Nanoscale

View Article Online

PAPER View Journal | View Issue

Doping of wide-bandgap titanium-dioxide

Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.

nanotubes: optical, electronic and magnetic

Cite this: Nanoscale, 2014, 6, 10839
properties†
Yahya Alivov,*a Vivek Singh,a Yuchen Ding,a Logan Jerome Cerkovnika
and Prashant Nagpal*abc

Doping semiconductors is an important step for their technological application. While doping bulk
semiconductors can be easily achieved, incorporating dopants in semiconductor nanostructures has
proven difficult. Here, we report a facile synthesis method for doping titanium-dioxide (TiO2) nanotubes
that was enabled by a new electrochemical cell design. A variety of optical, electronic and magnetic
dopants were incorporated into the hollow nanotubes, and from detailed studies it is shown that the
doping level can be easily tuned from low to heavily-doped semiconductors. Using desired dopants –
Received 5th May 2014
Accepted 7th July 2014
electronic (p- or n-doped), optical (ultraviolet bandgap to infrared absorption in co-doped nanotubes),
and magnetic (from paramagnetic to ferromagnetic) properties can be tailored, and these
DOI: 10.1039/c4nr02417f
technologically important nanotubes can be useful for a variety of applications in photovoltaics, display
www.rsc.org/nanoscale technologies, photocatalysis, and spintronic applications.

(carbon nanotubes, activated carbons, and graphene) which

Introduction
Titanium-dioxide (TiO2) is a wide-bandgap semiconductor that
is a key component of devices like solar cells, photocatalytic
reactors, photoelectrochemical cells, etc.1–4 While doping (elec-
tronic, optical, and magnetic) allows careful tuning of desired
properties, oxygen vacancies formed during the synthesis are
used as n-type dopants in these nominally “undoped” oxides.
Since desired doping of these semiconductor nanostructures
has proved challenging, most applications are focused on bulk
semiconductor lms.5–7 This can severely limit their applications
and device architectures, like fabricating p-type oxide lms or
changing the doping level of n-type dopants for optimized
conguration.1,3 Recent research in utilizing heavily-doped
semiconductors for low-loss plasmonics8,9 has also been gener-
ating interest in doping of these TiO2 nanostructures, already
used in thin lm solar-cells, for enhanced light harvesting.
Moreover, addition of desired dopants, like shallow or deep
donors, can also lead to the formation of heavily-doped trans-
parent oxide nanostructures, or co-doped photocatalysts10,11
which can absorb infrared light. Other efforts to improve the
performance of TiO2 photocatalytic activity12–19 involves hybrid
composites consisting of TiO2 and carbonaceous materials
cause a red-shi in the absorption spectrum and increase elec-
tron mobility.20–23 Since doping can simultaneously tune the
bandgap and increase charge conductivity, it can provide an
important alternative. Therefore, desired doping of these wide-
bandgap nanostructured lms can provide important materials
for a variety of applications in renewable energy, articial
displays, and other optoelectronic and magnetic processes.
One of the synthesis methods for the growth of TiO2 nano-
tubes is electrochemical oxidation, also termed as anodiza-
tion.24–28 Since this synthesis provides a facile route towards
large-scale fabrication of TiO2 nanotubes, simply incorporating
desired dopants during the growth can provide an important
method for the synthesis of doped wide-bandgap semi-
conductor nanostructures. However, simple addition of the
desired concentration of precursors, especially cationic, does
not lead to doping of nanotubes (see Experimental section).
Therefore, we designed a new electrochemical cell that enables
doping and co-doping of TiO2-nanotubes with desired cations
and anions, and characterized their resulting optical, electronic
and magnetic properties, to demonstrate the feasibility of this
new method. Using detailed studies, we demonstrate that this
method can provide an important route for the synthesis of
doped wide-bandgap TiO2 semiconductor nanotubes.

a
Department of Chemical and Biological Engineering, University of Colorado, Boulder,
USA. E-mail: y.alivov@colorado.edu; pnagpal@colorado.edu Experimental
b
Renewable and Sustainable Energy Institute, University of Colorado, Boulder, USA
The method for TiO2 nanotube growth by anodization
c
Materials Science and Engineering, University of Colorado, Boulder, USA
† Electronic supplementary information (ESI) available: See DOI: In a regular setting of an electrochemical cell, a negative-biased
10.1039/c4nr02417f platinum (Pt) electrode and positive-biased titanium (Ti) sheet

This journal is © The Royal Society of Chemistry 2014 Nanoscale, 2014, 6, 10839–10849 | 10839

Nanoscale
are immersed in the electrolyte 2–3 cm apart.24–28 In this
conguration negative ions in the electrolyte move toward the
positive Ti sheet and positive ions move toward the negative Pt
sheet. Addition of desired dopant metal cations into the elec-
trolyte results in the repulsion of the positively charged ions
from the positively biased Ti sheet (away from the growing TiO2
nanotubes). This prevents TiO2 nanotubes from doping with
metal cations.
Replacing the DC power supply with AC can provide a solu-
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
tion. During the “negative” cycle (switched polarities) of the
electrochemical growth, the desired cations are attracted by the
Ti sheet, thus incorporating cationic dopants uniformly in the
growing TiO2 nanotubes (Fig. 1a). However, simple addition of
cationic precursors in the growth solution leads to other prob-
lems. First, the added cations (especially in high doping
concentrations) dramatically increase conductivity, making it
impossible to apply high enough voltage to enable TiO2 nano-
tube growth (voltages higher than 5 V need to be applied to
initiate nanotube growth24–28). Second, during the positive bias
on the Ti sheet (negative bias on the Pt counter electrode),
cations deposit on the Pt electrode, rapidly covering it with a
reduced metal layer, quickly degrading its performance.
To overcome these problems we modied the electro-
chemical cell. The electrochemical cell was divided into two
sections, separated by a porous membrane (Fig. 1a). One part of
the cell contains the Pt electrode and is lled with an electrolyte
free of the dopant precursor. The second part of the cell contains
the Ti sheet and is lled with an electrolyte that contains the
dopant precursor. The porous membrane provides a barrier for
two different parts of the electrochemical cell to prevent from
mixing, while still allowing a ow of the electrolyte. The resulting
Fig. 1 (a) A sketch showing the design of a new electrochemical cell
developed in this work. It is divided into two parts separated with a
porous membrane to prevent from high currents and cation deposi-
tion onto the Pt electrode. (b) Schematic illustrating rectangular shape
AC signals with parameters employed for doping nanotubes.
10840 | Nanoscale, 2014, 6, 10839–10849
View Article Online
Paper
currents in this conguration are comparable to the currents in
regular electrochemical cells with the DC power supply, ranging
from 0.1–4.0 mA within the voltage range of 7–200 V. For the
separating membrane, we used a hydrophilic polymer lter
(pore size
0.1–0.4 mm). This membrane was attached in the
middle of the container using epoxy. While the membrane
provides a diffusion barrier, some cations move toward the Pt
electrode and over time the conductivity of the dopant-free
electrolyte will increase. The electrolyte should be changed if the
resistivity decreases by 50%.
Using an AC voltage source, rectangular shaped power pulses
were used in this present study. The shape of the AC voltage and
its parameters are shown in schematic Fig. 1b. The period of AC
voltage is characterized by the positive part V1, the negative part
1
V2, and the frequency f, which is equal to f ¼ , where t1
t1 þ t2
and t2 are durations for positive and negative parts of the cycle,
respectively. These parameters – t1 and t2 control the growth
and doping times, respectively. From a series of detailed
experiments, it was found that the dopant level, or dopant
concentration n, mostly depended on the negative bias V2,
frequency f, t1/t2 ratio (see Fig. 1b), and dopant precursor
amount b in the electrolyte. Our studies showed that the
dependence of n on f and t1/t2 at xed V2 and b is convoluted,
depending on a number of factors. Therefore, we studied the
dependence of doping (n) on each of these parameters sepa-
rately, keeping the other parameters constant. The dependence
of n on V2 (when other parameters were xed) was more
predictable; however, a strong sublinear relationship between
these two parameters was observed. The dependence between n
and the amount of precursor b in the electrolyte was nearly
linear (see Discussion below). Therefore, variation of the dopant
concentration was used for systematically controlling the
doping level in fabricated TiO2 nanotubes.
The electrolyte used in electrochemical growth solution
consisted of a solvent (glycerol or ethylene glycol) with 1%
ammonium uoride (NH3F), and 2% of water. As the starting
material commercial Ti sheets of 99.99% purity were used. The
growth rate of nanotubes in ethylene glycol was
4 mm per hour.
All as-grown samples were amorphous and annealed at 500  C
for 1 hour in air, to convert to the anatase phase. Three main
types of cationic dopants were used in this study – niobium (Nb),
iron (Fe), and copper (Cu). Among the prominent anion dopants,
we used nitrogen (N), while several combinations of co-doped
anion and cation co-dopants were prepared. As precursors nio-
bium(V) chloride, iron(III) chloride, and copper(II) sulfate were
used. Nitrogen doping was performed by using hexamethylene-
tetramine in the electrolyte. The nitrogen doping level was
controlled by changing the amount of the precursor (hexa-
methylenetetramine) in the electrolyte. While this method can
be applied to any cation, we focused on Nb, Fe, and Cu due to
their importance for a variety of applications. For example, Nb
was shown to be a shallow donor for TiO2, and therefore, a good
candidate for the growth transparent conducting oxide thin
lms.29–32 Fe is a good dopant for fabricating magnetic semi-
conductors.33–35 Cu and Nb co-doped TiO2 thin lms with N have
been shown to increase the photocatalytic activity of TiO2.36
This journal is © The Royal Society of Chemistry 2014

Paper
The morphology of grown TiO2 nanotubes was examined
using a eld-emission scanning electron microscope (FE-SEM)
JEOL 7401F. The crystal structures of doped and undoped TiO2
nanotube samples were analyzed by X-ray diffraction (XRD)
measurements, using a Scintag XDS 2000 X-ray diffractometer.
The measurements were performed using Cu Ka radiation at 45
kV and 40 mA. Compositional analysis was performed by energy
dispersive X-ray spectroscopy (EDS) embedded in an FE-SEM
JEOL 7401F instrument. The analysis was performed in
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
different modes (2D mapping, point scan, etc.) to test for
uniformity and possible secondary phases. The spatial resolu-
tion during EDS was limited by the focused spot size of the
electron beam, which was
10 nm.
Electrical characterization. I–V and I–V–T characterizations
were performed in two different ways: (1) from single nanotubes
using CS-AFM; and (2) ensemble measurements using a Keith-
ley source meter (Keithley 2612A, Tektronix Inc.). Temperature
dependent current–voltage characterization (I–V–T) was per-
formed in the temperature range of 20–300 K with 10 K
temperature steps, using a closed loop Helium Cryostat (ARS-
202AE with an ARS-2HW Helium compressor, Advanced
Research Systems Inc.). The voltage for these measurements
was varied in the range from 20 V to +20 V. Sheet resistance of
nanotubes R was calculated from I–V measurements using
V glass vial cut in half with a 0.64 cm2 cross-sectional area. The
Ohm's law I ¼ . Then, resistivity r was calculated using the
R vial was then enclosed in a 48 mL reactor and purged for 45
L minutes with CO2 (75 cm3 min1) humidied in a bubbler lled
equation R ¼ r , where L is the nanotube length, and A – the
A with D.I. water. Aer purging, the reactor was closed and irra-
total nanotube cross-section area. The cross-section area A was
estimated using the density of nanotubes, thickness of the
nanotube wall, and inner and outer diameters of the nanotube,
which were measured using SEM images.
Photoresponse measurements. Spectral photo I–V measure-
ments were performed in the wavelength range of 300 –1000
nm. Monochromatic light was obtained using a mono-
chromator from a light source. Photocurrent was calibrated to
the incident power of monochromatic light measured using an
Si-based detector (NIST calibration). The intensity of mono-
chromatic light was measured using an optical power meter
(Newport, 1830-C).
STM and STS measurements. Scanning tunneling micro-
scope images were obtained using a customized Molecular
Imaging PicoScan 2500 setup (with a PicoSPM II controller). An
STM nosecone (N9533A series, Agilent Technologies) was used
for scanning and spectroscopy, using chemically etched Pt–Ir
tips (80 : 20) purchased from Agilent Technologies, USA. The
measurements were done at room temperature under atmo-
spheric conditions. Tunneling junction parameters were set at
tunneling currents ranging between 100 and 500 pA and a
sample bias voltage between 5 and +5 V. Spectroscopy
measurements were obtained at a scan rate of 1 V s1. For STM
measurements, the pre-amp sensitivity was set to 1 nA V1. The
tunneling current as a function of the applied bias voltage (STS)
was recorded at multiple positions on the sample at room
temperature and atmospheric pressure.
SQUID measurements. The magnetic properties were
studied with a Superconducting Quantum Interference Device
This journal is © The Royal Society of Chemistry 2014
View Article Online
Nanoscale
magnetometer (SQUID). Magnetization measurements were
performed using a Quantum Design SQUID-VSM (Vibrating
Sample Magnetometer), and magnetization (M) was measured
as a function of the applied eld (H) and temperature (T) in the
magnetic eld range from 7 to 7 Tesla and a temperature
range of 1.8 to 300 K. Temperature dependent magnetization
(M–T) measurement was performed at the applied magnetic
eld of 1000 Oe. TiO2 nanotube materials for SQUID studies
were collected from the surface of a Ti sheet aer the growth.
The eld-cooled (FC) and zero-eld-cooled (ZFC) magnetization
measurements are performed from 5 to 300 K. TiO2 nanotube
samples were scraped from the Ti sheet aer the TiO2 nanotube
growth. Therefore, the scraped sample was a mixture of
randomly oriented nanotubes. As a result, induced magnetiza-
tion of the TiO2 nanotube sample was averaged from all nano-
tube orientations, and therefore, it was isotropic with respect to
the magnetic eld direction. Extreme precautions were taken
during these measurements to avoid any contamination of the
samples with possible magnetic dopants. Magnetization of the
sample holder was also measured and subtracted from that of
TiO2 nanotube samples to calculate the pure magnetization of
doped and undoped TiO2 nanotubes.
Photocatalysis – Gas-phase reduction of CO2 and H2O.
Twenty to thirty mg of the catalyst was deposited in a rounded
diated with 1 Sun (100 mW cm2) through a transparent
window using a solar simulator (ABET Technologies). One-
milliliter samples were extracted from the reactor and injected
into a gas chromatograph (GC) equipped with a thermal
conductivity detector (TCD) and a ame ionization detector
(FID) for measuring concentrations. Separation of the hydro-
carbons was done with a Hayesep D column or with Silica-gel
column. Reported data of the photocatalytic rate and quantum
yield were calculated based on the electron ux (mmol cm2 h1)
used to form H2, CH4, C2H6, hydrocarbons and others.
Results and discussion
SEM, XRD, STM, and absorption
Fig. 2 presents SEM images of doped TiO2 nanotubes fabricated
using our new electrochemical cell. The electron micrographs
clearly show that the tubular structure of these TiO2 nanotubes
is well preserved aer doping. Fig. S1 in the ESI† shows SEM
images of different diameter nanotubes grown at different
anodization voltages. XRD patterns of all doped samples shown
in Fig. 3a present X-ray diffraction peaks corresponding to the
anatase phase, with predominant peaks (101) and (200) at 2q ¼
25.3 and 47.95 , respectively.37 The XRD pattern of samples
does not change even with a heavy doping level up to 12%,
demonstrating the uniform incorporation of dopants and the
absence of secondary phases. At such high doping levels (>9%),
they can also be regarded as solid solutions. As an illustration in
Fig. S2a,† we show XRD patterns of Nb doped TiO2 nanotubes
with a doping level up to 12%, which shows only reection
Nanoscale, 2014, 6, 10839–10849 | 10841

Nanoscale
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
Fig. 2 Representative scanning electron micrographs (SEM) of TiO2 nanotube arrays grown by electrochemical anodization: (a) shows the top
view of densely packed nanotubes, (b) shows the bottom view after removing nanotubes from the Ti metal, and the side view of the vertical,
hollow, doped TiO2 nanotubes, and (c) and (d) show side profiles at different magnifications.
peaks corresponding to anatase TiO2 without any detectable
secondary phase. No shi of diffraction peaks was observed
with the increase of Nb concentration up to 12%, as shown in
Fig. S2a.† This result could be explained by close ionic radii of
Ti4+ and Nb5+, which are 0.605 Å and 0.64 Å, respectively.38 For
such similar lattice parameters, no signicant distortion of the
crystal lattice was expected. Energy dispersive X-ray spectros-
copy (EDS) conrmed the presence of doped elements in TiO2
nanotubes as shown in Fig. 3b. EDS analysis of undoped TiO2
nanotubes showed no detectable emission peaks from impuri-
ties (only Ti and O emission peaks were seen, Fig.3b, the red
curve). This indicates that the impurity concentration was
below the detection limit (<0.01%). The amount of incorporated
dopants in TiO2 nanotubes follows a linear relationship with
the concentration of the precursor in the electrolyte. To
demonstrate this, we show EDS spectra of Nb doped samples at
different NbCl5 wts% in the electrolyte (Fig. S2b†), from which a
nearly linear relationship between the detected EDS Nb signal
in TiO2 nanotubes and the wt% NbCl5 precursor is clearly seen
(Fig. 3c). The R2 value of the linear tting was calculated to be
0.995. Similar studies for other dopants (Nb, Fe, Cu, and N) also
showed a linear relationship between the incorporated dopant
and the dopant concentration in the electrolyte. The conduc-
tivity of Nb-doped TiO2 nanotubes increases (resistivity
decreases) with the wt% NbCl5 precursor (Fig. 3c), conrming
the incorporation of Nb into the TiO2 crystal. Nb is a well-known
shallow donor in TiO2 used for creating transparent conducting
thin lms.39,40 EDS point scans were also taken from different
points on individual doped TiO2 nanotubes. This scan was
performed by pointing a focused electron beam onto the surface
of a single nanotube (or mapping the elemental distribution
over a large area). Fig. S3† presents results for Cu-doped
samples. Highly pronounced Cu peaks seen at different points
(we are showing two points for clarity) indicate Cu is uniformly
incorporated in TiO2 nanotube crystals, rather than accumu-
lating as clusters or secondary phase oxides. These data further
10842 | Nanoscale, 2014, 6, 10839–10849
View Article Online
Paper
demonstrate that dopants (Cu here) were uniformly incorpo-
rated in the TiO2 crystal lattice. Similar scans for all dopants
(cations and anions) were performed to ensure uniform doping
in TiO2 nanotubes.
Absorbance spectra, of undoped and different doped
nanotubes are shown in Fig. 4a. The optical spectrum does
not reveal either the nature of the dopant, or the direction of
the shi of the Fermi-level (which was analyzed by Scanning
Tunnelling Spectroscopy (STS), as shown in Fig. 4b) towards
the conduction or valence band. But optical spectroscopy was
used to monitor the semiconductor bandgap by changing the
dopant and the concentration of the dopant in the TiO2
nanotube. Moreover, the electronic level of the dopant was
also estimated, since the Fermi-level (and the effective optical
bandgap) is also pinned at high dopant concentrations.
However, this analysis is complicated by the indirect
bandgap of the TiO2 semiconductor, large Rayleigh scat-
tering (f1/l4) observed in the nanotubes (shown as a black
dashed curve in Fig. 4a), and the presence of impurity donor
states (below the bandgap) in undoped TiO2. The band gaps
estimated from their respective optical spectra (aer
considering scattering) were 2.7 eV (due to impurity donors),
2.6 eV, 2.5 eV, 2.4 eV, 2.3 eV, 2.2 eV, and 2.0 eV for undoped,
Nb, Cu, Fe, N, Nb–N, and Cu–N doped and co-doped samples,
respectively (Fig. 4a). These bandgaps were in good agree-
ment with the bandgaps derived from STS (Fig. 4b), which
uses the xed Fermi-level of the Pt–Ir tip (at 0 V) as a probe of
the shi of the Fermi-level with doping. To further evaluate
the role played by these dopants on optoelectronic proper-
ties, and to ensure that the dopants are properly incorporated
in the TiO2 crystal (and not as recombination sites or traps on
the surface), we evaluated the photoresponse of these doped
nanotubes (Fig. 4c). We illuminated these lms with a small
ux of photons (to identify possible recombination from
dopants as traps), and observed the change in current (from
the dark) due to photogenerated charge carriers, normalized
This journal is © The Royal Society of Chemistry 2014

Paper
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
Fig. 3 (a). XRD patterns for undoped, Nb, Cu, Fe, and N doped, and
Nb/N and Cu/N co-doped and undoped samples. Labels on the right
guide to the corresponding samples. Only reflection peaks corre-
sponding to the anatase TiO2 phase were detected in all samples
indicating the lack of secondary phases. (b) Energy dispersive spec-
troscopy (EDS) for undoped, Fe-doped, Cu-doped, Nb-doped, and
N-doped, Nb and N co-doped, and Cu and N co-doped TiO2 nano-
tubes. Well pronounced characteristic X-ray emission peaks of
dopants can be seen along with the TiO2 matrix (Ti and O peaks). These
data prove the presence of desired dopants in TiO2 nanotube films; (c)
Nb concentration (blue) and resistivity (red) of Nb doped TiO2 nano-
tubes, as a function of wt% NbCl5 in the electrolyte.
Ilight  Idark
by the intensity of light. The photoresponse Iph ¼
Plight
for highly conductive samples (Fig. 4c) was within the error
bar for the measurement, due to the high concentration of
background electrons (photogenerated charge carriers are
much less than equilibrium carrier concentrations).
However, the acceptor dopants (Cu, N, and Cu–N) and
undoped samples showed good transport for photogenerated
charges in these devices, following exciton dissociation and
charge transport in these doped nanotubes. Spectrally
resolved normalized photocurrent (Fig. 4c) mimics the
absorption spectra of the respective nanotubes, shown in
Fig. 4a, with similar bandgaps. This was expected since the
photocurrent is limited by the thin lm absorption of
incident light.
This journal is © The Royal Society of Chemistry 2014
View Article Online
Nanoscale
Fig. 4 (a) STS spectrum of (a) undoped, (b) Cu-doped, (c) N-doped, (d)
Nb-doped, (e) Nb–N co-doped, and (f) Cu–N co-doped samples. (b)
Extinction spectra of undoped, monodoped and co-doped TiO2
samples. A clear change in the bandgap can be seen using the STS and
the optical spectrum of these doped wide-bandgap semiconductor
nanotubes. The dashed-black line represents the large Rayleigh
scattering observed in these nanotubes, which makes exact determi-
nation of bandgaps difficult in these indirect bandgap semiconductors,
using optical spectroscopy; (c) spectrally resolved photoresponse
(photocurrent normalized by incident light intensity) for undoped
(blue), Cu-doped (pink), N-doped (green), and Cu–N co-doped (red)
TiO2 nanotubes.
I–V, I–V–T, and photo-I–V
We also conducted current–voltage spectroscopy (I–V), on all
undoped and doped nanotubes, using ensemble and single
nanotube current sensing AFM (CS-AFM) measurements
(Fig. 5a and 7). The dopant concentrations for all cations and
anions were kept constant for all samples in Fig. 5a, such that
the differences in the charge carrier concentrations (electrons
or holes) arise from the number of ionized dopants
 
Ei
(n ¼ n0 exp  ), which depends on the dopant electronic
kT
level Ei and the available energy at room temperature (kBT ¼ 25
Nanoscale, 2014, 6, 10839–10849 | 10843
 View Article Online

Nanoscale Paper
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.

Fig. 5 Electronic n- and p-type doping of TiO2 nanotubes: (a) room-

temperature current–voltage (I–V) spectrum of TiO2 nanotubes doped
with 1 wt% of (a) Copper (Cu), (b) Copper–Nitrogen (Cu–N) co-dopant,
(c) Nitrogen (N), (d) Niobium–Nitrogen (Nb–N) co-dopant, (e) undoped,
(f) Iron (Fe), and (g) Niobium (Nb); (b) Temperature dependent current–
voltage characteristics presented in 3D mode; (b), conductance (I/V) of
Nb-doped (curve 1), Fe-doped (curve 2), undoped (curve 3), and Cu-
doped (curve 4) samples as a function of 1/kT. The slope at the higher Fig. 6 (a) Conductivity (green), carrier concentration (red), and
temperature region (>160 K) revealed activation energies 0.018 eV, mobility (blue) of N-doped samples as a function of N concentration.
0.037 eV, 0.093 eV, and 0.502 eV for Nb-doped, undoped, Fe-doped, (b) Scanning tunneling spectroscopy (STS) data for nitrogen-doped
and Cu-doped samples, respectively; (c) CS-AFM I–V curve and the TiO2 nanotubes, showing a clear shift of the Fermi-level (at 0 eV)
corresponding ln(I)–V plot for undoped TiO2 nanotubes to illustrate the towards the valence band (p-type doping).
calculation method for carrier concentration; (d) STS spectra for Nb-
doped and N-doped TiO2 nanotube; dashed line guides to the Fermi-
level (constant for our Pt–Ir tip) which is located near the bottom of the
conduction band for the Nb-doped sample (n-type), and near the top of conduction in these doped nanostructures, and further
the valence band for the nitrogen doped sample (p-type). evidence for electronic doping of TiO2 wide-bandgap nanotubes
using cationic and anionic dopants.

meV, where kB is the Boltzmann constant and T is the temper-

ature). To further understand ionization of dopants to generate
doped charge carriers, which determines the electronic property
of doped semiconductors, temperature dependent I–V charac-
teristics (I–V–T) of samples were measured. We plotted the
conductivity s ¼ ln(I/V) as a function of temperature T (Fig. 5b),
to measure the activation barrier for charge transport (or the
energy required for ionizing charge carriers in a doped semi-
conductor). Activation energies (Ea) for undoped and Nb, Fe,
and Cu doped samples were found to be 0.027 eV, 0.018 eV,
0.093 eV, and 0.502 eV, respectively (Fig. 5b and Table 1). The
activation energy for Nb-doped TiO2 nanotubes was close to the
reported ionization energy values for Nb donors in anatase TiO2
thin lms and bulk crystals, which was in the range of 10–40
meV.39,40 These values are close to the values derived using
Richardson plots (ln(I/T2) vs. 1/kT) (Fig. S5†). The mobility of
electron doped TiO2 nanotubes – oxygen vacancies (undoped),
Fe-doped, and Nb-doped; was very close with values 4.3, 2.7, and
3.8 cm2 V1 s1, respectively (Table 1). In this case the differ-
ences in conductivity of donor doped samples can be accounted
due to the difference in ionized donors. In the case of Cu, N, and
Cu–N dopants a drastic drop in mobility was observed (Table 1) Fig. 7 (a) I–V characteristics of low conductivity Cu-doped (1 wt%)
TiO2 nanotubes are well described by the typical space charge limited
that can be explained by the contribution of holes to transport
current (SCLC) model (I–V2 at higher bias). (b) Modified Fowler–
properties, since the hole mobility in TiO2 is lower than the Nordheim plot (explained in the text) for photogenerated electrons in
electron mobility.41 These ndings provide insights into charge Cu-doped TiO2 nanotubes.

10844 | Nanoscale, 2014, 6, 10839–10849 This journal is © The Royal Society of Chemistry 2014

Paper
Table 1 Summary of electrical parameters of doped TiO2 nanotubes
Undoped Fe
Resistivity, r, U cm 6.92 2.97
Carrier concentration, n, cm3 8.9  1017 2.5  1018
Mobility, m, cm2 V1 s1 4.3 2.7
Activation energy (eV) 0.027 0.093
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
We analyzed the transport properties of doped TiO2 nano-
tubes, such as carrier concentration (n) and mobility (m) for
these nanotubes, using single nanotube CS-AFM measure-
ments. We used a conductive gold coated AFM tip (known cross-
section of the contact), with a single nanotube (Schottky-barrier)
and measured the current–voltage (I–V) curves from a number
of nanotubes. We used the I–V curves (Fig. 5c) in the interme-
diate bias regime (where the reverse-biased Schottky barrier
dominates the total current, Fig. S6, details in the ESI†):42
 


e 1
ln I ¼ ln S þ  V þ lnJs (1)
kT E0
where J is the current density, S is the contact area, E0 is a
parameter depending on carrier concentration n, e is the
elementary charge, k – the Boltzmann constant, and Js is a
function of the applied bias (see ESI†). This logarithmic plot of
the current I as a function of the bias V gives approximately a
straight line, the slope of which is equal to e/(kT)  1/E0.
The electron mobility can be calculated using the relation
m ¼ 1/(ner), with r being the resistivity of the nanotubes. Table 1
summarizes the extracted values of resistivity, carrier concen-
tration, and mobility for doped, co-doped, and undoped
samples. The resistivity r of the samples was calculated using
sheet resistance R determined using the Ohmic law (see
Experimental section). There was a strong effect of dopants on
electrical properties, and the resistivity ranged from 6  104
(Fig. 9a) to 3  106 U cm (Table 1). The resistivity of the undoped
sample was 9.92 U cm. As-grown undoped TiO2 nanotubes were
of the n-type conductivity due to oxygen vacancies acting as
shallow donors.2,15 The n-type conductivity of as-grown TiO2
nanotubes is seen from the STM spectrum (Fig. 5d), where the
density of states (DOS) is seen near the conduction band and
the Fermi energy is also close to the conduction band. Nb doped
samples revealed the lowest resistivity 0.2 U cm. The increased
conductivity of Nb-doped TiO2 nanotubes can be explained by
the replacement of Ti atoms by Nb atoms in the crystal lattice.
Nb is known as a good shallow donor in the TiO2 crystal lattice,
with the activation energy in the range of 10–50 meV.29–32 The
presence of shallow donor states near the conduction band (CB)
of Nb doped TiO2 nanotubes samples was also conrmed by the
scanning tunneling spectrum, which showed a high density of
states near the conduction band (Fig. 4b). The iron (Fe)-doped
TiO2 sample also increased the conductivity, compared to the
undoped sample, which indicates donor-like substitution of
this dopant iron. This effect of Fe is less dramatic compared to
Nb, which can be explained by the higher activation energy for
Fe (0.097 eV) compared to Nb (0.018 eV). There have been few
reports on electrical properties of Fe-doped TiO2 nanotubes and
This journal is © The Royal Society of Chemistry 2014
View Article Online
Nanoscale
Nb Cu N Cu–N
0.201 906 2300 4.4  106
5  1019 1.6  1016 2.8  1016 1.3  1015
3.8 0.3 0.097 0.002
0.014 0.502 0.14
the available reports indicate that substitutional Fe produces
n-type conductivity.43–45 Roldan et al. showed, using periodic
density functional calculations, that the Fe-dopant stimulates
the generation of oxygen vacancies that act like shallow
donors.45 Nitrogen doped nanotubes revealed increased resis-
tivity (Table 1). Similar to cation dopants, doping of N anions
was also in good linear relationship with the added N precursor
in the electrolyte, with an R2 tting value of 0.988 (Fig. S7a†).
Nitrogen is known as a good acceptor in TiO2, with an energy
position of 0.14 eV above the top of the valence band.46 The
acceptor nature of the nitrogen dopant in our samples is
conrmed from STS studies, where the spectrum shows that the
Fermi level shis toward the valence band, indicating a p-type
effect of these dopants (Fig. 5d). By contrast, the Fermi-level of
Nb doped samples shis towards the conduction band
(compared with the Pt–Ir tip level at zero). The p-type behavior
of nitrogen-doped TiO2 nanotubes was also observed from
electrical characterization, which showed a clear injection of
holes, as opposed to the electron injection in n-type Nb-doped
TiO2 nanotubes (Fig. S7b†). Since nominally undoped nano-
tubes behave as n-doped semiconductors, the addition of an
increasing amount of p-type dopants (like nitrogen) in nano-
tubes rst results in a decrease in n-type conductivity ((from 0 to
1 wt%, Fig. 6a), followed by conversion to the p-type and an
increase in p-type conduction, as shown in Fig. 6a. The
conductivity rst decreases and then increases, with an increase
of N concentration (m) in TiO2, reaching a minimum at m ¼ 2%,
as shown. These phenomena can be easily explained by the shi
in the Fermi-level of the semiconductor with increasing N
concentration, monitored by the tunneling spectra using Pt–Ir
STM tips (the tip Fermi-level is xed). A clear shi of the Fermi-
level towards the valence band was observed with an increase in
N concentration, indicating an increasing p-type character for
doped TiO2 nanotubes shown in Fig. 6a.
While most dopants described so far have behaved as
substitutional dopants, copper incorporated in the TiO2 nano-
tube lattice shows an interesting behavior. A copper dopant also
behaves like an acceptor, leading to an enhancement of the
initial resistivity of Cu doped TiO2 nanotubes (Fig. 4a and b).
The resistivity of copper (Cu) doped samples increases (Fig. 5a
and Table 1), and can be explained by the creation of acceptor
centers that compensate for background impurity donors in a
nominally undoped sample. This behavior is surprising since
the substitution of the titanium cation (Ti4+) with copper (Cu+ or
Cu2+) is expected to show n-type doping.36 However, copper
doped nanotubes show clear p-type conductivity (conrmed
using STS spectra shown in Fig. 4b). This result can be
explained by interstitial doping of copper in TiO2 nanotubes,
Nanoscale, 2014, 6, 10839–10849 | 10845
 View Article Online

Nanoscale Paper

since it has been shown that Cu atoms can act as deep acceptors
when they are in interstitial positions in the crystal lattice.36
Therefore, Cu likely occupies interstitial positions in the TiO2
crystal lattice (interstitial dopant) in doped nanotubes fabri-
cated using our electrochemical cell. I–V characteristics of Cu
doped TiO2 nanotube samples at higher voltages showed a clear
nonlinear behavior of the form I–V2 (Fig. 7a), which can be
explained by the space-charge limited current model (SCLC).47
In this SCLC region the carrier concentration can be calculated
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.

3Vc
using equation n ¼ 2 ,47 where 3 is the dielectric constant, R- is
eR
the radius of the nanotubes, and Vc is the crossover voltage
presenting a transition from Ohmic to SCLC regimes. Extracting
charge carrier concentration using the SCLC model for Cu
doped TiO2 nanotubes shows a concentration of 5  1015 cm3,
which is very close to the value of 1.3  1016 cm3 that was
calculated using eqn (1). The Cu-dopant (Cu2+ precursor) exists
as the Cu2+ (thermodynamically stable) ion, as conrmed by the
electronic measurements of the number of charge carriers
added per dopant atom. To further evaluate the charge
conduction mechanism, we developed a modied Fowler–
Nordheim (F–N) plot for photogenerated charges. For identi-
fying different regimes of the charge transport and acceptor (or
donor) level below the bandgap, instead of the F–N plot of
 
Iphoto;l  Idark
ln(I/V2) vs. 1/V, we plotted V  ln 2 as a func-
V  Plight  Abs
tion of photon energy (Fig. 7b), where Plight is the incident light
intensity, and Abs is the absorption coefficient (for details, see
ESI Fig. S8†). The data for copper doped nanotubes showed a
photocurrent of
0.5–0.7 eV below the nominal TiO2 bandgap, Fig. 8 (a) Room-temperature magnetization (M) vs. magnetic field (H)
indicating the deep acceptor level for copper. These measure- hysteresis loop for Fe-doped TiO2 nanotube samples; (b) magnetiza-
ments are consistent with the measurements for activation tion–magnetic field loop for undoped TiO2 nanotube samples
energy from I–V–T and STM measurements (Fig. 4b and 5). measured using SQUID, which show clear paramagnetic behavior; (c)
presents zero-field cooled (ZFC) and field cooled (FC) temperature
Moreover, this indicates that even at low uence, photo-
dependent magnetization curves for the Fe-doped sample, which
generated charges are not trapped at the copper dopant sites demonstrates that the ferromagnetic behavior of Fe-doped nanotubes
(interstitial acceptor). These measurements clearly show that results from substitutional Fe magnetic dopants, rather than secondary
the dopant atoms are likely to be incorporated in the lattice, and phases.
contribute towards imparting the desired optical and electronic
properties to doped TiO2 nanotubes.
was parallel or perpendicular to nanotube axes49). For compar-
ison the M–H plot for undoped TiO2 nanotubes is also shown at
Magnetic properties 300 K, which shows nearly linear dependence indicating para-
Magnetic properties of Fe-doped TiO2 nanotube samples were magnetic behavior of undoped samples. It was shown earlier
studied as Fe is frequently chosen as the magnetic dopant when that the paramagnetic behavior of undoped nanotubes did not
pursuing magnetic semiconductors.34,35,48 Fig. 8 presents the change with cooling and stayed paramagnetic up to 1.8 K.50 The
magnetization vs. magnetic eld dependence (M–H) measured temperature dependence of magnetization (M–T) was
at 300 K aer subtracting the diamagnetic background. The measured, both under eld cooled (FC) and zero eld cooled
well-dened hysteresis loop is observed as typical to that of (ZFC) conditions, in order to better understand the origin of
ferromagnetic materials. The saturation magnetization (Ms), ferromagnetism in Fe-doped TiO2 nanotubes.51–53 The FC results
coercive eld (Hc), and remnant magnetization were extracted are obtained by measuring the magnetic moment of the sample
to be 0.22 emu g1, 402 Oe, and 0.028 emu g1, respectively. in a magnetic eld of 1000 Oe during cooling. The ZFC results
These values averaged along different nanotube orientations, as are obtained by rst cooling the sample to 5 K in zero elds and
our sample was a mixture of randomly oriented nanotubes (see then warming it in the same eld as that of the FC measure-
Experimental section), provide further evidence of the desired ment. The ZFC and FC curves are separated below 300 K and
magnetic doping (it was shown previously that magnetization, ZFC magnetization shows stronger temperature dependence
coercive eld, and remnant magnetization of magnetic nano- compared to the FC curve. The divergence between the FC and
tubes varied depending on whether the applied magnetic eld ZFC curves and the lack of blocking temperature peaks indicate

10846 | Nanoscale, 2014, 6, 10839–10849 This journal is © The Royal Society of Chemistry 2014

Paper
intrinsic magnetism of TiO2 nanotubes resulting from the
substitution of Ti atoms with Fe atoms, not from clusters of Fe
atoms.51–53 As was mentioned above, XRD studies revealed no
traces of secondary phases. Also it should be mentioned that the
M–H and M–T dependence of our Fe-doped TiO2 nanotubes are
similar to those in many other reports on Fe-doped TiO2 lms,
nanoparticles, and bulk crystals.34,35,48 These all together
strongly suggest that the ferromagnetic signal from our Fe-
doped TiO2 nanotube samples is intrinsic resulting from Fe
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
atoms substituting Ti atoms. The mechanism of ferromagne-
tism in doped transition-metal oxides is still under debate.
However, all existing models explaining ferromagnetism in
oxide semiconductors (superexchange, double exchange, free-
carrier-mediated exchange, and exchange through bound
magnetic polarons) involve the substitutional nature of
magnetic dopants replacing Ti atoms in the TiO2 crystal.54–58 In
addition, it was shown from density functional theory (DFT)
calculations that interstitial magnetic dopants destroy spin
polarization.59 Furthermore, using electronic measurements of
Fe-doped TiO2 nanotubes, the nanotubes show increased n-type
conductivity (doping created additional ionized electrons,
Fig. 5a), which can be expected by substitutional doping with
Fe2+ ions (of Ti4+ ions) in the TiO2 lattice. However, interstitial
doping with Fe2+ ions would have resulted in an opposite p-type
electronic behavior. Therefore, it is reasonable to conclude that
Fe2+ ions in our TiO2 nanotubes predominantly substitute Ti4+
ions in the crystal lattice.
Fig. 9 (a) The resistivity for Nb doped nanotubes as a function of
temperature with two different Nb doping levels – 1% (Ti0.99Nb0.01O2),
curve 2, and 10% (Ti0.9Nb0.1O2), curve 1. Ti0.99Nb0.01O2 shows an
increase of resistivity with temperature as typical to semiconductors.
By contrast Ti0.99Nb0.1O2 shows a decrease of resistivity with
temperature, revealing metallic behavior. (b) Transmittance of 10% Nb
doped TNT samples. Transmittance above 90% was observed in the
visible range for the 10% Nb doped TNT.
This journal is © The Royal Society of Chemistry 2014
View Article Online
Nanoscale
Applications
Metallic oxide nanotubes. By heavily doping TiO2 nanotubes
with a shallow dopant Nb, we fabricated transparent nanotube
lms (UV-bandgap
3.1 eV). Two samples with two different
concentrations of Nb were prepared – 1% and 10%. The resis-
tivity TiO2 sample doped with 1% Nb (Ti0.99Nb0.01O2) increased
with temperature, as a typical semiconductor (due to reduced
carrier concentration), as shown in Fig. 9a, curve 2. However,
the heavily doped Ti0.9Nb0.1O2 sample (10% Nb-doped TiO2),
showed a reduced resistance to electron transport with
temperature, typical for metals (due to reduced electron scat-
tering), Fig. 9a, curve 1. Thus, these results show that Nb-doped
TiO2 nanotubes can be excellent candidates for use as trans-
parent conducting oxide nanotube lms. At the high tempera-
ture region of 100–300 K, the r  T dependence is almost linear,
similar to metals, and can be described by r ¼ r0(1 + aDT),
where r0 is the residual resistivity approximately 9.15  104 U
cm, a – is the temperature coefficient 4.7  104 U K1, and DT
– the temperature difference. The carrier concentration n was
almost independent of temperature with only a slight decrease
by
10% when the temperature decreased from 300 K to 20 K
(Fig. S9†). Mobility, however, increased by a factor of 3.5 with a
decrease of temperature within the same temperature range
(from 2.9 to 10.3 cm2 V1 s1) which is explained by the
reduction of phonon scattering at lower temperatures. We
Fig. 10 (a) Co-doped TiO2 nanotubes (Cu–N and Nb–N) show
infrared absorption of incident light, shown using small bandgaps
determined by the STS spectrum. These photocatalysts can absorb
much more light than the undoped wide-bandgap nanotubes. (b)
Stability of undoped and nitrogen-doped TiO2 nanotube catalysts.
Undoped nanotubes show a strong drop in the catalytic activity, for
CO2–H2O reduction (artificial photosynthesis), with time. However,
nitrogen doped nanotubes show a relatively stable photocatalytic yield
of selective high-value products, on illumination with simulated solar
irradiation (AM1.5 spectrum).
Nanoscale, 2014, 6, 10839–10849 | 10847

Nanoscale
Table 2 Summary of photocatalytic reduction experiments for
carbon-dioxide to demonstrate higher selectivity and higher stability
for the doped TiO2 nanotubes
Production Rate (e mmol cm2 h1)
Sample Methane Ethane Acetaldehyde Selectivity
Undoped 0.014 0.023 0.088 70%
0.5 wt% nitrogen 0.003 0.006 0.060 87
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
conrmed that the transmission obtained in these lms is
above 90% for the entire visible wavelength range (Fig. 9b), and
demonstrate metal-like conductivity (Fig. 9a). The high trans-
mittance of these nanotubes (>90% above the bandgap and
10 wt% Nb dopant) allows the demonstration of TCO nano-
tubes. The transmittance h was estimated from reectivity
measurements using the following equation:
 
Rb  Rd
h¼ 1  100% (2)
Rs  Rd
where Rb, are measured signals from the blank Ti sheet, Rs – is
the signal from the sample, and Rd – is the background signal.
Photocatalysis. Using suitable (deep and shallow) mono-
dopants and co-dopants (Cu, N, Cu–N, and Nb–N), we fabricated
photocatalysts with high absorption throughout the visible
wavelengths (for enhanced light absorption, Fig. 10a). These
photocatalysts can absorb much more light than the undoped
wide-bandgap nanotubes. Besides the higher (infrared
bandgap) absorption, we can easily tune the electronic levels in
these doped wide-bandgap nanotubes, to match the desired
redox reactions for CO2/H2O for selective photocatalysis60
(Fig. 10a and S11†). Photocatalytic reduction of carbon-dioxide,
using simulated AM1.5 solar irradiation (Fig. S10†), was used to
demonstrate the higher selectivity (see Table 2) and higher
stability (Fig. 10b) of the doped TiO2 nanotubes, using nitrogen
doping (see ESI†). While the apparent quantum yield (AQY) of
these doped nanotubes decreases from 0.48% for undoped
nanotubes to 0.3% for nitrogen-doped TiO2, the improved
selectivity and stability can lead to important applications in the
generation of solar fuels. Moreover, measurements of the elec-
tronic properties of these doped nanotubes reveal no charge
trapping due to the addition of dopants, and high collection
efficiency for photogenerated charge carriers (discussion on
photo I–V, Fig. 7). Therefore, different light harvesting strategies
and solar concentrators can also be used to further improve the
quantum yield for xing CO2 using sunlight as desired solar
fuels.60 These results indicate the applicability of this new
doping method for improving the absorption and selectivity of
new TiO2 photocatalysts for applications in articial photosyn-
thesis, improved photocatalytic and photoelectrochemical cells,
and other devices for harvesting incident solar energy.
Conclusions
We presented a new method for nanotube doping during elec-
trochemical anodization using alternating current that was
10848 | Nanoscale, 2014, 6, 10839–10849
View Article Online
Paper
enabled by a new electrochemical cell design. A variety of
optical, electronic and magnetic dopants were successfully
incorporated into the hollow nanotubes leading to desired
physical properties. Detailed investigations shown here prove
that this versatile method is applicable for tailoring the dopant
(cationic, anionic, mono and co-doping) and the doping level
(from low doping to highly-doped semiconductors) over a wide
range of materials and physical properties (semiconductor to
metal-like transition). These results can have important impli-
cations for the development of new devices and device archi-
tectures for applications in the broad eld of optoelectronics,
catalysis, display technologies, and power electronics.
Acknowledgements
We would like to thank Zachary Fisk and Ted Grant from the
University of California–Irvine for their assistance in magneti-
zation measurements using SQUID. Financial support for this
work was provided by start-up funds from the University of
Colorado and NSF Career Award CBET-1351281.
References
1 B. Oregan and M. Gratzel, Nature, 1991, 353, 737.
2 A. L. Linsebigler, G. Q. Lu and J. T. Yates, Chem. Rev., 1995,
95, 735.
3 X. B. Chen, L. Liu, P. Y. Yu and S. S. Mao, Science, 2011, 331,
746.
4 Y. C. Nah, I. Paramasivam and P. Schmuki, ChemPhysChem,
2010, 11, 2698.
5 D. Mocatta, G. Cohen, J. Schattner, O. Millo, E. Rabani and
U. Banin, Science, 2011, 332, 77.
6 S. C. Erwin, L. J. Zu, M. I. Hael, A. L. Efros, T. A. Kennedy
and D. J. Norris, Nature, 2005, 436, 91.
7 D. J. Norris, A. L. Efros and S. C. Erwin, Science, 2008, 319,
1776.
8 A. Boltasseva and H. A. Atwater, Science, 2011, 331, 290.
9 J. M. Luther, P. K. Jain, T. Ewers and A. P. Alivisatos, Nat.
Mater., 2011, 10, 361.
10 Y. Q. Gai, J. B. Li, S. S. Li, J. B. Xia and S. H. Wei, Phys. Rev.
Lett., 2009, 102, 036402.
11 W. G. Zhu, X. F. Qiu, V. Iancu, X. Q. Chen, H. Pan, W. Wang,
N. M. Dimitrijevic, T. Rajh, H. M. Meyer, M. P. Paranthaman,
G. M. Stocks, H. H. Weitering, B. H. Gu, G. Eres and
Z. Y. Zhang, Phys. Rev. Lett., 2009, 103, 226401.
12 T. H. Zhou, Y. H. Du, A. Borgna, J. D. Hong, Y. B. Wang,
J. Y. Han, W. Zhang and R. Xu, Energy Environ. Sci., 2013,
6, 3229.
13 W. J. Ong, L. L. Tan, S. P. Chai, S. T. Yong and
A. R. Mohamed, Nanoscale, 2014, 6, 1946.
14 X. Y. Pan, M. Q. Yang, X. Z. Fu, N. Zhang and Y. J. Xu,
Nanoscale, 2013, 5, 3601.
15 Q. J. Xiang and J. G. Yu, J. Phys. Chem. Lett., 2013, 4, 753.
16 W. J. Ong, L. L. Tan, S. P. Chai, S. T. Yong and
A. R. Mohamed, ChemSusChem, 2014, 7, 690.
17 J. D. Hong, Y. S. Wang, Y. B. Wang, W. Zhang and R. Xu,
ChemSusChem, 2013, 6, 2263.
This journal is © The Royal Society of Chemistry 2014

Paper
18 J. Ran, J. Zhang, J. Yu, M. Jaroniec and S. J. Qiao, Chem. Soc.
Rev., 2014, DOI: 10.1039/c3cs60425j.
19 M. Ni, M. K. H. Leung, D. Y. C. Leung and K. Sumathy,
Renewable Sustainable Energy Rev., 2007, 11, 401.
20 G. Lui, J. Y. Liao, A. S. Duan, Z. S. Zhang, M. Fowler and
A. P. Yu, J. Mater. Chem. A, 2013, 1, 12255.
21 W. J. Ong, M. M. Gui, S. P. Chai and A. R. Mohamed, RSC
Adv., 2013, 3, 4505.
22 Q. J. Xiang, J. G. Yu and M. Jaroniec, Nanoscale, 2011, 3, 3670.
Published on 12 August 2014. Downloaded by Universitätsbibliothek Bern on 05/09/2014 14:04:08.
23 Y. H. Hu, H. Wang and B. Hu, ChemSusChem, 2010, 3, 782.
24 G. K. Mor, O. K. Varghese, M. Paulose, K. Shankar and
C. A. Grimes, Sol. Energy Mater. Sol. Cells, 2006, 90, 2011.
25 M. Paulose, K. Shankar, S. Yoriya, H. E. Prakasam,
O. K. Varghese, G. K. Mor, T. A. Latempa, A. Fitzgerald and
C. A. Grimes, J. Phys. Chem. B, 2006, 110, 16179.
26 C. A. Grimes, J. Mater. Chem., 2007, 17, 1451.
27 J. M. Macak, H. Hildebrand, U. Marten-Jahns and
P. Schmuki, J. Electroanal. Chem., 2008, 621, 254.
28 Y. Alivov, M. Pandikunta, S. Nikishin and Z. Y. Fan,
Nanotechnology, 2009, 20, 225602.
29 H. Tang, K. Prasad, R. Sanjines, P. E. Schmid and F. Levy, J.
Appl. Phys., 1994, 75, 2042.
30 Y. Furubayashi, N. Yamada, Y. Hirose, Y. Yamamoto,
M. Otani, T. Hitosugi, T. Shimada and T. Hasegawa, J.
Appl. Phys., 2007, 101, 093705.
31 T. Hitosugi, A. Ueda, S. Nakao, N. Yamada, Y. Furubayashi,
Y. Hirose, T. Shimada and T. Hasegawa, Appl. Phys. Lett.,
2007, 90, 212106.
32 T. Hitosugi, A. Ueda, Y. Furubayashi, Y. Hirose, S. Konuma,
T. Shimada and T. Hasegawa, Jpn. J. Appl. Phys., Part 2, 2007,
46, L86.
33 N. H. Hong, J. Sakai, W. Prellier, A. Hassini, A. Ruyter and
F. Gervais, Phys. Rev. B: Condens. Matter Mater. Phys., 2004,
70, 195204.
34 F. Lin, D. M. Jiang, Y. Lin and X. M. Ma, Phys. B, 2008, 403,
2193.
35 K. Y. S. Chan and G. K. L. Goh, Thin Solid Films, 2008, 516,
5582.
36 M. Ni, M. K. H. Leung, D. Y. C. Leung and K. Sumathy,
Renewable Sustainable Energy Rev., 2007, 11, 401.
37 O. K. Varghese, D. W. Gong, M. Paulose, C. A. Grimes and
E. C. Dickey, J. Mater. Res., 2003, 18, 156.
38 R. D. Shannon, Acta Crystallogr., Sect. A: Found. Crystallogr.,
1976, 32, 751.
39 Y. Furubayashi, T. Hitosugi, Y. Yamamoto, K. Inaba,
G. Kinoda, Y. Hirose, T. Shimada and T. Hasegawa, Appl.
Phys. Lett., 2005, 86, 252101.
This journal is © The Royal Society of Chemistry 2014
View Article Online
Nanoscale
40 J. Osorio-Guillen, S. Lany and A. Zunger, Phys. Rev. Lett.,
2008, 100, 036601.
41 P. Zawadzki, J. Rossmeisl and K. W. Jacobsen, Phys. Rev. B:
Condens. Matter Mater. Phys., 2011, 84, 121203.
42 Z. Y. Zhang, C. H. Jin, X. L. Liang, Q. Chen and L. M. Peng,
Appl. Phys. Lett., 2006, 88, 073102.
43 D. V. Wellia, Q. C. Xu, M. A. Sk, K. H. Lim, T. M. Lim and
T. T. Y. Tan, Appl. Catal., A, 2011, 401, 98.
44 A. R. Bally, E. N. Korobeinikova, P. E. Schmid, F. Levy and
F. Bussy, J. Phys. D: Appl. Phys., 1998, 31, 1149.
45 A. Roldan, M. Boronat, A. Corma and F. Illas, J. Phys. Chem.
C, 2010, 114, 6511.
46 C. Di Valentin, E. Finazzi, G. Pacchioni, A. Selloni,
S. Livraghi, M. C. Paganini and E. Giamello, Chem. Phys.,
2007, 339, 44.
47 A. A. Talin, F. Leonard, B. S. Swartzentruber, X. Wang and
S. D. Hersee, Phys. Rev. Lett., 2008, 101, 076802.
48 N. H. Hong, J. Sakai, W. Prellier and A. Ruyter, J. Phys. D:
Appl. Phys., 2005, 38, 816.
49 V. M. Prida, M. Hernandez-Velez, K. R. Pirota, A. Menendez
and M. Vazquez, Nanotechnology, 2005, 16, 2696.
50 Y. Alivov, T. Grant, C. Capan, W. Iwamoto, P. G. Pagliuso and
S. Molloi, Nanotechnology, 2013, 24, 275704.
51 G. J. Yang, D. Q. Gao, Z. H. Shi, Z. H. Zhang, J. Zhang,
J. L. Zhang and D. S. Xue, J. Phys. Chem. C, 2010, 114, 21989.
52 D. Y. Inamdar, A. D. Lad, A. K. Pathak, I. Dubenko, N. Ali and
S. Mahamuni, J. Phys. Chem. C, 2010, 114, 1451.
53 A. Manivannan, P. Dutta, G. Glaspell and M. S. Seehra, J.
Appl. Phys., 2006, 99, 08M110.
54 M. H. F. Sluiter, Y. Kawazoe, P. Sharma, A. Inoue, A. R. Raju,
C. Rout and U. V. Waghmare, Phys. Rev. Lett., 2005, 94,
187204.
55 J. M. D. Coey, A. P. Douvalis, C. B. Fitzgerald and
M. Venkatesan, Appl. Phys. Lett., 2004, 84, 1332.
56 A. C. Tuan, J. D. Bryan, A. B. Pakhomov, V. Shutthanandan,
S. Thevuthasan, D. E. McCready, D. Gaspar,
M. H. Engelhard, J. W. Rogers, K. Krishnan, D. R. Gamelin
and S. A. Chambers, Phys. Rev. B: Condens. Matter Mater.
Phys., 2004, 70, 054424.
57 P. V. Radovanovic and D. R. Gamelin, Phys. Rev. Lett., 2003,
91, 157202.
58 M. S. Park, S. K. Kwon and B. I. Min, Phys. Rev. B: Condens.
Matter Mater. Phys., 2002, 65, 161201.
59 W. T. Geng and K. S. Kim, Phys. Rev. B: Condens. Matter
Mater. Phys., 2003, 68, 125203.
60 V. Singh, I. J. C. Beltran, J. C. Ribot and P. Nagpal, Nano Lett.,
2014, 14, 597.
Nanoscale, 2014, 6, 10839–10849 | 10849