Available online at www.sciencedirect.

com

Journal of Power Sources 181 (2008) 46–55

Importance of the band gap energy and flat band potential for
application of modified TiO2 photoanodes in water photolysis
M. Radecka a,∗ , M. Rekas a , A. Trenczek-Zajac a , K. Zakrzewska b
a Faculty of Material Science and Ceramics, AGH-University of Science and Technology, al. Mickiewicza 30, 30-059 Krakow, Poland
b Faculty of Electrical Engineering, Automatics, Computer Science and Electronics, AGH-University of Science and Technology,

al. Mickiewicza 30, 30-059 Krakow, Poland

Received 8 September 2007; received in revised form 24 October 2007; accepted 26 October 2007
Available online 7 November 2007

Abstract
The influence on water photolysis of two important parameters of the electronic structure of photocatalytic semiconductors: the forbidden band
gap, Eg , that decides about the absorption spectrum and the flat band potential, VFb , that affects the recombination probability, was studied. The
photoelectrochemical experiments were performed in a three-electrode cell PEC with a TiO2 thin film photoanode immersed in liquid electrolyte
of variable pH. Titanium dioxide photoanodes doped with chromium (up to 16 at.%) and tin (up to 50 at.%) were prepared by rf reactive sputtering.
Different methods of flat band potential determination: Mott–Schottky plots and photocurrent versus voltage characteristics were used. The
energy band gap was derived from the spectrophotometric measurements of optical transmittance and reflectance coefficients of thin films. For
TiO2 + 7.6 at.% Cr high and negative flat band potential VFb = −0.72 eV (at pH 4) has been found but the recombination time τ = 8 s was the shortest
of all TiO2 modifications. Despite additional absorption feature at about 2.8 eV, i.e., at wavelength corresponding to visible range of the light
spectrum, the photoconversion efficiency of TiO2 + 7.6 at.% Cr was found to be much smaller (ηc = 0.1%) than that of undoped TiO2 (ηc = 1.8%)
and TiO2 doped with 8 at.% of Sn (ηc = 1.0%).
© 2007 Elsevier B.V. All rights reserved.

Keywords: Titanium dioxide; Flat band potential; Forbidden band gap; Photoelectrochemical cell; Water photolysis; Hydrogen generation

1. Introduction TiO2 (3.0 eV rutile, 3.2 eV anatase) before it was realized that
severe requirements imposed on the photoanode material could
Photoelectrochemical splitting of water is an environmentally not be met simultaneously by any existing semiconductor. These
friendly method of hydrogen generation based on renewable and requirements include:
apparently unlimited natural resources such as water and solar
energy. Assisted by solar radiation, direct splitting of water into (i) high stability and resistivity to corrosion and photocorro-
molecular hydrogen and oxygen was demonstrated for the first sion;
time in 1972 by Fujishima and Honda [1] in a photoelectrochem- (ii) low cost and availability;
ical cell PEC with an n-type semiconductor TiO2 photoanode. (iii) conduction band minimum, EC , above the H2 O/H2 electro-
Since that time, many efforts have been undertaken to chemical level of water reduction EC > EH2 O/H2 ;
improve the conversion efficiency of the process but more than (iv) valence band maximum EV below the O2 /H2 O electrochem-
30 years later this method is still far from commercialization. The ical level of water oxidation EV < EO2 /H2 O ;
reasons for this are fundamental and come as a consequence of (v) effective absorption of photons of the solar spectrum related
a considerable mismatch between the spectra of light absorption to the band gap in the photon energy range of 1.6–1.9 eV.
in TiO2 and that of solar radiation. Many other semiconductors
such as GaAs had been tried as a replacement for wide-band-gap
As titanium dioxide in the form of anatase fulfils all but the
last one condition, it has been admitted that the best way to the
∗ Corresponding author. Tel.: +48 126172526; fax: +48 126172493. improvement of the performance of the photoelectrochemical
E-mail address: [email protected] (M. Radecka). devices would be to modify the absorption spectrum of TiO2 .

0378-7753/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jpowsour.2007.10.082
 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55 47

This can be achieved by shifting the fundamental absorption photoelectrochemical process it is desired to have the flat band
edge to longer wavelengths or by creating additional absorption potential large and negative.
features within the band gap. The methods tried up till now In this work, we have used TiO2 thin films as photoanodes
include: cation doping [2], sensitization with organic dyes [3], in the process of water photolysis. Thin films were deposited by
composite materials [4], and recently proposed, anion doping rf reactive sputtering and modified by cation dopants such as Cr
with N, C or S [5]. However, whereas it is relatively easy to and Sn.
affect the absorption spectra of TiO2 by these methods, this is It is well-established, especially in the case of TiO2 thin
not in general true for the photocatalytic efficiency. The limiting films [6,7], that technological conditions of a deposition process
factor is the recombination rate of the photoexcited electrons affect the crystallographic structure, microstructure and surface
and holes. properties. The substrate temperature during film growth, oxy-
Therefore, there are two important parameters of the elec- gen partial pressure and doping have a tremendous impact on
tronic structure of photocatalytic semiconductors: the forbidden the interplay between two polymorphic forms of TiO2 , i.e.,
band gap that decides about the absorption spectrum and the flat anatase and rutile. A significant contribution from the amor-
band potential that affects the recombination probability. Fig. 1 phous phase dominates at lower growth temperatures [7]. It has
explains the meaning of these parameters. been demonstrated [8] that the anatase is the best form of all tita-
At the interface between the semiconducting photoanode and nium dioxide polymorphs as far as the photoactive properties are
the electrolyte, the band bending occurs (see Fig. 1). It is a concerned.
result of interface phenomena: solid-electrolyte. Moreover, the In our previous papers [9,10] we have reported on the com-
Fermi level EF in a semiconductor and electrochemical level prehensive structural and morphological studies of undoped and
Eredox in the electrolyte are equal at the equilibrium state. The modified by cation doping TiO2 thin films grown by sputtering.
space charge region is formed at the interface. This space charge Here, we concentrate on the importance of the flat band poten-
region provides a strong electric field that is indispensable for tial and band gap energy (forbidden band gap) for application
an effective separation of photoexcited electrons from holes. of these materials in the water photolysis. It should be pointed
On the other hand, if light is absorbed in the bulk of the pho- out that these parameters are very sensitive to the changes in the
toanode, the photoexcited electrons and holes are created but film microstructure and morphology.
there is a high probability they will recombine before being The aim of this work is to show how different modifications
used for water photolysis. Therefore, if light is absorbed in of TiO2 affect both the flat band potential and the forbidden band
this region, charge separation and field-assisted transport are gap. The influence of these parameters of the electronic structure
expected. on the photoelectrochemical efficiency of water splitting will be
The band bending is also affected by the external voltage demonstrated.
VB . For a given semiconductor and electrolyte, there exists a
unique potential for which the potential drop between the surface 2. Experimental
and the bulk is zero and there is no space charge layer. This
is the flat band potential VFb . From the point of view of the 2.1. Preparation of modified TiO2 thin film photoanodes

Titanium dioxide thin films were deposited by sputtering of

Ti target in Ar + O2 reactive gas atmosphere onto different sub-
strates such as Corning glass, amorphous silica, titanium and
carbon foils, NaCl, etc., depending on the requirements imposed
by further applications [9–14]. In the case of photoanodes, Ti
foil was used as a substrate. Modification of TiO2 properties was
achieved by cation doping with Cr and Sn. Chromium-doped
TiO2 photoanodes were obtained by sputtering from mosaic
Ti + Cr targets. It was possible to reach up to 16 at.% of Cr in
TiO2 thin films by changing the relative target area covered with
Cr discs. Deposition of TiO2 –SnO2 thin films required much
higher degree of technical sophistication related to the target
construction. This was a consequence of a large difference in
the melting points of Ti (1940 K) and Sn (500 K) as well as in
the sputtering rates of these two metals in the atmosphere con-
taining oxygen. Photoanodes of TiO2 containing up to 50 at.% of
Sn were obtained by sputtering from a dielectric Ti/SnO2 target
Fig. 1. Energy E diagram of the semiconducting photoanode/liquid elec- of varied surface coverage with SnO2 .
trolyte/metalic cathode system used for water photolysis. The flat band potential,
The influence of a variety of technological parameters such
VFb and band gap, Eg are shown. Energy levels separated by 1.23 eV and corre-
sponding to redox reactions E(H2 O/H2 ), E(O2 /H2 O) as well as the conduction as the substrate temperature TS and oxygen partial pressure on
and valence band edges are included. Upon the absorption of photon of the the film properties has been investigated. Detailed description of
energy hν a pair of electron e− and a hole h+ is created. rf and dc magnetron sputtering systems used for deposition of
48 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55

thin films of titanium dioxide can be found in our earlier papers

[9–14].

2.2. Material characterization

As-prepared thin films were subjected to a standard proce-

dure of material characterization. The chemical composition
of nonstoichiometric TiO2−x , TiO2 :Cr and TiO2 –SnO2 was
determined by X-ray electron microprobe (EMP) for 1 ␮m-
thick layers and Rutherford backscattering (RBS) for thinner
films. The details are given in [9,11]. The classical meth-
ods of morphology characterization such as scanning electron
microscopy (SEM) are not sensitive enough in the case of reac-
tively deposited thin oxide films due to the smoothness of their
surface. As determined from atomic force microscopy (AFM)
performed at contact mode (see for example [11,13]) the surface
roughness changes from 3 to 15 nm depending on the techno-
logical conditions of film deposition while the higher values are
obtained for thin films annealed at 1280 K. The results of trans-
mission electron microscopy (TEM) performed with JEM-1000
microscope indicate that thin films (up to 50 nm) of TiO2 are
amorphous to a large extent while those of SnO2 crystallize eas-
ily even when deposited at room temperature [14]. Annealing at
770 K induces a fast crystallization of TiO2 thin films deposited
by sputtering with precipitation of large anatase grains.
Fig. 2. The influence of substrate temperature TS on X-ray diffraction GID
Titanium dioxide thin films deposited onto amorphous silica
patterns of undoped TiO2 thin films deposited by rf sputtering. The film thickness
substrates were analyzed by X-ray diffraction at GID (grazing is of about 200 nm. A stands for anatase; R, rutile; fR represents the relative
incidence) geometry using X’Pert MPD Philips diffractometer. contribution of rutile calculated from Eq. (1).
The anatase-to-rutile relative composition as well as the contri-
bution from the amorphous phase were found to be a function of
the substrate temperature, as shown in Fig. 2. Deposition onto
unheated substrates (TS = 290 K) yields a mixture of anatase
and rutile with a large amount of amorphous phase. Starting
from TS = 520 K the anatase phase begins to dominate while at
TS = 670 K rutile almost completely disappears. The rutile rel-
ative contribution fR calculated as the intensity ratio of rutile
R(1 1 0) IR and anatase A(1 0 1) IA X-ray diffraction peaks [15]:

1.265 IIAR
fR = (1)
1 + 1.265 IIAR

decreases systematically with an increase in the substrate tem-

perature over this range of temperatures. At temperatures higher
than 770 K, the irreversible transformation from anatase to rutile
initiates spontaneously [12]. The temperature and rate of this
transformation depends on many factors among which the most
important is the type of dopants (acceptor or donor). Accep-
tors such as Cr3+ introduce oxygen vacancies and accelerate
anatase–rutile transition. As can be seen in Fig. 3, at the anneal-
ing temperature of Ta = 970 K the structures of undoped TiO2
and TiO2 + 2.8 at.% Cr are partially transformed into rutile while
anatase is still present.
Incorporation of Sn4+ into TiO2 lattice suppresses the growth
of anatase crystals [11]. The rutile phase only is observed for
TiO2 + 33 at.% Sn as demonstrated in Fig. 3. The same con- Fig. 3. X-ray diffraction GID patterns for TiO2 thin films doped with Cr and Sn,
clusion can be drawn for lower amount of Sn as well as for deposited by rf sputtering and annealed at Ta = 970 K. The results for undoped
as-sputtered and annealed films. TiO2 annealed at 970 K are given for comparison.
 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55 49

2.3. Band gap energy determination axis at α = 0 (Fig. 4b). As one can see here, deviation x from the
stoichiometric composition affects the band gap of TiO2−x thin
In order to determine the band gap energy Eg of trans- films.
parent oxide semiconductors, the optical uv–vis spectroscopy
has been performed. Optical measurements of the transmit- 2.4. Determination of the flat band potential and the
tance T and reflectance R coefficients for thin films deposited photoelectrochemical experiments
onto transparent, amorphous silica have been carried out
over λ = 250–2500 nm wavelength range with Lambda 19 There exist at least four methods of determination of the flat
PerkinElmer spectrophotometer. Example of T(λ) and R(λ) band potential [18]. In this work we have used two of them: the
spectra for a TiO2 thin film of 200 nm thickness is given in first one stems from the impedance spectroscopy in dark and
Fig. 4a. The oscillations clearly seen in T(λ) and R(λ) result the second employs the current–voltage characteristics under
from the interference of light at both air-film and film–substrate illumination.
interfaces and allow to calculate the absorption coefficient The impedance spectra and capacitance of the elec-
α by means of the envelope method [16]. The region over trode/electrolyte interface versus voltage (Mott–Schottky plots)
which the drastic drop of the transmittance coefficient occurs were recorded using a frequency response analyser (Solartron
(3.25 eV < hν < 4 eV) corresponds to the fundamental absorption Model 1260) equipped with an electrochemical interface
edge related to the forbidden band gap. The value of Eg can be (Solartron Model 1287).
found from the plot of (αhν)m as a function of the photon energy Photoelectrochemical experiments in dark and under illumi-
hν. Fig. 4b illustrates such plot for TiO2−x . The power coefficient nation were performed in a custom-made photoelectrochemical
m takes a value indicating the type of the predominant electronic cell, PEC described in detail in [13]. The three-electrode PEC
transition [17] and equals 1/2 for indirect allowed transitions. constructed for the experiments consists of TiO2 photoanode, a
This so-called Tauc plot, i.e, the following dependence: cathode made of Pt foil covered with Pt black and SCE as a refer-
ence electrode. The electrodes are immersed in liquid electrolyte
αhν = α0 (hν − Eg )1/m (2)
containing a buffer solution of 0.04 M CH3 COOH + 0.04 M
can be fitted to the experimental data. The band gap Eg is derived H3 PO3 + 0.004 M H3 BO3 + xKOH and aqueous solution of
from the intercept of this straight line with the photon energy KOH. The pH of the electrolyte can be varied from 4 to 10
by changing x in the buffer solution from 10 to 14.
The PEC was illuminated with a white or monochromatic
light provided by a 450 W Xe lamp and a TRIAX 180 Jobin
Yvon monochromator. The power density at the photoanode was
controlled by a power meter. The wavelength dependence of
the incident radiation was taken into account when calculating
the photoconversion efficiency and comparing the photocurrent
spectra. The area of the light spot on the photoanode surface
equals to 1 mm2 while the average power density is of about
85 kW m−2 .
ANKO potentiostat was used in the measurements of current
versus voltage (I–V) in order to keep a constant potential dif-
ferent. Keithley 6517 A electrometer recorded the photocurrent
values. Kinetics of the photocurrent rise and decay, i.e., IPh ver-
sus time t, were studied in order to gain some insight into the
mechanism of relaxation processes. The photocurrent spectra,
i.e., IPh (λ) were reported previously [19] while in this paper we
discuss mainly the results obtained with the white light.

3. Results and discussion

3.1. Band gap energy

Band gap energy Eg , i.e., the forbidden energy is a very

important parameter related to the electronic structure of a semi-
conducting photoanode. This parameter has to match at least the
energy difference 1.23 eV between the redox levels H2 O/H2 and
O2 /H2 O required for water splitting. In the case of TiO2 the band
Fig. 4. Determination of energy band gap Eg for TiO2 thin films. (a) Spectral gap is even too high, as it amounts to 3.0 eV for rutile and 3.2 eV
dependence of transmittance T(λ) and reflectance R(λ) coefficients. (b) Tauc’ for anatase and in reality depends on many factors such as a
plot (αhν)1/2 vs. photon energy hν; α is the absorption coefficient. degree of crystallization and doping type and level. Obviously,
50 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55

the relative contributions from anatase, rutile and amorphous and the conduction band occur at much lower photon energies
phase affect the band gap energy to a large extent in the case of (of about 2.4–2.8 eV) than those of the fundamental absorp-
thin films as shown in Fig. 5a. tion corresponding to the forbidden band gap Eg of TiO2 . As
Thin films of undoped TiO2 deposited by rf sputtering are expected, the Edop decreases considerably with the increasing
characterized by higher degree of amorphousness that the films concentration of Cr.
obtained in dc magnetron sputtering [9,13] hence the energy Fig. 6 shows the effect of TiO2 doping with isovalent Sn4+
band gap of rf sputtered TiO2 is larger. With the increasing rutile ions on the forbidden band gap. The ionic radius of Sn4+ equals
content fR the Eg decreases to about 3.0 eV at fR = 0.56 which to 0.071 nm and is slightly greater than that of Ti4+ (0.068 nm).
correlates well with the band gap of rutile. Fig. 5a illustrates However, the TiO2 –SnO2 forms a solid solution over a full com-
dependence of Eg versus rutile/anatase ratio, on the other hand positional range above a certain critical temperature that depends
Fig. 5b demonstrates the effect of doping of TiO2 with Cr+3 on on the system composition [21]. The optical spectra for thin
the optical transitions. The solubility limit of Cr3+ in TiO2 is of films indicate one transition only at the energy between 3.2 and
about 10 at.% but in the case of thin films even 16 at.% of Cr 3.4 eV, which is a function of the substrate temperature and post-
does not allow for precipitation of Cr-based phases [12]. deposition annealing temperature. The band gap energy changes
There are two optical transitions clearly seen (Fig. 5b). The only slightly with the increasing Sn4+ content up to 50 at.%. It
fundamental transition from the valence to the conduction band has been reported [22] that the SnO2 additions to TiO2 matrix
takes place at the energy Eg that obviously depends on the rutile affect the electronic structure of TiO2 by creating the admixture
content fR (as shown in Fig. 5a) similarly to the case of undoped of extended 5s Sn states to the conduction band predominantly
TiO2 . It has been reported previously [12] that the increased composed of localized 3d Ti orbitals. The continuous modi-
concentration of Cr3+ resulted in the increased rutile content in fication of the electronic structure from d-type (TiO2 ) to the
the films. Additionally, Cr3+ is known to create allowed states s-type conduction band characteristic for SnO2 is expected for
Edop deep in the forbidden band of TiO2 [20] as presented in TiO2 –SnO2 system. The energy band gap determined from the
the inset of Fig. 5b. The optical transitions between these states fundamental optical transitions has been found to increase con-
tinuously with the increasing SnO2 concentration from about
3.2–3.3. eV for TiO2 to 3.6–3.7 eV for SnO2 [23,24].

3.2. Flat band potential

The solid/electrolyte interface can be described by a three

layer model [25]. Within the electric double layer one can dis-
tinguish: the space charge region in the electrolyte, i.e., Gouy
layer (G) or diffuse part of the ionic layer 1–10 nm thick, an inter-
mediate region called a Helmholtz layer (H) of 0.4–0.6 nm and
the space charge layer in the semiconductor (SC) with thickness
of 10–100 nm. These layers contribute to the total capacitance
of the solid/electrolyte interface as if their capacitances where
connected in series 1/C = 1/CSC + 1/CH + 1/CG , but as the thick-
nesses of both Gouy and Helmholtz layers are small as compared

Fig. 5. (a) The energy band gap Eg as a function of the relative contribution of
rutile fR for undoped TiO2 and Cr-doped TiO2 . (b) Energies of the optical tran-
sitions over the band gap Eg and between Cr acceptor level and the conduction Fig. 6. Band gap energy Eg of Sn-doped TiO2 thin film photoanodes as a function
band Edop for TiO2 :Cr thin films deposited by rf sputtering as a function of Cr of the atomic composition Sn/(Sn + Ti); Ts , substrate temperature. Reference
concentration. data for anatase powder and rutile single crystal are given.
 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55 51

with that of the space charge layer in the semiconductor, their

contributions to the total capacitance C can be neglected.
From the impedance spectroscopy one can find the total
capacitance C of the solid/electrolyte interface and draw the so-
called Mott–Schottky plot, i.e., C−2 as a function of the applied
voltage VB . The total capacitance C can be approximated by
the capacitance CSC of the depleted semiconductor electrode.
This assumption is valid as long as the level of doping of the
semiconductor is low.
The CSC depends on the flat band potential in the following
way:
  
−2 2 kT
CSC = VB − VFb − (3)
εεo eND S 2 e
where e is the electron charge, ND denotes the donor density, ε
is the dielectric constant of the semiconductor, ε0 is the vacuum Fig. 8. Current–voltage I(VB ) characteristics of a TiO2 electrode in PEC in dark
permittivity, T is the absolute temperature, k is the Boltzman and under illumination with the white light at pH 4 and pH 8.
constant and S is the surface area of the electrode.
The Mott–Schottky plots for thin films of undoped TiO2 and The best VFb from the point of view of the photoelectrochem-
TiO2 doped with Cr are shown in Fig. 7. The results given in ical process is obtained at the doping level of 2.8 at.% of Cr in
Fig. 7 were obtained in the measurements performed at the fre- TiO2 . At this doping level the concentration of donors ND is the
quency of 10 kHz. It was verified in our earlier investigations that highest.
this frequency was within the region in which both the slope and The flat band potential has been determined from the pho-
the intercept did not depend on the frequency [26]. tocurrent measurements in the PEC with TiO2 photoanodes, as
From the best fit to the linear part of the C−2 (VB ) the flat well. Fig. 8 shows the typical current–voltage characteristics
band potential VFb , donor concentration ND and the width of the in dark and under illumination for an undoped thin film TiO2
depletion region W have been determined (see inset of Fig. 7). photoanode at different pH values. The magnitude and sign of
The width W of the depletion region cannot be determined the applied external potential VB affect the dark current and the
independently of the applied potential VB as shown in the fol- photocurrent. The band bending decreases upon illumination of
lowing equation: a semiconductor. The photocurrent approaches the dark current
  near VFb .
2εε0 (VB − VFb − kT/e) 1/2 The threshold of photocurrent at VB = VON can be treated as
W= (4)
e2 N D a measure of the flat band potential VFb . For TiO2 thin film pho-
toanodes, the flat band potential is negative and significantly
where the symbols have the meaning as in Eq. (3).
differs from zero. This means that even undoped titanium diox-
The values of W plotted in the inset of Fig. 7 were obtained
ide should have low enough recombination rate. As can be
for VB − VFb = 1 V.
seen in Fig. 8, if VB > VFb the anodic photocurrent, significantly
higher than the dark current, flows. Semiconductor electrode
works as a photoanode in the PEC and water photolysis takes
place as long as VB > VFb . On the contrary, at VB < VFb much
larger dark cathodic current flows and the system cannot work
as a photoelectrolizer.
In order to calculate more accurately the flat band potential
one has to note that in the vicinity of VFb the recombination
process tends to suppress the photocurrent, hence the threshold
of the photocurrent is not exactly equal to the flat band potential.
Better approach to the problem of VFb determination has been
proposed by Butler [27]:
  
ND IPh 2
VB − VFb = (5)
2eεε0 αI0
where I0 is intensity of the incident radiation and other symbols
have the same meaning as in Eq. (3).
Fig. 7. Capacitance C vs. applied voltage VB , C−2 (VB ) Mott–Schottky plots of To apply Eq. (5), the semiconductor electrode has to be
Cr-doped TiO2 electrode at pH 4 and pH 8. VB is determined against standard
illuminated with the monochromatic light at wavelength λ com-
calomel electrode, SCE. Density of donors, ND (left hand side axis) and thickness
of depleted layer W (right hand side axis) vs. Cr concentration are given in the parable with λG corresponding to the band gap excitations. Fig. 9
inset. presents how the flat band potential can be derived from the best
52 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55

Table 1
Comparison of the values of the flat band potential VFB
0
at pH = 0 for different
forms of TiO2

Photoanode materials VFb

0
(V) Reference

Rutile −0.20 [8]

Single crystal, rutile −0.16 [28]
Single crystal, rutile −0.27 [29]
Single crystal, rutile −0.26 [30]
Anatase −0.40 [8]
Anatase −0.32 [31]
Thin films, anatase −0.25 [32]
Thin films, anatase −0.37 [33]
This work
Thin films, rf, d = 300 nm, fR = 0.14 −0.19
d = 500 nm −0.17
Fig. 9. Normalized photocurrent (IPh /P)2 as a function of the applied voltage
Thin films dc magnetron, d = 700 nm, fR = 0.35 −0.14
VB at specific wavelengths λ in the vicinity of the fundamental absorption edge
of TiO2 photoanode; P radiation power. TiO2 :Cr, rf, d = 300 nm This work
2.8 at.% Cr −0.69
linear fit of Eq. (5) to the experimental IPh
2 (V ) dependence. 4.4 at.% Cr −0.60
B
7.6 at.% Cr −0.55
The photocurrent was normalized to the radiation power P. Each
plot has been recorded at a certain wavelength λ in the vicin-
ity of the fundamental absorption edge. The values of the flat potential at pH = 0, that can be compared for different photoan-
band potential VFb obtained this way are independent of the odes. According to Kalayanansundaram and Grätzel [8] VFb 0 is

wavelength when λ is exactly within the region of the funda- affected by the crystallographic structure of the TiO2 photoan-
mental absorption (λ = 300–350 nm) and correspond very well ode and amounts to about −0.4 V for anatase and −0.2 V for
with those calculated from the Mott–Schottky plots (Fig. 7). rutile. The form of the electrode material, i.e., thin film, sin-
According to Nernst equation: gle crystal, polycrystalline ceramics as well as doping are also
decisive factors as far as VFB
0 is concerned.
RT ln 10
VFb = VFb
0
− pH (6) Fig. 10 represents an overview of the experimental results
F with modified thin film photoanodes of TiO2 obtained in the
where R is universal gas constant, F is Faraday constant, T is the course of these studies. The straight line represents the best fit
temperature. of Nernst equation (Eq. (6)) to the experimental data of VFb . The
The flat band potential of a semiconducting photoanode is values of the VFb were calculated from the experimental data of
as a linear function of the pH of the electrolyte. This type of IPh (VB ) for TiO2 photoanodes of the same thickness measured
dependence along with the best fit described by Eq. (6) is given at different pH. Different series of photoanodes: of different Cr
in Fig. 10. In fact, it is the parameter VFb0 , i.e., the flat band content, different thickness for rf sputtering and dc magnetron
are also included. The values of VFb 0 estimated from the results

Fig. 10. The flat band potential VFb vs. pH of the electrolyte for different TiO2 Fig. 11. Photocurrent, IPh as a function of the applied voltage VB for TiO2 thin
photoanodes (undoped and modified) in PEC. film photoanodes (undoped and modified).
 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55 53

presented in Fig. 10 along with the values reported by other consequence of lower electrical conductivity of TiO2 doped
authors are listed in Table 1. It is clearly seen that the more with Cr [9].
negative values of VFb
0 are obtained for anatase. Doping with Cr Fig. 12 demonstrates three kinetics of the photocurrent, i.e.,
increases the absolute value of VFb
0 . the responses to the sudden switching on and switching off
the white light for the undoped, Cr-doped and Sn-doped TiO2
photoanodes.
3.3. Photoconversion kinetics and efficiency
Two different types of photocurrent responses are observed.
In the case of TiO2 and TiO2 :Cr the initial increase in IPh
The effect of doping on the photocurrent IPh is clearly seen
is followed by an exponential decrease with time. This initial
in Fig. 11. The threshold voltage depends on doping and was
maximum (anodic spike) is caused by a separation of the photo-
found to be −0.28 V versus SCE for undoped TiO2 photoanode,
generated electron–hole pairs at the semiconductor/electrolyte
−0.64 V for TiO2 + 2.8 at.%Cr and −0.42 V for TiO2 + 8 at.%
interface [34]. When the light is switched off, a cathodic spike
Sn.
is observed due to the recombination of the conduction band
It is shown that the photocurrent is the highest for the film
electrons with the holes trapped at the surface [35].
of TiO2 and the lowest for TiO2 :Cr. This is undoubtedly a
The kinetics of IPh of the undoped sample is about three
times slower than that of the Cr-doped sample as indicated by
the values of the recombination time τ calculated from the expo-
nential rise of IPh (for details see [26]). It means that Cr doping
accelerates the recombination process. It is well known that
incorporation of foreign ions into the TiO2 lattice increases the
concentration of point defects such as CrTi , V0 •• , Ti4+
i [36]. It
seems reasonable to assume that in the case of TiO2 the mech-
anism of recombination at point defects prevails.
Different shape of IPh versus time is observed for Sn-doped
sample (Fig. 12c). The value of the recombination time is higher
than that of undoped TiO2 . This suggests that tin does not create
additional recombination centres as chromium does.
The overall efficiency of a PEC, so-called solar conversion
efficiency ηc , has been defined by Parkinson [37] as:
output power
ηc = (7)
input power
In the case of water photo-electrolysis Eq. (7) assumes a form:
G◦ (H2 O)R(H2 ) − VB IPh IPh (1.23 − VB )
ηc = = (8)
PS PS
where Go (H2 O) is a standard free enthalpy of formation of
1 mole of liquid water (kJ mol−1 ), R(H2 ) is the rate of hydrogen

Fig. 12. Transient photocurrent IPh responses of (a) TiO2 , (b) TiO2 :Cr (7.6 at.%
Cr) and (c) TiO2 –SnO2 (8 at.% Sn) thin film electrodes in an electrolyte of pH 8
under illumination with white light at 0 V vs. SCE. Arrows indicate when light Fig. 13. Solar energy conversion efficiency ηc as a function of the applied
was turned on and off; τ is the recombination time. potential VB for TiO2 thin film photoanodes (undoped and modified) in PEC.
54 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55

Table 2
Summary of the results obtained in this work for undoped and modified thin film titanium dioxide photoanodes
Thin films photoanode Band gap energy, Eg (eV) Flat band potential Recombination Solar conversion efficiency
VFb (V) at pH 4 losses τ (s) ηc (%) at VB = 0 V

TiO2 rf, d = 300 nm 3.38 ± 0.05 −0.37 27 1.8

TiO2 + 7.6 at.%Cr Eg = 3.32 ± 0.05 −0.72 8 0.1
Edopt = 2.83 ± 0.05
TiO2 + 8 at.% Sn 3.29 ± 0.05 −0.43 35 1.0

generation (mol s−1 ), VB is the potential applied to the cell (V), undoped TiO2 . Further comparison shows that the recombina-
IPh is the photocurrent within a cell, P is radiation power density tion time is slightly higher while the conversion efficiency, not
(W m−2 ) and S is the irradiated electrode area (m2 ). as good as for undoped TiO2 , remains however much better than
The parameter ηc can be calculated from the results such as that of Cr-doped TiO2 .
those presented in Fig. 11 using Eq. (8). The analysis performed in this work, indicates that an
Fig. 13 shows the solar conversion efficiency ηc obtained in improvement of photoelectrochemical performance cannot be
this work for thin film photoanodes of titanium dioxide: undoped reached through optimization of one parameter of the electronic
and doped with chromium and tin. These results are comparable structure the semiconducting photoanode, only. The energy band
with those of other authors [1,38–42]. gap that governs the absorption process and photoelectron–hole
Application of undoped TiO2 thin film photoanodes in PEC pair generation as well as the flat band potential that affects
results in the highest value of ηc . The lowest ηc , obtained when charge transfer in PEC and the recombination process are impor-
Cr-doped TiO2 photoanodes are used, is related to high recom- tant from the point of view of the photoconversion efficiency of
bination losses. the PEC system but there certainly are others factors that make
this picture more complicated.
4. Conclusions
Acknowledgement
The photolysis of water, i.e., its decomposition into hydrogen
and oxygen as a result of the sunlight absorption in semicon- This work was supported by Polish Ministry of Science and
ductor electrode, is considered as the most promising source Higher Education grant no. 3T08D 054 30 (2006–2008).
of renewable energy. The efficiency of water photolysis is deter-
mined by the intrinsic properties of the semiconducting materials
applied as photoanodes. The most important parameters of the References
electronic structure, pertinent to the performance of the photo-
electrochemical cell, PEC are: the band gap energy Eg and the [1] A. Fujishima, K. Honda, Nature 238 (1972) 37–38.
[2] J.M. Hermann, J. Disdier, P. Pichat, Chem. Phys. Lett. 108 (1984) 618–622.
flat band potential VFb . The influence of cation dopants Cr3+ and [3] M.K. Nazeeruddin, A. Kay, I. Rodicio, R. Humphry-Baker, E. Müller,
Sn4+ on Eg and VFb of TiO2 thin film photoanodes was stud- P. Liska, N. Vlachopoulos, M. Grätzel, J. Am. Chem. Soc. 115 (1993)
ied. The resulting PEC performance quantitatively described by 6382–6390.
recombination time τ and solar conversion efficiency ηc were [4] G. Zhao, H. Kozuka, T. Yoko, Thin Solid Films 277 (1996) 147–154.
[5] H. Wang, J.P. Lewis, J. Phys. Condens. Mat. 18 (2006) 421–433.
derived from the dynamic changes of the photocurrent IPh and
[6] W.T. Pawlewicz, P.M. Martin, D.D. Hays, I.B. Mann, Proc. SPIE 325 (1982)
the photocurrent IPh versus voltage VB characteristics. All these 105–116.
parameters are listed in Table 2. [7] P. Löbl, M. Huppertz, D. Mergel, Thin Solid Films 251 (1994) 72–79.
The analysis of the results indicates that in spite of the ben- [8] K. Kalayanansundaram, M. Grätzel, Coordin. Chem. Rev. 77 (1998)
eficial influence of Cr3+ on the flat band potential, which is the 347–414.
[9] K. Zakrzewska, M. Radecka, M. Rekas, Thin Solid Films 310 (1997)
most negative from all TiO2 modifications, the recombination
161–166.
time is the shortest and the photoconversion efficiency is the [10] K. Zakrzewska, Thin Solid Films 391 (2001) 229–241.
lowest. This may be accounted for by the enhanced recombina- [11] K. Zakrzewska, M. Radecka, J. Przewoźnik, K. Kowalski, P. Czuba, Thin
tion of the photoelectrons and holes probably due to the fact that Solid Films 490 (2005) 101–107.
Cr3+ forms recombination centers. Acceptor level located deep [12] M. Radecka, K. Zakrzewska, M. Wierzbicka, A. Gorzkowska, S. Komor-
nicki, Solid State Ionics 157 (2003) 379–386.
in the forbidden bang gap gives additional absorption feature at
[13] A. Brudnik, A. Gorzkowska-Sobaś, E. Pamuła, M. Radecka, K.
longer wavelength than that corresponding to the fundamental Zakrzewska, J. Power Sources 173 (2007) 774–780.
absorption edge of TiO2 . However, this effect does not improve [14] F. Edelman, H. Hahn, S. Seifried, C. Alof, H. Hoche, A. Balogh, P. Werner,
the photoconversion efficiency. K. Zakrzewska, M. Radecka, P. Pasierb, A. Chack, V. Mikhelashvili, G.
On the contrary, the influence of Sn4+ on the spectral depen- Eisenstein, Mater. Sci. Eng. B69–70 (2000) 386–391.
[15] R.A. Spurr, H. Myers, Anal. Chem. 29 (1957) 760.
dence of the absorption in TiO2 is not as promising as in the
[16] J. Szczyrbowski, J. Phys. D: Appl. Phys. 11 (1978) 583–591.
case of Cr3+ . The energy band gap increases in TiO2 –SnO2 sys- [17] J. Tauc, Mater. Res. Bull. 5 (1970) 721–730.
tem with the increased SnO2 content. Surprisingly, the flat band [18] M. Radecka, M. R˛ekas, K. Zakrzewska, Trends Inorg. Chem. (2007), in
potential of TiO2 doped with Sn is more negative than that of print.
 M. Radecka et al. / Journal of Power Sources 181 (2008) 46–55 55

[19] A. Gorzkowska-Sobaś, E. Kusior, M. Radecka, K. Zakrzewska, Surf. Sci. [32] L. Kavan, M. Grätzel, Electrochim. Acta 40 (1995) 643–652.
600 (2006) 3964–3970. [33] C. da Fonseca, S. Boudin, M.C. Belo, J. Electroanal. Chem. 379 (1994)
[20] G. Campet, J. Verniolle, J.-P. Doumerc, J. Claverie, Mater. Res. Bull. 15 173–180.
(1980) 1135–1139. [34] A. Hagfeldt, H. Lindström, S. Södergren, S.-E. Lindquist, J. Electroanalyt.
[21] V.S. Stubican, A.H. Schultz, J. Am. Ceram. Soc. 55 (1968) 290–291. Chem. 381 (1995) 39–46.
[22] M. Radecka, P. Pasierb, K. Zakrzewska, M. Rekas, Solid State Ionics 119 [35] N.W. Duffy, L.M. Peter, R.M.G. Rajapakse, K.G.U. Wijayantha, Elec-
(1999) 43–48. trochem. Commun. 2 (2000) 658–662.
[23] K.M. Glassford, J.R. Chelikowsky, Phys. Rev. B 46 (1992) 1284–1289. [36] J. Nowotny, M. Radecka, M. Rekas, J. Phys. Chem. Solids 58 (1997)
[24] J. Melsheimer, D. Ziegler, Thin Solid Films 129 (1985) 35–41. 927–937.
[25] H. Gerischer, in: B.O. Seraphin (Ed.), Solar Energy Conversion, Solid State [37] B. Parkinson, Acc. Chem. Res. 17 (1984) 431–437.
Physics, Springer-Verlag, New York, 1979. [38] J.F. Houlihan, D.B. Armitage, T. Hoovler, D. Bonaquist, D.P. Madacsi,
[26] M. Radecka, M. Wierzbicka, S. Komornicki, M. Rekas, Phys. B 348 (2004) L.N. Mulay, Mater. Res. Bull. 13 (1978) 1205–1212.
160–168. [39] A.J. Nozik, Nature 271 (1978) 137.
[27] M.A. Butler, J. Appl. Phys. 48 (1977) 1914–1920. [40] A.K. Ghosh, H.P. Maruska, J. Electrochem. Soc. 124 (1977) 1516–
[28] E.C. Dutoit, F. Cardon, W.P. Gomes, J. Appl. Electrochem. 8 (1978) 1522.
247–252. [41] J. Akikusa, S.U.M. Khan, Int. J. Hydrogen Energy. 22 (1997) 875–
[29] L.J. Handley, A.J. Bard, J. Electrochem. Soc. 127 (1980) 338–343. 882.
[30] M.S. Wrighton, P.T. Wolczanski, A.B. Ellis, J. Solid State Chem. 22 (1977) [42] N. Giordano, V. Antonucci, S. Cavallaro, R. Lembo, J.C.J. Bart, Int. J.
17–29. Hydrogen Energy 7 (1982) 867–872.
[31] D. Duonghong, J. Ramsden, M. Grätzel, J. Am. Chem. Soc. 104 (1982)
2977–2985.