Materials Science and Engineering R 62 (2008) 135

Contents lists available at ScienceDirect

Materials Science and Engineering R

journal homepage: www.elsevier.com/locate/mser

Ferromagnetism and possible application in spintronics of

transition-metal-doped ZnO lms
F. Pan *, C. Song, X.J. Liu, Y.C. Yang, F. Zeng
Laboratory of Advanced Materials, Department of Materials Science and Engineering, Tsinghua University, Beijing 100084, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: This review article rst presents a summary of current understanding of the magnetic properties and
Diluted magnetic oxide intrinsic ferromagnetism of transition-metal (TM)-doped ZnO lms, which are typical diluted magnetic
ZnO oxides used in spintronics. The local structure and magnetic behavior of TM-doped ZnO are strongly
Local structure
sensitive to the preparation parameters. In the second part, we discuss in detail the effects of doping
Defects
elements and concentrations, oxygen partial pressure, substrate and its orientation and temperature,
Spintronics
deposition rate, post-annealing temperature and co-doping on the local structure and subsequent
ferromagnetic ordering of TM-doped ZnO. It is unambiguously demonstrated that room-temperature
ferromagnetism is strongly correlated with structural defects, and the carriers involved in carrier-
mediated exchange are by-products of defects created in ZnO. The third part focuses on recent progress in
TM-doped ZnO-based spintronics, such as magnetic tunnel junctions and spin eld-effect transistors,
which provide a route for spin injection from TM-doped ZnO to ZnO. Thus, TM-doped ZnO materials are
useful spin sources for spintronics.
2008 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Experimental procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Film growth of TM-doped ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. Characterization techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Local structure and ferromagnetism of TM-doped ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1. Effect of doping concentration on structure and magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.1. Effect of doping concentration on magnetic ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.2. Formation of a secondary phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2. Substrate effects on magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.1. Substrate-dependent local structure and magnetic behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.2. Substrate orientation-induced distinct magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3. Effects of substrate temperature and lm deposition rate on magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4. Inuence of oxygen partial pressure on magnetic ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.5. Strain-induced ferromagnetism enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.6. Inuence of post-annealing on magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.7. Effect of co-doping on magnetic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.8. Comparison of ZnO doped with Sc, Ti, V, Cr, Mn, Fe, Co, Ni and Cu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4. Origin of ferromagnetism in TM:ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1. Computational work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2. Experimental work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.1. Secondary phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.2. Carrier exchange interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

* Corresponding author. Tel.: +86 10 62772907; fax: +86 10 62771160.

E-mail address: panf@mail.tsinghua.edu.cn (F. Pan).

0927-796X/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.mser.2008.04.002
2 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

4.2.3. Defect-based BMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.3. Effect of bandgap on the Curie temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5. TM-doped ZnO-based prototype spintronics and multi-functional materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.1. Co-doped ZnO-based magnetic tunnel junctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.1.1. Spin-polarized transport in (Zn,Co)O/ZnO/(Zn,Co)O junctions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.1.2. Anomalous TMR in (Zn,Co)O-based junctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2. TM-doped ZnO-based spin eld-effect transistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3. Ferroelectricity and giant piezoelectric d33 in V and Cr-doped ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6. Conclusions and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

1. Introduction spintronics, but also owing to the excellent physical properties. So

A diluted magnetic semiconductor (DMS) is a compound with
properties intermediate between a non-magnetic semiconductor
and a magnetic element, and is obtained by doping a non-magnetic
semiconductor with transition-metal (TM) elements (e.g., Sc, Ti, V,
Cr, Mn, Fe, Co, Ni and Cu), including TM-doped IIIV (GaAs and
InAs) [14], IIVI (CdTe) [5] or group IV (Ge and Si) types [6].
However, most DMSs have a low Curie temperature (TC), which
limits their use in practical applications [7,8]. In 2000, Dietl et al.
[9] theoretically predicted that the TC of DMSs could be increased
above room temperature (RT) for p-type DMSs, and ferromagnet-
ism (FM) was stable in DMSs based on wide-bandgap semicon-
ductors, i.e., ZnO and GaN. Using rst-principle calculations, Sato
and Katayama-Yoshida [10] theoretically demonstrated that a ZnO
matrix doped with TM atoms such as V, Cr, Fe, Co and Ni exhibited
FM ordering, whereas doping with Ti and Cu resulted in a
paramagnetic state, which opened a window for experimental
attempts to prepare DMSs with room temperature ferromagnetism
(RTFM). FM semiconductors with O2 anions, such as TM-doped
ZnO, are also termed diluted magnetic oxides (DMOs).
Soon after that, consecutive experimental reports revealed that
TM-doped ZnO and TiO2 exhibited intrinsic RTFM [11,12]. These
observations prompted extensive experimental work and theore-
tical studies on RTFM DMOs, driven both by an urge to understand
the mechanisms involved and by the demand for better materials.
In fact, interest in TM-doped ZnO has increased because of
promising applications in the eld of semiconductor spintronics,
which seeks to extend the properties and applications of
established electronic devices by using the spin of electrons in
addition to their charge [7,1315]. Electron spin can potentially be
used to provide efcient injection of spin-polarized carriers for
spintronics and to fabricate transparent magnetic materials for
spintronics [1622]. There are two major criteria for selecting the
most promising materials for semiconductor spintronics. First, the
intrinsic FM should be retained at practical temperatures (i.e.
>300 K). Second, it is a major advantage if there is existing
technology for the material in other applications [14]. Fortunately,
TM-doped ZnO satises these two criteria. Following the
theoretical prediction of RTFM ascribed to a carrier-mediated
mechanism in ZnO-based DMOs [9], TM-doped ZnO has been
intensively investigated to achieve magnetic ordering above RT
and promising magneto-transport properties [11,14,1620]. As a
low-cost, wide-band gap (Eg = 3.37 eV) semiconductor, ZnO itself
has been the focus of renewed research for applications such as UV
light-emitters, transparent high-power electronics, surface acous-
tic wave devices, piezoelectric transducers and window materials
for display and solar cells [21,22].
Recently, a surge in research into TM:ZnO DMOs has been
observed, as shown in Fig. 1, indicating that this area is attracting
increasing interest not only because of promising applications in
far, RTFM has commonly been obtained in ZnO systems doped with
TM elements such as Sc [23,24], Ti [23,25], V [2628], Cr
[11,23,29,30], Mn [3140], Fe [4143], Co [8,23,4461], Ni
[62,63], and Cu [6468], and co-doping such as CoFe [69,70] and
MnCo [71,72], for which Co- and Mn-doped ZnO are the most
popular systems. Many groups have reported that TM-doped ZnO
lms exhibit FM ordering at a TC much lower than RT, such as 83 K
[73] and 110 K [74] for Mn-doped ZnO. Many previous publications
indicate that TM ions at Zn sites is a necessary but not sufcient
condition for FM in true TM:ZnO DMOs. Some ZnO lms doped
with Ti [75], V [27], Mn [42], Co [76,77] and Cu [78] only exhibited
paramagnetism or superparamagnetism as well as spin glass
behavior [79].
Although a considerable amount of experimental data and
corresponding mechanisms has been accumulated since the rst
report of RTFM [11,1620], the origin of FM ordering in DMOs,
including TM-doped ZnO, is still a matter of debate, i.e., whether
carrier-mediated exchange based on TM2+ replacement of Zn2+
[16,17] or the formation of secondary phases [18,69,80] is
involved. Thus, it is critical to distinguish true FM semiconductors
from those that merely show magnetic hysteresis. Even more
ambiguously, distinct results are frequently obtained in rather
similar TM-doped ZnO lms in the case of Co2+ substitution for
Zn2+ [11,81,82], indicating that the magnetization of TM-doped
ZnO is strongly dependent on the preparation parameters, and that
the preparation process is not easily reproduced from one growth
run to another.
In this rather contradictory context, several experiments and
calculations have been performed to elucidate the nature of RTFM
Fig. 1. Publications per year on TM:ZnO DMOs according to the Web of Science:
http://apps.isiknowledge.com/.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
[8,31,44,83]. Mechanism proposed include a bound magnetic
polaron (BMP) mechanism involving defects, including oxygen
vacancies (VO) and Zn interstitials (Zni) [8,44,57], hole doping
involving Zn vacancies [34,49,84], and the formation of FM
secondary phases (e.g., Mn2xZnxO3d via Zn diffusion into the
Mn oxide [31]) and TM-rich nanocrystals [83]. Moreover, ab initio
calculations predicted that competition between FM and anti-FM
coupling existed in DMOs [84]. However, there is ongoing debate
about these systems. In addition, a trace amount of an external
pollutant would also lead to a hysteresis loop during magnetiza-
tion measurements, further adding to the difculties in determin-
ing true DMOs [20,85,86]. So far, publications, including reviews
[1620], have commonly focused on a summary and comparison of
DMOs obtained by different groups and the different experimental
conditions used (e.g., preparation methods and parameters). The
resulting macroscopic magnetic properties, which are critically
dependent on growth and processing conditions, are diverse
without any common characterizations. It is therefore difcult to
identify a consistent mechanism to effectively understand the
origin of FM ordering merely by comparison of experimental
results achieved by different groups.
Half a decade has already passed since the rst report of RFTM
for TM-doped ZnO. Efforts are increasingly shifting not only to the
rational design of methods to evaluate this system, but also to the
development of TM-doped ZnO-based prototype spintronics
[14,85]. When successfully combined with semiconductor func-
tionalities, spintronics will have a considerable impact on future
electronic device applications [15]. Spin injection from a TM-doped
ZnO layer to pure ZnO and other conventional semiconductors is an
important long-term goal of the continuing research into ZnO
DMOs [87,88]. This concept has been realized in a representative
spintronics application, TM:ZnO-based magnetic tunnel junction
(MTJs), which exhibit superior spin-polarized transport, such as a
large half voltage (V1/2, at which the tunnel magnetoresistance
ratio becomes half of the maximum) [88,89], compared to
conventional magnetic metals and magnetic TM-oxide-based
MTJs with AlOx and MgO barriers [90]. Another typical application
is the spin eld effect transistor (spin-FET) proposed by Datta and
Das [91]. The core idea of this device is to induce spin precession by
the Rashba spinorbit interaction in a two-dimensional electron
gas and to use spin-dependent materials. Although a working
prototype of the DattaDas spin-FET has not yet been fabricated,
the relevant conditions necessary for the desired spin-FET
operation and simplied devices have been obtained [92,93].
In this review, to study the nature of FM ordering at high TC, our
main aim was to extract a single variable from a series of
experiments. An example is substrate variation [94], which is of
great interest owing to its potential to provide direct information
regarding the factors that affect FM ordering. Moreover, since ZnO
is a multi-functional material, recent signicant advances for this
system are described, including ferroelectric, piezoelectric, optical,
and electric properties, which play important roles in spintronics
applications of TM-doped ZnO [21]. We summarize recent progress
in TM-doped ZnO-based prototype spintronics, such as MTJs and
spin-FETs. It should be pointed out that it is almost impossible to
describe all the work performed around the world owing to the
rapid pace of studies in this eld, and therefore some references
may have been overlooked.
2. Experimental procedure
2.1. Film growth of TM-doped ZnO
TM-doped ZnO lms are commonly deposited by pulsed laser
deposition (PLD) [11,33,41,4649,95], magnetron co-sputtering,
3
including direct current (dc) reactive [44,45,94] and radio-
frequency (rf) [32,9698], combinatorial laser molecular-beam
epitaxy [50,51], chemical vapor deposition (metalorganic
[34,52,53,99,100], ultrasonic-assisted solution [57] and plasma-
enhanced [101]), ion-beam implantation [35,42], ion-beam
sputtering [54,55], and solgel methods [39,56]. Although the
magnetic behavior of TM-doped ZnO lms is sensitive to the
deposition conditions [1618], there is no conclusion on which
deposition method is best for FM ordering of the lms. For
example, Co-doped ZnO lms deposited by PLD [11,77] or
magnetron sputtering [74,94], two widely used deposition
methods, could alternatively exhibit RTFM and RT paramagnetism.
Therefore, magnetron sputtering is increasingly popular for
growing TM-doped ZnO lms owing to its low cost, high efciency,
and easy control, and its production of uniform lms of large size.
The substrate generally used is Al2O3(0 0 1) for low mismatch
between the lm and the substrate (2%) after a 308 in-plane
rotation [21], which is suitable for ZnO lms exhibiting high
crystallinity. Some other substrates suitable for the deposition of
TM-doped ZnO lms are Si [37,58,62,67,102], fused quartz [33],
glass [54,68,103,104], ZnO [51,105], ScAlMgO4 [106], LiNbO3 [44],
LiTaO3 [107], MgO [40], SiO2 [108] and NaCl [94]. Fig. 2 presents
typical cross-sectional high-resolution transmission electron
microscopy (HRTEM) images of Co-doped ZnO lms grown by
dc reactive magnetron co-sputtering on Al2O3(0 0 1), Si(1 1 1),
SiO2(1 0 1) and LiNbO3(1 0 4) substrates. In contrast to the
epitaxial growth of Co-doped ZnO lm on Al2O3 substrate
(Fig. 2a), the lms deposited on Si, SiO2 and LiNbO3 exhibit an
(0 0 2) preferred orientation with structural defects (e.g., edge
dislocations, marked by ?) in Fig. 2bd. Large- and small-angle
grain boundaries are observed in the lms on Si (Fig. 2b)/LiNbO3
(Fig. 2d) and SiO2 (Fig. 2c), respectively. Moreover, the (0 0 2)
planes are not parallel to the surface of the LiNbO3 substrate;
instead, the pillar-like grains tend to grow in two directions, as
highlighted by the arrows in Fig. 2d. These observations reveal that
the substrate affects the microstructure and the formation of
defects in TM-doped ZnO lm, which are expected to play an
important role in FM ordering in this system [94].
For PLD and rf magnetron sputtering techniques, lms were
deposited using a ceramic target prepared using standard ceramic
techniques from ZnO and TMO or TM2O3 powders of 4N purity [23].
For example, ZnO and MnO2 were mixed together, ground and
calcined for 8 h at 400 8C and sintered at 600900 8C for 12 h in air
to obtain Zn1xMnxO(0  x  0.1) ceramic pellets [33], whereas
ZnO and CoO/Co3O4 powders were used to prepare a
Zn1xCoxO(0  x  0.1) target. An alternative way to synthesize
TM-doped ZnO lms starts from pure metal targets of Zn and TM
[44,59,94] or alternate deposition of TM/ZnO [54,55]. The relative
sputtering area of TM chips attached to a Zn target was adjusted to
control the TM composition in TM-doped ZnO lms grown by dc
reactive magnetron co-sputtering [44]. Although it is hard to
precisely control the TM content in the lm to the same as that in
the target [59], the doping concentration in the lm can be
estimated using extrapolated data, i.e., 5 at.% Co in a ZnO target
corresponded to 6 at.% Co-doped ZnO in lms prepared by PLD
[25]. In general, the doping concentration ranges from 1 to 10 at.%,
producing typical DMOs exhibiting good crystal quality and strong
magnetization; lower doping amounts lead to weak magnetiza-
tion, whereas higher amounts tend to result in the appearance of
secondary phases, depending on the TM elements [50,56,69,80].
The typical deposition conditions for laser ablated lms are a
temperature of 400600 8C, oxygen partial pressure close to
1  105 to 5  105 torr and laser energy of 13 J cm2
[11,25,33,41,4651,95]. A relatively low (<600 8C) growth tem-
perature leads to homogeneous wurtzite TM-doped ZnO lms,
4 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 2. (a) HRTEM image of Zn0.96Co0.04O/Al2O3 in cross-section. The inset shows the corresponding SAED pattern. (b) Representative columnar crystal grain of 15 nm in
diameter along the ZnO(0 0 2) growth direction in Co:ZnO lm grown on Si substrate. (c) HRTEM image of the 1010 zone for Zn0.96Co0.04O grown on SiO2. (d) Cross-sectional
HRTEM image of Co:ZnO/128LN. The arrows show the growth direction along ZnO(0 0 2) planes. Typical edge dislocations are marked by white symbols (?).
whereas higher temperatures increase the possibility of an
inhomogeneous lm containing secondary phases, such as
hexagonal Co, CoO and Co3O4 phases [17,33,109,110]. For lms
grown by magnetron sputtering, the substrate temperature (200
400 8C) during deposition is typically lower than that for PLD
[44,68,111], and a high substrate temperature (650 8C) results in
the appearance of secondary phases [111]. The working gas is a
mixture of argon and oxygen at various pressures, which greatly
affects the local structure and the electric and magnetic properties
of ZnO lms [112114]. It was found that high oxygen partial
pressure (>3  103 torr) resulted in insulating Co-doped ZnO
lms (>105 V cm), which could provide a reference for elucidation
of the nature of RTFM [112,113].
2.2. Characterization techniques
Highly sensitive characterization techniques are critical for
understanding the local structure and the corresponding
magnetic behavior of TM-doped ZnO and to accurately detect
chemical information for doping elements at very low concen-
trations. X-ray diffraction (XRD) in the BraggBrentano, rocking
curve, and pole gure geometries is frequently applied to
characterize the lm structure and crystalline quality of TM-
doped ZnO, and to compare the position of diffraction peaks
between doped and undoped ZnO lms to predict the state and
site of doping elements [11,41,44,48,59,94]. Moreover, HRTEM,
eld-emission scanning electron microscopy (FE-SEM) and
selected-area electron diffraction (SAED) were used in most
DMO studies to identify interface bonding and structural
characteristics and to observe whether small nanoclusters were
present in the lms [36,44,48,94,110].
During the past 5 years of DMO research, many methods have
been introduced to identify the local structure of doping elements,
which is considered essential information for elucidating the
intrinsic FM and judging whether a system is a genuine DMO. (1)
Diffraction peaks due to secondary phases in a TM-doped ZnO
matrix can be observed from the XRD pattern. (2) Small
nanoclusters present in the lm can be observed in HRTEM
images. (3) Electron energy loss spectroscopy (EELS) combined
with HRTEM [33,108] and X-ray photoelectron spectroscopy (XPS)
[44,52] can be used to determine the valence state. (4) X-ray
absorption spectroscopy (XAS), including TM K-edge
[34,44,55,115], TM L-edge [97,116118] and O K-edge
[37,39,97,109], and its interpretation are used to elucidate the
local structure. This technique is currently attracting great interest
owing to its potential to provide local chemical information for
complex materials and its sensitivity to subtle differences in local
arrangement [119,120]. (5) Most importantly, DMOs should not
only have a strong magnetic circular dichroism (MCD) signal, but
should also exhibit an MCD spectral shape that reects the band
structure of the TM-doped ZnO, which is considered the most
useful tool to determine whether TM-doped ZnO is a real DMO or
not [37,47,52,60,85,121]. (6) Electron spin resonance (ESR)
[122,123] and Mossbauer spectroscopy [42,124] are carried out
to determine the local structure. (7) Measurement of optical
properties using Raman spectroscopy and optical transmittance
(absorption) spectrophotometry are effective in identifying the
local structure of the TM dopant [48,108,109,111]. If the state of a
trace amount of TM dopant cannot be determined using a single
characterization technique, the best way to investigate the local
TM structure may be to use several different methods.
Magnetization studies are carried out using a superconducting
quantum interference device (SQUID) magnetometer in the
temperature range 5350 K. High-temperature magnetization
(300 < T < 800 K) was recorded using a vibrating sample magnet-
ometer (VSM), which has also been used to measure TC
[11,44,59,125,126]. RTFM can also be examined using other
methods, such as alternating gradient magnetometry (AGM).
Furthermore, the average atomic magnetic moment is an inter-
esting parameter in evaluating and understanding DMOs. It is
therefore important to determine the accurate TM content in
samples. X-ray uorescence (XRF) [103], energy-dispersive X-ray
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
spectroscopy (EDS) and inductively coupled plasma (ICP) atomic
emission spectroscopy [44,127] have been used to determine the
TM content of samples after measuring the magnetic properties.
The average magnetic moment per Co atom was then calculated,
for which ICP provided more credible results for the TM content
[44,94,127]. To a certain extent, the characterization techniques
used are keys in studying DMOs. The rather contradictory situation
regarding DMO research means that suitable characterization
techniques are important. In contrast, incorrect measurements add
further confusion to the eld, which has already been dogged by
conicting papers published over the past several years. It is worth
pointing out that checking for basic experimental reproducibility,
in addition to correct selection of characterization methods, is
crucial in the search for high-TC DMOs and for reliable results and
discussion.
To elucidate the origin of RTFM and to identify relationships
between magnetization and the carriers in TM:ZnO, RT resistivity
and Hall effect measurements have been carried out using van der
Pauw geometry using indium metal for the contact electrodes and
current supplied by a dc-voltage source [109]. For some TM:ZnO
lms with quite high resistance, Hall effect measurements were
not suitable and the electric properties were determined using a TF
Analyser 2000 ferroelectric testing unit and impedance spectro-
scopy [44,108,128].
3. Local structure and ferromagnetism of TM-doped ZnO
Doped ZnO lms are either conducting or semiconducting, as
characterized by strong coupling between localized d electrons of
the TM ions and the extended s and p carriers of ZnO. Thus, the
carriers are spin-polarized and can mediate FM ordering of the
magnetic moments of TM ions doped into the oxide lattice, i.e., TM
ions replace Zn2+ sites. Furthermore, additional electron doping,
such as in (Co,Al)- [49,54,129,130], (Mn,Sn)- [131] and (Cu,Ga)-
doped ZnO [132], as well as hole doping, such as in (Co,Li)-
[133,134], (Co,N)- [135,136], (Mn,Cu)- [125], (Fe,Cu)- [126],
(Co,Cu)- [137] and (Mn,P)-co-doped ZnO [138], has been inves-
tigated to further examine carrier effects on RTFM. It was
importantly noted that additional carrier doping had no signicant
impact on RTFM enhancement [125,129,131,132], and some lms
even showed RTFM with an inverse correlation between magne-
tization and electron density [131,137,138]. Thus, inconsistent
phenomena and conclusions have been obtained, indicating that
the intrinsic FM of TM-doped ZnO systems remains an open
question. It is hence of vital importance to clarify the correlation, if
any, between carriers and the mechanism of FM inherent to this
class of DMOs.
On the other hand, defects are supposed to play a primary
role on FM ordering in TM-doped ZnO [8]. In fact, an increasing
number of studies have demonstrated that the formation of VO
and/or Zni [30,44,57,128,139] could lead to RTFM or FM
enhancement. For example, the production of VO by depositing
TM:ZnO lms in vacuum [44,111,117] or annealing lms in a
reducing atmosphere (such as vacuum [137,140,141] and Ar/H2
[55]) could greatly enhance RTFM. Annealing of Co-doped ZnO
lms in a Zn atmosphere also resulted in FM enhancement due
to Zni [57]. A similar phenomenon was observed by Schwartz
and Gamelin [142] for Co:ZnO nanocrystals. It is worth pointing
out that a large number of edge dislocations, which were
directly observed by HRTEM, enhanced the concentration of VO
and/or Zni point defects, and thus increased the RTFM of
TM-doped ZnO lms and nanowires [58,67,94,143]. Moreover,
charge-compensating structural defects can form polarons by
interaction between hydrogenic electrons and the defects,
leading to coupling of the exchange between polarons [8,139].
5
The mechanism related to BMP is discussed in detail in
Section 4.2.
Preparing TM-doped ZnO lms with low carrier concentrations
and measuring the RTFM of the samples obtained can provide
information on whether RTFM is a result of carrier-mediated
exchange and whether the TM-doped ZnO can actually be used in
spintronics applications. Thus it is worth studying TM-doped
oxides without carrier-mediated exchange (>105 V cm) to further
our understanding of the origin of RTFM. Insulating TM-doped
oxides with FM ordering have become a new topic of research,
namely diluted magnetic insulators (DMIs) [44,144]. On the other
hand, spin ordering in tunnel junctions made of barriers that are
DMIs such as EuS has interested researchers since Esaki et al. [145]
observed a large magnetic eld effect [146]. The topic has attracted
renewed interest because recent experiments revealed that the
transverse and longitudinal relaxation times for electron spins in
insulating quantum dots are in the nanosecond regime, indicating
potential as computing elements in quantum electronics
[147,148]. However, EuS failed in the practical sense, because its
TC is much lower than RT, with little hope of great improvement
[7]. RTFM with a TC of 350 K was rst observed for highly
insulating Co- and Cr-doped TiO2 [144,149,150] and TM-doped
SnO2 [41,151]. Interestingly, RTFM explained by the BMP
mechanism has been widely observed in TM-doped ZnO DMIs
[8,30,44,75,94,108,152,153]. For example, RTFM and a high TC of
790 K were observed in (4 at.%) Co-doped ZnO lms deposited on
1288 YX LiNbO3, which is not carrier-mediated, but co-exists with
the dielectric state [44]. This phenomenon can be also reproduced
in ZnO lms doped with different Co concentrations and grown on
various substrates. Measurement of RTFM in Zn1xFexO (x  0.04)
[43] and a lack of correlation between RTFM and conductivity
[154] support this observation. Moreover, Xu et al. [155,156]
observed that the saturated anomalous Hall resistivity increased
with decreasing electron concentration in Co-doped ZnO. Because
the spin splitting was proportional to the macroscopic magnetiza-
tion induced by the Co ions, the positive magnetoresistance
observed and the anomalous Hall effect in the insulating regime
suggested that FM could be realized in Co:ZnO with low electron
concentrations [157].
As indicated above, substitutional TM ions are a necessary but
not sufcient condition for FM ordering [27,42,47,79,158], and
large variations in carrier concentration and magnetic properties
have been reported for otherwise similar samples, indicating a
strong dependence of RTFM on the synthesis and processing
parameters [17,159]. Indeed, there is large variability in the
magnetic behavior of TM-doped ZnO lms, and the preparation
process is not easily reproduced from one growth run to another.
The intimate correlation between carrier concentrations and
point defects (i.e., VO and Zni) arising from a large number of
structural defects such as edge dislocations and stacking faults
leads to difculty in elucidating the relations among carriers,
structural defects and FM ordering. It is therefore important to
design a series of experiments to distinguish the variation of
carriers and structural defects and thus to identify the nature of
the RTFM. We discuss the inuence of parameters including the
doping elements and concentrations, oxygen partial pressure,
substrate and its orientation and temperature, deposition rate,
post-annealing temperature and co-doping on the local structure
and subsequent FM ordering of TM-doped ZnO in this section. This
is important for understanding the origin of RTFM in this system.
Such understanding can reveal the potential of a class of
promising magnetic materials with strong FM at the atomic
level, which is attractive for elds such as semiconductor
spintronics, quantum computation and magneto-optical devices
[160].
6 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

3.1. Effect of doping concentration on structure and magnetization
3.1.1. Effect of doping concentration on magnetic ordering
Incorporation of substitutional TM ions in IIVI ZnO should be
less difcult than incorporating them in other semiconductor
hosts, since the valence of Zn (2+) and the ionic radius can be
readily adopted by many 3d TM ions [19,50]. Different from TM-
doped TiO2 [20,144], TM-doped SnO2 [41,151] and similar systems,
this favors substitution of the TM ions at the cation site and helps in
achieving higher dopant concentrations. It also means that, in a
perfectly stoichiometric sample, substitutional TM introduces a
magnetic moment without contributing carriers. Thus, this system
can give direct insight into the question of whether the FM
observed is carrier-mediated or not [44]. Furthermore, TM atoms
have different thermal solubility limits in ZnO: in general, the
thermal solubility limits of Co and Mn are much higher than those
of Ti, V, Cr, Fe, Ni and Cu [50]. In contrast to the solubility of 15 at.%
for Co in ZnO [50,127], much lower limits (e.g., 5 at.%) have been
reported for V, Cr, Fe, Ni and Cu are detected [30,43,50], which is
consistent with Raman spectroscopic observations [161].
Table 1 presents a list of TM-doped ZnO DMOs, including the
substrate, TC, TM content and the corresponding magnetization
features at 300 K. As indicated above, the magnetization of TM-
doped ZnO exhibits no correlation with the preparation method,
which is therefore not listed in the table. There are many references
for Co- and Mn-doped ZnO owing to their strong magnetization
signals and the stability of Co and Mn ion congurations [8,23,44].
It is evident from Table 1 that Co-doped ZnO and Mn-doped ZnO
are the most popular systems in the literature [162166], not only
because of their high thermal solubility in ZnO lms, but also
because of their high magnetic moments at RT. The dopant
concentration has a denite inuence on the magnetic properties,
with more desirable behavior often obtained at comparatively
lower concentrations [23,25,37,62,66], with typical concentrations

Table 1
List of TM:ZnO DMOs

System TM content (at.%) Substrate TC (K) Magnetization features Reference

Co:ZnO 5 (525) Al2O3 (1 1 2 0) >280 0.01 mB/Co [11]

Co:ZnO 25 Al2O3(0 0 0 1) Spin glass [46]
Co:ZnO 3.5 (3.511.5) Glass >350 0.21 mB/Co [103]
Co:ZnO 5 Al2O3(0 0 0 1) 300 0.7 mB/mole Co [59]
Co:ZnO 25 Al2O3(0 0 0 1) >300 1.0 mB/Co [162]
Co:ZnO 35 Al2O3(0 1 2) >300 0.1 mB/Co [52]
Co:ZnO 7 (717) Si(0 0 1) >300 0.06 mB/Co [163]
Co:ZnO 10 Al2O3(0 0 0 1) >300 105 emu [48]
Co:ZnO 5 Al2O3 (1 1 0 2) >300 2.6 mB/Co [23]
Co:ZnO 25 Al2O3(0 0 0 1) 300 0.8 mB/Co [45]
Co:ZnO 1.3 (05) Al2O3(0 0 0 1) >300 0.32 mB/Co [53]
Co:ZnO 5 Si(1 0 0) >300 1.04 mB/Co [58]
Co:ZnO 0.30.5 Al2O3 >300 3 mB/Co [164]
Co:ZnO 2 Si(1 0 0) >300 0.4 mB/Co [154]
Co:ZnO 3 (125) Al2O3 (1 1 0 2) >300 5.9 mB/Co [165]
Co:ZnO 4 LiNbO3(1 0 4) 790 6.1 mB/Co [44]
Co:ZnO 5 ZnO/Al2O3(0 0 0 1) >300 1.5 mB/Co/very weak [51]
Co:ZnO 2 (213) Al2O3(0 0 0 1) >300 0.8 mB/Co [57]
Co:ZnO 12 (831) Al2O3(0 0 0 1) >300 0.04 mB/Co [111]
Co:ZnO 10 ZnO(0 0 0 1) AF [166]
Co:ZnO 5 Si(1 0 0) >300 0.5 mB/Co [134]
Mn:ZnO 36 Al2O3(0 0 0 1) Spin glass [79]
Mn:ZnO 2 (04) Glass 425 0.16 mB/Mn [33]
Mn:ZnO 7 Al2O3(1 1 2 0) Para [32]
Mn:ZnO 2 Al2O3(0 0 0 1) 980 0.04 mB/Mn [31]
Mn:ZnO 7 Al2O3(0 0 0 1) 78 20 emu/cm3 [98]
Mn:ZnO 1 (110) Al2O3(0 0 0 1) >300 0.6 mB/Mn [140]
Mn:ZnO 7 Al2O3(0 0 0 1) 83 1.17 mB/Mn [73]
Mn:ZnO 4 Si(1 0 0) >300 1.4 mB/Mn [34]
Mn:ZnO 2 Al2O3(0 0 0 1) >300 4.36 mB/Mn [95]
Mn:ZnO 3 (315) Si(1 0 0) >300 105 emu [37]
Mn:ZnO 10 Al2O3(0 1 2) >300 0.075 mB/Mn [125]
Mn:ZnO 2.6 Al2O3(0 0 0 1) >350 4.8 mB/Mn [40]
Mn:ZnO 2 (06) Si(0 0 1) >300 0.3 mB/Mn [101]
Sc:ZnO 5 Al2O3 (1 1 0 2) >300 0.3 mB/Sc [23]
Ti:ZnO 5 Al2O3 (1 1 0 2) >300 0.5 mB/Ti [23]
Ti:ZnO 9.9 Al2O3(0 0 0 1) Para [75]
V:ZnO 15 (515) Al2O3 (1 1 2 0) >350 0.4 mB/V [26]
V:ZnO 0.120 Al2O3(0 0 0 1) Para [27]
V:ZnO 13 Al2O3 (1 1 0 2) Spin glass [28]
Cr:ZnO 5 Al2O3 (1 1 0 2) 400 0.6 mB/Cr [29]
Cr:ZnO 1.1 395 0.73 mB/Cr [30]
Fe:ZnO 4 (07) LiNbO3(1 0 4) 400 0.15 mB/Fe [43]
Ni:ZnO 1 (125) Al2O3(0 0 0 1) Para [63]
Ni:ZnO 1 (17) Si(1 0 0) >300 0.37 mB/Ni [62]
Cu:ZnO 1 (117) Al2O3(0 0 0 1) 390 0.4 mB/Cu [64]
Cu:ZnO 1 (18.3) Glass 390 1.6 mB/Cu [65]
Cu:ZnO 2 (212) Glass 350 1.8 mB/Cu [66]
Cu:ZnO 5 Si >300 0.037 mB/Cu [67]
CoFe:ZnO <15 SiO2/Si >300 15 emu/cm3 [70]
MnCo:ZnO 10 Al2O3(0 0 0 1) >300 0.11 mB [72]

Magnetization features were recorded at 300 K for the TM content listed in column 2. Para, paramagnetism; AF, anti-ferromagnetism.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135 7

Note that Co dopants at a concentration of 35 at.% can lead to

giant magnetic moments [23,44,94,165]. If carrier-mediated
exchange were the origin of RTFM, there would be no correlation
between the dopant concentration and the average magnetic
moment, because carriers are not local; instead, carriers can play a
similar role in mediating RTFM in TM-doped ZnO with different
dopant concentrations. On the contrary, if RTFM stems from BMP
on the basis of defects, the polaron percolaton threshold requires a
minimum mismatch between the dopant content and the defect
content. In this case, a certain dopant concentration is a
precondition for the appearance of a giant magnetic moment.
Suitable dopant concentrations and defects effectively couple to
form a hydrogenic orbital arising from an electron associated with
a particular defect, and the hydrogenic orbital tends to sufciently
spread out to overlap with a large number of BMPs, so that few
isolated polarons exist. More remote dopants interact too weakly
with both polarons to be of any importance. This occurs in samples
with lower doping levels, e.g., 1 at.% Co-doped ZnO, which
Fig. 3. Reported magnetic moments measured in Co-doped ZnO for different dopant produces a much lower magnetic moment for weak interaction.
concentrations. The lines are guides for the eyes.
On the other hand, for lms doped with higher levels of Co, an
increase in dopant concentration (e.g., 15 at.% Co-doped ZnO) leads
to a rapid decrease in the moment due to enhanced dopantdopant
of 5 at.% for Co-doped ZnO [11,23,51,58,59,61,128,133,134,152,
association and thus progressive orbital moment quenching
167170], and 1 at.% for Ni-doped ZnO and Cu-doped ZnO
[44,172]. In this case, anti-FM or FM appears when there are
[62,64,65].
continuous paths throughout the crystal joining nearest-neighbor
Most groups have focused on doping concentrations below the
magnetic cations [8]. This reveals how RTFM can be explained by a
solubility limit to obtain true DMOs. Although distinct magnetic
BMP model based on defects.
behaviors have been obtained by different groups owing to the
It was reported that Zn0.97Mn0.02Al0.01O lms showed RTFM
effects of the preparation method and parameters, it is worth
with 4.36 mB/Mn, for which the 2 at.% Mn doping to yield a giant
evaluating the effect of dopant concentrations on magnetization
magnetic moment is lower than the level of Co dopant required
from a series of experiments. Magnetic moments measurements
(35 at.%). Interestingly, in contrast to the Co doping concen-
reported for Co-doped ZnO at different dopant concentrations are
tration range (125 at.%) reported in many publications
shown in Fig. 3. It is interesting that Co doping in the concentration
[11,45,46,162,165,176], Mn is generally used as a dopant at a
range 35 at.% shows much stronger RTFM than other 3d TM
much lower concentration of 110 at.% [32,140], as illustrated in
dopants at the same concentrations, as well as the large solubility
Table 1. Unlike those of Co- and Mn-doped ZnO, the average
limit of Co in ZnO, which explains the many reports on 5 at.%
magnetic moment of other TM-ZnO system, such as Cr-, Ni- and
Co:ZnO in the literature [11,23,5860,95,128,133,134,152]
Cu-doped ZnO, monotonically decreases with increasing dopant
(Table 1).
concentration from 1 at.% [30,54,62,6466,71,102,177], as
Fig. 3 illustrates the Co composition dependent magnetic
shown in Fig. 4. This phenomenon may be due to the low
moment of Co:ZnO. From Fig. 3 and Table 1, one can see that
solubility of these dopants in ZnO; thus, low concentrations
TM:ZnO materials exhibit giant magnetic moments, such as
(e.g., 1 at.%) correspond to the strongest magnetization and
5.9 mB/Co and 2.6 mB/Co in semiconducting Zn0.97Co0.03O and
higher concentrations correspond to lower magnetic moments.
Zn0.95Co0.05O lms, respectively [23,165]. These are much stronger
Therefore, 35 at.% Co, 2 at.% Mn and 1 at.% Cr, Ni, and Cu doping
than for the corresponding atom in bulk metal, such as 1.7 mB/Co in
levels in ZnO show the most robust magnetic moments, which
Co metal. Similarly, 4.36 mB/Mn [95], 4.8 mB/Mn [40] and 5 mB/Mn
[171] were obtained for semiconducting Mn-doped ZnO lms.
Several studies on DMOs such as 5 at.% Co- and Cr-doped SnO2
[172,173] and 5 at.% V-doped TiO2 [174] revealed very large
magnetic moment of 7.2 mB/Co, 6.0 mB/Cr and 4.2 mB/V, respec-
tively. On the other hand, 6.1 and 3.2 mB/Co were reported for
insulating Zn0.96Co0.04O/LiNbO3 [44], which was attributed to
magnetoelectric coupling between the Co:ZnO lm and the
ferroelectric substrate and a subsequent super-coupling mechan-
ism in terms of BMPs [94], indicating that carrier-mediated
exchange is not a necessary condition for giant magnetic moments.
Using the Lande g-factor and the total angular momentum
quantum number (J) of Co2+ to evaluate the effective magnetic
moment [175] according to

p
meff g mB JJ 1 (1)

leads to a value of 6.6 mB/Co. In contrast, the net spin moment

calculated from the spin-only formula mmax = gmBS is 3 mB. This
phenomenon can be explained by spin and orbital angular moment Fig. 4. Reported magnetic moments measured in Cu-, Ni- and Cr-doped ZnO for
contributions [44,172174]. different dopant concentrations. The lines are guides for the eyes.
8 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

are critically dependent on the different solubility of these TM

dopants in ZnO lms.

3.1.2. Formation of a secondary phase

It is considered necessary to achieve single-phase DMOs for use
in devices, in which the carriers can realize spin polarization. If
precipitates or clusters below detectable size limits are present and
responsible for the FM, and if the carriers do not mediate the FM
interaction, this then can limit the usefulness of spintronics
fabricated from such materials [18]. Therefore, it is important to
detect RTFM arising from TM2+ replacement of Zn2+ sites or
secondary phases, which is signicant for studying practical DMOs.
However, the detection of trace amounts of TM atoms requires
highly sensitive characterization techniques.
The formation of secondary phases mainly depends on the TM
solubility limit in ZnO. The thermal solubility limit for Co in ZnO is
less than 15 at.% and Co ions systematically replace Zn ions in the
lm without changing the wurtzite structure [50,127]. This is
supported by the observations of Prellier et al. [59], who reported
that the c-lattice parameter obeys Vegards law up to the solubility
limit of 10 at.% Co, after which a turning point suggests the
breakdown of Vegards law. However, different values of the
dopant solubility have been obtained; for example, limits in the
range 2550 at.% have been reported for Co-doped ZnO lms
grown by PLD. Kim et al. [178] claimed that the limit is 40 at.%,
close to the results reported by Ueda et al. [11]. This difference is
most likely attributable to the preparation process, which allows a
different distribution of Co within the ZnO wurtzite structure
[16,17,19]. The equilibrium solubility limit of Mn in ZnO was
determined by Jin et al. [50] to be 20 at.%. The high solubility of
dopants seems to result from the growth conditions. It was also Fig. 5. (a) XRD pattern of a Zn0.96Fe0.04O lm. The inset shows XPS spectra of
Zn0.96Fe0.04O (dashed line) and Zn0.93Fe0.07O (solid line). (b) XRD pattern of a
found that the formation energy of dopants can have a strong
Zn0.93Fe0.07O lm. The Fe3O4 secondary phase is marked by solid circles.
dependence on the local lattice constant, depending on the size
mismatch of the dopant and the host atom. For example, several
authors claimed that the non-equilibrium nature of the laser CoO peaks has been reported for doping concentrations above
material interaction allows doping at high Mn content (up to 25 and 22 at.% using solgel methods and reactive magnetron
36 at.%) [50,79]. Meanwhile, FM ordering has been reported for Mn co-sputtering, respectively [56,163]. The formation of Co metal
concentrations between 2.2 and 30 at.% [33,179,180], whereas nanoclusters for x  12 at.% in samples grown by solgel and rf
both the magnetic moment per Mn ion and the FM TC seem to sputtering techniques has also been observed [76]. The
decrease with increasing Mn concentration, i.e., the moment observation of a CoFe(1 1 0) diffraction peak for Zn1x(CoFe)xO
dropped from 0.71 to 0.25 emu cm3 when the Mn content O lms indicates the formation of CoFe metal particles, which
increased from 10 to 30 at.% [76]. In addition, the FM TC was above was further conrmed by a clear change in the slope of the
RT for all concentrations x  5 at.% [33,180], but was usually well d(0 0 2) variation with the CoFe content around x = 0.2 [70].
below 100 K (45 K) for higher concentrations [179]. Likewise, only a Ni(1 1 1) peak was detectable, revealing that
Besides RTFM arising from TM ion replacement of Zn2+, fcc Ni nanocrystals are highly oriented with respect to the ZnO
secondary phases such as TM metal and TM-based oxides are host matrix [181]. However, secondary phases in trace
frequently found to correspond to the magnetization measured. amounts and small clusters are hardly detected by XRD, so it
Several different methods have been used to determine the local should be remembered that the absence of evidence is not
structure of doping elements and to detect secondary phases, evidence of absence [139].
which is important for identifying whether a TM-doped ZnO is a (2) HRTEM and SEM are often performed to investigate the
genuine DMO or not. The methods for detecting secondary phases different phases that might have formed in the nanosize range
are summarized as follows. and to determine the state of Co atoms that cannot be detected
by XRD. For example, Co3O4 clusters were directly found by
(1) In the most simple approach, diffraction peaks of secondary both HRTEM [111] and FE-SEM [109] in Co-doped ZnO lms.
phases present in the TM-doped ZnO matrix can be observed in ZnCo2O4 and/or CoO clusters with a diameter of 5 nm have
the XRD pattern. In contrast to Zn1xFexO (x  0.04) samples, been observed by HRTEM [110], which was conrmed by SAD
which show a single phase with a wurtzite structure (Fig. 5a), results. Co metal clusters with an average diameter of 10 nm
as x increases to 0.07, a diffraction peak of a secondary phase have also been observed by HRTEM in 15 at.% Co-doped ZnO
appears in the XRD pattern, which is attributed to Fe3O4, as lms [76].
shown for Zn0.93Fe0.07O in Fig. 5b [43]. This observation is in (3) The valence state of a TM dopant can be determined by EELS
good agreement with the low solubility of Fe in ZnO, e.g., and XPS [33,43,44,52,108,182]. The advantage of EELS is that
<5 at.% [19,50]. In contrast, no secondary phases are found in the valence state can be measured in combination with
Co-doped ZnO, even for 10 at.%, for the same preparation observation of HRTEM images [33,108]. Wei et al. [43] observed
parameters and substrates [44], indicating the different that both the Fe 2p1/2 and 2p2/3 peaks of Zn0.93Fe0.07O were
solubility of Co and Fe in ZnO. The presence of Co3O4 and between those of Fe3+ and Fe2+, and closer to Fe3+, indicating a
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 6. Experimental (solid circles) and MS calculated (solid line) Fe K-edge XANES
spectra of (a) Zn0.96Fe0.04O and (b) Zn0.93Fe0.07O lms.
mix of 2+ and 3+ valence states with predominantly Fe3+, as
shown in the inset of Fig. 5a. In contrast, for a doping
concentration of x  0.04, such as in Zn0.96Fe0.04O, the positions
of the Fe 2p1/2 and 2p2/3 peaks are apparently different from
those of Zn0.93Fe0.07O (inset of Fig. 5a), revealing that Fe is in the
2+ valence state [43].
(4) TM K-edge [34,44,55,115], L-edge [78,97,116,117] and O K-
edge [37,97,109] XAS and its interpretation can provide a
structural ngerprint to determine TM sites in ZnO [8]. As
discussed above, for XRD and XPS changes for different doping
concentrations, Fe K-edge XANES can be used to determine the
local structure of Fe in the lms and conrm the preliminary
conclusion. Experimental (solid squares) and full multiple-
scattering (MS) ab initio calculated (solid line) XANES for
Zn0.96Fe0.04O and Zn0.93Fe0.07O are shown in Fig. 6a and b,
respectively [43]. The most interesting result here is that the
calculated spectrum has three main features that accurately
reect those of the experimental spectrum, denoted by A (A0 ), B
(B0 ) and C (C0 ) in order of increasing photon energy in Fig. 6a
and b. It should be pointed out that the calculations for Fig. 6a
and b were produced for a cluster within a sphere radius of 6 A
containing 77 atoms from the central Fe atom that replaces a
central Zn in the ZnO atomic arrangement and 89 atoms from
the Fe atom that occupies a tetrahedral site in the cubic
structure of Fe3O4. There are two main differences between
them: (i) the curve for Zn0.96Fe0.04O has an inection at
approximately EE0 = 5 eV (E0 = 7112 eV), whereas the curve
for Zn0.93Fe0.07O is at at that position (A and A0 ); (ii) peak C is
9
slightly more intense than peak B for Zn0.96Fe0.04O, whereas
peak B0 is apparently less intense than C0 for Zn0.93Fe0.07O.
Moreover, the curve for Zn0.93Fe0.07O has all the features of
Fe3O4 [183]. It is hence concluded that when x  0.04, Fe is in
the 2+ valence state and substitutes for Zn in the wurtzite
lattice of ZnO. When x increases beyond 0.04, Fe mainly exists
in the form of Fe3O4. Moreover, Fe2+ substituting for Zn2+ in
Zn0.96Fe0.04O and the Fe3O4 secondary phase in Zn0.93Fe0.07O
are responsible for 0.15 and 1.0 mB/Fe, respectively.
(5) Most importantly, MCD (including X-ray MCD and optical
MCD) is considered the best method for evaluating whether a
TM-doped ZnO lm is a real DMO or not [85,139]. X-ray MCD
(XMCD) is commonly used, and is the difference in absorption
of left- and right-hand-circularly polarized X-rays by a
magnetized sample. Thanks to simple sum rules, XMCD can
also provide quantitative information about the distribution of
spin and orbital angular momenta. Optical MCD measures the
relative difference in the circular polarization-dependent
optical absorption under an applied magnetic eld, i.e.,
2(K  K+)/(K + K+), where K and K+ are optical absorption
coefcient for s and s+ polarizations, respectively. The optical
absorption depends on the circular polarization ascribed to
Zeeman splitting of the band structures. MCD spectroscopy
sensitively detects DMOs and reects their electronic structure
and identies secondary phases, taking the strong spd
exchange interaction of DMOs and subsequent Zeeman
splitting of the optical transitions into account [85,158]. The
part of the MCD signal that linearly increases with the magnetic
eld (H) represents the paramagnetic component, whereas the
MCD signal that persists at H = 0 Oe is the FM component. MCD
spectra showing a multiplet structure, unlike spectra of TM
metals with a single peak, reveal that TM-doped ZnO shows FM
ordering that is not due to metallic TM secondary phases, but
arises from TM ion substitution in Zn sites [60,121]. For
example, Mn-, Co-, Ni- and Cu-doped ZnO lms prepared by
PLD show pronounced negative optical MCD peaks near 3.4 eV
attributed to the bandgap of the host ZnO semiconductor,
indicating that ZnO alloyed with Mn, Co, Ni, and Cu is a true
DMO with strong exchange interaction between sp-band
carriers and localized d electrons [184]. In contrast, the
magnetic moment was observed to be 0.1 mB/Co in Co-doped
TiO2, but the XMCD spectral line shape is nearly identical to
that of Co metal, showing that FM is induced by a small amount
of clustered Co, consistent with the Co metal cluster observed
by FE-SEM [185]. Likewise, the use of advanced element and
orbital selective techniques allows researchers to observe the
distinct spectral signature of Co in ionic or metallic states to
assign FM phases to the presence of Co metal clusters in lms
doped with Co concentrations >25 at.% [176].
(6) ESR [99,122,123] and Mossbauer spectra [42,124] also provide
information on the local structure of the TM dopant [131]. The
ESR spectrum of Zn1xMnxO (x = 0.001) exhibits the hyperne
and ne-structure lines of Mn2+, which broaden for higher Mn
content (x = 0.035) and change to a very broad single line due to
dipolar interactions between Mn ions at higher local concen-
trations [122]. This phenomenon indicates that ESR can reect
local structure variations for a TM dopant. However, a large
number of TM neighbors in the ZnO lattice (n  5) have no
effect on the line width observed [123]. Mossbauer spectra
cannot only reveal FM (sextet) and paramagnetic (doublet)
phases in Fe:ZnO, but are also useful in detecting secondary
phases, since the Mossbauer spectrum of Fe:ZnO is different
from the typical sextet of FeOx [124].
(7) Optical properties (e.g., Raman spectroscopy and optical
transmittance/absorption measurements) are commonly used
10 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

to characterize TM-doped ZnO because they are intimately

related to the defects and electronic structure of TM:ZnO
[48,53,108,109,111,161]. For example, spectra show character-
istic absorption edges around 567, 613 and 662 nm resulting
from the spin-orbit split ligand-eld transition of tetrahedrally
coordinated Co2+ ions due to 4A2(F) ! 2A1(G), 4A2(F) ! 4T1(P)
and 4A2(F) ! 2E(G), respectively. Once Co metallic clusters exist
in a lm, the spectrum exhibits an octahedral Co0
4
T1g(F) ! 4T1g(P) at approximately 513 nm [48,109]. Thakur
et al. [53] reported that for x  5 at.% in Zn1xCoxO, at which the
magnetic moment becomes negligible, broadening of the Elow 2
Raman mode was also negligible, suggesting rejection of Co
ions from Zn sites and the appearance of secondary phases. This
is consistent with the observation that a decrease in Raman
intensity indicates the formation of TM clusters [71,170].
Moreover, as the Cu concentration increased from 1 to 3 at.% in
Cu-doped ZnO, Raman spectra showed that besides a broad
asymmetric mode at 600 cm1 attributed to the E1 (LO) mode Fig. 7. Reported magnetic moment per Co atom for Co:ZnO lms grown on different
substrates.
of ZnO, a distinct peak at 634 cm1 arising from the Bg mode
of CuO emerged, indicating the formation of a secondary phase
of CuO [65]. Cu-doped ZnO lms exhibited 1.6 and 1.2 mB/Cu at 300 K on quartz
and sapphire, respectively [65]. RTFM with a magnetic moment of
These methods with different physical bases have been 1.5 mB/Co was observed for Zn0.95Co0.05O lms deposited on ZnO,
demonstrated to be effective in detecting second phases. However, whereas for the lms grown on Al2O3 only weak FM signals were
it is difcult to determine a trace amount of TM dopant using a detected that could not be unambiguously separated from those of
random selection of characterization techniques. The best way to the substrate. This difference can be explained by different
investigate the state of TM dopants may be to use several methods, structural properties and defect densities in the lms [51]. Thus,
and evidence of absence can only become a reality with the advent it seems that the substrate most likely impacts the local TM
of a combination of very sensitive tools. structure and magnetization of TM-doped ZnO.
To systematically study substrate effects on the local structure
3.2. Substrate effects on magnetization and magnetic behavior of TM-doped ZnO, a series of experiments
with substrate variation was designed [94]. Substrate-dependent
3.2.1. Substrate-dependent local structure and magnetic behavior magnetization of insulating Zn1xCoxO (x = 1, 4 and 10 at.%) lms
As discussed above, several reports and reviews reveal that the has been highlighted in Fig. 8, for which the substrates were 1288
magnetic behavior of DMOs appears to be dependent on YX LiNbO3 (128LN), 648 YX LiNbO3 (64LN) and 368 YX LiTaO3
preparation conditions such as the doping concentration, substrate (36LT) ferroelectric crystals, SiO2(1 0 1) piezoelectric crystal, some
temperature, and oxygen partial pressure [11,19,44,46]. However, other single crystals [e.g., Al2O3(0 0 1), Si(1 1 1), and NaCl(1 0 0)],
few studies have directly considered the substrate-dependent and amorphous glass. It is evident that the substrate has a
magnetic behavior of DMOs [94]. The best example is a review in profound inuence on the magnetic behavior of Co-doped ZnO
which statistical data imply that the magnetization of Co-doped lms. Zn0.96Co0.04O lms are taken as an example. Interestingly,
TiO2 lms is related to the crystal structure and the substrate [19]. whereas RTFM with giant magnetic moments of 6.1, 2.5, and
For Co-doped TiO2 lms grown on Si or Al2O3 substrates, the 3.2 mB/Co is observed for Co:ZnO lms on 128LN, 64LN and LT
average magnetic moment per Co atom is <1.0 mB in all cases. For ferroelectric crystals, respectively, the lm on SiO2 piezoelectric
samples using SrTiO3 as a substrate, values were scattered over
more or less the whole range of values reported, with a slight
accumulation in the lower region. However, for lms grown on
LaAlO3 substrate with an anatase structure, most of the values
accumulated in the upper region, i.e. >1.2 mB/Co, but with some
scatter between 0.23 and 1.72 mB/Co [19].
The substrates commonly used for deposition of TM:ZnO lms
are Al2O3 [11,23,45,46,52,111,162], Si [58,67,102,134,154,163],
fused quartz [33,94,103] and ZnO [51,105,186]. The reported
magnetic moment per Co atom for Co:ZnO lms grown on different
substrates is shown in Fig. 7. As the gure shows, the substrate
inuence on magnetic ordering is not clear. The data summarized
in Fig. 7 are from results obtained by different groups using
different preparation parameters. It is known that the local
structure and magnetization of DMOs are intimately related to
experimental conditions (e.g., preparation methods and para-
meters). Therefore, the preparation method, oxygen partial
pressure, substrate and temperature are all factors that can
explain the scatter from 0 to 5.9 mB/Co and from 0.1 to 1.04 mB/Co
on Al2O3 and Si substrates, respectively. However, TM-doped ZnO Fig. 8. Room-temperature magnetic moment expressed in Bohr magnetons (mB) per
lms deposited on different substrates with the same deposition Co atom for representative Zn1xCoxO (x = 1, 4 and 10 at.%) lms grown on 128LN,
parameters show distinct magnetization. For example, (0.6 at.%) 64LN, LT, SiO2, Al2O3, Si, NaCl, and glass substrates.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
crystal shows a much lower magnetic moment of 1.1 mB/Co.
Moreover, the magnetic moment decreases to 0.62, 0.66, and
0.24 mB/Co for lms deposited on Al2O3, Si, and NaCl single crystal,
respectively. The lm on the glass substrate does not exhibit any
FM signature [94]. The diamagnetic background of the substrate
has been subtracted from all of the magnetization data, indicating
that the difference in moments in Co:ZnO lms does not arise from
the substrate. A similar tendency was replicated in Co:ZnO lms
with other Co concentrations (e.g., 1 and 10 at.%).
From XPS and qualitative analysis of XANES spectra it is known
that the difference in moments cannot be ascribed to different Co
forms in Co:ZnO lms grown on different substrates, since all the
Co ions are in the 2+ valence state replacing Zn sites. The rather
robust moments on ferroelectric crystals compared to those on the
other substrates are attributed to modulation of the magnetic
moment of the Co dopant due to magnetoelectric coupling
between the Co:ZnO lm and ferroelectric crystals [94,187].
Capacitancevoltage (CV) measurements revealed that the typical
CV curve for Co:ZnO samples exhibits hysteresis with a large
memory window of 0.6 V, which can be caused by trapping/
detrapping of charges at defect sites. Evidence of defects with
trapped electrons in Co-doped ZnO would advance our under-
standing of the BMP mechanism.
The diffusion or motion of defects such as VO is then driven by
the presence of stress/strain and by an electric eld effect arising
from spontaneous polarization [149] to realize an appropriate
distance between Co2+ and defects and to achieve effective
interactions between them. This process of defect diffusion or
motion is referred as the reorganization of defects. Only this
condition is met, almost all doping Co2+ can effectively interact
with the defects producing few isolated Co2+, and the giant
magnetic moment of 6.1 mB/Co in Co:ZnO/128LN, which is slightly
lower than the ideal magnetic moment of 6.6 mB/Co calculated
using Eq. (1), can become a reality. In contrast, some Co2+ and
defects with inappropriate distances (e.g., large distances) in
Co:ZnO lms grown on non-ferroelectric substrates (without
reorganization of defects) are hard to achieve effective interac-
tions, leading to the existence of more isolated Co2+ and the
measured magnetic moment far below the ideal one. It is known
that donor spin from well-organized defects strongly correlated
with Co2+ within its orbit can mediate effective interactions
between them based on a Heisenberg exchange Hamiltonian. The
donors then tend to shape the BMP, coupling Co2+ ions within their
orbits, which try to sufciently spread out to overlap and interact
with adjacent BMPs to realize magnetic ordering, resulting in
rather robust moments that can be understood in terms of a
supercoupling mechanism [44].
It is worth pointing out that reorganization of the defects could
partly occur due to the comparatively weak polarization of the SiO2
piezoelectric substrate, leading to a lower moment per Co atom in
Co-doped ZnO lm, e.g., 1.1 mB/Co. Moreover, although the Co
dopant exists in a 2+ state when substituting for Zn2+, randomly
distributed defects would lead to very weak moments or no RTFM
in Co-doped ZnO lms on other single substrates (e.g., Al2O3, Si,
and NaCl) and glass substrate, respectively [94]. It is thus
concluded that the substrate strongly affects the magnetization
of TM-doped ZnO lms.
3.2.2. Substrate orientation-induced distinct magnetization
An interesting point in Fig. 8 is that signicant differences in
moments are observed for Co:ZnO lms grown on the same type of
substrate. For example, Zn0.96Co0.04O lm grown on 64LN exhibits a
comparatively lower moment of 2.5 mB/Co compared to 6.1 mB/Co
for lm grown on 128LN. The origin of magnetization differences on
the same type of ferroelectric substrate with only a difference in cut
11
Fig. 9. (a) Experimental and (b) MS calculated XANES spectra of Co:ZnO lms grown
on 1288 YX LiNbO3 (top) and 648 YX LiNbO3 (bottom).
direction could be explained by different CoO bond lengths [94].
Experimental Co K-edge XANES spectra of Co:ZnO lms are
compared in Fig. 9 with the ab initio calculated XANES using full
MS theory. The atomic arrangement difference between Co:ZnO on
128LN and on 64LN is the difference in CoO bond length on the
basis of the lattice parameters inferred from XRD results [94]. The
calculated spectrum displays three main features accurately
replicating those of experimental XANES spectra, denoted by A, B,
and C in order of increasing photon energy.
For Co-doped ZnO lm on 128LN, it was found that the Co
photoabsorber in tetrahedral coordination is surrounded by a rst
shell comprising one oxygen atom at 1.820 A and three oxygen
atoms at 2.069 A. This leads to a difference of 0.03 and 0.04
between features B and C in the normalized absorption experi-
mental and MS calculated spectra (upper spectra in Fig. 9a and b).
However, for Co-doped ZnO lm on 64LN, the center Co is
surrounded by a rst shell comprising one oxygen atom at 1.845 A
and three oxygen atoms at 2.097 A. This results in greater
differences of 0.08 and 0.07 for the experimental and calculated
spectra (lower spectra in Fig. 9a and b). These results demonstrate
that a slight change in CoO bond length in Co:ZnO lms has a
profound inuence on electron movement and scattering in the
vicinity of the Co atom. Subsequently, electron spin and the
corresponding moments are likely to be sensitive to the CoO bond
length. In fact, ab initio calculations predicted that FM coupling
would be affected by the MnO bond length in La0.67Ca0.33MnO3d
lms [94], and that energy differences between FM and anti-FM
alignment would be very small in Co-doped ZnO systems, with
competition between FM and anti-FM coupling [84,188].
ZnO lms doped with different Co concentrations were deposited
on LT substrates with three different orientations [LT(1 1 0),
LT(0 1 2) and LT(0 1 8)] by dc reactive magnetron co-sputtering.
The variation in magnetic moment on different orientations as a
function of the Co concentration is presented in Fig. 10.
Zn0.962Co0.038O lms deposited on LT(1 1 0), LT(0 1 2) and
LT(0 1 8) substrates exhibited different magnetization of 1.21,
2.42, 0.65 mB/Co, respectively, as shown in the inset of Fig. 10. It
is likely that the magnetization differences for Co-doped ZnO lms
on three oriented LT substrates can be attributed to the lm
microstructure and the CoO bond length, which primarily depends
on mismatch between the lm and the different oriented substrates
[107]. Therefore, as well as the substrate itself, its orientation has a
profound inuence on FM ordering of TM-doped ZnO.
12 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 10. Saturated magnetization of Co:ZnO as a function of Co concentration for
lms grown on LiTaO3 substrate in different orientations: LT(1 1 0), LT(0 1 2) and
LT(0 1 8). The inset shows hysteresis loops of Zn0.962Co0.038O lms deposited on
LT(1 1 0), LT(0 1 2) and LT(0 1 8).
3.3. Effects of substrate temperature and lm deposition rate on
magnetization
Sharma et al. [33] reported that substrate temperature affects
the magnetization of Mn-doped ZnO lms. Films prepared at a low
processing temperature exhibited RTFM, whereas samples at
standard high temperatures (T > 700 8C) showed clustering and no
RTFM. This is in agreement with work carried out by Kim et al. [46],
who found that a high substrate temperature of 700 8C led to the
appearance of secondary phases. In addition, Fe-doped ZnO
prepared by ion implantation at 253 K was a true DMO, whereas
preparation at 623 K led to the formation of particles inside the
host matrix responsible for the FM properties [42]. Very recently, it
was found that Co-doped ZnO lms grown at <200 8C exhibited
stronger FM than those deposited at higher temperatures [111].
These observations indicate that a high preparation temperature is
detrimental to RTFM of DMOs, and even results in the formation of
secondary phases within the host matrix. On the other hand,
different from other DMOs such as TM-doped TiO2 [189,190], the
local structure and magnetic properties of TM-doped ZnO lms are
more sensitive to the substrate temperature and lm deposition
rate, primarily due to temperature-dependent Zn loss during the
deposition process [58].
Song et al. [109] investigated how the electrical properties and
structural defects of TM:ZnO samples are related to FM ordering to
identify factors that affect RTFM. A suitable combination of
deposition rate and substrate temperature is key in determining
the crystallinity and corresponding magnetization of Co-doped
ZnO lms deposited by rf magnetron co-sputtering at 0.15 A/s
(slow growth) and 0.6 A/s (fast growth) at 573 K [high substrate
temperature (Tsub)] and RT (low Tsub), as presented in Table 2 [109].
For this series of lms, the sample prepared with slow growth and
Table 2
Resistivity (R), electron concentration (ne) and magnetic moments (M) for Co:ZnO samples prepared by rf magnetron co-sputtering at low/high deposition rate and low/high
substrate temperature (Tsub)
Sample Rate (A/s) Tsub (K)
Slow growth high Tsub 0.15 573
Slow growth low Tsub 0.15 RT
Fast growth high Tsub 0.6 573
Fast growth low Tsub 0.6 RT
high Tsub exhibits the lowest saturation magnetization of
approximately 0.25 mB/Co. For the lm grown at RT, the magnetic
moment slightly increases to 0.39 mB/Co. For faster deposition
rates, much larger magnetic moments of 0.73 and 0.91 mB/Co
are observed for the samples grown with high and low Tsub,
respectively. XRD results show that the intensity of the ZnO(0 0 2)
diffraction peak decreases with increasing deposition rate and/or
decreasing substrate temperature, reecting a decrease in ZnO
crystallinity. This was demonstrated by F scans, for which
comparison of the full width at half-maximum (FWHM) intensity
for peaks with sixfold symmetry revealed a decrease in crystal
quality and an increase in structural defects with increasing
deposition rate and decreasing Tsub [109].
Experimental and calculated Co K-edge XANES spectra are
presented in Fig. 11a and b, respectively. The height difference
between the characteristic B1 and C1 peaks increases as deposition
rate increases and Tsub decreases. This is reproduced by the MS
calculation (Fig. 11b). The introduction of oxygen vacancies in the
rst coordination sphere results in an obvious height difference
between B1 and C1 for the sample deposited with slow growth and
low Tsub, as well as the sample grown with fast growth and high
Tsub. For the sample prepared with fast growth and low Tsub, the
existence of oxygen vacancies in the rst coordination sphere and a
Zn interstitial between the rst two coordination spheres would
result in line shapes consistent with the experimental data. This is
supported by experimental and calculated O K-edge XAS spectra
(Fig. 11c and d, respectively). The calculations reect less of a Zn
loss during deposition for the sample prepared with fast growth
and low Tsub [58], whose spectrum is similar to that of an ideal
stoichiometric ZnO, and shoulder A2 at 531.2 eV appears and
increases in intensity from the sample prepared with slow growth
and low Tsub to the sample grown with fast growth and low Tsub,
corresponding to an increase in defect concentration in the lms, in
good agreement with microstructural characterizations [191].
The magnetic moment of Co-doped ZnO lm increases with
increasing structural defects and electron concentration (ne). It is
worth pointing out that this was the basis for the previous concept
of carrier-mediated RTFM in DMOs, because the correlation of FM
ordering and carrier concentration was emphasized. The true
origin of FM ordering, i.e., structural defects, was neglected in this
carrier-mediated hypothesis, which seriously misguided extensive
research into DMOs. To determine the intrinsic FM in a Co-doped
ZnO system, the sample prepared with slow growth and high Tsub
was annealed for 1 h at 400 8C, resulting in an increase in ne to
2.9  1019 cm3. The crystallinity of the Co-doped ZnO lm also
improved on annealing, revealing that the number of structural
defects such as edge dislocations and stacking faults decreased
[109]. The saturation magnetization slightly decreased from 0.25
to 0.19 mB/Co, revealing that an increase in ne cannot enhance the
magnetization of Co-doped ZnO after annealing at 400 8C. Instead,
the magnetic moment per Co atom decreased with increasing
crystallinity, which allows identication of a direct relationship
between the FM of Co-doped ZnO and the presence of structural
defects. This implies that the established theory of carrier-
mediated exchange, which works well for p-type GaMnAs, is not
R (V cm) ne (cm3) M (mB/Co)
>103 No Hall signal 0.25
5.6 2.3  1018 0.39
1.9  101 1.5  1019 0.73
1.7  102 5.2  1019 0.91
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135 13

Fig. 11. (a) Experimental and (b) calculated Co K-edge XANES spectra for Co:ZnO lms (E0 = 7708.8 eV). (c) Experimental and (d) calculated O K-edge XAS for Co:ZnO lms.
Spectra correspond to samples grown with slow growth and high Tsub (LH), slow growth and low Tsub (LL), fast growth and high Tsub (HH) as well as fast growth and low Tsub
(HL) from bottom to top.

generally applicable to TM-doped ZnO. Instead, structural defects corresponding to a deposition rate of 3.0 and 2.0 nm/min,
play a primary role in RTFM of TM-doped ZnO, and carriers respectively [112]. Fig. 12 shows the PO2 -dependent RTFM and
involved in carrier-mediated exchange are natural by-products of resistivity (R) of the lms. Both the magnetization and resistivity
the creation of defects, which is demonstrated in publications with decrease with decreasing PO2 . Insulating sample 1 (0.62 mB/Co)
increasing frequency [60,76]. exhibits more robust RTFM than semiconducting sample 2
(0.45 mB/Co), revealing that an increase in carrier concentration
3.4. Inuence of oxygen partial pressure on magnetic ordering does not enhance RTFM. A similar trend was observed by
Chakraborti et al. [137] and Ivill et al. [131,138]: an increase in
In general, both the oxygen partial pressure (PO2 ) and the conductivity did not increase the magnetic moment; instead, an
substrate temperature (Tsub) play important roles in the micro- increase in carrier concentration led to weak magnetization,
structure of DMO lms, which subsequently affects the magnetic suggesting that a mechanism mediated by free carriers is not a
properties. PO2 and Tsub affect the formation of secondary phases feasible explanation for FM in TM:ZnO lms.
[46,33]. Lee et al. [98] reported that ne dramatically decreased with To interpret the intrinsic RTFM in this system, microstructural
increasing PO2 up to an Ar/O2 ratio of 1:1 for the preparation of and local structural characterizations were conducted [112]. XRD
Mn:ZnO lms by rf magnetron sputtering, which showed intrinsic spectra in BraggBretano, rocking curve and F scan geometries
FM with TC of 78 K. However, lms grown under O2 overpressure suggested that the lms were highly crystalline, as shown in
showed various secondary phases, such as ZnMn3O4, MnO2 and Fig. 13. The XRD pattern of sample 1 (Fig. 13a) has a weaker
ZnMnO3. Films deposited at lower temperature (600 8C) and Co:ZnO(0 0 2) peak than that for sample 2 (Fig. 13b). Comparison of
higher oxygen pressures (105) showed a homogeneous dis-
tribution of Zn0.75Co0.25O, leading to spin-glass behavior, whereas
mixed phases of wurtzite ZnO with rock-salt CoO and hexagonal
Co phases coexisted at higher temperatures and lower pressures,
accompanied by RTFM [46]. On the other hand, variation of PO2
can lead to a change in magnetic behavior and electric properties.
On increasing PO2 from 1.0  104 to 5 Pa, the resistivity of
Zn0.95Co0.05O lms increased from 2.15  102 to 2.06  104 V cm,
corresponding to a change in magnetization from RTFM to
paramagnetism, revealing that oxygen vacancies are absolutely
necessary to induce FM coupling in TM-doped ZnO [168]. Philip
et al. [243] observed that the electric and magnetic behavior
of Cr-doped In2O3 lms from FM metal-like to FM semiconduct-
ing to paramagnetic insulators can be controllably tuned
by P O2 . It seems that PO2 plays an important role in the
magnetization of DMOs.
A transition from DMI to DMO was realized in Zn0.96Co0.04O Fig. 12. Magnetic moment and resistivity of Zn0.96Co0.04O lms as a function of P O2 .
lms grown by dc reactive magnetron co-sputtering by decreasing The inset shows magnetic hysteresis loops (MH) for Co:ZnO lms 1 and 2 prepared
P O2 from 4.2  103 to 1.0  103 torr for samples 1 and 2, at high and low P O2 , respectively.
14 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

Fig. 13. XRD spectra of Co:ZnO lm for (a) sample 1 and (b) sample 2. X-ray rocking curves from the wurtzite (0 0 2) plane for (c) sample 1 and (d) sample 2. F scan of {1 0 1}
family peaks for (e) sample 1 and (f) sample 2.

FWHM values for X-ray rocking curves for the wurtzite(0 0 2) RTFM in Co:ZnO lms by controlling the lm thickness from 15 to
plane for samples 1 and 2 reveals an improvement in crystallinity 900 nm (15, 30, 60, 100, 150, 300, 600 and 900 nm). The lms were
of sample 2 (0.188 decrease in the FWHM value; Fig. 13c and d). The deposited on Si(1 0 0) and LiNbO3(1 0 4) substrates by rf magne-
F scan for the (1 0 1) plane of sample 2 grown at lower PO2 is tron sputtering. The thinnest lms (15 nm) with greatest lattice
similar to that of sample 1 to a certain extent, but with more strain had the highest saturated magnetic moments of 2.96 and
symmetrical peaks. These results indicate that sample 2 is more 5.52 mB/Co for Co:ZnO/Si and Co:ZnO/LiNbO3, respectively (Fig. 15)
crystalline and has fewer structural defects such as edge The saturated FM rapidly decreased with increasing lm thickness;
dislocations and stacking faults due to the comparatively slow
rate of lm growth [189,190].
Fig. 14 presents Co K-edge XANES for samples 1 and 2 and for
reference materials (Co metal, CoO, Co2O3 and Co3O4). It is evident
that Co in Co:ZnO is not in the form of Co metal or Co-based oxides
such as CoO, Co2O3 and Co3O4. The calculated spectra displays
three main features accurately replicating those of the experi-
mental spectra, denoted by A, B, and C in order of increasing photon
energy (insets a and b of Fig. 14). These spectra reveal that for
sample 1 the Co photoabsorber in tetrahedral coordination is
surrounded by a rst shell comprising four oxygen atoms, with no
difference in normalized absorption between features B and C
(upper spectra in insets a and b). For sample 2, the Co center is
surrounded by a rst shell of three oxygen atoms and a VO, leading
to an obvious difference between features B and C for the
experimental (0.13) and calculated (0.11) spectra (lower spectra in
insets a and b). It is therefore concluded that there are more VO and
thus a higher carrier concentration in sample 2 due to lower P O2 .
However, sample 1 contains more structural defects than sample 2,
resulting in stronger RTFM for sample 1, demonstrating that
structural defects are responsible for RTFM in Co-doped ZnO.

3.5. Strain-induced ferromagnetism enhancement

Strain engineering, which offers the opportunity to enhance the

physical properties of a chosen material in thin lm form by
judicious use of growth parameters, represents a new direction for
nanoscale material research [192194]. It has been found that a
slight strain-induced break in symmetry has a great inuence on
the single spin and the corresponding magnetic behavior of typical
DMSs such as Mn:GaAs [195]. Given that growing lms of different Fig. 14. Co K-edge XANES spectra for Co:ZnO and references materials Co metal,
thickness is a simple and effective method to modulate the lattice CoO, Co2O3 and Co3O4. The insets show (a) experimental and (b) MS calculated
strain within lms, Liu et al. [196] studied the effects of strain on XANES spectra for samples 1 (square) and 2 (circle).
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 15. Saturated magnetization as a function of thickness for Co:ZnO lms
deposited on Si and LiNbO3 substrates measured at room temperature. The inset
shows the hysteresis loop for the 15-nm thick Co:ZnO/LiNbO3 lm.
the results for the 900-nm lm was 0.02 mB/Co, which is
approximately two orders of magnitude lower than for the 15-
nm thick lm.
Fig. 16a shows XRD patterns for Co:ZnO/Si lms of different
thickness. Comparison of the (0 0 2) peaks reveals a shift from
33.848 for the 15-nm lm to 34.458 for unstrained ZnO,
corresponding to a change in c-axis lattice constant from 5.207
to 5.553 A, indicating that the unit cell in the Co:ZnO lm is
elongated along the c axis. With increasing lm thickness, the
ZnO(0 0 2) peak position shifts towards the value for unstrained
ZnO. The c-axis lattice constant and calculated lattice strain are
plotted as a function of lm thickness in Fig. 16b and c. The c-axis
lattice constant decreases slightly with increasing thickness. Thus,
the initial stage of Co-doped ZnO growth represents a strongly
Fig. 16. (a) XRD patterns of Co:ZnO/Si lms of 15900 nm in thickness. (b) Lattice constant and (c) c-axis lattice strain as a function of lm thickness.
15
accumulated state of strain energy due to the large lattice
mismatch (40.10%) between the lm and the Si substrate, and
the 15-nm thick Co-doped ZnO lm exhibits the greatest tensile
strain of 6.65% (Fig. 16c). As the lm thickness increases, the
accumulated strain energy progressively relaxes, i.e., the c-axis
tensile strain continuously decreases, and is only 0.69% for the 900-
nm lm, corresponding to weak FM. The FM enhancement in Co-
doped ZnO lms originates from a combination of an increase in
CoO bond length and an increase in the number of defects
induced by lattice strain [196]. These results are consistent with
the observation that lattice deformation plays a signicant role in
the high-TC FM of DMOs [197].
3.6. Inuence of post-annealing on magnetization
Annealing has a great effect on the magnetization of TM-doped
ZnO, largely because it can change the microstructure and local
structure of lms [38,52,55,57,73,109,140,141,154]. It is com-
monly observed that annealing in vacuum or a reducing atmo-
sphere results in enhanced FM of the samples due to the
introduction of VO [52,55,140,141], with Zni critical for RTFM in
TM-doped ZnO [30,57,141,154,198]. For example, a study of
annealing effects on the local structure and magnetism of Co-
doped ZnO lms under air, Ar, and Ar/H2 atmospheres revealed
that lm annealed in Ar/H2 had a pronounced pre-edge Zn K-edge
XANES peak and contained more VO (1  1021 cm3) and stronger
RTFM, indicating that RTFM is strongly dependent on VO [55].
Furthermore, a change in Raman spectra (appearance/disappear-
ance of an additional broad vibrational mode) for Co-doped ZnO
lms on alternate annealing in air/high vacuum suggested that VO
is crucial in the development of RTFM [100]. Using alternate
annealing of Co-doped ZnO lms in air and Zn vapor, Kittilstved
et al. [198] demonstrated quantitative reversible cycling of
saturated magnetic moments, and direct correlation between
oxidative FM quenching and the diffusion and oxidation of Zni.
Similarly, annealing of lms in different oxidation atmospheres in
16 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

the presence or absence of Zn vapor revealed that RTFM is not

linked to conductivity and that a small level of Zni defects is
important in mediating RTFM in TM:ZnO [154,102]. Similar
phenomena have been observed for TM-doped ZnO nanocrystals
[20,199,142]. In contrast, a decrease in VO concentration due to
high-temperature annealing in oxygen had a large detrimental
effect on FM, which tends to disprove a mechanism involving free
carriers [200,177]. These experimental annealing results demon-
strate that the magnetic behavior observed is directly related to the
presence of intrinsic defects, notably VO and Zni. Electrons locally
trapped by these defects occupy an orbital overlapping with the d
shells of TM neighbors to form a BMP, which accounts for the
nature of intrinsic RTFM.
The microstructure and magnetic properties of Zn0.92Co0.08O
lms annealed at different temperatures (200, 400 and 600 8C) in
vacuum and air were systematically investigated by Liu et al. [141].
The magnetic moment increased after annealing in vacuum and
decreased after annealing in air at the same temperature. A typical
plot of hysteresis loops for as-grown lm and lms annealed in air
and vacuum at 400 8C is shown in Fig. 17. The vacuum-annealed
lm shows saturation magnetization (Ms) of approximately
0.45 mB/Co, which is slightly greater than for the as-grown lm
(0.34 mB/Co), whereas the value for the air-annealed lm (0.21 mB/
Co) is much lower than Ms of the as-grown lm. The value of Ms
increases with annealing temperature up to 400 8C due to the
creation of VO during annealing in vacuum [149]. However,
annealing at 600 8C greatly decreased the FM of Co-doped ZnO
lms, likely due to an improvement in crystal quality and a
corresponding decrease in structural defects in the lm [141]. This
is supported by the result indicating that FM ordering increased
with annealing of Zn0.88Co0.12O lms in vacuum up to 500 8C,
whereas a higher annealing temperature resulted in a decrease in
magnetization [111].
On annealing in air at high temperature (>700 8C), Mn2+ and
2+
Co ions may segregate from the crystal and form impurity phases
Fig. 18. (a) Co K-edge XANES spectra for as-deposited and annealed (750 8C) sample
or precipitates, which can lead to a decrease in Raman peak LH, Co metal, CoO and Co3O4 (E0 = 7708.8 eV). The inset shows an FE-SEM image of
intensity [71]. The XANES spectrum of Co:ZnO lm annealed at LH after annealing at 750 8C. (b) Magnetization of as-deposited and annealed
750 8C differs from that of an as-deposited lm, but closely (750 8C) LH.
matches that of Co3O4 (Fig. 18a) in terms of a similar absorption
edge, characteristic peaks, and normalized absorption intensity.
This can be ascribed to a transition from Co2+ replacement of Zn2+ in size from 60 to 120 nm. This transition in local structure (from
to the formation of Co3O4-based compounds. This is consistent Co2+ replacement of Zn2+ to Co3O4-based compounds) during
with a FE-SEM image showing small islands of precipitates 750 8C annealing means that the behavior of Co-doped ZnO lms
(Fig. 18a inset) that are almost randomly distributed and range changes from RTFM to paramagnetism, as illustrated in Fig. 18b.
These phenomena observed after annealing are consistent with the
principles described for substrate temperature-dependent mag-
netization. A combination of a comparatively low temperature
(400 8C) and a reducing atmosphere increases defects (VO or Zni)
and enhances FM ordering [33,94], whereas high-temperature
processing results in a decrease in RTFM.

3.7. Effect of co-doping on magnetic properties

Controlling the experimental parameters, e.g., decrease in P O2 ,

Fig. 17. Hysteresis loops for as-grown and vacuum- and air-annealed (400 8C)
Zn0.92Co0.08O thin lms measured at room temperature.
can increase the carrier concentration and thus enhances FM
[11,48,62,95,103,152,201]. However, some researchers have
reported that the creation of free carriers is not sufcient for FM
interactions [77,202]. Moreover, several groups observed RTFM in
the absence of free carriers [8,30,41,44,45,75,94,108,109,152,
153,186].
To elucidate whether the carrier concentration can enhance
RTFM or if there is a correlation between them, co-doping has been
applied to generate additional electron/hole doping, as presented
in Table 3. The possible effects of co-doping, i.e., simultaneous
doping of ZnO with two or more TM ions, are interesting. Typical
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135 17

Table 3
List of co-doped ZnO systems, including lms, bulk and nanocrystals

System Content (at.%) Methods Magnetism Reference

CoAl 9/1 PLD 0 ! 0.1 mB [52]

CoAl 10/(0,1.6) PLD 0.33 ! 0.11 mB [186]
CoAl 5/(00.5) rf sputtering 1.7 ! 3.36 mB [129]
CoAl 5/(03) Solgel 0.30 ! 0.48 mB [206]
CoAl 5/(02) PLD 0.88 ! 1.6 mB [95]
MnAl 2/(04) PLD 0.1 ! 4.36 mB [95]
MnSn (3,5)/ Ion implantation 5.5  105 emu [35]
MnSn 3/(00.1) PLD 0.1 ! 0.5 mB [131]
CuGa 5/1 PLD 1.45 ! 0 mB [132]
CoLi (510)/(010) Solgel 50 ! 375 emu/mole [133]
CoLi 5/5 Spin-coating 0 ! 0.5 mB [134]
CoLi 5/10 Solgel 0.009 ! 0.048 mB [203]
CoN 1.5/ rf sputtering 0.085 ! 0.129 mB [204]
CuN 10/ dc sputtering 0.36 ! 0.15 mB [66]
MnN 4/ ICP-CVD 0.3 ! 1.4 mB [205]
MnN 5/7 rf sputtering 0 ! 0.3 mB [135]
MnN 3/ rf sputtering 0 ! 0.92 mB [136]
MnP 3/2 PLD 0.2 ! 0.07 mB [138]
MnCu 10/(0,5) PLD 0.075 ! 0.1 mB [125]
FeCu 5/(01) Solid state reaction 0 ! 0.75 mB [126]
CoCu 5/(010) PLD 5 ! 15 emu/cm3 [137]

Content is presented as x/y for the main TM element/additional dopant. Magnetism results indicate the change in parameter before ! after co-doping.

electron doping elements in TM-doped ZnO are Al, Sn and Ga
[52,95,129,131,132,186,35], whereas N, Li, and Cu are used for hole
doping [125,126,129,203205]. It has been reported that an
increase in carrier concentration due to trace amounts of
additional dopants can greatly enhance RTFM [95,126,206].
Additional hole doping, using Li or N promotes strong hybridiza-
tion of Mn 3d/N 2p states and long-range FM coupling, as it brings
the d electron/TM closer to the optical values for double exchange
[133,134], although some authors reported that co-doping plays a
negligible role in enhancing the magnetic moment [125]. Ivill et al.
[131] designed a series of experiments on (Mn,Sn)-co-doped ZnO
with varying Sn concentration (00.1 at.%). They found that the
lms exhibited FM, with an inverse correlation between magne-
tization and electron density, as a function of the Sn doping. This
phenomenon is most consistent with the BMP model in which
bound acceptors mediate FM doping. In this case, an increase in
electron density decreases the acceptor concentration and thus
quenches the FM exchange. Similar conclusions have been
reported for (Mn,P)- and (Cu,Ga)-co-doped ZnO [132,138].
FM enhancement arising from an increase in donor defects or
electron carrier concentration is ambiguous because these two
factors coexist in most doped ZnO systems and are hard to
distinguish. It is hence of vital importance to design an experiment
that can separate the two factors [129,204]. Zn0.95xCo0.05AlxO
(x = 0, 0.001, 0.002, 0.004, 0.005) thin lms were deposited on
Si(1 0 0) substrates by rf magnetron co-sputtering [129]. Fig. 19
shows XRD spectra of Zn0.95xCo0.05AlxO (x = 0, 0.002) lms. The
peak corresponding to ZnO(0 0 2) changes little except for a
decrease in intensity after additional Al doping in the Zn0.95Co0.05O
thin lm, implying a decrease in crystalline quality and an increase
in structural defects on inclusion of a trace amount of Al. This was
conrmed by optical transmittance spectra of pure ZnO and
Zn0.95xCo0.05AlxO (0  x  0.005) [129].
Fig. 20 shows curves of the magnetization versus magnetic eld
(MH) for Zn0.95xCo0.05AlxO (0  x  0.005) thin lms. The Al
concentration dependence of Ms and ne is shown in the inset. It is
evident that a trace amount of Al can enhance the magnetic
moment per Co ion of Co:ZnO lms. The increase in Ms with

Fig. 20. Electron carrier concentration and saturation magnetization as a function of

Fig. 19. Comparison of XRD patterns for pure ZnO, Zn0.95Co0.05O and Al concentration. The inset shows magnetization hysteresis curves of
Zn0.95xCo0.05AlxO (x = 0, 0.002) thin lms on Si(1 0 0) substrates. Zn0.95xCo0.05AlxO thin lms measured at RT.
18 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

additional Al doping, combined with the increase in ne, reveals that
the free carrier concentration is important, as expected for DMOs.
However, it should be noted that ne (1.597  1020 cm3) is greater
and Ms (2.34 mB/Co) is smaller for the Zn0.95xCo0.05AlxO (x = 0.004)
lm than for the Zn0.95xCo0.05AlxO (x = 0.001) lm
(ne = 4.597  1019 cm3, Ms = 2.71 mB/Co), indicating that elec-
trons are not the real mediators of FM exchange. It is clear that
additional Al doping leads to a decrease in both the crystalline
quality and the number of crystal grains in Co:ZnO lms, indicating
that a large number of structural defects are introduced, mainly
located in grain boundary areas in the samples. In addition, the
concentration of donor defects (notably VO and Zni) increases to a
certain extent [58,109], accounting for the FM enhancement.
It is noted that the ne dramatically increases up to x = 0.002 and
then remains relatively constant with further increases in Al
concentration, implying that Al is not the only candidate
responsible for introducing electron carriers, as shown in
Fig. 21a (red circle). It is well known that ZnO is a natural n-
type semiconductor owing to the existence of native defects such
as VO and Zni. An increase in structural defects could enhance the
concentrations of point defects such as VO and Zni in the lms. Such
an increase in point defects, as well as the native defects, is
regarded as another primary donor source of electron carriers.
Since free electrons can be introduced into Co:ZnO lms not only
by additional Al doping, but also by an increase in donor defects,
the ratio of the carrier concentration to the Al density (ne/nAl) is an
indirect measure of the number of defects in a lm, with a larger ne/
nAl value indication a greater number of donor defects. For
Zn0.95xCo0.05AlxO (x = 0.001), ne/nAl = 1.09 is greater than the
value for x = 0.004 (ne/nAl = 0.95), implying a greater number of
donor defects in the former lm and corresponding to the stronger
Ms of 2.71 mB/Co, indicating that donor defects most likely play a
critical role in governing FM ordering.
A systematic study by Mokhtari et al. [207] revealed that
additional doping could ionize other defects, which played an
important role in FM enhancement. Fig. 21 presents the change in
carrier density as a function of the Al concentration in (a) Co- and
(b) Mn-doped lms. Data obtained for lms prepared by PLD at
0.05 mtorr and by magnetron sputtering are denoted by black and
red circles, respectively. The circles around the data points indicate
the magnitude of the magnetic moment measured at RT. The solid
lines correspond to addition of one carrier per Al atom. It is evident
that all data lie signicantly above this line, indicating that the
presence of Al has ionized other defects, perhaps some Co2+ and
Mn2+, which provides carriers in the lms.
Acceptor co-doping using Li and N has also been extensively
studied to determine the origin of RTFM [134,204]. Fig. 22 shows
unpolarized Raman scattering spectra measured at RT for
(Co,N):ZnO lms. Vibrational modes are evident at 273, 470,
639, 691 and 854 cm1, in addition to the host phonons of ZnO,
indicating the substitutional behavior of N3 for O2 and an
increase in donor defects for (Co,N)-co-doped ZnO lms [204]. The
higher intensity of the vibrational mode at 577 cm1 indicates a
higher concentration of donor defects in (Co, N):ZnO lms than in
Co:ZnO lms. Acceptor doping with nitrogen compensates the
electron carriers and hence decreases ne [135]. Magnetization
measurements reveal that co-doping increases Ms for (Co,N)-co-
doped ZnO (0.129 mB/Co) compared to Co:ZnO (0.085 mB/Co)

Fig. 21. Carrier density of (a) Zn0.95xCo0.05AlxO and (b) Zn0.98xMn0.02AlxO lms. The solid line corresponds to addition of one carrier per Al atoms. Films were grown by PLD at
0.05 mtorr (black) [207] and magnetron sputtering (red), respectively [129]. The thickness of the circles indicates the magnetic moment in Bohr magnetons. The circles lie
signicantly above the line, indicating that the presence of Al has ionized other defects and also possibly some of the Co2+. (For interpretation of the references to color in this
gure legend, the reader is referred to the web version of the article.)
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 22. Raman spectra of ZnO, Co:ZnO and (Co,N):ZnO lms. ZnO and silicon
substrate modes are marked by (*) and (#), respectively.
[204]. Combined with FM enhancement, donor defects other than
electron carriers are responsible for RTFM in TM-doped ZnO lms.
3.8. Comparison of ZnO doped with Sc, Ti, V, Cr, Mn, Fe, Co, Ni and Cu
More than 800 reports on TM-doped ZnO DMOs have been
published from 2000 to date. A large number of these focus on Co-
[8,11,23,4455,5860,97,103,111,126,141,152,154,164,165,198]
and Mn-doped ZnO lms [3137,39,40,73,98,117,140,186]. There
have also been numerous experiments on ZnO lms doped with
other TM elements, including Sc [23,24], Ti [23,25], V [2628], Cr
[11,23,29,30,208], Fe [4143,124], Ni [62,63] and Cu [6467,132].
Some of the magnetic moments measured at room temperature
reported for 5 at.% TM-doped ZnO lms are summarized in Fig. 23.
ZnO lms doped with 5 at.% Co commonly exhibit much stronger
RTFM than ZnO doped with other TM elements. This is consistent
with results obtained by Ueda et al. [11] and Venkatesan et al. [23]
for ZnO doped with a series of TM elements; such studies can
exclude the inuence of the preparation parameters on magnetic
ordering. It is evident that ZnO lms doped with different TMs
exhibit RTFM with different Ms values, as well as very small
magnetic moments indicating negligible FM. It is tempting to
Fig. 23. Magnetic moments reported for (5 at.%) TM:ZnO at room temperature.
19
relate this difference directly to the magnetization of ZnO doped
with different TM dopants.
The Lande g-factor can be expressed as [175]:
3 1 LL 1  SS 1
g  ; (2)
2 2 JJ 1
where J is the total angular momentum, including the spin (S) and
orbit (L) angular momentum. The Lande g-factor and J can be used
to evaluate the effective magnetic moment of TM ions in TM:ZnO
lms. For Co2+ in Co:ZnO, the Lande g-factor is 4/3, the effective
magnetic moment calculated using Eq. (1) is 6.6 mB/Co and the net
spin moment calculated from the spin-only formula mmax g mB S
is 3 mB. Similarly, Mn2+ has both an effective and net spin magnetic
moment of 5 mB. These different ideal magnetic values for TM ions
are most likely one of the reason for different moments of TM:ZnO.
On the other hand, it is generally considered that Co is more
soluble than other TM atoms in ZnO lms [11,44,105]. In terms of
concentration-dependent magnetization, the largest moments
have been observed for ZnO doped with different TM concentra-
tions, i.e., 35 at.% for Co [44,58,103], 2 at.% for Mn [33,95], and
1 at.% for Cr, Ni and Cu [62,64,66], and ZnO lms with other dopant
contents tend to show much lower magnetic moments. Therefore,
besides ideal moment discrepancies for these TM ions in ZnO lms,
different solubility and different maximum dopant concentration-
dependent magnetization would also account for different
magnetic moments in 5 at.% TM-doped ZnO lms.
4. Origin of ferromagnetism in TM:ZnO
As discussed in Section 3, almost all of the publications that
reported magnetization of TM:ZnO referred to the origin of
magnetic ordering. This is very important not only for under-
standing the nature of TM:ZnO, but also for applications in useful
spintronics. Many points and mechanisms have been proposed for
intrinsic magnetic ordering. However, it is hard to experimentally
characterize and quantitatively analyze the nature of RTFM
because so many factors affect the magnetic behavior of TM-
doped ZnO, including the microstructure, local structure, electro-
nic structure, coupling between localized magnetization, and trace
amounts of TM dopant atoms in complicated oxide lms. On the
other hand, FM values reported seem to be rather weak, often
many orders of magnitude smaller than the expected spin-only
saturation moments. The samples that exhibit FM often do so at
levels low enough so that no technique other than magnetization
itself can reveal whether the phenomenon is intrinsic or arises
from impurities [209]. Magnetization is measured using a SQUID
magnetometer with a high sensitivity of 108 emu; trace FM
contamination of the substrate can lead to lms with magnetiza-
tion signals, which means that measurements may be ambiguous.
Detection of the local structure and electronic structure of TM
elements in TM:ZnO lms is generally limited by the sensitivity
and efciency of the equipment used [85,139], which can lead to
false conclusions. Reports that cast doubt on the very existence of
any type of long-range magnetic order in clean samples of TM-
doped ZnO at low doping concentrations are also appearing with
increasing frequency [209]. There is ongoing debate about these
systems. Therefore, from the broad spread of experimental data it
is difcult to draw general conclusions about the nature of
magnetic interactions in TM-doped ZnO. The confusing situation
clearly calls for a critical and even subjective position to be taken
on this topic. So far, the most popular mechanisms suggested for
magnetic ordering in TM-doped ZnO are carrier-mediated
exchange (electron and hole carriers), and the BMP model with
20 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
respect to defects, including VO and Zni, and Zn vacancies, which
are described in detail in this section.
4.1. Computational work
Computational work based on ab initio calculations can provide
some initial predictions and explanations of the electronic
structure and the nature of FM ordering in TM-doped ZnO. To
simulate realistic dopant concentrations, calculation are required
for supercells containing a large number of atoms, in which the
doping concentrations are usually higher than 5 at.% [84,210].
Hence, computational results for TM:ZnO have been produced
using a range of electronic structure implementations, generally
using a density functional theory (DFT) approach based on
pseudopotentials with localized atomic-orbital basis sets, in which
the total energy of a many-electron system is described as a
function of the electron density [19,84,210,211]. This method
combines accuracy with low computational cost, particularly
compared to other approaches such as plane-wave pseudopoten-
tial or all-electron methods [84]. The most widely used approx-
imation is the local density approximation (LDA) exchange
correlation function [10]. Relaxation of the constraint of an equal
occupation of spin-up and spin-down states and minimization of
the total energy with respect to spin occupancy results in the local
spin-density approximation (LSDA) [19,133].
It is known that LDA or LDSA calculations fail in predicting the
insulating behavior of many TM oxides, but produce the metallic
ground state. Incorporation of the effective on-site Coulomb
interaction, characterized by Hubbard U (LDA + U), between TM 3d
electrons into LDA or LDSA calculations greatly improves the
interaction for TM oxides [212,213]. On the other hand, the highly
precise projected augmented wave (PAW) method within the
generalized gradient approximation (GGA) or GGA + U scheme,
which includes an explicit dependence of exchange-correlation
energy on the electron density gradient, has also been used in
calculations [33,10,214,215].
In 2000, Dietl et al. [9] theoretically predicted that p-type ZnO
doped with 5 at.% Mn and similar TM-doped wide-bandgap
semiconductors would show RTFM, which focused much attention
on TM:ZnO research. The prediction used a mean eld approach,
assuming hole-mediated FM interactions between Mn local
moments. TC was then determined from competition between
this long-range carrier-mediated FM coupling and short-range
MnMn anti-FM exchange interactions. Following this prediction,
Sato and Katayama-Yoshida [10] used the KorringaKohn
Rostoker (KKR) Green function method based on LDA of density
functional theory to calculate the properties of ZnO doped with
TMs such as V, Cr, Mn, Fe, Co, and Ni. Additional holes were
introduced by replacing O with N, and electrons were added by
replacing some of the Zn atoms with Ga. The energy difference
between FM and anti-FM ordering was taken as the upper TC limit.
This energy difference (EFM  EAFM) was in the region of 0.2 eV,
which corresponds to a high TC. An FM ground state would be
favored at a doping level of 25 at.% V, Cr, Fe, Co, or Ni in ZnO
without additional dopants to provide carriers, whereas Mn-doped
ZnO was anti-FM.
Spaldin [84] identied potential FM ground states within
(Zn,Co)O or (Zn,Mn)O using the DFT approach with an LSDA
exchange-correlation function. This is achieved for 32-atom
supercells containing two dopant ions in different positional
arrangements: a close conguration in which TM atoms were
separated by a single O ion, and a separated conguration in
which they were connected by an OZnO bond. The calcula-
tions show that the energy difference between the anti-FM and FM
spin orders was only a few meV, producing paramagnetic behavior
in the absence of free carriers. Since the total energy calculated is
very close for FM and anti-FM orders [84,210], actual TM-doped
ZnO systems observed in experiments may remain in different
metastable states that exhibit FM, anti-FM, and spin-glass
behavior, depending on subtle differences in experimental
conditions, even for similar lms.
More recently, Ye et al. [211] found that the FM state was
43 meV lower than the anti-FM state, and thus 12.5 at.% Cu-doped
ZnO was predicted to be in the ground state with a net magnetic
moment of 1 mB/Cu and an estimated TC of 380 K. In contrast, it
was found that the close case for Mn:ZnO favors an anti-FM state,
whereas the separated conguration tends to be an FM state, stable
by only 0.01 eV [216]. Furthermore, different geometries with
FM, ferrimagnetic and anti-FM congurations were found to be
nearly energetically degenerate for both bulk V:ZnO and subsur-
face layers of the lms, where V atoms couple ferromagnetically
when they occupy surface sites of the lms [210]. On the other
hand, Lee and Chang [188] calculated the energy difference
between FM and anti-FM alignments for two Co atoms substituted
in ZnO using the GGA approximation in supercells of 2  2  1,
2  1  2 and 4  1  1 to allow different CoCo distances. The
total energy difference between the FM and anti-FM states was
<3 meV per Co atom, indicating that competition between FM and
anti-FM coupling exists in Co-doped ZnO and that spontaneous
magnetization is not possible in intrinsic Co-doped ZnO, which
theoretically supports FM at 300 K in some Co-doped ZnO lms,
but the reproducibility was less than 10% [11]. Meanwhile, high
concentrations of both Co ions and electron carriers are needed to
achieve FM [188,217]. These diverse magnetic behaviors predicted
by calculations could be one of the main reasons for the conicting
experimental results reported.
However, for both Co- and Mn-doped ZnO, the FM state was
stabilized by the introduction of holes, such as the creation of Zn
vacancies in (Co,Cu)-doped ZnO. Following this prediction, the
effects of Zn vacancies on FM were experimentally and theore-
tically investigated [34,218,219]. Some groups demonstrated that
Zn vacancies promote the appearance of RTFM by rst-principle
calculations [34,218,219]. Comparison of experimental and
calculated O K-edge XANES revealed that Zn vacancies indeed
exist in Mn-doped ZnO lms and account for the RTFM. Likewise,
the presence of Fe3+ is a signature of hole doping induced by Zn
vacancies. A deep acceptor trap due to Zn vacancies with thermal
activation energy of 0.815 eV has also been detected by deep-level
transient spectroscopy [49]. Karmakar et al. [218] revealed
experimentally and theoretically that hole doping is crucial in
promoting FM in Fe-doped ZnO systems, and the presence of Fe3+ is
a signature of hole doping induced by Zn vacancies. Furthermore,
Yan et al. [34] recently demonstrated that Zn vacancies favor RTFM
in Mn-doped ZnO lms using simulations of O K-edge XANES and
rst-principle calculations; similar calculations and results were
reported by Iusan et al. [219]. To elucidate the effects of co-doping
and defects on magnetic ordering in TM-doped ZnO, Gopal and
Spaldin [213] systematically investigated four cases: (a) single-TM
(Cr, Mn, Fe, Co, Ni and Cu) substitution at Zn sites; (b)
substitutional magnetic TM ions combined with additional Cu
and Li dopants; (c) substitutional magnetic TM ions combined with
VO; and (d) pairs of magnetic ions (Co and Fe, Co and Mn). The
results indicate that FM ordering of TM ions is not induced by
addition of substitutional TM impurities or VO. Incorporation of
interstitial or substitutional Li is favorable for FM, as are Zn
vacancies, which is consistent with experimental results for
(Co,Li)-co-doped ZnO [133,134].
We cannot draw denite conclusions regarding intrinsic RTFM
from the computational work carried out. Electron carriers
[188,220] and hole doping [49,84,133,134,213,221] seem to be
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
necessary for FM ordering. Gopal and Spaldin [213] proposed that
the calculated magnetic behavior is strongly dependent both on
the computational details (with FM disfavored by improved
convergence) and on the choice of exchange correlation function
(with FM disfavored by the more appropriate LSDA + U method).
This observation explains the large spread of computational results
in the literature. Fortunately, there is some common ground: most
publications agree that intrinsic FM does not occur in TM-doped
ZnO at reasonable doping concentrations, but in many cases the
addition of carriers or point defects stabilizes the FM state
[222,223]. These additional carriers or point defects are in fact by-
products of structural defects in TM-doped ZnO, which have
frequently been demonstrated by recent experiments and appear
with increasing frequency.
4.2. Experimental work
Fueled by theoretical predictions, claims of FM at and above
room temperature in TM-doped ZnO have abounded. Unfortu-
nately, neither the true nature of these materials nor the physical
causes of the magnetism have been adequately determined. It is
known that it is difcult to introduce factors that may inuence
RTFM, such as defects and carriers, into calculations. Moreover, it is
not easy to realize low doping concentrations (<2 at.%) in TM-
doped ZnO supercells because of the limitation of the calculation
efciency of rst-principle calculations. So far, hundreds of studies
have experimentally investigated intrinsic RTFM by focusing on
correlations among the magnetization, microstructure, local
structure and electric properties. Besides the extrinsic origin of
RTFM arising from FM secondary phases, several mechanisms have
been proposed for intrinsic magnetic ordering, including carrier-
mediated exchange (electron and hole carriers) and a BMP model
in terms of structural defects (VO and Zni), which are described in
details in the following subsections.
4.2.1. Secondary phases
Before elucidating the mechanisms of intrinsic FM, it must rst
be conrmed whether the magnetic hysteresis loops measured are
intrinsic or extrinsic. Magnetic contamination [86], measurement
errors [139] and secondary phases [31,110] are all possibilities for
FM signals. Low levels of FM contamination in the substrate and
lm can yield magnetization signals detectable by magnetometers,
which have a high sensitivity of 108 emu. The mass of a 100-nm
thick lm on a substrate of 5 mm  5 mm  0.5 mm is approxi-
mately 1015 mg, whereas the substrate mass is 50 mg.
Approximately 200 ng of FM contaminant, equivalent to a speck
of radius of 20 mm, would be sufcient to produce the magnetic
moment observed for a typical FM thin-lm specimen [139]. A
common feature is that all samples tend to exhibit phase
segregation at high dopant concentrations and rather high
annealing temperatures (>700 8C in general). These secondary
phases are often likely to be responsible for the FM observed,
which are key points for the study of DMOs. A key requirement in
elucidating RFTM in TM-doped ZnO is to judge whether the
magnetism originates from substitutional dopants on Zn sites or
from the formation of a secondary FM phase.
A number of FM or ferrimagnetic Co-based phases have been
detected. For example, FM Mn2xZnxO3d has been proposed as the
origin of high-temperature FM in (Zn,Mn)O [31]. Both metals and
their oxides have been identied as being responsible for FM
behavior in TM-doped ZnO lms, such as Co [18,224,225], Mn
[122], CoFe [70], CoO [56,163], Co3O4 [56,109,111,163], Fe3O4 [43],
and CuO [65]. These types of TM-doped ZnO lms are not genuine
DMOs. However, these non-magnetic semiconductors with
embedded secondary FM phases may exhibit an anomalous Hall
21
effect (AHE) [226], which was previously considered an indicator of
whether a doped sample was a DMO or not. AHE was observed in
FM ZnO and in non-FM Cu-doped ZnO lms, indicating that AHE
does not uniquely demonstrate FM behavior [227]. Therefore,
checking for basic experimental reproducibility, in addition to
carrying out careful high-sensitivity characterizations, is highly
warranted.
4.2.2. Carrier exchange interactions
Since RTFM arising from carrier exchange interactions has been
predicted for several TM-doped ZnO systems [9,10,228], a great
deal of attention has been focused on doped ZnO systems as
potential DMOs to provide efcient injection of spin-polarized
carriers for spintronics devices. Meta-analysis reveals that early
publications (20012004) generally favored carrier exchange
interactions [11,81,48], which are characterized by strong coupling
between localized d electrons of TM ions and the extended sp
carriers of ZnO. The term carrier-mediated exchange refers to
interactions between localized magnetic moments that are
mediated by free carriers in the system, which can be divided
into three cases: the RudermannKittelKasuyaYosida (RKKY)
interaction; Zener carrier-mediated exchange; and Zener double
exchange. RKKY interaction formally describes the magnetic
exchange between a single localized magnetic moment and a free
electron gas. Zener carrier-mediated exchange proposes that
carriers can mediate FM interaction between local moments in
the lms and both local magnetic moments and itinerant carriers
[9]. FM ordering is driven by a decrease in carrier energy due to
redistribution between spin sub-bands split by the exchange
interaction. Finally, the Zener double-exchange model refers to
indirect coupling mediated by oxygen atoms between neighboring
FM ions in different states. The kinetic energy of the system
decreases if the magnetic moments align in parallel, since parallel
alignment allows electron transfer from ions in the low state to
those in the higher state [19].
In TM-doped ZnO, the addition of Al as a dopant, or a decrease in
PO2 during lm growth, seems to increase the carrier concentration
and enhance FM [46,206]. This suggests the presence of carrier-
mediated exchange. It was found that two distinct mechanisms
could give rise to FM in TM-doped ZnO prepared by PLD: magnetic
polarons and carrier-mediated exchange, giving rise to DMI and
DMO behavior, respectively [152]. Recently, some groups reported
that additional doping cannot signicantly enhance FM ordering
[109,112,125,129,131,138]. For ZnO, it is known that annealing in a
reducing atmosphere results in the creation of VO, with the
formation of one charge-compensating electron for every VO, thus
leading to an increase in conductivity. That is, an increase in
carriers generally accompanies the production of defects, leading
to misguided and shortsighted conclusion of a carrier-mediated
mechanism. By separating these two factors, several reports have
determined that FM enhancement is most likely due a decrease in
crystalline quality and an increase in structural defects with a trace
amount of additional doping (e.g., Al and Sn) [129,130,138].
4.2.3. Defect-based BMP
Note that conducting or semiconducting TM-doped ZnO lms
with a perfect single-crystal structure show negligible FM, which is
considered to represent the absence of structural defects
[27,77,79,153,158,184]. There is a dramatic increase in resistivity
for Co:ZnO lms on Cu doping, but this change in resistivity does
not affect the magnetic moment, suggesting that a carrier-
mediated mechanism is not a feasible explanation for FM in these
lms [137,139]. Cu-doped ZnO lms showed RTFM with 1.45 mB/
Cu, whereas FM completely vanished when additional carriers
were introduced into the lms [132]. Coey et al. [8] observed RTFM
22 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
in insulating TM:ZnO and elucidated a BMP mechanism. Further-
more, an increasing number of experimental results demonstrate
that carriers are not necessary for RTFM, and FM interaction can be
unambiguously observed in lms without free carriers but in a
dielectric state [8,30,41,44,45,75,94,108,109,152,153,186]. This
insulating nature is signicant, in that a magnetic coupling
interaction other than carrier-mediated exchange is apparently
operative.
In fact, many specically designed experiments appearing with
increasing frequency reveal that defects such as VO and Zni account
for RTFM in TM:ZnO [44,58,94,167]. For example, Co:ZnO lms
annealed in Ar/H2 with a pronounced Zn K-edge XANES pre-edge
peak showed an increase in VO (1  1021 cm3) that was responsible
for much stronger RTFM [55]. Annealing of Zn0.92Co0.08O lms in
vacuum and air also demonstrated that VO and Zni are responsible for
the RTFM of the lms [141]. More interestingly, quantitative
reversible cycling of magnetic moments was achieved by annealing
TM:ZnO lms in alternating air and Zn vapor (vacuum), demon-
strating a direct correlation between FM and Zni (VO) [102,154,198].
By annealing lms in different oxidation atmospheres and in the
presence or absence of Zn vapor, it has been demonstrated that
RTFM is not linked to conductivity and that a small level of Zni
defects and consequent BMP plays a crucial role in mediating RTFM
in TM:ZnO. Similarly, post-annealing leads to a decrease in the c
lattice parameter by affecting the Zni concentration, corresponding
to a decrease in RTFM [57].
Using XAS and XMCD, Matsumura et al. [197] obtained direct
evidence that the absence of anomalous dispersion of the atomic
scattering factor for Co is probably caused by signicant lattice
deformation of the local structure around Co due to VO, accounting
for RTFM in Co:ZnO, whereas Co ions indeed substitute in Zn sites
in paramagnetic Co:ZnO thin lms [197]. This observation also
demonstrates that VO is necessary for the appearance of RTFM.
Elimination of defects, as well as lling up VO, would degrade the
FM ordering [29,141,200]. For example, Cr:ZnO lms fabricated
under various conditions revealed that elimination of structural
defects and lling up of VO might lead to a certain level of FM
degradation [29]. Likewise, annealing of Cu-doped ZnO lms in
oxygen led to a decrease in VO and had a large detrimental effect on
FM [177]. The large amount of edge dislocations observed would
enhance the concentration of VO and Zni point defects, which
contribute to FM ordering [58,94]. Therefore, these results imply
Fig. 24. (a) Lattice structure of TM:ZnO. (b) Three-polaron subsystem representing supercoupling between polarons.
that structural defects have a signicant inuence on magnetism in
these systems.
Under conditions in which acceptor dopants are activated,
leading to a decrease in free-electron density, magnetization is
enhanced. The result is consistent with hole-mediated FM in Mn-
doped ZnO, in which bound acceptors mediate the FM ordering.
Increasing the electron density decreases the acceptor concentra-
tion, thus quenching the FM exchange [131,138]. In addition, ZnO
and similar TM oxides show paramagnetism at RT, whereas it has
been claimed that RTFM can be obtained in certain d0 TM oxides
without a dopant [229231]. It should be pointed out that pure
HfO2 powder develops a weak magnetic moment on heating in
vacuum, which is eliminated on annealing in oxygen [229]. Lattice
defects are proposed to be responsible for the FM properties in
these semiconducting and insulating lms without doping. It is
worth pointing out that the absence of free spin-polarized carriers
would restrict the application of these materials in RT spintronics
[139,154,232].
Fig. 24a shows the lattice structure of TM:ZnO, in which TM ions
are incorporated into the wurtzite lattice at Zn2+ ion sites. The
highly non-equilibrium process of lm deposition makes it
possible that structural defects in the doped lm are located
throughout the lattice at arbitrary distances with respect to Co
sites, which can play an important role in the system, as illustrated
in Fig. 24b. The formation of one charge-compensating electron for
every VO produces a polaron. An electron associated with a
particular defect is conned in a hydrogenic orbital of radius
rH = er(m/m*)a0 and g = er(m/m*), where er is the dielectric constant,
m the electron mass, m* the effective mass of the donor electrons,
and a0 is the Bohr radius (0.53 A). For Co:ZnO, m/m* = 3.57 and er is
in the range 437 [8,44,107,141], which is dependent on the
fabrication method and parameters. For example, er = 10 corre-
sponds to large rH and g, i.e., rH = 18.9 A and g = 35.7. Likewise, the
polaron size is estimated to be approximately 7.6 and 7.8 A in Co-
doped ZnO [8,100]. In contrast, the disappearance of RTFM after
hydrogen implantation can be ascribed to the enhanced effective
mass and subsequent decrease in g, which also supports the above
equation [38]. It should be noted that if the donor orbital extends
over a sufcient number of cations, there is nothing to be gained by
any further increase in er [8].
Interestingly, an electron trapped in a defect is demonstrated by
CV measurement, which exhibits hysteresis with a large memory
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135 23

window of 0.6 V [94], effectively providing evidence of a BMP

mechanism: the charge-compensating electron in each defect
forms a polaron. The formula for Co-doped ZnO can be written as
(Zn0.96Co0.04)(O&d), where & represents a donor (defect) and dp is
the defect percolation threshold [8]. For (4 at.%) Co:ZnO lms with
er = 22 and me/m  0.3, dp can be calculated from g3dp = 4.3 [8].
Using the value of dp calculated, the critical concentration of
oxygen vacancies (n&; defects) can be obtained from the relation
d = n&/no, where no is the oxygen density of no = 1.4  1022 for a
close-packed oxygen lattice in ZnO. In the present case, the defect
percolation threshold n& is thus approximately 1.6  1017 [108].
FM ordering is due to superexchange between complexes in DMI,
i.e., VO and magnetic impurities, which are stabilized by charge
transfer from vacancies to impurities. The percolation threshold for
magnetic ordering is determined by the radius of vacancy levels,
but the exchange mechanism does not require free carriers [228].
If there is a sufciently large orbital radius, overlap between a
hydrogenic electron and the cations within its orbit leads to FM Fig. 25. Theoretical Curie temperature of DMOs [9].
supercoupling between them, as shown in Fig. 24b. Interaction
between the hydrogenic electron and the cations is represented by
a Heisenberg exchange Hamiltonian [233]: increasing bandgap (Eg). It should be noted that high TC values
attributed to secondary phases are excluded from the table, such as
X
Hi j J i j Si s j ; (3) TC = 980 K for Mn-doped ZnO and 1180 K for Co:TiO2 due to the
ij formation of Co metal [226] and Mn2xZnxO3d [31], respectively.
In contrast to the confusion in the literature regarding magnetic
where S is the spin of Co2+ and s is the donor electron spin. The moments of DMOs, according to theoretical studies based on the
donors tend to form BMPs, coupling Co2+ within their orbits. The BMP mechanism [8,236], the TC of DMOs is commonly quantita-
Hamiltonian of a two-polaron subsystem is given by Eq. (3), where tively given by [8]:
the donor electron spin index j takes only two values, j1 and j2,
1=2 3
corresponding to the two polarons under consideration. The T C S 1s2 xd=3 Ji j f O rceff =r O =kB ; (4)
hydrogenic orbital tends to spread out sufciently to overlap with
a large number of BMPs, but some isolated polarons cannot be where S is the localized core spin, s the electron donor spin, x the
covered and thus cannot achieve macro-FM ordering. It is therefore doping concentration, d the donor concentration, Jij the sd exchange
concluded that the origin of FM in DMOs does not necessarily parameter (bandgap Eg), fO the oxygen packing fraction for the oxide,
involve free charge carriers and that RKKY-type indirect exchange rceff the effective cation radius, and kB is Boltzmanns constant.
interactions are not necessary for magnetic ordering. In contrast, Theoretical predictions according to Eq. (4) reveal that Eg and the
TM ions form deep levels in the bandgap and a certain doping and donor concentrations are important factors for TC of a
concentration of defects in the lms may play a more active role given system. For example, TC can be estimated from the following
in mediating FM exchange than in traditional RKKY systems, as parameters for Co:ZnO: S = 3/2, s = 1/2, x = 0.04, Jij = 3.6 eV [44],
increasingly supported by related publications. d = 0.01, rceff 0:20 nm 0:20 nm, and r0 = 0.14 nm [8]. The TC given

4.3. Effect of bandgap on the Curie temperature Table 4

Curie temperatures (TC) reported for DMOs

It is generally reported that TM-doped ZnO lms show FM Band gap (eV) System TC (K) Reference
ordering at or above RT (Table 1); these materials have been InAs 0.36 Mn:InAs 30 [237]
considered typical DMOs with high TC, and should be promising for
Ge 0.75 Mn:Ge 116 [238]
use in spintronics. In fact, using a mean eld approach, calculations
Mn:Ge 150 [6]
predicted TC for oxides and semiconductors, assuming hole-
mediated FM interactions between Mn local moments. TC was then GaAs 1.55 Mn:GaAs 150 [1]
Mn:GaAs 173 [2]
determined from competition between this long-range carrier-
mediated FM coupling and short-range TMTM anti-FM exchange ZnTe 2.38 Cr:ZnTe 300 [239]
interactions. The major breakthrough in such calculations is GaP 2.87 (Mn,C):GaP 300 [240]

magnetic ordering temperatures in excess of RT for Mn-doped GaN TiO2 3.0 (rutile) Fe/Co/Ni:TiO2 700 [234]
and ZnO (Fig. 25, plotted according to Ref. [9]). Given that TC is one 3.2 (anatase) Cr:TiO2 700 [190]
of the most important parameters for DMOs, many researchers ZnO 3.37 Mn:ZnO 425 [33]
have tried to enhance it and to elucidate the factors that inuence Co:ZnO 790 [44]
it. This has led to TC enhancement from much lower than RT [158] Cu:ZnO 390 [65]
to 790 K [44] for TM:ZnO and to 700 K for TM-doped TiO2 SnO2 3.4 Co: SnO2 650 [172]
[190,234], and to 110 K [3] to 173 K [235] for Mn:GaAs [236]. The Fe:SnO2 610 [41]
highest TC values for typical DMOs with intrinsic magnetization are CeO2 3.4 Co: CeO2 740875 [241]
listed in Table 4, including Mn:InAs [237], Mn:Ge [238,6], Mn:GaAs GaN 3.5 Mn:GaN 940 [242]
[1,2], Cr:ZnTe [239], (Mn,C):GaP [240], Fe/Co/Ni:TiO2 [234], In2O3 3.75 CrIn2O3 850930 [243]
Cr:TiO2 [190], Mn:ZnO [33], Co:ZnO [44], Cu:ZnO [65], Co: SnO2 LiNbO3 3.8 CoLiNbO3 710 [120]
HfO2 4.5 Co:HfO2 800 [244]
[172], Fe:SnO2 [41], Co:CeO2 [241], Mn:GaN [242], Cr:In2O3 [243],
AlN 6.2 Cr:AlN 900 [245]
Co:LiNbO3 [120], Co:HfO2 [244], Cr:AlN [245], in order of
24 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 26. Curie temperature reported for DMOs as a function of bandgap. The correlation between them is described by a simple linear t: TC = 226.3Eg  107.4.
by Eq. (4) can then be estimated as 822 K, which is close to the
experimental result of 790 K for (4 at.%) Co:ZnO [44]. TC values
reported for DMOs as a function of semiconductor Eg are
summarized in Fig. 26. Interestingly, the reported data reveal that
TC is proportional to the semiconductor Eg. This correlation can be
described by a simple linear t: TC = 226.3Eg  107.4, demonstrating
that Eg has a signicant inuence on the TC of TM-doped ZnO. This
relation provides an insight into the TC values measured for DMOs
lms and opens the possibility of increasing TC by modifying the Eg of
DMOs, especially for Mn-doped GaAs, which has potential applica-
tions in spintronics owing to its strong spinorbit coupling and high
spin polarization [2].
5. TM-doped ZnO-based prototype spintronics and multi-
functional materials
Spintronics is a topic that has attracted much interest in recent
years. Within the context of spintronics, the spin of electrons, and
not just their electrical charge, is controlled during the operation of
information circuits. The transport and manipulation of spin-
polarized electrons or holes in semiconductors offer huge potential
for novel devices that combine non-volatile information storage
with high processing speed at low power, which may even be
useful for quantum computation [7,246]. It was the discovery of
giant magnetoresistance (GMR) and tunnel magnetoresistance
(TMR) that initiated spintronics research and guided a rst-
generation device in the form of spin valves based on FM
multilayers, such as Co/Ru [247], which are now used in the read
heads of most hard-disk storage devices. Second-generation
spintronics integrating semiconductor materials and magnetic
elements involve new exible devices such as spin-FET and spin
logic devices, which not only improve the capabilities of electronic
transistors, but also have new functionalities [248]. Successful
realization of most spintronics applications depends critically on
the ability to create spin-polarized charge carriers in a conven-
tional semiconductor in a device structure [22]. To successfully
incorporate spin into existing semiconductor technology, technical
issues such as efcient injection, transport, control and manipula-
tion, and detection of spin polarization and spin-polarized currents
have to be solved [7].
A major issue that needs to be resolved is the transfer of spin-
polarized carriers from a magnetic contact into a non-magnetic
semiconductor. Highly efcient electrical spin injection was
realized for the rst time in 1999 [249]. Spin injection can be
accomplished under ambient conditions via optical pumping with
appropriately polarized laser light [249,250]. This method is well
known and has been used to achieve the drift transport of optically
induced spin-polarized carriers in semiconductors [251]. However,
ultimate device integration will require electrical spin injection,
which can be accomplished either by injection from a spin-
polarized source or by spin-ltering unpolarized carriers at the
interface [22,246]. Several experiments have successfully demon-
strated spin injection into semiconductors and additional concepts
for spin lters and spin aligners have been proposed [252]. It is
envisioned that the merging of electronics, photonics and
magnetics will ultimately result in new spin-based multifunctional
devices, e.g., MTJs, spin-FETs, spin light-emitting diodes (spin-
LEDs), spin resonant tunneling devices, optical switches operating
at terahertz frequency, and modulators and quantum bits for
quantum computation and communication [7]. For example, MTJs
and spin-FETs fabricated by DMO-based transparent heterostruc-
tures are not only effective models for studying spin transport and
spin injection, but are also promising in information processing
[237]. Spin-polarized electrons are injected via an n-contact to a p-
contact in a magnetic semiconductor-based LED with a TM-doped
ZnO n-contact and a conventional non-magnetic p-contact. ZnO is
expected to be one of the most promising materials for RT
polarized light emission, but optical spin polarization in ZnO has
not been detected to date owing to the short spin relaxation time,
which likely results from the Rashba effect. Possible solutions
involve either cubic-phase ZnO or the use of additional stressor
layers to create greater spin splitting to obtain polarized light
emission from these structures, as well as the consideration of
alternative semiconductors and fresh device approaches
[159,253,254].
Despite persistent efforts by many groups, spin injection from a
conventional FM metal into a semiconductor has proved highly
inefcient, and is limited to <0.1% [246,255]. Although promising
advances in spin-injection efciency from metal contacts into
semiconductors have been recently been achieved (using injection
from ohmic contacts or ballistic point contacts, tunnel injection or
hot electron injection), there are generally difculties in reprodu-
cibility [256]. Some experiments also yielded high spin-injection
efciencies; however, other experiments revealed no unambig-
uous results, and for many of the concepts proposed even a proof of
principle is still lacking [91]. In contrast, efcient spin injection has
recently been successfully demonstrated in all semiconductor
tunnel diode structures using either a spin-polarized DMO as the
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
injector or a paramagnetic semiconductor under high magnetic
eld as a spin lter [257259]. According to this concept, making
carriers ow from DMOs to a conventional semiconductor, such as
from TM:ZnO to ZnO, would be an effective way to realize high spin
injection efciency because of their minimum crystal structure
mismatch, good chemical compatibility and similar electric
properties.
5.1. Co-doped ZnO-based magnetic tunnel junctions
The main trigger for the search for spin injection was the
discovery of GMR in 1988 by Grunberg [260] and Fert [261], who
were awarded the 2007 Nobel Prize for Physics. In the simplest
case, a very thin, non-magnetic metal layer is sandwiched between
two FM metal layers. Later, CoFe/Al2O3/Co and NiFe junctions were
found to show TMR of 11.8% at RT [262]. These publications
prompted a focus on the study of GMR and TMR [247,263,264].
TM:ZnO and similar DMOs with RTFM have been prepared to
provide materials in which the spin freedom of electrons can be
manipulated for spin transport electronics [16]. Steady progress is
being made on this front, but few reports have been published on
spin-polarized transport in heterostructures made from doped
ZnO [88,89,265]. In fact, spin injection from a doping ZnO layer to
pure ZnO and other conventional semiconductors is a key long-
term goal of the extensive study of ZnO DMOs [87,88].
5.1.1. Spin-polarized transport in (Zn,Co)O/ZnO/(Zn,Co)O junctions
Large TMR at low temperature (e.g., 4 K) has been obtained in
(Ga,Mn)As-based MTJs [266269]. However, the operation of these
junctions is reduced to a rather low temperature (<30 K) owing to
the low TC (170 K) of (Ga,Mn)As [1,2,235]. It is important to
develop MTJs of high-TC DMOs (such as TM-doped ZnO) to enhance
the operation temperature, and eventually to achieve RT device
Fig. 27. (a) Schematic representation of a (Zn,Co)O/ZnO/(Zn,Co)O junction. (b) SEM image of an MTJ array. (c) AFM image of junctions for a scan area of 3 mm  3 mm. (d) XRD
spectrum of the MTJ. The inset displays a pole gure for the wurtzite(1 0 1) plane.
25
paradigms. In fact, besides TM:ZnO, Ti0.95Co0.05O and (Zn,Cr)Te
with TC above RT have been used as one of the magnetic electrodes
in MTJs, possessing TMR up to 180 and 250 K, respectively, but are
limited by inelastic tunneling conduction due to the low quality of
the amorphous AlOx barrier and/or the junction interface [270
272]. The TMR temperature and spin polarization could be
enhanced by optimization of the interface and barrier quality in
MTJs [264,270272], in analogy to single-crystal Fe/MgO/Fe MTJs
with large TMR of 180% at 300 K [273].
Based on the concepts described above, TM:ZnO-based
heterostructures have been developed to study spin injection
and spin-polarized transport [88,89,274]. In particular, (Zn,Co)O
was selected for fabrication of two FM electrodes, with a pure
insulating ZnO layer as the barrier, which is compatible with
(Zn,Co)O and allows epitaxial growth [88,89]. (Zn,Co)O/ZnO/
(Zn,Co)O epitaxial structures satisfy two primary requirements
for spin injection: a minimum conductance mismatch and high
spin polarized contacts [265]. In this scenario, TMR is expected to
persist up to RT and exhibit high V1/2 to overcome one of the main
shortcomings of MTJs for technical applications: a dramatic
decrease in TMR with bias. Moreover, this MTJ structure indeed
provides an approach for spin injection from a (Zn,Co)O electrode
to ZnO, detected by measuring SP tunneling into the other (Zn,Co)O
electrode.
Fig. 27a shows a cross-sectional view of (Zn,Co)O (50 nm)/ZnO
(4 nm)/(Zn,Co)O (25 nm) junctions. The two (Zn,Co)O layers serve
as the bottom (50 nm) and top (25 nm) magnetic electrodes
(r = 1.7  102 V cm), which are separated by an insulating ZnO
barrier with a lm thickness of 4 nm (107 V cm). For vertical
transport measurement, the ZnO barrier and the Zn0.94Co0.06O top
electrode were patterned using shadow mask into a circular
junction with a diameter of 300 mm (Fig. 27b). Fig. 27c shows a
typical atomic force microscopy image of the junction surface. The
26 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135

Fig. 28. (a) Low-magnication image of MTJ cross-section. The dashed line denotes the (Zn,Co)O/Al2O3 interface and areas b and c correspond to the images in (b) and (c),
respectively. (b) HRTEM image of (Zn,Co)O/ZnO/(Zn,Co)O heterostructure. (c) Cross-sectional HRTEM image of (Zn,Co)O/Al2O3 interface. (d) SAED pattern of the junction (F,
lm; S, substrate).

root-mean-square of the surface roughness is approximately (>1 T) does not reect any specic correlation with magnetization
0.9 nm for a scan area of 3 mm  3 mm, indicating that the data. This result is supported by work carried out by Ramachan-
junction has a smooth surface. The sharp and intense wurtzite dran et al. [265], who reported that the positive TMR apparently
ZnO(0 0 2) peak observed in Fig. 27d indicates that the trilayered saturates and starts to decrease at intermediate magnetic elds,
lms are highly crystalline. This is further conrmed by a pole and that the value of this magnetic eld decreases with decreasing
gure of the lms acquired from the (1 0 1) planes, which was temperature. The positive TMR can be attributed to spin splitting of
obtained by azimuthally rotating the sample from 08 to 3608 at tilt the conduction band and a subsequent redistribution of conduc-
angles from 208 to 908. Six poles are evident, in agreement with the tion electrons when an sd exchange interaction is present [265].
sixfold symmetry of hexagonal ZnO. The low-magnication image The diffusion of Co dopes the ZnO barrier near the interface and
in Fig. 28a reveals the uniform thickness of the junction of 80 nm. likely leads to alignment of paramagnetic Co states and a decrease
The HRTEM image in Fig. 28b shows the good crystallinity of the in spin-ip scattering; subsequent suppression of spin-ip
heterostructure acquired for area b in Fig. 28a. It was found that
both the (Zn,Co)O electrodes and the ZnO barrier behave as single-
crystal structures with well-dened interfaces. A typical cross-
sectional image of the (Zn,Co)O/Al2O3 interface is shown in Fig. 28c
for area c in Fig. 28a. The SAD pattern in Fig. 28d indicates that
the orientation relationship is wurtzite(0 0 2)jjsapphire(0 0 3). It is
thus concluded that epitaxial growth of (Zn,Co)O-based MTJs
occurs from the bottom (Zn,Co)O electrode via the ZnO barrier to
the top (Zn,Co)O electrode, although some defects exist.
The spin injection and spin transport characteristics are
manifested directly in the magnetotransport properties. Fig. 29
shows the TMR dependence on the magnetic eld at four typical
temperatures. The TMR ratio is dened as TMR% = [R(H)  R(0)]/
R(0), where R(H) and R(0) are the eld-dependent resistance and
zero-eld resistance, respectively. A common feature is that all
curves show positive TMR with symmetrical behavior and a high
saturation eld >2 T. On the other hand, the TMR curves do not
exhibit a buttery shape; instead, the TMR monotonously
increases with increasing magnetic eld. It is noted that Co:ZnO
junctions have a coercivity eld of 11 mT and saturation of
0.2 T. Although the magnetization is accompanied by an increase Fig. 29. TMR curves of junctions measured at a current of 1 mA at 4, 50, 100, 200, and
in parabolic resistance, it is clear that the high TMR saturation 300 K.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
scattering at or near the interface has been identied as a possible
source of high-eld TMR [88].
A possible explanation for the intrinsic positive TMR can be
found by invoking spin orientations in (Zn,Co)O. In the low eld,
the entire density of states near the Fermi energy (EF) arise from
the minority spin band for (Zn,Co)O [275]; the spin (S) is parallel in
the bottom and top electrodes, because both electrodes are
fabricated from the same (Zn,Co)O, as marked in Fig. 28. In this
parallel case, the electron can easily tunnel from one electrode to
the other, giving rise to low resistance in the junction. As the eld is
enhanced, domain rotation becomes prevalent, accompanied by
rotation of S. It is noted that the domains, as well as S, in magnetic
materials are rotated sequentially rather than simultaneously,
particularly for the bottom and top (Zn,Co)O electrodes of different
lm thickness in our scenario. Consequently, S in the two
electrodes would not be totally parallel when applying H, leading
to difcult transport and producing high resistivity. Moreover,
these descriptions concerning sequential rotation of S most likely
correspond to a monotonic increase rather than buttery TMR
curves.
Fig. 30. (a) MTJ IV characteristics under applied elds of H = 0 and 1 T at 4 K. Typical bias dependence TMR curves at (b) 4, (c) 100, (d) 200, and (e) 300 K.
27
The magnitude of a TMR ratio is commonly dened using
Jullieres formula, TMR = 2P1P2/(1  P1P2), where P1 and P2 are the
spin polarization of FM1 and FM2, respectively [276]. Thus, SP was
calculated as 31% at 4 K. This is apparently lower than the value for
(Ga,Mn)As-based MTJs [266], presumably because Co:ZnO is a
controllable n-type DMO with very low spinorbit coupling. This is
different to GaMnAs, which is a p-type DMO with very strong spin
orbit coupling [41]. Interestingly, the SP is estimated to be 4% at RT,
revealing an effective but very small spin polarization in fully
epitaxial (Zn,Co)O/ZnO/(Zn,Co)O junctions at RT. Clearly, higher SP
and a better understanding of the physical phenomena require
further efforts.
5.1.2. Anomalous TMR in (Zn,Co)O-based junctions
Fig. 30a shows non-linear currentvoltage (IV) characteristics
for MTJs up to 2 V under H = 0 and 1 T at 4 K, which is expected for a
tunneling device. This is indicative of carrier injection by tunneling,
taken the linear IV curve for a single (Zn,Co)O lm into account. On
the basis of IV curves obtained at H = 1 and 0 T, TMR versus bias
(TMR%V) at different temperatures is plotted in Fig. 30be, where
28 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
TMR% = [Rjun(1 T)  Rjun(0 T)]/Rjun(0 T). Although both the TMR ratio
and the signal-to-noise ratio drastically decrease with increasing
temperature, the parabolic TMR%V curve remains at 300 K, revealing
that the effects of bias on transport in the junction can persist up to
RT. The most outstanding property of all the TMR%V curves from 4 to
300 K is their weak bias dependence. V1/2, the voltage at which the
TMR ratio is half of the maximum, exceeds 2 V over a wide
temperature range (4300 K). These values are much larger than
for conventional magnetic metal and for DMO-based MTJs with AlOx,
MgO and similar barriers [270273]. The decrease in TMR with bias
can be explained by inelastic scattering by magnon excitations at the
FM/insulator interface [264]. The much lower bias dependence is
likely due to the coherence of the (Zn,Co)O/ZnO interface owing to
both the minimum conductance mismatch and the high spin-
polarized contacts.
The typical V1/2 value is approximately 0.2 V [262,277]. V1/2
enhancement has generally been achieved by fabrication of double
barriers [278] and by improving control of the single-crystal
barrier [273,279], yielding up to 1.44 V in fully epitaxial MgO
double-barrier MTJs [280]. As discussed above, all-epitaxial
(Zn,Co)O/ZnO/(Zn,Co)O junctions show a very large V1/2 far in
excess of 2 V [88]. The double-barrier junctions grown by coherent
sputtering on Al2O3(0 0 1) substrate are composed of two FM
(Zn,Co)O electrodes (50 and 25 nm for bottom and top, respec-
tively) separated by a ZnO (4 nm)/(Zn,Co)O (6 nm)/ZnO (4 nm)
trilayer [89]. On the basis of IV curves obtained at H = 2 and 0 T
(Fig. 31a), TMR%V at different temperature is plotted in Fig. 31.
Fig. 31. (a) MTJ currentbias voltage (IV) characteristics up to 5 V under applied elds of H = 0 and 2 T at 3 K. Typical bias voltage dependence of TMR curves at (b) 20, (c) 10, (d)
6, (e) 5, (f) 4, (g) 3 and (h) 2 K.
Four interesting phenomena are observed at different tempera-
tures: a very large V1/2 in excess of 4 V at T > 6 K; nearly constant
TMR up to 2 V at T = 5 K; a combination of higher (at low voltage)
and lower (at high voltage) TMR with increasing voltage at T = 3
4 K; and an anomalous increase in TMR with increasing voltage at
T = 2 K. The anomalous voltage-dependent TMR can be explained
by the large energy separation between the Fermi level and the
mobility edge [89]. A non-metal ! metal transition caused by the
magnetic eld can bring the Fermi level above the mobility edge, so
that the energy separation between the Fermi level and the
mobility edge is large enough to cause the V1/2 behavior observed
[281].
5.2. TM-doped ZnO-based spin eld-effect transistors
The most effective measurement of the quality of the DMO-
based FM materials is the operation of device structures such as the
MTJs described above, spin-FETs and photo-induced ferromagnets
[16,22,88,282]. The use of DMOs as the spin source in spintronics
could allow high spin injection efciency and long spin coherence.
The spin-FET proposed by Datta and Das [91] and subsequent
numerous clones have motivated extensive study of spin preci-
sion-controlled electronic semiconductor devices [237,282284].
Huang et al. [285] claimed that a silicon spin transport device
showed output current modulation through voltage control of spin
precision, with 37% electron current spin polarization after
transport through 10-mm undoped silicon, but no spin-FET based
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 32. Schematic diagram of an all ZnO-based spin-FET [14].
on FM semiconductors has been experimentally demonstrated yet.
The primary obstacle to experimental achievement is the inability
to achieve high spin injection efciency at the interface between
the source and channel and high spin detection efciency at the
interface between the channel and drain, which are key for spin-
FETs. It is known that the spin injection and detection efciencies
have to be of the order of 99.9995% to achieve an onoff
conductance ratio of 105, which is a typical value for modern
transistors [283]. However, the maximum spin injection efciency
demonstrated is only 90% at low temperature [249], which is far
below the requirement for spin-FETs.
Because any scattering at the interface between the ferro-
magnet and the semiconductor can cause spin relaxation, a
minimum conductance mismatch and high chemical compatibility
between these layers are necessary. Fortunately, ZnO-based DMO
heterostructures can be used to fabricate spin-FETs with potential
to show high spin injection efciency and spin transport coherence
[22,93]. Following the idea of the fully epitaxial growth of
(Zn,Co)O-based MTJs and the concept of ZnO-based spin-FETs by
Pearton et al. [22], a schematic diagram of a spin-FET on the basis of
ZnO DMOs is presented in Fig. 32. In this scheme, (Zn,Mn)O can be
a half-metallic ferromagnet, which serves as the channel, and
(Zn,Co)O is used to fabricate the source [30 nm (Zn,Co)O injector]
and drain [60 nm (Zn,Co)O detector], as well as a thin insulating
ZnO layer acting as a gate oxide. Using FM (Zn,Co)O as the source
Fig. 33. (a) Frequency dependence of the PE characteristics of Zn0.975V0.025O lms measured at room temperature. (b) PE characteristics of Ag/Zn0.94Cr0.06O/n-Si FET
measured at 2 kHz and room temperature. Piezoelectric hysteresis loop (black) and DV curve (red) for (c) Zn0.975V0.025O and (d) Zn0.94Cr0.06O lms. (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of the article.)
29
and drain contact material, it should be possible to have a 100%
spin-polarized electron ow in the (Zn,Mn)O channel [22]. These
heterostructures could be grown by magnetron sputtering to
realize fully epitaxial growth and shadow mask patterning. By
manipulating the relative orientation of the magnetization of the
injector and detector with a different in-plane external magnetic
eld and gate voltage, the current in the detector can be changed
correspondingly, in analogy to the in-plane spin valve [93,285].
P AP AP
The magnetocurrent ratio is MC% ISD  ISD =ISD , where ISD is the
magnetocurrent and the superscripts P and AP refer to parallel and
antiparallel injector/detector magnetization conguration, respec-
tively. Subsequently, this magnetocurrent ratio corresponds to a
P
conduction electron current spin polarization of P ISD 
AP P AP
ISD =ISD ISD with respect to Jullieres formula [276]. The
fabrication of TM-doped ZnO-based spin-FETs and the realization
of reliable spin detection are highly warranted and should be the
focus of intense efforts.
5.3. Ferroelectricity and giant piezoelectric d33 in V and Cr-doped ZnO
Although FM in TM-doped ZnO systems has been discussed in
detail, ferroelectricity, which usually exists in perovskites and
ilmenites, has also been found in these systems. Recently, Yang
et al. [286,287] reported the observation of ferroelectric behavior
in V- and Cr-doped ZnO lms. In their experiments, ferroelectric
measurements were performed on Ag/Zn1xTMxO/n-Si (TM = V,Cr)
sandwich structures in which Ag was deposited on lms as the top
electrodes and a conductive Si substrate served as the bottom
electrode. Fig. 33a and b show polarizationelectric eld (PE)
hysteresis loops for V- and Cr-doped ZnO lms, respectively.
Although the characteristics are not fully saturated, the hysteretic
behavior still indicates the presence of ferroelectricity. The shift of
the loop center observed for these Ag/Zn1xTMxO/n-Si structures is
considered to result from the asymmetric electrode structure. The
ferroelectric behavior in V- and Cr-doped ZnO lms was further
conrmed by well-shaped displacement-applied eld (DE)
buttery loops, as shown in Fig. 33c and d, respectively
30 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 34. (a) Room-temperature magnetic hysteresis of Zn0.985V0.015O lms. The inset
shows the same parameters on an expanded scale. (b) Temperature dependence of
(&) FC and (~) ZFC magnetizations for Zn0.985V0.015O lms. The inset shows the
temperature dependence of the difference between the FC and ZFC data.
[287,288]. It should be noted that ferroelectricity in ZnO-based
systems has also been reported for Li- and Mg-doped ZnO bulk
and/or lms [289291]. Piezoelectric hysteresis loops (d33-applied
eld) could be calculated from DE curves and clearly showed that
the Zn1xTMxO lms are switchable and the ferroelectricity is
retained. More importantly, high piezoelectric d33 coefcients
were observed for ferroelectric V- and Cr-doped ZnO lms.
The observation of ferroelectricity in V- and Cr-doped ZnO is
particularly intriguing in light of recent reports of FM in these
systems, suggesting the possibility of engineering a multiferroic
material with simultaneous FM and ferroelectricity. Fig. 34a
displays a typical magnetic hysteresis loop (M versus H) for a
Zn0.985V0.015O lm measured at 300 K. The sample exhibits weak
FM and saturation magnetization of 0.13 mB/V. A coercive eld of
8 mT can be observed on the expanded scale of the hysteresis
loop. Fig. 34b shows the temperature dependence of the eld-
cooled (FC) and zero-eld-cooled (ZFC) magnetization for the
Zn0.985V0.015O lm. The data were taken in a magnetic eld of 1500
Oe applied parallel to the lm surface. The inset shows the
magnetization difference between FC and ZFC data as a function of
temperature, i.e., DM(T) = MFC(T)  MZFC(T). The DMT curve
exhibits a decreasing trend overall, but the DM value is still
reasonably positive near RT. Other mechanisms such as spin glass
effects and superparamagnetic clusters could also present hyster-
esis, but they would exhibit a value of DM = 0 below 100 K [292].
Therefore, the hysteresis arises from intrinsic FM of insulating V-
doped ZnO, which is a special multiferroic. Multiferroics that show
Fig. 35. Effective d33 as a function of V concentration x in Zn1xVxO lms, bulk ZnO
and ZnO nanobelts.
simultaneous FM and ferroelectric ordering have attracted
extensive interest since they have potential applications in
important devices such as transducers and memory [292,293].
The piezoelectric properties of ZnO lms doped with different V
concentrations have been systematically investigated by SPM
characterization by Yang et al. [288] and Zhao et al. [294]. Fig. 35
shows the piezoelectric d33 coefcients reported for bulk ZnO and
quasi-one-dimensional nanobelts and the average d33 coefcient
as a function of vanadium concentration in Zn1xVxO lms. The d33
coefcient for bulk ZnO and a nanobelt is 9.9 and 26.7 pC/N,
respectively; the latter was the highest d33 value for ZnO-based
systems before ferroelectric TM-doped ZnO systems were found.
The d33 value measured for an undoped ZnO lm was 11.8 pC/N.
However, when 2.5 at.% V was introduced, a peak d33 value of
110 pC/N was obtained, approximately one order of magnitude
greater than the value for pure ZnO lm and fourfold greater than
for the ZnO nanobelt. A comparably high d33 value of 120 pC/N was
obtained in ferroelectric Zn0.94Cr0.06O lms. The intriguing piezo-
electric properties mean that V- and Cr-doped ZnO are promising
candidates for sensors, actuators, and transducers.
The origin of ferroelectric behavior in ZnO-based systems has
been the subject of much discussion, with two main interpreta-
tions: the so-called size-mismatch model and the electronic origin
model. It is commonly believed that the probable reason for the
ferroelectric behavior is that smaller substitutes can occupy off-
center positions and form permanent electric dipoles, which
results in spontaneous polarization. Four ZnO-based systems have
been found to exhibit ferroelectricity so far: Li:ZnO, Mg:ZnO,
V:ZnO, and Cr:ZnO. The ionic radius of Li+, Mg2+, V5+, and Cr3+ is 0.6,
0.65, 0.54, and 0.63 A, respectively, which are all smaller than that
of Zn2+ (0.74 A) [286,287,289,295]. Some researchers have
suggested that a change in electronic conguration could also
be a possible reason for this novel ferroelectricity [286,295]. For
instance, the replacement of Zn2+ (1s2 2s2 2p6 3s2 3p6 3d10) by V5+
(1s2 2s2 2p6 3s2 3p6) with no 3d electrons likely affects the 3d2p
hybridization between Zn and O in V-doped ZnO systems, which
could change the Zn(V)O bond length and induce local dipole
moments. However, both models are still assumptions and
convincing evidence is required for conrmation.
XANES provides advantageous local and chemical information
and is sensitive to subtle differences in local arrangement, which
can yield a better understanding of ferroelectric behavior. There-
fore, XANES was recently used to characterize ferroelectric V-
doped ZnO systems [286]. Qualitative analysis of XANES V K-edge
spectra revealed that V substitutes replacing Zn are in the 5+ state.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
Fig. 36. Experimental and MS calculated V K-edge XANES spectra for V:ZnO lms.
To verify the assumption that V5+ ions are off-center, V K-edge
XANES spectra were calculated via the full MS ab initio method. It
was found that three main features (peaks A, B and C) match the
experimental curve to a certain extent by simply replacing the
central Zn by V in a designed V-doped ZnO cluster (Fig. 36).
However, the relative height of peaks B and C differ, and there is a
much lower pre-edge peak compared to the experimental
spectrum. Constriction of local VO4 units (C) and displacement
of V5+ ions from the cell center (D) were then introduced as
parameters in the calculation. Introduction of VO4 constriction
improved the t to the experimental curve. When V5+ displace-
ment was also considered, the closest t was obtained. This
accurate replication of the characteristics of the experimental
spectrum implies that V5+ displacement likely occurs and is
responsible for ferroelectricity in the V:ZnO system. Thus the
XANES calculations provided a direct evidence for the size-
mismatch model and claried the ferroelectric origin in V:ZnO
system.
For ferroelectrics with a centrosymmetric paraelectric phase,
d33 can be expressed as follows:
d33 2Q eff e0 er Ps ; (5)
where Qeff is the effective electrostriction coefcient, and e0 and er
are the permittivity of free space and relative permittivity,
respectively. Eq. (5) indicates that the piezoresponse in ferro-
electrics is directly dependent on Ps via a factor linking the
electrostrictive characteristics and the dielectric constants. On the
other hand, this type of dependence also exists in ferroelectrics
with a non-centrosymmetric paraelectric phase [288]. Therefore,
the giant d33 in V- and Cr-doped ZnO was attributed to the
emergence of switchable Ps. Moreover, the relatively high er in V-
and Cr-doped ZnO lms induced by switchable Ps may also
contribute to the high d33. A microscopic explanation was also
31
proposed for a V-doped ZnO system [288]. Since the dominant
effect of the electric eld in wurtzite semiconductors is to rotate
bonds that are non-collinear with the polar c-axis, i.e., Zn2O1
bonds [287], towards the direction of the applied eld, thus
producing piezoelectric strain, the electromechanical response in
wurtzite materials should be mainly governed by the ease of bond
bending [296]. In V-doped ZnO lms, since V5+ ions have a higher
positive charge than Zn2+ ions, the non-collinear VO1 bonds ought
to have a stronger polarity than Zn2O1 bonds and hence can rotate
more easily in an applied eld. On the other hand, (2 at.%) Cu-, Ni-,
Co- and Fe-doped ZnO lms have much smaller d33 values of 13.5,
9.8, 10.7 and 5.8 pC/N, respectively, which may be primarily
because the radii of Cu2+ (0.72 A) and Ni2+ (0.72 A) are almost equal
to that of Zn2+ (0.74 A), whereas those of Co2+ (0.79 A) and Fe2+
(0.76 A) are greater than that of Zn2+ [297,298], different from the
scenario for Li+, Mg2+, V5+, and Cr3+ ions.
V- and Cr-doped ZnO show simultaneous FM and ferroelec-
tricity, and are thus promising multiferroic materials. TM-doped
ZnO is also considered as a promising optical material that could be
extensively used for transparent electrodes/devices and the
window of solar cells. Some optoelectronic applications of ZnO
overlap with those of GaN, which is now widely used for LEDs.
However, ZnO has some advantages over GaN: the large exciton
binding energy (60 meV) of ZnO could lead to lasing action based
on exciton recombination, even above RT, which is widely used for
production of green, blue-UV, and white LEDs, as discussed in
numerous reviews and reports [21,22,299,300,301]. It is worth
pointing out that ZnO has much simpler crystal-growth technol-
ogy, resulting in potential lower costs for ZnO-based devices. The
transmittance is usually greater than 80% for TM-doped ZnO in the
visible region [48,109,129], indicating that these materials are
suitable for transparent spintronics applications with integrated
TM-doped ZnO layers [88,89,302,303]. Thus, ZnO could be used as
materials with multi-functions, including FM, ferroelectric, piezo-
electric, optical, and electric features, which could play important
roles in spintronics applications of TM-doped ZnO.
6. Conclusions and outlook
In conclusion, we have presented a comprehensive review of
the preparation parameters that affect the local TM structure and
magnetization of TM-doped ZnO-based spintronics prototypes.
ZnO lms have been the subject of varying degrees of research
effort over recent decades. This effort has recently been intensied
for the preparation and elucidation of TM-doped ZnO DMOs with
RTFM and the development of spin-polarized injection, transport
and detection experiments on TM-doped ZnO-based heterostruc-
tures for spintronics applications. Although it is clear that TM-
doped ZnO systems exhibit RTFM, there are still many open
questions regarding its origin and use in spintronics. By extracting
a single variable from a series of experiments, carriers and defects
can be separated and direct relationships between FM and the
presence of structural defects can be identied. It has been
unambiguously demonstrated that structural defects are respon-
sible for RTFM in TM-doped ZnO, whereas carriers involved in
carrier-mediated exchange are natural by-products of the creation
of defects in ZnO. In contrast, an increase in electron density
decreases the acceptor concentration, thus quenching FM
exchange. We have described recent signicant advances in TM-
doped ZnO-based prototype spintronics, in which TM-doped ZnO-
based MTJ structures provide a compass for the spin injection from
(Zn,Co)O to ZnO at RT and shows a large half voltage (>4 V). In
addition, second generation spintronics integrating TM-doped ZnO
lms with promising ferromagnetic, ferroelectric, piezoelectric
and optical properties tend to fabricate new exible devices with
32 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
the ability to create and manipulate spin-polarized carriers in a
conventional semiconductor.
Although FM ordering arising from defects in TM-doped ZnO
have been determined, defects are difcult to characterize
experimentally and measure quantitatively, which is important
to profoundly understand the origin of magnetic behaviors in this
system. Considerably designing experiments to accurately control
the formation of defects and quantitatively build the correlations
between defects and magnetic moments are highly warranted. The
best way to investigate the innumerable possible defects in TM-
doped ZnO may be with the help of a computer. More importantly,
spin-polarized transport in TM-doped ZnO and the efciency of
spin injection from TM-doped ZnO to conventional semiconduc-
tors would draw extensive attention, and only in this way, DMOs
can move to the real applications in useful spintronics. Spin
polarized injection, transport and detection experiments in TM-
doped ZnO-based heterostructures are lacking at present, the spin
state at the interfaces of heterostructures would require a more
precise control and careful quantum analysis. In this case, efforts
are increasingly shifting not only to the design of TM-doped ZnO-
based prototype spintronics, but also to the task of thinking calmly
methods to evaluate these devices.
Acknowledgements
The authors are grateful to Prof. B.X. Liu of Tsinghua University
for helpful suggestions on this paper, and to Prof. G.A. Gehring of
the University of Shefeld for fruitful discussions. The authors also
acknowledge the cooperation of X.X. Wei in providing some data,
and of Dr. W.S. Yan and B. He of the National Synchrotron Radiation
Laboratory in local structure measurements. This work was
supported by the National Natural Science Foundation of China
(Grant nos. 50772055 and 50325105), the Ministry of Science and
Technology of China (Grant no. 2006CB605201-1) and a National
Hi-tech (R&D) Project of China (Grant no. 2006AA03Z435).
References
[1] A.H. Macdonald, P. Schiffer, N. Samarth, Nat. Mater. 4 (2005) 195.
[2] T. Jungwirth, J. Sinova, J. Masek, J. Kucera, A.H. MacDonld, Rev. Mod. Phys. 78
(2006) 809.
[3] H. Ohno, A. Shen, F. Matsukura, A. Oiwa, A. Endo, S. Katsumoto, Y. Iye, Appl. Phys.
Lett. 69 (1996) 363.
[4] H. Munekata, H. Ohno, S. von Molnar, A. Segmuller, L.L. Chang, L. Esaki, Phys. Rev.
Lett. 63 (1989) 1849.
[5] S. von Molnar, D. Read, J. Magn. Magn. Mater. 242245 (2002) 13.
[6] Y.D. Park, A.T. Hanbicki, S.C. Erwin, C.S. Hellberg, J.M. Sullivan, J.E. Mattson, T.F.
Ambrose, A. Wilson, G. Spanos, B.T. Jonker, Science 295 (2002) 651.
[7] S.A. Wolf, D.D. Awschalom, R.A. Buhrman, J.M. Daughton, S. von Molnar, M.L.
Roukes, A.Y. Chtchelkanova, D.M. Treger, Science 294 (2001) 1488.
[8] J.M.D. Coey, M. Venkatesan, C.B. Fitzgerald, Nat. Mater. 4 (2005) 173.
[9] T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand, Science 287 (2000) 1019.
[10] K. Sato, H. Katayama-Yoshida, Jpn. J. Appl. Phys. 39 (2000) L555.
[11] K. Ueda, H. Tabata, T. Kawai, Appl. Phys. Lett. 79 (2001) 988.
[12] Y. Matsumoto, M. Murakami, T. Shono, T. Hasegawa, T. Fukumura, M. Kawasaki,
P. Ahmet, T. Chikyow, S. Koshihara, H. Koinuma, Science 291 (2001) 854.
[13] I. Zutic, J. Fabian, S. Das Sarma, Rev. Mod. Phys. 76 (2004) 323.
[14] S.J. Pearton, C.R. Abernathy, D.P. Norton, A.F. Hebard, Y.D. Park, L.A. Boatner, J.D.
Budai, Mater. Sci. Eng. R. 40 (2003) 137.
[15] D.D. Awschalom, M.E. Flatte, Nat. Phys. 3 (2007) 153.
[16] S.J. Pearton, C.R. Abernathy, M.E. Overberg, G.T. Thaler, D.P. Norton, N. Theodor-
opoulou, A.F. Hebard, Y.D. Park, F. Ren, J. Kim, L.A. Boatner, J. Appl. Phys. 93
(2003) 1.
[17] W. Prellier, A. Fouchet, B. Mercey, J. Phys.: Condens. Matter. 15 (2003) R1583.
[18] S.J. Pearton, W.H. Heo, M. Ivill, D.P. Norton, T. Steiner, Semicond. Sci. Technol. 19
(2004) R59.
[19] R. Janisch, P. Gopal, N.A. Spaldin, J. Phys.: Condens. Matter 17 (2005) R657.
[20] S.A. Chambers, Surf. Sci. Rep. 61 (2006) 345.
[21] U. Ozgur, Ya.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, V. Avrutin, S.-J.
Cho, H. Morkoc, J. Appl. Phys. 98 (2005) 041301.
[22] S.J. Pearton, D.P. Norton, K. Ip, Y.W. Heo, T. Steiner, Prog. Mater. Sci. 50 (2005)
293.
[23] M. Venkatesan, C.B. Fitzgerald, J.G. Lunney, J.M.D. Coey, Phys. Rev. Lett. 93 (2004)
177206.
[24] J.M.D. Coey, M. Venkatesan, C.B. Fitzgerald, L.S. Dorneles, P. Stamenov, J.G.
Lunney, J. Magn. Magn. Mater. 290/291 (2005) 1405.
[25] L.S. Dorneles, D. OMahony, C.B. Fitzgerald, F. McGee, M. Venkatesan, I. Stanca,
J.G. Lunney, J.M.D. Coey, Appl. Surf. Sci. 248 (2005) 406.
[26] H. Saeki, H. Tabata, T. Kawai, Solid Solid Commun. 120 (2001) 439.
[27] S. Ramachandran, A. Tiwari, J. Narayan, J.T. Prater, Appl. Phys. Lett. 87 (2005)
172502.
[28] N.H. Hong, J. Sakai, A. Hassini, J. Appl. Phys. 97 (2005) 10D312.
[29] N.H. Hong, J. Sakai, N.T. Huong, N. Poirot, A. Ruyter, Phys. Rev. B 72 (2005)
045336.
[30] H. Liu, X. Zhang, L. Li, Y.X. Wang, K.H. Gao, Z.Q. Li, R.K. Zheng, S.P. Ringer, B. Zhang,
X.X. Zhang, Appl. Phys. Lett. 91 (2007) 072511.
[31] D.C. Kundaliya, S.B. Ogale, S.E. Loand, S. Dhar, C.J. Metting, S.R. Shinde, Z. Ma, B.
Varughese, K.V. Ramanujachary, L. Salamanca-riba, T. Venkatesan, Nat. Mater. 3
(2004) 709.
[32] X.M. Cheng, C.L. Chien, J. Appl. Phys. 93 (2003) 7876.
[33] P. Sharma, A. Gupta, K.V. Rao, Frank J. Owens, R. Sharma, R. Ahuja, J.M. Osorio
Guillen, B. Johansson, G.A. Gehring, Nat. Mater. 2 (2003) 673.
[34] W. Yan, Z. Sun, Q. Liu, Z. Li, Z. Pan, J. Wang, S. Wei, D. Wang, Y. Zhou, X. Zhang,
Appl. Phys. Lett. 91 (2007) 062113.
[35] D.P. Norton, S.J. Pearton, A.F. Hebard, N. Theodoropoulou, L.A. Boatner, R.G.
Wilson, Appl. Phys. Lett. 82 (2003) 239.
[36] M.A. Garca, M.L. Ruiz-Gonzalez, A. Quesada, J.L. Costa-Kramer, J.F. Fernandez, S.J.
Khatib, A. Wennberg, A.C. Caballero, M.S. Martn-Gonzalez, M. Villegas, F.
Briones, J.M. Gonzalez-Calbet, A. Hernando, Phys. Rev. Lett. 94 (2005) 217206.
[37] P. Thakur, K.H. Chae, J.-Y. Kim, M. Subramanian, R. Jayavel, K. Asokan, Appl. Phys.
Lett. 91 (2007) 162503.
[38] S.Y. Park, P.J. Kim, Y.P. Lee, S.W. Shin, T.H. Kim, J.H. Kang, Adv. Mater. 19 (2007)
3496.
[39] J.-H. Guo, A. Gupta, P. Sharma, K.V. Rao, M.A. Marcus, C.L. Dong, J.M.O. Guillen,
S.M. Butorin, M. Mattesini, P.A. Glans, K.E. Smith, C.L. Chang, R. Ahuja, J. Phys.:
Condens. Matter 19 (2007) 172202.
[40] N. Theodoropoulou, V. Misra, J. Philip, P. LeClair, G.P. Berera, J.S. Moodera, B.
Satpati, T. Som, J. Magn. Magn. Mater. 300 (2006) 407.
[41] J.M.D. Coey, A.P. Douvalis, C.B. Fitzgerald, M. Venkatesan, Appl. Phys. Lett. 84
(2004) 1332.
[42] S. Venkataraj, N. Ohashi, I. Sakaguchi, Y. Adachi, T. Ohgaki, H. Ryoken, H. Haneda,
J. Appl. Phys. 102 (2007) 014905.
[43] X.X. Wei, C. Song, K.W. Geng, F. Zeng, B. He, F. Pan, J. Phys.: Condens. Matter 18
(2006) 7471.
[44] C. Song, K.W. Geng, F. Zeng, X.B. Wang, Y.X. Shen, F. Pan, Y.N. Xie, T. Liu, H.T. Zhou,
Z. Fan, Phys. Rev. B 73 (2006) 024405.
[45] A. Dinia, G. Schmerber, V. Pierron-Bohnes, C. Meny, P. Panissod, E. Beaurepaire, J.
Magn. Magn. Mater. 286 (2005) 37.
[46] J.H. Kim, H. Kim, D. Kim, Y.E. Ihm, W.K. Choo, J. Appl. Phys. 92 (2002) 6066.
[47] K. Ando, H. Saito, Z. Jin, T. Fukumura, M. Kawasaki, Y. Matsumoto, H. Koinuma,
Appl. Phys. Lett. 78 (2001) 2700.
[48] S. Ramachandran, A. Tiwari, J. Narayan, Appl. Phys. Lett. 84 (2004) 5255.
[49] Q. Xu, H. Schmidt, L. Hartmann, H. Hochmuth, M. Lorenz, A. Setzer, P. Esquinazi,
C. Meinecke, M. Grundmann, Appl. Phys. Lett. 91 (2007) 092503.
[50] Z. Jin, T. Fukumura, M. Kawasaki, K. Ando, H. Saito, T. Sekiguchi, Y.Z. Yoo, M.
Murakami, Y. Matsumoto, T. Hasegawa, H. Koinuma, Appl. Phys. Lett. 78 (2001)
3824.
[51] K. Nielsen, S. Bauer, M. Lubbe, S.T.B. Goennenwein, M. Opel, J. Simon, W. Mader,
R. Gross, Phys. Stat. Sol. A 203 (2006) 3581.
[52] A.C. Tuan, J.D. Bryan, A.B. Pakhomov, V. Shutthanandan, S. Thevuthasan, D.E.
McCready, D. Gaspar, M.H. Engelhard, J.W. Rogers Jr., K. Krishnan, D.R. Gamelin,
S.A. Chambers, Phys. Rev. B 70 (2004) 054424.
[53] J.S. Thakur, G.W. Auner, V.M. Naik, C. Sudakar, P. Kharel, G. Lawes, R. Suryanar-
ayanan, R. Naik, J. Appl. Phys. 102 (2007) 093904.
[54] A.B. Pakhomov, B.K. Roberts, K.M. Krishnan, Appl. Phys. Lett. 83 (2003) 4357.
[55] H.S. Hsu, J.C.A. Huang, Y.H. Huang, Y.F. Liao, M.Z. Lin, C.H. Lee, J.F. Lee, S.F. Chen,
L.Y. Lai, C.P. Liu, Appl. Phys. Lett. 88 (2006) 242507.
[56] H.J. Lee, S.Y. Jeong, C.R. Cho, C.H. Park, Appl. Phys. Lett. 81 (2002) 4020.
[57] J.L. MacManus-Driscoll, N. Khare, Y. Liu, M.E. Vickers, Adv. Mat. 19 (2007)
2925.
[58] Y.B. Zhang, Q. Liu, T. Sritharan, C.L. Gan, S. Li, Appl. Phys. Lett. 89 (2006) 042510.
[59] W. Prellier, A. Fouchet, B. Mercey, Ch. Simon, B. Raveau, Appl. Phys. Lett. 82
(2003) 3490.
[60] J.R. Neal, A.J. Behan, R.M. Ibrahim, H.J. Blythe, M. Ziese, A.M. Fox, G.A. Gehring,
Phys. Rev. Lett. 96 (2006) 197608.
[61] Q. Liu, C.L. Gan, C.L. Yuan, G.C. Han, Appl. Phys. Lett. 92 (2008) 032501.
[62] X. Liu, F. Lin, L. Sun, W. Cheng, X. Ma, W. Shi, Appl. Phys. Lett. 88 (2006) 062508.
[63] T. Wakano, N. Fujimura, Y. Morinaga, N. Abe, A. Ashida, T. Ito, Physica E 10 (2001)
260.
[64] D.B. Buchholz, R.P.H. Chang, J.H. Song, J.B. Ketterson, Appl. Phys. Lett. 87 (2005)
082504.
[65] C. Sudakar, J.S. Thakur, G. Lawes, R. Naik, V.M. Naik, Phys. Rev. B 75 (2007)
054423.
[66] D.L. Hou, X.J. Ye, H.J. Meng, H.J. Zhou, X.L. Li, C.M. Zhen, G.D. Tang, Appl. Phys.
Lett. 90 (2007) 142502.
[67] T.S. Herng, S.P. Lau, S.F. Yu, H.Y. Yang, X.H. Ji, J.S. Chen, N. Yasui, H. Inaba, J. Appl.
Phys. 99 (2006) 086101.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
[68] D.L. Hou, X.J. Ye, X.Y. Zhao, H.J. Meng, H.J. Zhou, X.L. Li, C.M. Chen, J. Appl. Phys.
102 (2007) 033905.
[69] J.C.A. Huang, H.S. Hsu, Appl. Phys. Lett. 87 (2005) 132503.
[70] Y.M. Cho, W.K. Choo, H. Kim, D. Kim, Y.E. Ihm, Appl. Phys. Lett. 80 (2002) 3358.
[71] C.L. Du, Z.B. Gu, M.H. Lu, J. Wang, S.T. Zhang, J. Zhao, G.X. Cheng, H. Heng, Y.F.
Chen, J. Appl. Phys. 99 (2006) 123515.
[72] Z.-B. Gu, C.-S. Yuan, M.-H. Lu, J. Wang, D. Wu, S.T. Zhang, S.-N. Zhu, Y.-Y. Zhu, Y.-F.
Chen, J. Appl. Phys. 98 (2005) 053908.
[73] S. Lee, Y. Shon, S.-W. Lee, S.J. Hwang, H.S. Lee, T.W. Kang, D.Y. Kim, Appl. Phys.
Lett. 88 (2006) 212513.
[74] S.J. Lee, H.S. Lee, D.Y. Kim, T.W. Kim, J. Cryst. Growth 276 (2005) 121.
[75] H. Schmidt, M. Diaconu, H. Hochmuth, M. Lorenz, A. Setzer, P. Esquinazi, A. Poppl,
D. Spemann, K.-W. Nielsen, R. Gross, G. Wagner, M. Grundmann, Superlattices
Microstruct. 39 (2006) 334.
[76] J.H. Park, M.G. Kim, H.M. Jang, S. Ryu, Y.M. Kim, Appl. Phys. Lett. 84 (2004) 1338.
[77] T.C. Kaspar, T. Droubay, Y. Li, S.M. Heald, P. Nachimuthu, C.M. Wang, V. Shut-
thanandan, C.A. Johnson, D.R. Gamelin, S.A. Chambers, New J. Phys., submitted
for publication.
[78] D.J. Keavney, D.B. Buchholz, Q. Ma, R.P.H. Chang, Appl. Phys. Lett. 91 (2007)
012501.
[79] T. Fukumura, Z. Jin, M. Kawasaki, T. Shono, T. Hasegawa, S. Koshihara, H.
Koinuma, Appl. Phys. Lett. 78 (2001) 958.
[80] K.P. Bhatti, S. Kundu, S. Chaudhary, S.C. Kashyap, D.K. Pandya, J. Phys. D: Appl.
Phys. 39 (2006) 4909.
[81] T. Fukumura, Y. Yamada, H. Toyosaki, T. Hasegawa, H. Koinuma, M. Kawasaki,
Appl. Surf. Sci. 223 (2004) 62.
[82] T. Fukumura, H. Toyosaki, Y. Yamada, Semicond. Sci. Technol. 20 (2005) S103.
[83] T. Dietl, T. Andrearczyk, A. Lipinska, M. Kiecana, M. Tay, Y. Wu, Phys. Rev. B 76
(2007) 155312.
[84] N.A. Spaldin, Phys. Rev. B 69 (2004) 125201.
[85] K. Ando, Science 312 (2006) 1883.
[86] Y. Belghazi, G. Schmerber, S. Colis, J.L. Rehspringer, A. Dinia, A. Berrada, Appl.
Phys. Lett. 89 (2006) 122504.
[87] R. Mattana, J.M. George, H. Jaffres, F. Nguyen Van Dau, A. Fert, B. Lepine, A.
Guivarch, G. Jezequel, Phys. Rev. Lett. 90 (2003) 166601.
[88] C. Song, X.J. Liu, F. Zeng, F. Pan, Appl. Phys. Lett. 91 (2007) 042106.
[89] C. Song, Y.C. Yang, X.W. Li, X.J. Liu, F. Zeng, F. Pan, Appl. Phys. Lett. 91 (2007)
172109.
[90] E.Y. Tsymbal, O.N. Mryasov, P.R. LeClair, J. Phys.: Condens. Matter 15 (2003)
R109.
[91] S. Datta, B. Das, Appl. Phys. Lett. 56 (1990) 665.
[92] S.H. Noh, W. Choi, M.S. Oh, D.K. Huang, K. Lee, S. Im, S. Jang, E. Kim, Appl. Phys.
Lett. 90 (2007) 253504.
[93] K.C. Hall, M.E. Flatte, Appl. Phys. Lett. 88 (2006) 162503.
[94] C. Song, F. Zeng, K.W. Geng, X.J. Liu, F. Pan, B. He, W.S. Yan, Phys. Rev. B 76 (2007)
045215.
[95] X.H. Xu, H.J. Blythe, M. Ziese, A.J. Behan, J.R. Neal, A. Mokhtari, R.M. Ibrahim, A.M.
Fox, G.A. Gehring, New J. Phys. 8 (2006) 135.
[96] S. Lee, T.W. Kang, D.Y. Kim, J. Crystal Growth 284 (2005) 6.
[97] E. Biegger, M. Fonin, U. Rudiger, N. Janen, M. Beyer, T. Thomay, R. Bratschitsch,
Y.S. Dedkov, J. Appl. Phys. 101 (2007) 073904.
[98] S. Lee, H.S. Lee, S.J. Hwang, Y. Shon, D.Y. Kim, E.K. Kim, J. Crystal Growth 286
(2006) 223.
[99] A. Ben Mahmoud, H.J. von Bardeleben, J.L. Cantin, A. Mauger, E. Chikoidze, Y.
Dumont, Phys. Rev. B 74 (2006) 115203.
[100] C. Sudakar, P. Kharel, G. Lawes, R. Suryanarayanan, R. Naik, V.M. Naik, J. Phys.:
Condens. Matter 19 (2007) 026212.
[101] Y.B. Lin, J.P. Xu, W.Q. Zou, L.Y. Lv, Z.H. Lu, F.M. Zhang, Y.W. Du, Z.G. Huang, J.G.
Zheng, J. Phys. D: Appl. Phys. 40 (2007) 3674.
[102] T.S. Herng, S.P. Lau, S.F. Yu, J.S. Chen, K.S. Teng, J. Magn. Magn. Mater. 315 (2004)
107.
[103] S.G. Yang, A.B. Pakhomov, S.T. Hung, C.Y. Wong, IEEE Trans. Magn. 38 (2002)
2877.
[104] H.K. Yadav, K. Sreenivas, V. Gupta, J. Appl. Phys. 99 (2006) 083507.
[105] S.K. Mandal, A.K. Das, T.K. Nath, D. Karmakar, Appl. Phys. Lett. 89 (2006) 144105.
[106] T. Mizokawa, T. Nambu, A. Fujimori, T. Fukumura, M. Kawasaki, Phys. Rev. B 65
(2002) 085209.
[107] X.J. Liu, C. Song, P.Y. Yang, F. Zeng, F. Pan, Appl. Surf. Sci. 254 (2008) 3167.
[108] C. Song, F. Zeng, K.W. Geng, X.B. Wang, Y.X. Shen, F. Pan, J. Magn. Magn. Mater.
309 (2007) 25.
[109] C. Song, S.N. Pan, X.J. Liu, X.W. Li, F. Zeng, W.S. Yan, B. He, F. Pan, J. Phys.: Condens.
Matter 19 (2007) 176229.
[110] M. Tay, Y. Wu, G.C. Han, T.C. Chong, Y.K. Zheng, S.J. Wang, Y. Chen, X. Pan, J. Appl.
Phys. 100 (2006) 063910.
[111] X.-L. Li, Z.-L. Wang, X.-F. Qin, H.-S. Wu, X.-H. Xu, G.A. Gehring, J. Appl. Phys. 103
(2008) 023911.
[112] C. Song, X.J. Liu, K.W. Geng, F. Zeng, F. Pan, B. He, S.Q. Wei, J. Appl. Phys. 101
(2007) 103903.
[113] J.J. Chen, Y. Gao, F. Zeng, D.M. Li, F. Pan, Appl. Surf. Sci. 223 (2004) 318.
[114] J.J. Chen, F. Zeng, D.M. Li, J.B. Niu, F. Pan, Thin Solid Films 485 (2005) 257.
[115] J.C.A. Huang, H.S. Hsu, Y.M. Hu, C.H. Lee, Y.H. Huang, M.Z. Lin, Appl. Phys. Lett. 85
(2004) 3815.
[116] J. Okabayashi, K. Ono, M. Mizuguchi, M. Oshima, D.D. Subhra Sen Gupta, T.
Sarma, A. Mizokawa, M. Fujimori, C.T. Yuri, T. Chen, M. Fukumura, H. Kawasaki,
Koinuma, J. Appl. Phys. 95 (2004) 3573.
33
[117] S. Krishnamurthy, C. McGuinness, L.S. Dorneles, M. Venkatesan, J.M.D. Coey, J.G.
Lunney, C.H. Pattersonb, K.E. Smith, T. Learmonth, P.-A. Glans, T. Schmitt, J.-H.
Guo, J. Appl. Phys. 99 (2006), 08M111.
[118] G.S. Chang, E.Z. Kurmaev, D.W. Boukhvalov, L.D. Finkelstein, S. Colis, T.M.
Pedersen, A. Moewes, A. Dinia, Phys. Rev. B 75 (2007) 195215.
[119] A.L. Ankudinov, B. Ravel, J.J. Rehr, S.D. Conradson, Phys. Rev. B 58 (1998) 7565.
[120] C. Song, F. Zeng, Y.X. Shen, K.W. Geng, Y.N. Xie, Z.Y. Wu, F. Pan, Phys. Rev. B 73
(2006) 172412.
[121] M. Kobayashi, Y. Ishida, J.l. Hwang, T. Mizokawa, A. Fujimori, K. Mamiya, J.
Okamoto, Y. Takeda, T. Okane, Y. Saitoh, Y. Muramatsu, A. Tanaka, H. Saeki, H.
Tabata, T. Kawai, Phys. Rev. B 72 (2005) 201201(R).
[122] M. Diaconu, H. Schmidt, A. Poppl, R. Bottcher, J. Hoentsch, A. Rahm, H. Hoch-
muth, M. Lorenz, M. Grundmann, Superlattices Microstruct. 38 (2005) 413.
[123] P. Sati, A. Stepanov, V. Pashchenko, Low Temp. Phys. 33 (2007) 927.
[124] G.Y. Ahn, S.-I. Park, I.-B. Shim, C.S. Kim, J. Magn. Magn. Mater. 282 (2004) 166.
[125] N.H. Hong, V. Brize, J. Sakai, Appl. Phys. Lett. 86 (2005) 082505.
[126] S.-J. Han, J.W. Song, C.-H. Yang, S.H. Park, J.-H. Park, Y.H. Jeong, K.W. Rhie, Appl.
Phys. Lett. 81 (2002) 4212.
[127] T. Fukumura, Z. Jin, A. Ohtomo, H. Koinuma, M. Kawasaki, Appl. Phys. Lett. 75
(1999) 3366.
[128] H.S. Hsu, J.C.A. Huang, S.F. Chen, C.P. Liu, Appl. Phys. Lett. 90 (2007) 102506.
[129] X.J. Liu, C. Song, F. Zeng, F. Pan, J. Phys.: Condens. Matter 19 (2007) 296208.
[130] X.-C. Liu, E.-W. Shi, Z.-Z. Chen, H.-W. Zhang, B.-Y. Chen, L.-X. Song, S.-Q. Wei, B.
He, Z. Xie, J. Crystal growth 307 (2007) 14.
[131] M. Ivill, S.J. Pearton, D.P. Norton, J. Kelly, A.F. Hebard, J. Appl. Phys. 97 (2005)
053904.
[132] A. Tiwari, M. Snure, D. Kumar, J.T. Abiade, Appl. Phys. Lett. 92 (2008) 062509.
[133] M.H.F. Sluiter, Y. Kawazoe, P. Sharma, A. Inoue, A.R. Raju, C. Rout, U.V. Wagh-
mare, Phys. Rev. Lett. 94 (2005) 187204.
[134] Y.-H. Lin, M. Ying, M. Li, X. Wang, C.-W. Nan, Appl. Phys. Lett. 90 (2007) 222110.
[135] H.Y. Xu, Y.C. Liu, C.S. Xu, Y.X. Liu, C.L. Shao, R. Mu, Appl. Phys. Lett. 88 (2006)
242502.
[136] Z.-B. Gu, M.-H. Lu, J. Wang, D. Wu, S.-T. Zhang, X.-K. Meng, Y.-Y. Zhu, S.-N. Zhu, Y.-
F. Chen, X.-Q. Pan, Appl. Phys. Lett. 88 (2006) 082111.
[137] D. Chakraborti, S. Ramachandran, G. Trichy, J. Narayan, J.T. Prater, J. Appl. Phys.
101 (2007) 053918.
[138] M. Ivill, S.J. Pearton, Y.W. Heo, J. Kelly, A.F. Hebard, D.P. Norton, J. Appl. Phys. 101
(2007) 123909.
[139] J.M.D. Coey, Curr. Opin. Solid State Mater. Sci. 10 (2006) 83.
[140] S. Ramachandran, J. Narayan, J.T. Prater, Appl. Phys. Lett. 88 (2006) 242503.
[141] X.J. Liu, C. Song, F. Zeng, X.B. Wang, F. Pan, J. Phys. D: Appl. Phys. 40 (2007) 1608.
[142] D.A. Schwartz, D.R. Gamelin, Adv. Mater. 16 (2004) 2115.
[143] W.B. Jian, Z.Y. Wu, R.T. Huang, F.R. Chen, J.J. Kai, C.Y. Wu, S.J. Chiang, M.D. Lan, J.J.
Lin, Phys. Rev. B 73 (2006) 233308.
[144] K.A. Grifn, A.B. Pakhomov, C.M. Wang, S.M. Heald, K.M. Krishnan, Phys. Rev.
Lett. 94 (2005) 157204.
[145] L. Esaki, P.J. Stiles, S. von Molnar, Phys. Rev. Lett. 19 (1967) 852.
[146] T. Nagahama, T.S. Santos, J.S. Moodera, Phys. Rev. Lett. 99 (2007) 016602.
[147] J.M. Kikkawa, D.D. Awschalom, Nature 397 (1999) 139.
[148] M. Paillard, X. Marie, P. Renucci, T. Amand, A. Jbeli, J.M. Gerard, Phys. Rev. Lett. 86
(2001) 1634.
[149] T. Zhao, S.R. Shinde, S.B. Ogale, H. Zheng, T. Venkatesan, R. Ramesh, S. Das Sarma,
Phys. Rev. Lett. 94 (2005) 126601.
[150] T. Droubay, S.M. Heald, V. Shutthanandan, S. Thevuthasan, S.A. Chambers, J. Appl.
Phys. 97 (2005) 046103.
[151] X.F. Liu, Y. Sun, R.H. Yu, J. Appl. Phys. 101 (2007) 123907.
[152] A.J. Behan, A. Mokhtari, H.J. Blythe, D. Score, X.-H. Xu, J.R. Neal, A.M. Fox, G.A.
Gehring, Phys. Rev. Lett. 100 (2008) 047206.
[153] M. Ungureanu, H. Schmidt, H. von Wenckstern, H. Hochmuth, M. Lorenz, M.
Grundmann, M. Fecioru-Morariu, G. Guntherodt, Thin Solid Films 515 (2007)
8761.
[154] N. Khare, M.J. Kappers, M. Wei, M.G. Blamire, J.L. Macmanus-Driscoll, Adv. Mater.
18 (2006) 1449.
[155] Q. Xu, L. Hartmann, H. Schmidt, H. Hochmuth, M. Lorenz, R. Sdchmidt-Grund, C.
Sturn, D. Spemann, M. Grundmann, Y. Liu, J. Appl. Phys. 101 (2007) 063908.
[156] Q. Xu, L. Hartmann, H. Schmidt, H. Hochmuth, M. Lorenz, R. Sdchmidt-Grund, C.
Sturn, D. Spemann, M. Grundmann, J. Appl. Phys. 100 (2006) 013904.
[157] Q. Xu, L. Hartmann, H. Schmidt, H. Hochmuth, M. Lorenz, R. Sdchmidt-Grund, C.
Sturn, D. Spemann, M. Grundmann, Phys. Rev. B 73 (2006) 205342.
[158] K. Ando, A. Chiba, H. Tanoue, J. Appl. Phys. 83 (1998) 6545.
[159] S.J. Pearton, D.P. Norton, M.P. Ivill, A.F. Hebard, J.M. Zavada, W.M. Chen, I.A.
Buyanova, J. Electron. Mater. 36 (2007) 462.
[160] G.A. Gehring, private communication.
[161] H. Harima, J. Phys.: Condens. Matter 16 (2004) S5653.
[162] K. Rode, A. Anane, R. Mattana, J.-P. Contour, O. Durand, R. LeBourgeois, J. Appl.
Phys. 93 (2003) 7676.
[163] Z. Yin, N. Chen, C. Chai, F. Yang, J. Appl. Phys. 96 (2004) 5093.
[164] P. Sati, R. Hayn, R. Kuzian, S. Regnier, S. Schafer, A. Stepanov, C. Morhain, C.
Deparis, M. Laugt, M. Goiran, Z. Golacki, Phys. Rev. Lett. 96 (2006) 017203.
[165] C.B. Fitzgerald, M. Venkatesan, J.G. Lunney, L.S. Dorneles, J.M.D. Coey, Appl. Surf.
Sci. 247 (2005) 493.
[166] N. Jedrecy, H.J. von Bardeleben, Y. Zheng, J.-L. Cantin, Phys. Rev. B 69 (2004)
041308.
[167] M. Gacic, G. Jakob, C. Herbort, H. Adrian, T. Tietze, S. Bruck, E. Goering, Phys. Rev.
B 75 (2007) 205206.
34 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
[168] X.-C. Liu, E.-W. Shi, Z.-Z. Chen, B.-Y. Chen, T. Zhang, L.-X. Song, K.-J. Zhou, M.-Q.
Cui, W.S. Yan, Z. Xie, B. He, S.-Q. Wei, J. Phys.: Condens. Matter. 20 (2008) 025208.
[169] X. Wang, J. Xu, B. Zhang, H. Yu, J. Wang, X. Zhang, J. Yu, Q. Li, Adv. Mater. 18
(2006) 2476.
[170] K. Samanta, P. Bhattacharya, R.S. Katiyar, W. Iwamoto, P.G. Pagliuso, C. Rettori,
Phys. Rev. B 73 (2006) 245213.
[171] Z. Yang, J.L. Liu, M. Biasini, W.P. Beyermann, Appl. Phys. Lett. 92 (2008) 042111.
[172] S.B. Ogale, R.J. Choudhary, J.P. Buban, S.E. Loand, S.R. Shinde, S.N. Kale, V.N.
Kulkarni, J. Higgins, C. Lanci, J.R. Simpson, N.D. Browning, S. Das Sarma, H.D.
Drew, R.L. Greene, T. Venkatesan, Phys. Rev. Lett. 91 (2003) 077205.
[173] N.H. Hong, J. Sakai, W. Prellier, A. Hassini, J. Phys.: Condens. Matter 17 (2005)
1697.
[174] N.H. Hong, J. Sakai, A. Hassini, Appl. Phys. Lett. 84 (2004) 2602.
[175] C. Kittel, Introduction to solid state physics, John Wiley & Sons, New York, 1996.
[176] K. Rode, R. Mattana, A. Anane, V. Cros, E. Jacquet, J.-P. Contour, F. Petroff, A. Fert,
M.-A. Arrio, Ph. Sainctavit, P. Bencok, F. Wilhelm, N.B. Brookes, A. Rogalev, Appl.
Phys. Lett. 92 (2008) 012509.
[177] D. Chakraborti, G.R. Trichy, J.T. Prater, J. Narayan, J. Phys. D: Appl. Phys. 40 (2007)
7606.
[178] J.H. Kim, H. Kim, D. Kim, Y.E. Ihm, W.K. Choo, Physica B 327 (2003) 304.
[179] S.W. Jung, S.-J. An, G. Yi, C. Jung, S. Lee, S. Cho, Appl. Phys. Lett. 80 (2002)
4561.
[180] S.W. Lim, M.C. Jeong, M.H. Ham, J.M. Myoung, Jpn. J. Appl. Phys. 43 (2004) L280.
[181] S. Zhou, K. Potzger, J. von Borany, R. Grotzschel, W. Skorupa, M. Helm, J.
Fassbender, Phys. Rev. B 77 (2008) 035209.
[182] Y. Liu, Q. Xu, H. Schmidt, L. Hartmann, H. Hochmuth, M. Lorenz, M. Grundmann,
X. Han, Z. Zhang, Appl. Phys. Lett. 90 (2007) 154101.
[183] T. Stenberg, P. Vuoristo, J. Keranen, T. Mantyla, Buchler, S. Virtanen, P. Schmuki,
H. Bohni, Thin Solid Films 312 (1998) 46.
[184] K. Ando, H. Saito, Z. Jin, T. Fukumura, M. Kawasaki, Y. Matsumoto, H. Koinuma, J.
Appl. Phys. 89 (2001) 7284.
[185] J.-Y. Kim, J.-H. Park, B.-G. Park, H.-J. Noh, S.-J. Oh, J.S. Yang, D.-H. Kim, S.D. Bu, T.-
W. Noh, H.-J. Lin, H.-H. Hsieh, C.T. Chen, Phys. Rev. Lett. 90 (2003) 017401.
[186] Y. Zheng, J.C. Boulliard, D. Demaille, Y. Bernard, J.F. Petroff, J. Crystal Growth 274
(2005) 156.
[187] W. Eerenstein, M. Wiora, J.L. Prieto, J.F. Scott, N.D. Mathur, Nat. Mater. 6 (2007)
348.
[188] E.-C. Lee, K.J. Chang, Phys. Rev. B 69 (2004) 085205.
[189] T.C. Kaspar, S.M. Heald, C.M. Wang, J.D. Bryan, T. Droubay, V. Shutthanandan, S.
Thevuthasan, D.E. McCready, A.J. Kellock, D.R. Gamelin, S.A. Chambers, Phys. Rev.
Lett. 95 (2005) 217203.
[190] T.C. Kaspar, T. Droubay, V. Shutthanandan, S.M. Heald, C.M. Wang, D.E.
McCready, S. Thevuthasan, J.D. Bryan, D.R. Gamelin, A.J. Kellock, M.F. Toney,
X. Hong, C.H. Ahn, S.A. Chambers, Phys. Rev. B 73 (2006) 155327.
[191] V. Vaithianathan, B.T. Lee, C.H. Chang, K. Asokan, S.S. Kim, Appl. Phys. Lett. 88
(2006) 112103.
[192] K.J. Choi, M. Biegalski, Y.L. Li, A. Sharan, J. Schubert, R. Uecker, P. Reiche, Y.B.
Chen, X.Q. Pan, V. Gopalan, L.-Q. Chen, D.G. Schlom, C.B. Eom, Science 306 (2004)
1005.
[193] C. Thiele, K. Dorr, O. Bilani, J. Rodel, L. Schultz, Phys. Rev. B 75 (2007) 054408.
[194] J. Wang, J.B. Neaton, H. Zheng, V. Nagarajan, S.B. Ogale, B. Liu, D. Viehland, V.
Vaithyanathan, D.G. Schlom, U.V. Waghmare, N.A. Spaldin, K.M. Rabe, M. Wuttig,
R. Ramesh, Science 299 (2003) 1719.
[195] A.M. Yakunin, A.Yu. Silov, P.M. Koenraad, J.-M. Tang, M.E. Flatte, J.-L. Primus, W.
Van Roy, J. De Boeck, A.M. Monakhov, K.S. Romanov, I.E. Panaiotti, N.S. Averkiev,
Nat. Mater. 6 (2007) 512.
[196] X.J. Liu, C. Song, F. Zeng, F. Pan, B. He, W.S. Yan, J. Appl. Phys. 103 (2008) 093911.
[197] T. Matsumura, D. Okuyama, S. Niioka, H. Ishida, T. Satoh, Y. Murakami, H.
Toyosaki, Y. Yamada, T. Fukumura, M. Kawasaki, Phys. Rev. B 76 (2007) 115320.
[198] K.R. Kittilstved, D.A. Schwartz, A.C. Tuan, S.M. Heald, S.A. Chambers, D.R.
Gamelin, Phys. Rev. Lett. 97 (2006) 037203.
[199] S.A. Chambers, T.C. Droubay, C.M. Wang, K.M. Rosso, S.M. Heald, D.A. Schwartz,
K.R. Kittilstved, D.R. Gamelin, Mater. Today 9 (2006) 28.
[200] B. Huang, D. Zhu, X. Ma, Appl. Surf. Sci. 253 (2007) 6892.
[201] N.A. Theodoropoulou, A.F. Hebard, D.P. Norton, J.D. Budai, L.A. Boatner, J.S. Lee,
Z.G. Khim, Y.D. Park, M.E. Overberg, S.J. Pearton, R.G. Wilson, Solid State Electron.
47 (2003) 2231.
[202] J. Alaria, H. Bieber, S. Colis, G. Schmerber, A. Dinia, Appl. Phys. Lett. 88 (2006)
112503.
[203] O.D. Jayakumar, I.K. Gopalakrishnan, S.K. Kulshreshtha, Adv. Mater. 18 (2006)
1857.
[204] X.J. Liu, C. Song, F. Zeng, F. Pan, Thin Solid Films, submitted for publication.
[205] W. Yan, Z. Sun, Q. Liu, Z. Li, T. Shi, F. Wang, Z. Qi, G. Zhang, S. Wei, H. Zhang, Z.
Chen, Appl. Phys. Lett. 90 (2007) 242509.
[206] X.C. Liu, E.W. Shi, Z.Z. Chen, H.W. Zhang, B. Xiao, L.X. Song, Appl. Phys. Lett. 88
(2006) 252503.
[207] A. Mokhtari, A.J. Behan, D.S. Score, H.J. Blythe, W. Davies, J.R. Neal, A.M. Fox, G.A.
Gehring, New J. Phys., submitted for publication.
[208] H.J. Lee, S.Y. Jeong, J.Y. Hwang, C.R. Cho, Europhys. Lett. 64 (2003) 797.
[209] R. Seshadri, Curr. Opin. Solid State Mater. Sci. 9 (2005) 1.
[210] Q. Wang, Q. Sun, P. Jena, Z. Hu, R. Note, Y. Kawazoe, Appl. Phys. Lett. 91 (2007)
063116.
[211] L.-H. Ye, A.J. Freeman, B. Delley, Phys. Rev. B 73 (2006) 033203.
[212] S.-J. Hu, S.-S. Yan, M.-W. Zhao, L.-M. Mei, Phys. Rev. B 73 (2006) 245205.
[213] P. Gopal, N.A. Spaldin, Phys. Rev. B 74 (2006) 094418.
[214] C.-H. Chien, S.H. Chiou, G.Y. Guo, Y.-D. Yao, J. Magn. Magn. Mater. 282 (2004)
275.
[215] S.Y. Yun, G.-B. Cha, Y. Kwon, S. Cho, S.C. Hong, J. Magn. Magn. Mater. 272276
(2004) e1536.
[216] X. Feng, J. Phys. Chem. 16 (2004) 4251.
[217] E.-J. Kan, L.-F. Yuan, J. Yang, J. Appl. Phys. 102 (2007) 033915.
[218] D. Karmakar, S.K. Mandal, R.M. Kadam, P.L. Paulose, A.K. Rajarajan, T.K. Nath, A.K.
Das, I. Dasgupta, G.P. Das, Phys. Rev. B 75 (2007) 144404.
[219] D. Iusan, B. Sanyal, O. Eriksson, J. Appl. Phys. 101 (2007) 09H101.
[220] T. Zhang, L.-X. Song, Z.-Z. Chen, E.-W. Shi, L.-X. Chao, H.-W. Zhang, Appl. Phys.
Lett. 89 (2006) 172502.
[221] A. Debernardi, M. Fanciulli, Appl. Phys. Lett. 90 (2007) 212510.
[222] C.D. Pemmaraju, T. Archer, R. Hanan, S. Sanvito, J. Magn. Magn. Mater. 316
(2007) e185.
[223] A.L. Rosa, R. Ahuja, J. Phys.: Condens. Matter 19 (2007) 386232.
[224] M. Venkatesan, P. Stamenov, L.S. Dorneles, R.D. Gunning, B. Bernoux, J.M.D. Coey,
Appl. Phys. Lett. 90 (2007) 242508.
[225] Z.W. Zhao, B.K. Tay, J.S. Chen, J.F. Hu, B.C. Lim, G.P. Li, Appl. Phys. Lett. 90 (2007)
152502.
[226] S.R. Shinde, S.B. Ogale, S. Das Sarma, J.R. Simpson, H.D. Drew, S.E. Loand, C.
Lanci, J.P. Buban, N.D. Browning, V.N. Kulkarni, J. Higgins, R.P. Sharma, R.L.
Greene, T. Venkatesan, Phys. Rev. B 67 (2003) 115211.
[227] Q. Xu, H. Schmidt, S. Zhou, K. Potzger, M. Helm, H. Hochmuth, M. Lorenz, A.
Setzer, P. Esquinazi, C. Meinecke, M. Grundmann, Appl. Phys. Lett. 92 (2008)
082508.
[228] K. Kikoin, V. Fleurov, Phys. Rev. B 74 (2006) 174407.
[229] J.M.D. Coey, M. Venkatesan, P. Stamenov, C.B. Fitzgerald, L.S. Dorneles, Phys. Rev.
B 72 (2005) 024450.
[230] N.H. Hong, J. Sakai, N. Poirot, V. Brize, Phys. Rev. B 73 (2006) 132404.
[231] N.H. Hong, J. Sakai, V. Brize, J. Phys.: Condens. Matter 19 (2007) 036219.
[232] P. Stamenov, M. Venkatesan, L.S. Dorneles, D. Maude, J.M.D. Coey, J. Appl. Phys.
99 (2006), 08M124.
[233] P.A. Cox, Transition Metal Oxides, Clarendon Press, Oxford, 1992.
[234] N.H. Hong, J. Sakai, W. Prellier, A. Ruyter, J. Phys. D: Appl. Phys. 38 (2005)
816.
[235] T. Jungwirth, K.Y. Wang, J. Masek, K.W. Edmonds, J. Konig, J. Sinova, M. Polini,
N.A. Goncharuk, A.H. MacDonald, M. Sawicki, A.W. Rushforth, R.P. Campion, L.X.
Zhao, C.T. Foxon, B.L. Gallagher, Phys. Rev. B 72 (2005) 165204.
[236] A. Kaminski, S. Das Sarma, Phys. Rev. Lett. 88 (2002) 247202.
[237] H. Ohno, D. Chiba, F. Matsukura, T. Omiya, E. Abe, T. Dietl, Y. Ohno, K. Ohtani,
Nature 408 (2000) 944.
[238] A. Verna, L. Ottaviano, M. Passacantando, S. Santucci, P. Picozzi, F. DOrazio, F.
Lucari, M. De Biase, R. Gunnella, M. Berti, A. Gasparotto, G. Impellizzeri, F. Priolo,
Phys. Rev. B 74 (2006) 085204.
[239] H. Saito, V. Zayets, S. Yamagata, K. Ando, Phys. Rev. Lett. 90 (2003) 207202.
[240] N. Theodoropoulou, A.F. Hebard, M.E. Overberg, C.R. Abernathy, S.J. Pearton,
S.N.G. Chu, R.G. Wilson, Phys. Rev. Lett. 89 (2002) 107203.
[241] A. Tiwari, V.M. Bhosle, S. Ramachandran, N. Sudhakar, J. Narayan, S. Budak, A.
Gupta, Appl. Phys. Lett. 88 (2006) 142511.
[242] T. Sasaki, S. Sonoda, Y. Yamamoto, K. Suga, S. Shimizu, K. Kindo, H. Hori, J. Appl.
Phys. 91 (2002) 7911.
[243] J. Philip, A. Punnoose, B.I. Kim, K.M. Reddy, S. Layne, J.O. Holmes, B. Satpati, P.R.
Leclair, T.S. Santos, J.S. Moodera, Nat. Mater. 5 (2006) 298.
[244] M.S. Ramachandra Rao, C. Darshan, S.B. Kundaliya, L.F. Ogaleb, S.J. Fu, N.D. Welz,
V. Browning, B. Zaitsev, C.A. Varughese, A. Cardoso, S. Curtin, S.R. Dhar, T. Shinde,
S.E. Venkatesan, S.A. Loand, Schwarz, Appl. Phys. Lett. 88 (2006) 142505.
[245] D. Kumar, J. Antifakos, M.G. Blamire, Z.H. Barber, Appl. Phys. Lett. 84 (2004)
5004.
[246] G. Schmidt, J. Phys. D: Appl. Phys. 38 (2005) R107.
[247] C. Song, X.X. Wei, K.W. Geng, F. Zeng, F. Pan, Phys. Rev. B 72 (2005) 184412.
[248] Y.B. Xu, Curr. Opin. Solid State Mater. Sci. 10 (2006) 81.
[249] M. Fiederling, M. Keim, G. Reuscher, W. Ossau, G. Schmidt, A. Waag, L.W.
Molenkamp, Nature 402 (1999) 787.
[250] I. Malajovich, J.J. Berry, N. Samarth, D.D. Awschalom, Nature 411 (2001) 770.
[251] D. Hagele, M. Oestreich, W.W. Ruhle, N. Nestle, K. Eberl, Appl. Phys. Lett. 73
(2007) 1580.
[252] Y. Ohno, D.K. Young, B. Beschoten, F. Matsukura, H. Ohno, D.D. Awschalom,
Nature 402 (1999) 190.
[253] S.J. Pearton, D.P. Norton, M.P. Ivill, A.F. Hebard, J.M. Zavada, W.M. Chen, I.A.
Buyanova, IEEE Trans. Electron Devices 54 (2007) 1040.
[254] W.M. Chen, I.A. Buyanova, A. Murayama, T. Furuta, Y. Oka, D.P. Norton, S.J.
Pearton, A. Osinsky, J.W. Dong, Appl. Phys. Lett. 92 (2008) 092103.
[255] G. Schmidt, D. Ferrand, L.W. Molenkamp, A.T. Filip, B.J. van Wees, Phys. Rev. B 62
(2000) R4790.
[256] A.T. Hanbicki, B.T. Jonker, G. Itskos, G. Kioseoglou, A. Petrou, Appl. Phys. Lett. 80
(2002) 1240.
[257] J. Hubner, W.W. Ruhle, M. Klude, D. Hommel, R.D.R. Bhat, J.E. Sipe, H.M. van Driel,
Phys. Rev. Lett. 90 (2003) 216601.
[258] I. Malajovich, J.M. Kikkawa, D.D. Awschalom, J.J. Berry, N. Samarth, Phys. Rev.
Lett. 84 (2000) 1015.
[259] G. Kioseoglou, A.T. Hanbicki, J.M. Sullivan, O.M.J. van TErve, C.H. Li, S.C. Erwin, B.
Mallory, M. Yasar, A. Petrou, B.T. Jonker, Nat. Mater. 3 (2004) 799.
[260] G. Binasch, P. Grunberg, F. Saurenbach, W. Zinn, Phys. Rev. B 39 (1988) 4828.
[261] M.N. Baibich, J.M. Broto, A. Fert, F.N. Vandau, F. Petroff, P. Etienne, G. Creuzet, A.
Friedrich, J. Chazelas, Phys. Rev. Lett. 61 (1988) 2472.
 F. Pan et al. / Materials Science and Engineering R 62 (2008) 135
[262] J.S. Moodera, L.R. Kinder, T.M. Wong, R. Meservey, Phys. Rev. Lett. 74 (1995)
3273.
[263] H.F. Ding, W. Wulfhekel, J. Henk, P. Bruno, J. Kirschner, Phys. Rev. Lett. 90 (2003)
116603.
[264] E.Y. Tsymbal, K.D. Belashchenko, J.P. Velev, S.S. Jaswal, M. Van Schilfgaarde, I.I.
Oleynik, D.A. Stewart, Prog. Mater. Sci. 52 (2007) 401.
[265] S. Ramachandran, J.T. Prater, N. Sudhakar, D. Kumaer, J. Narayan, Solid State
Commun. 145 (2008) 18.
[266] C. Gould, K. Pappert, G. Schmidt, L.W. Molenkamp, Adv. Mater. 19 (2007)
323.
[267] D. Chiba, N. Akiba, F. Matsukura, Y. Ohno, H. Ohno, Appl. Phys. Lett. 77 (2000)
1873.
[268] S.H. Chun, S.J. Potashnik, K.C. Ku, P. Schiffer, N. Samarth, Phys. Rev. B 66 (2002)
100408(R).
[269] H. Saito, S. Yuasa, K. Ando, Phys. Rev. Lett. 74 (1995) 3273.
[270] H. Toyosaki, T. Fukumura, K. Ueno, M. Nakano, M. Kawasaki, Jpn. J. Appl. Phys.,
Part 2 44 (2005) L896.
[271] H. Toyosaki, T. Fukumura, K. Ueno, M. Nakano, M. Kawasaki, J. Appl. Phys. 99
(2006) 08M102.
[272] W.G. Wang, C. Ni, T. Moriyama, J. Wan, E. Nowak, John Q. Xiao, Appl. Phys. Lett.
88 (2006) 202501.
[273] S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, K. Ando, Nat. Mater. 3 (2004)
868.
[274] Q. Xu, L. Hartmann, H. Schmidt, H. Hochmuth, M. Lorenz, R. Schmidt-Grund, D.
Spemann, A. Rahm, M. Grundmann, Thin Solid Films 515 (2006) 2549.
[275] A.S. Risbud, N.A. Spaldin, Z.Q. Chen, S. Stemmer, R. Seshadri, Phys. Rev. B 68
(2003) 205202.
[276] M. Julliere, Phys. Lett. A 54 (1975) 225.
[277] J.M. De Teresa, A. Barthelemy, A. Fert, J.P. Contour, F. Montaigne, P. Seneor,
Science 286 (1999) 507.
[278] F. Montaigne, J. Nassar, A. Vaures, F. Nguyen Van Dau, F. Petroff, A. Schuhl, A. Fert,
Appl. Phys. Lett. 73 (1998) 2829.
[279] H. Boeve, E. Girgis, J. Schelten, J. De Boeck, G. Borghs, Appl. Phys. Lett. 76 (2000)
1048.
35
[280] T. Nozaki, A. Hirohata, N. Tezuka, S. Sugimoto, K. Inomata, Appl. Phys. Lett. 86
(2005) 082501.
[281] E. Dias Cabral, M.A. Boselli, A.T. da Cunha Lima, A. Ghazali, I.C. da Cunha Lima,
Appl. Phys. Lett. 90 (2007) 142118.
[282] D. Chiba, F. Matsukura, H. Ohno, J. Phys. D: Appl. Phys. 39 (2006) R215.
[283] J. Wan, M. Cahay, S. Bandyopadhyay, J. Appl. Phys. 102 (2007) 034301.
[284] J. Nitta, T. Akazaki, H. Takayanagi, T. Enoki, Phys. Rev. Lett. 78 (1997) 1335.
[285] B. Huang, D.J. Monsma, I. Appelbaum, Appl. Phys. Lett. 91 (2007) 072501.
[286] Y.C. Yang, C. Song, F. Zeng, F. Pan, Y.N. Xie, T. Liu, Appl. Phys. Lett. 90 (2007)
242903.
[287] Y.C. Yang, C. Song, X.H. Wang, F. Zeng, F. Pan, J. Appl. Phys. 103 (2008) 074107.
[288] Y.C. Yang, C. Song, X.H. Wang, F. Zeng, F. Pan, Appl. Phys. Lett. 92 (2008) 012907.
[289] A. Onodera, N. Tamaki, Y. Kawamura, T. Sawada, H. Yamashita, Jpn. J. Appl. Phys.,
Part 1 35 (1996) 5160.
[290] M. Joseph, H. Tabata, T. Kawai, Appl. Phys. Lett. 74 (1999) 2534.
[291] S.B. Dhananjay, Krupanidhi, Appl. Phys. Lett. 89 (2006) 082905.
[292] A.F. Hebard, R.P. Rairigh, J.G. Kelly, S.J. Pearton, C.R. Abernathy, S.N.G. Chu, R.G.
Wilson, J. Phys. D: Appl. Phys. 37 (2004) 511.
[293] R. Ramesh, N.A. Spaldin, Nat. Mater. 6 (2007) 21.
[294] M.H. Zhao, Z.L. Wang, S.X. Mao, Nano. Lett. 4 (2004) 587.
[295] A. Onodera, Ferroelectrics 267 (2002) 131.
[296] D. Karanth, H.X. Fu, Phys. Rev. B 72 (2005) 064116.
[297] X.B. Wang, C. Song, D.M. Li, K.W. Geng, F. Zeng, F. Pan, Appl. Surf. Sci. 253 (2006)
1639.
[298] X.B. Wang, D.M. Li, F. Zeng, F. Pan, J. Phys. D: Appl. Phys. 38 (2005) 4104.
[299] X.B. Wang, C. Song, K.W. Geng, F. Zeng, F. Pan, Appl. Surf. Sci. 253 (2007) 6905.
[300] R.F. Service, Science 276 (1997) 895.
[301] A. Tsukazaki, A. Ohtomo, T. Kita, Y. Ohno, H. Ohno, M. Kawasaki, Science 315
(2007) 1388.
[302] Z. Jin, M. Murakami, T. Fukumura, Y. Matsumoto, A. Ohtomo, M. Kawasaki, H.
Koinuma, J. Crystal Growth 214/215 (2000) 55.
[303] A. Tsukazaki, A. Ohtomo, T. Onuma, M. Ohtani, T. Makino, M. Sumiya, K. Ohtani,
S.F. Chichibu, S. Fuke, Y. Segawa, H. Ohno, H. Koinuma, M. Kawasaki, Nat. Mater.
4 (2005) 42.