ZnO nanowire growth and devices
Y.W. Heo
a,
*
, D.P. Norton
a
, L.C. Tien
a
, Y. Kwon
a
, B.S. Kang
b
, F. Ren
b
, S.J. Pearton
a
,
J.R. LaRoche
c
a
Department of Materials Science and Engineering, University of Florida, Gainesville, FL 32611, USA
b
Department of Chemical Engineering, University of Florida, Gainesville, FL 32611, USA
c
Raytheon, Waltham, MA 02451, USA
Accepted 23 September 2004
Available online 25 December 2004
Abstract
The large surface area of ZnO nanorods makes them attractive for gas and chemical sensing, and the ability to
control their nucleation sites makes them candidates for micro-lasers or memory arrays. In addition, they might be
doped with transition metal (TM) ions to make spin-polarized light sources. To date, most of the work on ZnO

nanostructures has focused on the synthesis methods and there have been only a few reports of the electrical
characteristics. We review fabrication methods for obtaining device functionality from single ZnO nanorods. A key
aspect is the use of sonication to facilitate transfer of the nanorods from the initial substrate on which they are grown to
another substrate for device fabrication. Examples of devices fabricated using this method are briefly described,
including metal-oxide semiconductor field effect depletion-mode transistors with good saturation behavior, a
threshold voltage of $À3 V and a maximum transconductance of order 0.3 mS/mm and Pt Schottky diodes with
excellent ideality factors of 1.1 at 25 8C and very low (1.5 Â 10
À10
A, equivalent to 2.35 A cm
À2
,atÀ10 V) reverse
currents. The photoresponse showed only a minor component with long decay times (tens of seconds) thought to
originate from surface states. These results show the ability to manipulate the electron transport in nanoscale ZnO
devices.
# 2004 Elsevier B.V. All rights reserved.
Keywords: Nanowires; Nanorods; ZnO; Bandgap
1. Introduction
In recent years, significant interest has emerged in the synthesis of nanoscale materials [1]. One of
the most attractive classes of materials for functional nanodevices are semiconductors. Various means
have been reported for the synthesis of semiconducting nanowires and nanorods [2–4]. Much effort
has focused on catalysis-driven bulk synthesis of nanomaterials using approaches that are neither
substrate site specific nor compatible with most planar device platforms. Nevertheless, nanodevice
functionality has been demonstrated with these materials in the form of electric field-effect switching
[5], single electron transistors [6], biological and chemical sensing [7], and luminescence [8] for one-
dimensional (1-D) semiconducting structures. Included in the semiconductors of interest are semi-
conducting oxides [9–12]. Of these, zinc oxide is particularly interesting for nanodevice applications.
ZnO is an n-type, direct bandgap semiconductor with E
g
= 3.35 eV [13,14].
Materials Science and Engineering R 47 (2004) 1–47

* Corresponding author. Tel.: +1 352 846 1091; fax: +1 352 846 1182.
E-mail address: fl.edu (Y.W. Heo).
0927-796X/$ – see front matter # 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.mser.2004.09.001
ZnO has been effectively used as a gas sensor material based on the near-surface modification of
charge distribution with certain surface-absorbed species [15]. ZnO nanorods would provide
significant enhancement in sensitivity due to high surface-to-volume ratio. ZnO is also piezoelectric,
and is used in surface acoustic wave devices [16]. Huang et al. have reported the site-specific
nucleation and growth of ZnO nanorods on deposited Au catalyst using a high-temperature vapor
transport process [8]. As with any semiconductor, 1-D ZnO nanostructures provide an attractive
candidate system for fundamental quantization and low-dimensional transport studies [17–19].
A large variety of ZnO one-dimensional structures have been demonstrated [20–47]. The large
surface area of the nanorods and bio-safe characteristics of ZnO makes them attractive for gas and
chemical sensing and biomedical applications, and the ability to control their nucleation sites makes
them candidates for micro-lasers or memory arrays. To date, most of the work on ZnO nanostructures
has focused on the synthesis methods. There have been only a few reports of the electrical
characteristics [20–24]. The initial reports show a pronounced sensitivity of the nanowire conductivity
to ultraviolet (UV) illumination and the presence of oxygen in the measurement ambient. There is
strong interest in developing solid-state ozone and hydrogen gas sensors for use in both industry and
domestic applications. Ideal sensors have the ability to discriminate between different gases and arrays
that contain different metal oxides (e.g. SnO
2,
ZnO, CuO, WO
3
) on the same chip can be used to obtain
this result. The fact that ZnO can be grown at low temperatures on cheap substrates such as glass also
makes it attractive for transparent electronics.
In this review, we report on the synthesis of both cored and radial heterostructure nanorods and
simple, reproducible methods for nanorod device fabrication and give some examples of device
functionality. The ability to control the synthesis of high quality ZnO nanowires leads to potential

applications in UV photodetection, gas sensing and transparent electronics.
2. ZnO nanorod synthesis
Previous effort in the synthesis of ZnO nanowires and nanorods have employed vapor-phase
transport via a vapor–liquid–solid (v–l–s) mechanism [46,47], gas reactions [48] and oxidation of
metal in the pores of anodic alumina membranes [49,50]. While these materials provide interesting
systems for investigating fundamental properties or for exploring device concepts via single prototype
device construction, the ability to synthesize nanorods at arbitrary locations at moderate temperatures
is needed for nanodevice integration. This requires site-specific nucleation of nanorods, as well as a
growth process that remains site specific and is compatible with the device platform of interest. It
would be advantageous to achieve nanorod growth from a flux source that could be controlled at the
atomic level, thus enabling compositional modulation along the rod length.
In this section, we report on the site-selective growth of ZnO nanorods using a catalysis-driven
molecular beam epitaxy (MBE) method. Low temperature MBE conditions are identified so that ZnO
nucleation and growth occurs only on the deposited metal catalyst. With this approach, site specific,
single crystal ZnO nanorod growth is achieved with nanorod diameters as small as 15 nm.
The growth experiments were performed using a conventional MBE system. The background
base pressure of the growth chamber was $5 Â 10
À8
mbar. An ozone/oxygen mixture was used as the
oxidizing source. The nitrogen-free plasma discharge ozone generator yielded an O
3
/O
2
ratio on the
order of 1–3%. No effort was made to separate the molecular oxygen from the ozone. The cation flux
was provided by a Knudsen effusion cell using high purity (99.9999%) Zn metal as the source. Cation
and O
2
/O
3

partial pressure was determined via a nude ionization gauge that was placed at the substrate
position prior to growth. The beam pressure of O
3
/O
2
mixture was varied between 5 Â 10
À6
and
2 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
5 Â 10
À4
mbar, controlled by a leak valve between the ozone generator and the chamber. The Zn
pressure was varied between 5 Â 10
À7
and 4 Â 10
À6
mbar. The substrates were Si wafers with native
SiO
2
layer terminating the surface.
Site-selective nucleation and growth of ZnO was achieved by coating Si substrates with Ag
islands. For thick Ag, a continuous ZnO film could be deposited. For nominal Ag film thicknesses of
20–200 A
˚
, discontinuous Ag islands are realized. On these small metal islands, ZnO nanorods were
observed to grow. Efforts to deposit ZnO on Ag-free SiO
2
/Si surface area under a variety of growth
conditions proved unsuccessful. Zn metal deposition could be achieved at substrate temperatures of
25–100 8C, but with no ZnO formation for a wide range of O

2
/O
3
partial pressure. Higher substrate
temperatures yield no deposition as the Zn metal vapor pressure rises quickly at moderate tempera-
tures. Typical growth times for ZnO on the Ag-coated silicon was 2 h with growth temperatures
ranging from T
g
= 300 to 500 8C. After growth, the samples were evaluated by X-ray diffraction,
scanning electron microscopy (SEM), transmission electron microscopy (TEM), and photolumines-
cence (PL).
Fig. 1 shows a scanning electron microscopy image of ZnO nanorods grown on a Si wafer that
was coated with a nominally 10 nm thick layer of Ag. The Ag was deposited using e-beam
evaporation. The images are for ZnO nanorods grown at 400 8C with a Zn pressure of 2 Â 10
À6
Torr
Torr and an oxygen/ozone pressure of 5 Â 10
À4
Torr. Under these conditions, ZnO deposition was
observed only on the Ag with no growth on regions of the SiO
2
-terminated Si surface that was devoid
of Ag. A dense entangled collection of ZnO nanorods is observed to grow from the surface. Both
cylindrical nanorods and faceted whiskers can be observed in the forest of ZnO nanostructures grown
at 400 8C. At higher temperatures, only nanorods are observed. In many cases, the length of ZnO
nanorods is in excess of 2 mm. Note also that multiple nanorods are observed to nucleate from the
relatively large Ag islands. As such, it does not appear that the diameter of the nanorods is determined
by the initial radii of the Ag islands. X-ray diffraction of the deposited materials confirms that the
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 3
Fig. 1. SEM image of ZnO nanorods nucleated on Ag-coated Si/SiO

2
substrate.
material is ZnO. X-ray diffraction patterns taken along the surface normal, indicating only ZnO peaks.
The diffraction pattern for the material grown at 400 8C is consistent with randomly oriented
polycrystalline material, although selected area electron diffraction (discussed below) indicates a
preferred c-axis orientation of individual nanorods along the long axis. A preferred (0 0 2) orientation
seen for nanorods obtained at 500 8C indicates a more vertically aligned growth at this temperature.
The most intriguing structures are those that result from isolated Ag nanoparticles. In depositing
the Ag catalyst films, certain regions of the SiO
2
/Si surface were shadowed from deposition, leading to
a gradient in Ag thickness, Ag nanoparticle coverage, and average nanoparticle diameter. Within these
areas, isolated Ag nanoparticles could be located, thus allowing direct imaging of nanorod formation
from individual Ag islands. Clusters of ZnO nanorods were observed to nucleate from these isolated
Ag islands. Fig. 2 shows field-emission SEM images of ZnO nanorod clusters, including a high-
resolution image of a single nanorod. Energy-dispersive spectrometry was used to determine the
nanorod composition (ZnO) in addition to confirming the absence of ZnO on regions of the substrate
surface that are devoid of Ag. In order to acquire these images, the sample was coated with a thin layer
of carbon to avoid charging effects. From the high-resolution image, the nanorod cross-section appears
to be cylindrical, although any faceting of the side walls would be obscured by the carbon coating. The
thickness of the nanorod shown in Fig. 2 is on the order of 30 nm, although the carbon coating may
exaggerate this thickness.
In addition to SEM, the nanorods were examined using transmission electron microscopy. Fig. 3
shows a transmission electron microscopy image of an individual ZnO nanorod. The rod was imaged
from a cross-sectioned sample. The nanorod shown in Fig. 3 was not carbon coated. An estimate of the
rod thickness is 20 nm. Selected area diffraction (SAD) of nanorod specimens indicates that the rods
are single crystal ZnO, with the c-axis oriented along the long axis of the rod. Also evident in the image
is a small particle embedded at the tip of the rod. This is similar to what is observed for other nanorod
synthesis that is driven by a catalytic reaction, where catalyst particles become suspended on the
nanorod tip [51,52]. Evidence for termination of the ZnO nanorods tips with catalyst particles is also

observed in field-emission SEM images. Local energy-dispersive spectrometry (EDS) measurements
indicate that the terminating particle is Ag, although more characterization is needed in order to
confirm this.
The mechanism for nanorod growth is catalysis driven, and appears to be related to the vapor–
liquid–solid model reported for the nanorod synthesis of other materials. ZnO nanoparticle formation
4 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
Fig. 2. Deterministic growth of ZnO nanorod clusters formed via catalytically driven MBE; nanorod diameter is 20–30 nm.
via the internal oxidation of Zn in Ag/Zn alloys has previously been reported [53]. In these studies,
oxygen is diffused into an Ag/Zn alloy, with nanoscale ZnO precipitates forming in the bulk of the
sample. For the present case of nanorod formation, the reaction between ozone/oxygen flux and the Ag
islands appears to result in surface and subsurface oxygen diffusion in the metal island, perhaps
involving the intermediate formation of Ag
2
O. Zn atoms impinging on the Ag island surface then
diffuse either on the surface or in the bulk of the island, where they react with the Ag
2
O to form ZnO.
The solid solubility of Zn in Ag is on the order of 25 wt.% for the temperatures considered in these
experiments. Zn addition to Ag significantly suppresses the melting point to 710 8C at 25 wt.% Zn.
Note that the melting point of Zn is rather low (420 8C). Based on these arguments, one might
anticipate rather high diffusion rates for Zn in Ag for the temperatures considered. Note that the
temperatures at which vapor–liquid–solid growth was previously reported for ZnO are significantly
higher than that used with the present work. It should also be noted that the addition of Ag during the
growth of complex oxide thin film has been reported to be effective in enhancing the oxidation process
for various oxide thin-film compounds [54].
In addition to examining the structure with microscopy and X-ray diffraction, the optical
properties of the nanorods were examined using photoluminescence. A He–Cd (325 nm) laser
was used as the excitation source. The room temperature luminescence reveals a robust near band
edge emission peak located at 375 nm indicating that that rods are highly crystalline. This is consistent
with luminescence reported for near-band edge emission in epitaxial films [29,30] and larger diameter

ZnO nanorods [55]. A broad, but weak, green emission peak is also observed at $520 nm that is
typically associated with trap-state emission attributed to singly ionized oxygen vacancies in ZnO [56].
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 5
Fig. 3. TEM and selected area diffraction image of a single crystal ZnO nanorod.
The emission is the result of the radiative recombination of photogenerated holes with electrons
occupying the oxygen vacancy. Similar results have been observed for ZnO nanorods formed via vapor
transport. Enhancement in the green emission in nanorods as compared to bulk may be attributed to a
higher density of vacancies in the rods. This may be due to the higher surface area to volume ratio for
nanorods as compared to bulk.
3. Structure and optical properties of cored wurtzite (Zn,Mg)O heteroepitaxial nanowires
Nanowire growth has been reported using several techniques [57,58], and has included numerous
semiconductors [59,60], including the oxides Ga
2
O
3
[61],In
2
O
3
[62], SnO
2
[63], and ZnO [64].
Despite significant progress, major challenges in the manipulation of nanowire materials remain. The
fabrication of integrated systems using nanowire material requires the site-specific growth or
placement of nanowires on relevant device platforms. In addition, the formation of complex,
multi-component structures and interfaces are needed for low-dimensional structures and electronic
devices. In thin-film semiconductor research, the formation of heteroepitaxial interfaces has proven to
be useful in the development of numerous device concepts, as well as in the investigation of low-
dimensional phenomena [65]. Unfortunately, such heterostructures have rarely been realized outside
of the conventional 2-D planar thin-film geometry [66]. Nevertheless, the synthesis of 1-D linear

heterostructures is scientifically interesting and potentially useful, particularly if a technique is
employed that allows for spatial selectivity in nanowire placement. Addressing these challenges could
prove useful in realizing integrated device functionality involving semiconducting nanowires for a
number of applications, including nanoscale electric field-effect transistors [67], single electron
transistors [68], biological and chemical sensors [69], electron emitters [70], optical emitters and
detectors [69,70].
In this section, the properties of 1-D heteroepitaxial structures are described. In particular, the
structural and optical properties of cored (Zn,Mg)O nanowires, formed via self-assembled bimodal
growth, are discussed. ZnO is among the more interesting and important semiconducting oxides [71].
ZnO is an n-type, direct bandgap semiconductor with E
g
= 3.35 eV. Electron doping via defects
originates from Zn interstitials in the ZnO lattice. The intrinsic defect levels that lead to n-type doping
lie 0.05 eV below the conduction band. The room temperature Hall mobility in ZnO single crystals is
among the highest for the oxide semiconductors, on the order of 200 cm
2
V
À1
s
À1
. The exciton
binding energy for ZnO is on the order of 60 meV, yielding efficient luminescence at room
temperature. The synthesis of ZnO nanowires and nanorods has been demonstrated using vapor-
phase transport via a vapor–liquid–solid mechanism [72], gas reactions [73], and oxidation of metal in
the pores of anodic alumina membranes [74]. Room temperature ultraviolet lasing via optical pumping
has been demonstrated with ZnO nanorods on deposited Au catalyst using a high-temperature vapor
transport process [70]. Recently, we reported on catalyst-driven molecular beam epitaxy of ZnO
nanorods [75]. The process is site specific, as single crystal ZnO nanorod growth is realized via
nucleation on Ag films or islands that are deposited on a SiO
2

-terminated Si substrate surface. Growth
occurs at relatively low substrate temperatures, on the order of 300–500 8C, making it amenable to
integration on numerous device platforms. With this approach, nanorod placement can be predefined
via location of metal catalyst islands or particles.
The heteroepitaxial cored nanostructures described here are based on the (Zn,Mg)O alloy system,
and were synthesized using the catalysis-driven molecular beam epitaxy method. Details of the growth
experiments are reported elsewhere. An ozone/oxygen mixture was used as the oxidizing source. The
cation flux was provided by Knudsen effusion cells using high purity (99.9999%) Zn metal and Mg
6 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
(99.95%) as the source materials. The substrates were Si wafers with native SiO
2
terminating the
surface. No effort was made to remove the native oxide or to terminate the surface with hydrogen.
Site-selective nucleation and growth of cored nanorods was achieved by coating Si substrates
with Ag islands. For a nominal Ag film thickness of 20 A
˚
, discontinuous Ag islands are realized. On
these small metal catalyst islands, (Zn,Mg)O nanorods were observed to grow. Fig. 4 shows an FE-
SEM micrograph of these nanorods, indicating a length approaching 1 mm. The growth temperature
was 400 8C. Energy-dispersive spectrometry measurement was performed on a single nanowire under
the Transmission electron microscopy. EDS confirmed the presence of Mg in the nanorod, as seen in
Fig. 5. Typical growth times for (Zn,Mg)O on the Ag-coated silicon was 2 h with growth temperatures
ranging from T
g
= 300–500 8C. The site specificity for nanowire growth using this technique is evident
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 7
Fig. 4. Field-emission scanning electron microscopy image of cored (Zn,Mg)O nanorods grown on Ag-coated Si. The
conditions for growth were T
g
: 400 8C; Zn pressure: 3 Â 10

À6
mbar, Mg pressure: 4 Â 10
À7
mbar, O
2
/O
3
pressure:
5 Â 10
À4
mbar. The Mg source was shuttered with a 60 s open/60 s closed cycle.
Fig. 5. Energy dispersive spectrometry data for (Zn,Mg)O nanorods.
in Fig. 6, showing FE-SEM images of (Zn,Mg)O nanorods on a Ag-patterned substrate grown in a Zn
pressure of 3 Â 10
À6
mbar, a Mg pressure of 4 Â 10
À7
mbar, and an O
2
/O
3
pressure of 5 Â 10
À6
mbar.
(Zn,Mg)O nanorods form only on the Ag-coated regions. The potential for growing single nanorods on
selected locations is exemplified in Fig. 7, where single ZnO nanorods are nucleated on Ag
nanoparticles dispersed on a SiO
2
-terminated Si surface. In order to acquire these images, the sample
was coated with a thin layer of carbon to avoid charging effects. From the high-resolution image, the

8 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
Fig. 6. Field-emission scanning electron microscopy image of cored (Zn,Mg)O nanorods grown on a patterned Ag-coated Si
substrate. The conditions for growth were T
g
: 400 8C; Zn pressure: 3 Â 10
À6
mbar; Mg pressure: 4 Â 10
À7
mbar; O
2
/O
3
pressure: 5 Â 10
À4
mbar. Note that (Zn,Mg)O nanorod nucleation occurs only on the catalyst-coated regions.
Fig. 7. Field-emission scanning electron microscopy image of individual ZnO nanorods grown on Ag nanoparticles
dispersed on the Si substrate.
nanorod cross-section appears to be cylindrical, although any faceting of the side walls may be
obscured by the carbon coating. The thickness of the nanorods shown is on the order of 30 nm,
although the carbon coating may exaggerate this thickness.
The formation of the (Zn,Mg)O nanorods includes the v–1–s mechanism described earlier,
although heteroepitaxial growth occurs as well as will be seen. Fig. 8 shows a Z-contrast scanning
transmission electron microscopy (Z-STEM) image of an individual (Zn,Mg)O nanorod grown at
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 9
Fig. 8. Z-contrast scanning transmission electron microscopy image (a) of a (Zn,Mg)O nanorod with (b) a Ag catalyst
particle at the rod tip.
400 8C, with a Zn pressure of 3 Â 10
À6
mbar, a Mg pressure of 4 Â 10
À7

mbar, and an O
2
/O
3
pressure
of 5 Â 10
À4
mbar. The Mg flux was cycled on and off every 60 s, which was inconsequential to the
nanorod structure. Evident in the image is a small particle embedded at the tip of the rod. This is
similar to what is observed for other nanorod synthesis that is driven by a catalytic reaction, where
catalyst particles become suspended on the nanorod tip. The diameter of the catalyst particle is $6 nm.
Note that, at the nanorod tip, the contrast in the Z-STEM image is relatively uniform, indicating
uniform cation distribution. However, the rod diameter is also tapered along the length, being thicker
at the base than on the tip, with an average diameter on the order of 10 nm.
As reported elsewhere, a radial segregation of the Zn and Mg occurs during growth [76]. A Zn-
rich core surrounded by a Mg-rich sheath is observed as seen in Fig. 9. As discussed elsewhere, in
bulk material, the solubility of Mg in ZnO is relatively low, on the order of 4 at % [77]. In contrast,
Mg content as high as Zn
0.67
Mg
0.33
O has been reported to be metastable in the wurtzite structure for
epitaxial thin films. For this composition, the bandgap of ZnO can be increased to $3.8 eV. For
(Zn,Mg)O nanorod growth, it appears that both (hence bimodal) growth modes are relevant, but for
different regions in the rod. Under low temperature MBE growth conditions, a solubility-driven
segregation occurs during the catalyst-driven core formation, with the core composition determined
by bulk solid solubility. Subsequently, an epitaxial sheath grows with Mg content and crystal
structure determined by epitaxial stabilization. The net result is the growth of (Zn,Mg)O nanorods
that are not uniform in composition across the diameter, but distinctly cored. Fig. 9 shows a high-
resolution Z-STEM image of a nanorod grown under the conditions described. The lattice image for

the nanorod specimen indicates that the rod is crystalline with the wurtzite crystal structure
maintained throughout the cross-section. The c-axis is oriented along the long axis of the rod.
The higher contrast for the center core region clearly indicates a higher cation atomic mass. The
structures consist of a zinc-rich Zn
1–x
Mg
x
O core (small x) surrounded by a Zn
1Ày
Mg
y
O (large y)
sheath containing higher Mg content.
While the nanorod imaged in Fig. 9 is crystalline across the entire cross-section, other rods exhibit
sheath properties that vary along the length. In particular, consider the nanorod shown in Fig. 10.In
this case, the core and sheath are both crystalline in one region of the rod. However, as one proceeds
along the length, the crystallinity changes. In particular, for the rod considered, the sheath region
becomes either polycrystalline or amorphous as one approaches the rod tip. Still further down the
nanowire, the image suggests a lack of crystallinity for both the core and sheath, although the lack of
crystallinity in the sheath may effectively obscure imaging of the core region. This change in
crystallinity change along the length of the rod may reflect the fact the temperature gradient will
develop along the nanowire length during growth. This occurs since the substrate is the source of heat
during nanorod formation. Sheath material deposited on the tip of longer rods during the latter part of
the synthesis process will do so at a lower local temperature than the material closer to the substrate.
In order to assess crystalline quality and investigate possible quantization effects, the optical
properties of the cored nanorods were examined using photoluminescene. Spectra were taken over the
temperature range 6–300 K. A He–Cd (325 nm) laser was used as the excitation source. For the low
temperature measurements, the sample was cooled using either a helium flow or closed cycle cryostat.
Fig. 11 shows the photoluminescence spectra taken at various temperatures for the cored nanorod
specimens. For ZnO rods (no Mg), the photoluminescene results are consistent with luminescence

reported for near band edge emission in crystals [78], epitaxial films [79], and larger diameter ZnO
nanorods [80]. The free exciton emission dominates luminescence, with a room temperature peak at
3.30 eV. At room temperature, the spectra for the cored (Zn,Mg)O nanorods is also dominated by the
free exciton luminescence. However, the peak in luminescence at room temperature is at 3.35 eV,
which is blueshifted relative to that seen in pure ZnO (peak at 3.30 eV). As the temperature is
10 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
decreased, the free exciton emission shifts to higher energy due to the temperature dependence of the
bandgap. At the lower temperatures, the donor-bound exciton (D
0
,X) dominates luminescence [79].
Also seen at low temperatures is a peak at 3.35 eV, which is assigned as the corresponding longitudinal
optical phonon replica (D
0
,X)-LO. At lower energies, at least four relatively weak peaks are also
observed in the visible spectrum, indicated by arrows in the figure. Similar peaks have been reported
elsewhere for bulk and thin-film ZnO materials [78,79].
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 11
Fig. 9. Z-contrast scanning transmission electron microscopy images of cored (Zn,Mg)O nanorods. The high contrast of the
core (a) indicates a significantly higher Zn content for the core relative to the sheath. The higher resolution image (b) of the
core indicates the wurtzite structure for the core and sheath material.
Blueshifts in the near band edge luminescence for nanoscale semiconducting structures can
reflect quantum confinement effects. With a bandgap range of 3.2–3.8 eV, the ZnO/(Zn,Mg)O
heterostructure system affords the opportunity to realize electron confinement due to band offsets.
In semiconductors, the onset of reduced dimensionality effects should be observable for structures
with dimensions below the relevant length scales. Quantization of the electron states should be
observable for low dimensional structures with length scales on the order of the exciton Bohr diameter
[81,82]. A significant blueshift in the near-band edge emission was previously observed for InP
nanorods with diameters approaching that of the exciton diameter [83]. For ZnO, the exciton Bohr
diameter is 34 A
˚

, which is approximately equal to the core diameters considered. Shifts in the near-
band edge photoluminescence peak energy have been observed for 0-D ZnO nanoparticles diameters
on the order of 2–6nm [84]. For ZnO/(Zn,Mg)O superlattices, quantum confinement effects are
predicted for well widths less than $5 nm, and are reported for ZnO layer thicknesses less than $4nm
[82]. ZnO/(Zn,Mg)O superlattices with a ZnO well width of 3.1 nm show a 100 meV blueshift in near-
band edge luminescent peak energy [82]. In the present work, a 50 meV shift is observed for the 1-D
nanorod cored structure, the core diameter (4 nm) being roughly equivalent to the bulk exciton
diameter (3.4 nm). While a blueshift in near edge emission is consistent with expected quantum
confinement in 4 nm diameter ZnO cores, one cannot easily differentiate this from a similar blueshift
that would result from a few percent doping of Mg in the core regions. However, from the Mg
dependence of the band edge peak, it appears that quantization effects on the PL shift are minimal.
12 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
Fig. 10. Z-contrast scanning transmission electron microscopy image of regions of a nanorod showing: (a) a crystalline core
and sheath; (b) an amorphous sheath; and (c) an amorphous sheath with no evidence of crystallinity in the core.
Fundamentally, these 1-D ZnO nanostructures provide an attractive system for probing low-
dimensional effects such as quantization, interface scattering, and ballistic transport. The relatively
low growth temperatures suggest that cored (Zn,Mg)O nanorods could be integrated on device
platforms for numerous applications, including chemical sensors, nano-optics, scanning probes, and
nanoelectronics. More generally, the spontaneous bimodal growth mode demonstrated for (Zn,Mg)O
may prove applicable in the synthesis of heteroepitaxial cored nanowire-materials where thermo-
dynamics and epitaxy impose different cation or anion solubility limits.
4. ZnO/cubic (Mg,Zn)O radial nanowire heterostructures
ZnO/(Zn,Mg)O heterostructures have been realized that exhibit quantum confinement in
quantum well structures. While the formation of planar semiconductor heterostructures is common
for thin films, the synthesis of one-dimensional heterostructures is difficult. Axial heterostructures, in
which the chemical modulation is imposed along the length of the nanowire axis, have been reported
for a few systems, such as InAs/InP and Si/SiGe nanowires [85,86] ZnO/Zn
1Àx
Mg
x

O quantum well
nanorods have been grown as axial heterostructures as well [28]. Little work has addressed the
synthesis of nanowire heterostructures in which the chemical modulation extends radially from the
wire center [87].
In this section, we discuss the formation of ZnO-based one-dimensional radial heterostructure
nanowires possessing a radial modulation in composition and structure. In particular, the nanowires
consist of wurtzite ZnO cores surrounded by a rock-salt structured (Mg,Zn)O sheath. Despite the
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 13
Fig. 11. Photoluminescence at various temperatures for (a) near band edge and (b) visible emission from cored (Zn,Mg)O
nanorods.
mismatch in lattice symmetry, the cubic (Mg,Zn)O sheath is found to be epitaxial on the wurtzite ZnO
core.
The growth experiments were performed using a conventional MBE system. The background
base pressure of the growth chamber was $5 Â 10
À8
mbar. An ozone/oxygen mixture was used as the
oxidizing source. The nitrogen-free plasma discharge ozone generator yielded an O
3
/O
2
ratio on the
order of 1–3%. No effort was made to separate the molecular oxygen from the ozone. The flux of Zn
and Mg were provided by Knudsen effusion cells using high purity Zn metal (99.9999%) and Mg
metal (99.995%) as the sources. Cation and O
2
/O
3
partial pressures were determined via a nude
ionization gauge that was placed at the substrate position prior to growth. The beam pressure of the O
3

/
O
2
mixture was varied between 5 Â 10
À6
and 5 Â 10
À4
mbar, controlled by a leak valve between the
ozone generator and the chamber. The Zn pressure was varied between 5 Â 10
À7
and 5 Â 10
À6
mbar
and for Mg, the pressure ranged between 1 Â 10
À7
and 1 Â 10
À6
mbar. Si wafers with a native SiO
2
layer terminating the surface were used as substrates. Ag served as the catalyst.
The ZnO/(Mg,Zn)O nanowires were nucleated and grown on Si substrates coated with Ag for
catalytic growth. Typical growth times for nanowires on the Ag-coated silicon were 2 h with growth
temperatures ranging from T
g
= 300–500 8C. Energy dispersive spectrometry spectra from single
nanowires were collected by TEM (Philips 420 EM). Compositional line-scans, profiled across the
nanowire, were measured by scanning transmission electron microscopy equipped with EDS.
Under continuous Zn, Mg, and O
3
/O

2
flux, the ZnO/(Mg,Zn)O nucleated uniformly on the
catalyst-coated substrate. Fig. 12 shows a SEM image of ZnO/(Mg,Zn)O nanowires grown on a Si
wafer that was coated with a nominally 2 nm thick layer of Ag. The Ag was deposited using e-beam
evaporation. The image is from radial heterostructured ZnO/(Mg,Zn)O nanowires grown with a Zn
pressure of 3 Â 10
À6
mbar, a Mg pressure of 4 Â 10
À6
mbar, and an O
2
/O
3
pressure of 5 Â 10
À4
mbar.
The growth temperature was 400 8C. Under these conditions, nanowires were observed only on the
Ag. No growth occurred on the substrate surface that was devoid of Ag. The length of the ZnO/
(Mg,Zn)O nanowires is in excess of 2 mm.
14 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
Fig. 12. SEM image of radial heterostructured ZnO/(Mg,Zn)O nanowires grown on an Ag-coated Si wafer.
The morphology and microstructure of ZnO/(Mg,Zn)O nanowires were analyzed by TEM.
Fig. 13(a) shows a TEM image of a nanowire, displaying a difference in brightness intensity between
the core and sheath regions. This image was obtained without the objective aperture in order to
minimize any diffraction contrast. The contrast across the diameter of the nanowire is predominantly
mass contrast, reflecting a difference in average atomic number (Z) of the core and sheath region. The
darker core region contains more Zn, the lighter sheath region more Mg. Selected area diffraction
taken from this single nanowire in Fig. 13(b) shows that the nanowire consists of two different crystal
structures. The single crystal diffraction pattern result corresponds to the hexagonal wurtzite structure.
Based on the mass contrast discussed earlier, this is a ZnO-rich core. The ring diffraction pattern that is

observed corresponds to a cubic rock salt structure with d-spacing similar to that for MgO. The first
circle is diffraction from the (Mg,Zn)O {1 1 1}, the second circle by the (Mg,Zn)O {2 0 0}, and the
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 15
Fig. 13. Radial heterostructured ZnO/(Mg,Zn)O nanowire. (a) TEM image of a nanowire showing a difference in brightness
intensity between core and sheath regions. (b) Selected-area diffraction (SAD) from the nanowire that consists of two
different crystal structures. (c) Dark field image taken from the diffraction spot of the wurtzite structure showing the ZnO core
of nanowire. In (d), a DF image from the ring diffraction pattern of the rock salt structure showing the (Mg,Zn)O sheath.
third circle by the (Mg,Zn)O {2 2 0} plane of the rock salt structure. A dark field (DF) image shown in
Fig. 13(c), formed from the (0 0 0 2) diffraction spot of the wurtzite structure, yields intensity
corresponding only to the core part of the nanowire. The discontinuous brightness along the nanowire
length results from bending of the nanowire. Clearly, the center core of the nanorod is hexagonal
wurtzite ZnO, with the sheath material not exhibiting this structure. Fig. 13(d) shows a DF image
formed from the circle diffraction spots of the rock salt structure. The image includes the entire volume
of the nanowire, indicating that it comes from the nanowire sheath. Therefore, the core of this wire is
ZnO having the wurtzite structure. The surrounding sheath is (Mg,Zn)O having the rock salt structure.
While the nanowire heterostructures considered here possess a rock-salt (Mg,Zn)O sheath, we have
also observed similar cored nanowires grown with lower Mg flux that display a wurtzite (Zn,Mg)O
epitaxial sheath.
High-resolution transmission electron microscopy was also performed on the interface region
between the ZnO core and (Mg,Zn)O sheath. The image shown in Fig. 14(a) indicates a radial
heterostructured nanowire in which the interface between the sheath and core is epitaxial despite the
discontinuity in crystal structure and symmetry. The lower-right region of the image is the core, while
the upper-left region is the sheath. Growth of the core yields the rod axis along the c-axis direction of
the hexagonal ZnO. For the (Mg,Zn)O sheath, the (2 0 0), (1 À1 1), and (1 1 À1) planes match those
of the rock salt structure, according to the lattice spacing and angle between planes. Note that the
lattice mismatch between the (Mg,Zn)O (2 0 0) planes and the ZnO (0 0 0 2) planes produces
regularly-spaced interfacial edge dislocations.
In order to further delineate the structure of these nanorod heterostructures, the compositional
distribution of Mg and Zn in a single ZnO/(Mg,Zn)O nanowire was examined by energy-dispersive
spectrometry using the electron beam in STEM. Fig. 14(b) plots the intensity of the Mg Ka

1
and Zn
Ka
1
peaks across the nanowire. The compositional line-scan clearly shows a radial segregation of the
Zn and Mg cation. Note that the spatially-varying spectra always include irradiation of the sheath
region. Hence, the Mg intensity does not drop to zero when irradiating the center. However, in the core
region, there is a clear drop in Mg intensity along with a corresponding increase in Zn intensity. This
result unambiguously supports the radial segregation of Zn and Mg, forming a ZnO core and a
(Mg,Zn)O sheath.
The growth of the ZnO/(Mg,Zn)O nanowires is a catalyst-driven reaction similar to ZnO
nanowire growth using catalysis-driven molecular beam epitaxy reported earlier [29]. Initially, the
ZnO core is nucleated on the Ag catalyst with the reaction between the Zn and the oxygen source. It
should be noted that uncored (Mg,Zn)O nanowires with the rock-salt structure throughout the rod
were not observed for growth on Ag-coated Si substrates. At the selected growth conditions, Ag does
not act as an effective catalyst for growth of cubic-cored (Mg,Zn)O nanowires. According to
the phase diagram between ZnO and MgO, cubic (Mg,Zn)O can accommodate a maximum of 56
at.% ZnO at 1600 8C and maintain its rock salt structure with a lattice constant close to that of pure
MgO [88]. In the case of ZnO, the solid solubility of Mg is limited to only 2 at.% maximum while
keeping its hexagonal structure. In pulsed laser deposited thin film, a non-equilibrium solid solution
can accommodate up to 36 at.% Mg solubility in ZnO [89]. However, the ZnO/(Mg,Zn)O
heterostructured nanowire synthesis appears to follow equilibrium growth according to the phase
diagram.
In order to further elucidate the optical properties, photoluminescence of the cored nanowires was
measured. This measurement was performed at room temperature with a He–Cd laser as the excitation
source. The comparison of band edge photoluminescence between pure ZnO nanorods and cored ZnO/
(Mg,Zn)O nanorods shows only a 50 meV blue shift. As previously reported, a 50 meV blue shift in
the near-band luminescent peak energy of cored ZnO compared to that of ZnO nanowires could
16 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
indicate approximately 2 at.% Mg solubility in the core-ZnO nanowire [90]. However, this radial

heterostructured nanowire could also produce quantum confinement, as the bandgap of (Mg,Zn)O is
much larger than that of ZnO. The diameter variation of the ZnO core is currently being explored to
investigate the quantum confinement effect of the cored ZnO/(Mg,Zn)O nanowire structures.
In summary, radial heterostructured nanowires of ZnO/(Mg,Zn)O were selectively grown on an
Ag-coated Si wafer. Structural and compositional analyses of the nanowires clearly indicate that,
under certain conditions, the core of nanowire is ZnO with the hexagonal wurtzite structure, while the
nanowire sheath is (Mg,Zn)O with a cubic rock salt structure.
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 17
Fig. 14. (a) High-resolution transmission electron microscopy of ZnO and (Mg,Zn)O interface; (Mg,Zn)O sheath in upper-
left region, ZnO core in lower-right region. (b) Compositional line-scan across the nanowire probed by STEM EDS
spectroscopy.
5. Ferromagnetism in ZnO
The manipulation of spin in semiconductors (so-called spintronics) presents a new paradigm for
functionality in electronic materials. Ideas currently under development offer intriguing opportunities
to pursue novel device concepts based on the discrimination and manipulation of spin distributions. Of
the semiconducting materials, ZnO offers significant potential in providing charge, photonic, and spin-
based functionality. ZnO is a direct, wide bandgap semiconductor with potential utility in UV
photonics and transparent electronics. For spintronics, theoretical predictions suggest that room
temperature carrier-mediated ferromagnetism should be possible in ZnO, albeit for p-type material.
Unfortunately, the realization of p-type ZnO has proven difficult. Ab initio calculations do predict
ferromagnetism in n-type ZnO doped with most transition metal (TM) ions, including Co and Cr, but
predict no ferromagnetism for Mn-doped ZnO. This is consistent with experimental results where
ferromagnetism is not observed in Mn-doped ZnO that is n-type due to group III impurities. However,
we recently observed ferromagnetism in n-type Mn-implanted, Sn-doped ZnO crystals, where Sn (a
group IV element) serves as a doubly ionized donor impurity. The Curie temperature is quite high,
approaching 250 K. It is unclear what special role the Sn dopant plays, as compared to group III
elements, in enabling ferromagnetism in the Mn-doped ZnO system. The Sn may simply provide
carriers, albeit electrons, that effectively mediate the spin interactions. The Sn dopant might
alternatively form complexes with Mn, resulting in both Mn
2+

and Mn
3+
sites that could yield a
ferrimagnetic ordering. Nevertheless, ferromagnetism is observed in Mn-doped ZnO (co-doped with
Sn).
In recent years, the injection and manipulation of spin-polarized electrons in semiconductor
materials has become the focus of intense research activity [91–93]. A functional spin offers an
opportunity to develop new device concepts based on the discrimination and manipulation of spin
distributions. Fundamental challenges related to the lifetime, manipulation, and detection of spin-
polarized electrons in a semiconductor host must be addressed in order to realize spintronic technology
based on semiconductors. Of particular importance is the realization of spin-polarized currents in a
semiconductor matrix. Efforts directed at investigating the injection of electrons through a ferro-
magnetic metal/semiconductor junction have shown this structure to be ineffective. However, recent
experiments indicate that injection of spin-polarized electrons from a ferromagnetic semiconductor
into a non-ferromagnetic semiconductor is possible without detrimental interface scattering. Unfor-
tunately, ferromagnetism in semiconductors is rare and poorly understood, with ferromagnetic
transition temperatures well below room temperature for most known materials. Clearly, the discovery
of ferromagnetism above 300 K in a semiconductor would prove enabling in realizing practical spin-
based electronic devices. More generally, the study of any semiconducting material that supports spin-
polarized electron distributions would be useful in understanding and developing spintronic concepts.
5.1. Ferromagnetism in semiconductors
Magnetism in semiconductor materials has been studied for many years, and includes spin glass
and anti-ferromagnetic behavior in Mn-doped II–VI compounds, as well as ferromagnetism in
europium chalcogenides and Cr-based spinels [93–99]. In recent years, ferromagnetism in semi-
conductors has received renewed attention, partly due to interest in spintronic device concepts [91].
Due to transition metal solubility and technological interest, contemporary research has primarily
focused on II–VI and III–V materials. Strong ferromagnetic interaction between localized spins has
been observed in Mn-doped II–VI compounds with high carrier densities. In Pb
1ÀxÀy
Sn

y
Mn
x
Te
(y > 0.6) possessing a hole concentration on the order of 10
20
–10
21
cm
À3
, ferromagnetism has been
18 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
achieved [88]. In addition, it has been shown that free holes in low-dimensional structures of II–VI
dilute magnetic semiconductors can induce ferromagnetic order [95]. For many semiconductor
materials, the bulk solid solubility for magnetic and/or electronic dopants is not conducive with
the coexistence of carriers and spins in high densities. However, the low solubility for transition metals
in semiconductors can often be overcome via low temperature epitaxial growth. This approach has
been used with Mn-doped GaAs [96–99] in achieving ferromagnetism with a transition temperature of
110 K, which is remarkable high compare to traditional dilute magnetic semiconductor material.
Behavior indicative of ferromagnetism at temperatures above 300 K has recently been reported for
GaN and chalcopyrite semiconductors doped with transition metals, illustrating the potential of
achieving room temperature spintronics technologies [100,101].
Despite recent experimental success, a fundamental description of ferromagnetism in semi-
conductors remains incomplete. Recent theoretical treatments have yielded useful insight into
fundamental mechanisms [102]. Dietl et al. [101,102] have applied Zener’s model for ferromagnetism,
driven by exchange interaction between carriers and localized spins, to explain the ferromagnetic
transition temperature in III–V and II–VI compound semiconductors. The theory assumes ferromag-
netic correlations mediated by holes from shallow acceptors in a matrix of localized spins in a
magnetically doped semiconductor. Specifically, Mn ions substituted on the group II or III site provide
the local spin. In the case of III–V semiconductors, Mn also provides the acceptor dopant. High

concentrations of holes are believed to mediate the ferromagnetic interactions among Mn ions. Direct
exchange among Mn is anti-ferromagnetic as observed in fully compensated (Ga,Mn)As that is donor-
doped. In the case of electron doped or heavily Mn doped materials, no ferromagnetism is detected.
Theoretical results suggest that carrier-mediated ferromagnetism in n-type material is relegated to low
temperatures, if it occurs at all, while it is predicted at higher temperatures for p-type materials [102].
In p-type GaAs grown by low temperature MBE, Mn doping in the concentration range
0.04 Â 0.06 results in ferromagnetism in GaAs. The model described has been reasonably
successful in explaining the relatively high transition temperature observed for (Ga,Mn)As.
Carrier-mediated ferromagnetism in semiconductors is dependent on the magnetic dopant
concentration as well as on the carrier type and carrier density. As these systems can be envisioned
as approaching a metal-insulator transition when carrier density is increased and ferromagnetism is
observed, it is useful to consider the effect of localization on the onset of ferromagnetism. As carrier
density is increased, the progression from localized states to itinerant electrons is gradual. On the
metallic side of the transition, some electrons populate extended states while others reside at singly
occupied impurity states. On crossing the metal-insulator boundary, the extended states become
localized, although the localization radius gradually decreases from infinity. For interactions on a
length scale smaller that the localization length, the electron wavefunction remains extended. In
theory, holes in extended or weakly localized states could mediate the long-range interactions between
localized spins. This suggests that for materials that are marginally semiconducting, such as in heavily
doped semiconducting oxides, carrier-mediated ferromagnetic interactions may be possible.
This theoretical treatment presents several interesting trends and predictions. For the materials
considered in detail (semiconductors with zinc-blende structure), magnetic interactions are favored in
hole-doped materials due to the interaction of Mn
2+
ions with the valence band. This is consistent with
previous calculations for the exchange interaction between Mn
2+
ions in II–VI compounds [94,103]
showing that the dominant contribution is from two-hole processes. This superexchange mechanism
can be viewed as an indirect exchange interaction mediated by the anions, thus involving the valence

band [104,105]. Note that valence band properties are primarily determined by anions in II–VI
compounds. The model by Dietl et al. predicts that the transition temperature will scale with a
reduction in the atomic mass of the constituent elements due to an increase in p–d hybridization and a
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 19
reduction in spin-orbit coupling. Most importantly, the theory predicts a T
c
greater than 300 K for p-
type GaN and ZnO, with T
c
dependent on the concentration of magnetic ions and holes. Recent
experimental evidence for ferromagnetism in GaN appears to substantiate the theoretical arguments
[104,105].
5.2. Spin polarization in ZnO
As pointed out earlier, the theory by Dietl predicts room temperature ferromagnetism for Mn-
doped p-type ZnO. In addition to Dietl’s prediction, ferromagnetism in magnetically doped ZnO has
been theoretically investigated by ab initio calculations based on local density approximation [106].
Again, the results suggest that ferromagnetic ordering of Mn is favored when mediated by hole doping.
However, for V, Cr, Fe, Co, and Ni dopants, ferromagnetic ordering in ZnO is predicted to occur
without the need of additional charge carriers. Several groups have investigated the magnetic
properties of TM-doped ZnO. In all of these studies, the ZnO material was n-type. The magnetic
properties of Ni-doped ZnO thin films were reported [107].Forfilms doped with 3–25 at.% Ni,
ferromagnetism was observed at 2 K. Above 30 K, superparamagnetic behavior was observed.
Fukumura et al. have shown that epitaxial thin films of Mn-doped ZnO can be obtained by
pulsed-laser deposition, with Mn substitution as high as 35% while maintaining the wurtzite structure
[108]. This is well above the equilibrium solubility limit of $13%, and illustrates the utility of low-
temperature epitaxial growth in achieving metastable solubility in thin films. Codoping with Al
resulted in n-type material with carrier concentration in excess of 10
19
cm
À3

. Large magnetoresistance
was observed in the films, but no evidence for ferromagnetism was reported. However, Jung et al.
recently reported ferromagnetism in Mn-doped ZnO epitaxial films, with a Curie temperature of 45 K
[109]. The discrepancy appears to lie in differing film-growth conditions.
Recently, researchers have utilized ion implantation to survey the magnetic properties of a
number of transition metal dopants in various semiconducting oxide materials. Among the system
investigated, the researchers have observed high temperature ferromagnetism in ZnO crystals that are
implanted with transition metal dopants, including Co and Mn, co-doped with Sn. In the case of Mn
co-doped with Sn, the Sn ions are provided as doubly ionized donors. This result differs from that
reported for Mn-doped ZnO doped n-type with Al or Ga, and suggests that group IV dopants may
behave differently than shallow group III donors in terms of interaction with magnetic dopants. If
carrier-mediated mechanisms are responsible, one must explain why the behavior depends on the
specific cation dopant species chosen (Sn versus Al,Ga). Insight into this issue may reside in the fact
that doping via a multi-ionized impurity likely introduces relatively deep donor levels in the energy
gap. Conduction from deep donors is often due to impurity band and/or hopping conduction, as
opposed to conventional free electrons excited to the conduction band. Any carrier-mediated processes
would be dependent on the relevant conduction mechanisms. This result appears to contradict the
expectation that, in the absence of a shallow acceptor level, the dominant exchange mechanism is
short-range superexchange which, for Mn
2+
in ZnO, should favor anti-ferromagnetic ordering. The
results will be discussed in detail in a later section. Despite the uncertainty in the mechanism, these
results (high temperature ferromagnetism in Co- and Mn,Sn-doped ZnO) indicate a pathway for
exploring spintronics in ZnO materials.
Table 1 shows the valence and ionic radii for a number of dopant candidates. The ionic radius of
Mn
2+
(0.66 A
˚
) is relative close to that for Zn (0.60 A

˚
), suggesting moderate solid solubility without
phase segregation. As such, the primary transition metal dopant of interest will be Mn. Chromium and
cobalt presents the possibility of achieving ferromagnetism in ZnO via doping with magnetic ions for
which the net superexchange coupling is ferromagnetic. Low temperature ferromagnetic behavior has
20 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
been observed in Cr-based spinel semiconductors. Theoretical results by Blinowski, Kacman, and
Majewski predict that Cr doping in II–VI semiconductors should result in ferromagnetism [110].Ab
initio calculations based on the local density of states approximation specifically predicts ferromag-
netism in Co- and Cr-doped ZnO without the need for additional doping [106].
Fig. 15 shows the magnetization versus field behavior at 10 K for Sn-doped ZnO samples
implanted with 3 and 5 at.% Mn. Hysteretic behavior is clearly observed, consistent with ferromagnet-
ism. At 10 K, the coercive field in the 3 at.% Mn-doped sample is 250 Oe. It must be noted that other
possible explanations for hysteretic M versus H behavior that are remotely possible include super-
paramagnetism and spin-glass effects [111–114]. Magnetization measurements were also performed
on Sn:ZnO crystals that were not subjected to the Mn implant. This was done to eliminate the
possibility that spurious transition metal impurities might be responsible for the magnetic response.
The Sn-doped ZnO crystals exhibit no magnetic hysteresis, showing that the Mn doping is responsible
for the behavior. To track the hysteretic behavior in the implanted samples as a function of
temperature, both field-cooled and zero field-cooled magnetization measurements were performed
from 4.2 to 300 K. By taking the difference between these two quantities, the para- and diamagnetic
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 21
Table 1
Valence and ionic radii for candidate dopant atoms
Atom Valence Ionic radius (A
˚
)
Zn +2 0.60
Sn +4 0.55
Li +1 0.59

Ag +1 1.00
Mn +2 0.66
Cr +3 0.62
Fe +2 0.63
Co +2 0.38
V +3 0.64
Ni +2 0.55
Mg +2 0.57
Fig. 15. M vs. H curve for Mn implanted ZnO:Sn single crystals showing ferromagnetism in (a) 3 at % Mn and (b) 5 at.% Mn
implantation doses.
contributions to the magnetization can be subtracted, leaving only a measure of the hysteretic
ferromagnetic regime.
When assigning the origin of ferromagnetism in transition metal doped semiconductors, one must
carefully consider the possibility that secondary phase formation is responsible. High-resolution
transmission electron microscopy is the most direct means of exploring this issue. These activities are
currently being pursued for the Mn-implanted ZnO samples. Nevertheless, one can also consider what
known ferromagnetic impurity phases are possible [111–114]. First, metallic Mn is anti-ferromag-
netic, with a Ne
´
el temperature of 100 K. In addition, nearly all of the possible Mn-based binary and
ternary oxide candidates are anti-ferromagnetic. The exception to this is Mn
3
O
4
, which is ferromag-
netic with a Curie temperature of 46 K in thin films [115]. X-ray diffraction measurements on the
implanted samples show no evidence for Mn-O phases, although it is recognized that diffraction is
limited in detecting secondary phases that may represent a fraction of a percent of total volume in the
implanted region. However, even if this phase were present, it could not account for the remarkably
high ferromagnetic transition temperature of $250 K observed for the Mn-implanted ZnO:Sn crystals.

Note also that the Mn concentrations used in this study were well below the solid solubility limit of Mn
in ZnO. It should also be noted that increasing the Mn content from 3 to 5 at.% resulted in a significant
decrease in the relative magnetization response as is seen in Fig. 15. This provides strong evidence,
albeit indirect, that the magnetization is not due to any precipitating secondary phase. If the formation
of a secondary Mn-related phase was responsible for the ferromagnetic behavior, an increase in Mn
concentration would presumably increase the secondary phase volume fraction and related magne-
tization signature. Instead, the opposite behavior is observed.
One can postulate as to the mechanism by which Sn doping yields ferromagnetic interactions
among the Mn ions. First, the Sn ions may simply provide carriers or bound donor states with extended
wavefunctions that mediate interactions among the Mn ions. The Sn dopants may alternatively form
complexes with the Mn ions, yielding a distribution of Mn
3+
sites. In this case, the mixture of Mn
2+
/
Mn
3+
sites could yield ferrimagnetic ordering through superexchange interaction. With this, the
magnetic moment/Mn ion should increase with Sn doping even if the measured conductivity changes
very little. It should be noted that very little is known about the behavior of Sn in ZnO. Identifying the
location of the energy state associated with Sn in ZnO would be most useful in the interpretation of
magnetic properties when co-doped with Mn. One could also look for shifts in the Sn or Mn dopant
state location as the other dopant is added.
An additional means of elucidating the mechanism for ferromagnetism in ZnO co-doped with a
deep donor is to consider an alternative transition metal. For this, Cr is particularly attractive. First, Cr
provides an opportunity to realize ferromagnetism in ZnO via ions for which the net superexchange
coupling is ferromagnetic. As with Mn, Cr itself is anti-ferromagnetic, thus eliminating any role of Cr
precipitates in yielding spurious ferromagnetism. Third, theory predicts that Cr-doped n-type ZnO
should be ferromagnetic. As such, experiments similar to those described above for ZnO co-doped
with Mn and Sn should be performed with Cr replacing Mn. Ferromagnetism in Co-doped

semiconducting oxides has been reported for several materials, including ZnO and TiO
2
.
The observation that a doubly ionized donor (Sn) leads to ferromagnetic interactions among Mn
atoms in ZnO raises the question as to whether deep states are, in general, effective in mediating
ferromagnetism. Of particular interest are the theoretical predictions that ferromagnetism above room
temperature should be possible in Mn-doped ZnO that is p-type. Given the difficulty in realizing a
shallow acceptor in ZnO, an obvious experiment, given the results for the Mn co-doped with a deep
donor (Sn), is to co-doped Mn with a deep acceptor. Attractive deep acceptors for addressing this
question are Cu and As. Copper doping introduces an acceptor level with an energy $0.17 eV below
the conduction band. There is a large ionic radii mismatch for As (2.22 A
˚
) on the O (1.38 A
˚
) site,
22 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
suggesting limited solid solubility for these anions [116]. Nevertheless, p–n junction-like behavior has
been reported between an n-type ZnO films on GaAs subjected to annealing [117]. In this case, a p-
type layer was reportedly produced at the GaAs/ZnO interface. Further understanding of group V
substitution in ZnO requires the study of doped materials that are free from complications of reactive
substrates or interfacial layers. The primary interest is to investigate whether the deep acceptor states
from Cu or As mediate ferromagnetic interactions in TM-doped ZnO.
For practical application in spintronic devices, the Curie temperature should be well above room
temperature [118,119]. As discussed earlier, theory suggests that the Curie temperature will tend to
increase with decreasing cation mass. In addition, there is phenomenological evidence that T
c
increases with increasing gap. The observed trend is for T
c
to increase with increasing bandgap.
While the equilibrium solid solubility of Mg in wurtzite ZnO is only $2%, epitaxial Zn

1Àx
Mg
x
O thin
films have been realized with x as large as 0.35 [120].
A key requirement in understanding ferromagnetism in transition metal doped semiconductors,
including ZnO, is to delineate whether the magnetism originates from substitutional dopants on cation
sites, or from the formation of a secondary phase that is ferromagnetic. The importance of this issue
cannot be understated. The concept of spintronics based on ferromagnetic semiconductors assumes
that the spin polarization exists in the distribution of semiconductor carriers. Localized magnetic
precipitates might be of interest in nanomagnetics, but is of little utility for semiconductor-based
spintronics. The question of precipitates versus carrier-mediated ferromagnetism is complex, and is a
central topic of discussion for other semiconducting oxides that exhibit ferromagnetism, in particular
the Co-doped TiO
2
system [121,122]. Several issues must be addressed in order to gain insight into the
possible role of secondary phase precipitates in the magnetic properties of transition metal doped
semiconductors, specifically for ZnO films. First, one should identify all candidate magnetic phases
possible from the assemblage of elements. The coincidence of T
c
with a known candidate secondary
ferromagnetic phase indicates a likely source of at least part of the magnetic signature. The UF group
has been investigating a number of TM-doped semiconducting oxide materials in order to understand
ferromagnetism in semiconductors, and to identify promising candidates for spintronics. As an
example, we recently investigated the magnetic properties of Mn-doped Cu
2
O. Cu
2
O is a p-type
semiconductor with a bandgap of 2.0 eVand a hole mobility of 100 cm

2
V
À1
s
À1
. The growth of Mn-
doped Cu
2
O films was achieved via pulsed laser deposition from a Mn-doped Cu-O target. These
epitaxial Cu
2
O films doped with Mn are clearly ferromagnetic with a T
c
of $48 K. However, some
question as to the origin of the ferromagnetism arises when it is recognized that the measured Curie
temperature close to that of Mn
3
O
4
, which has a T
c
of $46 K. Obviously, the simplest explanation for
ferromagnetic behavior in this material is Mn
3
O
4
precipitates. However, there is no evidence for this
phase in X-ray diffraction. Additional work is needed in order to delineate the possible formation of
Mn
3

O
4
in the Cu
2
O films. Nevertheless, knowledge of the candidate ferromagnetic secondary phases
is invaluable in sorting out the ferromagnetic response. Fortuitously, for Mn-doped ZnO:Sn, the only
ferromagnetic secondary phase candidate is the Mn
3
O
4
spinel. None of the other possible secondary
phases involving combinations of Zn, Mn, O, and Sn yield a known ferromagnetic material. The high
temperature ferromagnetism in Mn-implanted ZnO:Sn crystals cannot be attributed to Mn
3
O
4
as the T
c
is much higher in the Mn-doped ZnO than for the Mn
3
O
4
phase.
In addition to TEM, one can also use X-ray diffraction to search for secondary phases within the
films. Despite the obvious sensitivity limitations involved in detecting impurity phases that represent
only 1–5% vol.% of a thin-film sample, we have been able to detect nanoscale precipitates in transition
metal implanted samples. This is demonstrated specifically for another transition metal doped
semiconducting oxide that we have investigated, namely Co-implanted ZnO. It has been reported
that epitaxial (Zn
1Àx

Co
x
)O (x = 0.05–0.25) exhibit high temperature ferromagnetism with a T
c
greater
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 23
than 300 K [123]. The ferromagnetic behavior is assigned to substitutional Co on the Zn site. Monte
Carlo simulations on the indirect exchange interaction of Co-doped ZnO also predicts ferromagnetism
in these materials [124]. SQUID magnetometry measurements of Co-doped ZnO clearly indicate
ferromagnetism. However, a careful examination of the X-ray diffraction u–2u scan along the surface
normal, clearly indicates the presence of Co precipitates. A Co (1 1 0) peak is clearly evident. From the
width of the peak, the size of the cobalt precipitates can be estimated to be approximately 3.6 nm.
Clearly, the possibility of precipitate formation must be carefully considered for each dopant
investigated.
It should be noted that the presence of magnetic precipitates may mask an underlying carrier-
mediated ferromagnetism due to substitutional doping. In this case, a more direct means of measuring
the spin properties of the semiconductor carrier population is needed. Much of the research activity
being presently pursued in spintronics is directed at demonstrating device structures (spin-LED, spin-
FET) that differentiates spin polarization distribution among the electron/hole population in the
semiconductor. Future work would employ these materials in such device structures. Some techniques,
such as magnetic circular dichroism [125] and magneto-optical Kerr effect [126], also yield
information regarding global spin distributions.
The successful realization of most spintronics applications depends critically on the ability to
create spin-polarized charge carriers in a conventional semiconductor in a device structure. This can be
accomplished under ambient conditions via optical pumping with appropriately polarized laser light
[91]. However, ultimate device integration will require electrical spin injection. Electrical spin
injection can be accomplished either by injecting from a spin-polarized source or by spin-filtering
unpolarized carriers at the interface. Despite persistent efforts by many groups, spin injection from a
conventional ferromagnetic metal into a semiconductor has proven highly inefficient. By sharp
contrast, efficient spin injection has recently been successfully demonstrated in all-semiconductor

tunnel diode structures either by using a spin-polarized dilute magnetic semiconductor (DMS) as the
injector or by using a paramagnetic semiconductor under high magnetic field as a spin filter. These
experiments, coupled with the earlier discoveries of ferromagnetic ordering in (Ga,Mn)As at an
unprecedented temperature of 120 K and the long spin coherence length and time in a variety of
semiconductors offer the possibility of a breakthrough in spintronics. In order to implement these
applications, however, DMS with an ordering temperature over 300 K must be synthesized.
6. Ferromagnetism in ZnO nanorods
An SEM is shown of a single rod is shown in Fig. 16(top). The growth time was $2hat4008C.
The typical length of the resultant nanorods was $2 mm, with typical diameters in the range of 15–
30 nm. It is also possible to grow cored nanorods with ZnMgO composition varying across the rod
diameter (Fig. 16 bottom). The samples were subsequently implanted with Mn
+
and Co
+
ions at a fixed
energy of 250 keV and doses of 1–5 Â 10
16
cm
À2
, while the samples were held at $350 8C to avoid
amorphization. The projected range for both types of ion is $1500 A
˚
, with peak transition metal
concentrations corresponding to roughly 1–5 at.%. The samples were then annealed at 700 8C for
5 min under flowing N
2
to promote migration of the transition metal ions into substitutional positions.
Even at this highest dose condition, the nanorods are stable to the implant/anneal cycle. A comparison
of micrographs before and after this cycle did not show any observable change in the nanorods.
Fig. 17 shows the magnetization versus field behavior at 300 K for a Mn-implanted

(5 Â 10
16
cm
À2
dose) nanorod sample after the 700 8C, 5 min anneal. Hysteretic behavior is clearly
present, with a coercive field at 100 K of 100 Oe. The possible explanations for this data include
24 Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47
ferromagnetism, superparamagnetism or spin-glass effects [91,127]. The magnetization of unim-
planted, annealed nanorods was as much as three orders of magnitude lower than in the implanted,
annealed samples, demonstrating that the transition metals are responsible for the observed magnetic
properties. The calculated moment from the saturation magnetization was $2.2 Bohr magnetons per
Mn ion, showing that a significant fraction of the implanted Mn is contributing to the magnetization.
None of the potential second phases can account for the observed magnetic behavior. Thus, for
Mn-implanted nanorods, it does not appear that secondary phases plays a significant role in the
magnetic properties. Indeed, the magnetization results for the Mn-implanted nanorods are very similar
to those reported previously for Mn-implanted, bulk n-type ZnO single crystals, in which high-
resolution X-ray diffraction did not observe any secondary phases. We note also that solubility limits
Y.W. Heo et al. / Materials Science and Engineering R 47 (2004) 1–47 25
Fig. 16. TEM image of single-crystal ZnO nanorod(top) and TEM micrographs showing cored (Zn
1Àx
Mg
x
)O nanorods
having Zn-rich phase surrounded by another (Zn
1Àx
Mg
x
)O phase (bottom).