magnetic
moment of the thick one is in its former direction (case of Figure 4.20c).
The behavior of Figure 4.20c in a field range of a couple of Oe around
H eff = 0 Oe can be explained by the combined effects of the Joule heating
and temperature dependence of the magnetic properties. From SQUID and
FIGURE 4.20 Phase diagram displaying the different regime of magnetization

reversal vs. applied field that are thermal switching, irreversible switching (CIMS),
and reversible switching. The top abscise corresponds to the effective field equal
to the applied field added to the stray field equaling −13 Oe.
TMR measurements, we know that the anisotropy and coercive field of

the thin layer drops to a couple of Oe above 30 K or, equivalently, by heat-
ing the sample with a bias voltage value of about 700 mV [58]. Suppose,
for example, a P state with both magnetizations in the direction of posi-
tive H with H app = 9 Oe or, equivalently, H eff ≅ −4 Oe. If, at a current I*
smaller than the critical current for CIMS, the heating of the sample
reduces the magnitude of the coercive field of the thin layer to 4 Oe or
below, the effective field of −4 Oe switches the configuration to AP when
the current reaches I* or −I*. This behavior, with switching to AP by posi-
tive as well as by negative currents, is what we observe for Happ below the
window centered on H app = 13 Oe. Assume now a same parallel (P) state
with H app = 17 Oe or, equivalently H eff = +4 Oe. If the heating of the sample
reduces the anisotropy field below 4 Oe, we enter the regime of revers-
ible switching [77, 78, 172, 174, 176] that we cannot detect. Consistently,
we cannot detect any switching for H app above the window. We conclude
that, with respect to CIMS with magnetic metals, the dramatic variation of
the properties of (Ga,Mn)As with temperature, combined with the heating
of the sample, introduces significant complications in CIMS experiments
with (Ga,Mn)As and limits the observation of CIMS by spin transfer to a
narrow field range. These arguments give rise to the phase diagram shown
in Figure 4.20 and highlight the three different regions that are thermal
switching, irreversible switching (CIMS), and reversible switching. We
conclude that, with respect to CIMS with magnetic metals, the dramatic
variation of the properties of (Ga,Mn)As with temperature, combined with
the heating of the sample, introduces significant complications in CIMS
experiments with (Ga,Mn)As and limits the observation of CIMS by spin
transfer to a narrow field range.
The spin-transfer and spin-torques in MTJ can be described by the sum
of two different components: the so-called Slonczewski-like, or in-plane
torque, torque  IP , and the field-like, or out-of-plane torque  OOP . Contrary
to spin-transfer phenomena in metallic nanopillars, whereby the field-like
torque is generally smaller than the Slonczewski or antidamping-like torque
[185], the two components are of the same order of magnitude in tunnel
barriers because of the coherent nature of the spin-currents within the tun-
nel barriers [186–189]. However, these two types of torques differs by their
specific symmetry with the voltage variations according to:
 IP =  IP ,0 m × (s × m )  AIP ,0V m × (s × m ) (4.9)
 OOP =  OOP,0 m × s  AOOP,0V 2 (s × m ) (4.10)
The difference in the symmetry between the two components makes it pos-
sible to distinguish them experimentally in spin-diode experiments [190,
191] or via spin-transfer-torque spin-Hall effects (STT-SHE) experiments
[46, 47]. The result is that, under the action of these two torques, the general
magnetization dynamics of the local magnetization m follows the general-
ized Landau-Lifshitz equation according to [74, 78]:
∂m ∂m
= − γm × H + OOP ,0 m × s +  IP ,0 m × (s × m ) + αm × , (4.11)
∂t ∂t
where:
γ is the gyromagnetic factor
α is the phenomenological damping constant
Figure 4.21 displays our latest measurements on STT acquired on

400 × 200 nm2 (Ga,Mn)As (10 nm)/GaAs(6 nm)/(Ga,Mn)As(50 nm) tunnel
nanojunctions at 10 K prepared for STT experiments in current-perpendic-
ular-to-plane geometry. In these experiments, we have acquired the R(V)
switching cycles for different protocols (after positive and negative satura-
tion) and at fixed magnetic field. For example, let us comment first on the
first set of sub-figures at the top. The upper left figures corresponds to the
characteristic R(H) cycles after preparing the samples to positive saturation
(| H sat |= 7 KOe ) (negative saturation), then decreasing to zero field and apply-
ing a positive (negative) magnetic field, before applying the bias for describ-
ing the switching cycles. The R( H ) curves means that a positive moderate
bias (less than 2 V) has for effect to apply a reversible in-plane torque, via
 IP , linear in V switching the soft (Ga,Mn)As magnetization in its opposite
direction AP state. A larger positive bias (larger than 2 V) will have the effect
of switching back this magnetization in its formal direction though  OOP
(thanks to the field-like torque symmetry). A perfect symmetry exists when
the sample is prepared starting from the negative saturation (negative field
range of the corresponding sub-figures). The upper right figures corresponds
FIGURE 4.21 Phase Diagram for spin-transfer-torque (STT) experiments acquired on 400 × 200 nm2 (Ga,Mn)
As(10 nm)/GaAs(6 nm)/(Ga,Mn)As(50nm) tunnel nanojunctions. The current is injected in the perpendicular
direction for these STT-experiments. The TC of the trilayer structure is 150 K and the measurements are per-
formed at 10 K.
to the same protocol experiments, except that now the sample is prepared in
its relative AP state (by operating a minor loop) by the following preparation
field-cycle (± H sat → ∓100Oe → H where H is the fixed magnetic field while
swiping in bias) before applying the bias. The critical value of the field ±40
Oe from the main difference between the two types of cycles (critical point)
corresponds to the value of the antiparallel dipolar stray field from the thick
hard (Ga,Mn)As layer to the thinner soft (Ga,Mn)As layer.
More interestingly, the bottom panels of Figure 4.21 emphasize the role
of  OOP at larger bias in the reorientation of the hard (Ga,Mn)As layer along
the softer magnetization (at higher bias) to make the two magnets parallel in
any case. In other words, the ensemble of these figures demonstrates that the
external-field necessary to switch the hard (Ga,Mn)As layer along the softer
magnetization (that has already switched) in the P (parallel) state is reduced
due to the V2-dependence of the  OOP torques. The V 2 shape of the phase
diagram in the bottom part of the corresponding figures seems to reveal the
V 2 dependence of the field-like torque like discussed above.
4.4.4 Discussion
The order of magnitude of the switching current densities for MTJs
(~105A.cm−2) is consistent with the model of Slonczewski [74], when one
takes into account the magnitude of the (Ga,Mn)As magnetization, well
reduced compared to transition metals. This argument was already put for-
ward by Chiba et al. [79, 182]. In the simplest approximation, the switch-
ing currents at zero (or low) field are expected to be given by [173, 174]:
I P ( AP ) = G P ( AP )αtM[2πM + H K ), where G P ( AP ) is a coefficient depending on
the current spin polarization, α is the Gilbert damping coefficient, t is the
layer thickness, and H K is the anisotropy field. The main difference with
transition metals originates from the factor tM[2πM + H K ]. With t = 15 nm,
M = 0.035 T , and a negligible H K after heating, for the junctions of this
paper, and t = 2.5 nm, M = 1.78 T, Hk = 0.02 T for a typical Co/Cu/Co pil-
lar [171–173] we find that the factor tM[2πM + H K ] is larger by a factor of
500 for the metallic structure. This mainly explains the difference by two
orders of magnitude. Additional factors should come from the current spin
polarization (probably higher with (Ga,Mn)As) and the Gilbert coefficient
(probably larger by almost an order of magnitude for (Ga,Mn)As grown at
low temperature) [192], but a quantitative prediction is still a challenging
task. What about the sign of the switching currents in (Ga,Mn)As devices
which is the same as for standard metallic pillars [77, 171, 173, 174]? This is
consistent with what we expect from the valence bands structure of (Ga,Mn)
As, whereby the majority of electron spins (and no hole p state) at the Fermi
level are aligned parallel to the local Mn spins (d state) due to the nega-
tive p-d exchange interaction parametrized by the exchange integral β [29,
81]. Complementary experiments [81, 82] performed on double (Ga,Mn)
As-based MTJs in ms-pulsed current injection regime have demonstrated
the possibility of reducing the critical current densities to less than 3 × 104
A.cm−2, that is, by more than a factor of three compared to single (Ga,Mn)As
4.5 Summary 217
MTJs. This is due, in part, to the interfacial torque applied at both side of the
middle (Ga,Mn)As sensitive layer and imposed by the symmetric structure.
A question lies in the observation of switching phenomena by spin transfer
at bias (≈ 800 mV) corresponding to a quasi-vanishing TMR, and then to a loss
of the spin-polarization of holes with bias. Although in standard MTJs with
metallic electrodes, the decrease of the MR with the bias is generally ascribed
to electron-magnon scattering (emission and annihilation of magnons) [193]
and to other types of spin-dependent inelastic scatterings [194], calculations
performed by Sankowski et al. [162] put forward a natural decrease of TMR
with bias due to the band structure of (Ga,Mn)As. In the former case, even if
the spin current is strongly affected by spin-flip effects, the transverse compo-
nent of the spin-polarized tunneling current cannot be lost and should finally
be transferred to the magnetic moment of the layer, as proposed by Levy and
Fert [195]. On the other hand, concerning the band structure itself, recent the-
oretical investigations on torques and TMR in MTJs performed in the frame of
Keldysh formalism have demonstrated the possibility of conciliating increase
of spin transfer efficiency and decrease of TMR in a same device [187, 188].
The same kind of calculation remains to be investigated in a multiband for-
malism adapted to VB of semiconducting tunnel junctions, and, in particu-
lar, the role of respective Slonczewski and field contributions of the torque in
(Ga,Mn)As-based MTJs. The efficiency of the current induced torque exerted
by general angular momenta (J) and no more spin (S), including the role of
spin-orbit interactions as detailed in the following Appendix D, also remains
to be studied theoretically. However, a deeper description of spin-transfer phe-
nomena and spin-orbit torque with (Ga,Mn)As in both in-plane and current
perpendicular-to-plane geometry will require the extension of the generalized
spin-mixing conductance involving spin-orbit currents.
4.5 SUMMARY
To conclude, we have reviewed different properties of spin-polarized tun-
neling transport involving the (Ga,Mn)As DMS. It has been shown that a
relatively good agreement can be found between experimental TMR and
TAMR properties and Dietl theory for DMS in a molecular field approach
for (Ga,Mn)As ferromagnetism. The general description of spin-polarized
tunnel transport, including (Ga,Mn)As is fully described using standard k ⋅ p
theory and Laudauer Buttiker formalism. In particular, it has been possible
to understand the large modification of TMR properties as a function of
annealing procedure. Indeed, it is now commonly accepted that an optimized
annealing procedure, by the reduction of As interstitial at the free interface,
leads in parallel to an increase of the carrier density and an increase of the
spin exchange interaction. Further experiments exploring a wider region of
the phase diagram established theoretically are required to refine our under-
standing of the (Ga,Mn)As band structure. A more advanced technological
control of the material will be needed to extract precisely (Ga,Mn)As char-
acteristic parameters. More generally, our k ⋅ p framework seems to catch
the essential trends of the spin dependent tunneling phenomena involving
holes emitted from (Ga,Mn)As. As an example, the sizeable TAMR effects

originating from the anisotropy of the electronic band structure have been
finely reproduced. Exploring the TAMR signal on a resonant quantum well
has revealed the intricate competition of the different (Ga,Mn)As subbands.
The results indicate that a respective positive and negative contribution can
be expected from LH and HH bands. More generally, these different con-
tributions emphasize the band curvature of (Ga,Mn)As with HH and LH
characters near the Г point, assuming that the LH character is favorably
transmitted by the tunnel barrier.
In the last part we have demonstrated CIMS phenomena in nanoscale
MTJs integrating (Ga,Mn)As. Thanks to the particular magnetic properties
of (Ga,Mn)As, a smaller current density compared to transition metal is
required to reverse the magnetization. In the future, a deep understanding of
spin transfer phenomena involving holes will imply the development of more
sophisticated models, like the one developed in the standard k ⋅ p approach.
In parallel, time-resolved, or frequency-resolved, experiments performed in
the reversible regime of spin transfer will be certainly helpful for a com-
plete description of the inner mechanisms for spin transfer. Moreover, we
believe that the specificity of carrier-mediated ferromagnetism of (Ga,Mn)
As will give new opportunities to explore spin-dependent phenomena, tak-
ing advantage of semiconductor functionalities. Despite the difficulties in
the route for room temperature operations for this class of material, further
spin-related studies with ferromagnetic semiconducting heterojunctions
seem necessary to pursue fundamental physics, as well as to establish prom-
ising concepts in other systems.
Moreover, we have reported on theoretical investigations of a new type
of chiral-tunnel effects in semiconductors junctions and heterostructures
mediated by spin-orbit and exchange interactions. Our developments are
primary based on Green’s function formalism involved in the tunneling
transport at the first order of the perturbation theory. The analytical cal-
culations, performed for electrons in the CB with Dresselhaus interactions
and holes in the VB with core atomic spin-orbit interactions are favorably
compared with 6-band and advanced 30-band k ⋅ p tunneling models. This
study represents an extension of our previous contribution [51] dealing with
the role, on the electronic forward and backward transmission-reflection
asymmetry. Such forward scattering asymmetry also should lead to skew-
tunneling effects involving the branching of evanescent states in the barrier.
Recent experiments involving non-linear resistance variations vs. the trans-
verse magnetization direction or current direction in the in-plane current
geometry may be invoked by the phenomenon we discuss.
ACKNOWLEDGMENTS
J. Jaffrès gives special thanks to the students of the École Polytechnique tak-
ing part to the research laboratory project (P.R.L.), J. Boust, M. Dujany and
M. Le Dantec. T.H.D. acknowledges Idex Paris-Saclay and Triangle de la
Physique for funding.
Appendix A: Exchange Interactions in (Ga,Mn)As 219
APPENDIX A: EXCHANGE INTERACTIONS

IN (GA,Mn)AS: FROM THE Mn ISOLATED
ATOM TO THE METALLIC PHASE
From general group-theory arguments, in the effective mass approximation
[99, 101], nonmagnetic shallow acceptors can be described by hydrogenic states
of fundamental symmetry term 1S3/2 of binding energy equal to 28 meV for
GaAs. These are characterized by a total angular momentum F = L + J = 3 / 2 ,
which is a constant of motion where L is the angular momentum of the enve-
lope wave function. The result is that the fundamental 1S3/2 wave function
is Φ(S3/2 ) = f 0 (r ) L = 0, J = 3 / 2, F = 3 / 2, Fz + g 0 (r ) L = 2, J = 3 / 2, F = 3 / 2, Fz .
According to optical studies, Mn is known to form a shallow acceptor center
in GaAs of A0 = d 5 + h electronic configuration characterized by a binding
energy [110] of 110 meV, and an energy difference of the order of 10 (±3) meV
between the J = 1 and J = 2 h–d5 states [110]. In this picture, the J = S + j
quantum number constant of motion is the sum of the d5 Mn spin angular
momentum S = 5 / 2 and the j = 3 / 2 hole angular momentum [196, 197]. In
the S − j exchange coupling scheme where the exchange interaction is J exc S ⋅ j,
the energy difference between the extrema J = 1 and J = 4 states is equal to
9 J exc , whereas it gives 2 J exc between the two successive J = 1 and J = 2 states.
It follows that the energy difference between the states corresponding to the
spin of the bound hole parallel and antiparallel to the Mn spin can be esti-
mated to 45 meV. This relative small p-d exchange energy originates from
the relatively long extent of the bound hole wave function where the Mn d
states are mostly localized within an effective bohr radius a0* ∝ ε / m*  0.8
nm and corresponding to an effective volume of 3 nm3, as well as an effective
Mn concentration approaching xloc = 1.35% . A direct consequence is that the
average exchange integral ∆ exc is expected to be enhanced with increasing
the Mn content x above this threshold Mn concentration xloc = 1.35% .
In the metallic regime and in the S-s exchange coupling
scheme, the average exchange interaction in (Ga,Mn)As reads
∆ exc = −5 / 2 xN 0β, N 0 is the concentration of cations and N 0β = −(16 / S )
2
æ 1 1 ö æ1 2 3 ö
ç -D + U + D ÷ ´ çç 3 ( pds) - 9 ( pd p) ÷÷ < 0 is the exchange integral
è eff eff eff ø è ø
found by treating the p-d hybridization as a perturbation in the con-
figuration interaction picture [111] giving rise to antiferromagnetic
interactions between p and d shells. Here, S is the localized d spin,
( )
U eff = E(d n −1 ) + E d n +1 − 2 E(d n ) is the characteristic 3d-3d Coulomb inter-
( )
action, ∆ eff = E( Ld n ) − E d n −1 is the ligand-to-3d charge transfer energy. On
the other hand, ( pdσ) and ( pdπ) are characteristic Slater–Koster hopping
integrals. The value of N 0β = −1.2 eV (β = −54 meV nm 3 ) is generally admit-
ted from core level photoemission measurements for (Ga,Mn)As with a TC
close to 60 K [113] corresponding to xeff  4%. (Figure 4.13) display the 4-dif-
ferent exchange-splitted (Ga,Mn)As subbands calculated for a hole density
p = 1.7 × 1020 cm −3 and an average exchange energy of ∆ exc = 120 meV .
From a point of view of material properties, questions remain on the

general trends of TMR vs. exchange interactions ∆ exc = −5 / 2 xN 0β where
xN 0 is the concentration of Mn atoms as well as hole band filling within
(Ga,Mn)As. Also, what are the possible effects of the low-temperature (L-T)
growth procedure used for the synthesis of these MTJs on the band lineup,
valence band offset, and related barrier heights of heterostructures integrat-
ing (Ga,Mn)As? Remarkably, and as shown by non-linear I–V characteristics
recorded on junctions, both (Ga,As) and (In,Ga)As materials play the role of
a tunnel barrier for holes injected from/into (Ga,Mn)As [56–58, 128]. The
same qualitative features have been demonstrated through optical measure-
ment of the hole chemical potential in ferromagnetic (Ga,Mn)As/GaAs het-
erostructures by photoexcited resonant tunneling [129]. These results favor
a band-edge discontinuity due to the smaller band gap of (Ga,Mn)As com-
pared to GaAs [130], and indicate a pinning of the Fermi level deep inside
the band gap of the (Ga,As) host. A part of the answer lies in the incorpora-
tion of n-type double-donor As antisites during the L-T growth procedure
[27] that partly governs the pinning of the Fermi level at a higher energy
position than expected neighboring the midgap of GaAs. The second reason
is due to the positive coulombic-exchange potential experienced by holes,
and introduced by Mn species playing the role of hydrogen centers for holes
orbiting around it [26]. This is at the origin of an impurity-band formation
at smaller or intermediate doping level in the host bandgap [131], and an
intense debate remains about the position of the Fermi level relative to the
impurity band [132]. While infrared measurements [133] and dichroism
(MCD) [134] experimental and theoretical studies [135] seem to support the
scenario of a detached impurity band, recent low-temperature conductivity
measurements [136] validate the approach of a VB picture more compatible
with a k ⋅ p treatment of its electronic properties.
APPENDIX B: DESCRIPTION OF
THE 6-BAND k · p THEORY
The k ⋅ p methods: generalities.
We first give some general information about the k ⋅ p method we used
throughout this work. The k ⋅ p method is known to be very efficient at accu-
rately describing the properties of the electronic structure of a semiconduc-
tor near the Г point [198–202]. It may be implemented with a 2 × 2-band
model for the CB, a 6 × 6-band Luttinger model for the VB in an effective
Hamiltonian description via the Luttinger parameters γ i [203]. However,
an 8-band model is needed to describe the coupling between the CB and
the VB, whereas a 14-band model is mandatory to properly account for the
absence of inversion symmetry in the Td -symmetry group [204–206].
Beyond that, an extended 30-multiband treatment becomes necessary
[170, 207–210] to correctly describe the full BZ of indirect band gap semicon-
ductors like AlAs, Si, Ge, and related compounds. Its treatment requires the
inclusion of remote bands in the Hamiltonian basis, whereas the description
of the tunneling transport in a 14- or a 30-multiband approach requires

resolution of the issue of the spurious bands inherent to band truncation
[210–212].
The general k ⋅ p Hamiltonian in the crystal writes:
p2
H SC = + U(r ) + ( ∇U × p ) ⋅ σ , (4.12)
2m0 4m02c 2
where:

H SO = (∇U × p).σ is the atomic-SO term
4m02c 2
U is the lattice periodic potential
m0 is the free electron mass
c is speed of light
The wave function is the solution of the Schrödinger equation H SC Ψ = E Ψ ,

with the Bloch form Ψn ,k (r) = e ik .r ϕnk (r) satisfying
 2 2 
 H SC + k + m k ⋅ p + 4m2c 2 (∇U × k ) ⋅ σ  ϕnk = Enk ϕnk , (4.13)
 0 0 

where k 2 = (2 / 2m0 )k 2 is the free-electron energy. The atomic-like wave
function at k = 0, ϕnk (r), are assumed to be known as well as their symme-
try. We denote ϕn = ϕn( k =0)(r) and En = En( k =0) with H SC ϕn = Enϕn .
The wave functions at k ≠ 0 can be expanded as a series of ϕn
ϕ nk ( r ) = ∑C
k
nk ϕn . (4.14)
with Ω the crystal volume. Ω ∫

where {ϕn } is a set of basis functions, 〈ϕm | A | ϕn 〉 = (1 / L ) ϕ∗m (r) Aϕn (r)d r,
k · p Approach for Describing Metallic (Ga,Mn)As

The purpose of this section is to give some insights about the procedure to
calculate transport properties of spin-polarized holes in the valence band
of heterojunctions integrating the (Ga,Mn)As ferromagnetic semiconduc-
tor in the frame of the k ⋅ p theory. This includes the contribution of the p-d
exchange interaction in the molecular field approximation. For zinc-blende
semiconductors, we take into account explicitly the four Γ 8 and the two
Γ 7 valence subbands, for which we choose the basis functions in the form
according to Dietl et al. [29]:
3 3 1
u1 = J = , jz = = ( X + iY ) ↑
2 2 2
3 1 1
u2 = J = , jz = = i ( X + iY ) ↓ −2 Z ↑ 
2 2 6
3 1 1
u3 = J = , jz = − = i ( X − iY ) ↑ +2 Z ↓ 
2 2 6
3 3 1
u4 = J = , jz = − = i ( X − iY ) ↓
2 2 2
1 1 1
u5 = J = , jz = = ( X + iY ) ↓ + Z ↑ 
2 2 3
1 1 1
u6 = J = , jz = − = i  − ( X − iY ) ↑ + Z ↓  (4.15)
2 2 3
where X, Y and Z denote Kohn–Luttinger amplitudes, which for the sym-
metry operations of the crystal point group transform as px , p y and pz wave
functions of the hydrogen atom.
Added to the Kohn–Luttinger kinetic Hamiltonian [84], the hole
Hamiltonian H h includes a p-d exchange term introduced by the interaction
between the localized Mn magnetization and the holes derived in the mean-
field approximation thus giving [29, 83]:
H h = −( γ 1 + 4 γ 2 )k 2 + 6 γ 2 ∑L k
α
2 2
α α +
(4.16)
+6 γ 3 ∑
α ≠β
( Lα Lβ + Lβ Lα )kα kβ + λ so LS + ∆ excm̂S

α = { x , y , z }, Lα are l = 1 angular momentum operators, S is the vectorial
spin operator, m̂ the unit magnetization vector, and γ i are Luttinger param-
eters of the host semiconductor. ∆ exc represents the spin-splitting between
the heavy holes at the Γ 8 point. The detailed form of the 6 × 6 Hamiltonian
is given in Ref. [29]. We chose ∆ exc = 120 meV in agreement with the value
extracted from the magnetization saturation of (Ga,Mn)As [58]. Note that
3
the standard 4 × 4 k ⋅ p Hamiltonian projected on the Γ 8 ( J = ) basis can be
2
written in a cubic form as:
H h = ( γ 1 + 5 / 2 γ 2 ) κ 2 I 4 − 2 γ 2 (κ ⋅ J )
2
(4.17)
+2 ( γ 3 − γ 2 ) ∑i< j
κ iκ j ( J i J j + J j J i ) + ∆ excm̂S
 
2
in unit of 2m0 where J i is the 4 × 4 angular momentum operator in the
subspace J = 3 / 2 along the direction i (Table 4.3).
Figure 4.22 depicts the four different Fermi surfaces of (Ga,Mn)As cor-
responding to spin-polarized HH and LH-bands at the Г point, and calcu-
lated by the 6-band k ⋅ p method. These have been calculated for an exchange
parameter BG of 40 meV and a Fermi energy equaling −0.22 eV counted
from the top of the (Ga,Mn)As valence band. The magnetization is along the
[001] direction. The less-filled bands display a strong anisotropic character
due to the interplay between the spin-orbit and exchange interactions; i.e.
when the kinetic energy is small compared to the latter interactions.
TABLE 4.3
Luttinger Parameters γ, Spin-Orbit Splitting ∆0 (eV) and Valence-
Band Offset VBO (eV) Used
γ1 γ2 γ3 Δ0 VBO
GaAs 6.85 2.1 2.9 0.34 0
AlAs 3.25 0.64 1.21 0.275 −0.64
InAs 20.4 8.3 9.1 0.38
In0.25Ga0.75As 10.24 8.3 4.45 0.35 −0.7 (Low-T)
(Ga,Mn)As 6.85 2.1 2.9 0.34 +0.22
FIGURE 4.22 The four different Fermi surfaces of (Ga,Mn)As calculated in the
k × p formalism for a Fermi energy equal −135 meV counted from the top of the
valence band (p = 1× 1020 cm−3) and an exchange interaction ∆ exc = 120 MeV
(calculated without strain). The magnetization is along the [001] growth crystalline
axis. The color code scales the Fermi wavevactor (in nm−1) along the corresponding
crystalline axis. The corresponding band structure of (Ga,Mn)As in the reciprocal
space can be found in Refs. [12, 25] (After Tanaka, M. and Higo, Y., Phys. Rev. Lett. 87,
026602, 2001; Matsukura, F. at al., III–V ferromagnetic semiconductors, in Handbook
of Magnetic Materials, vol. 14, K. H. J. Buschow (Ed.), North Holland, Amsterdam, pp.
1–87, 2002. With permission.).
The 6-Band k · p Kane Model (Issued from the 14-Band)

In a 6-band model like that we use for the scattering perturbative treatment,
{ }
{X ,Y , Z } ⊗ ↑,↓ or linear combinations of them, represent the basis func-
tions in Equation 4.14. This effective 6-band model derives from a larger
{ }
14-band basis set {XC ,YC , ZC , S , X ,Y , Z } ⊗ ↑, ↓ from Lowdin downfolding.
The k ⋅ p term is diagonal in spin and allows (i) the coupling terms between
the Γ 6 and the {Γ 7 , Γ 8 } wave function as S px iX = ϖ; (ii) the coupling terms
between the {Γ 7 , Γ 8 } and the higher CBs of p-type symmetry {Γ 7C , Γ 8C } as
X p y iZC = ϖ X as well as (iii) the coupling terms, in the Td symmetry group,

between the Γ 6 and the {Γ 7C , Γ 8C } wave function as S px iXC = ϖ′ , and where
ϖ, ϖ′ , and ϖ X are real. ϖ′ = 0 in Oh as considered here. The k ⋅ p parameters

are introduced according to P = ϖ, P ′ = ϖ′ , PX = ϖ X with respective
m0 m0 m0
2m 2m 2m
characteristic energies E P = 2 0 P 2 , E P ′ = 2 0 P ′ 2 , E PX = 2 0 PX2 .

The SOI is introduced according to the work of Koster [201]. We resume
here the couplings which may differ from zero: (i) the core-SOI in the
æ 3 2 ö ¶U ¶U
higher CB, DC = ç 2 2 ÷
XC py - px iYC , (ii) the SOI in the VB,
è 4m0 c ø ¶x ¶y
æ 3 2 ö ¶U ¶U
D=ç 2 2 ÷
X py - px iY , and (iii) the SOI caused by the lack of inver-
è 4m0 c ø ¶x ¶y
æ 3 2 ö ¶U ¶U
sion center in the Td symmetry-group, D¢ = ç 2 2 ÷
X py - px iYC
è 4m0 c ø ¶x ¶y
(in Oh , ∆ ′ = 0).
The p-d exchange interactions in the VB one needs is introduced in a
mean-field approximation from Refs. [29, 83], like H exc = 3 BG sˆ ⋅ M ˆ , where
3BG represents the average energy interaction among holes, s the hole spin
and M the 3-d localized spin.
The 6-Band k · p Kane Model: Eigenvectors and

Effective Mass in III–V Heterostructures
One considers a magnetic tunnel barrier of thickness d and barrier height
(Φ = 2w ) equaling the exchange potential in order to avoid scattering of eva-
nescent waves in the barrier. What are the tunneling properties and related
asymmetry of transmission? One assumes the continuity of the effective
mass over the whole heterostructure. This can mimic the case of GaSb bar-
rier sandwiched between two (Ga,Mn)As ferromagnetic electrodes with a
larger SOI strength in the barrier. The true physical situation of SOI in the
III–V, electrodes (e.g. (Ga,Mn)As or (Ge,Mn)) and barrier, will be treated
numerically in a second step for comparison (see Ref. [168]).
The 6 × 6 Hamiltonian in the three different regions (left electrode, bar-
rier, and right electrode) then writes:
 Hˆ kp + Hˆ exc for x < 0 or x > a,

Hˆ 0 =  (4.18)
 Hˆ kp − V0 for 0 < x < a,
where Ĥ kp represents the kinetic energy, Hˆ exc = wσˆ x the exchange poten-
tial with w the exchange strength, and where V0 is the barrier height (band
discontinuity).
 Hˆ k↑↑. p 0 
Hˆ k . p =   , (4.19)
 0 Hˆ k↓↓. p 
with Hˆ k↓↓. p = Hˆ k↑↑. p , and

↑↓
Hk. p =
 X↑ Y ↑ Z↑ 
2
 2 2 
 (k x + k y ) 
2m0  E PX EP 
 X↑ −kxk y 
 E5C − E8
+  0 
E PX 2 EP 2 E 6 − E8 
 − ky − kx 
 E5C − E8 E 6 − E8 
2

 2 2
(k x + k y ) 
  E PX EP  2m0 .
Y ↑ −kxk y  +  0
  E5C − E8 E 6 − E8 
−
E PX 2
kx −
EP 2
ky

 E5C − E8 E 6 − E8 

2
2 2 
 2m0
(k x + k y ) 
 Z↑ 0 0
E PX 2 2

 − ( k x + k y )
 E5C − E8 
(4.20)
We then introduce the  and  parameters according to the notation of
Ref. [208]
 EP E 
M = + PX  , (4.21)
 E6 − E8 E5C − 8 
 EP E 
L= − PX , (4.22)
 E6 − E8 E5C − 8 
and then choose the parameters to make HH dispersion nearly flat,
2
≈ ( M − L) / 2. Within this approach, under oblique incidence, LH
2m0
and HH bands weakly mix together on reflection/transmission process.
Consequently, without SOI, the HH and LH states can be almost transmit-
ted like free carriers with respective effective masses m∗HH , m∗LH .

The core atomic SOI Hˆ SO =
4m02c 2
(∇U × pˆ ) ⋅ σ,
ˆ acting afterwards as a
perturbation, writes in the present basis:
Ĥ SO =
 X↑ Y↑ Z↑ X↓ Y↓ Z↓ 
X ↑ 0 −i ∆ / 3 0 0 0 ∆ / 3 

Y ↑ i∆ / 3 0 0 0 0 −i ∆ / 3
 
Z ↑ 0 0 0 −∆ / 3 i∆ / 3 0 ,
X ↓ 0 0 −∆ / 3 0 i∆ / 3 0 
 
Y ↓ 0 0 −i ∆ / 3 −i ∆ / 3 0 0 
 Z ↓ ∆/3 i∆ / 3 0 0 0 0 
(4.23)
In each layer, the bare Hamiltonian is diagonal within the spin index σ.
The condition for an infinite heavy-hole mass is (M − L) / 2 ≈ 1 where
 EP E   EP E 
= + PX  and  =  − PX  is, in that case, the
 E6 − E8 E5C − 8   E6 − E8 E5C − 8 
inverse of the LH effective mass [208]. One can show that the three eigenvec-
tors corresponding to the same eigenvalue E are of respective orbital-sym-
Mξ
metry Z L , R , Y L , R and X + Y in the left (L) and right (R) magnetic
LK L,R
Mξ
contacts and Z , Y and X − i Y in the tunnel barrier region
TB TB
LλK TB
( x̂ is the direction of the magnetization, ŷ the direction of k, and ẑ the
direction of tunneling).
With our chosen parameter of infinite HH-mass, HH 1 = Z and
HH 2 = Y are the two different HH-orbital states of same (infinite)
mass, carrying no current flux. On the other hand, the third eigenstate,
Mξ Mξ
LH = X + Y and LH = X − i Y are of a LH-symmetry with a
LK LλK
non-zero mass mlh =  −1. LH-to-HH is the only possible mixing after reflec-
tion-transmission upon barrier interface.
Under our assumption, it results in two main non-zero tunnel transi-
tions mediated by the SOI in the barrier, resp. 〈 Z | Lˆ + | LH 〉 and 〈 LH | Lˆ + | Z 〉
for respective | Z〉 and | LH 〉 ingoing waves, the latter process admitting both
LH → LH and LH → HH tunneling branching inside the barrier through
the non-perfect orthogonality between LH and HH states under oblique
incidence.
Mξ
Note that, among all states (eigenvectors), only X − i Y carries a
LλK TB
λK ξLM
non-zero orbital moment equaling 〈 LHTB | Lˆ z | LHTB 〉 = µB
(µ B is the Bohr magneton). ( λK L)2 + (ξM)2
APPENDIX C: THE 30-BAND FULL-BZ k · p

TREATMENT OF THE SKEW TUNNELING
We have implemented a robust tunneling 30-multiband k ⋅ p calculation
method, beyond the 14-band treatment, to calculate the properties of chiral
transmission in tunnel junction over the whole BZ if needed. We describe
here the method we employed and compare the results to standard tunnel-
ing codes.
The requirement to implement a 30-band full-BZ k ⋅ p treatment origi-
nates from the need to describe the tunneling transport in the valleys close
at the edge of the first BZ (1 − BZ ). In that context, Cardona and Pollak
[213] used a 15-function basis, without SOI, to describe the band disper-
sion throughout the whole BZ without supplementary Luttinger parameters.
Cavassilas [214] used a 20-function basis, including the spin, and introduced
two bands named s ∗ and pseudo-Luttinger parameters to mimic the d levels
along the ideas of Vogl et al. for LCAO. This 20-band model contains ten
Appendix C: The 30-Band Full-BZ k · p Treatment of the Skew Tunneling 227
adjustable parameters to describe the s∗ bands, nine coupling parameters

for the Td group (only six for Oh ) and six pseudo-Luttinger parameters, i.e.
25 adjustable parameters. However, this 20-band Hamiltonian do not give
access to the L valley of the second CB. Richard et al. [170] then proposed a
30-band model to describe Td or Oh groups involving SOI. The 15 states of
the real crystal correspond to the [000], (2π / a)[111], and (2π / a)[200] plane-
wave states of free electrons in the “empty” germanium lattice. The large gap
between the (2π / a)[200], and (2π / a)[200] plane waves (more than 15 eV)
suggests that these 15 states are sufficient to obtain a correct energy band
diagram. Details of the 30-band method can be found in Refs. [170, 208, 209]
for III–V and Ref. [207] for group IV semiconductors. The typical energy of
the bands one considers at the Г symmetry point are given in Table 4.4 for
GaAs.
The Oh symmetry-group (Si and Ge), admits ten k ⋅ p matrix elements
of interest and seven more compared to the 8-band description among:
PXd = XC p y iZ d , P3 = D1 px iX , P3 d = D1 px iX d , P2 = S2 px iX ,
P2 d = S2 px iX d , PS = Sv px iXC , PU = SU px iXC whereas, for the Td
symmetry group: P ′ , Pd′ = X d px iZ , P3′ = D1 px iXC , P2′ = S2 px iXC ,
PS′ = Sv px iX , PSd′ = Sv px iX d , PU′ = SU px iX , PUd ′ = SU px iX d ,
PS = Sv px iXC , PU = SU px iXC are also non-zero. The parameters for
GaAs and AlAs compounds we used are gathered in Table 4.5.
The supplementary atomic SOI is introduced via the following couplings:
3 ∂V ∂V
(i) ∆ d = Xd py − px iYd , (ii) the coupling between the two dif-
4m02c 2 ∂x ∂y
3 ∂V ∂V
ferent multiplets (Γ 7 , Γ 8 ) and (Γ 7 d , Γ 8 d ), ∆ dso = Xd py − px iY ,
4m02c 2 ∂x ∂y
(iii) the coupling between the (Γ 7C , Γ 8C ) multiplet and the (Γ 7 d , Γ 8 d )
which stems from Γ 5C levels and Γ 8 level which stem for Γ 3 ,
3 ∂V ∂V
∆ 3C = 2 2
D1 pz − p y iXC . For the Td group, there are some
4m0 c ∂x ∂y
additional SOI terms beyond ∆ ′ : (i) the coupling inside (Γ 7 , Γ 8 ) multi-
3 ∂V ∂V
plets ∆Cd ′ = Xd py − px iYC , (ii) the coupling inside (Γ 7 , Γ 8 )
4m02c 2 ∂x ∂y
TABLE 4.4
k = 0 Energy Level (eV) used in the 30-Band k·p Model
Γ8C Γ7C Γ6 Γ8 Γ7 Γ6v Γ6q Γ8d Γ7d Γ8−3 Γ6u
GaAs 4.569 4.488 1.519 0 −0.341 −12.55 13.64 11.89 11.89 10.17 8.56
AlAs 4.69 4.54 3.13 0 −0.30 −12.02 13.35 12.57 12.34 9.12 7.65
TABLE 4.5
Dipole Matrix Elements of GaAs and AlAs in a 30-Band k·p Model given in Unit of Energy in
Definition of Ref. [170]. Energies E and Matrix Components P are Linked by E = 2m0 /2 P 2
EP EPX EPd EPXd EP3 EP3d EP2 EP2d EPS EPU E P¢
GaAs 22.37 17.65 0.01 4.344 4.916 8.88 0.032 23.15 2.434 19.63 0.0656
AlAs 19.14 14.29 0.01 8.49 3.99 9.29 0.032 15.01 1.79 16.00 0.14
which stems from the Γ 5C levels and Γ 8 level and from the Γ 3 ,
3 ∂V ∂V 3 ∂V ∂V
∆ ′3 = 2 2
D1 py − px iX , ∆ ′3 d = 2 2
D1 pz − p y iX d , as
4m0 c ∂x ∂y 4m0 c ∂x ∂y
well as (iii) possible supplementary Luttinger parameters in order to account
for the effects of more remote bands.
Spurious states in the 14- and 30-bands k ⋅ p approach. A Novel ‘ghost-
band’ method.
The appearance of spurious states at large k, within (14-band) or out-
ward (30-band) the 1 − BZ , arises from the truncation procedure inherent
to the k ⋅ p method. Those generally crossing the gap, are detrimental for
the calculation of the elastic tunneling transmissions occurring at constant
energy E. As a result, they have to be removed from the gap without any
possibility to suppress them because of the matching rules. The method we
have adopted is the generalization of the procedure given by Ref. [212] to
the 14- and 30-band problems. In order to extend the procedure to a wider
region within the BZ away from the Г point, we propose the so-called ghost-
band method. The method consists of adding supplementary off-diagonal
coupling terms (Poff = α ij k z2 ) via a set of α ij parameter conveniently chosen,
and where k z is directed along the current flow (z). The idea is to utilize
the off-diagonal formalism [212], but operate at the edge of the 1 − BZ . We
report here the main issues:
1. New fictitious bands (the ghost-bands) of adequate symmetry are

introduced in order to branch-on Poff . These couplings are calculated
in order to leave unchanged the properties of the physical bands in
the vicinity of the valleys. These ghost-bands mimic, on average,
the other physical truncated remote bands. Their mean position in
energy, higher than the S-type CB, have to be set by error (on the
wave function components and energy difference) minimization pro-
cedures. Poff allows inversion of the concavity of the spurious bands at
large k, thus removing the gap-crossing.
2. The k ⋅ p Hamiltonian k . p involving the spurious state is changed
into a novel Hamiltonian  k . p with H k . p + S −1Voff S k z2 with
ij
α ij =   −1 Voff   and where S is the matrix of eigenstates at the point
where the off-diagonal coupling with ghost-bands operates (near the
edge of the 1 − BZ ).
3. The α ij parameters chosen are calculated to leave totally unchanged
the bare Hamiltonian near the Г point. This is performed via effec-
tive Hamiltonian analyses. The main argument is that α ij k z2  Pk z at
small k z .
4. The α ij parameters chosen are calculated to leave totally unchanged
the current-operator (1 / ) (∂Ĥ / ∂k z ) near the Г point from 2α ij k z  P
at small k z .
Appendix C: The 30-Band Full-BZ k · p Treatment of the Skew Tunneling 229
5. In order to optimize the transport in the X or L–Valleys of the CB

in the vicinity of the 1 − BZ edge, one needs to apply Voff between
spurious and ghost bands in the local basis where the Hamiltonian is
diagonal (matrix of eigenstates  αβ ) in order to leave unchanged the
remaining physical bands. It has the result that:
a. The electronic and transport properties are not affected at the
vicinity of the Г point for all the CB, VB, HH, LH, and SO bands
by Voff .
b. The electronic and transport properties of the CB are not affected
at the vicinity of the point where Voff is introduced (close to the
first BZ -edge). The tunneling current mediated by a finite eva-
nescent wave-vectors will be only poorly affected by the method.
Those ghost-band evanescent states correspond to very large eva-
nescent wave-vectors.
In order to check the validity of our 30 ghost-band approach, we have
calculated the in-plane energy dispersion of holes in an 5.1 nm AlAs/6.2 nm
GaAs/5.1 nm AlAs QW along the X direction with our 30-band k ⋅ p tunnel-
ing code with bulk 30-band parameters of III–V, GaAs and AlAs, extracted
from Richard et al. [170] (Figure 4.23 left). One compares the numerical
results obtained that way with effective 6-band k ⋅ p calculations displayed,
at right, for the two X and K directions. Results are in excellent agreement
which proves the relevance of our 30-ghost band approach for the tunneling
issues in heterostructures. We focus on the particular point that the AlAs
barriers admit an indirect gap along the X-valley which is perfectly taken
into account in our 30-band tunneling model.
FIGURE 4.23 Comparison between the in-plane hole energy dispersion for a
AlAs/6.21 nm GaAs/AlAs quantum well (QW) derived from our 30-band tunnel
k ⋅ p code (left) along the X direction of the BZ, and the one derived from Ref. [215]
obtained with a 6-band model along both X and K directions.
APPENDIX D: THEORY OF SPIN-TRANSPORT AND

SPIN-TRANSFER WITHIN THE k · p FRAMEWORK
Our calculations of the transmission coefficients are based on the multiband
transfer matrix technique developed in Ref. [85, 86] and applied to the hole
6 × 6 valence band k ⋅ p Hamiltonian H h . We use matching conditions for
the wave functions and wavecurrent given by Smith and Mailhiot [88]. To
calculate the transmission coefficient, we will use the transfer-matrix tech-
nique [85, 86] or the layer-by-layer S-scattering matrix formalism [216], and
represent the device as a stack of two-dimensional flat-band interior layers
of constant average electrostatic potential. The 0th (L) and ( N + 1)th (R) lay-
ers are semi-infinite emitter and collector. We will take into account elastic
processes only, i.e. assume that the hole energy ε h and lateral momentum k
are conserved during tunneling. Substituting k z → −i∂ / ∂z into the Kane
Hamiltonian matrix, we can find wave function eigenvectors ψ n,α depending
on k in the ath flat-band region (α) according to:
ψ n (z) = ∑ [A
α =1→ 6
+
n ,α υ(kn+,α )exp(ikn+,α z ) + An−,α υ(kn−,α )exp(ikn−,α z )] (4.24)
where ψ n,α and υn,α are six-component column vectors and we separated
+
all our solutions into two subsets with kn,α corresponding to either traveling
waves carrying the probability current from left to right or to evanescent
waves decaying to the right and kα,− n corresponding to their left counterparts.
Since the Kramers symmetry is broken in a magnetic region, kn− ≠ − kn+ for
real k (traveling waves), however, complex k always occur in complex con-
jugated pairs, i.e., kn− = kn+* . The latter condition is a consequence of the
Hermitian character of the hole Hamiltonian. This condition is necessary
to ensure that the current across the device is in the steady state. It results
also that, in the k ⋅ p approach, the current operator in the z direction writes
1
J = Ñ k Hˆ h [85, 86].

The boundary conditions to match at a single interface between two
different materials are:
1. the continuity of the six components of the envelope function accord-

+ − +
ing to ψ n + ∑n rn ,n ψ n = ∑n′ tn ,n′ ψ n′ where the subscript tn ,n ′ (rn , n ) refers
to the respective transmission (reflection) amplitude from incident
(n), reflected (n) and transmitted (n′ ) waves together with
2. the continuity of the six components of the current wave-vector
according to j ψ + ∑ r j ψ = ∑ t j ψ .
+ − +
n n n ,n n n′ n ,n′ n′
The boundary conditions at each interface zα (α = 0...N ), corresponding

to the continuity of the wave function and the current can be expressed in
matrix form as follows:
χα π α ( zα ) Aα = χα +1π α +1( zα +1 ) Aα +1 (4.25)

Appendix D: Theory of Spin-Transport and Spin-Transfer within the k · p Framework 231
where χα is the 12 × 12 matrix containing the 12 eigenvectors υ(kn , α ) as well

as the 12 current vectors ĵz (kn ,α )υ(kn ,α )
. υ(kn+,α ) . υ(kn−,α ) .
χα =   (4.26)
. ˆjz (kn+,α )υ(kn+,α ) . ˆjz (kn−,α )υ(kn−,α ) .
and where π αν,µ = exp(ikn ,α z )δ νµ is the propagation matrix within the layer α.
We adopt the following convention for enumeration of the lead layers
and all related quantities: 0 ≡ L (emitter) and N + 1 ≡ R (collector). The sys-
tem of linear Eqs. above can be solved as
 AL+   AR+   ++ +−   AR+ 

 −  =  = −  =    (4.27)
 AL   AR   −+  − −  AR− 
 
where the 12 × 12 transfer matrix
 = χ −L1  Π n=1.. N χα π α ( zn +1 − zn )χα−1  χ R π N +1( z N ) (4.28)
is partitioned according to the hole propagation direction in the emitter (L)

and collector (R) and where the boundary condition is given by AR− = 0. It fol-
lows that the respective 6 × 6 amplitude transmission and reflection matrices
are written respectively tn ,n¢ =[ ++ ]-1 and rn ,n ′ =[ + − ] × [+ + ]−1.
Nevertheless, one must be aware that the k ⋅ p theory of bulk semicon-
ductors is based on the consideration of the symmetry of the zone-center
eigenstates and a small number of interband matrix elements of the momen-
tum: it generally ignores the problem of the intracell shape and localization
of the Bloch functions. However, when the crystal potential is modified on
the scale of a unit cell, as, for example, in the case of an abrupt interface, it
is intuitively evident that the matrix elements of the perturbation will be
strongly dependent on the local properties of the Bloch functions [205, 217,
218]. In general, the reduced point symmetry C2v of a zinc-blende-based
(001) interface allows mixing between heavy-hole and light-hole states even
under normal incidence. However, this HH-LH non-diagonal mixing for the
wave function matching at one interface are calculated to be smaller than
10% for the elastic hole energy considered [205]. More refined calculations
of spin-polarized hole transmission trough zinc-blende (001) heterojunction
would require tight-binding treatment, as proposed by Sankowski et al. [162].
Assuming that the in-plane wave-vector k is conserved, the spin-polar-
ized transmission tn ,n ′ and reflection rn ,n ′ amplitudes through a specific het-
erojunction can be determined by a matrix numerical procedure detailed in
ˆ
k ,m
Refs. [85, 86]. The transmission coefficients Tn ,n ′ and elastic tunneling cur-
rent jTm̂ follows on the shell (ε,k) according to Landauer–Buttiker formula:
ˆ
k ,m ˆ
k ,m
ˆ
k ,m ˆ
k ,m ˆ
k ,m < ψ n′ | ˆj | ψ n′ >
Tn ,n ′ = [tn,n ′ )* tn,n ′ (4.29)
< ψ  | j | ψ  >
ˆ
k ,m ˆ
k ,m
n n
εm
∑∫
e ˆ
k ,m
jTmˆ = Tn ,n ′ (ε, eV )d ε (4.30)
hk εF
 ,n ,n ′
where ε is the hole energy considered whereas ε m refers to the lowest abso-
lute hole energy involved in the transport.
The TAMR = [ jT[001] − jT[100]) /[ jT[100]) is obtained numerically after summa-
tion of the tunneling current Equation 4.3 over n, n′ and k and integration
on the energy shell ε taking ε F = −180 meV and ε m = −140 meV (consider-
ing a typical energy broadening for (Ga,Mn)As of 40 meV). In the calcu-
lations performed throughout the article, when not explicitly written, the
Fermi level ε F of (Ga,Mn)As was fixed at −180 meV from the top of the
(Ga,Mn)As valence band, thus giving a reasonable hole concentration of
about 2 × 1020 cm−3. Concerning the heterostructure itself, the average VBO
between (Ga,Mn)As and AlAs (GaAs), varying linearly with bias, was fixed
at 0.7 eV (0.22 eV) at zero bias, corresponding to an effective barrier height
of 0.58 eV (0.1 eV). Note that the biaxial strain Hamiltonian H BS can be also
added in the Hamiltonian form H h when strain and stress effect need to be
considered.
General Arguments on Spin Transfer Phenomena

Concerning the calculation of spin torque in III–V heterojunctions, one
can start from the above equation giving the time derivative of the angular
momenta density operator µˆ i = ρˆ Jî at a certain position inside the hetero-
ˆ
structure where J i is the angular momentum operator along direction i and

J the velocity operator:
∂Jˆ

∂µˆ i
∂t
( )
= −∇ j Jî + ρˆ i
∂t

∂µˆ
∂t
∆
( )
= −∇ j Jî + exc m

ˆ × µˆ (4.31)
Summing the latter equation over the different hole states included in the
transport gives out (after having considered by the time derivative of the
angular moment operator is null in the limit of coherent elastic transport):

( )
∇ j Jî =
∆ exc

ˆ × µˆ
m (4.32)
This gives in fine the torque at a given interface (int.) after integration over
the semi half-infinite structure:

j Jî = exc × µ J (4.33)
int.
where:

exc is the average exchange field experienced by holes
Appendix D: Theory of Spin-Transport and Spin-Transfer within the k · p Framework 233
µ J is the magnetic moment vector (per unit surface) injected in the

thin layer subject to the torque by the spin-polarized current
We recover here the general results established by Kalitsov et al. [186] in

the case of a transport in the CB using the Keldysh framework.
This demonstrates the presence of two intertwined contributions

j J i (parallel to the layer) and j Jî (perpendicular to the layer) for the
ˆ
x z
spin torque ( ŷ is the magnetization direction) as derived for symmetric
[187, 188] and asymmetric barriers [189] and demonstrated experimen-
tally in epitaxial MgO barrier MTJs [190, 191]. These two components of
the transverse spin currents are also responsible for the particular angu-
lar dependence of the tunneling transmission, as in the case of the con-
duction band [219, 220]. These two components of the torque are linked to
the average value of the two components of the angular momenta carried
by the hole current [186]. Within the general circuit-theory formalism
developed by Braatas et al. [221] and adapted to metallic trilayers, one can
establish that these two contributions in magnetic tunnel junctions to the
torque can be linked to the magnetic moment in the barrier (generaliza-
tion of the spin accumulation) through the real and imaginary part of the
tunnel mixing conductance G↑↓ . These should be calculated in a future
work.
Spin Transfer Phenomena with SOI

We decompose now the total angular momentum J as a sum of spin and
orbital momentum to give J = l + s . In order to determine the efficiency
of spin-transfer and spin-orbit-torque, one needs to compute the dynam-
ics of the spin-polarized carriers with spin s and local 3d transition metal
magnetization M according to their dynamics. Starting with the overall
carrier-Hamiltonian:
 = 0 +
 exc +
 SO , (4.34)
where:
0 is the spin-independent part (e.g. kinetic part)
exc = − J excs ⋅ M is the exchange Hamiltonian between the carrier spin s
and the local magnetization M
J exc is the exchange constant
SO is the spin-orbit term
From a quantum mechanical picture, the dynamics of the spin s inside

the ferromagnetic layers is written:
∂s 1 s
= [s, Hexc ] + Ps n −
∂t i τs
∂s J exc s
= s × M + s n − , (4.35)
∂t τs
∂M ∂s J
= − = exc M × s (4.36)
∂t ∂t
if SO is neglected, and where Ps is the rate of the spin injected from an exter-
nal source (current, SHE spin-torque) at the interface with the ferromagnet, n
is the vectorial direction of the spin injected, and τ s is a characteristic longitu-
dinal spin-flip time inside the ferromagnet. The latter equation, derived from
general quantum mechanics arguments, gives the rule for the conservation of
the total angular momentum shared between s and M. Here, we have used:
¶<s> 1
= <[s, ] >, (4.37)
¶t i
where < ... > represents a certain quantum-mechanical averaging over the
nonequilibrium carrier states and < σ >= s and where σ is the corresponding
Pauli matrix.
In the case of a long spin-lifetime τ s compared to the precession length

τ exc = , one has:
J exc

s = s ( n × M ) , (4.38)
J exc
giving in the end:
∂M
= s M × ( n × M ) , (4.39)
∂t
One integrating over the total thickness t of the ferromagnet, one can express
that:
∂M
=  s M × (n × M ) , (4.40)
∂t
where  s is the spin-current injected from the external source. One recovers
then the standard expression of the Slonczewski-type like torque or anti-
damping torque.
In the case of a very short spin-lifetime τ s compared to the precession

length τ exc = , one has:
J exc
s = s t s n, (4.41)
giving in the end:
¶Mt J t
=  s exc s M ´ n , (4.42)
¶t
and one recovers the field-like torque component.

References 235
H SO introduces an additional precession term in the bulk or at interfaces

(e.g. Rashba) which can lead to local spin-memory loss [222] of the longitu-
dinal component (that is parallel to the magnetization) and to spin-decoher-
ence of its transverse component responsible for spin-transfer. According to
this, the SOI fields may have the effect, via local (interfacial) spin-precession
of incoming spin-polarized carriers, to decrease the efficiency of the spin-
torque (STT or SOT) reduced from the expected maximum amplitude, e.g.
∂Mt
=  s M × ( n × M ). This total transverse spin-current is generally divided
∂t
into a spin-torque current in the volume (ν) of the ferromagnet and at its
J exc
interface (S), and expressed as
∫
V +S
M × < σ > dz , and a spin-current dis-
J exc
sipated in the lattice (and then lost) because now J ∫ =
∫
V+ S
M × < σ > dz
is no longer fulfilled. The presence of the SOI at surface and in bulk requires
a re-examination of the generalized spin-mixing conductance as proposed
recently in the case of metallic multilayers [223].
REFERENCES
tronic structures, Rev. Mod. Phys. 86, 187–251 (2014).
2. T. Jungwirth, J. Wunderlich, V. Novak et al., Spin-dependent phenomena
and device concepts explored in (Ga,Mn)As, Rev. Mod. Phys. 86, 855–896
(2014).
3. H. Kurebayashi, J. Sinova, D. Fang et al., An antidamping spin-orbit torque
originating from the Berry curvature, Nat. Nanotechnol. 9, 211–217 (2014).
4. K. Olejnik, V. Novak, J. Wunderlich, and T. Jungwirth, Electrical detection of
magnetization reversal without auxiliary magnets, Phys. Rev. B 91, 180402(R)
(2015).
5. V. Novak, K. Oleijnik, J. Wunderlich et al., Curie point singularity in the tem-
perature derivative of resistivity in (Ga,Mn)As, Phys. Rev. Lett. 101, 077201
(2008).
6. M. Wang, R. P. Campion, A. W. Rushforth, K. W. Edmonds, C. T. Foxon, and
B. L. Gallagher, Achieving high curie temperature in (Ga,Mn)As, Appl. Phys.
Lett. 93, 132103 (2008).
7. L. Chen, S. Yan, P. F. Xu et al., Low-temperature magnetotransport behav-
iors of heavily Mn-doped (Ga,Mn)As films with high ferromagnetic transition
temperature, Appl. Phys. Lett. 95, 182505 (2015).
8. P. Nemec, V. Novak, N. Tesarova et al., The essential role of carefully optimized
synthesis for elucidating intrinsic material properties of (Ga,Mn)As, Nat.
Commun. 4, 1422 (2013).
9. L. D. Anh, N. Hai, and M. Tanaka, Observation of spontaneous spin-splitting
in the band structure of an n-type zinc-blende ferromagnetic semiconductor,
Nat. Commun. 7, 13810, (2016).
tions, Rev. Mod. Phys. 76, 323–410 (2004).
11. C. Chappert, A. Fert, and F. Nguyen Van Dau, The emergence of spin electron-
ics in data storage, Nat. Mater. 6, 813–823 (2007).
12. M. Tanaka and Y. Higo, Large tunneling magnetoresistance in GaMnAs/AlAs/
GaMnAs ferromagnetic semiconductor tunnel junctions, Phys. Rev. Lett. 87,
026602 (2001).
13. E. Rashba, Theory of electrical spin injection: Tunnel junction as a solution of

the conductivity mismatch problem, Phys. Rev. B 62, R16267–R16270 (2000).
14. A. Fert, H. Jaffrès, Conditions for efficient spin injection from a ferromagnetic
metal into a semiconductor, Phys. Rev. B 64, 184420 (2001).
15. H. Jaffrès and A. Fert, Spin injection from a ferromagnetic metal into a semi-
conductor, J. Appl. Phys. 91, 8111–8113 (2002).
16. A. Fert, J.-M. George, H. Jaffrès, and R. Mattana, Semiconductors between
spin-polarized sources and drains, IEEE Trans. Elec. Dev. 54, 921–932 (2007).
17. G. Schmidt, D. Ferrand, L. W. Molenkamp et al., Fundamental obstacle for
electrical spin injection from a ferromagnetic metal into a diffusive semicon-
ductor, Phys. Rev. B 62, R4790–R4793 (2000).
18. B. Eble, C. Testelin, P. Desfonds et al., Hole nuclear spin interaction in quan-
tum dots, Phys. Rev. Lett. 102, 146601 (2009).
19. D. Brunner, B. D. Gerardot, A. Dalgarno et al., A coherent single-hole spin in a
semiconductor, Science 325, 70–72 (2009).
20. M. Oltscher, M. Ciorga, M. Utz, D. Schuh, D. Bougeard, and D. Weiss,
Electrical spin injection into high mobility 2D systems, Phys. Rev. Lett. 113,
236602 (2014).
21. M. Tanaka, S. Ohya, and P. N. Hai, Recent progress in III-V based ferromag-
netic semiconductors: Band structure, Fermi level, and tunneling transport,
Appl. Phys. Rev. 1, 011101 (2014).
22. H. Ohno, Properties of ferromagnetic III–V semiconductors, J. Magn. Magn.
Mater. 200, 110–129 (1999).
1019–1022 (2000).
24. H. Ohno and F. Matsukura, A ferromagnetic III–V semiconductor: (Ga,Mn)
As, Solid State Commun. 117, 179–186 (2001).
25. F. Matsukura, H. Ohno, and T. Dietl, III–V ferromagnetic semiconductors,
in Handbook of Magnetic Materials, vol. 14, K. H. J. Buschow (Ed.), North
Holland, Amsterdam, pp. 1–87, 2002.
26. J. Konig, J. Schliemann, T. Jungwirth, and A. H. MacDonald, Ferromagnetism
in (III,V) semiconductors, in Electronic Structure and Magnetism of Complex
Material, D. J. Singh and D. A. Papaconstantopoulos (Eds.), Material Science,
Springer, Berlin, pp. 163–211, 2002.
27. T. Jungwirth, J. Sinova, J. Masek, J. Kucera, and A. H. M. MacDonald, Theory of
ferromagnetic (III,Mn)V semiconductors, Rev. Mod. Phys. 78, 809–864 (2006).
28. C. Zener, Interaction between the d shells in the transition metals, Phys. Rev.
81, 440–444 (1951).
hedrally coordinated semiconductors, Phys. Rev. B63, 195205 (2001).
30. S. Souma, L. Chen, R. Oszwaldowski et al., Fermi level position, Coulomb
gap, and Dresselhaus splitting in (Ga,Mn)As, Sci. Rep. (2016), doi:10.1038,
srep27266.
31. S. Ohya, I. Muneta, Y. Xin, K. Tanaka, and M. Tanaka, Valence-band structure
of ferromagnetic semiconductor (In,Ga,Mn)As, Phys. Rev. B 86, 094418 (2012).
32. M. Kobayashi, I. Muneta, Y. Takeda et al., Unveiling the impurity band induced
205204 (2014).
33. T. Ishii, T. Kawazoe, Y. Hashimoto et al., Electronic structure near the Fermi
level in the ferromagnetic semiconductor (Ga,Mn)As studied by ultrafast
time-resolved light-induced reflectivity measurements, Phys. Rev. B 93, 241303
(2016).
34. I. Muneta, S. Ohya, H. Terada, and M. Tanaka, Sudden restoration of the band
ordering associated with the ferromagnetic phase transition in a semiconduc-
tor, Nat. Commun. 7, 12013 (2016).
References 237
35. Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom, Observation of

the spin Hall effect in semiconductors, Science 306, 1910–1913 (2004).
36. J. Sinova, D. Culcer, Q. Niu, N. A. Sinitsyn, T. Jungwirth, and A. H. MacDonald,
Universal intrinsic spin Hall effect, Phys. Rev. Lett. 92, 126603 (2004).
37. J. Wunderlich, B. Kaestner, J. Sinova, and T. Jungwirth, Experimental observa-
tion of the spin-Hall effect in a two-dimensional spin-orbit coupled semicon-
ductor system, Phys. Rev. Lett. 94, 047204 (2005).
38. S. Murakami, N. Nagaosa, S. C. Zhang, Dissipationless quantum spin current
at room temperature, Science 301, 1348–1351 (2003).
39. H.-A. Engel, B. I. Halperin, and E. I. Rashba, Theory of spin Hall conductivity
in n-doped GaAs, Phys. Rev. Lett. 95, 166605 (2005).
40. N. Nagaosa, J. Sinova, S. Onoda, A. H. MacDonald, and N. P. Ong, Anomalous
Hall effect, Rev. Mod. Phys. 82, 1539–1592 (2010).
41. J. Sinova, S. O. Valenzuela, J. Wunderlich, C. H. Back, and T. Jungwirth, Spin
Hall effects, Rev. Mod. Phys. 87, 1213–1259 (2015).
42. J. C. Rojas Sanchez, L. Vila, G. Desfonds et al., Spin-to-charge conversion
using Rashba coupling at the interface between non-magnetic materials, Nat.
Commun. 4, 2944 (2013).
43. E. Lesne, Y. Fu, S. Oyarzun et al., Highly efficient and tunable spin-to-charge
conversion through Rashba coupling at oxide interfaces, Nat. Mater. 15, 1261–
1266 (2016).
44. S. Oyarzun, A. K. Nandy, F. Rortais et al., Evidence for spin-to-charge conver-
sion by Rashba coupling in metallic states at the Fe/Ge(111) interface, Nat.
Commun. 7, 873 (2016).
45. I. M. Miron, G. Gaudin, S. Auffret et al., Current-driven spin torque induced
by the Rashba effect in a ferromagnetic metal layer, Nat. Mater. 9, 230–234
(2010).
46. L. Liu, O. J. Lee, T. J. Gudmundsen, D. C. Ralph, and R. A. Buhrman, Current-
induced switching of perpendicularly magnetized magnetic layers using spin
torque from the spin Hall effect, Phys. Rev. Lett. 109, 096602 (2012).
47. K. Garello, I. M. Miron, C. O. Avci et al., Symmetry and magnitude of spin-
orbit torques in ferromagnetic heterostructures, Nat. Nanotechnol. 8, 587–593
(2013).
48. O. Krupin, G. Bihlmayer, K. Starke et al., Rashba effect at magnetic metal sur-
faces, Phys. Rev. B 71, 201403 (2005).
49. G. Bihlmayer, Y. Koroteev, Y. M. Echenique et al., The Rashba-effect at metallic
surfaces, Surf. Sci. 600, 3888–3891 (2006).
50. A. Matos-Abiague and J. Fabian, Tunneling anomalous and spin Hall effects,
Phys. Rev. Lett. 115, 056602 (2015).
51. T. H. Dang, H. Jaffrès, T. L. Hoai Nguyen, and H.-J. Drouhin, Giant forward-
scattering asymmetry and anomalous tunnel Hall effect at spin-orbit-split and
exchange-split interfaces, Phys. Rev. B 92, 060403(R) (2015); T. Huong Dang,
D. Quang To, E. Erina, H. Jaffrès, V. I. Safarov, H. Jaffres, and H.-J. Drouhin,
Theory of the anomalous tunnel Hall Effect at ferromagnet-semiconductor
junctions, J. Magn. Magn. Mater. 459, 37–42 (2018).
52. B. Scharf, A. Matos-Abiague, J.-E. Han, E. M. Hankiewicz, and I. Žutić,
Tunneling planar hall effect in topological insulators: Spin valves and ampli-
fiers, Phys. Rev. Lett. 117, 166806 (2016).
53. C. O. Avci, K. Garello, A. Ghosh, M. Gabureac, S. F. Alvarado, and P.
Gambardella, Unidirectional spin Hall magnetoresistance in ferromagnet/
normal metal bilayers, Nat. Phys. 11, 570–575 (2015).
54. K. Yasuda, A. Tsukazaki, R. Yoshimi, K. S. Takahashi, M. Kawasaki, and
Y. Tokura, Large unidirectional magnetoresistance in a magnetic topological
insulator, Phys. Rev. Lett. 117, 127202 (2016).
55. M. Jamet, A. Barski, T. Devillers, and V. Poydenot, High-Curie-temperature
ferromagnetism in self-organized Ge1-xMn x nanocolumns, Nat. Mater. 5, 653–
659 (2006).
56. D. Chiba, F. Matsukura, and H. Ohno, Tunneling magnetoresistance in

(Ga,Mn)As-based heterostructures with a GaAs barrier, Physica E 21, 966–
969 (2004).
57. R. Mattana, M. Elsen, J.-M. George et al., Chemical profile and magnetore-
sistance of Ga1-xMn xAs/GaAs/AlAs/GaAs/Ga1-xMn xAs tunnel junctions, Phys.
Rev. B 71, 075206 (2005).
58. M. Elsen, H. Jaffrès, R. Mattana et al., Spin-polarized tunneling as a probe
of the electronic properties of Ga1-xMn xAs heterostructures, Phys. Rev. B
76, 144415 (2007).
59. J.-M. George and H. Jaffrès, Spin-transfer with GaMnAs/InGaAs/GaMnAs
tunnel junction, (unpublished).
60. S. H. Chun, S. J. Potashnik, K. C. Ku, P. Schiffer, and N. Samarth, Spin-polarized
tunneling in hybrid metal-semiconductor magnetic tunnel junctions, Phys.
Rev. B 66, 100408 (2002).
61. H. Saito, S. Yuasa, K. Ando, Y. Hamada, and Y. Suzuki, Spin-polarized tunnel-
ing in metal-insulator-semiconductor Fe/ZnSe/Ga1-xMn x As magnetic tunnel
diodes, Appl. Phys. Lett. 89, 232502 (2006).
62. H. Saito, A. Yamamoto, S. Yuasa, and K. Ando, High tunneling magnetoresis-
tance in Fe/GaO x/Ga1-xMn xAs with metal/insulator/semiconductor structure,
Appl. Phys. Lett. 93, 172515 (2008).
63. S. Ohya, N. Hai, Y. Mizuno, and M. Tanaka, Quantum-size effect and
Tunneling Magnetoresistance in ferromagnetic-semiconductor quantum het-
erostructure, Phys. Rev. B 75, 155328 (2007).
64. S. Ohya, N. Hai, and M. Tanaka, Tunneling magnetoresistance in GaMnAs/
AlAs/InGaAs/AlAs/GaMnAs double-barrier magnetic tunnel junctions,
Appl. Phys. Lett. 87, 012105 (2005).
65. S. Ohya, I. Muneta, N. Hai, and M. Tanaka, valence-band structure of ferro-
magnetic semiconductor GaMnAs studied by spin-dependent resonant tun-
neling spectroscopy, Phys. Rev. Lett. 104, 167204 (2010).
66. C. Gould, C. Ruster, T. Jungwirth et al., Tunneling anisotropic magnetore-
sistance: A spin-valve-like tunnel magnetoresistance using a single magnetic
layer, Phys. Rev. Lett. 93, 117203 (2004).
67. C. Ruster, C. Gould, T. Jungwirth et al., Very Large Tunneling anisotropic
magnetoresistance of a (Ga,Mn)As/GaAs/(Ga,Mn)As stack, Phys. Rev. Lett.
94, 027203 (2005).
68. H. Saito, S. Yuasa, and K. Ando, Origin of the tunnel anisotropic magnetore-
sistance in Ga1-x Mn xAs/ZnSe/Ga1-x Mn xAs magnetic tunnel junctions of II–
VI/III–V heterostructures, Phys. Rev. Lett. 95, 086604 (2005).
69. R. Giraud, M. Gryglas, L. Thevenard, A. Lemaitre, and G. Faini, Voltage-
controlled tunneling anisotropic magnetoresistance of a ferromagnetic p++-
(Ga,Mn)As/n+-GaAs Zener-Esaki diode, Appl. Phys. Lett. 87, 242505 (2005).
70. A. Giddings, M. Khalid, T. Jungwirth et al., Large tunneling anisotropic mag-
netoresistance in (Ga,Mn)As nanoconstrictions, Phys. Rev. Lett. 94, 127202
(2005).
71. M. Elsen, H. Jaffrès, R. Mattana et al., Exchange-mediated anisotropy of
(Ga,Mn)As valence-band probed by resonant tunneling spectroscopy, Phys.
Rev. Lett. 99, 127203 (2007).
72. A. Matos Abiague and J. Fabian, Anisotropic tunneling magnetoresistance
and tunneling anisotropic magnetoresistance: Spin-orbit coupling in magnetic
tunnel junctions, Phys. Rev. B 79, 155303 (2009).
73. A. Matos Abiague, M. Gmitra, and J. Fabian, Angular dependence of the tunnel-
ing anisotropic magnetoresistance in magnetic tunnel junctions, Phys. Rev. B
80, 045312 (2009).
74. J. Slonczewski, Current-driven excitations of magnetic multilayers, J. Magn.
Magn. Mater. 159, L1–L7 (1996).
75. L. Berger, Emission of spin waves by a magnetic multilayer traversed by a cur-
rent, Phys. Rev. B 54, 9353–9358 (1996).
References 239
76. A. Fert, A. Barthelemy, J. Ben Youssef et al., Review of recent results on spin
polarized tunneling and magnetic switching by spin injection, Mater. Sci. Eng.
B 84, 1–9 (2001).
77. M. Stiles and J. Miltat, Spin Dynamics in Confined Magnetic Structures III,
B. Hillebrands and A. Thiaville (Eds.), Springer, vol. 101, pp 225–308 (2006).
78. D. C. Ralph and M. D. Stiles, Spin transfer torques, J. Magn. Magn. Mater. 30,
1190–1216 (2008).
79. D. Chiba, Y. Sato, T. Kita, F. Matsukura, and H. Ohno, Current-driven magne-
tization reversal in a ferromagnetic semiconductor (Ga,Mn)As/GaAs/(Ga,Mn)
As tunnel junction, Phys. Rev. Lett. 93, 216602 (2004).
80. M. Elsen, O. Boulle, J.-M. George et al., Spin transfer experiments on (Ga,Mn)
As/(In,Ga)As/ (Ga,Mn)As tunnel junctions, Phys. Rev. B 73, 035303 (2006).
81. J. Okabayashi, M. Watanabe, H. Toyao, T. Yamagushi, and J. Yoshino, Pulse-
width dependence in current-driven magnetization reversal using GaMnAs-
based double-barrier magnetic tunnel junction, J. Supercond. Nov. Magn. 20,
443–446 (2007).
82. M. Watanabe, J. Okabayashi, H. Toyao, T. Yamagushi, and J. Yoshino, Current-
driven magnetization reversal at extremely low threshold current density in
(Ga,Mn)As-based double-barrier magnetic tunnel junctions, Appl. Phys. Lett.
92, 082506 (2008).
83. M. Abolfath, T. Jungwirth, J. Brum, and A. H. MacDonald, Theory of magnetic
anisotropy in III1-xMn xV ferromagnets, Phys. Rev. B 63, 054418 (2001).
84. J.-M. Luttinger and W. Kohn, Motion of electrons and holes in perturbed peri-
odic fields, Phys. Rev. 97, 869–883 (1955).
85. A. G. Petukhov, A. N. Chantis, and D. O. Demchenko, Resonant enhancement
of tunneling magnetoresistance in double-barrier magnetic heterostructures,
Phys. Rev. Lett. 89, 107205 (2002).
86. A. G. Petukhov, D. O. Demchenko, and A. N. Chantis, Electron spin polariza-
tion in resonant interband tunneling devices, Phys. Rev. B 68, 125332 (2003).
87. R. Wessel and M. Altarelli, Resonant tunneling of holes in double-barrier het-
erostructures in the envelope-function approximation, Phys. Rev. B 39, 12802–
12807 (1989).
88. D. L. Smith and C. Mailhiot, Theory of semiconductor superlattice electronic
structure, Rev. Mod. Phys. 62, 173–234 (1990).
89. M. Jullière, Tunneling between ferromagnetic films, Phys. Lett. A 54, 225–226
(1975).
Rev. Lett. 74, 3273–3277 (1995).
91. S. S. P. Parkin, C. Kaiser, A. Panchula et al., Giant tunneling magnetoresistance
at room temperature with MgO(100) tunnel barriers, Nat. Mater. 3, 862–867
(2004).
92. S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Giant room-
temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel
junctions, Nat. Mater. 3, 868–871 (2004).
93. W. H. Butler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren, Spin-
dependent tunneling conductance of Fe/MgO/Fe sandwiches, Phys. Rev. B 63,
054416 (2001).
94. J. Mathon and A. Umerski, Theory of tunneling magnetoresistance of an epi-
taxial Fe/MgO/Fe(001) junction, Phys. Rev. B 63, 220403 (2001).
95. L. Brey, C. Tejedor, and J. Fernandez-Rossier, Tunnel magnetoresistance in
GaMnAs: Going beyond Julliere formula, Appl. Phys. Lett. 85, 1996 (2004).
96. L. Brey, J. Fernandez-Rossier, and C. Tejedor, Spin depolarization in the trans-
port of holes across Ga1-xMn xAs/GayAl1-y As/p-GaAs, Phys. Rev. B 70, 235334
(2004).
97. P. Krstajic and F. M. Peeters, Spin-dependent tunneling in diluted magnetic
semiconductor trilayer structures, Phys. Rev. B 72, 125350 (2005).
98. R. Mattana, J.-M. George, H. Jaffrès et al., Electrical detection of spin accumu-
lation in a p-Type GaAs quantum well, Phys. Rev. Lett. 90, 166601 (2003).
99. N. O. Lipari and A. Baldereschi, Angular momentum theory and localized
states in solids. Investigation of shallow acceptor states in semiconductors,
Phys. Rev. Lett. 25, 1660–1664 (1970).
100. A. Baldereschi and N. O. Lipari, Spherical model of shallow acceptor states in
semiconductors, Phys. Rev. B 8, 2697–2709 (1973).
101. A. Baldereschi and N. O. Lipari, Cubic contributions to the spherical model of
shallow acceptor states, Phys. Rev. B 9, 1525–1539 (1973).
102. J. M. George, H. Jaffrès, R. Mattana et al., Electrical spin detection in GaMnAs-
based tunnel junctions: Theory and experiments, Mol. Phys. Rep. 40, 23–33
(2004).
103. V. Garcia, H. Jaffrès, M. Eddrief, M. Marangolo, V. H. Etgens, and J.-M. George,
Resonant tunneling magnetoresistance in MnAs/III–V/MnAs junctions, Phys.
Rev. B 72, 081303(R) (2005).
104. V. Garcia, H. Jaffrès, J.-M. George, M. Marangolo, M. Eddrief, and V. H. Etgens,
Spectroscopic measurement of spin-dependent resonant tunneling through a
3D disorder: The case of MnAs/GaAs/MnAs junctions, Phys. Rev. Lett. 97,
246802 (2006).
105. G. Xiang, B. L. Sheu, M. Zhu, P. Schiffer, and N. Samarth, Noncollinear spin
valve effect in ferromagnetic semiconductor trilayers, Phys. Rev. B 76, 035324
(2007).
106. M. Ciorga, A. Einwanger, U. Wurstbauer, D. Schuh, W. Wegscheider, and D.
Weiss, Electrical spin injection and detection in lateral all-semiconductor
devices, Phys. Rev. B 79, 165321 (2009).
107. P. Van Dorpe, Z. Liu, W. Van Roy et al., Very high spin polarization in GaAs
by injection from a (Ga,Mn)As Zener diode, Appl. Phys. Lett. 84, 3495–3497
(2004).
108. M. Kohda, T. Kita, Y. Ohno, F. Matsukura, and H. Ohno, Bias voltage depen-
dence of the electron spin injection studied in a three-terminal device based
on a (Ga,Mn)As/n+-GaAs Esaki diode, Appl. Phys. Lett. 89, 012103 (2006).
109. P. Van Dorpe, W. Van Roy, J. De Boeck et al., Voltage-controlled spin injection
in a (Ga,Mn)As/(Al,Ga)As Zener diode, Phys. Rev. B 72, 205322 (2005).
110. J. Masek, F. Maca, J. Kudrnovsky et al., Microscopic analysis of the valence
band and impurity band theories of (Ga,Mn)As, Phys. Rev. Lett. 105, 227202
(2010).
111. M. Linnarsson, E. Janzén, B. Monemar, M. Kleverman, and A. Thilderkvist,
Electronic structure of the GaAs:MnGa center, Phys. Rev. B 55, 6938–6944
(1997).
112. B. E. Larsson, K. C. Hass, H. Ehrenreich, and A. E. Carlsson, Theory of exchange
interactions and chemical trends in diluted magnetic semiconductors, Phys.
Rev. B 37, 4137–4154 (1988).
113. J. Okabayashi, A. Kimura, O. Rader et al., Core-level photoemission study of
Ga1-xMn xAs, Phys. Rev. B 58, R4211–R4214 (1998).
114. F. Matsukura, H. Ohno, A. Shen, and Y. Sugarawa, Transport properties and
origin of ferromagnetism in (Ga,Mn)As, Phys. Rev. B 57, R2037–R2040 (1998).
115. S. J. Potashnik, K. C. Ku, R. Mahendiran et al., Saturated ferromagnetism and
magnetization deficit in optimally annealed Ga1-xMn xAs, Phys. Rev. B 66,
012408 (2002).
116. R. P. Campion, K. W. Edmonds, L. X. Zhao et al., The growth of GaMnAs
films by molecular beam epitaxy using arsenic dimers, J. Cryst. Growth 251,
311–316 (2003).
117. A. Richardella, P. Roushan, S. Mack, et al., Visualizing critical correlations near
the metal-insulator transition in Ga1-xMn xAs, Science 327, 665–669 (2010).
118. B. C. Chapler, R. C. Myers, S. Mack et al., Infrared probe of the insulator-
to-metal transition in Ga1-xMn xAs and Ga1-x Be xAs, Phys. Rev. B 84, 081203
(2011).
References 241
119. A. M. Nazmul, T. Amemiya, Y. Shuto, S. Sugahara, and M. Tanaka, High tem-

perature ferromagnetism in GaAs-based heterostructures with Mn-δ-doping,
Phys. Rev. Lett. 95, 017201 (2005).
120. P. W. Anderson, Antiferromagnetism. Theory of superexchange interaction,
Phys. Rev. 79, 350–356 (1950).
121. P. W. Anderson, Localized magnetic states in metals, Phys. Rev. 124, 41–53
(1961).
122. P. W. Anderson and H. Hasegawa, Considerations on double exchange, Phys.
Rev. 100, 675–681 (1955).
123. J. K. Glasbrenner, I. Žutić, and I. I. Mazin, Theory of Mn-doped II–II–V semi-
conductors, Phys. Rev. B 90, 140403(R) (2014).
124. T. O. Strandberg, C. M. Canali, and A. H. MacDonald, Magnetic interactions
of substitutional Mn pairs in GaAs, Phys. Rev. B 81, 054401 (2010).
125. A. W. Rushforth, N. R. S. Farley, R. P. Campion et al., Compositional depen-
dence of ferromagnetism in (Al,Ga,Mn)As magnetic semiconductors, Phys.
Rev. B 78, 085209 (2008).
126. T. Jungwirth K. Y. Wang, J. Mašek et al., Prospects for high temperature fer-
romagnetism in (Ga,Mn)As semiconductors, Phys. Rev. B 72, 165204 (2005).
127. M. Wang, K. W. Edmonds, B. L. Gallagher, et al., High Curie temperatures at
low compensation in the ferromagnetic semiconductor (Ga,Mn)As, Phys. Rev.
B 87, 121301 (2013).
128. Y. Ohno, I. Arata, F. Matsukura, and H. Ohno, Valence band barrier at (Ga,Mn)
As/GaAs interfaces, Physica E 13, 521–524 (2002).
129. O. Thomas, O. Makarovsky, A. Patanè et al., Measuring the hole chemical
potential in ferromagnetic Ga1-xMn xAs/GaAs heterostructures by photoex-
cited resonant tunneling, Appl. Phys. Lett. 90, 082106 (2007).
130. T. Tsuruoka, N. Tachikawa, S. Ushioda, F. Matsukura, K. Takamura, and
H. Ohno, Local electronic structures of GaMnAs observed by cross-sectional
scanning tunneling microscopy, Appl. Phys. Lett. 81, 2800–2802 (2002).
131. B. I. Shklovskii and A. L. Efros, Electronic Properties of Doped Semiconductors,
Springer Series in Solid-State Sciences, vol. 45, Springer–Verlag, Berlin, pp.
1–65, 1984.
132. T. Jungwirth, J. Sinova, A. H. MacDonald et al., Character of states near the
Fermi level in (Ga,Mn)As: Impurity to valence band crossover, Phys. Rev. B 76,
125206 (2007).
133. K. S. Burch, D. B. Shrekenhamer, E. J. Singley et al., Impurity band conduc-
tion in a high temperature ferromagnetic semiconductor, Phys. Rev. Lett. 97,
087208 (2006).
134. K. Ando, H. Saito, K. C. Agarwal, M. C. Debnath, and V. Zayets, Origin of
the anomalous magnetic circular dichroism spectral shape in ferromagnetic
Ga1-xMn xAs: Impurity bands inside the band gap, Phys. Rev. Lett. 100, 067204
(2008).
135. J. M. Tang and M. E. Flatté, Magnetic circular dichroism from the impurity
band in III–V diluted magnetic semiconductors, Phys. Rev. Lett. 101, 157203
(2008).
136. D. Neumaier, M. Turek, U. Wurstbauer et al., All-electrical measurement of the
density of states in (Ga,Mn)As, Phys. Rev. Lett. 103, 087203 (2009).
137. T. Wojtowicz, W. L. Lim, X. Liu et al., Correlation of Mn lattice location,
free hole concentration, and curie temperature in ferromagnetic GaMnAs, J.
Supercond. Nov. Magn. 16, 41–44 (2003).
138. S. J. Potashnik, K. C. Ku, R. F. Wang et al., Coercive field and magnetization
deficit in Ga1-xMn xAs epilayers, J. Appl. Phys. 93, 6784–6786 (2003).
139. K. Y. Wang, M. Sawiki, K. W. Edmonds et al., Control of coercivities in (Ga,Mn)
As thin films by small concentrations of MnAs nanoclusters, Appl. Phys. Lett.
88, 022510 (2006).
140. M. Adell, J. Adell, L. Ilver, J. Kanski, and J. Sadowski, Photoemission study of

the valence band offset between low temperature GaAs and (GaMn)As, Appl.
Phys. Lett. 89, 172509 (2006).
141. S. Tiwari and D. J. Frank, Empirical fit to band discontinuities and barrier
heights in III–V alloy systems, Appl. Phys. Lett. 60, 630–632 (1992). If one
takes into account explicitly the stress between (Ga,Mn)As and (In,Ga)As, this
shift should correspond to the valence band offset between (Ga,Mn)As and the
LH component of (In,Ga)As, that is preferentially transmitted through thick
barriers.
142. S. Lodha, D. B. Janes, and N.-P. Chen, Unpinned interface fermi-level in
Schottky contacts to n-GaAs capped with low-temperature-grown GaAs;
experiments and modeling using defect state distributions, J. Appl. Phys. 93,
2772–2779 (2003).
143. M. Malfait, J. Vanacken, W. V. Roy, G. Borghs, and V. Moshchalkov, Low-
temperature annealing study of Ga1-xMn xAs: Magnetic properties and Hall
effect in pulsed magnetic fields, J. Magn. Magn. Mater, 290, 1387–1390 (2005).
144. M. Sawiki, D. Chiba, A. Korbecka et al., Experimental probing of the interplay
between ferromagnetism and localization in (Ga,Mn)As, Nat. Phys. 6, 22–25
(2010).
145. Y. Nishitani, D. Chiba, M. Endo et al., Curie temperature vs. hole concentra-
tion in field-effect structures of (Ga,Mn) As, Phys. Rev. B 81, 045208 (2010).
146. A. Koeder, S. Frank, W. Schoch et al., Curie temperature and carrier concen-
tration gradients in epitaxy-grown Ga1-xMn xAs layers, Appl. Phys. Lett. 82,
3278–3280 (2003).
147. K. Pappert, M. J. Schmidt, S. Humpfner et al., Magnetization-switched metal-
insulator transition in a (Ga,Mn)As tunnel device, Phys. Rev. Lett. 97, 186402
(2006).
148. K. I. Bolotin, F. Kuemmeth, and D. C. Ralph, Anisotropic magnetoresistance
and anisotropic tunneling magnetoresistance due to quantum interference in
ferromagnetic metal break junctions, Phys. Rev. Lett. 97, 127202 (2006).
149. S. F. Shi, K. I. Bolotin, F. Kuemmeth, and D. C. Ralph, Temperature dependence
of anisotropic magnetoresistance and atomic rearrangements in ferromag-
netic metal break junctions, Phys. Rev. B 76, 184438 (2007).
150. M. Viret, M. Gabureac, F. Ott et al., Giant anisotropic magneto-resistance in
ferromagnetic atomic contacts, Eur. Phys. J. B 51, 1–4 (2006).
151. A. Bernand-Mantel, P. Seneor, K. Bouzehouane et al., Nat. Phys. 5, 920–924
(2009).
152. T. Uemura, Y. Imai, M. Harada, K. Matsuda, and M. Yamamoto, Tunneling
anisotropic magnetoresistance in epitaxial CoFe/n-GaAs junctions, Appl.
Phys. Lett. 94, 182502 (2009).
153. M. Wimmer, M. Lobenhofer, J. Moser, et al., Orbital effects on tunneling aniso-
tropic magnetoresistance in Fe/GaAs/Au junctions, Phys. Rev. B 80, 121301(R)
(2009).
154. J. Moser, A. Matos-Abiague, D. Schuh, W. Wegscheider, J. Fabian, and D. Weiss,
Tunneling anisotropic magnetoresistance and spin-orbit coupling in Fe/GaAs/
Au tunnel junctions, Phys. Rev. Lett. 99, 056601 (2007).
155. L. Gao, X. Jiang, S. H. Yang et al., Bias voltage dependence of tunneling aniso-
tropic magnetoresistance in magnetic tunnel junctions with MgO and Al2O3
tunnel barriers, Phys. Rev. Lett. 99, 226602 (2007).
156. B. G. Park, J. Wunderlich, D. A. Williams et al., Tunneling anisotropic mag-
netoresistance in multilayer-(Co/Pt)/AlOx/Pt structures, Phys. Rev. Lett. 100,
087204 (2008).
157. A. N. Chantis, K. D. Belashchenko, E. Y. Tsymbal, and M. van Schilfgaarde,
Tunneling anisotropic magnetoresistance driven by resonant surface states:
First-principles calculations on an Fe(001) surface, Phys. Rev. Lett. 98, 046601
(2007).
References 243
158. M. Zhu, M. J. Wilson, P. Mitra, P. Schiffer, and N. Samarth, Quasireversible

magnetoresistance in exchange-spring tunnel junctions, Phys. Rev. B 78,
195307 (2008).
159. M. J. Schmidt, K. Pappert, C. Gould, J. Schmidt, R. Oppermann, and L. W.
Molenkamp, Bound-hole states in a ferromagnetic (Ga,Mn)As environment,
Phys. Rev. B 76, 035204 (2007).
160. J. M. Tang and M. E. Flatté, Multiband tight-binding model of local magnetism
in Ga1-xMn xAs, Phys. Rev. Lett. 92, 047201 (2004).
161. J. M. Tang and M. E. Flatté, Spin-orientation-dependent spatial structure of
a magnetic acceptor state in a zinc-blende semiconductor, Phys. Rev. B 72,
161315(R) (2005).
162. P. Sankowski, P. Kacman, J. Majewski, and T. Dietl, Spin-dependent tunneling
in modulated structures of (Ga,Mn)As, Phys. Rev. B 75, 045306 (2007).
163. L. Thevenard, L. Largeau, O. Mauguin, A. Lemaitre, and B. Theys, Tuning
the ferromagnetic properties of hydrogenated GaMnAs, Appl. Phys. Lett. 87,
182506 (2005).
164. H. Ohno, N. Akira, F. Matsukura et al., Spontaneous splitting of ferromagnetic
(Ga, Mn)As valence band observed by resonant tunneling spectroscopy, Appl.
Phys. Lett. 73, 363–365 (1998).
165. R. K. Hayden, D. K. Maude, L. Eaves et al., Probing the hole dispersion curves
of a quantum well using resonant magnetotunneling spectroscopy, Phys. Rev.
Lett. 66, 1749–1752 (1991).
166. M. Tran, J. Peiro, H. Jaffrès, J.-M. George, and A. Lemaitre, Magnetization-
controlled conductance in (Ga,Mn)As-based resonant tunneling devices,
Appl. Phys. Lett. 95, 172101 (2009).
167. J. Wunderlich, T. Jungwirth, B. Kaestner et al., Coulomb blockade anisotropic
magnetoresistance effect in a (Ga,Mn)As single-electron transistor, Phys. Rev.
Lett. 97, 077201 (2006).
168. H. T. Dang, E. Erina, H. T. L. Nguyen, H. Jaffres, and H.-J. Drouhin, Spin-
orbit assisted chiral-tunneling at semiconductor tunnel junctions: Study
with advanced 30-band k.p methods, Spintronics IX, Proc. SPIE 9931, 993127
(2016), doi:10.1117/12.2238796.
169. J.-M. Jancu, R. Scholz, E. A. de Andrada e Silva, and G. C. La Rocca, Atomistic
spin-orbit coupling and k. p parameters in III–V semiconductors, Phys. Rev. B
72, 193201 (2005).
170. S. Richard, F. Aniel, and G. Fishman, Energy-band structure of Ge, Si, and
GaAs: A thirty-band k·p method, Phys. Rev. B 70, 235204 (2004).
171. J. A. Katine, F. J. Albert, R. A. Buhrman, E. B. Myers, and D. C. Ralph, Current-
driven magnetization reversal and spin-wave excitations in Co/Cu/Co pillars,
Phys. Rev. Lett. 84, 3149–3152 (2000).
172. S. I. Kiselev, J. Sankey, I. Krivorotov et al., Microwave oscillations of a nano-
magnet driven by a spin-polarized current, Nature 425, 380–383 (2003).
173. J. Grollier, V. Cros, A. Hamzic et al., Spin-polarized current induced switching
in Co/Cu/Co pillars, Appl. Phys. Lett. 78, 3663 (2001).
174. J. Grollier, V. Cros, H. Jaffrès et al., Field dependence of magnetization reversal
by spin transfer, Phys. Rev. B 67, 174402 (2003).
175. J. Sun, D. Monsma, D. Abraham, M. Rooks, and R. Koch, Batch-fabricated
spin-injection magnetic switches, Appl. Phys. Lett. 81, 2202 (2002).
176. S. Urazdhin, N. Birge, W. P. Pratt, and J. Bass, Switching current versus mag-
netoresistance in magnetic multilayer nanopillars, Appl. Phys. Lett. 84, 1516
(2004).
177. Y. Huai, F. Albert, P. Nguyen, M. Pakala, and T. Valet, Observation of spin-
transfer switching in deep submicron-sized and low-resistance magnetic tun-
nel junctions, Appl. Phys. Lett. 84, 3118–3121 (2004).
178. Y. Liu, Z. Zhang, P. Freitas, and J. L. Martins, Current-induced magnetization
switching in magnetic tunnel junctions, Appl. Phys. Lett. 82, 2871 (2003).
179. G. D. Fuchs, N. C. Emley, I. N. Krivorotov et al., Spin-transfer effects in

nanoscale magnetic tunnel junctions, Appl. Phys. Lett. 85, 1205–1208 (2005).
180. J. Slonczewski, Currents, torques, and polarization factors in magnetic tunnel
junctions, Phys. Rev. B 71, 024411 (2005).
181. J. A. Katine and E. E. Fullerton, Device implications of spin-transfer torques,
J. Magn. Magn. Mater. 320, 1217–1226 (2008).
182. D. Chiba, F. Matsukura, and H. Ohno, Electrical magnetization reversal in fer-
romagnetic III–V semiconductors, J. Phys. D Appl. Phys. 39, R215–R225 (2006).
183. A. Chernyshov, M. Overby, X. Liu, J. K. Furdyna, Y. Lyanda-Geller, and L. P.
Rokhinson, Evidence for reversible control of magnetization in a ferromagnetic
material by means of spin-orbit magnetic field, Nat. Phys. 5, 656–659 (2009).
184. M. Tsoi, A. G. M. Jansen, J. Bass et al., Excitation of a magnetic multilayer by
an electric current, Phys. Rev. Lett. 80, 4281–4284 (1998).
185. M. D. Stiles and A. Zangwill, Anatomy of spin-transfer torque, Phys. Rev. B 66,
014407 (2002).
186. A. Kalitsov, I. Theodonis, N. Kioussis, M. Chshiev, W. H. Butler, and A.
Vedyayev, Spin-polarized current-induced torque in magnetic tunnel junc-
tions, J. Appl. Phys. 99, 08G501 (2006).
187. I. Theodonis, N. Kioussis, A. Kalitsov, M. Chshiev, and W. H. Butler, Anomalous
bias dependence of spin torque in magnetic tunnel junctions, Phys. Rev. Lett.
97, 237205 (2006).
188. A. Kalitsov, M. Chshiev, I. Theodonis, N. Kioussis, and W. H. Butler, Spin-
transfer torque in magnetic tunnel junctions, Phys. Rev. B 79, 174416 (2009).
189. Y. H. Tang, N. Kioussis, A. Kalitsov et al., Controlling the nonequilibrium
interlayer exchange coupling in asymmetric magnetic tunnel junctions, Phys.
Rev. Lett. 103, 057206 (2009).
190. H. Kubota, A. Fukushima, K. Yakushiji et al., Quantitative measurement of
voltage dependence of spin-transfer torque in MgO-based magnetic tunnel
junctions, Nat. Phys. 4, 37–41 (2007).
191. J. C. Sankey, Y. T. Cui, J. Z. Sun et al., Measurement of the spin-transfer-torque
vector in magnetic tunnel junctions, Nat. Phys., 4, 67–71 (2007).
192. J. Sinova, T. Jungwirth, X. Liu et al., Magnetization relaxation in (Ga,Mn)As
ferromagnetic semiconductors, Phys. Rev. B 69, 085209 (2004).
193. S. Zhang, P. M. Levy, A. Marley, and S. S. P. Parkin, Quenching of magneto-
resistance by hot electrons in magnetic tunnel junctions, Phys. Rev. Lett. 79,
3744–3747 (1997).
194. A. M. Bratkovsky, Assisted tunneling in ferromagnetic junctions and half-
metallic oxides, Appl. Phys. Lett. 72, 2334–2336 (1998).
195. P. M. Levy and A. Fert, Spin transfer in magnetic tunnel junctions with hot
electrons, Phys. Rev. Lett. 97, 097205 (2006).
196. J. Schneider, U. Kaufmann, W. Wilkening, M. Baeumler, and F. Kohl, Electronic
structure of the neutral manganese acceptor in gallium arsenide, Phys. Rev.
Lett. 59, 240–243 (1987).
197. A. O. Govorov, Optical probing of the spin state of a single magnetic impurity
in a self-assembled quantum dot, Phys. Rev. B 70, 035321 (2004).
198. G. Dresselhaus, Spin-orbit coupling effects in zinc blende structures, Phys. Rev.
100, 580–586 (1955).
199. E. O. Kane, Band structure of indium antimonide, J. Phys. Chem. Solids 1,
249–261 (1957).
200. E. O. Kane, The k.p method, in Semiconductors and Semimetals, R. K.
Willardson and A. C. Beer (Eds.), Academic Press, New York, P. 1, 1966; E. O.
Kane, Energy band theory, in Handbook on Semiconductors, T. S. Moss and W.
Paul (Eds.), North-Holland, Amsterdam, P. 1, 1982.
201. G. F. Koster, J. O. Dimmock, R. G. Wheeler, H. Statz, Properties of the Thirty-
Two Point Groups, M.I.T. Press, Cambridge, MA, 1963.
202. R. Winkler, Spin-Orbit Coupling Effects in Two-Dimensional Electron and Hole
Systems, Springer–Verlag, Berlin, 2003.
References 245
203. J. M. Luttinger, Quantum theory of cyclotron resonance in semiconductors:

General theory, Phys. Rev. 102, 1030–1041 (1956).
204. M. Cardona, N. E. Christensen, and G. Fasol, Relativistic band structure
and spin-orbit splitting of zinc-blende-type semiconductors, Phys. Rev. B 38,
1806–1827 (1988).
205. E. L. Ivchenko, A. Yu. Kaminski, and U. Rössler, Heavy-light hole mix-
ing at zinc-blende (001) interfaces under normal incidence, Phys. Rev. B 54,
5852–5859 (1996).
206. M. V. Durnev, M. M. Glazov, and E. L. Ivchenko, Spin-orbit splitting of
valence subbands in semiconductor nanostructures, Phys. Rev. B 89, 075430
(2014).
207. D. Rideau, M. Feraille, L. Ciampolini et al., Strained Si, Ge, and Si1-x Ge x alloys
modeled with a first-principles-optimized full-zone k.p method, Phys. Rev. B
74, 195208 (2006).
208. G. Fishman, Semi-Conducteurs: Les Bases de la théorie k.p, Les Editions de
l’Ecole Polytechnique, Palaiseau, Paris, France, 2010.
209. S. Boyer-Richard, F. Raouafi, A. Bondi et al., 30-band k.p method for quantum
semiconductor heterostructures, Appl. Phys. Lett. 98, 251913 (2011).
210. N. A. Cukaric, M. Z Tadic, B. Partoens, and F. M. Peeters, 30-band k.p model
of electron and hole states in silicon quantum wells, Phys. Rev. B 88, 205306
(2013).
211. B. A. Foreman, Connection rules versus differential equations for envelope
functions in abrupt heterostructures, Phys. Rev. Lett 80, 3823–3826 (1998).
212. K. L. Kolokolov, J. Li, and C. Z. Ning, k.p Hamiltonian without spurious-state
solutions, Phys. Rev. B 68, 161308(R) (2003).
213. M. Cardona and F. Pollak, Energy-band structure of germanium and silicon:
The k·p method, Phys. Rev. 142, 530–543 (1966).
214. N. Cavassilas, F. Aniel, K. Boujdaria, and G. Fishman, Energy-band structure
of GaAs and Si: A sps* k.p method, Phys. Rev. B 64, 115207 (2001).
215. R. Eppenga, M. F. H. Schuurmans, and S. Colak, New k.p theory for GaAs/Ga1-x
Al xAs-type quantum wells, Phys. Rev. B 36, 1554–1564 (1987).
216. D. Y. K. Ko and J. C. Inkson, Matrix method for tunneling in heterostructures:
Resonant tunneling in multilayer systems, Phys. Rev. B 38, 9945–9951 (1988).
217. O. Krebs and P. Voisin, Giant optical anisotropy of semiconductor hetero-
structures with no common atom and the quantum-confined Pockels effect,
Phys. Rev. Lett. 77, 1829–1832 (1996).
218. S. Cortez, O. Krebs, and P. Voisin, Breakdown of rotational symmetry at semi-
conductor interfaces: a microscopic description of valence subband mixing,
Eur. Phys. J. B 21, 241–250 (2001).
219. J. Slonczewski, Conductance and exchange coupling of two ferromagnets sep-
arated by a tunneling barrier, Phys. Rev. B 39, 6995–7002 (1989).
220. H. Jaffrès, D. Lacour, F. Nguyen Van Dau et al., Angular dependence of the tun-
nel magnetoresistance in transition-metal-based junctions, Phys. Rev. B 64,
064427 (2001).
221. A. Braatas, G. E. W. Bauer, and P. J. Kelly, Non-collinear magnetoelectronics,
Phys. Rep. 427, 157–255 (2006).
222. J.-C. Rojas-Sanchez, N. Reyren, P. Laczkowski et al., Spin pumping and inverse
spin hall effect in platinum: The essential role of spin-memory loss at metallic
interfaces, Phys. Rev. Lett. 112, 106602 (2014).
223. V. P. Amin and M. D. Stiles, Spin transport at interfaces with spin-orbit cou-
pling: Phenomenology, Phys. Rev. B 94, 104420 (2016).
5
Spin Transport
in Organic
Semiconductors

5.2 Organic Semiconductors: General Background 249
5.3 Electrical Operation: Injection 251
5.4 Electrical Operation: Tunneling 256
5.5 Alternative Approaches 257
5.6 Interface Effects 259
5.7 Multifunctional Effects 260
5.8 Theoretical Models 261
5.9 Conclusions 263
References 264
247
248 Chapter 5. Spin Transport in Organic Semiconductors
5.1 INTRODUCTION
Recently, spintronics has expanded its choice of materials to the organic
semiconductors (OSCs) [1–3]. OSCs offer unique properties toward their
integration with spintronics, the main one being the weakness of the spin-
scattering mechanisms in OSC. Spin–orbit coupling is very small in most
OSCs; carbon has a low atomic number (Z) and the strength of the spin–
orbit interaction (SOI) is in general proportional to Z4 [4]. Typical SOI values
in OSC are less than about a few meV [5], well below any other characteristic
energy, including main vibrational modes. Small spin–orbit coupling implies
that the spin polarization of the carriers could be maintained for a very long
time. Indeed, spin relaxation times in excess of 10 µs have been detected by
various resonance techniques [6, 7], and these values compare very favorably
(at least 103 times larger) with those obtained with high-performing inor-
ganic semiconductors, such as GaAs [8].
The field of OSCs nowadays combines considerable achievements in
research and applications with challenging expectations. Over decades,
the interest toward OSCs was mainly sustained by their exceptional
optical properties, as the multicolor “offer” was backed by an enormous
choice of available or easily modifiable molecules as well as by significant
luminescence intensity. Indeed, these optical properties prompted a real
breakthrough in the optoelectronics field where, along with the achieve-
ment of a deep basic knowledge, a variety of important applications have
been developed. Display products based on hybrid light-emitting diodes
with organic emitters (OLEDs) have become available to consumers, and
organic photovoltaic devices are challenging existing commercial appli-
cations. These accomplishments are advancing a future strong demand
for high quality control and guiding circuitry based on novel hybrid
organic–inorganic devices. Organic field effect transistors (OFETs), in
which considerable improvements have already been achieved, together
with various elements based on magnetic and electrical resistance switch-
ing are foreseen as one of the most important candidates for these newly
emerging applications.
Organic molecules are also considered as a promising contender for the
future “beyond CMOS devices,” offering 1–10 nm-sized building blocks with
rich optical, electrical, and magnetic properties. OSCs have become a requi-
site of various roadmaps and research agendas, although many serious dif-
ficulties have still to be circumvented in order to meet these challenges. One
of the biggest obstacles for commercial applications is the resistance contact
problem [9], namely, the way to insert and extract electrically driven informa-
tion, interfacing fragile organic blocks with hard metallic or oxide electrodes.
However, it has to be noted, on the other hand, that the hybrid nature of most
organic-based devices also has positive aspects, providing routes for large
and fundamental modifications of the properties of the OSC, especially in
the interface region.
5.2 Organic Semiconductors: General Background 249
5.2 ORGANIC SEMICONDUCTORS:
GENERAL BACKGROUND
OSCs combine a strong intramolecular covalent carbon–carbon bonding
with a weak van der Waals interaction between molecules. This combina-
tion of different interactions generates unusual materials whose optical
properties are very similar to those of their constituent molecules, whereas
their transport properties are instead firmly governed by the intermolecular
interactions.
As a general rule, there are two very different classes of OSCs: small
molecules and polymers. These two cases differ not only on their physical
and chemical properties, but also on their technological processing. Small-
molecule OSCs are rigid objects of about 1 nm size, with very different shapes
and symmetries, but able to form highly ordered van der Waals layers and
polycrystalline films. The small-molecule film growth involves either high
or ultrahigh vacuum molecular beam deposition. A few of the main repre-
sentatives of this kind of materials are the classes of oligothiophenes (4T, 6T,
and others), metal chelates (Alq3, Coq3, and others), metal-phthalocyanines
(CuPc, ZnPc, and others), acenes (tetracene, pentacene, and others), rubrene,
and many more. Polymers, on the other hand, are long chains formed from a
given monomer, and usually feature connection of oligo-segments of various
lengths leading to a random conjugation length distribution. Polymer films
are strongly disordered and generally cannot be produced in UHV condi-
tions; instead, totally different processing technologies have to be applied,
including ink-jet printing, spin coating, drop casting, and other extremely
cheap methods. This potentially inexpensive fabrication constitutes one
of the main technological advantages of the polymers. However, these
growth approaches generally preclude a real interface control, as the elec-
trode surface is not as clean as in UHV conditions and cannot be directly
characterized by such important surface techniques as x-ray photoemission
spectroscopy (XPS), ultraviolet photoemission spectroscopy (UPS), x-ray
magnetic dichroism (XMCD), and others.
The electrical properties of OSC are usually depicted by two main com-
ponents: charge injection and charge transport. The charge injection process
from a metal into an OSC is conceptually different from the case of metals/
inorganic semiconductors, mainly because of the vanishingly low density of
intrinsic carriers—about 1010 cm−3—in OSC. The external electrodes provide
the electrical carriers responsible for the transport, as the OSC molecules
can easily accommodate an extra charge, although this charge modifies con-
siderably the molecular spatial conformation and creates strong polaronic
effects. Carrier injection into OSC is best described [10] in terms of thermally
and field-assisted charge tunneling across the inorganic–organic interface,
followed by carrier diffusion into the bulk of OSC. The carriers propagate
via a random site-to-site hopping between pseudo-localized states distrib-
uted within approximately a 0.1 eV energy interval. This interval describes
the spatial distribution of the localized levels and is not to be confused with
the bandwidth. OSCs are mainly families of π-conjugated molecules, which
combine a considerable carrier delocalization inside the molecules with a
weak van der Waals intermolecular interaction that limit considerably the
carrier mobility. The electron mean free path is usually of about a molecule
size, namely, about 1 nm, while standard mobilities are well below 1 cm2/V s.
Two conducting channels are usually considered active: lowest unoccupied
molecular level (LUMO) for n-type and highest occupied molecular level
(HOMO) for p-type carriers. However, defects and interface states have to be
considered depending on the material and its structural quality. These states
usually emerge via one of the most important characteristics of OSC—
trapping centers. Puzzlingly, in OSCs, the impurities do not induce addi-
tional carriers, as in inorganic semiconductors, but mainly generate trap
states. Energetically distributed shallow (about/below 0.1 eV) and deep
(about 0.5 eV, but even up to 1 eV) traps are known to characterize the elec-
trical properties (especially the mobility) of many OSCs, their role being
stronger in low-order materials like Alq3, polymers, etc.
The mobility of OSC has a strong dependence on the electric field, and
in the absence of traps, it is usually described by the so-called Poole–Frenkel
dependence [11]
 β 
µ( E ) = µ 0exp  (5.1)
 √ E 
where µ0 is the field-independent mobility. Including trapping compli-
cates the equations and numerical solutions for different cases have to be
calculated.
In order to complete this very brief description of electrical properties of
OSC, we would like to add two more concepts.
The first concept highlights the difference between the two main trans-
port regimes present for most organic-based devices: injection-limited cur-
rent (ILC) and space charge-limited current (SCLC). In the first case, the
interface resistance dominates the current–voltage curves and the flow-
ing current is basically independent on the device thickness. However, the
SCLC current (JSCLC) requires at least one ohmic electrode and corresponds
to the case when the injected carrier density reaches the charging capacity
of the material. This SCLC current is characterized by a strong thickness (d)
dependence, which for a trap-free OSC is given by [11]
µ( E )V 2
J SCLC ∼ (5.2)
d3
where V is the voltage across the device and, at that given voltage, the electric
field (E) value inside OSC depends as well on the thickness.
The second concept is the distinction between unipolar and bipolar
transport in OSCs (the later featuring significant recombination effects).
These two different transport regimes depend on the intrinsic properties of
the OSC as well as on the matching of the electrode work function with
either the LUMO or the HOMO levels. For more extensive reading on the
electrical properties, we suggest the excellent reviews of Coropceanu et al.
[12] and Brütting et al. [11].
The optical gap in many OSCs is about 2–3 eV, which conveniently fits
very well in the visible range. The simultaneous injection of electrons and
holes from two independent electrodes generates Frenkel excitons with a 3:1
triplet/singlet ratio. The S1 → S0 optical transition is usually the only one
allowed as a consequence of low spin–orbit coupling and Franck–Condon
selection. The former rules out the singlet–triplet mixing, while the latter
points out once again on the strong role of the polaronic effects in OSC [13,
14], which excludes the transitions where the equilibrium positions of the
nuclei in both the ground and the excited states would be identical.
5.3 ELECTRICAL OPERATION: INJECTION

In the following discussion on the electrically operated organic spintronic
devices (OSPDs), it is necessary to distinguish between two different cases:
tunneling devices characterized by zero residence time of charges/spins
in the organic barrier, and injection devices in which there is a net flow of
spin-polarized carriers directly into the electronic levels of the OSC. While
tunneling devices can feature large magnetoresistance (MR) values and are
useful as 0–1 switching elements or magnetic sensors, injection devices are
characterized by a finite lifetime of the generated spin polarization in OSC,
and offer such opportunities as spin manipulation and control of the exciton
statistics in OSC. These options may well open the way to magnetically con-
trolled OFETs or OLEDs.
In the description of the spin injection process into OSC, we will need
to mention numerous times two materials with exceptional importance
in the field: manganites, mostly La0.7Sr0.3MnO3 (LSMO), and Alq3 (Tris(8-
hydroxyquinoline) aluminum(III)). Importantly, about half of all the OSC
spintronic devices available in the literature involve either one or both of
these materials. This fact has some historical grounds, as the first successful
experiments dealt with these materials, but it also indicates some outstand-
ing spintronic properties of these materials, which have not been understood
and developed in detail yet.
La1−xSrxMnO3 (with 0.2 < x < 0.5) is a ferromagnetic material with 100%
spin-polarized carriers at T ≪ Tc. The Curie temperature is around 370 K in
bulk samples, while thin films have Tc around 330 K, changing with thick-
ness and deposition method. The carrier hopping between Mn3+ and Mn4+
ions is governed by a strong on-site Hund interaction (nearly 1 eV), which
makes La1−xSrxMnO3 an unusual metal, characterized by a narrow band
(∼1–2 eV) and small carrier density (1021–1022 cm−3), very much like the high-
temperature cuprate superconductors and some organic metals.
A critical issue for the spintronic applications of manganites is the
difference between their bulk and surface magnetic properties: the latter,
being responsible for the injected spin-polarized current, decays much faster
with temperature than the former. This leads to the fact that in the room
temperature region, the spin polarization is much smaller than the expected
magnetization ratio M(T)/M(0) [15]. Nevertheless, special surface treat-
ments have succeeded in improving the thin film surface properties, and
samples with reproducible XMCD signals at room temperature are currently
routinely obtained [16]. Well before raising of organic spintronics, LSMO
was already well-known for its spin injection properties, and has proved to
be highly successful in diverse inorganic spintronic devices, such as tunnel
junctions [17] and artificial grain boundaries devices [18].
Tris(8-hydroxyquinoline aluminum(III) or Alq3 is a popular OSC from
the early stages of xerographic and optoelectronics applications [19]. This
material might look somewhat of a strange choice for transport-oriented appli-
cations, as Alq3 has very low mobility values and a permanent electric dipole
[20, 21]. These two factors are, in principle, able to compromise spin trans-
port, but in spite of the expected problems with Alq3, strong spin-mediated
effects have been repeatedly reported by different groups.
The first experimental report on injection in OSPD featured a lateral
device that combined LSMO electrodes and OSC conducting channels
between 100 and 500 nm long (Figure 5.1a). Sexithiophene (6T), a rigid con-
jugated oligomer rod, a pioneer in organic thin film transistors [22], was
chosen for the spin transport channel. A strong magnetoresistive response
was recorded up to room temperature in 100 and 200 nm channels, and
was explained as a result of the conservation of the spin polarization of the
injected carriers. Using the time-of-flight approach, a spin relaxation time of
the order of 1 µs was found.
This work introduced OSCs as very appealing materials for long-
distance spin transport, although some important questions were left open.
For example, the experiment did not provide a straightforward demonstra-
tion that the MR was related to the magnetization of the electrodes, and
hence to its spin polarization, as no comparison between parallel and anti-
parallel magnetization of electrodes was available.
Following this report, an important step forward was the fabrication
of a vertical spin-valve device consisting of LSMO and cobalt electrodes,
sandwiching a thick (100–200 nm) layer of Alq3 [23]. The spin valve showed
MR up to a temperature around 200 K, and for voltages below 1 V. Unlike
standard spin-valve devices, in which lower resistance corresponds to paral-
lel electrode magnetization, the LSMO/Alq3/Co devices showed lower resis-
tance for an antiparallel magnetization configuration.
Starting from this work, the vertical geometry has acquired a prevailing
role in organic spintronics, mainly because of its technological simplicity
with respect to planar devices, requiring a substantial lithography involve-
ment. Several groups have confirmed the so-called “inverse spin-valve effect”
in LSMO/Alq3/Co devices [24–28]. Nevertheless, its nature has not yet been
clearly understood, although at least two different explanations have been
proposed [23, 26]. One of these models requires the injection of a spin-down-
oriented current from the cobalt electrode [23] (i.e. the spin polarization of
carriers antiparallel to the applied magnetic field), in a serious disagreement
with many experimental and theoretical reports [29, 30]. The second model
6T
LSMO LSMO
w = 70 – 500 nm
V
I
0.00
(a)
–0.05
MR (%)
Co
–0.10
Barrier
–0.15
LSMO
OS –100 –75 –50 –25 0 25 50 75 100
Substrate
(b) (c) Field (Oe)
FIGURE 5.1 (a) Spin lateral device presented by Dediu et al. [32]: two LSMO electrodes obtained by elec-
tron beam lithography were connected by an organic thin film (T6). The channel length ranged between 70
and 500 nm. (b) Vertical spin-valve device consisting of LSMO electrode as bottom electrode and cobalt as
the top one. The typical MR measurements are performed by applying a constant biasing voltage V across
the electrodes while measuring the resulting current as a function of an in-plane sweeping external mag-
netic field. (c) GMR loop of an LSMO (20 nm)/Alq3 (100 nm)/AlOX (2 nm)/Co (20 nm) vertical spin-valve device
measured at room temperature. The antiparallel configurations of the magnetizations correspond to the low
resistance state.
introduces a very narrow hopping channel in OSCs, connecting at some

given voltage the spin-up bands of one of the electrodes with the spin-down
bands of the opposite one [26]. While in principle feasible, this model can
be applied for the n-type carriers only and works mainly at voltages about
1 V or higher, that is, values at which spintronics effects in LSMO/Alq3/Co
devices generally vanish. This situation clearly indicates the limits of the
overall comprehension of the main laws governing both the spin injection
and transport in organic spintronics.
As explained above, Alq3 is characterized by its very low mobility: about
10 cm2/V s for electrons and less than 10−6 cm2/V s for holes [31]; while the
−5
full relevance of the mobility has yet to be understood for OSPD, the wide
use of Alq3 is probably due to the good-quality thin films that can be grown
on various ferromagnetic substrates by standard UHV evaporation.
Nevertheless, electrical detection of spin injection and transport has
been actively pursued in a number of other OSCs (see Table 5.1). Thus, α-
NPD (N,N-bis 1-naphtalenyl-N,N-bis phenylbenzidiane) and CVB (4, 4′-(Bis
(9-ethyl-3-carbazovinylene)-1,1′-biphenyl) [24] were demonstrated to behave
very similarly to Alq3 with LSMO and Co set of magnetic electrodes: inverse
spin-valve effect (around 10% at low temperatures) and fast decrease of the
MR with increasing temperature, vanishing between 210 and 240 K. The
same couple of electrodes was used to investigate spin injection in tetra-
phenylporphyrin (TTP)—the detected MR in this case was attributed,
TABLE 5.1
Main Properties of OSCs Investigated in Organic Spintronics.
Rubrene Pentacene
µ = 1 cm2/V s µ = 10–1 cm2/V s
Oligoacenes p-type p-type
Sexitiophene RRP3HT
Tiophenes µ = 10–1 cm2/V s C6H13 C6H13 µ = 10–1 cm2/V s
s s s p-type S p-type
s s s S S n
C6H13
NPD CVB
µ = 10–5 cm2/V s
Triarylamines p-type N µ = 10–3 cm2/V s
N
TPP
N
µ = 10–5 cm2/V s
Porphyrines n-type
NH HN
N
CuPC Alq3
N
Metal N N N –2 2
µ = 10 cm /V s O Al O N µ = 10–5 cm2/V s
complexes N Cu N
n-type O n-type
N N N N
however, to tunneling effects [28]. A planar geometry with two LSMO elec-
trodes, similar to the first publication in the field [32], provided about 30%
MR at low temperature in pentacene and BTQBT [33].
Moving from low to room temperature operation in vertical devices has
required an advanced control of the interface quality. Thus, room temperature
MR in LSMO/Alq3/Al2O3/Co devices was demonstrated [26] by improving
the quality of both injecting interfaces, especially the top one, in which a 2 nm
thick Al2O3 insulating layer was added (see Figure 5.1b and c). Similarly, about
1% room temperature MR across thick Alq3 and CuPc layers was achieved in
vertical devices with both metallic electrodes and excellent interface quality
(Figure 5.2c) [34]. Importantly, the substitution of the LSMO electrode by Fe
has removed the inversion of the spin-valve effect [34], indicating that inver-
sion is strongly caused by the manganite properties [26].
All the experiments described above were performed using UHV condi-
tions for the OSC growth. Spin-coated polymers such as regio-random and
regio-regular poly(3-hexylthiophene), (RRaP3HT and RRP3HT, respectively)
have also shown noticeable MR [27, 35]. In these reports, the spin-valve effect
was positive and a small value was recorded even at room temperature.
We can summarize the main findings obtained, thanks to the study of
electrically operated injection devices. First, they unambiguously demon-
strated both spin injection and spin transport in OSPDs. The available data
confirms so far spin injection and transport only for voltages not exceed-
ing approximately 1 V. Moreover, highly efficient MR are constrained in a
narrower interval of about 0.1–0.2 V. Current MR measurements provide
1.0 2.0
Normalized MR
0.8 CuPc
Normalized MR
N 1.5
Alq3
0.6 O O
AL N
1.0
0.4 O
O
0.2 N 0.5
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0 100 200 300
(a) T/Tc (b) (c) T (K)
FIGURE 5.2 (a) Temperature dependence of normalized MR of vertical spin-valve devices consisting of
LSMO and Co as electrodes and Alq3 as spacer (dots) with Tc = 325 K. The line represents the LSMO surface
magnetization (b) Alq3 molecular structure. (c) Temperature dependence of maximum observed MR for two
junctions based on Alq3 and CuPc, as organic spacer and Fe and Co as metallic contact. From Liu et al., the
MRs for each sample are normalized to their values at T = 80 K. (After Liu, Y. et al., J. Appl. Phys. 105, 07C708,
mainly qualitative information and in the absence of well-defined models for

spin injection and transport in OSC, it is difficult to extract the values of the
spin polarization or its decay inside the organic material. The investigation
of the thickness and temperature dependence of the MR in OSPDs provides
important indications anyway for the basic physics underlying spintronics
effects.
Surprisingly, very important information comes from the “confusing”
set of data on the thickness dependence. Except for a few reported sequences
where an expected decrease of MR upon increasing thickness was reported,
many published [28, 36] as well as unpublished data (various conference dis-
cussions) indicate quite casual dependences.
A reasonable and, in our opinion, credible explanation for this comes,
amazingly, from the analysis of the temperature behavior for the most arche-
typal LSMO-Alq3-Co OSPD [24, 26]. Figure 5.2a presents the normalized
MR versus temperature for a device with a 100 nm thick Alq3 layer and a 2 nm
thick Al2O3 barrier separating Alq3 and top Co electrode [26]. Remarkably,
the data extrapolate to zero at the Curie temperature of the manganite [26]
(325 K in this particular case). Moreover, the temperature dependence of MR
data agrees with that of the surface magnetization (SM) curve for LSMO
(solid line in Figure 5.2a) obtained by Park et al. [15]. The SM represents the
magnetization from the top 5 Å in a standard LSMO film, as determined
by spin-polarized photoemission spectroscopy [15], and it is effectively the
parameter of interest for spin injection in LSMO-based devices [37]. The
results summarized above are in agreement with the previous claim that the
temperature dependence of MR in Alq3 spintronic devices is governed by the
manganite electrode [24].
The temperature behavior described above emphasizes a very important
statement: the LSMO-Alq3-Co devices behave in an “injection limited”-like
regime, as the electrode spin polarization nearly dominates the tempera-
ture dependence. The spin losses at the interface significantly prevail over
the transport bulk losses. By analogy with the pure charge injection case, it
seems coherent to expect a weak or even a random thickness dependence in

such devices: for example, small fluctuations of the interface quality should
influence the device behavior much more than a thickness variation. Also,
we believe this regime can be expanded to other materials and interfaces,
although there are no a priori reasons to suppose that all OSPDs should
work in injection-limited mode.
5.4 ELECTRICAL OPERATION: TUNNELING

While the interest in spin injection and long-distance spin transport in
OSCs was growing, several groups also started to explore the possibilities
of these materials as spin tunnel barriers. Spin-polarized tunneling in OSPD
was initially demonstrated in quite complex geometries. Petta et al. [38] fab-
ricated Ni-octanethiol-Ni vertical tunneling devices in a nanopore geometry
with an octanethiol self-assembled monolayer. These devices showed MR up
to 16% at 4.2 K, although it vanished at about 30 K. The sign of the MR was
observed to switch from positive to negative for different voltage values, but
also from sample to sample at the same voltage. While evidence of the spin-
polarized tunneling through a monolayer of organic was given, it has also
been noticed that such devices are subject to strong intrinsic and extrinsic
noises.
An interesting approach for the investigation of tunneling effects is the
utilization of composite systems featuring magnetic nanoparticles embed-
ded into an organic matrix. Depending on the separation between nanopar-
ticles, a switching from interparticle tunneling to hopping conductivity
between particles occurs. Such effects have been detected in Co:C60 nano-
composite [39].
An important step forward for organic spin tunneling was the fab-
rication of simple devices via direct in situ UHV organic deposition with
shadow masking. Vertical tunneling devices (Co/Al2O3-Alq3/NiFe [40] and
Co/Al2O3-rubrene/Fe [41]) made use of hybrid inorganic–organic ultrathin
barriers to produce positive TMR when the electrode magnetizations were
switched from parallel to antiparallel. TMR around 15% was recorded at
4.2 K [41], and, importantly, both OSC showed few percent TMR at room
temperatures, opening the door to possible commercial applications.
Another important aspect is that Al2O3-Alq3 devices had the hybrid tun-
nel barrier well inside the typical quantum mechanical tunneling regime
(0.6–1.6 nm). However, the Al2O3-rubrene hybrid barriers were considerably
thicker, with rubrene layers ranged from 4.6 to 15 nm. These thicknesses are
well over the tunneling regime and probably imply spin injection in rubrene
layers. A transition from tunneling into injection regime was in principle
tackled for the first time for organic–inorganic hybrid devices [41]. At the
present state of the art, the organic spintronic community is not yet ready
to provide a thickness value where such a transition can take place, nor is it
ready to describe convincingly the system behavior at the two sides of transi-
tion. Further investigation is required in order to face these key fundamen-
tal questions. One has to keep firmly in mind, nevertheless, that OSCs of
5.5 Alternative Approaches 257
about 20–30 nm (e.g. Alq3) are already good light emitters, indicating that
this thickness is already well beyond tunneling limit.
The experiments reported above clearly show the possibility of achiev-
ing spin-polarized tunneling across organic or organic–inorganic hybrid
barriers. Although the control of spin tunneling across organic layers is still
relatively poor, it is also clear that there is a great margin for improvement,
especially in the fields of interface control and engineering of single mol-
ecule studies.
5.5 ALTERNATIVE APPROACHES
In this section, we will outline different important experiments that have
been reported recently in the field of organic spintronics, but not with stan-
dard electronic spin transport in devices. These alternative experiments are
extremely important, as they allow us to have access to different physical
parameters than the ones restricted by simple electronic transport data. Our
overview here does not pretend to be systematic, but simply to highlight dif-
ferent recent approaches that are shedding light on different aspects of spin
injection and transport in OSCs.
One important piece of information recently revealed is the spin polar-
ization profile inside an OSC. Such information has deep implications, as
it would allow the research community to evaluate the true potential of
organic spintronics for carrying information at long distances. The data has
been obtained by two powerful nonstandard techniques: two-photon photo-
emission spectroscopy and muon spin rotation technique.
Two-photon photoemission spectroscopy experiments allow us to
obtain the spin polarization of injected electrons in the first monolayers of
organic materials grown on top of a spin-polarized ferromagnetic material
[42]. Spin-polarized electrons from the metal are excited by the first energy
pulse into an intermediate energy state. Some of the electrons propagate into
the OSC where, with a certain probability, they can be excited by the second
photon causing the photoemission. The spin-polarized injection efficiency
from cobalt into the first monolayer of the copper phthalocyanine was esti-
mated to be close to 100% [42]. An obvious drawback in these measurements
is that the energy requested for the electron injection process is much higher
than the energy gap in the OSC. For those energies, the spin scattering is
higher than for the (lower) voltages at which electron/spin injection usually
takes place, making the comparison between photoemission spectroscopy
and standard electron devices far from straightforward.
In a totally independent article, electrical injection was combined with
low-energy muon spin rotation to study the spin diffusion length inside Alq3
in a standard vertical spin-valve device [43]. The stopping distribution of the
spin-polarized muons was varied in the range 3–200 nm by controlling the
implantation energy, while a quantitative analysis determined the spin dif-
fusion length (lS) and its temperature dependence. On one hand, the diffu-
sion length reached a low temperature value of 35 nm: a lower value than
the one estimated from electrical measurements (>100 nm), a discrepancy
that could nevertheless be explained, bearing in mind the imperfect injec-

tion efficiency and very small MR of the explored device. On the other hand,
the weak temperature dependence of the lS is in good qualitative agreement
with independent MR characterizations in Alq3-based devices [26], and also
in agreement with a recently proposed mechanism for the hyperfine-driven
spin scattering in OSC.
In a different subset, we could position the articles that explore the
spin transport properties in OSC in relation to their electroluminescence
(EL) properties. In fact, EL has been key in the advancement of the OSC
research, since it opens the door to multiple commercial applications, such
as OLEDs. Similarly, the careful study of the emitted light could give us very
useful information about the electron/hole recombination processes inside
the organic. In OLEDs, due to low spin–orbit coupling, the dominant EL
channel involves singlet–singlet transitions [44]. In OLED, the singlet/triplet
ratio is determined by quantum statistics as 1:3 when considering a similar
formation probability for one singlet and three triplet states (Figure 5.3). The
injection of carriers with a controlled spin state could therefore enable the
amplification of the singlet (triplet) exciton density, leading to a significant
increase of efficiency for the EL [45], rather than a polarization in the emit-
ted light as in other inorganic semiconductors (such as GaAs) [46–48]. This
theoretically expected increase in the emitted light attracted many research-
ers into organic spintronics a few years ago, but unfortunately, the efficiency
amplification has not been reported in the literature [49, 50]. The absence of
amplification of light in spin-polarized OLED might be due to the fact that
Antiparallel spins Parallel spins

Electron injector
ETL
HTL
Hole injector
Singlet states Singlet states

Triplet states
Triplet states
Ground state Ground state
FIGURE 5.3 Schematic structure of a bilayered OLED. Electrons and holes are
injected by cathode and anode, respectively, into the electron-transport layer (ETL)
and the hole-transport layer (HTL). Exciton formation and recombination eventu-
ally take place at the interface. Spin statistic of carrier recombination in case of
spin injection. Antiparallel spin states for electrons and holes generate singlet and
triplet state in ratio 1:1. Parallel spin states for electrons and holes’ spin state of car-
riers give rise only to the triplet state.
5.6 Interface Effects 259
light emission can typically only be detected for applied voltages in excess of
a few volts, while spin injection in organic materials has been demonstrated
so far for voltages not exceeding 1 V [23, 45]. While this limitation does not
seem to be a fundamental obstacle, it has yet to be understood whether it can
be experimentally circumvented.
Irrespective of the current lack of successful results in light amplifica-
tion in spin-polarized OLED, they are very useful anyway for the study of
the electron spin conversion, that can be studied by electron spin resonance
(ESR) [51]. In this chapter, the authors achieved a magnetically guided
transfer between singlet (PPS) and triplet (PPT) polaron pairs caused by
the spin precession. The two conversion channels give completely differ-
ent inputs into the photoconductivity, and monitoring the later provides a
way to control the PP dynamics. A periodical modification of the PPS → PPT
transfer (Rabi oscillations) was observed in the conjugated polymer MEH-
PPV, from which an extremely long spin coherence time above 0.5 µs [51]
was extracted.
5.6 INTERFACE EFFECTS
In a strict sense, for the MR in a spin-valve organic device to be present,
both bottom and top ferromagnetic/OSC interfaces would have to be spin
selective (such as, e.g. a tunnel barrier [52]). In this context, the perception of
the complex nature of the interfaces between magnetic materials and OSCs
is motivating several groups to undertake detailed studies of the organic–
inorganic interfaces. Important issues such as the interfacial energy levels
and the degree of chemical interaction or the different interfacial states are
being clarified with this kind of research. In general, spectroscopic tech-
niques are being used to study these systems. UPS/XPS are usually the pre-
ferred methods to study these interfaces.
Two groups reported almost parallel independent research on what, in
principle, should be the optimum combination of materials, a high Curie
temperature ferromagnetic material (cobalt) and a high-mobility OSC (pen-
tacene, Pc) (see Figure 5.4a). However, it is important to remember that Co,
as with any transition metal, is chemically very active, and this activity could
result in a very complex interfacial structure. In general, interfacial dipoles
of around 1 eV are observed between Co and Pc [53]; however, it is claimed
that the Pc/Co interface (with the Co film deposited on top of the organic)
behaves differently, showing vacuum-level alignment (no dipole) and high-
lighting again the importance of careful characterization of these complex
devices [54].
The ubiquitous LSMO/Alq3/Co device has been fully investigated for
interface electronic properties (Figure 5.4b). Detailed XPS and UPS investi-
gations have revealed a 0.9 eV interfacial dipole at LSMO/Alq3 interface [55]
and a 1.5 eV dipole at direct Co/Alq3 interfaces [56]. The latter is reduced
from 1.5 to 1.3 eV when a 2 nm Al2O3 barrier is inserted between Alq3 and
Co layer [57]. This knowledge provides a complete set of barriers across the
whole device, describing the current injection into LUMO and HOMO levels
Al2O3
1.05 eV ~0 eV 0.9 eV 1.5 – 1.3 eV
5.0 eV 4 eV 5.1 eV
4.92 eV 4.9 eV
LUMO
N
O
LUMO
H
Al
C
–+ + –
0.96 eV 0.96 eV –+ + –
1.7 eV 2.1 eV + –
HOMO
HOMO HOMO HOMO
(a) Co Co (b) LSMO Co
FIGURE 5.4 (a) Schematic band diagram for an LSMO/Alq3 and Alq3/Co structure presented by Zhang et al.
[56, 57]. (b) Schematic band diagram for a layered Co/pentacene/Co structure presented by Popinciuc [54].
of OSC. This information is not, however, sufficient for the spin injection
efficiency, as no clear data on the spin transfer is revealed. Thus, the neces-
sary evolution of the interface studies consists in moving toward monitoring
of the spin selectivity and injecting interfaces. For example, proximity effects
at organic–inorganic interface can induce new states in the gap of OSCs with
non-trivial spin polarization [58].
Recently, Grobosch et al. [59] have questioned whether interfacial
energy values measured in UHV conditions can be used to explain the elec-
tronic transport behavior of actual devices measured in standard laboratory
conditions.
5.7 MULTIFUNCTIONAL EFFECTS
A new field that is currently attracting much attention, and also one in which
OSC could have a significant impact, is that of multifunctional or multipur-
pose devices. In general terms, these are devices that react either simultane-
ously or separately under different external stimuli, such as electric field,
magnetic field, light irradiation, and pressure. In this area of research, the
possibility of combining magnetic and electrical resistance switching has
been realized recently [60], profiting from the unique combination of prop-
erties in OSPDs. Nonvolatile electrical memory effect in hybrid organic–
inorganic materials has been extensively studied (see e.g. the extended recent
review of Scott and Bozano [61]). Under the application of an electric field,
resistance changes in excess of one order of magnitude, and retention times
of tens of hours have been routinely detected. Following those successes,
the International Technology Roadmap for Semiconductors has identified
molecular-based memories among the emerging technology lines (http://
www.itrs.net).
In organic spintronics, hybrid vertical devices with LSMO and Co as
charge/spin-injecting electrodes show both spin-valve and nonvolatile elec-
trical memory effects on a same chip (see Figure 5.5). This result was con-
firmed for two different OSCs, Alq3 and 6T, and consists of the observation of
5.8 Theoretical Models 261
100 2 W
ITO/xBP9F: AuNP/Al
25
Voltage (V)
0
VON R
Current density (A/cm2)
ON 20
∆R/R %
10–2 –1 1
15
Current (mA)
–2
10 0
VOFF 0
10–4 –3.0–1.5 0.0 1.5 3.0 0.02
Vth 5
Current (A)
Magnetic field (KOe)
OFF 0
0.01
10–5 –5 ON
–10 0.00
OFF
10–6 –15
–8–7–6–5–4–3–2–1 0 1 2 3 4 5 6 7 8 –3 –2 –1 0 1 2 3 830 835 840 845 850
(a) Voltage (V) (b) Voltage (V) (c) Time (s)
FIGURE 5.5 (a) Typical current density–voltage characteristic for a nonvolatile memory cell based on com-
posite of metallic nanoparticles and polymers embedded in an organic semiconducting host. (After Bozano,
L.D. et al., Adv. Funct. Mater. 15, 1933–1939, 2005. With permission). (b) Current–voltage room temperature
characteristics in T6-based vertical device. The irreversibility at both positive and negative voltages is clearly
visible, and it is associated to the classical spin-valve effect presented in the inset. (c) Switching effect for the
OrgSV memory cell. Writing (W), erasing (E), and reading (R) voltages have been used. Two very clear and
reproducible ON and OFF states can be observed multiple times even if the voltage has been set to zero
between different reading events.
10% negative spin-valve magnetic switching at voltages below 1 V, and 100%
electrical switching at voltages above 2–3 V. Importantly, the resistance val-
ues achieved by electrical switching are not altered when the devices go back
to zero voltage (nonvolatility; see Figure 5.5b and c). With an initial writ-
ing voltage pulse, the device moves to a resistance state (called 1), and this
resistance state can be recovered multiple times with a reading voltage pulse
(reading voltage < writing voltage), even if the voltage has been set to zero
after the writing pulse. If we later apply a negative erasing pulse, the new
resistance value at the reading voltage will be different (called 0), and again
this value can be recovered multiple times even if the voltage is set to zero
between reading pulses. This behavior, explained above, implies that a set
resistance value (either 1 or 0) can be recovered even if the memory cell has
been left without any voltage for a set amount of time, allowing an operation
with much lower power and avoiding the need of constant refreshing of cur-
rent RAM cells.
We believe this combination will become a subject of extensive inves-
tigations and can provide a basis for the generation of two bits/cell memory
elements. As a future development, we envisage multifunctional devices
with magnetic and electric field control of the resistance, and that in addition
to these responses, will profit from the luminescent properties of the OSCs.
5.8 THEORETICAL MODELS
Many theoretical groups have done a considerable amount of work of differ-
ent aspects of organic spintronics. Much of the current research is clearly
progressing from previous works on inorganic devices; however, there are
great new ideas and unique approaches that are helping the organic spin-
tronics field to move forward. Probably one of the most active topics is the
one related to the theoretical description of tunneling devices with a single
(or a few) molecules placed between two spin-polarized electrodes [62–64].

Tunneling is perhaps the less-complicated organic spintronics process to
describe, and very interesting results have been obtained by applying density
functional theory (DFT) combined with nonequilibrium Green’s functions
to calculate the transport characteristics [64, 65]. Rocha et al. [64] consid-
ered two prototypical molecules, 8-alkane-dithiolate (octane-dithiolate) and
1,4-3-phenyl-dithiolate (tricene-dithiolate), sandwiched between nickel con-
tacts. Very large MR values have been calculated for both cases: about 100%
for the octane-dithiolate, and up to 600% for the tricene-dithiolate molecule
(Figure 5.6).
In close contact with all the experimental research dealing with the
characterization of organic–inorganic interfaces, many authors have
developed different theoretical approaches to understand the optimum
conditions for spin injection into an OSC. A few authors have indicated
that an ohmic contact would be more efficient for spin injection into
OSC [66]. By contrast, it has been suggested in many other papers that
spin injection into OSC can only be achieved across an interfacial tunnel
or Schottky barrier [67, 68]. In all the cases in which spin injection into
OSC has been demonstrated, the organic–inorganic interfaces presented
natural Schottky-like barriers, thus showing that experimentally this kind
of interfacial contact seems more adequate for the realization of OSPD.
A Su–Schrieffer–Heeger model was applied by Xie et al. [69] to study
the importance of the electron–lattice interactions, considering inter-
faces featuring strong polaronic effects at both sides of them. The authors
Nickel Sulfur Carbon Hydrogen
40.0
600
MR (%)
400
20.0 200
0
I (µA)
–2 –1 0 1 2
0
Parallel configuration
–20.0 Antiparallel configuration
–40.0
–2 –1 0 1 2
V (V)
FIGURE 5.6 Schematic representation of a tricene molecule attached to a (001)

Ni surface from Rocha et al. (upper figure). Theoretical calculation of the current–
voltage (I–V) characteristic of a tricene–nickel spin valve. The inset shows the
calculated MR ratio with values up to 600%. (After Rocha, A.R. et al., Nat. Mater. 4,
5.9 Conclusions 263
showed that under certain biases, spin-polarized injection can be achieved

for direct manganite–OSC interface, while substitution of the manganite
with a conventional (non-polaronic) ferromagnetic metal resulted in the
absence of a spin-polarized flow.
If the injection process has attracted much attention, it is clear as well
that the decoherence mechanisms in organic are being subject of intense
research. In a naive approach, it is clear that spin–orbit coupling in OSC
is very small, but a detailed and predictive description has been lacking.
Currently, alternative mechanisms for spin decoherence are being consid-
ered. Recently, Bobbert et al. have advanced the idea that the precession
of the carrier spins in the random hyperfine fields of hydrogen nuclei [70]
(around 5 mT) represents the dominant source for spin scattering in OSC.
The authors have derived the spin diffusion length lS in a typical OSC and
they obtain a very weak dependence of lS on temperature, in good agree-
ment with recent experimental data [26]. Surprisingly, it was also shown
that the spin-scattering strength is much less sensitive to disorder than the
carrier mobility, a result that could shed light on the apparent discrepancy
between low mobility and long spin diffusion lengths in OSC. It has to be
noted that spin scattering by hyperfine fields is also considered as one of
the most probable mechanisms [71] responsible for the so-called organic
magnetoresistance (OMAR) effect in OSC [72], that is, MR without mag-
netic electrodes.
5.9 CONCLUSIONS
In spite of the important achievements described above, an understanding of
the basic rules governing spin injection and transport in OSC is still lacking.
It is also difficult to predict now the future evolution of organic spintron-
ics, or any possible implementation into applications. Organic spintronics
remains, so far, a fascinating scientific problem, which requires facing new
physics conceptually and developing innovative technological procedures.
In addition to this stimulating fundamental interest, the results obtained
in this novel field are already encouraging. It cannot be excluded that OSC
will compete with other materials for the leadership in the spintronics field,
or at least in some selected niche applications.
In order to link spintronics to such important organic applications like
OLEDs and OFETs, it is necessary nevertheless to overcome some revealed
limitations. Among these, we could cite the low voltages at which spin-
polarized effects occur in electrically driven devices. While low operation
voltages are required in order to reduce energy consumption, this limitation
prevents the practical application of spin-polarized currents in OLEDs and
OFETs, which are operated at V > 1 V (generally at few volts). The available
data for spintronic effects in organic materials confirm so far spin injection
and transport only for voltages not exceeding approximately 1 V. Expanding
this working interval to about a few volts represents thus a challenging appli-
cative goal. On the other hand, the available operation voltages do not pre-
vent the utilization of OSPDs in sensor or memory applications.
As a final remark, we would like to point out on the immense poten-

tial of OSCs for the spintronics field. An almost infinite list of materials,
easily modifiable interfaces, and cheap technologies for the growth of films
on rigid and flexible substrates make us look optimistically toward the next
steps of this young field.
REFERENCES
1. V. A. Dediu, L. E. Hueso, I. Bergenti, and C. Taliani, Spin routes in organic
semiconductors, Nat. Mater. 8, 707–716 (2009).
2. F. Wang and Z. V. Vardeny, Organic spin valves: The first organic spintronics
devices, J. Mater. Chem. 19, 1685–1690 (2009).
3. W. J. M. Naber, S. Faez, and W. G. van der Wiel, Organic-spintronics, J. Phys.
D-Appl. Phys. 40, R205–R228 (2007).
4. D. S. McClure, Spin-orbit interaction in aromatic molecules, J. Chem. Phys. 20,
682–686 (1952).
5. D. Beljonne, Z. Shuai, G. Pourtois, and J. L. Bredas, Spin-orbit coupling and
intersystem crossing in conjugated polymers: A configuration interaction
description, J. Phys. Chem. A 105, 3899–3907 (2001).
6. C. B. Harris, R. L. Schlupp, and H. Schuch, Optically detected electron spin
locking and rotary echo trains in molecular excited states, Phys. Rev. Lett. 30,
1019–1022 (1973).
7. V. I. Krinichnyi, S. D. Chemerisov, and Y. S. Lebedev, EPR and charge-
transport studies of polyaniline, Phys. Rev. B 55, 16233–16244 (1997).
tions, Rev. Mod. Phys. 76, 323–410 (2004).
10. V. I. Arkhipov, E. V. Emelianova, Y. H. Tak, and H. Bassler, Charge injection
into light-emitting diodes: Theory and experiment, J. Appl. Phys. 84, 848–856
(1998).
11. W. Brütting, S. Berleb, and A. G. Mückl, Device physics of organic light-emit-
ting diodes based on molecular materials, Org. Electron. 2, 1–36 (2001).
12. V. Coropceanu, J. Cornil, D. A. da Silva, Y. Olivier, R. Silbey, and J. L. Bredas,
Charge transport in organic semiconductors, Chem. Rev. 107, 926–952
(2007).
13. T. W. Hagler, K. Pakbaz, K. F. Voss, and A. J. Heeger, Enhanced order and
electronic delocalization in conjugated polymers oriented by gel processing in
polyethylene, Phys. Rev. B 44, 8652–8666 (1991).
14. W. P. Su, J. R. Schrieffer, and A. J. Heeger, Solitons in polyacetylene, Phys. Rev.
Lett. 42, 1698–1701 (1979).
15. J.-H. Park, E. Vescovo, H.-J. Kim, C. Kwon, R. Ramesh, and T. Venkatesan,
Magnetic properties at surface boundary of a half-metallic ferromagnet
La0.7Sr0.3MnO3, Phys. Rev. Lett. 81, 1953–1956 (1998).
16. M. P. de Jong, I. Bergenti, V. A. Dediu, M. Fahlman, M. Marsi, and C. Taliani,
Evidence for Mn2+ ions at surfaces of La0.7Sr0.3MnO3 thin films, Phys. Rev. B 71,
014434 (2005).
17. J. M. De Teresa, A. Barthelemy, A. Fert, J. P. Contour, F. Montaigne, and
P. Seneor, Role of metal-oxide interface in determining the spin polarization of
magnetic tunnel junctions, Science 286, 507–509 (1999).
18. J. E. Evetts, M. G. Blamire, N. D. Mathur et al., Defect-induced spin disorder
and magnetoresistance in single-crystal and polycrystal rare-earth manga-
nite thin films, Philos. Trans. R. Soc. Lond. Ser. A-Math. Phys. Eng. Sci. 356,
1593–1613 (1998).
References 265
19. C. W. Tang and S. A. VanSlyke, Organic electroluminescent diodes, Appl.

Phys. Lett. 51, 913–915 (1987).
20. A. Curioni, M. Boero, and W. Andreoni, Alq(3): ab initio calculations of its
structural and electronic properties in neutral and charged states, Chem. Phys.
Lett. 294, 263–271 (1998).
21. A. N. Caruso, D. L. Schulz, and P. A. Dowben, Metal hybridization and elec-
tronic structure of Tris(8-hydroxyquinolato) aluminum (Alq(3)), Chem. Phys.
Lett. 413, 321–325 (2005).
22. G. Horowitz, D. Fichou, X. Z. Peng, Z. G. Xu, and F. Garnier, A field-effect
transistor based on conjugated alpha-sexithienyl, Solid State Commun. 72,
381–384 (1989).
23. Z. H. Xiong, D. Wu, Z. V. Vardeny, and J. Shi, Giant magnetoresistance in
organic spin-valves, Nature 427, 821–824 (2004).
24. F. J. Wang, C. G. Yang, Z. V. Vardeny, and X. G. Li, Spin response in organic
spin valves based on La2/3Sr1/3MnO3 electrodes, Phys. Rev. B 75, 245324 (2007).
25. A. Riminucci, I. Bergenti L. E. Hueso et al., Negative spin valve effects in
manganite/organic based devices, arXiv:cond-mat/0701603v1 (2006).
26. V. Dediu, L. E. Hueso, I. Bergenti, et al., Room-temperature spintronic effect in
Alq3-based hybrid devices, Phys. Rev. B 78, 115203 (2008).
27. S. Majumdar, H. S. Majumdar, R. Laiho, and R. Osterbacka, Comparing small
molecules and polymer for future organic spin-valves, J. Alloys Compd. 423,
169–171 (2006).
28. W. Xu, G. J. Szulczewski, P. LeClair et al., Tunneling magnetoresistance
observed in La0.67Sr0.33MnO3/organic molecule/Co junctions, Appl. Phys. Lett.
90, 072506 (2007).
tance at room-temperature in ferromagnetic thin-film tunnel-junctions, Phys.
Rev. Lett. 74, 3273–3276 (1995).
30. E. Y. Tsymbal, O. N. Mryasov, and P. R. LeClair, Spin-dependent tunnelling in
magnetic tunnel junctions, J. Phys. Condens. Matter 15, R109–R142 (2003).
31. S.-W. Liu, J.-H. Lee, C.-C. Lee, C.-T. Chen, and J.-K. Wang, Charge carrier
mobility of mixed-layer organic light–emitting diodes, Appl. Phys. Lett. 91,
142106 (2007).
32. V. Dediu, M. Murgia, F. C. Matacotta, C. Taliani, and S. Barbanera, Room
temperature spin polarized injection in organic semiconductor, Solid State
Commun. 122, 181–184 (2002).
33. T. Ikegami, I. Kawayama, M. Tonouchi, S. Nakao, Y. Yamashita, and H. Tada,
Planar-type spin valves based on low–molecular-weight organic materials
with La0.67Sr0.33MnO3 electrodes, Appl. Phys. Lett. 92, 153304 (2008).
34. Y. Liu, T. Lee, H. E. Katz, and D. H. Reich, Effects of carrier mobility and mor-
phology in organic semiconductor spin valves, J. Appl. Phys. 105, 07C708
(2009).
35. N. A. Morley, A. Rao, D. Dhandapani, M. R. J. Gibbs, M. Grell, and
T. Richardson, Room temperature organic spintronics, J. Appl. Phys. 103,
07F306 (2008).
36. H. Vinzelberg, J. Schumann, D. Elefant, R. B. Gangineni, J. Thomas, and
B. Büchner, Low temperature tunneling magnetoresistance on (La,Sr)MnO3/
Co junctions with organic spacer layers, J. Appl. Phys. 103, 093720 (2008).
37. Y. Ogimoto, M. Izumi, A. Sawa et al., Tunneling magnetoresistance above
room temperature in La0.7Sr0.3MnO3/SrTiO3/La0.7Sr0.3MnO3 junctions, Jpn. J.
Appl. Phys. Part 2-Lett. 42, L369–L372 (2003).
38. J. R. Petta, S. K. Slater, and D. C. Ralph, Spin-dependent transport in molecular
tunnel junctions, Phys. Rev. Lett. 93, 136601 (2004).
39. S. Miwa, M. Shiraishi, S. Tanabe, M. Mizuguchi, T. Shinjo, and Y. Suzuki,
Tunnel magnetoresistance of C-60-Co nanocomposites and spin-dependent
transport in organic semiconductors, Phys. Rev. B 76, 214414 (2007).
40. T. S. Santos, J. S. Lee, P. Migdal, I. C. Lekshmi, B. Satpati, and J. S. Moodera,

Room-temperature tunnel magnetoresistance and spin-polarized tunneling
through an organic semiconductor barrier, Phys. Rev. Lett. 98, 016601 (2007).
41. J. H. Shim, K. V. Raman, Y. J. Park et al., Large spin diffusion length in an amor-
phous organic semiconductor, Phys. Rev. Lett. 100, 226603 (2008).
42. M. Cinchetti, K. Heimer, J. P. Wustenberg et al., Deter-mination of spin injec-
tion and transport in a ferromagnet/organic semiconductor heterojunction by
two-photon photoemission, Nat. Mater. 8, 115–119 (2009).
43. A. J. Drew, J. Hoppler, L. Schulz et al., Direct measurement of the electronic
spin diffusion length in a fully functional organic spin valve by low-energy
muon spin rotation, Nat. Mater. 8, 109–114 (2009).
44. R. H. Friend, R. W. Gymer, A. B. Holmes et al., Electroluminescence in conju-
gated polymers, Nature 397, 121–128 (1999).
45. I. Bergenti, V. Dediu, E. Arisi et al., Spin polarised electrodes for organic light
emitting diodes, Org. Electron. 5, 309–314 (2004).
46. R. Fiederling, M. Keim, G. Reuscher et al., Injection and detection of a spin-
polarized current in a light-emitting diode, Nature 402, 787–790 (1999).
erostructure, Nature 402, 790–792 (1999).
49. G. Salis, S. F. Alvarado, M. Tschudy, T. Brunschwiler, and R. Allenspach,
Hysteretic electroluminescence in organic light-emitting diodes for spin injec-
tion, Phys. Rev. B 70, 085203 (2004).
50. A. H. Davis and K. Bussmann, Organic luminescent devices and magnetoelec-
tronics, J. Appl. Phys. 93, 7358–7360 (2003).
51. D. R. McCamey, H. A. Seipel, S. Y. Paik et al., Spin Rabi-flopping in the photo-
current of a polymer light-emitting diode, Nat. Mater. 7, 723–728 (2008).
52. E. I. Rashba, Theory of electrical spin injection: Tunnel contacts as a solution of
the conductivity mismatch problem, Phys. Rev. B 62, R16267–R16270 (2000).
53. M. V. Tiba, W. J. M. de Jonge, B. Koopmans, and H. T. Jonkman, Morphology
and electronic properties of the pentacene on cobalt interface, J Appl. Phys.
100, 093707 (2006).
54. M. Popinciuc, H. T. Jonkman, and B. J. van Wees, Energy level alignment sym-
metry at Co/pentacene/Co interfaces, J. Appl. Phys. 100, 093714 (2006).
55. Y. Q. Zhan, I. Bergenti, L. E. Hueso, and V. Dediu, Alignment of energy levels
at the Alq(3)/La0.7Sr0.3MnO3 interface for organic spintronic devices, Phys. Rev.
B 76, 045406 (2007).
56. Y. Q. Zhan, M. P. de Jong, F. H. Li, V. Dediu, M. Fahlman, and W. R. Salaneck,
Energy level alignment and chemical interaction at Alq(3)/Co interfaces for
organic spintronic devices, Phys. Rev. B 78, 045208 (2008).
57. Y. Q. Zhan, X. J. Liu, E. Carlegrim et al., The role of aluminum oxide buffer
layer in organic spin-valves performance, Appl. Phys. Lett. 94, 053301 (2009).
58. Y. Q. Zhan and M. Fahlman, Private communication (2009).
59. M. Grobosch, K. Dorr, R. B. Gangineni, and M. Knupfer, Energy level align-
ment and injection barriers at spin injection contacts between La0.7Sr0.3MnO3
and organic semiconductors, Appl. Phys. Lett. 92, 023302 (2008).
60. L. E. Hueso, I. Bergenti, A. Riminucci, Y. Q. Zhan, and V. Dediu, Multipurpose
magnetic organic hybrid devices, Adv. Mater. 19, 2639–2642 (2007).
61. J. C. Scott and L. D. Bozano, Nonvolatile memory elements based on organic
materials, Adv. Mater. 19, 1452–1463 (2007).
62. E. G. Emberly and G. Kirczenow, Molecular spintronics: Spin-dependent elec-
tron transport in molecular wires, Chem. Phys. 281, 311–324 (2002).
63. R. Pati, L. Senapati, P. M. Ajayan, and S. K. Nayak, First-principles calculations
of spin-polarized electron transport in a molecular wire: Molecular spin valve,
Phys. Rev. B 68, 100407(R) (2003).
References 267
64. A. R. Rocha, V. M. Garcia-Suarez, S. W. Bailey, C. J. Lambert, J. Ferrer, and

S. Sanvito, Towards molecular spintronics, Nat. Mater. 4, 335–339 (2005).
65. D. Waldron, P. Haney, B. Larade, A. MacDonald, and H. Guo, Nonlinear spin
current and magnetoresistance of molecular tunnel junctions, Phys. Rev. Lett.
96, 166804 (2006).
66. J. F. Ren, J. Y. Fu, D. S. Liu, L. M. Mei, and S. J. Xie, Spin polarized injection and
transport in organic polymers, Synth. Met. 155, 611–614 (2005).
67. P. P. Ruden and D. L. Smith, Theory of spin injection into conjugated organic
semiconductors, J. Appl. Phys. 95, 4898–4904 (2004).
68. J. H. Wei, S. J. Xie, L. M. Mei, J. Berakdar, and W. Yan, Conductance switching,
hysteresis, and magnetoresistance in organic semiconductors, Org. Electron. 8,
487–497 (2007).
69. S. J. Xie, K. H. Ahn, D. L. Smith, A. R. Bishop, and A. Saxena, Ground-state
properties of ferromagnetic metal/conjugated polymer interfaces, Phys. Rev. B
67, 125202 (2003).
70. P. A. Bobbert, W. Wagemans, F. W. A. van Oost, B. Koopmans, and M.
Wohlgenannt, Theory for spin diffusion in disordered organic semiconduc-
tors, Phys. Rev. Lett. 102, 156604 (2009).
71. P. A. Bobbert, T. D. Nguyen, F. W. A. van Oost, B. Koopmans, and M.
Wohlgenannt, Bipolaron mechanism for organic magnetoresistance, Phys.
Rev. Lett. 99, 216801 (2007).
72. T. L. Francis, O. Mermer, G. Veeraraghavan, and M. Wohlgenannt, Large mag-
netoresistance at room temperature in semiconducting polymer sandwich
devices, New J. Phys. 6, 185 (2004).
73. L. D. Bozano, B. W. Kean, M. Beinhoff, K. R. Carter, P. M. Rice, and J. C. Scott,
Organic materials and thin-film structures for cross-point memory cells
based on trapping in metallic nanoparticles, Adv. Funct. Mater. 15, 1933–1939
(2005).
6
Spin Transport in
Ferromagnet/III–V
Semiconductor
Heterostructures
6.1 Spin Transport and Dynamics in III–V

Semiconductors 271
6.2 Spin Injection from Ferromagnetic Metals into
Semiconductors 273
6.3 Fe/GaAs Heterostructures and Lateral Spin-Transport
Devices 276
6.4 Scanning Kerr Microscopy and Local Hanle Effect
Studies 278
6.4.1 Kerr Set-Up 278
6.4.2 Imaging Spin Injection, Local Hanle-Kerr
Studies, and Modeling 280
269
270 Chapter 6. Spin Transport in Ferromagnet/III–V Semiconductor Heterostructures
6.4.3 Imaging Spin Injection and Extraction,

Both Upstream and Downstream of the
Current Path 283
6.5 Electrical Detection of Spin Accumulation 286
6.5.1 Electrical Detection of Optically Injected Spins 286
6.5.2 Electrical Detection by the Three-Terminal
Technique 288
6.6 Non-Local Detection of Spin Accumulation 290
6.6.1 Bias Dependence of the Accumulated Spin
Polarization 294
6.6.2 Hyperfine Effects 299
6.6.3 Detector Bias Dependence 303
6.7 Recent Developments 305
6.7.1 Spin-Orbit Effects 305
6.7.2 Heusler Alloy-Based Heterostructures 306
6.8 Conclusions 307
Acknowledgments 307
References 308
T
he processes of electrical spin injection, transport, and detection
in nonmagnetic semiconductors are inherently nonequilibrium
phenomena. A nonequilibrium spin polarization, which we will refer
to in this chapter as spin accumulation, may be generated and probed by a
variety of optical and transport techniques. We will focus on spin accumula-
tion in heterostructures of ferromagnetic metals (particularly Fe) and III–V
semiconductors (particularly GaAs). Much of the essential physics of these
systems for both metals and semiconductors has been addressed in other
chapters in this volume. The basic phenomena of spin accumulation in met-
als have been covered in Chapter 5, Volume 1 and Chapter 9, Volume 3. This
chapter will provide a review of a sequence of experiments that have advanced
the understanding of spin transport in semiconductors through their sensi-
tivity to electron spin dynamics. To a great extent, this has been a matter of
careful experimental design and execution, and a good part of the discussion
will address tests of well-known models. The benefit of this approach has
been a reliable demonstration of all-electrical devices incorporating a fer-
romagnetic (FM) injector, semiconductor (SC) channel, and FM detector. In
this context, we will show how quantitative electrical spin detection mea-
surements can be made with these devices using both optical imaging and
transport techniques, and we will conclude with some discussion of the new
physics that can be addressed. Although the emphasis of the chapter is on
the progression of experiments leading up to the demonstration of a III–V
6.1 Spin Transport and Dynamics in III–V Semiconductors 271
based lateral spin valve, some discussion of more recent developments as

well as additional references are included in this revision.
6.1 SPIN TRANSPORT AND DYNAMICS

IN III–V SEMICONDUCTORS
The basic physics of electron spin dynamics in III–V semiconductors was
established through optical pumping experiments, which were reviewed
authoritatively more than three decades ago [1]. Polarized photolumines-
cence measurements identified many of the important spin relaxation mech-
anisms in GaAs, with the significant caveat that most experiments were
carried out on p-type, or compensated materials, in order to enhance the
recombination rate of spin-polarized electrons. As a result, the electron spin
dynamics observed in these experiments were influenced by the presence
of holes and by recombination dynamics. The advent of absorption-based
techniques, such as time-resolved Faraday rotation in the 1990s, brought
particular attention to the case of n-type GaAs doped in close proximity
to the metal-insulator transition (MIT) [2]. As will be evident below, the
relatively long spin lifetime ( 100ns) of electrons [2–4] in this doping range
and at low temperatures provides an enormous technical advantage if the
semiconductor can be integrated with an appropriate source and detector
of spin-polarized electrons. Before considering the role of the source and
detector, however, it will be useful to discuss spin transport and dynamics in
GaAs in the doping range near the MIT.
The interpretation of the spin transport measurements discussed below
will be based on the standard spin drift-diffusion model, which provided the
framework for the original proposal of electrical spin injection into semi-
conductors in the 1970s [5]. A detailed discussion of the elementary physics,
elucidating some of the important differences with respect to spin transport
in metals, is provided in Ref. [6]. An electron spin density S in a semiconduc-
tor is treated as a vector field evolving according to the dynamical equation
∂S S
= D∇r2S + µ(E ⋅ ∇r )S + g e µ B −1(Beff × S) − + G(r) = 0, (6.1)
∂t τs
where:
D is the diffusion constant
E is the electric field
Beff is the effective magnetic field acting on the electron spin
τ s is the spin lifetime
G(r) is a term representing the spatial distribution of spins generated
by an external source
The electron mobility is μ, and we will use g e = −0.44 for the Landé
g-factor of an electron in GaAs. The corresponding gyromagnetic ratio
can be expressed conveniently as γ =| g e | µ B / = 3.8 × 106 ⋅ s −1 G−1 or
γ / 2π = 0.61 MHz/Gauss. Three features of semiconductors which make
the spin dynamics non-trivial are (a) the relaxation mechanisms determin-
ing τ s , (b) the distinct contributions to the effective field Beff (for example,
applied fields, spin-orbit fields, or nuclear hyperfine fields), and (c) the rela-
tive importance of the drift term in the presence of internal electric fields.
In combination with ordinary diffusive transport theory, knowledge
of the relevant spin lifetimes allows us to identify the critical length scales
for a spin transport device. The relevant spin relaxation mechanisms are
reviewed by Fabian in this volume. In the doping range of interest for us, the
dominant channel of spin relaxation is the Dyakonov–Perel (DP) mechanism
[7], which corresponds to the randomization of spin during diffusive trans-
port by precession around the instantaneous spin-orbit field. Further dis-
cussion of DP and other spin relaxation mechanisms is given in Chapter 1,
Volume 2. The bulk Dresselhaus spin-orbit coupling (due to the absence of
inversion symmetry in GaAs), leads to a rapid decrease, proportional to ne−2 ,
in the spin lifetime with increasing electron concentration ne [2, 4, 7]. The
strong dependence on carrier density reflects the k 3 dependence in the spin-
orbit Hamiltonian. A maximum in the spin lifetime of order 100–200 ns is
typically observed near the MIT ( 2 ´ 1016 cm -3 ) at low temperatures [4], and
even longer spin lifetimes (600ns) are found at elevated temperatures (10K)
in samples doped slightly below the MIT [8, 9]. The characteristic length
scales for spin transport in the semiconductor are the spin diffusion length
λ s = Dτ s and drift length ld = µE τ s . The characteristic mobility of n-GaAs
at the MIT is of order 5000 cm2/V sec. Assuming a barely degenerate elec-
tron gas one finds that λ s is of the order of several microns.
We now consider the effective fields Beff that couple to the spin through
the torque term in Equation 6.1. In this case, the relevant magnetic field
scale is determined by the gyromagnetic ratio and the spin lifetime. For
example, the magnetic field required in order for a spin to precess a full
cycle before it decays can be estimated from 2π / γτ s  17 G (for τ s = 100ns).
One immediate consequence of this relatively small field scale is that it is
feasible to design a transport experiment that is extremely sensitive to the
torque term in Equation 6.1. It then becomes possible to probe the effects
of other “fields” that couple to spin, particularly those due to hyperfine and
spin-orbit coupling. Hyperfine effects in GaAs near the MIT have been stud-
ied extensively using optical pumping [10, 11], and it suffices for now to note
that the effective field acting on an electron spin due to dynamically polar-
ized nuclei can easily exceed 1 kG. As will be seen below, this hyperfine
field can influence the electron spin dynamics in a transport device pro-
foundly. In contrast, the effective spin-orbit fields in bulk GaAs are small,
and their effects on spin transport and dynamics are more subtle. The reader
is referred to the monograph of Winkler [12] for a complete discussion. The
only case that applies to the discussion of bulk GaAs in this chapter is the
effect of strain [13], which leads to spin-orbit fields that depend linearly on
momentum. We will not consider heterostructures, in which spin-orbit fields
linear in momentum appear due to electric fields from structural inversion
asymmetry (e.g. an applied gate voltage) [14, 15], or by modification of the
Dresselhaus Hamiltonian by confinement [16].
6.2 Spin Injection from Ferromagnetic Metals into Semiconductors 273
For the case of a uniform electric field, the drift velocity v d = µE , which
for modest electric fields (E  1V/cm) is on the order of 104 cm/s. In this
range, the spin drift length v d ts is comparable to the spin diffusion length,
and it is apparent that both the drift and diffusion terms in Equation 6.1
need to be considered. This is completely different from the situation in ordi-
nary metals, where drift is negligible. One might guess that the drift length
could become very large for transport near the MIT. In fact, however, τ s
drops precipitously due to impact ionization of donors [8] at modest electric
fields above 10 V/cm. This indicates a maximum spin drift length in n-GaAs
in the range of tens of microns. The physics near interfaces, where electric
fields can be both large and nonuniform, is potentially much richer [6, 17].
6.2 SPIN INJECTION FROM FERROMAGNETIC

METALS INTO SEMICONDUCTORS
As the previous section makes clear, the conditions for spin transport in
bulk GaAs near the MIT are rather favorable. The length scales required are
readily accessible with photolithography, and the magnetic fields required to
probe spin dynamics are modest. Nonetheless, significant progress on the
integration of ferromagnetic sources and detectors with semiconductors has
occurred only since the turn of the century. The discussion here will focus
on work since 2000. For a review of the status of the field at about that time,
see Ref. [18].
As some stock was taken of the situation around 2000, a few important
observations emerged. First, the standard description of spin transport in
metals [19–22], when applied in a straightforward manner to semiconduc-
tors [23], appeared to be consistent with the absence of spin accumulation
when electrons were injected from a ferromagnetic metal into a semicon-
ductor. The essence of this argument was that the spin current flowing into
the semiconductor is limited by the semiconductor’s relatively large resistiv-
ity, while the low resistance of the ferromagnetic metal is itself an efficient
sink for any spin accumulation at the interface.
It was immediately recognized that this obstacle to spin injection, often
referred to as the “conductivity mismatch problem”, could be circumvented
by inserting a tunnel barrier between the ferromagnetic metal and the
semiconductor [24–26]. In fact, this principle had already been exploited
in the earliest demonstration of spin injection into GaAs using a scanning
tunneling microscope with a ferromagnetic tip [27]. For FM/SC devices
based on Al xGa1−xAs, in which a Schottky contact is present anyway, the
only problem would appear to be engineering a barrier of the appropriate
thickness and height. Two general approaches have been followed: either
using a pure “tunneling Schottky” barrier [28–30], or creating an additional
artificial barrier by inserting an insulator at the FM/SC interface [31–33].
For the case of a high resistance barrier, ignoring the effects of dispersion,
and assuming that that spin polarization of the ferromagnet at the interface
with the semiconductor is equal to the bulk value, the polarization of the
tunneling current should be limited only by the spin polarization of the fer-
romagnetic metal.
Although adopting a tunnel barrier for the injection contact was an obvi-
ous choice, it did not provide an immediate solution to the problem of elec-
trical spin detection. At the same time, however, the community working
on the problem of spin injection from magnetic semiconductors into their
nonmagnetic counterparts developed a new version of the reliable lumines-
cence polarization technique employed in optical pumping experiments. By
making the magnetic semiconductor one contact in a p-i-n junction light-
emitting diode (LED), it was possible to convert a spin-polarized current of
either electrons [34] or holes [35] into polarized light. This so-called “spin-
LED” (originally proposed in the 1970s [36]) has since been adapted by the
community working on spin injection from FM metals and is discussed in
greater detail in Chapter 2, Volume 2. By placing the Schottky tunnel bar-
rier in series with the p-i-n diode, spin-polarized electrons flowing from the
metal recombine with unpolarized holes. For a quantum well, the degree of
polarization of the emitted light is equal to the spin polarization of the elec-
tron component of the recombination current, provided that the component
of the spin being measured is along the growth axis. Using this principle,
spin injection from Fe into GaAs was first reported by Zhu et al. in 2001 [28].
A significant improvement in the efficiency of the Fe/AlxGa1−xAs spin-LED
was then reported by the NRL group using a graded Schottky barrier [29].
In parallel, the development of spin-LEDs based on insulating tunnel bar-
riers inserted between the ferromagnet and semiconductor was being car-
ried out by the IMEC and IBM groups [31, 32]. In all of these experiments,
the extraction of the spin polarization of the tunneling current was based on
rate equation arguments that we will not discuss here [37–39]. By 2005, the
Schottky spin-LED was sufficiently well understood that it was being used for
the evaluation of new potential spin injection materials, including Heusler
alloys [40, 41], perpendicular anisotropy materials [42, 43], and others.
At the same time as spin-LEDs were being used to probe the polariza-
tion of electrons tunneling from a ferromagnet metal into a semiconductor,
a different line of research was devoted to the study of spin accumulation
due to unpolarized electrons incident on a FM/SC interface from the semi-
conductor. In 2001, Kawakami et al. [44] reported on spin accumulation at
the interface between (100) GaAs and MnAs (a ferromagnetic semimetal)
under optical pumping by linearly polarized light. This experiment as well
as a subsequent one on Fe/GaAs [45] utilized time-resolved Faraday rota-
tion, and hence could be carried out in simple bilayers of the ferromagnetic
metal with an n-type GaAs epilayer. Ciuti, McGuire, and Sham interpreted
this process using a tunneling approach [46, 47]. The spin accumulation
was enhanced in experiments in which a forward electrical bias was applied
while optically pumping [48], consistent with the tunneling hypothesis. The
tunneling argument implied that it should be possible to generate a spin
accumulation in semiconductors under forward bias even in the absence
of optical pumping, as was then observed by Stephens et al. [49] in MnAs/
GaAs bilayers.
6.2 Spin Injection from Ferromagnetic Metals into Semiconductors 275
For technical reasons, the majority of the spin-LED experiments dis-

cussed above probed the component of the spin accumulation parallel to an
applied magnetic field. In contrast, the experiments under optical pumping
and/or a forward-bias current probed the transverse component of the accu-
mulated spin polarization, which could be detected unambiguously through
its precession about the applied field. The exceptions among the spin-LED
experiments were those carried out by the IMEC [31, 50, 51] and Minnesota
groups [30, 37, 52], which were based on the precession of spins in the semi-
conductor after injection from the ferromagnet. Given the relatively short
spin lifetimes in quantum wells (several hundred picoseconds), significant
precession could be observed in these experiments only because of the large
hyperfine contributions to the effective magnetic field. Although somewhat
more complicated in its implementation, a steady-state measurement that is
sensitive to electron spin precession allows for a much stronger quantitative
analysis of the spin dynamics. We will call experiments of this type “Hanle
measurements”, as they are analogs of the Hanle effect commonly observed
in photoluminescence polarization under optical pumping by circularly
polarized light [1].
The experiments introduced in this section demonstrated that a spin
accumulation could be generated in GaAs under either forward or reverse
electrical bias across a FM/SC interface. There remained, however, signifi-
cant limitations. First, the effects demonstrated were local. That is, the spin
polarization was measured in the semiconductor at the location of the fer-
romagnetic source. Second, the detection techniques were strictly optical.
Although there were some earlier attempts to detect optically induced spin
accumulations electrically [53–55], they were not supported by the demon-
stration of a Hanle effect. A key point is that the tunnel barriers used in spin-
LEDs, which were perfectly adequate for spin injection, were not particularly
transparent to electrons at low bias voltages. The resistance-area product
of a typical Schottky barrier of the design of Hanbicki et al. [29] is about
105 - 106 W mm 2 , which is several orders of magnitude larger than the values
achieved in typical magnetic tunnel junctions. This means that in simple
two-terminal (FM-SC-FM) devices, the overall resistance is dominated by
the interfaces. Except for the limit in which the entire semiconductor acts
as a tunnel barrier [56, 57], the ordinary magnetoresistance (R↑↑ − R↑↓ ) / R↑↑
will be extremely small, as noted by Fert and Jaffrès [25]. This is in fact the
other side of the two-edged sword of the “conductivity mismatch” argument.
High resistivity barriers permit efficient spin injection, but the effectiveness
of a such a barrier for spin detection is limited by its capacity for sinking
the charge current. As detailed by Fert and Jaffrès, there is a narrow range
of optimal barrier resistances for two-terminal FM-SC-FM structures, but
these lie away from the practical values for Fe/GaAs devices with Schottky
barriers.
In the next three sections, we address how these shortcomings have
been addressed in lateral FM/GaAs devices, a process that has entailed three
principal steps. The first of these was establishing that spin accumulations
could be detected remotely from the ferromagnetic source (or drain) contact.
The experimental approach for this aspect of the project was similar to that
used by Stephens et al. [49]. Scanning Kerr microscopy and local Hanle–Kerr
effect studies were used to detect electron spins injected into a GaAs channel
from Fe contacts under reverse bias, as expected from spin-LED experiments
[9, 58, 59]. Also, spin accumulations generated by spin extraction at forward-
biased contacts was demonstrated. As discussed below, these optical stud-
ies were used to directly measure the relevant spin transport parameters
( ts , D , v d ), and established that the drift-diffusion model of Equation 6.1
provides a satisfactory description of the spin transport.
The second step was to show that Fe/GaAs Schottky barriers, although
highly resistive, could be used as electrical spin detectors [58, 60]. Electrical
detection of both optically-injected as well as electrically-injected spin were
demonstrated. These latter experiments were based on the serendipitous
discovery that it was possible to electrically generate and detect a spin accu-
mulation at a single FM/SC interface. As will be evident from the discussion
below, this was possible only because the measurement was sensitive to the
suppression of the spin accumulation by the Hanle effect.
The third step was the implementation of a completely electrical injec-
tion and detection scheme in a lateral geometry [61]. This experiment was
based on the non-local approach in which spin and charge currents are sepa-
rated in space. The overall philosophy follows that developed originally by
Johnson and Silsbee for metals [19, 20], and implemented by many other
groups in metallic nanostructures. The essential physics of the non-local
measurement and its interpretation are covered in Chapter 5, Volume 1 and
its application in more recent experiments on metals and graphene are cov-
ered in the chapters by Otani and van Wees. We will emphasize only those
aspects that are of critical importance in semiconductors.
6.3
Fe/GaAs HETEROSTRUCTURES AND
LATERAL SPIN-TRANSPORT DEVICES
In spite of the rapid refinement of spin-LEDs and the achievement of cor-
respondingly high spin injection efficiencies, the appropriate modification of
these devices for lateral transport experiments was not immediately appar-
ent. For some period of time, the conductivity mismatch argument and its
extension by Fert and Jaffrès [25] cast a pessimistic light over efforts to fabri-
cate a lateral FM-SC-FM spin valve. In cases where successes were reported
[62], the means for extracting information about dynamics (e.g. spin life-
times) quantitatively were not demonstrated. In particular, no Hanle effect
was observed.
Our own approach to this problem evolved from a purely technical
question: is it possible to image an electrically injected spin accumulation
in n-GaAs in a manner similar to that demonstrated for optical pumping
experiments in the presence of an electric field [63–65]? In contrast to a
spin-LED, in which electrons recombine with holes, luminescence is not a
viable detection technique in this case. In contrast, Faraday or Kerr rota-
tion are suitable probes of spin accumulation in n-type material. Given the
6.3 Fe/GaAs Heterostructures and Lateral Spin-Transport Devices 277
time and length scales for spin transport in n-GaAs samples doped near the
MIT, a relatively simple magneto-optical polar Kerr effect microscope is suf-
ficient, provided that the injected spins can be tipped perpendicular to the
sample plane (thereby enabling detection by the polar Kerr effect) without
perturbing the magnetization of the Fe contacts. This can be accomplished
by exploiting the large uniaxial anisotropy of the epitaxial Fe/GaAs (001)
interface [37]. This anisotropy, derived from the surface electronic structure,
leads to a magnetic easy axis along the [011] direction and allows for small
“tipping” magnetic fields to be applied along the [011] (hard) direction with-
out inducing significant rotation of the Fe magnetization. On the other hand,
once injected into the semiconductor, an electron spin experiences a torque
due to the small applied field and therefore precesses into the [100] direction,
allowing for detection by polar Kerr rotation [49]. The spin accumulation,
driven by lateral drift and diffusion, can then be imaged as it emerges from
underneath a ferromagnetic contact. (An alternative approach avoids the use
of precession by imaging the spin accumulation on the cleaved edge of the
sample. This has been implemented by Kotissek et al. [66]).
A cross-section of a typical FM/SC heterostructure is shown in
Figure 6.1a. The principal consideration in the design of these samples is the
integration of a Fe/GaAs Schottky tunnel barrier with a conducting chan-
nel, while keeping the spin diffusion length long enough so that fabrication
by photolithography and detection by optical Kerr microscopy are practical.
The doping profile chosen for the Schottky barrier was originally perfected
by Hanbicki et al. for spin LEDs [29], and the conduction band structure
near the Fe/GaAs interface is shown in Figure 6.1b. The heterostructures we
have used in all measurements described in this chapter are grown on semi-
insulating GaAs (100) wafers and typically consist of (from bottom to top) a
2.5 μm thick n-doped channel (Si donor concentration n  3 − 7 × 1016 cm −3 ),
a 15 nm transition region in which the doping is increased from the chan-
nel doping up to 5 × 1018 cm −3 and an additional 15 nm thick heavily doped
(5 × 1018 cm −3 ) region. After growth of the semiconductor layers, the sample
FIGURE 6.1 (a) A schematic cross-section of the heterostructures discussed in

this chapter. See the text for a discussion of the dopings, which vary slightly from
sample to sample. (b) Conduction band structure near the Fe/GaAs interface for
the doping profile used in this work [29].
is cooled to approximately room temperature, and the epitaxial Fe layer

is deposited. Typically, an As-rich c(4 × 4) reconstructed surface is used.
Finally, the structures are capped with Al and Au. Transmission electron
microscopy (TEM) studies have established that the interface between the
Fe and GaAs is atomically abrupt (within one or two monolayers) [67], and
all films have the required interfacial anisotropy, with the Fe easy mag-
netization axis along the [011] direction. The devices are prepared using
standard wet and dry etching techniques. Gold vias are deposited over SiN
isolation layers to contact the ferromagnetic electrodes. Micrographs of
samples used in the three types of experiments discussed in this chapter
are shown in Figure 6.2.
6.4 SCANNING KERR MICROSCOPY AND

LOCAL HANLE EFFECT STUDIES
6.4.1 Kerr Set-up
Although any of these Fe/GaAs structures shown in Figure 6.2 can be (and
have been) imaged with Kerr microscopy, the discussion in this section
focuses first on the simple two-terminal device shown in Figure 6.2a. The
“source” electrode is reverse-biased, meaning that electrons flow from the
FM into the SC. The “drain” electrode at the opposite end of the n-GaAs
channel is forward-biased. Figure 6.3 shows a schematic of the scanning Kerr
microscopy experiment. The Fe/GaAs spin transport devices are mounted,
nominally strain-free, on the vacuum cold finger of a small optical cryostat,
which in turn is mounted on a x-y positioning stage. The out-of-plane com-
ponent of electron spin polarization in the semiconductor, Sz , is measured
at a particular location via the polar Kerr rotation θ K of a linearly-polarized
probe beam that is focused onto (and reflected from) the sample at normal
incidence. The probe beam is derived from a narrow-band continuous-wave
Ti:sapphire ring laser, which is typically tuned just below the semiconduc-
tor bandgap in order to minimize the influence of the probe beam on the
resident electrons in the n-GaAs while maintaining high sensitivity to spin
polarization [59, 64]. The final probe focusing lens is also mounted on a
positioning stage, and two-dimensional images of Sz are obtained by raster-
scanning either the cryostat or (more typically) the probe focusing lens.
Electrically-injected spins are studied by modulating, at kilohertz fre-
quencies, the voltage bias applied to the Fe source/drain contacts, and then
measuring the induced change in θ K using lock-in amplifiers. As discussed
above, the spins are injected along the magnetic easy axis of the Fe, which
is the crystallographic [011] direction, or the x̂-direction in Figure 6.3. They
precess about the applied field B, and θ K is proportional to Sz at the position
of the focused laser spot. External coils control the applied tipping fields
Bx , By , and Bz . These coils are also used together with a fixed (local) probe
beam to study the relaxation and dephasing of spins at a particular spot on
the sample as a continuous function of an applied transverse magnetic field–
the “local Hanle effect”.
6.4 Scanning Kerr Microscopy and Local Hanle Effect Studies 279
FIGURE 6.2 Micrographs of representative Fe/GaAs spin transport devices used

for the experiments discussed in this chapter. (a) A simple two-terminal device
used in the original Kerr imaging studies of Ref. [58, 59]. (b) Hall bars used for the
detection of spin accumulation at a single Fe/GaAs interface [60]. Note how the
source, drain, and channel contributions to the two-terminal voltage Va−b can be
measured independently. (c) A typical non-local device, with a schematic of the
connections to a current source and voltmeter [61]. Kerr imaging of non-local
devices [9] is also discussed in the text.
The samples may also be held by a small cryogenic vise machined into
the cold finger [64]. The uniaxial stress applied to the sample by the vise
is uniform and can be varied in situ by a retractable actuator. For devices
grown on GaAs (100) substrates and cleaved along the usual 〈011〉 crystal
axes, this uniaxial stress along 〈011〉 leads to non-zero off-diagonal elements
FIGURE 6.3 (a) A schematic of the scanning Kerr microscope used to image electron spin transport in Fe/
GaAs devices. The measured polar Kerr rotation θK of the reflected probe laser beam is proportional to the
out-of-plane ẑ component of the conduction electron spin polarization, Sz . External coils (not drawn) control
the applied magnetic fields B x , B y , and Bz . (b) A representation of a Fe/GaAs spin transport structure in cross-
section. In this drawing, electrons tunnel from Fe into the n-GaAs with initial spin polarization S0 antiparallel
to the Fe magnetization M. A small applied transverse magnetic field ±B y is used to precess the injected spins
out-of-plane (along ± ẑ) so that they can be measured by the probe laser via the polar Kerr effect.
of the GaAs crystallographic strain tensor, ε xy . For electrons moving in the

x-y sample plane, ε xy couples directly to electron spin σ and momentum k
via a spin-orbit coupling energy ε xy (σ y k x − σ x k y ), which has the same form
[13, 64] as the well-known Rashba spin-orbit Hamiltonian. The resulting
effective magnetic field B ε lies in-plane and is orthogonal to the electron
momentum k.
6.4.2 Imaging Spin Injection, Local Hanle-Kerr

Studies, and Modeling
In Figure 6.4, we show Kerr-rotation images and the associated local Hanle
curves for electrically-injected spins. Figure 6.4a shows a spin transport
device having rectangular Fe/GaAs source and drain contacts at either end
of a 300 μm long n-GaAs channel. We image the 80 × 80 μm region shown
by the dotted square, which includes part of the Fe injection contact and the
FIGURE 6.4 (a) A lateral spin transport device consisting of a 300 µm long n-GaAs channel separating Fe/
GaAs source and drain contacts. The dotted square shows the 80 × 80 µm region imaged. (b) An image of
the reflected probe laser power, showing device features. (c) Images of electrical spin injection and transport
in the n-GaAs channel. The electron current Ie = 92µA. Electrons are injected with initial spin polarization
S0  − xˆ and B y is varied from −8.4 to +8.4 G (left to right), causing spins to precess out of and into the page
(± ẑ), respectively. (d) A series of local Hanle curves (θK vs. B y) acquired with the probe laser fixed at different
distances (8–120 µm) from the source contact (offset for clarity). (e) Simulated local Hanle curves using a 1D spin
drift-diffusion model Equation 6.2, with τ s = 125ns, v d = 2 .8 × 10 4 cm / s, and D = 10 cm2 / s.
bottom edge of the n-GaAs channel. An image of the reflected probe laser
power (see Figure 6.4b) clearly shows these device features. With this contact
under reverse voltage bias, so that spin-polarized electrons are injected from
the Fe into the channel, Figure 6.4c shows a series of Kerr-rotation images as
the applied field By is varied from −8.4 G to +8.4 G. Injected electrons, ini-
tially polarized along the − x̂ direction, precess into the + ẑ direction when
By is oriented along − ŷ. When By inverts sign and is oriented along + ŷ, the
spins are tipped into the opposite direction (along − ẑ ), reversing the sign of
the measured Kerr rotation. These injected electrons flow down the channel
with an average drift velocity v d that is the same in all the images. The drift-
ing spins precess at a rate proportional to | By |; thus, the spatial period of the
observed spin precession is shorter when | By | is larger.
When the probe laser is fixed at a point in the n-GaAs channel and
Sz ∝ θ K is measured as a continuous function of By , we obtain local
Hanle curves having the characteristic antisymmetric lineshape shown in
Figure 6.4d. These local Hanle curves are acquired in the n-GaAs channel
at different distances from the Fe source contact (8–120 μm), as labeled.
With increasing distance, the amplitudes of the curves fall, their character-
istic widths decrease, and they develop multiple oscillations associated with
multiple precession cycles. The detailed structure of this family of Hanle
curves contains considerable information about the dynamics of electron
spin transport, including the electron spin lifetime τ s , diffusion constant D,
and drift velocity v d [64, 68]. The Hanle curves invert when the magnetiza-
tion M of the Fe contacts is reversed, as expected.
To model these local Hanle data we consider a simple one-dimensional
picture of spin drift, diffusion, and precession in the GaAs channel. Spin-
polarized electrons, injected with S0 = Sx (at x = 0 and t = 0), precess as they
flow down the channel with a drift velocity v d , arriving at the point of detec-
tion x0 at a later time t = x0 / v d . At this point, Sz = S0exp( −t / τ s )sin(Ω Lt ),
where Ω L = g e µ B By / is the Larmor precession frequency. The actual signal
is therefore computed by averaging the spin orientations of the precessing
electrons over the Gaussian distribution of their arrival times (which has a
half-width determined by diffusion).
x0 + w ¥
S0
ò ò
2
S z ( By ) = e -( x - vd t ) /4 Dt ´ e -t / ts sin(W Lt )dtdx , (6.2)
x0 0 4 pDt
where:
v d = mE is the drift velocity
μ is the electron mobility
E is the electric field in the channel
The spatial integral accounts for the width w of the source contact. This
type of averaging is the basis of the Hanle effect observed in optical pumping
experiments and in previous spin transport experiments in metals [20, 69]
and semiconductors [30, 49, 50]. Using this model, Figure 6.4e shows that good
overall agreement with the local Hanle data is obtained using D = 10 cm2/s,
v d = 2.8 ´ 10 4 cm/s, and τ s = 125 ns . The large drift velocity and spin lifetime
in these devices allow access to a spatial regime far from the contacts and
well beyond a spin diffusion length (x0 > Dτ s ), in which the average time-
of-flight from the source to the point of detection, T = ( x0 + w / 2) / v d , deter-
mines the characteristic “age” of the measured spins. In this limit, the first
peak in the data (By = Bpeak ) is the field at which electrons have precessed
through one-quarter Larmor cycle, so that T = π / (2Ω L ) = π / (2 g e µ B Bpeak ) .
Thus, in Figure 6.4d, Bpeak = 1.05 G when x0 = 88 µm, indicating that
T  380 ns and v d  2.8 ´ 104 cm / s .
Simulating lateral spin flows in more complicated device geometries
requires more sophisticated models of spin drift and diffusion. Two-
dimensional models generally suffice for planar devices in which spin
transport occurs in epilayers that are thinner than the characteristic spin
diffusion lengths. For the case of applied in-plane magnetic and electric
fields (Bx , y , E x , y ), the steady-state spin polarization S(r) can then be derived
from the two-dimensional (2D) version of the spin drift-diffusion equation.
Although Equation 6.1 in 2D can be solved analytically in certain limits [9],
it is often easier to compute Sz from a set of coupled spin drift-diffusion

equations in two dimensions using numerical Fourier transform methods,
particularly when including spin-orbit coupling terms [64, 68]. For in-plane
electric and magnetic fields, and for spin-orbit coupling due to off-diago-
nal strain ε xy in bulk GaAs, the three equations determining the steady-
state spin densities Sx , y , z (r) are O1Sx + O2 Sz = −Gx , O1S y + O3 Sz = −G y , and
O4 Sz − O2 Sx − O3 S y = −Gz , where
O1 = D∇r2 + µE ⋅ ∇r − C s2 D − 1 / τ s , (6.3)
O2 = g e µ B By / + C s (2 D∇ x + µE x ), (6.4)
O3 = − g e µ B Bx / + C s (2 D∇ y + µE y ), (6.5)
O4 = D∇r2 + µE ⋅ ∇r − 2C s2 D − 1 / τ s . (6.6)
Gx , y , z (r) are the generation terms, and C s = C3mε xy / 2 is the spin-orbit term
that couples spin to the off-diagonal elements of the strain tensor in GaAs,
ε xy [13, 68].
The presence of effective magnetic fields (Beff = B ε ) due to spin-orbit
coupling causes moving spins to precess, even in the absence of applied
magnetic fields [64, 70]. The effective magnetic fields are directly revealed
in local Hanle measurements. For example, for electrons flowing in the x̂
direction as shown in Figure 6.4, B ε is directed along ± ŷ. B ε therefore either
directly augments or opposes the applied (real) magnetic field By that is used
to acquire the local Hanle curve. The net effect is therefore to shift the Hanle
curve to the left or right, so that it is no longer antisymmetric with respect
to By . An example of this is shown in Figure 6.5a, which shows local Hanle
data acquired with increasing uniaxial stress applied to the device substrate
along the [011] direction, which effectively turns on a Rashba-like spin-orbit
interaction in the bulk n-GaAs channel of the device. The corresponding
simulations shown in Figure 6.5b were performed using the numerical
approach described above, and qualitatively reproduce the data when the
applied strain ε xy = 0, 1, and 2 × 10−4.
6.4.3 Imaging Spin Injection and Extraction, Both

Upstream and Downstream of the Current Path
Scanning Kerr miroscopy can also be used to image the spin currents that
exist when electrically-injected spins diffuse outside of the charge-current
path. This is demonstrated on a “non-local” Fe/GaAs spin transport device
of the type shown in Figure 6.2c, and again (with contact labels and wiring
diagram) in Figure 6.6a. These devices have five Fe contacts in a n-GaAs
channel, with easy-axis magnetization M along the ŷ direction. We focus
on spin injection and spin extraction at contact 4. The dotted square in
Figure 6.6a shows the 65 × 65 µm region around contact 4 that was imaged.
The reflected probe power in this region (Figure 6.6b) clearly shows contact 4,
FIGURE 6.5 (a) Local Hanle effect (Sz vs. B y ) from electrically-injected spins in
the presence of spin-orbit coupling and effective magnetic field B ε due to strain.
The device is the same as shown in Figure 6.4. These data are acquired at a point
4 µm from the source contact, for three values of uniaxial stress applied along the
[011] substrate axis. Curves shift to the left with increasing stress, as the effective
magnetic field B ε  ŷ increases due to the induced shear strain ε xy . (b) Simulated
local Hanle data, for ε xy = 0,1,2 × 10 −4.
FIGURE 6.6 (a) A spin transport device with five Fe/GaAs contacts (#1–5). A 65 × 65 µm region around
contact 4 is imaged (dotted square). (b) The reflected probe power in this region, showing device features.
(c, d) Images of the majority spin polarization that is electrically injected from contact 4 (B x = 2 .4 G). Arrows
show the direction and magnitude of net electron current, Ie. In (c), spins are injected and drift/diffuse to the
right (Ie4→5 = 25 µA ). In (d), spins are injected and drift to the left (Ie4→1 = 25 µA); the spins to the right of contact 4
are diffusing outside of the current path. (e, f) Images of minority spin accumulation due to spin extraction at
contact 4, both within and outside of the charge current path.
the n-GaAs channel, and the edges of the SiN layer. Under bias, the electron
current I e flows within the imaged region when electrons flow between con-
tacts 4 and 5. In this case the electric field in the channel, E x , is non-zero in
the imaged region and the drift velocity v d = mE x may either augment or
oppose electron diffusion away from contact 4, depending on the direction
of I e . Conversely, when I e flows between contacts 4 and 1, the current path is
outside of the imaged region. In this case, E x is nominally zero in the imaged
region regardless of I e , and spin transport in this region should be purely
diffusive.
Figure 6.6c and d show the imaged spin polarization for the case of spin
injection (that is, the Fe/GaAs Schottky contact is reverse-biased and elec-
trons flow from Fe into n-GaAs). In this device, majority spins having spin
orientation antiparallel to the Fe magnetization M are injected for all reverse
biases. The images show two cases, in which charge current path lies within
or outside of the imaged region. For clarity, contact 4 in these images is out-
lined by a dotted rectangle, and the black arrows indicate the direction of
the electron current I e . In Figure 6.6c, injected electrons drift slowly to the
right along + x̂. The electron current I e4 →5 =25 µA and the voltage across the
Fe/GaAs interface is Vint  53mV . All of these images show the difference
between Kerr images acquired at Bx = +2.4 and −2.4 G. In this way, field-inde-
pendent backgrounds are subtracted off, and only the signal that depends
explicitly on electron spin precession remains. Figure 6.6c clearly shows a
cloud of spin-polarized electrons emerging from and flowing away from con-
tact 4, with an apparent spatial extent of order 15 µm in the n-GaAs chan-
nel. Note that the apparent extent of Sz in these images depends on | Bx |,
as shown earlier in Figure 6.4. In this case, both drift and diffusion drive a
net flow of spins to the right.
A remarkably similar image, however, is observed in Figure 6.6d for the
case in which spins are injected out of contact 4 under similar bias condi-
tions (I e4 →1 = 25 µA) but are drifting to the left, along − x̂. In this case the
current path is outside the imaged region, and E x in the imaged region is
nominally zero. The cloud of spins seen to the right of contact 4 are those
spins that have diffused out from under the contact, and which are now flow-
ing to the right due to diffusion alone. The similarity of both the magnitude
and the spatial extent of Sz in Figure 6c and d indicates that diffusion, rather
than drift, plays the major role in determining the overall transport of spins
under these low injection bias conditions. These images show that robust
spin currents, spatially separated from any charge currents, can be generated
in the n-GaAs channel. It is precisely these “upstream” spin currents that we
will measure using all-electrical non-local spin detection, as discussed later
in the chapter.
In addition to electrical spin injection from a ferromagnet into a semi-
conductor, a spin polarization can also accumulate in a semiconductor
when electrons flow from the semiconductor into the FM at a forward-
biased Schottky barrier [49, 58]. Figure 6e and f show the corresponding
set of images for the case of spin accumulation due to spin extraction out of
contact 4 in this Fe/GaAs device. At this bias, the preferential tunneling of

majority electrons from n-GaAs into contact 4 leaves behind an accumula-
tion of minority-spin electrons in the channel, as evidenced by the opposite
sign of the Kerr rotation signal. This accumulated minority spin polariza-
tion diffuses out from under the contact, either into the current path (where
drift now opposes diffusion), or outside of the current path (where spins
diffuse freely).
Figure 6e and f show spin accumulation due to spin extraction both
within and outside of the current path under low forward bias condi-
tions (I e5→ 4 = I e1→ 4 = 25 µA). In both cases, the apparent spatial extent of
the accumulated Sz is similar to the case of purely diffusive spin trans-
port. Local Hanle data confirm this, showing decay lengths of order 10 µm.
Using these images and the associated local Hanle data (not shown), the
important spin-transport parameters—τ s , D , and v d —can be accurately
determined as a function of temperature and bias conditions. Knowledge
of these parameters is especially useful for interpreting and understanding
all-electrical measurements of spin injection and spin detection in similar
devices.
6.5 ELECTRICAL DETECTION OF
SPIN ACCUMULATION
6.5.1 Electrical Detection of Optically Injected Spins
As shown by the Kerr imaging, a spin accumulation due to either spin injec-
tion or extraction forms at the Fe/GaAs interface under reverse or forward
bias respectively. A rough calibration of the sensitivity of the Kerr micro-
scope indicates that spin polarizations of 5–10% exist within a few microns
of the injection electrode. We now turn to the question of how this spin
accumulation can be detected electrically.
An obvious experiment to carry out along these lines is to inject spin-
polarized electrons into GaAs by optical pumping [1] and then detect the
correspond change in the conductance of a nearby FM/SC interface. As
noted previously, related ideas had been pursued in earlier experiments on
vertical FM/SC structures [53–55]. However, because in these studies the
semiconductor was optically excited through the semitransparent FM con-
tact, the observed signals were always comparable to background effects due
to hot electrons, and from magneto-absorption of the pump light by the FM
contact. Moreover, no Hanle effect was observed. These shortcomings can
be overcome in lateral devices by injecting spin-polarized electrons into the
n-GaAs channel, away from the FM contact [58, 71]. Doing so avoids mag-
neto-absorption effects and allows the spins to cool before they drift and dif-
fuse under the Fe contact. A schematic showing the implementation of this
experiment is shown in Figure 6.7a, where we optically inject spin-polarized
electrons, initially oriented along ± ẑ , into the channel using circularly polar-
ized modulated light as depicted. Spins are injected at a small spot approxi-
mately 40 µm from the edge of the forward-biased Fe drain contact. A Kerr
FIGURE 6.7 (a) A schematic diagram of an experiment showing electrical detection of optically injected
spins. Modulated circularly polarized light produced by passing a linearly polarized pump beam (50 µW at
λ = 785 nm) through a photoelastic modulator (PEM) is incident on the semiconductor channel. The spin-
dependent change in device voltage is measured using lock-in techniques. (b) A Kerr image of these optically-
injected spins diffusing and drifting towards the drain contact. B y = 0. The spins are generated in the small
dot approximately 40 µm from the electrode. (c) Spin-dependent voltage change as a function of B y , which
induces precession of the optically injected spins parallel or antiparallel to M. Data for the two different mag-
netization directions of the detector are shown.
image of these spins in zero field is shown in Figure 6.7b. Small applied mag-
netic fields By cause these spins to precess as they drift and diffuse towards
the drain, tipping them into a direction parallel or antiparallel to the Fe mag-
netization M. The spin-dependent change in the conductance is then mea-
sured by lock-in techniques. Figure 6.7c shows, as a continuous function of
By , the spin-dependent change in the voltage across the device that is explic-
itly due to the presence of the optically-injected and spin-polarized electrons
at the drain contact. The signal, which reverses when M is reversed, shows
the same antisymmetric local-Hanle lineshape observed in the imaging of
electrical injection, as expected. These data demonstrate that these Schottky
tunnel barrier contacts are sensitive to the spin polarization of the electrons
in the GaAs, and therefore can be used as electrical spin detectors. In these
studies, the spin-dependent voltages are on the order of microvolts, which
represents a change of only a few parts per million of the total voltage across
the device.
6.5.2 Electrical Detection by the Three-

Terminal Technique
When the Kerr imaging experiments discussed above were originally being
performed, the two-terminal resistance of these Fe/GaAs/Fe lateral devices
was also measured, but we found that most of the magnetic field dependence
was due to anisotropic magnetoresistance (AMR) in the Fe electrodes, as
well as a dependence of the effective tunneling resistance on the orientation
of the magnetization of the contact with respect to the crystalline axes. The
latter effect, which is now known as tunneling anisotropic magnetoresis-
tance (TAMR) [72, 73], could be as large as 1% of the device resistance. While
attempting to characterize these backgrounds, we carried out measurements
on devices with the electrodes several hundred microns apart, for which we
reasoned that only the background AMR and TAMR effects should appear.
A micrograph of one of these structures is shown in Figure 6.2b. When a
small magnetic field was applied perpendicular to the sample, a peak in
the two-terminal resistance appeared at zero magnetic field, as shown in
Figure 6.8. At this field scale ( 100 G), the magnetization of the contact was
unchanged, and so the effect could therefore be attributed to some effect in
the semiconductor. Using the additional contacts on these devices, it was
also possible to measure the voltage drops across the reverse-biased source
contact, the forward biased drain contact, and the channel individually. In
the vast majority of cases, the peak occurred only in a measurement across
the forward-biased drain, and never appeared in the channel voltage.
Given the correspondence with the field scales observed in the Kerr
imaging experiments, we hypothesized that the signature is due to dephasing
of the spin accumulation at the Fe/GaAs interface, and subsequent experi-
ments have verified that this is the case. The magnitude of the peak is the
FIGURE 6.8 (a) The voltage measured across a forward-biased Fe/GaAs interface as a function of perpen-
dicular magnetic field. A constant offset of 0.36 V has been subtracted from the data. The measurement
geometry is shown in the inset. (b) The magnitude of the zero-field peak is shown as a function of tempera-
ture for several different forward bias current densities.
enhancement of the voltage drop across the tunnel junction due to the pres-
ence of a spin accumulation, which is destroyed (dephased) as the magnetic
field is applied. This measurement and its interpretation are discussed in
more detail in Ref. [60]. We represent the Fe/GaAs interface by a one-dimen-
sional strip of length w. Spins polarized along x̂ are generated at a point x1 on
the interface and diffuse to another point x2 , at which they are detected. The
average steady-state spin accumulation is obtained by integrating the steady-
state solutions of the drift diffusion equation over x1 and x2 :
w w ¥
S0
òòò
2
Sx ( Bz ) = e -( x1 - x2 - v d t ) /4 Dt ´ e -t / ts cos(W Lt )dtdx1dx2 , (6.7)
0 0 0 4 pDt
where S0 is the spin generation rate (per unit contact length) and the other
parameters are the same as in Equation 6.2. As with the imaging measure-
ments, most of the parameters describing this Hanle curve can be obtained
by independent transport or optical characterization. A typical fit is shown
in Figure 6.8.
This single-contact measurement, often called a “three-terminal” exper-
iment, might appear to be a special case, particularly when, as has now been
demonstrated in many FM/GaAs heterostructures, measurements with a
separate source and detector are possible. There are some cases, however,
such as that discussed by Tran et al. [74], in which this approach allows the
experimentalist to work with highly resistive tunnel barriers that cannot be
probed by other techniques. Outside of the formalities of the drift-diffusion
model, this type of measurement has a relatively simple interpretation in
terms of two time scales: a residency time τ r between spin generation and
detection, as well as the spin lifetime τ s . Some simple statements can be
made by direct analogy with a traditional optical Hanle effect experiment
based on polarized photoluminescence [1] and the corresponding rate equa-
tions. The magnitude of the spin accumulation at zero field is proportional
to 1 / (1 + τ r / τ s ) , while the width of the Hanle curve is proportional to the
total relaxation rate 1 / τ r + 1 / τ s . Expressed in this way, the tradeoff in the
case of a resistive tunnel barrier is clear: the signal becomes small because τ r
is long, but the width of the Hanle curve can still be quite narrow (as it will
depend only on τ s ). This makes a small spin accumulation easier to measure
in the sense that (a) it can still be inferred from the magnetic field depen-
dence, even if its magnitude is small; and (b) the transport properties of the
bulk semiconductor are not important provided that it does not function as
a strong source of spin relaxation.
The three-terminal technique has continued to be applied to various
other FM/SC heterostructures, due in large part to the fact that it is easy
to implement. It has played a central, but controversial, role in the case of
silicon-based devices with artificial tunnel barriers, in which observation of
a Hanle effect at room temperature was first reported [75]. On the other
hand, the observation of a peak in the three-terminal resistance in zero (per-
pendicular) field has now been observed in so many different cases that its
utility as a probe of spin transport in the semiconductor channel has been
called into question [76, 77, 78, 79, 80]. Some of this discussion has focused
on the effect observed when the field is applied in the plane, for which the
standard interpretation based only on the Hanle effect and a uniform in-
plane magnetization does not apply [81]. In the case of silicon and germa-
nium, similar spin lifetimes were extracted for p and n-type material [82, 83],
which contradicts standard expectations. Nonetheless, the ubiquitous char-
acter of this effect, including its observation for junctions with and with-
out artificial tunnel barriers as well as other hybrid structures such as FM/
SrTiO3 [84, 85] and FM/LaAlO3/SrTiO3 [79], has led to continued interest
in the three-terminal approach as a simple means to probe spin-dependent
transport, whether or not it is a signature of spin transport in the bulk of
the semiconductor. In the case of Schottky barrier-based FM/III–V semi-
conductors, where non-local measurements of the type discussed below are
now carried out routinely, and for which hyperfine effects (discussed below)
serve as a fingerprint of bulk spin transport, the origin of the three-terminal
signal is less controversial.
6.6 NON-LOCAL DETECTION OF
SPIN ACCUMULATION
We now turn to the overall goal of a device in which a spin accumulation is
generated at one FM electrode, transported, and then detected with a sepa-
rate electrode. If the reader consults previous reviews of transport in fer-
romagnet-semiconductor heterostructures [18], a few clear themes emerge.
First, backgrounds from local Hall effects, anisotropic magnetoresistance,
and tunneling anisotropic magnetoresistance complicate virtually any mea-
surement in which a charge current flows through a ferromagnetic detec-
tion electrode. Second, the most definitive way to establish the existence of
a spin accumulation is to probe its dynamics. The first of these points can be
addressed by separating the charge current in a device from the spin current.
The second is addressed naturally by the observation of a Hanle effect.
The classic means by which spin and charge currents are separated is
the non-local geometry of Johnson and Silsbee [19, 20], which is illustrated
in Figure 6.9. Spins are introduced at a FM injector by an ordinary spin-
polarized charge current. The charge current is sunk at an electrode at one
end of the channel, but the spin accumulation created at the injector dif-
fuses in both directions. A second FM contact, the detector, is located on
the opposite branch of the channel, outside the path of the charge current.
The lower part of Figure 6.9 shows a spatial profile of the spin-dependent
chemical potentials for the ideal case in which no electric field (and hence
no charge current) flows in the right-hand branch of the channel. The detec-
tor itself acts as a spin-dependent voltmeter, with chemical equilibrium
established by the flow of a pure spin current between the FM detector and
the channel. If the FM detector were a half-metal (with the Fermi level in
a single spin band and therefore unity polarization), the electrochemical
potentials for each of the two magnetization states ↑ and ↓ of the detector
FIGURE 6.9 A schematic representation of a non-local spin transport device.

F1 is the ferromagnetic injector, F2 is the detector, and SC is the semiconductor
channel. Spin S diffuses in both directions from the injector contact. A representa-
tion of the spin-dependent chemical potentials µ ↑ and µ ↓ is shown for the ideal
case in which the electric field vanishes to the right of the injector. The voltage V is
measured for both magnetization states of the detector.
would be equal to µ ↑ and µ ↓ . The voltage difference ∆V = (µ ↑ − µ ↓ ) / e would

then provide a direct measurement of the spin accumulation. In practice,
the polarization PFM of the FM at the Fermi level is less than one, and spin
flips can occur as carriers flow back and forth across the detector interface.
The latter process is accounted for by an efficiency factor η, which we can
estimate from spin-LED measurements to be of order 0.5 for Fe/GaAs [38].
We make the assumption that the spin current flowing into (or out of) the
detector does not impact the measurement, which corresponds to the limit
of high-interface resistances. (In practice, the resistance of all of the devices
discussed here are dominated by the interfaces. This is not necessarily the
case in metallic devices, as discussed in Chapter 5, Volume 1).
The spin-dependent electrochemical potential is inferred from two non-
local voltage measurements V↑↑ and V↑↓ obtained with the detector and
injector magnetizations parallel and antiparallel, respectively. The difference
∆V NL = V↑↓ − V↑↑ can be related to the spin-dependent chemical potential
difference ∆µ by
e∆V NL
∆µ = , (6.8)
ηPFM
where PFM is the spin polarization of the ferromagnet at the Fermi level. The
electron spin polarization PGaAs in the semiconductor can be obtained from
∆µ, the carrier density n, and the density of states ∂n / ∂µ :
n↑ − n↓ ∆µ  ∂n 
PGaAs = = . (6.9)
n↑ + n↓ n  ∂µ 
Some indication of the overall efficiency is the non-local resistance ∆V NL / I inj,
which is proportional to the polarization per unit carrier flowing across
the interface, which we denote Pj . We emphasize here that our samples are
not truly degenerate semiconductors, and we therefore do not have direct
knowledge of the density of states from the carrier density. One can, how-
ever, make a crude estimate of the order of magnitude of ∆V NL if we assume
a Pauli-like density of states for a degenerate semiconductor of carrier den-
sity n  3 × 1016 cm −3 and an effective mass m* = 0.07me . Given the injected
polarizations from spin-LED measurements, and the approximate spatial
dependence PGaAs ( x) ∝ e − x / λ s , a conservative estimate of the spin polarization
at a detector several microns from the source is PGaAs  0.01. We estimate the
efficiency factor η  0.5 from spin-LED measurements [38]. Using a litera-
ture value PFM = 0.4 for the spin-polarization of iron [86], we find
ηPFM 2 (3π 2n)2/3

∆V NL  PGaAs = 10 µV. (6.10)
3m*e
This is a large signal by the standards of a typical transport measurement. In
our best devices, a noise floor of order 10 nV / Hz can be reached, although
occasional devices will simply be too noisy to resolve the spin-dependent
chemical potential shift. In practice, most of the difficulties in the experi-
ments discussed below derive from the non-ideal aspects of the non-local
measurement. In contrast to the cartoon representation of Figure 6.9, a
significant voltage drop (up to several mV) will develop at the non-local
detector due to spreading of the charge current from the source. This spin-
independent background depends relatively strongly on magnetic field and
temperature. Careful cryogenic practice, particularly temperature regula-
tion, is necessary in order to obtain a stable background voltage throughout
a measurement.
We now turn briefly to a discussion of the Hanle effect. Since the spin
lifetimes and diffusion constants for these samples are known from the opti-
cal measurements discussed previously, it is possible to solve Equation 6.1
in steady-state, given the geometry of the source and detector. As was noted
above, the typical field scale (width of a Hanle curve) is on the order of 100 G.
For fields applied in the vertical direction, this will have no impact on the
magnetizations of the contacts. It should therefore be possible to prepare
the source and detector in either parallel or antiparallel states, and obtain
Hanle curves in both cases. The large interfacial anisotropy of our samples
facilitates this measurement, since the Fe contacts remain uniformly mag-
netized at remanence.
A micrograph of a typical non-local device is shown in Figure 6.2c.
Photolithography and a combination of wet and dry etching are used to pat-
tern a large mesa and a series of FM electrodes. Two of these are located over
150 µm from the center of the device and serve as the sink for the charge
current, and the reference contact for the non-local voltage measurement.
The bias current is generated with a current source and the non-local volt-
age V NL is measured with a nanovoltmeter. Data for a device with channel
doping 5 × 1016 cm −3 obtained at 60 K are shown in Figure 6.10. Figure 6.10
shows raw data obtained in the longitudinal (or “spin-valve”) geometry, in
which the spin-dependent non-local voltage depends only on the relative
orientations of the source and detector. The expected discontinuities when
FIGURE 6.10 Non-local electrical detection of spin accumulation at T = 60 K under a forward bias current of
4 mA ( J = 1 kA / cm2 ). The injection contact is 5 × 80 µm2 and the detector contact is 5 × 50 µm2. The center to
center spacing of the contacts is 8 µm. (a) Raw data obtained in the longitudinal (or spin-valve) geometry. (b)
Raw data obtained in the perpendicular (or Hanle) geometry for both parallel and antiparallel orientations of
the detector and injector. The dashed line is a fit of the background obtained by a parabolic fit of the data at
high fields. (c) The spin-valve data after subtraction of the spin-independent background. (d) Hanle data after
subtraction of the parabolic background. The solid curves are fits to the drift-diffusion model. The lifetime
τ s = 4 ns and the spin diffusion length λ s = 4 µm.
the source and detector switch between parallel and antiparallel states are
superimposed on a background voltage (approximately −165 µV in this case)
due to the small amount of unwanted charge current spreading in the non-
local geometry. This background is removed in Figure 6.10c. Data obtained
in a perpendicular field (Figure 6.10b) show the Hanle effect when the source
and detector are prepared in either the parallel or antiparallel states. In this
case, there is a parabolic background due to the Lorentz forces acting on
the carriers, but, significantly, the data for the two spin states merge at high
fields, where the spins are completely dephased. After subtraction of the
spin-independent background (dashed curve in Figure 6.10b), the Hanle
curves of Figure 6.10d are obtained. As expected, these two curves are equal
and opposite, and the difference in their peak values is equal to the mag-
nitude of the jump in the spin-valve data. Curves showing fits to the drift-
diffusion model are also shown in Figure 6.10d. In this case, we find a spin
lifetime τ s = 4 nsec and a spin diffusion length λ s = 4 µm.
It is also possible to determine the spin diffusion length by measuring
the dependence of ∆V NL = V↑↓ − V↑↑ on contact separation in multitermi-
nal devices. By putting the detection contact in the current path (and sub-
tracting out the additional background due to the ohmic voltage drop), it is
FIGURE 6.11 The spin-dependent voltage V↑↓ − V↑↑ is shown as a function of

the separation between the injection and detection electrodes in the device with
several injection/detection contacts. Positive separations correspond to the pure
non-local geometry, in which the detection contact is outside the current path. For
negative separations, the detector is located in the path of the charge current, and
the signal is enhanced by drift.
also possible to measure the voltages V↑↓ and V↑↑ as well as the Hanle effect
“down-stream” from the injector, where both diffusion and drift contribute
to the spin transport. An example is shown in Figure 6.11, which shows ∆V NL
as a function of the injector/detector separation for a spin detector located
outside the path of the charge current (contact separation d > 0), and in the
current path (d < 0). The enhancement by drift in the latter case is clearly
evident.
Demonstrations of the non-local detection of spin accumulation have
been made by other groups in Fe/GaAs [87, 88] as well in other semiconduc-
tor-based systems, including MnAs/GaAs [89], Si [90], GaMnAs/GaAs [91],
NiFe/InAs [92], and graphene [93, 94]. In many (but not all) cases, claims
in these experiments have been supported by the observation of a Hanle
effect, which has also been invoked extremely effectively in spin transport
measurements on other spin-transport devices, particularly Si spin-valve
transistors [95, 96]. We now turn to some aspects of these measurements
that are less well-understood.
6.6.1 Bias Dependence of the Accumulated

Spin Polarization
The overall phenomenology of the non-local measurements is in general
agreement with expectations based on the drift-diffusion model for spin
transport in the semiconductor. We now turn to aspects of the data that
reflect the important and less understood role of the Fe/GaAs interface
itself. Figure 6.12a shows a series of non-local spin-valve measurements
as a function of longitudinal magnetic field for different injection cur-
rents. As expected, there are jumps at the transitions between parallel and
FIGURE 6.12 (a) The non-local spin-valve voltage is shown as a function of the longitudinal magnetic field
for several different bias voltages. Note that the signal inverts sign twice. (b) The spin-dependent non-local
voltage ∆VNL = V↑↓ − V↑↑ (closed squares, left axis) obtained from field sweeps is shown as a function of the
interfacial voltage Vint across the Schottky barrier injection contact. The Hanle–Kerr rotation signal θK (open
circles, right axis) measured on this device as described in Section 4 is shown for the same bias conditions.
(c) The polarization Pj of the interfacial tunneling current is proportional to ∆VNL (Vint ) / I(Vint ), which is shown as
a function of Vint . Note the change of sign at approximately −0.1 V.
antiparallel states of the injector and detector. (The additional peak at zero
field is due to hyperfine interactions and will be discussed below). If the
polarization of the tunneling current were fixed (e.g. by the bulk value of
PFe ), then the sign of the spin accumulation should change at zero bias. This
is indeed the case, but the sign of ∆V NL reverses again at a modest bias
current of −60 µA (forward bias), equivalent to a voltage drop Vint = −0.1 V
measured between the GaAs channel and the Fe injector. A more complete
picture can be seen in Figure 6.12b, which shows ∆V NL as a function of Vint .
The same figure also shows the Kerr rotation θ K measured in the same
device (in the channel “upstream” of the injection contact) as a function of
Vint using the methods of Section 4. Up to a scale factor, the two measure-
ments agree, confirming that the sign change at −0.1 V is due to a change
in the polarity of the spin accumulation. The Kerr data also determine the
absolute sign of the spin accumulation, which is positive (or majority) for
positive values of θ K . Another representation of the data is provided in
Figure 6.12c, in which we show the effective spin polarization of the cur-
rent: Pj = ( j↑ − j↓ ) / ( j↑ + j↓ ) ∝ ∆V NL (Vint ) / I (Vint ). In this form, it is evident
that the spin polarization of the current depends very strongly on the injec-
tor bias voltage, with a peak near zero bias and a reversal in sign at −0.1 V.
The bias dependences of ∆V NL and the Kerr rotation θ K are shown for
two additional samples in Figure 6.13, which illustrates that for a given sign
of the injector bias voltage, the magnitude and sign of ∆V NL (closed squares)
FIGURE 6.13 Bias dependences of ∆VNL (closed squares, left axis) and Kerr
rotation θK (open circles, right axis) for two different samples. As discussed in the
text, the bias dependences of θK and ∆VNL in (b) agree only after the sign of ∆VNL is
inverted (open squares).
can vary from heterostructure to heterostructure. In contrast, all devices

grown from a single heterostructure (i.e. those based on a single growth
of an epitaxial Fe/GaAs interface) will show approximately the same bias
dependence. Interestingly, the bias dependence can be changed by anneal-
ing, which can even invert the sign of the non-local signal [97]. Regardless
of annealing history, however, majority spin accumulation is observed for
large forward and reverse bias currents in all cases. At first glance, this wide
variation among heterostructures, particularly the overall sign, appears
counterintuitive. In spin-LEDs, for example, the polarization of the current
at reverse bias is always observed to have the correct sign for majority spin
injection [38]. A first clue as to what is going on can be found in the Kerr
rotation data of Figure 6.13. As was observed for the data of Figure 6.12b,
the bias dependences of θ K and ∆V NL for the heterostructure of Figure 6.13b
overlap nearly exactly, but for the sample of Figure 6.13c, θ K overlaps with
−∆V NL . In other words, the sign of the non-local voltage for a given spin
accumulation is reversed between the two samples. Otherwise, the ∆V NL is
proportional to the spin accumulation, as expected.
Focusing on the Kerr rotation, from which we can unambiguously
determine the absolute sign of the spin accumulation, we always observe
majority spin accumulation under large reverse bias (greater than +50 mV).
In other words, a majority spin polarization flows from Fe into GaAs, as
expected. Under forward bias, we observe majority spin accumulation for
biases greater than (i.e. more negative than) −100 mV. This corresponds to
the flow of minority spin from the semiconductor into the ferromagnet.
In other words, the relative transmission probabilities for minority and
majority spins always reverses near zero bias. This can be seen explicitly in
Figure 6.14, which shows the polarization of the current Pj ∝ ∆V NL (Vint ) / I int
for the same two samples as in Figure 6.13. Based on the Kerr rotation
measurements, we have chosen the sign of the polarization to be positive
at large positive (reverse) bias voltages. Although the polarization of the
current at zero bias can be either positive or negative, the polarization at
large reverse bias is always positive, and it is always negative at large for-
ward bias.
The absolute sign of the non-local signal, or, equivalently, the sign of Pj
at zero bias, is determined by the tunneling density of states along with the
exact energy at which the Fermi level is pinned at the Fe/GaAs interface.
The most direct access we have to this information comes from the Kerr
rotation, which measures the spin accumulation directly (to the extent that
θ K ∝ PGaAs ), and hence the quantity δPGaAs / δVint for a small change δVint in
the bias. Hence, the slope of θ K vs. Vint at zero bias should serve as predictor
for the signs of both ∆V NL (which is always measured at zero bias across the
detector) and Pj . This is in fact the case. Although this self-consistency is
reassuring, the fact remains that we cannot predict the spin polarization at
FIGURE 6.14 The spin polarization of the tunneling current flowing from the
semiconductor into the ferromagnet for the two samples of Figure 6.13. Note they
have opposite sign at zero bias. Positive voltage (reverse bias) corresponds to elec-
trons flowing from the ferromagnet into the semiconductor.
zero bias a priori. This remains a frustrating point in the materials physics
of these systems, particularly because the interfaces appear to be remarkably
well-controlled when evaluated by spin-LED or scanning transmission elec-
tron microscopy (STEM) measurements [67, 98].
While sample-specific effects may be regarded as problematic, it is
important to emphasize the universal observation that for Fe/GaAs (100),
Pj always reverses sign over a narrow window ±0.1 V around zero bias [97].
Similar observations have been made by other groups [99]. For epitaxial Fe/
GaAs (100), the minority spin current always dominates at large forward
bias. Similar anomalous behavior (i.e. a crossover to minority-dominated
tunneling) has been observed in Fe/GaAs/Fe tunnel junctions in which one
interface is grown epitaxially [57]. In the case of metallic magnetic tunnel
junctions, a reversal in the current polarization (and hence the sign of the
tunneling magnetoresistance) has been observed in systems with crystal-
line barriers, as in Fe/MgO/Fe [100]. The association with crystalline barriers
suggests that the effects of dispersion are important. On the other hand, the
sign reversal is not necessarily observed for other ferromagnets grown epi-
taxially on GaAs. For example, Kotissek et al. found minority spin accumu-
lation in Fe0.32Co0.68/GaAs structures under forward bias [66], and we find
minority accumulation under forward bias in heterostructures (otherwise
identical to those under discussion here) in which Fe is replaced by epitaxial
Fe3Ga. Experiments on Heusler alloys also show non-trivial injector bias
dependence, depending on the degree of structural order [101].
Addressing the bias dependence requires moving beyond a simple
Jullière-type model, in which the tunneling density of states and matrix
elements are independent of energy. An extension of the Jullière model,
accounting for the bulk Fe density of states, does produce a sign reversal,
but only at much higher bias. The simplest approach to modifying the
tunneling matrix elements, discussed by Smith and Ruden [102], accounts
for the bias dependence of the Schottky barrier profile, while modeling
the ferromagnet as an s-band metal with distinct Fermi wave-vectors
for majority and minority spins. Other approaches emphasize different
aspects of the electronic structure of the Fe/GaAs (001) Schottky bar-
rier [103, 104, 105]. Chantis and coworkers [103] focus on the tunneling
density of states at the Fe/GaAs interface. By calculating the surface den-
sity of states and then integrating the tunneling current for both majority
and minority bands in k-space, they find an enhancement of the minor-
ity current at small forward bias, leading to a crossover from majority
to minority transmission. Dery and Sham [104] focus on the conduction
band near the surface of the semiconductor, arguing that the minority
current is enhanced by tunneling from bound states formed in the shallow
potential well just inside the Schottky barrier. It is impossible to distin-
guish between these models based only on existing data, although Li and
Dery [106] have proposed inserting a barrier (of higher bandgap material)
just inside the semiconductor. This would allow for an explicit test of the
Dery–Sham model.
6.6.2 Hyperfine Effects
A second set of anomalies in the non-local measurements are associated
with the magnetic field dependence and are illustrated in Figure 6.15. First,
in the longitudinal field sweep at low temperatures (Figure 6.15a), a peak is
observed at zero field in addition to the two expected jumps when the mag-
netizations of the source and detector are in the antiparallel state. (This peak
can also be observed in the fields sweeps of Figure 6.12.) The zero-field peak
depends very strongly on the field sweep rate and vanishes if the measure-
ment time is sufficiently long (requiring wait times over a minute at each
data point), or if the temperature is sufficiently high (typically above 60 K).
Various checks can be used to ascertain that the magnetizations of the con-
tacts are indeed fixed as the field is swept through zero, and so we conclude
that it is due to some change in the spin dynamics in the semiconductor.
Similar observations of a “zero-field peak” have been made by Salis et al. [87]
and Ciorga et al. [91].
The second anomaly, illustrated in Figure 6.15b is the extreme sensi-
tivity of the Hanle curves to the magnetic field orientation, an effect that
becomes so strong at low temperatures that it is difficult to obtain a mean-
ingful Hanle curve in an ordinary field sweep. As shown in Figure 6.15b,
there is also a clear difference between the two different parallel configura-
tions of the contacts. Instead of being identical, the Hanle curves for V↑↑ and
V↓↓ are both distorted, and they appear to be mirror images with respect to
inversion of the magnetic field axis. Unlike the zero-field peak in the longi-
tudinal field sweeps, this effect persists in steady-state, and the distortion
as well as the apparent shift with respect to zero field increase with the
degree of field misalignment as well as with the bias current. Distortion of
non-local Hanle curves has also been reported by the NRL and Regensburg
groups [88, 91].
FIGURE 6.15 Effects of hyperfine interactions. (a) Data from a longitudinal field sweep (field along easy axis
of contacts) at 10 K. Note the peak at zero field in addition to the switching features. (b) Hanle data at 50 K
obtained with the magnetic field oriented a few degrees away from vertical. Solid lines are a fit to the drift
diffusion model with a constant hyperfine field with a sign determined by the magnetization of the injector.
The samples and bias conditions are different for the two panels, and so the magnitudes of the signals should
not be compared.
These effects, which are ubiquitous in Fe/GaAs devices, are due to the
dynamic polarization of nuclei by electron spins localized on donor sites.
This phenomenon was originally observed in optical pumping experi-
ments [11] and has been reviewed extensively by Fleisher and Merkulov [10].
Discussions of similar effects in GaAs-based spin-LEDs appear in Refs. [37]
and [51]. From the standpoint of electron spin transport and dynamics, the
main consequence of the dynamic nuclear polarization (DNP) is an effective
mean field B N which acts on the electron spins. A proper calculation of B N
requires a self-consistent treatment of the electron and nuclear spin dynam-
ics. In a relatively general form:
(B + beS) ⋅ S(B + beS)

B N = bn , (6.11)
(B + beS)2 + ξBl2
where:
bn and be , which are both negative in GaAs [11], represent effective fields
due to the polarized nuclei and electrons
S is the average electron spin (| S |= 1 / 2 for electron spin polariza-
tion PGaAs = 1)
Bl is the local dipolar field experienced by the nuclei
ξ parameterizes the assisting processes which allow energy to be
conserved in mutual spin flips between electrons and nuclei. (We
have simplified the notation of Paget [11] by incorporating prefac-
tors into bn and be that cannot be determined independently).
The electron spin precesses in a total field Btot = B + B N , which can be

much larger than the applied field B.
If we ignore the exchange field (Knight field) be acting on the nuclei, B N
is parallel to the applied field, but its magnitude and sign are determined by
the electron spin accumulation. This is the origin of the distortion of the
Hanle curves in Figure 6.15b as well as their dependence on the sign of the
injected spins. A more extreme case is shown in Figure 6.16a, obtained for a
misalignment of 15° in the direction of the applied field with respect to the
sample normal. The satellite peak in the Hanle curve at ~300 G is due to
the approximate cancelation of the applied field by the z-component of the
hyperfine field. Given the dot product in Equation 6.11, the magnitude of the
hyperfine field in this case is approximately 1000 G.
The sensitivity of the electron spin accumulation to B N implied by
Figure 6.16 can be regarded either positively or negatively depending on
one’s point of view, and it is important to note that it can be suppressed
to a large extent by inverting the magnetization of the injector on a times-
cale that is fast relative to the timescale for the nuclei to become polarized,
which is of the order of one second. This is feasible if small electromagnets
are used. On the other hand, an optimist might be inclined to use the elec-
tron spin accumulation as a nuclear spin detector. For example, depolar-
izing a particular isotope (either 69Ga, 71Ga, or 75As) by resonant excitation
suppresses its contribution to B N , resulting in a corresponding shift of Btot .
As a result, the Hanle curve will shift along the field axis, and the value of
FIGURE 6.16 (a) Non-local Hanle curve obtained with the field applied at 15° from the vertical direction. An
expansion of the region around zero field is shown in the inset. The solid curve is obtained from the modi-
fied drift diffusion model with a nuclear field of the form in Equation 6.11. Note the double peak in both the
model and the experimental data in the inset, which is due to the non-zero Knight field be . (b) NMR spectrum
obtained in an applied field of 320 G for the same conditions in (a) as the frequency of the current in a coil
over the sample is swept. The resonances for NMR transitions of 75As, 69Ga, and 71Ga are shown using circles,
closed squares, and open squares, respectively.
the spin accumulation at a fixed field will change. This means of detecting
nuclear magnetic resonance is illustrated in Figure 6.16b, which shows the
Hanle signal measured at an applied field of 320 G as a function of the fre-
quency of current passing through a small coil over the sample. The primary
resonances for the three main isotopes are indicated by symbols.
Other manifestations of the hyperfine interaction are less amenable to
simple interpretation. The zero-field feature in the longitudinal field sweeps
has been studied carefully by Salis et al. [87], who were able to model it in
terms of the formalism presented here by carrying out longitudinal field
sweeps in a fixed perpendicular magnetic field. Its ubiquitous presence in
longitudinal field sweeps is due to the rotation of B N as the applied field is
swept through zero.
It is evident that nuclear spins have a profound influence on electron
spin dynamics. The influence of the electronic field be , also known as the
Knight field, is more subtle, since it acts directly only on the nuclei. An effect
on the electron spin accumulation occurs only because the exchange field
beS changes the magnitude and direction of the effective field in which the
nuclei precess. If, however, an experiment is designed to be sensitive to the
Knight field, it then allows for a measurement of the magnitude and sign of
the electron spin polarization without requiring knowledge of the density of
states in the semiconductor. This comes about because the field beS, which is
proportional to the electron spin polarization, modifies the magnetic field
dependence of the spin accumulation signal, whereas S alone only impacts
the magnitude.
Chan et al. [107] have solved the problem of the coupled dynamics of
the electron and nuclear spin system by incorporating the effective field of
Equation 6.11 into the drift-diffusion model of Equation 6.1, solving the two
equations self-consistently in one dimension. A model curve based on this
result is shown in Figure 6.16a. The region around zero field is shown in
the inset. The small double-peak near zero is due to the effective field beS.
A more sensitive measurement of beS described in detail in Ref. [107], has

allowed for a direct quantitative measurement of the spin accumulation S
as a function of the injector bias. The resulting spin polarization PGaAs is
shown in Figure 6.17 superimposed on the ordinary measurement of the
spin-dependent non-local voltage ∆V NL = V↑↑ − V↑↓ . These two very different
measurements show the same dependence on the bias voltage of the injector.
As can be seen on the right-hand axis, the Knight field measurement gives
an electron spin polarization of about 6% at the maximum injector bias. The
sign corresponds to accumulation of majority spins, as expected based on
the Kerr measurements of other structures at large forward bias. Given this
polarization, the magnitude of V NL is in fact quite close to the very crude
estimate of Equation 6.10, although the agreement is probably serendipi-
tous. In subsequent work, the actual Knight shift in the electrically detected
nuclear magnetic resonance frequency was also observed [108], and (within
factors of order unity), the magnitudes of the shifts agreed with expectations
based on the average Knight field beS, as well as the magnitude of the spin
accumulation as inferred from the non-local voltage. It is also possible to
use the coupled electron-nuclear system to measure the nuclear spin relax-
ation time T1 . Kölbl et al. used spin valves to probe the electron-nuclear spin
coupling down to temperature T ~ 0.1 K, observing a departure from the
usual Korringa law (T1−1 ∝ T ) at the lowest temperatures [109]. Uemura et al.
have even developed a pulse sequence that allows for the observation of Rabi
oscillations of the nuclear polarization in a Co2MnSi/GaAs spin valve [110].
The success of these measurements reflects the extreme sensitivity of the
polarized electron system to small changes in the hyperfine field. In effect,
the FM/GaAs spin valves allow for the study of nuclear spin dynamics over
a very small volume. On these scales, inhomogeneities of the hyperfine field
also impact the electron spin dynamics [109], including the electron spin
relaxation rate [111, 112].
FIGURE 6.17 The non-local voltage ∆VNL = V↑↑ − V↑↓ from spin-valve studies
(open squares) compared with the spin polarization PGaAs determined from the
measurement of be (Knight field) as described in the text. Note the two indepen-
dent y-scales.
6.6.3 Detector Bias Dependence

Given the difficulties described above in Section 6.1 in achieving a quan-
titative interpretation of the injector bias dependence, it might appear
premature to consider applying an additional bias voltage to the detector.
A biased detector would seem to remove some of the purported advantages
of the non-local approach, particularly from the standpoint of traditional
objections to two-terminal measurements [18]. In the context of the above
measurements, however, we now have an excellent grasp of the behavior of
spin-dependent signals (and artifacts) in Fe/GaAs devices. Furthermore, in
the presence of the semiconductor channel, it is easy to bias the FM source
and detector electrodes individually. The actual experimental phase space
is therefore much larger than considered in some of the more pessimistic
assessments of two-terminal devices with high-resistivity barriers [25, 113].
In fact [71], the application of separate injector and detector bias voltages
allows one to exploit the advantages of semiconductor-based spin transport
devices in ways that are not possible in similar metallic devices. The poten-
tial of multiterminal FM/SC devices has only begun to be explored, and the
reader is referred to Chapter 17, Volume 3 for a discussion of some of the
possible applications.
Given the complications addressed in Section 6.1, we separate into two
contributions the effects of an additional detector bias voltage Vd on the
sensitivity of electrical spin detection [71]. The first contribution, discussed
above, is simply the effect of Vd on the polarization Pj of the tunneling cur-
rent that flows across the Fe/GaAs detector interface. Tuning the magnitude
and/or sign of Pj with Vd will tune the magnitude and/or sign of the spin-
detection sensitivity. The second contribution, which we exploit to further
enhance the spin-dependent sensitivity of electrical spin detectors, origi-
nates from the electric fields in the semiconductor channel near the detector
[6, 17] that arise because the detector is biased, including the enhancement
of the spin transport by drift [61, 114], as well as the modification of the
(spin-dependent) carrier densities near the detector interface.
We consider a situation where a bias voltage Vd is applied between the
FM detector contact and the SC channel. As a measure of the detector’s
spin-dependent sensitivity, we then consider the voltage change ∆Vd at the
detector that is induced by a remotely-injected spin polarization (analogous
to conventional non-local studies). If there is no electric field in the semicon-
ductor (e.g. if the current is purely diffusive), simple reciprocity consider-
ations require that for a given unit of spin-polarized current, the change ∆Vd
must be proportional to the change ∆µ in the spin accumulation that occurs
when the same unit of spin-polarized current is injected from an identical
contact at injection bias Vint . The sensitivity of the detector at a voltage V is
therefore proportional to Pj = ∆V NL (V ) / I (V ). As noted in Section 6.1, this is
proportional to the polarization Pj of the current flowing through the injec-
tor. From a conventional non-local experiment (in which V is the injector
interface voltage), it is therefore possible to predict the electrical detection
sensitivity of the contact in the absence of electric field effects.
In Ref. [71], we have addressed the reasoning of the previous paragraph

by using optical pumping to generate a small spin-polarized current that
is incident on a detector contact biased at a voltage Vd . The change ∆Vd in
the detector voltage was measured using a standard modulation technique
and compared with ∆V NL (V ) / I (V ) for the same device. For detector bias
voltages as small as 0.1 V, there are significant departures from the expec-
tations of the simple reciprocity argument. In particular, the sensitivity of
the detector at forward bias was observed to be dramatically enhanced. A
simple model [115] showed that the enhancement is due to the modification
of spin-dependent carrier densities n↑ ,↓ (and their gradients) by the electric
field at the interface.
The implication of this result is that the sensitivity of a non-local spin-
valve can be tuned over a wide range by application of a bias voltage at
the detector. Figure 6.18 shows that this is indeed the case. Figure 6.18a
shows the magnitude of the spin-valve signal versus detector bias. We con-
sider one of the sets of data (the black points) obtained at a fixed injector
bias current (−1.5 mA). At zero detector bias—where non-local studies are
usually performed—the spin-valve signal is quite small (<2 µV). However,
when the non-local spin detector is forward-biased by −150 mV, the mag-
nitude of the spin-valve signal increases by over an order of magnitude to
approximately 20 µV. Furthermore, with the detection electrode under a
small reverse bias, the spin valve signal actually inverts sign. Note that in
all three cases the injected spin density remains unchanged. Thus, elec-
trons in GaAs with a fixed spin polarization can be made to induce either
positive or negative voltage changes at a detection electrode, simply by
tuning the detector bias Vd by a few tens of meV. The detector sensitiv-
ity in this voltage range is completely independent of the injector bias,
FIGURE 6.18 (a) Inset: All-electrical lateral spin-valve setup with separately-biased injector and detector
electrodes. Black (gray) points show the spin-valve signal ∆Vd versus detector bias Vd for a forward (reverse)
biased spin injector. (b) Raw spin-valve data at three detector biases [arrows in (a)], using a fixed −1.5 mA
injector bias (curves offset). Note sign switching and ten-fold enhancement of the detected signal.
6.7 Recent Developments 305
as demonstrated by the second set of data (gray points) in Figure 6.18a,

which were obtained with a totally different injection current of +1.5 mA.
After correcting for the different spin injection efficiencies, the two sets of
data collapse onto a single curve. Figure 6.18b shows spin-valve data for
these three scenarios (Vd = −150, 0, and +82 mV), explicitly showing that
spin-detection sensitivities are freely tunable in both sign and magnitude
in all-electrical devices. This approach to enhancing the sensitivity of the
detector has also been used to enable operation of FM/n-GaAs non-local
spin valves at room temperature [116]. It is possible to enhance the detec-
tion sensitivity by over a factor of 10, while maintaining linearity of the
output with respect to the injector bias current.
6.7 RECENT DEVELOPMENTS
6.7.1 Spin-Orbit Effects
Another outstanding class of problems involves the detection of spin Hall
effects (see Chapter 8, Volume 2) [117, 118, 119], an area which has seen
activity over the last several years (since the previous edition of this chap-
ter) on many different fronts [120]. The first obvious candidate is the elec-
trical detection of the spin Hall effect in a manner analogous to the Hanle
experiment of Kato et al. [118], who used Kerr microscopy to detect the spin
accumulation at the edges of a Hall bar biased with an unpolarized cur-
rent. In principle, this only requires placing ferromagnetic Hall contacts on
a semiconductor channel. Given the small values of the spin accumulation
(~0.1%), the technical requirements are more stringent than for the experi-
ments introduced above. This experiment has been carried out [121], and
it has allowed for a much more detailed comparison with theory than has
been possible with optical probes. In particular, it was possible to measure
the spin Hall conductivity (inferred from the spin accumulation and the
drift-diffusion model) as a function of the ordinary charge conductivity. The
skew and side jump scattering contributions to the spin Hall conductivity
were extracted and compared with predictions for the case of ionized impu-
rity scattering [122]. A similar experiment was subsequently carried out in
n-GaAs based Esaki diodes using Ga1−xMnxAs as the spin detector [123]. It
has also been possible to detect the transverse charge current generated in a
FM/GaAs device by a spin current that is generated non-locally (the inverse
spin Hall effect (ISHE)) [124]. The latter experiment is analogous to the origi-
nal detection of the ISHE in a metallic non-local spin valve [125].
The classic Datta–Das spin-FET [15], in which the role of the magnetic
field in a Hanle measurement is assumed by the effective spin-orbit mag-
netic field, requires a set of experiments, parallel to those discussed above,
on a 2D electron system. This has been a remarkably difficult problem to
address, in large part because the types of tunneling contacts developed for
efficient spin injection into bulk GaAs have not been easily adaptable to two-
dimensional electron gas (2DEG) heterostructures. Koo et al. did report on
gate modulation of the spin valve signal in an InAs 2DEG [92], although
no Hanle effect was observed. They later reported on a gate modulation of

the ISHE in the same device [126]. A very clear demonstration of a 2D non-
local spin valve was executed for the case of (Al,Ga)As/GaAs by Oltscher
and coworkers [127, 128], who adapted their Esaki-diode approach in order
to inject spin-polarized electrons into a 2DEG. These devices, however, were
not gated, although, as in the bulk case [129], it has been possible to verify
the spin accumulation using Kerr microscopy [130]. It remains to be seen
whether all of the components of a spin-FET can be demonstrated in a FM/
III–V device, although numerous alternatives are now being explored.
The presence of spin-orbit coupling, even in bulk III–V semiconduc-
tors, has been a key component of a series of experiments which have
attempted to exploit spin pumping from a ferromagnet, in lieu of spin
injection using a charge current. The essential idea of spin pumping is to
drive the FM out of equilibrium using ferromagnetic resonance (FMR), and
the system relaxes towards equilibrium through the flow of a spin current
into an adjacent material [131]. This concept was demonstrated in metallic
multilayers [132] and then non-local spin valves [133], but the measure-
ment is much easier when the generated spin current can be detected using
the ISHE [134]. In principle, the spin-pumping approach is complementary
to spin injection by a charge current, in that it only requires an interface
with a sufficiently high mixing conductance [131], and would hence appear
to be a means to circumvent the conductivity mismatch problem [23].
Ando et al. tested this proposition by carrying out spin pumping measure-
ments with ISHE detection on very simple devices consisting of permalloy
(Ni0.81Fe 0.19 sputtered on either p- or n-GaAs [135]. Remarkably, both elec-
tron and hole-doped systems showed ISHE signals on resonance, and these
reversed sign with the reversal of the magnetization of the FM. Similar
spin pumping signals have been observed in FM/Si [136, 137] and FM/Ge
[138, 139] devices, and more recently in FM/GaN [140], for which the bulk
spin-orbit coupling is weak. Shikoh et al. demonstrated that a spin current
generated by spin pumping could be detected non-locally in Si [137], and
Yamamoto et al. have carried out a similar experiment in GaAs [141]. In
both of these experiments, the non-local detection was carried out using
an ordinary heavy metal and the ISHE. Both ordinary magnetotransport
(through the tunneling anistoropic magnetoresistance [72, 73]) and spin
accumulation in FM/SC hybrids with tunnel barriers are sensitive to FMR
[142, 143]. Given other difficulties in extracting spin transport parame-
ters in traditional spin pumping devices [144], a direct comparison of the
spin-pumping and non-local spin valve techniques would be a major step
forward.
6.7.2 Heusler Alloy-Based Heterostructures

Since the early days of semiconductor spintronics, it has been realized that
a large class of ternary alloys in the Heusler family are ferromagnets and
lattice-matched to GaAs [145]. Moreover, many of these are predicted to be
half-metallic, although most of these predictions have been made for bulk
Acknowledgments 307
materials and not interfaces. Felser et al. [146] and Palmstrøm [145] have
recently reviewed Heusler alloys in the context of spintronics. Although this
chapter will not provide a review of the materials physics of the Heuslers
(see, for example, Ref. [147]), it is important to note that their incorpora-
tion into III–V based devices has led to significant improvements, includ-
ing operation at room temperature. Spin injection from Heusler alloys
into GaAs was observed in spin-LED measurements [40, 148–150], and in
recent years they have received considerably more attention in the context
of non-local spin valves. Bruski and coworkers fabricated Co2FeSi/n-GaAs
spin valves demonstrating both the local and non-local transport signatures
introduced above [151]. Although there has not been an explicit demonstra-
tion of half-metallicity of a Heusler/GaAs interface, there have now been
several demonstrations of Heusler/III–V based lateral spin valves operating
at room temperature [116, 152], and spin injection from Co2MnSi has been
shown to be a particularly efficient source of dynamic nuclear polarization
[110]. We anticipate additional improvements as the materials physics of this
important class of ferromagnets, including their integration with semicon-
ductors, is explored further.
6.8 CONCLUSIONS
Although this chapter has focused primarily on the early days of FM/
III–V semiconductor heterostructures, devices based on this materials
system have now advanced to the point that they can function as quan-
titative tools for studying spin transport. In particular, it is possible to
probe spin transport and dynamics using combinations of optical and
transport techniques that allow for detailed tests of theory. Although out-
standing problems remain, particularly in the area of integration of FMs
with quantum well heterostructures with strong spin-orbit coupling, the
experiments described in this chapter have provided a foundation for this
ongoing effort.
ACKNOWLEDGMENTS
It is a pleasure to thank our collaborators who have made this work so
enjoyable: Christoph Adelmann, Mun Chan, Athanasios Chantis, Kevin
Christie, Madalina Furis, Chad Geppert, Eric Garlid, Qi Hu, Tsuyoshi
Kondo, Changjiang Liu, Xiaohua Lou, Chris Palmstrøm, Sahil Patel, Tim
Peterson, Madhukar Reddy, Darryl Smith, Gordon Stecklein, and Jianjie
Zhang. The work discussed in this review has been supported by NSF under
DMR-0804244 and DMR-1104951, the Office of Naval Research, the MRSEC
Program of the National Science Foundation under Award Numbers DMR-
0212302 and DMR-0819885, the Los Alamos LDRD program, the National
High Magnetic Field Laboratory (which is supported by NSF DMR-1157490
and the State of Florida), the National Science Foundation NNIN and NNCI
programs, and by C-SPIN, one of the six centers of STARnet, a Semiconductor
Research Corporation program sponsored by MARCO and DARPA.
REFERENCES
1. F. Meier and B. P. Zakharchenya (Eds.), Optical Orientation, North Holland,
Amsterdam, 1984.
2. J. M. Kikkawa and D. D. Awschalom, Resonant spin amplification in n-type
GaAs, Phys. Rev. Lett. 80, 4313–4316 (1998).
3. R. I. Dzhioev, B. P. Zakharchenya, V. L. Korenev, and M. N. Stepanova, Spin
diffusion of optically oriented electrons and photon entrainment in n-gallium
arsenide, Phys. Solid State 39, 1765–1768 (1997).
5. A. G. Aronov and G. E. Pikus, Spin injection into semiconductors, Sov. Phys.
Semicond. - USSR 10, 698–700 (1976).
6. Z. G. Yu and M. E. Flatté, Spin diffusion and injection in semiconductor struc-
tures: Electric field effects, Phys. Rev. B 66, 235302 (2002).
7. M. I. D’yakonov and V. I. Perel’, Spin relaxation of conduction electrons in
noncentrosymmetric semiconductors, Sov. Phys. Solid State 13, 3023–3026
(1972).
8. M. Furis, D. L. Smith, S. A. Crooker, and J. L. Reno, Bias-dependent electron
spin lifetimes in n-GaAs and the role of donor impact ionization, Appl. Phys.
Lett. 89, 102102 (2006).
9. M. Furis, D. L. Smith, S. Kos et al., Local Hanle-effect studies of spin drift and
diffusion in n:GaAs epilayers and spin-transport devices, New J. Phys. 9, 347
(2007).
10. V. G. Fleisher and I. A. Merkulov, Optical orientation of the coupled electron-
nuclear spin system of a semiconductor, in Optical Orientation, F. Meier and
B. P. Zakharchenya (Eds.), North Holland, Amsterdam, pp. 173–258, 1984.
11. D. Paget, G. Lampel, B. Sapoval, and V. I. Safarov, Low field electron-nuclear
spin coupling in gallium arsenide under optical pumping conditions, Phys.
Rev. B 15, 5780–5796 (1977).
Systems, Springer-Verlag, Berlin, 2003.
13. G. E. Pikus and A. N. Titkov, Spin relaxation under optical orientation in semi-
conductors, in Optical Orientation, F. Meier and B. P. Zakharchenya (Eds.),
North Holland, Amsterdam, pp. 73–132, 1984.
14. Y. A. Bychkov and E. I. Rashba, Oscillatory effects and the magnetic suscepti-
bility of carriers in inversion layers, J. Phys. C: Solid State Phys. 17, 6039–6045
(1984).
15. S. Datta and B. Das, Electronic analog of the electro-optic modulator, Appl.
Phys. Lett. 56, 665–667 (1990).
16. M. I. D’yakonov and V. Y. Kachorovskii, Spin relaxation of two-dimensional
electrons in noncentrosymmetric semiconductors, Sov. Phys. Semicond. -
USSR 20, 110–112 (1986).
17. I. Žutić, J. Fabian and S. Das Sarma, Spin injection through the depletion layer:
A theory of spin-polarized p-n junctions and solar cell, Phys. Rev. B 64, 121201
(2001).
18. H. X. Tang, F. G. Monzon, F. J. Jedema, A. T. Filip, B. J. Van Wees, and M.
L. Roukes, Spin injection and transport in micro- and nanoscale devices, in
Semiconductor Spintronics and Quantum Computation, D. D. Awschalom, D.
Loss, and N. Samarth (Eds.), Springer-Verlag, Berlin, pp. 31–87, 2002.
19. M. Johnson and R. Silsbee, Interfacial charge-spin coupling: Injection and
detection of spin magnetization in metals, Phys. Rev. Lett. 55, 1790–1793
(1985).
20. M. Johnson and R. H. Silsbee, Spin-injection experiment, Phys. Rev. B 37,
5326–5335 (1988).
References 309
21. P. C. van Son, H. van Kempen, and P. Wyder, Boundary resistance of the fer-
romagnetic-nonferromagnetic metal interface, Phys. Rev. Lett. 58, 2271–2273
(1987).
22. T. Valet and A. Fert, Theory of the perpendicular magnetoresistance in mag-
netic multilayers, Phys. Rev. B 48, 7099–7113 (1993).
23. G. Schmidt, D. Ferrand, L. W. Molenkamp, A. T. Filip, and B. J. van Wees,
Fundamental obstacle for electrical spin injection from a ferromagnetic metal
into a diffusive semiconductor, Phys. Rev. B 62, 4790–4793 (2000).
24. E. I. Rashba, Theory of electrical spin injection: Tunnel contacts as a solu-
tion of the conductivity mismatch problem, Phys. Rev. B 62(24), 16267–16270
(2000).
26. D. L. Smith and R. N. Silver, Electrical spin injection into semiconductors,
Phys. Rev. B 64, 045323 (2001).
27. S. F. Alvarado and P. Renaud, Observation of spin-polarized-electron tun-
neling from a ferromagnet into GaAs, Phys. Rev. Lett. 68, 1387–1390 (1992).
28. H. J. Zhu, M. Ramsteiner, H. Kostial, M. Wassermeier, H. P. Schönherr, and K.
H. Ploog, Room temperature spin injection from Fe into GaAs, Phys. Rev. Lett.
87, 016601 (2001).
29. A. T. Hanbicki, B. T. Jonker, G. Itskos, G. Kioseoglou, and A. Petrou, Efficient
electrical injection from a magnetic metal/tunnel barrier contact into a semi-
conductor, Appl. Phys. Lett. 80, 1240–1242 (2002).
30. J. Strand, B. D. Schultz, A. F. Isakovic, C. J. Palmstrøm, and P. A. Crowell,
Dynamic nuclear polarization by electrical spin injection in ferromagnet-
semiconductor heterostructures, Phys. Rev. Lett. 91, 036602 (2003).
31. V. F. Motsnyi, J. De Boeck, J. Das et al., Electrical spin injection in a ferro-
magnet/tunnel barrier/semiconductor heterostructure, Appl. Phys. Lett. 81,
265–267 (2002).
injection into GaAs from a magnetic tunnel transistor source, Phys. Rev. Lett.
90, 256603 (2003).
33. X. Jiang, R. Wang, R. M. Shelby et al., Highly spin-polarized room-temper-
ature tunnel injector for semiconductor spintronics using MgO(100), Phys.
Rev. Lett. 94, 056601 (2005).
34. R. Fiederling, M. Kelm, G. Reuscher et al., Injection and detection of a spin-
polarized current in a light-emitting diode, Nature 402, 787–789 (1999).
erostructure, Nature 402, 790–792 (1999).
36. D. R. Scifres, B. A. Huberman, R. M. White, and R. S. Bauer, A new scheme for
measuring itinerant spin polarizations, Solid State Commun. 13, 1615–1617
(1973).
37. J. Strand, X. Lou, C. Adelmann et al., Electron spin dynamics and hyperfine
interactions in Fe/Al0.1Ga0.9As/GaAs spin injection heterostructures, Phys.
Rev. B 72, 155308 (2005).
38. C. Adelmann, X. Lou, J. Strand, C. J. Palmstrøm, and P. A. Crowell, Spin injec-
tion and relaxation in ferromagnet-semiconductor heterostructures, Phys. Rev.
B 71, 121301 (2005).
39. G. Salis, R. Wang, X. Jiang et al., Temperature independence of the spin-
injection efficiency of a MgO-based tunnel spin injector, Appl. Phys. Lett. 87,
262503 (2005).
40. X. Y. Dong, C. Adelmann, J. Q. Xie et al., Spin injection from the Heusler
alloy Co2MnGe into Al0.1Ga0.9As/GaAs heterostructures, Appl. Phys. Lett. 86,
102107 (2005).
41. M. C. Hickey, C. D. Damsgaard, I. Farrer, D. A. Ritchie, R. F. Lee, G. A. C. Jones,

M. Pepper, Spin injection between epitaxial Co2.4Mn1.6Ga and an InGaAs
quantum well, Appl. Phys. Lett. 86, 252106 (2005).
42. N. C. Gerhardt, S. Hovel, C. Brenner et al., Electron spin injection into GaAs
from ferromagnetic contacts in remanence, Appl. Phys. Lett. 87, 032502 (2005).
43. C. Adelmann, J. L. Hilton, B. D. Schultz et al., Spin injection from perpendicu-
lar magnetized ferromagnetic δ-MnGa into (Al,Ga)As heterostructures, Appl.
Phys. Lett. 89, 112511 (2006).
44. R. K. Kawakami, Y. Kato, M. Hanson et al., Ferromagnetic imprinting of
nuclear spins in semiconductors, Science 294, 131–134 (2001).
45. R. J. Epstein, I. Malajovich, R. K. Kawakami et al., Spontaneous spin coherence
in n-GaAs produced by ferromagnetic proximity polarization, Phys. Rev. B 65,
121202(R) (2002).
46. C. Ciuti, J. P. McGuire, and L. J. Sham, Spin polarization of semiconductor car-
riers by reflection off a ferromagnet, Phys. Rev. Lett. 89, 156601 (2002).
47. J. P. McGuire, C. Ciuti, and L. J. Sham, Theory of spin transport induced by
ferromagnetic proximity on a two-dimensional electron gas, Phys. Rev. B 69,
115339 (2004).
48. R. J. Epstein, J. Stephens, M. Hanson et al., Voltage control of nuclear spin in
ferromagnetic Schottky diodes, Phys. Rev. B 68, 41305 (2003).
49. J. Stephens, J. Berezovsky, J. P. McGuire, L. J. Sham, A. C. Gossard, and D.
D. Awschalom, Spin accumulation in forward-biased MnAs/GaAs Schottky
diodes, Phys. Rev. Lett. 93, 097602 (2004).
51. P. Van Dorpe, W. Van Roy, J. De Boeck, and G. Borghs, Nuclear spin orienta-
tion by electrical spin injection in an AlxGa1-x As/GaAs spin-polarized light-
emitting diode, Phys. Rev. B 72, 035315 (2005).
52. J. Strand, A. F. Isakovic, X. Lou, P. A. Crowell, B. D. Schultz, and C. J. Palmstrøm,
Nuclear magnetic resonance in a ferromagnet-semiconductor heterostruc-
ture, Appl. Phys. Lett. 83, 3335–3337 (2003).
53. M. W. J. Prins, H. van Kempen, H. van Leuken, R. A. de Groot, W. Van Roy, and
J. De Boeck, Spin-dependent transport in metal/semiconductor tunnel junc-
tions, J. Phys. Condens. Matter 7, 9447–9464 (1995).
54. A. Hirohata, Y. B. Xu, C. M. Guertler, J. A. C. Bland, and S. N. Holmes, Spin-
polarized electron transport in ferromagnet/semiconductor hybrid structures
induced by photon excitation, Phys. Rev. B 63, 104425 (2001).
55. A. F. Isakovic, D. M. Carr, J. Strand, B. D. Schultz, C. J. Palmstrøm, and P. A.
Crowell, Optical pumping in ferromagnet-semiconductor heterostructures:
Magneto-optics and spin transport, Phys. Rev. B 64, 161304 (2001).
56. S. Kreuzer, J. Moser, W. Wegscheider, D. Weiss, M. Bichler, and D. Schuh, Spin
polarized tunneling through single-crystal GaAs(001) barriers, Appl. Phys.
Lett. 80, 4582–4584 (2002).
57. J. Moser, M. Zenger, C. Gerl, W. Wegscheider, and D. Weiss, Bias dependent
inversion of tunneling magnetoresistance in Fe/GaAs/Fe tunnel junctions,
Appl. Phys. Lett. 89, 162106 (2006).
magnet/semiconductor structures, Science 309, 2191–2195 (2005).
59. S. A. Crooker, M. Furis, X. Lou et al., Optical and electrical spin injection and
spin transport in hybrid Fe/GaAs devices, J. Appl. Phys. 101, 081716 (2007).
60. X. Lou, C. Adelmann, M. Furis, S. A. Crooker, C. J. Palmstrøm, and P. A.
Crowell, Electrical detection of spin accumulation at a ferromagnet-semicon-
ductor interface, Phys. Rev. Lett. 96, 176603 (2006).
61. X. Lou, C. Adelmann, S. A. Crooker et al., Electrical detection of spin transport
in lateral ferromagnet-semiconductor devices, Nat. Phys. 3, 197–202 (2007).
62. P. R. Hammar and M. Johnson, Detection of spin-polarized electrons injected
into a two-dimensional electron gas, Phys. Rev. Lett. 88, 066806 (2002).
References 311
63. J. M. Kikkawa and D. D. Awschalom, Lateral drag of spin coherence in GaAs,

Nature 397, 139–141 (1999).
64. S. A. Crooker and D. L. Smith, Imaging spin flows in semiconductors subject to
electric, magnetic, and strain fields, Phys. Rev. Lett. 94, 236601 (2005).
65. M. Beck, C. Metzner, S. Malzer, and G. H. Döhler, Spin lifetimes and strain-
controlled spin precession of drifting electrons in GaAs, Europhys. Lett. 75,
597–603 (2006).
66. P. Kotissek, M. Bailleul, M. Sperl et al., Cross-sectional imaging of spin injec-
tion into a semiconductor, Nat. Phys. 3, 872–877 (2007).
67. J. M. LeBeau, Q. O. Hu, C. J. Palmstrøm, and S. Stemmer, Atomic structure of
postgrowth annealed epitaxial Fe/(001)GaAs interfaces, Appl. Phys. Lett. 93,
121909 (2008).
68. M. Hruška, Š Kos, S. A. Crooker, A. Saxena, and D. L. Smith, Effects of strain,
electric, and magnetic fields on lateral electron-spin transport in semiconduc-
tor epilayers, Phys. Rev. B 73, 075306 (2006).
69. F. J. Jedema, H. R. Heersche, A. T. Filip, J. J. A. Baselmans, and B. J. van Wees,
Electrical detection of spin procession in a metallic mesoscopic spin valve,
Nature 416, 713–716, (2002).
70. Y. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom, Coherent spin
manipulation without magnetic fields in strained semiconductors, Nature 427,
50–53 (2004).
71. S. A. Crooker, E. S. Garlid, A. N. Chantis et al., Bias-controlled sensitivity of
ferromagnet/semiconductor electrical spin detectors, Phys. Rev. B 80, 041305
(2009).
72. C. Gould, C. Rüster, T. Jungwirth et al., Tunneling anisotropic magnetore-
sistance: A spin-valve-like tunnel magnetoresistance using a single magnetic
layer, Phys. Rev. Lett. 93, 117203 (2004).
73. J. Moser, A. Matos-Abiague, D. Schuh, W. Wegscheider, J. Fabian, and D. Weiss,
Tunneling anisotropic magnetoresistance and spin-orbit coupling in Fe/GaAs/
Au tunnel junctions, Phys. Rev. Lett. 99, 056601 (2007).
74. M. Tran, H. Jaffrès, C. Deranlot et al., Enhancement of the spin accumulation
at the interface between a spin-polarized tunnel junction and a semiconduc-
tor, Phys. Rev. Lett. 102, 036601 (2009).
75. S. P. Dash, S. Sharma, R. S. Patel, M. P. de Jong, and R. Jansen, Electrical cre-
ation of spin polarization in silicon at room temperature, Nature 462, 491–494
(2009).
76. Y. Song and H. Dery, Magnetic-field-modulated resonant tunneling in fer-
romagnetic-insulator-nonmagnetic junctions, Phys. Rev. Lett. 113, 047205
(2014).
77. A. G. Swartz, S. Harashima, Y. Xie et al., Spin-dependent transport across Co/
LaAlO3/SrTiO3 heterojunctions, Appl. Phys. Lett. 105, 032406 (2014).
78. I. Appelbaum, H. N. Tinkey, and P. Li, Self-consistent model of spin accumula-
tion magnetoresistance in ferromagnet/insulator/semiconductor tunnel junc-
tions, Phys. Rev. B 90, 220402 (2014).
79. H. Inoue, A. G. Swartz, N. J. Harmon et al., Origin of the magnetoresistance in
oxide tunnel junctions determined through electric polarization control of the
interface, Phys. Rev. X 5, 041023 (2015).
80. O. Txoperena and F. Casanova, Spin injection and local magnetoresistance
effects in three-terminal devices, J. Phys. D: Appl. Phys. 49, 133001 (2016).
81. S. P. Dash, S. Sharma, J. C. Le Breton et al., Spin precession and inverted Hanle
effect in a semiconductor near a finite-roughness ferromagnetic interface,
Phys. Rev. B 84, 054410 (2011).
82. K.-R. Jeon, B.-C. Min, Y.-H. Jo et al., Electrical spin injection and accumula-
tion in CoFe/MgO/Ge contacts at room temperature, Phys. Rev. B 84, 165315
(2011).
83. S. Iba, H. Saito, A. Spiesser et al., Spin accumulation in nondegenerate and
heavily doped p-type germanium, Appl. Phys. Express 5, 023003 (2012).
84. N. Reyren, M. Bibes, E. Lesne et al., Gate-controlled spin injection at LaAlO3/

SrTiO3 interfaces, Phys. Rev. Lett. 108, 186802 (2012).
85. W. Han, X. Jiang, A. Kajdos, S.-H. Yang, S. Stemmer, and S. S. P. Parkin, Spin
injection and detection in lanthanum- and niobium-doped SrTiO3 using the
Hanle technique, Nat. Commun. 4, 2134 (2013).
86. R. J. Soulen Jr., J. M. Byers, M. S. Osofsky, C. T. Tanaka, J. Nowak, J. S. Moodera,
A. Barry, J. M. D. Coey, Measuring the spin polarization of a metal with a
superconducting point contact, Science 282, 85–88 (1998).
87. G. Salis, A. Fuhrer, and S. F. Alvarado, Signatures of dynamically polarized
nuclear spins in all-electrical lateral spin transport devices, Phys. Rev. B 80,
115332 (2009).
88. C. Awo-Affouda, O. M. J. van’t Erve, G. Kioseoglou et al., Contributions to
Hanle lineshapes in Fe/GaAs nonlocal spin valve transport, Appl. Phys. Lett.
94, 102511 (2009).
89. D. Saha, M. Holub, P. Bhattacharya, and Y. C. Liao, Epitaxially grown MnAs/
GaAs lateral spin valves, Appl. Phys. Lett. 89, 142504 (2006).
90. O. M. J. van’t Erve, A. T. Hanbicki, M. Holub et al., Electrical injection and
detection of spin-polarized carriers in silicon in a lateral transport geometry,
Appl. Phys. Lett. 91, 212109 (2007).
91. M. Ciorga, A. Einwanger, U. Wurstbauer, D. Schuh, W. Wegscheider, and D.
Weiss, Electrical spin injection and detection in lateral all-semiconductor
devices, Phys. Rev. B 79, 165321 (2009).
92. H. C. Koo, J. H. Kwon, J. Eom, J. Chang, S. H. Han, and M. Johnson, Control of
spin precession in a spin-injected field effect transistor, Science 325, 1515–1518
(2009).
temperature, Nature 448, 571–574 (2007).
94. W. Han, K. Pi, W. Bao et al., Electrical detection of spin precession in single
layer graphene spin valves with transparent contacts, Appl. Phys. Lett. 94,
222109 (2009).
95. I. Appelbaum, B. Q. Huang, and D. J. Monsma, Electronic measurement and
control of spin transport in silicon, Nature 447, 295–298 (2007).
97. Q. O. Hu, E. S. Garlid, P. A. Crowell, and C. J. Palmstrøm, Spin accumulation
near Fe/GaAs (001) interfaces: The role of semiconductor band structure, Phys.
Rev. B 84, 085306 (2011).
density-functional theory, Phys. Rev. Lett. 96, 196101 (2006).
99. G. Salis, S. F. Alvarado, and A. Fuhrer, Spin-injection spectra of CoFe/GaAs
contacts: Dependence on Fe concentration, interface, and annealing condi-
tions, Phys. Rev. B 84, 041307 (2011).
100. C. Tiusan, J. Faure-Vincent, C. Bellouard, M. Hehn, E. Jouguelet, and A. Schuhl,
Interfacial resonance state probed by spin-polarized tunneling in epitaxial Fe/
MgO/Fe tunnel junctions, Phys. Rev. Lett. 93, 106602 (2004).
101. P. Bruski, S. C. Erwin, J. Herfort, A. Tahraoui, and M. Ramsteiner, Probing the
electronic band structure of ferromagnets with spin injection and extraction,
Phys. Rev. B 90, 245150 (2014).
102. D. L. Smith and P. P. Ruden, Spin-polarized tunneling through potential barri-
ers at ferromagnetic metal/semiconductor Schottky contacts, Phys. Rev. B 78,
125202 (2008).
103. A. N. Chantis, K. D. Belashchenko, D. L. Smith, E. Y. Tsymbal, M. van
Schilfgaarde, and R. C. Albers, Reversal of spin polarization in Fe/GaAs (001)
driven by resonant surface states: First-principles calculations, Phys. Rev. Lett.
99, 196603 (2007).
References 313
104. H. Dery and L. J. Sham, Spin extraction theory and its relevance to spintronics,
Phys. Rev. Lett. 98, 046602 (2007).
105. S. Honda, H. Itoh, J. Inoue et al., Spin polarization control through resonant
states in an Fe/GaAs Schottky barrier, Phys. Rev. B 78, 245316 (2008).
106. P. Li and H. Dery, Tunable spin junction, Appl. Phys. Lett. 94, 192108 (2009).
107. M. K. Chan, Q. O. Hu, J. Zhang, T. Kondo, C. J. Palmstrøm, and P. A. Crowell,
Hyperfine interactions and spin transport in ferromagnet-semiconductor het-
erostructures, Phys. Rev. B 80, 161206 (2009).
108. K. D. Christie, C. C. Geppert, S. J. Patel, Q. O. Hu, C. J. Palmstrøm, and P. A.
Crowell, Knight shift and nuclear spin relaxation in Fe/n-GaAs heterostruc-
tures, Phys. Rev. B 92, 155204 (2015).
109. D. Kölbl, D. M. Zumbühl, A. Fuhrer, G. Salis, and S. F. Alvarado, Breakdown of
the Korringa law of nuclear spin relaxation in metallic GaAs, Phys. Rev. Lett.
109, 086601 (2012).
110. T. Uemura, T. Akiho, Y. Ebina, and M. Yamamoto, Coherent manipulation of
nuclear spins using spin injection from a half-metallic spin source, Phys. Rev. B
91, 140410 (2015).
111. N. J. Harmon, T. A. Peterson, C. C. Geppert et al., Anisotropic spin relax-
ation in n-GaAs from strong inhomogeneous hyperfine fields produced by the
dynamical polarization of nuclei, Phys. Rev. B 92, 140201 (2015).
112. Y.-S. Ou, Y.-H. Chiu, N. J. Harmon et al., Exchange-driven spin relaxation
in ferromagnet-oxide-semiconductor heterostructures, Phys. Rev. Lett. 116,
107201 (2016).
113. R. Jansen and B. C. Min, Detection of a spin accumulation in nondegenerate
semiconductors, Phys. Rev. Lett. 99, 246604 (2007).
114. C. Jozsa, M. Popinciuc, N. Tombros, H. T. Jonkman, and B. J. Van Wees,
Electronic spin drift in graphene field-effect transistors, Phys. Rev. Lett. 100,
236603 (2008).
115. A. N. Chantis and D. L. Smith, Theory of electrical spin-detection at a ferro-
magnet/semiconductor interface, Phys. Rev. B 78, 235317 (2008).
116. T. A. Peterson, S. J. Patel, C. C. Geppert et al., Spin injection and detection
up to room temperature in Heusler alloy/n-GaAs spin valves, Phys. Rev. B 94,
235309 (2016).
117. M. I. D’yakonov and V. I. Perel’, Possibility of orienting electron spins with cur-
rent, JETP Lett. 13, 467–469 (1971).
the spin Hall effect in semiconductors, Science 306, 1910–1913 (2004).
tion of the spin-Hall effect in a two-dimensional spin-orbit coupled semicon-
Hall effects, Rev. Mod. Phys. 87, 1213–1260 (2015).
121. E. S. Garlid, Q. O. Hu, M. K. Chan, C. J. Palmstrøm, and P. A. Crowell, Electrical
measurement of the direct spin Hall effect in Fe/InxGa1-x As heterostructures,
Phys. Rev. Lett. 105, 156602 (2010).
122. H.-A. Engel, B. I. Halperin, and E. I. Rashba, Theory of spin Hall conductivity
in n-doped GaAs, Phys. Rev. Lett. 95, 166605 (2005).
123. M. Ehlert, C. Song, M. Ciorga et al., All-electrical measurements of direct spin
Hall effect in GaAs with Esaki diode electrodes, Phys. Rev. B 86, 205204 (2012).
124. K. Olejnk, J. Wunderlich, A. C. Irvine et al., Detection of electrically modu-
lated inverse spin Hall effect in an Fe/GaAs microdevice, Phys. Rev. Lett. 109,
076601 (2012).
125. S. O. Valenzuela and M. Tinkham, Direct electronic measurement of the spin
Hall effect, Nature 442, 176–179 (2006).
126. W. Y. Choi, H.-J. Kim, J. Chang, S. H. Han, H. C. Koo, and M. Johnson,
Electrical detection of coherent spin precession using the ballistic intrinsic
spin Hall effect, Nat. Nanotechnol. 10, 666–670 (2015).
127. M. Oltscher, M. Ciorga, M. Utz, D. Schuh, D. Bougeard, and D. Weiss,

Electrical spin injection into high mobility 2D systems, Phys. Rev. Lett. 113,
236602 (2014).
128. M. Ciorga, Electrical spin injection and detection in high mobility 2DEG sys-
tems, J. Phys. Condens. Matter 28, 453003 (2016).
129. B. Endres, M. Ciorga, R. Wagner et al., Nonuniform current and spin accumu-
lation in a 1 µm thick n-GaAs channel, Appl. Phys. Lett. 100, 092405 (2012).
130. M. Buchner, T. Kuczmik, M. Oltscher, C. Schüller, D. Bougeard, D. Weiss, and
C. H. Back, Optical investigation of electrical spin injection into an inverted
two-dimensional electron gas structure, Phys. Rev. B 95, 035304 (2017).
131. Y. Tserkovnyak, A. Brataas, and G. E. W. Bauer, Enhanced Gilbert damping in
thin ferromagnetic films, Phys. Rev. Lett. 88, 117601 (2002).
132. R. Urban, G. Woltersdorf, and B. Heinrich, Gilbert damping in single and mul-
tilayer ultrathin films: Role of interfaces in nonlocal spin dynamics, Phys. Rev.
Lett. 87, 217204 (2001).
133. M. V. Costache, M. Sladkov, S. M. Watts, C. H. van der Wal, and B. J. van Wees,
Electrical detection of spin pumping due to the precessing magnetization of a
single ferromagnet, Phys. Rev. Lett. 97, 216603 (2006).
134. E. Saitoh, M. Ueda, H. Miyajima, and G. Tatara, Conversion of spin current
into charge current at room temperature: Inverse spin-Hall effect, Appl. Phys.
Lett. 88, 182509 (2006).
135. K. Ando, S. Takahashi, J. Ieda et al., Electrically tunable spin injector free from
the impedance mismatch problem, Nat. Mater. 10, 655–659 (2011).
136. K. Ando and E. Saitoh, Observation of the inverse spin Hall effect in silicon,
Nat. Commun. 3, 629 (2012).
137. E. Shikoh, K. Ando, K. Kubo, E. Saitoh, T. Shinjo, and M. Shiraishi, Spin-pump-
induced spin transport in p-type Si at room temperature, Phys. Rev. Lett. 110,
127201 (2013).
138. M. Koike, E. Shikoh, Y. Ando et al., Dynamical spin injection into p-type ger-
manium at room temperature, Appl. Phys. Express 6, 023001 (2013).
139. J. C. Rojas-Sanchez, M. Cubukcu, A. Jain, A. Marty, L. Vila, J. P. Attane, E.
Augendre, G. Desfonds, S. Gambarelli, H. Jaffrès, J. M. George, and M. Jamet,
Spin pumping and inverse spin Hall effect in germanium, Phys. Rev. B 88,
064403 (2013).
140. R. Adhikari, M. Matzer, A. T. Martin-Luengo, M. C. Scharber, and A. Bonanni,
Rashba semiconductor as spin Hall material: Experimental demonstration of
spin pumping in wurtzite n-GaN:Si, Phys. Rev. B 94, 085205 (2016).
141. A. Yamamoto, Y. Ando, T. Shinjo, T. Uemura, and M. Shiraishi, Spin transport
and spin conversion in compound semiconductor with non-negligible spin-
orbit interaction, Phys. Rev. B 91, 024417 (2015).
142. C. Liu, Y. Boyko, C. C. Geppert et al., Electrical detection of ferromagnetic
resonance in ferromagnet/n-GaAs heterostructures by tunneling anisotropic
magnetoresistance, Appl. Phys. Lett. 105, 212401 (2014).
143. C. Liu, S. J. Patel, T. A. Peterson et al., Dynamic detection of electron spin
accumulation in ferromagnet/semiconductor devices by ferromagnetic reso-
nance, Nat. Commun. 7, 10296 (2016).
144. L. Chen, F. Matsukura, and H. Ohno, Direct-current voltages in (Ga,Mn)As
structures induced by ferromagnetic resonance, Nat. Commun. 4, 2055 (2013).
145. C. Palmstrøm, Heusler compounds and spintronics, Prog. Cryst. Growth
Charact. 62, 371–397 (2016).
146. C. Felser, L. Wollmann, S. Chadov, G. H. Fecher, and S. S. P. Parkin, Basics and
prospective of magnetic Heusler compounds, APL Mater. 3, 041518 (2015).
147. C. Felser and A. Hirohata (Eds.), Heusler Alloys–Properties, Growth,
Applications, Springer, Heidelberg, 2016.
148. M. C. Hickey, C. D. Damsgaard, S. N. Holmes et al., Spin injection from
Co2MnGa into an InGaAs quantum well, Appl. Phys. Lett. 92, 232101 (2008).
References 315
149. M. Ramsteiner, O. Brandt, T. Flissikowski et al., Co2FeSi/GaAs/(Al,Ga)As spin

light-emitting diodes: Competition between spin injection and ultrafast spin
alignment, Phys. Rev. B 78, 121303 (2008).
150. R. Farshchi and M. Ramsteiner, Spin injection from Heusler alloys into semi-
conductors: A materials perspective, J. Appl. Phys. 113, 191101 (2013).
151. P. Bruski, Y. Manzke, R. Farshchi, O. Brandt, J. Herfort, and M. Ramsteiner,
All-electrical spin injection and detection in the Co2FeSi/GaAs hybrid system
in the local and non-local configuration, Appl. Phys. Lett. 103, 052406 (2013).
152. T. Saito, N. Tezuka, M. Matsuura, and S. Sugimoto, Spin injection, trans-
port, and detection at room temperature in a lateral spin transport device
with Co2FeAl0.5Si0.5/n-GaAs Schottky tunnel junctions, Appl. Phys. Express 6,
103006 (2013).
7
Spin Polarization
by Current

7.2 Model 320
7.3 Anisotropy of the Inverse Spin-Galvanic Effect 326
7.4 Concluding Remarks 331
Acknowledgments 331
References 332
7.1 INTRODUCTION
Spin generation and spin currents in semiconductor structures lie at the
heart of the emerging field of spintronics and are a major and still growing
direction of solid-state research. Among the plethora of concepts and ideas,
current-induced spin polarization has attracted particular interest from both
experimental and theoretical points of view; for reviews, see Refs. [1–14]. In
nonmagnetic semiconductors or metals belonging to the gyrotropic point
317
318 Chapter 7. Spin Polarization by Current
groups,* see Refs. [7, 16–20], dc electric current is generically accompanied

by a non-zero average nonequilibrium spatially homogeneous spin polariza-
tion and vice versa. The latter phenomenon is referred to as the spin-galvanic
effect observed in GaAs QWs [21] and other two-dimensional systems; see
e.g. reviews [5, 7, 10, 11, 13, 14, 22–24, 58]. In low-dimensional semiconductor
structures these effects are caused by asymmetric spin relaxation in systems
with lifted spin degeneracy due to k-linear terms in the Hamiltonian, where
k is the electron wave-vector. In spite of the terminological resemblance,
spin polarization by electric current fundamentally differs from the spin
Hall effect [4, 6, 8, 11, 12, 13, 25–31], which refers to the generation of a pure
spin current transverse to the charge current, and causes spin accumula-
tion at the sample edges. The distinctive features of the current-induced spin
polarization are: that this effect can be present in gyrotropic media only, that
it results in non-zero average spin polarization, and that it does not depend
on the real-space coordinates. Thus, it can be measured in the whole sample
under appropriate conditions. The spin Hall effect, in contrast, does not yield
average spin polarization, and does not require gyrotropy, at least for the
extrinsic spin Hall effect. Related discussion on spin Hall effect can be found
in Chapter 8, Volume 2, and Chapter 7, Volume 3.
The ability of charge current to polarize spins in gyrotropic media was
predicted more than thirty years ago by Ivchenko and Pikus [32]. The effect
has been considered theoretically for bulk tellurium crystals, where, almost
at the same time, it was demonstrated experimentally by Vorob’ev et al. [33].
In bulk tellurium, current-induced spin polarization is a consequence of the
unique valence band structure of tellurium with hybridized spin-up and spin-
down bands (”camel back” structure) and, in contrast to the spin polarization
in quantum well structures, is not related to spin relaxation. In zinc-blende
structure based QWs, this microscopic mechanism of the current-induced
spin polarization is absent [34, 35]. Vas’ko and Prima [35], Aronov and Lyanda-
Geller [36], and Edelstein [37] demonstrated that spin orientation by electric
current is also possible in two-dimensional electron systems and is caused
by asymmetric spin relaxation. Two microscopic mechanisms, namely scat-
tering mechanism and precessional mechanism, based on Elliott–Yafet and
D’yakonov–Perel spin relaxation, respectively, were developed. The first direct
experimental proofs of this effect were obtained in semiconductor QWs by
Ganichev et al. in (113)-grown p-GaAs/AlGaAs QWs [38, 39], by Silov et al.
* We remind readers that the gyrotropic point group symmetry makes no difference between
certain components of polar vectors, like electric current or electron momentum, and axial
vectors, like a spin or magnetic field, and is described by the gyration tensor [7, 15, 16].
Gyrotropic media are characterized by the linear in light or electron wave-vector k spatial
dispersion resulting in optical activity (gyrotropy) or Rashba/Dresselhaus band spin-splitting
in semiconductor structures [7, 16–20], respectively. Among 21 crystal classes lacking inver-
sion symmetry, 18 are gyrotropic, from which 11 classes are enantiomorphic (chiral) and do
not possess a reflection plane or rotation-reflection axis [7, 18, 19]. Three non-gyrotropic non-
centrosymmetric classes are Td, C3h and D3h. We note that it is often, but misleadingly, stated
that gyrotropy (optical activity) can be obtained only in non-centrosymmetric crystals having
no mirror reflection plane. In fact seven non-enantiomorphic classes groups (CS, C2v, C3v, S4,
D2d, C4v and C6v) are gyrotropic, also allowing spin orientation by the electric current.
in (001)-grown p-GaAs/AlGaAs heterojunctions [40, 41], by Sih et al. in (110)-

grown n-GaAs/AlGaAs QWs [42], and by Yang et al. in (001)-grown InGaAs/
InAlAs QWs [43], as well as in strained bulk (001)-oriented InGaAs and ZnSe
epilayers by Kato et al. [44, 45], and Stern et al. [46], respectively. The experi-
ments include the range of optical methods, such as Faraday rotation and lin-
ear-circular dichroism in transmission of terahertz radiation, time-resolved
Kerr rotation, and polarized luminescence in near-infrared up to visible spec-
tral range. We emphasize that current-induced spin polarization was observed
even at room temperature [38, 39, 45]. It has been demonstrated that depend-
ing on the point group symmetry electric current in 2D system may result in
the in-plane spin orientation, as in the case of e.g. (001)-grown structures [40,
41, 43–50], or may additionally align spins normal to the 2DEG’s plane. The
latter take place in [llh]- or [lmh]-oriented structures, e.g. [113]-, [110]- and
[013]-grown QWs (see [10, 38, 39, 42, 51]). We also would like to note, that
investigating spin injection from a ferromagnetic film into a two-dimensional
electron gas, Hammar et al. [52, 53] used the concept of a spin orientation by
current in a 2DEG (see also [54, 55]) to interpret their results. Though a larger
degree of spin polarization was extracted, the experiment’s interpretation
is complicated by other effects [56, 57]. Later experiments on ferromagnet/
(Ga,Mn)As bilayers [58], uniaxial (Ga,Mn)As epilayers [59], metallic interfaces
and surfaces [60–65] and interfaces between ferromagnetic films and topo-
logical insulators [66, 67] clearly demonstrated the ability of spin polarization
by electric current in ferromagnetics- and metal-based structures. Chapters 4
and 9, Volume 2 provide some background material on (Ga,Mn)As.
The experimental observation of current-induced spin polarization has
given rise to extended theoretical studies of this phenomenon in various
systems using various approaches and theoretical techniques. These include
the semiclassical Boltzmann Equation [51, 68–73] derived from the quan-
tum mechanical Liouville Equation [74] and other diffusion-type equations
describing the dynamics of spin expectation values [6, 75–83]. Other authors
have performed important and fruitful studies of the same issues using vari-
ous Green’s function techniques of many-body physics [2, 84–91]. The effect
of external contacts to the system and boundaries was studied explicitly in
Refs. [55, 92, 93], and in Refs. [94, 95] the spin response of an electron gas to a
microwave radiation was calculated. The influence of four terminal geometry
has been studied in Ref. [96] using a numerical Landauer-Keldysh approach,
and weak localization corrections for current-induced spin polarization have
been calculated in [97]. The current-induced spin polarization has also been
investigated in hole systems [89, 98, 99]. A search for efficient spin generation
and manipulation by all electrical means has given rise to theoretical analysis
of current induced spin polarization in exotic regimes, like streaming caused
by high electric fields [73, 100] or very weak electron-impurity interaction
[101], in one-dimensional channels [102, 103] and combined structures
with metal/insulator [104], ferromagnets/topological insulators [105–
108], ferromagnets/graphene [109, 110], or metal/semiconductor [62, 111]
interfaces. Magnetic heterostructures with topological insulators and gra-
phene are discussed in Chapter 5, Volume 2, and Chapter 5, Volume 3.
Searching for publications on current-induced spin polarization, one

can be confused by the fact that several different names are used to describe
it. Besides current-induced spin polarization (CISP) this phenomenon is
often referred to as the inverse spin-galvanic effect (ISGE), current-induced
spin accumulation (CISA), the magnetoelectric effect (MEE), or kinetic mag-
netoelectric effect (KMEE) (this term was first used to describe the effect in
nonmagnetic conductors by Levitov et al. [112]), or electric-field mediated
in-plane spin accumulation. This variety of terms was extended further after
the observation of the current-induced spin polarization at the interface
between nonmagnetic metals [60]. The authors introduce the new term—
Edelstein effect (EE)—which in following works was modified to Rashba–
Edelstein effect (REE), as well as to its inversion (IEE) corresponding to the
spin-galvanic effect. Despite the enormous diversity of labels, in all these
cases we deal with one and the same microscopic effect: appearance of non-
equilibrium spin polarization due to dc electric current in the gyrotropic
media with Rashba/Dresselhaus spin splitting of the bands. In our chapter
we will use two of these terms: current-induced spin polarization and the
inverse spin-galvanic effect.
7.2 MODEL
Phenomenologically, the electron’s averaged nonequilibrium spin S can be
linked to an electric current j by
jλ = ∑Q
µ
S ,
λµ µ (7.1)
Sα = ∑R
γ
j ,
αγ γ (7.2)
where Q and R are second rank pseudotensors. Equation 7.1 describes the
spin-galvanic effect and Equation 7.2 represents the effect inverse to the spin-
galvanic effect: an electron spin polarization induced by a dc electric cur-
rent. We note the similarity of Equations 7.1 and 7.2 characteristic for effects
due to gyrotropy: both equations linearly couple a polar vector with an axial
vector. The phenomenological Equation 7.2 shows that the spin polarization
can only occur for those components of the in-plane components of j which
transform as the averaged nonequilibrium spin S for all symmetry opera-
tions. Thus the relative orientation of the current direction and the average
spin is completely determined by the point group symmetry of the structure.
This can most clearly be illustrated by the example of a symmetric (110)-
grown zinc-blende QW where an electric current along x [110] results in
a spin orientation along the growth direction z: see Figure 7.1a. These QWs
belong to the point-group symmetry C2ν and contain, apart from the identity
and a C2-axis, a reflection plane m1 normal to the QW plane and x-axis, and
a reflection plane m2 being parallel to the interface plane, see Figure 7.1b.
The reflection in m1 transforms the current component jx and the average
7.2 Model 321
FIGURE 7.1 (a) Electric current-induced spin polarization in symmetric (110)-

grown zinc-blende structure based QWs. (b) symmetry elements of symmetrical
QW grown in the z || [110] direction. Arrows in the sketch (b) show reflection of the
components of polar vector j x and axial vector Sz by the mirror reflection plane
m1 . An additional reflection by the mirror reflection plane m2 does not modify the
components of an in-plane polar vector j x as well as not changing the polarity of
an out-plane axial vector Sz . Thus, the linear coupling of j x and Sz is allowed for
structures of this symmetry.
spin component Sz in the same way ( jx → − jx , S y → − S y ): see Figure 7.1b.

Also the reflection in m2 transforms these components in an equal way,
and their sign remains unchanged for this symmetry operation. Therefore,
a linear coupling of the in-plane current and the out-plane average spin is
allowed, demonstrating that a photocurrent jx can induce the average spin
polarization Sz . In asymmetric (110)-grown QWs or (113)-grown QWs, the
symmetry is reduced to Cs and additionally to the z-direction, spins can be
oriented in the plane of QWs.* Similar arguments demonstrate that in (001)-
grown zinc-blende structure based QWs, an electric current can result in an
in-plane spin orientation only. In this case, the direction of spins depends
on the relative strengths of the structure inversion asymmetry (SIA) [113]
and bulk inversion asymmetry (BIA) [114], resulting in an anisotropy of the
current-induced spin polarization.
A microscopic model of the current-induced spin polarization [38, 39] is
sketched in Figure 7.2a. To be specific, we consider an electron gas in sym-
metric (110)-grown zinc-blende QWs. The explanation of the effect mea-
sured in structures of other crystallographic orientation or in a hole gas can
be given in a similar way. In the simplest case, the electron’s (or hole’s) kinetic
energy in a quantum well depends quadratically on the in-plane wave-vector
components k x and k y . In equilibrium, the spin degenerated k x and k y states
are symmetrically occupied up to the Fermi energy E F . If an external elec-
tric field is applied, the charge carriers drift in the direction of the resulting
force. The carriers are accelerated by the electric field and gain kinetic energy
until they are scattered. A stationary state forms where the energy gain and
* Note that for the lowest symmetry C1, which contains no symmetry elements besides iden-
tity, the relative direction between current and spin orientation becomes arbitrary.
FIGURE 7.2 Current-induced spin polarization (a) and spin-galvanic effect (b) due to spin-flip scattering
in symmetric (110)-grown zinc-blende structure based QWs. In this case, only β zx σ z k x term are present in
the Hamiltonian. The conduction subband is split into two parabolas with spin-up | +1/ 2〉 z and spin-down
| −1/ 2〉 z pointing in the z-direction. In (a) biasing along the x-direction causes an asymmetric in k-space occu-
pation of both parabolas. In (b) nonequilibrium spin orientation along z-direction causes an electric current
in x-direction. (After Ganichev, S.D. et al., cond-mat/0403641, 2004; Ganichev, S.D. et al., J. Magn. and Magn.
Mater. 300, 127, 2006. With permission.)
the relaxation are balanced, resulting in a non-symmetric distribution of

carriers in k-space yielding an electric current. The electrons acquire the
average quasimomentum
∆k = eEτ ps (7.3)
where E is the electric field strength, τ ps is the momentum relaxation time

for a given spin split subband labeled by s, and e is the elementary charge. As
long as spin-up and spin-down states are degenerated in k-space the energy
bands remain equally populated and a current is not accompanied by spin
orientation. In QWs made of zinc-blende structure material like GaAs or
strained bulk semiconductors, however, the spin degeneracy is lifted due to
SIA and BIA [113, 114] and dispersion reads
2k 2
Eks = + βlmσ l km (7.4)
2m*
with the spin-orbit pseudotensor β, the Pauli spin matrices σ l and effective
mass m* (note the similarity to Equations 7.1 and 7.2). The parabolic energy
band splits into two subbands of opposite spin directions shifted in k-space
symmetrically around k = 0 with minima at ±k0 . For symmetric (110)-grown
zinc-blende structure based QWs, the spin-orbit interaction results in a
Hamiltonian of the form β zx σ z k x ; the corresponding dispersion is sketched
in Figure 7.2. Here the energy band splits into two subbands of sz = 1 / 2 and
sz = −1 / 2, with minima symmetrically shifted in the k-space along the k x axis
from the point k = 0 into the points ±k0 , where k0 = m*β z ′x / 2 . As long as
the spin relaxation is switched off, the spin branches are equally populated
and contribute equally to the current. Due to the band splitting, spin-flip
7.2 Model 323
relaxation processes ±1 / 2 → ∓1 / 2 , however, become different because of

the difference in quasimomentum transfer from initial to final states. In
Figure 7.2a the k-dependent spin-flip scattering processes are indicated
by arrows of different lengths and thicknesses. As a consequence, differ-
ent amounts of spin-up and spin-down carriers contribute to the spin-flip
transitions, causing a stationary spin orientation. In this picture, we assume
that the origin of the current-induced spin orientation is, as sketched in
Figure 7.2a, exclusively due to scattering and hence dominated by the
Elliott–Yafet spin relaxation (scattering mechanism) [115]. The other pos-
sible mechanism resulting in the current-induced spin orientation is based
on the D’yakonov–Perel spin relaxation [115] (precessional mechanism). In
this case, the relaxation rate depends on the average ∆k -vector equal to
k1/2 = − k0 + 〈 ∆k x 〉 for the spin-up and k −1/2 = k0 + 〈 ∆k x 〉 for the spin-down
subband [114]. Hence, for the D’yakonov–Perel mechanism also, the spin
relaxation becomes asymmetric in k-space, and a current through the elec-
tron gas causes spin orientation.
Figure 7.2b shows that not only phenomenological equations, but also
microscopic models of the current-induced spin polarization and the spin-
galvanic effect are similar. Spin orientation in the x-direction causes the
unbalanced population in spin-down and spin-up subbands. As long as the
carrier distribution in each subband is symmetric around the subband mini-
mum at k x± , no current flows. The current flow is caused by k-dependent
spin-flip relaxation processes [7, 21, 22]. Spins oriented in the z-direction are
scattered along k x from the higher filled—e.g. spin subband | +1 / 2〉 z —to the
less filled—e.g. spin subband | −1 / 2〉 z . The spin-flip scattering rate depends
on the values of the wave-vectors of the initial and the final states [115]. Four
quantitatively different spin-flip scattering events exist. They preserve the
symmetric distribution of carriers in the subbands, and thus do not yield
a current. While two of them have the same rates, the other two, sketched
in Figure 7.2b by bent arrows, are inequivalent, and generate an asym-
metric carrier distribution around the subband minima in both subbands.
This asymmetric population results in a current flow along the x-direction.
Within this model of elastic scattering, the current is not spin polarized,
since the same number of spin-up and spin-down electrons move in the
same direction with the same velocity. Like current-induced spin polariza-
tion, spin-galvanic effect can also result from the precessional mechanism
[7] based on the asymmetry of the Dyakonov–Perel spin relaxation.
In order to foster the above phenomenological arguments and models,
let us discuss a microscopic description of such processes as introduced in
Ref. [71]. This theory is based on an analytical solution to the semiclassical
Boltzmann equation, which does not include the spin relaxation time explic-
itly. Instead, one utilizes the so-called quasiparticle life time τ 0 , which is
defined via the scattering probability alone. This is in contrast to the momen-
tum relaxation time τ ps , which, in addition, depends on the distribution
function itself and, therefore, should be self-consistently deduced from the
Boltzmann equation written within the relaxation time approximation. In
the framework of a simplest model dealing with the elastic delta-correlated
scatterers, the quasiparticle life time τ 0 depends neither on carrier momen-

tum nor spin index, and in that way, it essentially simplifies our description.
Before we proceed, let us first discuss the applicability of the quasiclassi-
cal Boltzmann kinetic equation to the description of the spin polarization in
quantum wells observed in [40, 41, 42, 43, 44, 45, 46]. There are two restric-
tions which are inherited by the Boltzmann equation due to its quasiclassical
origin. The first one is obvious: the particle’s de Broglie length must be much
smaller than the mean free path. At low temperatures—compared to the
Fermi energy—the characteristic de Broglie length λ relates to the carrier
m* E F
concentration ne  approximately as λ  2π / ne , whereas the mean
π2

free path l can be estimated as l  * 2πne τ 0 . Thus, the first restriction can
m
be written as
ne  m* / τ 0 . (7.5)
The second one is less obvious and somewhat more specific to our systems
here, but still it is important: the smearing of the spin-split subband due
to the disorder / τ 0 must be much smaller than the spin splitting energy
E+ k − E− k . The latter depends strongly on the quasimomentum, and there-
fore at the Fermi level, it is defined by the carrier concentration. As a conse-
quence, the concentration must fulfill the following inequality
ne  2 / 8πβ2 τ 20 , (7.6)
where β is the spin-orbit coupling parameter. This restriction can also be

reformulated in terms of the mean free path l and spin precession length
λ s  π / k0 . Namely, l must be much larger than λ s so that an electron ran-
domizes its spin orientation due to the spin-orbit precession between two
subsequent scattering events. This restriction corresponds to the approxi-
mation which neglects the off-diagonal elements of the nonequilibrium dis-
tribution function in the spin space. Indeed, an electron spin cannot only
be in one of two possible spin eigen states of the free Hamiltonian, but in
an arbitrary superposition of them, the case described by the off-diagonal
elements of the density matrix. The latter is not possible in classical physics
where a given particle always has a definite position in the phase space. Thus,
we could not directly apply Boltzmann equation in its conventional form for
the description of the electron spin in 2DEGs with spin-orbit interactions as
long as the inequality (7.6) is not fulfilled. Note, on the other hand, that all
the quantum effects stemming from the quantum nature of the electron spin
can be smeared out by a sufficient temperature larger than the spin splitting
energy E+ k − E− k . Thus, the room temperature Troom = 25 meV being much
larger than the spin-orbit splitting energy of the order of 3meV (which is
relevant for InAs quantum wells) makes the Boltzmann equation applicable
for sure. In order to verify whether the conditions given by Equations 7.5
and 7.6 are indeed satisfied, one can deduce the quasiparticle life time τ 0
from the mobility µ = eτ 0 / m* . Then, using spin-orbit coupling parameters in
7.2 Model 325
the range usually found in experiments [116, 118, 119], the above inequalities
turn out to be fulfilled. In a case when inequality (7.6) is not fulfilled, the
theory based on spin-density matrix formalism yields the result for CISP
degree, depending on the ratio between energy and spin relaxation times
[51, 120]. If energy relaxation is slower than Dyakonov–Perel spin relaxation,
then the electrically induced spin density is given by Equation 7.11 derived
for well-split spin density subbands. This regime takes place at low tempera-
tures. By contrast, spin density is twice as large in the opposite case of fast
energy relaxation which is realized at moderate and high temperatures [51].
In general, the Boltzmann equation describes the time evolution of the
particle distribution function f (t , r , k) , in the coordinate r and momentum k
space. To describe the electron kinetics in presence of a small homogeneous
electric field E in a steady state, one usually follows the standard procedure
widely spread in the literature on solid state physics. The distribution func-
tion is then represented as a sum of an equilibrium f 0 ( Esk ) and nonequilib-
rium f1( s, k) contributions. The first one is just a Fermi-Dirac distribution,
and the second one is a time and coordinate independent nonequilibrium
correction linear in E. This latter contribution should be written down as
a solution of the kinetic equation; however, it might be also deduced from
qualitative arguments in what follows.
Let us assume that the scattering of carriers is elastic, which means,
above all, that spin flip is forbidden. Then, the average momentum ∆k which
the carriers gain due to the electric field can be estimated relying on the
momentum relaxation time approximation from Equation 7.3. If the electric
field is small (linear response), then to get the nonequilibrium term f1( s, k)
one has to expand the Fermi–Dirac function f 0 ( Es( k − ∆k ) ) into the power
series for small ∆k up to the term linear in E. Recalling v = −∂ ∆k E( sk − ∆k ) |∆k =0 ,
the nonequilibrium contribution f1( s, k) can be written as
 ∂f 0 ( Esk ) 
f1( s, k) = −eEvτ ps  − . (7.7)
 ∂Esk 
Since f1( s, k) is proportional to the derivative of the step-like Fermi–Dirac
distribution function, the nonequilibrium term substantially contributes to
the total distribution function close to the Fermi energy only, and the sign
of its contribution depends on the sign of the group velocity v. Note that
the group velocity v for a given energy and direction of motion is the same
for both spin split subbands and, therefore, nonequilibrium addition to the
distribution function would be the same for both branches as long as τ ps
were independent of the spin index. The latter is, however, not true, and τ ps
is different for two spin split subbands, as depicted in Figure 7.2. This is the
microscopic reason why the nonequilibrium correction to the distribution
function gives rise to the spin polarization.
The nonequilibrium correction can be also found as an analytical solu-
tion of the Boltzmann Equation [71], and has the form
τ 0  ∂f 0 ( Esk ) 
g 1( s, k) = −eEk − . (7.8)
m*  ∂Esk 
FIGURE 7.3 This plot shows schematically the nonequilibrium term of the distribution function vs. quasi-
momentum for (a) spin-degenerate and (b) spin-split subbands. The electric field is directed along the x-axis.
(a) Without spin splitting the nonequilibrium distribution function just provides a majority for right mov-
ing electrons at the expense of the ones with opposite momentum. (b) The spin-orbit splitting leads to the
additional electron redistribution between two spin split subbands. As one can see from the plot, an amount
of electrons belong to the outer spin split subband is larger then for the inner one. Thus, the spins from inner
and outer subbands are not compensated with each other, and spin accumulation occurs.
In contrast to Equations 7.7 and 7.8 contains momentum k and quasiparticle

life time as a prefactor.* Since the quasiparticle life time does not depend
on the spin index, and the quasiparticle momentum calculated for a given
energy is obviously different for two spin split subbands, Equation 7.8, imme-
diately allows us to read out the very fact that the nonequilibrium contribu-
tion depends on the spin index; see also Figure 7.3. Thus, the spins are not
compensated with each other as long as the system is out of equilibrium.
Therewith one can see that the efficiency of the current-induced spin polar-
ization is governed by the splitting between two subbands, i.e. it is directly
proportional to the spin-orbit constants. One can also prove the later state-
ments just calculating the net spin density
d 2k
〈 Sx , y , z 〉 = ∑∫
s
(2π)2
Sx , y , z (k , s) g 1( s, k), (7.9)
with Sx , y , z being the spin expectation values.

To conclude this section, we would like to emphasize that the two
Equations 7.7 and 7.8 are just two ways to describe the same physical content.
One can think about spin accumulation either in terms of the momentum
relaxation time τ ps dependent on the spin split subband index, or one may
rely on the exact solution Equation 7.8 which relates the spin accumulation
to the quasimomentum difference between electrons with the opposite spin
orientations.
7.3 ANISOTROPY OF THE INVERSE

SPIN-GALVANIC EFFECT
Similar to the spin-galvanic effect [7, 116, 118] the current-induced spin polar-
ization can be strongly anysotropic, due to the interplay of the Dresselhaus
* Note that the well-known textbook relation v=ћk/m* holds only in the absence of spin-orbit
coupling.
and Rashba terms. The relative strength of these terms is of general impor-
tance because it is directly linked to the manipulation of the spin of charge
carriers in semiconductors, one of the key problems in the field of spintron-
ics. Both Rashba and Dresselhaus couplings result in spin splitting of the
band and give rise to a variety of spin-dependent phenomena which allow
us to evaluate the magnitude of the total spin splitting of electron subbands
[22, 23, 29, 115, 116, 118, 119, 121–130]. Dresselhaus and Rashba terms can
interfere in such a way that macroscopic effects vanish, though the individ-
ual terms are large [115, 124, 125]. For example, both terms can cancel each
other, resulting in a vanishing spin splitting in certain k-space directions [10,
14, 22, 125]. This cancellation leads to the disappearance of an antilocaliza-
tion [122, 131], circular photogalvanic effect [118], magneto-gyrotropic effect
[129, 130], spin-galvanic effect [116] and current-induced spin polarization
[71], the absence of spin relaxation in specific crystallographic directions
[115, 123], the lack of Shubnikov–de Haas beating [124], and has also given
rise to a proposal for spin field-effect transistor operating in the nonballistic
regime [125].
While the interplay of Dresselhaus and Rashba spin splitting may play
a role in QWs of different crystallographic orientations, we focus here on
anisotropy of the inversed spin-galvanic effect in (001)-grown zinc-blende
structure based QWS. For (001)-oriented QWs linear in wave-vector part of
Hamiltonian for the first subband reduces to
k(1) = α(σ x0 k y0 − σ y0 k x0 ) + β(σ x0 k x0 − σ y0 k y0 ), (7.10)
where the parameters α and β result from the structure-inversion and bulk-
inversion asymmetries, respectively, and x0 , y0 are the crystallographic axes
[100] and [010].
To study the anisotropy of the spin accumulation, one can just cal-
culate the net spin density from Equation 7.9, where the integral over k
can be taken easily making the substitution ε = E( s, k ) and assuming that
−∂f 0 (ε) / ∂ε = δ( E F − ε). The rest integrals over the polar angle can be taken
analytically, and after some algebra we have
em*τ 0  β α
〈S 〉 = E. (7.11)
2π3  −α −β
This relation between the spin accumulation and electrical current can also
be deduced phenomenologically applying Equation 7.2 (Figure 7.4).
The magnitude of the spin accumulation 〈 S 〉 = 〈 Sx 〉2 + 〈 S y 〉2 depends
on the relative strength of β and α and varies after
eEm*τ 0 x ).
〈S 〉 = α 2 + β2 + 2αβ sin(2Ee (7.12)
2π3
It is interesting to note, that 〈 S 〉 depends on the direction of the electric field
(see Figure 7.4), i.e. the spin accumulation is anisotropic. This anisotropy
A B C
FIGURE 7.4 Anisotropy of spin accumulation (in arbitrary units) vs. direction
of the electric field in polar coordinates for different Rashba and Dresselhaus
constants: A — α = 3β , B — α = 2 .15β, C — α  β. (After Žutić, I., Nat. Phys. 5, 630,
2009. With permission.) The curve B corresponds to the n-type InAs-based QWs
investigated in [116]. (After Trushin, M. and Schliemann, J., Phys. Rev. B 75, 155323,
reflects the relation between Rashba and Dresselhaus spin–orbit constants.

If either α = 0 or β = 0 then the anisotropy vanishes. In the opposite case of
α  β , the anisotropy reaches its maximum. Note, that the case of α being
exactly equal to β requires special consideration, and the nonequilibrium
distribution function turns out to be different from Equation 7.8. Here, the
effective magnetic field does not depend on the direction of motion, and
Dyakonov–Perel spin relaxation vanishes [10, 14, 123, 125]. A similar effect
has also been found in rolled-up heterostructures [132], where Dyakonov–
Perel spin relaxation also vanishes at a certain radius of curvature. In such
situations, spin orientation by electric current becomes only possible due to
other mechanisms of spin relaxation. Nevertheless, the anisotropy of spin
relaxation will reflect the anisotropy of the band structure and in our case of
(001)-oriented QW will qualitatively correspond to the curve C in Figure 7.4.
The simple Drude-like relation between electrical current and field
allows us to write down a useful relation between the spin accumulation and
charge current density
m*2  β α
〈S 〉 = j. (7.13)
2π3ene  −α −β
Note that, in the coordinate system with the coordinate axes parallel to the
mirror crystallographic planes x [110] and y [110], the Hamiltonian k(1)
gets the form β xy σ x k y + β yx σ y k x with β xy = β + α , β yx = β − α , and the relation

between 〈S〉 and j simplifies to
m*2  0 β + α
〈S 〉 = j. (7.14)
2π3ene  β − α 0 
Hence, for the spin accumulation 〈 S[110] 〉 and 〈 S[110] 〉 provided by the electri-
cal current along [110] and [110] crystallographic directions, respectively we
obtain
〈 S[110] 〉 α + β
= . (7.15)
〈 S[110] 〉 β − α
From Equations 7.15 or 7.14, one can see that the spin accumulation is
strongly anisotropic if the constants α and β are close to each other. As was
discussed earlier, the reason for such an anisotropy can be understood either
from the spin–orbit splitting differently for different direction of the carrier
motion, or from the anisotropic spin relaxation times [123].
Equations 7.14 and 7.15 show that by measuring the spin polarization
for current flowing in particular crystallographic directions, one can map
the spin–orbit constants and deduce their magnitudes. To find the absolute
values of the spin–orbit constants, one needs to know the carrier effective
mass m* and quasiparticle life time τ 0 . The latter can be roughly estimated
from the carrier mobility. Indeed, though the band structure described by
the Hamiltonian (7.10) is anisotropic, the electrical conductivity remains
isotropic which can be verified directly computing the electrical current
∑∫ (2dπk) v g (s, k).

2
j = −e 2 s 1
s
Taking this integral in the same manner as in Equation 7.9, one can see that
the anisotropy of the group velocity v s is compensated by the one stemming
from the distribution function g 1( s, k). Therefore, the electrical conductivity
is described by the Drude-like formula σ = e 2ne τ 0 / m* with
2
m* E F  m*  α 2 + β 2
ne = +
π2  2  π
being the carrier concentration. Using this Drude-like relation, one can
estimate the quasiparticle life time as τ 0  m*µ / e. To give an example, an
n-type InAs quantum well [116] with µ = 2 ⋅ 104 cm 2 /(V ⋅ s) and m*  0.032m0 ,
with m0 being the bare electron mass, yields the quasiparticle life time
τ 0  4 ⋅ 10 −13 s.
We considered the regime linear in the electric current above. Stronger
electric fields provide a needle-shaped electron distribution known as a
“streaming” regime, in which each free charge carrier accelerates quasibal-

listically in the “passive” region until it reaches the optical-phonon energy,
then it emits an optical phonon and starts the next period of acceleration
[117]. The inclusion of a spin degree of freedom into the streaming-regime
kinetics gives rise to rich and interesting spin-related phenomena. In par-
ticular, the current-induced spin-orientation remarkably increases, reach-
ing a high value  2% in the electric field  1kV/cm. The spin polarization
enhancement is caused by squeezing of the electron momentum distribution
in the direction of drift [73, 100]. Note that with further increase in the field
spin polarization falls.
Finally, we would like to address another mechanism of the current-
induced spin polarization and the spin-galvanic effect proposed in Refs.
[69], [133] and [134], respectively. It does not require the spin splitting of
electron spectrum, and is based on spin-dependent electron scattering by
static defects or phonons which is asymmetric in the momentum space in
gyrotropic structures [115, 135]. The spin polarization then occurs due to
asymmetric spin-dependent scattering, followed by the processes of spin
relaxation, which can be of both the Elliott–Yafet and the Dyakonov–Perel
types, discussed further in Chapter 1, Volume 2. The scattering-related
mechanism is expected to dominate at room temperature and/or electron
gas of high density. It may be responsible for the observed current-induced
spin polarization in n-doped InGaAs epilayers [49], where the anisotropy of
both current-induced spin polarization and spin splitting, was studied in the
same structure, and the smallest spin polarization was surprisingly observed
for the electric current applied in the crystal directions corresponding to
larger spin splitting. While it might be important in some particular cases,
a detailed consideration of this mechanism is out of scope of our chapter.
So far we have discussed spin- polarized electric transport in homogeneous
two-dimensional structures. The family of these effects can be extended by
using inhomogeneous structures. For example, spin solar cell and spin pho-
todiode (spin-voltaic effect), proposed in Ref. [136], were recently observed in
GaAs-based p-n junctions [137]. These effects occurred due to illumination
of a p-n junction with circularly polarized interband light resulting in spin
polarization of a charge current. Circular polarization generates spin-polar-
ized electrons and holes. Due to the fast relaxation of hole spin polarization
in the bulk and the long spin lifetime of electrons, the photocurrent becomes
spin polarized.
At last but not least, so far we discussed spin polarized electric transport
in homogeneous two-dimensional structures. The family of these effects
can be extended by using inhomogeneous structure. An example are the
spin solar cell and spin-voltaic effect suggested in [136] recently observed
in n-GaAlAs/p-GaInAs/p-GaAs junction [137], see Chapter 8, Volume 2.
These effects effect occur due to uniform illumination of a p−n junction with
circularly polarized inter-band light resulting in spin polarization of a charge
current. Circular polarization generates spin-polarized electrons and holes.
Due to the fast relaxation of hole spin polarization in the bulk and the long
spin lifetime of electrons, the photocurrent becomes spin polarized.
Acknowledgments 331
7.4 CONCLUDING REMARKS
We have given an overview of current-induced spin polarization in gyrotropic
semiconductor nanostructures. Such a spin polarization as a response to a
charge current may be classified as the inverse of the spin-galvanic effect, and
is sometimes called a magneto-electrical effect or Edelstein (Rashba–Edelstein)
effect. Apart from reviewing the experimental status of affairs, we have provided
a detailed theoretical description of both effects in terms of a phenomenologi-
cal model of spin-dependent relaxation processes, and an alternative theoreti-
cal approach based on the quasiclassical Boltzmann equation. Two microscopic
mechanisms of this effect, both involving k-linear band spin splitting, are
known so far: scattering mechanism based on Elliott–Yafet spin relaxation and
precessional mechanism due to D’yakonov–Perel spin relaxation. We note that
recently, another mechanism of the current-induced spin polarization [133]
and spin-galvanic effect [134] which may dominate at room temperature has
been proposed. It is based on spin-dependent asymmetry of electron scattering,
and is out of the scope of the present chapter. The relative direction between
electric current and nonequilibrium spin for all mechanisms is determined by
the crystal symmetry. In particular, we have discussed the anisotropy of the
inverse spin-galvanic effect in (001)-grown zinc-blende structure based QWs in
the presence of spin-orbit interaction of both the Rashba and the Dresselhaus
type. Combined with the theoretical achievements derived from the Boltzmann
approach, the precise measurement of this anisotropy allows in principle the
determination of the absolute values of the Rashba and the Dresselhaus spin-
orbit coupling parameter. Moreover, these interactions can be used for the
manipulation of the magnitude and direction of electron spins by changing the
direction of current, thereby enabling a new degree of spin control. We focused
here on the fundamental physics underlying the spin-dependent transport of
carriers in semiconductors, which persists up to room temperature [38, 39, 45],
and, therefore, may become useful in future semiconductor spintronics.
As of December 2018, the field still attracts a considerable amount of
attention as reflected by the number of recent papers devoted to the prob-
lem of current induced spin polarization. The major theoretical efforts have
been recently applied towards understanding the interfaces between ferro-
magnetic films [138, 139, 140, 141], leave alone a few Keldysh-based descrip-
tions of the current-induced spin polarization [142, 143, 144] and a unified
description of current-induced spin orientation, spin-galvanic, and related
spin-Hall effects [7, 145, 146], see Chapter 4, Volume 2. The current induced
spin polarization has also been found in a few novel materials including two-
dimensional transition metal dichalcogenides [147], oxide interfaces [148],
p-doped tellurium [149, 150, 151], and topological insulators [152].
ACKNOWLEDGMENTS
We thank E.L. Ivchenko, V.V. Bel’kov, D. Weiss, L.E. Golub, and S.A. Tarasenko
for helpful discussions. This work is supported by the DFG (SFB 689 and
1277) and the Elite Network of Bavaria (K-NW-2013-247). M.T. is supported
by the Director’s Senior Research Fellowship from CA2DM at NUS (NRF

Singapore Medium Sized Centre Programme R-723-000-001-281).
REFERENCES
1. R. H. Silsbee, Spin-orbit induced coupling of charge current and spin polariza-
tion, J. Phys. Condens. Matter 16, R179 (2004).
2. E. I. Rashba, Spin currents, spin populations, and dielectric function of non-
centrosymmetric semiconductors, Phys. Rev. B 70, 161201(R) (2004).
3. E. I. Rashba, Spin dynamics and spin transport, J. Supercond. Incorporating
Novel Magn. 18, 137 (2005).
4. E. I. Rashba, Semiconductor spintronics: Progress and challenges, in
Future Trends in Microelectronics. Up to Nano Creek, S. Luryi, J. M. Xu,
and A. Zaslavsky (Eds.), Wiley-Interscience, Hoboken, pp. 28–40, 2007,
arXiv:cond-mat/0611194.
5. R. Winkler, Spin-Dependent Transport of Carriers in Semiconductors, in Hand
book of Magnetism and Advanced Magnetic Materials, vol. 5, H. Kronmuller and
S. Parkin (Eds.), John Wiley & Sons, New York, 2007, arXiv:cond-mat/0605390.
6. I. Adagideli, M. Scheid, M. Wimmer, G. E. W. Bauer, and K. Richter, Extracting
current-induced spins: Spin boundary conditions at narrow Hall contacts,
New J. Phys. 9, 382 (2007).
7. E. L. Ivchenko and S. D. Ganichev, Spin Photogalvanics, in Spin Physics in
Semiconductors M. I. Dyakonov (Ed.), Springer, Berlin, pp. 245–277, 2008, sec-
ond edition pp. 281–328 (2017).
8. Y. K. Kato and D. D. Awschalom, SPECIAL TOPICS, Advances in spintronics,
electrical manipulation of spins in nonmagnetic semiconductors, J. Phys. Soc.
Jpn. 77, 031006 (2008).
9. S. D. Ganichev, Spin-galvanic effect and spin orientation by current in non-
magnetic semiconductors – A review, Int. J. Mod. Phys. B 22, 1 (2008).
10. S. D. Ganichev and L. E. Golub, Interplay of Rashba/Dresselhaus spin splittings
probed by photogalvanic spectroscopy – A review, Phys. Stat. Solidi B 251,
1801 (2014).
Hall effects, Rev. Mod. Phys. 87, 1213 (2015).
12. A. Manchon, H. C. Koo, J. Nitta, S. M. Frolov, and R. A. Duine, New perspec-
tives for Rashba spin-rbit coupling, Nat. Mater. 14, 871 (2015).
13. T. Jungwirth, X. Marti, P. Wadley, and J. Wunderlich, Antiferromagnetic spin-
tronics, Nat. Nanotechnol. 11, 231 (2016).
14. J. Schliemann, Persistent spin textures in semiconductor nanostructures, Rev.
Mod. Phys. 89, 011001 (2017).
15. L. D. Landau, E. M. Lifshits, and L. P. Pitaevskii, Electrodynamics of Continuous
Media, vol. 8, Elsevier, Amsterdam, 1984.
16. J. F. Nye, Physical Properties of Crystals: Their Representation by Tensors and
Matrices, Oxford University Press, Oxford, 1985.
17. V. M. Agranovich and V. L. Ginzburg, Crystal Optics with Spatial Dispersion,
and Ex-citons in Springer Series, in Solid-State Sciences, vol. 42, Springer,
Berlin, 1984.
18. V. A. Kizel’, Yu. I. Krasilov, and V. I. Burkov, Experimental studies of gyrotropy
of crystals, Usp. Fiz. Nauk 114, 295 (1974) [Sov. Phys. Usp. 17, 745 (1975)].
19. J. Jerphagnon and D. S. Chemla, Experimental studies of gyrotropy of crystals,
J. Chem. Phys. 65, 1522 (1976).
20. B. Koopmans, P. V. Santos, and M. Cardona, Optical activity in semiconduc-
tors: stress and confinement effects, Phys. Stat. Sol. (B) 205, 419 (1998).
21. S. D. Ganichev, E. L. Ivchenko, V. V. Bel’kov et al., Spin-galvanic effect, Nature
417, 153 (2002).
References 333
22. S. D. Ganichev and W. Prettl, Spin photocurrents in quantum wells, J. Phys.

Condens. Matter 15, R935 (2003).
23. S. D. Ganichev and W. Prettl, Intense Terahertz Excitation of Semiconductors,
Oxford University Press, Oxford, 2006.
24. E. L. Ivchenko, Optical Spectroscopy of Semiconductor Nanostructures, Alpha
Science International, Harrow, UK, 2005.
25. J. Schliemann, Int. J. Mod. Phys. B 20, 1015 (2006).
26. H.-A. Engel, E. I. Rashba, and B. I. Halperin, in Handbook of Magnetism and
Advanced Magnetic Materials, H. KronmÜller and S. Parkin (Eds.), John Wiley,
2007.
27. E. M. Hankiewicz and G. Vignale, Spin-Hall effect and spin-Coulomb drag in
doped semiconductors, J. Phys. Condens. Matter 21, 253202 (2009).
28. M. I. Dyakonov and V. I. Perel, Possibility of orienting spins with current,
Pis’ma Zh. Eksp. Teor. Fiz. 13, 657 (1971) [JETP Lett. 13, 467 (1971)].
spintronics, Acta Phys. Slov. 57, 565 (2007), arXiv:cond-mat/0711.1461.
30. J. Sinova, and A. MacDonald, Theory of spin-orbit effects in semiconductors,
in Spintronics, vol. 82, T. Dietl, D.D. Awschalom, M. Kaminska and H. Ohno,
(Eds.), Semiconductors and Semimetals, Elsevier, Amsterdam, pp. 45–89, 2008.
31. M. I. Dyakonov and A. V. Khaetskii, Spin Hall effect, in Spin Physics in
Semiconductors, M. I. Dyakonov (Ed.), Springer, Berlin, pp. 211–244, 2008,
second edition pp. 241–280.
32. E. L. Ivchenko and G. E. Pikus, New photogalvanic effect in gyrotropic crys-
tals, Pis’ma Zh. Eksp. Teor. Fiz. 27, 640 (1978) [JETP Lett. 27, 604 (1978)].
33. L. E. Vorob’ev, E. L. Ivchenko, G. E. Pikus, I. I. Farbstein, V. A. Shalygin, and
A. V. Sturbin, Optical activity in tellurium induced by a current, Pis’ma Zh.
Eksp. Teor. Fiz. 29, 485 (1979) [JETP Lett. 29, 441 (1979)].
34. A. G. Aronov, Yu. B. Lyanda-Geller, and G. E. Pikus, Spin polarization of elec-
trons by an electric current, Zh Eksp. Teor. Fiz. 100, 973 (1991) [Sov. Phys. JETP
73, 537 (1991)].
35. F. T. Vas’ko and N. A. Prima, Spin splitting of the spectrum of two-dimen-
sional electrons, Fiz. Tverdogo Tela 21, 1734 (1979) [Sov. Phys. Solid State 21,
994 (1979)].
36. A. G. Aronov and Yu. B. Lyanda-Geller, Nuclear electric resonance and orien-
tation of carrier spins by an electric field, Pis’ma Zh. Eksp. Teor. Fiz. 50, 398
(1989) [JETP Lett. 50, 431 (1989)].
37. V. M. Edelstein, Spin polarization of conduction electrons induced by elec-
tric current in two-dimensional asymmetric electron systems, Solid State
Commun. 73, 233 (1990).
38. S. D. Ganichev, S. N. Danilov, P. Schneider et al., Can electric current orient
spins in quantum wells? cond-mat/0403641 (2004).
39. S. D. Ganichev, S. N. Danilov, P. Schneider et al., Electric current induced
spin orientation in quantum well structures, J. Magn. Magn. Mater. 300, 127
(2006).
40. A. Yu. Silov, P. A. Blajnov, J. H. Wolter, R. Hey, K. H. Ploog, and N. S. Averkiev,
Currentinduced spin polarization at a single heterojunction, Appl. Phys. Lett.
85, 5929 (2004).
41. N. S. Averkiev, and A. Yu. Silov, Circular polarization of luminescence caused
by the current in quantum wells, Semiconductors 39, 1323 (2005).
42. V. Sih, R. C. Myers, Y. K. Kato, W. H. Lau, A. C. Gossard, and D. D. Awschalom,
Spatial imaging of the spin Hall effect and current-induced polarization in
two-dimensional electron gases, Nat. Phys. 1, 31 (2005).
43. C. L. Yang, H. T. He, L. Ding et al., Spin photocurrent and converse spin polar-
ization induced in a InGaAs/InAlAs two-dimensional electron gas, Phys. Rev.
Lett. 96, 186605 (2006).
44. Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom, Current-induced
spin polarization in strained semiconductors, Phys. Rev. Lett. 93, 176601 (2004).
45. Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom, Current-induced

spin polarization in strained semiconductors, Appl. Phys. Lett. 87, 022503
(2005).
46. N. P. Stern, S. Ghosh, G. Xiang, M. Zhu, N. Samarth, and D. D. Awschalom,
Current-induced polarization and the spin Hall effect at room temperature,
Phys. Rev. Lett. 97, 126603 (2006).
47. S. Kuhlen, K. Schmalbuch, M. Hagedorn et al., Electric field-driven coher-
ent spin reorientation of optically generated electron spin packets in InGaAs,
Phys. Rev. Lett. 109, 146603 (2012).
48. I. Stepanov, S. Kuhlen, M. Ersfeld, M. Lepsa, and B. Beschoten, All-electrical
timeresolved spin generation and spin manipulation in n-InGaAs, Appl. Phys.
Lett. 104, 062406 (2014).
49. B. M. Norman, C. J. Trowbridge, D. D. Awschalom, and V. Sih, Current-induced
spin polarization in anisotropic spin-orbit fields, Phys. Rev. Lett. 112, 056601
(2014).
50. F. G. G. Hernandez, S. Ullah, G. J. Ferreira, N. M. Kawahala, G. M. Gusev,
and A. K. Bakarov, Macroscopic transverse drift of long current-induced spin
coherence in two-dimensional electron gases, Phys. Rev. B 94, 045305 (2016).
51. L. E. Golub and E. L. Ivchenko, Spin orientation by electric current in (110)
quantum wells, Phys. Rev. B 84, 115303 (2011).
52. P. R. Hammar, B. R. Bennett, M. J. Yang, and M. Johnson, Observation of spin
injection at a ferromagnet-semiconductor interface, Phys. Rev. Lett. 83, 203
(1999).
53. P. R. Hammar, B. R. Bennett, M. J. Yang, and M. Johnson, A Reply to the
Comment by F. G. Monzon, H. X. Tang, and M. L. Roukes, and also to
the Comment by B. J. van Wees, Phys. Rev. Lett. 84, 5024 (2000).
54. M. Johnson, Theory of spin-dependent transport in ferromagnet-semiconduc-
tor heterostructures, Phys. Rev. B 58, 9635 (1998).
55. R. H. Silsbee, Theory of the detection of current induced spin polarization in a
two-dimensional electron gas, Phys. Rev. B 63, 155305 (2001).
56. F. G. Monzon, H. X. Tang, and M. L. Roukes, Magnetoelectronic phenomena
at a ferromagnet-semiconductor interface, Phys. Rev. Lett. 84, 5022 (2000).
57. B. J. van Wees, Phys. Rev. Lett. 84, 5023 (2000).
58. T. D. Skinner, K. Olejník, L. K. Cunningham et al., Complementary spin-Hall
and inverse spin-galvanic effect torques in a ferromagnet/semiconductor
bilayer, Nat. Commun. 6, 6730 (2015).
59. K. Olejník, V. Novák, J. Wunderlich, and T. Jungwirth, Electrical detection of
magnetization reversal without auxiliary magnets, Phys. Rev. B 91, 180402(R)
(2015).
60. J. C. Rojas Sánchez, L. Viva, G. Desfonds et al., Spin-to-charge conversion
using Rashba coupling at the interface between non-magnetic materials, Nat.
Commun. 4, 2944 (2013).
61. H. J. Zhang, S. Yamamoto, Y. Fukaya et al., Current-induced spin polarization
on metal surfaces probed by spin-polarized positron beam, Sci. Rep. 4, 4844
(2014).
62. W. Zhang, M. B. Jungfleisch, W. Jiang, J. E. Pearson, and A. Hoffmann, Spin
pumping and inverse Rashba-Edelstein effect in NiFe/Ag/Bi and NiFe/Ag/Sb,
J. Appl. Phys. 117, 17C727 (2015).
63. S. Sangiao, J. M. De Teresa, L. Morellon, I. Lucas, M. C. Martinez-Velarte, and
M. Viret, Control of the spin to charge conversion using the inverse Rashba-
Edelstein effect, Appl. Phys. Lett. 106, 172403 (2015).
64. A. Nomura, T. Tashiro, H. Nakayama, and K. Ando, Temperature dependence
of inverse Rashba-Edelstein effect at metallic interface, Appl. Phys. Lett. 106,
212403 (2015).
65. M. Isasa, M. C. Martínez-Velarte, E. Villamor et al., Origin of inverse Rashba-
Edelstein effect detected at the Cu/Bi interface using lateral spin valves, Phys.
Rev. B 93, 014420 (2016).
References 335
66. Y. Shiomi, K. Nomura, Y. Kajiwara et al., Spin-electricity conversion induced

by spin injection into topological insulators, Phys. Rev. Lett. 113, 196601 (2014).
67. A. R. Mellnik, J. S. Lee, A. Richardella et al., Spin-transfer torque generated by
a topological insulator, Nature 511, 449 (2014).
68. M. G. Vavilov, Giant magneto-oscillations of electric-field-induced spin polar-
ization in a two-dimensional electron gas, Phys. Rev. B 72, 195327 (2005).
69. S. A. Tarasenko, Scattering induced spin orientation and spin currents in gyro-
tropic structures, JETP Lett. 84, 199 (2006).
70. H.-A. Engel, E. I. Rashba, and B. I. Halperin, Out-of-plane spin polarization
from in plane electric and magnetic fields, Phys. Rev. Lett. 98, 036602 (2007).
71. M. Trushin and J. Schliemann, Anisotropic current-induced spin accumula-
tion in the two-dimensional electron gas with spin-orbit coupling, Phys. Rev. B
75, 155323 (2007).
72. O. E. Raichev, Frequency dependence of induced spin polarization and spin
current in quantum wells, Phys. Rev. B 75, 205340 (2007).
73. L. E. Golub and E. L. Ivchenko, Spin-dependent phenomena in semiconductors
in strong electric fields, New J. Phys. 15, 125003 (2013).
74. D. Culcer and R. Winkler, Generation of spin currents and spin densities in
systems with reduced symmetry, Phys. Rev. Lett. 99, 226601 (2007).
75. A. A. Burkov, A. S. Nunez, and A. H. MacDonald, Theory of spin-charge-cou-
pled transport in a two-dimensional electron gas with Rashba spin-orbit inter-
actions, Phys. Rev. B 70, 155308 (2004).
76. O. Bleibaum, Spin diffusion equations for systems with Rashba spin-orbit
interaction in an electric field, Phys. Rev. B 73, 035322 (2006).
77. V. L. Korenev, Bulk electron spin polarization generated by the spin Hall cur-
rent, Phys. Rev. B 74, 041308 (2006).
78. P. Kleinert and V. V. Bryksin, Spin polarization in biased Rashba –Dresselhaus
two-dimensional electron systems, Phys. Rev. B 76, 205326 (2007).
79. M. Duckheim and D. Loss, Loss Mesoscopic fluctuations in the spin-electric
susceptibility due to Rashba spin-orbit interaction, Phys. Rev. Lett. 101, 226602
(2008).
80. P. Kleinert and V. V. Bryksin, Electric-field-induced long-lived spin excitations
in two-dimensional spin-orbit coupled systems, Phys. Rev. B 79, 045317 (2009).
81. M. Duckheim, D. Loss, M. Scheid, K. Richter, İ. Adagideli, and P. Jacquod, Spin
accumulation in diffusive conductors with Rashba and Dresselhaus spin-orbit
interaction, Phys. Rev. B 81, 085303 (2010).
82. Ka Shen, G. Vignale, and R. Raimondi, Microscopic theory of the inverse
Edelstein effect, Phys. Rev. Lett. 112, 096601 (2014).
83. Ka Shen, R. Raimondi, and G. Vignale, Theory of coupled spin-charge trans-
port due to spin-orbit interaction in inhomogeneous two-dimensional elec-
tron liquids, Phys. Rev. B 90, 245302 (2014).
84. F. T. Vasko and O. E. Raichev, Quantum Kinetic Theory and Application:
Electrons, Photons, Phonons, Springer, Berlin, pp. 662–663, 2005.
85. A. V. Chaplik, M. V. Entin, and L.I. Magarill, Phys. E 13, 744 (2002).
86. J. Inoue, G. E. W. Bauer, and L. W. Molenkamp, Diffuse transport and spin
accumulation in a Rashba two-dimensional electron gas, Phys. Rev. B 67,
033104 (2003).
87. Y. J. Bao and S. Q. Shen, Electric-field-induced resonant spin polarization in a
88. R. Raimondi, C. Gorini, M. Dzierzawa, and P. Schwab, Current-induced spin
polarization and the spin Hall effect: A quasiclassical approach, Solid State
Commun. 144, 524 (2007).
89. C. X. Liu, B. Zhou, S. Q. Shen, and B.-f. Zhu, Current-induced spin polarization
in a two-dimensional hole gas, Phys. Rev. B 77, 125345 (2008).
90. C. Gorini, P. Schwab, M. Dzierzawa, and R. Raimondi, Spin polarizations and
spin Hall currents in a two-dimensional electron gas with magnetic impuri-
ties, Phys. Rev. B 78, 125327 (2008).
91. R. Raimondi and P. Schwab, Tuning the spin hall effect in a two-dimensional
electron gas, Europhys. Lett. 87, 37008 (2009).
92. I. Adagideli, G. E. W. Bauer, and B. I. Halperin, Detection of current-induced
spins by ferromagnetic contacts, Phys. Rev. Lett. 97, 256601 (2006).
93. Y. Jiang and L. Hu, Kinetic magnetoelectric effect in a two-dimensional semi-
conductor strip due to boundary-confinement-induced spin-orbit coupling,
Phys. Rev. B 74, 075302 (2006).
94. J. H. Jiang, M. W. Wu, and Y. Zhou, Kinetics of spin coherence of electrons in
n-type InAs quantum wells under intense terahertz laser fields, Phys. Rev. B
78, 125309 (2008).
95. M. Pletyukhov and A. Shnirman, Spin density induced by electromagnetic
waves in a two-dimensional electron gas with both Rashba and Dresselhaus
spin-orbit coupling, Phys. Rev. B 79, 033303 (2009).
96. M.-H. Liu, S.-H. Chen, and C.-R. Chang, Current-induced spin polarization in
spin-orbit-coupled electron systems, Phys. Rev. B 78, 165316 (2008).
97. D. Guerci, J. Borge, and R. Raimondi, Spin polarization induced by an electric
field in the presence of weak localization effects, Phys. E. 75, 370 (2016).
98. A. Zakharova, I. Lapushkin, K. Nilsson, S. T. Yen, and K. A. Chao, Spin polar-
ization of an electron-hole gas in InAs/GaSb quantum wells under a dc cur-
rent, Phys. Rev. B 73, 125337 (2007).
99. I. Garate and A. H. MacDonald, Influence of a transport current on magnetic
anisotropy in gyrotropic ferromagnets, Phys. Rev. B 80, 134403 (2009).
100. L. E. Golub and E. L. Ivchenko, Advances in Semiconductor Research: Physics of
Nanosystems, Spintronics and Technological Applications, D. Persano Adorno
and S. Pokutnyi (Eds.), Nova Science Publishers, pp 93–104, 2014.
101. G. Vignale and I. V. Tokatly, Theory of the nonlinear Rashba-Edelstein effect:
The clean electron gas limit, Phys. Rev. B 93, 035310 (2016).
102. I. A. Kokurin and N. S. Averkiev, Orientation of electron spins by the current
in a quasi-one-dimensional system, JETP Lett. 101, 568 (2015).
103. M. P. Trushin and A. L. Chudnovskiy, A curved one-dimensional wire with
Rashba coupling as a spin switch, JETP Lett. 83, 318 (2006).
104. I. V. Tokatly, E. E. Krasovskii, and G. Vignale, Current-induced spin polariza-
tion at the surface of metallic films: A theorem and an ab-initio calculation,
Phys. Rev. B 91, 035403 (2015).
105. I. Garate and M. Franz, Inverse spin-galvanic effect in the interface between a
topological insulator and a ferromagnet, Phys. Rev. Lett. 104, 146802 (2010).
106. T. Yokoyama, J. Zang, and N. Nagaosa, Theoretical study of the dynamics of
magnetization on the topological surface, Phys. Rev. B 81, 241410(R) (2010).
107. T. Yokoyama, Current-induced magnetization reversal on the surface of a
topological insulator, Phys. Rev. B 84, 113407 (2011).
108. W. Luo, W. Y. Deng, H. Geng et al., Perfect inverse spin Hall effect and inverse
Edelstein effect due to helical spin-momentum locking in topological surface
states, Phys. Rev. B 93, 115118 (2016).
109. H. Li, X. Wang, and A. Manchon, Valley-dependent spin-orbit torques in two-
dimensional hexagonal crystals, Phys. Rev. B 93, 035417 (2016).
110. A. Dyrdał, J. Barnaś, and V. K. Dugaev, Current-induced spin polarization in
graphene due to Rashba spin-orbit interaction, Phys. Rev. B 89, 075422 (2014).
111. J. Borge, C. Gorini, G. Vignale, and R. Raimondi, Spin Hall and Edelstein effects
in metallic films: From two to three dimensions, Phys. Rev. B 89, 245443 (2014).
112. L. S. Levitov, Y. V. Nazarov, and G. M. Eliashberg, Magnetoelectric effects in
conductors with mirror isomer symmetry, Zh Eksp. Teor. Fiz. 88, 229 (1985)
[Sov. Phys. JETP 61, 133 (1985)].
113. Y. A. Bychkov and E. I. Rashba, Properties of a 2D electron gas with lifted spectral
degeneracy, Pis’ma Zh. Eksp. Teor. Fiz. 39, 66 (1984) [JETP Lett. 39, 78 (1984)].
114. M. I. D’yakonov and V. Yu. Kachorovskii, Spin relaxation of two-dimensional
electrons in noncentrosymmetric semiconductors, Fiz. Tekh. Poluprovodn. 20,
178 (1986) [Sov. Phys. Semicond. 20, 110 (1986)].
References 337
115. N. S. Averkiev, L. E. Golub, and M. Willander, Spin relaxation anisotropy in two-

dimensional semiconductor systems, J. Phys. Condens. Matter 14, R271 (2002).
116. S. D. Ganichev, V. V. Bel’kov, L. E. Golub et al., Experimental separation of
Rashba and Dresselhaus spin-splittings in semiconductor quantum wells,
Phys. Rev. Lett. 92, 256601 (2004).
117. I. I. Vosilyus and I. B. Levinson, Galvanomagnetic effects in strong electric
fields during nonelastic electron scattering, Sov. Phys. JETP 25, 672 (1967).
118. S. Giglberger, L. E. Golub, V. V. Bel’kov et al., Rashba and dresselhaus spin-
splittings in semiconductor quantum wells measured by spin photocurrents,
Phys. Rev. B 75 035327 (2007).
119. L. Meier, G. Salis, E. Gini, I. Shorubalko, and K. Ensslin, Two-dimensional imag-
ing of the spin-orbit effective magnetic field, Phys. Rev. B 77, 035305 (2008).
120. L. E. Golub, Spin transport in heterostructures, Phys.-Usp. 55, 814 (2012).
121. J. Luo, H. Munekata, F. F. Fang, and P. J. Stiles, Effects of inversion asymmetry
on electron energy band structures in GaSb/InAs/GaSb quantum wells, Phys.
Rev. B 41, 7685 (1990).
122. W. Knap, C. Skierbiszewski, A. Zduniak et al., Weak antilocalization and spin
precession in quantum wells, Phys. Rev. B 53, 3912 (1996).
123. N. S. Averkiev and L. E. Golub, Giant spin relaxation anisotropy in zinc-blende
heterostructures, Phys. Rev. B 60, 15582 (1999).
124. S. A. Tarasenko and N. S. Averkiev, Interference of spin splittings in magneto-
oscillation phenomena in two-dimensional systems, Pis’ma Zh. Èksp. Teor. Fiz.
75, 669 (2002) [JETP Lett. 75, 552 (2002)]; N. S. Averkiev, M. M. Glazov, and S.
A. Tarasenko, Solid State Commun. 133, 543 (2005).
125. J. Schliemann, J. C. Egues and D. Loss, Nonballistic spin-field-effect transistor,
Phys. Rev. Lett. 90, 146801 (2003).
Systems, vol. 191, Springer Tracts in Modern Physics, Springer, Berlin, 2003.
127. W. Zawadzki and P. Pfeffer, Spin splitting of subbands energies due to inversion
asymmetry in semiconductor heterostructures, Semicond. Sci. Technol. 19, R1
(2004).
tions. Rev. Mod. Phys. 76, 323 (2004).
129. V. V. Bel’kov, P. Olbrich, S. A. Tarasenko et al., Symmetry and spin dephasing
in (110)-grown quantum wells, Phys. Rev. Lett. 100, 176806 (2008).
130. V. Lechner, L. E. Golub, P. Olbrich et al., Tuning of structure inversion asym-
metry by the δ-doping position in (001)-grown GaAs quantum wells, Appl.
Phys. Lett. 94, 242109 (2009).
131. M. Kohda, V. Lechner, Y. Kunihashi et al., Gate-controlled persistent spin helix
state in InGaAs quantum wells, Phys. Rev. B 86, R081306 (2012).
132. M. Trushin and J. Schliemann, Spin dynamics in rolled-up two-dimensional
electron gase, New J. Phys. 9, 346 (2007).
133. S. A. Tarasenko, Spin orientation of free carriers by dc and high-frequency
electric field in quantum wells, Phys. E 40, 1614 (2008).
134. L. E. Golub, JETP Lett. 85, 393 (2007).
135. E. L. Ivchenko and S. A. Tarasenko, New mechanism of the spin-galvanic
effect, JETP 99, 379 (2004).
136. I. Žutić, J. Fabian, and S. Das Sarma, Spin-polarized transport in inhomoge-
neous magnetic semiconductors: Theory of magnetic/nonmagnetic p−n junc-
tions, Phys. Rev. Lett. 88, 066603 (2002); I. Žutić, J. Fabian, and S. Das Sarma,
Spin injection through the depletion layer: A theory of spin-polarized p–n
junctions and solar cells, Phys. Rev. B 64, 121201 (2001).
137. B. Endres, M. Ciorga, M. Schmid, M. Utz, D. Bougeard, D. Weiss, G. Bayreuther,
and C.H. Back, Demonstration of the spin solar cell and spin photodiode effect,
Nat. Commun. 4, 2068 (2013).
pling: Formalism, Phys. Rev. B 94, 104419 (2016).
pling: Phenomenology, Phys. Rev. B 94, 104420 (2016).
140. T. H. Dang, D. Q. To, E. Erina, T. L. Hoai Nguyen, V. I. Safarov, H. Jaffrès, H.-J.
Drouhin, Theory of the anomalous tunnel Hall effect at ferromagnet-semicon-
ductor junctions, J. Magn. Magn. Mater. 459, 37 (2018).
141. S. Tölle, U. Eckern, and C. Gorini, Spin-charge coupled dynamics driven by a
time-dependent magnetization, Phys. Rev. B 95, 115404 (2017).
142. C. Gorini, A. M. Sheikhabadi, K. Shen, I. V. Tokatly, G. Vignale, and R.
Raimondi, Theory of current-induced spin polarization in an electron gas,
Phys. Rev. B 95, 205424 (2017).
143. A. M. Sheikhabadi, I. Miatka, E. Y. Sherman, and R. Raimondi, Theory of the
inverse spin galvanic effect in quantum wells, Phys. Rev. B 97, 235412 (2018).
144. I. V. Tokatly, Usadel equation in the presence of intrinsic spin-orbit coupling:
A unified theory of magnetoelectric effects in normal and superconducting
systems, Phys. Rev. B 96, 060502(R) (2017).
145. D. S. Smirnov and L. E. Golub, Electrical spin orientation, spin-galvanic, and
spin-Hall effects in disordered two-dimensional systems, Phys. Rev. Lett. 118,
116801 (2017).
146. A. V. Shumilin, D. S. Smirnov, and L. E. Golub, Spin-related phenomena in the
two-dimensional hopping regime in magnetic field, Phys. Rev. B 98, 155304
(2018).
147. X.-T. An, M. W.-Y. Tu, V. I. Fal’ko, and W. Yao, Realization of valley and spin
pumps by scattering at nonmagnetic disorders, Phys. Rev. Lett. 118, 096602
(2017).
148. G. Seibold, S. Caprara, M. Grilli, and R. Raimondi, Theory of the spin galvanic
effect at oxide interfaces, Phys. Rev. Lett. 119, 256801 (2017).
149. V. A. Shalygin, A. N. Sofronov, L. E. Vorob’ev, and I. I. Farbshtein, Current-
induced spin polarization of holes in tellurium, Phys. Solid State 54, 2362
(2012).
150. C. Sahin, J. Rou, J. Ma, and D. A. Pesin, Pancharatnam-Berry phase and kinetic
magnetoelectric effect in trigonal tellurium, Phys. Rev. B 97, 205206 (2018).
151. S. S. Tsirkin, P. A. Puente, and I. Souza, Gyrotropic effects in trigonal tellurium
studied from first principles, Phys. Rev. B 97, 035158 (2018).
152. J. Tian, S. Hong, I. Miotkowski, S. Datta, and Y. P. Chen, Observation of cur-
rent-induced, long-lived persistent spin polarization in a topological insulator:
A rechargeable spin battery, Sci. Adv. 3 (4), e1602531 (2017).
8
Anomalous and
Spin-Injection
Hall Effects

8.2 Spin-Dependent Hall Effects in Spin–Orbit-Coupled
Systems 341
8.2.1 Anomalous Hall Effect 341
8.2.2 Mechanism of the Spin-Dependent Hall Effects 343
8.2.2.1 Intrinsic Contribution, s AH-int
xy 345
8.2.2.2 Skew-Scattering Contribution, s AH-skew
xy 346
8.2.2.3 Side-Jump Contribution s xy 347
AH-sj
8.2.3 Recent Developments on the Anomalous Hall

Effect 349
8.2.4 Spin Hall Effect 350
8.2.5 Spin-Injection Hall Effect 351
8.3 Spin-Injection Hall Effect: Experiment 352
339
340 Chapter 8. Anomalous and Spin-Injection Hall Effects
8.4 Theory of Spin-Injection Hall Effect 356

8.4.1 Nonequilibrium Polarization Dynamics along
–
the [110] Channel 357
8.4.2 Hall Effect 359
8.4.3 Combined Spin–Orbit and Zeeman Effects in
External Magnetic Fields 361
8.4.4 Prospectives of Spin-Injection Hall Effect 362
References 364
8.1 INTRODUCTION
This Chapter provides a valuable introduction to a growing family of Hall
effects and the importance of the spin-orbit coupling. Subsequent develop-
ments in Hall effects, spurred by the discovery of novel two-dimensional
materials (Chapter 5, Volume 3) and topological insulators (Chapter 14,
Volume 2), are discussed throughout this book. For example, recent advances
in fundamental phenomena and possible applications of Hall effects in a
wide class of materials are addressed in Chapter 9, Volume 1, Chapters 4
and 15, Volume 2, and Chapters 7–9, Volume 3. Throughout this book we
have learned about many aspects of spin dynamics in both metal and semi-
conducting systems involving the diagonal transport of spin and charge
and their interactions, in some cases, with the collective ferromagnetic-
order parameter driven out of equilibrium. In this chapter, we venture into
another aspect of spin-charge dynamics, in which an applied electric field
induces a spin or a spin-charge response transverse to the field. These are the
spin-dependent Hall effects.
In metallic spintronics, many studies have centered around the injec-
tion of spin and its manipulation via external magnetic fields, or nonequi-
librium current effects such as spin-torque effects. Within these studies, the
spin–orbit (SO) coupling in the system becomes an important source of spin
dephasing and, in many instances, it is important to minimize it in order to
enhance the observed effects. On the other hand, in semiconductor spin-
tronics, SO coupling comes to the forefront as a source of spin manipulation
and spin generation, becoming a necessary element in physical phenomena
of the spin-dependent Hall effects, such as the anomalous Hall effect, spin
Hall effect, inverse spin Hall effect, and spin-injection Hall effect. These
spin-dependent Hall effects are, of course, governed by common mecha-
nisms that we describe in this chapter.
We focus primarily on the anomalous Hall effect and the spin-injection
Hall effect. The former is the primary spin-dependent Hall effect that has
been studied for many decades, but it is one that continues to fascinate, and
is at the center of many theoretical and experimental studies. The latter is
one of the recent members of the spintronics Hall family [1], and its phenom-
enology combines many interesting physics observed in semiconducting
8.2 Spin-Dependent Hall Effects in Spin–Orbit-Coupled Systems 341
spintronics, and can be a good playground to study in detail these mecha-

nisms and other spin-dynamics effects controlled by particular choices of
SO-coupling strengths that can be tuned experimentally. The other princi-
pal spin-dependent Hall effects, the spin Hall effect and the inverse spin Hall
effect, have also been intensively studied over the past five years and their
recent progress has been presented in several reviews. We will therefore only
provide a simple and short qualitative description of these reviews. Other
chapters that are pertinent to these physics are Chapter 5, Volume 1 on spin-
injection and detection through non-local measurements by Mark Johnson,
and Chapter 7, Volume 3 on spin Hall effects in metallic system by Takahashi
and Maekawa.
This chapter is hence structured as follows: In the remainder of this chap-
ter, we introduce the different spin-dependent Hall effects, which are most
closely related. In the discussion of the anomalous Hall effect (Section 8.2.1),
we introduce the basic mechanisms that give rise to these effects within
the metallic systems (Section 8.2.2). In Section 8.2.5, we discuss the experi-
mental phenomenology of the spin-injection Hall effect, and in Section 8.4,
we discuss the theoretical aspects of this phenomenology, which incorpo-
rates anomalous Hall effect physics as well as persistent spin-helix physics.
Throughout, we focus more on qualitative descriptions rather than detailed
derivations, and refer to the list of references for further details.
8.2 SPIN-DEPENDENT HALL EFFECTS IN

SPIN–ORBIT-COUPLED SYSTEMS
Within the family of spin-dependent Hall effects, the anomalous Hall effect,
the spin Hall effect, the inverse spin Hall effect, and the spin-injection Hall
effect are among the most closely related, as sketched in Figure 8.1. The more
extensively studied of these is the anomalous Hall effect from which many
of the mechanisms have been identified. Several extensive reviews have been
written over the years on these subjects. The anomalous Hall effect has been
recently reviewed by Nagaosa et al. [2], on which we base part of our chapter.
Within the spin Hall effect physics, several recent reviews have appeared,
such as the one by Schliemann [3], focusing on some aspect of the intrinsic
spin Hall effect [4,5], and Engel et al. [6] and Hankiewicz and Vignale [7],
focusing on the extrinsic spin Hall effect and quantum spin Hall effect. This
particular field is evolving rapidly, generating many new fields of interest
such as the spin Hall insulators following the studies of the quantum spin
Hall effect [8–10], and extending even to spin thermodynamic Hall effects.
nomalous Hall Effect

8.2.1 A
In 1879, Edwin Hall ran a current through a gold foil and discovered that
a transverse voltage was induced when the film was exposed to a perpen-
dicular magnetic field. The ratio of this Hall voltage to the current is the
Hall resistivity. For paramagnetic materials, the Hall resistivity is propor-
tional to the applied magnetic field, and Hall measurements can be shown
AHE SIHE
Mz Mz = 0
Magnetic Nonmagnetic
Majority
Fso Fso
Fso Fso
I I
Minority
V V
SHE SHE–1
Mz = 0 Mz = 0
Fso Fso
Fso Fso
I Ispin
I=0
V=0 V
Nonmagnetic Nonmagnetic
optical detection
FIGURE 8.1 Schematic of the family of spin-dependent Hall effects: Anomalous

Hall effect (AHE), spin-injection Hall effect (SIHE), spin Hall effect (SHE), and inverse
spin Hall effect (SHE−1). A common thread is their existence at zero external
magnetic field (i.e. no ordinary Hall effect present), with the key difference on the
nature of the currents involved and the method of detection. Within the spin-
injection Hall effect, the Hall response is overlaid on a new type of spin excitation
termed persistent spin-helix.
to be equivalent to a measurement of the concentration of free carriers

and determination whether they are holes or electrons. Magnetic films
were found to exhibit both this ordinary Hall response and an extraordi-
nary or anomalous Hall effect response that may persist at zero magnetic
fields, which is proportional to the internal magnetization perpendicular
to the plane:
RHall = Ro H + Rs M z , (8.1)
where
RHall is the Hall resistance
Ro and Rs are the ordinary and anomalous Hall coefficients
Mz is the magnetization perpendicular to the plane
H is the applied magnetic field
Based on this phenomenology, the anomalous Hall effect has been an

important tool to measure electrically the magnetization in many systems.
The key player in this effect is the SO coupling, without which one would
not observe the effect in a homogeneously magnetized system. The role of
the broken time-reversal symmetry is mainly to create an asymmetric spin
population that leads to a charge Hall voltage, but would not produce a mea-
surable anomalous Hall signal without the SO coupling.
The field of the anomalous Hall effect and its recent progress have been
recently reviewed by Nagaosa et al. [2], focusing particularly on the new
understanding of the underlying physics based on the topological nature of
the effect. Let us first embark on a qualitative description of the different
mechanisms.
8.2.2 Mechanism of the Spin-Dependent Hall Effects

The usual descriptions of the mechanisms of the anomalous Hall effect are
shown in Figure 8.2. Many of these, at least as explained in many texts, are
based on semiclassical descriptions of the different scattering and deflec-
tion mechanisms. Although we are going to retain the historical labeling
of each contribution, in what follows, we recast their definition in a more
physical and phenomenological way, anticipating their correct connection
to the actual microscopic physics and to a precise experimental definition
of each.
A natural classification of contributions to the anomalous Hall effect
and other spin-dependent Hall effects, which is guided by experiment and
by the microscopic theory of metals, is to separate them according to their
dependence on the Bloch state transport lifetime τ, which is equivalent to
a 1/kF l expansion, where kF is the Fermi wave-vector and l is the mean free
d r Electrons have an anomalous velocity perpendicular to

∂E
= + e E × bn the electric field related to their Berry’s phase curvature
dt ћ∂k ћ
(a) Intrinsic deflection
(b) Side-jump
(c) Skew-scattering
FIGURE 8.2 Schematic description of the mechanisms of the anomalous Hall

effect. (a) Intrinsic deflection arises from interband coherence induced by an exter-
nal electric field, giving rise to a velocity contribution perpendicular to the field
direction. (b) Side-jump arises from the fact that electron velocity is deflected in
opposite directions by the opposite electric fields experienced upon approaching
and leaving an impurity. (c) Skew-scattering arises from asymmetric scattering due
to the effective SO coupling of the electron or the impurity.
path. It is relatively easy to identify contributions to the anomalous Hall con-

ductivity, σ xy
AH
, which vary as τ1 and as τ0. In experiment, a similar separation
can sometimes be achieved by plotting σ xy AH
versus the longitudinal conduc-
tivity σxx ∝ τ, when τ is varied by altering disorder or varying temperature.
More commonly (and equivalently), the Hall resistivity is separated into
contributions proportional to ρxx and ρxx2 . Since in almost all cases σxx ≫ σ xy
AH
,
it means that an anomalous Hall resistivity, ρxyAH = σ AH (
yx
2
σ2xx + σ xy
AH
)
, propor-
tional to ρxx implies a dominating mechanism that is scattering independent,
2
that is, σ xy AH
≈ ρyx /ρxx ~ σ0xx , and an anomalous Hall resistivity proportional
to ρxx implies a dominating mechanism that is very dependent on scatter-
ing, that is, σ xy AH
≈ ρyxAH/ρxx ~ σ1xx . The illustration of this scaling is shown in
Figure 8.3a and b. This already highlights the counterintuitive nature of the
anomalous Hall effect, which implies that a “dirtier” system will be domi-
nated by scattering-independent mechanisms whereas a “cleaner” system
will be dominated by impurity-scattering mechanisms.
This partitioning seemingly gives only two contributions to σ xy AH
, one
∼τ and the other ∼τ°. The first contribution is known as the skew-scattering
contribution, σ xy AH -skew
. Note that in this parsing of anomalous Hall effect con-
tributions, it is the dependence on τ (or σxx) that defines it, not a particu-
lar mechanism linked to a microscopic or semiclassical theory. The second
contribution proportional to τ° (or independent of σxx) is further separated
into two parts: intrinsic and side-jump. It has long been believed that this
partitioning is only a theoretical exercise. However, although these two
contributions cannot be separated experimentally by dc measurements,
they can be separated experimentally by defining the intrinsic contribution,
σ xy
AH -int
, as the extrapolation of the ac-interband Hall conductivity to zero fre-
quency in the limit of τ → ∞, with 1/τ → 0 faster than ω → 0. This then leaves
a unique definition for the third and last contribution, termed side-jump,
as σ xyAH -sj
≡ σ xy
AH
− σ xy
AH -skew
− σ xy
AH -int
. This possible experimental extraction of
σ xy and σ xy
AH -sj AH -int
is illustrated in Figure 8.3c.
AH-int
σxy
AH
AH
σxy σxx ρxy ρxx
)
AH)
AH
AH AH 2 AH
0 σxy (ω)
Log (σxy
AH-sj AH
ρxy ρxx (0)–σ AH-int
Log (ρxy
σxy σxx σxy = σxy xy
(a) Log (σxx) (b) Log (ρxx) (c) ω
Material with dominant scattering-independent mechanism

Material with dominant skew scattering mechanism
FIGURE 8.3 Scaling of anomalous Hall conductivity (a) and resistivity (b) versus diagonal conductivity and
resistivity, respectively, for a material whose anomalous Hall effect is dominated by skew scattering (dot-
dashes) and for a material dominated by scattering-independent mechanisms. (c) Illustration of the scaling of
ac-anomalous Hall conductivity for samples with decreasing impurity concentration (increasing τ) to deter-
mine σ AH-int
xy and, from this, obtain σ AH-sj
xy .
We examine these three contributions in the following. It is important

to note that the above definitions have not relied on identifications of semi-
classical processes such as side-jump scattering [11] or skew scattering from
asymmetric contributions to the semiclassical scattering rates [12] identi-
fied in earlier theories. Not surprisingly, the contributions defined earlier
contain these semiclassical processes. However, it is now understood that
other contributions are present in the fully generalized semiclassical theory,
which were not precisely identified previously, and are necessary to be fully
consistent with microscopic theories [2, 13, 14].
8.2.2.1 Intrinsic Contribution, s AH-int

xy
We have defined the intrinsic contribution microscopically as the dc limit

of the interband ac conductivity, a quantity that is not zero in ferromag-
nets when the SO-coupling interactions are included. There is, however, a
direct link to semiclassical theory in which the induced interband coherence
is captured by a momentum space Berry phase–related contribution to the
anomalous velocity. We show this equivalence in the following.
Historically, this contribution to the anomalous Hall effect was first
derived by Karplus and Luttinger [15] but its topological nature was not fully
appreciated until recently [16–18]. The concept of intrinsic anomalous Hall
effect was emphasized by Jungwirth et al. [17], motivated by the experimen-
tal importance of the anomalous Hall effect in ferromagnetic semiconduc-
tors and also by the earlier analysis of the relationship between momentum
space Berry phases and anomalous velocities in semiclassical transport the-
ory by Niu and coworkers [18, 19]. The frequency-dependent interband Hall
conductivity, which reduces to the intrinsic anomalous Hall conductivity in
the dc limit, had been evaluated earlier for a number of materials by Mainkar
et al. [20] and by Guo and Ebert [21].
The intrinsic contribution to the conductivity is dependent only
on the band structure of the perfect crystal. It can be calculated directly
from the simple Kubo formula for the Hall conductivity for an ideal lattice,
given the eigenstates |n, k〉 and eigenvalues εn(k) of a Bloch Hamiltonian H:
∑ ∫ (2π) [ f (ε (k)) − f (ε
dk
σijAH -int = e 2 d n n′ (k ))]
n ≠ n′
(8.2)
Im  n, k vi (k ) n′, k n′, k v j (k ) n, k 
×  ,
(εn (k ) − εn′ (k ))2
and the velocity operator is defined by
1 1
v(k ) = [r , H (k )] = ∇k H (k ), (8.3)
i 
where H is the k-dependent Hamiltonian for the periodic part of the Bloch
functions. Note the restriction n ≠ n′ in Equation 8.2.
What makes this contribution quite unique is that it is proportional to
the integration over the Fermi sea of the Berry’s curvature of each occupied
band, or equivalently [22, 23] to the integral of Berry phases over cuts of
Fermi surface segments. This result can be derived by noting that
n, k Ñk H (k ) n′, k
n, k Ñk n′, k = . (8.4)
εn′ (k ) − εn (k )
Using this expression, Equation 8.2 reduces to
e2
∑ ∫ (2π)
dk
σijAH -int = − εijl d
f (εn (k ))bnl (k ), (8.5)
 n
where:
εijl is the antisymmetric tensor
an(k) is the Berry-phase connection an(k) = i〈n,k|∇k|n,k〉,
bn(k) is the Berry-phase curvature
bn (k ) = Ñk × an (k ) (8.6)
corresponding to the states {|n,k〉}.

This same linear response contribution to the anomalous Hall effect
conductivity can be obtained from the semiclassical theory of wave-packet
dynamics [18, 19, 24]. It can be shown that the wave-packet group veloc-
ity has an additional contribution in the presence of an electric field:
rc = ∂En (k )/∂k − (E / ) × bn (k ). The intrinsic Hall conductivity formula,
Equation 8.5, is obtained simply by summing the second (anomalous) term
over all occupied states.
One of the motivations for identifying the intrinsic contribution σ xyAH -int
is that it can be evaluated accurately, even for relatively complex materials,

using first-principles electronic structure theory techniques. In many mate-
rials that have strongly SO-coupled bands, the intrinsic contribution seems
to dominate the anomalous Hall effect.
8.2.2.2 Skew-Scattering Contribution, s AH-skew

xy
This contribution is simply the component of σ xy AH

proportional to the Bloch
state transport lifetime. It therefore tends to dominate in nearly perfect crys-
tals. It is the only contribution to the anomalous Hall effect that appears
within the confines of traditional Boltzmann transport theory in which
interband coherence effects are completely neglected. Skew-scattering is due
to chiral features that appear in the disorder scattering of SO-coupled fer-
romagnets [12, 25].
Treatments of semiclassical Boltzmann transport theory found in
textbooks often appeal to the principle of detailed balance, which states
that the transition probability Wn→m from n to m is identical to the tran-
sition probability in the opposite direction (Wm→n). Although these two
transition probabilities are identical in a Fermi’s golden rule approxima-
2
tion, since Wn→n′ = (2π/ ) n V n′ δ(En − En′ ), where V is the perturbation
inducing the transition, detailed balance in this microscopic sense is not
generic. In the presence of SO coupling, either in the Hamiltonian of the

perfect crystal or in the disorder Hamiltonian, a transition that is right-
handed with respect to the magnetization direction has a different transi-
tion probability than the corresponding left-handed transition. When the
transition rates are evaluated perturbatively, asymmetric chiral contribu-
tions appear first at third order. In simple models, the asymmetric chiral
contribution to the transition probability is often assumed to have the
form [12, 25]
WkkA ′ = − τ −1
A k × k ′ ⋅ M s . (8.7)
When this asymmetry is inserted into the Boltzmann equation, it leads

to a current proportional to the longitudinal current driven by E and per-
pendicular to both E and Ms. When this mechanism dominates, both the
anomalous Hall conductivity σ xy AH
and the conductivity σ xx are approxi-
mately proportional to the transport lifetime τ, and the anomalous Hall
resistivity ρAH -skew = σ AH -skewρxx2 is therefore proportional to the longitudinal
resistivity ρ xx.
There are several specific mechanisms for skew scattering. Evaluation
of the skew-scattering contribution to the Hall conductivity or resistivity
requires simply that the conventional linearized Boltzmann equation be
solved using a collision term with accurate transition probabilities, since
these will generically include a chiral contribution. We emphasize that skew-
scattering contributions to σ xy AH
are present not only because of SO coupling
in the disorder Hamiltonian, but also because of SO coupling in the perfect
crystal Hamiltonian combined with purely scalar disorder. Either source of
skew scattering could dominate σ xy AH -skew
, depending on the host material and
also on the type of impurities, and, as we will see, it dominates the physics of
the present experiments in spin-injection Hall effect.
One important thing to note is that we have not defined the skew-
scattering contribution to the anomalous Hall effect as the sum of all the
contributions arising from the asymmetric scattering rate present in the
collision term of the Boltzmann transport equation. We know from micro-
scopic theory that this asymmetry also makes an anomalous Hall effect con-
tribution of order τ°. There exists a contribution from this asymmetry, which
is actually present in the microscopic theory treatment associated with the
so-called ladder diagram corrections to the conductivity, and therefore of
order τ°. In our experimentally practical parsing of anomalous Hall effect
contributions, we do not associate this contribution with skew scattering but
place it under the umbrella of side-jump scattering, even though it does not
physically originate from any sidestep type of scattering.
AH - sj
8.2.2.3 Side-Jump Contribution s xy
Given the sharp definition we have provided for the intrinsic and skew-scat-
tering contributions to the anomalous Hall effect conductivity, the equation
σ xy
AH
= σ xy
AH -int
+ σ xy
AH -skew
+ σ xy
AH -sj
(8.8)
defines the side-jump contribution as the difference between the full Hall
conductivity and the two simpler contributions. In using the term side-
jump for the remaining contribution, we are appealing to the historically
established taxonomy. Establishing this connection mathematically has
been the most controversial aspect of anomalous Hall effect theory, and
the one that has taken the longest to clarify from a theory point of view.
Although this classification of Hall conductivity contributions is often use-
ful, it is not generically true that the only correction to the intrinsic and
skew contributions can be physically identified with the side-jump process
defined as in the earlier studies of the anomalous Hall effect [26].
The basic semiclassical argument for a side-jump contribution can be
stated straightforwardly: when considering the scattering of a Gaussian
wave-packet from a spherical impurity with SO interaction (HSO = (1/2m2c2)
(r−1∂V/∂r)SzLz), a wave-packet with incident wave-vector k will suffer a dis-
placement transverse to k equal to (1/6)kħ2/m2c2. This type of contribution
was first noticed, but discarded, by Smit [25] and reintroduced by Berger
[26], who argued that it was the key contribution to the anomalous Hall
effect. This kind of mechanism clearly lies outside the bounds of traditional
Boltzmann transport theory in which only the probabilities of transitions
between Bloch states appear, and not microscopic details of the scattering
processes. This contribution to the conductivity ends up being independent
of τ and therefore contributes to the anomalous Hall effect at the same order
as the intrinsic contribution in an expansion in powers of scattering rate. The
separation between intrinsic and side-jump contributions, which cannot be
distinguished by their dependence on τ, has been perhaps the most argued
aspect of anomalous Hall effect theory [2].
Side-jump and intrinsic contributions have different dependences on
more specific system parameters, especially in systems with complex band
structures [14]. Some of the initial controversy that surrounded side-jump
theories was associated with the physical meaning ascribed to quantities
that were plainly gauge dependent, like the Berry’s connection, which in
early theories is typically identified as the definition of the side step upon
scattering. Studies of simple models, for example, models of semiconduc-
tor conduction bands, also gave results in which the side-jump contribu-
tion seemed to be the same size, but opposite in sign compared to the
intrinsic contribution [27]. We now understand [13] that these cancel-
ations are unlikely, except in models with a very simple band structure,
for example, one with a constant Berry’s curvature. It is only by a careful
comparison between fully microscopic linear response calculations, such
as Keldysh (nonequilibrium Green’s function) or Kubo formalisms, and
the systematically developed semiclassical theory that the specific contri-
bution due to the side-jump mechanism can be separately identified with
confidence [13].
A practical approach, which is followed at present for materials in which
σ xy
AH
seems to be independent of σxx, is to first calculate the intrinsic contri-
bution to the anomalous Hall effect. If this explains the observation (and it
appears that it usually does), then it is deemed that the intrinsic mechanism
dominates. If not, we can take some comfort from understanding on the basis
of simple model results, that there can be other contributions to σ xy
AH
, which
are also independent of σxx and can for the most part be identified with the
side-jump mechanism. It seems extremely challenging, if not impossible,
to develop a predictive theory for these contributions, partly because they
require many higher-order terms in the perturbation theory that must be
summed, but more fundamentally because they depend sensitively on details
of the disorder in a particular material, which are normally unknown and dif-
ficult to model accurately.
8.2.3 Recent Developments on the

Anomalous Hall Effect
The earlier description of the different mechanisms of the anomalous Hall
effect in ferromagnetic metals has been refined over the past few years, with
some of the recent research focused on unifying the different theories that
disagreed. We finalize this section by pointing out some of the key recent
developments, which are important to know when discussing the anomalous
Hall effect:
1. When the anomalous Hall conductivity, σ xy AH

, is independent of σ xx,
the anomalous Hall effect can often be understood in terms of the
geometric concepts of Berry phase and Berry curvature in momen-
tum space. This anomalous Hall effect mechanism is responsible
for the intrinsic anomalous Hall effect (see following text). In this
regime, the anomalous Hall current can be thought of as the unquan-
tized version of the quantum Hall effect.
2. Three broad regimes have been identified when surveying a large
body of experimental data for diverse materials [2]: (i) A high-
conductivity regime (σ xx > 10 6 (Ω cm)−1) in which a linear contri-
bution to σ xyAH
∼ σ xx due to skew scattering dominates σ xy AH
. In this
regime, the normal Hall conductivity contribution can be signifi-
cant and even dominate σ xy. (ii) An intrinsic or scattering-inde-
pendent regime in which σ xy AH
is roughly independent of σ xx (10 4
(Ω cm) < σ xx < 10 (Ω cm) ). (iii) A bad-metal or insulating
−1 6 −1
regime (σ xx < 10 4 (Ω cm)−1) in which σ xy AH

decreases with decreas-
ing σ xx at a rate faster than linear. This regime remains still a chal-
lenge, theoretically.
3. The band structure of the ferromagnetic materials plays a key role in
these systems. In particular, the band (anti-)crossings near the Fermi
energy has been identified using first-principles Berry curvature cal-
culations as a mechanism that can lead to a large intrinsic anomalous
Hall effect. The effect of scattering on these crossings is still not well
understood.
4. A semiclassical treatment by a generalized Boltzmann equation tak-
ing into account the Berry curvature and coherent interband mix-
ing effects due to band structure and disorder has been formulated.
This theory provides a clearer physical picture of the anomalous Hall

effect than early theories by identifying correctly all the semiclas-
sically defined mechanisms [13, 14]. This generalized semiclassical
picture has been verified by comparison with controlled microscopic
linear response treatments for identical models.
8.2.4 Spin Hall Effect

A natural progression of the physics of the anomalous Hall effect to systems
with time-reversal symmetry and with SO coupling is the spin Hall effect.
In describing the different mechanisms giving rise to anomalous Hall effect,
the exchange field is only important in creating a spin population imbalance
that will give rise to a finite voltage that one can measure. Hence, since the
mechanisms that give rise to the anomalous Hall effect are also present in
these systems, the “spin” separation will supposedly take place, but it will
not be directly measurable through a voltage perpendicular to the flow of
the current in the sample, since there is no exchange field that induces the
population asymmetry between spin-up and spin-down states. However,
consistent with the global time-reversal symmetry, a spin current will be
generated perpendicular to an applied charge current, which in turn should
lead to spin accumulations with opposite magnetization at the edges. This is
the spin Hall effect first predicted over three decades ago by simply invoking
the phenomenology of the anomalous Hall effect in ferromagnets and apply-
ing it to semiconductors [28].
Motivated by studies of intrinsic anomalous Hall effect in strongly
SO-coupled ferromagnetic materials, where scattering contributions do
not seem to dominate, the possibility of an intrinsic spin Hall effect was put
forward by Murakami et al. [4] and by Sinova et al. [5], with scattering play-
ing a minor role. This proposal generated an extensive theoretical debate
motivated by its potential as a spin-injection tool in low dissipative devices.
The interest in the spin Hall effect has also been dramatically enhanced by
experiments by two groups reporting the first observations of the spin Hall
effect in n-doped semiconductors [29, 30] and in 2D hole gases (2DHG)
[31]. These observations, based on optical experiments, have been followed
by metal-based experiments in which both the spin Hall effect and the
inverse spin Hall effect have been observed [32–36], and other experiments
in semiconductors at other regimes, even at room temperature [37–40].
In the case of the inverse spin Hall effect, it is a spin current that gener-
ates a charge current in the so-called non-local measurements. These spin
currents can be utilized in the spin-torque experiments and are being very
actively studied. Given the extent of detail and still unsettled debate in this
field, and the several reviews already mentioned in the introduction, we will
not go further into the experiment and theory of spin Hall effect and inverse
spin Hall effect. Instead, we will focus on the new phenomenology of the
spin-injection Hall effect that combines several effects, is at present well
understood in the diffusive regime, and is perhaps the most promising from
a technology standpoint.
pin-Injection Hall Effect

8.2.5 S
The spin-injection Hall effect was observed in a photovoltaic cell device
that opens direct application potential in opto-spintronics, for example,
as a solid-state polarimeter. Importantly, the spin-injection Hall effect can
be also used to measure spin dynamics of the electrons propagating in
the semiconducting channel, in particular the spin-helix effect. The spin-
injection Hall effect can, therefore, be used as an efficient spin-detection tool
for basic studies of SO coupling phenomena in semiconductors and can also
be directly incorporated in a number of spintronic devices, including the
spin-field-effect transistor of the Datta–Das type [63].
The electrical detection of spin-polarized transport in semiconductors
is a key requisite for the possible incorporation of spin in semiconductor
microelectronics. In the spin-injection Hall effect, a polarized current is
injected (optically in the present case) and its polarization can be detected by
transverse electrical signals, via the anomalous Hall effect, directly along the
semiconducting channel. One key practical feature is that this is achieved
without disturbing the spin-polarized current or using magnetic elements,
thus opening the door for possible spin-based sequential logic schemes in
logic circuits.
Spin-polarized transport phenomena in semiconductors have been
studied by a range of techniques used in ferromagnets and novel approaches
developed specifically for semiconductor spintronics. The techniques
include magneto-optical scanning probes [29, 37, 41–43], spin-polarized
electroluminescence [31, 44–47], and magnetoelectric measurements uti-
lizing spin-valve effects, magnetization dependence of nonequilibrium
chemical potentials, and SO coupling phenomena [32, 42, 43, 48–54]. Two
of these spintronic research developments have been particularly impor-
tant for the observation of the spin-injection Hall effect. First, it has been
demonstrated that electrons carrying electrical current in a ferromagnet
align their spins with the local direction of magnetization, and that the
resulting electrical signals due to the anomalous Hall effect can be used
as an alternative to magneto-optical scanning probes to measure the local
magnetization [55, 56]. The second is the observation of the spin Hall effect
that verified that the physics of the anomalous Hall effect translate directly
to semiconducting systems without broken time-reversal symmetry, that
is, the spin Hall effect. Combining these two facts, one can surmise that
injecting spin-polarized electrical currents into nonmagnetic semiconduc-
tors should also generate a Hall effect, which, as long as the spins of the
charge carriers remain coherent, yields transverse charge accumulation
and is therefore detectable electrically. In Wunderlich et al. [1], the device
diode is operated in the reverse regime as a photocell in order to inject
spin-polarized electrical currents into the semiconductor from a spatially
confined region.
The effect, as discussed in the theory section, is well understood by
microscopic calculations that reflect spin dynamics induced by an internal
SO field. The spin-injection Hall effect is observed up to high temperatures
σ+ – + –
2DHG
2DEG
+
– +
FIGURE 8.4 Sketch of the spin Hall effect device. Polarized light injects spin-
polarized electron–hole pairs. The coplanar p–n junction [31] is connected in
reverse bias, and the electrons, moving down the semiconductor channel in the
(110) direction, experience a precessing motion in their spin, which is recorded via
Hall probes. (After Wunderlich, J. et al., Nat. Phys. 5, 675, 2009. With permission.)
and the devices can also be thought of as a realization of a nonmagnetic spin-

photovoltaic polarimeter, which directly converts polarization of light into
transverse voltage signals (Figure 8.4).
8.3 SPIN-INJECTION HALL EFFECT: EXPERIMENT

In this section, we first describe the experimental findings of Wunderlich
et al. [1] and follow it up with the theoretical analysis in the subsequent and
final section of this chapter.
Figure 8.5a shows lateral micrographs of the planar 2D electron-hole
gas (2DEG-2DHG) photodiodes with the p-region and n-region patterned
into unmasked 1 µm wide Hall bars (similar experiments were performed
in masked samples) in the (110) direction. The effective width of each Hall
contact is 50–100 nm and separation between two Hall crosses is 2 µm. The
sketch of the device and its experimental setup is also illustrated in Figure 8.1.
As in the spin Hall effect experiments [31], the semiconductor wafer con-
sists of a modulation p-doped AlGaAs/GaAs heterojunction on top of the
structure separated by 90 nm of intrinsic GaAs from an n-doped AlGaAs/
GaAs heterojunction underneath. Without etching, the top heterojunction
in the wafer is populated by the 2DHG while the 2DEG at the bottom het-
erojunction is depleted. The n-side of the coplanar p–n junction is formed
by removing the p-doped surface layer from a part of the wafer through
wet etching. The removal of this top heterojunction results in populating
the 2DEG. At zero or reverse bias, the device is not conductive in dark, due
to charge depletion of the p–n junction. Counterpropagating electron and
hole currents can be generated by illumination at sub-gap wavelengths.
Because of the optical selection rules, the spin-polarization of injected
electrons and holes is determined by the sense and degree of the circular
polarization of vertically incident light. The optical spin-injection area is
controlled by bias-dependent p–n junction depletion and, additionally, by
50
25
0
H0 H1 H2 –25
Vb = 0 V
–50
RH2 (Ω)
50
25
5.5 µm 0
–25 Vb = –10 V
(a)
–50
Vb
20
RL (kΩ)
h h
h H0
h h
h H1
e VL 10
e
e
e
e H2 σ– σ0 σ+
e
VH2
H2
2DHG 0
2DEG 0 25 50 75
(b) (c) tm (s)
FIGURE 8.5 Devices, schematics of the experiment, and basic observation of

the spin-injection Hall effect. (a) Micrograph of the coplanar p–n junction device.
(b) Schematic diagram of the wafer structure of the 2DEG-2DHG p–n diode and of
the spin-injection Hall effect measurement setup. (c) Steady-state spin-injection
Hall effect signals changing sign for opposite helicities (σ− and σ+) of the incident
light beam, that is, for opposite spin-polarizations of the injected electrons. RH2
is the second Hall bar resistance and RL is the longitudinal resistance. For linearly
polarized light (σ0), the injected electron current is spin-unpolarized and the spin-
injection Hall effect vanishes. Measurements were performed at the 2DEG Hall
cross H2 at a temperature of 4 K. (After Wunderlich, J. et al., Nat. Phys. 5, 675, 2009.
With permission.)
the position and focus of the laser spot, or by including metallic masks on
top of the Hall bars [1].
Figure 8.5c shows the measurements at the Hall cross bar H2 in the
n-channel at 4 K, laser wavelength of 850 nm, and for 0 and −10 V reverse bias
with the laser spot fixed on top of the p–n junction and focused on approxi-
mately 1 µm in diameter. While the longitudinal resistance RL is insensitive
to the polarization, the transverse signal RH is observed only for polarized
spin-injection into the electron channel and reverses sign upon reversing the
polarization. With the laser spot focused on the p–n junction, the transverse
voltage is only weakly bias dependent. Note that the large signals of tens
of microvolts, corresponding to transverse resistances of tens of ohms, are
detected outside the spin-injection area at a Hall cross separated by 3.5 µm
from the p–n junction.
Figure 8.6a shows the simultaneous electrical measurements at Hall
cross H0, which is wider and partially overlaps with the injection area, and at
remote Hall crosses H1, H2, and H3. To confirm that the transverse signals
10 2
H0 (×3)
H2 H1
5 1
αH (10–3)
αH (10–3)
0 0
H1 (×3)
H3 (×3)
–5 –1
σ– σ+
–10 –2
0 10 20 30 40 50 –1.0 –0.5 0.0 0.5 1.0
(a) tm (s) (b) σ +− σ–
FIGURE 8.6 (a) Spin-injection Hall effect Hall probe signals are plotted for Hall
crosses H0H3 in the 2DEG channel. Gray region corresponds to the manual rota-
tion of the λ/2 wave plate, which changes the helicity of the incident light and,
therefore, the spin-polarization of injected electrons. (b) Hall angles measured
at n-channel Hall crosses H1 (similar results with opposite sign occur for H2). The
spin-injection Hall effect angles are linear in the degree of polarization, hence
behaving phenomenologically, as in the anomalous Hall effect. The laser spot is
focused on the p–n junction and the bias voltage, laser wavelength, and measure-
ment temperature are also the same as in Figure 8.5. (After Wunderlich, J. et al.,
Nat. Phys. 5, 675, 2009. With permission.)
do not result from spurious effects but originate from the polarized spin-
injection, the helicity of the incident light is reversed by manually rotating a
λ/2 wave plate by 45°. The signals are recalculated to Hall angles, αH = RH/RL,
whose magnitude 10−3–10−2 is comparable to the anomalous Hall effect in
conventional metal ferromagnets. The spatial variation of the sign of the Hall
signals observed suggests a modulated spin dynamics in the length scale of
microns. The scattering mean free path in this system is on the order of ∼10–
100 nm and the typical SO-coupling length to ∼1 µm. The first length scale
determines the onset of the anomalous Hall effect–like transverse charge
accumulation due to skew scattering off SO-coupled impurity potentials
(see theory section). The second length scale governs the spin precession
about the internal SO field, which in this configuration tends to point in-
plane and perpendicular to the electron momentum. In quasi 1D channels,
electrons injected with out-of-plane spins would precess coherently about
this internal SO field, but in wider channels (diffusive transport), the spin-
coherence length is expected to be much shorter. As we will see in the next
section, the coherence can also exceed micrometer scales in wider channels
due to the additional Dresselhaus SO field originating from inversion asym-
metry of GaAs. In Figure 8.6b, the spin-injection Hall effect as a function of
the degree of polarization at Hall cross H1 is shown. The linear dependence
of the Hall angle on the polarization of injected electrons is analogous to
the linear dependence on magnetization of the anomalous Hall effect in a
ferromagnet.
Within the experiment, other important checks on the Hall symme-
tries were made in order to discard external effects for the observed phe-
nomenology. Masked samples (only injection area open) show the same
Vb
Vb
VH2
VH2
VH3 VH3
20 20
σ0 σ– σ– σ0 0 σ– σ– σ0
10 10 σ
VH (µV)
0 VH (µV) 0
–10 H2 –10 H2
H3 H3
–20 –20
4 4
2 2
I (µA)
I (µA)
0 0
Vb = –5 V Vb = +5 V Vb = –5 V Vb = +5 V
–2 –2
–4 –4
0 20 40 60 80 0 20 40 60 80
(a) tm (s) (b) tm (s)
FIGURE 8.7 (a) Spin-injection Hall effect measurements in a masked sample

with linearly polarized light and circularly polarized light of a fixed helicity for
opposite polarities of the optical current. Schematics of each experimental setup
are shown in the upper panel. Middle panel shows that spin-injection Hall effect
voltages are detected only at negative bias when spin-polarized electrons move
from the illuminated aperture toward the measured Hall crosses H2 and H3 in the
n-channel. The optical current is plotted in the lower panel. (b) Complementary
measurements to (a) in an unmasked sample with only the n-channel biased and
Hall crosses H2 and H3 directly illuminated with a 10× stronger light intensity as
compared to (a). Weak anomalous Hall effect signals are detected in this case,
which are antisymmetric with respect to the polarity of the current. Lower panel
shows the optically generated part of the current. (After Wunderlich, J. et al., Nat.
phenomenology (shown in Figure 8.7a as the unmasked ones. Hence, the

possibility that the observed effects arises from stray light from the focused
polarized beam seems unlikely. Complete Hall symmetries of the transverse
signals measured in our devices are confirmed by the experiments presented
in Figure 8.7. Figure 8.7a shows data recorded in a sample with Hall crosses
H2 and H3 fully covered by an insulating thin film and a metallic mask.
The spin-injection Hall effect is observed only at reverse bias when electrons
move from the illuminated aperture toward the n-channel Hall crosses. At
forward bias, optically generated electrons are accelerated in the opposite
direction and no Hall signals are detected independent of polarization.
The distinct features of the spin-injection Hall effect are highlighted
by comparison to complementary data shown in Figure 8.7b. After etching
the surface layers in the n-side of the p–n diode, Wunderlich et al. [1] were
able to select wafers with residual sub-gap optical activity in an unmasked

n-channel. The dark current generates zero Hall voltage at crosses H2 and
H3 while clear (albeit weak) Hall signals are detected upon directly illumi-
nating the crosses with intense circularly polarized light. The signals are
attributed to the anomalous Hall effect since they occur inside the spin-
generation area and, as expected, they are antisymmetric with respect to
the current polarity. Also consistent with the anomalous Hall effect, they
observe in these experiments Hall voltages with polarity depending only
on current orientation and spin-polarization, but with the same signs on all
irradiated Hall crosses measured. It contrasts the spin-injection Hall effect
measurements of spin-injection from the p–n junction in which the sign can
alternate among the Hall crosses.
It is also interesting that when measuring within the p-region, con-
sistent with expectations, no clear Hall signal was measured at the first
p-channel Hall cross with the laser spot focused on top of the p–n junc-
tion when the measurements were taken simultaneously with the n-channel
measurements. This confirms that it is the spin-decoherence of propagating
holes rather than an inherent absence of the Hall effect in the 2DHG, which
explains the negative result of measurements.
Wunderlich et al. [1] also found that the measurements in a sample
showed rectifying p–n junction characteristics at temperatures up to 240 K,
hence demonstrating that the spin-injection Hall effect is readily detect-
able at high temperatures. Together with the zero-bias operation shown in
Figure 8.5c and linearity of the spin-injection Hall effect in the degree of
circular polarization of the incident light, these characteristics represent
the realization of the spin-photovoltaic effect in a nonmagnetic struc-
ture and demonstrate the utility of the device as an electrical polarimeter
[52, 57, 58].
8.4 THEORY OF SPIN-INJECTION HALL EFFECT

The previous section was limited to presenting the experimental phenome-
nology of the spin-injection Hall effect. As may have become apparent to the
reader, it is rather surprising that such large spin-coherence lengths are also
associated with an apparent coherent precession across enormous length
scales beyond what is expected in normal 2DEG. On the other hand, because
of the particular parameters chosen in the experiments, this is a system that
we can model rather precisely and, as we will see in the following text, it is
highly anisotropic in this coherent behavior.
The theoretical approach is based on the observation that the microm-
eter length scale governing the spatial dependence of the nonequilibrium
spin-polarization is much larger than the ∼10–100 nm mean free path in
our 2DEG, which governs the transport coefficients. This allows us to dis-
regard the Hall signal at first instance and first calculate the steady-state
spin-polarization profile along the channel, and only afterward consider the
spin-injection Hall effect as a response to the local out-of-plane component
of the polarization. Note that this would not be possible when these length
scales are of similar order, as they will be when mobilities increase and we
exit the diffusive regime.
Our starting point is the description of GaAs near the Γ point with the
effect of the valence bands taken into account through a two-band effective
Hamiltonian; this can be achieved through the so-called Lowin transfor-
mation. In the presence of an electric potential V(r), the corresponding 3D
SO-coupling Hamiltonian reads
H 3D -SO =  λ* σ ⋅ (k × ∇V (r ))
(8.9)
( )
+ kx k 2y − kz2 σ x + cyclic permutations  ,
 
where:
σ are the Pauli spin matrices
k is the momentum of the electron
 ≈ 10eV Å3
λ* = 5.3 Å2 for GaAs [59, 60]
Equation 8.9 together with the 2DEG confinement yields an effective 2D

Rashba and Dresselhaus SO-coupled Hamiltonian [61, 62],
 2k 2
H 2 DEG = + α(k y σ x − kx σ y ) + β(kx σ x − k y σ y )
2m (8.10)
+ Vdis (r ) + λ* σ ⋅ (k × ∇Vdis (r )),
where:
m = 0.067me
β = −  kz2 ≈ − 0.02 eV Å
α = eλ*Ez ≈ 0.01 − 0.03 eV Å for the strength of the confining electric field
eEz ≈ 2–5 × 10−3 eV/Å pointing along the [001] direction
The strength of the confinement is obtained from a self-consistent

Poisson–Schrödinger simulation of the conduction band profile of our
GaAs/AlGaAs heterostructure [31]. Typically α can be tuned whereas β is a
material-dependent parameter fixed by the choice of growth direction and,
to a lesser extent, the degree of confinement.
onequilibrium Polarization Dynamics

8.4.1 N
–
along the [110] Channel
The realization of the original Datta–Das device concept in a purely Rashba
SO-coupled system has been unsuccessful until recently due to spin-
coherence issues; that is, in the required length scales in which transport is
diffusive, no oscillating persistent precession states are present [63]. However,
in a 2DEG where both Rashba and Dresselhaus SO coupling have similar
strengths, a long-lived precessing excitation of the system has been shown to
exist along a particular direction [37, 62]. When α and β are equal in magni-
tude, the component of the spin along the [110] direction for α = −β or along
the [110] direction for α = β is a conserved quantity [61], as well as a precess-
ing spin-wave (spin-helix) of wavelength λspin-helix = πħ2/(2mα) in the direction
perpendicular to the conserved spin component [62]. This spin-helix state
has been observed through optical transient spin-grating experiments [37].
In the strong scattering regime of the structure considered, with αkF and
βkF ∼ 0.5 meV, much smaller than the disorder scattering rate ħ/τ ∼ 5 meV,
the system obeys a set of spin-charge diffusion equations [62] for arbitrary
ratio of α and β:
∂t n = D∇2n + B1∂ x + Sx − − B2∂ x − Sx + ,
∂t Sx + = D∇2∂t Sx + − B2∂ x −n − C1∂ x + Sz − T1Sx + ,
∂t Sx − = D∇2∂t Sx − + B1∂ x +n − C2∂ x − Sz − T2Sx − ,
∂t Sz = D∇2∂t Sz + C2∂ x − Sx − + C1∂ x + Sx + − (T1 + T2 )Sz ,
where:
x+ and x− correspond to the [110] and [110] directions
B1/2 = 2(α  β)2 (α ± β)kF2 τ 2
T1/2 = (2/m)(α ± β)2 (kF2 τ/ 2 )
D = v F2 τ/2
C12/2 = 4DT1/2
For the present device, where α ≈ −β, the 2DEG channel is patterned
along the [110] direction, which corresponds to the direction of the spin-helix
propagation. Within this direction, the dynamics of S x− and Sz couple through
the diffusion equations above. Seeking steady-state solutions of the form exp
[qx−] yields the transcendental equation (Dq 2 + T2 )(Dq 2 + T1 + T2 ) − C22q 2 = 0,
which can be reduced to q 4 + (Q12 − 2Q22 )q 2 + Q12Q22 + Q24 = 0 , where
Q1/2 ≡ T1/2 /D = 2m α ± β 2 . This yields a physical solution for q = |q| exp[iθ]
of
 2Q
12Q22 − Q 14/4 
( ) 1
1/ 4
12Q
q= Q 22 + Q
24 and θ = arctan   . (8.11)
 
 Q2 − Q1/2
2 2
2 
The resulting damped spin precession of the out-of-plane polarization com-

ponent for the parameter range of the device, where we have set β = −0.02 eV
Å and varied α from −0.5β to −1.5β, is shown in Figure 8.8 in the main text.
These results are in agreement with Monte Carlo calculations on similar sys-
tems (modeling an InAs 2DEG), but with higher applied biases [64]. In the
Monte Carlo calculations, longer decaying lengths were observed at higher
biases. However, the present device is well within the linear regime with very
low driving fields; this results in shorter decay length scales of the oscilla-
tions as compared to Ref. [64].
1.0
0.5
p 0.0
z
–0.5
–1.0
0
0.5
1
X (1
1.5
10
2
)
1.4
(µ
1.2
m
2.5 1
)
0.8
3 0.6 α/|β|
FIGURE 8.8 Steady-state solution of the spin precession for α/|β|=0.7, 1.0, 1.3.
The bottom panel indicates the precession of the polarization out-of-plane com-
ponent for the parameter range of our device, where we have β = −0.02 eV Å and
varied α from −0.5β to −1.5β.
These results show good agreement with the steady-state variations

(changes in the length scale of 1–2 µm) in the out-of-plane nonequilibrium
polarization of our system observed indirectly through the Hall signals. We
note that Monte Carlo simulations including temperature broadening of the
quasiparticle states confirm the validity of the analytical results up to the
high temperatures used in the experiment.
8.4.2 Hall Effect

The Hall effect signal can be understood within the theory of the anomalous
Hall effect. The contributions to the anomalous Hall effect in SO-coupled
systems with non-zero polarization can be classified into two types: the
first type arises from the SO-coupled quasiparticles interacting with the
spin-independent disorder and the electric field, and the second type arises
from the non-SO-coupled part of the quasiparticles scattering from the
SO-coupled disorder potential (last term in Equation 8.10). The contribu-
tions of the first type have recently been studied in 2DEG ferromagnetic
systems with Rashba SO coupling [65–68]. These have shown that, within
the regime applicable to our present devices, the anomalous Hall effect
contribution due to the intrinsic and side-jump mechanisms vanish in
the presence of even moderate disorder. In addition, the skew-scattering
contribution from this type of contribution is also small (with respect to
the contribution shown in the following) and furthermore is not linear in
polarization [65]. Hence, we can disregard the contributions of the first type
within our devices.
This is not surprising, since in our devices αkF, βkF ≪ ħ/τ, and hence these
contributions arising from the SO coupling of the bands are not expected to
dominate. Instead, the observed signal originates from contributions of the
second type, that is, from interactions with the SO-coupled part of the dis-
order [27, 69]. Within this type, one contribution is due to the anisotropic
scattering, the extrinsic skew scattering, and is obtained within the second
Born approximation treatment of the collision integral in the semiclassical
linear transport theory [27, 69]:
skew 2πe 2 λ*
σ xy = V0 τn(n↑ − n↓ ), (8.12)
2
where n = n↑ + n↓ is the density of photoexcited carriers with polarization
pz = (n↑ − n↓)/(n↑ + n↓). Using the relation for the mobility µ = eτ/m and the
relation between ni, V0, and τ,  /τ = niV02m/ 2 , the extrinsic skew-scattering
contribution to the Hall angle, αH ≡ ρxy/ρxx ≈ σxy/σxx, can be written as
e
a skew
H = 2pl* npz (x[1 10] )
nim
o
l *[A 2 ](n- - n¯ ) éë1011 cm -2 ùû
= 2.44 ´ 10-4 (8.13)
m éë103 cm 2 /Vs ùû ni éë1011 cm -2 ùû
~ 1.1 ´ 10 -3 pz ,
where we have used n = 2 × 1011 cm−2, pz is the polarization, µ = 3 × 103 cm2/V s,

and ni = 2 × 1011 cm−2; the last estimate is introduced to give a lower bound to
the Hall angle contribution within this model. In addition to this contribu-
tion, there also exists a side-jump scattering contribution in the SO-coupled
disorder term given by
2e 2 λ*
α sH− j = (n↑ − n↓ ) (8.14)
σ xx
−4
λ * Å2 
= 3.0 × 10 pz ~ 5.3 × 10 −4 pz . (8.15)
µ 103 cm2 /Vs 
As expected, this is an order of magnitude lower than the skew-scattering

contribution within this system.
We can then combine this result with the results from the previous sec-
tion to predict, in this diffusive regime, the resulting theoretical αH along the
[110] direction for the relevant range of Rashba and Dresselhaus parameters
corresponding to the experimental structure [1]. This is shown in Figure 8.9.
We have assumed a donor impurity density ni on the order of the equilib-
rium density n2DEG = 2.5 × 1011 cm−2 of the 2DEG in dark, which is an upper
bound for the strength of the impurity scattering in our modulation-doped
heterostructure and, therefore, a lower bound for the Hall angle. For the
mobility of the injected electrons in the 2DEG channel, one can consider
1.0
β = –0.02 Å eV 0.5
pz 0.0
–0.5
1.1
–1.0
0.5
0.0
αH (10–3)
–0.5
–1.1
0 1.4
0.5 1.2
1 1 |
X[
1.5
2 0.8 α/|β
110] (
µm) 2.5 0.6
3
FIGURE 8.9 Microscopic theory of the spin-injection Hall effect assuming

SO-coupled band structure parameters of the experimental 2DEG system. The
calculated spin-precession and spin-coherence lengths and the magnitude of
the Hall angles are consistent with experiment. The grayscale surface shows
the proportionality between the Hall angle and the out-of-plane component
of the spin-polarization. (After Wunderlich, J. et al., Nat. Phys. 5, 675, 2009. With
permission.)
the experimental value determined from ordinary Hall measurements with-

out illumination, µ = 3 × 103 cm2/V s. The density of photoexcited carriers
of n ≈ 2 × 1011 cm−2 was obtained from the measured longitudinal resistance
between successive Hall probes under illumination, assuming constant
mobility.
The theory results shown in Figure 8.9 provide a semiquantitative
account of the magnitude of the observed spin-injection Hall effect angle
(∼10−3), and explain the linear dependence of the spin-injection Hall effect on
the degree of spin-polarization of injected carriers. The calculations are also
consistent with the experimentally inferred precession length on the order of
a micrometer, and the spin coherence exceeding micrometer length scales.
We emphasize that the 2DEG in the strong disorder, weak SO-coupling
regime realized in our experimental structures is a particularly favorable
system for establishing theoretically the presence of the spin-injection Hall
effect. In this regime, and for the simple band structure of the archetypal
2DEG, the spin-diffusion equations and the leading skew-scattering mecha-
nism of the SO-coupling-induced Hall effect are well understood areas of the
physics of quantum-relativistic spin-charge dynamics.
ombined Spin–Orbit and Zeeman

8.4.3 C
Effects in External Magnetic Fields
The above-mentioned theoretical considerations help explain the observed
phenomenology and predict other expectations, such as higher damped
oscillations observed in the orthogonal direction for the given device.
Another aspect that will affect these oscillations is the application of an

external magnetic field whose consequences we consider next.
The spin dynamics at high temperatures and in external magnetic field
can be studied theoretically by Monte Carlo–Boltzmann simulations, assum-
ing the additional Zeeman term in the Hamiltonian, geµBB·σ/2, where ge is
the g-factor of GaAs. The orbital effects of the magnetic field are included on
a classical level by considering the Lorentz force acting on injected electrons
between scattering events. To model the device, the system is chosen to have
an L × W = 3 µm × 0.5 µm rectangular 2DEG connected to a source at x = 0 and
a drain at x = L. The x-axis is along the [110] direction and the y-axis is along the
[110] direction. The electron density is n2DEG = 2.5 × 1011 cm−2. Fermi energy is
EF = 9 eV and Fermi momentum is kF = 0.125 nm−1. The Dresselhaus parameter
β = −0.02 eV Å with the temperature of injected electrons is T = 300 K. Note
that temperature enters our calculations only through the Fermi distribution
of electrons; we do not include temperature-dependent scattering effects like,
for example, electron-phonon scattering. The results we obtain are only very
weakly temperature-dependent when comparing simulations at 1 and 300 K.
Electrons are injected along the [110] direction, with initial spin-polarization
along the [001] axis. In the simulations, only point-like impurities such that
the mean free path is Lmf = 40 nm were considered. The details of the Monte-
Carlo–Boltzmann calculations can be found in Ref. [1].
Figure 8.10 shows the calculated spin-polarizations in the channel in
magnetic fields oriented along the [001], [110], and [110] directions for the
Rashba coupling parameter α = −0.5β (the effects of the magnetic field for
α = −β are even less apparent than for the case of |α| ≠ |β| shown in the figure).
Since the calculations do not include quantum orbital effects of the magnetic
field, their validity is limited to fields <~ 1 T . In agreement with expectations
for the present 2DEG with relatively large SO-coupling strength, the Zeeman
term has only a minor effect on the spin-polarization profile along the chan-
nel even at ∼1 T fields. For the out-of-plane and [110]-oriented in-plane field,
the Zeeman coupling causes a weak suppression of the spin coherence, while
for the [110]-oriented field, it leads to a weak suppression of the spin coher-
ence, and a weak reduction or extension of the precession length depending
on the sign of the magnetic field. The variations at magnetic fields <~ 1 T are
not large enough to provide practical means for controlling the spin dynam-
ics in the 2DEG channels. At higher magnetic fields, we expect an increasing
role of the quantum orbital effects, which, due to the SO coupling, cannot
be simply disentangled from the spin effects, and their description requires
more sophisticated theoretical modeling. Furthermore, the field depen-
dence of the spin-current generation characteristics of the p–n junctions is
expected to be significant at fields of several Tesla.
8.4.4 Prospectives of Spin-Injection Hall Effect

The spin-injection Hall effect in nonmagnetic semiconductors and the
ability to tune independently the strengths of disorder and SO coupling
in semiconductor structures open up new opportunities for resolving
B=0 B1–10 = 0.5T

1 1
0 0
pz
pz
–1
0 0.25 –1
1 0 0.25
2 0 1
X110
– (µ
m ) 3 –0.25 X 110 (µm) X110
2 0
– (µ
m ) 3 –0.25 X 110 (µm)
B001 = 1 T B1–10 = 1 T
1 1
pz
0
pz
–1 –1
0 0.25 0 0.25
1 1 0
2 0 2 m)
m) X110
– (µ 3 –0.25 X 110 (µ
X110
– (µ
m ) 3 –0.25 X 110 (µ m)
B001 = 2 T B1–10 = 2 T
1 1
0 0
pz
pz
–1 –1
0 0.25 0 0.25
1 1 0
2 0 2
m) X110 (µm)
X110 X 110 (µ
– (µ – (µ 3 X 110
m) 3 –0.25 m) –0.25
(a) (b)
B110 = +0.5 T
1
0
pz
–1
0 0.25
1
2 0
X110
– (µ
m) 3 –0.25 X 110 (µm)
B110 = –0.5 T
0
pz
–1
0 0.25
1 0
2 m)
X110
– (µ
m) 3 –0.25 X 110 (µ
B110 = –1 T
0
pz
–1
0 0.25
1 0
2 m)
X110
– (µ
m) 3 –0.25 X 110 (µ
(c)
FIGURE 8.10 Monte Carlo modeling of the out-of-plane component of the spin-polarization in the chan-
nel for α = −0.5β at 300 K. (a) Zero magnetic field and magnetic field oriented perpendicular to the plane of
the 2DEG. (b) Magnetic field along the in-plane direction (the 2DEG channel direction). (c) Magnetic field
along the [110] in-plane direction (the 2DEG channel direction). Magnetic field causes a weak suppression of
the spin coherence for these two field orientations. (After Wunderlich, J. et al., Nat. Phys. 5, 675, 2009. With
permission.)
long-standing debates on the nature of spin-charge dynamics in the intrigu-

ing strong SO-coupling, weak disorder regime [66]. From the application
perspective, the spin-injection Hall effect devices can be directly imple-
mented as spin-photovoltaic cells and polarimeters, switches, invertors,
and, due to the non-destructive nature of the spin-injection Hall effect,
also as interconnects. We also foresee application of the spin-injection
Hall effect in the Datta–Das [63], and other proposed spintronic transistor
concepts [70]. An important next step toward a practical implementation
in nanotechnology, as pointed out in Ref. [71], is the replacement of opti-
cal spin-injection by other solid-state means of spin-injection. These light-
less devices utilizing the spin-injection Hall effect can be fabricated in a
broad range of materials, including indirect-gap SiGe semiconductors [72].
Since the magnitude of the spin-injection Hall effect scales linearly with
the SO-coupling strength, we expect ∼100× weaker signals in the SiGe
2DEGs as compared to our measurements in the GaAs/AlGaAs, which is
still readily detectable.
REFERENCES
1. J. Wunderlich, A. C. Irvine, J. Sinova et al., Spin-injection Hall effect in a planar
photovoltaic cell, Nat. Phys. 5, 675 (2009).
Hall effect, Rev. Mod. Phys. 82, 1539 (2010).
3. J. Schliemann, Spin Hall effect, Int. J. Mod. Phys. B 20, 1015 (2006).
4. S. Murakami, N. Nagaosa, and S. C. Zhang, Dissipationless quantum spin cur-
rent at room temperature, Science 301, 1348 (2003).
6. H.-A. Engel, E. I. Rashba, and B. I. Halperin, Theory of spin Hall effects in semi-
conductors, in Handbook of Magnetism and Advanced Magnetic Materials,
vol. 5, H. K. S. Parkin (Ed.), John Wiley & Sons Ltd., Chichester, UK, p. 2858,
2007.
7. E. M. Hankiewicz and G. Vignale, Spin-Hall effect and spin-Coulomb drag in
doped semiconductors, J. Phys. Condens. Matter 21, 253202 (2009).
8. M. Koenig, H. Buhmann, L. W. Molenkamp et al., The quantum spin Hall
effect: Theory and experiment, J. Phys. Soc. Jpn. 77, 310 (2008).
9. B. Bernevig and S. Zhang, Quantum spin Hall effect, Phys. Rev. Lett. 96, 0680
(2006).
10. C. L. Kane and E. J. Mele, Quantum spin hall effect in graphene, Phys. Rev. Lett.
95, 226801 (2005).
11. L. Berger, Side-jump mechanism for the Hall effect of ferromagnets, Phys. Rev.
B 2, 4559 (1970).
12. J. Smit, The spontaneous Hall effect in ferromagnetics-I, Physica 21, 877 (1955).
13. N. A. Sinitsyn, A. H. MacDonald, T. Jungwirth, V. K. Dugaev, and J. Sinova,
Anomalous Hall effect in a two-dimensional Dirac band: The link between
the Kubo-Streda formula and the semiclassical Boltzmann equation approach,
Phys. Rev. B 75, 045315 (2007).
14. N. A. Sinitsyn, Semiclassical theories of the anomalous Hall effect, J. Phys.
Condens. Matter 20, 023201 (2008).
15. R. Karplus and J. M. Luttinger, Hall effect in ferromagnetics, Phys. Rev. 95,
1154 (1954).
16. M. Onoda and N. Nagaosa, Topological nature of anomalous Hall effect in fer-
romagnets, J. Phys. Soc. Jpn. 71, 19 (2002).
References 365
17. T. Jungwirth, Q. Niu, and A. H. MacDonald, Anomalous Hall effect in ferro-

magnetic semiconductors, Phys. Rev. Lett. 88, 207208 (2002).
18. G. Sundaram and Q. Niu, Wave-packet dynamics in slowly perturbed crystals:
Gradient corrections and Berry-phase effects, Phys. Rev. B 59, 14915 (1999).
19. M. Chang and Q. Niu, Berry phase, hyperorbits, and the Hofstadter spectrum:
Semi-classical dynamics in magnetic Bloch bands, Phys. Rev. B 53, 7010 (1996).
20. N. Mainkar, D. A. Browne, and J. Callaway, First-principles LCGO calculation
of the magneto-optical properties of nickel and iron, Phys. Rev. B 53, 3692
(1996).
21. G. Y. Guo and H. Ebert, Band-theoretical investigation of the magneto-optical
Kerr effect in Fe and Co multilayers, Phys. Rev. B 51, 12633 (1995).
22. F. D. M. Haldane, Berry curvature on the Fermi surface: Anomalous Hall effect
as a topological Fermi-liquid property, Phys. Rev. Lett. 93, 206602 (2004).
23. X. Wang, D. Vanderbilt, J. R. Yates, and I. Souza, Fermi-surface calculation of
the anomalous Hall conductivity, Phys. Rev. B 76, 195109 (2007).
24. M. P. Marder, Condensed Matter Physics, Wiley, New York, 2000.
25. J. Smit, The spontaneous Hall effect in ferromagnetics–II, Physica 24, 39
(1958).
26. L. Berger, Influence of spin–orbit interaction on the transport processes in
ferromagnetic nickel alloys, in the presence of a degeneracy of the 3d band,
Physica 30, 1141 (1964).
27. P. Nozieres and C. Lewiner, Simple theory of anomalous Hall effect in semi-
conductors, J. Phys. (Paris) 34, 901 (1973).
28. M. I. D’yakonov and V. I. Perel’, Possibility of orienting electron spins with cur-
rent, Sov. Phys. JETP 467 (1971).
the spin Hall effect in semiconductors, Science 306, 1910 (2004).
30. V. Sih, R. Myers, Y. Kato, W. Lau, A. Gossard, and D. Awschalom, Spatial imag-
ing of the spin Hall effect and current-induced polarization in two-dimen-
sional electron gases, Nat. Phys. 1, 31 (2005).
tion of the spin-Hall effect in a two dimensional spin–orbit coupled semicon-
32. S. Valenzuela and M. Tinkham, Electrical detection of spin polarized currents:
The spin-current induced Hall effect, J. Appl. Phys. 101, 09B103 (2007).
33. S. Valenzuela and M. Tinkham, Direct electronic measurement of the spin
Hall effect, Nature 442, 176 (2006).
34. S. Valenzuela, Nonlocal spin detection, spin accumulation and the spin Hall
effect, Int. J. Mod. Phys. B 23, 2413 (2009).
35. E. Saitoh, M. Ueda, H. Miyajima, and G. Tatara, Conversion of spin current
into charge current at room temperature: Inverse spin-Hall effect, Appl. Phys.
Lett. 88, 8250 (2006).
36. T. Kimura, Y. Otani, T. Sato, S. Takahashi, and S. Maekawa, Room temperature
reversible spin Hall effect, Phys. Rev. Lett. 98, 156601 (2007).
37. C. P. Weber, J. Orenstein, B. A. Bernevig, S.-C. Zhang, J. Stephens, and D. D.
Awschalom, Non-diffusive spin dynamics in a two-dimensional electron gas,
Phys. Rev. Lett. 98, 076604 (2007).
38. N. P. Stern, S. Ghosh, G. Xiang, M. Zhu, N. Samarth, and D. D. Awschalom,
Current-induced polarization and the spin Hall effect at room temperature,
Phys. Rev. Lett. 97, 126603 (2006).
39. N. P. Stern, D. W. Steuerman, S. Mack, A. C. Gossard, and D. D. Awschalom,
Drift and diffusion of spins generated by the spin Hall effect, Phys. Rev. Lett.
91, 6210 (2007).
40. B. Liu, J. Shi, W. Wang et al., Experimental observation of the inverse spin Hall
effect at room temperature, arXiv:cond-mat/0610150 (2006).
41. J. M. Kikkawa and D. D. Awschalom, Lateral drag of spin coherence in gallium
arsenide, Nature 397, 139 (1999).
magnet/semiconductor structures, Science 309, 2191 (2005).
43. X. Lou, C. Adelmann, S. A. Crooker et al., Electrical detection of spin trans-
port in lateral ferromagnet-semiconductor devices, Nat. Phys. 3, 197 (2007).
44. R. Fiederling, M. Keim, G. Reuscher et al., Injection and detection of a spin-
polarized current in a light-emitting diode, Nature 402, 787 (1999).
Mater. 200, 110 (1999).
46. H. J. Zhu, M. Ramsteiner, H. Kostial, M. Wassermeier, H. P. Schönherr, and K.
H. Ploog, Room-temperature spin injection from Fe into GaAs, Phys. Rev. Lett.
87, 016601 (2001).
47. X. Jiang, R. Wang, R. M. Shelby et al., Highly spin-polarized room-temperature
tunnel injector for semiconductor spintronics using MgO(100), Phys. Rev. Lett.
94, 056601 (2005).
48. J. N. Chazalviel, Spin-dependent Hall effect in semiconductors, Phys. Rev. B 11,
3918 (1975).
49. H. Ohno, H. Munekata, T. Penney, S. von Molmar, and L. L. Chang,
Magnetotransport properties of p-type (In,Mn)As diluted magnetic III–V
semiconductors, Phys. Rev. Lett. 68, 2664 (1992).
50. J. Cumings, L. S. Moore, H. T. Chou et al., Tunable anomalous Hall effect in a
nonferromagnetic system, Phys. Rev. Lett. 96, 196404 (2006).
51. M. I. Miah, Observation of the anomalous Hall effect in GaAs, J. Phys. D: Appl.
Phys. 40, 1659 (2007).
52. S. D. Ganichev, E. L. Ivchenko, S. N. Danilov et al., Conversion of spin into
directed electric current in quantum wells, Phys. Rev. Lett. 86, 4358 (2001).
53. P. R. Hammar and M. Johnson, Detection of spin-polarized electrons injected
into a two-dimensional electron gas, Phys. Rev. Lett. 88, 066806 (2002).
55. J. Wunderlich, D. Ravelosona, C. Chappert et al., Influence of geometry on
domain wall propagation in a mesoscopic wire, IEEE Trans. Magn. 37, 2104
(2001).
56. M. Yamanouchi, D. Chiba, F. Matsukura, and H. Ohno, Current-induced
domain-wall switching in a ferromagnetic semiconductor structure, Nature
428, 539 (2004).
57. I. Žutić, J. Fabian, and S. D. Sarma, Spin-polarized transport in inhomoge-
neous magnetic semiconductors: Theory of magnetic/nonmagnetic p–n junc-
tions, Phys. Rev. Lett. 88, 066603 (2002).
58. T. Kondo, J. ji Hayafuji, and H. Munekata, Investigation of spin voltaic effect in
a pn heterojunction, J. Appl. Phys. 45, L663 (2006).
59. W. Knap, C. Skierbiszewski, A. Zduniak et al., Weak antilocalization and spin
precession in quantum wells, Phys. Rev. B 53, 3912 (1996).
60. R. Winkler, Spin–Orbit Coupling Effects in Two-Dimensional Electron and
Hole Systems, Springer-Verlag, New York, 2003.
61. J. Schliemann, J. C. Egues, and D. Loss, Nonballistic spin-field-effect transistor,
Phys. Rev. Lett. 90, 146801 (2003).
62. B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Quantum spin Hall effect and
topological phase transition in HgTe quantum wells, Science 314, 1757 (2006).
63. S. Datta and B. Das, Electronic analog of the electro-optic modulator, Appl.
Phys. Lett. 56, 665 (1990).
64. M. Ohno and K. Yoh, Datta-Das-type spin-field-effect transistor in the non-
ballistic regime, Phys. Rev. B 77, 045323 (2008).
65. A. A. Kovalev, K. Vyborny, and J. Sinova, Hybrid skew scattering regime of the
anomalous Hall effect in Rashba systems: Unifying Keldysh, Boltzmann, and
Kubo formalisms, Phys. Rev. B 78, 41305 (2008).
References 367
66. M. Borunda, T. Nunner, T. Luck et al., Absence of skew scattering in two-

dimensional systems: Testing the origins of the anomalous Hall effect, Phys.
Rev. Lett. 99, 066604 (2007).
67. T. S. Nunner, N. A. Sinitsyn, M. F. Borunda et al., Anomalous Hall effect in a
68. S. Onoda, N. Sugimoto, and N. Nagaosa, Quantum transport theory of anoma-
lous electric, thermoelectric, and thermal Hall effects in ferromagnets, Phys.
Rev. B 77, 165103 (2008).
69. A. Crépieux and P. Bruno, Theory of the anomalous Hall effect from the Kubo
formula and the Dirac equation, Phys. Rev. B 64, 014416 (2001).
71. I. Žutić, Spins take sides, Nat. Phys. 5, 630 (2009).
Rev. Lett. 97, 026602 (2006).
Section V
Magnetic
Semiconductors,
Oxides and
Topological Insulators
Chapter 9 Magnetic Semiconductors: III–V
Semiconductors371
Carsten Timm
Chapter 10 Magnetism of Dilute Oxides 413
J.M.D. Coey
Chapter 11 Magnetism of Complex Oxide Interfaces 457
369
370 Section V. Magnetic Semiconductors, Oxides and Topological Insulators
Chapter 12 LaAlO3/SrTiO3: A Tale of Two Magnetisms 489

Chapter 13 Electric-Field Controlled Magnetism 519
Chapter 14 Topological Insulators: From Fundamentals
to Applications 543
Chapter 15 Quantum Anomalous Hall Effect in
Topological Insulators 573
9
Magnetic
Semiconductors
III–V Semiconductors
Carsten Timm

9.2 III–V DMS with Narrow to Intermediate Gaps 373
9.2.1 Properties of Isolated Mn Dopants:
Single-Electron Picture 374
Many-Particle Picture 376
9.2.3 Other Isolated Defects 377
9.2.4 Effects of Weak Mn Doping 379
9.2.5 Effects of Heavy Mn Doping 381
9.2.6 Carrier-Mediated Magnetic Interaction 386
371
372 Chapter 9. Magnetic Semiconductors
9.2.7 Disorder and Transport 390

9.2.8 Disorder and Magnetism 393
9.3 III–V DMS with Wide Gaps 395
References 398
9.1 INTRODUCTION
The present chapter discusses the properties of III–V diluted magnetic
semiconductors (DMS) and their theoretical understanding. By DMS, we
mean semiconducting compounds doped with magnetic ions and show-
ing ferromagnetic order below a Curie temperature TC. Due to the semi-
conducting properties of DMS, the carrier concentration can be changed
significantly by doping, and even in situ by the application of a gate voltage,
or by optical excitations. Since the carriers couple strongly to the magnetic
moments, this leads to unique properties, and makes DMS promising for
spintronics. In particular, magnetotransport and spin transfer in DMS-
based devices are discussed in Chapter 4, Volume 2, while DMS quantum
dots are discussed in Chapter 6, Volume 3. Electric-field control of magne-
tism in DMS has also revealed important opportunities in ferromagnetism
metals, a discussed in Chapter 13 of Volume 2. The strong spin-orbit cou-
pling in most DMS is responsible for the anomalous Hall effect discussed
in Chapter 8, Volume 2.
Of the many excellent reviews on III–V DMS we only mention a few.
Jungwirth et al. [1] focus on the theory of III–V DMS, in particular for the
prototypical compound (Ga,Mn)As, and cover both ab initio calculations
and the Zener kinetic-exchange theory discussed below. More recently, Dietl
[2], as well as Dietl and Ohno [3], have reviewed experimental and theo-
retical aspects of DMS in general, where the latter article focuses mostly
on (Ga,Mn)As and also discusses device concepts. Jungwirth et al. [4] dis-
cuss the physical properties of (Ga,Mn)As in view of spintronics applica-
tions. A book chapter by Jungwirth [5] more generally discusses Mn-doped
III–V DMS and spintronics. Sato et al. [6] review ab initio calculations as
well as model-based calculations and simulations for III–V and II-VI DMS.
In a contribution for a broader audience, Zunger et al. [7] review theoreti-
cal approaches, contrasting model-based and ab initio calculations, and dis-
cussing potential pitfalls. More specialized reviews discuss wide-gap III–V
DMS and dilute oxides [8, 9], disorder effects in DMS [10], ferromagnetic
resonance (FMR) in (Ga,Mn)As [11], carrier localization in DMS [12], opti-
cal properties of III–V DMS [13], synthesis and epitaxial growth [14], the
anomalous Hall effect [15], and the nanomorphology of DMS [16]. The basic
picture put forward in Ref. [13] has been challenged by Jungwirth et al. [17],
as discussed below.
Two classes of III–V DMS have emerged: On the one hand, com-
pounds with narrow to intermediate gaps, such as antimonides and arse-
nides, which show Curie temperatures below room temperature, and on the
other wide-gap materials, such as nitrides and phosphides, often with higher
apparent TC. The first class appears to be much better understood, and will
be discussed first and in more detail. At least some of the wide-gap materi-
als are more similar to the dilute magnetic oxides discussed in the following
Chapter 10, and we will restrict ourselves to a number of comments perti-
nent to wide-gap III–V compounds.
9.2 III–V DMS WITH NARROW TO

INTERMEDIATE GAPS
The equilibrium solubility of Mn in DMS with narrow to intermediate
gaps lies below one percent of cations, too low for ferromagnetic order to
occur. This led to efforts to achieve higher Mn concentrations by means of
nonequilibrium molecular beam epitaxy (MBE) at low substrate tempera-
tures. With this technique, ferromagnetism in (In,Mn)As with TC » 7.5K
was achieved in 1992 [18], and in (Ga,Mn)As with TC » 60K in 1996 [19].
The Curie temperature of (Ga,Mn)As has been pushed above 100 K [20] and
recently up to around 190 K [21–24] by improvements in the MBE technique
and careful post-growth annealing, which allow higher Mn concentrations
up to nominally x » 16% per cation site [3, 21–25]. In GaAs δ-doped with
Mn, i.e. containing an essentially two-dimensional Mn-rich layer, TC » 250K
has been reached [26]. (Ga,Mn)As with large Mn concentration has also
been produced by Mn-ion implantation of GaAs followed by pulsed-laser
annealing [27]. The magnetization curves and the temperature-dependent
resistivity are comparable to annealed MBE-grown samples, while the values
of TC are somewhat lower, but still exceeding 100 K [27]. Meanwhile, TC for
(In,As)Mn has been increased to 90 K [28]. Recently, Tu et al. have observed
room-temperature ferromagnetism with TC ≈ 340 K in p-type (Ga,Fe)Sb thin
films [29] and with TC ≈ 335 K in n-type (In,Fe)Sb thin films [30], all grown
by low-temperature MBE.
(Ga,Mn)As is the most extensively studied III–V DMS, not the least
because the host material GaAs is technologically important and well-
understood. It is known that ferromagnetism in (Ga,Mn)As requires Mn
concentrations of x  1% [20, 31, 32]. There is a zero-temperature insulator-
to-metal transition at a somewhat higher concentration of about x = 1.5% for
well annealed samples [31–33]. The narrow-gap antimonides (Ga,Mn)Sb and
(In,Mn)Sb show lower TC [34–36], but are otherwise similar to the arsenides.
Recently, ferromagnetism with Curie temperatures of up to 40 K for x » 5%
Mn concentration has been observed in Mn-ion-implanted and pulsed laser
annealed InP [37], which is insulating for all concentrations. InP is of inter-
est, since its band gap and hole effective masses are similar to those of GaAs,
while the Mn acceptor level is much deeper [38].
At present, two central questions regarding (Ga,Mn)As remain partially
open. It is clear that for relatively light doping, bound acceptor states form a
narrow impurity band (IB) in the gap. It is also obvious that for heavy doping,
this IB will become very broad, merge with the valence band (VB), and lose
its identity [39]. There is an ongoing debate whether in high-TC (Ga,Mn)As
one can still identify an IB. A related, but separate, question is whether the
states close to the Fermi energy, which are responsible for charge transport,
and for the carrier-mediated magnetic interaction, are localized or extended.
In the following sections, we discuss the physics of (III,Mn)V DMS, pro-
gressing from light to heavy Mn-doping.

Single-Electron Picture
Manganese in narrow and intermediate-gap III–V semiconductors forms
Mn2+ ions at cation sites of the zinc blende lattice, as seen in electron para-
magnetic resonance (EPR) [40, 41], and FMR [42]. The lattice with substitu-
tional Mn impurities is depicted for the example of (Ga,Mn)As in Figure 9.1a.
Since Mn2+ replaces a 3+ ion, manganese provides one fewer electron or,
equivalently, donates a hole and thus acts as an acceptor. Since it has a nega-
tive charge relative to the cation sublattice, it can bind the hole. The neu-
tral acceptor-hole complex has been observed in infrared (IR) spectroscopy
[43, 44] and EPR [45]. The hole binding energy is found to be 112.4 meV
[43–45], in agreement with the thermal activation energy seen in transport
[46]. Mn acceptors close to the surface have been imaged and switched
between the ionized (no bound hole) and neutral states with a scanning tun-
neling microscope (STM) at room temperature [47].
In contrast to simple shallow acceptors, Mn2+ has a half-filled d-shell.
Photoemission experiments [48, 49] find the maximum weight of occupied
d-orbitals about 4.5 eV below the Fermi energy and do not find evidence
for a mixed-valence state, suggesting that the unoccupied d-levels also do
not lie close to the Fermi energy. The absence of d-level signatures in the
optical conductivity shows that they lie above the conduction-band (CB)
minimum. The d-shell thus mostly acts as a local spin S = 5/2, consistent
Mn interstitial
MnGa
MnGa
As antisite
(a) (b)
FIGURE 9.1 (a) Zinc blende lattice structure of GaAs with two substitutional Mn impurities. Ga3+ ions are
shown as small red spheres, As3− as large, semitransparent blue spheres, and Mn2+ as small green spheres. The
gray cube denotes the conventional unit cell. The ionic radii are used in the plot, which is an exaggeration,
since the bonds are partially covalent. (b) Lattice structure of GaAs with a Mn2+ interstitial in the As-coordinated
tetrahedral position and an As antisite defect (small blue sphere with the As5+ ionic radius).
with EPR experiments [45]. It is useful to notionally divide the Mn impurity

into a simple acceptor and the half-filled d-shell. Ignoring the d-shell for the
moment, the attractive potential should lead to hydrogenic energy levels for
the hole. A spherical approximation for the band structure of GaAs gives a
binding energy of 25.7 meV [50, 51]. Binding energies of typical hydrogenic
acceptors are larger, for example 30.6 meV for Zn2+ [52], since the dielectric
screening of the impurity charge becomes less effective close to the impurity.
This can be taken into account by a central-cell correction [51], i.e. an addi-
tional short-range attractive potential.
A more complex picture emerges if we take into account that the hole
also experiences the periodic potential of the host lattice. The bound accep-
tor states are derived from small wave-vector states close to the VB top,
which mostly consist of anion p-orbitals. The point group at the impurity
(cation) site in zinc blende semiconductors is the tetrahedral group Td [53].
Neglecting spin-orbit coupling, the anion p-orbitals correspond to the G5
irreducible representation of Td, i.e. they have t2 symmetry [53]. Since G5 is
three-dimensional, the bound acceptor state is six-fold degenerate, including
the spin degeneracy.
In the presence of spin-orbit coupling, the theory of double groups [53]
shows that the level splits into a G 7 (e¢¢ ) doublet and a G8 (u¢) quartet. The
doublet can be thought of as resulting from the split-off band, which for GaAs
is separated from the heavy-hole and light-hole bands by ∆ SO = 341 meV at
the G point. The doublet is therefore expected to form a resonance in the
VB. The remaining four-fold degeneracy can be understood in terms of a
total angular momentum j = 3 / 2 resulting from orbital angular momentum
l = 1 and spin s = 1 / 2 . STM experiments [33, 47, 54, 55] find a pronounced
spin-dependent spatial anisotropy of these states, in agreement with theory
[47, 54–57].
We now turn to the Mn d-electrons. In the Td crystal field, the five
d-orbitals split into three orbitals with t2 character and two with e char-
acter. All five are singly occupied and the electron spins are aligned (high-
spin state), showing that the Hund’s-rule coupling is larger than the
crystal-field splitting. Importantly, the t2-type d-orbitals hybridize with the
p-orbitals of the neighboring anions. This hybridization strongly affects the
bound acceptor states, which are also of t2 type, if spin-orbit coupling is
neglected, and have a large amplitude at the neighboring sites [58, 59]. The
occupied d-orbitals, which all have the same spin, can only hybridize with
the bound acceptor states with the same spin. Consequently, level repul-
sion shifts the occupied d-levels down in energy and the bound acceptor
states with the same spin up. The resulting bound acceptor states, called
the dangling-bond hybrid [58, 59], contain some d-level admixture, and
are deeper in the gap, and thus more strongly bound than for a hydrogenic
impurity [51, 60]. The occupied d-levels with some admixture of anion
p-orbitals are called the crystal-field resonance [58, 59]. The resulting spin-
up single-particle levels are sketched in the left-hand part of Figure 9.2.
On the other hand, the bound acceptor state with spin opposite to
the Mn d-shell hybridizes with the unoccupied t2 orbitals and is shifted
t2 Mn d
E
CFR
CB e
DBH
acceptor acceptor
states t 2 t 2 states
VB DBH
e
CFR
Mn d t2
FIGURE 9.2 Sketch of single-particle states introduced by substitutional Mn acceptors in GaAs. The spin of
the Mn d-shell is assumed to point upward. Spin-orbit coupling is ignored. Bold (dashed) arrows indicate occu-
pied (unoccupied) states. The VB and CB edges are indicated. Left: level repulsion between spin-up d-states of
t2 symmetry and bound acceptor states also of t2 symmetry leads to the occurrence of a spin-up dangling-bond
hybrid (DBH) in the gap and a crystal-field resonance (CFR) deep in the VB. Right: for spin-down states, the unoc-
cupied d-states are high above the Fermi energy, and the DBH is shifted downward in energy. In the ground
state, the single hole introduced by the Mn acceptor occupies one of the most strongly bound spin-up DBHs.
downward, as shown in the right-hand part of Figure 9.2. This leads to spin
splitting of the bound state, which depends on the orientation of the Mn spin
[58, 59, 61]. In the ground state, a hole will occupy the most strongly bound
state, which has the same spin as the d-electrons. Thus an electron with this
spin is missing, and the total spin of the system is reduced. In this sense, the
exchange coupling Jpd between the Mn d-shell and the VB-derived states is
antiferromagnetic.

Many-Particle Picture
For several reasons, the single-particle picture is not sufficient for the under-
standing of Mn dopants [17]: First, the splitting between occupied and unoc-
cupied Mn d-levels is large, mostly due to the on-site Coulomb repulsion
U ≈ 3.5 eV [49]. Second, there is a large Hund’s-rule coupling J H ≈ 0.55 eV
[62]. The resulting strong correlations are not well-captured by density-
function theory (DFT) using standard approximations such as the local (spin)
density approximation (LDA, LSDA). There has been progress in incorpo-
rating correlations either by phenomenologically introducing a Hubbard-U
(LDA+U, LSDA+U) [6, 63, 64] or by correcting for self-interaction artifacts
[6, 65, 66]. These approaches generically push Mn d-orbital weight away from
the Fermi energy, leading to a smaller Mn d-orbital weight in the bound
acceptor states. Since the hybridization between the anion p-orbitals and
the Mn d-orbitals is small compared to their energy separation, it can be
described perturbatively. Canonical perturbation theory [67, 68] maps the

d-electrons onto a local spin and a short-range potential scatterer. The lat-
ter is a typically negligible correction to the Coulomb attraction. One finds
an antiferromagnetic pd-exchange interaction Jpd(r) between the hole and
Mn spins, which falls off on the length scale of the anion-cation separa-
tion. In reciprocal space, this leads to a k-dependent interaction Jpd(k) [68].
The exchange coupling quoted in the literature is the limit J pd = J pd (k ® 0).
Third, each Mn acceptor can bind a single hole, but not a second one [59]. Thus
the Coulomb repulsion Uacc in the bound acceptor state is at least of the same
order as the binding energy. Impurity states in doped semiconductors are
strongly correlated systems [69]. Naive single-particle approaches predict
too many bound acceptor states, seen for example in Figure 9.2.
The emerging picture describes an impurity state containing a single
hole with total angular momentum j = 3/2 antiferromagnetically coupled to
a local spin S = 5/2. In the four-dimensional subspace spanned by the j = 3/2
impurity states, one finds the operator identity s = j / 3 [1, 51], which allows
one to write the exchange interaction either as ej × S or as 3es × S . Taking the
factor of three into account, one can express the exchange constant ε in
terms of the exchange coupling Jpd and the envelope function f(r) describing
the spatial extent of the bound acceptor state [1, 51],
1 3 1
e=
3 ò
d rJ pd (r)| f (r)|2 » J pd | f (r)|2 .
3
(9.1)
The approximation in the second step is valid if the envelope function

changes much more slowly in space than the exchange interaction. The aver-
age is taken over the range of Jpd(r). From IR spectroscopy [44], a value of
ε ≈ 5 meV is obtained. Photoemission experiments [48] independently give
J pd ≈ 54 meV nm 3. (In the literature, Jpd is often denoted by β and JpdnMn by
βN0, where nMn is the Mn concentration [1]). The exchange interaction can
be rewritten as e j × S = (e / 2)[ F ( F + 1) - j( j + 1) - S(S + 1)], where F = j + S is
the total angular momentum. The splitting between the ground state triplet
(F = 1) and the first excited quintet (F = 2) is thus 2e » 10 meV [44].
9.2.3 Other Isolated Defects

III–V DMS contain other defects apart from substitutional magnetic dop-
ants. For (Ga,Mn)As, the most important ones are As ions at Ga sites (arse-
nic antisites), and Mn ions in interstitial positions. The positions of these
two defects are shown in Figure 9.1b. Vacancy defects are present in much
smaller concentrations, but can strongly enhance the mobilities of other
defects [70].
Even undoped GaAs grown by low-temperature MBE can contain a rela-
tively high concentration of As antisites [71], which act as double donors, as
expected for As5+ cations. In (Ga,Mn)As, antisites should thus lead to par-
tial compensation: There are fewer holes than Mn acceptors. Variation of
the As:Ga ratio during growth has been found to lead to a corresponding
change in the hole concentration, confirming that antisites are also present
in (Ga,Mn)As [72]. Cracking As4 clusters to As2 during growth can substan-
tially reduce the compensation [73], probably by reducing the concentration
of antisites. The mobility of antisites in GaAs is likely small at relevant tem-
peratures [74, 75] so that their distribution is not affected by post-growth
annealing. However, due to their Coulomb attraction, antisite donors could
preferentially be incorporated close to Mn acceptors during growth [76].
During low-temperature growth of (Ga,Mn)As, some Mn2+ ions are
incorporated in interstitial positions. Channeling Rutherford backscatter-
ing experiments [77] have provided evidence for about 17% of Mn residing in
interstitial positions for as-grown samples with a Mn concentration of x » 7% .
DFT calculations find a lower formation energy of the As-coordinated tetra-
hedral Mn interstitial than of the Ga-coordinated one [59, 78, 79], as expected
for a cation, but the predicted energy differences do not agree well. Channeling
Rutherford backscattering cannot distinguish between these two positions.
Mn2+ interstitials in GaAs are double donors contributing to compensa-
tion. Since the mobility of Mn interstitials at typical growth and annealing
temperatures around 250°C is substantial [74, 77, 78, 80, 81], they can lower
their electrostatic energy by assuming positions close to substitutional Mn
acceptors during annealing, not just during growth [76, 82]. Mn interstitials
also diffuse to the film surface during annealing, where they form a magneti-
cally inactive layer [78, 80, 81, 83–85]. This is thought to be the main origin
of the observed decrease of compensation during annealing [70, 83, 86] and
of the higher TC of thinner films [87, 88]. This is nicely demonstrated by cap-
ping the surface with a thin GaAs layer, which strongly suppresses the effect
of annealing [83–85]. Mn2+ interstitials also carry a spin S = 5/2. For a Mn
interstitial and a substitutional Mn dopant in nearest-neighbor positions,
one expects a strong direct antiferromagnetic exchange interaction [79].
X-ray magnetic circular dichroism experiments [89] support this picture.
Substitutional-interstitial pairs can thus form strongly bound spin singlets,
effectively removing active spins. The contribution of these pairs is sup-
pressed by annealing [89]. On the other hand, the question of the exchange
interaction between VB holes and Mn-interstitial spins is still open, with
DFT calculations giving conflicting results [79, 82].
Since As antisites and Mn interstitials are double donors, the electrostatic
interaction favors their formation for higher concentrations of substitutional
Mn acceptors. This self-compensation is a limiting factor in the growth of heav-
ily Mn-doped (Ga,Mn)As. Yuan et al. [61] have reported that ion implantation
of Mn in GaAs and InAs followed by pulsed laser melting leads to samples that
are essentially free of both As antisite defects and Mn interstitials.
A promising idea for how to overcome this problem as well as the limita-
tion of the same Mn dopants supplying both magnetic moments and carriers
is to dope III–V semiconductors with iron [29, 30, 90]. Iron mostly replaces
the cation in the Fe3+ state so that it only contributes magnetic moments,
while the carrier are introduced by independent doping. This allows to grow
both p-type and n-type DMS and room-temperature ferromagnetism has
been reported for the representative compounds (Ga,Fe)Sb [29] and (In,Fe)
Sb [30], respectively.
9.2.4 Effects of Weak Mn Doping

A non-zero concentration of Mn impurities does not have a strong effect
on the exchange interaction Jpd, since Jpd is mainly due to hybridization
between Mn and neighboring anion orbitals. In contrast, it does have impor-
tant effects on the electronic system. For small Mn concentrations, the total
potential seen by the holes consists of the sum of the Coulomb potentials
of the individual acceptors. The impurity potentials fall off like 1/r in the
absence of metallic screening, whereas the bound-state wave function falls
off exponentially. Hence, the overlap of the potentials remains relevant for
arbitrarily small concentrations, unlike the overlap of the bound states. This
situation of very weak doping is sketched in Figure 9.3a. The acceptor binding
energy in this regime is reduced, since an acceptor can be ionized by excit-
ing the hole to roughly the average potential. This leads to a reduction of the
binding energy proportional to x1/3 [46, 91], which is observed in Mn-doped
GaAs [17]. In this regime, IR spectroscopy shows a broad mid-IR peak in the
ac conductivity [43, 44, 92]. This peak is broad in spite of the sharp energy
of bound acceptor states, since the bound impurity wave functions contain
many wave vector components, so that transitions from large-k states deep
within the VB to the impurity states are possible [17].
With increasing Mn doping x, the bound acceptor states start to over-
lap. Let us for the moment assume a periodic superlattice of Mn dopants,
N ( E)
N ( E)
acceptor states
VB CB VB IB CB
(a) EF E (b) EF E
N ( E)
N ( E)
localized
VB IB states CB VB CB
(c) EF E (d) EF E
FIGURE 9.3 Density of states of III–V DMS with narrow to intermediate gaps. The Fermi energy EF is in all
cases shown assuming no compensation. (a) For very weak doping, the overlap and thus the hybridization
of bound acceptor states is negligible. (b) For larger doping, but assuming a regular superlattice of dopants,
a sharply delimited IB emerges. If the doping level is not too high, there is a true gap between IB and VB.
For zero compensation, EF falls into this gap. (c) A quite different picture emerges if disorder is included. The
bands develop tails of localized states (shown cross-hatched), and the IB becomes asymmetric. Note that it
is at present unknown whether extended states exist in the center of the IB before it merges with the VB, as
shown here, or not. (d) For heavy doping, VB and IB merge completely, and the Fermi energy lies deep in the
region of extended states.
ignoring disorder. Tunneling of electrons between the acceptor states leads

to a narrow IB. Consequently, the mid-IR conductivity peak broadens, and
shifts slightly to lower energies [92]. The total weight of the IB increases
linearly with x, whereas the tunneling matrix elements, and thus the band
width, increase exponentially. Thus the density of states does not become
very large. This might explain why the Stoner enhancement of ferromagne-
tism, which is governed by the product of local Coulomb repulsion Uacc and
density of states, remains small [93–96].
Since we have so far ignored disorder, the density of states shows sharp
edges of IB and VB and a true gap between the bands. A density of states of
this type is predicted by theories that effectively average over disorder [60,
97, 98]. The doping also affects the host VB, for example since the impurity
states take away spectral weight from the VB. The density of states in this
regime is sketched in Figure 9.3b.
Due to the random positions of Mn dopants, the system is not peri-
odic. Hence, band theory is not strictly valid, whereas the density of states
remains well-defined. Due to disorder, the band edges in the density of
states develop tails [56, 99], which always overlap, so that there is no true
gap. However, it is reasonable to say that the IB still significantly affects
the physics, if there is a clear minimum in the density of states in the over-
lap region with the VB. In addition, since the tunneling matrix elements
decay exponentially with separation, the density of states of the IB becomes
highly asymmetric [100]. The disorder also leads to strong (Anderson–
Mott) localization of the states in the band tails, which are separated from
extended states by mobility edges [12, 101–103]. This situation is sketched
in Figure 9.3c.
Ab initio calculations suggest that Mn dopants have a relatively large
attractive interaction [3, 104]. Hence, they are likely incorporated in posi-
tively correlated positions. They might further cluster during annealing but
only in the presence of vacancies [74]. While the qualitative picture is not
changed by correlated disorder, for a quantitative description detailed mod-
eling of dopant positions is desirable. With compensation, EF can shift to the
extended states close to the IB center. While it is controversial up to which
Mn concentration x this picture applies in (Ga,Mn)As, for (In,Mn)P, which
has a similar electronic structure but a much larger acceptor binding energy
of 220 meV, all studied samples with x £ 5% are expected to be in the IB
regime [37, 38].
It remains true that the IB contains one state per dopant, whereas
single-particle approaches taking the angular momentum j = 3/2 into
account find four states. Such approaches predict the Fermi energy EF
to lie deep within the IB, even for vanishing compensation, leading to a
relatively large density of states at EF [105, 106]. However, taking many-
particle effects into account, EF falls into the minimum of the density of
states between IB and VB, leading to very different physics. A simplistic
model with a separate IB containing too many states can evidently mimic
the large density of states at EF found if the bands have merged.
9.2.5 Effects of Heavy Mn Doping

For further increased Mn doping, the IB peak in the density of states grows
in weight and width. Consequently, the minimum in the density of states
between IB-type and VB-type states is gradually filled in. Krstajić et al. [39]
estimate that the energy of the IB center minus half its width—a rough mea-
sure of its lower edge—crosses the unperturbed VB top already for a Mn
concentration below 1%.
In the heavily doped limit, the IB completely merges with the VB. Due to
disorder, the states in the band tail are expected to be localized, see Section
9.2.7, but the hole concentration is relatively high so that EF lies deep in the
band, unless compensation is nearly complete. Both the impurity potential
and the Coulomb repulsion between the holes are screened by the itinerant
holes close to EF. In this regime, it is reasonable to start from the unper-
turbed VB with the chemical potential located in the band and to incorpo-
rate disorder perturbatively in the second step [1]. The density of states for
this case is sketched in Figure 9.3d.
The picture emerging for heavily doped DMS—holes in a weakly per-
turbed VB coupled to local spins by an antiferromagnetic exchange inter-
action—is the Zener kinetic-exchange or pd-exchange model [1, 39, 93, 94,
101, 107, 108]. Taking the pd-exchange interaction to be local, the model is
characterized by a Hamiltonian of the form
H = H holes + J pd åS × s ,
i
i i (9.2)
where
H holes = åe
nks
c†
c
nks nks nks (9.3)
describes non-interacting holes in the VB and Si and si are the impurity

and electronic spins, respectively, at Mn position Ri. In this basic form, the
Coulomb potentials due to acceptors and donors are neglected. Since prop-
erties of the host band structure are thought to be crucial, a realistic descrip-
tion is called for. The band structure of GaAs is sketched in Figure 9.4. In
DFT one obtains the (Kohn–Sham) band structure as a matter of course,
which is in fair agreement with the real (Landau-quasiparticle) bands if spin-
orbit coupling, which is strong in arsenides and antimonides, is included.
The main difference is the well-understood underestimation of the band gap.
Adding a local approximation to the Coulomb potential of the Mn acceptors,
VMn ∑i ni, where ni is the electronic charge density at Mn position Ri, one
arrives at the so-called V-J model [109, 110]. As discussed further in Section
9.2.8, Bouzerar and Bouzerar [109, 110] obtain reasonable results from a V-J
model with Hholes describing holes on a simple cubic lattice, suggesting that
while the Coulomb disorder potential is crucial, details of the band structure
are not as important as is often believed.
There are two main model-based approaches to a realistic band
structure enks . On the one hand, in Luttinger–Kohn k × p theory [57, 95,
111–113], one restricts the Hilbert space to the relevant bands (usually
the heavy-hole, light-hole, and split-off valence bands, perhaps adding the
conduction band) and expands enks in the wave vector k around the G
point (k = 0). The approach incorporates spin-orbit coupling and the non-
spherical k-dependence of enks and includes the correct effective masses
at G. The bands become increasingly inaccurate away from k = 0. If only
the light-hole and heavy-hole bands are taken into account and the non-
spherical k-dependence is averaged over all directions, one arrives at the
spherical approximation, which is described by the first-quantized 4 × 4
Hamiltonian
2 éæ 5 ö 2 2ù
H sph (k) = ê ç g1 + g 2 ÷ k - 2 g 2 (k × j) ú (9.4)
2m ëè 2 ø û
in terms of the 4 × 4 total-angular-momentum operator j with j = 3/2. g1
and g 2 are parameters. The last term stems from spin-orbit coupling, and
is responsible for the splitting between the light-hole and heavy-hole bands
away from the G point.
The other approach, Slater–Koster tight-binding theory [56, 68, 99,
114–116], is formulated in real space by specifying on-site energies and hop-
ping amplitudes for the relevant orbitals. Spin-orbit coupling is included
[115]. These parameters are chosen so as to match the real band structure at
high-symmetry points in the Brillouin zone. Therefore, the overall shape of
the band structure agrees well with experiments—see Figure 9.4—but local
properties, such as effective masses, are not well described. To resolve this
problem, Yildirim et al. [117] have introduced a tight-binding model that
reduces to the k × p Hamiltonian for small k.
CB
2
band gap
energy (eV)
0
HH
LH
-2
SO
-4
L Γ X K Γ
FIGURE 9.4 Sketch of the band structure of GaAs (more precisely, the result of a
tight-binding calculation using parameters from Ref. [115]). The band structure of
other zinc blende semiconductors is topologically identical. The conduction band
(CB) and the heavy-hole (HH), light-hole (LH), and spin-off (SO) valence bands are
indicated.
Returning to the controversy regarding the range of applicability of the

VB picture, model-based theories tend to predict that the IB and VB have
merged at Mn concentrations x of a few percent in GaAs [1, 39, 60, 100,
118]. In particular, Mašek et al. [118] discuss various proposed models with
separate IBs, and argue that none of them can be derived from a plausible
microscopic model. The VB picture has been very successful in explaining
many experiments, for example, magnetization curves, magnetic anisot-
ropies, magnetic stiffness, Gilbert damping, magnetotransport properties,
the anomalous Hall effect, photoreflectance spectroscopy, Faraday and Kerr
effect, and Raman spectroscopy [1, 3, 15, 119]. Experiments that appear to be
inconsistent with this picture are discussed in the following [1, 3, 13, 119].
(1) Some angle-resolved photoemission spectroscopy (ARPES) experi-
ments show a flat (momentum-independent) band merging with the VB top
close to the G point for (Ga,Mn)As with x » 3.5% [120] and x » 2.5% [121].
(The flat band has not been observed in other ARPES experiments with
x » 3% [122]). The flat band is naturally interpreted as an IB that in the den-
sity of states has merged with the VB. However, it can also be described in
terms of strongly localized—and thus momentum non-conserving—states at
the top of the VB. Recent off- and on-resonance photoemission for x » 0.1% ,
1%, and 6% [123], and hard x-ray photoemission experiments for both x » 1%
and x » 13% [124], also show merged IB and VB. The latter work [124] does
not show any dip in the density of states even at the low concentration x » 1%,
in agreement with photoemission results at larger doping. Hence, the IB pic-
ture as understood in Section 9.2.4 does not apply. Care should be taken when
interpreting the additional density of states close to the Fermi energy as being
due to the merged IB. These states derive from the DBH, i.e. they mostly have
anion p-orbital character with some Mn d-orbital admixture (see Section
9.2.1). It is thus misleading to speak of a “Mn 3d-derived impurity band” [121].
Off- and on-resonance difference photoemission spectroscopy [123] is sensi-
tive to the d-orbital weight. According to this work, d-orbital weight exists up
to the merged VB edge and holes are predominantly located close to the Mn
acceptors, as expected for DBHs. Nevertheless, the states at the Fermi energy
have only about 1% Mn d-orbital character [123]. We note in passing that the
comparison between experiment and theory [123] suggests that correlation
effects have to be taken into account to achieve quantitative agreement. This
can be understood in terms of a Coulomb anomaly at the Fermi energy [3].
(2) The effects of intentionally increased disorder are interpreted in
terms of the IB picture. Stone et al. [125] find that in the quaternary DMS
(Ga,Mn)(As,P) and (Ga,Mn)(As,N) already a concentration of phosphorus
or nitrogen in the percent range makes the samples insulating and signifi-
cantly reduces TC . This is in agreement with estimates based on the IB pic-
ture. Furthermore, Sinnecker et al. [126] have irradiated highly compensated
(Ga,Mn)As with Li ions to induce disorder, finding that the transport and
magnetic properties are rather robust. This is argued to indicate that trans-
port takes place in an IB, since the VB should be much more strongly affected
by the disorder [126]. Note that high and low sensitivity of the DMS against
disorder are both cited as support for the IB picture [125, 126], indicating
that further work is needed. Recently, Zhou et al. [127] have performed He
ion implantation for nearly uncompensated (Ga,Mn)As. The magnetization
curves remain Brillouin-function like but TC decreases continuously with
increasing He fluence, which is interpreted in terms of the VB picture.
(3) Various IR spectroscopy experiments are interpreted as favoring the
IB picture. There are contradictory results from ellipsometry concerning the
shift of critical points with Mn concentration x [128–130]. These experi-
ments are sensitive to the evolution of the VB-CB band gap in various parts
of the Brillouin zone. Gluba et al. [130] also perform photoreflectance exper-
iments, which have a higher energy resolution, and find, in agreement with
Ref. [129] but contradicting Ref. [128], no significant blueshift of the critical
points. While the blueshift reported in Ref. [128] was interpreted in terms
of the IB picture, its absence is argued to favor the VB picture [129, 130]. It
should be noted that no detailed comparison with theory has been made.
Singley et al. [131] and Burch et al. [132] have studied (Ga,Mn)As with
x = 1.7% − 7.3% using IR absorption. Except for x = 1.7%, the samples are metal-
lic and show a broad mid-IR peak in the ac conductivity [131, 132]. This peak
increases in weight for increasing x and also with annealing. The peak maxi-
mum lies at around 250 meV for x = 2.8% and shifts to lower energies (red-
shifts) for increasing x and with annealing. In a newer study over a broader
doping range [25], the mid-IR peak is argued to develop adiabatically from
insulating to very heavily doped (x = 16%) samples. The main tenet of Refs. [13,
25, 131, 132] is to interpret this peak in terms of VB-to-IB transitions. The
redshift with increasing x is then naturally explained in terms of increased
electronic screening, which moves the IB closer to the VB, and growing IB
width [13, 131, 132]. These features are incorporated in a recent theoretical
study [106], which treats the disorder by exact diagonalization in a supercell.
The authors consider x ≤ 1.5%, mostly outside of the controversial region. Note
that the calculation is based on a single-particle picture and therefore over-
estimates the number of states in the IB. The calculated peak shows a slight
redshift or blueshift with increasing x depending on the compensation [106].
A systematic experimental study of the mid-IR peak vs. Mn concentra-
tion x for samples with low compensation [133] mostly shows a blueshift with
increasing x. This agrees with the VB picture under the assumption of weak
disorder [134]. In this picture, the mid-IR peak is due to transitions between
the light-hole and heavy-hole bands. For increasing x, the Fermi energy
moves deeper into the VB, the typical k vectors of these transitions grow,
and the energy difference between light-hole and heavy-hole bands increases.
Jungwirth et al. [17] have challenged the IB interpretation of IR experiments
showing a redshift [13, 131, 132]. They point out that the maximum of the
mid-IR peak is at a higher energy than expected from extrapolating the weak-
doping data [92], and that the VB picture can also explain a redshift of the
mid-IR peak, if disorder is treated beyond the first-order Born approximation
(see Section 9.2.7). The main idea is that transitions into localized states in
the tail of the merged VB do not have to conserve momentum, and can there-
fore contribute with large weight at relatively high energies [17]. For increas-
ing x, the samples become more metallic and the high-energy contribution
from localized states decreases, which could lead to a redshift. This view is
supported by calculations by Yang et al. [135] within the VB picture, which
incorporate the electron–electron interaction in the Hartree approximation,
and the disorder by exact diagonalization for supercells.
For some of the samples, the ac conductivity also shows a smaller maxi-
mum at zero frequency [131, 132], which is interpreted as a Drude peak. By
fitting a two-component model to the data and applying the conductivity
sum rule to the Drude part alone, Burch et al. [132] extract the effective
mass of the carriers. The resulting large effective mass is interpreted in terms
of strong electron–electron interactions, as expected for an IB [132]. The
analysis should be viewed with caution, since it relies on a single sample and
neglects that the Drude peak can loose weight by disorder scattering [17].
Hot-electron photoluminescence experiments [136] are interpreted by the
authors in terms of transitions of hot electrons from the CB into an IB.
However, Jungwirth et al. [17] argue that a VB picture including disorder
can explain the results, along with the shift of the mid-IR peak. Detailed
calculations based on the VB or IB picture do not exist, however.
Finally, magnetic circular dichroism experiments [137] on (Ga,Mn)As
with x  3% have been interpreted in terms of multiple IBs. The data and in
particular the positive sign of the dichroism signal are claimed to be inconsis-
tent with the VB picture [137]. A calculation that is based on a tight-binding
model for the VB, and includes disorder in real space [138], can explain the
positive signal. This model still predicts a clear IB, only slightly overlapping
with the VB for x = 2% [139]. Turek et al. [139] show that the positive-circular-
dichroism result in Ref. [138] mainly comes from VB-to-CB transitions, not
from VB-to-IB transitions. They also find that a tight-binding model based
on ab initio calculations [140], which does not contain an IB, predicts a posi-
tive dichroism signal over most of the energy range. Note that IR magneto-
optical experiments have been explained within the VB picture [141].
(4) Resonant tunneling spectroscopy [142, 143] for (Ga,Mn)As and
(In,Ga,Mn)As suggests that Mn doping affects the VB only very weakly and
that the Fermi energy is pinned above the VB edge, presumably in an IB.
This interpretation has been challenged by Dietl and Sztenkiel [3, 144], who
explain the observations in terms of the VB picture for (Ga,Mn)As and invok-
ing the formation of hole subbands in the GaAs:Be electrode. They also point
out that the tunneling spectra do not show tunneling into the purported IB.
The alternative interpretation has been rejected by the original authors [145].
Detailed calculations for heterostructures should help to resolve the issue.
On the other hand, several rather general arguments support the VB pic-
ture for not too low Mn concentrations. First, for carrier-mediated magnetic
interactions, a minimum in the density of states as sketched in Figure 9.3c
would lead to a suppression of TC at weak compensation. Experimentally,
there is no sign of this for metallic (Ga,Mn)As [1, 21, 22, 127, 146–150].
Indeed, this question was specifically addressed in a study by Wang et al.
[150] employing hydrogen-co-doping. The authors find a monotonic increase
in Tc with carrier concentration, and observe the largest values of Tc for the
least compensated samples, in agreement with the VB picture. Furthermore,
in the He-implantation study [127], the main effect of irradiation appears

to be the reduction of the hole concentration through the creation of deep
traps. The monotonic decrease of Tc for increasing fluence is then inconsis-
tent with the IB picture.
Second, the temperature-dependent dc conductivity should show an
upturn when thermal activation of holes from the Fermi energy, supposed
to lie in an IB, to extended states in the VB becomes possible. Indeed, this
is observed for weakly Mn-doped GaAs [17, 151]. In metallic (Ga,Mn)As
with x  2%, no activated behavior is observed [17]. Neumaier et al. [152]
have extracted the density of states from the temperature-dependent cor-
rection to the conductivity due to the electron–electron interaction [153]
for samples of various geometries with x » 4% and x » 6% . They do not find
any evidence for a dip in the density of states and do find effective masses
on the order of the bare mass. Their results are close to the prediction of a
tight-binding model including disorder [56], and suggest merged bands and
itinerant holes at EF . Taken together, these results indicate that there is no
pronounced dip in the density of states and no mobility gap, i.e. an energy
range with localized states.
9.2.6 Carrier-Mediated Magnetic Interaction

For heavily doped (Ga,Mn)As, we have arrived at the Zener kinetic-exchange
model [107]. This is different from the Zener double-exchange model [154],
which is concerned with the interaction between mixed-valence ions. The
two pictures are opposite limiting cases on a continuum, though. For heavy
doping, superexchange between neighboring Mn dopants might also be rel-
evant [63, 96, 155, 156]. For Mn2+ in III–V DMS, superexchange is antiferro-
magnetic [157]. Dietl et al. [101] have argued that it is small. Within standard
perturbation theory, the superexchange interaction is indeed smaller than
the RKKY interaction discussed below by a factor on the order of EF divided
by the Coulomb repulsion U in the d-shell [68].
There are a number of smoking-gun experiments showing that the mag-
netic interaction is indeed carrier-mediated: in field-effect transistors based
on (In,Mn)As [158] and (Ga,Mn)As [146, 148, 149, 159–161], (In,Fe)As
[162], as well as on GaAs δ-doped with Mn [163], the hole concentration can
be changed by a gate voltage (see also Chapter 13, Volume 2). A scheme of
the device is shown in Figure 9.5. A positive (negative) voltage, which repels
(attracts) the holes, leads to a reversible decrease (increase) of the magneti-
zation and of Tc, as expected for a hole-mediated interaction. A particularly
large change in Tc of 42% has been achieved by using a thin (In,Fe)As layer
within a broader InAs quantum well [162]. The effect of the electric field is
large since it tunes the overlap of the relevant electronic wave functions with
the DMS layer, while the carrier concentration is nearly unaffected, realizing
an earlier theoretical proposal [164]. A change of the magnetic anisotropy by
a gate voltage has also been demonstrated [159, 161]. This mechanism can be
used to electrically switch the magnetization direction. The reverse effect, i.e.
using a magnetic field close to Tc to change the magnetization, and thereby
Vg
gate
S D
( M,Mn)As
FIGURE 9.5 Field-effect transistor with (M,Mn)As, M = Ga,Mn, active region
(schematically). By applying a voltage Vg to the gate electrode, which is electrically
isolated from the (M,Mn)As channel, the hole concentration in the DMS can be
changed, leading to a change in the magnetic interactions and thus in the mag-
netization. Source (S) and drain (D) electrodes can be used to measure transport
properties of the gated DMS film.
the carrier concentration [165] has not yet been observed. In an nonuniform
magnetic field, this should lead to a voltage drop in the direction of the field
gradient [165].
Furthermore, Koshihara et al. [166] have shown that irradiation of origi-
nally paramagnetic (In,Mn)As/GaSb heterostructures with deep red to IR
light induces magnetic hysteresis. The results are easily understood in terms
of photoinduced holes transferred from the GaSb layer to the (In,Mn)As layer.
In ferromagnetic (Ga,Mn)As, Wang et al. [167] have observed an ultrafast
enhancement of ferromagnetism after irradiation. The light pulse is thought to
generate additional holes in the VB, which increase the magnetic interaction.
Measurements of the anomalous Hall effect in (Ga,Mn)As [15, 168]
also support the picture of carrier-mediated magnetic interactions, since
the magnetization inferred from the anomalous Hall effect agrees well with
the one obtained from direct magnetometer measurements (see Chapter 8,
Volume 2). In addition, experiments show a large magnetic anisotropy in
(In,Mn)As and (Ga,Mn)As films [11, 169–172], which for example leads to
reorientation transitions of the easy axis as a function of carrier concentra-
tion, temperature etc. In view of the tiny spin-orbit splitting [173] and the
resulting tiny magnetic anisotropy of the Mn d5 configuration, this would be
hard to understand based on direct exchange interactions.
The kinetic-exchange model predicts a Ruderman–Kittel–Kasuya–
Yosida (RKKY) interaction [174] between the Mn spins. The interaction is
2
essentially proportional to J pd c(r1 - r2 ), where c is the non-local magnetic
susceptibility of the carriers. The Stoner mechanism due to carrier–carrier
interactions leads to an enhancement of the susceptibility by on the order of
20% [3, 93, 94]. If the Zeeman splitting of the VB is small, as is the case close
to TC, it is sufficient to assume unpolarized bands. The RKKY interaction
calculated for realistic bands [68, 105, 175–177] is quite different from the
textbook expression for a parabolic band [174]. Using a tight-binding band
structure of GaAs, Timm and MacDonald [68] find an RKKY interaction
(a) (b)
FIGURE 9.6 Sketch of regions with ferromagnetic (pink) and antiferromagnetic
(blue) RKKY interaction with regard to a given impurity spin, denoted by the
arrow in the center. In (a) several neighboring impurity spins are ferromagnetically
coupled, favoring ferromagnetic order, whereas in (b) the nearest-neighbor inter-
action is essentially random in sign.
that is highly anisotropic, both in real space, and in spin space. The interac-
tion is ferromagnetic and large at small separations, and shows the charac-
teristic Friedel oscillations with period π/kF at larger separations, albeit with
anisotropic Fermi wavenumber kF. In compensated samples, typically, sev-
eral neighboring impurities lie within the first ferromagnetic maximum, as
sketched in Figure 9.6a, favoring ferromagnetic order [93, 105, 175, 176, 178].
Interactions at larger distances have random signs, but are small and aver-
age out for most impurity spins. However, new high-quality samples are only
weakly compensated so that the period of the Friedel oscillations is compa-
rable to the impurity separation, as shown in Figure 9.6b. In this case it might
be important that the first minimum in the (110) direction does not reach
negative—i.e. antiferromagnetic—values [68], e.g. in Figure 9.6b, the inner-
most antiferromagnetic (blue) region is absent in some directions. Similar
“avoided frustration” is also found in DFT calculations [155, 179].
Kitchen et al. [54] have performed STM experiments for Mn ions on
the (110) surface of Zn-doped p-type GaAs. The observed effective magnetic
interactions are also highly anisotropic in real space. They agree well with
tight-binding calculations using the bulk GaAs band structure [54].
The effective interaction between impurity spins has also been calcu-
lated within DFT from the total energy vs. Mn spin orientation and the
corresponding torques [155, 179, 180]. The DFT and model-based results
agree in that they show strong anisotropy in real space. DFT results typi-
cally lead to an overestimation of TC, while model-based theories underesti-
mate TC [68]. Probably contributing to this discrepancy are the neglect of the
Coulomb potential of the acceptors in the model calculations, the neglect
of spin-orbit coupling in the DFT results, and the unphysically large weight
of Mn d-orbitals close to the Fermi energy in DFT. Indeed, LDA+U usually
leads to a reduction of the predicted TC [63]. Disorder is also expected to play
a major role, as discussed below.
The conceptually simplest treatment of the kinetic-exchange model
involves two distinct mean-field-type approximations: First, the pd-exchange
interaction is decoupled at the mean-field (MF) level, and, second, the ran-
dom distribution of impurity spins is replaced by a smooth magnetization
(virtual-crystal approximation, VCA). Note that the same result for TC is
obtained by first calculating the RKKY interaction between impurity spins,
and then making the VCA/MF approximations [93]. The result reads [1, 93,
94, 178, 181, 182]
2
S(S + 1) J pd nMn cPauli
k BTC = , (9.5)
3 g 2m2B
where:
nMn is the concentration of Mn impurities
g is their g-factor
μB is the Bohr magneton
cPauli is the Pauli susceptibility of the carriers
2
For a single parabolic band, one obtains k BTC = S(S + 1) N ( E F ) J pd nMn / 6
[178], where N(EF ) is the density of states per spin direction at EF. This leads
to TC µ a0-6 p1/3 x 4/3 , where a0 is the lattice constant, x is the Mn doping level
related to the Mn concentration by nMn = 2.21 ´ 1022 cm -3 x , and p is the num-
ber of carriers per impurity. For increasing x, the Curie temperature should
thus increase without any saturation. An increase is indeed observed for
high-quality samples [1, 21, 22, 146–150]. The present model does not pre-
dict a reduction of TC close to vanishing compensation, p = 1, in contrast to
the IB picture.
Dietl et al. [101] have predicted the Curie temperatures of 5% Mn-doped
III–V and II–VI DMS based on the k × p Hamiltonian. The main trend is
a strong increase of TC for lighter elements [101], which can already be
understood from the parabolic-band model: Assuming that Jpd is approxi-
mately independent of the host semiconductor [101], the strong dependence
TC µ a0-6 favors high TC for hosts with small lattice constant a0. This depen-
dence stems, in equal parts, from the Mn concentration and from the elec-
tronic density of states by way of the Pauli susceptibility in Equation 9.5.
Reinforcing this trend, the electronic effective mass is typically larger for
lighter elements, further increasing the density of states.
Several routes have been taken to go beyond the MF approximation. We
here restrict ourselves to equilibrium properties, spin dynamics is discussed in
Chapter 1, Volume 2. A common approach is to first map the system onto a (Mn)
spin-only model [155, 177, 183]. While Gaussian fluctuations (non-interacting
spin waves) do not affect TC, higher-order fluctuations can appreciably reduce
it. This has been shown within the Tyablikov (random-phase) approximation
combined with a treatment of random spin positions in a supercell [177, 183],
called the “selfconsistent local random-phase approximation” (LRPA) [110,
177]. The results for TC(x) agree quantitatively with experiments for high-
quality, annealed (Ga,Mn)As.
Calculations of the spin-wave stiffness find a much higher energy of
spin waves for realistic band structures incorporating spin-orbit coupling
compared to a parabolic band [1, 108]. This is partly due to the splitting
between heavy-hole and light-hole bands, which can be understood as mag-
netic anisotropy of the total angular momentum j. The large stiffness sta-
bilizes ferromagnetism and is consistent with FMR experiments [184]. The
theoretical situation is not quite clear, though, since the spin-only model
mentioned above [177, 183] does not include spin-orbit coupling, but nev-
ertheless within the LRPA yields a stiffness in good agreement with experi-
ments [185]. In addition, a calculation going beyond RPA but neglecting
disorder [186] finds a strong enhancement of the spin-wave stiffness com-
pared to the RPA. The relative importance of spin-orbit coupling, disorder,
and higher-order many-particle corrections for the spin-wave stiffness is
thus not clear.
9.2.7 Disorder and Transport

The resistivity of high-TC (Ga,Mn)As generically shows a finite residual value
at low temperatures, sometimes with a small upturn for T ® 0 [20, 21, 27,
70, 86, 187, 188]. A finite limit of the resistivity is the defining property of a
metal. (Ga,Mn)As is metallic due to the heavy doping. In addition, there is a
broad and high peak in the resistivity in the vicinity of TC [17, 20, 21, 27, 70,
86, 131, 187, 188], also seen in (In,Mn)As [189]. In high-quality samples, its
slope on the T > TC side is relatively flat [17, 21, 86].
DMS are highly disordered, since the magnetic dopants and the com-
pensating defects are randomly distributed. Due to the strong interactions
between charged defects, they are expected to be incorporated in corre-
lated positions, likely as defect clusters, or to achieve such positions during
annealing [76, 82, 96, 190–196]. Random-telegraph noise in the charge trans-
port in (Ga,Mn)As [197] is interpreted in terms of the random switching of
the magnetic moments of small impurity clusters.
There are two main contributions to the disorder felt by the carriers.
First, the Mn acceptors attract the holes. In metallic DMS, the long-range
part of this attraction is screened by itinerant holes, but screening is not very
efficient due to the small hole concentration. The strong short-range attrac-
tion, which includes the central-cell correction, is not appreciably screened.
The acceptor potentials, together with the repulsive potentials from Mn
interstitials and antisites, form the Coulomb disorder potential. Clustering
of defects of course has a large effect on this potential [76, 195]. Second, the
pd-exchange interaction leads to disorder. Kyrychenko and Ullrich [196] have
found that dynamical spin fluctuations are important for transport. Most
theoretical approaches have nevertheless assumed the Mn spins to be frozen
on the time scale relevant for electronic transport. In this limit, they act like
a partially random Zeeman term, which is small compared to the Coulomb
disorder [1]. The metallic behavior of high-quality samples suggests that dis-
order can be treated perturbatively in the first-order Born approximation
[198, 199], which leads to a finite lifetime of quasiparticle states and to the
exponential decay of the electronic Green function on the length scale of
the elastic mean free path l. This causes a finite conductivity at T = 0. For
annealed (Ga,Mn)As, the product kF l, which is a measure for the strength
of disorder, is experimentally found to increase from k F l » 1 for x » 1.5% to
k F l » 3 - 5 for x » 7% [17, 132], consistent with a bad metal.
Going beyond the weak-disorder limit, disorder can lead to localized
electronic states [12, 153, 199–201]: Electrons close to band minima tend to
be trapped in the lowest depressions of the random potential landscape. The
analog holds for holes close to band maxima. If disorder is not too strong,
only states in the band tails are localized, whereas states in the band center
are extended, with mobility edges separating the two types (see Figures 9.3c
and d). If the states in the vicinity of EF are localized, conduction occurs
through thermal activation into extended states, leading to an Arrhenius
form of the conductivity, s µ exp(-DE / T ), where ΔE is the difference
between EF and the nearest mobility edge [202]. In (Ga,Mn)As, EF seems
to pass through a mobility edge for x » 1.5% [17, 31–33], corresponding to
an insulator-to-metal transition. Note that there is one conflicting report
of metallic conduction in an IB in (Ga,Mn)As with x » 0.3% [203]. A higher
dopant concentration is required to make the system metallic compared to
GaAs with shallow dopants, since the potential of a single Mn acceptor, and
consequently the disorder, are stronger [17].
Moca et al. [188] find that measurements of the magnetoresistivity of
(Ga,Mn)As with x  6.7% and small k F l  1 are reasonably well-fitted by
expressions derived from the scaling theory of localization [200]. The com-
parison shows that the drop in resistivity below TC is strongly correlated with
the magnetization and suggests that the upturn of the resistivity at low T
[20, 21, 70, 86, 187, 188] results from disorder scattering. STM and scanning
tunneling spectroscopy experiments [33] exhibit a rich spatial structure
of the density of states at the surface of (Ga,Mn)As close to the metal-
insulator transition. The correlations of the density of states at EF, but not
at other energies, decay as a power law with distance [33]. If this power-law
behavior is indeed pinned to EF and not to the mobility edge, this indicates
that the electron–electron interaction plays a crucial role [33]. It is interest-
ing in this context that photoemission experiments also suggest strong cor-
relation effects [3, 123]. How the pronounced spatial dependence [33] can be
reconciled with the homogeneous ferromagnetism observed by muon spin
relaxation (μSR) [204] is not fully clear.
As noted above, the question of localization of states is separate from
the question of the survival of the IB [12]. In particular, an insulator-to-metal
transition has been observed in p-doped GaAs quantum wells with EF in an
IB [205]. Metallic conduction in an IB is also proposed for (Ga,Mn)As with
x » 0.3% [203], contradicting other transport measurements [31, 32]. A pic-
ture of localized and extended states within an IB, as sketched in Figure 9.3c,
is corroborated by exact-diagonalization studies [105, 106]. At present, it is
unclear whether such a scenario pertains to any III–V DMS.
In the regime of weak doping, where an IB exists and the states at EF are
localized, it is reasonable to start from a picture of holes hopping between
bound acceptor states [105, 106, 191, 206–210]. The bound-hole spin is
exchange-coupled to one or more impurity spins, forming a hole dressed with
a spin-polarization cloud, called a bound magnetic polaron (BMP) [207, 208].

The exchange coupling to the same hole spins leads to an effective ferro-
magnetic coupling of the impurity spins in the BMP. The confinement of
carriers in DMS quantum dots increases the tendency of BMP formation
(see Chapter 6, Volume 3).
In metallic DMS, where strong localization is not relevant, disorder can
still lead to subtle effects, in particular in reduced dimensions. Closed mul-
tiple-scattering paths such as the ones sketched in Figure 9.7 are important
here: in the absence of a magnetic field and of spin-orbit coupling, the wave
functions of carriers traversing a closed loop in opposite directions accumu-
late the same quantum phase, and therefore interfere constructively. This
increases the probability for carriers to return to the same point and, con-
sequently, decreases the conductivity. This weak localization [153, 198, 199]
leads to a logarithmic increase in the resistivity for T ® 0, as observed in
(Ga,Mn)As nanowires [211].
If a magnetic field is applied, the quantum phases accumulated for the
two directions become different by an amount proportional to the magnetic
flux Φ enclosed by the loop (see Figure 9.7). For a single loop, the conductivity
shows Aharonov–Bohm oscillations as a function of Φ. In typical nanoscopic
samples, loops of different sizes contribute, which have different oscillation
periods. This leads to a characteristic quasi-random dependence of the con-
ductivity on magnetic field with universal amplitude (universal conductance
fluctuations) [212]. Both the Aharonov–Bohm effect and universal conduc-
tance fluctuations have been observed in nanoscopic (In,Mn)As [189] and
(Ga,Mn)As [211, 213, 214] devices. The reason why these effects can occur at
all is that DMS are both diluted ferromagnets and bad metals, as discussed
by Garate et al. [215]. In normal metallic ferromagnets, they are suppressed
by the stronger internal magnetic field. In larger samples, so many loops of
different sizes contribute that the net effect is the suppression of weak local-
ization by a magnetic field, leading to negative magnetoresistance. Negative
FIGURE 9.7 Scattering paths traversed in opposite directions. The black dots
represent random scatterers. The constructive interference of the wave functions
of carriers following the solid and dashed paths leads to weak localization. A mag-
netic flux Φ threading the loop leads to a phase difference between the two paths.
magnetoresistance has been attributed to weak localization for (Ga,Mn)

As films with up to x » 8% Mn and non-optimal TC [216]. The effect was
observed at low temperatures, T  3K , and in weak magnetic fields on the
order of 20 mT [216].
Since spin-orbit coupling in (Ga,Mn)As is strong, its effect on weak
localization has to be taken into account. For a parabolic band with strong
spin-orbit coupling one expects positive magnetoresistance (weak antilo-
calization) [199, 215]. Rokhinson et al. [216] argue that their observation of
negative magnetoresistance in (Ga,Mn)As with x » 8% therefore indicates
that EF is not located in the VB, but in an IB. However, it is unclear why
spin-orbit effects should be significantly smaller in an IB derived from VB
states. Garate et al. [215] show that within a k × p approach, the complex
band structure leads to negative magnetoresistance and weak localization, as
observed. Weak antilocalization has also been invoked to explain the broad
resistivity peak around TC [217]. However, it is now thought that the peak
maximum is located a few Kelvin above TC, and that the peak is mostly due to
temperature-dependent changes in the band structure and the scattering
rate [1, 188]. On the other hand, a singularity in the temperature deriva-
tive of the resistivity, dρ/dT, has been observed precisely at TC [21, 24]. This
singularity is coupled to the one observed in the specific heat [218] and is
consistent with the Fisher–Langer theory [219] for the resistive anomaly in
metallic ferromagnets [4]. Deviations from Fisher–Langer theory are more
likely to be observed if the resistivity at TC is dominated by disorder scatter-
ing [217].
9.2.8 Disorder and Magnetism

Disorder affects the magnetic properties in two principal ways. First, the
carrier-mediated magnetic interaction is sensitive to carrier scattering.
Second, the Mn local moments occupy random positions in the DMS lattice.
We will discuss these mechanisms in turn.
The Curie temperature of (Ga,Mn)As decreases with decreasing x, but
remains non-zero in the doping range 1%  x  1.5% , where the material is
insulating at low temperatures. It has been suggested that an RKKY descrip-
tion remains valid in the presence of strong localization [93, 101–103].
In this regime, the RKKY interaction between two typical impurity spins
decays exponentially on the scale of the localization length [102, 103].
Ferromagnetic order is still possible, as long as the localization length is
larger than the separation between impurity spins. While the RKKY para-
digm might hold in this regime, disorder effects are very strong, precluding
a VCA/MF description. Indeed, Bouzerar et al. [109, 110, 220] find that the
RKKY interaction calculated including electronic disorder in the coherent-
potential approximation strongly deviates from the clean case and leads to
enhanced TC. Myers et al. [72] have measured the dependence of TC on the
number p of holes per Mn in this regime by systematically varying the As:Ga
ratio across their samples during growth. They indeed find a strong deviation
from the prediction TC µ p1/3 of VCA/MF theory [72].
In this regime, a BMP picture may be more appropriate [105, 106, 191,
206–209]. For strong compensation, a single BMP comprises a hole and sev-
eral ferromagnetically aligned impurity spins. For larger impurity concen-
trations, the BMPs overlap, leading to a ferromagnetic interaction between
impurity spins within the same cluster of overlapping BMPs. At zero tem-
perature, ferromagnetic long-range order emerges if the BMPs percolate. At
finite temperatures, thermal fluctuations destroy the alignment between
weakly coupled clusters. Kaminski and Das Sarma [207] have calculated TC
based on this picture. A more recent theory based on variable-range hopping
[209] is similar in spirit. Above the percolation transition at TC, rare but large
ordered regions have been predicted to lead to Griffiths singularities [221],
which would result in characteristic scaling behavior of the susceptibility
and specific heat. Such ordered regions have been observed in (In,Mn)Sb
[222], but Griffiths singularities have not yet been found in III–V DMS.
Carrier scattering has another important consequence: One cannot
simply grow quaternary DMS such as (Ga,Mn)(As,P) and hope to achieve
properties interpolating between the end points. The alloying introduces
additional disorder, which dramatically reduces the mean free path and
TC [125].
We now turn to the effects of disorder due to the random distribution
of magnetic dopants. In the extreme case, certain regions of the sample may
be effectively decoupled from the rest and thus not take part in the sponta-
nous magnetization. There is evidence that such regions are present even
in high-quality samples [148], whereas a recent study using μSR as the local
probe found homogeneous ferromagnetism in the metallic regime and even
through the metal-insulator transition [204]. A random distribution can lead
to non-collinear ground states, since the RKKY interaction changes sign as
a function of separation, which leads to frustration, and is anisotropic both
in real space and in spin space [10, 68, 105, 175, 176, 223, 224]. The resulting
reduction of the low-temperature magnetization has been estimated to be
small [1, 68, 225].
In various theoretical approaches, disorder is generically found to lead
to an anomalous temperature dependence of the magnetization, character-
ized by an upward curvature or nearly linear behavior in a broad intermedi-
ate temperature range [103, 117, 178, 191, 206, 224, 226–228]. Magnetization
curves of this type have been observed for insulating (Ga,Mn)As [229–231]
and also for early metallic samples [168, 229, 230]. Note that an upward cur-
vature can also result from a temperature-induced change of the easy axis
[1]. New high-quality samples show normal, Brillouin function-like magne-
tization curves [21].
Simultaneously treating both effects of disorder—the carrier scat-
tering and the spatial disorder of spins—in a realistic calculation is com-
putationally demanding. Extending earlier work [191, 232, 233], Yildirim
et al. [117] perform large-scale Monte Carlo simulations of (Ga,Mn)As,
where the local spins are treated classically, and an electronic tight-bind-
ing Hamiltonian is diagonalized for each spin configuration. For reason-
ably large x and weak compensation, the magnetization curves are fairly
Brillouin function-like, in agreement with experiments [21]. However, to

achieve realistic values of TC, Jpd had to be assumed to be much smaller
than usually thought [234]. Moreover, the Coulomb disorder potential was
neglected, although it is likely the larger contribution to the disorder seen
by the holes [1]. Its inclusion would probably reduce TC, counteracting the
effect of a larger Jpd.
The disorder due to the random distribution of spins has also been stud-
ied in spin-only models within the LRPA [110, 177, 183], which includes disor-
der by employing a large supercell. As noted above, the results for TC(x) are in
good agreement with experiments. Disorder is crucial for this. Remarkably,
TC(x) calculated based on a spin-only model derived from a V-J model using
a simple, unrealistic electronic band agrees well with DFT calculations [110].
This suggests that details of the band structure are not crucial for certain
magnetic properties. However, this is not the case for properties relying on
anisotropic exchange.
9.3 III–V DMS WITH WIDE GAPS

Wide-gap III–V DMS such as (Ga,Mn)N are interesting, since for several
of these compounds Curie temperatures above room temperature have
been reported [235–244]. Dietl [103, 147] has discussed several scenarios
for apparent ferromagnetic response of DMS. For quasi-uniform DMS with
merged VB and IB, the kinetic-exchange model can be applied, leading to
an RKKY-type magnetic interaction. Indeed, both the arguments based on
a parabolic band given in Section 9.2.6 and more sophisticated k × p theory
predict high TC values [101]. Deep Mn-derived levels in the gap likely invali-
date the underlying kinetic-exchange model for wide-gap DMS, though.
Ferromagnetic response can also result from precipitates of known mag-
netic compounds [103, 147]. In this case, the semiconducting and magnetic
properties result from different parts of the sample. A more interesting pos-
sibility is chemical nanoscale phase separation into regions rich and poor in
magnetic ions [103, 147]. In contrast to precipitates, the crystal structure of
the rich regions is in this case determined by the host, which can stabilize
phases that are unstable in the bulk [103, 147]. Nanoscale phase separation
has been observed in (Ga,Mn)N [245] and (Ga,Fe)N [242] as well as in the
intermediate-gap DMS (Ga,Mn)As [246]. Note that magnetic and trans-
port properties can still be strongly coupled in such systems. In both cases
of phase separation, an apparently ferromagnetic response is seen below
a blocking temperature TB: For T < TB, the thermal energy is too small to
overcome the size-dependent magnetic anisotropy energy of the magnetic
regions, which therefore show frozen magnetic moments. This superpara-
magnetic behavior has been observed for Gd-ion-implanted GaN [247] and
for (Ga,Fe)N [242]. Ney et al. [248] have pointed out that a large fraction
of the apparently ferromagnetic signals could result from metastable states.
Careful analysis is required to check that static, quasi-uniform ferromagne-
tism is actually present in a wide-gap DMS. In the following, we discuss such
a still partially hypothetical state.
We concentrate on (Ga,Mn)N. Like in (Ga,Mn)As, substitutional Mn2+

introduces one hole, which is attracted by the Mn2+ core. Neglecting hybrid-
ization, one would find a hydrogenic acceptor state. However, the crucial dif-
ference between (Ga,Mn)N and (Ga,Mn)As is that in the nitride, the d 5 ® d 4
(Mn 2 + ® Mn 3 + ) transition lies in the band gap [65, 66, 249]. It is probably
deeper than the bound acceptor state, becoming a d 4 ® d 5 transition. Due
to the smaller energy difference, the effect of hybridization, and, in particu-
lar, the pd-exchange interaction Jpd are larger, as seen in photoemission as
well as in optical and x-ray spectroscopy for (Ga,Mn)N [250, 251]. Jpd may
even become ferromagnetic [252, 253]. Hybridization and level repulsion
lead to an unoccupied level with more than 50% d-orbital character deep in
the gap, and an occupied level with more than 50% VB character, which may
occur as a resonance in the VB.
The strong pd-hybridization agrees with LDA+U and LSDA+U calcula-
tions [63, 64, 254], which find large Mn d-orbital weight at the Fermi energy
in (Ga,Mn)N unlike in (Ga,Mn)As [63, 64]. Employing self-interaction cor-
rected LSDA, Schulthess et al. [65, 66] also find a large Mn d-orbital weight
in the gap, close to EF, and a predominantly d4 state of the Mn dopants for
both (Ga,Mn)N and (Ga,Mn)P, but not for (Ga,Mn)As. This picture is sup-
ported by optical spectroscopy [255], x-ray spectroscopy [250], and transport
and magnetization measurements [256]. However, in other x-ray spectros-
copy experiments [257], (Ga,Mn)P appears to be more similar to (Ga,Mn)
As, suggesting that its properties are intermediate between (Ga,Mn)As and
(Ga,Mn)N.
The overlap between the strongly d-orbital-like states in the gap is small.
The IB formed for given Mn doping x is thus narrower than in (Ga,Mn)As.
Furthermore, it is much deeper in the gap. Therefore, a significant overlap,
let alone merging, of IB and VB is unlikely [258]. The density of states is
sketched in Figure 9.8. Dynamical mean-field theory calculations assuming
local Coulomb and exchange interactions between holes and Mn impurities
support this view [60]. The kinetic-exchange picture with the Fermi energy
in a weakly perturbed VB is thus not applicable Figure 9.8.
N ( E)
IB
VB CB
EF E
FIGURE 9.8 Sketch of the density of states for a typical wide-gap III–V DMS.
Compared to Figure 9.3c, the IB is both narrower and deeper in the gap. The Fermi
energy is shown for partial compensation.
The high observed TC [235, 240, 241] at concentrations of magnetic dop-

ants below the percolation threshold suggests a carrier-mediated mecha-
nism. The suppression of the remanent magnetization in (Ga,Mn)N below
about 10 K, accompanied by an increase of the resistivity attributed to
localization supports a carrier-mediated scenario [241]. This picture is also
favored by the observation that exposure to atomic hydrogen has a strong
effect on the magnetic order [259], presumably by changing the compensa-
tion. Since in the absence of compensation, the narrow IB would be com-
pletely empty, a carrier-mediated mechanism requires partial compensation
[65]. A high concentration of nitrogen vacancies, which are donors with one
weakly bound electron [260], is typical for GaN and thus provides a likely
source of compensation. Transport and x-ray spectroscopy experiments
[261] on (Ga,Mn)N with x  1%, grown by metal-organic chemical vapor
deposition, suggest a strong increase in the concentration of nitrogen vacan-
cies with x (self-compensation). Due to the large d-orbital weight in the IB,
strong compensation leads to strong charge fluctuations in the Mn d-shell.
This mixed-valence nature of Mn might lead to interactions through the
double-exchange mechanism [64, 154, 240, 241, 254, 262], which can lead to
high TC for large x.
Dhar et al. have observed ferromagnetic order in MBE-grown [237]
and ion-implanted [239] GaN with very low concentrations of Gd, nGd,
down to 7 ´ 1015 cm -3 . Dividing the measured magnetization by nGd, they
obtain magnetic moments per Gd of up to 5000 μB. This, of course,
implies that most of the magnetization is not carried by the Gd dopants.
Giant moments per Gd have been confirmed for Gd-implanted GaN by
Wang et al. [244]. The natural candidates are defects [263]. This interpre-
tation is supported by the observations that (a) similar magnetic proper-
ties are found if the bare sapphire substrate or Si wavers are implanted
with Gd, showing that the effect is not material specific [244], and (b) the
magnetic signal vanishes if the implanted sample is kept at room tem-
perature for a few weeks, presumably due to removing part of the defects
by annealing [244]. Since the concentrations of unwanted impurities are
small [237], native defects, most likely nitrogen vacancies [260], are prob-
ably responsible. Electrons weakly bound to nitrogen-vacancy donors
[260] would each contribute a spin 1/2. Indeed, the total magnetization is
found to depend only weakly on nGd for small concentrations, followed by
a crossover to a roughly linear behavior for larger nGd with a coefficient
consistent with the f 7 configuration of Gd3+ [237]. However, it is not yet
understood how very weak Gd doping could induce magnetic ordering
of the defect spins. It has been found that in some Gd-doped GaN sam-
ples the magnetic effects are metastable [248], perhaps due to electrons
trapped by defects. Magnetic moments of defects have also been invoked
to explain magnetic signals in other semiconductors, including samples
without transition-metal or rare-earth dopants [264–267]. This so-called
d0 magnetism [264], which is particularly relevant for oxides, is discussed
in Chapter 10, Volume 2.
REFERENCES
1. T. Jungwirth, J. Sinova, J. Mašek, J. Kučera, and A. H. MacDonald, Theory of
ferromagnetic (III,Mn)V semiconductors, Rev. Mod. Phys. 78, 809–864 (2006).
2. T. Dietl, A ten-year perspective on dilute magnetic semiconductors and oxides,
Nat. Mater. 9, 965–974 (2010).
tronic structures, Rev. Mod. Phys. 86, 187–251 (2014).
4. T. Jungwirth, J. Wunderlich, V. Novák et al., Spin-dependent phenomena and
device concepts explored in (Ga,Mn)As, Rev. Mod. Phys. 86, 855–896 (2014).
5. T. Jungwirth, III–V based magnetic semiconductors, Handbook of Spintronics:
Part V., Y. Xu, D. D. Awschalom, and J. Nitta (Eds), Springer, Dordrecht, pp.
465–521, (2016).
6. K. Sato, L. Bergqvist, J. Kudrnovský et al., First-principles theory of dilute mag-
netic semiconductors, Rev. Mod. Phys. 82, 1633–1690 (2010).
7. A. Zunger, S. Lany, and H. Raebiger, Trend: The quest for dilute ferromagne-
tism in semiconductors: Guides and misguides by theory, Physics 3, 53 (2010).
8. S. J. Pearton, C. R. Abernathy, M. E. Overberg et al., Wide band gap ferromag-
netic semiconductors and oxides, J. Appl. Phys. 93, 1–12 (2003).
9. C. Liu, F. Yun, and H. Morkoç, Ferromagnetism of ZnO and GaN: A review,
J. Mater. Sci. Mater. Electron. 16, 555–597 (2005).
10. C. Timm, Disorder effects in diluted magnetic semiconductors, J. Phys.
Condens. Matter 15, R1865–R1896 (2003).
11. X. Liu and J. K. Furdyna, Ferromagnetic resonance in Ga1-xMn x As dilute mag-
netic semiconductors, J. Phys. Condens. Matter 18, R245–R279 (2006).
12. T. Dietl, Interplay between carrier localization and magnetism in diluted
magnetic and ferromagnetic semiconductors, J. Phys. Soc. Jpn. 77, 031005
(2008).
13. K. S. Burch, D. D. Awschalom, and D. N. Basov, Optical properties of III–
Mn–V ferromagnetic semiconductors, J. Magn. Magn. Mater. 320, 3207–3228
(2008).
14. A. Bonanni and T. Dietl, A story of high-temperature ferromagnetism in semi-
conductors, Chem. Soc. Rev. 39, 528–539 (2010).
hall effect, Rev. Mod. Phys. 82, 1539–1592 (2010).
16. T. Dietl, K. Sato, T. Fukushima et al., Spinodal nanodecomposition in semicon-
ductors doped with transition metals, Rev. Mod. Phys. 87, 1311–1377 (2015).
17. T. Jungwirth, J. Sinova, A. H. MacDonald et al., Character of states near the
Fermi level in (Ga,Mn)As: Impurity to valence band crossover, Phys. Rev. B 76,
125206 (2007).
18. H. Ohno, H. Munekata, T. Penney, S. von Molnár, and L. L. Chang,
Magnetotransport properties of p-type (In,Mn)As diluted magnetic III–V
semiconductors, Phys. Rev. Lett. 68, 2664–2667 (1992).
19. H. Ohno, A. Shen, F. Matsukura et al., (Ga,Mn)As: A new diluted magnetic
semiconductor based on GaAs, Appl. Phys. Lett. 69, 363–365 (1996).
20. F. Matsukura, H. Ohno, A. Shen, Y. Sugawara, Transport properties and origin
of ferromagnetism in (Ga,Mn)As, Phys. Rev. B 57, R2037–R2040 (1998).
21. V. Novák, K. Olejnk, J. Wunderlich et al., Curie point singularity in the tem-
perature derivative of resistivity in (Ga,Mn)As, Phys. Rev. Lett. 101, 077201
(2008).
22. M. Wang, R. P. Campion, A. W. Rushforth, K. W. Edmonds, C. T. Foxon, and
B. L. Gallagher, Achieving high curie temperature in (Ga,Mn)As, Appl. Phys.
Lett. 93, 132103 (2008).
23. L. Chen, S. Yan, P. F. Xu et al., Low-temperature magnetotransport behav-
iors of heavily Mn-doped (Ga,Mn)As films with high ferromagnetic transition
temperature, Appl. Phys. Lett. 95, 182505 (2009).
References 399
24. P. Nĕmec, V. Novák, N. Tesařová et al., The essential role of carefully opti-
mized synthesis for elucidating intrinsic material properties of (Ga,Mn)As,
Nat. Commun. 4, 1422 (2013).
25. B. C. Chapler, R. C. Myers, S. Mack et al., Infrared probe of the insulator-to-
metal transition in Ga1-xMn x As and Ga1-x Bex As, Phys. Rev. B 84, 081203(R)
(2011).
26. A. M. Nazmul, T. Amemiya, Y. Shuto, S. Sugahara, and M. Tanaka, High tem-
perature ferromagnetism in GaAs-based heterostructures with Mn δ-Doping,
Phys. Rev. Lett. 95, 017201 (2005).
27. M. A. Scarpulla, R. Farshchi, P. R. Stone et al., Electrical transport and fer-
romagnetism in Ga1-xMn x As synthesized by ion implantation and pulsed-laser
melting, J. Appl. Phys. 103, 073913 (2008).
28. T. Schallenberg and H. Munekata, Preparation of ferromagnetic (In,Mn)As with
a high Curie temperature of 90 K, Appl. Phys. Lett. 89, 042507 (2006).
29. N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, High-temperature ferromagne-
tism in heavily Fe-doped ferromagnetic semiconductor (Ga,Fe)Sb, Appl. Phys.
Lett. 108, 192401 (2016).
30. N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, High-temperature ferromag-
netism in new n-type Fe-doped ferromagnetic semiconductor (In,Fe)Sb, Appl.
Phys. Express 11, 063005 (2018).
31. S. J. Potashnik, K. C. Ku, R. Mahendiran et al., Saturated ferromagnetism and
magnetization deficit in optimally annealed Ga1-xMn x As epilayers, Phys. Rev. B
66, 012408 (2002).
32. R. P. Campion, K. W. Edmonds, L. X. Zhao et al., The growth of GaMnAs films
by molecular beam epitaxy using arsenic dimers, J. Cryst. Growth 251, 311–316
(2003).
33. A. Richardella, P. Roushan, S. Mack et al., Visualizing critical correlations near
the metal-insulator transition in Ga1-xMn x As, Science 327, 665–669 (2010).
34. E. Abe, F. Matsukura, H. Yasuda, Y. Ohno, and H. Ohno, Molecular beam epi-
taxy of III–V diluted magnetic semiconductor (Ga,Mn)Sb, Phys. E 7, 981–985
(2000).
35. T. Wojtowicz, G. Cywi n’ski, W. L. Lim et al., In1-xMn xSb—a narrow-gap fer-
romagnetic semiconductor, Appl. Phys. Lett. 82, 4310–4312 (2003).
36. M. Csontos, G. Mihály, B. Jankó, T. Wojtowicz, X. Liu, and J. K. Furdyna,
Pressure-induced ferromagnetism in (In,Mn)Sb dilute magnetic semiconductor,
Nat. Mater. 4, 447–449 (2005); G. Mihály, M. Csontos, S. Bordács et al.,
Anomalous hall effect in the (In,Mn)Sb dilute magnetic semiconductor, Phys.
Rev. Lett. 100, 107201 (2008).
37. M. Khalid, E. Weschke, W. Skorupa, M. Helm, and S. Zhou, Ferromagnetism
and impurity band in a magnetic semiconductor: InMnP, Phys. Rev. B 89,
121301(R) (2014); M. Khalid, K. Gao, E. Weschke et al., A comprehensive study
of the magnetic, structural, and transport properties of the III–V ferromagnetic
semiconductor InMnP, J. Appl. Phys. 117, 043906 (2015).
38. R. Bouzerar, D. May, U. Löw, D. Machon, P. Melinon, and G. Bouzerar,
Effective Spin Models and Critical Temperatures for Diluted Magnetic
Semiconductors, Verhandlungen der DPG, Frühjahrstagung Regensburg, HL
96.3 (2016).
39. P. M. Krstajič, F. M. Peeters, V. A. Ivanov, V. Fleurov, and K. Kikoin, Double-
exchange mechanisms for Mn-doped III–V ferromagnetic semiconductors,
Phys. Rev. B 70, 195215 (2004).
40. N. Almeleh and B. Goldstein, Electron paramagnetic resonance of manganese
in gallium arsenide, Phys. Rev. 128, 1568–1570 (1962).
41. J. Szczytko, A. Twardowski, K. Świątek et al., Mn impurity in Ga1-xMn x As epi-
layers, Phys. Rev. B 60, 8304–8308 (1999).
42. Y. Sasaki, X. Liu, J. K. Furdyna, M. Palczewska, J. Szczytko, and A. Twardowski,
Ferromagnetic resonance in GaMnAs, J. Appl. Phys. 91, 7484–7486 (2002).
43. R. A. Chapman and W. G. Hutchinson, Photoexcitation and photoionization of

neutral manganese acceptors in gallium arsenide, Phys. Rev. Lett. 18, 443–445
(1967).
44. M. Linnarsson, E. Janzén, B. Monemar, M. Kleverman, and A. Thilderkvist,
Electronic structure of the GaAs:MnGa center, Phys. Rev. B 55, 6938–6944
(1997).
45. J. Schneider, U. Kaufmann, W. Wilkening, M. Baeumler, and F. Köhl, Electronic
structure of the neutral manganese acceptor in gallium arsenide, Phys. Rev.
Lett. 59, 240–243 (1987).
46. J. S. Blakemore, W. J. Brown Jr., M. L. Stass, and D. A. Woodbury, Thermal
activation energy of manganese acceptors in gallium arsenide as a function of
impurity spacing, J. Appl. Phys. 44, 3352–3354 (1973).
47. A. M. Yakunin, A. Yu. Silov, P. M. Koenraad, W. Van Roy, J. De Boeck, and J. H.
Wolter, Imaging of the (Mn2+3d5 + hole) complex in GaAs by cross-sectional
scanning tunneling microscopy, Phys. E 21, 947–950 (2004); A. M. Yakunin, A.
Yu. Silov, P. M. Koenraad et al., Spatial structure of an individual Mn acceptor
in GaAs, Phys. Rev. Lett. 92, 216806 (2004).
48. J. Okabayashi, A. Kimura, O. Rader et al., Core-level photoemission study of
Ga1-xMn x As, Phys. Rev. B 58, R4211–R4214 (1998).
49. J. Okabayashi, A. Kimura, T. Mizokawa, A. Fujimori, T. Hayashi, and M.
Tanaka, Mn 3d partial density of states in Ga1-xMn x As studied by resonant
photoemission spectroscopy, Phys. Rev. B 59, R2486–R2489 (1999).
50. A. Baldereschi and N. O. Lipari, Spherical model of shallow acceptor states in
51. A. K. Bhattacharjee and C. B. à la Guillaume, Model for the Mn acceptor in
GaAs, Solid State Commun. 113, 17–21 (2000).
52. R. F. Kirkman, R. A. Stradling, and P. J. Lin-Chung, An infrared study of the
shallow acceptor states in GaAs, J. Phys. C 11, 419–433 (1978).
53. M. S. Dresselhaus, G. Dresselhaus, and A. Jorio, Group Theory: Application to
the Physics of Condensed Matter, Springer, Berlin, 2008.
54. D. Kitchen, A. Richardella, J.-M. Tang, M. E. Flatté, and A. Yazdani, Atom-
by-atom substitution of Mn in GaAs and visualization of their hole-mediated
interactions, Nature 442, 436–439 (2006).
55. C. Çelebi, J. K. Garleff, A. Yu. Silov et al., Surface induced asymmetry of accep-
tor wave functions, Phys. Rev. Lett. 104, 086404 (2010).
56. J.-M. Tang and M. E. Flatté, Multiband tight-binding model of local magne-
tism in Ga1-xMn x As, Phys. Rev. Lett. 92, 047201 (2004).
57. M. J. Schmidt, K. Pappert, C. Gould, G. Schmidt, R. Oppermann, and L. W.
Molenkamp, Bound-hole states in a ferromagnetic (Ga,Mn)As environment,
Phys. Rev. B 76, 035204 (2007).
58. A. Zunger in H. Ehrenreich and D. Turnbull (eds), Solid State Physics, Vol. 39.
Orlando: Academic (1986), p. 275.
59. P. Mahadevan and A. Zunger, Ferromagnetism in Mn-doped GaAs due to
substitutional-interstitial complexes, Phys. Rev. B 68, 075202 (2003).
60. F. Popescu, C. Şen, E. Dagotto, and A. Moreo, Crossover from impurity to
valence band in diluted magnetic semiconductors: Role of Coulomb attraction
by acceptors, Phys. Rev. B 76, 085206 (2007).
61. Y. Yuan, C. Xu, R. Hübner et al., Interplay between localization and magnetism
in (Ga,Mn)As and (In,Mn)As, Phys. Rev. Mater. 1, 054401 (2017).
62. L. Craco, M. S. Laad, and E. Müller-Hartmann, Ab initio description of the
diluted magnetic semiconductor Ga1-xMn x As: Ferromagnetism, electronic
structure, and optical response, Phys. Rev. B 68, 233310 (2003).
63. L. M. Sandratskii, P. Bruno, and J. Kudrnovský, On-site Coulomb interac-
tion and the magnetism of (GaMn)N and (GaMn)As, Phys. Rev. B 69, 195203
(2004).
64. M. Wierzbowska, D. Sánchez-Portal, and S. Sanvito, Different origins of the fer-
romagnetic order in (Ga,Mn)As and (Ga,Mn)N, Phys. Rev. B 70, 235209 (2004).
References 401
65. T. C. Schulthess, W. M. Temmerman, Z. Szotek, W. H. Butler, and G. M.

Stocks, Electronic structure and exchange coupling of Mn impurities in III–V
semiconductors, Nat. Mater. 4, 838–844 (2005).
66. T. C. Schulthess, W. M. Temmerman, Z. Szotek, A. Svane, and L. Petit, First-
principles electronic structure of Mn-doped GaAs, GaP, and GaN semicon-
ductors, J. Phys. Condens. Matter 19, 165207 (2007).
67. J. R. Schrieffer and P. A. Wolff, Relation between the Anderson and Kondo
Hamiltonians, Phys. Rev. 149, 491–492 (1966); K. A. Chao, J. Spałek, and
A. M. Oleś, Canonical perturbation expansion of the Hubbard model, Phys.
Rev. B 18, 3453–3464 (1978); B. E. Larson, K. C. Hass, H. Ehrenreich, and A.
E. Carlsson, Theory of exchange interactions and chemical trends in diluted
magnetic semiconductors, ibid. 37, 4137–4154 (1988).
68. C. Timm and A. H. MacDonald, Anisotropic exchange interactions in III–V
diluted magnetic semiconductors, Phys. Rev. B 71, 155206 (2005).
69. E. Nielsen and R. N. Bhatt, Search for ferromagnetism in doped semiconduc-
tors in the absence of transition metal ions, Phys. Rev. B 82, 195117 (2010).
70. S. J. Potashnik, K. C. Ku, S. H. Chun, J. J. Berry, N. Samarth, and P. Schiffer,
Effects of annealing time on defect-controlled ferromagnetism in Ga1-xMn x As,
Appl. Phys. Lett. 79, 1495–1497 (2001).
71. M. Luysberg, H. Sohn, A. Prasad et al., Electrical and structural properties of
LT-GaAs: Influence of As/Ga flux ratio and growth temperature, Mater. Res.
Soc. Symp. Proc. 442, 485–490 (1997).
72. R. C. Myers, B. L. Sheu, A. W. Jackson et al., Antisite effect on hole-mediated
ferromagnetism in (Ga,Mn)As, Phys. Rev. B 74, 155203 (2006).
73. R. P. Campion, K. W. Edmonds, L. X. Zhao et al., High-quality GaMnAs films
grown with arsenic dimers, J. Cryst. Growth 247, 42–48 (2003).
74. B. Tuck, Atomic Diffusion in III–V Semiconductors, IOP Publishing, Bristol,
1988.
75. M. Stellmacher, R. Bisaro, P. Galtier, J. Nagle, K. Khirouni, and J. C. Bourgoin,
Defects and defect behaviour in GaAs grown at low temperature, Semicond.
Sci. Technol. 16, 440–446 (2001).
76. C. Timm, F. Schäfer, and F. von Oppen, Correlated Defects, Metal-Insulator
Transition, and Magnetic order in ferromagnetic semiconductors, Phys. Rev.
Lett. 89, 137201 (2002).
77. K. M. Yu, W. Walukiewicz, T. Wojtowicz et al., Effect of the location of Mn
sites in ferromagnetic Ga1-xMn x As on its Curie temperature, Phys. Rev. B 65,
201303(R) (2002).
78. K. W. Edmonds, P. Bogusławski, K. Y. Wang et al., Mn interstitial diffusion in
(Ga,Mn)As, Phys. Rev. Lett. 92, 037201 (2004).
79. J. Mašek and F. Máca, Interstitial Mn in (Ga,Mn)As: Binding energy and
exchange coupling, Phys. Rev. B 69, 165212 (2004).
80. V. Holý, Z. Matĕj, O. Pacherová et al., Mn incorporation in as-grown and
annealed (Ga,Mn)As layers studied by x-ray diffraction and standing-wave
fluorescence, Phys. Rev. B 74, 245205 (2006).
81. G. S. Chang, E. Z. Kurmaev, L. D. Finkelstein et al., Post-annealing effect on the
electronic structure of Mn atoms in Ga1-xMn x As probed by resonant inelastic
x-ray scattering, J. Phys. Condens. Matter 19, 076215 (2007).
82. J. Blinowski and P. Kacman, Spin interactions of interstitial Mn ions in ferro-
magnetic GaMnAs, Phys. Rev. B 67, 121204 (2003).
83. M. B. Stone, K. C. Ku, S. J. Potashnik, B. L. Sheu, N. Samarth, and P. Schiffer,
Capping-induced suppression of annealing effects on Ga1-xMn x As epilayers,
Appl. Phys. Lett. 83, 4568–4570 (2003).
84. J. Adell, I. Ulfat, L. Ilver, J. Sadowski, K. Karlsson, and J. Kanski, Thermal diffu-
sion of Mn through GaAs overlayers on (Ga,Mn)As, J. Phys. Condens. Matter
23, 085003 (2011).
85. L. Horák, J. Matĕjová, X. Mart et al., Diffusion of Mn interstitials in (Ga,Mn)
As epitaxial layers, Phys. Rev. B 83, 245209 (2011).
86. K. W. Edmonds, K. Y. Wang, R. P. Campion et al., Hall effect and hole densities
in Ga1-xMn x As, Appl. Phys. Lett. 81, 3010–3012 (2002).
87. B. S. Sørensen, J. Sadowski, S. E. Andresen, and P. E. Lindelof, Dependence of
Curie temperature on the thickness of epitaxial (Ga,Mn)As film, Phys. Rev. B
66, 233313 (2002).
88. K. C. Ku, S. J. Potashnik, R. Wang et al., Highly enhanced Curie temperature
in low-temperature annealed [Ga,Mn]As epilayers, Appl. Phys. Lett. 82, 2302–
2304 (2003).
89. K. W. Edmonds, N. R. S. Farley, T. K. Johal et al., Ferromagnetic moment and anti-
ferromagnetic coupling in (Ga,Mn)As thin films, Phys. Rev. B 71, 064418 (2005).
90. P. N. Hai, L. D. Anh, S. Mohan et al., Growth and characterization of n-type
electron-induced ferromagnetic semiconductor (In,Fe)As, Appl. Phys. Lett.
101, 182403 (2012); N. T. Tu, P. N. Hai, L. D. Anh, and M. Tanaka, (Ga,Fe)Sb: A
p-type ferromagnetic semiconductor, ibid. 105, 132402 (2014)..
91. G. L. Pearson and J. Bardeen, Electrical properties of pure silicon and silicon
alloys containing boron and phosphorus, Phys. Rev. 75, 865–883 (1949).
92. W. J. Brown Jr. and J. S. Blakemore, Transport and photoelectrical properties
of gallium arsenide containing deep acceptors, J. Appl. Phys. 43, 2242–2246
(1972).
93. T. Dietl, A. Haury, and Y. Merle d’Aubigné, Free carrier-induced ferromag-
netism in structures of diluted magnetic semiconductors, Phys. Rev. B 55,
R3347–R3350 (1997).
94. T. Jungwirth, W. A. Atkinson, B. H. Lee, and A. H. MacDonald, Interlayer
coupling in ferromagnetic semiconductor superlattices, Phys. Rev. B 59, 9818–
9821 (1999).
96. T. Jungwirth, K. Y. Wang, J. Mašek et al., Prospects for high temperature fer-
romagnetism in (Ga,Mn)As semiconductors, Phys. Rev. B 72, 165204 (2005).
97. A. Chattopadhyay, S. Das Sarma, and A. J. Millis, Transition temperature of
ferromagnetic semiconductors: A dynamical mean field study, Phys. Rev. Lett.
87, 227202 (2001).
98. C. Santos and W. Nolting, Ferromagnetism in the Kondo-lattice model, Phys.
Rev. B 65, 144419 (2002); J. Kienert and W. Nolting, Magnetic phase diagram
of the Kondo lattice model with quantum localized spins, ibid. 73, 224405
(2006).
99. M. Turek, J. Siewert, and J. Fabian, Electronic and optical properties of ferro-
magnetic Ga1-xMn x As in a multiband tight-binding approach, Phys. Rev. B 78,
085211 (2008).
100. C. Timm, F. Schäfer, and F. von Oppen, Comment on “Effects of disorder on
ferromagnetism in diluted magnetic semiconductors”, Phys. Rev. Lett. 90,
029701 (2003).
1019–1022 (2000).
102. J. A. Sobota, D. Tanasković, and V. Dobrosavljević, RKKY interactions in the
regime of strong localization, Phys. Rev. B 76, 245106 (2007).
103. T. Dietl, Origin of ferromagnetic response in diluted magnetic semiconductors
and oxides, J. Phys. Condens. Matter 19, 165204 (2007).
104. M. van Schilfgaarde and O. N. Mryasov, Anomalous exchange interactions in
III–V dilute magnetic semiconductors, Phys. Rev. B 63, 233205 (2001).
105. G. A. Fiete, G. Zaránd, and K. Damle, Effective hamiltonian for Ga1-xMn x As
in the dilute limit, Phys. Rev. Lett. 91, 097202 (2003); G. A. Fiete, G. Zaránd, K.
Damle, and C. P. Moca, Disorder, spin-orbit, and interaction effects in dilute
Ga1-xMn x As, Phys. Rev. B 72, 045212 (2005).
106. C. P. Moca, G. Zaránd, and M. Berciu, Theory of optical conductivity for dilute
Ga1-xMn x As, Phys. Rev. B 80, 165202 (2009).
References 403
107. C. Zener, Interaction between the d shells in the transition metals, Phys. Rev.
81, 440–444 (1951).
108. J. König, H.-H. Lin, and A. H. MacDonald, Theory of diluted magnetic semi-
conductor ferromagnetism, Phys. Rev. Lett. 84, 5628–5631 (2000); J. König,
T. Jungwirth, and A. H. MacDonald, Theory of magnetic properties and spin-
wave dispersion for ferromagnetic (Ga,Mn)As, Phys. Rev. B 64, 184423 (2001).
109. R. Bouzerar, G. Bouzerar, and T. Ziman, Non-perturbative V-Jpd model and
ferromagnetism in dilute magnets, EPL 78, 67003 (2007); R. Bouzerar and
G. Bouzerar, Unified picture for diluted magnetic semiconductors, ibid. 92,
47006 (2010).
110. G. Bouzerar and R. Bouzerar, Unraveling the nature of carrier-mediated fer-
romagnetism in diluted magnetic semiconductors, Comptes. Rendus. Phys. 16,
731 (2015).
111. J. M. Luttinger and W. Kohn, Motion of electrons and holes in perturbed peri-
odic fields, Phys. Rev. 97, 869–883 (1955).
112. I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, Band parameters for III–V
compound semiconductors and their alloys, J. Appl. Phys. 89, 5815–5875 (2001).
114. J. C. Slater and G. F. Koster, Simplified LCAO method for the periodic potential
problem, Phys. Rev. 94, 1498–1524 (1954); G. F. Koster and J. C. Slater, Wave
functions for impurity levels, ibid. 95, 1167–1176 (1954).
115. D. J. Chadi, Spin-orbit splitting in crystalline and compositionally disordered
116. T. Jungwirth, J. Mašek, J. Sinova, and A. H. MacDonald, Ferromagnetic tran-
sition temperature enhancement in (Ga,Mn)As semiconductors by carbon
codoping, Phys. Rev. B 68, 161202 (2003).
117. Y. Yildirim, G. Alvarez, A. Moreo, and E. Dagotto, Large-scale monte carlo study
of a realistic lattice model for Ga1-xMnxAs, Phys. Rev. Lett. 99, 057207 (2007).
118. J. Mašek, F. Máca, J. Kudrnovská et al., Microscopic analysis of the valence
band and impurity band theories of (Ga,Mn)As, Phys. Rev. Lett. 105, 227202
(2010).
119. H. Ohno and T. Dietl, Spin-transfer physics and the model of ferromagnetism
in (Ga,Mn)As, J. Magn. Magn. Mater. 320, 1293–1299 (2008).
120. J. Okabayashi, A. Kimura, O. Rader, et al., Angle-resolved photoemission study
of Ga1-xMn x As, Phys. Rev. B 64, 125304 (2001).
121. M. Kobayashi, I. Muneta, Y. Takeda et al., Unveiling the impurity band induced
205204 (2014).
122. A. X. Gray, J. Minár, S. Ueda et al., Bulk electronic structure of the dilute mag-
netic semiconductor Ga1-xMn x As through hard X-ray angle-resolved photo-
emission, Nat. Mater. 11, 957–962 (2012).
123. I. Di Marco, P. Thunström, M. I. Katsnelson et al., Electron correlations in
Mn xGa1-x As as seen by resonant electron spectroscopy and dynamical mean
field theory, Nat. Commun. 4, 2645 (2013).
124. J. Fujii, B. R. Salles, M. Sperl et al., Identifying the electronic character and role
of the Mn States in the valence band of (Ga,Mn)As, Phys. Rev. Lett. 111, 097201
(2013).
125. P. R. Stone, K. Alberi, S. K. Z. Tardif et al., Metal-insulator transition by isova-
lent anion substitution in Ga1-xMn x As: Implications to ferromagnetism, Phys.
Rev. Lett. 101, 087203 (2008).
126. E. H. C. P. Sinnecker, G. M. Penello, T. G. Rappoport et al., Ion-beam modi-
fication of the magnetic properties of Ga1-xMn x As epilayers, Phys. Rev. B 81,
245203 (2010).
127. S. Zhou, L. Li, Y. Yuan et al., Precise tuning of the Curie temperature of
(Ga,Mn)As-based magnetic semiconductors by hole compensation: Support
for valence-band ferromagnetism, Phys. Rev. B 94, 075205 (2016).
128. K. S. Burch, J. Stephens, R. K. Kawakami, D. D. Awschalom, and D. N. Basov,

Ellipsometric study of the electronic structure of Ga1-xMn x As and low-tem-
perature GaAs, Phys. Rev. B 70, 205208 (2004).
129. T. D. Kang, G. S. Lee, H. Lee, D. Koh, and Y. J. Park, Optical properties of Ga1-x
Mn x As (0≤x≤0.09) studied using spectroscopic ellipsometry, J. Korean Phys.
Soc. 46, 482–486 (2005).
130. L. Gluba, O. Yastrubchak et al., On the nature of the Mn-related states in the
band structure of (Ga,Mn)As alloys via probing the E1 and optical transi-
tions, Appl. Phys. Lett. 105, 032408 (2014); O. Yastrubchak, J. Sadowski, H.
Krzyżanowska et al., Electronic- and band-structure evolution in low-doped
(Ga,Mn)As, J. Appl. Phys. 114, 053710 (2013).
131. E. J. Singley, R. Kawakami, D. D. Awschalom, and D. N. Basov, Infrared probe
of itinerant ferromagnetism in Ga1-xMn x As, Phys. Rev. Lett. 89, 097203 (2002);
E. J. Singley, K. S. Burch, R. Kawakami, J. Stephens, D. D. Awschalom, and D. N.
Basov, Electronic structure and carrier dynamics of the ferromagnetic semi-
conductor, Phys. Rev. B 68, 165204 (2003).
132. K. S. Burch, D. B. Shrekenhamer, E. J. Singley et al., Impurity band conduc-
tion in a high temperature ferromagnetic semiconductor, Phys. Rev. Lett. 97,
087208 (2006).
133. T. Jungwirth, P. Horodyská, N. Tesařová et al., Systematic study of Mn-doping
trends in optical properties of (Ga,Mn)As, Phys. Rev. Lett. 105, 227201 (2010).
134. J. Sinova, T. Jungwirth, S.-R. E. Yang, J. Kučera, and A. H. MacDonald, Infrared
conductivity of metallic (III,Mn)V ferromagnets, Phys. Rev. B 66, 041202
(2002).
135. S.-R. E. Yang, J. Sinova, T. Jungwirth, Y. P. Shim, and A. H. MacDonald, Non-
drude optical conductivity of (III,Mn)V ferromagnetic semiconductors, Phys.
Rev. B 67, 045205 (2003).
136. V. F. Sapega, M. Moreno, M. Ramsteiner, L. Däweritz, and K. H. Ploog,
Polarization of valence band holes in the (Ga,Mn)As diluted magnetic semi-
conductor, Phys. Rev. Lett. 94, 137401 (2005); V. F. Sapega, M. Ramsteiner, O.
Brandt, L. Däweritz, and K. H. Ploog, Hot-electron photoluminescence study of
the (Ga,Mn)As diluted magnetic semiconductor, Phys. Rev. B 73, 235208 (2006).
137. K. Ando, H. Saito, K. C. Agarwal, M. C. Debnath, and V. Zayets, Origin of
the anomalous magnetic circular dichroism spectral shape in ferromagnetic
Ga1-xMn x As: Impurity bands inside the band gap, Phys. Rev. Lett. 100, 067204
(2008).
138. J.-M. Tang and M. E. Flatté, Magnetic circular dichroism from the impurity
band in III–V diluted magnetic semiconductors, Phys. Rev. Lett. 101, 157203
(2008).
139. M. Turek, J. Siewert, and J. Fabian, Magnetic circular dichroism in Ga1-x
Mn x As: Theoretical evidence for and against an impurity band, Phys. Rev. B
80, 161201(R) (2009).
140. J. Mašek, J. Kudrnovský, F. Máca et al., Mn-doped Ga(As,P) and (Al,Ga)As fer-
romagnetic semiconductors: Electronic structure calculations, Phys. Rev. B 75,
045202 (2007).
141. G. Acbas, M.-H. Kim, M. Cukr et al., Electronic structure of ferromagnetic
semiconductor Ga1-xMn x As probed by subgap magneto-optical spectroscopy,
Phys. Rev. Lett. 103, 137201 (2009).
142. S. Ohya, I. Muneta, P. N. Hai, and M. Tanaka, Valence-band structure of the
ferromagnetic semiconductor GaMnAs studied by spin-dependent resonant
tunneling spectroscopy, Phys. Rev. Lett. 104, 167204 (2010); S. Ohya, K. Takata,
and M. Tanaka, Nearly non-magnetic valence band of the ferromagnetic semi-
conductor GaMnAs, Nat. Phys. 7, 342–347 (2011).
143. S. Ohya, I. Muneta, Y. Xin, K. Takata, and M. Tanaka, Valence-band structure
of ferromagnetic semiconductor (In,Ga,Mn)As, Phys. Rev. B 86, 094418 (2012).
144. T. Dietl and D. Sztenkiel, Reconciling results of tunnelling experiments on
(Ga,Mn)As, arXiv:1102.3267 (2011).
References 405
145. S. Ohya, K. Takata, I. Muneta, P. N. Hai, and M. Tanaka, Comment on [140],

arXiv:1102.4459 (2011).
146. D. Chiba, F. Matsukura, and H. Ohno, Electric-field control of ferromagnetism
in (Ga,Mn)As, Appl. Phys. Lett. 89, 162505 (2006).
147. T. Dietl, Origin and control of ferromagnetism in dilute magnetic semicon-
ductors and oxides, J. Appl. Phys. 103, 07D111 (2008).
148. M. Sawicki, D. Chiba, A. Korbecka et al., Experimental probing of the interplay
(2010).
149. Y. Nishitani, D. Chiba, M. Endo et al., Curie temperature versus hole con-
centration in field-effect structures of Ga1-x Mn x As, Phys. Rev. B 81, 045208
(2010).
150. M. Wang, K. W. Edmonds, B. L. Gallagher et al., High Curie temperatures at
low compensation in the ferromagnetic semiconductor (Ga,Mn)As, Phys. Rev.
B 87, 121301(R) (2013).
151. D. A. Woodbury and J. S. Blakemore, Impurity conduction and the metal-
nonmetal transition in manganese-doped gallium arsenide, Phys. Rev. B 8,
3803–3810 (1973).
152. D. Neumaier, M. Turek, U. Wurstbauer et al., All-electrical measurement of
the density of states in (Ga,Mn)As, Phys. Rev. Lett. 103, 087203 (2009).
153. P. A. Lee and T. V. Ramakrishnan, Disordered electronic systems, Rev. Mod.
Phys. 57, 287–337 (1985).
154. C. Zener, Interaction between the d-Shells in the Transition Metals. II.
Ferromagnetic compounds of manganese with perovskite structure, Phys. Rev.
82, 403–405 (1951).
155. J. Kudrnovský, I. Turek, V. Drchal, F. Máca, P. Weinberger, and P. Bruno,
Exchange interactions in III–V and group–IV diluted magnetic semiconduc-
tors, Phys. Rev. B 69, 115208 (2004).
156. G. Bouzerar, R. Bouzerar, and O. Cépas, Superexchange induced canted fer-
romagnetism in dilute magnets, Phys. Rev. B 76, 144419 (2007).
157. J. B. Goodenough, An interpretation of the magnetic properties of the
perovskite-type mixed crystals La1-xSrxCoO3-λ, J. Phys. Chem. Solids 6, 287–
297 (1958); J. Kanamori, Superexchange interaction and symmetry properties
of electron orbitals, ibid. 10, 87–98 (1959).
158. H. Ohno, D. Chiba, F. Matsukura et al., Electric-field control of ferromagne-
tism, Nature 408, 944–946 (2000); D. Chiba, M. Yamanouchi, F. Matsukura,
and H. Ohno, Electrical manipulation of magnetization reversal in a ferromag-
netic semiconductor, Science 301, 943–945 (2003).
Magnetization vector manipulation by electric fields, Nature 455, 515–518
(2008); D. Chiba, T. Ono, F. Matsukura, and H. Ohno, Electric field control of
thermal stability and magnetization switching in (Ga,Mn)As, Appl. Phys. Lett.
103, 142418 (2013).
160. M. H. S. Owen, J. Wunderlich, V. Novák et al., Low-voltage control of ferro-
magnetism in a semiconductor p-n junction, New J. Phys. 11, 023008 (2009).
161. T. Niazi, M. Cormier, D. Lucot et al., Electric-field control of the magnetic
anisotropy in an ultrathin (Ga,Mn)As/(Ga,Mn)(As,P) bilayer, Appl. Phys. Lett.
102, 122403 (2013).
162. L. D. Anh, P. N. Hai, Y. Kasahara, Y. Isawa, and M. Tanaka, Modulation of fer-
romagnetism in (In,Fe)As quantum wells via electrically controlled deforma-
tion of the electron wave functions, Phys. Rev. B 92, 161201(R) (2015).
163. A. M. Nazmul, S. Kobayashi, S. Sugahara, and M. Tanaka, Electrical and opti-
cal control of ferromagnetism in III–V semiconductor heterostructures at
high temperature (~100K), Jpn. J. Appl. Phys. Part 2, 43, L233–L236 (2004).
164. E. Dias Cabral, M. A. Boselli, R. Oszwałdowski, I. Žutić, and I. C. da Cunha
Lima, Electrical control of magnetic quantum wells: Monte Carlo simulations,
Phys. Rev. B 84, 085315 (2011)
165. C. Timm, Charge and magnetization inhomogeneities in diluted magnetic

semiconductors, Phys. Rev. Lett. 96, 117201 (2006); Possible magnetic-field-
induced voltage and thermopower in diluted magnetic semiconductors, Phys.
Rev. B 74, 014419 (2006).
166. S. Koshihara, A. Oiwa, M. Hirasawa et al., Ferromagnetic order induced by
photogenerated carriers in magnetic III–V semiconductor heterostructures of
(In,Mn)As/GaSb, Phys. Rev. Lett. 78, 4617–4620 (1997); H. Munekata, T. Abe,
S. Koshihara et al., Light-induced ferromagnetism in III–V-based diluted mag-
netic semiconductor heterostructures, J. Appl. Phys. 81, 4862–4864 (1997).
167. J. Wang, I. Cotoros, K. M. Dani, X. Liu, J. K. Furdyna, and D. S. Chemla,
Ultrafast enhancement of ferromagnetism via photoexcited holes in GaMnAs,
Phys. Rev. Lett. 98, 217401 (2007).
Mater. 200, 110–129 (1999); H. Ohno and F. Matsukura, A ferromagnetic
III–V semiconductor: (Ga,Mn)As, Solid State Commun. 117, 179–186 (2001).
169. X. Liu, Y. Sasaki, and J. K. Furdyna, Ferromagnetic resonance in Ga1-xMn x As:
Effects of magnetic anisotropy, Phys. Rev. B 67, 205204 (2003); X. Liu, W. L.
Lim, L. V. Titova et al., External control of the direction of magnetization in
ferromagnetic InMnAs/GaSb heterostructures, Phys. E 20, 370–373 (2004).
170. M. Sawicki, F. Matsukura, A. Idziaszek et al., Temperature dependent magnetic
anisotropy in (Ga,Mn)As layers, Phys. Rev. B 70, 245325 (2004); M. Sawicki,
K.-Y. Wang, K. W. Edmonds et al., In-plane uniaxial anisotropy rotations in
(Ga,Mn)As thin films, ibid. 71, 121302 (2005).
171. S. C. Masmanidis, H. X. Tang, E. B. Myers et al., Nanomechanical measurement
of magnetostriction and magnetic anisotropy in (Ga,Mn)As, Phys. Rev. Lett. 95,
187206 (2005).
172. K. W. Edmonds, G. van der Laan, N. R. S. Farley et al., Strain dependence of the
Mn anisotropy in ferromagnetic semiconductors observed by x-ray magnetic
circular dichroism, Phys. Rev. B 77, 113205 (2008).
173. Y. Wan-Lun and T. Tao, Theory of the zero-field splitting of 6S(3d5) -state ions
in cubic crystals, Phys. Rev. B 49, 3243–3252 (1994).
174. K. Yosida, Theory of Magnetism. Berlin: Springer (1996).
175. G. Zaránd and B. Jankó, Ga1-xMn x As: A frustrated ferromagnet, Phys. Rev. Lett.
89, 047201 (2002).
176. L. Brey and G. Gómez-Santos, Magnetic properties of GaMnAs from an effec-
tive Heisenberg Hamiltonian, Phys. Rev. B 68, 115206 (2003).
177. G. Bouzerar, T. Ziman, and J. Kudrnovský, Calculating the Curie temperature
reliably in diluted III–V ferromagnetic semiconductors, Europhys. Lett. 69,
812–818 (2005); Compensation, interstitial defects, and ferromagnetism in
diluted ferromagnetic semiconductors, Phys. Rev. B 72, 125207 (2005).
178. S. Das Sarma, E. H. Hwang, and A. Kaminski, Temperature-dependent magne-
tization in diluted magnetic semiconductors, Phys. Rev. B 67, 155201 (2003).
179. P. Mahadevan, A. Zunger, and D. D. Sarma, Unusual directional dependence
of exchange energies in GaAs diluted with Mn: Is the RKKY description rel-
evant?, Phys. Rev. Lett. 93, 177201 (2004).
180. H. Ebert and S. Mankovsky, Anisotropic exchange coupling in diluted mag-
netic semiconductors: Ab initio spin-density functional theory, Phys. Rev. B
79, 045209 (2009); S. Mankovsky, S. Polesya, S. Bornemann et al., Spin-orbit
coupling effect in (Ga,Mn)As films: Anisotropic exchange interactions and
magnetocrystalline anisotropy, ibid. 84, 201201(R) (2011).
181. A. A. Abrikosov and L. P. Goŕkov, On the nature of impurity ferromagne-
tism, Sov. Phys. JETP 16, 1575–1577 (1963) [Zh. Eksp. Teor. Fiz. 43, 2230–2233
(1962)].
182. T. Jungwirth, J. König, J. Sinova, J. Kučera, and A. H. MacDonald, Curie tem-
perature trends in (III,Mn)V ferromagnetic semiconductors, Phys. Rev. B 66,
012402 (2002).
References 407
183. S. Hilbert and W. Nolting, Magnetism in (III,Mn)-V diluted magnetic semi-

conductors: Effective Heisenberg model, Phys. Rev. B 71, 113204 (2005).
184. S. T. B. Goennenwein, T. Graf, T. Wassner et al., Spin wave resonance in Ga1-x
Mn x As, Appl. Phys. Lett. 82, 730–732 (2003).
185. G. Bouzerar, Magnetic spin excitations in diluted ferromagnetic systems: The
case of Ga1-xMn x As, EPL 79, 57007 (2007).
186. M. D. Kapetanakis and I. E. Perakis, Spin dynamics in (III,Mn)V ferromag-
netic semiconductors: The role of correlations, Phys. Rev. Lett. 101, 097201
(2008).
187. A. Van Esch, L. Van Bockstal, J. De Boeck et al., Interplay between the mag-
netic and transport properties in the III–V diluted magnetic semiconductor
Ga1-xMn x As, Phys. Rev. B 56, 13103–13112 (1997).
188. C. P. Moca, B. L. Sheu, N. Samarth, P. Schiffer, B. Jankó, and G. Zaránd, Scaling
theory of magnetoresistance and carrier localization in Ga1-xMn x As, Phys. Rev.
Lett. 102, 137203 (2009).
189. S. Lee, A. Trionfi, T. Schallenberg, H. Munekata, and D. Natelson, Mesoscopic
conductance effects in InMnAs structures, Appl. Phys. Lett. 90, 032105 (2007).
190. J. Mašek, I. Turek, V. Drchal, J. Kudrnovský, and F. Máca, Correlated doping
in semiconductors: The role of donors in III–V diluted magnetic semiconduc-
tors, Acta Phys. Pol. A 102, 673–678 (2002); J. Mašek, I. Turek, J. Kudrnovský,
F. Máca, and V. Drchal, Compositional dependence of the formation energies
of substitutional and interstitial Mn in partially compensated (Ga,Mn)As,
ibid. 105, 637–644 (2004).
191. G. Alvarez, M. Mayr, and E. Dagotto, Phase diagram of a model for diluted
magnetic semiconductors beyond mean-field approximations, Phys. Rev. Lett.
89, 277202 (2002).
192. G. Alvarez and E. Dagotto, Clustered states as a new paradigm of condensed
matter physics, J. Magn. Magn. Mater. 272–276, 15–20 (2004).
193. P. Mahadevan, J. M. Osorio-Guillén, and A. Zunger, Origin of transition metal
clustering tendencies in GaAs based dilute magnetic semiconductors, Appl.
Phys. Lett. 86, 172504 (2005).
194. H. Raebiger, M. Ganchenkova, and J. von Boehm, Diffusion and clustering of
substitutional Mn in (Ga,Mn)As, Appl. Phys. Lett. 89, 012505 (2006).
195. F. V. Kyrychenko and C. A. Ullrich, Enhanced carrier scattering rates in dilute
magnetic semiconductors with correlated impurities, Phys. Rev. B 75, 045205
(2007).
196. F. V. Kyrychenko and C. A. Ullrich, Transport and optical conductivity in
dilute magnetic semiconductors, J. Phys. Condens. Matter 21, 084202 (2009);
Temperature-dependent resistivity of ferromagnetic Ga1-xMn x As: Interplay
between impurity scattering and many-body effects, Phys. Rev. B 80, 205202
(2009).
197. M. Zhu, X. Li, G. Xiang, and N. Samarth, Random telegraph noise from mag-
netic nanoclusters in the ferromagnetic semiconductor (Ga,Mn)As, Phys. Rev.
B 76, 201201(R) (2007).
198. H. Bruus and K. Flensberg, Many-Body Quantum Theory in Condensed Matter
Physics, Oxford University Press, Oxford, 2004.
199. J. Rammer, Quantum Transport Theory, Westview Press, Boulder, 2004.
200. E. Abrahams, P. W. Anderson, D. C. Licciardello, and T. V. Ramakrishnan,
Scaling theory of localization: Absence of quantum diffusion in two dimen-
sions, Phys. Rev. Lett. 42, 673–676 (1979).
201. B. I. Shklovskii and A. L. Efros, Electronic Properties of Doped Semiconductors,
Springer, Berlin, (1984).
202. D. M. Basko, I. L. Aleiner, and B. L. Altshuler, Metal-insulator transition in a
weakly interacting many-electron system with localized single-particle states,
Ann. Phys. 321, 1126–1205 (2006); On the problem of many-body localization,
cond-mat/0602510 (2006).
203. T. Słupiński, J. Caban, and K. Moskalik, Hole transport in impurity band and
valence bands studied in moderately doped GaAs:Mn single crystals, Acta
Phys. Pol. A 112, 325–330 (2007).
204. S. R. Dunsiger, J. P. Carlo, T. Goko et al., Spatially homogeneous ferromagne-
tism of (Ga,Mn)As, Nat. Mater. 9, 299–303 (2010).
205. N. V. Agrinskaya, V. I. Kozub, D. V. Poloskin, A. V. Chernyaev, and D. V.
Shamshur, Transition from strong to weak localization in the split-off impurity
band in two-dimensional p-GaAs/AlGaAs structures, JETP Lett. 80, 30–34
(2004) [P. Zh. Eksp. Teoret. Fiz. 80, 36–40 (2004)]; Crossover from strong to
weak localization in the split-off impurity band in two-dimensional p-GaAs/
AlGaAs structures, Phys. Stat. Sol. (C) 3, 329–333 (2006).
206. M. Berciu and R. N. Bhatt, Effects of disorder on ferromagnetism in diluted
magnetic semiconductors, Phys. Rev. Lett. 87, 107203 (2001); Mean-field
approach to ferromagnetism in (III,Mn)V diluted magnetic semiconductors at
low carrier densities, Phys. Rev. B 69, 045202 (2004).
207. A. Kaminski and S. Das Sarma, Polaron percolation in diluted magnetic semi-
conductors, Phys. Rev. Lett. 88, 247202 (2002); Magnetic and transport perco-
lation in diluted magnetic semiconductors, Phys. Rev. B 68, 235210 (2003).
208. A. C. Durst, R. N. Bhatt, and P. A. Wolff, Bound magnetic polaron interactions
in insulating doped diluted magnetic semiconductors, Phys. Rev. B 65, 235205
(2002).
209. B. L. Sheu, R. C. Myers, J.-M. Tang et al., Onset of ferromagnetism in low-
doped Ga1-xMn x As, Phys. Rev. Lett. 99, 227205 (2007).
210. A. N. Andriotis and M. Menon, The synergistic character of the defect-induced
magnetism in diluted magnetic semiconductors and related magnetic materials,
J. Phys. Condens. Matter 24, 455801 (2012).
211. D. Neumaier, K. Wagner, S. Geißler et al., Weak localization in ferromagnetic
(Ga,Mn)As nanostructures, Phys. Rev. Lett. 99, 116803 (2007).
212. P. A. Lee, A. D. Stone, and H. Fukuyama, Universal conductance fluctuations
in metals: Effects of finite temperature, interactions, and magnetic field, Phys.
Rev. B 35, 1039–1070 (1987).
213. K. Wagner, D. Neumaier, M. Reinwald, W. Wegscheider, and D. Weiss,
Dephasing in (Ga,Mn)As nanowires and rings, Phys. Rev. Lett. 97, 056803
(2006).
214. L. Vila, R. Giraud, L. Thevenard et al., Universal conductance fluctuations in
epitaxial GaMnAs ferromagnets: Dephasing by structural and spin disorder,
Phys. Rev. Lett. 98, 027204 (2007).
215. I. Garate, J. Sinova, T. Jungwirth, and A. H. MacDonald, Theory of weak local-
ization in ferromagnetic (Ga,Mn)As, Phys. Rev. B 79, 155207 (2009).
216. L. P. Rokhinson, Y. Lyanda-Geller, Z. Ge et al., Weak localization in Ga1-x
Mn x As: Evidence of impurity band transport, Phys. Rev. B 76, 161201(R) (2007).
217. C. Timm, M. E. Raikh, and F. von Oppen, Disorder-induced resistive anomaly
near ferromagnetic phase transitions, Phys. Rev. Lett. 94, 036602 (2005).
218. S. Yuldashev, K. Igamberdiev, S. Lee et al., Specific heat study of GaMnAs, Appl.
Phys. Express 3, 073005 (2010).
219. M. E. Fisher and J. S. Langer, Resistive anomalies at magnetic critical points,
Phys. Rev. Lett. 20, 665–668 (1968).
220. G. Bouzerar, J. Kudrnovský, and P. Bruno, Disorder effects in diluted ferromag-
netic semiconductors, Phys. Rev. B 68, 205311 (2003).
221. V. M. Galitski, A. Kaminski, and S. Das Sarma, Griffiths phase in diluted mag-
netic semiconductors, Phys. Rev. Lett. 92, 177203 (2004).
222. A. Geresdi, A. Halbritter, M. Csontos et al., Nanoscale spin polarization in the
dilute magnetic semiconductor (In,Mn)Sb, Phys. Rev. B 77, 233304 (2008).
223. J. Schliemann and A. H. MacDonald, Noncollinear Ferromagnetism in (III,Mn)
V Semiconductors, Phys. Rev. Lett. 88, 137201 (2002); J. Schliemann, Disorder-
induced noncollinear ferromagnetism in models for (III,Mn)V semiconduc-
tors, Phys. Rev. B 67, 045202 (2003).
References 409
224. G. A. Fiete, G. Zaránd, B. Jankó, P. Redliński, and C. P. Moca, Positional dis-

order, spin-orbit coupling, and frustration in Ga1-xMn x As, Phys. Rev. B 71,
115202 (2005).
225. T. Jungwirth, J. Mašek, K. Y. Wang et al., Low-temperature magnetization of
(Ga,Mn)As semiconductors, Phys. Rev. B 73, 165205 (2006).
226. M. Mayr, G. Alvarez, and E. Dagotto, Global versus local ferromagnetism in
a model for diluted magnetic semiconductors studied with Monte Carlo tech-
niques, Phys. Rev. B 65, 241202(R) (2002).
227. D. J. Priour Jr., E. H. Hwang, and S. Das Sarma, Disordered RKKY lattice mean
field theory for ferromagnetism in diluted magnetic semiconductors, Phys.
Rev. Lett. 92, 117201 (2004).
228. K. Aryanpour, J. Moreno, M. Jarrell, and R. S. Fishman, Magnetism in semi-
conductors: A dynamical mean-field study of ferromagnetism in Ga1-xMn x As,
Phys. Rev. B 72, 045343 (2005).
229. B. Beschoten, P. A. Crowell, I. Malajovich et al., Magnetic circular dichroism
studies of carrier-induced ferromagnetism in (Ga1-xMn x)As, Phys. Rev. Lett. 83,
3073–3076 (1999).
230. R. Mathieu, B. S. Sørensen, J. Sadowski et al., Magnetization of ultrathin
(Ga,Mn)As layers, Phys. Rev. B 68, 184421 (2003).
231. W. Limmer, A. Koeder, S. Frank et al., Waag, Electronic and magnetic proper-
ties of GaMnAs: annealing effects, Phys. E 21, 970–974 (2004).
232. J. Schliemann, J. König, and A. H. MacDonald, Monte Carlo study of ferromag-
netism in (III,Mn)V semiconductors, Phys. Rev. B 64, 165201 (2001).
233. G. Alvarez and E. Dagotto, Single-band model for diluted magnetic semicon-
ductors: Dynamical and transport properties and relevance of clustered states,
Phys. Rev. B 68, 045202 (2003).
234. G. Bouzerar and R. Bouzerar, Comment on “Large-Scale Monte Carlo Study of
a Realistic Lattice Model for Ga1-xMn x As”, Phys. Rev. Lett. 100, 229701 (2008);
A. Moreo and E. Dagotto, Reply, ibid. 100, 229702 (2008).
235. S. Sonoda, S. Shimizu, T. Sasaki, Y. Yamamoto, and H. Hori, Molecular beam
epitaxy of wurtzite (Ga,Mn)N films on sapphire(0001) showing the ferromag-
netic behaviour at room temperature, J. Cryst. Growth 237–239, 1358–1362
(2002); T. Sasaki, S. Sonoda, Y. Yamamoto et al., Magnetic and transport char-
acteristics on high Curie temperature ferromagnet of Mn-doped GaN, J. Appl.
Phys. 91, 7911–7913 (2002).
236. N. Theodoropoulou, A. F. Hebard, M. E. Overberg et al., Unconventional
carrier-mediated ferromagnetism above room temperature in ion-Implanted
(Ga,Mn)P:C, Phys. Rev. Lett. 89, 107203 (2002).
237. S. Dhar, O. Brandt, M. Ramsteiner, V. F. Sapega, and K. H. Ploog, Colossal
magnetic moment of Gd in GaN, Phys. Rev. Lett. 94, 037205 (2005); S. Dhar, L.
Pérez, O. Brandt et al., Gd-doped GaN: A very dilute ferromagnetic semicon-
ductor with a Curie temperature above 300 K, Phys. Rev. B 72, 245203 (2005).
238. R. Rajaram, A. Ney, G. Solomon, J. S. Harris Jr., R. F. C. Farrow, and S. S. P. Parkin,
Growth and magnetism of Cr-doped InN, Appl. Phys. Lett. 87, 172511 (2005).
239. S. Dhar, T. Kammermeier, A. Ney et al., Ferromagnetism and colossal mag-
netic moment in Gd-focused ion-beam-implanted GaN, Appl. Phys. Lett. 89,
062503 (2006).
240. S. Sonoda, I. Tanaka, H. Ikeno et al., Coexistence of Mn2+ and Mn3+ in ferro-
magnetic GaMnN, J. Phys. Condens. Matter 18, 4615–4621 (2006).
241. S. Yoshii, S. Sonoda, T. Yamamoto et al., Evidence for carrier-induced high-Tc
ferromagnetism in Mn-doped GaN film, EPL 78, 37006 (2007).
242. A. Bonanni, M. Kiecana, C. Simbrunner et al., Paramagnetic GaN:Fe and fer-
romagnetic (Ga,Fe)N: The relationship between structural, electronic, and
magnetic properties, Phys. Rev. B 75, 125210 (2007).
243. J. M. Zavada, N. Nepal, C. Ugolini et al., Optical and magnetic behavior of
erbium-doped GaN epilayers grown by metal-organic chemical vapor depo-
sition, Appl. Phys. Lett. 91, 054106 (2007).
244. X. Wang, C. Timm, X. M. Wang et al., Metastable giant moments in

Gd-Implanted GaN, Si, and Sapphire, J. Supercond. Novel Magn. 24, 2123–
2128 (2011).
245. G. Martnez-Criado, A. Somogyi, S. Ramos et al., Mn-rich clusters in GaN:
Hexagonal or cubic symmetry? Appl. Phys. Lett. 86, 131927 (2005).
246. M. Yokoyama, H. Yamaguchi, T. Ogawa, and M. Tanaka, Zinc-blende-type
MnAs nanoclusters embedded in GaAs, J. Appl. Phys. 97, 10D317 (2005).
247. S. Y. Han, J. Hite, G. T. Thaler et al., Effect of Gd implantation on the structural
and magnetic properties of GaN and AlN, Appl. Phys. Lett. 88, 042102 (2006).
248. A. Ney, R. Rajaraman, T. Kammermeier et al., Metastable magnetism and
memory effects in dilute magnetic semiconductors, J. Phys. Condens. Matter
20, 285222 (2008).
249. P. Mahadevan and A. Zunger, First-principles investigation of the assumptions
underlying model-Hamiltonian approaches to ferromagnetism of 3d impuri-
ties in III–V semiconductors, Phys. Rev. B 69, 115211 (2004).
250. J. I. Hwang, Y. Ishida, M. Kobayashi et al., High-energy spectroscopic study
of the III–V nitride-based diluted magnetic semiconductor Ga1-x Mn x As,
Phys. Rev. B 72, 085216 (2005).
251. S. Marcet, D. Ferrand, D. Halley et al., Magneto-optical spectroscopy of
(Ga,Mn)N epilayers, Phys. Rev. B 74, 125201 (2006).
252. T. Dietl, Hole states in wide band-gap diluted magnetic semiconductors and
oxides, Phys. Rev. B 77, 085208 (2008).
253. W. Pacuski, P. Kossacki, D. Ferrand et al., Observation of strong-coupling
effects in a diluted magnetic semiconductor Ga1-xFexN, Phys. Rev. Lett. 100,
037204 (2008).
254. B. Sanyal, O. Bengone, and S. Mirbt, Electronic structure and magnetism of
Mn-doped GaN, Phys. Rev. B 68, 205210 (2003).
255. T. Graf, M. Gjukic, M. S. Brandt, M. Stutzmann, and O. Ambacher, The Mn3+/2+
acceptor level in group III nitrides, Appl. Phys. Lett. 81, 5159–5161 (2002).
256. M. A. Scarpulla, B. L. Cardozo, R. Farshchi et al., Ferromagnetism in Ga1-x
Mn xP: Evidence for inter-Mn exchange mediated by localized holes within a
detached impurity band, Phys. Rev. Lett. 95, 207204 (2005).
257. P. R. Stone, M. A. Scarpulla, R. Farshchi et al., Mn L3,2 x-ray absorption and
magnetic circular dichroism in ferromagnetic Ga1-xMn xP, Appl. Phys. Lett. 89,
012504 (2006).
258. L. Kronik, M. Jain, and J. R. Chelikowsky, Electronic structure and spin polar-
ization of Mn xGa1-xN, Phys. Rev. B 66, 041203 (2002).
259. S. Sonoda, I. Tanaka, F. Oba et al., Awaking of ferromagnetism in GaMnN
through control of Mn valence, Appl. Phys. Lett. 90, 012504 (2007).
260. J. Neugebauer and C. G. Van de Walle, Atomic geometry and electronic struc-
ture of native defects in GaN, Phys. Rev. B 50, 8067–8070 (1994); P. Bogus-
ławski, E. L. Briggs, and J. Bernholc, Native defects in gallium nitride, ibid. 51,
17255–17258 (1995); S. Limpijumnong and C. G. Van de Walle, Diffusivity of
native defects in GaN, ibid. 69, 035207 (2004).
261. X. L. Yang, Z. T. Chen, C. D. Wang et al., Effects of nitrogen vacancies induced
by Mn doping in (Ga,Mn)N films grown by MOCVD, J. Phys. D: Appl. Phys. 41,
125002 (2008).
262. K. Sato and H. Katayama-Yoshida, First principles materials design for semi-
conductor spintronics, Semicond. Sci. Technol. 17, 367–376 (2002).
263. G. M. Dalpian and S.-H. Wei, Electron-induced stabilization of ferromagne-
tism in Ga1-xGd xN, Phys. Rev. B 72, 115201 (2005).
264. J. M. D. Coey, M. Venkatesan, and C. B. Fitzgerald, Donor impurity band
exchange in dilute ferromagnetic oxides, Nat. Mater. 4, 173–179 (2005).
265. T. Dubroca, J. Hack, R. E. Hummel, and A. Angerhofer, Quasiferromagnetism
in semiconductors, Appl. Phys. Lett. 88, 182504 (2006).
References 411
266. D. M. Edwards and M. I. Katsnelson, High-temperature ferromagnetism of

sp electrons in narrow impurity bands: Application to CaB6, J. Phys. Condens.
Matter 18, 7209–7225 (2006).
267. J. M. D. Coey, K. Wongsaprom, J. Alaria, and M. Venkatesan, Charge-transfer
ferromagnetism in oxide nanoparticles, J. Phys. D: Appl. Phys. 41, 134012
(2008).
10
Magnetism of
Dilute Oxides
J.M.D. Coey

10.2 Dilute Oxides 420
10.3 Magnetic Measurements 423
10.4 Results on Specific Oxides 426
10.4.1 Dioxides 426
10.4.1.1 TiO2 426
10.4.1.2 SnO2 427
10.4.1.3 HfO2 428
10.4.1.4 CeO2 429
10.4.2 Sesquioxides 430
10.4.2.1 In2O3 430
10.4.3 Monoxides 432
10.4.3.1 ZnO 432
10.5 Discussion 437
413
414 Chapter 10. Magnetism of Dilute Oxides
10.5.1 Models 440

10.6 Conclusions 446
Acknowledgment 447
References 447
R
eports of Curie temperatures well above room temperature for wide-
bandgap oxides doped with a few percent of transition-metal cations
have triggered intense interest in these materials as potential magnetic
semiconductors. The origin of the magnetism is debated; in some systems,
the ferromagnetism can be attributed to nanoparticles of a ferromagnetic
secondary phase, but in others, properties are found that are incompat-
ible with any secondary phase, and an intrinsic origin related to structural
defects is implicated. The magnetic interactions in these materials are quali-
tatively different from those in magnetically concentrated compounds, and
the standard “m-J paradigm” of localized magnetism is unable to explain
their behavior. In this chapter, we first summarize the established view of
magnetism in the oxides, before reviewing data on a selection of the most
widely studied materials; then we consider the physical models that have
been advanced to explain the magnetism, and raise issues that still have to be
addressed. As noted in Chapter 9, Volume 2, some of the concepts also apply
to magnetically doped wide-gap III–V compounds.
10.1 INTRODUCTION
Transition-metal oxides have long been regarded as an important and
well-understood class of magnetic materials. We have built up an excellent
working knowledge of the spinel ferrites TFe2O4, garnets M3Fe5O12, and hex-
agonal ferrites MFe12O19, which have widespread industrial applications, and
represent about 40% of the global market for hard and soft magnets. This
knowledge can be regarded as a familiar landmark in our understanding of
magnetism in solids, epitomized by the award of the 1970 Nobel Prize in
physics to Louis Néel.
The crystal structures of these oxides are usually based on a close-packed
lattice of oxygen anions where the 3d cations mostly occupy octahedral inter-
stices with six oxygen neighbors or tetrahedral interstices with four oxygen
neighbors. The larger 3d cations show some preference for the octahedral
sites. Rare earth or alkali ions such as Ba or Sr have a greater coordination
number, and may sometimes replace an anion in the close-packed oxygen
lattice. A list of the ionic radii of common 3d cations is given in Table 10.1,
based on a radius of 140 pm for O2−.
An ionic picture of the bonding, with formal cation valences of 2+, 3+,
or 4+ and an oxygen valence of 2− leading to a 2p6 closed shell, is the tra-
ditional starting point for looking at the electronic structure. The number
of electrons per cation is an integer, and the magnetic moments are well-
localized. The oxides are usually insulators or wide-gap semiconductors.
TABLE 10.1
harge State, Electronic Configuration, and Ionic Radius in
C
Octahedral Coordination for Common 3d Cations in Oxides
Cation Charge State Configuration Ionic Radius (pm)
Sc 3+ 3d0 83
Ti 3+/4+ 3d0/3d1 61/69
V 2+ 3d3 72
Cr 3+ 3d3 64
Mn 2+/3+/4+ 3d5/3d4/3d3 83/65/53
Fe 2+/3+ 3d6/3d5 82/65
Co 2+/3+ 3d7/3d6 82/61
Ni 2+/3+ 3d8/3d7 78/69
Cu 2+ 3d9 72
4s (T ) 4s (T )
3d8 (T) E0 eg U
3F ∆
t2g
2p (O) 2p (O)
(a) (b)
FIGURE 10.1 Electronic structure diagrams for a 3d metal oxide (a) shows the 3d
electrons as a highly correlated level, (b) shows the 3d one-electron states in the
crystal field. The example illustrated is a 3d8 cation with S = 1 in an octahedral site.
They owe their colors—green for yttrium iron garnet, brown for barium
hexaferrite—to electronic transitions involving the 3d levels, which nor-
mally lie in the primary bandgap E0 between a valence band with largely
oxygen 2p character and a conduction band formed from unoccupied metal
4s orbitals. The magnitude of E0 decreases on moving along the 3d series
from 8.1 eV for TiO2 or 5.7 eV for Sc2O3, to 4.9 eV for Ga2O3 or 3.4 eV for ZnO.
The measured bandgap Eg can be much less than E0 because of the possibility
of p–d, or d–s transitions.
According to Hund’s rules, the 3d ions are in either an A (d5), D (d1, d4,
d , d9), or F (d2, d3, d7, d8) orbital ground state, depending on the number
6
of d electrons they possess. The ground-state levels are split by the crystal/
ligand field interaction [1, 2] and there are higher energy levels, which can
be excited optically. A schematic representation of the electronic structure
is given in Figure 10.1a.
It is often helpful to think of one-electron d states, even though the
d shell is strongly correlated. The five d orbitals split into a group of three
(dxy, dyz, dzx) and a group of two (dx2−y2, dz2) in cubic symmetry—undistorted
octahedral and tetrahedral sites have cubic symmetry. These are labeled as
t2g and eg states in octahedral symmetry, and as t2 and e states in tetrahe-
dral symmetry where there is no inversion center. The eg and t2 overlap and
hybridize strongly with the oxygen 2p ligand orbitals (px, py, px) but the t2g
and e do not. This produces the ligand field splitting of the triplet and the
doublet, which reinforces the purely electrostatic crystal field effect of the
oxygen charge on the d orbitals to give a splitting Δ. The d–d correlations are
roughly taken into account in this picture by a separation of the ↑ and ↓ lev-
els by an energy U, as shown in Figure 10.1b. When U > Δ, Hund’s first rule
applies to the ion, and it is in a high-spin state. Otherwise, when U < Δ, it is in
a low-spin state, with a spin moment determined by the lowest orbital occu-
pancy. The crystal field quenches the orbital moments for 3d ions in solids,
so their magnetic moments are, to a first approximation, spin-only moments
of just one Bohr magneton, µB, per unpaired electron. S is the good quantum
number for the 3d ion. The one-electron picture works well for D-state ions,
where there is a single electron or hole outside a filled or half-filled shell. It is
not nearly as good for the F-state ions.
Three other points complete the beginner’s toolkit for 3d ions in oxides.
(a) Filled shells and half-filled shells are particularly stable. In other cases, a
crystal field stabilization energy can be calculated for the ion. For Cr3+ d3, for
example, this is 3 × (2/5)Δ in octahedral sites. (The crystal field splitting of
the energies of the five d orbitals preserves their center of gravity). (b) Jahn–
Teller distortion of the octahedron or tetrahedron may stabilize the energy
of the ion. The effect is strongest for a singly occupied eg or e level—in other
words, d4 or d9 ions in octahedral coordination, and d1 and d6 ions in tetra-
hedral coordination. (c) There is a tendency for the d-levels to become more
stable as we move across the series from Sc to Zn because of the increasing
nuclear charge. For example, the d levels of Ti are close to the conduction
band, whereas those of Cu may overlap the 2p valence band.
Electrons or holes, introduced into oxides by doping or non-stoichiom-
etry, are often trapped. If they are at all mobile, they tend to form polarons
with a large effective mass. Electron doping can populate states in the 4s
conduction band, or 3dn+1 states in the primary gap, but the extra electrons
do not part easily from their dopant atom. Tin in In2O3 is an exception, and
indium tin oxide (ITO) is a good example of a transparent conducting oxide.
Hole doping, by oxygen substoichiometry, for example, usually creates donor
states near the bottom of the conduction band, which also tend to remain
localized. The insulating character of the oxides effectively resists tinkering
with the stoichiometry. Transport properties are quite different from those
of the classical covalent semiconductors, where electrons or holes diffuse
freely in the conduction or valence bands. The best oxide semiconductors are
those with the most covalent bonding, especially those with the 3d cations in
tetrahedral coordination—ZnO and CuAlO2, for example.
Electronic structure calculations based on density functional theory
have become very popular. These calculations work well for metals, and can
accurately predict the magnetic moments. There are difficulties in applying
them to oxides, where the primary gap E0 and the unoccupied electronic
levels may be difficult to reproduce.
The sources of the magnetism in oxides are the metal cations, which
bear a moment due to unpaired electrons in their 3d shells. Any 3d cation
with 1 ≤ n ≤ 9 has a spin moment, with very few exceptions—3d6 cations in
octahedral coordination and 3d4 cations in tetrahedral coordination may
be in a nonmagnetic low-spin state. The wave functions of the 3d cations
have exponentially decaying tails, which ensure that there is negligible over-
lap between cations that are not nearest-neighbors. These neighbors share
one or more common oxygen ligands. Hence, the magnetic superexchange
coupling essentially involves a threesome of two nearest-neighbor cations
and an intervening oxygen, as shown schematically for the Mn2+–O2−–Mn2+
bond in Figure 10.2. The Mn has a half-filled 3d shell, so only minority-spin
electron transfer from the oxygen to the manganese is possible. A 2p elec-
tron with ↓ is transferred to the Mn on the left, leaving a 2p hole that can
be filled by an electron from the other Mn, provided it is ↓. Superexchange
theory leads to a Heisenberg-type Hamiltonian
 = − 2JSi ⋅ S j (10.1)
where the exchange constant J = J0 cos2 θ, where θ is the superexchange bond
angle, and J0 = −t2/U′, where t is the M–O transfer integral and U′ is the
on-site Coulomb interaction for adding an electron to the M ion. Typically,
t ≈ 0.1 eV and U ≈ 5 eV. Hence, the order of magnitude of the exchange con-
stant is 2 meV or −20 K.
The expression for the Curie temperature in mean-field theory is
2Z0 JS(S + 1)
TC = (10.2)
3k B
where k B is the Boltzmann constant and S is the spin. Hence, if S = 5/2, the
largest value possible in the 3d series, and the cation coordination num-
ber Z0 = 8, we find TC ≈ 800 K. In practice, the superexchange interactions
are often frustrated—the antiparallel coupling of all nearest-neighbors
cannot be achieved for geometric reasons imposed by the topology of the
lattice. Furthermore, Equation 10.2 is known to overestimate TC by about
30%, because it takes no account of spin-wave excitations. In practice, the
magnetic-ordering temperatures of oxides almost never exceed 1000 K.
Hematite—αFe2O3—has a Néel temperature of 960 K. In most cases, the
ordering temperatures are a few hundred Kelvin, at best, as indicated in
Hole
O2–
FIGURE 10.2 Superexchange interaction between ions with more than half-
filled d shells via an O2− anion.
FIGURE 10.3 Histogram of the magnetic-ordering temperature of a selection of

magnetic oxides. (After Ackland, K. et al., Phys. Rep. 746, 1, 2018. With permission.)
Figure 10.3. When the oxide is diluted with nonmagnetic cations, the coor-
dination number Z = Z0 x is reduced, and TC varies as x above the percolation
threshold as shown in Figure 10.4.
The validity of the Heisenberg model in oxides has been amply demon-
strated by fitting the exchange constants to the spin-wave dispersion rela-
tions determined by inelastic neutron scattering. Complete data sets have
been obtained for Fe2O3, Cr2O3, and Fe3O4, among others [3–5]. The val-
ues of the exchange constants range from −30 to +6 K in α-Fe2O3. Negative
1.0
0.8
Tc/Tc (x= 1)
0.6
0.4
0.2 xp
0.0
0.0 0.2 0.4 0.6 0.8 1.0
x
FIGURE 10.4 Sketch of the magnetic-ordering temperature in a dilute magnetic

oxide with nearest-neighbor superexchange interactions. The dashed line is the
prediction of mean-field theory for long-range interactions, xp is the percolation
threshold.
(antiferromagnetic) interactions are for superexchange bonds involving the

same ions, with large bond angles. Positive (ferromagnetic) values are for
bonds involving ions in different valence states, or near-90° bonds between
ions in the same valence state.
During the 1960s, a wealth of information was accumulated on the nature
of the exchange interactions between different cations for different superex-
change bond angles. An extensive set of rules was formulated by Goodenough
[6] and Kanamori [7]. These rules were simplified by Anderson [8], as follows:
1. 180° exchange between half-filled orbitals is strong and

antiferromagnetic.
2. 90° exchange between half-filled orbitals is ferromagnetic, and rather
weak.
3. Exchange between a half-filled orbital and an empty orbital of differ-
ent symmetry is ferromagnetic and rather weak.
The stronger interactions tend to be negative (antiferromagnetic), cou-

pling the two cation spin moments antiparallel, and leaving no net moment
on the intervening oxygen. The three main families of magnetic insulators
mentioned in the opening paragraph are all ferrimagnetic, with dominant
nearest-neighbor antiferromagnetic exchange between ions on different,
unequal sublattices.
Not all transition-metal oxides are ferrimagnetic or antiferromagnetic,
and not all are insulators. The extended 3d orbitals of the elements at the
beginning of the series can form band states. There exist a few ferromagnetic
oxides, such as CrO2, which is the only simple oxide that is a ferromagnetic
metal (TC = 390 K).
In an effort to increase the magnetization of oxides, beginning in the
1950s, ways were sought to create strong ferromagnetic coupling. A success-
ful approach, implemented in mixed-valence manganites, was to impose
two different charge states on the manganese on the octahedral sites of
the perovskite structure by including cations on the cubic sites, which had
well-defined, but different charge states [9]. A famous example is (La0.7Sr0.3)
MnO3. Formally, 70% of the manganese is Mn3+ 3d4, while 30% is Mn4+ 3d3.
The fourth manganese d electrons can hop around among the 3d3 ion cores,
preserving their spin memory and providing ferromagnetic coupling via a
mechanism known as double exchange [10]. The delocalization of the d elec-
trons is conditional on parallel alignment of the ion cores, which gives rise to
the characteristic colossal magnetoresistance in these oxides near the Curie
point, which can be as high as 370 K. Another route to ferromagnetism is
illustrated by the double perovskite Sr2FeMoO6, where the spins of the iron
3d5 ion cores are coupled antiparallel to the spins of the electrons in a spin-
polarized 4d1 molybdenum band, giving ferromagnetic alignment of the iron.
To summarize, concentrated magnetic oxides are well described in
terms of the “m-J paradigm”: there are localized magnetic moments m on the
cations, and superexchange or double-exchange interactions to couple them.
The paradigm provides a good account of both the magnetic order and the
spin waves. Exchange interactions are normally antiferromagnetic, and

magnetic-ordering temperatures do not exceed 1000 K.
10.2 DILUTE OXIDES
So much for magnetic general knowledge. Our concern in this chapter is
dilute magnetic oxides with general formula
(M1− xTx )On (10.3)
where:
M is a nonmagnetic cation
T is a transition-metal cation
n is an integer or rational fraction
x < 0.1 (10%)
Typical values of x are a few percent.
An important limit is xp, the percolation threshold [11], where continu-

ous nearest-neighbor paths first appear, which link M cations throughout
the crystal. We are interested in magnetic bond percolation, where cations
are regarded as neighbors when their spins are coupled via a superexchange
bond. Below xp there are only isolated cations, and small clusters of nearest-
neighbor pairs, triplets, etc. Above xp there is a bulk cluster, which encom-
passes most of the magnetic cations. Provided the exchange interactions only
involve nearest neighbors, there is no possibility of long-range order below xp,
which is approximately 2/Z0 [12], where Z0 is the cation coordination number.
Depending on the structure, Z0 lies between 6 and 12, which means that xp
is in the range 16%–33%. Our dilute oxides fall below the percolation thresh-
old, and should not be expected to order magnetically. A random distribution
of magnetic dopant ions is illustrated schematically in Figure 10.5a. At low
concentrations, most of the dopants are isolated, with no magnetic nearest
neighbors, and a Curie-law susceptibility is expected for them:
x1C
χ1 = (10.4)
T
Here x1 is the fraction of isolated ions, and C is the Curie constant
µ0Ng2µB2S(S + 1)/3kB, where N is the number of cations per unit volume, and
g = 2 for spin-only moments. As the concentration increases, there will be an
increasing proportion of dimers and larger groups. The dimers (and other
even-membered groups) have a Curie–Weiss susceptibility
x 2C
χ2 = (10.5)
(T − θ)
where:
x2 is the fraction ions present as dimers
θ is a negative constant of order −10 to −100 K, which is proportional
to the antiferromagnetic exchange coupling J0
10.2 Dilute Oxides 421
Isolated polaron
Antiferromagnetic pair
Isolated ion
Overlapping polarons
(a) (b)
FIGURE 10.5 (a) Schematic representation of distribution of dopant ions in a DMS. (b) The same, but with
donor defects (◻) that create magnetic polarons where the dopant ions are coupled ferromagnetically.
The odd-membered groups with a net moment contribute a modified

Curie-law susceptibility. With all this, deviations from a linear response to
an applied field of up to 1 T, for example, are only perceptible below about
10 K. The room-temperature susceptibility is of order 10−3 x (x is defined in
the formula (10.3)). This is the behavior expected for a dilute magnetic oxide,
and indeed it is exactly what is found in bulk crystalline materials and in
well-crystallized, defect-free thin films. There are numerous examples in the
literature, for example, Refs. [13–17].
The conventional picture of magnetism in insulators or magnetic semi-
conductors is at a loss to account for high-temperature ferromagnetism in
dilute magnetic oxides. The Curie or Néel temperatures of oxides were plot-
ted on a histogram in Figure 10.3. None exceeded the Néel point of αFe2O3,
which is 960 K. The ordering temperatures in solid solutions vary as x or as
x1/2, depending on the nature of the interactions [18]. No order is expected at
room temperature when x < 10%, or in insulators at any temperature when
x < xp. The magnetic behavior we would normally expect from the m-J para-
digm is a superposition of Curie-law paramagnetism for isolated ions and
Curie–Weiss behavior for the small, antiferromagnetically coupled groups
of two, three, or more ions. Indeed, this is precisely what is found in well-
crystallized, bulk material [15].
It therefore came as a surprise when, at the beginning of 2001,
Matsumoto et al. published their claim that anatase TiO2 doped with 7%
cobalt was ferromagnetically ordered at room temperature. There followed a
deluge of similar reports on other materials. The first samples were all thin
films, but soon reports of high-temperature ferromagnetism were appearing
for nanoparticles and nanocrystalline material as well. Table 10.2 lists some
of these early reports. There are now well over 2000 publications on dilute
ferromagnetic oxides, about half of them on ZnO, and new ones continued
TABLE 10.2
Early Reports of Ferromagnetic Oxide Thin Films with TC above
Room Temperature
Moment
Material Eg (eV) Doping (μB) TC(K) Reference
TiO2 3.2 V—5% 4.2 >400 Hong et al. [37]
Co—7% 0.3 >300 Matsumoto et al. [38, 39]
Co—7% 1.4 >650 Shinde et al. [40]
Fe—2% 2.4 300 Wang et al. [41]
SnO2 5.5 Fe—5% 1.8 610 Coey et al. [42]
Co—5% 7.5 650 Ogale et al. [43]
HfO2 3.2 Co—2% 9.4 >400 Coey et al. [44]
Fe—1% 2.2 >400 Hong et al. [45]
CeO2 Co—3% 6.1 725 Tiwari et al. [46]
In2O3 2.9 Mn—5% Sn 0.8 >400 Philip et al. [47]
Fe—5% 1.4 >600 He et al. [48]
Cr—2% 1.5 900 Philip et al. [49]
ZnO 3.4 V—15% 0.5 >350 Saeki et al. [50]
Mn—2.2% 0.16 >300 Sharma et al. [51]
Fe 5%, Cu 1% 0.75 550 Han et al. [52]
Co—10% 2.0 280–300 Ueda et al. [53]
Ni—0.9% 0.06 >300 Radovanovic and Gamelin [54]
to appear every month. Many of these are electronic structure calculations

for dilute magnetic oxides, with various crystalline defects.
The experimental results were astonishing for three reasons:
1. The magnetism appears in oxide materials where the doping is far

below xp.
2. The magnetism appears to be ferromagnetic, whereas superexchange
in oxides is usually antiferromagnetic.
3. The samples appear to be ferromagnetic at room temperature and
above, although no dilute magnetic material had ever been found to
be magnetically ordered at room temperature.
In these circumstances, the claims of ferromagnetism were met with

deep skepticism, and the conviction that the high-temperature ferromag-
netism must somehow be associated with a segregated, magnetically con-
centrated impurity phase, or else it was simply a measurement artifact.
Otherwise, one is led to believe that the high-temperature ferromagnetic-
like behavior of dilute oxides is a new and significant magnetic phenomenon.
Although the field is fraught with irreproducibility, and impurity phases
are often implicated, such is the view of the present author. It is difficult
to digest the flood of often-contradictory experimental information, that is
now beginning to slacken. There are some helpful reviews, for example by
Ogale (2010) [19], Fukumura and Kawasaki (2013) [20], and Hong (2016) [21].
Our approach here is to present experimental data rather than electronic
structure calculations on six main oxide systems, and then to focus on physi-
cal models that could account for important aspects of the phenomena.
10.3 MAGNETIC MEASUREMENTS
Before considering the experimental evidence, some words of caution are in
order. We are dealing mainly with films, about 100 nm thick. If x = 1%, there
may the equivalent of just four layers of ferromagnetic cations in the film. The
ferromagnetic volume fraction of a typical 5 × 5 mm2 sample on its substrate,
which is usually about 0.5 mm thick, is just two parts per million. Assuming
1 µB per cation, the ferromagnetic moment is approximately 1.5 × 10−8 A m2
(1.5 × 10−5 emu). While this is comfortably above the sensitivity of a super-
conducting quantum interference device magnetometer (SQUID), an alternat-
ing-gradient force magnetometer (AGFM), or even a good vibrating-sample
magnetometer (VSM), it does represent a very small moment. A 0.1 µg speck
of iron or magnetite has a moment of this magnitude. These contaminants are
ubiquitous, and some care is needed to be sure that what is being measured
is the sample, and not magnetic contamination. Co nanoparticles are diffi-
cult to detect, but Fe and Fe3O4 are readily picked up by Mössbauer spectros-
copy. Element-specific x-ray absorption is also useful. Secondary phases are
an important consideration, and are probably the complete explanation of the
ferromagnetism in certain cases. Other high-temperature ferromagnets are
spinel phases, where the antiferromagnetic interactions lead to ferromagnetic
structures where the moment does not exceed 1.5 µB per magnetic cation. Low
concentrations of nanoscale ferromagnetic inclusions may elude detection by
x-ray diffraction, but should be detectable by careful transmission-electron
microscopy. In magnetic measurements, they are expected to behave super-
paramagnetically, with the characteristic signature of anhysteretic magnetiza-
tion curves that superpose perfectly when plotted as a function of H/T.
Several publications have emphasized how bogus magnetic signals may
arise from markers, kapton tape, silver paint, contamination from steel twee-
zers, unsuitable substrate holders, and other sources of external pollution
[22–27]. It is essential, as part of the experimental protocol, to run blank,
control samples that have been treated in exactly the same way as those con-
taining the magnetic cations. In order to make a case that magnetic cations
or unpaired electrons in the oxide film are the source of the observed mag-
netism, precautions in sample preparation and handling are imperative, to
do justice to the sensitivity of the magnetometer.
The problems for nanocrystalline material are different. Here the
moments reported are of order 2 × 10−4 A m2 kg−1 (2 10−4 emu g−1). The issue
now is sample purity; the presence of ferromagnetic metals or oxides at the
ppm level has to be discounted. The specific moments of Fe, Co, Ni, and
Fe3O4 are, respectively, 217, 162, 55, and 92 A m2 kg−1 (emu g−1 is an equiva-
lent unit); their magnetizations are 1710, 1440, 490, and 480 kA m−1 (emu cc−1
is an equivalent unit).
For thin films deposited on substrates, the huge mismatch between the
mass of the thin film sample—typically some tens of micrograms—and that
of the substrate—which is about a thousand times greater—presents a chal-

lenge. Commonly used substrates, sapphire (Al2O3), MgO, SrTiO3, LaAlO3,
and Si, are all diamagnetic with susceptibilities of −4.8, −3.1, −1.3, −2.7, and
−1.5 × 10−9 m3 kg−1. The problem is illustrated in Figure 10.6. The substrate
gives a diamagnetic signal, which is perhaps reproducible to within 1%, given
the uncertainties in positioning, and centering the substrate in the mag-
netometer used to measure the hysteresis loop. It is therefore practically
impossible to determine the diamagnetic susceptibility of the undoped film.
A Curie-law signal due to paramagnetic dopant ions can be readily detected
at low temperatures, but it is difficult to measure any high-field slope that
may be associated with the thin film.
The measured response for a ferromagnetic thin film is shown in
Figure 10.6b. The magnetization appears to saturate in a field of order 1 T,
and the high-field slope becomes that of the substrate, within experimental
error. The procedure is then to suppose the magnetization is indeed satu-
rated, and subtract the high-field diamagnetic slope as the substrate correc-
tion, yielding the magnetization curve for the film shown in Figure 10.6c.
Typically, these ‘ferromagnetic’ hysteresis loops for dilute magnetic oxides
exhibit little or no coercivity (≈10 mT), and a small remanence ratio
(Mr/Ms ≈ 5%–10%). Trapped flux in a SQUID, or the remanence of an iron-
cored electromagnet create fields of this magnitude, so it is important to
measure precisely the field actually acting on the sample. A flaw in the
additivity argument of the moments of thin film and substrate is that it
ignores the possible emergence of a new electronic or magnetic state on
one side or the other of the interface [28]. Best known is the example of a
thin film of LaAlO3 on (100) SrTiO3 when a two-dimensional electron gas
20
10
0
–10
(a)
Moment (10–8 Am2)
–20
10
0
–10
(b)
5
0
–5
(c)
–10
–1.0 –0.5 0.0 0.5 1.0
µ0H (T)
FIGURE 10.6 Data reduction for magnetization measurements on a thin film of a

typical dilute magnetic oxide: (a) diamagnetic substrate, (b) ferromagnetic thin film
with diamagnetic substrate, (c) ferromagnetic component isolated from (b).
that shows evidence of magnetic order at liquid-helium temperatures forms

in the SrTiO3 provided the thickness of the LaAlO3 overlayer exceeds four
unit cells [29–31]. The common explanation is charge transfer of about 0.5
electrons per unit cell into 3d(Ti) states just below the interface, although
oxygen vacancies in SrTiO3 just below the interface may also contribute [32,
33]. The charge transfer tends to occur towards the oxide with the lower
bandgap, so in films of LaMnO3 on SrTiO3, the charge buildup in the man-
ganite lead to an abrupt switch from antiferromagnetic to ferromagnetic
order when the cap layer thickness 6 unit cells [34].
A more recent study of polished (100), (110) and (111) SrTiO3 crystal
slices from four different suppliers has revealed a new magnetic phenome-
non related to oxygen vacancies near the surface. This is quite different from
the effects of the magnetic impurities that are present to a greater or lesser
degree these substrates — paramagnetic 3d impurities, usually present in
ppm or sub ppm amounts, that give a characteristic Curie-law upturn in the
susceptibility at low temperature and hysteretic ferromagnetic Fe-Ni inclu-
sions in the surface, probably introduced by the polishing process. The new
effect is a ferromagnetic-like response with little or no hysteresis or tempera-
ture dependence, at least between 4 K and 400 K. The effect is increased by
heating the substrate under vacuum, and it is increased by heating the sub-
strate under vacuum, and reduced by treating the surface with a Ti-specific
electron-donor catechol molecule [35]. These studies illustrate the potential
difficulties in disentangling the magnetic response of a thin film from that
of the substrate on which it rests. However SrTiO3 may be an extreme case.
Other substrates such as c- or r-cut sapphire appear to be magnetically inert.
Units of µB nm–2 are useful when the surface or interface is the source of
the magnetic signal. The oxides may even be blank substrates [35, 36, 204].
In SrTiO3, for example, anhysteretic, saturating magnetic moments of up to
60 µB nm–2 are found [35].
At first, it seemed reasonable to quote the magnetization of the samples in
units of Bohr magnetons per 3d cation, normalizing the measured magnetic
moment by the number of transition-metal cations thought to be in the sample.
This unit is useful when comparing different doping concentrations and also
when comparing different host materials, provided the transition-metal con-
centration is accurately known. However, the practice is misleading when the
oxides turn out not to be dilute magnetic semiconductors (DMSs), and the
dopant cations are not the primary source of the magnetic order. Units of µB
nm−2 are useful when the surface or interface is the source of the magnetic sig-
nal. The oxides may even be blank substrates [34–36]. In SrTiO3, for example,
anhysteretic, saturating magnetic moments of up to 30 µB nm−2 are found [36].
We now present experimental results on a selection of dilute oxides,
which can exhibit ferromagnetic behavior when they are in thin film or
nanocrystalline form. The crystal structures of host materials are presented
in Figure 10.7. Some structural details are included in Table 10.3, including
the percolation threshold xp, the cation site coordination, and bandgap Eg.
The systems are now discussed in turn.
(a) (b) (c)
(d) (e) (f)
FIGURE 10.7 Crystal structures of host semiconductors: (a) TiO2 (anatase), (b)
SnO2, TiO2 (rutile), (c) HfO2, (d) CeO2 (fluorite), (e) In2O3 (bixbyite), and (f) ZnO
(wurtzite), oxygen are the large dark ions.
TABLE 10.3
Parameters for Some Oxides
Material Structure ε m*/m Coordination xp
TiO2 Anatase 37 0.4 Octahedral 0.25
SnO2 Rutile 25 0.4 Octahedral 0.25
HfO2 Monoclinic 26 0.5 Sevenfold 0.18
CeO2 Fluorite 9 0.3 Cubic 0.18
In2O3 Bixbyite 8 0.3 Octahedral 0.18
ZnO Wurtzite 9 0.3 Tetrahedral 0.18
ε, static dielectric constant; m*/m, effective electron mass
10.4 RESULTS ON SPECIFIC OXIDES

ioxides
10.4.1 D
10.4.1.1 TiO2
Anatase with 7% Co-doping was the first example of high-temperature
ferromagnetism in a dilute oxide film [38]. The same group reported fer-
romagnetism in rutile films [39]. The solubility of Co in these oxides is low,
and there is the possibility of cobalt-metal clustering [55, 56]. But films with
1%–2% of Co can be prepared, which are apparently free of clusters, and

exhibit ferromagnetism with a moment of ≈1 µB/Co up to 700 K [40]. Other
evidence that Co-doped TiO2 can be intrinsically ferromagnetic, and carrier-
mediated include: electric-field modulation of the magnetization [57], obser-
vation of band-edge optical dichroism [58], and anomalous Hall effect [59],
as well as tunnel magnetoresistance [60].
Other cation dopants reported to make TiO2 ferromagnetic include
V–Ni [37], Cr [16], Mn [41], Fe [61], and Cu [62, 63]. In some highly perfect,
Fe-doped films, an Fe3O4 secondary phase forms at the surface [64]. It has
been shown that highly perfect Cr-doped TiO2 films are not ferromagnetic,
whereas films with structural defects can be [65]. The Cr is trivalent in TiO2
films, and the magnetization of (001) anatase films grown on LaAlO3 can be
highly anisotropic [66], but the magnetism again disappears as the structural
quality improves [67]. The order of magnitude of the magnetization of the
doped TiO2 films is 10–20 kA m−1.
TiO2, unlike ZnO or SnO2, is usually quite insulating in thin film form.
Griffin et al. [68–70] have established that free carriers are not necessary
for the high-temperature ferromagnetic order in Co-TiO2, and that it does
not depend critically on the uniformity of the cobalt distribution. The mate-
rials may be dilute magnetic dielectrics, rather than DMSs. Nevertheless,
Fukumura et al. [71] concluded from a detailed study of magnetization,
anomalous Hall effect, and magnetic circular dichroism of epitaxial films
with different cobalt contents and carrier densities, that the ferromagne-
tism is related to the electron density. Many studies indicate that a critical
requirement for ferromagnetism in TiO2 films is an abundance of oxygen
vacancies, Ov. These form easily in titanium oxide, as is evident from the
existence of a sequence of Magneli phases TinO2n−1.
Giant moments of 4.2 µB/V were reported for 5% V-doped TiO2 [37], and
moments as large as 22.9 µB/Co have been reported for 0.15% Co-doping [72].
These numbers imply that the transition-metal cation cannot be the car-
rier of the moment. Ferromagnetism with a magnetization of ≈10–20 kA m−1
at temperatures up to 880 K has also been reported in 200–300 nm thick
films of undoped anatase-TiO2 deposited on LaAlO3 [73, 74]. It has been
shown by Zhou et al. [75] that a TiO2 crystal irradiated with a dose of 5 × 1015
cm−2 of 2 MeV oxygen ions develops a tiny magnetization of 2 A m−1. The
induced defects are Ti3+–Ov complexes. Oxygen vacancies are also thought
to be responsible for the magnetism of TiO2 nanoribbons [76]. Other proce-
dures that have the effect of introducing cation vacancies by nitrogen doping
(1 µB/N) [77], Ta doping [78], or simply by Ti substoichiometry in anatase [79]
all produce material with room-temperature ferromagnetism, in the absence
of substitution of any magnetic 3d cation for Ti. These results may be inter-
preted in terms of the magnetism of strongly-correlated 2p holes [80].
10.4.1.2 SnO2
First reports of high-temperature ferromagnetism in SnO2 films came from
Ogale et al. [43], who found a Curie temperature of 650 K, and an extraordi-
nary moment of 7.5 Bohr magnetons per cobalt atom for films doped with
5% Co. High-temperature ferromagnetism was subsequently observed for

films doped with V [81], Cr [82], Fe [42], Mn [83], and elements from Cr–Ni.
[84]; in some of these films, the moments again exceed the cation spin-only
values [18]. In the case of V-doping, for example, the results depend rather
critically on the choice of substrate [85, 86]. Co-doped films exhibit Faraday
rotation of 570° cm−1, and show some sign of an anomalous Hall effect [87].
The magnetization of the 3d doped SnO2 films is typically 20 kA m−1.
There is a remarkable anisotropy of the magnetization when the films
grow with a (101) texture, which is found in films with V, Mn, Fe, and
Co-doping [84, 86], and even in undoped films [88]. The effect cannot be
explained by magnetocrystalline anisotropy, because the ferromagnetic
moments measured in different directions do not converge in high fields.
The anisotropy becomes very pronounced at low temperature, and it has
been modeled in terms of orbital current loops [89].
There is much evidence that the magnetism is somehow associated with
defects related to the oxygen stoichiometry. It was originally suggested that
these could be F-centers—an oxygen vacancy that traps an electron, which
serves to couple the magnetic dopant ions [42]. The defects are not necessar-
ily adjacent to the cation dopant, which are usually octahedrally coordinated
in substitutional sites. SnO2 forms dilute magnetic oxides not only in thin
film form, but also in nanoparticles [90–92] and nanocrystalline ceramics
[93, 94]. A significant experiment by Archer and Gamelin [95] demonstrated
that the magnetism of colloidal, 2 nm Ni-doped SnO2 nanoparticles could
be switched on and off as they aggregate under gentle thermal annealing at
100°C. The ferromagnetism is controlled by grain-boundary oxygen defects,
as was shown in similar experiments on colloidal particles of Ni-doped ZnO
[54] and Co-doped TiO2 [96], which become strongly ferromagnetic when
spin coated into thin films. Magnetic force microscopy of Fe-doped SnO2
thin films shows magnetic contrast at the grain boundaries [84].
There are also reports of magnetic moments in undoped SnO2 films [88],
nanoparticles [97], and nanowires [98]. Magnetization is of order 10 kAm−1,
10 Am−1 and 10 kAm−1, respectively. In the latter case, ferromagnetism is
enhanced by UV irradiation. Cation doping with Li in films [99], or nanopar-
ticles [100], or with Mg in films [101] produces magnetism that is associated
with 2p holes. Magnetization is of order 10 kA m−1 and 10 A m−1, respectively.
10.4.1.3 HfO2
A landmark in the study of ferromagnetism in oxide thin films was the
report by Venkatesan et al. [102] of high-temperature ferromagnetism in
undoped HfO2 films. These results helped shift the focus from dopants to
defects, and led to coining the term d0 magnetism for the weak, anhyster-
etic high-temperature magnetism of a wide range of materials that are not
normally expected to be magnetic, because they contain no 3d ions [18]. A
subsequent investigation [44] of more than 40 HfO2 films of different thick-
ness, deposited on different substrates by pulsed-laser deposition, with and
without doping, established that moments were largest on c-cut sapphire
and smallest on Si; they showed no systematic variation with film thickness
(except that the thinnest films were not magnetic). Dopants, whether mag-
netic or nonmagnetic, tended to reduce the moment. The moments were
widely scattered around an average of ~200 µB nm−2, but they were unsta-
ble, decaying over times of order six months. Magnetization of the thinnest
films was comparable to that of Ni (400 kA m−1), and in some samples, it was
very anisotropic, being larger when the field was applied perpendicular to
the film, than when it was applied parallel. The results were substantially
confirmed in a series of papers by Hong and coworkers [74, 45, 103, 104]
who looked at HfO2 undoped, or with Fe or Ni dopants. Oxygen annealing
reduced the moment, which was associated with defects near the film/sub-
strate interface. However, other studies could find no intrinsic ferromagnetic
signal in similarly prepared HfO2 films, irrespective of oxygen pressure dur-
ing deposition [105, 106]. It was shown by Abraham et al. [22] that similar
signals could be observed in samples manipulated with stainless steel twee-
zers, and it is now standard practice to manipulate thin film samples with
plastic or wooden tools.
A study of colloidal HfO2 nanorods 2.6 nm in diameter [107] found small
magnetic moments ~100 A m−1. Furthermore, a magnetization of comparable
magnitude can be reversibly induced by heating pure HfO2 powder in vacuum
or air at 750°C [44]. Oxygen vacancy-related magnetism has also been iden-
tified in nanorods [108], and amorphous HfO2 nanostructures [109]. It has
been shown that magnetism and oxygen vacancy content can be controlled
by magnetic field annealing [110]. However, a detailed study by Hildebrandt
et al. of monoclinic HfO2-x grown on c-cut sapphire by reactive molecular
beam epitaxy for a range of oxygen stoichiometry found no sign of room-
temperature ferromagnetism, even in films that exhibit p-type conductivity
[111]. Oxygen vacancies may be necessary, but they cannot be the whole story.
10.4.1.4 CeO2
The first report of ferromagnetism in CeO2 was by Tiwari et al. [46] who found
a moment of 6.1 µB/Co in 3% Co-doped films that increased with temperature
up to 725 K; it corresponds to a magnetization of 40 kA m−1, and is too big
to be explained by the presence of Co metal in the films. The magnetization
depends on the substrate and oxygen pressure during growth [112, 113]. Very
much smaller magnetization has been found in CeO2 nanoparticles that are
undoped [97, 114–116] or doped with Ni [117] or Ca [118]. Values range from
1 to 5000 A m−1, and magnetization is greatest in sub-20 nm particles with no
3d doping [114]. The high-temperature ferromagnetism in the nanoparticles
is associated with Ce3+–Ce4+ surface pairs [116], rather than surface oxygen
vacancies [115]. Trivalent rare-earth doping has a decided influence on the
magnetism of CeO2. Pr first increases the magnetism of thin films, and then
decreases it when the Pr content > 10% [119]. In nanoparticles, the moment
declines when the Pr content > 1% [120].
Cerium dioxide nanoparticles are perhaps the oxide materials that most
reliably exhibit d0 ferromagnetism, although the magnitude of the room-
temperature moment depends on the form of the particles and the prep-
aration method [121]. A thorough study of the magnetism of 4 nm CeO2
FIGURE 10.8 Magnetic properties of 4 nm CeO2 nanoparticles. (a) Effect of La doping on the magnetization,
(b) Temperature-independent anhysteretic magnetization curves of a sample containing 1 wt% La, (c) Effect
of physical dilution of a 4 mg sample with similar masses of different magnetically-inert powders, showing
a characteristic length scale of order 100 nm for the appearance of magnetism in agglomerates of the CeO2
nanoparticles, and (d) Comparison of UV absorption spectra of magnetic (upper two curves) and nonmag-
netic (lower two curves) CeO2 powders. (After Coey, J.M.D. et al., Nature Phys. 12, 694, 2016. With permission.)
nanoparticles [121] revealed two unexpected results. First was the need for a
small amount of trivalent rare-earth doping to switch on the ferromagnetic-
like response. The dopant does not have to be magnetic; 1% of La produces
the best response (Figure 10.8a). More remarkable is the observation that the
CeO2 nanoparticles have to be in contact to form subassemblies greater than
a certain size for the magnetism to appear. There is a characteristic length
scale associated with the magnetism of about 300 nm, which is associated
with the appearance of UV absorption around this wavelength in the mag-
netic samples (Figure 10.8d). The explanation proposed in terms of giant
orbital paramagnetism is discussed in § 10.5.1. There is a comprehensive
review of the room-temperature magnetism in CeO2 [206].
10.4.2 Sesquioxides
10.4.2.1 In2O3
Indium oxide, In2O3, has the cubic bixbyite structure, with indium in
undistorted 8b and highly distorted 24d octahedral sites (Figure 10.7e).
The structure is related to fluorite (Figure 10.7d) with a quarter of the oxy-
gen atoms removed. The bandgap E0 is 2.93 eV [122]. The oxide is usually
oxygen-deficient, creating donor levels near the bottom of the conduction
band. Electron concentrations of order 1024 –1026 m−3 lead to resistivities of
about 10 µΩ m, a high mobility, and an effective mass of 0.3me. The electrons
tend to accumulate at the oxide surface. Substituting some of the In by Sn is
a convenient way of controlling the carrier concentration in In2O3. Indium
tin oxide (In1−xSnx)2O3 (ITO) with x ≈ 0.1 is a widely used transparent n-type
conducting oxide.
The first report of ferromagnetism in Mn-doped ITO by Philip et al. [47]
showed a moment of about 1 µB per manganese, and evidence of an anoma-
lous Hall signal. It was followed by indications of ferromagnetism in In2O3,
for bulk material doped with Fe and Cu [123], and for oxygen-poor thin films
doped with Fe [48, 124] and other dopants [125–127]. Subsequent investiga-
tions of ferromagnetic Mn- and Fe-doped ITO films [128] found that the man-
ganese behaved paramagnetically at low temperature, and that the moment
was not the primary source of the magnetism. The presence of paramagnetic
manganese in mixed valence states (Mn2+ and Mn3+) [129] points to charge
transfer ferromagnetism (§10.5.1). The anhysteretic magnetization curve,
shown in Figure 10.9, may be regarded as typical of many ferromagnetic dilute
magnetic oxide thin films. Hysteretic ferromagnetism of the iron-doped sam-
ples was due to an Fe3O4 secondary phase. Anhysteretic ferromagnetism was
observed in Fe-doped films thought to be intrinsically ferromagnetic [124].
A study of Cr-doped In2O3 by Philip et al. [49] demonstrated that the
ferromagnetism was related to carrier concentration, appearing at a con-
centration ne = 2 × 1025 m−3, which would correspond to the formation of an
impurity band [130]. Moments of up to 1.5 µB/Cr were found for 2% doping.
Chromium is a good dopant, because there is little prospect of ferromag-
netic secondary phases; CrO2 is ferromagnetic, with TC = 390 K, but TC was
2
300 K
200 K
150 K
1 100 K
Moment (10–8 A m2)
50 K
4K
–1
–2
–1.0 –0.5 0.0 0.5 1.0
µ0H (T)
FIGURE 10.9 Ferromagnetic magnetization curves of a thin film of ITO doped

with 5% Mn, at different temperatures. The manganese in these films is paramag-
netic. (After Venkatesan, M. et al., J. Appl. Phys. 103, 07D135, 2008. With permission.)
measured in this study to be ≈900 K. Domains were observed by magnetic

force microscopy (MFM). Further investigations by Panguluri et al. [131] on
both undoped and Cr-doped films found paramagnetic Cr coexisting with
a film moment of about 500 A m−1. Both materials exhibited a spin polar-
ization of 50% in measurements of point-contact Andreev reflection. The
implication is that the moment of the 3d dopant is inessential for the fer-
romagnetism, which is associated with a narrow defect-related band. This
picture is consistent with the dependence of the anomalous Hall effect on
carrier concentration [127].
There are also reports of ferromagnetism of similar magnitude in
undoped thin films [126], with smaller values in undoped nanoparticles
[129, 205]. Sundaresan et al. [97] found very small moments in 12 nm
nanoparticles of In2O3 (<5 A m−1), but in another study by Bérardan et al.
[132], In2O3 nanoparticles prepared in a variety of different conditions were
found to be completely diamagnetic, and transition-metal doped samples
were paramagnetic. There is a recent summary of the magnetic properties of
doped and undoped In2O3 and its derivatives [133].
onoxides
10.4.3 M
10.4.3.1 ZnO
Although the first report of ferromagnetism in thin films of a dilute mag-
netic oxide was for TiO2, it is on ZnO that the majority of the investigations
have been carried out. Zinc oxide is a promising semiconductor material with
a bandgap of 3.37 eV. It crystallizes in the wurtzite structure (Figure 10.7f),
where Zn occupies tetrahedral sites. The oxide has a natural tendency to be
n-type on account of oxygen vacancies or interstitial zinc atoms. It can be
doped n-type with atoms such as Al, and it has been possible to make nitro-
gen-doped p-type material, which opens the way to producing light-emitting
diodes and laser diodes. Anion doping with Bi is a new approach to creating
spin-polarized p-states in ZnO [134]. Extensive reviews of the semiconduct-
ing properties of ZnO are available [135, 136].
Various cations can replace zinc in the structure, and small dopings
of all the 3d elements have somewhere been reported to make ZnO ferro-
magnetic. There is an admirable review of the voluminous early literature
on doped ZnO by Pan et al. [137]. The largest moments are found for Co2+.
The Co2+ ion gives a characteristic pattern of optical absorption in the band-
gap, due to crystal field transitions of the 4F ion in tetrahedral coordination.
Studies of the magnetic properties of bulk material doped with Co [15, 138]
or Mn [139] show only the paramagnetism expected for isolated ions, and
small antiferromagnetically coupled nearest-neighbor clusters that arise sta-
tistically at a low doping level, as discussed above (Figure 10.5a).
The range of solid solubility of cations of the 3d series in ZnO films
prepared by pulsed-laser deposition was established by Jin et al. [140];
solubility limits as high as 15% were found for Co, but most other cations
could be introduced at the 5% level. None of these films turned out to be
to be magnetic, but following Ueda et al. [53], who were the first to report
high-temperature ferromagnetism in their films containing cobalt at the

10% level, there have been a great many reports of ferromagnetic moments
in films doped with Co (see among others Janisch et al. [141], Prellier et al.
[142], Pearton et al. [143], and references therein), as well as every other 3d
dopant from Sc to Cu [50, 51, 54, 144]. The variation of magnetic moment
for a series of films nominally doped with 5% of various transition-metal
ions prepared by pulsed-laser deposition in the same conditions is shown
in Figure 10.10. There are also numerous counterexamples where no room-
temperature moment has been found in doped films, or an extrinsic origin
has been established. Negative results tend to be underreported in the lit-
erature, so the observation of saturating magnetism in ZnO and other oxide
films is probably less prevalent than one might imagine. It is very sensitive
to deposition conditions such as substrate temperature and oxygen pres-
sure [135, 137]. However, nanoparticles and nanorods of ZnO with cobalt
and other dopants have also been found to be magnetic under certain pro-
cess conditions [145], although the moments per cobalt atom are one or
two orders of magnitude less than those found for the thin films, which are
typically 0.1–1 µB/Co.
Some progress has been made toward providing a systematic experi-
mental account of the phenomenon. Narrow process windows have been
delimited where the magnetism can be observed, which differ depending
on the deposition method—pulsed-laser deposition, sputtering, evaporation,
organometallic chemical vapor deposition, and others. With Cr and Mn, for
example, it is difficult to obtain ferromagnetic moments in n-type material,
whereas p-type samples can exhibit the symptoms [112] (Figure 10.11). The
substrate temperature required for the magnetism is often around 400°C,
where the films are not of the best crystalline quality. This implies a point-,
line-, or planar-defect-related origin [146]. Other evidence in this sense
comes from the appearance of magnetism in Zn-doped and oxygen-deficient
2
σ (µB/M)
0
Sc Ti V Cr Mn Fe Co Ni Cu Zn
3d Dopant (5 at%)
FIGURE 10.10 Magnetic moment measured at room temperature for ZnO thin
films with various 3d dopants (After Venkatesan, M. et al., Appl. Phys. Lett. 90,
p-type n-type
M (µB/Mn2+)
0
IMCD
IMCD
Mn
–1 S = 3/2 S = 5/2
4 0 0.2 0.4 0 0.2 0.4
MCD
M (×10–2 µB/Co2+)
βH/2kT βH/2kT
0
0.5° 0.02°
Co
–4 Co Mn
–0.3 0 0.3 –0.3 0 0.3 28 24 20 16 28 24 20 16
(a) µ0H (T) (b) Energy (103 cm–1)
FIGURE 10.11 (a) Development of magnetism in n-type ZnO with Co or p-type ZnO with Mn. (After
Kittilstved, K.R. et al., Nat. Mater. 5, 291, 2006. With permission.) (b) MCD spectra and the magnetic field
dependence of the intensity of the MCD signal (insets) recorded at different energies in ZnO doped with Co
(left) and Mn (right).
films [147, 148], and the creation of a large magnetization of ~200 kA m−1 in
the volume implanted by 100 keV Ar ions [149].
Evidence that Co-doped ZnO is an intrinsically ferromagnetic semicon-
ductor is relatively sparse. No magnetoresistance has been observed at room
temperature, despite the high TC, nor are there clear signs of an anomalous
Hall effect. Semiconductor-like magnetoresistance can, however, be observed
below about 30 K [150]; there is a positive component, as expected in the clas-
sical two-band model, and a negative component due to ionized impurity scat-
tering. There is also quite a large anisotropic magnetoresistance below 4 K
(≈10%), which is a band effect in the semiconductor, not ferromagnetic AMR.
Furthermore, there is band-edge magnetic dichroism with a paramagnetic
dichroic response from the dilute Co2+ ions in tetrahedral sites in the wurtzite
lattice, as well as a component with a ferromagnetic response [151, 152]. The
ferromagnetic dichroism has been observed for Ti, V, Mn, and Co-doping [152].
Magnetic circular dichroism experiments on single nanoparticles in the elec-
tron microscope by Zhang et al. [153] indicate that their particles are intrinsi-
cally ferromagnetic at room temperature when doped with Co, but not with Fe.
Some reports [154, 155] have related the carrier concentration to the
existence of ferromagnetism, as would be expected if ZnO were a DMS.
Results of Behan et al. [155], based on a systematic investigation of samples
with 5% Co and carrier densities ranging from 1016 to 1021 cm−3 that the mate-
rial is a dilute magnetic insulator at low carrier concentrations and a DMS
at high concentrations, with an intermediate concentration range around
1019 cm−3 where the ferromagnetism is quenched. However, the interaction
between the conduction electrons and the Co spins seems to be rather weak,
<20 K, and in other studies the magnetic properties are little influenced by
changing the carrier concentration by Al doping, for example [156]. It has
been shown that the sign of the weak Mn–Mn exchange in colloidal Mn–
ZnO nanocrystals can be changed from antiferromagnetic to ferromagnetic,
by exciting carriers optically [157], although the strength of the interaction
is only a few Kelvin. It seems that there may be a correlation between carrier
concentration and ferromagnetism, but this is not a cause and effect relation.
A problem in interpreting the magnetic data for some dopants is the
tendency to form ferromagnetic impurity phases, which may escape detec-
tion by x-ray diffraction. For cobalt in particular, absence of evidence can-
not be taken as evidence of absence. There is a well-documented tendency
for nanometric coherent cobalt precipitates to appear in the ZnO films
[158–161], and single crystals [162], which exhibit ferromagnetic proper-
ties, especially at doping levels above 5%. An impurity-phase-based expla-
nation (Mn2−xZnxO3) has been advanced also for Mn-doped material [163].
For other dopants, such as vanadium or copper, contamination by high-
temperature ferromagnetic impurity phases is improbable.
In some cases, there are features of the data, which make an impurity-
phase explanation untenable. These are (a) observation of a moment per
transition-metal ion that is greater than that of any possible ferromagnetic
impurity phase based on the dopant ion, and which may exceed the spin-
only moment of the cation, and (b) observation of an anisotropy of the mag-
netization, measured in different directions relative to the crystal axes of
the film, which is not a feature of any known ferromagnetic phase at room
temperature [102, 144, 164]. An example is shown in Figure 10.12. Both the
features have occasionally been reported in the other thin film oxide systems
we have been considering.
A common feature of the ferromagnetism in ZnO and the other oxides
is its tendency to decay over times of order of a few months [24, 164]. This has
led to the name phantom ferromagnetism, for the fickle syndrome presented
by dilute magnetic oxides.
The tunnel spin polarization of transition-metal-doped ZnO has been
measured [165]. Although poorly reproducible, these results are in line with
the observation of MCD at the band-edge as discussed above. The possibility
to obtain and manipulate a spin current in ZnO could be of great interest, as
the spin lifetime in ZnO is rather long [166].
There is much evidence to suggest that ferromagnetism in transition-
metal-doped ZnO is unambiguously correlated with structural defects,
8
Field perpendicular to film plane 0.6 µB/V
6 Field along diagonal
Field along edge
4
Moment (10–8 Am2)
2 0.2 µB/V
0
–2
–4
–6
–8
–1.0 –0.5 0.0 0.5 1.0
µ0H (T)
FIGURE 10.12 Anisotropy of the magnetization of a thin film of (Zn0.95V0.05)O.

which depend on the substrate temperature and oxygen pressure used during
film deposition [26, 137]; the carriers are by-products of these defects. Ney
et al. [17] have demonstrated that highly perfect Co-doped epitaxial films,
for example, are purely paramagnetic, with cobalt occupying zinc sites, and
no sign of anything other than weak antiferromagnetic Co–O–Co nearest-
neighbor superexchange. Droubay et al. [167] show that Mn-doped epitaxial
films are paramagnetic, with Mn substituting for Zn, and no moments on
either oxygen or zinc. Barla et al. [168] showed that the cobalt in ferromag-
netic Co-doped ZnO films was paramagnetic, and there was no moment on
the zinc. A study by Tietze et al. [169] who investigated 5% Co-doped films
produced by pulsed laser deposition (PLD), which exhibited a moment of ≈1
µB/Co, corresponding to a magnetization of ≈20 kA m−1 came to a remark-
able conclusion: extensive x-ray magnetic circular dichroism (XMCD)
measurements on the Co and O edges with a very good signal/noise ratio
showed only the signatures of paramagnetic cobalt and oxygen. There is no
sign of element-specific ferromagnetism. By a process of elimination, only
oxygen vacancies remained as the possible intrinsic source of the ferromag-
netism observed in the ZnO films. The influence of oxygen vacancies can be
enhanced by doping ZnO with Co in mixed valence states [170], in line with
the CTF model discussed below.
Thin films and tunnel junctions made of Cr-doped material, which is a
magnetic insulator, show that defects near the surface rather than through-
out the bulk are responsible for the magnetism [171]. The studies of many
systems as a function of thickness, which indicate magnetic moments of
100–200 µB nm−2 argue in the same sense [24, 164]. The idea then is that 3d
ions may enhance the magnetism of ZnO, but that the magnetic moments
are not the direct cause of it. Indeed, as in TiO2 [172], the 3d dopants may
not even be magnetically coupled to the ferromagnetic phase. [169, 173]. An
optimized doping strategy uses one dopant to augment the concentration of
defects, and another, which is in a mixed-valence state, to promote charge
transfer to a spin-split impurity band, has produced magnetizaion as high
as 20 kA/m [170]. As in the other oxides, there are reports of magnetism in
undoped films [174–176], where the magnetization is greatest in the thinnest
ZnO films, and is sensitive to strain [177]. The magnetism is very sensitive
to the choice of substrate and processing conditions [177–179]. Interfaces
are important, and the magnetism of ZnO may be modified in thin-film
heterostructures [180, 181] and capped ZnO nanoparticles [182, 183], where
the moment in organic-capped ZnO detected by Zn K-edge XMCD is found
to reside within 0.8 nm of the interface [180]. Undoped nanocrystals [184,
185], nanorods [173], and other nanostructures [186] are also reported to be
ferromagnetic, but with a very small magnetization ~1–100 A m−1. A compi-
lation of data from several groups by Straumal et al. [187] indicates that weak
ferromagnetism is only found in undoped bulk ceramic material with a grain
size below 60 nm, and in Mn-doped ceramic samples with a grain size of less
than about 1.5 µm (Figure 10.13). The grain boundaries are interstitial amor-
phous layers surrounding 20 nm ZnO grains or 1000 nm Mn-doped ZnO
grains [188] analysis of the nanocrystalline ceramic samples of undoped and
10.5 Discussion 437
Equivalent size of equiaxial grains, m Equivalent size of equiaxial grains, m

10–2 10–4 10–6 10–8 10–2 10–4 10–6 10–8
2000 Single crystals and s.c. films
1400 Pure ZnO Mn-doped ZnO Equiaxial grains, dense
1800 Equiaxial grains, porous
Elongated and flattened
1200 1600 grains, dense
Elongated grains, porous
1000 1400
Temperature (°C)
Temperature (°C)
1200
800
1000
600 800
400 600
400
200
200
0 0
102 104 106 108 102 104 106 108
(a) Grain boundary area to volume ratio SGB (m2/m3) (b) Grain boundary area to volume ratio SGB (m2/m3)
FIGURE 10.13 Plot of magnetic moment versus grain-boundary area for undoped (a) and Mn-doped
(b) ZnO ceramics. (After Straumal, B.B. et al., Phys. Rev. B 79, 205206, 2009. With permission.)
Co- or Mn-doped ZnO by Straumal et al. [187] suggests that the magnetism
is associated with the grain boundaries, and they envisage the ferromagne-
tism as a grain-boundary foam, which only occupies a few percent of the
volume of the samples. This picture, supported by µsr measurements [189]
captures the ideas that defects are somehow responsible, but the magnetism
requires an extended structure of interfaces to appear.
Although there is as yet no consensus, the current view of transition-
metal-doped ZnO is moving far away from the initial expectations that the
materials are DMSs.
10.5 DISCUSSION
The experimental picture is confused by the difficulty in reproducing many
of the results, the often ephemeral nature of the magnetism, the problem of
characterizing defects and interface states in thin films and nanoparticles, as
well as the difficulty in detecting small amounts (≈1 wt%) of secondary phases
in the films [26]. Despite the tendency for experiments where nothing unusual
is found to pass unreported in the literature, there does seem to be a prima
facie case that there is something to explain in the numerous reports of fickle
ferromagnetism in dilute magnetic oxides, and d0 systems which cannot be
attributed to sloppy experimentation. If so, it seems better to seek a general
explanation of the phenomenon, rather than to rely on a different ad hoc expla-
nation for each material. For this reason, we will not discuss the numerous
electronic structure calculations for specific defects in the six oxides, which
we have been considering, but focus instead on generic physical models.
It is now quite clear that the anomalous ferromagnetism is not found in
highly perfect films or well-crystallized bulk material. Those dilute magnetic
oxides behave paramagnetically down to liquid helium temperatures, in the
way that was described in Section 10.1, albeit with a nonstatistical distribution
of dopants [167]. There is accumulating evidence, reviewed in Section 10.4,
that the phenomenon is closely associated with defects of some sort, which
are present in thin films and nanocrystals, but are absent in perfectly crys-
talline material [190]. The poor reproducibility of the experimental data may
be due to the difficulty in precisely recapturing the process conditions that
lead to a specific defect distribution and density. The samples as prepared are
not in equilibrium, and they evolve with time and temperature.
At first, it was taken for granted that the magnetism was spatially homo-
geneous, and due to the spin moments of the 3d ions. Both these assumptions
increasingly suspect. The difficulty with any spatially homogeneous explana-
tion based on localized 3d moments is that the magnetic interactions have to
be ferromagnetic, long-range, and extraordinarily strong—at least one or two
orders of magnitude greater than anything previously encountered in oxides
(Figure 10.3). The view [191] that dilute magnetic oxides were DMSs, similar to
the canonical example of (Ga1−xMnx)As, discussed in Chapter 9, Volume 2, was
influential. There the manganese introduces occupied 3d states deep below the
top of the As 4p valence band, and the ferromagnetism is mediated by 4p holes
that are polarized by exchange with the 3d5 Mn2+ ion cores. Unfortunately, the
model does not work for most 3d dopants in oxides, because the 3d level lies
above the top of the oxygen 2p valence band. Exchange via the 4s electrons in
the conduction band introduced by donor doping, as in Gd-doped EuO, for
example [192], provides weaker coupling than exchange via the 2d holes.
A third possibility illustrated in Figure 10.14 is to couple the 3d ion cores
via electrons in a spin-polarized impurity band [193]. The impurity band in
a semiconductor is made up of large Bohr-like orbitals associated with the
dopant electrons. In dilute magnetic oxides, it could be derived from defect
states such as F-centers: the idea is that the impurity band becomes spin-
split, either spontaneously if it is narrow enough to satisfy the Stoner crite-
rion, or else on account of s–d interaction with the magnetic dopant, when
it is formed of overlapping magnetic polarons (Figure 10.5b). Calculations
of TC in the molecular field approximation [193, 194] lead to the expression
1/ 2
 S(S + 1)s 2 xδ  J sd ω c
TC =   (10.6)
 3  kB
where:
S and s are the cation core spin and the donor spin, respectively
δ is the donor or acceptor defect concentration
ωc ≈ 8% is the cation volume fraction in the oxide
cb cb cb
EF
EF
ib
EF
vb vb
vb
FIGURE 10.14 Schematic models for DMSs left; spin-split conduction band cen-
ter; spin-split valence band and right; spin-split impurity band.
10.5 Discussion 439
The difficulty is that, knowing the values of the parameters in the

model, especially Jsd ≈ 1 eV, the predicted Curie temperatures are of order
10 K, which is one or two orders of magnitude too low. The model can be
modified by introducing hybridization and charge transfer from the donor
orbitals to those of the dopants, but at some point in the dilute limit, it has to
fail, and a different approach is needed. Small concentrations of conduction
electrons do provide ferromagnetic coupling via the RKKY interaction, but
an estimate of the magnitude of the magnetic-ordering temperature due to
this interaction is [193]
Z0n5/ 3m * x1/ 3δJ sf2 S(S + 1)

TC ≈ (10.7)
(96π2nc2/ 3 k B )
The Curie temperature at high carrier concentrations nc ~ 1021 cm−3 does not
exceed a few tens of Kelvin.
One way of increasing the exchange energy density is to look for an inho-
mogeneous distribution of magnetic ions. At one extreme, they may form a
segregated secondary phase, which should have a high TC if it is to explain
the magnetism. Phase segregation is expected when the solubility limit is
exceeded, but the tolerance for dissolved ionic species may be greater in thin
films or nanoparticles, which are not in thermodynamic equilibrium, than
it is in the bulk. Magnetically ordered transition-metal oxides such as CoO
or Fe3O4 have predominantly antiferromagnetic superexchange coupling,
and are usually antiferromagnetic or ferrimagnetic. However, it is possible
to find the reduced metal as a secondary phase, especially after deposition
or thermal treatment in vacuum at high temperature. Fe, Co, and Ni are
ferromagnetic metals, and the presence of segregated ferromagnetic metal
or ferromagnetic oxide inclusions is undoubtedly the explanation of the fer-
romagnetism (with telltale hysteresis) of some of the dilute magnetic oxide
samples. It is a pity that so much effort has been invested in studying these
dopants, and especially cobalt, which is the ferromagnet with the highest
Curie temperature of all. It is easier to make a case for an intrinsically new
magnetic phenomenon when ferromagnetism is observed in oxides that are
undoped, or doped with elements such as Sc, Ti, V, Cr, or Cu, which form no
phases with high-temperature magnetic order.
The near-anhysteretic magnetization curves of dilute ferromagnetic
oxides are practically temperature-independent, and emphatically do not
imply an explanation in terms of superparamagnetism, where anhysteretic
magnetization curves taken at different temperatures must superpose when
plotted as a function of H/T. In the dilute ferromagnetic oxides, there is almost
no difference between the curves measured at 4 and 300 K (Figure 10.9).
It may be that in some cases the macrospin moments of cobalt or iron
nanoinclusions are coupled by electrons in the host oxide. Such high-
temperature-ferromagnetic conducting nanocomposites could possibly find
applications in electronic devices.
Even when substituting for cations in the host oxide, the magnetic
dopant ions may experience a tendency to cluster. Spontaneous chemical
segregation in a solid solution is known as spinodal decomposition, and it

often takes the form of a composition density wave with alternating regions
rich and poor in one or other element. Spinodal decomposition has been
suggested as a mechanism for high-temperature ferromagnetism in dilute
oxides [195]. It has the virtue of segregating regions of the sample, which
still looks like a solid solution, with a higher magnetic energy density than
the average. Since these regions percolate as sheets throughout the mate-
rial, there is no expectation of independent superparamagnetic clusters. The
problem is to explain why the regions rich in transition-metal ions should be
ferromagnetic, granted the propensity for 3d ions in oxides to couple antifer-
romagnetically by superexchange. There may be an unbalanced moment in
a lamella related to certain antiferromagnetic arrangements of the spins, but
this could not be a general explanation, especially for the 3d ions from the
beginning of the series, which have a small spin moment, and do not form
any oxide with a high magnetic-ordering temperature.
The accumulating evidence that defects play a central role in the mag-
netism of dilute magnetic oxides does not mean that the 3d cations are
not involved, but it suggests that they do not contribute in the way at first
assumed. Beyond a threshold thickness, the moments do not scale with
film thickness, but seem to depend on surface area, with values of 100–
400 µB nm−2. Unphysically large moments have been inferred “per cation” at
low doping levels in almost all the oxide systems. Then there are the reports
of d0 magnetism in some samples of undoped films and nanoparticles, and
at some oxide crystal surfaces [35], and interfaces. Although they are par-
ticularly important for getting to the root of d0 magnetism, these reports are
rather uncommon. It is much more usual to find that the x = 0 end-member
is nonmagnetic, and this then serves as a credibility check of the experimen-
tal protocol, establishing the 3d dopant’s role in initiating the magnetism,
and incidentally confirming that the laboratory is not irredeemably contam-
inated with airborne magnetic dust.
odels
10.5.1 M
Having discussed the messages that the experimental data seem to convey,
we now summarize the generic models that have been proposed for ferro-
magnetism in dilute magnetic oxides. There are six of them. Five depend on
whether the spin magnetic moments are localized or delocalized, and the
nature of the electrons involved in the exchange. The sixth does not involve
spin ferromagnetism, but collective orbital paramagnetism.
1. The DMS model [191]. There are well-defined local moments on the
dopant cations, which are coupled ferromagnetically via 2p holes, or
by RKKY interactions with 4s electrons.
2. The bound magnetic polaron (BMP) model [193]. Here again, there
are well-defined local moments on the dopants, but they interact
with electrons associated with defects, which form an impurity band.
Each defect electron occupies a large orbit and interacts with several
10.5 Discussion 441
dopant cations to form a magnetic polaron, and ferromagnetism sets

in at the percolation threshold for these large objects.
3. A variant of the model (BMP), [131] dispenses with the magnetic dop-
ants, but retains the idea of localized moments, which are now asso-
ciated with electrons trapped at defects. Triplet pairs could form to
give S = 1 moments, which are then coupled by electrons in an impu-
rity or conduction band.
4. The spin-split impurity band (SIB) model [196], which is based on a
local density of states associated with defects where the density of
states at the Fermi level is sufficient to satisfy the Stoner criterion.
5. The charge-transfer ferromagnetism (CTF) model [172, 197] is a
related model with a defect-based impurity band, but there is another
charge reservoir in the system that allows for the facile transfer of
electrons to or from the impurity band to create a filling that leads to
spontaneous spin splitting. In dilute magnetic oxides, this reservoir
is associated with the dopant ions.
6. Lastly, there is the possibility that the effects are not essentially
related to collective spin magnetism at all, but that the saturating
magnetic signal is due to giant orbital paramagnetism (GOP) associ-
ated with a new collective orbital state of the electrons. The magnetic
moments are then induced rather than aligned by the field.
The first three of the models are Heisenberg models, with localized spin
moments that fit the m-J paradigm. The first two of them require roughly
uniform ferromagnetism of the sample, since the moments are borne by the
3d dopants that are distributed more or less uniformly throughout the mate-
rial. We have seen little proof that the dopants in the dilute magnetic oxides
are magnetically ordered, and growing evidence that they are actually para-
magnetic. The models also require long-range ferromagnetic exchange inter-
actions of a strength not found in any concentrated magnetically ordered
material. Curie temperatures of order 10–50 K might be possible for materi-
als with a few percent doping, but not values of order 800 K. Neither model
accounts for d0 ferromagnetism in undoped, defective oxides. The third
model associates the local moments with defects. They can occur in oxides
with no 3d ions [80]. The defects could be distributed near interfaces or grain
boundaries, and so the model can account for both inhomogeneous and d0
ferromagnetism. The drawback is that, as with any Heisenberg model, spin-
wave excitations are inevitable, and there is no reason to expect high Curie
temperatures.
The next two models are Stoner models. Localized magnetic moments
are not involved; all that is needed is a defect-related density of states with
the right structure and occupancy. The magnetism is not homogeneous,
but is confined to percolating regions such as surfaces, interfaces, or grain
boundaries. Furthermore, Edwards and Katsnelson [196], who developed
the SIB model for the d0 ferromagnet CaB6, show that spin waves are sup-
pressed for the defect-related impurity band, so Curie temperatures may be
very high. The main problem with the model is its lack of flexibility. There is
no specific role for the 3d dopants. The number of electrons in the impurity
band is determined by the defect structure of the sample, and it would be
something of a fluke for the electron occupancy to coincide with a peak in
the density of states where the Stoner criterion is satisfied:
I N (E F ) > 1 (10.8)
where I is the Stoner exchange integral, which is of order 1 eV. The model
can have ferromagnetic metallic, half-metallic, and insulating ground states.
The difficulty arising from a fixed number of electrons in the impurity
band is resolved in the CTF model. In this model, there is a proximate
electron reservoir, which can feed or drain the impurity band as required
to meet the Stoner criterion. This makes it possible for a wider range of
materials to become magnetically ordered. In the case of dilute magnetic
oxides, the 3d dopants can serve as the reservoir, provided they are able to
coexist in different valence states such as Ti3+/Ti4+, Mn3+/Mn4+, Fe2+/Fe3+,
Co2+/Co3+, or Cu+/Cu 2+. The point of doping with 3d cations is not that
they carry a magnetic moment, but that their mixed valence allows them
to act as the electron reservoir. The cations can remain paramagnetic. It
is possible that they tend to migrate to the surfaces or grain boundaries,
where the defects lie. The defect-based models can explain the inhomoge-
neous nature of magnetism. The magnetic regions are those that are rich
in defects, which tend to concentrate at interfaces and grain boundaries
(Figure 10.15).
It is a challenge to account for the observations that the dilute mag-
netic oxides may be insulating, semiconducting, or metallic. The Heisenberg
models account for a nonconducting magnetic state, by supposing that the
carriers that propagate the exchange are immobile; the Fermi energy in the
conduction band lies below the mobility edge [155]. However, the Stoner
models have the option of a half-filled SIB, which can arise from single-
electron defect states. The CTF model has a rich electronic phase diagram,
with insulating, half-metallic, and metallic ferromagnetic states, as well
as nonmagnetic states (Figure 10.16). The undoped end-member is most
often nonmagnetic, which shows that the doping is critically important.
Ferromagnetic d0 end-member oxide films remain very much an exception,
rather than the rule.
(a) (b) (c) (d)
FIGURE 10.15 Inhomogeneous ferromagnetism in a dilute magnetic oxide. The

ferromagnetic regions are distributed (a) at random, (b) in spinodally segregated
regions, (c) at the surface/interface of a film, and (d) at grain boundaries.
10.5 Discussion 443
Mn+ M(n+1)+ 0.4
U/w
0.2
U/w
0.0
0.0 1.0 2.0
Ntot/W
Nonmagnetic semiconductor Ferromagnetic half metal
Nonmagnetic metal Ferromagnetic metal
FIGURE 10.16 Phase diagram for the CTF model. Electron transfer from the 3d charge reservoir into the
defect-based impurity band (a “double-peak” structure is chosen) can lead to spin splitting, as shown on the
left. Axes are the total number of electrons in the system Ntot and the 3d coulomb energy Ud, each normalized
by the impurity bandwidth w. The Stoner integral I is taken to be 0.6. In the unshaded nonmagnetic semicon-
ductor regions, the impurity band is empty or full. The curved gray line marks half-filling. In the gray regions,
it is partly filled, but not spin-split. In the blue regions, the Stoner criterion is satisfied, and it is ferromagnetic.
(After Coey, J.M.D. et al., New J. Phys. 12, 053025, 2010. With permission.)
Finally, there are two features of the ferromagnetism, which have received
less attention than they deserve. One is the magnetization process—the fact
that the ferromagnetism exhibits very little hysteresis, and the magnetiza-
tion curve is practically independent of temperature. Magnetization curves
like that in Figure 10.9 are found in all the systems we have been discussing.
It may be asked whether a material with a magnetization curve showing no
hysteresis can really be ferromagnetic; where is the broken symmetry? But
the magnetization curves of iron, permalloy, and other soft ferromagnets
often fail to show hysteresis, at least at room temperature, because they form
domains or other flux-closed structures to reduce their dipolar self-energy.
Hysteresis is normally related to magnetocrystalline anisotropy, which
falls off rapidly with increasing temperature. The virtual lack of hysteresis
below room temperature indicates that dipolar interactions are controlling
the approach to saturation. A uniformly magnetized soft ferromagnetic film
has a demagnetizing factor of zero when the magnetization lies in-plane and
unity when the magnetization is perpendicular to the plane; the magneti-
zation curve saturates in a small field in the first case and it is linear up to
saturation at H = M0 in the second. The measured magnetization curves are
usually not like this; they saturate in almost the same way regardless of field
direction, with a magnetization curve that can be approximated by a tanh
function
 H 
M = M0tanh  (10.9)
 H d 
where Hd ≈ 70–130 kA m−1. Isolated micron and submicron iron particles

(M0 = 1720 kA m−1) behave similarly, but with Hd ≈ 300 kA m−1. The iron par-
ticles adopt a vortex magnetic configuration to minimize the demagnetiz-
ing energy, which progressively unwinds as the magnetization approaches
saturation.
A spin-split defect-based s-like impurity band would not be expected
to exhibit magnetocrystalline anisotropy. Assuming the ferromagnetic
regions are thin layers following the grain boundaries, the ferromagnetic
grain-boundary foam will tend to have a wandering ferromagnetic axis that
lies in the plane of the grain boundary. On applying an increasing magnetic
field, the wandering axis will unwind, and straighten up as saturation is
approached. Modeling the magnetization process of an ensemble of macro-
spins subjected to a randomly oriented demagnetizing field gives a magne-
tization curve with Hd ≈ 0.16M0 [198], where M0 is the magnetization of the
ferromagnetic regions. From the plot in Figure 10.17, based on the magneti-
zation curves of 250 dilute ferromagnetic oxide films, and the summary in
Table 10.4, it can be deduced that M0/Ms ≈ 100, where Ms is the saturation
magnetization of the sample. We infer that the ferromagnetic volume frac-
tion in the dilute magnetic oxides is about 1%. When data on the nanocrys-
talline ceramics is analyzed in the same way, it turns out that the volume
fraction there is around one part in 105 or 106.
The other point to flag is the anisotropy of the saturation magnetization
of oriented films, either between parallel and perpendicular to the plane or
within the plane. The effect has been reported for HfO2 [44, 102, 144], TiO2
[66], SnO2 [84, 86, 88, 89], and ZnO [144, 164] (Figure 10.12); the anisotropy
may be as large as a factor 10. Reports of this anisotropy of d-zero magnetism
are intriguing, but rather uncommon. They merit further investigation.
The last model we consider, the GOP model, has an entirely different
basis, namely that the temperature-independent “ferromagnetic” response
1000
100
10
ZnO
Ms (kA m–1)
SnO2
1
Iron
TiO2
0.1 Graphite
HO2
0.01
0.001
0.0001
10 100 1000
Hd (kA m–1)
FIGURE 10.17 Plot of magnetization Ms versus internal field Hd for thin films and
nanoparticles of doped and undoped oxides. For films of the thin films of oxides,
Ms clusters around 10 kA m−1, whereas for nanoparticles, it is three orders of magni-
tude lower. Hd in both cases is about 100 kA m−1. Also included on the scatter plot
are data on graphite and dispersed iron nanoparticles.
10.5 Discussion 445
TABLE 10.4
Local Dipole Field Hd
Hd (kA m−1)
TiO2 125 (40)
SnO2 79 (30)
HfO2 94 (35)
ZnO 83 (30)
Graphite 68 (42)
Fe 275 (40)
to an applied field is not fundamentally related to any sort of collective,

exchange-driven ferromagnetic spin ordering, but is entirely field-induced and
originates from coherent mesoscopic conduction currents. In other words,
the effect is a novel type of temperature-independent, saturating orbital para-
magnetism. The possibility of surface orbital moments has been evoked in the
context of Au nanoparticles [199, 200] and nanorods, and also of zinc oxide
nanoparticles [201, 202]. A theory has recently been developed that envisages
the possibility of coherent electronic states in quasi-two-dimensional systems
with a large surface to volume ratio, due to coupling with zero-point fluctua-
tions of the vacuum electromagnetic field [203]; the magnetic consequences
of the theory have been developed in the context of CeO2 nanoparticles [121].
The theoretical form for the magnetization curve is
M = Ms x / (1 + x 2 )1/2 (10.10)
where x = C B. electromagnetic field. The form of the saturation resembles

Equation 10.9, but fitting experimental data gives the values of C, which
yields the characteristic wavelength
1/ 4
l = éë(C / Ms )( 6c f c ) ùû (10.11)
where fc is the volume fraction of the sample that is magnetically coherent.

In the case of CeO2 nanoparticles, the analysis gives λ = 300 nm as the
coherence length, which corresponds to a maximum in the UV absorption
for magnetic nanoparticles at 4.1 eV. The remarkable disappearance of the
magnetism when the nanoparticles are segregated by dilution (Figure 10.8c)
is interpreted in terms of a minimum lengthscale of order needed to establish
the coherent electronic state. The model may be applicable to a range of granu-
lar nanoscale oxides with conducting interfaces, and to oxide surfaces such as
that of SrTiO3 [35] or nanoporous amorphous alumina membranes [207]. We
emphasize the essentially anhysteretic nature of the temperature-independent
magnetic signal. By including the electron spin, and Rashba spin-orbit coupling
of spin moments to the field-induced orbital moments in the GOP model, it may
be possible to account for much of the behavior of nanocrystalline and nano-
granular d0 oxides, including the weak hysteresis that is occasionally observed.
10.6 CONCLUSIONS
The magnetism of dilute and undoped magnetic oxides remains one of the
most puzzling and potentially important open questions in magnetism
today. There is now sufficient evidence that the observations in many cases
are not simply attributable to measurement artifacts or trivial impurities.
There is definitely something to explain.
Two possible sources of the magnetism are defects and dopants. Both
may be involved, but in the d0 systems and perhaps in most of the dilute
magnetic oxide systems that have been investigated, the defects suffice. The
nature of the magnetic order and the coupling mechanism have to be bet-
ter understood, but it seems that the m-J paradigm of magnetism in solids
is unable to encompass these materials. On present evidence, the coupling
between the 3d dopants is weak. They are not DMSs.
A new picture of high-temperature defect-related magnetism in oxides
is emerging, where there are two contending explanations, based on quite
different physical premises [208]. One model is Stoner spin ferromagnetism
involving a spin-split defect-based impurity band. Charge transfer from the
3d cations can help stabilize the spin splitting. There is much to investigate
in the CTF model, and to connect with specific materials systems – spin
waves, micromagnetism, orbital moments, anisotropy, magnetotransport,
and other spin-orbit coupling effects. The other model is giant orbital para-
magnetism, where there is no symmetry-breaking magnetic order, but an
unusual reversible orbital paramagnetic response of electrons that exhibit
coherence on a mesoscopic scale as a result of a collective Lamb shift [203].
Here too there is much to investigate, especially in thin film structures arti-
ficially patterned on a mesoscopic scale.
The lack of reproducibility of magnetic properties between and within lab-
oratories, and the lack of appropriate physical characterization tools that oper-
ate in ambient conditions for the aperiodic defects, such as oxygen vacancies
and grain boundaries that appear to be important in many cases may be less
of an obstacle than first thought. These hurdles can be circumvented to a great
extent by experimental methods that create or destroy the d-zero magnetism
in a particular sample. Specific chemical surface treatments with catechols are
useful for strontium titanate surfaces [209] and nanoporous amorphous alu-
mina membranes [207], CeO2 nanoparticles come closest to a dilute magnetic
‘fruitfly’ system that is reproducible and easy to prepare [206].
The Co-doped TiO2 or ZnO films produced by pulsed laser deposition
that launched the field, have proved misleading. We need to tie down the
correlation between optical, especially UV, absorption and magnetism, and
better understand the significance of a critical size for the appearance of
magnetism.
Device applications can follow by design or by serendipity. Once the
basic spin or orbital character of the magnetism is firmly established, the
challenge will be to generate stable and controllable defect structures where
we can reap the benefit of this unusual high-temperature magnetism, whose
occurrence is by no means limited to dilute magnetic oxides.
References 447
ACKNOWLEDGMENT
This work was supported by Science Foundation Ireland, grant 16/IA/4534.
REFERENCES
1. R. G. Burns, Mineralogical Applications of Crystal Field Theory, 2nd edn.,
Cambridge University Press, Cambridge, UK, 1993.
2. D. J. Newman and B. Ng, Crystal Field Handbook, Cambridge University Press,
Cambridge, UK, p. 290, 2000.
3. E. J. Samuelsen and G. Shirane, Inelastic neutron scattering investigation of
spin waves and magnetic interactions in alpha-Fe2O3, Phys. Status Solidi 42,
241–256 (1970).
4. E. J. Samuelsen, M. T. Hutchings, and G. Shirane, Inelastic neutron scattering
investigation of spin waves and magnetic interactions in Cr2O3, Physica 48,
13–42 (1970).
5. H. Bourdonnay, A. Bousquet, J. P. Cotton et al., Experimental determination of
exchange integrals in magnetite, J. Phys. 32(2–3), C1–C1182 (1970).
6. J. Goodenough, Theory of the role of covalence in the perovskite-type manga-
nites [La, M(II)]MnO3, Phys. Rev. 100, 564–573 (1955).
7. J. Kanamori, Superexchange interaction and symmetry properties of electron
orbitals, J. Phys. Chem. Solids 10, 87–98 (1959).
8. P. W. Anderson, Theory of magnetic exchange interactions: Exchange in insu-
lators and semiconductors, Solid State Phys. 14, 99–214 (1963).
9. J. M. D. Coey, M. Viret, and S. von Molnar, Mixed-valence manganites, Adv.
Phys. 48, 167–293 (1999).
10. C. Zener, Interaction between the d-shells in the transition metals II:
Ferromagnetic compounds of manganese with the perovskite structure, Phys.
Rev. 82, 403–405 (1951).
11. D. Stauffer, Introduction to Percolation Theory, Taylor & Francis, London, UK,
1985.
12. G. Deutscher, R. Zallen, and J. Adler (Eds), Percolation Structures and
Processes, Adam Hilger, Bristol, UK, 1983.
13. P. Sati, R. Hayn, R. Kuzian et al., Magnetic anisotropy of Co2+ as signature of
intrinsic ferromagnetism in ZnO:Co, Phys. Rev. Lett. 96, 017203 (2006).
14. W. Pacuski, D. Ferrand, J. Cibert et al., Effect of the s,p-d exchange interaction
on the excitons in Zn1−xCoxO epilayers, Phys. Rev. B 73, 035214 (2006).
15. C. N. R. Rao and F. L. Deepak, Absence of ferromagnetism in Mn- and
Co-doped ZnO, J. Mater. Chem. 15, 573–578 (2005).
16. S. A. Chambers and R. F. Farrow, New possibilities for ferromagnetic semicon-
ductors, MRS Bull. 28, 729–733 (2003).
17. A. Ney, K. Ollefs, S. Ye et al., Absence of intrinsic ferromagnetic interactions
with isolated and paired Co dopant atoms in Zn1−xCoxO with high structural
perfection, Phys. Rev. Lett. 100, 157201 (2008).
18. J. M. D. Coey, d0 ferromagnetism, Solid State Sci. 7, 660–667 (2005).
19. S. B. Ogale, Dilute doping, defects and ferromagnetism in metal oxide systems,
Adv. Mater. 22, 3125–3155 (2010).
20. T. Fukumura and M. Kawasaki, Chapter 3: Magnetic oxide semiconductors: on
the high-temperature ferromagnetism in TiO2- and ZnO-based compounds,
in Functional Metal Oxides: New Science and Novel Applications. S. B. Ogale,
T. Venkatesen and M. Blamire (eds.), Wiley-VCH Verlag, Weinheim, Germany,
2013.
21. N. H. Hong, Magnetic oxide semiconductors, in Handbook of Spintronics, Y.
Xu, D. D. Awschalom and J. Nitta (eds.), Springer, Dordrecht, the Netherlands,
pp. 563–583, 2016.
22. D. W. Abraham, M. M. Frank, and S. Guha, Absence of magnetism in hafnium

oxide films, Appl. Phys. Lett. 87, 252502 (2005).
23. Y. Belghazi, G. Schmerber, S. Colis et al., Extrinsic origin of ferromagnetism in
ZnO and Zn0.9Co0.1O magnetic semiconductor films prepared by sol-gel tech-
nique, Appl. Phys. Lett. 89, 122504 (2006).
24. J. M. D. Coey, Dilute magnetic oxides, Curr. Opin. Solid State Mater. Sci. 10,
83–92 (2007).
25. M. A. Garcia, E. Fernandez Pinel, J. de la Venta et al., Sources of experimental
errors in the observation of nanoscale magnetism, J. Appl. Phys. 105, 013925
(2009).
26. K. Potzger and S. Q. Zhou, Non-DMS related ferromagnetism in transition-
metal doped zinc oxide, Phys. Status Solidi B 246, 1147–1167 (2009).
27. M. Khalid, A. Setzer, M. Ziese et al., Ubiquity of ferromagnetic signals in com-
mon diamagnetic oxides, Phys. Rev. B 81, 214414 (2010).
28. J. M. D. Coey, Ariando, and W. E. Pickett, Magnetism at the edge; New phe-
nomena at oxide interfaces, MRS Bull. 38, 1040–1047 (2013).
29. A. Ohtomo, H. Y. Hwang, A high-mobility electron gas at the LaAlO3/SrTiO3
heterointerface, Nature 427, 423–426 (2004).
30. N. Nakagawa, H. Y. Hwang, D. A. Muller, Why some interfaces cannot be
sharp, Nat. Mater. 5, 204–209 (2006).
31. G. Herranz, M. Basletic, M. Bibes et al., High mobility in LaAlO3/SrTiO3 het-
erostructures: Origin, dimensionality, and perspectives, Phys. Rev. Lett., 98,
216803 (2007).
32. Z. Q. Liu, C. J. Li, W. M. Lü et al., Origin of the two-dimensional electron gas at
LaAlO3/SrTiO3 interfaces: The role of oxygen vacancies and electronic recon-
struction, Phys. Rev. X 3, 021010 (2013).
33. N. C. Bristowe, P. Ghosez, P. B. Littlewood, and E. Artacho, The origin of two-
dimensional electron gases at oxide interfaces; Insights from theory, J. Phys.
Condens. Matter 26, 143201 (2014).
34. X. Renshaw Wang, C. J. Li, W. M. Lu ̈ et al., Imaging and control of ferromag-
netism in LaMnO3/SrTiO3 heterostructures, Science 349, 716–719 (2015).
35. J. M. D. Coey, M. Venkatesan, and P. Stamenov, Surface Magnetism of
Strontium Titanate, J. Phys. Condens. Matter 28, 485001 (2016).
36. M. Khalid, A. Setzer, M. Ziese et al., Ubiquity of ferromagnetic signals in com-
mon diamagnetic oxide crystals, Phys. Rev. B 81, 214414 (2009).
37. N. H. Hong, J. Sakai, W. Prellier et al., Ferromagnetism in transition-metal
doped TiO2 thin films, Phys. Rev. B 70, 195204 (2004).
38. Y. Matsumoto, M. Murakami, T. Shono et al., Room-temperature ferromag-
netism in transparent transition metal-doped titanium dioxide, Science 291,
854–856 (2001).
39. Y. Matsumoto, R. Takahashi, M. Murakami et al., Ferromagnetism in
Co-doped TiO2 rutile thin films grown by laser molecular beam epitaxy, Jpn. J.
Appl. Phys. 40, L1204–L1206 (2001).
40. S. R. Shinde, S. B. Ogale, S. Das Sarma et al., Ferromagnetism in laser deposited
anatase Ti1−xCoxO2−δ films, Phys. Rev. B 67, 115211 (2003).
41. Z. J. Wang, J. K. Tang, Y. X. Chen et al., Room-temperature ferromagnetism in
manganese-doped reduced rutile titanium dioxide thin films, J. Appl. Phys. 95,
7384 (2004).
42. J. M. D. Coey, A. P. Douvalis, C. B. Fitzgerald, and M. Venkatesan, Ferromagnetism
in Fe-doped SnO2 thin films, Appl. Phys. Lett. 84, 1332–1334 (2004).
43. S. B. Ogale, R. J. Choudhary, J. P. Buban et al., High temperature ferromagne-
tism with a giant magnetic moment in transparent Co-doped SnO2−δ, Phys.
Rev. Lett. 91, 077205 (2003).
44. J. M. D. Coey, M. Venkatesan, P. Stamenov et al., Magnetism in hafnium diox-
ide, Phys. Rev. B 72, 024450 (2005).
45. H. N. Hong, N. Poirot, and J. Sakai, Evidence for magnetism due to oxygen
vacancies in Fe-doped HfO2 thin films, Appl. Phys. Lett. 89, 042503 (2006).
References 449
46. A. Tiwari, V. M. Bhosle, R. Ramachandran et al., Ferromagnetism in Co-doped

CeO2: Observation of a giant magnetic moment with a high Curie tempera-
ture, Appl. Phys. Lett. 88, 142511 (2006).
47. J. Philip, N. Theodoropoulou, G. Berera et al., High-temperature ferromag-
netism in manganese-doped indium–tin oxide films, Appl. Phys. Lett. 85(5),
777–779 (2004).
48. J. He, S. F. Xu, Y. K. Yoo et al., Room-temperature ferromagnetic n-type semi-
conductor (In1−xFex)2O3−σ, Appl. Phys. Lett. 87, 052503 (2005).
49. J. Philip, A. Punnoose, B. I. Kim et al., Carrier-controlled ferromagnetism in
transparent oxide semiconductors, Nat. Mater. 5, 298–304 (2006).
50. H. Saeki, H. Tabata, and T. Kawai, Magnetic and -electric properties of vana-
dium doped ZnO films, Solid State Commun. 120, 439–442 (2001).
51. P. Sharma, A. Gupta, K. V. Rao et al., Ferromagnetism above room temperature
in bulk and transparent thin films of Mn-doped ZnO, Nat. Mater. 2, 673–677
(2003).
52. S. J. Han, J. W. Song, C. H. Yang et al., A key to room-temperature ferromagne-
tism in Fe-doped ZnO: Cu, Appl. Phys. Lett. 81, 4212–4214 (2002).
53. K. Ueda, H. Tabata, and T. Kawai, Magnetic and electric properties of transi-
tion-metal-doped ZnO films, Appl. Phys. Lett. 79, 988–990 (2001).
54. P. V. Radovanovic and D. R. Gamelin, High-temperature ferromagnetism in
Ni2+-doped ZnO aggregates prepared from colloidal dilute magnetic semicon-
ductor quantum dots, Phys. Rev. Lett. 91, 157302 (2003).
55. J. Y. Kim, J. H. Park, B. G. Park et al., Ferromagnetism induced by clustered Co
in Co-doped anatase TiO2 thin films, Phys. Rev. Lett. 90, 017401 (2003).
56. P. A. Stampe, R. J. Kennedy, Y. Xin, and J. S. Parker, Investigation of the cobalt
distribution in the room temperature ferromagnet TiO2:Co, J. Appl. Phys. 93,
7864–7866 (2003).
57. T. Zhao, S. R. Shinde, S. B. Ogale et al., Electric field effect in diluted magnetic
insulator anatase Co: TiO2, Phys. Rev. Lett. 94, 126601 (2005).
58. H. Toyosaki, T. Fukumura, Y. Yamada, and M. Kawasaki, Evolution of
f
erromagnetic circular dichroism coincident with magnetization and
anomalous Hall effect in Co-doped rutile TiO2 , Appl. Phys. Lett. 86, 182503
(2005).
59. H. Toyosaki, T. Fukumura, Y. Yamada et al., Anomalous Hall effect governed
by electron doping in a room-temperature transparent ferromagnetic semi-
conductor, Nat. Mater. 3, 221–224 (2004).
60. H. Toyosaki, T. Fukumura, K. Ueno et al., A ferromagnetic oxide semiconduc-
tor as spin injection electrode in a magnetic tunnel junction, Jpn. J. Appl. Phys.
44, L896–L898 (2005).
61. Z. Wang, W. Wang, J. Tang et al., Extraordinary Hall effect and ferromagne-
tism in Fe-doped reduced rutile, Appl. Phys. Lett. 83, 518–520 (2003).
62. S. Duhalde, M. F. Vignolo, F. Golmar et al., Appearance of room-temper-
ature ferromagnetism in Cu-doped TiO2−δ films, Phys. Rev. B 72, 161313
(2005).
63. D. L. Hou, R. B. Zhao, H. J. Meng et al., Room-temperature ferromagnetism in
Cu-doped TiO2 thin films, Thin Solid Films 516, 3223–3226 (2008).
64. Y. J. Kim, S. Thevuthasan, T. Droubay et al., Growth and properties of molec-
ular beam epitaxially grown ferromagnetic Fe-doped TiO2 rutile films on
TiO2(110), Appl. Phys. Lett. 84, 3531–3533 (2004).
65. T. C. Kaspar, S. M. Heald, C. M. Wang et al., Negligible magnetism in excellent
structural quality CrxTi1−xO2 anatase: Contrast with high-TC ferromagnetism
in structurally defective CrxTi1−xO2, Phys. Rev. Lett. 95, 217203 (2005).
66. J. Osterwalder, T. Droubay, T. Kaspar et al., Growth of Cr-doped TiO2 films in
the rutile and anatase structures by oxygen plasma assisted molecular beam
epitaxy, Thin Solid Films 484, 289–298 (2005).
67. S. A. Chambers, Ferromagnetism in doped thin film oxide and nitride semi-
conductors and dielectrics, Surf. Sci. Rep. 61, 345–381 (2006).
68. K. A. Griffin, A. B. Pakhomov, C. M. Wang et al., Intrinsic ferromagnetism in

insulating cobalt doped anatase TiO2, Phys. Rev. Lett. 94, 157204 (2005).
69. K. A. Griffin, A. B. Pakhomov, C. M. Wang et al., Cobalt-doped anatase TiO2:
A room-temperature dilute magnetic dielectric material, J. Appl. Phys. 97,
10D320 (2006).
70. K. A. Griffin, M. Varela, S. J. Pennycook et al., Atomic-scale studies of cobalt
distribution in Co-TiO2 anatase TiO2 thin films: Processing, microstructure
and the origin of ferromagnetism, J. Appl. Phys. 97, 10D320 (2006).
71. T. Fukumura, H. Toyosaki, K. Ueno et al., Role of charge -carriers for ferro-
magnetism in Co-doped rutile TiO2, New J. Phys. 10, 055018 (2008).
72. A. F. Orlov, L. A. Balagurov, A. S. Konstantinova et al., Giant magnetic
moments in dilute magnetic semiconductors, J. Magn. Magn. Mater. 320,
895–897 (2008).
73. S. D. Yoon, Y. Chen, A. Yang et al., Oxygen-defect-induced magnetism at 880
K in semiconducting anatase TiO2−δ films, J. Phys. Condens. Matter 18, L355–
L361 (2006).
74. N. H. Hong, J. Sakai, N. Proirot et al., Room-temperature ferromagnetism
observed in undoped semiconducting and insulating oxide thin films, Phys.
Rev. B 73, 132404 (2006).
75. Z. Q. Zhou, C. Cizmar, K. Potzger et al., Origin of magnetic moments in defec-
tive TiO2 single crystals, Phys. Rev. B 79, 113201 (2009).
76. B. Santora, P. K. Giri, K. Imakita and M. Fujii, Evidence of oxygen-vacancy-
induced room-temperature ferromagnetism in solvothermally synthesised
indoped TiO2 nanoribbons, Nanoscale 5, 5476–5480 (2013).
77. N. N. Bao, H. M. Fan, J. Ding, and J. B. Yi, Toom-temperature ferromagnetism
in N-doped rutile TiO2 films, J. Appl. Phys. 109, 07C302 (2011).
78. A. Rusydi, S. Dhar, A. R. Barman et al., Cation-vacancy-induced room-tem-
perature ferromagnetism in transparent, conducting anatase Ti1-xTa xO2 thin
films, Phil. Trans. Roy. Soc. A 370, 4926–4943 (2012).
79. S. Wang, L. Pan, J. J. Song et al., Titanium-defected undoped anatase TiO2
with p-type conductivity, room-temperature ferromagnetism and remarkable
photocatalytic properties, J. Am. Chem. Soc. 137, 2975–2983 (2015).
80. I. S. Elfimov, S. Yunoki, and G. A. Sawatzky, Possible path to a new class of
ferromagnetic and half-metallic ferromagnetic materials, Phys. Rev. Lett. 89,
216403 (2002).
81. N. H. Hong and J. Sakai, Ferromagnetic V-doped SnO2 thin films, Phys. B 358,
265–268 (2005).
82. N. H. Hong, J. Sakai, W. Prellier, and A. Hassini, Transparent Cr-doped SnO2
thin films: Ferromagnetism beyond room temperature with a giant magnetic
moment, J. Phys. Condens. Matter 17, 1697–1702 (2005).
83. K. Gopinadhan, S. C. Kashyap, D. K. Pandya et al., High temperature ferro-
magnetism in Mn-doped SnO2, J. Appl. Phys. 102, 113513 (2007).
84. C. B. Fitzgerald, M. Venkatesan, L. S. Dorneles et al., Magnetism in dilute mag-
netic oxide thin films based on SnO2, Phys. Rev. B 74, 11530 (2006).
85. N. H. Hong, J. Sakai, N. T. Huong et al., Role of defects in tuning ferromagne-
tism in diluted magnetic oxide thin films, Phys. Rev. B 72, 045336 (2005).
86. J. Zhang, R. Skomski, L. P. Yue et al., Structure and magnetism of V-doped
SnO2 thin films: Effect of the substrate, J. Phys. Condens. Matter 19, 256204
(2007).
87. H. S. Kim, L. Bi, G. F. Dionne et al., Structure, magnetic and optical properties,
and Hall effect of Co- and Fe-doped SnO2 films, Phys. Rev. B 77, 214436 (2008).
88. N. H. Hong, N. Poirot, and J. Sakai, Ferromagnetism observed in pristine SnO2
thin films, Phys. Rev. B 77, 033205 (2008).
89. J. Zhang, R. Skomski, Y. F. Lu et al., Temperature-dependent orbital-moment
anisotropy in dilute magnetic oxides, Phys. Rev. B 75, 214417 (2007).
References 451
90. P. I. Archer, P. V. Radovanovic, S. M. Heald et al., Low-temperature activation

and deactivation of high-curie-temperature ferromagnetism in a new diluted
magnetic semiconductor: Ni2+-doped SnO2, J. Am. Chem. Soc. 127, 14479–
14487 (2005).
91. C. Van Komen, A. Thurber, K. M. Reddy et al., Structure-magnetic prop-
erty relationship in transition metal (M = V, Cr, Mn, Fe, Co, Ni) doped SnO2
nanoparticles, J. Appl. Phys. 103, 07D141 (2008).
92. A. Thurber, K. M. Reddy, V. Shutthanandan et al., Ferromagnetism in chem-
ically-substituted CeO2 nanoparticles by Ni doping, Phys. Rev. B 76, 165206
(2007).
93. C. B. Fitzgerald, M. Venkatesan, L. S. Dorneles et al., SnO2 doped with Mn, Fe
or Co; room-temperature dilute magnetic semiconductors, J. Appl. Phys. 95,
7390–7395 (2004).
94. A. F. Cabrera, A. M. N. Mudarra, C. E. T. Rodriguez et al., Mechanosynthesis
of Fe-doped SnO2 nanoparticles, Phys. B 398, 215–218 (2007).
95. P. I. Archer and D. R. Gamelin, Controlled grain-boundary defect formation
and its role in the high-Tc ferromagnetism on Ni2+: SnO2, J. Appl. Phys. 99,
08M107 (2006).
96. J. D. Bryan, S. M. Heald, S. A. Chambers, and D. R. Gamelin, Strong room-
temperature ferromagnetism in Co2+-doped TiO2 made from colloidal nano-
crystals, J. Am. Chem. Soc. 126, 11640–11647 (2004).
97. A. Sundaresan, R. Bhargavi, N. Rangarajan et al., Ferromagnetism as a univer-
sal feature of nanoparticles of otherwise nonmagnetic oxides, Phys. Rev. B 74,
161306(R) (2006).
98. S. Bhaumik, A. K. Sinha, S. K. Ray, and A. K. Das, Defect-iduced room-temper-
ature ferromagnetism in SnO2 nanowires controlled by UV light irradiation,
IEEE Trans. Magn. 50, 2400206 (2014).
99. J. Wang, W. Zhou, and P. Wu, Band gap narrowig and d0 ferromagnetism in
epitaxial Li-doped SnO2 films, Appl. Surf. Sci. 314, 188–192 (2014)
100. N. Wang, W. Zhou, and P. Wu, Ferromagnetic order in SnO2 nanoparticles
with nonmagnetic Li doping, J. Mater. Sci. 26, 4132–4137 (2015).
101. P. Wu, B. Zhou, and W. Zhou, Room-temperature ferromagnetism in epitaxial
Mg-doped SnO2 thin films, Appl. Phys. Lett. 100, 182405 (2012).
102. M. Venkatesan, C. B. Fitzgerald, and J. M. D. Coey, Unexpected magnetism in
a dielectric oxide, Nature 430, 630 (2004).
103. H. N. Hong, Magnetism due to defects/oxygen vacancies in HfO2 thin films,
Phys. Status Solidi C 4, 1270–1275 (2007).
104. N. H. Hong, J. Sakai, N. Poirot, and A. Ruyter, Laser ablated Ni-doped HfO2 thin
films: Room temperature ferromagnets, Appl. Phys. Lett. 86, 242505 (2005).
105. M. S. R. Rao, D. C. Kundaliya, S. B. Ogale et al., Search for ferromagnetism in
undoped and cobalt-doped HfO2−∂, Appl. Phys. Lett. 88, 142505 (2006).
106. N. Hadacek, A. Nosov, L. Ranno et al., Magnetic properties of HfO2 thin films,
J. Phys. Condens. Matter 19, 486206 (2007).
107. E. Tirosh and G. Markovich, Control of defects and magnetic properties in
colloidal HfO2 nanorods, Adv. Mater. 19, 2608–2612 (2007).
108. X. Liu, Y. Chen, L. Wang, and D. L. Peng, Transition from paramagnetism to
ferromagnetinm in HfO2 nanorods, J. Appl. Phys. 113, 076102 (2013).
109. Q. Xie, W. P. Mang, Z. Xie et al., Room-temperature ferromagnetism in
undoped amorphous HfO2 nanohelix arrays, Chin. Phys. B 24, 057503 (2015).
110. Q. Xie, W. Wang, Z. Xie et al., High-magnetic field annealing effect on room-
temperature ferromagnetism enhancement of undoped HfO2 thin films, App.
Phys. A 119, 917921 (2015).
111. E. Hildebrandt, J. Kurian, and L. Alff, Physical properties and band structure
of reactive molecular beam epitaxy grown oxygen engineered HfO2-x, J. Appl.
Phys. 112, 114112 (2012).
112. Y. Q. Song, H. W. Zhang, Q. Y. Wen et al., Room-temperature ferromagne-

tism in Co-doped CeO2 thin films on Si (111) substrates, Chin. Phys. Lett. 24,
218–221 (2007).
113. R. Vodungbo, Y. Zheng, F. Vidal et al., Room-temperature ferromagnetism of
Co-doped CeO2−δ diluted magnetic oxide: Effect of oxygen and anisotropy,
Appl. Phys. Lett. 90, 162510 (2007).
114. Y. L. Liu, Z. Lockmann, A. Aziz et al., Size-dependent ferromagnetism in
cerium oxide (CeO2) nanostructures independent of oxygen vacancies, J. Phys.
Condens. Matter 20, 165201 (2008).
115. M. Y. Ge, H. Wang, E. Z. Liu et al., On the origin of ferromagnetism in CeO2
nanocubes, Appl. Phys. Lett. 93, 062505 (2008).
116. M. L. Li, S. H. Ge, W. Qiao et al., Relationship between the surface chemical
states and magnetic properties of CeO2 nanoparticles, Appl. Phys. Lett. 94,
152511 (2009).
117. A. Thurber, K. M. Reddy, and A. Punnoose, Influence of oxygen level on struc-
ture and ferromagnetism in Sn0.95Fe0.05O2 nanoparticles, J. Appl. Phys. 105,
07E706 (2009).
118. X. B. Chen, G. S. Li, Y. G. Su et al., Synthesis and room--temperature ferromag-
netism of CeO2 nanocrystals with nonmagnetic Ca2+ doping, Nanotechnology
20, 115600 (2009).
119. G. Niu, E. Hildebrandt, M. A. Schubert et al., Oxygen vacancy induced room
temperature ferromagnetism in Pr-doped CeO2 thin films on silicon, ACS
Appl. Mater. Interfaces 4, 17496–17505 (2014).
120. N. Paunovic, Z. Dohčević-Mitrović et al., Suppression of inherent ferromagne-
tism in Pr-doped CeO2 Nanocrystals, Nanoscale 4, 5469–5476 (2012).
121. J. M D. Coey, K. Ackland, M. Venkateran, and S. Sen, Collective magnetic
response of CeO2 nanoparticles, Nat. Phys. 12, 694–699 (2016).
122. P. D. C. King, T. D. Veal, F. Fuchs et al., Band gap, electronic structure and
surface electron accumulation in cubic and rhombohedral In2O3, Phys. Rev. B
79, 205211 (2009).
123. Y. K. Yoo, Q. Z. Xue, H. C. Lee et al., Bulk synthesis and high-temperature fer-
romagnetism of (In1−xFex)2O3−s with Cu co-doping, Appl. Phys. Lett. 86, 042506
(2005).
124. X. H. Xu, F. X. Jiang, J. Zhang et al., Magnetic properties of n-type Fe-doped
In2O3 ferromagnetic thin films, Appl. Phys. Lett. 94, 212510 (2009).
125. G. Peleckis, X. L. Wang, and S. X. Dou, Room-temperature ferromagnetism in
Mn and Fe codoped In2O3, Appl. Phys. Lett. 88, 132507 (2006).
126. N. H. Hong, J. Sakai, N. T. Huong et al., Magnetism in transition-metal-doped
In2O3 thin films, J. Phys. Condens. Matter 18, 6897–6905 (2006).
127. Z. G. Yu, J. He, S. F. Xu et al., Origin of ferromagnetism in semiconducting
(In1−x−yFexCuy)2O3−σ, Phys. Rev. B 74, 165321 (2006).
128. M. Venkatesan, R. D. Gunning, P. Stamenov et al., Room-temperature ferro-
magnetism in Mn- and Fe-doped indium tin oxide thin films, J. Appl. Phys.
103, 07D135 (2008).
129. S. S. Farvid, T. Sabergharesou, L. N. Hutfluss et al., Evidence of charge-transfer
ferromagnetism in transparent diluted magnetic oxide nanocrystals: Switching
the mechanism of magnetic interactions, J. Am. Chem. Soc. 136, 7669–7679 (2014).
130. H. Raebiger, S. Lam, and A. Zunger, Control of ferromagnetism via electron
doping in In2O3: Cr, Phys. Rev. Lett. 101, 027203 (2008).
131. R. P. Panguluri, P. Kharel, C. Sudakar et al., Ferromagnetism and spin-polar-
ized charge carriers in In2O3 thin films, Phys. Rev. B 79, 165208 (2009).
132. D. Bérardan, E. Guilmeau, and D. Pelloquin, Intrinsic magnetic properties of
In2O3 and transition-metal doped In2O3, J. Magn. Magn. Mater. 320, 983–989
(2008).
133. S. H. Babu, S. Kaleemulla, N. M. Rao, and C. Krishnamoorthi, Indium oxide: A
transparent, conducting ferromagnetic semiconductor for spintronic applica-
tions, J. Magn. Magn. Mater. 416, 66–74 (2016).
References 453
134. J. Lee, N. G. Subramaniam, I.A. Kowalik et al., Towards a new class of heavy
ion doped magnetic semiconductors for room temperature applications, Sci.
Rep. 5, 17053 (2015).
135. U. Ozgur, Y. I. Alivov, C. Liu et al., A comprehensive review of ZnO materials
and devices, J. Appl. Phys. 98, 041301 (2005).
136. A. Janotti and C. G. van de Walle, Fundamentals of Zinc oxide as a semicon-
ductor, Rep. Prog. Phys. 76, 126501 (2009).
137. F. Pan, C. Song, X. J. Liu, Y. C. Yang, and F. Zeng, Ferromagnetism and pos-
sible application in spintronics of transition-metal doped ZnO films, Mater.
Sci. Eng. R 62, 1–35 (2008).
138. M. Bouloudenine, N. Viart, S. Colis et al., Antiferro-magnetism in bulk Zn1−x
CoxO magnetic semiconductors prepared by the coprecipitation technique,
Appl. Phys. Lett. 87, 052501 (2005).
139. J. Alaria, P. Turek, M. Bernard et al., No ferromagnetism in Mn-doped ZnO
semiconductors, Chem. Phys. Lett. 415, 337–341 (2005).
140. Z. Jin, T. Fukumura, M. Kawasaki et al., High throughput fabrication of tran-
sition-metal-doped epitaxial ZnO thin films: A series of oxide-diluted mag-
netic semiconductors and their properties, Appl. Phys. Lett. 78(24), 3824–3826
(2001).
141. R. Janisch, P. Gopal, and N. A. Spaldin, Transition metal-doped TiO2 and ZnO-
present status of the field, J. Phys. Condens. Matter 17, R657–R689 (2005).
142. W. Prellier, A. Fouchet, and B. Mercey, Oxide-diluted magnetic semiconduc-
tors: A review of the experimental status, J. Phys. Condens. Matter 15, R1583–
R1601 (2003).
143. S. J. Pearton, W. H. Heo, M. Ivill et al., Dilute magnetic semiconducting oxides,
Semicond. Sci. Technol. 19, R59–R74 (2004).
144. M. Venkatesan, C. B. Fitzgerald, J. G. Lunney, and J. M. D. Coey, Anisotropic
ferromagnetism in substituted zinc oxide, Phys. Rev. Lett. 93, 177206 (2004).
145. B. Martinez, F. Sandiumenge, L. Balcells et al., Role of the microstructure on
the magnetic properties of Co-doped ZnO nanoparticles, Appl. Phys. Lett. 86,
103113 (2005).
146. N. Khare, M. J. Kappers, M. Wei et al., Defect induced ferromagnetism in
Co-doped ZnO, Adv. Mater. 18, 1449–1452 (2006).
147. L. E. Halliburton, N. C. Giles, N. Y. Garces et al., Production of native donors
in ZnO by annealing at high temperature in Zn vapor, Appl. Phys. Lett. 87,
172108 (2005).
148. D. A. Schwartz and D. R. Gamelin, Reversible 300 K ferromagnetic ordering in
a diluted magnetic semiconductor, Adv. Mater. 16, 2115–2119 (2004).
149. R. P. Borges, R. C. da Silva, S. Magalhaes et al., Magnetism in Ar-implanted
ZnO, J. Phys. Condens. Matter 19, 476207 (2007) (see also ibid 20, 429801).
150. P. Stamenov, M. Venkatesan, L. S. Dornales et al., Magnetoresistance of
Co-doped ZnO thin films, J. Appl. Phys. 99, 08M124 (2006).
151. K. R. Kittilstved, W. K. Liu, and D. R. Gamelin, Electronic origins of polarity-
dependent high TC ferromagnetism in oxide diluted magnetic semiconductors,
Nat. Mater. 5, 291 (2006).
152. J. R. Neal, A. J. Behan, R. M. Ibrahim et al., Room temperature magneto-optics
of ferromagnetic transition-metal doped ZnO thin films, Phys. Rev. Lett. 96,
107208 (2006).
153. Z. H. Zhang, X. F. Wang, J. B. Xu et al., Evidence of intrinsic ferromagnetism in
individual dilute magnetic semiconducting nanostructures, Nat. Nanotechnol.
4, 523–527 (2009).
154. X. H. Xu, H. J. Blythe, M. Ziese et al., Carrier-induced ferromagnetism in
ZnMnAlO and ZnCoAlO thin films at room temperature, New J. Phys. 8, 135–
144 (2006).
155. A. J. Behan, H. Mokhtari, H. Blythe et al., Two magnetic regimes in doped
ZnO corresponding to a dilute magnetic semiconductor and a dilute magnetic
insulator, Phys. Rev. Lett. 100, 047206 (2008).
156. M. Venkatesan, R. D. Gunning, P. Stamenov et al., Magnetic, magnetotrans-

port and optical properties in Al doped Zn0.95Co0.05 O thin films, Appl. Phys.
Lett. 90, 242508 (2006).
157. S. T. Oschenbein, Y. Feng, K. M. Whittaker et al., Charge-controlled mag-
netism in colloidal doped semiconductor nanocrystals, Nat. Nanotechnol. 4,
681–687 (2009).
158. J. H. Park, M. G. Kim, H. M. Jang et al., Co metal clustering as the origin of
ferromagnetism in Co-doped ZnO thin films, Appl. Phys. Lett. 84, 1338–1341
(2004).
159. L. S. Dorneles, M. Venkatesan, R. Gunning et al., Magnetic and structural
properties of Co-doped ZnO thin films, J. Magn. Magn. Mater. 310, 2087–
2088 (2007).
160. K. Rode, R. Mattana, A. Anane et al., Magnetism of (Zn,Co)O thin films probed
by x-ray absorption spectroscopies, Appl. Phys. Lett. 92, 012509 (2008).
161. M. Ivill, S. J. Pearton, S. Rawal et al., Structure and magnetism of cobalt-doped
ZnO thin films, New J. Phys. 10, 065002 (2008).
162. D. P. Norton, M. E. Overberg, S. J. Pearton et al., Ferromagnetism in VCo-
implanted ZnO, Appl. Phys. Lett. 83, 5488–5490 (2003).
163. D. C. Kundaliya, S. B. Ogale, S. E. Lofland et al., On the origin of high-temper-
ature ferromagnetism in the low-temperature processed Mn–Zn–O system,
Nat. Mater. 3, 709–712 (2004).
164. A. Zhukova, A. Teiserskis, S. van Dijken et al., Giant moment and magnetic
anisotropy in Co-doped ZnO films grown by pulse-injection metal organic
chemical vapor deposition, Appl. Phys. Lett. 89, 232503 (2006).
165. K. Rode, Contribution à l’étude des semiconducteurs ferromagnétiques: Cas
des films minces d’oxyde de zinc dopé au cobalt, PhD thesis, Université Paris
XI, France, 2006.
166. S. Ghosh, V. Sih, W. H. Lau et al., Room-temperature spin coherence in ZnO,
Appl. Phys. Lett. 86, 232507 (2005).
167. T. C. Droubay, D. J. Keavney, T. C. Kaspar et al., Correlated substitution in
paramagnetic Mn2+-doped ZnO epitaxial films, Phys. Rev. B 79, 155203 (2009).
168. A. Barla, G. Schmerber, E. Beaurepaire et al., Paramagnetism of the Co sublat-
tice in ferromagnetic Zn1−xCoxO films, Phys. Rev. B 76, 125201 (2007).
169. T. Tietze, M. Gacic, G. Schütz et al., XMCD studies on Co and Li doped ZnO
magnetic semiconductors, New J. Phys. 10, 055009 (2008).
170. J. J. Beltrán, C. A. Barrero, and A. Punnoose, Combination of defects plus mixed
valence of transition metals: A strong strategy for ferromagnetic enhancement
in ZnO Nanoparticles, J. Phys. Chem. C 120, 8969–8978 (2016).
171. B. K. Roberts, A. B. Pakhomov, P. Voll et al., Surface scaling of magnetism
in Cr-ZnO dilute magnetic dielectric thin films, Appl. Phys. Lett. 92, 162511
(2008).
172. J. M. D. Coey, R. D. Gunning, M. Venkatesan, P. Stamenov, and K. Paul,
Magnetism in defect-ridden oxides, New J. Phys. 12, 053025 (2010).
173. M. V. Limaye, S. B. Singh, R. Das, P. Poddar, and S. K. Kulkarni, Room tempera-
ture ferromagnetism in undoped and Fe doped ZnO nanorods: Microwave-
assisted synthesis J. Solid State Chem. 184, 391–400 (2011).
174. Q. Y. Xu, H. Schmidt, S. Q. Zhou et al., Room temperature ferromagnetism in
ZnO films due to defects, Appl. Phys. Lett. 92, 082508 (2008).
175. N. H. Hong, E. Chikoidze, and Y. Dumont, Ferromagnetism in laser-ablated
ZnO and Mn-doped ZnO thin films: A comparative study from magnetization
and Hall-effect measurements, Phys. B 404, 3978–3981 (2009).
176. H. N. Hong, A. Barla, J. Sakai et al., Can undoped semiconducting oxides be
ferromagnetic? Phys. Status Solidi C 4(12), 4461–4466 (2007).
177. C. S. Ong, T. S. Herng, X. L. Huang, Y. P. Feng, and J. Ding, Strain-induced
ZnO spinterfaces, J. Phys. Chem. C 116, 610–617 (2012).
178. M. Khalid, M. Ziese, A. Setzer et al., Defect-induced magnetic order in pure
ZnO films, Phys. Rev. B 80, 035331(2009).
References 455
179. P. Zhan, W. P. Wang, Z. Xie et al., Substrate effect on the room-temperature

ferromagnetism in undoped ZnO films, Appl. Phys. Lett. 101, 031913 (2012).
180. D. Gao, Z. Zhang, Y. Li et al., Abnormal room temperature ferromagnetism
in CuO–ZnO heterostructures: Interface related or not? Chem. Commun 51,
1151–1153 (2015).
181. T. Taniguchi, K. Yamaguchi, A. Shigeta, Y. Matsuda, and S. Hayami, Enhanced
and engineered d0 ferromagnetism in molecularly-thin zinc oxide nanosheets,
Adv. Func. Mater. 13, 2140–2145 (2013).
182. J. Chaboy, R. Boada, C. Piquer et al., Evidence of intrinsic magnetism in capped
ZnO nanoparticles, Phys. Rev. B 82, 064411 (2010).
183. C. Guglieri, M. A. Laguna-Marco, M. A. García, N. Carmona, and E. Ce ́spedes,
XMCD Proof of Ferromagnetic Behavior in ZnO Nanoparticles, J. Phys Chem.
C 116, 6608–6614 (2012).
184. A. Sundaresan and C. N. R. Rao, Implications and consequences of ferromag-
netism universally exhibited by inorganic nanoparticles, Solid State Commun.
140, 1197–2000 (2009).
185. X. Y. Xu, C. X. Xu, J. Dai et al., Size dependence of defect-induced room tem-
perature ferromagnetism in undoped ZnO nanoparticles. J. Phys. Chem. C
116, 8813–8818 (2012).
186. X. F. Bie, C. Z. Wang, H. Ehrenberg, Y. J. Wei, and G. Chen, Room-temperature
ferromagnetism in pure ZnO nanoflowers, Solid State Sci. 12, 1365–1367
(2010).
187. B. B. Straumal, A. A. Mazilkin, S. G. Protasova et al., Magnetization study of
nanograined pure and Mn-doped ZnO films: Formation of a ferromagnetic
grain-boundary foam, Phys. Rev. B 79, 205206 (2009).
188. B. B. Straumal, A. A. Mazilkin, S. G. Protasova et al., Ferromagnetism of nano-
structured zinc oxide films, Phys. Metal. Metallog. 113, 1244–1256 (2012).
189. T. Tietze, P. Audehm, Y. C. Chen et al., Interfacial dominated ferromagnetism
in nanograined ZnO: A μSR and DFT study, Sci. Rep. 5, 8871 (2015).
190. J. M. D. Coey, High-temperature ferromagnetism in dilute magnetic oxides,
J. Appl. Phys. 97, 10D313 (2004).
191. T. Dietl, H. Ohno, F. Matsukura et al., Zener model description of ferromagne-
tism in zinc-blende magnetic semiconductors, Science 287, 1019 (2000).
192. S. Methfessel and D. C. Mattis, Magnetic semiconductors, in Handbuch der
Physik, vol. 18(1), S. Flügge (Ed.), Springer-Verlag, Berlin, 1968.
193. J. M. D. Coey, M. Venkatesan, and C. B. Fitzgerald, Donor impurity band
exchange in dilute ferromagnetic oxides, Nat. Mater. 4, 173–179 (2005).
194. D. J. Priour Jr. and S. Das Sarma, Clustering in disordered ferromagnets:
The Curie temperature in diluted magnetic semiconductors, Phys. Rev. B 73,
165203 (2006).
195. T. Dietl, Origin and control of magnetism in dilute magnetic semiconductors
and oxides, J. Appl. Phys. 103, 07D111 (2008).
196. D. M. Edwards and M. I. Katsnelson, High-temperature ferromagnetism of
sp electrons in narrow impurity bands: Application to CaB6, J. Phys. Condens.
Matter 18, 7209–7225 (2006).
197. J. M. D. Coey, K. Wongsaprom, J. Alaria et al., Charge transfer ferromagnetism
in oxide nanoparticles, J. Phys. D: Appl. Phys. 41, 134012 (2008).
198. J. M. D. Coey, J. T. Mlack, M. Venkatesan, and P. Stamenov, Magnetization
process in dilute magnetic oxides, IEEE Trans. Magn. 46, 2501–2504 (2010).
199. R. Gréget, G. L. Nealon, B. Vileno et al., Magnetic properties of gold nanoparti-
cles: A room-temperature quantum effect, Chem. Phys. Chem. 13, 3092–3097
(2012).
200. G. L. Nealon, B. Donnio, R. Gréget et al., Magnetism in gold nanoparticles,
Nanoscale 4, 5244–5258 (2012).
201. A. Hernando, P. Crespo, and M. A. Garcia, Origin of orbital ferromagnetism
and giant magnetic anisotropy on the nanoscale, Phys. Rev. Lett. 96, 057206
(2006).
202. A. Hernando, P. Crespo, M. A. Garcia, M. Coey, A. Ayuela, and P. M. Eschnique,

Revisiting magnetism of capped Au and ZnO nanoparticles: Surface band
structure and atomic orbitals with giant magnetic moment, Phys. Status Solidi
B 248, 2352–2360 (2011).
203. S. Sen, K. S. Gupta, and J. M. D. Coey, Mesoscopic structure formation in con-
densed matter due to vacuum fluctuations, Phys. Rev. B 92, 155115 (2015).
204. M. Venkatesan, P. Kavle, S. B. Porter, K. Ackland, and J. M. D. Coey, Magnetic
analysis of polar and nonpolar oxide substrates, IEEE Trans. Magn. 50,
2201704 (2014).
205. S. K. S. Patel, K. Dewangan, S. K. Srivastav, and N. S. Gajbhiye, Synthesis of
monodisperse In2O3 nanoparticles and their d0 ferromagnetism, Curr. Appl.
Phys. 14, 905–908 (2014).
206. K. Ackland, and J. M. D. Coey, Room-temperature magnetism in CeO2 – a
review. Phys. Rep. 746, 1–40 (2018).
207. A. S. Esmaeily, M. Venkatesan, S. Sen and J. M. D. Coey, d-zero magnetism
in nanoporous amorphous alumina membranes, Phys. Rev. Mater. 2, 054404
(2018).
208. J. M. D. Coey, d-zero magnetism in oxides, Nat. Mater. 18 (2019), in press.
209. J. M. D. Coey, M. Venkatesan, and P. Stamenov, Surface magnetism of stron-
tium titanate, J. Phys.: Condens. Matter 28, 485001 (2016).
11
Magnetism of
Complex Oxide
Interfaces

11.2 Band Diagram 459
11.3 Ferroelectric-Manganite Heterostructures 461
11.3.1 Interface Magnetic Interaction: Molecular
Orbital Picture 462
11.3.2 Interfaces with Ferroelectrics 465
11.4 Manganite Superlattices and Interfaces 468
11.5 Exchange Bias Effects across the Ferromagnetic/
G-Type Antiferromagnetic Interface 472
11.6 Cobaltite Thin Films 475
11.7 Theoretical Predictions for Interfaces Grown in the
(111) Direction 477
11.8 Conclusion 480
Acknowledgments 480
References 480
457
458 Chapter 11. Magnetism of Complex Oxide Interfaces
11.1 INTRODUCTION
Controlling the magnetic properties of magnetic materials in bulk crystals
and at interfaces is crucial for spintronic applications and related phenom-
ena. In this chapter, we will review theoretical studies focused on interfaces
or heterostructures of complex oxides such as perovskite transition-metal
oxides (TMOs). TMOs have been a focus of materials science for decades
because of their exotic behavior originating from strong electron-electron
or electron-lattice interactions, including high critical temperature (Tc )
superconductivity in cuprates [1] and colossal magnetoregistance effects in
manganites [2, 3]. In fact, after the discovery of high-Tc cuprates, crystal
growth and measurement techniques have been improved dramatically, and
a variety of novel phenomena have been discovered, such as novel metal-
insulator transitions and spin charge orbital ordering [4, 5]. Further, the
recent progress in thin-film growth techniques led to the discovery of metal-
lic states at interfaces between dissimilar insulators [6, 7]. Thus, while bulk
complex TMOs can provide interesting functionalities, heteroengineering
of TMOs could also provide additional useful and novel functionalities for
applications.
One potential application of such TMOs is a tunneling magneto-
resistance (TMR) [8, 9] device using ferromagnetic metallic manganites.
Currently, popular materials used for such devices are ferromagnetic
metallic Fe with MgO as tunneling barriers [10–14] (see also Chapter 12,
Volume 1 of this book). Replacing Fe with manganites such as La1−xSrxMnO3
is expected to improve the TMR performance because of the higher spin
polarization. While such efforts have been reported [15–17], manganite-
based TMR devices still do not outperform Fe-MgO-based devices [18–21];
while the TMR ratio at low temperatures could exceed more than 1000%, it
becomes extremely small at room temperature. A possible mechanism for
such a rapid decrease in the TMR ratio for manganite-based devices is that
the interface magnetization decreases more rapidly than that in the bulk
region [22, 23]. Thus, to improve the TMR performance of manganite-based
devices, it is necessary to carefully engineer their interfaces [19].
This principle is applied to all other TMOs; for practical applications,
we need to have full understanding of their interface properties. However,
this task is not easy because strong correlation effects sometimes render the
interfacial properties incompatible with the standard density functional the-
ory (DFT), which is widely used to describe a wide range of materials prop-
erties. Therefore, one should take multiple approaches, such as advanced
theoretical techniques to solve model Hamiltonians, phenomenological
calculations, DFT calculations, and the combination of some of them. This
chapter describes examples of such theoretical studies of TMO interfaces
employing a variety of methods.
This Chapter is organized as follows: in Section 11.2, the band diagram
of many TMOs is presented. From experimental estimations of the work-
functions, we are able to align the bands of several complex oxides allowing
for estimations of the direction of charge flow when they are combined in
11.2 Band Diagram 459
heterostructures. Section 11.3 describes theoretical work at the interfaces of

a ferroelectric and a magnetic manganite. We provide rules to judge the type
of coupling between the spins at opposite sides of the interfaces. We also
study how the electric field of the ferroelectric component affects the elec-
tronic density of the other component, typically a metallic manganite, thus
altering its transport properties and sometimes even inducing a phase tran-
sition. In Section 11.4, manganite superlattices are studied. Depending on
the type of manganite components used in the superlattices and their thick-
ness, a variety of states can be found, sometimes involving metal to insulator
transitions. The possibility of a novel type of spin frustration only present
at interfaces is also discussed. In Section 11.5, the puzzling exchange bias
effect is addressed for the interfaces involving a ferromagnet on one side and
a G-type (staggered) antiferromagnetic state on the other. A rationalization
of this effect based on the spin-orbit coupling is provided. In Section 11.6,
cobaltite thin films are studied, with emphasis on the dimensionality
dependence of the spin state that can be high or low, depending on the
geometry. Finally, in Section 11.7, the intriguing growth of interfaces along
the unusual (111) direction is described. It is argued that, by this procedure,
a physical realization of the quantum spin Hall state at room temperature
can be achieved.
Overall, this chapter is presented in such a manner that the most impor-
tant physical aspects of the many problems addressed are described, while
the technicalities are only briefly mentioned. We encourage readers to con-
sult the literature cited for more details.
11.2 BAND DIAGRAM
For the success of the emergent field of oxide heterostructures, it is crucial
to properly determine the relative work functions of the materials involved,
since these work functions control the bending of the valence and conduc-
tion bands (VB and CB) of the constituent materials at the interfaces, and
ultimately the carrier concentration at those interfaces. Work functions of
conventional metals and semiconductors have been studied for decades,
establishing the fundamental background for current electronics. As a con-
sequence, for the next-generation electronic devices utilizing the complex
properties of correlated-electron systems, such as high-Tc cuprates and CMR
manganites, determining the relative work functions of a variety of transi-
tion-metal oxides is equally important.
In this section, we provide a crude estimation of the band diagram of
perovskite transition-metal oxides using experimental data currently avail-
able. This is achieved by combining information from chemical potential
shifts obtained using photoemission spectroscopy (PES) with diffusion volt-
age measurements on heterostructures.
PES is a powerful technique in this context and has provided fun-
damental information to uncover the properties of complex oxides. The
workfunction of a target material could be directly measured by this tech-
nique. On the other hand, the diffusion voltage Vd (or built-in potential)
measurement requires a junction consisting of two materials. By measur-

ing the diffusion voltage Vd (or built-in potential) of the junction, one can
also extract the chemical potential differences between the constituents.
If there are no extra effects, such as interface polarities, impurities, or
lattice reconstructions, the diffusion voltage Vd is equivalent to the work-
function difference between the two members of the junction. If the work
function of one of the materials is known, the work function of the other
can be estimated. In Ref. [24], we started investigating this interesting
subject by constructing the band diagram for the case of cuprates and
manganites.
In addition to cuprates and manganites, iron-based oxides, ruthenium-
based oxides, and aluminum oxides are also important materials for spin-
tronics applications. In the cases of LaFeO3 and BiFeO3, the Fe+3 ion has a
large moment, formally 5m B , and BiFeO3 is the only known material show-
ing multiferroic behavior at room temperature [25, 26]. SrRuO3 is an itin-
erant ferromagnet with TC  160 K [27]. Finally, LaAlO3 is a nonmagnetic
wide band gap insulator. This material has been used as a tunneling barrier
for devices for several years. Moreover, since the discovery of high-mobility
electron gases at an interface between LaAlO3 and SrTiO3, considerable
attention has been paid to LaAlO3 [7]. In this section, we will complement
the work that started in Ref. [24] by adding these other oxides into the pre-
viously established band diagram with a brief summary of experimental
results.
In Ref. [28, 29], the workfunction of cubic LSMO was estimated to be
~4.8 eV. This value is slightly smaller than that of SrRuO3 ~5.2 eV [30]. In
Ref. [31], the diffusion voltage, i.e. the workfunction difference, between
SrRuO3 and both LaFeO3 and BiFeO3 was studied. The electron affinity
and the band gap were determined to be ~3.3 eV and ~2.7 eV, respectively,
for LaFeO3 [32] and ~3.3 eV and ~2.8 eV, respectively, for BiFeO3 [31, 33,
34]. From optical absorption experiments, the band gap of LaAlO3 has
been estimated to be 5.6 eV [35]. Furthermore, the electron affinity was
deduced to be 2.5eV [36]. From these numbers, we can construct the band
diagram covering cuprates, manganites, ferrites, SrRuO3, and LaAlO3 as
shown in Figure 11.1. It is remarkable that all of the valence band maxima
and the conduction band minima of the listed compounds are located
inside the band gap of LaAlO3. This is why LaAlO3 can work as a nice
tunneling barrier for many materials, as long as the polarity discontinuity
induces interfacial conduction electron gases [7]. It should be also noted
that the Fermi level of LSMO is expected to lie in the middle of the band
gap of LaFeO3 and BiFeO3. Therefore, we do not expect a considerable
charge transfer between the two compounds when an interface is formed
between them.
In summary, band diagrams as discussed in this section are important
to guide in the fabrication of complex oxide interfaces, particularly with
regard to the possibility of charge transfer between the components that
may induce novel states at those interfaces.
Vacuum level
2.5
3.3 3.3
0.8
x=0
EF CB 0.15 x=0.6 5.2
0.35 2.7 2.8
0.65
0.3—0.5 1.3 <0.1
~ 1.5 × 0.4
1.5eV 5.6
1 0.7 x=0.2 0.8 0.9
0.4
VB
FIGURE 11.1 Schematic band diagrams of La2CuO4, Sm2CuO4 (Nd2CuO4), Nb0.01-SrTiO3, Nb0.05-SrTiO3, YBa2Cu3Oy,
La1−x SrxMnO3, SrRuO3, LaFeO3, BiFeO3, and LaAlO3 based on diffusion voltage measurements [30, 31, 37–40],
photoemission spectroscopy [28, 29, 41–43], and optical absorption [32, 34, 35]. Tops of valence bands (VB)
and bottoms of conduction bands (CB) are indicated by solid lines, while chemical potentials are indicated by
dashed lines. The vacuum level is indicated by a dotted line. Dark (light) hatched regions are occupied (unoc-
cupied) states.
11.3 FERROELECTRIC-MANGANITE
HETEROSTRUCTURES
As in the case of ferromagnetism, the presence of ferroelectricity is also an
important property of functional materials. Moreover, the combination of
ferromagnetism and ferroelectricity gives rise to the so-called “multiferroic-
ity”, which is physically interesting due to the coexistence of two types of
orders. Multiferroic materials are important for novel applications because
electric fields that are usually associated with ferroelectricity, may be used
to manipulate magnetic properties if the two orders (ferromagnetism and
ferroelectricity) are coupled as expected in multiferroics [44, 45]. However,
in general for single phase materials, it is very hard to host these two ferroic
orders simultaneously [46]. For this reason, heterostructures constructed by
combining strong ferroelectric and strong ferromagnetic (FM) oxides have
received considerable attention, because they are good candidates to real-
ize the above-mentioned magnetoelectric functions, namely, electric field
control of magnetism or vice versa. To pursue these magnetoelectric appli-
cations, it is necessary to perform both theoretical and experimental stud-
ies involving the interfaces between ferroelectric and FM materials, such as
BiFeO3/La1−xSrxMnO3, and Pb(Zr,Ti)O3/La1−xSrxMnO3.
11.3.1 Interface Magnetic Interaction:

Molecular Orbital Picture
Several complex transition metal oxide are intrinsically magnetic. Therefore,
when different compounds are combined to form interfaces, novel magnetic
ordering patterns could emerge. In this section, we describe how different
TMOs interact when an interface is formed.
In bulk magnetic materials, magnetic ordering is induced through a
variety of interactions. It is well-established that, in strongly correlated Mott
insulators, local moments interact with each other via superexchange (SE)
interactions [47]. The sign of exchange interaction between two magnetic
sites would depend on the angle formed by the magnetic sites and ligand
ions [48, 49] or orbital ordering [50]. In the presence of itinerant electrons,
the interaction between localized spins is mediated by those electrons. Such
interactions include the double-exchange (DE) interaction [51] and the
RKKY interaction [52–55].
At an interface between different materials, the symmetry is lowered,
and its natural consequence is the appearance of the Dzyaloshinskii–Moriya
interaction [55, 56]. In addition, constituent materials could interact with
each other via interfacial magnetic interactions. As discussed in Section 2,
different TMOs have different chemical potentials in general. Therefore,
electrons could be transfered across interfaces and a purely localized picture
may not be entirely true, even if the constituent TMOs of the heterostructure
are Mott insulators. For such a complicated situation, a simple description
based on molecular orbitals turned out to be useful [57]. While it is qualita-
tive, this provides a physically transparent picture by which both experiment
and theory may greatly benefit.
Here we describe a general consideration for the magnetic interaction at
an interface involving manganites [57]. We focus on the interfacial interac-
tion derived by d3 z2 -r 2 orbitals which have the largest hybridization along
the z layer-stacking direction. We see that the sign of the magnetic interac-
tion via the d3 z2 -r 2 orbitals is naturally fixed based on the molecular orbit-
als formed at the interface and the generalized Hund’s rule. The molecular
orbitals effectively lift the degeneracy between the d3 z2 -r 2 and dx2 - y2 orbitals
by an energy of order the hopping amplitude. The t2 g electrons are assumed
to be electronically inactive and considered as localized spins when a finite
number of electrons occupy the t2 g orbitals; for example, S = 3 / 2 in a man-
ganite component. The TM region is specified by the number of electrons
occupying a d3 z2 -r 2 orbital. Reference [57] also discussed how to general-
ize the molecular-orbital-based argument for more complicated situations
including t2 g orbitals.
(d3 z2 -r 2 )0 system: Let us start from the simplest case, an interface between
Mn and a (d3 z2 -r 2 )0 system (Figure 11.2, top figure). In this case, the bonding
(B) orbital is occupied by an electron whose spin is parallel to the localized
t2 g spin of Mn. When there are other unpaired electrons in the (d3 z2 -r 2 )0
system, their spins align parallel to that of the electron in the B orbital, due
to the Hund coupling. Thus, the FM coupling is generated between Mn and

(d3 z2 -r 2 )0 systems. When the orbital dx2 - y2 is much lower in energy than
d3 z2 -r 2 in the interfacial Mn layer, the Mn and (d3 z2 -r 2 )0 systems are virtually
decoupled. Thus, the magnetic coupling becomes antiferromagnetic (AFM)
due to the superexchange (SE) interaction between t2 g electrons.
(d3 z2 -r 2 )1,2 normal: This simple consideration can be easily generalized
to (d3 z2 -r 2 )1 and (d3 z2 -r 2 )2 systems. First, we consider that the d3 z2 -r 2 and
dx2 - y2 orbitals are nearly degenerate in the interfacial Mn layer, and the
unoccupied dx2 - y2 level in the TM region is much higher than the d3 z2 -r 2
level (Figure 11.2, second top figures). We call this configuration “normal”
(N) configuration. In the lowest energy configuration, the B orbitals and the
Mn dx2 - y2 orbital are occupied by electrons. For the (d3 z2 -r 2 )1 system, the FM
interaction is favored, as in the (d3 z2 -r 2 )0 system. On the other hand, for the
(d3 z2 -r 2 )2 system, the down electron orbital is hybridized with the minority
band in the Mn region. Thus, the “down” B orbital is higher in energy and
has larger weight on the TM than the “up” B orbital. Because of the Hund
(d 3 −r 2
) 0
z2
d3 Mn
−r 2 z2
d
TM −y2 x2
d3 −r 2 z2
t2 g
(d 3 −r 2
) 1
z2
(d 3 −r 2
) 2
z2
Normal Normal
(d 3 −r 2
) 1
z2
(d 3 −r 2
) 2
z2
Anomalous Anomalous
d −y2 x2
(d 3 −r 2
) 1
z2
(d 3 −r 2
) 2
z2
JT JT
FIGURE 11.2 Molecular orbitals formed by d3 z 2 -r 2 orbitals on Mn and the

(d )
0,1,2
3 z 2 -r 2
systems. In the normal (anomalous) configurations, d3 z 2 -r 2 and d x 2 - y 2
orbitals are nearly degenerate in the interfacial Mn, and the unoccupied d x 2 - y 2
orbital in the neighboring TM is higher in energy (lower in energy than the occu-
pied Mn d x 2 - y 2 ). In the Jahn–Teller case, the d3 z 2 -r 2 level is significantly lower than
the d x 2 - y 2 level. Black (light) lines indicate the level of majority (minority) spins.
The up level and down level are exchange split, resulting in the level scheme as
indicated. The minority levels are neglected where these are irrelevant. (After
Okamoto, S., Phys. Rev. B 82 024427, 2010. With permission.)
coupling with the “down” electron in the B orbital, other unpaired electrons,
if they exist, tend to be antiparallel to the Mn spin.
(d3 z2 -r 2 )1,2 anomalous: When the dx2 - y2 level in the TM region becomes
lower than the Mn dx2 - y2 level, the charge transfer occurs. We shall call
this configuration “anomalous” (AN) (Figure 11.2, third figures). The elec-
tron transferred to the TM dx2 - y2 orbital has the same spin as the higher
energy B orbital due to the Hund coupling (indicated by arrows). Therefore,
the sign of the magnetic coupling between the Mn and (d3 z2 -r 2 )1,2 systems is
unchanged.
The magnetic interactions discussed so far are insensitive to the elec-
tronic density in the interfacial Mn because the interactions are mainly
derived from the virtual electron excitation from the occupied d3 z2 -r 2 orbital
in the TM region to the unoccupied counterpart in the Mn region.
Next, we consider the case where the d3 z2 -r 2 level is much lower than the
dx2 - y2 in the interfacial Mn layer due to either the local Jahn–Teller distor-
tion, or compressive strain originating from the substrate (Figure 11.2, lower
figures denoted by “JT”).
(d )
1
JT: When the Mn d3 z2 -r 2 density is close to 1, the superex-
3 z2 -r 2
change interaction between the occupied d3 z2 -r 2 orbitals becomes AFM. On
the other hand, when the density is much less than 1, the FM interaction
( ) ( )
0 1
between d3 z2 -r 2 configuration on Mn and d3 z2 -r 2 becomes dominant.
( )
2
d3 z2 -r 2 JT: When the Mn d3 z2 -r 2 occupancy is close to 1, “up” elec-
trons are localized on both sites while “down” electrons can be excited from
( )
2
the d3 z2 -r 2 system to the Mn minority level, i.e. the “down” electron den-
sity is virtually reduced in the TM region. As a result, unpaired spins, if
they exist, become parallel to the “up” spin, i.e. FM coupling. When the Mn
d3 z2 -r 2 density becomes much less than 1, the “up” AB orbital becomes less
occupied while keeping the occupancy of B orbitals relatively unchanged.
Eventually, the “down” density in the TM region becomes larger than the
( )
2
“up” density, and the magnetic coupling between the Mn and d3 z2 -r 2
regions becomes AFM.
It is instructive to see the relation between the molecular-orbital picture
described here, and other known mechanisms, although those are for bulk
( )
0
magnetic interactions. An interaction with a d3 z2 -r 2 system with a t2 g
spin is ferromagnetic. This corresponds to the double exchange interaction
originally proposed by Zener [51]. When the e g degeneracy is lifted by the JT
effect and there is one e g electron per site, one should expect the antiferro-
magnetic superexchange interaction [47]. This situation corresponds to the
(d )
1
3 z2 -r 2
Jahn–Teller case. Indeed, we have antiferromagnetic interaction
in this case. On the other hand, when the e g degeneracy remains, one could
expect the ferromagnetic interaction with antiferro-type orbital ordering,
( )
1
i.e. d3 z2 -r 2 / dx2 - y2 [50]. This situation corresponds to the d3 z2 -r 2 normal
case. Indeed, we have ferromagnetic interaction in this situation.
For what systems can we apply the above discussion? In Ref. [58], the
orbital and magnetic states at a cuprate/manganite interface were studied.
In cuprates, the d3 z2 -r 2 orbital is completely filled while the dx2 - y2 orbital is
less-than-half filled. As discussed in Section 2, cuprates have larger work-
function than manganites, and the electron transfer from the latter to the
( )
2
former is expected. Thus, this situation corresponds to d3 z2 -r 2 anomalous.
In agreement with these observations the antiferromagnetic spin alignment
was observed in Ref. [58]. Such antiferromagnetic coupling between cuprate
and manganite causes an interesting consequence called inverse spin switch
behavior, the increase in the superconducting transition temperature, in
manganite/cuprate/manganite trilayer systems [59, 60]. In Ref. [61], the mag-
netic states at a BiFeO3/(La,Sr)MnO3 interface were analyzed. Since an Fe3+
ion has the high spin state with S = 5 / 2 and the minimal electron transfer is
expected across such an interface according to the band diagram shown in
( ) normal. While we expect
1
Section 11.2, this situation corresponds to d3 z2 -r 2
ferromagnetic coupling across the interface, experimental results show the
antiferromagnetic spin arrangement. Reference [57] argued that, because
of the presence of a robust dx2 - y2 orbital ordering at an interface Mn layer,
this layer and the second Mn layer develop an antiferromagnetic coupling,
resulting in the antiferromagnetic spin arrangement between an interface Fe
layer and the bulk manganite region.
11.3.2 Interfaces with Ferroelectrics

In addition to the famous case of BiFeO3/La1−xSrxMnO3 briefly addressed
in the previous subsection, several other ferroelectrics have been used at
interfaces, as described in this subsection. As a general motivation, and as
described before, note that the most striking effect in these ferroelectric/fer-
romagnetic heterostructures is expected to be the ferroelectric field effect.
Physically, the presence of a ferroelectric polarization can be mimicked by
the presence of charge at the surface or interface. For example, a polarization
o o
of 10 µC/cm2 equals to 0.1 elementary charge per 4 A × 4 A surface. To screen
this surface charge, mobile electrons or holes will be attracted to or repelled
away from the interface, due to the bending of energy bands by electrostatic
potentials, as sketched in Figure 11.3a.
Considering the large polarization of Pb(Zr,Ti)O3 (up to ~80 µC/cm2),
this screening effect and effective charge accumulation at the interface is a
strong force to modulate the carrier density of ferromagnetic materials on
the other side of that same interface, which may alter their physical proper-
ties if correlated electrons systems are used as the ferromagnetic component.
For example, it is well-known that La1−xSrxMnO3 has the best ferromagnetic
metallic behavior when x = 3 / 8. If the interfacial carrier density is modu-
lated away from this region, the ferromagnetism and transport will become
worse [62, 63]. In contrast, if the original doping x is not in the optimal
region, a proper modulation by field effect may enhance the ferromagne-
tism as well as transport [62, 63], as sketched in Figure 11.3b. In fact, there
FIGURE 11.3 Illustrations and results corresponding to a metal-ferroelectric-correlated electron oxide

(CEO) heterostructure studied both experimentally and theoretically. (a) Schematic of tunneling current and
energy bands modulated by ferroelectric polarization. The carrier population is controlled by the direc-
tion of the polarization, which yields either a hole accumulated (top) or depleted (bottom) state in the CEO
layer. (b) Phase diagram of a typical CEO: La1−x SrxMnO3. The field effect modulation will give different results
for x = 0 .20 , 0.33, and 0.50 according to this phase diagram. (c) The theoretical depth profile showing the
changes in the eg electronic density for upward and downward polarizations, at different electronic compo-
sitions. Two different dielectric constants (Ɛ) have been used to compare the degree of ferroelectric control.
These idealized calculations show that the electronic density modifications in the first and second layers can
indeed be very large, compatible with the experimental results. (After Jiang, L. et al., Nano Letters. 13, 5837,
have been many experimental and first-principles calculations addressing

this important issue [64-69]. Because of the rich phase diagrams of man-
ganites, it is expected that the modifications in the carrier density at the
interface induced by the ferroelectric effective interfacial charge may even
induce complete phase transitions from metals to insulators on the manga-
nite side, restricted to the interfacial zone, vastly increasing the modifica-
tions in transport properties. Readers are referred to related topical reviews
for additional details on these interesting ideas [70–73].
Model studies as described before can also be applied to investigate the
field effect modulated magnetism in ferroelectric-manganite heterostruc-
tures. By solving self-consistently the Poisson equation for the electro-
static potential, the distribution of carrier density can be finally obtained,
as shown in Figure 11.3c for example. The modifications in the magnetic
properties can also be simulated, which only occurs within one or two
unit cells of the interface, because of the short screening length in metal-
lic systems such as La1−xSrxMnO3 [75]. Such short range screening limits
the magnetoelectricity effects to occurring within one or two unit cells
[68, 69, 75], although the tunneling current across the interface can still
be significantly modified [62]. To overcome this barrier and improve the
magnetoelectricity, a theoretical design was recently proposed based on a
manganite bilayer [74], as sketched in Figure 11.4a. The asymmetric inter-
faces (n-type vs p-type terminations) will break the symmetry between the
two layers of Mn, then the ferroelectric polarization can tune this asym-
metry, as shown in Figure 11.4b. Accompanying the tuning of the charge
density, a magnetic phase transition may occur between ferromagnetism
and A-type antiferromagnetism. Indeed, both the model simulation and
first-principles calculations confirmed the possibility of ferroelectic field
tuned magnetic transition [74]. The ideal on/off ratio of magnetization can
be as high as 93%.
Summarizing this section for the non-expert readers, it is clear that
the wide variety of possible combinations of complex oxides at interfaces,
with both magnetic and ferroelectric components, can lead to a plethora
of spin arrangements, charge redistributions, phase transitions, and other
interesting effects. This is why oxide interfaces are considered a “play-
ground” to create new artificial materials. Even if the potential promise
of new functionalities and oxide electronics are not realized in the future,
at a minimum these interfaces can provide fertile ground to create novel
states of correlated electronic systems that do not appear in bulk transition
metal oxides.
FIGURE 11.4 Schematic of a ferroelectric (e.g. BaTiO3) and bilayer manganite (e.g. La0.75Sr0.25MnO3) hetero-
structure. (a) Crystal structure. Green = Ba; red = O; cyan = Ti; purple = Mn; yellow = La0.75Sr0.25. The n-/p-type
interfaces are indicated. Left/right are the negative/positive polarization cases, with switched magnetic
orders. (b) The cases of three eg electron densities (spheres) and potential (bars) modulated by asymmetric
interfaces (bricks) and ferroelectric polarization (arrows). (After Dong, S. et al., Phys. Rev. B 88, 140404(R), 2013.
With permission.)
11.4 MANGANITE SUPERLATTICES
AND INTERFACES
As typical correlated electronic materials, the manganese oxides widely
known as manganites have received considerable attention in the past two
decades for their unique physical properties, especially the colossal magne-
toresistance where relatively small magnetic fields induce very large modi-
fications in the resistivity, usually by several orders of magnitude [76–78].
Moreover, these materials are known to have a very rich phase diagram when
varying the chemical composition and the electronic bandwidth, and after
introducing external fields. Several insulating phases arise, with antiferro-
magnetic spin states, including the A-type state in the undoped compound
and the CE-type state at half doping with spin, charge, and orbital order. In
the past decade, these materials have received even more attention because of
remarkable experimental advances that have allowed for the precise control
of the digital growth of manganite heterostructures, leading to layer-by-layer
grown manganite ultrathin films and/or superlattices [79–84]. Interesting
physics emerges in these digital superlattices, such as particular metal-insu-
lator transitions, that do not appear in the materials when in bulk form.
Manganite heterostructures were theoretically studied using the
famous two-orbital double-exchange model [85, 86], which has been repeat-
edly confirmed to be a successful description of manganites [76, 77]. Both
the kinetic double-exchange hopping term, the antiferromagnetic super-
exchange among the localized t2g sping, and the electron-lattice coupling
involving Jahn–Teller modes, have been properly incorporated in the model
Hamiltonian. In addition, the electrostatic potential VMn , which raises from
the polar discontinuity and charge transfer across the interface, should also
be taken into account in the theoretical study of heterostructures. As a first
order approximation, VMn can be assumed to be a function of the eight sur-
rounding A-site ions [87]. By adopting this approximation, Dong and col-
laborators studied the (LaMnO3)2n/(SrMnO3)n superlattices [85]. Despite the
well-known physical properties of the bulk alloy-mixed La2/3Sr1/3MnO3, with
a dominant state that is ferromagnetic and metallic, these superlattices dis-
play a novel n-dependent metal-insulator transition and a spatial modulation
of the magnetization [81, 82, 84]. The model studied by Dong and coworkers
can successfully reproduce the observations reported in experiments. More
specifically, these superlattices are metallic when n < 3, as it occurs for the
material in bulk from, but the artificial structures become insulating once
n ³ 3, as sketched in Figure 11.5. When n ³ 3, the LaMnO3 layers and inter-
facial layers are ferromagnetic, while the inner portion of the SrMnO3 com-
ponent is antiferromagnetic, which also agrees with intuitive expectations
(note that for large enough n, each building block has to display properties
resembling their bulk counterparts) as well as with neutron measurements
[84]. In addition, first-principles calculations also revealed the presence of
layer-modulated magnetism in (LaMnO3)2n/(SrMnO3)n [88–90].
A subsequent experiment now on (LaMnO3)n/(SrMnO3)2n (note the sub-
indexes n and 2n are switched with regard to the previous case) revealed
FIGURE 11.5 Theoretical results for (LaMnO3)2n/(SrMnO3)n superlattices. (a-b) The calculated in-plane and
out-of-plane conductivity, respectively. L2S1 denotes the n = 1 case and other cases are indexed following the
same rule. (c) Sketch of spin (denoted by arrows) arrangements found at interfaces for the cases n = 2 (left) and
n = 3 (right). (d) Sketch of the in-plane current flowing through superlattices. (e) The orbital orders expected to
be found at the interfaces. (After Dong, S. et al., Phys. Rev. B 78, 201102(R), 2008. With permission.)
even more exotic phenomena, more specifically a much enhanced Néel tem-
perature (TN ) for the cases n = 1 or 2 as compared with the corresponding
results for bulk La1/3Sr2/3MnO3 [91]. This enhanced A-type antiferromag-
netism looks non-trivial, since the strained LaMnO3 has a ferromagnetic
tendency [85, 92], while SrMnO3 displays G-type (i.e. spin staggered) antifer-
romagnetism. In fact, the A-type antiferromagnetic ordering in doped man-
ganites has been known for La1−xSrxMnO3 [93], Pr1−xSrxMnO3 [93, 94], and
Nd1−xSrxMnO3 [95, 96]. Early theoretical studies for bulk manganites incor-
porating strong correlation effects already clarified for this magnetic order-
ing the importance of x 2 - y 2 orbital ordering by which itinerant e g electrons
could gain the kinetic energy [97–99]. The aforementioned double-exchange
model with an on-site potential modulation (accompanying the A-site ions)
was also adopted to study the superlattice (RMnO3)n/(AMnO3)2n (R: rare

earth; A: alkaline earth) [100]. As shown in Figure 11.6b–c, the A-type anti-
ferromagnetism was found to be the most favorable ground state phase
in a particular region of the phase diagram (namely, when varying model
parameters), corresponding to the case of large-bandwidth manganites (e.g.
R = La and A = Sr). The A-type antiferromagnetic region is much larger for
the n = 1 case (Figure 11.6b) than for the n = 2 one (Figure 11.6c), in agree-
ment with the experimental observation that TN is higher in the n = 1 super-
lattice. The theoretical study emphasized the crucial role of the Q3 mode of
Jahn–Teller distortions, imposed by the tensile strain that is induced by the
substrate material SrTiO3 into the SrMnO3 layers. This Q3 mode of the Jahn–
Teller distortions prefers the x 2 - y 2 orbital ordering (Figure 11.6f–g), which
enhances/suppresses the double-exchange tendencies in-plane/out-of-plane,
and thus leads to the enhanced A-type antiferromagnetism.
FIGURE 11.6 Theoretical results for the (RMnO3)n/(AmnO3)2n superlattices. (a) Sketch of layers in a finite cluster
with six Mn-oxide layers (gray). “Mn” stands for the MnO2 layers,“R” for the RO layers, and “A” for the AO layers.
The only active layers for the mobile electrons are the Mn layers, but the other layers influence the Mn layers by
their electrostatic potential. (b)–(e) Possible ground state phase diagrams arising from the theoretical calcula-
tions (see text): (b)–(c) Using a SrTiO3 substrate; (d)–(e) Using a LaAlO3 substrate; (b) and (d): n = 1. (c) and (e): n = 2
. The magnetic configurations can be simple (e.g. A-type or C-type antiferromagnetism), or complex (combina-
tion of different spin patterns): see Ref. [100] for more details. The vertical axis is the on-site potential modulation
by the La and Sr layer structure. The horizontal axis is the superexchange intensity between Mn’s t2g spins: the
narrower the bandwidth, the larger the superexchange. As the Jahn–Teller coupling constant, λQ=0.9t is used.
(f)–(g) Orbital ordering patterns using SrTiO3 for the n = 1 and n = 2 cases, respectively. (h)–(i) Orbital ordering
patterns using LaAlO3 for the n = 1 and n = 2 cases, respectively. The cyan bars denote the positions of the SrO
sheets. (After Dong, S. et al., Phys. Rev. B 86, 205121, 2012. With permission.)
Not only the Q3 mode or the tensile strain, but also the lower dimension-
ality could stabilize the A-type antiferromagnetic ordering [101, 102]. This
effect is intensively studied in layered manganites such as La2−2xSr1+2xMn2O7.
It was established that the A-type antiferromagnetic ordering is stabilized in
a wider-range of hole concentration than in the cubic manganites [103, 104].
Interestingly, the C-type antiferromagnetic ordering at large hole concentra-
tions also appears [104], as predicted in Ref. [102]. In this ordering, ferromagnet-
ically-ordered Mn chains lie along an x or y direction, not along the z direction,
which requires compressive strain as discussed below. While the heteroepitaxy
of layered compounds is more difficult than that of cubic perovskite systems,
the idea of “dimensionality” is another useful route to control the electronic
property of complex oxides. We will review such a study in Section 6.
An interesting extension of this physical mechanism is provided by
the opposite Q3 mode, namely, by using compressive strain that arises from
small lattice constant substrates such as LaAlO3. In this case, the calculation
predicted a different result; in particular, the ground state is probably given
by the C-type antiferromagnetism, with 3 z 2 - r 2 orbital ordering as shown in
Figure 11.6h–i. This type of orbital ordering prefers the ferromagnetic dou-
ble-exchange to be out-of-plane, but it displays antiferromagnetic coupling
in-plane. Thus, both the A-type antiferromagnetic and C-type antiferromag-
netic order of (LaMnO3)n /(SrMnO3)2n are basically caused by the substrate
strains. Similar conclusions were also obtained by using first-principles cal-
culations [105].
Contrary to the most studied wide-bandwidth case of La1−xSrxMnO3,
the narrow-bandwidth manganites in bulk form, such as for the case of
Pr1−xCaxMnO3, usually show more complex magnetic/charge/orbital orders
[76, 78]. This situation also occurs in manganite superlattices. Still consid-
ering the case of (RMnO3)n/(AMnO3)2n as an example, the ground state for
narrow bandwidth manganites (e.g. R = Pr and A = Ca) could become quite
different from the above described simple cases of A-type or C-type antifer-
romagnetism. According to the phase diagram Figure 11.6b–e, the possible
ground state can be much complex, with spatial modulation of magnetic
orders, which certainly need experimental confirmation.
A theoretical study using a single interface between narrow bandwidth
manganites can reveal the true characteristics of a non-trivial interfacial
state [86]. In this particular study, the electrostatic potential was solved
self-consistently accompanying the charge transfer, using the Poisson equa-
tion (as opposed to a phenomenological potential increasing the difficulty
of the calculation). The magnetic ground state was obtained by optimizing
all classical spins and lattice distortions of the double exchange model. The
magnetism emergent from this highly non-trivial calculation revealed a
layer-dependent evolution, from the original A-type antiferromagnetism on
the RMnO3 side, through an exotic (canted) CE-type charge/orbital-ordered
antiferromagnetism in the vicinity of the interface, to a normal collinear
CE-type state, and finally to the G-type antiferromagnetism in the AMnO3
end side of the superstructure, as summarized in Figure 11.7. This calcu-
lation predicts a new state, the canted CE state, and moreover provides a
FIGURE 11.7 The case of a narrow bandwidth RMnO3/AMnO3 interface to illustrate a possible new source of
spin frustration. Z labels layers in the vicinity of an interface: Z £ 4 is nominally RMnO3, while Z ³ 5 is AMnO3.
(a) Optimized spin configurations near the interface. The dashed lines highlight the spin zigzag chains that
are characteristic of CE states with ferromagnetic order within each zigzag chain. (b) Layer dependence of the
spin structure factor for the optimized spin configurations. (c) Layer dependence of the real-space orbital pat-
tern. (After Yu, R. et al., Phys. Rev. B 80, 125115, 2009. With permission.)
concrete example of a novel type of frustration that emerges in superlat-

tices: for layers “sandwiched” between very stable states, such as the A-type
and CE-type states, the spin arranges in new patterns that are not stabilized
in manganites when in bulk form. This theoretically proposed novel form
of frustration deserves more work, and for concrete predictions to be con-
trasted against experiments.
11.5 EXCHANGE BIAS EFFECTS ACROSS

THE FERROMAGNETIC/G-TYPE
ANTIFERROMAGNETIC INTERFACE
Exchange bias is a widely observed unusual effect corresponding to ferro-
magnetic-antiferromagnetic interfaces [106]. In principle, the pinning effect
by the antiferromagnetic interface is expected to bias the hysteresis loop
of the attached ferromagnetic layer, as sketched in Figure 11.8a. However,
the origin of this effect is still under much discussion. For example, for a
fully compensated antiferromagnetic interface, such as the (001) surface
of a G-type antiferromagnetic perovskite (Figure 11.8b–c), it intuitively
appears that the antiferromagnetic moments are symmetrically distributed
with respect to their orientations, and thus they cannot bias the neighbor-
ing ferromagnetic moments. Several possible mechanisms have been pro-
posed to understand the exchange bias effect. Extrinsic factors are often
11.5 Exchange Bias Effects across the Ferromagnetic/G-Type Antiferromagnetic Interface 473
FIGURE 11.8 Sketches that illustrate new ideas proposed to understand exchange bias effects. (a)–(c)
Schematic of various ferromagnetic-antiferromagnetic configurations at an interface. Local magnetic
moments are denoted by the arrows. (a) corresponds to fully uncompensated antiferromagnetic interface,
as it occurs for example at the (001) surface of an A-type antiferromagnetic state or at the (111) surface of
a G-type antiferromagnetic state. (b) and (c) are fully compensated antiferromagnetic interfaces, as they
occur in a G-type antiferromagnetic state. In (a) and (b), the ferromagnetic and antiferromagnetic magnetic
moments are collinear. In (c), the magnetic moments are non-collinear, namely, the magnetic easy axes of
the ferromagnetic and antiferromagnetic materials are different. ((a)–(c) After Dong, S. et al., Phys. Rev. B 84,
224437, 2011. With permission.) (d) The (mutually perpendicular) relationship between the Mi-O-Mj
bond, oxygen displacement, and Di , j vector. (e) Schematic of the interface between ferromagnetic and
G-type antiferromagnetic perovskites, including the oxygen octahedral tilting. The staggered

directions of the Dij vectors at the interface are marked as in- and out-arrows, while the uniform AFM hD vectors

are also shown near the oxygens. (f) The resulting uniform hD should be perpendicular to S and D .
(g) Ferroelectric-polarization-driven asymmetric bond angles and modulated normal superexchange at the

interface. (h) A switch of the FE polarization will switch the biased field hJ . ((d)–(g) After Dong, S. et al., Phys.
Rev. Lett. 103, 127201, 2009. With permission.)
considered, such as interface roughness, spin canting near the interface, as

well as frozen interfacial and domain pinning, but there is no universally
accepted explanation for this exotic phenomenon [107–111]. Recent experi-
ments demonstrated the presence of exchange bias in BiFeO3/La0.7Sr0.3MnO3
heterostructures [61, 112], as well as in other BiFeO3/ferromagnetic alloys
[113]. More interestingly, it was observed that this exchange-bias can be
affected by the switching of the ferroelectric polarization of BiFeO3. This
new ingredient cannot be well understood by any of the traditional theories
of exchange bias.
As an alternative, Dong and collaborators proposed two other related
mechanisms to enlarge the existing possible theories of exchange bias [115].
These new mechanisms are based on the Dzyaloshinskii–Moriya interac-
tion, and on the ferroelectric polarization. The latter mechanism is only
active in those heterostructures that involve multiferroics (e.g. BiFeO3).
To understand these new ideas, consider that the spin-spin interaction in
perovskites can be described by a simplified effective Hamiltonian:

H= å[ J S × S + D × (S ´ S )], (11.1)
<ij >
ij i j ij i j
where J ij is the standard superexchange coupling between two nearest-

neighbor spins, while Dij is the Dzyaloshinskii–Moriya interaction vector,
which, at a more fundamental level, arises from the spin-orbit coupling.

In perovskites, Dij is determined by the bending of the Mi-O-Mj bond (M:
metal cation), which is induced by the oxygen octahedral rotations and tilt-
ing. Moreover, the vector is perpendicular to the Mi-O-Mj bond, as shown
in Figure 11.8d. Since the tilting and rotations are collective, the nearest-
neighbor oxygens in the same direction will move away from midpoint in
opposite directions, namely, the nearest-neighbor displacements are them-

selves staggered, as shown in Figure 11.8e. As a consequence, the Dij vectors
between nearest-neighbor bonds along the same direction are also stag-
gered. But note that the spins in a G-type antiferromagnetic state SAFM are

staggered as well. Combining these two staggered components, namely Dij
AFM
and S , will give rise to a uniform Dzyaloshinskii–Moriya effect at the
interface, as shown in Figure 11.8d–e, which can be described by an effec-
tive Hamiltonian:
FM FM
H Dinterface
M = åD × (S
<ij >
ij i ´ S jAFM ) = -hD × åS i
i , (11.2)

where hD = D ´ S jAFM can be regarded as the effective biasing magnetic field
which is fixed by the field-cooling process, and then assumed to be frozen at
low temperatures during the hysteresis loop measurement.
Furthermore, if one component has a ferroelectric polarization which
induces a uniform displacement between cations and anions, the bond
angles at the interface become no longer symmetric. Since the magnitude
of normal superexchange coupling depends on the bond angle, these modu-
lated bond angles induce staggered interfacial superexchange couplings,
which are denoted as J L and J S , as shown in Figure 11.8f. As in the previous
example, once again the staggered superexchange couplings at the interface
will also induce a uniform bias field when in the presence of the G-type anti-
ferromagnetic spin order, which can be described by:
FM
H Dinterface
M = åJ
<ij >
ij × (SiFM × S jAFM ) = -hJ × åS
i
i , (11.3)
(J - J )
where hJ = S L S jAFM arises from the modulation of the superexchange J,
2
which can be switched by polarization (Figure 11.8g).
These new ideas emphasize the interactions between the lattice dis-
tortion and magnetism, rather than continuing to rely on the existence of
uncompensated antiferromagnetic moments, which are often used in other
models, as previously explained. Using the first principles theory, Dong and
collaborators chose the SrRuO3/SrMnO3 system to verify these two pro-
posals for the exchange bias effect [114]. Needless to say, it is important to
continue exploring these and other new ideas to explain the exchange bias
effect, and to design experiments to confirm them.
11.6 Cobaltite Thin Films 475
11.6 COBALTITE THIN FILMS

In this section, we discuss a more complicated situation where both e g states
and t2 g states are electronically active, as realized in perovskite cobalbite
LaCoO3 (LCO) [116, 117]. While LCO has not been actively utilized for spin-
tronics applications, as discussed below, this could be useful for controllable
spin filter devices.
LCO is a unique perovskite compound in which two quantum many-
body states are close in energy; in bulk, a low spin (LS) state (t26 g ; S = 0) is
the ground state and a high spin (HS) state (t24g e 2g ; S = 2) is the first excited
state. The pure LS state is realized below T  100K and a mixed spin state
of HS and LS is realized at T  100K , due to thermal excitation [118]. This
indicates that the LS and HS states are in a delicate balance [119], and tun-
ing between these two states with a small energy cost would be possible, for
example, by applying strain. It is also important to note that the electrical
resistivity shows an insulating behavior at all temperature regime, while it
becomes small at T  400K, indicating the importance of correlation effects.
In Ref. [120], an alternative approach was taken to control the spin
state in LSO, i.e. the dimensionality, by fabricating superlattices of LCO
and LaAlO3, as schematically shown in Figure 11.9a. Since LaAlO3 is a large
bandgap insulator, SLs with thin LCO could be regarded as a two-dimen-
sional limit. As shown in Figure 11.9b, the orbital level scheme in Co3+ is
expected to be changed by reducing the dimensionality.
The spin-state transition in LCO thin films are experimentally sug-
gested from optical conductivity and x-ray absorption spectroscopy (XAS)
measurements. For the bulk LCO, the optical conductivity s(w) consists of
three features: strong absorption features at w  3eV and ~1.5 eV and a
weak feature at w  0.5 eV . By reducing the thickness of LCO in SLs, the
weak feature at w  0.5eV is systematically diminished and the strong fea-
ture at ~1.5 eV showed a blue shift. In XAS measurements, bulk LCO shows
two features: a strong absorption peak at w  529 eV , and a relatively weak
shoulder at w  528 eV . With reducing dimensionality, the weak shoulder
feature is systematically diminished, and the strong peak shows a blue shift.
These features are consistent with the dynamical mean-field theory
(DMFT) calculations [121], which support the spin-state transition in the
LCO/LAO superlattices. Numerical results for the orbital-resolved density
of states (DOS) for the 3D and 2D systems are presented in Figure 11.9c. The
energy distribution of the orbital states shows broad Co e g (narrow Co t2 g )
bands located between +1 and +4 eV (1 and +1 eV). O p bands are mostly
located below −2 eV. For 3D LCO, a portion of the e g orbital is occupied
near E = 0.8 eV and a portion of the t2 g orbital is unoccupied near E = 0.5 eV.
This configuration manifests a mixed spin state in 3D LCO at room tem-
perature. By contrast, for 2D LCO, the e g (3 z 2 - r 2 and x 2 - y 2 ) bands are
empty, and the t2 g (xy, yz and xz) bands are fully occupied, suggesting an
electron transfer from the e g orbitals to the t2 g orbitals by reducing the
dimensionality. Based these results, the characteristic features in s(w) for
3D LAO can be attributed to a O p–Co d charge transfer transition (labeled
(a) 3D Quasi 2D 2D
(b) 3D: HS+LS 2D: LS

3z2−r2
eg x2−y2
xz, yz
t2g xy
(c)
0.6 0.8
Co t2g 3D Co xz, yz 2D
Co xy
Co eg 0.6 2 2 O pπ
DOS (eV-1)
0.4 Co 3z -r
O pπ O pσ
γ 2 2
Co x -y
γ
O pσ 0.4
β
0.2
α 0.2
β
0.0 0.0
-8 -6 -4 -2 0 2 4 -8 -6 -4 -2 0 2 4
Energy (eV) Energy (eV)
FIGURE 11.9 Dimensional crossover of the spin state of cobaltites. Schematic geometry of LaCoO3/LaAlO3
superlattices, bulk cubic (3D) cobaltite, artificial superlattice of cobaltite having quasi 2D structure, and
monolayer of cobaltite which is purely 2D. Only B sites of perovskite structure ABO3 are shown; red circles are
Co sites and blue circles are Al. (b) Spin states. In 3D, a mixture of the high-spin (HS) state and the low-spin (LS)
state is realized, while in 2D the pure LS state is realized. (c) The DMFT results for the orbital-resolved spec-
tral function for 3D cobaltite and 2D cobaltite. (After Jeong, D.W. et al., Scientific Reports. 4, 6124, 2014. With
permission.)
γ with w  3eV ), a Co t2 g - e g transition (labeled β with w  1.5 eV ), and

a Co t2 g − t2 g transition (labeled α with w  0.5 eV ). For 2D LCO, because
the t2 g orbitals are fully filled, the weak absorption feature α seen for 3D
LCO at w  0.5 eV is absent. The optical gap is thus characterized by the
lowest excitations α mode for 3D and β mode for 2D. Two features in XAS,
w  528 eV and  529 eV, are ascribed to the electron excitations from
core states to unoccupied t2 g states at w  0.5 eV and to e g states with the
low-energy peak at w  1.5 eV . So, the absence of the low energy absorption
edge at w  528 eV for 2D LCO is naturally understood.
By reducing the dimensionality from 3D to 2D, the insulating nature is
changed from a Mott insulator to a band insulator with the gap amplitude
11.7 Theoretical Predictions for Interfaces Grown in the (111) Direction 477
increased from  0.5 eV to  1 eV . Since a Co3+ ion has finite spin in the
HS+LS mixed state for 3D LCO, it is expected to be easy to align the Co
moments by an external magnetic field or attaching a ferromagnet. Therefore,
when LCO is prepared at the critical regime between the mixed HS-LS state
and the LS state, and used as a tunneling barrier between magnetic metals,
this would work as a efficient and controllable spin filter; for the HS state,
the band gap for electrons with parallel spin arrangement to the Co moment
is smaller than those with antiparallel spin, allowing one of two spins to
transmit, and for the LS state, the band gap is large and independent of spin,
prohibiting the spin transport across the barrier.
11.7 THEORETICAL PREDICTIONS FOR

INTERFACES GROWN IN THE
(111) DIRECTION
In this section, we will discuss interesting novel phenomena, the topological
insulators, that could be realized in oxide heterostructures.
The study of novel electronic states driven by the non-trivial band topol-
ogy of electrons has been a major subject of interest in condensed matter
physics since the discovery of the integer quantum Hall effect (QHE) [122,
123]. After the proposal by Haldane [124] that the QHE could be realized in
an electronic system with a honeycomb lattice without Landau levels, Kane
and Mele proposed that graphene could show the spin Hall effect when the
spin-orbit coupling (SOC) strength is strong [125]. While the SOC in gra-
phene turned out to be too small to realize the spin Hall effect at non-zero
temperatures, the spin Hall effect or topological insulator (TI) state in HgTe
quantum wells was theoretically proposed [126] and experimentally verified
[127]. Further theoretical predictions [128, 129] and experimental realiza-
tions [130, 131] have established on a firm ground the area of TIs in materi-
als with strong SOC. For potential applications, and for exploring further
novel phenomena, inducing magnetism by doping magnetic ions or interfac-
ing with magnetic and/or superconducting systems has also been proposed
[61, 132–134] Once realized in real materials, these phenomena could lead
to entirely new device paradigms for spintronics and quantum computing.
However, until now most efforts on TIs have focused on narrow bandgap
semiconductors involving heavy elements such as Hg and Bi where the elec-
tronic properties are dominated by s and p orbitals.
Here, we will consider another class of materials—heterostructures of
transition-metal oxides (TMOs) involving d electrons. In terms of mag-
netism and superconductivity, TMOs are ideal playgrounds as they have
already demonstrated a wide variety of symmetry breaking effects arising
from strong correlations between electrons. However, only a few studies
addressing TIs based on the transition-metal oxides have appeared so far
[135]. As we have discussed in early sections, experimentally the quality of
oxide-based heterostructures is becoming extremely high. Recent highlights
of TMO heterostructures research include the realization of the integer
quantum Hall effect [136] and the fractional quantum Hall effect [137] in Zn
oxide heterostructures, as well as the integer quantum Hall effect in δ-doped
SrTiO3 [138]. We believe that our results may open new directions focusing
on topological phenomena in the rapidly growing field of oxide electronics.
We will focus on two-dimensional (2D) TIs realized using cubic
perovskite TMOs. Our design principle starts from the simple observation
that, when viewed from the [111] crystallographic axis, threefold symme-
try is realized in cubic perovskite structure (Figure 11.10a). Furthermore,
when we focus on a bilayer, it forms a buckled honeycomb lattice, which
resembles graphene (Figure 11.10b). Therefore, Dirac dispersions are natu-
rally expected. One could also adjust the Fermi level at those Dirac points
by choosing appropriate elements. Finally, when the SOC is turned on, the
system would become a 2D TI.
In order to find possible candidate systems, Ref. [139] started from a
tight-binding model for either t2 g or e g electrons and examined the z2 topo-
logical index [140] and edge spectra using finite-thick slabs. For t2 g systems,
this study revealed that t21 g , t22 g , t23 g and t25 g TMOs are possible candidates
to realize TIs when the SOC is much larger than the crystal field splitting
between the a1 g level and e′g level, and t22 g , t24g and t25 g are candidates when the
SOC is much weaker than the crystal field. From similar analyses for e g elec-
tron systems, e1g , e 2g and e 3g systems were found to be all possible candidates
for 2D TIs. One might find this rather strange, because the SOC is supposed
to be quenched in the e g multiplet. However, even in the e g multiplet, the
SOC is found to be active because of the symmetry lowering from octahe-
dral (Oh ) to trigonal (C3v ) in our (111) bilayers.
In practice, the (111) bilayer of perovskite TMO ABO3 has to be stabi-
lized by being sandwiched between insulating TMO A’B’O3. Based on the
tight-binding results, Ref. [139] examined the t24g system LaReO3, t25 g sys-
tems LaRuO3, LaOsO3, SrRhO3, SrIrO3, and e 2g systems LaCuO3, LaAgO3
and LaAuO3 using DFT methods. For A = La, LaAlO3 is used for an insulator
A’B’O3, and for A = Sr, SrTiO3 is used. In this research it was found that, in
the cases of LaOsO3, SrIrO3, LaAgO3 and LaAuO3, the Fermi level is located
inside the gap. Therefore, (111) bilayers of these materials would become two-
dimensional topological insulators. However, for LaReO3, SrRhO3, LaRuO3
and LaCuO3, the Fermi level crosses more than one band. Furthermore, in
LaCuO3, antiferromagnetic ordering is realized within DFT. As a conse-
quence, the LaReO3, SrRhO3, LaRuO3 (111) bilayers are classified as topologi-
cal metals rather than TIs, and in particular LaCuO3 is an antiferromagnetic
trivial metal. The dispersion relation of the SrIrO3 (111) and LaAuO3 (111)
bilayers are shown in Figure 11.10c–f, respectively. Figure 11.10c,e are for
bulk, and Figure 11.10d,f are for finite-thick zigzag slabs computed by using
Wannier parametrization. The gapless edge modes crossing the Fermi level
(E = 0 ) in the latter two panels indicate that these are indeed 2D TIs.
It is remarkable that the non-trivial gap amplitude of the LaAuO3 (111)
bilayer is about 150 meV. For this reason, this system should remain a 2D
TI even at room temperature. However, since the extension of the electron
wave functions shrinks when moving from 5d TMOs to 4d TMOs, and then
11.7 Theoretical Predictions for Interfaces Grown in the (111) Direction 479
(a) (b)
B2
B B
B B’ B1 B1
A A
B’ B
B2 B2
B B
B1
z z
x y x y
(c) (d)
0.5 0.5
0.0 0.0
-0.5 -0.5
E (eV)
E (eV)
-1.0 -1.0
-1.5 -1.5
-2.0 -2.0
Γ K M Γ 0.0 0.2 0.4 0.6 0.8 1.0

(e) (f) kzigzag [ π 3a ] 2
1.6 1.6
1.2 1.2
0.8 0.8
0.4 0.4
E (eV)
E (eV)
0.0 0.0
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
Γ K M Γ -0.4 -0.2 0.0 0.2 0.4
kzigzag [ π 3a ] 2
FIGURE 11.10 Crystal structure of perovskite (111) bilayer of ABO3 sandwiched by AB’O3, conventional view
(a). (b) (111) bilayer of ABO3 viewed from the [111] direction. Wannier band structure of SrIrO3 (111) bilayer for
bulk (c) and finite-thick zigzag slab (d) Wannier band structure of LaAuO3 (111) bilayer for bulk (e) and finite-
thick zigzag slab (f). (After Okamoto, S. et al., Phys. Rev. B 89, 195121, 2014. With permission.)
to 3d TMOs, some instability caused by stronger local Coulomb interactions

could appear. Reference [141] reexamined the stability of the 2D TI states
of SrIrO3 and LaAuO3 (111) bilayers using DFT+DMFT [121]. It was found
that the SrIrO3 (111) bilayer is indeed unstable against an antiferromagnetic
trivial insulating state, result consistent with the strong coupling approach
[142]. Recently, Hirai and coworkers successfully fabricated (111) bilayers of
perovskite iridate, Ca0.5Sr0.5IrO3, and found that it is strongly insulating with

magnetic ordering (the particular magnetic order pattern remains unclear)
[143]. This result appears to be consistent with the DFT+DMFT results. On
the other hand, the LaAuO3 (111) bilayer remains a 2D TI even after includ-
ing correlation effects, at least within the DFT+DMFT approximation. The
band gap is reduced somewhat, but remains larger than 100 meV. Therefore,
if synthesized properly, the LaAuO3 (111) bilayer is expected to become a
physical realization of the quantum spin Hall effect at room temperature.
The growth direction [111] described in this section opens a new subfield
of research within complex oxide interfaces, providing yet another interest-
ing variable to produce novel materials and states. In particular, topological
insulators could be realized by this procedure. While the theoretical basis
for these ideas appears to be on a firm ground, the next natural challenge
is to realize these ideas experimentally in real systems exploiting the [111]
crystalographic axis.
11.8 CONCLUSION
In this chapter, we have reviewed theoretical studies involving transition
metal oxide interfaces and heterostructures that might have potential rel-
evance for spintronic applications of complex oxide materials. Because of
the subtleties associated with strong correlation effects, including phase
competition among many spin, charge, orbital, and lattice ordered states, a
variety of approaches must be considered depending on the situation being
analyzed. In the many sections of this chapter, we have described (phenome-
nological) models for bulk and interfacial oxides, advanced numerical calcu-
lations using model Hamiltonians, and the implementation of some of them
together with DFT methodologies. We envision that, from the combination
of all these points of view, we could help to understand or predict novel phe-
nomena in TMO heterostructures. We are also convinced that TMOs have
a great potential for novel spintronic applications not only within magne-
tism, as intensively discussed here, but also in the context of non-trivial band
topology and correlations. It is clear that the field of oxide interfaces provides
a new and exciting playground to create new states that cannot be observed
in bulk compounds. We hope this article motivates further research to real-
ize efficient electronic devices based on TMOs.
ACKNOWLEDGMENTS
S.O. and E.D. were supported by the U.S. Department of Energy, Office of
Science, Basic Energy Sciences, Materials Sciences and Engineering Division.
S.D. was supported by the National Natural Science Foundation of China
(Grant No. 11274060).
REFERENCES
1. J. G. Bednorz and K. A. MÜ ller, Possible high Tc superconductivity in the
BaLaCuO system, Z. fur Phys. B Condens. Matter 64 , 189 (1986).
References 481
2. S. Jin, T. H. Tiefel, M. McCormack, R. A. Fastnacht, R. Ramesh, and L. H.

Chen, Thousandfold change in resistivity in magnetoresistive La-Ca-Mn-O
films, Science 264 , 413 (1994).
3. Y. Tokura, A. Urushibara, Y. Moritomo et al., Giant magnetotransport phe-
nomena in filling-controlled kondo lattice system: La1-x Srx MnO3, J. Phys.
Soc. Jpn . 63 , 3931 (1994).
4. M. Imada, A. Fujimori, and Y. Tokura, Metal-insulator transitions, Rev. Mod.
Phys . 70 , 1039 (1998).
5. Y. Tokura and N. Nagaosa, Orbital physics in transition-metal oxides, Science
288 , 462 (2000).
6. A. Ohtomo, D. A. Muller, J. L. Grazul, and H. Y. Hwang, Artificial charge-
modulation in atomic-scale perovskite titanate superlattices, Nature 419 , 378
(2000).
7. A. Ohtomo and H. Y. Hwang, A high-mobility electron gas at the LaAlO3/
SrTiO3 heterointerface, Nature 427 , 423 (2004).
8. M. Julliere, Tunneling between ferromagnetic films, Phys. Lett. 54A , 225 (1975).
9. S. Maekawa and U. Gafvert, Electron tunneling between ferromagnetic films,
IEEE Trans. Magn . 18 , 707 (1982).
10. W. H. Buttler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren, Spin-
dependent tunneling conductance of Fe/MgO/Fe sandwiches, Phys. Rev. B 63 ,
054416 (2001).
11. J. Mathon and A. Umerski, Theory of tunneling magnetoresistance of an epi-
taxial Fe/MgO/Fe (001) junction, Phys. Rev. B 63 , 220403 (2001).
12. M. Bowen, V. Cros, F. Petroff et al., Large magnetoresistance in Fe/MgO/
FeCo(001) epitaxial tunnel junctions on GaAs(001), Appl. Phys. Lett . 79 , 1655
(2001).
13. S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Giant room-
temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel
junctions, Nat. Mater. 3 , 868 (2004).
14. S. S. P. Parkin, C. Kaiser, A. Panchula et al., Giant tunnelling magnetoresis-
tance at room temperature with MgO (100) tunnel barriers, Nat. Mater. 3 , 862
(2004).
15. M. Viret, M. Drouet, J. Nassar, J. P. Contour, C. Fermon, and A. Fert, Low-field
colossal magnetoresistance in manganite tunnel spin valves, Europhys. Lett .
39 , 545 (1997).
16. M.-H. Jo, N. D. Mathur, N. K. Todd, and M. G. Blamire, Very large magnetore-
sistance and coherent switching in half-metallic manganite tunnel junctions,
Phys. Rev. B 61 , R14905(R) (2000).
17. M. Bowen, M. Bibes, A. Barthelemy, J.-P. Contour, Y. Lemaitre, and A. Fert,
Nearly total spin polarization in La2/3Sr1/3MnO3 from tunneling experiments,
Appl. Phys. Lett . 82 , 233 (2003).
18. Y. Ogimoto, M. Izumi, A. Sawa et al., Tunneling magnetoresistance above
room temperature in La0.7Sr0.3MnO3/SrTiO3/La0.7Sr0.3MnO3 junctions, Jpn. J.
Appl. Phys . 42 , L369 (2003).
19. H. Yamada, Y. Ogawa, Y. Ishii et al., Engineered interface of magnetic oxides,
Science 305 , 646 (2004).
20. Y. Ishii, H. Yamada, H. Sato et al., Improved tunneling magnetoresistance in
interface engineered (La,Sr)MnO3 junctions, Appl. Phys. Lett. 89 , 042509
(2006).
21. M. Bibes and A. Barthelemy, Oxide spintronics, IEEE Trans. Electron Devices
54 , 1003 (2007).
22. J.-H. Park, E. Vescovo, H.-J. Kim, C. Kwon, R. Ramesh, and T. Venkatesan,
Magnetic properties at surface boundary of a half-metallic ferromagnet
La0.7Sr0.3MnO3, Phys. Rev. Lett . 81 , 1953 (1998).
23. J. J. Kavich, M. P. Warusawithana, J. W. Freeland et al., Nanoscale suppression
of magnetization at atomically assembled manganite interfaces: XMCD and
XRMS measurements, Phys. Rev. Lett . 76 , 014410 (2007).
24. S. Yunoki, A. Moreo, E. Dagotto, S. Okamoto, S. S. Kancharla, and A. Fujimori,

Electron doping of cuprates via interfaces with manganites, Phys. Rev. B 76 ,
064532 (2007).
25. G. A. Smolenskii and I. E. Chupis, Ferroelectromagnets, Soviet Physics Uspekhi
25 , 475 (1982).
26. J. Wang, J. B. Neaton, H. Zheng et al., Epitaxial BiFeO3 multi-ferroic thin film
heterostructures, Science 299 , 1719 (2003).
27. A. Callaghan, C. W. Moeller, and R. Ward, Magnetic interactions in ternary
ruthenium oxides, Inorg. Chem. 5 , 1572 (1966).
28. M. P. de Jong, V. A. Dediu, C. Taliani, and W. R. Salaneck, Electronic structure
of La0.7Sr03MnO3 thin films for hybrid organic/inorganic spintronics applica-
tions, J Appl. Phys . 94 , 7292 (2003).
29. M. Minohara, I. Ohkuho, H. Kumigashira, and M. Oshima, Band diagrams
of spin tunneling junctions La0.6Sr0.4MnO3Nb:SrTiO3 and SrRuO3Nb:SrTiO3
determined by in situ photoemission spectroscopy, Appl. Phys. Lett. 90 , 2123
(2007).
30. C. Yoshida, A. Yoshida, and H. Tamura, Nanoscale conduction modulation in
AuPb(Zr,Ti)O3/SrRuO3 heterostructure, Appl. Phys. Lett. 75 , 1449 (1999).
31. A. Tsurumaki-Fukuchi, H. Yamada, and A. Sawa, Resistive switching artifi-
cially induced in a dielectric/ferroelectric composite diode, Appl. Phys. Lett.
103 , 152903 (2013).
32. M. D. Scafetta, Y. J. Xie, M. Torres, J. E. Spanier, and S. J. May, Optical
absorption in epitaxial La1-XSrxFeO3 thin films, Appl. Phys. Lett. 102 , 081904
(2013).
33. S. J. Clark and J. Robertson, Band gap and Schottky barrier heights of multifer-
roic BiFeO3, Appl. Phys. Lett. 90 , 132903 (2007).
34. J. F. Ihlefeld, N. J. Podraza, Z. K. Liu et al., Optical band gap of BiFeO3 grown
by molecular-beam epitaxy, Appl. Phys. Lett. 92 , 142908 (2008).
35. S. G. Lim, S. Kriventsov, T. N. Jackson et al., Dielectric functions and opti-
cal bandgaps of high-K dielectrics for metal-oxide-semiconductor field-effect
transistors by far ultraviolet spectroscopic ellipsometry, J. Appl. Phys . 91 , 4500
(2002).
36. P. W. Peacock and J. Robertson, Band offsets and Schottky barrier heights of
high dielectric constant oxides, J. Appl. Phys . 92 , 4712 (2002).
37. Y. Muraoka, T. Muramatsu, J. Yamaura, and Z. Hiroi, Photogenerated hole car-
rier injection to YBa2Cu3O7-x in an oxide heterostructure, Appl. Phys. Lett. 85 ,
2950 (2004).
38. T. Muramatsu, Y. Muraoka, and Z. Hiroi, Photocarrier injection and current
voltage characteristics of La0.8Sr0.2MnO3/SrTiO3:Nb heterojunction at low
temperature, Jpn. J. Appl. Phys . 44 , 7367 (2005).
39. M. Nakamura, A. Sawa, H. Sato, H. Akoh, M. Kawasaki, and Y. Tokura, Optical
probe of electrostatic-doping in an n-type Mott insulator, Phys. Rev. B 75 ,
155103 (2007).
40. T. Fujii, M. Kawasaki, A. Sawa, Y. Kawazoe, H. Akoh, and Y. Tokura, Electrical
properties and colossal electroresistance of heteroepitaxial SrRuO3/SrTi1-x
Nbx O3 (0.0002 ≤ x ≤ 0.02) Schottky junctions, Phys. Rev. B 75 , 165101 (2007).
41. H. Namatame, A. Fujimori, Y. Tokura et al., Resonant-photoemission study of
Nd2-x Cex CuO4, Phys. Rev. B 41 , 7205 (1990).
42. A. Fujimori, A. Ino, J. Matsuno, T. Yoshida, K. Tanaka, and T. Mizokawa, Core-
level photoemission measurements of the chemical potential shift as a probe
of correlated electron systems, J. Electron Spectrosc. Relat. Phenom . 124 , 127
(2002).
43. J. Matsuno, A. Fujimori, Y. Takeda, and M. Takanoi, Chemical potential shift in
La1-xSrxMnO3: Photoemission test of the phase separation scenario, Europhys.
Lett . 59 , 252 (2002).
44. S.-W. Cheong and M. Mostovoy, Multiferroics: A magnetic twist for ferro-
electricity, Nat. Mater. 6 , 13– 20 (2007).
References 483
45. S. Dong, J. M. Liu, S. W. Cheong, and Z. F. Ren, Multiferroic materials and

magnetoelectric physics: Symmetry, entanglement, excitation, and topology,
Adv. Phys . 64 , 519– 626 (2015).
46. N. A. Hill, Why are there so few magnetic ferroelectrics? J. Phys. Chem. B 104 ,
6694– 6709 (2000).
47. P. W. Anderson, Antiferromagnetism. Theory of superexchange interaction,
Phys. Rev . 79 , 350 (1950).
48. J. B. Goodenough, Magnetism and the Chemical Bond , Interscience, New York,
1963.
49. J. Kanamori, Superexchange interaction and symmetry properties of electron
orbitals, J. Phys. Chem. Solids 10 , 87 (1959).
50. K. I. Kugel and D. I. Khomskii, The Jahn– Teller effect and magnetism:
Transition metal compounds, Soviet Physics Uspekhi 25 , 231 (1982).
51. C. Zener, Interaction between the d-Shells in the transition metals. II.
Ferromagnetic compounds of manganese with perovskite structure, Phys. Rev .
82 , 403 (1951).
52. M. A. Ruderman and C. Kittel, Indirect exchange coupling of nuclear magnetic
moments by conduction electrons, Phys. Rev . 96 , 99 (1954).
53. T. Kasuya, A theory of metallic ferro- and antiferromagnetism on zener’ s
model, Prog. Theor. Phys . 16 , 45 (1956).
54. K. Yoshida, Magnetic properties of Cu-Mn alloys, Phys. Rev . 106 , 893 (1957).
55. I. Dzyaloshinskii, A thermodynamic theory of “ weak” ferromagnetism of anti-
ferromagnetics, J. Phys. Chem. Solids 4 , 241 (1958).
56. T. Moriya, Anisotropic superexchange interaction and weak ferromagnetism,
Phys. Rev . 120 , 91 (1960).
57. S. Okamoto, Magnetic interaction at an interface between manganite and
other transition-metal oxides, Phys. Rev. B 82 , 024427 (2010).
58. J. Chakhalian, J. W. Freeland, H.-U. Habermeier et al., Orbital reconstruction
and covalent bonding at an oxide interface, Science 318 , 1114 (2007).
59. N. M. Nemes, M. Garcia-Hernandez, S. G. E. te Velthuis et al., Origin of the
inverse spin-switch behavior in manganite/cuprate/manganite trilayers, Phys.
Rev. B 78 , 094515 (2008).
60. J. Salafranca and S. Okamoto, Unconventional proximity effect and inverse
spin-switch behavior in a model manganite-cuprate-manganite trilayer sys-
tem, Phys. Rev. Lett . 105 , 256804 (2010).
61. P. Yu, J.-S. Lee, S. Okamoto et al., Interface ferromagnetism and orbital recon-
struction in BiFeO3-La0.7Sr0.3MnO3 heterostructure, Phys. Rev. Lett . 105 ,
027201 (2010).
62. L. Jiang, W. S. Choi, H. Jeen et al., Tunneling electroresistance induced by
interfacial phase transitions in ultrathin oxide heterostructures, Nano Lett .
13 , 5837– 5843 (2013).
63. L. Jiang, W. S. Choi, H. Jeen, T. Egami, and H. N. Lee, Strongly coupled phase
transition in ferroelectric/correlated electron oxide heterostructures, Appl.
Phys. Lett. 101 , 042902 (2012).
64. C. H. Ahn, J. M. Triscone, N. Archibald et al., Ferroelectric field effect in epi-
taxial thin film oxide SrCuO2/Pb(Zr0.52Ti0.48)O3 heterostructures, Science 269 ,
373– 376 (1995).
65. J. D. Burton and E. Y. Tsymbal, Prediction of electrically induced magnetic
reconstruction at the manganite/ferroelectric interface, Phys. Rev. B 80 ,
174406 (2009).
66. J. D. Burton and E. Y. Tsymbal, Giant tunneling electroresistance effect driven
by an electrically controlled spin valve at a complex oxide interface, Phys. Rev.
Lett . 106 , 157203 (2011).
67. J. Hoffman, X. A. Pan, J. W. Reiner et al., Ferroelectric field effect transistors for
memory applications, Adv. Mater . 22 , 2957– 2961 (2010).
68. H. J. A. Molegraaf, J. Hoffman, C. A. F. Vaz et al., Magnetoelectric effects in
complex oxides with competing ground states, Adv. Mater . 21 , 3470 (2009).
69. C. A. F. Vaz, J. Hoffman, Y. Segal et al., Origin of the magnetoelectric coupling

effect in Pb(Zr0.2 Ti0.8)O3/La0.8Sr0.2MnO3 multiferroic heterostructures, Phys.
Rev. Lett . 104 , 127202 (2010).
70. C. H. Ahn, A. Bhattacharya, M. Di Ventra et al., Electrostatic modification of
novel materials, Rev. Mod. Phys . 78 , 1185– 1212 (2006).
71. C. A. F. Vaz, Electric field control of magnetism in multiferroic heterostruc-
tures, J. Phys. Condens. Matter 24 , 333201 (2012).
72. J. P. Velev, S. S. Jaswal, and E. Y. Tsymbal, Multi-ferroic and magnetoelectric
materials and interfaces, Philos. Trans. R. Soc. A-Math. Phys. Eng. Sci 369 ,
3069– 3097 (2011).
73. X. Huang and S. Dong, Ferroelectric control of magnetism and transport in
oxide heterostructures, Mod. Phys. Lett. B 28 (23), 1430010 (2014).
74. S. Dong and E. Dagotto, Full control of magnetism in a manganite bilayer by
ferroelectric polarization, Phys. Rev. B 88 , 140404(R) (2013).
75. S. Dong, X. T. Zhang, R. Yu, J. M. Liu, and E. Dagotto, Microscopic model for
the ferroelectric field effect in oxide heterostructures, Phys. Rev. B 84 , 155117
(2011).
76. E. Dagotto, Nanoscale Phase Separation and Colossal Magnetoresistance ,
Springer, Berlin, 2003.
77. E. Dagotto, T. Hotta, and A. Moreo, Colossal magnetoresistant materials: The
key role of phase separation, Phys. Rep. Rev. Sect. Phys. Lett . 344 , 1– 153 (2001).
78. Y. Tokura, Critical features of colossal magnetoresistive manganites, Rep. Prog.
Phys . 69 , 797– 851 (2006).
79. S. Smadici, B. B. Nelson-Cheeseman, A. Bhattacharya, and P. Abbamonte,
Interface ferromagnetism in a SrMnO3/LaMnO3 superlattice, Phys. Rev. B 86 ,
174427 (2012).
80. A. Bhattacharya, X. Zhai, M. Warusawithana, J. N. Eckstein, and S. D Bader,
Signatures of enhanced ordering temperatures in digital superlattices of
(LaMnO3)m/(SrMnO3)2m, Appl. Phys. Lett. 90 , 222503 (2007).
81. A. Bhattacharya, S. J. May, S. G. E. te Velthuis et al., Metal-insulator transition
and its relation to magnetic structure in (LaMnO3)2n/(SrMnO3)n superlattices,
Phys. Rev. Lett . 100 , 257203 (2008).
82. C. Adamo, C. A. Perroni, V. Cataudella, G. De Filippis, P. Orgiana, and L.
Maritato, Tuning the metal-insulator transitions of (SrMnO3)n/(LaMnO3)2n
superlattices: Role of interfaces, Phys. Rev. B 79 , 045125 (2009).
83. C. Adamo, X. Ke, P. Schiffer et al., Electrical and magnetic properties of
(SrMnO3)n/(LaMnO3)2n superlattices, Appl. Phys. Lett. 92 , 112508 (2008).
84. S. J. May, A. B. Shah, S. G. E. te Velthuis et al., Magnetically asymmetric inter-
faces in a LaMnO3/SrMnO3 superlattice due to structural asymmetries, Phys.
Rev. B 77 , 174409 (2008).
85. S. Dong, R. Yu, S. Yunoki, G. Alvarez, J. M. Liu, and E. Dagotto, Magnetism,
conductivity, and orbital order in (LaMnO3)2n/(SrMnO3)n super-lattices, Phys.
Rev. B 78 , 201102 (2008).
86. R. Yu, S. Yunoki, S. Dong, and E. Dagotto, Electronic and magnetic properties
of RMnO3/AMnO3 heterostructures, Phys. Rev. B 80 , 125115 (2009).
87. G. Bouzerar and O. Cepas, Effect of correlated disorder on the magnetism of
double exchange systems, Phys. Rev. B 76 , 020401 (2007).
88. B. R. D. Nanda and S. Satpathy, Effects of strain on orbital ordering and magne-
tism at perovskite oxide interfaces: LaMnO3/SrMnO3, Phys. Rev. B 78 , 054427
(2008).
89. B. R. K. Nanda and S. Satpathy, Effects of strain on orbital ordering and magne-
tism at perovskite oxide interfaces: LaMnO3/SrMnO3, Phys. Rev. B 78 , 054427
(2008).
90. B. R. K. Nanda and S. Satpathy, Polar catastrophe, electron leakage, and mag-
netic ordering at the LaMnO3/SrMnO3 interface, Phys. Rev. B 81 , 224408 (2010).
91. S. J. May, P. J. Ryan, J. L. Robertson et al., Enhanced ordering temperatures in
antiferromagnetic manganite superlattices, Nat. Mater. 8 , 892– 897 (2009).
References 485
92. Y. S. Hou, H. J. Xiang, and X. G. Gong, Intrinsic insulating ferromagnetism in

manganese oxide thin films, Phys. Rev. B 89 , 064416 (2014).
93. O. Chmaissem, B. Dabrowski, S. Kolesnik, J. Mais, J. D. Jorgensen, and S. Short,
Structural and magnetic phase diagrams of La1-xSrxMnO3 and Pr1-ySr yMnO3,
Phys. Rev. B 67 , 094431 (2003).
94. Z. Jirak, J. Hejtmanek, E. Pollert et al., Magnetic ground states in Pr1-x SrxMnO3
(x = 0.48 – 0.75), J. Appl. Phys. 89 , 7404 (2001).
95. H. Kuwahara, T. Okuda, Y. Tomioka, A. Asamitsu, and Y. Tokura, Two-
dimensional charge-transport and spin-valve effect in the layered anti-ferro-
magnet Nd0.45Sr0.55MnO3, Phys. Rev. Lett . 82 , 4316 (1999).
96. R. Kajimoto, H. Yoshizawa, H. Kawano et al., Hole-concentration-induced
transformation of the magnetic and orbital structures in Nd1-xSrxMnO3, Phys.
Rev. B 60 , 9506 (1999).
97. R. Maezono, S. Ishihara, and N. Nagaosa, Orbital polarization in manganese
oxides, Phys. Rev. B 57, R13993 (1998).
98. R. Maezono, S. Ishihara, and N. Nagaosa, Phase diagram of manganese oxides,
Phys. Rev. B 58 , 11583 (1998).
99. S. Okamoto, S. Ishihara, and S. Maekawa, Phase transition in perovskite man-
ganites with orbital degree of freedom, Phys. Rev. B 61 , 14647 (2000).
100. S. Dong, Q. F. Zhang, S. Yunoki, J. M. Liu, and E. Dagotto, Magnetic and orbital
order in (RMnO3)n/(AMnO3)2n superlattices studied via a double-exchange
model with strain, Phys. Rev. B 86 (20), 205121 (2012).
101. S. Okamoto, S. Ishihara, and S. Maekawa, Orbital structure and magnetic
ordering in layered manganites: Universal correlation and its mechanism,
Phys. Rev. B 63 , 104401 (2001).
102. S. Ishihara, S. Okamoto, and S. Maekawa, Roles of electron correlation and orbital
degree of freedom in manganese oxides, Trans. Mater. Res. Soc. Jpn . 26 , 963 (2001).
103. M. Kubota, H. Fujioka, K. Ohoyama et al., Neutron scattering studies on mag-
netic structure of the double-layered manganite La2-2x Sr1+2x Mn2O7 (0.30 ≤ x ≤
0.50), J. Phys. Chem. Solids 60 , 1161 (2000).
104. C. D. Ling, J. E. Millburn, J. F. Mitchell, D. N. Argyriou, J. Linton, and H. N.
Bordallo, Interplay of spin and orbital ordering in the layered colossal mag-
netoresistance manganite La2-2x Sr1+2x Mn2O7(0.5 ≤ x ≤ 1.0), Phys. Rev. B 62 ,
15096 (2000).
105. Q. F. Zhang, S. Dong, B. L. Wang, and S. Yunoki, Strain-engineered magnetic
order in (LaMnO3)n /(SrMnO3)2n superlattices, Phys. Rev. B 86 , 094403 (2012).
106. J. Nogues, J. Sort, V. Langlais et al., Exchange bias in nanostructures, Phys. Rep.
Rev. Sect. Phys. Lett . 422 , 65– 117 (2005).
107. A. P. Malozemoff, Random-field model of exchange-anisotropy at rough ferro-
magnetic-antiferromagnetic interfaces, Phys. Rev. B 35 , 3679– 3682 (1987).
108. N. C. Koon, Calculations of exchange bias in thin films with ferromagnetic/
antiferromagnetic interfaces, Phys. Rev. Lett . 78 , 4865– 4868 (1997).
109. M. Kiwi, J. Mejia-Lopez, R. D. Portugal, and R. Ramirez, Exchange-bias sys-
tems with compensated interfaces, Appl. Phys. Lett . 75 , 3995– 3997 (1999).
110. T. C. Schulthess and W. H. Butler, Consequences of spin-flop coupling in
exchange biased films, Phys. Rev. Lett . 81 , 4516– 4519 (1998).
111. M. Kiwi, Exchange bias theory, J. Magn. Magn. Mater. 234 , 584– 595 (2001).
112. S. M. Wu, S. A. Cybart, P. Yu et al., Reversible electric control of exchange bias
in a multiferroic field-effect device, Nat. Mater . 9 , 756– 761 (2010).
113. P. Borisov, A. Hochstrat, X. Chen, W. Kleemann, and C. Binek, Magneto-
electric switching of exchange bias, Phys. Rev. Lett . 94 , 117203 (2005).
114. S. Dong, Q. Zhang, S. Yunoki, J.-M. Liu, and E. Dagotto, Ab initio study of
the intrinsic exchange bias at the SrRuO3/SrMnO3 interface, Phys. Rev. B 84 ,
224437 (2011).
115. S. Dong, K. Yamauchi, S. Yunoki et al., Exchange bias driven by the
Dzyaloshinskii-Moriya interaction and ferroelectric polarization at G-type
antiferromagnetic perovskite interfaces, Phys. Rev. Lett . 103 , 127201 (2009).
116. M. Raccah and J. B. Goodenough, First-order localized-electron collective-

electron transition in LaCoO3, Phys. Rev . 155 , 932 (1967).
117. J. B. Goodenough, An interpretation of the magnetic properties of the
perovskite-type mixed crystals La1− x Srx CoO3− λ , J. Phys. Chem. Solids 6 , 287
(1958).
118. Y. Tokura, Y. Okimoto, S. Yamaguchi, H. Taniguchi, T. Kimura, and H. Takagi,
Thermally induced insulator-metal transition in LaCoO3: A view based on the
Mott transition, Phys. Rev. B 58 , R1699(R) (1998).
119. M. W. Haverkort, Z. Hu, J. C. Cezar et al., Spin state transition in LaCoO3 stud-
ied using soft X-ray absorption spectroscopy and magnetic circular dichroism,
Phys. Rev. Lett . 97 , 176405 (2006).
120. D. W. Jeong, W. S. Choi, S. Okamoto et al., Dimensionality control of d-orbital
occupation in oxide superlattices, Sci. Rep . 4 , 6124 (2014).
121. A. Georges, G. Kotliar, W. Krauth, and M. J. Rozenberg, Dynamical mean-field
theory of strongly correlated fermion systems and the limit of infinite dimen-
sions, Rev. Mod. Phys . 68 , 13 (1996).
122. R. E. Prange and S. M. Girvin (Eds.), The Quantum Hall Effect , Springer– Verlag,
New York, 1987.
123. D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs, Quantized
hall conductance in a two-dimensional periodic potential, Phys. Rev. Lett . 49 ,
405 (1982).
124. F. D. M. Haldane, Model for a quantum Hall effect without Landau levels:
Condensed-matter realization of the “ parity anomaly” , Phys. Rev. Lett . 61 ,
2015 (1988).
125. C. L. Kane and E. J. Mele, Z2 topological order and the quantum spin hall
effect, Phys. Rev. Lett . 95 , 146802 (2005).
126. B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Quantum spin Hall effect and
topological phase transition in HgTe quantum wells, Science 314 , 1757 (2006).
127. M. Konig, S. Wiedmann, C. BrÜ ne et al., Quantum spin Hall insulator state in
HgTe quantum wells, Science 318 , 766 (2007).
128. J. E. Moore and L. Balents, Topological invariants of time-reversal-invariant
band structures, Phys. Rev. B 75 , 121306 (2007).
129. L. Fu, C. L. Kane, and E. J. Mele, Topological insulators in three dimensions,
Phys. Rev. Lett . 98 , 106803 (2007).
130. D. Hsieh, D. Qian, L. Wray et al., A topological Dirac insulator in a quantum
spin Hall phase, Nature 452 , 970 (2008).
131. Y. Xia, D. Qian, D. Hsieh et al., Observation of a large-gap topological-insula-
tor class with a single Dirac cone on the surface, Nat. Phys . 5 , 398 (2009).
132. X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Topological field theory of time-rever-
sal invariant insulators, Phys. Rev. B 78 , 195424 (2008).
133. L. Fu and C. L. Kane, Superconducting proximity effect and Majorana fer-
mions at the surface of a topological insulator, Phys. Rev. Lett . 100 , 096407
(2008).
134. X.-L. Qi, R. Li, J. Zang, and S.-C. Zhang, Inducing a magnetic monopole with
topological surface states, Science 323 , 1184 (2009).
135. A. Shitade, H. Katsura, J. Kunes, X.-L. Qi, S.-C. Zhang, and N. Nagaosa,
Quantum spin Hall effect in a transition metal oxide Na2IrO3, Phys. Rev. Lett .
102 , 256403 (2009).
136. A. Tsukazaki, A. Ohtomo, T. Kita, Y. Ohno, H. Ohno, and M. Kawasaki,
Quantum Hall effect in polar oxide heterostructures, Science 315 , 1388 (2007).
137. A. Tsukazaki, S. Akasaka, K. Nakahara et al., Observation of the fractional
quantum Hall effect in an oxide, Nat. Mater. 9 , 889 (2010).
138. Y. Matsubara, K. S. Takahashi, M. S. Bahramy et al., Observation of the quan-
tum Hall effect in 5-doped SrTiO3, Nat. Commun . 7 , 11631 (2016).
139. D. Xiao, W. Zhu, Y. Ran, N. Nagaosa, and S. Okamoto, Interface engineering of
quantum Hall effects in digital transition-metal oxide heterostructures, Nat.
Commun. 2 , 596 (2011).
References 487
140. L. Fu and C. L. Kane, Time reversal polarization and a z2 adiabatic spin pump,
Phys. Rev. B 74 , 195312 (2006).
141. S. Okamoto, W. Zhu, Y. Nomura, R. Arita, D. Xiao, and N. Nagaosa, Correlation
effects in (111) bilayers of perovskite transition-metal oxides, Phys. Rev. B 89 ,
195121 (2014).
142. S. Okamoto, Doped Mott insulators in (111) bilayers of perovskite transition-
metal oxides with the strong spin-orbit coupling, Phys. Rev. Lett . 110 , 066403
(2013).
143. D. Hirai, J. Matsuno, and H. Takagi, Fabrication of (111)-oriented Ca0.5Sr0.5IrO3/
SrTiO3 superlattices A designed playground for honeycomb physics, APL
Mater. 3 , 041508 (2015).
12
LaAlO3/SrTiO3
A Tale of Two Magnetisms

12.1.1 Organization of This Chapter 492
12.2 Experimental Signatures of Magnetism 492
12.2.1 Magnetometry 492
12.2.1.1 SQUID Magnetometry 492
12.2.1.2 Cantilever Magnetometry 493
12.2.1.3 Scanning SQUID Microscopy 493
12.2.1.4 β-NMR 494
12.2.1.5 Magnetic Force Microscopy (MFM) 495
12.2.2 Transport 496
12.2.2.1 Anisotropic Magnetoresistance 496
12.2.2.2 Non-Linear and Anomalous Hall Effects 497
12.2.2.3 Rashba Spin-Orbit Coupling 498
12.2.2.4 Electron Pairing Without
Superconductivity 499
489
490 Chapter 12. LaAlO3/SrTiO3
12.2.3 Coexistence of Magnetism and

Superconductivity 501
12.2.3.1 Superconductivity at LaAlO3/SrTiO3
Interface 501
12.2.3.2 Magnetic effects 501
12.2.4 Null results, Inconsistencies and Artifacts 502
12.2.4.1 Artifacts 502
12.2.4.2 Null Results 502
12.2.4.3 Scanning SQUID 502
12.3 Theories of Magnetism at LaAlO3/SrTiO3 Interface 502
12.4 Main Thesis: Two Distinct Classes of Magnetic
Behavior in LaAlO3/SrTiO3 504
12.5 Ferromagnetic and Kondo Regimes 505
12.5.1 What are the Local Moments? 506
12.5.1.1 Magnetic Impurities, Oxygen Vacancies
(Vacancy Clusters), and F-Centers 507
12.6 Metamagnetism 508
12.6.1 Lifshitz Transition 508
12.6.2 Anisotropic Transport versus n, B 508
12.6.3 Anomalous Hall Effect 509
12.6.4 Correlated Transport Properties 509
12.7 Future Directions 510
12.7.1 Empirical Tests 510
12.7.2 Where Are the Oxygen Vacancies? 511
12.7.3 What Is the Origin of Electron Pairing? 512
Acknowledgments 512
References 513
12.1 INTRODUCTION
Complex oxides have been a continual source of discoveries in solid state
physics. Recent advances in thin-film growth techniques have enabled unit
cell-level control over their composition, resulting in the creation of lay-
ered, complex oxide heterostructures. Emergent phenomena—properties or
phases observed only within heterostructures and not found in the parent
compounds—provide a recurring theme for this active interdisciplinary field
that spans materials science, physics, chemistry, and engineering. One of the
most striking, and controversial, examples of an emergent property is the
observation of magnetism at the interface formed between the two com-

plex oxides LaAlO3 and SrTiO3. Both materials are separately nonmagnetic,
and yet there are many reports of magnetic phenomena associated with the
LaAlO3/SrTiO3 interface.*
The properties of SrTiO3 have fascinated researchers for generations.
This simple perovskite material has been widely utilized as a substrate for
the growth of other materials, but has its own wide-ranging properties that
span “all of solid state physics”—save for magnetism [1] (see Figure 12.1).
SrTiO3 has also inspired important discoveries such as high-temperature
superconductivity [2]. Interest in the fundamental properties of SrTiO3 was
renewed in the last decade due to the 2004 report by Ohtomo and Hwang of
high-mobility electron transport at the interface between LaAlO3 and TiO2-
terminated SrTiO3 [3]. The discovery of magnetism at the LaAlO3/SrTiO3
interface was first reported by Brinkman et al. [4] in 2007. Following that ini-
tial report, a cascade of new signatures of magnetism at the LaAlO3/SrTiO3
interface were obtained using a variety of measurement techniques.
While most published findings are internally consistent, some reports
apparently contradict the existence of interfacial magnetism, and some mag-
netic signatures have not always been reproduced under nominally identical
conditions. The goal of this chapter is to help organize and classify the some-
times contradictory observations that have been made on LaAlO3/SrTiO3
heterostructures, and closely related systems, which have been investigated
extensively, as well as explored theoretically. We argue that there is not a
FIGURE 12.1 Properties of LaAlO3/SrTiO3. Properties that have been observed

for SrTiO3 from superconductivity via doping, ferroelectricity, and piezoelectricity.
Magnetism, however, is found at the interface of the LaAlO3/SrTiO3, even though
neither of the two materials are magnetic. Quotes are from Ref. [1] and Ref. [101].
* It should be noted that there are also many complex-oxide heterostructures with magnetic
properties that are derived from bulk behavior. For a more general discussion of magnetism
at complex oxide interfaces, readers are referred to Chapter 11, Volume 2.
single form of magnetism, but rather that there are two main types of mag-
netic behavior with distinct properties and origins.
12.1.1
Organization of This Chapter
The purpose of this chapter is twofold. First, we will review the experimen-
tal evidence for magnetic phenomena at the LaAlO3/SrTiO3 interface. We
will then argue that essentially all of the signatures of magnetism can be
sorted into two distinct categories: (1) magnetic phases (e.g. ferromagnetic
or Kondo) involving local magnetic moments and their coupling to itinerant
electrons; (2) metamagnetic effects that are mediated by attractive electron-
electron interactions that do not involve local moments. We will review pos-
sible candidates for the local moments that give rise to the ferromagnetic
phases and focus on arguments for one potential source: oxygen vacancies.
For the metamagnetic transport signatures, band-structure effects (e.g.
Lifshitz transition) and strong attractive electron-electron interaction can
help consolidate disparate experimental findings. The coexistence of mag-
netism and superconductivity is discussed briefly.
While the main focus will be on experimental evidence, we will also
briefly summarize various theoretical approaches that have been taken so
far. However, we argue that the starting point (assumptions) for many theo-
ries could be hindered by attempts to combine these two distinct (and weakly
interacting) forms of magnetism. To make better progress theoretically, we
argue that it will be simpler to initially restrict the domain of inquiry to one
or the other phases. Toward the end of the article, we discuss open questions
as well as some tests for this framework for understanding the two classes of
magnetic phenomena.
12.2 EXPERIMENTAL SIGNATURES OF MAGNETISM

The first signature of magnetism at the LaAlO3/SrTiO3 interface was reported
in 2007 by Brinkman et al. [4]. They reported a hysteresis loop in the magne-
toresistance (Figure 12.2a) and a characteristic Kondo minimum (followed
by a saturation at lower temperature) in the resistance versus temperature
that was attributed to magnetic impurities (Figure 12.3a). The Kondo signa-
ture was later also observed by Hu et al. [5], as well as by Lee et al. [6] in
electrolyte-gated SrTiO3. Several other signatures of magnetism followed,
including observations of anisotropic magnetoresistance [5, 7–13], anoma-
lous Hall effect [11, 14–16], and direct observation of magnetization with tools
from cantilever magnetometry [17], scanning SQUID magnetometry and sus-
ceptometry [18–20], magnetic force microscopy [21, 22], and x-ray magnetic
circular dichroism [23, 24]. Below, we consider these findings in more detail.
12.2.1 Magnetometry
12.2.1.1 SQUID Magnetometry
In 2011, Ariando et al. [9] reported a direct measurement of LaAlO3/SrTiO3
magnetization using SQUID magnetometry. As shown in Figure 12.3b, a
FIGURE 12.2 Inconsistencies and null results. Even though the signatures of magnetism are reported by
many, there have also been controversies. (a) A hysteresis loop in the magnetoresistance previously thought
as purely magnetic, but later identified as magnetocaloric effect [55]. (After Brinkman, A. et al., Nat. Mater.
6, 493, 2007. With permission.) (b) A null result reported by Wijnands et al. [57]: scanning SQUID on samples
nominally equally to those reported in Bert et al. [18]. No ferromagnetic patches are found on the area of
the sample explored. (After Wijnands, T. Scanning Superconducting Quantum Interference Device Microscopy,
Master’s thesis, 2013. With permission.) (c) Using neutron spin-dependent reflectivity, where the magneti-
zation is computed from asymmetry in the spin-dependent superlattice Bragg reflections as a function of
wave vector transfer and neutron beam polarization, Fitzsimmons et al. [56] reported magnetization for
LaAlO3/SrTiO3 superlattice close to the noise level through the presented six samples grown by two indepen-
dent groups. (After Fitzsimmons, M.R. et al., Phys. Rev. Lett. 107, 217201, 2011. With permission.)
series of hysteresis loops of the magnetization in external magnetic field is

observed. This hysteresis persists up to room temperature. It is worth noting
that this ferromagnetic hysteresis was only reported for the samples grown
at P(O2) = 10−2 mbar pressures, which exist at the boundary of the metal-
insulator transition. A similar finding for Nb-doped SrTiO3 has also been
reported by Liu et al. [25].
12.2.1.2 Cantilever Magnetometry
Using cantilever-based torque magnetometry (Figure 12.3c), Li et al. [17]
reported an in-plane magnetization of the 2DEG at LaAlO3/SrTiO3 with
magnitude corresponding to ~0.3 μB per interface unit cell (if all the mag-
netization is assigned to the interface). The magnetization was found to be
independent of temperature up to 40 K and persists beyond 200 K. The same
sample was also found to be superconducting below 120 mK.
12.2.1.3 Scanning SQUID Microscopy

Scanning SQUID microscopy has been used to image inhomogeneous
magnetism in LaAlO3/SrTiO3. With a 3 μm-diameter SQUID loop serv-
ing as a local sensor of magnetic flux, this technique can detect magnetic
moments as small as ~ 102 m B / Hz [26]. Bert et al. [18] reported dipole-
like magnetic patches at the interface of LaAlO3/SrTiO3 (Figure 12.3d) [18].
Subsequent investigations by Kalisky et al. [19] determined that the patches
were highly nonuniform in size, orientation, and physical placement, and
only observed at or above a critical thickness for LaAlO3 layer (3 u.c.), the
same as the threshold for the insulator-to-metal transition in LaAlO3/
FIGURE 12.3 Experimental evidence for ferromagnetism at the LaAlO3/SrTiO3 interface. (a) Kondo resis-
tance minimum. (After Brinkman, A. et al., Nat. Mater. 6, 493, 2007. With permission.) (b) SQUID measurement:
hysteresis loops taken at different temperature superimposed on top of the temperature dependence of
magnetic moments. (After Ariando, X. et al., Nat. Commun. 2, 188, 2011. With permission.) (c) Cantilever-based
magnetometry: the magnetic moment induced a torque under external magnetic field; the magnetization
of the sample is inferred accordingly. (After Li, L. et al., Nat. Phys. 7, 762–766, 2011. With permission.) (d) With a
micrometer-sized SQUID on a probe tip, microscopic magnetization can be imaged. Dipole-shaped patches
are observed. (After Bert, J.A. et al., Nat. Phys. 7, 767–771, 2011. With permission.) (e) Electrically-controlled
ferromagnetism observed with magnetic force microscope. The magnetism signal is observed only when the
interface is insulating. (After Bi, F. et al., Nat. Commun. 5, 5019, 2014. With permission.)
SrTiO3 [27]. By touching the LaAlO3 surface with the SQUID sensor, it was
discovered that the magnetic moment and orientation of the ferromag-
netic patches could be manipulated (Figure 12.4f) [20]. The same scanning
SQUID system can also simultaneously image diamagnetic susceptibility
associated with superconductivity [18, 28]. Correlations between magnetic
and superconducting order were not found within the spatial resolution of
this technique.
12.2.1.4 β-NMR
Salman et al. [29] measured magnetic properties of LaAlO3/SrTiO3 using β-
NMR. In this technique, spin-polarized radioactive 8Li atoms are shot at the
LaAlO3/SrTiO3 sample. The spins of Li nuclei are inferred from the spins
of electrons emitted via β decay. The presence of magnetic moments in the
sample increases the spin decoherence rate. A “weak” (~ 1.8 ´ 10 -3 m B / u.c.)
magnetization is obtained for both 6 and 8 u.c samples. If a high degree of
spatial nonuniformity is assumed, then a local density ~ 1012 m B / cm 2 may be
present [29], which is on the same order of magnitude as Bert et al. [18].
FIGURE 12.4 Ferromagnetism, oxygen vacancies, and ferroelastic domains. (a) Using magnetic circular
dichroism (MCD), Rice et al. [71] reported a persistent magnetic signal for oxygen deficient SrTiO3 samples.
The signal disappears upon further re-oxygenation and reappears after introduction of oxygen vacancies
. (After Rice, W.D. et al., Nat. Mater. 13, 481, 2014. With permission.) (b) The persistent MCD signal created
in real space. (After Rice, W.D. et al., Nat. Mater. 13, 481, 2014. With permission.) (c) With DFT calculation
for ferroelectric perovskite BaTiO3, Goncalves-Ferreira et al. [95] found oxygen vacancies do have lower
energy when around twin wall of the ferroelastic domains. (After Goncalves-Ferreira, L. et al., Phys. Rev.
B. 81, 024109, 2010. With permission.) (d) and (e) Experimentally observed ferroelastic domains using (d)
a Scanning SET. (After Honig, M. et al., Nat. Mater. 12, 1112, 2013. With permission.) (e) scanning SQUID.
(After Kalisky, B. et al., Nat. Mater. 12, 1091, 2013. With permission.) (f) Mechanical manipulation of the
magnetization using the probe of scanning SQUID. (After Kalisky, F.B. et al., Nano Lett., 2012, 12, 2012. With
permission.)
12.2.1.5 Magnetic Force Microscopy (MFM)

Bi et al. [21] reported electronically-controlled magnetism at the interface
of LaAlO3/SrTiO3 using magnetic force microscopy (Figure 12.3e). When
an in-plane polarized magnetic tip in an atomic force microscope (AFM)
is brought to the vicinity of the sample surface, the magnetization of the
sample and its interaction with the MFM tip shifts the oscillation ampli-
tude, phase, and frequency, which are recorded as a function of position.
Magnetic interactions between the MFM tip and the interface take place
through a thin layer of gold deposited on the surface of LaAlO3 to serve
as a top gate for the sample. The 2DEG at the interface can be tuned from
insulating to conducting, which is monitored through the capacitance of
the two-terminal device. Magnetic interactions are observed only when the
tip is magnetized in-plane, and when the device is gated into the insulating
regime. Related measurements show a pronounced magnetoelectric effect,
which indicates that free carriers rapidly screen (on <10 μs time scales) fer-
romagnetic domains [21]. Follow-up measurements revealed that this elec-
tronically-controlled ferromagnetism is observed only within a thickness
window 8–25 u.c. [22]. Outside this region, the samples cannot be gated into
the insulating regime: samples with thinner LaAlO3 layers are prone to leak-
age via direct tunneling, while samples with LaAlO3 layer thicker than 25
u.c. are subjected to Zener tunneling.
12.2.2 Transport
12.2.2.1 Anisotropic Magnetoresistance
Anisotropy in the magnetotransport measurements is a widely reported sig-
nature of magnetism at the LaAlO3/SrTiO3 interface [5, 7–13]. When the
magnetic field is (i) oriented out-of-plane, the magnetoresistance is posi-
tive [30]. When the magnetic field is (ii) in-plane and perpendicular to the
direction of transport, the magnetoresistance is positive for small fields and
eventually becomes negative at larger field [31]. When the magnetic field is
(iii) parallel to the direction of transport, the magnetoresistance is negative
[7, 9]. As the applied magnetic field rotates in plane, as shown in Figure 12.5a,
a sinusoidal dependence of magnetoresistance as a function of the angle φ
between the field and the direction of current (Figure 12.5b). However, when
the carrier concentration ne is above a critical value nL , a Lifshitz transition
FIGURE 12.5 Transport metamagnetism: anisotropic magnetoresistance. (a) Setup for measuring in-plane
anisotropy of the magnetoresistance and the metamagnetism transition critical field Bp. The anisotropy
appears only when carrier concentration ne is greater than nL and the external magnetic field is greater than
Bp (which is also a function of ne). A characteristic change in the magnetoresistance can be clearly seen from
(b) to (c), when the carrier concentration is going above nL, or (d) to (e), when the external field is exceeding Bp.
((b) and (c) after Joshua, A. et al., Proc. Natl. Acad. Sci. 110, 9633, 2013. With permission; (d) and (e) after Annadi,
A. et al., Phys. Rev. B 87, 201102, 2013. With permission.)
occurs and this two-fold symmetry develops crystalline anisotropy with

high harmonics (Figure 12.5c,e).
12.2.2.2 Non-Linear and Anomalous Hall Effects

Nonlinearity in the Hall coefficient is widely reported for SrTiO3-based het-
erostructures [8, 11, 14–16, 30, 32–38]. Figure 12.6a [36] shows the most com-
mon nonlinearity in the Hall resistance Rxy , using a standard Hall geometry
with an out-of-plane external magnetic field. This results in an overall para-
bolic shape of the Hall coefficient RH (the slope of the Hall resistance Rxy )
in the non-linear regime, as shown in Figure 12.6b [16]. This feature is com-
monly attributed to the multiband nature of SrTiO3 [11, 30, 38].
The attribution of non-linear Hall effect to the existence of multiple
bands cannot explain the upturn of RH at smaller fields. Gunkel et al. [16]
postulated an additional anomalous Hall effect (AHE) with Langevin-type
dependence on the external magnetic field [16]. While the physical origin for
this AHE was not identified, it was considered to be a manifestation of some
magnetizable component in the system, e.g. due to the release and alignment
of the spin degree of freedom.
Using a different configuration, namely a large in-plane magnetic field
with small out-of-plane component, as shown in the inset of Figure 12.6c,
Joshua et al. [15] reported a relatively sharp onset of an anomalous Hall effect
a critical in-plane field H c|| that depends strongly on carrier concentration
ne . This critical field diverges near the Lifshitz density nL : H c|| ~ ( ne - nL ) ,
-1
increasing from 2 T to values as large as 14 T. This offset in Hall resistance

was interpreted as some type of magnetization that has been “released”
above a density-dependent critical magnetic field.
FIGURE 12.6 Transport metamagnetism: non-linear and anomalous Hall effect at the LaAlO3/SrTiO3 inter-
face. (a) Commonly reported nonlinearity in the Hall resistance: with standard Hall geometry and an out-of-
plane external magnetic field, the magnitude of the Hall coefficient RH (the slope of the Hall resistance R xy)
decreases with the increased field and settles at some value smaller than the zero field RH. (After Ben Shalom,
M. et al., Phys. Rev. Lett. 104, 126802, 2010. Copyright the American Physical Society. With permission.) (b)
Hall coefficient RH as function of applied field. While the overall parabolic-like shape can be explained by the
multiband nature of SrTiO3, the small upturn around zero field cannot (adapted from [16]). (c) When applying
a large in-plane field with a small out-of-plane field, the dependence of the Hall component on the Hall resis-
tivity can be manipulated with the large in-plane field. There is a sudden release of the magnetic moments
whenever the field exceeds a critical value. (After Joshue, A. et al., Proc. Natl. Acad. Sci. 110, 9633, 2013. With
permission.)
12.2.2.3 Rashba Spin-Orbit Coupling

Strong gate-tunable Rashba spin-orbit coupling was first reported by
Caviglia et al. [39] and Ben Shalom et al. [7]. Due to the polar nature of the
LaAlO3 layer, there is a built-in electric field in the direction perpendicular
to the interface (along the c-axis for [001] the SrTiO3). This field breaks inver-
sion symmetry and introduces Rashba spin-orbit coupling (Figure 12.7) that
mixes the dxy band and dxz / d yz bands and spin-splitting. The spin-splitting is
at its largest (several meV) near the avoided crossings of the bands.
Rashba spin-orbit coupling was probed via magnetotransport. As the
gating voltage increases, Caviglia et al. found that the magnetoconductance
changes from positive (associated with weak localization and negligible
Rashba spin-orbit coupling) to negative (associated with weak antilocaliza-
tion and strong Rashba spin-orbit coupling). By fitting the magnetoconduc-
tance to the Maekawa–Fukuyama theory of spin-relaxation, the coupling
FIGURE 12.7 Band structure of Ti 3d t2g orbitals. The 3d orbitals of Ti split into a less energetic triplet t2g (dxy,
( )
dyz, dxz) and a more energetic doublet eg d3 z 2 -r 2 , d x 2 - y 2 . dxy has the lowest energy of the three t2g orbitals, and
is populated first. As the carrier concentration increases, d yz / d xz start to be populated at Lifshitz point (nL).
strength was found to be strongly dependent on gate, with a sharp rise near
the gating voltage corresponding to insulator to superconductor transition
of the same sample.
Ben Shalom et al. [36], on the other hand, reported a field-independent
longitudinal resistance for in-plane magnetic fields greater than the super-
||
conducting upper critical field H c2 but below a larger critical field H * , fol-
lowed by a significant drop of the resistance above H * . Ben Shalom et al.
[36] associated H * with a spin-orbit field, and obtain a dependence of Rashba
spin-orbit coupling on gate voltage opposite to that reported by Caviglia
et al. [39].
A current-driven Rashba field was reported by Narayanapillai et al. [40].
Using strong Rashba spin-orbit coupling at LaAlO3/SrTiO3 for spin-charge
conversion has been demonstrated recently by Lesne et al. [41]. The reported
spin-charge conversion efficiency is comparable to non-2D materials such as
Pt, Ta, W [41]. Additionally, a sign change of the Rashba coefficient is also
reported [41].
Attempts to resolve the spin-dependent band structure using spin-
polarized ARPES (SARPES) have been made, with conflicting reports. A
large spin-splitting of 90 meV was reported by Santander-Syro et al. on
doped SrTiO3 [42], followed by a null result reported by Walker et al. [43],
and a possible reconciliation of the two contradictory reports [44].
12.2.2.4 Electron Pairing Without Superconductivity

The superconducting state is associated with a variety of non-trivial mag-
netic properties. A conventional (type I) superconductor is a perfect diamag-
net and the superconducting state consists of a condensate of Cooper pairs,
which (in most cases) have a spin-singlet configuration. As will be described
below, LaAlO3/SrTiO3 exhibits a robust phase in which electrons remain
paired, but are not superconducting. The pairing transition itself involves
significant changes in magnetization and orbital character that are associ-
ated with the complex metamagnetic behavior observed in this system.
Superconductivity in SrTiO3 was first reported in 1964 by Schooley et al
[45]. Apart from being one of a few semiconductors to exhibit superconduct-
ing behavior, it was suspected that its nature might be different from con-
ventional BCS superconductivity. For example, the superconducting “dome”
observed for SrTiO3 [46] bears a striking resemblance to high-temperature
superconductors, which were discovered two decades later [2]. One myste-
rious and controversial phase of high-temperature superconductors is the
so-called “pseudogap” regime in which superconductivity is not observed,
but a gap in the single-particle spectrum remains, as observed through tun-
neling experiments and other measurements [47]. While it is far from clear
that SrTiO3 shares a similar pairing mechanism—it lacks, for example, an
antiferromagnetic parent phase that is believed to contribute to pairing
in high-Tc compounds—planar tunneling experiments on LaAlO3/SrTiO3
structures by Richter et al. [48] show a pseudogap feature past the boundary
of the superconducting dome. The physical origin of the pseudogap phase
in high-Tc compounds continues to be debated. In particular, a seemingly
straightforward question—whether the pseudogap phase is related to pre-

formed Cooper pairs—has been challenging to resolve experimentally in the
cuprates.
Cheng et al. [49] investigated the phenomenon of electron pairing
using a single-electron transistor (SET) geometry (Figure 12.8a). A SET is
a three-terminal device with a quantum dot (QD) that is tunnel-coupled to
source and drain leads, with a gate that can change the number of carriers
in the QD [50]. The conductance of a conventional SET is generally low if
the energy of the lowest available state in the QD is not aligned with the
chemical potential of the leads. When there is such an alignment, the energy
of the state with N and N+1 electrons in the QD becomes degenerate, and
electrons can tunnel resonantly through the device. Cheng et al. created a
SET device using conductive-AFM (c-AFM) at the LaAlO3/SrTiO3 inter-
face (Figure 12.8a). They found that the differential conductance of the SET
(Figure 12.8b) departed significantly from what one would expect for a SET
formed from ordinary semiconductors. Three distinct regimes of behavior
were observed as a function of applied magnetic field (out of plane): (i) at
very low magnetic fields (below B < m 0 H c 2 ~ 0.1 T ), a Josephson supercur-
rent (marked SC in Figure 12.8b) is found to flow through the device; (ii)
at intermediate fields (m 0 H c 2 < B < BP ~ 2 - 11 T), a series of vertical (mag-
netic field-independent) lines are observed; (iii) B > BP , the vertical lines
bifurcate, leading to a doubling of the number of conductance peaks which
Zeeman shift as the magnetic field is further increased. Significantly, while
Bp varies from device to device, it generally is found to increase as the carrier
FIGURE 12.8 Electron pairing without superconductivity. Using a c-AFM lithography-defined single elec-
tron transistor, Cheng et al. [49] reported a phase of LaAlO3/SrTiO3 in which the electrons are paired, but not
condensed into the superconducting phase. (a) Schematics for the c-AFM lithography process and the single
electron transistor device, consisting of a quantum dot (QD) and a sidegate . (After Cheng, G. et al., Nature 521,
196, 2015. With permission.) (b) Evolution of the pairing transition as a function of external out-of-plane mag-
netic field and sidegate voltage. (After Cheng, G. et al., Nature 521, 196, 2015. With permission.)
FIGURE 12.9 Metamagnetism and pairing transition. (a) Pairing field (Bp) as
a function of voltage on sidegate (Vsg) (replotted from Ref [49]). (b) Density-
dependent critical field HC|| that marks the onset of AMR, AHE, and giant negative
magnetoresistance. (After Joshue, A. et al., Proc. Natl. Acad. Sci. 110, 9633, 2013.
With permission.)
density decreases (Figure 12.9a). Furthermore, this value is more than an

order of magnitude greater than the upper critical field for the supercon-
ductivity, µ 0 H c 2 (Figure 12.9a), and for temperatures that greatly exceed the
highest observed superconducting state in SrTiO3. That is to say, it is a state
in which there is electron pairing without superconductivity.
12.2.3 Coexistence of Magnetism and Superconductivity

12.2.3.1 Superconductivity at LaAlO3/SrTiO3 Interface
The 2DEG exhibits superconductivity [51, 52], at low carrier concentrations
(ne ~ 1013 cm -2) and critical temperatures in the range T » 200 - 300mK [51].
Like most other characteristics of this interface, the superconducting prop-
erties can be tuned by a backgate [52]. By tracing out the critical temperature
as a function of gating voltage (and therefore carrier concentration), a dome-
shaped outline enclosing the superconducting phase is seen, similar to high-
Tc superconductors as well as the doped bulk SrTiO3.
12.2.3.2 Magnetic effects
While a true spatial coexistence of ferromagnetism and superconductivity at
LaAlO3/SrTiO3 is not confirmed, coexistence at the sample level is reported
by Li et al. [17] and Dikin et al. [53] and within the resolution of the scanning
SQUID (~3 µm) by Bert et al. [18]. Dikin et al. [53] observed a clear hys-
teresis in magnetotransport measurements as a function of external mag-
netic field by mapping the critical temperature as a function of out-of-plane
magnetic field while controlling the temperature of the sample to maintain
a constant resistance near the superconducting transition. If, alternatively,
the temperature is kept constant, hysteresis is seen in both the sheet resis-
tance Rs and the Hall resistance Rxy . The interplay of magnetic effects and
superconducting state was later also probed by Ron et al. [54] in shadow-
masked one dimensional nanowire with width ~50 nm, comparable to the
superconducting coherence length. Ron reported a critical field H s (8mT to
12.5 mT) at the onset of a drop in the in-plane (both parallel and perpendic-
ular to the current) magnetoresistance. Hysteresis loops in the magnetore-
sistance are reported for all directions with magnetization field close to H s .
12.2.4 Null results, Inconsistencies and Artifacts

12.2.4.1 Artifacts
The first report of magnetism at the LaAlO3/SrTiO3 interface [4], in addi-
tion to reporting a Kondo resistance minimum (Figure 12.9a), also shows
a hysteretic magnetoresistance curve that depends on the direction of the
sweeping of the external field (Figure 12.2a), as well as the sweep rate. The
hysteresis is stronger at lower temperatures, or at higher sweep rate. After
subsequent analyses, the authors determined that the hysteresis loops in
those magnetoresistance curves can be ascribed to a magnetocaloric effect
within the ceramic chip carrier used to hold the sample [55]. On the other
hand, the other magnetic signature reported in [4], the Kondo resistance
minimum, is not susceptible to this magnetocaloric effect.
12.2.4.2 Null Results
While the 2DEG of LaAlO3/SrTiO3 is teeming with evidence of magnetism,
there have been reports that are seemingly at odds, either quantitatively or
qualitatively, with some of the primary reports. Using polarized neutron
reflectometry, Fitzsimmons et al. [56] found essentially no evidence for
magnetism in a variety of LaAlO3/SrTiO3 samples grown by two indepen-
dent groups. The magnetization, which is calculated from the asymmetry
in specular reflectivity as a function of wave-vector transfer and neutron
beam polarization (Figure 12.2c), is close to the signal-to-noise limit of the
measurement.
12.2.4.3 Scanning SQUID
Follow-up scanning SQUID observations at the LaAlO3/SrTiO3 interface
[18–20, 28] by Wijnands [57] failed to reproduce observed magnetic patches
(Figure 12.2b). The experiments were performed using a scanning SQUID
microscope of similar design and sensitivity, with samples prepared under
nominally identical conditions, as Bert et al. [18].
12.3 THEORIES OF MAGNETISM AT
LaAlO3/SrTiO3 INTERFACE
Theories of magnetism at the LaAlO3/SrTiO3 interface usually begin with
a theory of the ferromagnetic phase. This phase was first attributed, by
Pentcheva and Pickett [58], to the presence of oxygen vacancies at the inter-
face that localize electrons in nearby Ti 3d states. Similar oxygen vacancies
at the surface of SrTiO3 may also introduce ferromagnetic order [59]. These
localized states are located either in the Ti d xy orbitals [60, 61], which lie very
12.3 Theories of Magnetism at LaAlO3/SrTiO3 Interface 503
close to the interface and are thus easily localized by interface defects, or the
Ti eg orbitals by a restructuring of the Ti 3d orbitals near oxygen vacancies
[62–64]. In the former case, the ferromagnetic state is assumed to arise from
oxygen vacancies introducing localized magnetic moments, while in the lat-
ter case, the localized eg electrons interact with itinerant electrons leading to
Stoner magnetism.
With the discovery of localized ferromagnetic patches separated by a
paramagnetic phase, additional detail was necessary. By taking the view that
oxygen vacancies lead to an orbital reconstruction in nearby Ti, Pavlenko
et al. [64] showed that ferromagnetism induced by oxygen vacancies would
only occur above a critical vacancy density. This would then allow for the
phase separation of superconductivity and ferromagnetism [65], which has
alternatively been explained by a spiral-spin state in the dxz / d yz bands by
Banerjee et al. [66], while superconductivity is present in the dxy band. The
patches [18–20, 28] are possibly the broken spirals due to defects [66, 67].
However, in this picture, the spiral states are made possible by the Rashba
spin-orbit coupling, whose magnitude can be controlled via electric field
effect [36, 39], the spiral states and, accordingly, ferromagnetic patches, are
expected to be tunable via the electric field effect [66, 67], but this is not
observed by Bert et al. [28].
Alternative explanations of the coexistence of superconductivity and
ferromagnetism rely upon a similar phase-separation picture. Michaeli
et al. [61] argue that Fulde–Ferrell–Larkin–Ovchinnikov (FFLO) pair-
ing can occur in the dxz / d yz bands while the ferromagnetism arises from
a Ruderman–Kittel–Kasuya–Yosida (RKKY) interaction amongst localized
dxy electrons. Alternatively, Fidkowski et al. [68] argue that superconduc-
tivity in the interface is of a hybrid s- and p-wave pairing induced by the
proximity effect with superconducting grains in the SrTiO3 bulk, while the
ferromagnetism arises from Kondo interactions between localized moments
and itinerant electrons, to which the hybrid pairing is insensitive.
In addition to the coexistence of ferromagnetism and superconductiv-
ity, there have also been several theoretical attempts to explain anisotropic
magnetic effects. Fischer et al. [69] shows that Rashba spin-orbit coupling in
a system with Stoner magnetism can result in an anisotropic spin suscep-
tibility, as well as a nematic phase arising from unequal occupation of the
dxz / d yz bands. Also using spin-orbit coupling, Fete et al. [12] presented a
model where anisotropic magnetoconductance arises from an orbital recon-
struction in which, at low carrier densities, the dxy band is responsible for
the electron transport, and, as density increases, the dxz / d yz bands begin to
dominate and introduce anisotropic effects due to their nearly one-dimen-
sional nature at the interface.
A recent attempt to explain many of these features simultaneously has
been presented by Ruhman et al. [70]. In this theory, the magnetic features
arise as a result of competition between Kondo screening in the dxy band
and RKKY interactions in the dxz / d yz bands. In addition to having a para-
magnetic (dominated by Kondo screening) and a ferromagnetic (dominated
by localized magnetic moments/ordering of the RKKY interaction), there is
also a spin-glass phase where the localized magnetic moments are insuffi-
cient to order the random RKKY interactions in the dxz/d yz bands.
More recently, this many-body explanation of the magnetic effects in
LaAlO3/SrTiO3 interfaces has been questioned due to the temperature and
density dependence of the giant magnetoresistance. In the case of Kondo
screening, this effect should be very sensitive to thermalization effects,
but this is not observed experimentally. Thus, Diez et al. [31] put forward
a model based upon the quasiclassical Boltzmann diffusion equation with
spin-orbit coupling and band anisotropy to explain this phenomenon as a
single-particle effect.
12.4 MAIN THESIS: TWO DISTINCT CLASSES OF

MAGNETIC BEHAVIOR IN LaAlO3/SrTiO3
Here we argue that the collection of experimental evidence for magne-
tism at the LaAlO3/SrTiO3 interface can be meaningfully sorted into two
categories. The first is a ferromagnetic or Kondo liquid phase derived from
local magnetic moments that are coupled via exchange interactions with
itinerant electrons (Figure 12.10b). The ferromagnetic phase is favored
(a) (d)
I N
(b) (e)
FM P
(f)
(c)
K SC
local moment itinerant electron
FIGURE 12.10 Schematics for various phases for LaAlO3/SrTiO3. (a) Insulating state
(I): the concentration of itinerant electrons is not high enough to support conduction
at the interface. (b) Ferromagnetic (FM): local moments are antiferromagnetically
coupled to the itinerant electrons. When the concentration of the itinerant electrons
is high enough to support long-range ferromagnetic order, but not too high to
screen the local moments. (c) Kondo-screened (K) phase: when the concentration of
itinerant electrons is further increased, the spins of itinerant electrons are coupled to
localized moment. (d) Normal state (N): when the itinerant electrons are sufficient to
support conduction at the interface. (e) Paired state (P): where pairs of electrons are
formed. (f) Superconducting (SC) phase.
when the interface is insulating, weakly conductive, or exhibiting locally

insulating regions. The Kondo regime is favored when the density of itin-
erant electrons exceeds the density of local magnetic moments. We argue
this is the magnetic phase that has been observed by scanning SQUID
[18–20], SQUID [5, 9], cantilever magnetometry [17], MFM [21, 22], and
MCD [71].
The second type of magnetism is a metamagnetic phase, resulting from
an agglomeration of effects from the band structure of the d manifold of
the Ti atoms in SrTiO3, spin-orbit coupling, and electronic pairing. This
second type of magnetism manifests itself in a variety of magnetotransport
signatures including magnetoresistance anisotropy [5, 7–12], sign changes
of the magnetoresistance [15, 31], and anomalous Hall effects [11, 12, 14,
15]. Notably, these effects do not appear to interact with localized magnetic
moments. Regarding interactions between ferromagnetism and supercon-
ductivity, we argue that the evidence suggests that they may indicate prox-
imity effects, but not necessarily true coexistence [72].
12.5 FERROMAGNETIC AND KONDO REGIMES

Ferromagnetism at the LaAlO3/SrTiO3 interface results from localized
moments and their antiferromagnetic (RKKY) exchange with itinerant
electrons [73]. The detail of the exchange interaction depends on the con-
centration of the itinerant carriers, as well as the concentration of localized
moments. Variations involving superexchange [66] and double exchange
have also been described theoretically [74, 75]. As the local moments are
antiferromagnetically exchange-coupled to itinerant electrons, ferromag-
netic ordering between the local moments is expected [24, 76] as long
as the density of local moments nm exceeds that of the itinerant electrons
ne. The ferromagnetic order disappears as the interface become conductive.
The dynamic magnetic screening of magnetization by itinerant carriers was
directly observed by Bi et al. [21] using top-gated LaAlO3/SrTiO3 structures
[21]. In the regime where ne > nm , we expect Kondo physics to dominate
(Figure 12.10c).
The fact that ferromagnetism can be tuned electronically is crucial for
reconciling many of the apparent inconsistencies—including null results—
described earlier. For example, the experiments of Fitzsimmons [56] did not
actively control electron density at the interface (Figure 12.11a) and may have
taken place in the Kondo regime. Inconsistencies in observations of ferro-
magnetic patches—either the patches themselves or sample-to-sample varia-
tions—can possibly be reconciled in two ways. If a sample is locally insulating,
or simply has an excess of magnetic moments relative to the itinerant carrier
density (ne < nm ), then a ferromagnetic phase is favored. Local variations in
the impurity density (or patches of magnetic moments) could fulfill this con-
dition. Alternatively, local variations in the itinerant carrier density—with
magnetic-impurity density presumably held constant—could produce a local
ferromagnetic region. Surface adsorbates are known to strongly influence the
local electron density [77–79].
(a) Δx-1 (b) 4
H # LaAlO3 u.c. 8
1 7
Lifshitz transistion (nL)

c.
30
u.
T
6
40
B
n
3 2 Bp
300K 5
T 10-6 mbar 3 mbar
P(O2)
??
1. Ref. [9] ?
2. Ref. [18]
3. Ref. [21, 22]
4. Ref. [4]
ne stress
ne
0.3 μB/interface u.c. 1. Ref. [14, 15, 70] 5. Ref. [12]
5. Ref. [56]
2. Ref. [11] 6. Ref. [31]
6. Ref. [17]
7. Ref. [29]
nm 3. Ref. [39] 7. Ref. [7]
4. Ref. [36] 8. Ref. [49, 93, 94]
FIGURE 12.11 Approximate parameter space and phase diagrams. The parameter space explored by
various research groups discussed in the main text. (a) Ferromagnetism. The knobs commonly tuned,
are: magnetic field (B), temperature (T), carrier concentration (ne), moment concentration (nm), mechani-
cal stress, oxygen partial pressure for sample growth (P(O2)), number of LaAlO3 layers (# LaAlO3 u.c.), and
resolution (Δ× −1). We argue that the null results reported by Fitzsimmons et al. [56] could potentially be
reconciled if the carrier concentration is controlled. (b) Metamagnetism parameter space explored by vari-
ous works discussed in the text, and the carrier concentration dependent critical field Bp for the metamag-
netic transition.
The magnetic patches were not tuned within the backgate range
explored (−70 V to 390 V) by Bert et al. [28], suggesting that the local
density of the itinerant electrons n e and magnetic moments nm inside
the magnetic patches must be substantially different from the rest of the
sample and are therefore difficult to control via backgating, unlike the
gate-tuned superfluid density for the rest of the sample. This explanation,
however, does not seem to be compatible with the fact that no clear phase
competition between the superfluid density and ferromagnetic patches is
observed [80]. The independence of the magnetic effect on the backgate
is also reported by Ron et al. [54] in shadow-mask created 1D nanowires.
One possible explanation is that magnetic ordering exists at the intrin-
sically insulating boundary of the nanowire, which makes it harder to
manipulate using the electric field effect. However, the low critical tem-
perature (1 K) reported by Ron et al. [54] suggests a different origin or
regime altogether.
12.5.1 What are the Local Moments?

Perhaps one of the outstanding questions relates to the physical origin of
the local moments. They can either be extrinsic in nature, e.g. derived from
the interface, or simply magnetic impurities. Alternatively, they can have an
intrinsic origin, such as defect states from vacancies of composite atoms, cat-
ion substitution, or interdiffusion [81]. Here we discuss both possibilities and
then narrow the discussion to the case for oxygen vacancies.
12.5.1.1 Magnetic Impurities, Oxygen Vacancies

(Vacancy Clusters), and F-Centers
Dilute magnetic impurities are generally found in SrTiO3 substrates. Impurity
levels in the order of one part per million have been reported [82–84], with
large vendor-dependent variations [84]. For impurity concentrations of
this level, it is possible to induce magnetism, as in the case of many dilute
magnetic semiconductors [85]. However, observation of magnetic signals
depends strongly on growth conditions, such as the oxygen partial pressure
during growth of LaAlO3 [9, 86], post-growth annealing temperature, and
oxygen partial pressure [24, 25, 87]. This makes oxygen related defects such
as vacancies and clusters the most common suspects. Both of these growth
“knobs” can also tune the electron density. As mentioned, ferromagnetic
hysteresis was only observed by Ariando [9] for samples with high oxygen
pressure (P(O2) = 10−2 mbar), close to the metal-insulator transition. The
more highly conductive samples grown at lower oxygen pressures did not
show ferromagnetic hysteresis.
Oxygen vacancies are intrinsic, pervasive point defects for SrTiO3.
Oxygen vacancies and vacancy clusters modify the band structure and cre-
ate deep, localized in-gap states [59, 63], at, or in the vicinity of, defect sites.
Each oxygen vacancy, in principle, donates two electrons. Depending on the
configuration of vacancy clustering, these electrons can remain localized on
the vacancy site, or fill the d orbitals of nearby Ti atoms. The defect sites
with unpaired electrons are referred to as F-centers or color centers. Both
F-centers and Ti 3d orbital can possess a local moment [88, 89].
Strong evidence linking magnetic moments to oxygen vacancies comes
from a report by Rice et al. [71] which used magnetic circular dichroism
(MCD) to optically excite long-lived magnetic states of reduced SrTiO3 sin-
gle crystals. The sample was magnetized using a circularly polarized pump
(Figure 12.4a) and probed via differential absorption of right- and left-circu-
larly polarized light. The observed magnetization persists after the pump-
ing beam is blocked, enabling long-lived magnetic states to be imprinted
(Figure 12.4b). Samples doped with niobium did not show magnetization
induced by MCD until oxygen vacancies were subsequently induced. The
samples with oxygen vacancies did not exhibit magnetization after being re-
oxygenated. A similar dependence on oxygen vacancies was also reported by
Liu et al. [25] using SQUID magnetometry.
Following Rice et al. [71], Choi et al. [59] provided DFT calculations for
in-gap states induced by oxygen vacancies of different valence, VO, VO+, and
VO2+, consistent with Rice et al [71]. Moments due to unpaired spins of Ti are
only induced from VO and VO+. Furthermore, VO2+ dominate the population
in the bulk, suggesting that magnetic moments may cluster around inter-
faces or ferroelastic domain boundaries. Calculations by Cuong et al. [90]
suggest that vacancies tend to cluster into apical divacancy pairs, creating
in-gap state of eg character. Altmeyer et al. [91] discuss magnetic moments
of Ti 3d as a function of distance between the Ti atom and single oxygen
vacancy or apical divacancies.
TABLE 12.1
Parameter Space Explored by the Works Discussed in the Text
Max Backgate #LaAlO3 Unit Device Size/ Post
Field (T) Temperature (K) ne (1013 cm−2) Voltage (V) Cells (u.c.) Dimensionality P(O2) (mbar) Anneal?
Joshua et al. [15] 14 T 50 mK, 2 K, 4.2 K 1–3 −50 to 450 6, 10 mm/2D 10−4 o
Ben Shalom 18 T 20 mK-100 K 0.1–5 −50 to 50 8, 15 mm/2D 5 × 10−4 − 10−3 x
et al. [36]
Caviglia et al. [39] 12 T 30 mK - 20 K 1.85–2.2 −340 to 320 12 mm/2D 6 × 10−5, 10−4 o
Fete et al. [12] 7T 1.5 K >4 mm/2D 10−4 o
Ariando et al. [9] 9T 2 K–300 K 10 mm/2D 10−6 −5 × 10−2 x
Ron et al. [54] 18 T 20 mK, 60 mK −5 to 10 10, 6, 16 50 nm 10−4 (Torr) o
Maniv et al. [104] mm/2D
Cheng et al. [49] 9T 50 mK 3.4 nm/1D 10−3 o
12.6 METAMAGNETISM
Here we describe a complementary set of effects, most of which are related
to transport, that we label with a single term “metamagnetism”. These effects
generally take place well below room temperature (~35 K) but above the
superconducting transition. Many different effects take place within the same
range of tunable parameters (Figure 12.11b and Table 12.1); we argue below
that these effects are manifestations of the same underlying phenomena.
12.6.1 Lifshitz Transition
Among all the reports from giant negative magnetoresistance, anisotropy
in magnetoresistance and the anomalous Hall effect, the Lifshitz transi-
tion appears to play a central role. Its importance within the context of
magnetotransport phenomena was first stressed by Joshua et al. [14]. The
critical (Lifshitz) density (referred as n L in the following text) is where the
d xy-derived branches of the t2 g bands are partially filled, and the d xz /dyz-
derived bands begin to be populated (Figure 12.7). Giant negative mag-
netoresistance, magnetotransport anisotropy, and anomalous Hall effect
are all observed only above the Lifshitz transition. The magnetoresistance
near the Lifshitz transition evolves from being weak and weakly-depen-
dent on angle, to strong and highly anisotropic, with a crossover that
depends on both magnetic field and electron density. The Lifshitz tran-
sition is also strongly linked to an observed tunable spin-orbit coupling
[39] and takes place close to the maximum superconducting transition
temperature [14].
12.6.2 Anisotropic Transport versus n, B

As discussed previously, the anisotropy of the magnetoresistance undergoes
a vivid transition from sinusoidal 2-fold oscillation (Figure 12.5b) to a four-
fold, or irregular oscillation (Figure 12.5c), when the carrier concentration
ne is increased above the Lifshitz transition nL. This transition between low
and high anisotropy is also found to be a function of the external magnetic
12.6 Metamagnetism 509
field (Figure 12.5d,e). The critical field for this transition depends on the car-
rier concentration, and a phase diagram (Figure 12.9b) was mapped out by
Joshua et al. [15], showing that the critical magnetic field can vary from ~2
T to as large as 15 T as the carrier density is reduced to nL. In separate mea-
surements, Diez et al. [31] showed a large negative magnetoresistance that
exhibits a strong backgate dependence, becoming much more pronounced
at large backgate values (Figure 12.12b). In a separate magnetotransport
study of Ben Shalom et al. [36] (Figure 12.12a), a large negative magneto-
resistance effect takes place at a critical magnetic field m 0 H * that ranges
from 2 T to >18 T as the carrier density is reduced from n = 7.8 ´ 1013 cm -2
to 3.0 ´ 1013 cm -2. It is worth noting that all of these experiments are prob-
ing a similar range of magnetic field and carrier density, and all of them are
observing two distinct phases that are separated by a curve similar to that
shown in Figure 12.11b and Table 12.1.
12.6.3 Anomalous Hall Effect

As mentioned previously, there is also an critical carrier density nL as well
as a carrier ne dependent critical in-plane H c|| and out-of-plane field H c^ , as
mapped out by Joshua et al. [15]. The (model-independent) interpretation is
that the offset and increase in Hall angle represents an excess magnetization
that is “released” above a critical magnetic field, indicating a “gate-tunable
polarized phase” [15]. Now the question becomes: what is this magnetiz-
able degree of freedom, and what causes the sudden release as the field is
increased? This magnetization does not behave as a single-component
Langevin-type paramagnetism, because it lacks the anticipated temperature
dependence, i.e. it is temperature-independent [16]. As discussed, by adding
an anomalous Hall term with Langevin-type paramagnetic field dependence,
Gunkel et al. [16] was able to capture the small upturn in the Hall coefficient
RH missed by the simple multiband model. However, Gunkel et al. [16] also
pointed out that (i) a temperature-independent saturation field BC for the
Hall coefficient RH, and (ii) a temperature dependent R0AHE deviates from the
expected scaling of the Langevin-type spin-1/2 paramagnetic system.
12.6.4 Correlated Transport Properties

The anisotropic transport, anomalous Hall effect, and electron pairing with-
out superconductivity all share features that change with carrier density and
magnetic field in similar ways (see Figure 12.11b). Hence it is natural to con-
sider the possibility that they are different manifestations of the same underly-
ing phenomena. Here we discuss this possibility explicitly. First, let us consider
the AHE and its connection to electron pairing without superconductivity
in a SET reported by Cheng et al. [49]. Based on the stability of the paired
state conductance peaks with respect to magnetic field (region B < BP » 2T
in Figure 12.8b) it is known that the electron pairs exist in a spin-singlet con-
figuration, i.e. Stot = 0. In the unpaired state (i.e. for B > BP ), electrons are now
unpaired, with spin states S , Sz ñ = 1 / 2, ± 1 / 2ñ that can subsequently Zeeman
split and polarize in the excess magnetic field B - BP . This polarization of

unpaired spins should produce an anomalous Hall effect, similar to what is
observed in bulk 2D experiments. The phase diagram for H C|| from Joshua et
al. [15] (Figure 12.9b) shows a monotonically decreasing dependence on the
carrier density ne. In the SET experiments, pairing fields as large as 5 T were
reported; in follow-up experiments on ballistic quantum channels, the pairing
transition has been observed at magnetic fields as large as 11 T [92]. The SET
measurements only probe pairing at discrete values of carrier density, which
itself is not directly measurable on the QD. For a single device, a few (2–5) data
points are clearly observable. Figure 12.9a shows the pairing energy for a set
of devices, plotted against a normalized gate voltage. While the scaling is not
always monotonic, there is a general trend that favors larger pairing fields (in
a particular device) with lower electron density. This behavior (Figure 12.9b)
is similar to the crossover curve separating regions of large (anisotropic) and
small (isotropic) magnetoresistance and anomalous Hall effect.
Regarding the transition between low and high anisotropy in transport,
again, the collective evidence indicates that the paired state is most likely com-
posed of dxy carriers, which, near the Lifshitz transition, is the preferred ground
state due to a presumed attractive interaction shared only between dxy carriers.
The dxz and dyz carriers, which have highly anisotropic band structure, become
occupied when the pairing energy is overcome by sufficiently large magnetic
fields. As the electron density increases above the Lifshitz point, the energy
difference between the (unoccupied) dxz / d yz states and the (occupied) dxy
states reduces and eventually vanishes. That is to say, a monotonically decreas-
ing pairing field is expected, regardless of the physical origin of the electron
pairing itself. The crossover between attractive (U < 0) and repulsive (U > 0)
regimes was observed by Cheng et al. [93] in transport measurements on
devices similar to those depicted in Figure 12.8a, but at higher carrier density.
Differences between the paired and unpaired phases are also manifested
as negative magnetoresistance effects, notably the results from Ben Shalom
et al. [36] (Figure 12.12a). In that experiment, a critical field H * was identified
as the onset for giant negative magnetoresistance. The value of H * changes
from <2 T to >18 T as the carrier density decreases toward what appears to
be the Lifshitz transition (although about 2x higher than the value measured
by Joshua et al. [15]). This behavior can also be incorporated into a frame-
work in which changes in resistance may arise due to qualitative differences
in transport between electron pairs and unpaired electrons. At sufficiently
low carrier density, the electron pairs may propagate ballistically [94] or
become localized.
12.7 FUTURE DIRECTIONS
12.7.1 Empirical Tests
How would one test the proposed picture of two magnetisms? Regarding
ferromagnetism and related phases, it is necessary to establish the origin of
the magnetic moments. As discussed previously, oxygen vacancies are prime
12.7 Future Directions 511
FIGURE 12.12 Giant negative longitudinal (Rxx) magnetoresistance. When an external magnetic field
exceeds Hc|| and a carrier-concentration-dependent critical field H * , the longitudinal magnetoresistance Rxx
drops significantly. (a) At T = 20mK, the evolution of the sheet resistance is plotted as a function of in-plane
magnetic field (parallel to the current) at different carrier concentration. The resistance is restored at Hc|| ,
followed by a field-independent interval from Hc|| to H * , and then a sudden drop at H * . (After Shalom, B. et al.,
Phys. Rev. Lett. 104, 126802, 2010. With permission.) (b) At T = 1 .4 K (above Tc), a drop of 70% in Rxx is reported
by Diez et al. (After Diez, M. et al., Phys. Rev. Lett. 115, 016803, 2015. With permission.)
suspects for producing the required magnetic moments. Engineering oxygen

vacancies via growth conditions and/or annealing can control the concentra-
tion of local moments [87], but it is not clear where exactly they are stabilized
and whether they are single vacancies or clusters. To capture the physics,
more detailed experimental and theoretical studies need to be performed.
Concerning experiments that probe ferromagnetism, it is important to be
able to control the concentration of itinerant electrons, a parameter that has
been shown to be critical to the stabilization of ferromagnetic phases [21].
Topgating, as opposed to backgating through the SrTiO3 substrate, is more
effective in controlling carrier density and reaching a properly insulating
phase. Finally, spatially resolved measurements correlating oxygen vacancies
and the magnetic moments, together with any long-range order they induce,
could unequivocally establish the mechanism for dilute ferromagnetism at
this interface.
12.7.2 Where Are the Oxygen Vacancies?

Here we turn specifically to the suspected role of oxygen vacancies. If
they are indeed responsible for the magnetic moments, the question
becomes: where exactly are these vacancies located? Are they pinned to
the interfaces? Do they cluster around ferroelastic domain walls? While
there is no direct experimental evidence showing that the oxygen vacan-
cies are pinned within ferroelastic domain walls in SrTiO3, in general,
for perovskites, the movement of ferroelastic domain walls is restricted

by defects, and could be easily pinned by them. Conversely, mobile
oxygen vacancies may become trapped at ferroelastic domain walls, as
pointed out by Goncalves-Ferreira et al. [95] for CaTiO3 (Figure 12.4c),
oxygen vacancies have lower energy when they are placed at a twin wall.
Measurements using soundwaves, suggesting that domain walls are polar,
has been reported by Scott et al. [96].
Ferroelastic domains in SrTiO3 have been imaged in a variety of ways,
including optical microscopy [97], scanning SET (Figure 12.4d), scanning
SQUID (Figure 12.4e), and low temperature SEM [98]. Ferroelastic domains
are found to modulate the flow of the current [99], as well as the critical tem-
perature of the superconductivity [100], and they play an important role in
the gating of essentially all LaAlO3/SrTiO3 heterostructures at low tempera-
tures. Scanning SQUID experiments, (Figure 12.4f ) in which the probe gently
touches the LaAlO3 surface, have been demonstrated to produce significant
changes in the magnitude and direction of ferromagnetic patches [20]. The
force magnitude (~1 μN) is too small to create damage or otherwise depart
from the elastic regime; however, it is plausible that this kind of force may
result in a rearrangement of ferroelastic domains and therefore the spatial
arrangement of magnetic moments. Further experimentation would be nec-
essary to establish a clearer connection and to relate the ferroelastic domain
structure to the presence of stabilized magnetic moments.
12.7.3 What Is the Origin of Electron Pairing?

In all of the experiments related to electron pairing thus far, the picture
appears to be that there is a strongly attractive pairing interaction (Hubbard
U < 0) at low electron densities that is responsible for electron pairing and
superconductivity. As the electron density increases, the sign of the elec-
tron-electron interaction changes (U > 0) and superconductivity is ulti-
mately suppressed. The paired, non-superconducting phase, close to the
Lifshitz transition, becomes the dominant mode of transport, and yet it is
far from being a Fermi liquid. Perhaps the biggest open question relates to
the underlying physical mechanism for electron pairing. So far, there is no
clear experiment (or theory) that favors a particular mechanism. A deeper
understanding of electron pairing (and, as a consequence, superconductiv-
ity) in SrTiO3-based systems would represent an important milestone in our
overall ability to describe the physics of one of the richest material systems
known.
ACKNOWLEDGMENTS
We would like to acknowledge Ariando, A. D. Caviglia, B. Kalisky, K. A.
Moler, K. Nowack, and J. M. Triscone for helpful feedback on the manu-
script. We thank NSF DMR (1124131, 1609519, 1124131) and ONR N00014-
15-1-2847 for financial support.
References 513
REFERENCES
1. M. L. Cohen, Superconductivity in low-carrier-density systems: Degenerate
semiconductors, Superconductivity: Part 1 615 1969.
2. J. G. Bednorz and K. A. Muller, Perovskite-type oxides-the new approach to
High Tc superconductivity, Nobel Lecture (1987).
3. A. Ohtomo and H. Y. Hwang, A high-mobility electron gas at the LaAlO3 /
SrTiO3 heterointerface, Nature 427 , 423– 426 (2004).
4. A. Brinkman, M. Huijben, M. Van Zalk et al., Magnetic effects at the interface
between non-magnetic oxides, Nat. Mater. 6 , 493– 496 (2007).
5. H.-L. Hu, R. Zeng, A. Pham et al., Subtle interplay between localized magnetic
moments and itinerant electrons in LaAlO3 /SrTiO3 heterostructures, ACS
Appl. Mater. Interfaces 13630– 13636 (2016).
6. M. Lee, J. R. Williams, S. Zhang, C. D. Frisbie, and D. Goldhaber-Gordon,
Electrolyte gate-controlled Kondo effect in SrTiO3 , Phys. Rev. Lett. 107 , 256601
(2011).
7. M. Ben Shalom, C. W. Tai, Y. Lereah et al., Anisotropic magnetotransport at
the SrTiO3 /LaAlO3 interface, Phys. Rev. B 80 , 140403 (2009).
8. S. Seri and L. Klein, Antisymmetric magnetoresistance of the SrTiO3 /LaAlO3
interface, Phys. Rev. B 80 , 180410 (2009).
9. X. Ariando, G. Wang, Z. Q. Baskaran et al., Electronic phase separation at the
LaAlO3 /SrTiO3 interface, Nat. Commun. 2 , 188 (2011).
10. X. Wang, W. M. Lu, A. Annadi et al., Magnetoresistance of two-dimensional
and three-dimensional electron gas in LaAlO3 /SrTiO3 heterostructures:
Influence of magnetic ordering, interface scattering, and dimensionality, Phys.
Rev. B 84 , 075312 (2011).
11. S. Seri, M. Schultz, and L. Klein, Interplay between sheet resistance increase and
magnetotransport properties in LaAlO3 /SrTiO3 , Phys. Rev. B 86 , 085118 (2012).
12. A. Fê te, S. Gariglio, A. D. Caviglia, J. M. Triscone, and M. Gabay, Rashba
induced magnetoconductance oscillations in the LaAlO3 -SrTiO3 heterostruc-
ture, Phys. Rev. B 86 , 201105 (2012).
13. A. Annadi, Z. Huang, K. Gopinadhan et al., Fourfold oscillation in anisotropic
magnetoresistance and planar Hall effect at the LaAlO3 /SrTiO3 heterointer-
faces: Effect of carrier confinement and electric field on magnetic interactions,
Phys. Rev. B 87 , 201102 (2013).
14. A. Joshua, S. Pecker, J. Ruhman, E. Altman, and S. Ilani, A universal critical
density underlying the physics of electrons at the LaAlO3 /SrTiO3 interface,
Nat. Commun. 3 , 1129 (2012).
15. A. Joshua, J. Ruhman, S. Pecker, E. Altman, and S. Ilani, Gate-tunable polar-
ized phase of two-dimensional electrons at the LaAlO3 /SrTiO3 interface, Proc.
Natl. Acad. Sci. 110 , 9633 (2013).
16. F. Gunkel, C. Bell, H. Inoue et al., Defect control of conventional and anoma-
lous electron transport at complex oxide interfaces, Phys. Rev. X 6 , 031035
(2016).
17. L. Li, C. Richter, J. Mannhart, and R. C. Ashoori, Coexistence of magnetic
order and two-dimensional superconductivity at LaAlO3 /SrTiO3 interfaces,
Nat. Phys. 7 , 762– 766 (2011).
18. J. A. Bert, B. Kalisky, C. Bell et al., Direct imaging of the coexistence of ferro-
magnetism and superconductivity at the LaAlO3 /SrTiO3 interface, Nat. Phys.
7 , 767– 771 (2011).
19. B. Kalisky, J. A. Bert, B. B. Klopfer et al., Critical thickness for ferromagnetism
in LaAlO3 /SrTiO3 heterostructures, Nat. Commun. 3 , 922 (2012).
20. F. B. Kalisky, J. A. Bert, C. Bell et al., Scanning probe manipulation of magne-
tism at the LaAlO3 /SrTiO3 heterointerface, Nano Lett. 12 , 4055– 4 059 (2012).
21. F. Bi, M. Huang, S. Ryu et al., Room-temperature electronically-controlled fer-
romagnetism at the LaAlO3 /SrTiO3 interface, Nat. Commun. 5 , 5019 (2014).
22. F. Bi, M. Huang, H. Lee, C.-B. Eom, P. Irvin, and J. Levy, LaAlO3 thickness
window for electronically controlled magnetism at LaAlO3 /SrTiO3 heteroin-
terfaces, App. Phys. Lett. 107 , 082402 (2015).
23. J. S. Lee, Y. W. Xie, H. K. Sato et al., Titanium d xy ferromagnetism at the
LaAlO3 /SrTiO3 interface, Nat. Mater. 12 , 703– 706 (2013).
24. M. Salluzzo, S. Gariglio, D. Stornaiuolo et al., Origin of interface magnetism
in BiMnO3 /SrTiO3 and LaAlO3 /SrTiO3 heterostructures, Phys. Rev. Lett. 111 ,
087204 (2013).
25. Z. Q. Liu, W. M. Lü , S. L. Lim et al., Reversible room-temperature ferromagne-
tism in Nb-doped SrTiO3 single crystals, Phys. Rev. B 87 , 220405 (2013).
26. M. E. Huber, N. C. Koshnick, H. Bluhm et al., Gradiometric micro-SQUID
susceptometer for scanning measurements of mesoscopic samples, Rev. Sci.
Instrum. 79 , 053704 (2008).
27. S. Thiel, G. Hammerl, A. Schmehl, C. W. Schneider, and J. Mannhart, Tunable
quasi-two-dimensional electron gases in oxide heterostructures, Science 313 ,
1942– 1945 (2006).
28. J. A. Bert, K. C. Nowack, B. Kalisky et al., Gate-tuned superfluid density at the
superconducting LaAlO3 /SrTiO3 interface, Phys. Rev. B 86 , 060503 (2012).
29. Z, Salman, O. Ofer, M. Radovic et al., Nature of weak magnetism in SrTiO3 /
LaAlO3 multilayers, Phys. Rev. Lett. 109 , 257207 (2012).
30. M. Ben Shalom, A. Ron, A. Palevski, and Y. Dagan, Shubnikov-De Haas oscil-
lations in SrTiO3 /LaAlO3 interface, Phys. Rev. Lett. 105 , 206401 (2010).
31. M. Diez, A. M. R. V. L. Monteiro, G. Mattoni et al., Giant negative magneto-
resistance driven by spin-orbit coupling at the LaAlO3 /SrTiO3 interface, Phys.
Rev. Lett. 115 , 016803 (2015).
32. D. Stornaiuolo, C. Cantoni, G. M. De Luca et al., Tunable spin polarization and
superconductivity in engineered oxide interfaces, Nat. Mater. 15 , 278– 283
(2016).
33. A. Fê te, C. Cancellieri, D. Li et al., Growth-induced electron mobility enhance-
ment at the LaAlO3 /SrTiO3 interface, App. Phys. Lett. 106 , 051604 (2015).
34. J. S. Kim, S. S. A. Seo, M. F. Chisholm et al., Nonlinear Hall effect and multi-
channel conduction in LaTiO3 /SrTiO3 superlattices, Phys. Rev. B 82 , 201407
(2010).
35. P. Gallagher, M. Lee, T. A. Petach et al., A high-mobility electronic system at
an electrolyte-gated oxide surface, Nat. Commun. 6 , 6437 (2015).
36. M. Ben Shalom, M. Sachs, D. Rakhmilevitch, A. Palevski, and Y. Dagan, Tuning
spin-orbit coupling and superconductivity at the SrTiO3 /LaAlO3 interface: A
magnetotransport study, Phys. Rev. Lett. 104 , 126802 (2010).
37. Y. Lee, C. Clement, J. Hellerstedt et al., Phase diagram of electrostatically
doped SrTiO3 , Phys. Rev. Lett. 106 , 136809 (2011).
38. C. Bell, S. Harashima, Y. Kozuka et al., Dominant mobility modulation by the
electric field effect at the LaAlO3 /SrTiO3 interface, Phys. Rev. Lett. 103 , 226802
(2009).
39. A. D. Caviglia, M. Gabay, S. Gariglio, N. Reyren, C. Cancellieri, and J. M.
Triscone, Tunable Rashba spin-orbit interaction at oxide interfaces, Phys. Rev.
Lett. 104 , 126803 (2010).
40. K. Narayanapillai, K. Gopinadhan, X. Qiu et al., Current-driven spin orbit field
in LaAlO3 /SrTiO3 heterostructures, App. Phys. Lett. 105 , 162405 (2014).
41. E. Lesne, Y. Fu, S. Oyarzun et al., Highly efficient and tunable spin-to-charge
conversion through Rashba coupling at oxide interfaces, Nat. Mater. (advance
online publication) (2016).
42. A. F. Santander-Syro, F. Fortuna, C. Bareille et al., Giant spin splitting of
the two-dimensional electron gas at the surface of SrTiO3 , Nat. Mater. 13 ,
1085– 1090 (2014).
43. S. McKeown Walker, S. Riccò , F. Y. Bruno et al., Absence of giant spin splitting
in the two-dimensional electron liquid at the surface of SrTiO3 (001), Phys. Rev.
B 93 , 245143 (2016).
References 515
44. A. C. Garcia-Castro, M. G. Vergniory, E. Bousquet, and A. H. Romero, Spin

texture induced by oxygen vacancies in strontium perovskite (001) surfaces: A
theoretical comparison between SrTiO3 and SrHfO3 , Phys. Rev. B 93 , 045405
(2016).
45. J. F. Schooley, W. R. Hosler, and M. L. Cohen, Superconductivity in semicon-
ducting SrTiO3 , Phys. Rev. Lett. 12 , 474– 475 (1964).
46. C. S. Koonce, M. L. Cohen, J. F. Schooley, W. R. Hosler, and E. R. Pfeiffer,
Superconducting transition temperatures of semiconducting SrTiO3 , Phys.
Rev. 163 , 380– 390 (1967).
47. T. Timusk and B. Statt, The pseudogap in high-temperature superconductors:
an experimental survey, Rep. Prog. Phys. 62 , 61– 122 (1999).
48. C. Richter, H. Boschker, W. Dietsche et al., Interface superconductor with
gap behaviour like a high-temperature superconductor, Nature 502 , 528– 531
(2013).
49. G. Cheng, M. Tomczyk, S. C. Lu et al., Electron pairing without superconduc-
tivity, Nature 521 , 196– 199 (2015).
50. M. A. Kastner, The single-electron transistor, Rev. Mod. Phys. 64 , 849– 858
(1992).
51. N. Reyren, S. Thiel, A. D. Caviglia et al., Superconducting interfaces between
insulating oxides, Science 317 , 1196– 1199 (2007).
52. A. D. Caviglia, S. Gariglio, N. Reyren et al., Electric field control of the LaAlO3 /
SrTiO3 interface ground state, Nature 456 , 624– 627 (2008).
53. D. A. Dikin, M. Mehta, C. W. Bark, C. M. Folkman, C. B. Eom, and V.
Chandrasekhar, Coexistence of superconductivity and ferromagnetism in two
dimensions, Phys. Rev. Lett. 107 , 056802 (2011).
54. A. Ron, E. Maniv, D. Graf, J. H. Park, and Y. Dagan, Anomalous magnetic
ground state in an LaAlO3 /SrTiO3 interface probed by transport through
nanowires, Phys. Rev. Lett. 113 , 216801 (2014).
55. V. K. Guduru, Surprising magnetotransport in oxide heterostructures,
Master’ s thesis, Radboud University, Nijmegen, the Netherlands, 2014.
56. M. R. Fitzsimmons, N. W. Hengartner, S. Singh et al., Upper limit to magne-
tism in LaAlO3 /SrTiO3 heterostructures, Phys. Rev. Lett. 107 , 217201 (2011).
57. T. Wijnands, Scanning superconducting quantum interference device micros-
copy, Master’ s thesis, University of Twente, Enschede, the Netherlands, 2013,
http://essay.utwente.nl/62800/1/Master_Th esis_Tom_Wijnands_openbaar.pdf.
58. R. Pentcheva and W. E. Pickett, Charge localization or itineracy at
LaAlO3 ∕ SrTiO3 interfaces: Hole polarons, oxygen vacancies, and mobile elec-
trons, Phys. Rev. B 74 , 035112 (2006).
59. H. Choi, J. D. Song, K.-R. Lee, and S. Kim, Correlated visible-light absorption
and intrinsic magnetism of SrTiO3 due to oxygen deficiency: Bulk or surface
effect? Inorg. Chem. 54 , 3759– 3765 (2015).
60. R. Pentcheva and W. E. Pickett, Ionic relaxation contribution to the electronic
reconstruction at the n-type LaAlO3 /SrTiO3 interface, Phys. Rev. B 78 , 205106
(2008).
61. K. Michaeli, A. C. Potter, and P. A. Lee, Superconducting and ferromagnetic
phases in SrTiO3 /LaAlO3 oxide interface structures: Possibility of finite
momentum pairing, Phys. Rev. Lett. 108 , 117003 (2012).
62. G. Chen and L. Balents, Ferromagnetism in itinerant two-dimensional t 2g sys-
tems, Phys. Rev. Lett. 110 , 206401 (2013).
63. N. Pavlenko, T. Kopp, E. Y. Tsymbal, J. Mannhart, and G. A. Sawatzky, Oxygen
vacancies at titanate interfaces: Two-dimensional magnetism and orbital
reconstruction, Phys. Rev. B 86 , 064431 (2012).
64. N. Pavlenko, T. Kopp and J. Mannhart, Emerging magnetism and electronic
phase separation at titanate interfaces, Phys. Rev. B 88 , 201104 (2013).
65. N. Pavlenko, T. Kopp, E. Y. Tsymbal, G. A. Sawatzky, and J. Mannhart,
Magnetic and superconducting phases at the LaAlO3 /SrTiO3 interface: The
role of interfacial Ti 3d electrons, Phys. Rev. B 85 , 020407(R) (2012).
66. S. Banerjee, O. Erten, and M. Randeria, Ferromagnetic exchange, spin-orbit

coupling and spiral magnetism at the LaAlO3 /SrTiO3 interface, Nat. Phys. 9 ,
626– 630 (2013).
67. M. Gabay and J.-M. Triscone, Oxide heterostructures: Hund rules with a twist,
Nat. Phys. 9 , 610– 611 (2013).
68. L. Fidkowski, H. C. Jiang, R. M. Lutchyn, and C. Nayak, Magnetic and super-
conducting ordering in one-dimensional nanostructures at the LaAlO3 /SrTiO3
interface, Phys. Rev. B 87 (2013).
69. M. H. Fischer, S. Raghu, and E. A. Kim, Spin-orbit coupling in LaAlO3 /SrTiO3
interfaces: Magnetism and orbital ordering, New J. Phys. 15 , 023022 (2013).
70. J. Ruhman, A. Joshua, S. Ilani, and E. Altman, Competition between Kondo
screening and magnetism at the LaAlO3 /SrTiO3 interface, Phys. Rev. B 90 ,
125123 (2013).
71. W. D. Rice, P. Ambwani, M. Bombeck et al., Persistent optically induced mag-
netism in oxygen-deficient strontium titanate, Nat. Mater. 13 , 481– 487 (2014).
72. G. Cheng, J. P. Veazey, P. Irvin et al., Anomalous transport in sketched nano-
structures at the LaAlO3 /SrTiO3 interface, Phys. Rev. X 3 , 011021 (2013).
73. M. A. Ruderman and C. Kittel, Indirect exchange coupling of nuclear magnetic
moments by conduction electrons, Phys. Rev. 96 , 99– 102 (1954).
74. D. Odkhuu, S. H. Rhim, D. Shin, and N. Park, La displacement driven dou-
ble-exchange like mediation in titanium dxy ferromagnetism at the LaAlO3 /
SrTiO3 , J. Phys. Soc. Jpn. 85 , 043702 (2016).
75. F. Lechermann, L. Boehnke, D. Grieger, and C. Piefke, Electron correlation and
magnetism at the LaAlO3 /SrTiO3 interface: A DFT+DMFT investigation, Phys.
Rev. B 90 , 085125 (2014).
76. M. Salluzzo in P. Mele et al. (Eds.), Oxide Thin Films, Multilayers, and
Nanocomposites , Springer International Publishing, Switzerland, pp. 181– 211,
2015.
77. Y. Xie, Y. Hikita, C. Bell, and H. Y. Hwang, Control of electronic conduction at
an oxide heterointerface using surface polar adsorbates, Nat. Commun. 2 , 494
(2011).
78. K. A. Brown, S. He, D. J. Eichelsdoerfer et al., Giant conductivity switching
of LaAlO3 /SrTiO3 heterointerfaces governed by surface protonation, Nat.
Commun. 7 , 10681 (2016).
79. W. Dai, S. Adhikari, A. C. Garcia-Castro et al., Tailoring LaAlO3 /SrTiO3
interface metallicity by oxygen surface adsorbates, Nano Lett. 16 , 2739– 2743
(2016).
80. J. A. Bert, Superconductivity in reduced dimensions, Doctoral thesis, Stanford
University, CA, USA, 2012, http://www.stanford.edu/group/moler/theses/be
rt_thesis.pdf.
81. G. Salvinelli, G. Drera, A. Giampietri, and L. Sangaletti, Layer-resolved cation
diffusion and stoichiometry at the LaAlO3 /SrTiO3 heterointerface probed by
X-ray photoemission experiments and site occupancy modeling, ACS Appl.
Mater. Interfaces 7 , 25648– 25657 (2015).
82. J. M. D. Coey and S. A. Chambers, Oxide dilute magnetic semiconduc-
tors— Fact or fiction? MRS Bull. 33 , 1053– 1058 (2008).
83. M. A. Garcia, E. Fernandez Pinel, J. de la Venta et al., Sources of experimental
errors in the observation of nanoscale magnetism, J. Appl. Phys. 105 , 013925
(2009).
84. J. M. D. Coey, M. Venkatesan, and P. Stamenov, surface magnetism of stron-
tium titanate, J. Phys. Condens. Matter 28 , 485001 (2016).
tronic structures, Rev. Mod. Phys. 86 , 187– 251 (2014).
86. M. Huijben, A. Brinkman, G. Koster, G. Rijnders, H. Hilgenkamp, and D. H. A.
Blank, Structure-property relation of SrTiO3 /LaAlO3 interfaces, Adv. Mater.
21 , 1665– 1677 (2009).
References 517
87. Z. Q. Liu, L. Sun, Z. Huang et al., Venkatesan and Ariando, dominant role of
oxygen vacancies in electrical properties of unannealed LaAlO3 /SrTiO3 inter-
faces, J. Appl. Phys. 115 , 054303 (2014).
88. A. Lopez-Bezanilla, P. Ganesh, and P. B. Littlewood, Magnetism and metal-
insulator transition in oxygen-deficient SrTiO3 , Phys. Rev. B 92 , 115112 (2015).
89. A. Lopez-Bezanilla, P. Ganesh, and P. B. Littlewood, Research update: Plentiful
magnetic moments in oxygen deficient SrTiO3 , APL Mater. 3 , 100701 (2015).
90. D. D. Cuong, B. Lee, K. M. Choi, H.-S. Ahn, S. Han, and J. Lee, Oxygen vacancy
clustering and electron localization in oxygen-deficient SrTiO3 : LDA+U study,
Phys. Rev. Lett. 98 , 115503 (2007).
91. M. Altmeyer, H. O. Jeschke, O. Hijano-Cubelos et al., Magnetism, spin texture,
and in-gap states: Atomic specialization at the surface of oxygen-deficient
SrTiO3 , Phys. Rev. Lett. 116 , 157203 (2016).
92. Cheng et al., unpublished.
93. G. Cheng, M. Tomczyk, A. B. Tacla et al., Tunable electron-electron interac-
tions in LaAlO3 /SrTiO3 nanostructures, Phys. Rev. X 6 , 041042 (2016).
94. M. Tomczyk, G. Cheng, H. Lee et al., Micrometer-scale ballistic transport
of electron pairs in LaAlO3 /SrTiO3 nanowires, Phys. Rev. Lett. 117 , 096801
(2016).
95. L. Goncalves-Ferreira, S. A. T. Redfern, E. Artacho, E. Salje, and W. T. Lee,
Trapping of oxygen vacancies in the twin walls of perovskite, Phys. Rev. B 81 ,
024109 (2010).
96. J. F. Scott, E. K. H. Salje, and M. A. Carpenter, Domain wall damping and
elastic softening in SrTiO3 : Evidence for polar twin walls, Phys. Rev. Lett. 109 ,
187601 (2012).
97. Z. Erlich, Y. Frenkel, J. Drori et al., Optical study of tetragonal domains in
LaAlO3 /SrTiO3 , J. Supercond. Novel Magn. 28 , 1017– 1020 (2015).
98. H. J. H. Ma, S. Scharinger, S. W. Zeng et al., Local electrical imaging of tetrago-
nal domains and field-induced ferroelectric twin walls in conducting SrTiO3 ,
Phys. Rev. Lett. 116 , 257601 (2016).
99. Y. Frenkel, N. Haham, Y. Shperber et al., Anisotropic transport at the LaAlO3 /
SrTiO3 interface explained by microscopic imaging of channel-flow over
SrTiO3 domains, ACS Appl. Mater. Interfaces 8 , 12514– 12519 (2016).
100. H. Noad, E. M. Spanton, K. C. Nowack et al., Enhanced superconducting tran-
sition temperature due to tetragonal domains in two-dimensionally doped
SrTiO3 , Phys. Rev. B 94 , 174516 (2016).
101. H. Kroemer, Nobel lecture: Quasielectric fields and band offsets: Teaching
electrons new tricks, Rev. Mod. Phys. 73 , 783– 793 (2001).
102. M. Honig, J. A. Sulpizio, J. Drori, A. Joshua, E. Zeldov, and S. Ilani, Local elec-
trostatic imaging of striped domain order in LaAlO3 /SrTiO3 , Nat. Mater. 12 ,
1112– 1118 (2013).
103. B. Kalisky, E. M. Spanton, H. Noad et al., Locally enhanced conductivity due
to the tetragonal domain structure in LaAlO3 /SrTiO3 heterointerfaces, Nat.
Mater. 12 , 1091– 1095 (2013).
104. E. Maniv, M. B. Shalom, A. Ron et al., Strong correlations elucidate the elec-
tronic structure and phase diagram of LaAlO3 /SrTiO3 interface, Nat. Commun.
6 , 8239 (2015).
13
Electric-Field
Controlled
Magnetism

13.2 Electric Field Effects in Magnetic Semiconductors 520
13.2.1 Carrier-Induced Ferromagnetism in
Semiconductors 520
13.2.2 Electric Field Modulation of the Curie
Temperature 522
13.2.3 Electric Field Modulation of Magnetization
Process and Anisotropy 524
13.2.4 Electric Field Modulation of Other Magnetic
Parameters 526
13.3 Electric Field Effects in Ferromagnetic Metals 528
13.3.1 Electric Field Modulation of Magnetic
Anisotropy and Other Properties 528
519
520 Chapter 13. Electric-Field Controlled Magnetism
13.3.2 Electric Field-Induced Magnetization Dynamics 531

13.3.3 Electric Field-Induced Magnetization Switching 532
13.4 Conclusion 534
Acknowledgments 534
References 535
13.1 INTRODUCTION
As learned in the previous chapters, especially Chapters 9, 10, and 17, Volume
1, many researchers are now studying intensively the interplay between
electricity and magnetism. The electrical input to the magnets can induce
magnetization dynamics, and even switch its direction via spin torque, spin-
orbit torque, and electric-field-induced torque. This chapter focuses on the
electric-field effects on magnetism, as well as the electric-field control of
magnetization dynamics and direction. The effect was first demonstrated
using ferromagnetic semiconductors, and is now observed in other mate-
rial systems, e.g. metals and multiferroics. Here, we will restrict ourselves to
describing the electric-field effect-related phenomena in magnetic semicon-
ductors and metals. We will not cover the control of magnetic anisotropy in
magnets through stress adjacent to the piezoelectric material to which an
electric-field is applied. Readers who are interested in other material systems
and electric-field-induced phenomena than those treated here may consult
other reviews [1–4].
13.2 ELECTRIC FIELD EFFECTS IN

MAGNETIC SEMICONDUCTORS
One of the most important characteristics of magnetic semiconductors is
that the magnetic order is brought about by the presence of carriers. Hence,
their magnetism is expected to be controlled by conventional means that
vary the carrier concentration without changing temperature. This can be
done electrostatically by applying an electric field onto the semiconductors
using capacitor structures like a field effect transistor. The idea of electric
field controlled magnetism was already proposed in the 1960s, in the con-
text of research on rare earth magnetic semiconductors [5], and was dem-
onstrated experimentally in 2000 for a III–V compound-based magnetic
semiconductor, (In,Mn)As [6].
13.2.1 Carrier-Induced Ferromagnetism
in Semiconductors
As discussed in Chapters 9 and 10, Volume 2, many magnetic semiconduc-
tors are confirmed to exhibit carrier-induced ferromagnetism with the car-
rier concentration-dependent Curie temperature TC [7, 8]. In typical III–V
magnetic semiconductors, (In,Mn)As and (Ga,Mn)As, the majority of Mn

substituting for a III-group cation act as an acceptor, and thus one does not
need additional carrier doping to induce the ferromagnetism [9, 10]. In II–VI
counterparts, such as (Cd,Mn)Te and (Zn,Mn)Te, Mn ions are electrically
neutral, and thus one needs additional hole doping to observe the carrier-
induced ferromagnetism [11, 12]. There has been a long discussion on the
position of the Fermi level in (Ga,Mn)As, i.e. whether it lies in the valence
band or the impurity band [13, 14]. Recent experiments confirmed that it is
in the valence band [15, 16], consistent with the electronic structure obtained
from the first principles calculation [17, 18]. In such a case, the ferromagne-
tism is stabilized by the energy gain resulting from hole-repopulation in the
valence band with spin splitting induced by the p-d exchange interaction
(p-d Zener model) [13, 19]. The p-d exchange interaction is the exchange
interaction among p electrons in the valence band of the host semiconduc-
tor and d electrons of the guest magnetic ions. According to the model, TC of
these materials for three-dimensional case is given by,
TC = xeff N 0 S ( S + 1)β2 AFρs (TC ) / 12kB − TAF (13.1)
where:
xeff is the effective Mn spin composition
N0 the cation density
S the Mn spin
β the p-d exchange integral
ρS the spin density of states
AF the Fermi-liquid parameter
kB the Boltzmann constant
TAF the reduction of the Curie temperature due to the short-range
antiferromagnetic superexchange interaction among Mn spins
One can take TAF = 0 for Mn doped III–V compounds due to the forma-
tion of bound magnetic porlarons around close pairs of ionized Mn accep-
tors. By calculating ρS from a 6 × 6 k·p matrix with p-d exchange interaction,
the experimentally observed TC can be reproduced reasonably well without
any adjustable parameter. The magnitude of TC increases with the increase
of hole concentration p, because ρS increases with p. This carrier-induced
nature of ferromagnetism is consistent with the experimental observation
of higher TC of (Ga,Mn)As with higher conductivity [20]. The conductiv-
ity dependence of TC was also investigated by changing the conductivity
through the co-doping of donors and the growth condition, and was shown to
be consistent again with the carrier-induced mechanism of ferromagnetism
[21, 22]. It is known that there exist a number of interstitial Mn in (Ga,Mn)
As, which act as a double donor compensating holes, and whose spin couples
antiferromagnetically with a spin of substitutional Mn [23, 24]. The post-
growth annealing at ~200°C results in the diffusing out of interstitial Mn to
the surface, and thus increases p and magnetic moments [25, 26]. The sys-
tematic study utilizing the annealing effect confirmed also the monotonic
increase of TC with increasing conductivity [25]. The p-d Zener model pre-
dicts higher TC for materials with wider bandgap due to their larger ρS and
p-d exchange interaction, if similar values of xeff and p to those in (Ga,Mn)
As are realized in them [13]. The prediction encouraged many researchers to
work on related materials, and as a result, we now have a lot of information
to help understand the physics of magnetic semiconductors from compre-
hensive points of view [27].
The p-d Zener model explains also the experimental observation of the
strain-direction-dependent magnetic easy axis direction [28]. The magnetic
anisotropy is brought about by the anisotropic carrier-mediated exchange
interaction associated with the spin-orbit interaction (SOI) in the host mate-
rials. The model predicts that the direction and magnitude of the magnetic
anisotropy are hole-concentration dependent [13, 19, 29].
13.2.2 Electric Field Modulation of

the Curie Temperature
In this subsection, we introduce the electric field modulation of TC of fer-
romagnetic semiconductors, mainly focusing on the experimental results
obtained for III–V compounds. Typical ferromagnetic III–V semiconductors
contain a relatively large amount of holes, ranging from 1019 to 1021 cm−3.
Hence, the electric field screening length is the order of a nanometer, and
thus most of experiments were done on thin films with several nanometers
to observe sizable electric field effects.
The electric field modulation of TC was observed first in 2000 for (In,Mn)
As thin layers in a field effect transistor as a channel layer [6]. The magneti-
zation curves were probed by Hall resistance RHall thanks to the anomalous
Hall effect, as shown in schematically in Figure 13.1a. The increase of p by
the application of negative voltage VG to a gate electrode increases TC deter-
mined from the Arrott plots, and the decrease of p by positive VG decreases
TC. The effect is reversible with the change in p, and consistent with the p-d
Zener model. Similar observation of the modulation of TC has been reported
with various kinds of ferromagnetic semiconductors, such as MnGe [30],
(Cd,Mn)Te [31], (Ga,Mn)As [32], (Ga,Mn)Sb [33], Co doped TiO2 [34], and
so on. Among them, the effect in (Ga,Mn)As has been most extensively
investigated.
For (Ga,Mn)As thin layers, the electric field modulation of TC was con-
firmed by both magnetization and magnetotransport measurements [32, 35,
36]. The magnetization measurements revealed that the electric fields modu-
late the magnitude of spontaneous magnetic moment, in addition to TC due
to the change in the surface depletion layer thickness [35]. The Mn spins
in the depleted region do not participate in the ferromagnetic order due to
the absence of holes. The experimentally obtained relationship between TC
and p from transport measurements follows TC ∝ pγ with γ ~ 0.2 as shown
in Figure 13.1b. On the other hand, Equation 13.1 gives γ = 1/3 for the free-
electron model with quadratic band dispersion ε = ħ2k2/(2m*), where ε is the
kinetic energy, k the wavenumber, and m* the effective mass of the carriers,
V
V
S Gate metal
Gate insulator
D
Magnet
(a)
(b)
FIGURE 13.1 (a) Schematic of electric field device and configuration for trans-
port measurements. S, D, and G denote source, drain, and gate, respectively. (b)
Logarithm plots of experimentally determined hole concentration p dependence
of the Curie temperature TC for (Ga,Mn)As and (Ga,Mn)Sb, where p is modulated by
the application of electric fields. Numbers in parentheses are nominal Mn compo-
sition x and the exponent γ in TC ∝ pγ. Solid lines are linear fitting. (After Chang,
H.W. et al., Appl. Phys. Lett. 103, 142402, 2013. With permission.)
and ħ the Dirac constant. More realistic valence-band dispersions with

warps calculated by 6 × 6 k·p matrix give the value of γ between 0.6 and 0.8
depending on p. This indicates that Equation 13.1 should be adapted for thin
film cases with nonuniform hole distribution along the film normal (z direc-
tion) due to the presence of depletion layers at the interfaces of (Ga,Mn)As
and the short screening length of the electric field. The adapted expression
for TC with TAF = 0 is given by:

TC =  xeff N 0 S ( S + 1)β2 AFρsheet / 12kB 
∫ ( p ( z ) / p ) dz
sheet
(13.2)
2
∫
= TC 3D
( z )dz ∫ ( p ( z ) / psheet ) dz
where:
ρsheet is sheet density of states given by ρsheet = ∫ρSdz, p(z) is z-dependent
hole concentration, psheet is the sheet hole concentration given by
psheet = ∫p(z)dz
T C3 D is the Curie temperature for three-dimensional case given by
Equation 13.1 [35, 36]
The p(z) is calculated from the one-dimensional Poisson equation, in

which material parameters, the net concentration of Mn acceptors, the con-
centration of donor-like interface states at insulator/(Ga,Mn)As boundary,
and the concentration of As antisite donors in a low-temperature grown
GaAs buffer are treated as adjustable parameters. The calculated hole con-
centration p dependence of TC by Equation 13.2, where p is modulated by
applying electric fields, reproduces well the experimental observation of TC
∝ pγ with γ ~ 0.2 [36]. The electric field dependence of the saturation moments
is shown to be proportional to the thickness of a magnetically active layer
(thickness of (Ga,Mn)As subtracted by the calculated depletion layer thick-
ness), consistent with the nature of the carrier-induced ferromagnetism [35].
Because the value of γ is related to the hole distribution in the channel,
it is expected to depend on the channel materials in the electric field effect
devices through the material-dependent Fermi-level pinning position at the
interface. The pinning results in a surface depletion layer for p-GaAs, but an
accumulation layer for p-GaSb [37]. The relation of TC ∝ pγ with γ = 1.4–1.6
was obtained for (Ga,Mn)Sb from the electric field effect measurements
probed by RHall, as shown in Figure 13.1b. The larger value of γ than that of
(Ga,Mn)As was shown to be explained by the presence of the surface accu-
mulation layer in (Ga,Mn)Sb [38].
When the magnetic ions are located in an only partial channel region
along the film normal, one can change a degree of overlap of the magnetic
ions with the wave function of carriers by the application of electric fields
[39], which is expected to also result in the change in TC through the modu-
lation of the carrier-mediated exchange coupling [6]. This scheme was uti-
lized to demonstrate the electric field modulation of TC in heterostructures
of Mn delta-doped GaAs/p-(Al,Ga)As, as well as InAs/(In,Fe)As/InAs quan-
tum well [40, 41].
The modulation of TC of (Ga,Mn)As has been observed in a variety of
field effect structures, which include a back gate structure with a p-n junc-
tion of (Ga,Mn)As/n-GaAs for low-voltage operation [42], ferroelectric-
gate structure with nonvolatile functionality [43], and electric-double layer
structure to apply higher electric fields [44]. It also demonstrated that the
electric field formation of magnetic nanodots from a continuous (Ga,Mn)As
layer using a meshed gate electrode, in which nanowindow regions without
electric fields correspond to nanomagnets surrounded by the paramagnetic
region created by depleting holes under the meshed gate [45].

Magnetization Process and Anisotropy
As described above, in the vicinity of TC, one can change the magnetic phases
either ferromagnetic or paramagnetic, by applying electric fields without
changing temperature. In addition, one can change coercive force HC and
magnetic anisotropy well below TC.
The electric field control of HC was first demonstrated for (In,Mn)As
channels with perpendicular magnetic easy axis in an electric field effect
structure [46]. The increase of p was found to result in the increase in HC

below TC. The same tendency was observed for (Ga,Mn)As [32]. The domain
structure in (In,Mn)As, observed by magneto-optical Kerr microscope
(MOKE), indicated that the magnetization process takes place through
domain nucleation and its expansion [47], consistent with the magnetiza-
tion process from the virgin state [46]. In addition, the dependence of HC on
temperature T and electric field E was shown to follow the scaling relation-
ship of HC ∝ [1 − (T/TC*)]η with η ~ 2, where TC* is defined as the temperature
at which HC becomes zero (Figure 13.2) [47]. These results suggest that the
modulation of HC is related to the modulation of the exchange stiffness con-
stant AS. The electric field modulation of HC can be used for electric field-
assisted magnetization switching; the reduction in HC enables us to induce
the magnetization reversal by applying an electric field pulse in a small
external magnetic field [46].
The presence of the SOI in ferromagnetic materials manifests itself as
the presence of magnetic anisotropy. According to the p-d Zener model,
the magnetic anisotropy in (Ga,Mn)As is expected to depend on the strain
(direction and magnitude), hole concentration, and the magnitude of the
magnetization through SOI in the valence band. The expectation was cor-
roborated by the experimental observation, the change of direction of the
magnetic easy axis, either in-plane or perpendicular, by changing the direc-
tion of strain, annealing condition (thus hole concentration and magnetiza-
tion), or temperature (thus magnetization) [28, 48, 49]. Hence, one expects
that one can control the direction and magnitude of magnetic easy axis by
applying electric fields.
The electric field modulation of the magnitude of the perpendicular
magnetic anisotropy in (Ga,Mn)As with in-plane easy axis was detected
through the anisotropic magnetoresistance (AMR) effect under a perpen-
dicular magnetic field [50]. The saturation magnetic field of the AMR was
shown to increase with increasing p by applying a negative electric field. The
result indicates that the in-plane easiness increases with increasing p, and
is consistent qualitatively with the description of the p-d Zener model for
FIGURE 13.2 Dependence of coecivities HC on the reduced temperature T/T * as

a function of applied electric fields E for a 5-nm-thick In0.937 Mn0.063As, where TC* is
defined as the temperature at which HC becomes zero. Line is guide for the eye.
(After Chiba, D. et al., J. Phys. D: Appl. Phys. 39, R215, 2006. With permission.)
(Ga,Mn)As with compressive strain. The AMR effect is the phenomenon in

which the resistance depends on a relative angle between the magnetiza-
tion and current as a result of the presence of SOI. Clear AMR effect was
observed for (Ga,Mn)As [51], and its strain-dependent sign was explained by
the Boltzmann transport theory with the p-d Zener model [52, 53].
The electric field modulation of the in-plane magnetic anisotropy was
also observed through the transverse component of AMR effect [50], some-
times referred as the planar Hall effect (PHE) [54]. The application of E
ranging from −4 to 4 MV/cm to a 4-nm-thick (Ga,Mn)As through a gate
insulator modulates the in-plane magnetic field angle dependence of the
planar Hall resistance, reflecting the modulation of the in-plane anisotropy.
The (Ga,Mn)As possesses relatively complicated in-plane magnetic anisot-
ropies; fourfold in-plane easy axis along 〈100〉, as well as two uniaxial easy
axes along [110] or [1‒10] and [100] or [010] [55]. The analysis of the angular
dependence of the planar Hall resistance indicated that the sign and mag-
nitude of the uniaxial anisotropy along [110] is modulated, and results in
the change in in-plane magnetization direction by ~10° by applying E in the
absence of an external magnetic field [50]. The origin of the two uniaxial
easy axes remains to be elucidated: one possible explanation for the [110]
anisotropy is the formation of Mn dimers along [1‒10] orientation during
epitaxial growth [56]. The theoretical work indicated that the anisotropic
Mn distribution can serve as effective shear strain [56], and thus results in
the carrier-concentration dependent uniaxial anisotropy along [110] orien-
tation [57].
The control of magnetic anisotropies by external means provides new
schemes for magnetization reversal; the reversal through ratchet-type
motion of magnetization by rotating the minima in the potential land-
scape, and the reversal through the magnetization precession induced by
the abrupt change of the anisotropy direction [58–60]. Electric field-induced
stochastic magnetization switching was observed for (Ga,Mn)As, in which
the magnetocrystalline anisotropy energy barrier was reduced by E to switch
the magnetization through thermal fluctuation of magnetization [61].

Other Magnetic Parameters
Because (Ga,Mn)As is in the vicinity of metal-insulator transition (MIT)
[20], it contains two types of regions; one is populated by extended holes and
the other by localized holes (two fluids model) [62]. The region populated
by extended holes exhibits ferromagnetism, whereas that by localized holes
exhibits superparamagnetic-like behavior. The presence of two-phase is hard
to detect by transport measurements due to much higher conductivity in the
ferromagnetic region than in the superparamagnetic region. However, it can
be detected by history-dependent remanent magnetization measurements
[35]. It was shown that the portion of the superparamagnetic-like region can
be changed by the application of the electric fields; when p in a (Ga,Mn)As
layer is decreased, the portion increases [35].
Ferromagnetic resonance (FMR) spectra (resonance fields HR and spec-

tral linewidths ΔH) of (Ga,Mn)As were shown to be described well by the
Landau–Lifshitz–Gilbert (LLG) equation with perpendicular and in-plane
magnetic anisotropies, without considering extrinsic contributions such as
magnetic anisotropy dispersion and two-magnon scattering [63]. The FMR
spectra as a function of external magnetic-field angles θH were measured for
a thin metallic (Ga,Mn)As film in a capacitor structure under E between
−4 and 4 MV/cm [64]. Hole concentration p ~ 3.2 × 1020 cm−3 at E = 0 was
modulated by ~25%, and thus the resistivity ρ ~ 9 mΩcm at E = 0 by ~25% by
the application of E. The analysis of the angular dependence of HR indicated
that the anisotropy of this particular (Ga,Mn)As is nearly independent of E.
On the other hand, the analysis of the angular dependence of ΔH indicated
that the damping constant α ~ 0.054 at E = 0 is modulated by ~10% by E; the
increase in p results in the decrease of α. The theoretical work based on the
relaxation mechanism due to the SOI predicted the carrier-concentration
dependent α [65], however, the observed tendency was opposite to the predic-
tion. To understand the discrepancy, the relation among α, ρ, and magnetic
properties was investigated by changing p through E and the post annealing.
As described above, the magnetic disorder defined as the ratio MSP/Mtot of
the superparamagnetic-like magnetization component to total magnetiza-
tion, increases with decreasing p. A clear correlation between α and MSP/Mtot
was obtained as shown in Figure 13.3, suggesting strongly that the electric
field modulation of α in (Ga,Mn)As is determined by the magnetic disorder
rather than the SOI.
(Ga,Mn)As usually shows a positive anomalous Hall coefficient [66].
However, it was found that thin (Ga,Mn)As with relatively high conductivity
σ > 200 S/cm sometimes shows a negative anomalous Hall coefficient. For
such (Ga,Mn)As, the sign of the coefficient can be changed by the application
of E [67]. The origin of the behavior is not understood yet [68], but is probably
FIGURE 13.3 Dependence of damping constant α of (Ga,Mn)As on the ratio

MSP/Mtot of superparamagnetic-like component to total magnetization. Stars are
the result for a 4-nm-thick (Ga,Mn)As with Mn composition x of 0.13 in an electric
field effect structure, in which the resistivity is changed by the application of elec-
tric fields. Triangles and circles are results for a 20-nm-thick metallic (Ga,Mn)As with
x = 0.068 and a 200-nm-thick insulating (Ga,Mn)As with x = 0.075, respectively, for
which the resistivity is changed by post annealing. (After Chen, L. et al., Phys. Rev.
Lett. 115, 057204, 2015. With permission.)
related to the interplay between carrier confinement and topology of band

structure.
The electric field modulation of the magnetic domain wall (DW) veloc-
ity was also reported for (In,Mn)As [69].
13.3 ELECTRIC FIELD EFFECTS IN

FERROMAGNETIC METALS
Stimulated by studies on magnetic semiconductors, researchers began to
investigate electric field effects in metals. The first observation of an electric
field effect on magnetism in metals was reported for thin layers of FePt and
FePd, where HC was modulated at room temperature using an electric double
layer [70]. Prior to this study, it was assumed that a large electric field effect
in metals would be difficult to observe owing to the associated short screen-
ing length. Subsequent theoretical work showed that electric field effects on
surface magnetization and surface magnetocrystalline anisotropy can be
sizable, even for metal ferromagnets such as Fe, Co, and Ni, because of the
spin-dependent screening of the electric field [71]. The studies encouraged
many to investigate electric field effects in a number of materials, including
Fe(Co)/MgO system, and led to the observation of the electric field-induced
change of magnetic anisotropy at room temperature, whose origin is attrib-
uted to the modulation of interfacial magnetic anisotropy [72, 73]. The effect
is now being utilized to induce the magnetization switching [74, 75].
13.3.1 Electric Field Modulation of Magnetic

Anisotropy and Other Properties
Experimental works indicated the presence of the interfacial perpendicu-
lar magnetic anisotropy at a Fe/MgO, Co/AlO, and CoFeB/MgO interfaces
[72, 73, 76–79]. These findings were followed by the fabrication of nanoscale
CoFeB/MgO magnetic tunnel junctions (MTJs) with a perpendicular easy
axis [80]. For the Fe/MgO junction, theoretical studies showed that the
hybridization of Fe 3d and O 2p orbitals modifies the electronics structure
at the interface, and the resultant occupancy of electrons among 3d orbit-
als with different magnetic quantum numbers determines the direction and
magnitude of magnetic anisotropy through the SOI [81, 82]. This theoretical
description was shown to be consistent with the results of x-ray magnetic
circular dichroism (XMCD) measurement [83]. The formation of Fe(Co)-O
bonds was also confirmed experimentally for the CoFeB/MgO interface
[84]. Application of an electric field alters the Fermi energy position at the
interface, which changes the occupancy of the orbitals and hence the inter-
face magnetic anisotropy [81, 82]. The results of x-ray absorption (XAS) and
XMCD for Fe/MgO suggested that an electric field can reversibly change the
degree of oxidation of Fe at the interface, which may influence the magne-
tism [85]. However, XAS measurement of Fe/MgO revealed the presence of
the electric field effect on the anisotropy without redox, indicating that the
effect is associated with the electron doping and/or redistribution induced
by the electric field at the interface [86]. Some experimental works showed
that the efficiency and the polarity of the effect depend on the composition
of materials and stack structure, as well as deposition and annealing con-
dition [87–91]. Theoretical works predicted that they depend on the coun-
ter-interface material, strain, and the Rashba SOI [92–95]. The electric field
modulation ratio ξ is defined as ξ = dKefft/dE, where Keff is an effective per-
pendicular magnetic anisotropy density, and t is the thickness of a ferromag-
netic material. Typical reported values of ξ for CoFeB/MgO range from ~10
to ~100 fJ/Vm, which is consistent with those obtained from the first prin-
ciples calculations for (Co)Fe/MgO (or vacuum) systems [81, 82]. The applied
electric field results in an atomic-scale inhomogeneous electric field at the
metal/insulator interface due to electrostatic screening in the metal, which
induces electric quadrupoles at the interface. The quadrupoles produce the
magnetic dipole moments, which change the magnetic anisotropy through
SOI. This mechanism based on the induced dipoles was confirmed by the
combination of XMCD experiments and first principles calculations on an
L10-FePt system [96]. The electric field-induced change in the anisotropy due
to the magnetic dipole moments could be large, but is largely compensated
by the anisotropy with the opposite sign, due to the electric field effect on the
orbital moments in the FePt system [96].
The modulation of TC and magnetic moments have also been observed
in ultrathin layers of Fe and Co at room temperature [97–99]. The observa-
tion may be related to the modulation of the exchange constant as indicated
by the first principles calculation [100, 101]. The magnetic phase transition
between ferromagnetic and antiferromagnetic phases was observed through
the structural phase transition between body-centered and face-centered
cubic phases induced by the application of the local electric field by the tip of
a scanning tunneling microscope [102].
Ferromagnetic resonance (FMR) spectra were measured for Ta/CoFeB/
MgO junctions with different CoFeB thickness t ranging from 1.4 to 1.8
nm as functions of E and external magnetic field angle θH measured from
the film normal [103]. As shown in Figure 13.4 for Co0.2Fe0.6B0.2 with t = 1.5
FIGURE 13.4 Ferromagnetic resonance spectra for a 1.5-nm-thick Co0.2Fe0.6B0.2

sandwiched by Ta and MgO as a function of applied electric fields E.
nm, the resonance field HR is shifted by the application of E. The analysis

of theangular dependence of the HR indicated that the first order effective
perpendicular magnetic anisotropy field HK1eff increases with decreasing t,
indicating its interfacial origin, whereas the second order anisotropy field
HK2 is virtually independent of E. The magnitude of HK1eff is modulated by
E, while that of HK2 is not. The magnitude of the modulation of the areal
anisotropy energy density does not depend virtually on t, indicating that the
electric field effect is the interfacial effect. The analysis of the FMR spectral
linewidth ΔH indicated that the damping constant α increases with decreas-
ing t, most probably due to the spin pumping at the Ta/CoFeB interface [104].
It was found that the FMR spectral linewidths ΔH can be also modulated by
E. The analysis of the angular dependence indicated that the magnitude of
α can be modulated by a few percent by E of 1 MV/cm for CoFeB/MgO with
thin CoFeB possessing perpendicular magnetic easy axis [103].
The electric field effect on magnetic domain structures was observed
by a magneto-optical Kerr effect microscopy. It was shown that CoFeB/
MgO junctions with perpendicular magnetic anisotropy show an in-plane
isotropic maze domain pattern at demagnetized state [105]. The applica-
tion of E modulates the domain periods DP from 1.5 to 2.0 μm by applying
E of ±1.1 MV/cm for the particular 1.8-nm-thick CoFeB used in the study
[106] (Figure 13.5). The electric field dependence of DP can be described
well by the existing model with the modulation of AS by a few percent in
addition to that of the magnetic anisotropy [107]. The electric field modula-
tion of DP was observed also for Co [108], and the modulation can be used
for magnetization reversal in a local area [109]. The modulation of AS was
also shown by the modulation of the distance between the resonance fields
between the Kittel and spin-wave modes in a nanoscale CoFeB/MgO MTJ
detected by homodyne technique [110].
Electric field effect can also be used to control DW motion under an
external magnetic field, as shown in Co wires. Here, the applied electric field
modifies the creep velocity of the DW by over an order of magnitude, with
changes most probably related to the modification of the magnetic anisotropy
[111, 112]. By using patterned double-gate insulators, the DW pinning could
FIGURE 13.5 Domain structures at demagnetized state for a 1.5-nm-thick

Co0.2Fe0.6B0.2 sandwiched by Ta and MgO under (a) negative and (b) positive electric
fields E. (After Dohi, T. et al., AIP Adv. 6, 075017, 2016. With permission.)
be controlled by an applied electric field, in which the electrode perimeter

acts as a barrier for the DW motion after the application of the negative elec-
tric field. Relatively large pinning fields of 65 mT can be removed by applying
a positive electric field. Once the electric field was applied, the DW pinning
was maintained at zero electric field for several days, suggesting that the
observed nonvolatile nature of the effect is related to the ionic transport and/
or charge trap in the insulator [112]. The theoretical work predicted that the
current-induced Walker breakdown can be suppressed by the application of
the electric field through the spin flexoelectric interaction, leading to a faster
maximum DW velocity [113]. There have also been many interesting topics
on the electric field control of the DW motion in multiferroic structures [114].
Electric field effects on antiferromagnetic materials and the modula-
tion of the anomalous Hall effect in a paramagnetic metal Pt were reported
[115–117].
13.3.2 Electric Field-Induced Magnetization Dynamics

The electron spin resonance and nuclear magnetic resonance can be induced
in nonmagnetic materials by applying a high-frequency electric field through
the modulation of the SOI or quadrupole moments [118, 119]. It was shown
that FMR in ferromagnetic metals can be excited also by the application
of rf electric fields through the rf-modulation of the magnetic anisotropy
[120, 121]. The FMR is usually excited in a free layer in an MTJ severed as
a pseudocapacitor structure fabricated on a coplanar waveguide by apply-
ing rf electric fields. The FMR detection is done by homodyne technique, in
which rectified dc voltage Vdc is measured thanks to the tunnel magnetore-
sistance (TMR) effect (see Chapters 11–13, Volume 1), which results in the
change in junction resistance R synchrony with the input rf signal [122]. The
homodyne-detected FMR is a powerful tool to determine magnetic param-
eters, such as anisotropy constant, its electric field modulation ratio, and
damping constant, of nanomagnets in MTJs [123–126]. The excitation of the
FMR by electric fields are confirmed from its spectral shape and magnetic
field angle dependence, which are different from those for the FMR excited
by spin-transfer torque (STT) [120]. One can observe electric field-induced
non-linear FMR [127], in which spectral shape is largely distorted from anti-
symmetric Lorentzian due to the foldover effect by increasing input rf power
Prf as shown in Figure 13.6 [128]. From the rf power dependence of the FMR
spectra, one can determine the value of ξ. The behavior was reproduced by
micromagnetic simulations [127]. The electric field effect can also excite
spin-wave resonance in nanoscale MTJs [110, 129]. The electric field-induced
FMR was shown to be used as a microwave detector with higher sensitivity
than that induced by STT [130].
The spin-wave propagation in magnetic insulator and metals can be
modulated by the application of electric fields through the modification of
the SOI or magnetic anisotropy [131, 132]. The analysis of spin-wave reso-
nance frequency in an Au/Fe/MgO structure showed that the interfacial
Dzyaloshinskii–Moriya interaction (DMI) is modulated by the electric
FIGURE 13.6 Normalized homodyne-detected ferromagnetic resonance spectra

Vdc/Prf for a CoFeB/MgO-based magnetic tunnel junction with perpendicular mag-
netic easy axis under in-plane magnetic field of 0.1 T as a function of input rf power
Prf. (After Hirayama, E. et al., Appl. Phys. Lett. 107, 132404, 2015. With permission.)
fields in addition to the modulation of the interfacial anisotropy [133]. The

presence of the DMI sometimes stabilizes a topologically non-trivial spin
texture, skyrmion [134]. Micromagnetic simulation showed that one can
create and annihilate a skyrmion as well as switch its core polarity in a
magnetic nanodisk by applying E through the change in magnetic anisot-
ropy [135].
13.3.3 Electric Field-Induced Magnetization Switching

Electric field-induced magnetization switching that involves interface
anisotropy was demonstrated in MTJs, which serve as a pseudocapacitor.
The magnetization configuration is detected by the TMR effect. So far, two
schemes have been demonstrated: one utilizes the electric field-induced
change of coercivity [136], and is essentially the same as the electric field-
assisted magnetization switching demonstrated for a single layer of a ferro-
magnetic semiconductor and ferromagnetic metal in a field effect structure
[46, 137]; the other utilizes magnetization precession induced by a temporal
change in the direction of magnetic anisotropy [74, 75]. The first scheme
achieves unipolar switching under a constant magnetic field, and was dem-
onstrated for a CoFeB/MgO MTJ in a small external magnetic field [136].
The field is needed to ensure that the potential wells—otherwise degenerate
bistable magnetic states—are asymmetric, because the electric field does not
break time-reversal symmetry. The application of an electric field reduces
the coercivity of the CoFeB electrode, and switches the magnetization direc-
tion to align it with an applied magnetic field. The magnetization direction
is detected by the junction resistance. STT switching was utilized to induce
switching in the opposite direction; without STT, the polarity of the external
magnetic field must be reversed to induce the electric field-assisted switch-
ing in the opposite direction.
In the second scheme, an initial study used a FeCo/MgO/Fe MTJ
with an in-plane magnetic easy axis [74] was followed by a CoFeB/MgO/
CoFeB with a perpendicular magnetic easy axis [75]. For the case of an
MTJ with a perpendicular easy axis, the application of an electric field
pulse temporarily aligns the easy axis in-plane, thus inducing magneti-
zation precession in the free layer about the new easy axis. Applying an
external constant magnetic field assists the change in direction for easy
axis, fixes the precessional axis and changes the precessional period (the
inverse of the Larmor frequency) as shown in Figure 13.7. The precessional
period is determined by the magnitude of the effective magnetic field (sum
of anisotropy, stray, and external fields). Magnetization switching takes
place when the pulse duration is a half-integer multiple of the precession
period. The obtained switching probabilities therefore show an oscillatory
behavior with the pulse duration, which requires fine control of the pulse
duration for switching. The amplitude of oscillation decays relatively fast
as the pulse duration increases, which is due to thermal agitation enhanced
by the distribution of effective fields [138]. The effect of thermal agitation
results in randomization of the precessional phase, as confirmed by the
real-time observation of the precession in Figure 13.8 [139]. The real-time
observation was conducted by measuring the transmission voltage V T dur-
ing the application of the bias voltage, which induces the precession in a
free layer in an MTJ. The random phase shift becomes larger with time,
and thus results in rapid decay in the probability. More reliable switching
is possible by combining both the electric field-induced and STT-induced
switching, with two successive voltage pulse applied to utilize the advan-
tages of the two schemes [140]. The first pulse, VE, induces the magneti-
zation precession by the electric field effect, and the second pulse, VSTT,
determines the final magnetization direction by STT. This combined
scheme achieves higher probabilities than those of STT-induced or electric
field-induced switching alone at shorter pulse duration regime, and gives it
potential for applications.
Easy direction
Free layer
Barrier layer
Reference layer
H
Hz
qH
Hy
FIGURE 13.7 Schematics of operation scheme of electric field induced magne-

tization switching in a magnetic tunnel junction. Upper arrows show the direction
of magnetic easy direction in a free layer. The application of an electric field pulse
induces magnetization precession through the temporal change of the easy axis
direction. Magnetization switching takes place when the pulse duration is a half-
integer multiple of the precession period. The resultant direction is the same as the
initial direction when the duration is an integer multiple of the period. A perpen-
dicular component Hz of an external magnetic field H compensates a perpendicu-
lar component of a stray field from a reference layer, and an in-plane component
Hy determines the precessional axis.
FIGURE 13.8 Time evolution of transmitted voltage V T from a CoFeB/MgO-based

magnetic tunnel junction under bias voltage. Upper curve and lower curves are V T
for parallel and antiparallel magnetization configuration, respectively. Two middle
oscillatory curves are for situations with magnetization precession from parallel
and antiparallel magnetization configuration. (After Kanai, S. et al., Jpn. J. Appl.
Phys. 58, 0802A3, 2017. With permission.)
Because the MTJ is a pseudocapacitor, tunnel current flows through the

barrier layer during the application of electric fields. As a result, the switch-
ing energy is dominated by the Joule heating rather than the charging energy
of the capacitor, although the STT is not important for the switching. In
order to reduce the Joule energy, the electric field-induced switching was
investigated for CoFeB/MgO-based MTJs with high junction resistance by
increasing the MgO barrier-layer thickness [141, 142]. It was shown that the
switching energy can be reduced to ~6 fJ/bit, which is the smallest reported
value for the magnetization switching in MTJs, but is still dominated by the
Joule heating. To reduce the switching energy further, one needs to improve
the electric field modulation ratio ξ.
13.4 CONCLUSION
Curiosity about the interplay between electricity and magnetism brought
about initial study of the electric field effect on magnetism. Nowadays, the
effect is also drawing attention from a technological viewpoint, because
of its compatibility with semiconductor device technology, and capability
for realizing nonvolatile devices with low power consumption. To put the
electric field devices to practical use, a number of challenges must be over-
come; for instance, one needs to enhance the effect to reduce the opera-
tional energy, and improve the switching reliability, as well as to ensure
thermal stability at small dimensions. Meeting these challenges will have
a big impact on future nanoelectronics. In addition, the effect is vital as a
powerful probe of condensed matter physics to explore the mechanism of
cooperation phenomena.
ACKNOWLEDGMENTS
The authors are grateful for many collaborations, especially with T. Dietl,
D. Chiba, and L. Chen on ferromagnetic semiconductors, as well as with
References 535
S. Kanai, A. Okada, E. Hirayama, and T. Dohi on ferromagnetic metals. The

authors acknowledge the support from MEXT, a Grant-in-Aid for Scientific
Research (No. 26103002), and R&D Project for ICT Key Technology.
REFERENCES
1. Y. Tokura, S. Seki, and N. Nagaosa, Multiferroics of spin origin, Rep. Prog. Phys.
77, 075501 (2014).
2. F. Matsukura, Y. Tokura, and H. Ohno, Nat. Nanotech. 10, 209–220 (2015).
3. K. Roy, Ultra-low-energy straintronics using multiferroic composites, SPIN 3,
1330003 (2013).
4. M. Barangi and P. Mazumder, Straintronics, IEEE Nanotech. Magazine, 15–24
(2015).
5. S. Methfessel, Potential application of magnetic rare earth compounds, IEEE
Trans. Magn. 1, 144–155 (1965).
6. H. Ohno, D. Chiba, F. Matsukura et al., Electric field control of ferromagne-
tism, Nature 408, 944–946 (2000).
7. F. Matsukura, H. Ohno, A. Shen, and Y. Sugawara, Transport properties and
origin of ferromagnetism in (Ga,Mn)As, Phys. Rev. B 57, R2037–R2040 (1998).
8. K. W. Edmonds, K. Y. Wang, R. P. Campion et al., High-Curie-temperature
Ga1-xMn xAs obtained by resistance-monitored annealing, Appl. Phys. Lett. 81,
4991–4993 (2002).
9. H. Ohno, H. Munekata, T. Penny, S. von Molnár, and L. L. Chang,
Magnetotransport properties of p-type (In,Mn)As diluted magnetic III-V
semiconductors, Phys. Rev. Lett. 68, 2664–2667 (1992).
10. H. Ohno, A. Shen, F. Matsukura et al., (Ga,Mn)As: A new diluted magnetic
semiconductor based on GaAs, Appl. Phys. Lett. 69, 363–365 (1996).
11. A. Haury, A. Wasiela, A. Arnoult et al., Observation of ferromagnetic tran-
sition induced by two-dimensional hole gas in modulation-doped CdMnTe
quantum wells, Phys. Rev. Lett. 79, 511–514 (1997).
12. D. Ferrand, J. Cibert, A. Wasiela et al., Carrier-induced ferromagnetism in
p-Zn1-xMn xTe, Phys. Rev. B 63, 085201 (2001).
1019–1022 (2000).
14. M. Tanaka, S. Ohya, and P. N. Hai, Recent progress in III–V based ferromag-
netic semiconductors: Band structure, Fermi level, and tunneling transport,
Appl. Phys. Rev. 1, 011102 (2014).
15. S. Souma. L. Chen. R. Oszwałdowski et al., Fermi level position, Coulomb gap,
and Dresselhaus splitting in (Ga,Mn)As, Sci. Rep. 6, 277266 (2016).
16. J. Kanski, L. Ilver, K. Karlsson et al., Electronic structure of (Ga,Mn)As revis-
ited, New J. Phys. 19, 023006 (2017).
17. J. H. Park, S. K. Kwon, and B. I. Min, Electronic structures of III–V based ferromag-
netic semiconductors; Half-metallic phase, Physica B 281&282, 703–704 (2000).
18. M. Toyoda, H. Akai, K. Sato, and H. Katayama-Yoshida, Curie temperature of
GaMnN and GaMnAs from LDA-SIC electronic structure calculations, Phys.
Stat. Sol. (c) 3, 4155–4159 (2006).
20. A. Oiwa, S. Katsumoto, A. Endo et al., Nonmetal-metal-nonmetal transi-
tion and large negative magnetoresistance in (Ga,Mn)As/GaAs., Solid State
Commun. 103, 209–213 (1997).
21. Y. Satoh, D. Okazawa, A. Nagashima, and J. Yoshino, Carrier concentration
dependence of electronic and magnetic properties of Sn-doped GaMnAs,
Physica E, 10, 196–200 (2001).
22. H. Shimizu, T. Hayashi, T. Nishinaga, and M. Tanaka, Magnetic and transport

properties of III–V based magnetic semiconductor (Ga,Mn)As: Growth condi-
tion dependence, Appl. Phys. Lett. 74, 398–400 (1999).
23. K. M. Yu, W. Walukiewicz, T. Wojtowicz et al., Effect of the location of Mn
sites in ferromagnetic Ga1-xMn xAs on its Curie temperature, Phys. Rev B 65,
201303 (R) (2002).
24. J. Blinowski and P. Kacman, Spin interactions of interstitial Mn ions in ferro-
magnetic GaMnAs, Phys. Rev. B 67, 121204 (R) (2003).
25. K. W. Edmonds, P. Bogusławski, K. Y. Wang et al., Mn interstitial diffusion in
(Ga,Mn)As, Phys. Rev. Lett. 92, 037201 (2004).
26. K. W. Wang, K. W. Edmonds, R. P. Campion et al., Influence of the Mn intersti-
tial on the magnetic and transport properties of (Ga,Mn)As, J. Appl. Phys. 95,
6512–6514 (2004).
27. T. Dietl, A ten-year perspective on dilute magnetic semiconductors and oxides,
Nat. Mater. 9, 965–974.
28. A. Shen, H. Ohno, F. Matsukura et al., Epitaxy of (Ga,Mn)As, a new diluted
magnetic semiconductor based on GaAs, J. Cryst. Growth 175/176, 1069–1074
(1997).
30. Y. D. Park, A. T. Hanbicki, S. C. Erwin et al., A group IV ferromagnetic semi-
conductor; MnxGe1-x, Science 295, 651–654 (2002).
31. H. Boulari, P. Kossacki, M. Bertolini et al., Light and electric field control of
ferromagnetism in magnetic quantum structures, Phys. Rev. Lett. 88, 207204
(2002).
32. D. Chiba, F. Matsukura, and H. Ohno, Electric field control of ferromagnetism
in (Ga,Mn)As, Appl. Phys. Lett. 89, 162505 (2006).
33. Y. Nishitani, M. Endo, F. Matsukura, and H. Ohno, Magnetic anisotropy in a
ferromagnetic (Ga,Mn)Sb thin film, Physica E 42, 2681–2684 (2010).
34. Y. Yamada, K. Ueno, T. Fukumura et al., Electrically induced ferromagnetism
at room temperature in cobalt-doped titanium dioxide, Science 332, 1065–
1067 (2011).
35. M. Sawicki, D. Chiba, A. Korbecka et al., Experimental probing of the interplay
(2010).
36. Y. Nishitani, D. Chiba, M. Endo et al., Curie temperature versus hole concen-
tration in field-effect structures of Ga1-xMn xAs, Phys. Rev. B 81, 045208 (2010).
37. H. Hasegawa, H. Ohno, and T. Sawada, Hybrid orbital energy for heterojunc-
tion band lineup, Jpn. J. Appl. Phys. Part 2 25, L265–L268 (1986).
38. H. W. Chang, S. Akita, F. Matsukura, and H. Ohno, Hole concentration depen-
dence of the Curie temperature of (Ga,Mn)Sb in a field-effect structure, Appl.
Phys. Lett. 103, 142402 (2013).
39. B. Lee, T. Jungwirth, and A. H. MacDonald, Theory of ferromagnetism in
diluted magnetic semiconductor quantum wells, Phys. Rev. B 61, 15606–15609
(2000).
40. A. M. Nazmul, S. Kobayashi, S. Sugahara, and M. Tanaka, Electrical and opti-
cal control of ferromagnetism in III–V semiconductor heterostructures at
high temperature (~100 K), Jpn. J. Appl. Phys. Part 2 43, L233–L236 (2004).
41. L. D. Anh, P. N. Hai, Y. Kasahara, Y. Iwasa, and M. Tanaka, Modulation of fer-
romagnetism (In,Fe)As quantum wells via electrically controlled deformation
of the electron wave functions, Phys. Rev. B 92, 161201(R) (2015).
42. K. Olejnik, M. H. S. Owen, V. Novák et al., Enhanced annealing, high Curie
temperature, and low-voltage gating in (Ga,Mn)As; A surface oxide control
study, Phys. Rev. B 78, 054403 (2008).
43. I. Stolichnov, S. W. E. Riester, H. J. Trodahl et al., Non-volatile ferroelectric
control of ferromagnetism in (Ga,Mn)As, Nat. Mater. 7, 464–467 (2008).
References 537
44. M. Endo, D. Chiba, H. Shimotani, F. Matsukura, Y. Iwasa, and H. Ohno,

Electric double layer transistor with a (Ga,Mn)As channel, Appl. Phys. Lett.
96, 022515 (2000).
45. D. Chiba, F. Matsukura, and H. Ohno, Electrically defined ferromagnetic
nanodots, Nano Lett. 10, 4505–4508 (2010).
46. D. Chiba, M. Yamanouchi, F. Matsukura, and H. Ohno, Electrical manipula-
tion of magnetization reversal in a ferromagnetic semiconductor, Science 301,
943–945 (2003).
47. D. Chiba, F. Matsukura, and H. Ohno, Electrical magnetization reversal in fer-
romagnetic III–V semiconductors, J. Phys. D: Appl. Phys. 39, R215–R225 (2006).
48. K. Takamura, F. Matsukura, D. Chiba, and H. Ohno, Magnetic properties of
(Al,Ga,Mn)As, Appl. Phys. Lett. 81, 2590–2592 (2002).
49. M. Sawicki, F. Matsukura, A. Idziaszek et al., Temperature dependent mag-
netic anisotropy in (Ga,Mn)As layers, Phys. Rev. B 70, 245325 (2004).
Magnetization vector manipulation by electric fields, Nature 455, 515–518
(2008).
51. D. V. Baxter, D. Ruzumetov, J. Scherschiligh et al., Anisotropic magnetoresis-
tance in Ga1-xMn xAs, Phys. Rev. B 65, 212407 (2002).
52. T. Jungwirth, M. Abolfath, J. Sinova, J. Kučera, and A. H. MacDonald,
Boltzmann theory of engineered anisotropic magnetoresistance in (Ga,Mn)
As, Appl. Phys. Lett. 81, 4029–4031 (2002).
53. F. Matsukura, M. Sawicki, T. Dietl, D. Chiba, and H. Ohno, Magnetotransport
properties of metallic (Ga,Mn)As films with compressive and tensile strain,
Physica E 21, 1032–1036 (2004).
54. H. X. Tang, R. K. Kawakami, D. D. Awschalom, and M. L. Roukes, Giant planar
Hall effect in epitaxial (Ga,Mn)As devices, Phys. Rev. Lett. 90, 107201 (2003).
55. K. Pappert, S. Hümpfner, J. Wenisch et al., Transport characterization of the
magnetic anisotropy of (Ga,Mn)As, Appl. Phys. Lett. 90, 062109 (2007).
56. M. Birowska, C. Śliwa, J. A. Majewski, and T. Dietl, Origin of bulk uniaxial
anisotropy in zinc-blende dilute magnetic semiconductors, Phys. Rev. Lett.
108, 237203 (2012).
57. M. Sawicki, K.-Y. Wang, K. W. Edmonds et al., In-plane uniaxial anisotropy
rotations in (Ga,Mn)As thin films, Phys. Rev. B 71, 121302(R) (2005).
58. Y. Iwasaki, Stress-driven magnetization reversal in magnetostrictive films with
in-plane magnetocrystalline anisotropy, J. Magn. Magn. Mater. 240, 395–397
(2002).
59. D. Chiba, Y. Nakatani, F. Matsukura, and H. Ohno, Simulation of magneti-
zation switching by electric field manipulation of magnetic anisotropy, Appl.
Phys. Lett. 96, 192506 (2010).
60. P. Balestriere, T. Devolder, J. Wunderlich, and C. Chappert, Electric field
induced anisotropy modification in (Ga,Mn)As: A strategy for the precessional
switching of the magnetization, Appl. Phys. Lett. 96, 142504 (2010).
61. D. Chiba, T. Ono, F. Matsukura, and H. Ohno, Electric field control of thermal
stability and magnetization switching in (Ga,Mn)As, Appl. Phys. Lett. 103,
142418 (2013).
62. T. Dietl, Interplay between carrier localization and magnetism in diluted mag-
netic and ferromagnetic semiconductors, J. Phys. Soc. Jpn. 77, 031005 (2008).
63. L. Chen, F. Matsukura, and H. Ohno, Direct-current voltages in (Ga,Mn)As
structures induced by ferromagnetic resonance, Nat. Commun. 4, 2055 (2013).
64. L. Chen, F. Matsukura, and H. Ohno, Electric field modulation of damping
constant in a ferromagnetic semiconductor (Ga,Mn)As, Phys. Rev. Lett. 115,
057204 (2015).
65. J. Sinova, T. Jungwirth, X. Liu et al., Magnetization relaxation in (Ga,Mn)As
ferromagnetic semiconductors, Phys. Rev. B 69, 085209 (2004).
66. T. Jungwirth, Q. Niu, and A. H. MacDonald, Anomalous Hall effect in ferro-
magnetic semiconductors, Phys. Rev. Lett. 88, 207208 (2002).
67. D. Chiba A. Werpachowska, M. Endo et al., Anomalous Hall effect in field-

effect structures of (Ga,Mn)As, Phys. Rev. Lett. 104, 106601 (2010).
68. A. Werpachowska and T. Dietl, Effect of inversion asymmetry on the intrinsic
anomalous Hall effect in ferromagnetic (Ga,Mn)As, Phys. Rev. B 81, 155205
(2010).
69. M. Yamanouchi, D. Chiba, F. Matsukura, and H. Ohno, Current-assisted
domain wall motion in ferromagnetic semiconductors, Jpn. J. Appl. Phys. 45,
3854–3859 (2006).
70. M. Weisheit, S. Fähler, A. Marty, Y. Souche, C. Poinsignon, and D. Givord,
Electric field-induced modification of magnetism in thin-film ferromagnets,
Science 315, 349–351 (2007).
71. C.-G. Duan, J. P. Velev, R. F. Sabirianov et al., Surface magnetoelectric effect in
ferromagnetic metal films, Phys. Rev. Lett. 101, 137201 (2008).
72. T. Maruyama, Y. Shiota, T. Nozakim et al., Large voltage-induced magnetic
anisotropy change in a few atomic layers of iron, Nat. Nanotech. 4, 158–161
(2009).
73. M. Endo, S. Kanai, S. Ikeda, F. Matsukura, and H. Ohno, Electric field effects
on thickness dependent magnetic anisotropy of sputtered MgO/Co40Fe40B20/
Ta structures, Appl. Phys. Lett. 96, 212503 (2010).
74. Y. Shiota, T. Nozaki, F. Bonell, S. Murakami, T. Shinjo, and Y. Suzuki, Induction
of coherent magnetization switching in a few atomic layers of FeCo using volt-
age pulses, Nat. Mater. 11, 39–43 (2012).
75. S. Kanai, M. Yamanouchi, S. Ikeda, Y. Nakatani, F. Matsukura, and H.
Ohno, Electric field-induced magnetization reversal in perpendicular-
anisotropy CoFeB-MgO magnetic tunnel junction, Appl. Phys. Lett. 101,
122403 (2012).
76. T. Shinjo, S. Hine, and T. Takada, Mössbauer spectra of ultrathin Fe films
coated by MgO, J. Phys.-Paris 40, C2-86–87 (1979).
77. M. Klaua, D. Ulmann, J. Barthel et al., Growth, structure, electronic, and mag-
netic properties of MgO/Fe(001) bilayers and Fe/MgO/Fe(001) trilayers, Phys.
Rev. B 64, 134411 (2001).
78. A. Manchon, C. Ducret, L. Lombard et al., Analysis of oxygen induced anisot-
ropy crossover in Pt/Co/MOx trilayers, J. Appl. Phys. 104, 043914 (2008).
79. S. Yakata, H. Kubota, Y. Suzuki et al., Influence of perpendicular magnetic
anisotropy on spin-transfer switching current in CoFeB/MgO/CoFeB mag-
netic tunnel junctions, J. Appl. Phys. 105, 07D131 (2009).
80. S. Ikeda, K. Miura, H. Yamamoto et al., A perpendicular-anisotropy CoFeB-
MgO magnetic tunnel junction, Nat. Mater. 9, 721–724 (2010).
81. K. Nakamura, R. Shimabukuro, Y. Fujiwara, T. Akiyama, T. Ito, and A. Freeman,
Giant modification of the magnetocrystalline anisotropy in transition-metal
monolayers by an external electric field, Phys. Rev. Lett. 102, 187201 (2009).
82. R. Shimabukuro, K. Nakamura, T. Akiyama, and T. Ito, Electric field effects on
magnetocrystalline anisotropy in ferromagnetic Fe monolayers, Physica E 42,
1014–1017 (2010).
83. S. Kanai, M. Tsujikawa, Y. Miura, M. Shirai, F. Matsukura, and H. Ohno,
Magnetic anisotropy in Ta/CoFeB/MgO investigated by x-ray magnetic cir-
cular dichroism and first-principles calculation, Appl. Phys. Lett. 105, 222409
(2014).
84. Z. Wang, M. Saito, K. P. McKenna et al., Atomic-scale structure and local
chemistry of CoFeB-magnetic tunnel junctions, Nano Lett. 16, 1530–1536
(2016).
85. F. Bonell, Y. T. Takahashi, D. D. Lam et al., Reversible change in the oxidation
state and magnetic circular dichroism of Fe driven by an electric field at the
FeCo/MgO interface, Appl. Phys. Lett. 102, 152401 (2013).
86. S. Miwa, K. Matsuda, K. Tanaka et al., Voltage-controlled magnetic anisot-
ropy in Fe|MgO tunnel junctions studied by x-ray absorption spectroscopy,
Appl. Phys. Lett. 107, 162402 (2015).
References 539
87. S. Kanai, M. Endo, S. Ikeda, F. Matsukura, and H. Ohno, Magnetic anisotropy

modulation in Ta/CoFeB/MgO structure by electric fields, J. Phys.: Conf. Ser.
266, 012092 (2011).
88. T. Nozaki, K. Yakushiji, S. Tamaru et al., Voltage-induced magnetic anisotropy
changes in an ultrathin FeB layer sandwiched between two MgO layers, Appl.
Phys. Express 6, 073005 (2013).
89. T. Koyama, A. Obinata, Y. Hibino, and D. Chiba, Sign reversal of electric field
on coercivity in MgO/Co/Pt system, Appl. Phys. Express 6, 123001 (2013).
90. W. Skowroński, T. Nozaki, D. D. Lam et al., Underlayer material influence on
electric field controlled perpendicular magnetic anisotropy in CoFeB/MgO
magnetic tunnel junctions, Phys. Rev. B 91, 188410 (2015).
91. Y. Hibino, T. Koyama, A. Obinata et al., Peculiar temperature dependence of
electric field effect on magnetic anisotropy in Co/Pd/MgO system, Appl. Phys.
Lett. 109, 082403 (2016).
92. S. E. Barnes, J. Ieda, and S. Maekawa, Rashba spin-orbit anisotropy and the
electric field control of magnetism, Sci. Rep. 4, 4105 (2014).
93. D. Yoshikawa, M. Obata, Y. Taguchi, S. Haraguchi, and T. Oda, Possible origin
of nonlinear magnetic anisotropy variation in electric field effect in a double
interface system, Appl. Phys. Express 7, 111305 (2014).
94. P. V. Ong, N. Kioussis, D. Odkhuu, P. K. Amiri, K. L. Wang, and G. P. Carman,
Giant voltage modulation of magnetic anisotropy in strained heavy metal/
magnet/insulator heterostructures, Phys. Rev. B 92, 020407(R) (2015).
95. X. W. Guan, X. M. Cheng, T. Huang, S. Wang, K. H. Xue, and X. S. Mio, Effect
of metal-to-metal interface states on the electric field modified magnetic
anisotropy in MgO/Fe/non-magnetic metal, J. Appl. Phys. 119, 133905 (2016).
96. S. Miwa, M. Suzuki, M. Tsujikawa et al., Voltage controlled interfacial magne-
tism through platinum orbits, Nat. Commun. 8, 15848 (2017).
97. D. Chiba, S. Fukami, K. Shimamura, N. Ishiwata, and T. Ono, Electrical con-
trol of the ferromagnetic phase transition in cobalt at room temperature, Nat.
Mater. 10, 853–856 (2011).
98. K. Shimamura, D. Chiba, S. Ono et al., Electrical control of Curie temperature
in cobalt using an ionic liquid, Appl. Phys. Lett. 100, 122402 (2012).
99. M. Kawaguchi, K. Shimamura, S. Ono et al., Electric field effect on magnetiza-
tion of an Fe ultrathin film, Appl. Phys. Express 5, 063007 (2012).
100. M. Oba, K. Nakamura, T. Akiyama, T. Ito, M. Weinert, and A. J. Freeman,
Electric field modification of the magnon energy, exchange interaction, and
Curie temperature of transition-metal thin films, Phys. Rev. Lett. 114, 107202
(2015).
101. A.-M. Pradipto, T. Akiyama, T. Ito, and K. Nakamura, Mechanism and electric
field induced modification of magnetic exchange stiffness in transition metal
thin films on MgO (001), Phys. Rev. B 96, 014425 (2017).
102. L. Gerhard, T. K. Yamada, T. Balashov et al., Magnetoelectric coupling at metal
surfaces, Nat. Nanotech. 5, 792–797 (2010).
103. A. Okada, S. Kanai, M. Yamanouchi, S. Ikeda, F. Matsukura, and H. Ohno,
Electric field effects on magnetic anisotropy and damping constant in Ta/
CoFeB/MgO investigated by ferromagnetic resonance, Appl. Phys. Lett. 105,
052415 (2014).
104. A. Okada, S. He, B. Gu et al., Magnetization dynamics and its scattering mech-
anism in thin CoFeB films with interfacial anisotropy, Proc. Natl. Acad. Sci.
114, 3815–3820 (2017).
105. M. Yamanouchi, A. Jander, P. Dhagat, S. Ikeda, F. Matsukura, and H. Ohno,
Domain structure in CoFeB thin films with perpendicular magnetic anisot-
ropy, IEEE Magn. Lett. 2, 3000304 (2011).
106. T. Dohi, S. Kanai, A. Okada, F. Matsukura, and H. Ohno, Effect of electric field
modulation of magnetic parameters on domain structure in MgO/CoFeB, AIP
Adv. 6, 075207 (2016).
107. A. L. Sukstanskii and K. I. Primak, Domain structure in an ultrathin ferromag-

netic film, J. Magn. Magn. Mater. 169, 31–38 (1997).
108. F. Ando, H. Kakizakai, T. Koyama et al., Modulation of the magnetic domain
size induced by an electric field, Appl. Phys. Lett. 109, 022401 (2016).
109. H. Kakizakai, F. Ando, T. Koyama et al., Switching local magnetization by elec-
tric field-induced domain wall motion, Appl. Phys. Express 9, 963004 (2016).
110. T. Dohi, S. Kanai, F. Matsukura, and H. Ohno, Electric field effect on spin-wave
resonance in a nanoscale CoFeB/MgO magnetic tunnel junction, Appl. Phys.
Lett. 111, 072403 (2017).
111. D. Chiba, M. Kawaguchi, S. Fukami et al., Electric field control of magnetic
domain-wall velocity in ultrathin cobalt with perpendicular magnetization,
Nat. Commun. 3, 888 (2012).
112. U. Bauer, S. Emori, and S. D. Beach, Voltage-controlled domain wall traps in
ferromagnetic nanowires, Nat. Nanotech. 8, 411–416 (2013).
113. H.-B. Chen and Y.-Q. Li, Electric field-controlled suppression of Walker break-
down and chirality switching in magnetic domain wall, Appl. Phys. Express 9,
073004 (2016).
114. K. J. A. Franke, B. Van de Wiele, Y. Shirahata, S J. Hämäläinen, T. Taniyama,
and S. van Djiken, Reversible electric field-driven magnetic domain wall
motion, Phys. Rev. X 5, 011010 (2015).
115. M. Goto, K. Nawaoka, S. Miwa, S. Hatanaka, N. Mizuochi, and Y. Suzuki,
Electric field modulation of tunneling anisotropic magnetoresistance in tun-
nel junctions with antiferromagnetic electrodes, Jpn. J. Appl. Phys. 55, 080304
(2016).
116. P. X. Zhang, G.F. Yin, Y.Y. Wang, B. Cui, F. Pan, and C. Song, Electrical control
of antiferromagnetic metal up to 15 nm, Sci. China-Phys. Mech. Astron. 59,
687511 (2016).
117. S. Shimizu, K. S. Takahashi, T. Hatano, M. Kawasaki, Y. Tokura, and Y. Iwasa,
Electrically tunable anomalous Hall effect in Pt thin films, Phys. Rev. Lett. 111,
216803 (2013).
118. K. C. Nowak, F. H. Koopens, Y. V. Nazarov, and L. M. K. Vandersypen, Coherent
control of a single electron spin with electric fields, Science 318, 1430-1433
(2007).
119. M. Ono, J. Ishihara, G. Sato, Y. Ohno, and H. Ohno, Coherent manipulation of
nuclear spins in semiconductors with an electric field, Appl. Phys. Express 6,
033002 (2013).
120. T. Nozaki, Y. Shiota et al., Electric field-induced ferromagnetic resonance exci-
tation in an ultrathin ferromagnetic metal layer, Nat. Phys. 8, 491–496 (2012).
121. J. Zhu, J. A. Katine, G. E. Rowlands et al., Voltage-induced ferromagnetic reso-
nance in magnetic tunnel junctions, Phys. Rev. Lett. 108, 197203 (2012).
122. A. A. Tulapurkar, Y. Suzuki, A. Fukushima et al., Spin-torque diode effect in
magnetic tunnel junctions, Nature 438, 339–342 (2005).
123. K. Mizunuma, M. Yamanouchi, H. Sato et al., Size dependence of magnetic
properties of nanoscale CoFeB-MgO magnetic tunnel junctions with perpen-
dicular magnetic easy axis observed by ferromagnetic resonance, Appl. Phys.
Express 6, 063002 (2013).
124. S. Kanai, M. Gajek, D. C. Worledge, F. Matsukura, and H. Ohno, Electric field-
induced ferromagnetic resonance in a CoFeB/MgO magnetic tunnel junction
under dc bias voltages, Appl. Phys. Lett. 105, 242409 (2014).
125. E. Hirayama, S. Kanai, H. Sato, F. Matsukura, and H. Ohno, Ferromagnetic
resonance in nanoscale CoFeB/MgO magnetic tunnel junctions, J. Appl. Phys.
117, 17B708 (2015).
126. M. Shinozaki, E. Hirayama, S. Kanai, H. Sato, F. Matsukura, and H. Ohno,
Damping constant in a free layer in nanoscale CoFeB/MgO magnetic tunnel
junctions investigated by homodyne-detected ferromagnetic resonance, Appl.
Phys. Express 10, 013001 (2017).
References 541
127. E. Hirayama, S. Kanai, J. Ohe, H. Sato, F. Matsukura, and H. Ohno, Electric

field induced nonlinear ferromagnetic resonance in a CoFeB/MgO magnetic
tunnel junction, Appl. Phys. Lett. 107, 132404 (2015).
128. M. Harder, Y. Gui, and C.-M. Hu, Electrical detection of magnetization
dynamics via spin rectification effect, Phys. Rep. 661, 1–59 (2016).
129. H. Mazraati, T. Q. Le, A. A. Awad et al., Free- and reference-layer magnetiza-
tion modes versus in-plane magnetic field in a magnetic tunnel junction with
perpendicular magnetic easy axis, Phys. Rev. B 94, 104428 (2016).
130. Y. Shiota, S. Miwa, S. Tamaru et al., High-output microwave detector using
voltage-induced ferromagnetic resonance, Appl. Phys. Lett. 105, 192408 (2014).
131. X. Zhang, T. Liu, M. E. Flatté, and H. X. Tang, Electric field coupling to spin
waves in a centrosymmetric ferrite, Phys. Rev. Lett. 113, 037202 (2014).
132. K. Nawaoka, Y. Shiota, S. Miwa et al., Voltage modulation of propagation of
spin waves in Fe, J. Appl. Phys. 117, 17A905 (2015).
133. K. Nawaoka, S. Miwa, Y. Shiota, N. Mizuochi, and Y. Suzuki, Voltage induc-
tion of interfacial Dzyaloshinskii-Moriya interaction in Au/Fe/MgO artificial
multilayer, Appl. Phys. Express 8, 063004 (2015).
134. S. Mühlbauer, B. Binz, F. Jonietz et al., Skyrmion lattice in a chiral magnet,
Science, 323, 915–919 (2009).
135. Y. Nakatani, M. Hayashi, S. Kanai, S. Fukami, and H. Ohno, Electric field con-
trol of Skyrmions in magnetic nanodisks, Appl. Phys. Lett. 108, 152403 (2016).
136. W.-G. Wang, M. Li, S. Hageman, and C. L. Chien, Electric field assisted switch-
ing in magnetic tunnel junctions, Nat. Mater. 11, 64–68 (2012).
137. Y. Shiota, T. Maruyama, T. Nozaki, T. Shinjo, M. Shiraishi, and Y. Suzuki,
Voltage-assisted magnetization switching in ultrathin Fe80Co20, Appl. Phys.
Express 2, 063001 (2009).
138. S. Kanai, Y. Nakatani, M. Yamanouchi, S. Ikeda, F. Matsukura, and H. Ohno,
In-plane magnetic field dependence of electric field-induced magnetization
switching, Appl. Phys. Lett. 103, 074208 (2013).
139. S. Kanai, F. Matsukura, and H. Ohno, Electric field-induced magnetization
switching in CoFeB/MgO magnetic tunnel junctions, Jpn. J. Appl. Phys 58,
0802A3 (2017).
140. S. Kanai, Y. Nakatani, M. Yamanouchi et al., Magnetization switching in a
CoFeB/MgO magnetic tunnel junction by combining spin-transfer torque and
electric field-effect, Appl. Phys. Lett. 104, 212406 (2014).
141. C. Grezes, F. Ebrahimi, J. G. Alzate et al., Ultra-low switching and scaling in
electric field-controlled nanoscale magnetic tunnel junctions with high resis-
tance-area product, Appl. Phys. Lett. 108, 012403 (2014).
142. S. Kanai, F. Matsukura, and H. Ohno, Electric field-induced magnetization
switching in a CoFeB/MgO magnetic tunnel junctions with high junction
resistance, Appl. Phys. Lett. 108, 192406 (2014).
14
Topological
Insulators
From Fundamentals
to Applications

14.2 Topological Insulators in 2D: Quantum Spin Hall
Systems 549
14.2.1 Effective Bulk Hamiltonian for 2D TIs 549
14.2.2 Helical Edge States in 2D TIs and Their
Protection 552
14.3 Three-Dimensional Topological Insulators 557
14.4 Towards Topological Quantum Computation Using
3D Time-Reversal Invariant Topological Insulators 559
14.5 Majorana Fermions in 3D Time-Reversal Invariant
Topological Insulators 560
543
544 Chapter 14. Topological Insulators
14.6 Experimental Progress in Proximity-Coupled 3D

Time-Reversal Invariant Topological Insulators 564
Acknowledgments 566
References 567
14.1 INTRODUCTION
One of the biggest challenges of physics is the classification of different
states of matter. Until recently, the phases of matter could be understood
using Landau–Lifshitz theory [1], which characterizes states in terms of
their underlying symmetries that are spontaneously broken. For example, a
magnet spontaneously breaks rotation symmetry, although the fundamental
interactions within the magnet itself are isotropic in nature. Starting from
1980 with the discovery of the quantum Hall effect (QHE) [2, 3], new phases
have emerged that are not characterized by broken symmetry, but rather
by the presence of a global, or topological, invariant that is contributed to
by all of the states in the system. The QHE appears in large magnetic fields
when the two-dimensional density of states becomes broken into successive,
highly degenerate Landau levels. As is the case in the traditional Hall effect,
the presence of an applied in-plane electric field drives a current from one
side of the system to the other. In the case of the QHE, the position of the
Fermi level relative to the Landau levels determines the type of transport that
is observed. When the Fermi level is inside a Landau level, there are many
states available to carry current across the system, both within the bulk and
at the edge. However, when the Fermi level is between two successive Landau
levels, then the kinetic energy of the bulk is quenched and becomes insulat-
ing. The resulting conduction becomes quantized, and appears through chi-
ral (unidirectional) edge states observed at the boundary of the sample that
are immune against backscattering. The connection between the QHE and
topology is made via the TKKN (after Thouless, Kohomoto, Nightingale, and
den Nijs) invariant, in which the first Chern number describes the winding
of the corresponding Bloch wave functions over the magnetic Brillouin zone
[4]. This quantized number corresponds exactly to the number of propagat-
ing edge states in the QHE, and is invariant to the details of the underly-
ing system, so long as the two edge states remain spatially segregated. The
TKNN invariant thus provides the means of calculating the topological
invariant for the QHE.
While it is clear from the relationship between the TKNN invariant
and the number of edge states in the system that the QHE is topological in
nature, there is no explicit symmetry that protects the edge states. Therefore,
the question then becomes whether topological phases in nature exist which
have a non-trivial, or non-zero, topological invariant, but are characterized
by particular symmetries that are preserved. In an early effort to answer
this question, Haldane constructed an artificial model on the hexagonal
lattice, characterized by a non-zero Chern number, but without any net
magnetization [5]. This development of a topological phase that has broken

time-reversal symmetry, yet preserved inversion symmetry, inspired Kane
and Mele to define quantum spin Hall (QSH) systems, commonly referred
to as two-dimensional (2D) topological insulators (TIs), that preserve time-
reversal symmetry [6, 7]. Indeed, in a more general sense, one can define
TIs as materials that possess an insulating gap in the bulk, while on the
boundary, there are gapless metallic spin-polarized edge states whose gap-
less nature is stabilized by the presence of an underlying symmetry.
Specifically, within 2D TIs [6–11], the observed metallic electric con-
duction is associated with propagating states that occur only near the sample
edges, while the conduction in the interior is suppressed by a band gap, like
in ordinary band insulators. These edge states originate from intrinsic spin-
orbit (SO) coupling, and are profoundly different from those appearing in
quantum Hall systems in a strong perpendicular magnetic field [12, 13]. The
key distinction between the edge states observed in the QHE, and the edge
states of 2D TIs lies in the role of time-reversal symmetry. In 2D TIs, the SO
coupling preserves time-reversal symmetry, resulting in a pair of counter-
propagating channels on the same edge, as opposed to the chiral edge states
in integer quantum Hall systems. The strong SO interactions in the 2D TI
force the spin and momentum components of the two independent TI edge
channels to be locked in opposite directions, so that one spin-polarized edge
state propagates in one direction, while the other oppositely spin-polarized
edge state flows in the opposite direction. Such spin-momentum locked edge
states are referred to as helical. The helical edge state consisting of a single
massless Dirac fermion is “holographic”, in the sense that it cannot exist in
a purely 1D system, but it can only exist as the boundary of a two-dimen-
sional system [14]. Moreover, as a result of the presence of the time-reversal
symmetry, the helical edge states have a nodal band dispersion (see also
Figure 14.1) that is topologically protected against any structural or sample
imperfections that do not break time-reversal symmetry [15, 16].
Nonetheless, TIs are not strictly limited to 2D systems, and we now
introduce the three-dimensional (3D) analogues of 2D TIs. In 3D TIs, the
topologically protected electronic states appear on the surface of a bulk
material. These surface states have a nodal band dispersion in the form of a
2D Dirac-like cone reflecting a continuum of momentum directions on the
surface, as shown in Figure 14.2. The family of materials and heterostruc-
tures which can host 2D surface states is large, and continues to increase.
2D surface states were first predicted in inverted semiconductor contacts,
but without realizing their topological protection [17, 18]. The coexistence of
metallic surface states and a bulk gapped band structure, normally referred
to as the 3D TI phase, has been established theoretically for many materials
including: the semiconducting alloy Bi1-xSbx [19], strained 3D layers of α-Sn
and HgTe [19], the tetradymite semiconductors Bi2Se3, Bi2Te3, and Sb2Te3
[20], thallium-based ternary chalcogenides TlBiTe2 and TlBiSe2 [21–23], as
well as Pb-based layered chalcogenides [24, 25]. Experimentally, topological
surface states in 3D materials have been observed by means of angle-resolved
photoemission spectroscopy (ARPES) in many of the theoretically predicted
FIGURE 14.1 Band structure of (a) an ordinary insulating material, and (b) a 2D
topological insulator with inverted band order based on HgTe QWs. In (b), the
helical edge states of 2D TIs and their dispersion (for a single edge) are shown
(depicted as solid and dashed lines, respectively).
materials, such as: Bi1-xSbx [26, 27], Bi2Se3 [28], Bi2Te3 [29], TlBiSe2 [30–32],
TlBiTe2 [32], strained HgTe [33, 34], Pb(Bi1-xSbx)2Te4 [35], and PbBi2Te4 [36].
Subsequent to the work on TIs in both 2D and 3D, the notion of topolog-
ical protection was then extended to encompass topological phases of mat-
ter that are protected by symmetries beyond time-reversal. An example of
topological protection beyond that provided by the presence of time-reversal
symmetry, are symmetries such as crystalline symmetries. These crystalline
symmetries include, for example: rotations [37], reflections [38], and glide
planes [39]. The most common topological phases stabilized by the presence
of crystalline symmetry are mirror-symmetric TIs [40–42]. As in the case of a
TI that has preserved time-reversal symmetry, a topological crystalline insu-
lator is generally defined as a bulk insulator with gapless edge or surface states
that cannot be removed so long as the preserving crystal symmetry is intact.
Recently, further examples of topological phases have emerged in mate-
rials that are not insulating in their bulk, but rather semimetallic. Much like
their 2D cousin graphene, Dirac semimetals contain degenerate 3D gapless
Dirac points that are centered in the bulk of the material, rather than just
on the surface. In contrast to graphene, however, Dirac semimetals have
FIGURE 14.2 Two-dimensional gapless Dirac cone which is the dispersion of the
surface state of 3D TIs. The helicity of the upper and lower parts of the cone is opposite.
Hamiltonians that are comprised of two counterpropagating Weyl fermi-

ons. Nevertheless, as two counterpropagating Weyl fermions may annihilate
one another if they come into contact within momentum space, resulting
in an avoided band crossing, there must be an additional symmetry pres-
ent in the system that stabilizes the band crossing. Early theoretical pre-
dictions stated that A3Bi (A = Na, K, Rb) compounds would be candidate
materials to host the requisite 3D Dirac dispersion in the bulk [43]. These
materials crystalize in the simple honeycomb lattice in-plane stacked in the
c-axis direction orthogonal to the plane. The heavy atoms associated with
the A3 compound were predicted to invert the respective conduction and
valence bands. However, the mere presence of band inversion is not suffi-
cient to be able to predict topological behavior, as the band crossing will
simply form an avoided level crossing without the presence of an additional
stabilizing symmetry. In the case of Na3Bi, the stabilization that allows for
the preservation of the bulk band crossing comes from the time-reversal and
inversion symmetries. Following the theoretical predictions, the first Dirac
semimetal that has been experimentally observed is, naturally, Na3Bi [44].
Dirac semimetals with additional symmetries beyond those present in Na3Bi
have also been theoretically predicted and subsequently experimentally
confirmed. Examples of additional Dirac semimetals include: Cd3As2, that
has preserved time-reversal symmetry and C4 rotational symmetry [45], and
CuMnAs [46, 47]. CuMnAs is a particularly interesting example of a Dirac
semimetal, in that it is an antiferromagnetic semimetal that breaks both
time-reversal symmetry and inversion symmetry. At first glance, it would
seem that CuMnAs would then possess a gap due to the broken symmetries,
yet CuMnAs preserves the combination of time-reversal and inversion sym-
metry, thereby allowing a protected bulk 3D band crossing point to persist.
As we have mentioned, Dirac semimetals are characterized by a low-

energy effective Hamiltonian that consists of two copies of counterpropa-
gating Weyl fermions. Weyl fermions are massless fermionic quasiparticle
excitations, as in the case of a Dirac electron. However, Weyl fermions have
a definite chirality in real space associated with them, essentially rendering
them similar to that of the edge state encountered within the integer QHE.
Another manner within which one can understand the relationship between
Weyl fermions and Dirac fermions is that the Weyl fermion two component
spinor is exactly half of the normal four component Dirac spinor. For many
years, these excitations have been sought as another example of the intimate
connection between high-energy physics and condensed matter physics. To
be more specific, neutrinos had been assumed to be Weyl fermions; however,
it was later realized that they do, in fact, possess a small mass that invali-
dates them as candidate particles that possess Weyl fermion characteristics.
While finding high-energy realizations of Weyl fermions has not yet resulted
in confirmed examples, the search has been fruitful within the context of
condensed matter realizations of quasiparticle excitations that have char-
acteristics consistent with Weyl fermions in semimetallic materials. Weyl
semimetals are characterized by exotic Fermi open surface projections, or
Fermi arcs, that connect sources and sinks of Berry curvature, known as
Weyl nodes, that are characterized by an integer Chern number. To date
there have been several different mechanisms through which Weyl fermi-
ons have been realized in semimetals by breaking different symmetries that
allow the typical Dirac spinor to be broken into a Weyl spinor. Specifically,
one can break: time-reversal symmetry, as in the case of topological hetero-
structures [48] or HgCr2Se4 [49], inversion symmetry as in the case of TaAs
and related compounds [50, 51], or within compounds that break Lorentz
invariance such as W1−xMoxTe2 and LaAlGe that are also referred to as
Type-II Weyl semimetals [52].
The remainder of the chapter is constructed as follows: In Section 14.2,
we describe in detail the construction of the bulk Hamiltonian for 2D TIs, as
this forms the foundation for the understanding of the 3D TI and allows one
to extrapolate to the construction of more complicated Hamiltonians con-
taining additional symmetries. In our construction, we focus on the salient
features associated with 2D TIs, including the formation of the metallic spin-
polarized edge states and their topological protection. Using this formula-
tion as a base, in Section 14.3, we move on to discuss the properties of 3D
TIs, building on the knowledge that we gained in Section 14.2. The unique
physics of TIs naturally leads to the development of many new proposals
for applications. From the spintronics point of view [53], the spin-polarized
edge channels in 2D TIs could be used to inject spin currents into doped
semiconductors or metals [54]. Further, ferromagnet/3D TI junctions have
shown a significant promise for potential spintronics applications due to: the
large spin-orbit torque that the Dirac surface can generate [55, 56], electri-
cal manipulation of spin-polarized currents in ferromagnet/3D TI junctions
[57–59], as well as an unusual in-plane tunneling Hall effect which could
be possibly used for spin-valves [60]. Nevertheless, while there are many
14.2 Topological Insulators in 2D: Quantum Spin Hall Systems 549
proposals for applications of TIs, one of the most intriguing is their possible
application in the field of topological quantum information processing. In
Section 14.4, we address this interesting application through an introduc-
tion to the formation of the requisite states. In Section 14.5, we present a
mathematical formulation of the process through which the combination
of normal s-wave superconductivity and time-reversal invariant TIs can be
utilized to form the excitations, known as Majorana-bound states, that form
the basis of popular implementations of the topological quantum computing
introduced in Section 14.4. We conclude with Section 14.6, where we discuss
some of the recent progress that has been achieved in the field of topologi-
cal superconductivity within TIs, so as to realize the necessary components
for topological quantum information processing. Some related discussion on
topological insulators can be found in Chapter 15, Volume 2, and on triplet
superconducting pairing needed for Majorana-bound states in Chapter 16,
Volume 1.
14.2 TOPOLOGICAL INSULATORS IN 2D:

QUANTUM SPIN HALL SYSTEMS
14.2.1 Effective Bulk Hamiltonian for 2D TIs
We begin by describing the construction of the model Hamiltonian for 2D
TIs. In this endeavor, there are two main models describing 2D TIs: the
Bernevig/Hughes/Zhang (BHZ) model [8], which describes HgTe/CdTe and
InAs/GaSb quantum wells (QWs), and the extensions of the Kane–Mele
model which describes 2D TIs on hexagonal lattices [6, 7]. From a historical
perspective, the Kane–Mele model of graphene is the first material that was
proposed to be a 2D TI [7]. However, as the main orbitals at the Fermi energy
are pz-like, the atomic-SO interaction is between next-nearest neighbors, the
induced gap is small, only around 20 μeV [61, 62], and the predicted effect
is not experimentally measurable. Recently, in the context of new types of
purely 2D materials on hexagonal lattices similar to that of graphene, there
are proposals that may be understood via extensions of the Kane–Mele
model that purport to show larger SO interaction, and, therefore, may yield
topological phases that are more experimentally accessible. These materials
include: silicene [63, 64] (with a band gap of around a few meV), germanene
[63] (with a band gap of around 20 meV), the heterostructure of germanene
on MoS2 [65], functionalized stanene [66] (with a band gap of around 0.3
eV), and single layers of WTe2 [67, 68, 69] (with a band gap of around 0.1
eV). Additionally, bismuthene on SiC has recently gained a lot of attention
as a potential candidate for room-temperature topological behavior [70–73]
where the on-site SO interactions give a band gap of the order of 0.7 eV.
As most of the physics surrounding the Kane–Mele model has yet to
be experimentally verified, we will focus our attention on the BHZ model,
and discuss 2D TIs in the context of their first experimental realization in
HgTe/CdTe QW [9–11, 54]. We start our examination by considering the
structure of CdTe. CdTe is a zinc-blende-type semiconductor whose band
structure can be described by the eight-band k × p model. CdTe has nor-

mal band structure, i.e. its conduction band (G 6 ) consists of s-wave orbit-
als, while the valence band (G8 ) consists of p-like orbitals. Therefore, a thin
layer of CdTe produces the well-known semiconducting band structure that
is similar to the one shown in Figure 14.1a with the E1 band (conduction
band) having mainly electron-like properties, and the H1 band (valence
band) having heavy hole-like character and positive band gap of value 2.
However, this is not the situation that one finds in bulk HgTe. Unlike con-
ventional zinc-blende semiconductors, due to large relativistic corrections,
including relativistic velocity corrections and the SO interaction, HgTe has
an inverted band structure, by which we mean that the G 6 band, that origi-
nates from metallic s-orbitals and usually forms the conduction band in
normal semiconductors, has a lower energy than the G8 band which derives
from chalcogenide p-orbitals [75]. Consequently, in the HgTe layer, the ener-
getic subbands are also inverted and the band gap 2 at k = 0 is negative, as
shown in Figure 14.1b. In this case, the conduction band has heavy hole-like
(H1) character, while the valence band has mainly electron-like character.
Now, let us build heterostructures where HgTe is sandwiched between
CdTe layers to form a HgTe/CdTe QW. Intuitively, one can say that as
long as the layer of HgTe is very thin, the full heterostructure has normal
band ordering, like in Figure 14.1a, while for thick enough HgTe layers,
the band structure can be inverted, like in Figure 14.1b. Indeed, this intui-
tive picture is confirmed in Figure 14.3, where the band structure of the
QW as a function of the thickness, d, of the HgTe layer is shown. Here, G8 -
derived subbands are denoted by H 1, H 2,... as they correspond to heavy-hole
subbands in non-inverted semiconductors, whereas the E1, E 2,... subbands
are mainly derived from G 6 subbands and, therefore, are mainly electron-
like subbands. Indeed, with decreasing HgTe thickness d, the energies of
the E subbands shift to positive energies as a result of quantum confine-
ment, whereas those of the H subbands shift to negative energies, as shown
in Figure 14.1a. This results in the normal band ordering for thin QWs. The
different d-dependence of the E1 and H1 subbands, shown in Figure 14.3,
demonstrates this principle that there exists a critical thickness, dc, at which
point the subbands change their energetic order, leading to a positive band
gap in CdTe and a negative band gap in HgTe (see Figure 14.1b). The effect
of band ordering that we have been describing here implies something inter-
esting should happen as one moves from one material that has normal band
ordering (band gap  larger than zero), and HgTe that has inverted band
ordering (band gap  smaller than zero).
To better understand the situation where we have a heterostructure that
contains a transition between non-inverted and inverted materials, as we
have in the case of CdTe/HgTe QW, we now begin to construct a simple
model. Assuming that d » dc and near the G point, and integrating growth
direction (z-direction), we can derive an effective four-band model for the
HgTe/CdTe QWs involving double degenerate E1 and H1 subbands from the
eight-band k × p model [8, 76]. We now introduce the relevant basis states
| E1, jz = 1 / 2ñ, | H 1, jz = 3 / 2ñ, | E1, jz = -1 / 2ñ and | H 1, jz = -3 / 2ñ , where
80
c
40
[meV] 0
−40
−80
4 6 8 10 12 14
[nm]
FIGURE 14.3 Ordering of electron E1, E 2, . . . and heavy-hole H1, H 2, . . . subband

energies versus well thickness d from k × p calculations for a HgTe/Hg0.3Cd0.7Te
quantum well. One can see the transition from the topologically trivial to the topo-
logically non-trivial regime for a critical thickness of dc = 6.3 nm, which corresponds
to the inversion of the bands between E1 and H1 [74]. (After Büttner, B. et al., Nat.
jz denotes z-component of the total angular momentum. Now using these

basis states, one can write the effective four-band Hamiltonian for the QW
system as follows [8, 76]:
éh(k) 0 ù
H HgTe = ê * ú . (14.1)
êë 0 h ( - k )ú
û
In Equation 14.1,
h(k) = A(σ x k x − σ y k y ) + Mk σ z + Dk 2σ 0 , (14.2)
and
Mk = M + Bk 2 . (14.3)
We must now define the individual contributions to the Hamiltonian. We

note that the two diagonal blocks of H HgTe (14.1) describe pairs of states
related to one another by time-reversal symmetry, in other words, Kramers
partners. Indeed, for a half-integer spin, as in our case, every energy level is at
least double degenerate (Kramers theory) and the two states corresponding
to this degeneracy are related by time-reversal symmetry and called Kramers
partners. Each of the blocks has a 2 × 2 matrix structure, with Pauli matrices
s x , y , z and unit matrix s0 representing the two lowest-energy subbands E1
and H1. k is an Einstein mass term whose value can be manipulated to
yield the band gap 2 at the G (k = 0) point of the Brillouin zone (corre-
sponding to the gap between electron and positron bands 2mc2 for a relativ-
istic electron in vacuum). The linear terms in Equation 14.2, proportional to
the constant  and in-plane wave-vectors k x , y , describe the E1–H1 hybrid-

ization. The positive quadratic terms k 2 and k 2 are related to the effec-
tive (Newtonian) masses (band curvature) of the E1 and H1 bands in HgTe
QWs [8]. The Hamiltonian (14.1) can be extended to include the SO coupling
between the Kramers partners [10, 76].
æ 0 s ö
Using the unitary transformation H ® UHU † with U = ç -is y 0z ÷ , we
can recast the Hamiltonian (14.1) into a Dirac-like form è ø
H = τ z σ⋅ (Ak + Mk z) + Dk 2 τ 0σ 0 , (14.4)
where k = (k x, k y, 0), z = (0, 0, 1) and the Pauli matrix tz and the unit matrix
t0 act on the Kramers partners. We note that despite the Einstein mass
term k τ z σ z , the Hamiltonian (14.4) is invariant under time reversal, i.e.
 † H  = H , where T = it y s x C is the time-reversal operator, with  denot-
ing complex conjugation. The band dispersion for both normal and topologi-
cally non-trivial insulators is described by
E = Dk 2 ± A 2 k 2 + (M + Bk 2 )2 , (14.5)
where the energy E is double degenerate. Therefore, to identify the topologi-

cally non-trivial regime, one needs to calculate a topological invariant of the
system, or introduce a boundary in the problem to check the formation of
edge states. Since time-reversal symmetry is preserved in 2D TIs, the Chern
number, C, (winding number of the Bloch wave functions over the magnetic
Brillouin zero) is zero, due to the cancellation of the Chern numbers for two
time-reversal related blocks (C↑ + C↓ = 0) for Equation 14.1, where C↑ is the
Chern number for a block with the positive jz and C¯ is the Chern number
for the block with the negative jz. However, the spin Chern number describ-
ing the difference between the winding number of two corresponding Bloch
wave functions for the Kramers partners over the whole Brillouin zone is
non-zero i.e. C s = (C↑ − C↓ ) / 2. This spin Chern number defines the topologi-
cal invariant of the BHZ model.
14.2.2 Helical Edge States in 2D TIs

and Their Protection
Edge states can be realized when a QW with an inverted gap  < 0 borders
with a normal insulator with  > 0 (see Figure 14.1b). They originate from
the fact that one cannot deform the insulator with the negative band gap
(topological insulator) with spin Chern number Cs = 1 to the one with the
positive band gap (trivial insulator) with spin Chern number Cs = 0, without
going through gapless states at the boundary (edge states). Through the bulk
boundary correspondence [15, 16], the number of pairs of edge states (called
also Z2 invariant [6, 15, 16]) is exactly equal to n = C s mod 2. Therefore for
the BHZ model, we expect one pair of edge states at the interface between
a topologically trivial and non-trivial insulator, as shown in Figure 14.1b.
In this subsection, we explicitly show the existence of such edge states and
discuss the following properties:
ªª the edge-state spectrum is gapless and merges into the bulk spec-
trum above the band gap;
ªª the edge states are the orthogonal eigenstates of the helicity operator
S = tz s× k̂, where k̂ is the unit vector in the direction of the edge-state
momentum. For this reason, the QSH edge states are called helical;
ªª a local static perturbation H ¢ preserving time-reversal symmetry
does not couple the QSH edge states.
In order to illustrate these properties, we will make further simplifi-

cations. To get a simple analytical model, we will omit all terms µ k 2 in
Hamiltonian (14.4). This is justified since the edge states occur in the vicin-
ity of the G(k = 0) point. Although without quadratic terms, one cannot
define a topological invariant uniquely, for the particle-hole symmetric case,
the edge states only exist in the topologically non-trivial regime [77]. Then
Hamiltonian (14.4) takes the form H = tz s× (Ak + Mz), which in position
representation corresponds to the following equation for the four-compo-
nent wave function Y(r) :
[es0 - tz s× (-iAÑ + Mz)]Y(r) = 0. (14.6)
Second, one cannot confine the Dirac spectrum with hard wall boundary
conditions when the quadratic terms (lattice normalization) are missing. For
a linear differential equation, like the Dirac equation, only one boundary
condition is needed, and one requires that particles do not leak through the
boundary, i.e. the normal component of the particle current at the boundary
vanishes [78]. In our case, we assume vanishing of the current for y = 0. This
yields the following condition:
J y ( x , y = 0) = (e / )Y † ( x , y = 0)t z s y Y( x , y = 0) = 0 (14.7)
and correspondingly for the wave function:
Y( x , y = 0) = t0s x Y( x , y = 0). (14.8)
We seek solutions to Equation 14.6 in the form of the two eigenstates, Y k ,± (r ),

of the diagonal matrix tz s0 propagating along the edge (in x-direction) of the
TI and decaying exponentially away from it in the y-direction. The decaying
solution is enforced since there is a finite band gap on both sides of an inter-
face in y-direction (see Figure 14.1).
 1  Ψ1k +  ikx − y / λ (14.9)

Ψ k , + (r) =   ⊗  e ,
 0  Ψ2 k + 
 0  Ψ1k −  ikx − y / λ (14.10)

Ψ k , − (r) =   ⊗  e ,
 1  Ψ 2 k − 
with a real positive decay length l > 0. The symbol Ä denotes a tensor prod-
uct of an eigenstate of tz (first column) and the wave function in s space
(second column). The conditions for the non-trivial solutions for the coef-
ficients Y 1k ± and Y 2k ± follow from Equations 14.6 and 14.8, yielding two
equations for l and e:
1 / l 2 - M 2 /  2 = k 2 - e2 /  2 , (14.11)
1 / l + M /  = k - e / t, t = ±1. (14.12)
We notice that the left-hand side of Equation 14.12 does not contain the
index t, whereas the right-hand side does. This can only be true if both
sides of Equation 14.12 (and those Equation 14.11) vanish independently,
which yields a solution with a gapless linear dispersion and a real decay
length:
e k t = Ak t, l = -A / M, M < 0. (14.13)
Since l must be positive, the edge states exist only in a system with an
inverted negative gap, disappearing when  turns positive. Their propaga-
tion velocity v =  / coincides with that of the bulk states above the gap
(see also Figure 14.1).
The edge-state wave functions normalized to the half-space 0 £ y < ¥
are given by
 1  1 | M | ikx −|M|y / A (14.14)

Ψ k , + (r) =   ⊗   e ,
 0  1 A
 0  1 | M | ikx −|M|y / A (14.15)

Ψ k , − (r) =   ⊗   e .
 1  1 A
The key feature of the edge states (14.14) and (14.15) is that they are orthogo-
nal eigenstates of the helicity operator S = t z s x :
SY k ,t (r) = t Y k ,t (r), t = ±1. (14.16)
The helicity S is defined as the projection of the vector S= t z s on the direc-

tion of the edge-state momentum k̂  x. Since the matrix structure of S
derives from the SO-split energy bands, it is also called the spin helicity.
Equation 14.16 is a manifestation of time-reversal symmetry, and the fact
that the QSH state is generally characterized by a Z2 topological invariant,
as mentioned at the beginning of this subsection [6]. Therefore, indeed the
Z2 topological invariant can only have two values: n = 1 if the number of
pairs of Kramers partners is odd, or n = 0 if the number of Kramers pairs is

even. Further, using properties of the time-reversal operator, one can easily
show why the topological protection appears for an odd number of pairs of
Kramers partners.
It is easy to see that one helical channel can be obtained from the other
by simply applying the time-reversal operator  = it y s x C :
Y k ,+ =  Y - k ,- , Y - k ,- = - Y k ,+ . (14.17)
The helicity (see Equation 14.16) protects the edge states from local pertur-
bations that do not break time-reversal symmetry. Concretely, let us con-
sider a perturbation H ¢ which is invariant under time reversal:
[ H ¢,  ]= 0. (14.18)
Using Equations 14.17 and 14.18, we can transform the matrix element
á Y - k ,- | H ¢ | Y k ,+ ñ invoking the antiunitary properties of the time-reversal
operator as follows:
〈 Ψk ,+ | H ′ | Ψ − k , − 〉 = 〈 (Ψ − k , − )| H ′ | Ψ − k , − 〉 = 〈 Ψ − k , − | H ′ |  2 Ψ − k , − 〉
(14.19)
For an odd number of pairs of Kramers partners, i.e for half-integer spin of
electrons  2 = -1 , and using the hermiticity of H ¢ one gets immediately:
á Y k ,+ | H ¢ | Y - k ,- ñ = (-1)á Y k ,+ | H ¢ | Y - k ,- ñ , (14.20)
i.e. for a Hermitian perturbation H ¢ , its matrix element is zero:
á Y - k ,- | H ¢ | Y k ,+ ñ = 0. (14.21)
Therefore, the helical edge states of 2D TIs are protected against elastic and
time-reversal symmetry protecting single-particle perturbations. In particu-
lar, the spin-independent disorder potential cannot cause scattering between
the helical edge states. However, two-particle scattering involving electron-
electron interactions, or perturbations which break time-reversal symme-
try, like magnetic fields or magnetic impurities, can cause backscattering
between the helical edges. In the two-terminal geometry, this protection
induces a quantized conductance G = 2e 2 / h in the gap of 2D TIs. The quan-
tized value in the band gap of 2D TIs originates from the spin-momentum
locking enforcing, under an applied chemical potential difference, that one
Kramers partner propagates at the upper edge, while the other one can only
propagate at the lower edge. In contrast, in the normal insulator G = 0 in the
band gap [9, 10]. Still, if the edge-state spin polarization is out-of-plane and
one omits the terms which mix Kramers partners blocks, an out-of-plane
magnetic field commutes with the Hamiltonian, and does not open a gap
in the edge-state spectrum [79–81]. Only perturbations which can mix the
Kramers blocks, like charge puddles where electrons lose the orientation of
their spin, or inversion symmetry breaking terms can cause scalar disorder
to mix the edge states for out-of-plane magnetic fields. Indeed, in experi-
ments on HgTe QWs, the longitudinal two-terminal conductance in a per-
pendicular magnetic field and in the diffusive regime decays to zero when
Rashba SO interactions and puddles are present [9, 10, 82].
Still, experiments in magnetic fields do not directly confirm the helicity
of the edge states. The helical edge transport in a QSH system can be seen in
four-terminal devices shown in Figure 14.4 and distinguished from the usual
QH effect. Using the Landauer–Bütikker approach, we express the current,
Ii, injected through contact i in terms of voltages Vj induced on all contacts
as follows [11, 83]:
N
e2
Ii =
h å(T V - T V ), (14.22)
j =1
ji i ij j
where T ji is the transmission probability from contact i to contact j. For a

chiral QH edge channel, T ji connects the neighboring contacts only in one
propagation direction (see Figure 14.4a), such that
T (QH )i +1,i = 1, i = 1,..., N , (14.23)
where N is the number of terminals, for example, N = 4 in Figure 14.4a, with

the convention that TN +1, N = T1, N describes the transmission from terminal
N to terminal 1. Assuming, for concreteness, that the current flows from
terminal 1 to terminal 4, while leads 2 and 3 are used as voltage probes,
(a) (b) (c)

1 2 1 2
1
QSH insulator 2
Gate 1
Gate 2
3 SHE-1 4
4 3 4 3
FIGURE 14.4 Schematic of (a) chiral and (b) helical edge transport in a four-terminal device based on 2D TIs.
(c) Verifying the spin polarization of edge channels in double gate H-bar structures. In (c) the upper part of
H-bar structure is in the quantum spin Hall state, while the lower part is in the metallic phase characterized by
the inverse spin Hall effect (SHE−1). The current is driven between contacts 1 and 2 leading to the injection of
a spin current into the middle part of the H-bar structure. This spin current is further converted into a voltage
difference due to SHE−1. (After Brüne, C. et al., Nat. Phys. 8, 486, 2012. With permission.)
14.3 Three-Dimensional Topological Insulators 557
V1 - V4
yields a finite two-terminal resistance R14,14 = = h / e 2 and zero four-
I14
terminal (non-local) resistances R14,12 = R14,23 = R14,13 = 0 . In contrast, in a
QSH system, the helical edge channels connect the neighboring contacts in
both propagation directions (see Figure 14.4b), such that
T (QSH )i +1,i = T (QSH )i ,i +1 = 1, i = 1,..., N , (14.24)
with the conventions TN +1, N = T1, N and TN, N +1 = TN,1 . Consequently,

for a current flowing from 1 to 4, we find the two-terminal resis-
V - V4 3 h
tance [11] R14,14 = 1 = , and the four-terminal resistances [11]
I14 4 e2
1 h 1 h
R14,12 = R14,23 = and R14,13 = 2
4 e2 2e
The non-zero non-local resistances are unique to the 2D TI state, allow-
ing its unambiguous experimental detection [11]. The universality of the
non-local resistances is just a consequence of time-reversal symmetry, and,
therefore, is also expected for other proposed realizations of 2D TIs, for
example, inverted InAs/GaSb QWs [84].
Yet other experiments studied the spin polarization of the edge states
in QSH systems [54, 85]. The first experiment confirming the spin polar-
ization of edge states involved an H-bar structure with two gates [54]. The
upper part of the structure was in the QSH regime, while the lower one
was in the metallic regime, which, due to strong SO interaction, exhibits an
inverse spin Hall effect (SHE−1) (see Figure 14.4c). The current was injected
(contacts 1 and 2) into the QSH regime, while a non-local voltage drop
was measured across the metallic leg (contacts 3 and 4). In this configu-
ration, the spin-polarized helical edge channels injected a spin-polarized
current into the metallic leg, causing a local imbalance in the chemical
potential of spin-up and spin-down polarized carriers. Due to the inverse
spin Hall effect (generation of a transverse voltage difference due to a spin
current) [86–90], the spin current in the metallic region induced a volt-
age between contacts 3 and 4. This voltage can only develop provided the
helical edge channels are spin polarized, and the metallic leg exhibits the
inverse spin-Hall effect. To interpret the experiments, semiclassical Monte
Carlo calculations were provided for the geometry shown in Figure 14.4c,
demonstrating that indeed the experimental non-local resistance signal
originated from the all-electrical detection of the spin polarization of the
edge states [54].
14.3 THREE-DIMENSIONAL
TOPOLOGICAL INSULATORS
Having discussed the case of 2D TIs in detail, we will now also provide a
brief overview of their 3D counterparts. Similarly to 2D TIs, 3D TIs are
characterized in the bulk by a gapped 3D Dirac Hamiltonian [20, 91]. When
this 3D Hamiltonian is then projected into the (001) plane, the following
Hamiltonian for the 2D gapless Dirac state emerges (see Figure 14.2):

ò ( )
H surf = d 2r y†v F px s x + p y s y y, (14.25)
where y† and y are electron field operators with ψ = ( ψ ↑ , ψ ↓ ) including

T
( )
two spin indices, s = s x , s y , while px and py are electron momenta, and
v F is the Fermi velocity. By examining the commutator éë , H surf ùû , we can
observe that the Hamiltonian does indeed obey time-reversal symmetry.
Similarly, as in 2D TIs, the spin and momentum of 2D surface state are
locked, i.e. at every point of the Fermi surface spin is parallel to the momen-
tum (see Figure 14.2), defining the helicity of the conduction band opposite
to the helicity of the valence band. Further, any perturbation which is elastic
and does not mix the spin cannot connect states | s, pñ with states | - s, - pñ
(see Figure 14.2), and therefore backscattering (180-degree scattering), for
example, by scalar disorder is prohibited. This lack of backscattering has
many physical consequences, i.e. transport scattering in this system is two
times larger than elastic scattering, giving rise to a non-monotonic mobil-
ity dependence as a function of the density of electrons [92], as well as weak
antilocalization [93]. Further, the spin-momentum locking for surface states
of 3D TIs gives rise to a revised Hikami–Larkin–Nagaoka formula when SO
impurities and the Dirac dispersion are combined [94].
To further characterize 3D TIs, we will now introduce the concept of
topological invariant for these materials. As we explained in Section 14.2,
the Z2 invariant for 2D TIs can have only two values, i.e. n = 0 or n = 1.
Correspondingly, the extension of this definition for 3D TIs with inversion
symmetry includes four Z2 invariants (n 0 ; n1n 2n 3 ) [95, 96], especially n 0
which defines whether the TI is weak (consisting of many 2D TIs weakly
coupled), or strong can be determined by calculating the product of di = ±1
in eight time-reversal symmetric points of the Brillouin zone Gi , i.e.
(-1)n0 = Õd (14.26)
i
i
∏
N
with δ i = ζ 2m (Γ i ) , where z 2m (Gi ) is the parity eigenvalue of ±1 of the
m=1
2m-th occupied band. Translating more explicitly Equation 14.26, if the sys-
tem possesses inverted bands at an odd number of high symmetry points
(every inversion is related to the change in the parity eigenvalue) in their
bulk 3D Brillouin zone, then it should also exhibit an odd number of sur-
face state crossings. This gives n 0 = 1, and prohibits adiabatic continuation
of the band structure to the ordinary insulator. Since these crossings can be
observed in the photoemission experiment, this gives direct confirmation of
the non-trivial band structure of 3D TIs. Indeed, angle-resolved photoemis-
sion directly confirmed that Bi1-xSbx [26, 27], Bi2Se3[28], Bi2Te3 [29], TlBiSe2,
[30–32], TlBiTe2 [32], Pb(Bi1-xSbx)2Te4 [35], PbBi2Te4 [36], and strained
HgTe [33] are 3D TIs. Further, recent experiments on the spin-resolved
14.4 Towards Topological Quantum Computation 559
photoemission confirm the spin-momentum locking for the surface states

of 3D TIs.
Another way to characterize the unusual properties of 3D TIs is
using transport experiments. Indeed, applying an out-of-plane magnetic
field causes an opening of the gap in the Hamiltonian (Equation 14.25)
inducing the first Chern number corresponding to the Hall conductance
s xy = 2(n + 1 / 2) due to the Dirac dispersion. Therefore, one should expect an
odd Hall quantization for surface states of 3D TIs in magnetic fields. This has
indeed been observed in transport measurements in strained HgTe [33, 34]
and in BiSbTeSe2 [97]. Further consequences of this non-trivial quantization
are, for example, the Faraday and magnetoelectric effects [98–102].
14.4 TOWARDS TOPOLOGICAL QUANTUM

COMPUTATION USING 3D TIME-REVERSAL
INVARIANT TOPOLOGICAL INSULATORS
Armed with an understanding of the basic properties of time-reversal invari-
ant TIs, it is now interesting to begin to explore the various possibilities
associated with these fascinating new materials, in particular as a result of
their strong SO interactions. In this chapter, we do so by examining the pos-
sibility of finding Majorana fermions on the surfaces of 3D TIs when they are
coupled through the proximity effect with a conventional s-wave supercon-
ductor. As mentioned previously, Majorana fermions have been predicted
to be the fundamental basis upon which topological quantum computing
has been built. Therefore to be able to find and manipulate these quasipar-
ticles within 3D TIs would indeed be quite interesting. Prior to exploring
how these excitations may be manifested in TIs when paired with s-wave
superconductors, it is important to mention some of the basic properties of
Majorana fermions. First and foremost, these quasiparticles are comprised
of an equal superposition of particle and anti-particle and, as such, may
be described by a real wave equation. In terms of operators, we may write
this equivalence as gˆ = gˆ † , where ĝ † is the Majorana creation operator and
ĝ is the Majorana annihilation operator. The presence of superconducting
materials is important mathematically, as it doubles the size of the Hilbert
space, and, therefore, quasiparticles in the superconductor naturally come
with their anti-particles. This particle-hole redundancy is referred to as an
imposed, rather than inherent, symmetry that is present in the superconduc-
tor and is known as particle-hole symmetry. Thus, from a naive standpoint,
it may be possible to find non-Abelian anyonic excitations in superconduct-
ing materials. This supposition has subsequently been theoretically proven
for a variety of different systems, such as the fractional quantum Hall state
5
at filling factor n = [103], semiconductor quantum wires coupled to a con-
2
ventional s-wave superconductor when placed in a parallel magnetic field
[104–107], one-dimensional strongly SO-coupled Fe atoms on an s-wave
superconductor [108], an array of magnetic tunnel junctions coupled to a
proximity induced superconductivity in a two-dimensional electron gas

[109], and spinless px + ip y superconductors [110], where the Majorana fer-
mions exist as zero energy bound states at the core of vortex excitations
in these unconventional superconductors. Put another way, the existence of
Majorana fermions as bound states at the vortex core implies that [ gˆ , Hˆ ] = 0 ,
where Ĥ is the Hamiltonian for the system of interest. The preceding com-
mutation relation is interesting, in that it further states that the action of
Majorana fermions does not change the ground state energy of the system.
N
The presence of N such vortices leads to a 2 2 degenerate ground state, which
may be non-trivially operated on through braiding processes in which the
vortices are adiabatically rearranged. Majorana fermions have been prof-
fered as the backbone of a potentially revolutionary type of computing, topo-
logical quantum computing, that offers significant potential technological
advances, such as disruptive increases in computational efficiency at greatly
reduced energy consumption through the manipulation and production of
non-Abelian anyons [111]. Yet, beyond the possibility of technological rel-
evance, the emergence of these new particles points to the fact that proxim-
ity-coupled topological systems provide a unique playground for exploring
issues central to the foundation of quantum mechanics, such as supersym-
metry and entanglement.
In the remainder of this chapter, we will explore the physics surround-
ing the initial theoretical predictions and current work surrounding the
observation of Majorana fermions, and the underlying conditions that are
necessary for their existence in proximity-coupled TIs. To do so, we will
use the example system of a 3D time-reversal invariant topological insula-
tor that is proximity-coupled to a conventional s-wave superconductor. In
Section 14.5, we give a simple derivation following the pioneering work of Fu
and Kane [112] to give a clear understanding of the equations and conditions
that may lead to the observation of Majorana fermions, and the associated
unconventional superconductivity. In Section 14.6, we discuss the current
research into the pairing of TIs with superconductors to elucidate the prog-
ress achieved. In this final section of the chapter, we focus our attention on
the experimental developments within 3D TIs in keeping with our discus-
sion, though it should be noted that extensive theoretical and experimental
progress has also been made within proximity-coupled 2D TIs [113–120].
14.5 MAJORANA FERMIONS IN 3D TIME-REVERSAL

INVARIANT TOPOLOGICAL INSULATORS
We begin by considering the 2D surface state Hamiltonian for the 3D time-
reversal invariant TI, for which the strength of the SO interaction has led
to an inversion of the band gap at an odd number of time-reversed points
within the Brillouin zone, written in real space as

ò
H surf = d 2r y† ( -iv F s ×Ñ - m ) y. (14.27)
14.5 Majorana Fermions in 3D Time-Reversal Invariant Topological Insulators 561
where in Equation 14.27, ψ = ( ψ ↑ , ψ ↓ ) are the electron field operators, in

T
( )
which we keep two spin indices, s = s x , s y , and m and v F are the chemical
potential and Fermi velocity, respectively. We can see that Equation 14.27 is
indeed time-reversal invariant by defining the time-reversal operator, here
defined as  = is yC , where  represents complex conjugation. By examin-
ing the commutator, [  , H surf ], we can observe that the Hamiltonian does
indeed obey time-reversal symmetry. To proceed, we can rewrite Equation
14.27 in a more familiar form where we have replaced the partial deriva-
tives with the momentum within the surface as in Equation 14.25 (see
Section 14.3).
As we are explicitly concerned with a situation in which we are cou-
pling the surface states of the 3D time-reversal invariant TI with an s-wave
superconductor, then we must add the superconductive pairing that is
associated with the presence of a superconductor. Formally, this is accom-
plished by adding a potential term of Vsc = ∆ψ ↑† ψ ↓† to the Hamiltonian of
Equation 14.25. The Hamiltonian that results from this coupling must be
rewritten in the expanded particle-hole basis using the Bogoliubov–de
Gennes equation as
H BdG =
 ψ↑ 
v F ( px σ + p y σ ) − µ
x y
i∆ σ* y
  ψ↓ 
∫d r (ψ
2 †
↑ )
ψ ↓† ψ ↑ ψ ↓ 
 −i∆σ y
 † .
v F ( px σ x − p y σ y ) + µ   ψ ↑ 
 
 ψ ↓† 
(14.28)
The seemingly complicated equation above, which comprises the s-wave
superconductor and the TI surface state Hamiltonian, can be simplified by
( ↑ ↓ ↓ ↑ )
= ψ , ψ , ψ † − ψ † . This allows one to eliminate the
changing the basis to Ψ
off-diagonal terms present in the pairing blocks of Equation 14.28 to obtain
[112, 121],
 v F ( px σ x + p y σ y ) − µ ∆ *σ 0 
∫

H BdG = d r ψ 
† 2
 ∆σ 0 x
−v F ( px σ + p y σ ) + µ 

y

 ψ, (14.29)
where in Equation 14.29 s0 is the 2 × 2 identity matrix. Equation 14.29 can
be expanded into a more useful form by representing the superconducting
pairing as both a magnitude, D 0 , and a phase, f, in the following manner,
D = D 0e if , to obtain
H BdG = v F px s x Ä t z + v F p y s y Ä tz - ms0 Ä tz
(14.30)
(
+ D 0s0 Ä cos f t x + sin f t y , )
where t = ( t x , t y , t z ) are Pauli matrices that mix the blocks of y and y†

within the Hamiltonian. By defining the particle-hole symmetry that
superconductors possess as an operator, defined as X = s y Ä t yC , we can
show that the Hamiltonian in Equation 14.30 obeys {X, HBdG } = 0 , thereby
showing the system to be particle-hole symmetric, as one expects. If we
assume that the superconducting order parameter is spatially homo-
geneous over the entire surface of the TI, then the spectrum of the
Hamiltonian is
( ±vF k - m )
2
| E( k )|= + D 20 . (14.31)
Assuming that m  D 0 , then the low-energy spectrum of Equation 14.31

resembles that of a spinless px + ip y superconductor that obeys time-rever-
sal symmetry. This can best be seen by rewriting Equation 14.28 in the
single-particle eigenstates of Equation 14.27 and projecting away the low-
est-energy band. The resulting Hamiltonian is then formally equivalent to
a spinless px + ip y superconductor [112, 122]. As we are interested in deter-
mining the eigenvalues of this Hamiltonian in order to determine the zero
energy modes associated with the presence of Majorana fermions, we first
ask where the zero energy modes should exist. As Majorana fermions are
topological objects, they cannot be defined by any local order parameter.
Yet the surface of the proximity-coupled TI clearly has a superconducting
order parameter on the surface that is well-defined everywhere. Thus, we
look in the one place where the superconducting order is not well-defined:
within the vortices proliferated on the surface of the TI. Furthermore, as
we have made the case that the surface of the TI is proximity-coupled to
a superconductor, we expect the presence of h / 2e vortex-bound states on
the surface. Figure 14.5 shows a schematic view of the envisioned setup.
To proceed, we assume that there is a vortex present at a real-space posi-
tion of (0,0) with the other partner positioned off at (¥, ¥) . It is easiest to
solve for the zero modes in Equation 14.30 by rewriting the momentum
p = ( px , p y , pz ) as -iÑ = -i ( ¶ / ¶ x , ¶ / ¶ y , ¶ / ¶ z ), and changing the coordinates
FIGURE 14.5 Schematic picture of the proximity-induced superconductor on

the top of the surface state of a 3D TI with an inserted h / 2e vortex.
14.5 Majorana Fermions in 3D Time-Reversal Invariant Topological Insulators 563
from Euclidian to polar. With these transformations, the original surface

state Hamiltonian becomes
(
H surf = -iv F cos J s x + sin J s y ) ¶¶r
(14.32)
(
-iv F - sin J s x + cos J s y
¶
r ¶J
- ms0 . )
In Equation 14.32, J is the angle in position space that sweeps around the
vortex that is located at position, r. Nonetheless, we are interested in solv-
ing for the eigenvalues of the combined TI and s-wave superconductor case,
in which the Bogoliubov–de Gennes Hamiltonian is represented in polar
coordinates as
(
H BdG = -iv F cos J s x + sin J s y Ä t z ) ¶
¶r
(
-iv F - sin J s x + cos J s y Ä t z ) ¶
r ¶J
(14.33)
(
+ D 0s0 Ä cos J t x ± sin J t y . )
In Equation 14.33, it should be noted that we have explicitly set the chemi-
cal potential, m, to be zero and the superconducting pairing potential here
is expressed in polar coordinates as D 0e ± iJ . As we are particularly interested
in the bound states of H BdGx = Ex associated with the vortices, we postulate
that the wave functions have the form
r
Y ( r , J) = a ± ( J) e
-
ò 0
D0 ( r ¢ ) dr ¢/ v F
, (14.34)
where a ± depends only on J and the superconducting pairing magni-

tude only depends on the radial position, r. The four component vectors α+
and α– denote solutions if the pairing potential is Δ0 eiϑ and Δ0 e–iϑ, respec-
tively. Due to the fact that Y is an analytic function at r = 0 and D 0 = 0,
we expect a ± to be single valued and
H BdGY = 0

( ) (
= D 0 éêi cos J s x + sin J s y Ä t z + s0 Ä cos J t x ± sin J t y ùú Y.
ë û )
(14.35)
In Equation 14.35, we have assumed that a ± is J-independent, as we are

only interested in the lowest angular momentum state, or pJ = 0. By setting
both the superconducting pairing amplitude, D 0 , and the position, r, to zero,
we are now capable of solving the eigenvalue problem at hand. Rewriting
Equation 14.35 slightly we have
( ) (
éi cos J s x + sin J s y Ä t z + s0 Ä cos J t x ± sin J t y ù a ± = 0. (14.36)
ëê úû )
In examining Equation 14.36, we can find that—up to a phase—the only pos-

sible J-independent eigenvectors for a ± are
a + = ( 0,1, -i,0 ) (14.37)

T

and
a - = (1,0,0, -i ) (14.38)
T

for D(r , J) = D 0 (r )e ± iJ . Therefore, we arrive at the final form of the zero energy
Majorana mode for D(r , J) = D 0 (r )e + iJ as
Y = eip/4 ( 0,1, -i,0 ) e

T
-
ò
0
D0 ( r ¢ )/ v F dr ¢
(14.39)
which is not spin-polarized, as the particle and hole possess the same spin.
In Equation 14.39, we have added the phase p / 4, such that XY = Y . Hence,
the zero energy mode described by Y is its own charge conjugate, and indeed
obeys the properties associated with Majorana bound states. The field operator
ĝ that is comprised of fermionic creation and annihilation operators and con-
structed from the wave function given by Equation 14.39 then satisfies gˆ = gˆ † .
14.6 EXPERIMENTAL PROGRESS IN PROXIMITY-

COUPLED 3D TIME-REVERSAL INVARIANT
TOPOLOGICAL INSULATORS
One may easily infer that when considering both topological materials and
conventional s-wave superconductors, there are many available choices. The
recent focus on the proximity effects, however, has naturally been on 3D
time-reversal invariant topological materials, such as Bi2Se3. Prior to being
able to actually observe Majorana fermions, one must be able to identify the
underlying unconventional superconductivity that is necessary to produce
vortex excitations capable of harboring Majorana bound states. Given the
task of observing induced unconventional superconductivity in time-rever-
sal invariant TIs, the initial experimental efforts have focused on depositing
s-wave superconductors directly onto TI crystals, thin films, and nanostruc-
tures with the intent of observing the topology of the bands in the pres-
ence of superconductivity. This focus is due to the fact that it is difficult to
produce insulating TIs i.e. where the Fermi energy lies within the bulk gap.
Efforts quickly focused on thin films, nanowires, and exfoliated TIs where
the intrinsic doping may be mitigated. In order to reach the topological
regime, these devices were produced on substrates that allowed for backgat-
ing between superconducting contacts.
Nonetheless, an important first step towards the observation of
Majorana fermions in proximity-coupled topological systems is to observe
14.6 Experimental Progress in Proximity-Coupled 3D 565
superconducting behavior in these systems. To this end, numerous groups

have now demonstrated the presence of a proximity-induced superconduct-
ing state with transport behavior consistent with Cooper pair tunneling into
the topological surface state of Bi2Se3 and Bi2Te3 [123–125], and strained
HgTe [119, 126, 127]. These systems show clear evidence of superconduc-
tivity, such as a Fraunhofer pattern [123, 125], and the appearance of only
even Shapiro steps under microwave radiation [117, 119], as well as univer-
sal skewness (non-sinusoidal current-phase relation) over a large range of
parameters [127], which could indicate transport through the topological
Andreev or Majorana bound states.
Often, however, in Bi2Se3 and Bi2Te3, the Fermi energy is well outside of
the topological regime in the bulk bands. On the other hand, by combina-
tions of electrical and chemical gating, the Fermi level may be pulled into
the topological regime, as is evidenced by the observation of the Dirac point,
as revealed through the ensuing peak in the resistance and change in the
sign of the Hall voltage as the gate voltage is modulated in the presence of
electrical transport [128]. With the ability to access the topological regime
within the 3D TI, a supercurrent is observed whose critical current follows
the gate-induced changes in the position of the chemical potential within the
topological and non-topological bands. Thus, when the chemical potential
is below the bottom of the conduction band, the supercurrent is carried by
the surface states, until the chemical potential enters the bulk bands and the
surface becomes sufficiently disordered that the surface orbitals hybridize
with bulk orbitals, thereby removing the topological protection of the sur-
face states. The topological and non-topological contributions to transport
measurements are explained by 3D quantum transport calculations. These
confirm that the supercurrents are carried by the topological surface states
by demonstrating the robustness of the critical current within the topologi-
cal regime in the presence of time-reversal preserving impurity disorder
which disappears as the chemical potential enters the bulk band structure.
More recent advances in proximity-coupled systems have been enabled
by the use of materials growth in which the TI is grown directly on top of
a superconductor, as is the case with the TI Bi2Se3 and the s-wave super-
conductor NbSe2 [129]. In this case, clear superconducting gaps appear on
the surface of the heterostructure, indicating that superconducting Cooper
pairs have indeed tunneled into the TI, while the band crossing in the sur-
face states is still preserved, as the addition of s-wave superconductivity
does not break time-reversal symmetry. Subsequent studies seeking signs of
unconventional superconductivity examined the measured superconduct-
ing gap as a function of the angle at the Fermi surface. However, no signs
of unconventional superconductivity were observed [130]. This is unusual,
as one expects to see a clear angular dependence of the pairing magnitude
as the Fermi surface is traversed. Therefore, in this case, it is proper to ask
which order parameter develops in the TI when paired with an s-wave super-
conductor. To answer this question, we can make a very simple symmetry
argument as to the correct form of the order parameter: within the TI, we
have broken SU(2) symmetry due to the presence of strong SO coupling,
which allows both the singlet and triplet order to coexist within the TI.
Furthermore, as we have broken inversion symmetry at both interfaces,
inter-orbital and intra-orbital pairing is allowed by symmetry. Additionally,
the symmetry constructed order parameter must obey time-reversal sym-
metry as this has not been broken. Thus, the order parameter that accu-
mulates in proximity-coupled TI heterostructures is s + px ± ip y , which can
essentially be understood as two copies of time-reversal breaking p-wave
superconductivity with different chiralities, as discussed in the previous sec-
tion, and one would expect to see isotropy in the Fermi surface, as the angle
around the Fermi surface is changed as the s-wave order parameter is isotro-
pic at the Fermi surface.
One potential way of eliminating the isotropic s-wave component so that
the unconventional p-wave part may be more clearly observed is by break-
ing the underlying time-reversal symmetry. A simple method of breaking the
underlying time-reversal symmetry is to add magnetic dopants to the TI.
Yet this still does not produce purely p-wave superconductivity, as broken
time-reversal symmetry solely guarantees that the two components of the
superfluid are no longer equal to one another, and this is no guarantee that
the isotropic s-wave component will disappear. Recent theoretical work has
illustrated that when magnetic dopants are added to a thin-film proximity-
coupled TI, the bulk states disappear as the magnetization increases, as the
singlet state becomes energetically unfavorable with increasing magnetiza-
tion. However, due to the strong SO coupling, this is not the case for the sur-
face states. In the presence of the Zeeman field from the magnetic impurities,
the spin of the surface electrons begins to move from being purely in-plane
to out-of-plane, as one expects. However, there remains a finite projection
of the spin onto the surface of the TI and this causes the s-wave compo-
nent to persist up to larger Zeeman energies [131]. Yet, some of the most
promising signatures of not only unconventional superconductivity, but also
of Majorana fermions in TIs, arise when the proximity-coupled TI is in the
quantum anomalous Hall insulator phase, which occurs when the system is
both sufficiently magnetically doped and has broken translational invariance.
When the chemical potential is in the gap in the energy spectrum opened
by the magnetic impurities, the TI is in the quantum anomalous Hall phase
(see also Chapter 15, Volume 2), and there exists a chiral state at the edge of
the system. When this chiral edge state is proximity coupled with an s-wave
superconductor, quantum transport measurements have shown conductance
plateaus that appear at close to 0.5 e2/h consistent with theoretical predic-
tions and the existence of Majorana fermions [132].
ACKNOWLEDGMENTS
E.M.H. thanks G. Tkachov for discussions and B. Scharf for a careful reading
of the text. E.M.H. acknowledges financial support from the DFG via SFB
1170 ToCoTronics and the ENB Graduate School on Topological Insulators.
M.J.G. acknowledges financial support from the NSF under CAREER Award
ECCS-1351871 and DMR 17-20663 COOP.
References 567
REFERENCES
1. L. D. Landau and E. M. Lifshitz, Statistical Physics, Pergamon Press, Oxford,
1980.
2. K. Von Klitzing, G. Dorda, and M. Pepper, New method for high-accuracy
determination of the fine-structure constant based on quantized Hall resis-
tance, Phys. Rev. Lett. 45, 494 (1980).
3. R. B. Laughlin, Quantized Hall conductivity in two dimssensions, Phys. Rev. B
23, 5632 (1981).
4. D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs, Quantized
Hall conductance in a two-dimensional periodic potential, Phys. Rev. Lett. 49,
405 (1982).
Condensed-matter realization of the “Parity Anomaly”, Phys. Rev. Lett. 61,
2015 (1988).
6. C. L. Kane and E. J. Mele, Z2 topological order and the quantum spin Hall
effect, Phys. Rev. Lett. 95, 146802 (2005).
7. C. L. Kane and E. J. Mele, Quantum spin Hall effect in graphene, Phys. Rev.
Lett. 95, 226801 (2005).
8. B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Quantum spin hall effect and
topological phase transition in HgTe quantum wells, Science 314, 1757 (2006).
9. M. König, S. Wiedmann, C. Brüne, A. Roth, H. Buhmann, L. W. Molenkamp,
X.-L. Qi, and S.-C. Zhang, Quantum spin hall effect and topological phase
transition in HgTe quantum wells, Science 318, 766 (2007).
10. M. König, H. Buhmann, L. W. Molenkamp, T. Hughes, C.-X. Liu, X.-L. Qi, and
S.-C. Zhang, The quantum spin Hall effect: Theory and experiment, J. Phys.
Soc. Jpn. 77, 031007 (2008).
11. A. Roth, C. Brüne, H. Buhmann, L. W. Molenkamp, J. Maciejko, X.-L. Qi, and
S.-C. Zhang, Nonlocal transport in the quantum spin hall state, Science 325,
294 (2009).
12. B. Halperin, Quantized Hall conductance, current-carrying edge states, and
the existence of extended states in a two-dimensional disordered potential,
Phys. Rev. B 25, 2185 (1982).
13. A. MacDonald and P. Streda, Quantized Hall effect and edge currents, Phys.
Rev. B 29, 1616 (1984).
14. C. Wu, B. A. Bernevig, and S.-C. Zhang, Helical Liquid and the Edge of
Quantum Spin Hall Systems, Phys. Rev. Lett. 96, 106401 (2006).
15. M. Z. Hasan and C. L. Kane, Colloquium: Topological insulators, Rev. Mod.
Phys. 82, 3045 (2010).
16. X. L. Qi and S.-C. Zhang, Topological insulators and superconductors, Rev.
Mod. Phys. 83, 1057 (2011).
17. B. A. Volkov and O. A. Pankratov, Two-dimensional massless electrons in an
inverted contact, Pis’ma Zh. Eksp. Teor. Fiz. 42, 145 (1985).
18. O. Pankratov, S. V. Pakhomov, and B. A. Volkov, Supersymmetry in hetero-
junctions: Band-inverting contact on the basis of Pb1-x Sn x Te and Hg1-x Cd x Te
Sol. State Commun. 61, 93 (1987).
19. L. Fu and C. L. Kane, Topological insulators with inversion symmetry, Phys.
Rev. B 76, 045302 (2007).
20. H. Zhang, C.-X. Liu, X.-L. Qi, X. Dai, Z. Fang, and S.-C. Zhang, Topological
insulators in Bi2Se3, Bi2Te3 and Sb2Te3 with a single Dirac cone on the surface,
Nat. Phys. 5, 438 (2009).
21. B. Yan, C.-X. Liu, H.-J. Zhang, C.-Y. Yam, X.-L. Qi, T. Frauenheim, and S.-C.
Zhang, Theoretical prediction of topological insulators in thallium-based
III-V-VI2 ternary chalcogenides, EPL 90, 37002 (2010).
22. H. Lin, R. S. Markiewicz, L. A. Wray, L. Fu, M. Z. Hasan, and A. Bansil, Single-
Dirac-Cone Topological Surface States in the TlBiSe2 Class of Topological
Semiconductors Phys. Rev. Lett. 105, 036404 (2010).
23. S. Eremeev, Y. Koroteev, and E. Chulkov, Ternary thallium-based semimetal chal-

cogenides Tl-V-VI2 as a new class of three-dimensional topological insulators,
Piśma Zh. Eksp. Teor. Fiz. 91, 664 (2010).
24. S. Eremeev, Y. Koroteev, and E. Chulkov, On possible deep subsurface states in
topological insulators: The PbBi4Te7 system, Piśma Zh. Eksp. Teor. Fiz. 92, 183
(2010).
25. H. Jin, J. Song, A. Freeman, and M. Kanatzidis, Candidates for topological
insulators: Pb-based chalcogenide series, Phys. Rev. B 83, 041202 (2011).
26. Hsieh, D., Qian, D., Wray, L., Xia, Y., Hor, Y. S., Cava, R. J., and Hasan, M. Z.,
A topological Dirac insulator in a quantum spin Hall phase, Nature 452, 970
(2008).
27. D. Hsieh, Y. Xia, L. Wray et al., Observation of Unconventional Quantum Spin
Textures in Topological Insulators, Science 323, 919 (2009).
tor class with a single Dirac cone on the surface, Nat. Phys. 5, 398 (2009).
29. Y. L. Chen, J. G. Analytis, J.-H. Chu et al., Experimental Realization of a Three-
Dimensional Topological Insulator, Bi2Te3, Science 325, 178 (2009).
30. T. Sato, K. Segawa, H. Guo et al., Direct Evidence for the Dirac-Cone
Topological Surface States in the Ternary Chalcogenide TlBiSe2, Phys. Rev.
Lett. 105, 136802 (2010).
31. K. Kuroda, M. Ye, A. Kimura et al., Experimental Realization of a Three-
Dimensional Topological Insulator Phase in Ternary Chalcogenide TlBiSe2,
Phys. Rev. Lett. 105, 146801 (2010).
32. Y. L. Chen, Z. K. Liu, J. G. Analytis et al., Single Dirac Cone Topological
Surface State and Unusual Thermoelectric Property of Compounds from a
New Topological Insulator Family, Phys. Rev. Lett. 105, 266401 (2010).
33. C. Brüne, C. X. Liu, E. G. Novik et al., Quantum Hall effect from the topologi-
cal surface states of strained bulk HgTe, Phys. Rev. Lett. 106, 126803 (2011).
34. C. Brüne, C. Thienel, M. Stuiber et al., Dirac-screening stabilized surface-state
transport in a topological insulator, Phys. Rev. X 4, 041045 (2014).
35. S. Souma, K. Eto, M. Nomura et al., Topological surface states in lead-based
ternary telluride Pb(Bi1−xSbx)2Te4, Phys. Rev. Lett. 108, 116801 (2012).
36. K. Kuroda, H. Miyahara, M. Ye et al., Experimental verification of PbBi2Te4 as
a 3D topological insulator, Phys. Rev. Lett. 108, 206803 (2012).
37. C. Fang, M. J. Gilbert, and B. A. Bernevig, Bulk topological invariants in non-
interacting point group symmetric insulators, Phys. Rev. B 86, 115112 (2012).
38. T. L. Hughes, E. Prodan, and B. A. Bernevig, Inversion-symmetric topological
insulators, Phys. Rev. B 83, 245132 (2011).
39. C.-X. Liu, R. X. Zhang, and B. K. VanLeeuwen, Topological nonsymmorphic
crystalline insulators, Phys. Rev. B 90, 085304 (2014).
40. L. Fu, Topological crystalline insulators, Phys. Rev. Lett. 106, 106802 (2011).
41. S.-Y. Xu, C. Liu, N. Alidoust et al., Observation of topological crystalline
insulator phase in the lead tin chalcogenide Pb1–xSn xTe material class, Nat.
Commun. 3, 1192 (2012).
42. Y. Ando and L. Fu, Topological crystalline insulators and topological super-
conductors: From concepts to materials, Ann. Rev. Conden. Mat. Phys. 6, 361
(2015).
43. Z. Wang, Y. Sun, X.-Q Chen et al., Dirac semimetal and topological phase
transitions in A3Bi (a = Na, k, Rb), Phys. Rev. B 85, 195320 (2012).
44. Z. K. Liu, B. Zhou, Y. Zhang et al., Discovery of a three-dimensional topologi-
cal Dirac semimetal, Na3Bi, Science 343, 864 (2014).
45. Z. K. Liu, J. Jiang, B. Zhou et al., A stable three-dimensional topological Dirac
semimetal Cd3As2, Nat. Mater. 13, 677 (2014).
46. P. Tang, Q. Zhou, G. Xu, and S.-C. Zhang, Dirac fermions in an antiferromag-
netic semimetal, Nat. Phys. 12, 1100 (2016).
47. P. Wadley, B. Howells, J. Zelezny et al., Electrical switching of an antiferromag-
net, Science 351, 587 (2016).
References 569
48. A. A. Burkov and L. Balents, Weyl semimetal in a topological insulator, Phys.

Rev. Lett. 107, 127205 (2011).
49. C. Fang, M. J. Gilbert, X. Dai, and B. A. Bernevig, Multi-Weyl topological semi-
metals stabilized by point group symmetry, Phys. Rev. Lett. 108, 266802 (2012).
50. B. Q. Lv, H. M. Weng, B. B. Fu et al., Experimental discovery of Weyl semimetal
TaAs, Phys. Rev. X 5, 031013 (2015).
51. L. X. Yang, Z. K. Liu, H. Peng et al., Weyl semimetal phase in the non-centro-
symmetric compound TaAs, Nat. Phys. 11, 728 (2015).
52. A. A. Soluyanov, D. Gresch, Z. Wang et al., Type-II Weyl semimetals, Nature
527, 495 (2015).
54. C. Brüne, A. Roth, H. Buhmann et al., Spin polarization of the quantum spin
Hall edge states, Nat. Phys. 8, 485 (2012).
55. A. R. Mellnik, J. S. Lee, A. Richardella et al., Spin-transfer torque generated by
a topological insulator, Nature 511, 449 (2014).
56. Y. Fan, P. Upadhyaya, X. Kou et al., Magnetization switching through giant
spin-orbit torque in a magnetically doped topological insulator heterostruc-
ture, Nat. Mater. 13, 699 (2014).
57. Z. Wu and J. Li, Spin-related tunneling through a nanostructured electric-
magnetic barrier on the surface of a topological insulator, Nanoscale Res. Lett.
7, 90 (2012).
58. C. H. Li, O. M. J. van’t Erve, J. T. Robinson, Y. Liu, L. Li, and B. T. Jonker,
Electrical detection of charge-current-induced spin polarization due to spin-
momentum locking in Bi2Se3, Nat. Nanotechnol. 9, 218 (2014).
59. J. Tian, I. Miotkowski, S. Hong, and Y. Chen, Electrical injection and detection
of spin-polarized currents in topological insulator Bi2Te2Se, Sci. Rep. 5, 14293
(2015).
60. B. Scharf, A. Matos-Abiague, J. E. Han, E. M. Hankiewicz, and I. Žutić,
Tunneling Planar Hall Effect in Topological Insulators: Spin Valves and
Amplifiers, Phys. Rev. Lett. 117, 166806 (2016).
61. M. Gmitra, S. Konschuh, C. Ertler, C. Ambrosch-Draxl, and J. Fabian, Band-
structure topologies of graphene: Spin-orbit coupling effects from first prin-
ciples, Phys. Rev. B 80, 235431 (2009).
62. S. Konschuh, M. Gmitra, and J. Fabian, Tight-binding theory of the spin-orbit
coupling in graphene, Phys. Rev. B 82, 245412 (2010).
63. C.-C. Liu, W. Feng, and Y. Yao, Quantum spin Hall effect in silicene and two-
dimensional germanium, Phys. Rev. Lett. 107, 076802 (2011).
64. P. Vogt, P. De Padova, C. Quaresima et al., Silicene: Compelling experimen-
tal evidence for graphenelike two-dimensional silicon, Phys. Rev. Lett. 108,
155501 (2012).
65. L. Zhang, P. Bampoulis, A. N. Rudenko et al., Structural and electronic proper-
ties of germanene on MoS2, Phys. Rev. Lett. 116, 256804 (2016).
66. Y. Xu, B. Yan, H.-J. Zhang et al., Large-gap quantum spin Hall insulators in tin
films, Phys. Rev. Lett. 111, 136804 (2013).
67. Z. Fei, T. Palomaki, S. Wu et al., Edge conduction in monolayer WTe2, Nat.
Phys. 13, 677 (2017).
68. Z.-Y. Jia, Y.-H. Song, X.-B. Li et al., Direct visualization of a two-dimensional
topological insulator in the single-layer 1T′−WTe2, Phys. Rev. B 96, 041108
(2017).
69. S. Wu, V. Fatemi, Q. D. Gibson et al., Observation of the quantum spin Hall
effect up to 100 kelvin in a monolayer crystal, Science 359, 76 (2018).
70. F. Reis, G. Li, L. Dudy et al., Bismuthene on a sic substrate: A candidate for a
high-temperature quantum spin Hall material, Science 357, 287 (2017).
71. F. Dominguez, B. Scharf, G. Li, J. Schäfer, R. Claessen, W. Hanke, R. Thomale,
and E. M. Hankiewicz, Testing topological protection of edge states in hexago-
nal quantum spin Hall candidate materials, Phys. Rev. B 98, 161407(R) (2018).
72. G. Li, W. Hanke, E. M. Hankiewicz, F. Reis, J. Schäfer, R. Claessen, C. Wu, and

R. Thomale, Theoretical paradigm for the quantum spin Hall effect at high
temperatures, Phys. Rev. B 98, 165146 (2018).
73. T. Zhou, J. Zhang, H. Jiang, I. Žutić, and Z. Yang, Giant spin-valley polarization
and multiple Hall effect in functionalized bismuth monolayers, npj Quantum
Materials 3, 39 (2018).
74. B. Büttner, C. X. Liu, G. Tkachov et al., Single valley Dirac fermions in zero-
gap HgTe quantum wells, Nat. Phys. 7, 418 (2011).
75. J. Chu and A. Sher, Physics and Properties of Narrow Gap Semiconductors,
Springer, New York, 2008.
76. D. G. Rothe, R. W. Reinthaler, C.-X. Liu, L. W. Molenkamp, S.-C. Zhang, and
E. M. Hankiewicz, Fingerprint of different spin-orbit terms for spin transport
in HgTe quantum wells. New J. Phys. 12, 065012 (2010).
77. D. R. Candido, M. Kharitonov, C. Egues, and E. M. Hankiewicz, Paradoxical
extension of the edge states across the topological phase transition due to
emergent approximate chiral symmetry in a quantum anomalous Hall system,
Phys. Rev. B 98, 161111(R) (2018).
78. M. V. Berry and R. J. Mondragon, Neutrino billiards: time-reversal symmetry-
breaking without magnetic fields, Proc. R. Soc. Lond. A Math. Phys. Sci. 412,
53 (1987).
79. G. Tkachov and E. M. Hankiewicz, Ballistic Quantum Spin Hall State and
Enhanced Edge Backscattering in Strong Magnetic Fields, Phys. Rev. Lett. 104,
166803 (2010).
80. B. Scharf, A. Matos-Abiague, and J. Fabian, Magnetic properties of HgTe quan-
tum wells, Phys. Rev. B 86, 075418 (2012).
81. B. Scharf, A. Matos-Abiague, I. Žutić, and J. Fabian, Probing topological transi-
tions in HgTe/CdTe quantum wells by magneto-optical measurements, Phys.
Rev. B 91, 235433 (2015).
82. J. Maciejko, X.-L. Qi, and S.-C. Zhang, Magnetoconductance of the quantum
spin Hall state, Phys. Rev. B 82, 155310 (2010).
83. S. Datta, Electronic Transport in Mesoscopic Systems, Cambridge University
Press, Cambridge, UK, 1995.
84. I. Knez and R.-R. Du, Quantum spin Hall effect in inverted InAs/GaSb quan-
tum wells, Front. Phys. 7, 200 (2012).
85. K. C. Nowack, E. M. Spanton, M. Baenninger et al., Imaging currents in
HgTe quantum wells in the quantum spin Hall regime, Nat. Mater. 12, 787
(2013).
86. S. Murakami, N. Nagaosa, and S.-C. Zhang, Dissipationless quantum spin cur-
rent at room temperature, Science 301, 1348 (2003).
88. E. M. Hankiewicz, L. W. Molenkamp, T. Jungwirth, and J. Sinova, Manifestation
of the spin Hall effect through charge-transport in the mesoscopic regime,
Phys. Rev. B 70, 241301(R) (2004).
89. E. M. Hankiewicz, J. Li, T. Jungwirth, Q. Niu, S.-Q. Shen, and J. Sinova, Charge
Hall effect driven by spin-dependent chemical potential gradients and Onsager
relations in mesoscopic systems, Phys. Rev. B 72, 155305 (2005).
90. C. Brüne, A. Roth, E. G. Novik et al., Evidence for the ballistic intrinsic spin
Hall effect in HgTe nanostructures, Nat. Phys. 6, 448 (2010).
91. C.-X. Liu, X.-L. Qi, H. J. Zhang, X. Dai, Z. Fang, and S.-C. Zhang, Model
Hamiltonian for topological insulators, Phys. Rev. B 82, 045122 (2010).
92. G. Tkachov, C. Thienel, V. Pinneker et al., Backscattering of Dirac fermions in
HgTe quantum wells with a finite gap, Phys. Rev. Lett. 106, 076802 (2011).
93. G. Tkachov and E. M. Hankiewicz, Weak antilocalization in HgTe quantum
wells and topological surface states: Massive versus massless Dirac fermions,
Phys. Rev. B 84, 035444 (2011).
References 571
94. P. Adroguer, W. E. Liu, D. Culcer, and E. M. Hankiewicz, Conductivity cor-

rections for topological insulators with spin-orbit impurities: Hikami-larkin-
nagaoka formula revisited, Phys. Rev. B 92, 241402 (2015).
95. L. Fu, C. L. Kane, and E. J. Mele, Topological Insulators in Three Dimensions,
Phys. Rev. Lett. 98, 106803 (2007).
96. J. E. Moore and L. Balents, Topological invariants of time-reversal-invariant
band structures, Phys. Rev. B 75, 121306(R) (2007).
97. Y. Xu, I. Miotkowski, C. Liu et al., Observation of topological surface state
quantum Hall effect in an intrinsic three-dimensional topological insulator,
Nat. Phys. 10, 956 (2014).
98. X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Topological field theory of time-rever-
sal invariant insulators. Phys. Rev. B 78, 195424 (2008).
99. W.-K. Tse and A. H. MacDonald, Giant magneto-optical Kerr effect and uni-
versal faraday effect in thin-film topological insulators, Phys. Rev. Lett. 105,
057401 (2010).
100. G. Tkachov and E. M. Hankiewicz, Anomalous galvanomagnetism, cyclotron
resonance, and microwave spectroscopy of topological insulators, Phys. Rev. B
84, 035405 (2011).
101. L. Wu, M. Salehi, N. Koirala, J. Moon, S. Oh, and N. P. Armitage, Quantized
Faraday and Kerr rotation and axion electrodynamics of a 3D topological insu-
lator, Science 354, 1124 (2016).
102. V. Dziom, A. Shuvaev, A. Pimenov et al., Observation of the universal magne-
toelectric effect in a 3D topological insulator, Nat. Commun. 8, 15197 (2017).
103. G. Moore and N. Read, Nonabelions in the fractional quantum Hall effect,
Nuc. Phy. B 360, 362 (1991).
104. V. Mourik, K. Zuo, S. M. Frolov, S. R. Plissard, E. P. A. M. Bakkers, and L.
P. Kouwenhoven, signatures of Majorana fermions in hybrid superconductor-
semiconductor nanowire devices, Science 336, 1003 (2012).
105. J. D. Sau, R. M. Lutchyn, S. Tewari, and S. Das Sarma, Generic new platform
for topological quantum computation using semiconductor heterostructures,
Phys. Rev. Lett. 104, 040502 (2010).
106. Y. Oreg, G. Refael, and F. von Oppen, Helical liquids and Majorana bound
states in quantum wires, Phys. Rev. Lett. 105, 177002 (2010).
107. S. Albrecht, A. Higginbotham, M. Madsen et al., Exponential protection of
zero modes in Majorana islands, Nature 531, 206 (2016).
108. S. Nadj-Perge, I. K. Drozdov, J. Li, H. Chen, S. Jeon, J. Seo, A. H. MacDonald,
B. A. Bernevig, and A. Yazdani, Observation of Majorana fermions in ferro-
magnetic atomic chains on a superconductor, Science 346, 602 (2014).
109. G. L. Fatin, A. Matos-Abiague, B. Scharf, and I. Žutić, Wireless Majorana bound
states: from magnetic tunability to braiding, Phys. Rev. Lett. 117, 077002 (2016).
110. N. Read and D. Green, Paired states of fermions in two dimensions with break-
ing of parity and time-reversal symmetries and the fractional quantum Hall
effect, Phys. Rev. B 61, 10267 (2000).
111. C. Nayak, S. H. Simon, A. Stern, M. Freedman, and S. Das Sarma, Non-Abelian
anyons and topological quantum computation, Rev. Mod. Phys. 80, 1083
(2008).
112. L. Fu and C. L. Kane, Superconducting proximity effect and Majorana fermions
at the surface of a topological insulator, Phys. Rev. Lett. 100, 096407 (2008).
113. F. Crépin, B. Trauzettel, and F. Dolcini, Signatures of Majorana bound states in
transport properties of hybrid structures based on helical liquids, Phys. Rev. B
89, 205115 (2014).
114. S. Mi, D. I. Pikulin, M. Wimmer, and C. W. J. Beenakker, Proposal for the
detection and braiding of Majorana fermions in a quantum spin Hall insula-
tor, Phys. Rev. B 87, 241405 (2013).
115. S. Hart, H. Ren, T. Wagner et al., Induced superconductivity in the quantum
spin Hall edge, Nat. Phys. 10, 638 (2014).
116. G. Tkachov, P. Burset, B. Trauzettel, and E. M. Hankiewicz, Quantum interfer-

ence of edge supercurrents in a two-dimensional topological insulator, Phys.
Rev. B 92, 045408 (2015).
117. E. Bocquillon, R. S. Deacon, J. Wiedenmann et al., Gapless Andreev bound
states in the quantum spin Hall insulator HgTe, Nat. Nanotechnol. 12, 137
(2016).
118. S. Hart, H. Ren, M. Kosowsky et al., Controlled finite momentum pairing and
spatially varying order parameter in proximitized HgTe quantum wells, Nat.
Phys. 13, 87 (2017).
119. R. S. Deacon, J. Wiedenmann, E. Bocquillon et al., Josephson radiation from
gapless Andreev bound states in HgTe-based topological junctions, Phys. Rev.
X 7, 021011 (2017).
120. F. Dominguez, O. Kashuba, E. Bocquillon et al., Josephson junction dynamics
in the presence of 2π- and 4π-periodic supercurrents, Phys. Rev. B 95, 195430
(2017).
121. G. Tkachov and E. M. Hankiewicz, Spin-helical transport in normal and
superconducting topological insulators, Phys. Status Solidi 250, 215 (2013).
122. J. Alicea, New directions in the pursuit of Majorana fermions in solid state
systems, Rep. Prog. Phys. 75, 076501 (2012).
123. M. Veldhorst, M. Snelder, M. Hoek et al., Josephson supercurrent through a
topological insulator surface state, Nat. Mater. 11, 417 (2012).
124. B. Sacépé, J. B. Oostinga, J. Li et al., Gate-tuned normal and superconducting
transport at the surface of a topological insulator. Nat. Commun. 2, 575 (2011).
125. J. R. Williams, A. J. Bestwick, P. Gallagher et al., Unconventional Josephson
effect in hybrid superconductor-topological insulator devices, Phys. Rev. Lett.
109, 056803 (2012).
126. L. Maier, J. B. Oostinga, D. Knott et al., Induced superconductivity in the three-
dimensional topological insulator HgTe, Phys. Rev. Lett. 109, 186806 (2012).
127. I. Sochnikov, L. Maier, C. A. Watson et al., Nonsinusoidal current-phase rela-
tionship in Josephson junctions from the 3D topological insulator HgTe, Phys.
Rev. Lett. 114, 066801 (2015).
128. S. Cho, B. Dellabetta, A. Yang et al., Symmetry protected Josephson super-
currents in three-dimensional topological insulators, Nat. Commun. 4, 1689
(2013).
129. M.-X. Wang, C. Liu, J.-P. Xu et al., The coexistence of superconductivity and
topological order in the Bi2Se3 thin films, Science 336, 52 (2012).
130. S.-Y. Xu, N. Alidoust, I. Belopolski et al., Momentum-space imaging of Cooper
pairing in a half-Dirac-gas topological superconductor, Nat. Phys. 10, 943
(2014).
131. Y. Kim, T. M. Philip, M. J. Park, and M. J. Gilbert, Topological superconductiv-
ity in an ultrathin, magnetically-doped topological insulator proximity cou-
pled to a conventional superconductor, Phys. Rev. B 94, 235434 (2016).
132. Q. L. He, L. Pan, A. L. Stern et al., Chiral Majorana fermion modes in a quan-
tum anomalous Hall insulator–superconductor structure, Science 357, 294
(2017).
15
Quantum
Anomalous Hall
Effect in Topological
Insulators
15.1 Background and Theory 574

15.2 Development of QAHE Materials 576
15.3 Structure, Magnetism and Temperature Scale 581
15.4 Ferromagnetic Insulator/TI Heterostructures 586
Acknowledgments 587
References 588
573
574 Chapter 15. Quantum Anomalous Hall Effect in Topological Insulators
15.1 BACKGROUND AND THEORY

The flow of an electrical current in a typical metal experiences dissipation
due to scattering of electrons from defects in the crystal lattice. Systems
where electricity can flow without resistance are rather unique, and require
a mechanism that results in phase coherent transport. The best known
examples are superconductors, where electrons form pairs and condense
into a Bose–Einstein condensate (see Chapter 16, Volume 1). The quantum
Hall effect (QHE) is another example of dissipationless electrical transport
that can occur when a large magnetic field is applied to a high mobility
two-dimensional (2D) electron gas at absolute zero. Here, the Landau levels
caused by the magnetic field result in edge states with a Hall conductance
quantized in integer multiples of e 2 / h and where the longitudinal 4-point
resistance goes to zero, though the 2-point resistance does not. The integer
multiples of the Hall conductance arise from the topology of the wave func-
tions of the quantum Hall state, and are known as Chern numbers [1]. The
realization that the QHE could be understood by using concepts of topology
led to a profound reconceptualization of quantum phases of matter. It is now
understood that distinct phases can be defined solely by differences in their
topology, in the absence of any symmetry breaking.
As QHE depends on a large applied magnetic field, it was natural to ask
if a material could possess similar quantized transport inherently, without
Landau levels from external fields. This possibility was first raised in the
context of a conventional semiconductor heterostructure [2]. Haldane pro-
vided another, better-recognized theoretical prediction of this possibility,
showing that a zero-field quantized Hall conductivity could be achieved on
a graphene-like lattice, if time reversal symmetry was broken by applying
an opposite magnet flux on each sublattice, with the total flux summing
to zero [3]. In graphene, there are two inequivalent Dirac cones at the K
and K′ = -K points in the Brillouin zone. The opposite fields on each sub-
lattice would open a gap at each Dirac point, adding an oppositely signed
mass term to the Hamiltonian each cone. This in turn leads to each cone
contributing a half integer quantum Hall conductance that sums to a total
Hall conductance of e 2 / h . The key to observing this in a realistic system
is breaking time reversal symmetry in a system with a topologically non-
trivial band structure.
Kane and Mele explored what happens in graphene when spin-orbit
coupling is considered [4]. The spin-orbit term in the Hamiltonian was found
to have a form similar to the staggered field in the Haldane model, except
that it was spin-dependent. As in the spin Hall effect, spin-orbit coupling
acts like an effective magnetic field that depends on the direction of the spin
polarization. This takes the chiral edge state of the spinless Haldane model,
and splits it into two oppositely rotating edge states with opposite spins. This
was termed the quantum spin Hall (QSH) effect. Further, this state is topo-
logically distinct from an ordinary band insulator, which guarantees that
the edge states are robust against disorder (see Chapter 14, Volume 2). It can
be characterized by a topological invariant called Z2, which is similar to the
15.1 Background and Theory 575
Chern number for the QHE state [5]. QSH was also predicted in strained
semiconductors with large spin-orbit coupling, and was soon after experi-
mentally observed in the HgTe/CdTe quantum wells [6, 7].
The QSH state can only be observed at low temperatures, because it
is possible for spin non-conserving scattering, such as inelastic scattering,
to scatter carriers from one edge channel into the other. The longitudinal
resistance measured in a two-terminal device is determined by the quantum
conductance of two channels, 2e 2 / h . It also differs from the Haldane model.
Any non-zero Hall conductance breaks time reversal, so a system displaying
it must break time reversal. Spin-orbit coupling preserves time reversal, so
the Hall conductance is zero for QSH, in the absence of an applied external
field.
The QSH state is also called a 2D topological insulator (TI), to distin-
guish it from 3D TIs which were identified soon afterwards [8]. In a 3D TI,
there are surface states within the bulk bandgap that exist on all the sur-
faces of the crystal. These states have a Dirac dispersion, and a spin polariza-
tion that is locked perpendicular to their momentum. For a 3D TI, such as
Bi2Se3, these surface states form a single Dirac cone centered in k-space at
the Γ point, where the spin rotates in a left-handed sense for energies above
the Dirac point [9]. The degeneracy of the Dirac point in Bi2Se3 is protected
by time-reversal and inversion symmetry. When time reversal is broken,
however, a gap can be opened at the Dirac point which can be described
by a massive Dirac Hamiltonian, where the sign of the mass is determined
by whether the magnetization perpendicular to the surface is pointing out-
ward, or inward into the surface [10].
Jackiw and Rebbi predicted long ago that the mass domain wall of a 1D
Dirac system carries a bound state [11]. A simple generalization to a 2D Dirac
system predicts the presence of a 1D chiral (one-way propagating) mode at
the mass domain wall. The 2D Dirac surface states of 3D TIs are a natural
test bed for these predictions. This is illustrated in Figure 15.1. Consider a
3D TI thin film in proximity with two oppositely oriented, perpendicular-
to-plane magnetized domains of a ferromagnet. The surface states under the
magnetic domains can be gapped by exchange coupling, and acquire a mass,
whose sign is dependent on the direction of magnetization of the overly-
ing domain. Therefore, the region underneath the magnetic domain wall
is expected to create a mass domain wall that carries a chiral mode. This
1D mode exists within the magnetic gap, moves in one direction, protected
from backscattering, and is therefore dissipationless. Practically, however,
this requires that the chemical potential be placed inside the magnetic gap,
and that there are no other states at the chemical potential to scatter into.
Obviously, using a metallic ferromagnet on top of the TI is therefore prob-
lematic. One way to accomplish this is to make the TI itself ferromagnetic
by magnetic doping, and to control the carrier density so that chemical
potential lies inside the magnetic gap. If the magnetization points up, it will
point outward from the top surface, and inward through the bottom surface,
resulting in a 1D edge mode around the outside edges of the sample. This is
the origin of the quantum anomalous Hall effect (QAHE), which was first
FIGURE 15.1 A chiral state at the domain wall between two oppositely oriented
ferromagnets on a TI surface. In the absence of a domain wall the whole surface is
gapped and no state exists.
observed in Cr doped (Bi1-xSbx)2Te3 [12, 13]. When two oppositely oriented

magnetic domains are created in a ferromagnetic TI, as expected, quantized
transport can be observed through a chiral mode localized at the domain
wall [14, 15]. Another approach is using an insulating ferromagnet. As dis-
cussed later, much recent research has focused on interfacing 3D topological
insulators with insulating ferromagnets, though to date such a chiral mode
has not yet been observed using this method.
15.2 DEVELOPMENT OF QAHE MATERIALS

Bi2Se3, Bi2Te3, and Sb2Te3 are the most widely used 3D TI materials. All are
narrow bandgap semiconductors, with a bulk bandgap up to ~0.3 eV for
Bi2Se3, and have a single Dirac surface state cone centered at the Γ point in
k-space [16]. They are layered materials with a rhombohedral crystal struc-
ture and van der Waals bounds between layers. A major drawback of MBE
grown TI thin films of Bi2Se3 (Bi2Te3) is unintentional doping arising from
Se (Te) vacancies that results in the chemical potential lying in the bulk con-
duction band. These bulk carriers mask signatures of transport through the
surface states. Furthermore, the Dirac point for Bi2Te3 lies submerged below
the bulk valence band maximum. In contrast, thin films of Sb2Te3 have hole-
type carriers due to Sb-Te antisite defects that place the chemical poten-
tial in the bulk valence band. Additionally, the Dirac point for Sb2Te3 lies
exposed within its bulk band gap (Figure 15.2). Taking advantage of the simi-
lar lattice constants of Bi2Te3 and Sb2Te3, growth of thin films of the alloy
(Bi1-xSbx)2Te3 showed that the non-trivial topology of the band structure
FIGURE 15.2 (a) In Bi2Te3 the Fermi energy of tends to be n-type, while the posi-
tion of the Dirac point is below the valence band edge. Sb2Te3 is p-type and the
Dirac point is in the bulk bandgap. By tuning the ratio of Bi to Sb, EF can be placed
at the Dirac point, which can be above the valence band edge. (b) An out-of-plane
magnetization, such as from magnetically doping a TI, can open a gap at the Dirac
point. If the magnetization is sufficiently large it can create a chiral edge state
around the outside edge of the sample that is the origin of the quantum anoma-
lous Hall effect.
could be preserved over the entire composition range [17]. Importantly, it

was shown that, by adjusting the Bi:Sb ratio, one could achieve ultra-low car-
rier density thin films with the chemical potential in the bulk band gap, with
an exposed Dirac point. Further, tuning of the chemical potential could then
be done using conventional gating, which is difficult to do effectively with
materials like Bi2Se3. Gating a (Bi,Sb)2Te3 sample from the conduction band
across the bulk bandgap and into the valence band reveals a gate dependence
of the longitudinal and transverse resistivity that is reminiscent of the classic
Dirac system—graphene. The longitudinal resistance reaches a maximum in
the vicinity of the Dirac point and the carrier density changes sign [18]. This
ability to control the carrier density was an important step towards separat-
ing surface state effects in electrical transport, and was critical to achieving
the QAHE.
To measure a quantum of resistance associated with the QAHE 1D
chiral mode requires freezing out any potential dissipative channels, which
implies that the chemical potential should be tuned into the surface state
magnetic gap. Growing these thin films on SrTiO3 (STO) substrates is very
helpful in this regard. The STO substrate presents itself as a natural backgate
dielectric, due to STO’s huge dielectric constant at cryogenic temperatures.
Despite the large lattice mismatch between the film and STO substrate, the
van der Waals nature of the bonds on the TI surface means that reason-
able quality TI films can be grown. This is referred to as van der Waals epi-
taxy. Also, for the QAHE, the carrier density of the TI films needs to be
low, despite the magnetic doping. Initial experiments in our group with Mn
doping of Bi2Se3 and Bi2Te3 thin films revealed an unfavorable increase in
the electron carrier density, and a propensity toward formation of second-
ary phases [19, 20]. This was in contrast to the p-type doping typically seen
with Mn doping using bulk growth techniques. Another key requirement is
the realization of a ferromagnetic phase with out-of-plane magnetic anisot-
ropy. This has been shown with doping of Bi2Te3 and Sb2Te3 thin films with
Mn, Cr and V. However, Mn doping of Bi2Se3 revealed a ferromagnetic phase
with in-plane magnetic anisotropy [19]. The extent of magnetic doping is
yet another important consideration. A large substitution of the “heavy” Bi
has been shown to cause a transition to a topologically trivial band struc-
ture [21]. Finally, yet another important requirement is an exposed Dirac
point. A Dirac point that is submerged in the bulk valence band would lead
to back-scattering through these bulk states, and therefore a deviation from
perfect quantization. In this context, band-engineering by controlling the
Bi:Sb ratio is an important step. If the magnetic dopant introduces acceptors
that requires too high of a Bi ratio to compensate for this, then the Dirac
point could end up below the valence band edge.
The choice of magnetic dopant is thus crucial for the realization of the
QAHE. Our group’s initial attempts with Mn doping of Bi2Se3 thin films
did not reveal an AHE in transport measurements, while Mn doped Bi2Te3
thin films demonstrated clear ferromagnetism with out-of-plane magnetic
anisotropy [20]. Other attempts with Gd doping of Bi2Te3 or Cr doping of
Bi2Se3 either did not show signatures of ferromagnetism or were in-plane
easy axis [22–25]. Revisiting past work from U. Mich [26] that demonstrated
the strong out-of-plane magnetic anisotropy of Cr-Sb2Te3 was a key step in
the hunt for the QAH, creating the hope that Cr doping of (Bi,Sb)2Te3 could
lead to low carrier density ferromagnetic thin films with an out-of-plane easy
axis, and satisfying key requirements for the QAH. However, the extent of
Cr doping is yet another important variable. Work from our group showed
that in the high Cr doping limit, ultrathin films of Cr-Sb2Te3 grown on
InP (111)A revealed Tc’s ~150 K, and square-shaped AHE curves, indicative of
the strong out-of-plane magnetic anisotropy [27] (Figure 15.3). Interestingly,
at He3 temperatures, these films showed longitudinal sheet resistances in the
MΩ range, raising the possibility of their use as low temperature ferromag-
netic insulators. Similar results were seen more recently from the group at
Tsinghua, in 5QL thick thin films of Cr y(Bi xSb1-x)2-yTe3 grown on STO. For a
Cr concentration of y = 0.44, they observed a Tc ~ 77 K, with a gate tunable
electrical resistivity taking maximum values over 10 MΩ [28]. In the limit of
such large doping, the strength of the spin-orbit coupling is reduced, leading
to a transition to a topologically trivial phase.
FIGURE 15.3 Transport of highly Cr doped Sb2Te3. (a) At low temperatures the resistance of the sample
becomes strongly insulating. (b) A TC of ~150 K was observed in this sample from the anomalous Hall signal.
(C) Square hystersis loops are observed, but with saturation values far from quantization. (After Kandala, A.
Transport studies of mesoscopic and magnetic topological insulators, Ph.D. thesis, 2015. With permission.)
The requirements for realizing the QAHE were first met by the group
from Tsinghua University, using a Cr doping concentration that was signifi-
cantly lower than that discussed in the previous paragraph [13]. At 30 mK,
5 QL thick films of Cr0.15(Bi0.1Sb0.9)1.85Te3 on STO revealed a zero-field pla-
teau in the transverse resistance around h/e2, when the chemical potential
was tuned to charge neutrality. This was the first experimental realization of
the QAHE. The temperature dependence of the AHE revealed a Tc ~ 15 K.
A crucial point to note is that the regime of quantization is at temperatures
almost two orders of magnitude lower than the onset of bulk ferromagne-
tism. We shall revisit this in later sections. When the chemical potential is
tuned to the Dirac point, flipping the magnetization switches the Hall pla-
teau between +h/e2 and −h/e2. The switching of the magnetization is accom-
panied by a sharp rise (2250%) in the longitudinal resistance, which, as in
the QHE, is interpreted to arise from backscattering between adjacent edge
modes via dissipative pathways in the bulk. Typical QAHE data, observed in
magnetic field sweeps, is shown in (Figure 15.4). Similar quantization of the
Hall resistance is seen in gate sweeps at zero field, with a plateau around h/e2
near charge neutrality. The quantized Hall plateau is accompanied by a drop
in the longitudinal resistance, which is another familiar signature of chiral
edge transport from the QHE. This gate dependence is in stark contrast to
the nonmagnetic TI thin films where the longitudinal resistance reaches a
maximum at charge neutrality. In these samples there was a finite zero-field
ρxx of ~0.098 h/e2, which is indicative of some dissipative current pathways.
However, these dissipative channels were localized by the application of a
strong external magnetic field (>10 T), that led to a vanishing longitudinal
resistance. As the temperature is raised, the Hall plateau deviates from per-
fect quantization, and the drop in ρxx is increasingly shallow. These are typical
signatures of edge transport in the presence of increased dissipation. These
signatures persisted up to 400 mK, indicating that the edge states persist to
temperatures well beyond the regime of perfect quantization. Similar results
were soon reproduced by several groups [29–31].
FIGURE 15.4 QAHE in Cr doped (Bi,Sb)2Te3. Despite mobilities on the order of a

few hundred cm2/Vs, quantization to parts in 104 can be achieved at dilution fridge
temperatures. (After Liu, M. et al., Sci. Adv. 2, e1600167, 2016. With permission.)
First principles calculations predicted that doping of 3D TI thin films

with Vanadium (V) would result in d-orbital impurity bands that lie in the
band gap [12]. This would preclude the realization of a QAHE, due to the
absence of an insulating ferromagnetic state. However, the experimental
work of Chang et al. showed otherwise, demonstrating a robust QAHE at
25 mK, in 4 QL thick films of (Bi0.29Sb0.71)1.89V0.11Te3 grown on STO [32]. Key
differences with Cr doping were a significantly higher Tc and lower carrier
densities for comparable doping concentrations, and a coercive field that is
larger by an order of magnitude. Furthermore, key signatures of dominant
edge state transport, such as a drop in ρxx, are seen at temperatures as high
as 5 K, making it an important first step towards enhancing the tempera-

ture scale of the QAHE. Similar results with V doping were reproduced with
growth on hydrogen passivated (111) Si substrates [33].
15.3 STRUCTURE, MAGNETISM AND

TEMPERATURE SCALE
In both Cr doped and V doped films, precise quantization has been observed,
with ρxy deviating from 1 by a few 10 thousandths and ρxx exceeding 0 also by
a few ten thousandths, in units of h/e2 [33, 34, 35, 36]. Deviations from these
values arise from transport though parallel dissipative channels in the sam-
ple. As the system is tuned away from the ideal conditions, for instance, by
changing the gate voltage or raising the temperature, dissipation increases
rapidly, and can be understood as a quantum phase transition to an insulat-
ing state [37]. Tilted field measurements provide a convenient method to
quantify ballistic edge vs. dissipative contributions in temperature and gate-
voltage regimes that do not display full quantization [31]. When the external
field rotates the sample magnetization from out-of-plane towards the in-
plane direction, the magnetic gap that stabilizes the QAHE edge state closes,
and the edge state is completely destroyed when the magnetization is forced
in-plane. This field dependence is distinct from the typical anisotropic mag-
netoresistance (AMR) observed in conventional ferromagnets, and can be
modeled using the Landauer–Büttiker formalism to quantify the proportion
of ballistic and dissipative transport (Figure 15.5).
Despite the high degree of quantization that is achievable, remarkably,
these materials are structurally far from perfect and contain a number of
defects [38]. The QAHE has been observed in TI thin films grown on a
variety of substrates, including InP, Si, and STO, but in all cases, struc-
tural defects such as dislocations and twin domains are common. Figure
15.6 shows the structural characterization of these thin films on STO by
atomic force microscopy (AFM) and transmission electron microscopy
(TEM). The AFM image of Figure 15.6 shows the presence of blobs that are
almost ~20 nm in height, in addition to the well-studied layered structure
of triangular QL’s with twinned domains. The crystallinity of these fea-
tures is seen in TEM; EDS reveals a chemical composition that is similar
to the rest of the film, suggesting that the blobs are misoriented grains
of Cr-(Bi,Sb)2Te3. Furthermore, TEM also reveals an amorphous layer at
the substrate-film interface. Unsurprisingly, electrical transport measure-
ments of these films show poor mobilities, in the range 100–200 cm 2/Vs
at cryogenic temperatures. Similar mobilities have been reported by other
groups. This is in stark contrast to traditionally studied systems of 1-D
transport in the QH regime, such as graphene and 2D electron gases that
have required orders of magnitude larger mobilities. This is a remarkable
feature of the QAHE, and highlights the difference in its origins from the
QH effect, which requires high mobilities for the creation of Landau levels
at modest magnetic fields.
FIGURE 15.5 Giant anisotropic magnetoresistance in a QAHE sample. (a) The magnetoresistance as the field
is swept from out-of-plane to in-plane can be modeled by a Landauer–Büttiker formalism to extract edge
and bulk contributions to the conduction. (b) In contrast, when the field is swept in-plane, the typical AMR
dependence of a trivial ferromagnet is observed. (c) The spatial dependence of the edge state is shown sche-
matically as the field is tilted into the plane of the sample where it is eventually destroyed as the magnetic gap
closes. (After Kandala, A. et al., Nat. Commun. 6, 7434, 2015. With permission.)
Typically, ferromagnetism arises in dilute magnetic semiconductors

via the carrier-mediated RKKY exchange interaction (see Chapters 4 and 9,
Volume 2). For QAHE samples, which are purposely tuned and gated into
the magnetic gap to eliminate bulk and surface state carriers, an interesting
question arises: why are these samples ferromagnetic? Theoretically, it was
found that the spin susceptibility of these TI materials is so large it can sta-
bilize ferromagnetic order: this is known as van Vleck ferromagnetism [12].
When the carrier density is large, when the chemical potential is near the
bulk band edges, RKKY also likely plays a role in mediating ferromagnetic
order [39, 40].
Another interesting question to address is the reason for the rela-
tively low temperatures required to access the QAH regime, despite a
FIGURE 15.6 On the left, AFM image of the surface of a Cr-(Bi,Sb)2Te3 sample grown on STO and capped
with Al. Despite roughness from step edges and defects, QAHE can be observed. On the right, corresponding
high angle annular dark field scanning TEM images. Quintuple layer structure of the TI on STO is clearly seen,
as is an amorphous region at the interface. Below, a defect in the film which appears to be a misoriented grain
which may correspond with the type of tall defects seen in the AFM.
significantly higher Tc for the onset of ferromagnetic order. Understanding

this issue is critical for developing strategies for increasing the temperature
scale. Initial samples required access to dilution fridge temperatures. The
first results from Tshingua University were reported at 30 mK, from thin
films with a Tc ~ 13 K, and similar temperature scales were observed in
subsequent experiments [13]. This is suggestive of a significantly smaller
energy gap for activation of dissipative channels. As in studies of QHE,
an activation gap may be extracted from an Arrhenius fit to the metallic
temperature dependence of the longitudinal conductivity. Measurements
on our thin films with a Tc ~ 15 K reveal an activation gap corresponding to
a temperature scale of ~190 mK, and similar numbers have been reported
elsewhere [32, 35].
Some clues to the origin of this discrepancy are provided by nano
scale measurements. Scanning tunneling spectroscopy measurements of
V-(Bi,Sb)2Te3 showed gaps around the Dirac point that varied widely in
energy on the nanometer scale spatially [41]. These gaps were interpreted as
magnetic gaps due to the V doping and implied that the distribution of the
smallest gaps could allow percolative transport paths through the sample.
Additionally, scanning nano-superconducting quantum interference device
(nano-SQUID) measurements of Cr doped samples at 250 mK revealed that
the switching behavior of the magnetization looks superparamagnetic in
character [42]. On the 10s of nm scale, the switching is granular, with island-
like domains switching independently of their neighbors, without any evi-
dence of domain wall motion, but also displaying long timescale relaxation
dynamics consistent with superparamagnetic or glassy behavior (Figure 15.7).

At higher temperatures (5 K), magnetic force microscopy measurements of
V-Sb2Te3 have reported disordered bubble-like domains but with more con-
ventional domain wall propagation [43]. Note, though, that V-Sb2Te3 is heav-
ily p-type and may be more representative of RKKY mediated magnetism
than the situation when the chemical potential is near the Dirac point.
The picture that arises is one dominated by disorder, both in terms of
local potentials due to random doping and defects, but also in magnetic
inhomogeneity. One result that stands in contrast to this is the observation
of large instantaneous jumps in the local and non-local transport resistiv-
ity when sweeping through the coercive field in the quantized regime [32].
At the lowest mK temperatures, the switching behavior appears smooth in
both ρxx and ρxy. However, in a temperature range of ~100 to 170 mK, jumps
on the order of h/e2 were seen in large macroscopic samples (Figure 15.7).
FIGURE 15.7 (a)–(h) Scanning SQUID measurements showing the superpara-

magnetic-like nature of the local magnetization during the switching transition
of Cr-(Bi,Sb)2Te3 at 250 mK [42]. (i)–(j) Instantaneous jumps in the Hall resistance
when sweeping through the the coercive field observed between ~70 and 140 mK
[34]. These jumps involve magnetization reversal in regions on the mm scale and
contrast strongly with the local magnetization seen in the scanning SQUID. (After
Liu, M. et al., Sci. Adv. 2, e16000167, 2015; Lachman, E.O. et al., Sci. Adv. 1, e1500740,
From non-local measurements, it is clear the switching involved moving the

magnetization domain wall over distances of hundreds of microns. This was
interpreted as possible evidence of macroscopic quantum tunneling into the
magnetic ground state as the field was swept through the transition. Such
quantum coherence is difficult to reconcile with the superparamagnetic-like
local magnetization seen ~100 mK higher in temperature, but could point
to strongly temperature-dependent dynamics that again may be limiting the
temperature scale of QAHE.
Another way to characterize the QAHE transition is looking at the scal-
ing behavior of the flow trajectory diagram of the σ xx vs. σ xy [30]. In general,
it is found that there is a very large increase in ρxx during the transition when
sweeping through the coercive field. This drives σ xx towards zero and, in
some samples, results in a plateau at zero in σ xy, which has been called the
C=0 or zero Hall plateau [44]. In these samples, the flow diagram resembles
that of the integer QHE following a semicircle running from (0,0) to (0, e2/h)
in σ xx and σ xy. Tilted field measurements have also been used to show the
same behavior [45]. We caution that the observation of σ xy = 0 can also result
as an arithmetical artifact of converting measured resistivity into a deduced
conductivity when a large value of ρxx coincides with ρxy approaching zero
during the transition between the integer Chern states. Others observed that
if ρxx does not exceed h/e2 during the transition, then a different flow diagram
is obtained where the semicircle connects (0, −e2/h) to (0, e2/h) approaching
(e2/h, 0) during the transition [46]. Whether this implies a different phase of
the QAHE is currently under debate.
This is relevant in the context of one fairly successful method to improve
the temperature scale of QAHE, magnetic modulation doping near the sur-
faces of TI films. Tokura’s group has shown that confining the magnetic
dopants to the surfaces, or near surface regions, of the films can result in
QAHE persisting to temperature as high ~2 K [47]. This appears to work by
confining the magnetic doping to the surfaces, where it is needed to gap the
surfaces’ states, while reducing the disorder in the bulk that could lead to
unwanted conduction pathways. Interestingly, since the magnetization of
the top and bottom surfaces are now decoupled, it has been demonstrated
that it is possible to switch only one surface, so that the magnetization is
pointing outward on both surfaces, destroying the QAHE edge state while
still leaving the surfaces gapped [48]. This has been called the axion insula-
tor state and could be relevant for a host of exotic physics. The debate about
the scaling flow diagrams is a question of whether this state has truly been
achieved. Recently, using a Cr doped bottom layer and a V doped top layer
with very different coercivities, it has been shown that the scaling behavior
of the top and bottom surfaces can be plotted independently and that they
each show the expected half integer Hall conductivity of a single gapped
surface, which cancels to zero when each surface’s magnetization is oppo-
sitely oriented [49, 50].
Another route towards increasing the temperature scale could deal with
improved control of the chemical potential. TI materials are sensitive to expo-
sure to air and water vapor, typically resulting in higher carrier densities as
the films age and oxidize. To protect them from degradation, films are often
capped in situ after growth to help preserve the properties of their surfaces.
Al (which oxidizes on exposure to air), Al2O3, and Te are common capping
materials. Ex situ atomic layer deposition (ALD) of Al2O3, to encapsulate the
films prior to lithography, is also common if top gates are to be fabricated. To
avoid possible deteriorating effects of exposure during conventional lithog-
raphy, the Hall bars are often fashioned by mechanical scratching of the thin
films. When STO is used as the substrate, a gated device can be made in this
way with a minimum of steps. With most QAHE experiments to date, thin
films with a single gate have been used. This effectively depletes the carriers
near the closer interface, but can create a potential gradient across the film.
The fabrication of dual-gated magnetic TI devices would enable independent
control of the top and bottom surface states, and thereby improved control
over the chemical potential, which would be useful given the small size of
the magnetic gap. Our group’s attempts in this direction led to the fabrica-
tion of thin film Hall devices of Cr-(Bi,Sb)2Te3 grown on STO, capped with
a top gate dielectric of ALD deposited HfO2 dielectric and evaporated Au
metal. As shown in Figure 15.8, while the thin film could be tuned to its
Dirac point with both gates, the films were likely degraded by the fabrication
process, and the maximum gate-tuned AHE was ~3 KΩ. Others have seen
similar reductions in the AHE signal after fabricating dual gates on samples
on STO [51].
15.4 FERROMAGNETIC INSULATOR/TI
HETEROSTRUCTURES
In our group, initial experiments exploring the magnetic heterostructure
route were focused on the insulating ferromagnet GdN [52]. This is a low
Curie temperature ferromagnet that was deposited on Bi2Se3 thin films by
reactive ion sputtering at ambient temperatures. Controlling the nitrogen
composition has been previously shown to be an excellent knob for manipu-
lating its electrical conductivity. The ability to deposit GdN at ambient tem-
peratures was of particular interest, since this prevents the diffusion of the
magnetic species into the TI thin film, as confirmed by high-resolution elec-
tron energy loss spectroscopy measurements. This is an important consid-
eration to differentiate proximity exchange effects with effects attributed to
magnetic doping, which may have adverse effects on the topologically non-
trivial band structure. The magnetic characterization of the heterostructures
by SQUID magnetometry revealed a ferromagnetic phase with Tc ~ 13 K and
an in-plane easy axis. We were also able to fabricate Hall bars for electri-
cal characterization, with bare and GdN capped channels. This enabled a
direct comparison of transport, and revealed the suppression of weak anti-
localization in the magnetically capped layer. A key drawback of these initial
attempts was the use of Bi2Se3 thin films which had their chemical potential
up in the bulk conduction band, due to unintentional doping in the as-grown
films. For transport experiments, this precludes the ability to observe clean
signatures of time-reversal symmetry breaking effects in the surface states.
15.4 Ferromagnetic Insulator/TI Heterostructures 587
FIGURE 15.8 Dual gating of a Cr-(Bi, Sb)2Te3 TI film. (a) R xx vs. top and backgate
voltages. (b) Maximum in R xx crossing the Dirac point. (c) R xy hysteresis vs. top gate
voltage at a fixed backgate. (After Richardella, A. et al. APL Mater. 3, 83303, 2015.
With permission.)
Our efforts since have focused on lowering the carrier density in TI thin
films, and the use of alternate high resistivity magnets such as (Ga,Mn)As
[53] and YIG. Spin pumping measurements on TI/YIG heterostructures have
been useful to characterize the nature of the spin interaction at the interface
[54]. Other groups have explored the use of insulating ferromagnets such
as barium ferrite [55], EuS [56, 57], Cr2Ge2Te6 [58], and thulium iron gar-
net (TIG). In particular, TI/EuS and TI/TIG heterostructures have recently
shown high temperature, proximity induced ferromagnetism at the interface
[59, 60]. However, a convincing signature of the dissipationless chiral modes
in such heterostructures remains lacking, and in this context, magnetic dop-
ing has been a far more successful route.
ACKNOWLEDGMENTS
This research is supported by grants from ONR (N00014-15-1-2370), NSF
(DMR-1306510 and the Pennsylvania State University Two-Dimensional
Crystal Consortium Materials Innovation Platform through NSF coopera-
tive agreement DMR-1539916) and ARO MURI (W911NF-12-1-0461).
REFERENCES
1. D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. Den Nijs, Quantized
Hall conductance in a two-dimensional periodic potential, Phys. Rev. Lett. 49 ,
405– 4 08 (1982).
2. O. A. Pankratov, Supersymmetric inhomogeneous semiconductor structures
and the nature of a parity anomaly in (2+1) electrodynamics, Phys. Lett. A 121 ,
360– 366 (1987).
Condensed-matter realization of the ‘ parity anomaly’ , Phys. Rev. Lett . 61 ,
2015– 2018 (1988).
4. C. L. Kane and E. J. Mele, Quantum spin Hall effect in graphene, Phys. Rev.
Lett. 95 , 226801 (2005).
5. C. L. Kane and E. J. Mele, Z2 topological order and the quantum spin Hall
effect, Phys. Rev. Lett . 95 , 146802 (2005).
6. B. A. Bernevig, S. Zhang, and C. Wu, Quantum spin Hall effect, Phys. Rev. Lett.
96 , 1– 4 (2006).
7. M. Kö nig, S. Wiedmann, C. Brü ne et al., Quantum spin Hall insulator state in
HgTe quantum wells, Science 318 , 766– 770 (2007).
8. L. Fu, C. L. Kane, and E. J. Mele, Topological insulators in three dimensions,
Phys. Rev. Lett . 98 , 106803 (2007).
tor class with a single Dirac cone on the surface, Nat. Phys. 5 , 398– 4 02 (2009).
10. M. Z. Hasan and C. L. Kane, Colloquium: Topological insulators, Rev. Mod.
Phys. 82 , 3045– 3067 (2010).
11. R. Jackiw and C. Rebbi, Solitons with fermion number ½ , Phys. Rev. D 13 ,
3398– 3409 (1976).
12. R. Yu, W. Zhang, H.-J. Zhang, S.-C. Zhang, X. Dai, and Z. Fang, Quantized
anomalous Hall effect in magnetic topological insulators, Science 329 , 61– 64
(2010).
13. C. Z. Chang, X. Feng, J. Zhang et al., Experimental observation of the quan-
tum anomalous Hall effect in a magnetic topological insulator, Science 340 ,
167– 170 (2013).
14. K. Yasuda, M. Mogi, R. Yoshimi et al., Quantized chiral edge conduction on
reconfigurable domain walls of a magnetic topological insulator, Science 358 ,
1311 (2017).
15. I. T. Rosen, E. J. Fox, X. Kou, L. Pan, K. L. Wang, and D. Goldhaber-Gordon,
Chiral transport along magnetic domain walls in the quantum anomalous
Hall effect, NPJ Quantum Mater. 2 , 69 (2017).
References 589
16. H. Zhang, C.-X. Liu, X.-L. Qi, X. Dai, Z. Fang, and S.-C. Zhang, Topological
insulators in Bi2 Se3 , Bi2 Te3 and Sb2 Te3 with a single Dirac cone on the surface,
Nat. Phys . 5 , 438– 4 42 (2009).
17. J. Zhang, C.-Z. Chang, Z. Zhang et al., Band structure engineering in (Bi1-xSbx)2
Te3 ternary topological insulators, Nat. Commun. 2 , 574 (2011).
18. D. Kong, Y. Chen, J. J. Cha et al., Ambipolar field effect in the ternary topo-
logical insulator (Bi xSb1-x)2Te3 by composition tuning, Nat. Nanotechnol. 6 ,
705– 709 (2011).
19. D. Zhang, A. Richardella, D. W. Rench et al., Interplay between ferromagne-
tism, surface states, and quantum corrections in a magnetically doped topo-
logical insulator, Phys. Rev. B 86 , 205127 (2012).
20. J. S. Lee, A. Richardella, D. W. Rench et al., Ferromagnetism and spin-depen-
dent transport in n-type Mn-doped bismuth telluride thin films, Phys. Rev. B
89 , 174425 (2014).
21. J. Zhang, C. Z. Chang, P. Tang et al., Topology-driven magnetic quantum
phase transition in topological insulators, Science 339 , 1582– 1586 (2013).
22. S. E. Harrison, L. J. Collins-McIntyre, S. Li et al., Study of Gd-doped Bi2 Te3
thin films: Molecular beam epitaxy growth and magnetic properties, J. Appl.
Phys. 115 (2014).
23. M. Liu, L. He, X. Kou et al., Crossover between weak antilocalization and weak
localization in a magnetically doped topological insulator, Phys. Rev. Lett. 108 ,
36805 (2012).
24. P. P. J. Haazen, J. B. Laloe, T. J. Nummy et al., Ferromagnetism in thin-film
Cr-doped topological insulator Bi2 Se3 , Appl. Phys. Lett. 100 , 82404 (2012).
25. L. J. Collins-McIntyre, S. E. Harrison, P. Schö nherr et al., Magnetic ordering in
Cr-doped Bi2 Se3 thin films, EPL , 107 , 57009 (2014).
26. Z. Zhou, Y.-J. Chien, and C. Uher, Thin film dilute ferromagnetic semiconduc-
tors Sb2−x Cr x Te3 with a Curie temperature up to 190 K, Phys. Rev. B 74 , 224418
(2006).
27. A. Kandala, Transport studies of mesoscopic and magnetic topological insula-
tors, Ph.D. thesis, The Pennsylvania State University, 2015.
28. Y. Ou, C. Liu, L. Zhang et al., Heavily Cr-doped (Bi,Sb)2Te3 as a ferromagnetic
insulator with electrically tunable conductivity, APL Mater. 4 , 086101 (2016).
29. X. Kou, S.-T. Guo, Y. Fan et al., Scale-invariant quantum anomalous Hall effect
in magnetic topological insulators beyond the two-dimensional limit, Phys.
Rev. Lett. 113 , 137201 (2014).
30. J. G. Checkelsky, R. Yoshimi, A. Tsukazaki et al., Trajectory of the anomalous
Hall effect towards the quantized state in a ferromagnetic topological insula-
tor, Nat. Phys. 10 , 731 (2014).
31. A. Kandala, A. Richardella, S. Kempinger, C. Liu, and N. Samarth, Giant
anisotropic magnetoresistance in a quantum anomalous Hall insulator, Nat.
Commun. 6 , 7434 (2015).
32. C.-Z. Chang, W. Zhao, D. Y. Kim et al., High-precision realization of robust
quantum anomalous Hall state in a hard ferromagnetic topological insulator,
Nat. Mater. 14 , 434. (2015).
33. S. Grauer, S. Schreyeck, M. Winnerlein, K. Brunner, C. Gould, and L. W.
Molenkamp, Coincidence of superparamagnetism and perfect quantization in
the quantum anomalous Hall state, Phys. Rev. B 92 , 201304 (2015).
34. M. Liu, W. Wang, A. R. Richardella et al., Large discrete jumps observed in the
transition between Chern states in a ferromagnetic topological insulator, Sci.
Adv. 2 , e1600167 (2016).
35. A. J. Bestwick, E. J. Fox, X. Kou, L. Pan, K. L. Wang, and D. Goldhaber-Gordon,
Precise quantization of the anomalous Hall effect near zero magnetic field,
Phys. Rev. Lett. 114 , 187201 (2015).
36. E. J. Fox, I. T. Rosen, Y. Yang et al., Part-per-million quantization and current-
induced breakdown of the quantum anomalous Hall effect, Phys. Rev. B 98 ,
075145 (2018).
37. C. Z. Chang, W. Zhao, J. Li et al., Observation of the quantum anomalous

Hall insulator to Anderson insulator quantum phase transition and its scaling
behavior, Phys. Rev. Lett. 117 , 126802 (2016).
38. A. Richardella, A. Kandala, J. S. Lee, and N. Samarth, Characterizing the struc-
ture of topological insulator thin films, APL Mater. 3 , 83303 (2015).
39. X. Kou, M. Lang, Y. Fan et al., Interplay between different magnetisms in
Cr-doped topological insulators, ACS Nano 7 , 9205– 9212 (2013).
40. Z. Zhang, X. Feng, M. Guo et al., Electrically tuned magnetic order and mag-
netoresistance in a topological insulator, Nat. Commun. 5 , 4915 (2014).
41. I. Lee, C. H. Kim, J. Lee et al., Imaging Dirac-mass disorder from magnetic
dopant atoms in the ferromagnetic topological insulator Crx (Bi0.1Sb0.9)2-xTe3,
Proc. Natl. Acad. Sci. 112 , 1316– 1321 (2015).
42. E. O. Lachman, A. F. Young, A. Richardella et al., Visualization of superpara-
magnetic dynamics in magnetic topological insulators, Sci. Adv. 1 , e1500740
(2015).
43. W. Wang, C.-Z. Chang, J. S. Moodera, and W. Wu, Visualizing ferromagnetic
domain behavior of magnetic topological insulator thin films, NPJ Quantum
Mater. 1 , 16023 (2016).
44. Y. Feng, X. Feng, Y. Ou et al., Observation of the zero Hall plateau in a quan-
tum anomalous Hall insulator, Phys. Rev. Lett. 115 , 126801 (2015).
45. X. Kou, L. Pan, J. Wang et al., Metal-to-insulator switching in quantum anom-
alous Hall states, Nat. Commun. 6 , 8474 (2015).
46. S. Grauer, K. M. Fijalkowski, S. Schreyeck et al., Scaling of the quantum anom-
alous Hall effect as an indicator of axion electrodynamics, Phys. Rev. Lett. 118 ,
246801 (2017).
47. M. Mogi, R. Yoshimi, A. Tsukazaki et al., Magnetic modulation doping in
topological insulators toward higher-temperature quantum anomalous Hall
effect, Appl. Phys. Lett. 107 , 182401 (2015).
48. M. Mogi, M. Kawamura, R. Yoshimi et al., A magnetic heterostructure of
topological insulators as a candidate for an axion insulator, Nat. Mater. 16 ,
516– 521 (2017).
49. D. Xiao, J. Jiang, J.-H. Shin et al., Realization of the axion insulator state in
quantum anomalous Hall sandwich heterostructures, Phys. Rev. Lett. 120 ,
056801 (2018).
50. M. Mogi, M. Kawamura, A. Tsukazaki et al., Tailoring tricolor structure of
magnetic topological insulator for robust axion insulator, Sci. Adv. 3 , eaao1669
(2017).
51. C.-Z. Chang, M. Kawamura, A. Tsukazaki et al., Simultaneous electrical-field-
effect modulation of both top and bottom dirac surface states of epitaxial thin
films of three-dimensional topological insulators, Nano Lett. 15 , 1090– 1094
(2015).
52. A. Kandala, A. Richardella, D. W. Rench, D. M. Zhang, T. C. Flanagan, and
N. Samarth, Growth and characterization of hybrid insulating ferromagnet-
topological insulator heterostructure devices, Appl. Phys. Lett. 103 , 202409
(2013).
53. J. S. Lee, A. Richardella, R. D. Fraleigh, C. Liu, W. Zhao, and N. Samarth,
Engineering the breaking of time-reversal symmetry in gate-tunable hybrid
ferromagnet/topological insulator heterostructures, Quantum Materials, 3,
51 (2018).
54. H. Wang, J. Kally, J. S. Lee et al., Surface state dominated spin-charge current
conversion in topological insulator/ferromagnetic insulator heterostructures,
Phys. Rev. Lett. 117 , 076601 (2016).
55. W. Yang, S. Yang, Q. Zhang et al., Proximity effect between a topological insu-
lator and a magnetic insulator with large perpendicular anisotropy, Appl. Phys.
Lett. 105 , 92411 (2014).
56. P. Wei, F. Katmis, B. A. Assaf et al., Exchange-coupling-induced symmetry
breaking in topological insulators, Phys. Rev. Lett. 110 , 186807 (2013).
References 591
57. Q. I. Yang, M. Dolev, L. Zhang et al., Emerging weak localization effects on

a topological insulator– insulating ferromagnet (Bi2 Se3 -EuS) interface, Phys.
Rev. B 88 , 81407 (2013).
58. H. Ji, R. A. Stokes, L. D. Alegria et al., A ferromagnetic insulating substrate for
the epitaxial growth of topological insulators, J. Appl. Phys. 114 , 114907 (2013).
59. F. Katmis, V. Lauter, F. S. Nogueira et al., A high-temperature ferromagnetic
topological insulating phase by proximity coupling, Nature 533 , 513– 516
(2016).
60. C. Tang, C.-Z. Chang, G. Zhao et al., Above 400-K robust perpendicular fer-
romagnetic phase in a topological insulator, Sci. Adv. 3 , e1700307 (2017).
Index
Ab initio theory, 49 Charge-transfer ferromagnetism (CTF) model, 436,

AFM, see Atomic force microscope (AFM) 441– 4 43, 446
AGFM, see Alternating gradient force magnetometer Chern number, 544, 548, 552, 559, 574– 575, 585
(AGFM) CIMS, see Current-induced magnetization switching
AHE, see Anomalous Hall effect (AHE) (CIMS)
Al, see Aluminum (Al) Circuit-theory, 233
Alternating gradient force magnetometer (AGFM), 423 CISA, see Current-induced spin accumulation (CISA)
Aluminum (Al), 71, 74– 75, 96, 109, 151, 156, 278, 432, CISP, see Current-induced spin polarization (CISP)
434, 586 CMOS, see Complementary metal-oxide-semiconductor
Angle-resolved photoemission spectroscopy (ARPES), (CMOS), 61
188, 383, 545 Cobaltite thin films, 459, 475– 477
Anomalous Hall effect (AHE) CoFe/MgO/AlGaAs/GaAs, TEM image, 106
recent developments, 349– 350 Complementary metal-oxide-semiconductor (CMOS),
scattering and deflection mechanism, 343– 3 49 61, 248
skew-tunneling phenomena, 202– 205 Conduction band (CB), 62– 63, 65, 67, 72, 75, 81– 82,
spin dependency, 341– 3 43 86– 87, 92– 93, 95, 101, 108– 109, 177, 179, 204,
Anti-parallel (AP), 115– 116, 134– 135, 158, 179, 205, 206, 218, 220, 224, 227– 229, 233, 374, 376, 379,
209– 210, 212, 214– 216 382, 385, 459, 461
AP, see Anti-parallel (AP) Conductivity mismatch, FM metal
ARPES, see Angle-resolved photoemission spectroscopy discrete layer as tunnel barrier, 96– 108
(ARPES) fundamental issue, 83– 85
Atomic force microscope (AFM), 463– 4 64, 495, 581, 583 Schottky barrier as tunnel barrier, 86– 96
Axion insulator state, 585 Coulomb-blockade magnetoresistive (CB-TAMR), 202
Crystal-field resonance, 375– 376
Ballistic Hot Electron Transport, 150, 163 CTF model, see Charge-transfer ferromagnetism (CTF)
Ballistic transport model
in germanium, 161– 164 Current-induced magnetization switching (CIMS), 68,
in silicon, 158– 161 176, 189, 211– 214, 218
BiFeO3 , band diagram, 461 Current-induced spin accumulation (CISA), 320
Bloch equations, 10, 26– 27, 29, 31– 32, 35, 37– 38, 40– 4 6, Current-induced spin polarization (CISP), 317– 323,
221, 231, 345– 3 46, 348 325– 327, 330– 331
BMP, see Bound magnetic polaron (BMP)
Boltzmann equation, 203, 319, 323– 325, 331, 347, 359 Dangling-bond hybrid (DBH), 375– 376, 383
Boltzmann transport theory, 346, 348, 526 DBH, see Dangling-bond hybrid (DBH)
Born– Markov approximation, 8, 12, 16 Decay law, 25
Bound magnetic polaron (BMP), 392, 394, 440, 441 Density functional theory (DFT), 92– 94, 262, 376, 378,
Brinkman, Dynes, and Rowell (BDR) model, 88, 102, 103 381, 388, 395, 416, 458, 478, 480, 495, 507
Bulk insulator, 546 Density of states (DOS), 25, 49, 63, 133, 157, 160, 181,
291– 292, 297– 298, 301, 379– 381, 383,
Canonical two-channel model, 83 385– 386, 389, 391, 396, 441– 4 42, 475, 521,
CB, see Conduction band (CB) 523, 544
CB-TAMR, see Coulomb-blockade magnetoresistive Dery– Sham model, 298
(CB-TAMR) DFT, see Density functional theory (DFT)
593
594 Index
Dietl theory, 188, 217, 221, 372, 385– 386, 389, 395, 534 Faraday geometry, 71, 76– 78, 90, 107, 109, 153
Diluted magnetic semiconductors (DMS), 67– 68, Fe/AlGaAs Schottky barrier
186– 187, 217, 372– 379, 434, 440; see also Dietl HRTEM images, 93
theory interface model, 95
Dilute oxides, general formula, 420– 423 low energy structure, 94
d 0 magnetism, 397 spin injection, 90– 96
Domain wall (DW), 511– 512, 528, 530, 575– 576, transport mechanism, 87– 88
583– 585 Z -contrast TEM images, 95
DOS, see Density of states (DOS) Fe/Al2 O3 /Si
Double exchange, 186, 386, 397, 419, 462, 464, 468– 471, EL spectra, 100, 104
505 spin-LED structures, 98
DMS, see Dilute magnetic semiconductor (DMS) TEM image, 103
DNP, see Dynamic nuclear polarization (DNP) Fe/GaAs heterostructures, 276– 278
Drift-diffusion model, 41, 45, 164, 271, 276, 281– 283, 289, Fe/MgO, spin-LED samples, 105
293–294, 301, 305 Ferroelectric-manganite heterostructures, 461
D’ yakonov– Perel’ mechanism interfaces with ferroelectrics, 465– 4 68
kinetic equation, spin, 19– 21 molecular orbital picture, 462– 4 65
persistent spin helix, 21– 24 Ferromagnetic and Kondo regimes, 505– 506
spin– orbit field, 18– 19 Ferromagnetic insulator, 578, 586– 588
Dynamic nuclear polarization (DNP), 300 Ferromagnetic metal, 63, 150, 270, 461, 504
DW, see Domain wall (DW) spin injection, 273– 276
Ferromagnetic resonance (FMR) spectra, 527, 529– 531
Edelstein (Rashba– Edelstein) effect, 320,331, 350 Ferromagnetic semiconductors (FMS), 66
Edelstein effect (EE), 178, 320 atomic constituents, 66– 67
EDMR, see Electrically detected magnetic resonance band symmetries, 81– 83
(EDMR) magnetic semiconductors, 66– 69, 80– 81
Electrical detection non-local detection, 110– 121
optically injected spins, 286 spin transmission, 81– 83
spin accumulation, 286 Ferromagnetism
three-terminal technique, 288– 290 CeO2 , 429– 430
Electrically detected magnetic resonance (EDMR), 154 In2 O3 , 430– 432
Electrical operation SnO2 , 427– 429
injection, 251– 256 thin film, magnetic measurements, 423– 426
tunneling, 256– 257 TiO2 , 426– 427
Electric field effect ZnO, 432– 437
carrier-induced ferromagnetism, 520– 522 Fisher– Langer theory, 393
Curie temperature, 522– 524 FM/GaAs heterostructures, single-contact measurement,
in ferromagnetic metals, 528– 534 275, 289, 302, 305
induced magnetization dynamics, switching, FM metal, see Ferromagnetic (FM) metal
531– 534 FMS, see Ferromagnetic semiconductors (FMS)
magnetic anisotropy, 528– 531
magnetic semiconductors, 520 GaAs
magnetization process and anisotropy, 524– 526 electronic and ferromagnetic properties, 186– 187
modulation of magnetic anisotropy, 528– 531 Mn exchange interaction, 185
other magnetic parameters, 526– 528 three terminal geometry, 121– 122
Electric field-induced magnetization tunneling transport properties, 188– 190
dynamics, 531– 532 (Ga,Mn)As
switching, 532– 534 CIMS curves, 213– 216
Electroluminescence (EL) spectra, 71, 258 spin transfer, 211– 212
field dependence, 79 TMR phenomena, 181– 185
selected values, 77 Ga1-x Mn x A s, spin-LED studies, 78– 79
Electron-beam lithography, 128, 131 Ge, see Germanium (Ge)
Electron-impurity scattering, 25– 26, 31, 34, 45, 319 Germanium (Ge); see also Ballistic transport
Electron spin resonance (ESR), 126, 132, 154, 159, 164, spin-orbit effects, 161– 164
259, 531 Ghost-band approach, 209– 210, 228– 229
Elliott– Yafet mechanism Giant orbital paramagnetism (GOP), 430, 441, 444– 4 46
electron-impurity scattering, 25 GOP, see Giant orbital paramagnetism (GOP)
electron– phonon scattering, 27– 29 Graphene
Elliott processes, 29 four-terminal non-local spin valve geometry, 127– 131
spin-flip momentum scattering, 17– 18 spin relaxation, 47– 51
spin relaxation, 30 spin transport in silicon nanowires, 131– 135
EL spectra, see Electroluminescence (EL) spectra three-terminal device geometry, 123– 127
ESR, see Electron spin resonance (ESR) three terminal geometry, 123– 127
Exchange bias effects, 459, 472– 475 as tunnel barriers, 122– 123
Index 595
Haldane model, 477, 544, 574– 575 relaxation time versus temperature, 47

Hall effect theory, 348, 359– 361 spin relaxation, 32– 41
Hamiltonian model, 209 transport direction, 41– 4 4
Hanle effect, 159 KMEE, see Kinetic magneto-electric effect (KMEE)
Kerr microscopy, 278– 283 Kramers partners theory, 551– 552, 555– 556
spin injection, 125
spin lifetime, 126 LaAlO3 , band diagram, 461
Hartree– Fock (HF) term, 35– 37 La 2 CuO4 , band diagram, 461
Heavy hole (HH), 6– 63, 69, 71, 73, 76– 77, 81, 87, 194, 196, Ladder diagram corrections to conductivity, 347
198, 200– 201, 206– 208, 210, 218, 225– 226, LaFeO3 , band diagram, 461
229, 382, 550– 551 Landau– Lifshitz theory, 544
Heisenberg model, 417– 418, 441– 4 42 La1-x 1Srx MnO3 , band diagram, 461
Helical edge states, 545– 546, 555– 557 LH, see Light holes (LH)
Heusler alloy-based heterostructures, 274, 298, 306– 307 Landau–Lifshitz– Gilbert (LLG) equation, 527
HF, see Hartree– Fock (HF) term Light-emitting diode (LED)
HH, see Heavy-hole (HH) doping profile, 87
Highest occupied molecular level (HOMO), 250– 251, EL spectra, 77
259– 260 with organic emitters (OLEDs), 248
HOMO, see Highest occupied molecular level (HOMO) schematic cross-section, 76
Hot electron generation and collection, 150– 151 semiconductor electrical spin injection, 69– 75
Hot electron transport spin polarization, 69– 75
ballistic spin-filtering effect, 152, 156– 158 Light hole (LH), 39, 62, 119, 156, 191, 196, 201, 210, 231,
band diagrams, 152 375, 382, 384, 390
non-SMS SVT-related designs, 153 LLG equation, see Landau–Lifshitz– Gilbert (LLG) equation
in silicon, 153– 155 LO, see Longitudinal-optical (LO)
tunnel junctions, 152, 156– 158 Longitudinal-optical (LO), 89
Lowest unoccupied molecular level (LUMO), 250– 251,
ILC, see Injection-limited current (ILC) 259– 260,
Inhomogeneous magnetic field, spin random walk, 14– 15 Lowin transformation, 357
Injection-limited current (ILC), 250 LUMO, see Lowest unoccupied molecular level (LUMO)
In-plane injection, spin transfer phenomena, 211– 212
Interfaces grown direction, 477– 480 Maekawa– Fukuyama theory, 498
Internal photoemission (IPE), 151, 152, 153 Majorana-bound states, 549
International Technology Roadmap for Semiconductors Majority and minority spin, 85, 92, 116, 133, 298
(ITRS), 61, 260 Magnetic anisotropy, electric field effect, 528– 531
Inverse spin Magnetic behavior, classification, 504– 505
switch behavior, 465 Magnetic circular dichroism (MCD), 378, 385, 427,
valve effect, 252 434– 436, 492, 495, 505, 507
Inverse spin-galvanic effect (ISGE), 320 Magnetic force microscopy (MFM), 428, 432, 489, 492,
anisotropy, 326– 330 495, 505
Inverse spin Hall effect (ISHE), 178, 305– 306, 557 Magnetic random access memories (MRAMs), 181
non-local measurement, 350 Magnetic semiconductors
IPE, see Internal photoemission (IPE), 151, 152 electrical spin injection, 65– 69
ISGE, see Inverse spin-galvanic effect (ISGE) ferromagnetic semiconductors (FMS), 66– 69
ISHE, see Inverse spin Hall effect (ISHE) spin light-emitting diode, 69– 75
ITRS, see International Technology Roadmap for Magnetic tunnel junctions (MTJs), 142– 143, 178– 179,
Semiconductors (ITRS) 181– 182, 193, 208– 209, 216– 217, 531– 534
Magnetic tunnel transistor (MTT), 102, 108– 110, 153
Jullière model, 181, 298 band diagram, 109
Magnetism
Kane–Mele model, 549 coexistence of magnetism, 501
Kerr microscopy, 278– 280 empirical tests, 510– 511
Hanle effect, 278– 283 magnetometry, 492
imaging spin injection and extraction, 283– 286 null results, inconsistencies and artifacts, 502
local Hanle effect studies, 280– 283 origin of electron pairing, 512
set up, 278– 280 oxygen vacancies, 511– 512
spin accumulation, 286– 288 theories, 502– 504
spin injection and extraction, 283– 286 III– V DMS with narrow to intermediate gaps,
Kinetic equation, 19– 21, 31, 201 393– 395
Kinetic-exchange model, 372, 381, 386– 388, 395– 396 transport, 496
Kinetic magneto-electric effect (KMEE), 320 Magneto-electric effect (MEE), 320
Kinetic-spin-Bloch-equation approach Magnetoresistance (MR), 127, 251
equation, 31 resistance area, 127
inelastic scattering, 46 Manganite superlattices and interfaces, 459, 468– 472
596 Index
Master equation, 16 multifunctional effects, 260– 261

MBE, see Molecular beam epitaxy (MBE) theoretical models, 261– 263
MCD, see Magnetic circular dichroism (MCD) OSCs, see Organic semiconductors (OSCs)
Mean-field theory, 178, 18, 186– 187, 224, 388– 389, 396, OSPDs, see Organic spintronic devices (OSPDs)
417– 418, 475 Out-of-plane injection, spin transfer phenomena,
Mean-free-path (MFP), 150– 152, 157 211– 212
MEE, see Magneto-electric effect (MEE) Oxygen plasma etching, 128, 212
Metal-insulator transition (MIT), 40, 97, 99, 112, 116,
118, 120– 121, 193, 271, 394, 468, 507, 526 Particle-hole symmetry, 553, 559, 561– 562
Metamagnetism, 508 p-d Zener model, 178, 186, 521– 522, 525– 526
anisotropic transport versus n, B, 508– 509 Perovskite transition-metal oxides, band diagram,
anomalous Hall effect, 509 458– 4 61
correlated transport properties, 509– 510 Persistent spin helix, 21– 24
Lifshitz transition, 508 Perturbation theory, 7, 32, 154, 162, 180, 185, 206,
MFM, see Magnetic force microscopy (MFM) 218– 219, 225, 231, 346, 349, 377, 386, 553,
MFP, see Mean-free-path (MFP) 555– 556, 558
MF theory, see Mean-field (MF) theory PHE, see Planar Hall effect (PHE)
Microscopic theory, 51, 318, 320– 321, 323, 325, 331, Photoluminescence (PL), 67, 73, 77– 78, 90– 91, 271, 275,
343– 3 48, 350– 351, 361, 383, 494 289, 385
Microscopic theory of metals PL, see Photoluminescence (PL)
intrinsic contribution, 345– 3 46 Planar Hall effect (PHE), 178, 526
side-jump contribution, 347– 3 48 Poole– Frenkel dependence, 250
skew-scattering contribution, 346– 3 47 Pseudogap, 499– 500
Molecular beam epitaxy (MBE), 66, 68– 69, 75– 76, 97, P -type semiconductors
105, 112, 156, 182, 186, 373, 377, 397, 429, 576 tunneling anisotropic magnetoresistance (TAMR),
Moore’ s Law, 61 192– 202
Motional narrowing, 11, 12, 17– 18 tunneling magnetoresistance (TMR), 180– 192
MR, see Magnetoresistance (MR)
MRAMs, see Magnetic random access memories QAHE, see Quantum anomalous Hall effect (QAHE)
(MRAMs) QD, see Quantum dot (QD)
MTJs, see Magnetic tunnel junctions (MTJs) QHE, see Quantum Hall effect (QHE)
MTT, see Magnetic Tunnel Transistor (MTT) QSH, see Quantum spin Hall (QSH)
Multiferroicity, 461 Quantum anomalous Hall effect (QAHE)
background and theory, 574– 576
Nb0.01 -SrTiO3 and Nb 0.05 -SrTiO3 , band diagram, 461 magnetism, 581– 586
NiFe/monolayer graphene/Si, zero bias resistance, 125 materials development, 576– 581
NLSV, see Non-local spin valve (NLSV) structure, 581– 586
Non-local detection temperature scale, 581– 586
bias dependence, 294– 298 Quantum dot (QD), 32, 177– 178, 372, 392, 500, 510
detector bias dependence, 303– 305 Quantum Hall effect (QHE), 349, 477– 478, 544– 545,
hyperfine effects, 299– 302 548, 574– 575, 579, 583, 585
Non-local spin valve (NLSV), 60, 112, 122, 127, 305– 307 Quantum mechanical description, 16
edge-to-edge separation, 130 Quantum spin Hall (QSH), 545, 553– 554, 556, 574– 575
four terminal geometry, 127– 131 Quantum well (QW), 31– 32, 34– 35, 37– 38, 42– 43, , 63,
resistance measurement, 129 70– 81, 91, 153, 179, 183– 184, 197, 218, 229,
schematic layout, 111 274– 275, 307, 318, 320– 322, 324, 329, 386,
spin current in silicon, 112– 116 391, 477, 524, 549– 552
Quasiparticle life time, 323– 324, 326, 329, 359, 381, 390,
OFETs, see Organic field effect transistors (OFETs) 548, 559
OMAR effect, see Organic magnetoresistance (OMAR) QW, see Quantum well (QW)
effect
Optical orientation, 61, 154 RA, see Resistance-Area (RA)
Optical pumping, spin polarization, 61– 63, 70, 73, 77, Random magnetic field
96– 97, 101, 271– 272, 274– 276, 282, 286, 300, reversible spin dephasing, 12– 13
304 Random walk
Organic field effect transistors (OFETs), 248, 251, 263 in inhomogeneous fields, 14– 15
Organic magnetoresistance (OMAR) effect, 263 motional narrowing, 12
Organic semiconductors (OSCs), 248 Rashba– Edelstein effect, 320, 331, 350
general background, 249– 251 Resistance-Area (RA), 120, 122, 127– 128, 155, 189–191, 275
Organic spintronic devices (OSPDs) Resonant tunneling diodes (RTD), 68– 69, 184, 196
injection, 251– 256 Reversible dephasing, 12– 13
tunneling, 256– 257 RKKY, see Ruderman– K ittel– Kasuya– Yosida
Organic spintronics (RKKY) model
alternative approaches, 257– 259 Rowell criteria
interface effects, 259– 260 transport mechanism, 88– 89
Index 597
RTD, see Resonant tunneling diodes (RTD) Spin-dependent Hall effects

Ruderman– K ittel– Kasuya– Yosida (RKKY) model, anomalous conductivity, 341– 3 43, 349– 350
186– 187, 386– 389, 393– 395, 439– 4 40, 462, microscopic theory of metals, 343– 3 49
503– 505, 582– 584 spin– orbit-coupled systems, 341
Spin dephasing, 159
Scaling theory, 61, 166, 344, 391, 394, 509, 510, 525, 585 Bir– A ronov– Pikus mechanism, 39
Scanning electron microscope (SEM), 128, 129, 512 kinetic-spin-Bloch-equation approach, 32– 41
Scanning transmission electron microscopy (STEM), 298 time and temperature measurement, 36– 37
Scanning tunneling microscope (STM), 62, 153, 273, Spin diffusion
374– 375, 388, 391, 529 approach, 129, 135
SCLC, see Space charge-limited current (SCLC) polarization and spin injection direction, 44– 45
Secular broadening, 11 transport properties, 41– 47
Self-consistent local random-phase approximation, Spin echo, 14
300– 301, 323, 389, 466, 471 Spin ensemble, random magnetic field, 12– 13
SEM, see Scanning electron microscope (SEM) Spin extraction, 95, 114, 116, 276, 283– 286
Semiclassical theory, 343– 3 46, 348– 350, 360 Spin ferromagnetism
Semiconductor electrical spin injection bound magnetic polaron (BMP) model, 440– 4 41
ferromagnetic metals, 80– 110 charge-transfer ferromagnetism (CTF) model, 441
initial transport studies, 69 DMS model, 440
light-emitting diode, 69– 75 spin-split impurity band (SIB) model, 441
magnetic properties, 65– 69 Spin Hall effect, 318, 341, 350– 351, 352, 477, 480, 574
spin-LED studies, 75– 79 spin orbit effect, 305
Semiconductor-metal-semiconductor (SMS), 67, 151– 152 Spin-Hall magnetoresistance (SMR), 177
Semimagnetic semiconductors, 67 Spin helicity, 554
SET, see Single-electron transistor (SET) Spin-injecting contact
SF, see Stacking faults (SF) band symmetries, 81– 83
Si, see Silicon (Si) conductivity mismatch, 83– 110
SIB model, see Spin-split impurity band (SIB) model, 441 electrical detection, 110
SIHE, see Spin-injection Hall effect (SIHE) magnetic semiconductors as, 75– 80
Silicon (Si) optical detection, 80– 110
graphene as a tunnel barrier, 122– 123 Spin injection
nanowires, spin transport, 131– 135 Al2 O3 /AlGaAs-GaAs, 96
non-local detection, 112– 116 Al2 O3 /Si, 96– 102
spin-polarized electrons, 153– 156 fe/AlGaAs Schottky barrier, 90– 96
spin-valve measurement, 158– 161 ferromagnetic metals into semiconductors, 273– 276
three terminal geometry, 117– 121 GaOx/AlGaAs-GaAs, 107– 108
Single-electron transistor (SET), 495, 500, 509– 510, 512 Hanle spin precession measurements, 125
6-band kp theory, 190– 192, 207– 209, 220 hot electron, 108– 110
transmission and chirality, 207– 208 MgO/AlGaAs-GaAs, 104– 107
tunneling spin transport, 190– 192 outlook and critical research issues, 135– 136
6-band Luttinger model, 208, 209, 220 SiO2 /Si, 102– 104
Skew-scattering contribution, 343, 344, 346– 3 47, three terminal geometry, 117– 122
359– 361 Spin-injection Hall effect (SIHE), 351– 352
Skew-tunneling combined spin– orbit and Zeeman effects,
6-band kp Kane model, 207– 211 361– 362
unidirectional character (U-SMR), 202– 205 experiment, 352– 356
VB from atomic SOI, 205– 207 Hall effect, 359– 361
in VB semiconductor, 205– 207 nonequilibrium polarization, 357– 359
Slater– Koster tight-binding theory, 382 prospectives of, 362– 364
Slonczewski spin-transfer torque model, 211 theory of, 356– 364
SOC, see Spin-orbit coupling (SOC) Spin-LED structures
SOI, see Spin-orbit interactions (SOI) band symmetries, 81– 83
SOT, see Spin-orbit torques (SOT) conductivity mismatch, 83– 102
Sm2 CuO4 (Nd 2 CuO4 ), band diagram, 461 Ga1-x Mn x A s, 78– 79
SMR, see Spin-Hall magnetoresistance (SMR) GaOx/AlGaAs-GaAs, 107– 108
SMS, see Semiconductor-metal-semiconductor (SMS) MgO/AlGaAs-GaAs, 104– 107
Space charge-limited current (SCLC), 250 SiO2 /Si, 102– 104
Spin accumulation SPI transmission, 81– 83
bias dependence, 294–298 Zn1-x Mn xSe, 75– 78
detector bias dependence, 303– 305 Spinodal decomposition, 440, 442
hyperfine effects, 299– 302 Spin-orbit coupling (SOC), 477
Kerr microscopy, 286– 288 Spin-orbit effects
non-local detection, 290 Datta– Das spin-FET, 305
optically injected spins, 286–288 germanium, 161– 164
three-terminal technique, 288– 290 Hall effect, 305
598 Index
Hanle measurement, 305– 306 TA, see Transverse acoustic (TA)

ISHE detection, 306 TAMR, see Tunneling anisotropic magnetoresistance
Spin-orbit field (TAMR)
electron-impurity scattering, 25– 26 TEM, see Transmission electron microscopy (TEM)
energy splitting, 18– 19 3D time-reversal invariant topological insulators
host lattice coupling, 26– 27 Majorana fermions, 560– 564
Spin-orbit torques (SOT), 177– 178, 211, 217, 233, 235, overview, 557– 559
520, 548 proximity effects, 564– 566
Spin polarization quantum computing, 559– 560
ballistic hot electron injection, 156– 158 3D topological insulators, 557– 559
current-induced model, 320– 325 III– V DMS
electrical injection, 63– 65 carrier-mediated magnetic interaction, 386– 390
hot electron transport, 151– 153 disorder and transport, 390– 393
inverse spin-galvanic effect, 326– 330 heavy Mn doping, 381– 386
light-emitting diode (LED) structure, 69– 75 magnetism, 393– 395
optical pumping, 62– 63 many-particle picture, 376– 377
optical transitions, 62– 63 narrow to intermediate gaps, 373– 374
solid state research, 317– 320 other isolated defects, 377– 379
Spin-polarized hot electron transport, 151– 153 single-electron picture, 374– 376
Spin relaxation weak Mn doping, 379– 381
conduction electrons, 16– 18 wide gaps, 395– 397
Elliott and Yafet processes, 30 III– V semiconductors, 271– 273
in graphene, 47– 51 Three-terminal experiment, 288– 290, 500
kinetic-spin-Bloch-equation approach, 32– 41 Three terminal geometry
out-of-plane to in-plane ratio, 49– 50 GaAs, 121– 122
in silicon, 29 graphene, 123– 127
single and bilayer graphene, 49 Si at room temperature, 117– 121
temperature dependence, 33 Tight-binding theory, 183, 186– 188, 196, 231, 382,
toy model, 5 385– 388, 394, 478
Spin scattering, interface defects, 79– 80 TI heterostructures, 586– 588
Spin-split impurity band (SIB) model, 441– 4 42 TJ, see Tunnel junction (TJ)
Spin transfer, 211 TMDC, see Two-dimensional transition-metal
with holes, 212 dichalcogenides (TMDC)
out of plane, 211 TMOs, see Transition-metal oxides (TMOs)
in plane current injection, 211 TMR, see Tunneling magneto-resistance (TMR)
Spin-transfer-torque (STT), 215, 235, 531– 534 TO, see Transverse optical (TO)
Spin transmission, band structures, 81– 83 Topological insulator (TI), 202, 477, 546, 552, 575
Spin transport Topological insulators in 2D
dynamics in III– V semiconductor, 271– 273 effective bulk Hamiltonian, 549– 552
FM/SC heterostructure, 276– 278 helical edge states in 2D TIs and protection, 552– 557
two channel model, 84 quantum spin hall systems, 549
Spin-up and down, 92, 191 Topological quantum computation
Spin-valve effect Majorana fermions, 560– 564
magnetic fields, 158– 159 3D time-reversal invariant topological insulators,
Si transport layer, 158 559– 560
time information, 160– 161 Toy model
Spin-valve transistor (SVT), 152– 153, 156, 158 motional narrowing, 12
SrRuO3 , band diagram, 461 quantum mechanical description, 16
SrTiO3 , superconductivity, 499 reversible dephasing, 12– 13
Stacking faults (SF), 79– 80 reversible spin dephasing, 13
STEM, see Scanning transmission electron microscopy spin dephasing, 5
(STEM) spin relaxation, 5
STM, see Scanning tunneling microscope (STM) spin relaxation of conduction electrons, 16– 18
Stoner models, 187, 380, 387, 438, 441– 4 43, 446, 503 TPHE, see Tunneling planar Hall effect (TPHE)
Structure, Magnetism and Temperature Scale, Transition-metal oxides (TMOs), 414, 419, 458– 459, 462,
581– 586 477– 479
STT, see Spin-transfer-torque (STT) Transmission electron microscopy (TEM), 79, 92, 94– 95,
Superconducting quantum interference device 97– 98, 102– 103, 105– 106, 278, 581, 583
magnetometer (SQUID), 90, 182, 213, Transverse acoustic (TA), 98, 100, 101, 104
423– 424, 492– 495, 501– 502, 505, 507, 512, Transverse optical (TO), 98, 100, 101, 104
583– 584, 586 Transverse times, 10
Super exchange theory, 417 Tunneling anisotropic magnetoresistance (TAMR), 180,
Su– Schrieffer– Heeger model, 262 189, 288
SVT, see Spin-valve transistor (SVT) AlAs/GaAs/AlAs RTDs, 196– 202
Index 599
classification, 193– 194 overview of TMR, 181– 185

with holes, 194– 196 phase diagram for CIMS, 213– 216
p -type semiconductors, 192– 202 TAMR holes, 194
Tunneling magnetoresistance (TMR), 179, 182– 183, 217, tunneling anisotropic magnetoresistance, 192
220, 256, 458, 531 tunneling magnetoresistance, 180
electronic and ferromagnetic properties, 186– 187 van der Waals epitaxy, 47, 249– 250, 576, 578
exchange interactions vs, 188 van Vleck ferromagnetism, 582
Mn in GaAs, 185 VBO, see Valence band offset (VBO)
phenomenological model, 180– 185 VB semiconductor
Tunneling planar Hall effect (TPHE), 178 asymmetrical transmission, 208– 211
Tunnel junction (TJ), 151– 153, 178– 179, 181– 183, skew-tunneling, 205– 207
195– 196, 209– 211, 532– 534 Vibrating-sample magnetometer (VSM), 423
2D topological insulators V-J model, 381, 395
Hamiltonian model, 549– 552
helical edge states, 552– 557 Wedge-type ferromagnetic semiconductor, schematic
Two-dimensional transition-metal dichalcogenides illustration, 182
(TMDC), 49 Weyl nodes, 547– 548
Type-II Weyl semimetals, 548
Yafet processes, 4, 29– 30, 160– 161; see also Elliott– Yafet
Ultraviolet photoemission spectroscopy (UPS), 249, 259 mechanism
UPS, see Ultraviolet photoemission spectroscopy (UPS) electron-phonon scattering, 29– 30
YBa2 Cu3 O y , band diagram, 461
Valence band (VB), 62– 63, 65, 67, 69– 70, 72, 74– 75, 87,
176– 177, 179– 180, 188, 191, 198, 204– 206, ZBR, see Zero bias resistance (ZBR)
208– 211, 217– 218, 220, 224, 229, 373, Zener kinetic-exchange theory, 372, 381, 386
375– 376, 378– 381, 383– 387, 393, 395– 396, Zero bias resistance (ZBR), 88– 89, 102– 104,
459, 461 124– 125, 132
Valence band offset (VBO), 190– 191, 220, 232 NiFe/monolayer graphene/Si, 125
Valence-band of p -type semiconductors Zero Hall plateau (C=0), 356, 585
experiments vs. 6-band kp theory, 190– 192 Z2 invariant, 552, 558
material trends, 185 Zn1-x Mn x Se, spin-LED studies, 75– 78

Tsymbal, Zutic - SpintronicsHandbook (2019)

Uploaded by

Copyright:

Available Formats

Tsymbal, Zutic - SpintronicsHandbook (2019)

Uploaded by

Document Information

Original Title

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

Tsymbal, Zutic - SpintronicsHandbook (2019)

Uploaded by

Copyright:

Available Formats

Spintronics Handbook:

Spin Transport and

© 2019 by Taylor & Francis Group, LLC

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-1-4987-6960-0 (Hardback)

Visit the Taylor & Francis web site at

and the CRC Press web site at

Section IV—Spin Transport and

Chapter 1. Spin Relaxation and Spin Dynamics in Semiconductors and

Chapter 6. Spin Transport in Ferromagnet/III–V Semiconductor

Section V—Magnetic Semiconductors,

Chapter 9. Magnetic Semiconductors: III–V Semiconductors. . . . . . . . . . . . . . . . 371

together with a more conventional spintronic memory, magnetic random

ferromagnetism expected from the Mermin-Wagner theorem. These two-

emerging in two-dimensional electron gases at oxide interfaces composed of

Evgeny Y. Tsymbal is a George Holmes University Distinguished Professor

Igor Ž utić received his Ph.D. in theoretical physics at the University of

Professor of Physics and got promoted to an Associate Professor in 2009

Ian Appelbaum Elbio Dagotto

Paul A. Crowell Valentin Dediu

Shuai Dong Ewelina M. Hankiewicz

Jaroslav Fabian Henri Jaffrès

Jeremy Levy John Schliemann

Yun-Yi Pai Anthony Tylan-Tyler

Chapter 4 Tunneling Magnetoresistance, Spin-Transfer

1.1 Toy Model: Electron Spin in a Fluctuating

1.1.2.3 Spin Random Walk in

Before we discuss the two main mechanisms, we introduce a toy model

1.1 TOY MODEL: ELECTRON SPIN IN A

w(t ) = 0, wa (t )wb (t ¢) = dab w2ae -|t -t ¢|/ tc . (1.1)

S y = ω 0 Sx − ω x (t )Sz + ω z (t )Sx , (1.3)

Sz = ω x (t )S y − ω y (t )Sx . (1.4)

The inverse relations are

S− = −iω 0 S− − iω z S− + iω − Sz , (1.10)

s+ = iω z s+ − iω + sz e − iω0t , (1.13)

s− = −iω z s− + iω − sz e iω0t , (1.14)

ω(t )ω(t ′)s(t ′) ≈ ω(t )ω(t ′) s(t ′) (1.20)

As the second assumption, we consider a “coarse-grained” time evolu-

Using the rules of Equation 1.1, Equation 1.22 simplifies to

Since we consider the times t ≫ τc, we can approximate

Finally, going back to S x and Sy

We can give the above equation a more conventional form by introducing

Then, from Equations 1.35 through 1.37, it follows that

to describe the generic spin relaxation.

1.1.1 Motional Narrowing

 eversible Dephasing, Spin Ensemble,

1.1.2.1 Reversible Spin Dephasing: Spin Ensemble

S± (t ) = ∑S (t) = S (0)e ∫ dω P(ω )e

Evaluating the Gaussian integral we get

* This discussion applies to nuclear spins as well.

1.1.2.3 Spin Random Walk in Inhomogeneous Magnetic Field

* The inhomogeneity could be due to the magnetic field or to the g-factor.

Here ω′ is the gradient of ω. As the electron diffuses, it precesses with the

We are interested in the averaged quantities over many realizations of the

According to the central limit theorem, ϕN is distributed normally with the

1.1.3 Quantum Mechanical Description

in which H0 is our system per se and V(t) is the time-dependent random

V (t ) = V (t ) = 0, V (t )V (t ′) ~ e −|t − t ′|/ τc . (1.63)

Chapter 4 Tunneling Magnetoresistance, Spin-Transfer

1.1 TOY MODEL: ELECTRON SPIN IN A

S y = ω 0 Sx − ω x (t )Sz + ω z (t )Sx , (1.3)

Sz = ω x (t )S y − ω y (t )Sx . (1.4)

S− = −iω 0 S− − iω z S− + iω − Sz , (1.10)

s+ = iω z s+ − iω + sz e − iω0t , (1.13)

s− = −iω z s− + iω − sz e iω0t , (1.14)

1.1.1 Motional Narrowing

eversible Dephasing, Spin Ensemble,

1.1.2.1 Reversible Spin Dephasing: Spin Ensemble

1.1.2.3 Spin Random Walk in Inhomogeneous Magnetic Field

1.1.3 Quantum Mechanical Description

pin Relaxation of Conduction Electrons

1.2 D’YAKONOV–PEREL’ MECHANISM

W k = 2 éëê(a BR + g D )k y ,-(a BR - g D )k x ùûú . (1.75)

1.2.2 Kinetic Equation for the Spin

The symbol denotes averaging over different momenta. Our goal is to

sa = WkaW kb tsb - W 2kt sa . (1.84)

ersistent Spin Helix

s = − 2iτ ( q ⋅ v k ) (Wk × s ) + τ Wk × (Wk × s ) − τ ( q ⋅ v k )2 s. (1.98)

sα = −2iτε αβγ qδ vkδ Ω kβ sγ + Ω kα Ω kβ τsβ − W2kτ sα − Dq 2 sα . (1.100)

1.3.1.1 Spin–Orbit Coupling by the Impurity

1.3.1.2 Spin–Orbit Coupling by the Host Lattice

1.3.2 Electron–Phonon Scattering

1.4 RESULTS BASED ON KINETIC-SPIN-

1.4.1 Kinetic Spin Bloch Equations

ρ k(r, t ) = ρ k(r, t ) dr + ρ k(r, t ) dif + ρ k(r, t ) coh + ρ k(r, t ) scat. (1.143)

1.4.3 Spin Diffusion/Transport