Electrochimica Acta 83 (2012) 113124

Contents lists available at SciVerse ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

(Photo)Electrochemical characterization of nanoporous TiO2 and Ce-doped TiO2

solgel lm electrodes
Sutasinee Kityakarn a , Yingyot Pooarporn b , Prayoon Songsiriritthigul b , Attera Worayingyong a ,
Simone Robl c , Andr M. Braun d , Michael Wrner e,
a
Department of Chemistry, Faculty of Science, Kasetsart University, Bangkok 10903, Thailand
b
Synchrotron Light Research Institute (Public Organization), Nakhon Ratchasima 30000, Thailand
c
Institute for Technical Thermodynamics and Refrigeration (ITTK) Karlsruhe Institute of Technology, Karlsruhe D-76131, Germany
d
Engler-Bunte-Institute, Karlsruhe Institute of Technology, Karlsruhe D-76131, Germany
e
Institute of Engineering in Life Sciences, Karlsruhe Institute of Technology, Karlsruhe D-76131, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Focusing on effects of charge separation limitations and doping on the photoactivity of TiO2 , dif-
Received 19 January 2012 ferent nanoporous TiO2 lm electrodes were prepared by the solgel or the suspension methods,
Received in revised form 6 July 2012 respectively. X-ray diffraction (XRD) characterization revealed for all undoped TiO2 electrodes (TiO2 -
Accepted 8 July 2012
SG and TiO2 -P25) mixed phases of anatase and rutile, whereas the 2%CeTiO2 -SG electrode showed only
Available online 16 August 2012
anatase phase pattern, revealing that the incorporation of Ce ions prevented phase transformation. The
(photo)electrochemical characterization of the nanoporous TiO2 lm electrodes was performed by cyclic
Keywords:
voltammetry (CV) and electrochemical impedance spectroscopy (EIS) in the dark and under UV irradi-
Nanoporous TiO2 lm electrode
Ce-doped TiO2
ation, respectively. A photocurrent maximum depending on the lm thickness was observed at about
Photoelectrochemistry 13 m for all TiO2 electrodes investigated. The photocurrent responses of Ce-doped solgel TiO2 elec-
Electrochemical impedance spectroscopy trodes were comparably low, indicating an enhanced electronhole recombination at the Ce ion levels
(EIS) within the band gap of TiO2 . A porous electrode model was adapted for the tting of the experimen-
Porous electrode model tal EIS data. The potential and incident radiant power density dependence of the heterogeneous charge
transfer reaction (RCT ) and the coupled transport of photogenerated electrons to the collector electrode
(ZS ) were ascribed to promoted charge carrier separation and migration, i.e. electron drift dominated
over diffusion. Consequently, the dependence of different discrete impedance elements was discussed,
supposing a macroscopic electric eld across the 3D array of interconnected TiO2 nanoparticles that form
the lm electrodes. Besides photocurrent doubling, addition of methanol as a hole scavenger revealed
the limitation of the overall reaction by the heterogeneous charge transfer reaction, in particular at high
radiant power densities.
2012 Elsevier Ltd. All rights reserved.

1. Introduction contrast to the pure dispersions, the observed increased photo-

Titanium dioxide (TiO2 ) is a promising material in photo-
catalytic applications for environmental technologies, due to its
semiconductor properties and the generation of electronhole
pairs under UV irradiation. These electronhole pairs migrate or dif-
fuse to the TiO2 surface, where they react with adsorbed molecules.
A lot of effort was made to improve the photocatalytic perfor-
mance [14]. The photocatalytic properties of titania are sensitive
to their polymorphism and microstructure. Anatase is widely used
as a photocatalyst and considered to be the most active form [5,6],
but high photocatalytic activity was also obtained using mixtures
of aqueous suspensions of anatase and rutile particles [3,7]. In
Corresponding author. Tel.: +49 721 608 46235; fax: +49 721 608 46240.
E-mail address: michael.woerner@kit.edu (M. Wrner).
catalytic activity was assumed to be due to inter-particle contacts
formed between anatase and rutile particles in water and, conse-
quently, to band bending upon a Fermi level line-up. A potential
drop within the contact zone would prevent an electron ow from
anatase to rutile particles, whereas holes created in anatase may
migrate to rutile particles. In addition, for mixed-phase TiO2 -P25
particles, an intrinsic charge separation was proposed to explain
the high photocatalytic activity in particulate dispersive aqueous
systems [8].
Cerium and transition metals are added to increase the photo-
catalytic activity of TiO2 and to extend the absorption range into
the visible. The presence of such metal ions provides trap sites
for electrons and holes, in addition to those located at the parti-
cle surface (e.g. adsorbed O2 and OH ), that may lead to higher
rates of charge separation and interfacial charge transfer reac-
tions. The photocatalytic activity of aqueous TiO2 suspensions for

0013-4686/$ see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.electacta.2012.07.129
114 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124
mercaptobenzothiazole degradation was signicantly enhanced by
incorporation of Ce(III) [9]. However, contradictory results were
published, e.g. the doping with different metal ions leading to an
increased rate of recombination of electronhole pairs and to lower
rates of substrate (4-nitrophenol) degradation [10].
Photoelectrochemical (PEC) systems were also used to achieve
an increase of the photocatalytic efciency. Different methods
were used to immobilize TiO2 at conductive substrates in form
of nanoporous lms to permit large semiconductor surface areas
[2,1113]. The effect of the applied potential was shown for
degradation of 4-chlorophenol and revealed that depletion layer
conditions were needed for high rates of photocatalytic degrada-
tion [2].
For the improvement of the performance of PEC systems based
on semiconducting TiO2 , a detailed mechanistic understanding is
needed. Electrochemical impedance spectroscopy (EIS) provides
specic information about double layer properties, charge transfer
phenomena, charge carrier diffusion and charge generation-
recombination processes [14]. The results of EIS analyses are
usually discussed in terms of equivalent circuits allowing the eval-
uation of different physicochemical parameters of the reaction
system [15].
A number of equivalent circuits for semiconductor/electrolyte
interfaces may be assembled by combinations of resistors and
capacitors (i.e. Debye elements) [16]. However, more recent stud-
ies show that it is necessary to employ constant phase elements
(CPE) under conditions of the observed frequency dispersion [17].
CPEs are associated with disorder, either in the electrode structure
or in the diffusion dynamics [18,19]. In addition, a transmis-
sion line model (TLM) is introduced to describe the small signal
ac-impedance of nanoporous TiO2 lm electrodes in aqueous elec-
trolytes [20,21]. This model accounts for two transport channels,
in the pores (lled with electrolyte) and in the solid (intercon-
nected/sintered TiO2 particles), as well as for the cross reaction
at the interface (interfacial charge transfer at the semiconduc-
tor/electrolyte boundary).
Under UV irradiation, several coupled processes such as carrier
generation and diffusion, recombination, charge transfer and mass
transport in the solution have to be considered. Nevertheless, the
analysis of different simultaneous processes by EIS requires ade-
quately differing time constants.
The aim of this work was to study the (photo)electrochemical
properties of nanoporous TiO2 lms prepared on ITO (indium tin
oxide) coated glass substrates for the improvement of the photo-
catalytic performance of such electrodes. A specically developed
solgel (SG) method was used to prepare nanoporous TiO2 -SG lm
electrodes with the same anatase to rutile ratio as the one found
in TiO2 -P25 (75/25), providing the possibility to compare different
preparation methods and to evaluate intrinsic charge separation
effects. Furthermore, the effect of Ce-doping on the anatase to
rutile transformation and on the photo-electrochemical proper-
ties was studied. A model to describe the impedance behavior of
nanoporous TiO2 electrodes (under UV irradiation) is presented in
this work. It provides the possibility to optimize the photocatalytic
activity of TiO2 -electrodes, considering morphology modication
and doping aspects.
2. Experimental
2.1. Electrode preparation
2.1.1. Nanoporous TiO2 -SG lm electrodes (TiO2 -SG and
2%CeTiO2 -SG)
Nanoporous TiO2 electrodes were prepared with a solgel
method [22] using titianium(IV)bis(ethyl acetoacetato)
diisopropoxide (Aldrich, technical grade) as precursor and cerium
nitrate hexahydrate (Fluka, 99.0%) for doping (Table 1).
The titanium alkoxide precursor solution (0.01 mol) was added
drop-wise into a stirred aqueous nitric acid (pH 1.0, 36.0 cm3 )
at 296 1 K. The stirred suspension was heated to 358 K and
kept at this temperature for 100 min. ITO-coated oat glass (PGO,
Germany) with SiO2 passivation layers, 2 cm 3 cm (<10 /cm2 )
was used as a support for TiO2 . This support was immersed in boil-
ing acetone (348 K) for 10 min and afterwards in 30% (v/v) HNO3
for 30 min. The pre-treated ITO glass was stored in distilled water
for 1 h before coating. An area of 2 cm 2 cm of the pretreated ITO
glass was dipped into the sol for 30 s, withdrawn (within 10 s)
and kept at 473 K for 10 min. This procedure was repeated sev-
eral times to increase the thickness of the resulting TiO2 lm, the
number of repetitions representing the number of layers. The TiO2 -
coated substrate was calcinated at 823 K for 4 h in air at ambient
pressure. Compared to TiO2 -P25 electrodes, a higher temperature
was required to obtain TiO2 -SG electrodes with the same anatase
to rutile ratio as for P25 (75:25). For Ce-doping (2% CeTiO2 -SG),
the corresponding amount of cerium was added to the nitric acid
solution. The procedure was the same as described above for the
preparation of TiO2 -SG electrodes.
2.1.2. Nanoparticulate TiO2 -P25 lm electrodes
Vinodgopals method [2] to prepare nanoparticulate thin lm
electrodes was modied as follows: an aqueous suspension of
TiO2 -P25 (Degussa, 1.00 g in 50 cm3 water/methanol, 1/4, v/v) was
dispensed on pre-treated ITO glass. The solvent was evaporated at
308 K for 2 h. The procedure was repeated to increase the number
of layers, and the electrode was nally calcinated at 723 K for 4 h in
air at ambient pressure.
The electrical contact of all prepared electrodes was made by
conductive adhesive bonding of a silver wire over the whole width
of the uncovered upper part of the ITO glass substrate, and all con-
ductive parts of the electrodes were protected by insulating epoxy
glue, cutting edges included.
2.2. Characterization of TiO2 lm electrodes
2.2.1. Structure and morphology
X-ray diffraction (XRD) analyses were performed with a Philips
XPert diffractometer using Cu K -radiation with an anode cur-
rent of 30 mA and an accelerating voltage of 40 kV with a scanning
step of 0.02 . The diffraction patterns were indexed by comparison
with the JCPDS (Joint Committee on Powder Diffraction Standards)
les number 83-2243, 78-1510 and 81-0792 for anatase, rutile and
cerium oxide, respectively.
The morphology of the electrodes was analyzed by eld
emission scanning electron microscopy (SEM Philips XL30). An
accelerating voltage in the range of 1030 keV was used. The spec-
imens were prepared by Au sputtering to increase the conductivity
of the samples.
2.2.2. (Photo)Electrochemical characterization
The (photo)electrochemical experiments were carried out with
three electrode technique using the prepared TiO2 lm elec-
trodes as working electrode, a platinum counter electrode (ag
of 1 cm 1 cm) and a Ag/AgCl (in aqueous KCl, 3.0 M) reference
electrode. All electrode potentials are given with respect to the
reference electrode. A rectangular large cuvette (Optical Glass,
40 mm 40 mm 40 mm, Hellma, Germany) equipped with a self-
made Teon top for positioning and xing the electrodes was used
as measuring cell. Cyclic voltammograms (CV, scan rate: 25 mV/s)
and electrochemical impedance spectra (EIS) were recorded with
an IM6 Electrochemical Workstation (ZAHNER-Elektrik, Germany).
An aqueous sodium sulfate solution (Fluka, 99.0%, 0.50 M) was
 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124 115

Table 1
Sample codes and composition of different prepared TiO2 lm electrodes.

Sample code Source Preparation Ti (%w/w) Ce (%w/w)

TiO2 -P25 Degussa Suspension 100

TiO2 -SG Titanium(IV)bis(ethyl acetoacetato)diisopropoxide Solgel 100
2%CeTiO2 -SG Titanium(IV)bis(ethyl acetoacetato)diisopropoxide Solgel 97.7 2.3

used as electrolyte. It was degassed with Argon (5.0) before and
kept under an Argon atmosphere during experiments. Methanol
(Aldrich, 98%) was added to the electrolyte solution (2.5 M of
methanol) to obtain more detailed information about the effect of
a hole acceptor.
For EIS studies, a 5 mV ac-perturbation was applied within a fre-
quency range of 30 mHz to 100 kHz. For comparative experiments
with undoped and Ce-doped TiO2 -SG electrodes, a 250 W halogen
lamp was used as irradiation source (5025AF-S, Braun, Germany).
Further experiments were carried out with a halogen light source
equipped with a ber light guide (cold light source, KL 2500 LCD,
Schott, Germany) permitting radiant power density variation. Inci-
dent radiant power densities, P0 [23], were determined with a
calibrated ber optic spectroradiometer (USB2000+, Ocean Optics,
USA) and a 400 nm short pass lter (Laser 2000 GmbH, Germany)
allowing to quantify the part of the emitted wavelength range that
will be absorbed by the TiO2 lm electrodes. The optical ber was
placed at the same distance at which the TiO2 electrodes are being
irradiated. Therefore, the measured values of P0 are not corrected
for adsorption and reection losses in cuvette glass and electrolyte
solution.
3. Results and discussion
3.1. Structural and morphological characterizations
3.2. Electrochemical and photoelectrochemical characterizations
3.2.1. Cyclic voltammetry (CV)
3.2.1.1. Experiments in the dark. The voltammetric responses of a
bare ITO substrate and of TiO2 -SG electrodes in 0.50 M aqueous
Na2 SO4 (pH 6.3) are presented in Fig. 3a. In a degassed electrolyte
solution, the bare ITO electrode could be polarized in the poten-
tial range of 1.3 to 1.1 V, the result indicating only double layer
charging.
For TiO2 -SG coated electrodes, a Faradaic current was observed
below 0.3 V (vs. Ag/AgCl) that may be attributed to the reduc-
tion/oxidation of the Ti(IV)/Ti(III) sites of the TiO2 layer. The
reduction of Ti(IV) at the hydroxylated surface is in accordance with
Eq. (1) [26]:
[Ti(OH)2 ]2+ + H+ + e [Ti(OH)]2+ + H2 O (1)
[Ti(OH)2 ]2+
and [Ti(OH)]2+ represent
the Ti(IV) and Ti(III) sur-
face species of TiO2 lms, respectively. The observed current was
linearly dependent on the number of layers of nanoporous TiO2 -SG
(Fig. 3a). Considering redox reaction occurring only at hydroxylated
surface sites, it may be assumed that the inner surface area of the
porous TiO2 lm is proportional to the amount of deposited TiO2
and that the electrolyte solution may penetrate the whole pore
volume of the nanoporous TiO2 lm. To verify, the charge of the
anodic branch was calculated by integration of the corresponding

The XRD patterns of ITO conducting glass, TiO2 -P25, TiO2 -SG
and 2%CeTiO2 -SG lm electrodes are presented in Fig. 1.
The XRD of the TiO2 -P25 lm electrode revealed characteris-
tic peaks of anatase at 2 25.33, 48.10 and 55.14 (JCPDS PDF No.
83-2243) and of rutile at 2 27.51, 36.16 and 54.46 (JCPDS PDF
No. 78-1510). The XRD pattern of the TiO2 -SG lm electrode exhib-
ited the same pattern and almost the same intensity distribution as
TiO2 -P25. The molar ratios of anatase and rutile were determined
by the method described by Spurr and Myers [24]. The rutile phase
of TiO2 -SG lm electrodes was determined to be 23%, in compar-
ison to 24% of that of the TiO2 -P25 lm electrode. It should be
mentioned that calcination of TiO2 -P25 at the same temperature as
for the preparation of TiO2 -SG electrodes (823 K) does not change
the anatase to rutile ratio. The XRD pattern of 2%CeTiO2 -SG lm
electrode was identied to be pure anatase (Fig. 1). The anatase to
rutile transformation requiring temperatures above 873 K for Ce-
doped TiO2 [25], anatase was stabilized by Ce ions at a calcination
temperature of 823 K.
Low resolution SEM images showed signicant differences
between the TiO2 lm electrodes investigated (Fig. 2ac). The SEM
image of the TiO2 -P25 electrode (Fig. 2a) showed a microstructured
lm of TiO2 -P25 at low resolution, but a network of agglomerated
particles, appearing as a nanoparticulate layer, became visible at
high resolution (inset Fig. 2a). The particle diameter was deter-
mined to be about 50 nm. The SEM images of undoped and doped
TiO2 -SG resembled compact lms at low resolution (Fig. 2b and
c), but at higher resolution, tightly packed nanoparticles, forming
nanoporous TiO2 lms were observed (see inset Fig. 2b and c). The
particle sizes at undoped and doped TiO2 -SG electrodes were deter-
mined to be 2030 and 1020 nm, respectively. SEM was also used
to determine the thickness of the lm to be 1215 m for 20 layers Fig. 1. Diffraction patterns of ITO conducting glass and different TiO2 lm electrodes
of TiO2 -SG and for 5 layers of TiO2 -P25, respectively. (TiO2 -P25, TiO2 -SG and 2%CeTiO2 -SG; A: anatase, R: rutile, I: indium oxide).
116 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124
Fig. 2. SEM images of different TiO2 lm electrodes: (a) TiO2 -P25, (b) TiO2 -SG and (c) 2%CeTiO2 -SG.
i vs. t-plots, and thereby, the charge per mass of deposited TiO2
was plotted vs. the number of layers (Fig. 3b). Only TiO2 -SG elec-
trodes showed the ideal behavior represented by a constant charge
per mass (charge density), or in other terms, a linear increase
of the total charge with the number of layers. For TiO2 -P25 and
2%CeTiO2 -SG, a decrease of the charge density was observed with
increasing lm thickness. This deviation from the ideal behavior
may be explained by gradients of active surface sites and/or by
different lm morphologies.
The doping with Ce ions affected the charge per mass of the
2%CeTiO2 -SG electrodes considerably. It was found to be about
10 times higher than that of the TiO2 -SG electrodes (Fig. 3b). The
result might be due to the larger total surface area of 2%CeTiO2 -
SG as the particles were found to be smaller and forming a
nanoporous lm electrode. The result is also in agreement with
Alves et al. [27] who observed that the replacement of TiO2 by CeO2
Fig. 3. (a) Cyclic voltammograms of nanoporous TiO2 -SG lm electrodes in depen-
dence on the number of deposited layers (in aqueous 0.5 M Na2 SO4 supporting
electrode (pH 6.3)). (b) Anodic peak charge of nanoporous TiO2 electrodes in depen-
dence of the number of deposited layers of TiO2 -P25, TiO2 -SG and 2%CeTiO2 -SG.
in IrO0.3 Ti(0.7x) Cex O2 led to a linear increase of the total charge.
The authors explained their ndings by an increased effective sur-
face area due to the decreased probability of homogeneous crystal
growth. The replacement of TiO2 by CeO2 resulted in fact in the
formation of a very ne crystalline structure [27,28]. However, the
calculated charge per mass of TiO2 -P25 lm electrodes, consist-
ing of a network of relatively large particles, was (also) found to
be about 10 times higher than that of undoped TiO2 -SG electrodes
and is therefore comparable to the charge per mass of Ce-doped
lm electrodes (Fig. 3b). These ndings might imply that TiO2 -
P25 lm electrodes exhibit a signicantly higher number of active
(hydroxylated) sites per surface area than electrodes prepared by
the solgel method, as might be expected, when a lower calcination
temperature is applied [29].
3.2.1.2. Experiments under irradiation. Irradiation of TiO2 lm elec-
trodes with energies exceeding the band gap of TiO2 generates
electronhole pairs. The holes diffuse or migrate to the TiO2 surface
and oxidize adsorbed water or surface hydroxide groups, while the
promoted electrons have to cross particle grain boundaries to be
withdrawn at the TiO2 network/ITO interface during anodic polar-
ization. The observed photocurrent depends on the thickness of the
TiO2 lm electrodes, as shown in Fig. 4 for TiO2 -SG electrodes under
chopped irradiation. A photocurrent maximum was observed for
different numbers of layers depending on the type of lm electrode:
20 layers for TiO2 -SG (12 m thick), 15 layers for 2%CeTiO2 -SG and
5 layers for TiO2 -P25 electrodes (15 m thick), respectively. The
CVs of the different TiO2 lm electrodes at maximum photocurrent
are presented in Fig. 5.
During irradiation, a fast hole transfer to the adsorbed accep-
tor and into the solution takes place, whereas the promoted
electrons move through a network of interconnected particles
Fig. 4. Anodic photocurrent in dependence on the number of deposited layers of
nanoporous TiO2 -SG lm electrodes under chopped irradiation.
 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124 117

Fig. 5. Anodic photocurrent (maximum) of different TiO2 lm electrodes under

chopped irradiation; 20 layers for TiO2 -SG, 15 layers for 2%CeTiO2 -SG and 5 layers
of TiO2 -P25.

to the (ITO) collector electrode, where they are withdrawn as

photocurrent. As long as the depth of penetration of the inci-
dent radiation exceeds the thickness of the TiO2 layer, the
photocurrent increases with the layer thickness due to the
higher number of photons absorbed. The diffusion length of the
promoted electrons being limited mainly by the charge sepa-
ration efciency and the structure of the nanoporous lm, the
probability of recombination processes increases with the lm
thickness related to electron transport across an increased num-
ber of particles and grain boundaries. At highest photocurrents,
the lm thickness (e.g. 20 layers of TiO2 -SG electrode) may
reach the diffusion length of the electrons and/or the distance
of penetration depth of the incident radiation. The decrease of
the photocurrent for thicker lms may therefore be explained
with a higher loss of promoted electrons by recombination and
by charge transfer processes into the solution during diffusion
across the interconnected particles to the collector electrode
(back contact).
It should be noticed that for 2%CeTiO2 -SG electrodes, the
photocurrent was rather low and a pronounced delay in the pho-
tocurrent progression was observed (Fig. 5). Assuming that Ce4+
interstitial sites would act as effective electron scavenger for
photogenerated electrons of TiO2 [10], the delayed photocurrent
response may be explained by a fast lling of these (deep) trap sites
and a subsequent diffusion of electrons via the conduction band to
the back contact [30]. Nevertheless, the low photocurrent observed
for 2%CeTiO2 -SG lm electrodes might indicate an enhanced pho-
tocurrent loss, and it may be assumed that the trapped electrons
efciently recombine with the photogenerated holes at the Ce ion
levels within the band gap of TiO2 .
3.2.2. Electrochemical impedance spectroscopy (EIS)
3.2.2.1. Equivalent circuits for EIS analysis. Impedance spectra of
TiO2 -SG and 2%CeTiO2 -SG electrodes at 0.30 V in the dark and under
irradiation are presented as Bode plots (amplitude, |Z|, and phase,
||, vs. frequency) in Fig. 6a and b, respectively.
For experiments in the dark, the phase angle of both electrodes
reaches almost 90 in the mid frequency range and remains
almost constant at low frequencies (Fig. 6a). The ndings indi-
cate a blocking (fully capacitive) behavior of the nanoporous
TiO2 electrolyte interface. For the analysis of the EIS results, a
porous electrode model, corresponding to a transmission line
with two transport channels and a crosswise interfacial reac-
tion, was employed. The homogeneous porous electrode model
proposed by Ghr [18,31] is implemented in the simulation and
evaluation program of the used software package (Thales, ZAHNER-
Elektrik, Germany). This model is based on a porous electrode with
Fig. 6. Impedance spectra (Bode plots) of different nanoporous TiO2 -SG lm elec-
trodes at a potential of 0.3 V (vs. Ag/AgCl ref. electrode) (a) in the dark, (b) under
irradiation (symbols: experimental data; solid lines: tted data).
uniform cylindrical pores, but assuming that the porous TiO2 lm
electrode is a random network of TiO2 nanoparticles, a simplied
model for the simulation procedure may be used [20,32,33]. The
total impedance of the porous layer is expressed in terms of differ-
ent macroscopic impedance elements (partial equivalent circuits)
(Fig. 7):
Zq for the porous TiO2 layer/pore electrolyte interface.
Zs for the porous TiO2 layer of interconnected TiO2 nanoparticles.
Zp for the pore electrolyte.
Zn the porous TiO2 layer/ITO layer interface.
Zo for the porous TiO2 surface/bulk electrolyte interface.
For nanoporous TiO2 electrodes, the impedance of the outer
porous TiO2 /bulk electrolyte interface (Zo ) is thought to be very
low compared to the impedance of the interior porous TiO2 /pore
electrolyte interface (Zq ) due to the huge size difference of the
interfacial areas and can be neglected.
With this simplication, best tting results of EIS data (recorded
in the dark) were obtained by using the equivalent circuit pre-
sented in Fig. 7c. It contains partial schemes for the interfacial
impedances Zn and Zq , respectively. Zn comprises two RC circuits
in serial connection, the rst representing the conductive ITO bulk
layer [element 1 (CPE) and 2 (RITO )], the second corresponds to the
ITO/electrolyte interface at the bottom of the pores [element 3 (C3 )
and 4 (R4 )], taking into account the surface oxidation of the ITO
layer by thermal treatment [22]. The impedance of the (porous)
TiO2 /pore electrolyte interface (Zq ) is described by a capacitance
118 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124
Fig. 7. Schematic and equivalent circuits of nanoporous TiO2 lm electrodes. (a) Array of TiO2 nanoparticles and corresponding distributed impedance elements according
to the porous electrode model [18,31]. (b) Impedance of the porous electrode in terms of macroscopic impedance elements (Zs , Zp , Zq , Zo and Zn , as described in the main
text). (c) Equivalent circuit applied for TiO2 electrodes under dark conditions. (d) Equivalent circuit for UV irradiated TiO2 electrodes.
only (element 5, CSC ). Element 6 represents the network of the
differential (distributed) impedances Zpi and Zsi (i = 1, 2, 3, . . ., n,
Fig. 7a). The simulation procedure allows determination of the
integral pore electrolyte resistance (Zp ) and the integral solid bulk
resistance, Zs (i.e. resistance of the porous TiO2 lm consisting of
the interconnected TiO2 particles). Finally, the impedance of the
bulk electrolyte is included as a pure ohmic resistance, element 7
(R ).
The results of the tting procedure based on the proposed equiv-
alent circuit are listed in Table 2. The resulting impedance values
for Zn (elements 14) are different for doped and undoped TiO2 -
SG electrodes. This may be explained by the different interfacial
structures, assuming different particle sizes and particle surface
structures for undoped and doped materials. In addition, ITO sur-
face oxidation during the calcination procedure may be affected by
different colloidal structures of undoped and doped gels.
Equal amounts of TiO2 being deposited on both electrodes, the
higher interfacial capacitance of the pore electrolyte/TiO2 interface
(CSC ) for the 2%CeTiO2 -SG may be related to the Ce-doping and/or
to a larger surface area. An increased surface area for the Ce-doped
material would be in accord with the results obtained from the CV
experiments (3.2.1). The impedance of the 2%CeTiO2 -SG electrode
is two times higher than that of the undoped material (impedance
element ZS , Table 2). The argument seems to be contradictory,
because the doping of the material should lead to a higher con-
ductivity and therefore to a smaller overall impedance. However,
considering that ZS is attributed to a porous TiO2 layer consisting
of an assembly of interconnected TiO2 nanoparticles, the effective
bulk conductivity of the porous lm depends on the intrinsic con-
ductivity of the TiO2 particles, the geometry of the contact zone and
the packing structure. In addition, the impedance of the sintered
semiconductor particle network may depend primarily on discon-
tinuities at the grain boundaries. However, the different particle
surface structures of undoped and doped TiO2 -SG lm electrodes
imply that ZS depends mainly on the structure of the particle sinter
bridges.
Under UV irradiation, the generated photocurrent leads to a sig-
nicant decrease of the total impedance of the porous electrode in
the frequency range below 10 Hz. That was also accompanied by
a drop of the phase angle (Fig. 6b). Both effects were more pro-
nounced for the undoped TiO2 -SG than for the doped 2%CeTiO2 -SG
electrode. The higher photoconductivity of the TiO2 -SG electrode
hence is indicated by smaller impedance and phase angle at low
frequencies. These ndings are in accordance with the results of
CV measurements that revealed a two times higher photocurrent
for the undoped material. The EIS results were simulated using
a modied equivalent circuit, where charge carrier transfer and
recombination processes were taken into account (Fig. 7d). The par-
tial equivalent circuit for Zn was adjusted, because the impedance
element 4 had to be eliminated, i.e. the ITO/pore ground interface
was xed to be insulating (innite resistance). In fact, element
4 could no longer be tted after implementation of additional
elements for Zq . Consequently and in consistence with the the-
ory, a charge transfer reaction at this interface was excluded in
the investigated potential range, and the observed current was
assigned exclusively to the photogenerated electrons crossing the
TiO2 particle/ITO interface. The interfacial impedance element Zq
was modied by addition of elements modeling the photo-induced
reactions: (i) an ohmic resistance, element 8 (RCT ) for the heteroge-
neous charge transfer reaction at the porous TiO2 layer/electrolyte
interface, (ii) a capacitive element 10 (CIF ) for the accumulation
of photoinduced charge carriers at the semiconductor surface, in
parallel to the heterogeneous charge transfer reaction resistance
and (iii) an ohmic resistance, element 9 (RREC ) for the charge car-
rier recombination. Under consideration of different models and
the corresponding theoretical (physicochemical) background, sev-
eral attempts to lower the number of distinct impedance elements
were made. However, the equivalent circuit presented here was so
far the only one to yield good tting results, especially for studies
under electrode potential and radiant power density variation, be
aware, that a good t does not, in itself, validate the model used.
However, the use of a simplied partial equivalent circuit for Zq ,
 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124 119

Table 2
Fitting results of different impedance elements for various TiO2 lm electrodes in the dark and under irradiation at an applied potential of 0.3 V (vs. Ag/AgCl ref. electrode).

Dark UV irradiation

Electrode (steady state photocurrent)/impedance element TiO2 -SG (38.7 nA) 2%Ce TiO2 -SG (51.1 nA) TiO2 -SG (51.0 A) 2%Ce TiO2 -SG (39.5 A)

1. CPE (uF) 14.7 11.7 14.5 12.3

0.70 0.74 0.76 0.76
2. RITO ( ) 11.1 16.9 9.5 17.2
3. C3 (F) 30.5 28.1 30.6 31.9
4. R4 (M ) 3.9 2.7
5. CSC (F) 2.4 3.3 45.0 30.4
6. Porous impedance
Zp (Ohm) 8.9 7.6 8.8 8.2
Zs (k ) 240.6 524.5 33.8 146.4
7. R (Ohm) 16.4 14.0 16.5 15.3
8. RCT (k ) 2.1 100.8
9. RREC (k ) 1.2 8.2
10. CIF (mF) 1.58 0.49

often applied for mesoporous semiconductor electrodes in form

of a so called chemical capacitance parallel to a resistance (which
accounts for charge transfer or recombination) [34], yielded simu-
lation results of low signicance and failed. Finally, Zq resulted in an
equivalent circuit representative for coupled electrochemical reac-
tions [35], taking into account, that heterogeneous charge transfer
and recombination may take place at different interfacial electronic
states. The associated capacities are attributed to (i) charge accu-
mulation at the sc surface, CIF (associated to surface trapped holes),
and to (ii) the capacitance of the semiconductor matrix consider-
ing a space charge layer (CSC ). An alternative approach for Zq would
be two parallel RC combinations in serial alignment, regarding to
identical electronic states for charge transfer and recombination
[36].
When comparing the modeling results of electrodes in the dark
and under irradiation, elements for the interface ITO/pore ground
(Zn , including elements 13), pore electrolyte resistance (Zp ) and
electrolyte resistance (element 7, R ) remained constant (Table 2).
As expected, these elements are not affected by UV irradiation, i.e.
photoinduced processes. It has to be noticed that the impedances
RCT , RREC and ZS were found to be signicantly smaller for the
TiO2 -SG electrodes, due to their higher photoactivity. Taking into
account that the same mass of TiO2 was deposited for TiO2 -SG
and 2%CeTiO2 -SG electrodes, it may be concluded that the TiO2
nanoparticle network of the undoped lm electrode exhibits a
higher rate constant for the heterogeneous charge transfer reaction,
a smaller recombination rate and a higher conductivity. The most
pronounced difference was found for the charge transfer resistance,
and it seems that the overall reaction for the doped TiO2 mate-
rial was limited by the rates of the hole reactions at the surface.
However, the charge transfer reaction is linked to the recombina-
tion process and, therefore, cannot be seen independently. A more
detailed discussion of the EIS results would require yet unavailable
data concerning the inner surface area, the size and the intercon-
nection structure of the particles.
3.2.2.2. Photoelectrochemical characterization of nanoporous TiO2 -
SG electrodes by EIS. TiO2 -SG electrodes yielding high photocur-
rents in CV experiments were prepared to study the potential
dependence, the effects of the incident radiant power density and
a hole acceptor/scavenger (addition of methanol), employing EIS.
The irradiation system used for CV experiments was replaced by a
halogen lamp coupled to a ber optics to avoid emitting perturbing
noise in the low frequency range of EIS measurements. In addition,
by this measure the emitting power of the installation could be
varied and heating of the electrolyte solution could be prevented.
3.2.2.2.1. Potential dependence of the impedance (under UV irra-
diation). The potential dependence of the impedance of TiO2 -SG
Fig. 8. Impedance spectra (Bode plots) of a TiO2 -SG electrode (20 layers) in depen-
dence of the applied electrode potential (vs. Ag/AgCl ref. electrode in 0.5 M Na2 SO4
aqueous solution; symbols: experimental data; solid lines: tted data).
lm electrodes was investigated in the potential range from 0.0
to 0.5 V vs. reference electrode, at a xed incident radiant power
density of 5.2 W/m2 . The results are presented in Fig. 8 and demon-
strate the effect of the electrode potential on total impedance and
phase angle in the frequency range below 1 Hz.
Modeling results, presented in Fig. 9, were obtained using the
equivalent circuit already described (Fig. 7d). If the anodic polariza-
tion was increased, the heterogeneous charge transfer resistance
(RCT ) decreased as well as the resistive element for the recombi-
nation (RREC ) and the interconnected TiO2 particle network (ZS ).
The observed strong potential dependence of the impedance of
TiO2 -SG electrode shows rather the characteristics of a compact
solid semiconductor lm electrode under depletion conditions than
of a nanoporous electrode penetrated by the electrolyte solution,
for which only a small effect is expected [37]. The characteristics
of a classical electrolytesemiconductor junction under irradi-
ation can be described by theoretical derivations, as established
by Grtner for semiconductor pn junctions or Schottky contacts
[3840]. During anodic polarization of solid thin lm semicon-
ductors, a space charge layer is formed that leads to an efcient
separation of the photogenerated charge carrier in the electric eld
within the depletion zone. The photo-generated holes migrate to
the TiO2 /electrolyte interface where they can be withdrawn by a
heterogeneous charge transfer reaction (redox reaction), while the
remaining electrons are directed to the collector electrode produc-
ing a photocurrent. When the thickness of the TiO2 lm electrode
is larger than the width of the depletion layer (space charge layer),
120 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124
Fig. 9. Fitting results of different impedance elements in dependence of the applied
potential for a nanoporous TiO2 -SG lm electrode (20 layers) in pure electrolyte
solution (aqueous 0.5 M Na2 SO4 ) under UV irradiation.
an efcient electron/hole recombination during the diffusion of
the photogenerated electrons to the back contact must be con-
sidered. Under these conditions and for the given incident radiant
power density, the photocurrent can be increased with increasing
anodic polarization yielding a thicker depletion layer, where ef-
cient charge separation may take place. However, in the present
study, the electrode is composed by very small TiO2 particles
forming a nanoporous layer that is penetrated by the electrolyte
solution. Assuming individual nanoparticles, the width of the space
charge layer cannot exceed the particle radius, and for maximum
depletion conditions, the voltage difference between the surface
and the center of the nanosized TiO2 particles cannot exceed several
tenths of millivolts [41].
The results obtained with TiO2 -SG electrodes support the
hypothesis of the formation of a macroscopic electric eld between
the ITO collector electrode and the outer part of the nanoporous
TiO2 lm. It seems that for these nanoporous lm electrodes, an
apparent space charge layer becomes operative. With increas-
ing anodic polarization (i.e. extended macroscopic electric eld),
migration of photogenerated electrons to the electrode contact and
the redox reactions of holes at the semiconductor/electrolyte inter-
face are enhanced, and a decrease of the charge transfer resistance
(RCT ) and of the impedance element for the recombination reaction
(RREC ) is expected. In fact, RCT decreased steadily with increasing
anodic potential (Fig. 9a). The drop of RREC was less pronounced,
most probably due to the characteristics of the applied equivalent
circuit, as discussed below.
A macroscopic electric eld across a semiconductor nanopar-
ticulate network was also suggested by Hagfeldt et al. to explain
the high photocurrent quantum efciencies of colloidal TiO2 lm
electrodes [42]. In contrast to this picture, models of the poten-
tial distribution in semiconducting porous electrodes, penetrated
by an electrolyte solution, showed a potential drop within a spa-
tial resolution of few particle layers next to the collector electrode
(ITO layer surface) [37,43]. Alternatively, the observed potential
dependence might be due to an electric eld induced by photogen-
erated charges. Assuming very fast hole removal reactions, negative
charges may shift the band edges to higher energy, due to remaining
electrons trapped in interfacial band-gap states. Taking the spatial
absorption characteristic into account, such a shift of band edges
would result in an electric eld component driving electrons to
the collector electrode, when the nanoparticulate TiO2 lm would
be irradiated from the electrolyte (front) side. Nevertheless, it has
been published that light-induced electric eld effects should not
alter the electron-transport characteristics in a fundamental way
[44]. Both effects, potential drop next to the collector electrode
and light-induced electric eld, should have a strong impact on the
photo-response of the nanoparticulate TiO2 lm, when irradiation
is changed from electrolyte (front) to back side. The experimental
results of this work with nanoparticulate TiO2 -SG lm electrodes
(20 layers) revealed only small changes for all tted impedance ele-
ments (mean error 1.0 0.5%) and cannot support these alternative
approaches [37,43,44].
Compared to dark conditions, the impedance of the particle
network (ZS ) dropped by a factor of about 8 during UV irradia-
tion. This can be explained by a high (excess) concentration of
electrons in the transport state (conduction band) due to exci-
tation of the porous TiO2 electrode in combination with the fast
removal of the holes. In contrast, the observed potential depen-
dence of ZS in the dark was small, and varied only by a factor
of 1.5 in the potential range between 0.0 and 0.5 V vs. ref. elec-
trode. Only biasing TiO2 SG electrodes in the cathodic potential
region towards the conduction band edge potential induced a pro-
nounced drop of ZS as predicted by the theory, which based on
an exponential increase of the density of states (DOS) approach-
ing the conduction band [21,45]. Focusing on the photocatalytic
performance of nanoporous electrodes, we excluded the cathodic
potential region (vs. ref) for the experiments of UV irradiated elec-
trodes. Hence, under idealized conditions, ZS would only depend on
the incident radiant power density, considering the low conductiv-
ity (DOS) und the small potential dependency for the dark situation,
respectively. Therefore, the observed potential dependence (see
Fig. 9a) of the impedance of the particle network (ZS ), i.e. the resis-
tive channel of electron transport to the collector electrode, was
unanticipated. It seems that the applied equivalent circuit is over-
simplied with respect to RREC , implemented as an element of the
interfacial impedance Zq only. Charge carrier recombination may
also take place in the bulk material (nanoparticle array) and at grain
boundaries. In fact, lower charge carrier recombination rates at the
grain boundaries would affect the impedance of the nanoparticle
network (ZS ) resulting in the observed decrease of ZS with increas-
ing anodic polarization. This interdependence might be ascribed to
an extended diffusion length of photo-generated electrons.
The partial equivalent circuit for Zq comprises two capacitive
elements, CIF and CSC (see Fig. 7d). The twofold increase of CIF with
increasing anodic polarization implies that a higher number of sur-
face trapped holes cannot be removed by heterogeneous charge
transfer reactions leading to an accumulation of photogenerated
charge at the semiconductor surface. This seems to contradict the
enhanced charge transfer conditions when increasing the anodic
bias, but the steady state photocurrent increased by a factor of
5 simultaneously. CSC also depends on the applied potential and
decreases steadily with increased anodic biasing. A MottSchottky
plot yielded a linear dependence of C2 vs. electrode potential, see
Fig. 10. The at band potential under irradiation was determined to
be 0.22 V vs. ref. electrode. The results are consistent with respect
to the theoretical prediction for a wide band gab bulk semicon-
ductor. The Schottky behavior of CSC underlines that an apparent
space charge layer could be responsible for the enhanced charge
separation efciency by increasing the anodic bias, indicated by
 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124 121

Fig. 10. MottSchottky plot of CSC for a nanoporous TiO2 -SG lm electrode (20
layers) in pure electrolyte solution (aqueous 0.5 M Na2 SO4 ) under UV irradiation.

the decrease of RCT and ZS , and the increase of the steady state
photocurrent by a factor of 5 in the investigated potential range. It
seems that CSC behaves not as a chemical capacitance, often used to
describe the (exponential) potential dependence of charge accumu-
lation in mesoporous semiconductors [34]. Nevertheless, it should
be taken into account that a switch between different mechanism in
dependence of the applied potential may occur, and in addition that
the generation of charge carriers (electron and holes) by excitation
of the semiconductor material itself may require a more complex
model than established for dye sensitized solar cells [21,46].
3.2.2.2.2. Effect of incident radiant power density (P0 ). Photoin-
duced processes at (nanoporous) TiO2 semiconductor electrodes
depend on various parameters such as morphology and absorp-
tion characteristics of the lm, the thickness of space charge layer
with respect to the lm dimension and the spectral emission of
the irradiation source. Investigating the rate determining process,
as e.g. limitations by the heterogeneous charge transfer or by the
recombination process, the incident radiant power density (P0 ) was
varied.
Bode plots of TiO2 -SG electrodes for different incident radiant
power densities applying an electrode potential of 0.3 V are pre-
sented in Fig. 11. Obviously, the variation of the radiant power
density affected only the impedance below 1 Hz. With increas-
ing P0 , the decrease of the overall impedance was accompanied
by a pronounced phase angle bending. Impedance spectra taken
at different incident radiant power densities were simulated with
Fig. 12. Fitting results of different impedance elements and steady state photocur-
the same high accuracy employing the outlined equivalent circuit rent in dependence of the incident radiant power density (P0 ) for a TiO2 -SG electrode
(Fig. 7d). Elements, supposed not to be affected by photo-induced (20 layers) in pure electrolyte solution (aqueous 0.5 M Na2 SO4 ) and after addition
processes, remain almost constant: (i) Zn (elements 13), (ii) Zp of methanol (2.5 M) at an applied potential of 0.3 V (vs. Ag/AgCl ref. electrode).

(pore electrolyte resistance) and (iii) element 7, R (bulk electrolyte

Fig. 11. Impedance spectra (Bode plots) of a TiO2 -SG electrode (20 layers) in depen-
dence of the incident radiant power density at 0.3 V vs. Ag/AgCl ref. electrode in 0.5 M
Na2 SO4 aqueous solution (symbols: experimental data; solid lines: tted data).
resistance). Impedance elements which showed a corresponding
dependence of the incident radiant power density are presented in
Fig. 12.
The observed photocurrent is due to the coupling of the inter-
facial charge transfer (RCT ) and the resistive electron transport
channel (ZS ). Under idealized conditions, the photoconductivity
of TiO2 -SG lm electrodes is expected to be proportional to the
rate of absorbed photons. Assuming a constant absorption coef-
cient across the TiO2 lm and, a xed correlation between the
incident radiant power density and the rate of absorbed pho-
tons, RCT and ZS should show a linear reciprocal dependence on
the incident radiant power density, if the efciency of charge
separation and the electron diffusion length are not affected by
irradiation. In fact, the charge transfer resistance RCT (element 8)
decreased with increasing incident radiant power density up to
3.0 W/m2 (58% of maximum power). At higher values of P0 , RCT
remained constant within the experimental error of 5%, implying
that the heterogeneous charge transfer reaction starts to limit the
122 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124

photocurrent (overall reaction). When the electrode is irradiated on
the front side (semiconductor/electrolyte interface), the penetra-
tion depth of the photons, i.e. the space, where the charge carriers
are produced, has to be related to the width and the position of
the supposed space charge layer, where an efcient charge sepa-
ration can take place. The experimental results may be interpreted
in a way that the penetration depth reaches the thickness of the
apparent space charge layer at an incident radiant power density
of about 3.0 W/m2 . Based on these idealized assumptions, at low
radiant power densities all charge carriers will be produced within
the zone of efcient charge separation. Through this, the number
of charge carriers would depend linearly on the incident radiant
power density.
Assuming that holes can be removed quantitatively by surface
redox reactions, RCT should drop reciprocally with the incident
radiant power density. For values 3.0 W/m2 the experimental
results corroborate with the assumed interdependence. Never-
theless, this hypothesis does not imply the effect of charge
recombination reactions (RREC ) that may also depend on P0 . In
fact, the impedance element for recombination (RREC ) decreases
with increasing incident radiant power density, showing a lin-
ear reciprocal dependence (Fig. 12a). The observed dependence
suggests that a certain percentage of photo-generated charge car-
riers are lost by recombination. Nevertheless, this nding has
to be related to the implementation of RREC , see discussion
below.
A reciprocal dependence on the incident radiant power density
(for idealized conditions) is also observed for ZS , again limited to
values 3.0 W/m2 . At higher P0 , the tting procedure resulted in
higher values of ZS , as displayed by the drop of the slope in the
plot of the reciprocal resistance of the particle network (ZS1 ) vs.
the incident radiant power density (P0 ) (Fig. 12b). Interestingly,
the deviation of ZS from the idealized behavior was observed at
the same incident radiant power density, where RCT became con-
stant (>3.0 W/m2 ), implying that the same process is decisive for
the observed dependence on the incident radiant power density. In
accord with the discussion concerning RCT , at moderate P0 the pen-
etration depth of the irradiation does not exceed the width of the
supposed space charge layer. Under these considerations, when the
charge carriers are generated within an electric eld, the expected
reciprocal dependency can be observed. Then again, when at high
P0 the irradiated volume exceeds the zone of supported charge
carrier separation, a higher rate of recombination is expected. A
reduced electron diffusion length due to enhanced charge car-
rier recombination in the bulk and at the grain boundaries would
affect ZS negatively. As a consequence, this suppressed incident
radiant power density dependency would result in a decreased
slope in the ZS1 vs. P0 plot at high power densities as displayed
in Fig. 12b.
3.2.2.2.3. Effect of methanol addition. The experiments aimed
at possible limitations of the overall reaction (photocurrent)
by heterogeneous charge transfer processes under conditions
of high incident radiant power densities. Further, the pos-
sible effect of a hole-scavenger was studied. The addition
of e.g. citrate, formic acid or methanol to the supporting
electrolyte solution accelerates the hole reaction at the semi-
conductor surface, leading to a faster removal of the holes
and, therefore, to a more efcient charge separation process
[13].
In this study, the steady state photocurrent increased by a fac-
tor of about two, when 2.5 M methanol in 0.5 M Na2 SO4 aqueous
solution was used (Fig. 12c). This result may be explained by an
effect introduced as current doubling [47,48] that was not a topic
of our experimental studies. Current doubling was described to
a reaction sequence, where methanol is oxidized by a hole (Eq.
(2)) (or by a surface hydroxyl radial), the intermediate C-centered
radical injecting subsequently an electron into the conduction band
of TiO2 (Eq. (4)):
+
CH3 OH + h+ CH3 OH (2)
+
CH3 OH CH3 O + H+ (3)
CH3 O CH2 O + H+ + e (4)
To determine the impedance elements, the EIS data were pro-
cessed as before. The addition of methanol affected primarily the
heterogeneous charge transfer resistance (RCT ) that dropped by a
factor of about 2.2. This effect is more pronounced at high radiant
power densities (Fig. 12a). Expectedly, heterogeneous charge trans-
fer reactions are, in the case of nanoporous TiO2 -SG lm electrode
in indifferent aqueous electrolyte solution, a key limiting factor of
the efciency of the overall reaction.
As the hole reaction rate at the semiconductor surface is
increased, the coupled charge recombination rate is expected to
drop. In fact, RREC decreased by a factor of about 1.8 due to the
addition of methanol.
For the resistive pathway of electrons to the electrode back
contact (impedance of the particle array, ZS ), an almost linear radi-
ant power dependence was found. The addition of methanol led
to experimental results from which a better reciprocal correlation
between ZS and the incident radiant power density (idealized con-
ditions) could be obtained. Compared to the results in the pure
electrolyte solution, the most pronounced effect of methanol addi-
tion was found at high radiant power densities (Fig. 12b).
The evolution of the ZS1 vs. P0 plot for pure electrolyte solutions
was explained by a higher loss of electrons due to an enhanced
charge carrier recombination under irradiation, where the penetra-
tion depth of light exceeds the width of the (supposed) macroscopic
electric eld. Under such conditions, it may be assumed that an
improved photocurrent is mainly due to a very efcient removal
of holes by addition of methanol, independent of where the charge
carriers are generated, and the effect can be dragged down to the
level of individual nanoparticles.
Besides the current doubling effect of methanol, the results
support the hypothesis that the photoelectrochemical response
of TiO2 -SG electrodes may be increased by hole scavenging. For
environmental applications, e.g. waste water remediation, a strong
dependence of the process efciency on chemical nature and
adsorption characteristics of substrates and intermediates is to be
expected.
4. Conclusion
Nanoporous TiO2 -SG and 2%CeTiO2 -SG lm electrodes were
prepared by solgel method, and TiO2 -P25 electrodes were pre-
pared by suspension method. XRD characterization revealed for all
undoped TiO2 electrodes (TiO2 -SG and TiO2 -P25) a phase mixture
of anatase and rutile, whereas 2%CeTiO2 -SG electrodes exhibited
only anatase phase pattern. The result indicates that the incor-
poration of Ce ions prevented the transformation from anatase to
rutile.
The (photo)electrochemical properties of nanoporous TiO2 lm
electrodes were studied by CV and EIS. The Faradaic current under
cathodic polarization in the dark increased with the number of TiO2
layers, and the result may be linked to the reduction of hydrated
Ti(IV) surface sites within the lm electrodes that are penetrated
by the electrolyte solution. The doped 2%CeTiO2 electrodes showed
higher dark currents per deposited mass compared to TiO2 -SG elec-
trodes, a result generally explained by a higher interior surface area
due a decreased probability of homogeneous crystal growth, caused
by Ce addition.
 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124
For UV irradiated electrodes, a photocurrent maximum was
observed for lm electrodes consisting of 20 layers of TiO2 -SG, 15
layers of 2%CeTiO2 -SG and 5 layers of TiO2 -P25, respectively. The
dependence of the photocurrent on the number of deposited layers
may be a consequence of a combination of the penetration depth
of light and the free diffusion length of electrons in relation to the
lm thickness.
In the presence of Ce ions in titania, the observed photocurrent
was comparably small with a delayed signal response. The addition
of Ce ions may lead to an enhanced charge carrier trapping at the Ce
levels within the band gap of TiO2 , resulting in an unwanted higher
electronhole recombination rate.
Comparing irradiated undoped and doped TiO2 -SG lm elec-
trodes, modeling results revealed a higher rate constant for the
heterogeneous charge transfer, a smaller recombination rate and
a higher conductivity for undoped TiO2 -SG lm electrodes. In
addition, the potential dependence, the effect of the incident radi-
ant power density and methanol addition on the photocurrent
response of TiO2 -SG electrodes (20 layers) were studied. The poten-
tial dependence led to the hypothesis, that a macroscopic electric
eld across the nanoporous lm electrode is required to explain
the tting results concerning the heterogeneous charge transfer
(RCT ) and electron transportation to the electrode back contact (ZS )
which favors drift over diffusion. Such a hypothesis would ask for a
macroscopic space charge layer promoting the drift of electrons to
the back contact that would not be restricted to individual particles
or few particle layers next to the contact (ITO collector electrode).
To receive more information about the rate determining pro-
cess, the incident radiant power density (P0 ) was varied. Under
idealized conditions, it may be expected that the photocurrent
density of TiO2 -SG lm electrodes is proportional to the rate of
absorbed photons. In fact, a reciprocal dependence of ZS , the resis-
tive pathway of electron motion to the back contact, on moderate
values of radiant power densities were determined. At high radiant
power density, the tting procedure yielded higher values for ZS ,
explained by enhanced recombination losses of excited electrons
when the penetration depth of light exceeds the zone of promoted
charge separation as a consequence of implementation of RREC as
surface related impedance element only.
The same dependence on the incident radiant power density
was obtained for the heterogeneous charge transfer reaction (RCT ).
For P0 of up to 3.0 W/m2 , a drop of RCT with increasing values of
P0 implies that holes could be removed by surface redox reaction
in an efcient way. At high radiant power densities, the heteroge-
neous charge transfer reaction rate leveled off. This was ascribed
to conditions, where the electronhole recombination affected
the photocurrent response notably. Again, when at high P0 the
irradiated volume exceeds the zone of supported charge carrier
separation, a higher rate of recombination is expected, which would
lead to the observed dependency of RCT .
Addition of methanol as hole-scavenger doubled the steady-
state photocurrent of TiO2 -SG electrodes. The linear reciprocal
dependency of RCT on the photon ux was much more consistent,
in particular at high incident radiant power densities. Under these
conditions, the heterogeneous charge transfer reaction plays a deci-
sive role in the limitation of the overall reaction under anodic bias.
The removal of holes by their reaction with methanol suppressed
the competing electronhole recombination in an efcient way, as
shown by a remarkable drop of RREC , in particular at high radiant
power densities.
In summary, the photochemical characterization gave evidence
that nanoporous TiO2 -SG lm electrodes can be considered as a
hybrid of individual nanoparticles and solid crystalline semicon-
ducting material. Current doubling disregarded, the pronounced
effect of methanol addition, especially at high incident radiant
power densities, on the heterogeneous charge transfer reaction
123
is a feature of individual nanoparticles. The high photocurrent of
relatively thick lm electrodes at moderate radiant power densi-
ties would then be the result of nanoparticles surrounded by the
electrolyte solution which favors a fast hole removal by surface
reactions. Alternatively, the observed dependence of the heteroge-
neous charge transfer reaction (RCT ) on the electrode potential, and
the radiant power density dependence of RCT and the ZS (impedance
of the TiO2 particle network) could be explained by a 3D array
of nanoparticles that behave as a collective like a solid crystalline
electrode.
Future work will deal with the validation of the hypotheses by
EIS characterization of TiO2 -SG electrodes with different lm thick-
ness, investigating potential dependence, direction and power of
incident irradiation, as well as hole scavenging reactions. Focusing
at the possible effect of intrinsic charge separation on the photoac-
tivity of TiO2 , nanoporous lm electrodes with different anatase
to rutile ratios will be prepared by solgel method. Furthermore, a
new photoelectrochemical set-up featuring monochromatic irradi-
ation with tunable output will enable investigations on the spectral
response of the semiconductor electrodes and the determination of
photocurrent quantum yields.
Acknowledgments
The authors acknowledge gratefully support from Synchrotron
Light Research Institute (Public Organization), Thailand (Grant
Number: 2548/05) and the former Chair of Umweltmesstechnik,
University of Karlsruhe, now KIT (Karlsruhe Institute of Technol-
ogy), Germany. Further, we acknowledge support from Deutsche
Forschungsgemeinschaft (DFG).
References
[1] M.R. Hoffmann, S.T. Martin, W. Choi, D.W. Bahnemann, Chemical Reviews 95
(1995) 69.
[2] K. Vinodgopal, S. Hotchandani, P.V. Kamat, Journal of Physical Chemistry 97
(1993) 9040.
[3] B. Sun, P.G. Smirniotis, Catalysis Today 88 (2003) 49.
[4] U.I. Gaya, A.H. Abdullah, Journal of Photochemistry and Photobiology C 9 (2008)
1.
[5] S.-I. Nishimoto, B. Ohtani, H. Kajiwara, T. Kagiya, Journal of the Chemical Society,
Faraday Transactions 81 (1985) 61.
[6] Z. Ding, G.Q. Lu, P.F. Greeneld, Journal of Physical Chemistry 104 (2000) 4815.
[7] T. Ohno, K. Sarukawa, K. Tokieda, M. Matsumura, Journal of Catalysis 203 (2001)
82.
[8] B. Sun, A.V. Vorontsov, P.G. Smirniotis, Langmiur 19 (2003) 3151.
[9] F.B. Li, X.Z. Li, M.F. Hou, K.W. Cheah, W.C.H. Choy, Applied Catalysis A 285 (2005)
181.
[10] K. Nagaveni, M.S. Hegde, G. Madras, Journal of Physical Chemistry B 108 (2004)
20204.
[11] A. Fujishima, K. Honda, Nature 238 (1972) 37.
[12] H. Bartkova, P. Kluson, L. Bartek, M. Drobek, T. Cajthaml, J. Krysa, Thin Solid
Films 515 (2007) 8455.
[13] D. Jiang, H. Zhao, Z. Jia, J. Cao, R. John, Journal of Photochemistry and Photobi-
ology A 144 (2001) 197.
[14] M. Radecka, M. Wierzbicka, M. Rekas, Physica B 351 (2004) 121.
[15] I.D. Raistrick, D.R. Franceschetti, J.R. Macdonald, in: E. Barsoukov, J.R. Mac-
donald (Eds.), Impedance Spectroscopy: Theory, Experiment, and Applications,
Second Edition, John Wiley & Sons, Inc, Hoboken, NJ, USA, 2005.
[16] P.J. Boddy, Journal of Electroanalytical Chemistry 10 (1965) 199.
[17] M. Tomkiewicz, Electrochimica Acta 35 (1990) 1631.
[18] J.R. MacDonald, Journal of Electroanalytical Chemistry 223 (1987) 25.
[19] J.-B. Jorcin, M.E. Orazem, N. Pbre, B. Tribollet, Electrochimica Acta 51 (2006)
1473.
[20] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago, N.S. Ferriols, P. Bodganoff,
E.C. Pereira, Journal of Physical Chemistry B 104 (2000) 2287.
[21] F. Fabregat-Santiago, G. Garcia-Belmonte, J. Bisquert, A. Zaban, P. Salvador,
Journal of Physical Chemistry B 106 (2002) 334.
[22] Y. Pooarporn, A. Worayingyong, M. Wrner, P. Songsiriritthigul, A.M. Braun,
Water Science and Technology 55 (2007) 153.
[23] Radiant power density (P0 ) equates to emitted, transferred, or received radiant
energy per time and area.
[24] R.A. Spurr, H. Myers, Analytical Chemistry (Washington, DC, U. S.) 29 (1957)
760.
[25] B. Liu, X. Zhao, N. Zhang, Q. Zhao, X. He, J. Feng, Surface Science 595 (2005) 203.
124 S. Kityakarn et al. / Electrochimica Acta 83 (2012) 113124
[26] J. Zhang, C. Yang, G. Chang, H. Zhu, M. Oyama, Materials Chemistry and Physics
88 (2004) 398.
[27] V.A. Alves, L.A. Da Silva, J.F.C. Boodts, S. Trasatti, Electrochimica Acta 39 (1994)
1585.
[28] C. Angelinetta, S. Trasatti, Journal of Electroanalytical Chemistry 214 (1986)
535.
[29] A.V. Vorontsov, A.A. Altynnikov, E.N. Savinov, E.N. Kurkin, Journal of Photo-
chemistry and Photobiology A 144 (2001) 193.
[30] M. Pavlovic, B. Santic, D.I. Desnica-Frankovic, N. Radic, T. Smuc, U.V. Desnica,
Journal of Electronic Materials 32 (2003) 1100.
[31] H. Ghr, Electrochemical Applications 1/97 (1997) 2, <http://www.
zahner.de/pdf/ea1997.pdf>.
[32] J.-P. Candy, P. Fouilloux, M. Keddam, H. Takenouti, Electrochimica Acta 26
(1981) 1029.
[33] I.D. Raistrick, Electrochimica Acta 35 (1990) 1579.
[34] J. Bisquert, Physical Chemistry Chemical Physics 5 (2003) 5360.
[35] M.E. Orazem, P. Agarwal, L.H. Garcia-Rubio, Journal of Electroanalytical Chem-
istry 378 (1994) 51.
[36] J. Schefold, Journal of the Electrochemical Society 142 (1995) 850.
[37] A. Zaban, A. Meier, B.A. Gregg, Journal of Physical Chemistry B 101 (1997) 7985.
[38] W.W. Grtner, Physical Review 116 (1959) 84.
[39] M.A. Butler, Journal of Applied Physics 48 (1977) 1914.
[40] P. Lemasson, A. Etcheberry, J. Gautron, Electrochimica Acta 27 (1982) 607.
[41] B. ORegan, J. Moser, M. Anderson, M. Grtzel, Journal of Physical Chemistry 94
(1990) 8720.
[42] A. Hagfeldt, U. Bjrkstn, S.E. Lindquist, Solar Energy Materials and Solar Cells
27 (1992) 293.
[43] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago, Journal of Solid State Elec-
trochemistry 3 (1999) 337.
[44] D. Vanmaekelbergh, P.E. de Jongh, Physical Chemistry B 103 (1999) 747.
[45] F. Fabregat-Santiago, H. Randriamahazaka, A. Zaban, J. Garcia-Canandas, G.
Garcia-Belmonte, J. Bisquert, Physical Chemistry Chemical Physics 8 (2006)
1827.
[46] S. Gimenez, H.K. Dunn, P. Rodenas, F. Fabregat-Santiago, S.G. Miralles, E.M.
Barea, R. Trevisan, A. Guerrero, J. Bisquert, Journal of Electroanalytical Chem-
istry 668 (2012) 119.
[47] N. Hykaway, W.M. Sears, H. Morisaki, S.R. Morrison, Journal of Physical Chem-
istry 90 (1986) 6663.
[48] F. Forouzan, T.C. Richards, A.J. Bard, Journal of Physical Chemistry 100 (1996)
18123.