Band-gap engineering of metal oxides for dye-sensitized solar cells

Article abstract of DOI:10.1021/jp063857c

Mixed oxides Ti_1-xZr_xO₂ with 0 < x ≤ 0.2 were synthesized by means of thermal hydrolysis for use in dye-sensitized solar cells. The lattice parameter d is observed to increase linearly with Zr content x. The band gap of th

Full text of DOI:10.1021/jp063857c

J. Phys. Chem. B 2006, 110, 21899-21902
21899
Band-Gap Engineering of Metal Oxides for Dye-Sensitized Solar Cells
M. D u1 rr,* S. Rosselli, A. Yasuda, and G. Nelles†
Materials Science Laboratory, Sony Deutschland GmbH, 70327 Stuttgart, Germany
ReceiVed: June 20, 2006; In Final Form: September 1, 2006
x 2
Mixed oxides Ti1-xZr O with 0 < x e 0.2 were synthesized by means of thermal hydrolysis for use in
dye-sensitized solar cells. The lattice parameter d is observed to increase linearly with Zr content x. The band
gap of the mixed oxides was measured to increase by up to 0.2 eV. The respective shift of the conduction
band edge leads to an increase of the open circuit voltage (VOC) by up to 0.1 V. Among others, temperature-
dependent measurements of VOC clearly identify the correlation between band edge shift and change in VOC
.
Introduction
Metal-oxide semiconductors, especially TiO2, are intensively
used both in electrochemical and in photoelectrochemical
applications.1 Because of their large internal surface area, por-
ous media are of special interest for high turnover applications.
With respect to photoelectrochemical applications, light is either
directly absorbed by the TiO2, or the semiconductor material
serves as an electron conductor for electrons injected from dye
molecules, which are adsorbed on its surface. The latter concept
finds its application, for example, in dye-sensitized solar cells
-3
(DSSC). In these cells, the dye molecule is regenerated from a
second electrode via a redox couple in electrolyte, and a
photovoltaic current is established between the TiO2- and the
counter-electrode.
Figure 1. Schematics of the dependence of conduction band edge and
valence band (VB) edge as a function of zirconium content x in a mixed
oxide. Band offsets are chosen similarly to the results in ref 5.
4
In both kinds of application, that is, direct absorption or dye-
sensitization, the exact positions of the band edges play a
decisive role for the efficiency of the respective process. For
direct excitation in the semiconductor, the band positions
determine the band gap and therefore the threshold energy of
the absorbed photons. In the case of dye-sensitization, the con-
duction band (CB) edge has to match the lowest excited state
of the dye molecule to enable effective electron transfer from
the molecule’s excited state to the semiconductor’s conduction
band. If the CB edge is too high in energy, that is, higher than
the excited state of the dye molecules, electron injection from
the dye into the conduction band is suppressed. If, on the other
hand, the CB level is low in energy, the electron injection might
be good. However, the electrons lose energy by means of
thermalization in the CB, which is then lost for further photo-
voltaic energy conversion. In the case of DSSC, this shows up
as a reduced open circuit voltage (VOC) because VOC is, among
others, determined by the energy difference between CB edge
and redox potential of the ion conductor.
It is therefore highly desirable to design semiconductors with
tunable band edge positions and band gap. In this contribution,
we describe such band-gap engineering by means of the synthe-
sis of nanoporous titanium-zirconium mixed oxides. As indi-
cated in Figure 1, the band edge positions of Ti1-xZrxO2 are
expected to change with the content of zirconium, x, between
the positions of TiO2 and ZrO2, similar to earlier observed band
edge movement in mixed oxide systems, for example, in ref 6.
An increase of zirconium content should therefore allow for a
higher CB edge and, in consequence, for a higher open circuit
voltage when used in DSSC. Indeed, very recently it was shown
for one fixed value of zirconium content, x ) 5%, that an
7
increase in open circuit voltage can be achieved.
In this contribution, we studied systematically the electrical
properties of Ti1-xZrxO2 as a function of x and especially their
influence on the photovoltaic properties of the dye-sensitized
solar cells. Additionally, the effect of zirconium content on
lattice structure and the optical properties of the nanoparticles
have been studied. Indeed, a monotone increase of lattice
constant and open circuit voltage was observed with increasing
zirconium content. The latter was traced back to the change of
the band edge positions. A widened band gap was directly
observed by means of UV-vis spectroscopy.
Experimental Section
Materials. The synthesis of mixed oxide particles was
performed as follows: (1 - x) mole of titanium isopropoxide,
i
Ti( PrO)4 (Aldrich), was mixed with an equimolar amount of
acetic acid, CH3COOH (VWR). The mixture was stirred until
a pale yellow color appeared. Then x mole of zirconium iso-
i
propoxide, Zr( PrO)4 (Aldrich), was added. The mixture was
poured under continuous stirring into a beaker containing
distilled water. The resulting milky mixed oxide suspension was
heated at 80 °C in the presence of HNO3 (0.1 M, VWR). Finally,
the mixture was poured into a Teflon inlet inside a stainless
steel reactor and heated at 240 °C for 12 h.
*
Corresponding author. Present address: University of Applied Sciences
DSSC Preparation. The DSSC were assembled as follows:
A 30-nm-thick bulk TiO2 blocking layer was formed on FTO
Esslingen. E-mail: Michael.Duerr@hs-esslingen.de.
†
E-mail: Nelles@sony.de.
1
0.1021/jp063857c CCC: $33.50 © 2006 American Chemical Society
Published on Web 10/10/2006

21900 J. Phys. Chem. B, Vol. 110, No. 43, 2006
D u¨ rr et al.
Figure 3. XRD diffractograms of TiO
2
anatase and mixed oxides with
2
various Zr content x. For comparison, a diffractogram of ZrO is shown.
Figure 2. Zr content as measured by EDX for 12 single particles
The peaks of the latter one can be attributed to the monoclinic (M)
and tetragonal (T) phase as exemplified for 2θ ≈ 30°.
i
synthesized with a Zr( PrO)
4
content of x ) 5.7%. Inset: TEM image
of particles made by the same synthesis.
(approximately 100 nm on glass). An approximately 10-µm-
thick porous layer of semiconductor particles was screen printed
on the blocking layer and sintered at 450 °C for half an hour.
If not mentioned otherwise, red-dye (cis-bis(isothiocyanato)-
bis(2,2′-bipyridyl-4,4′-dicarboxylic acid) ruthenium(II)) mol-
ecules were adsorbed to the particles via self-assembling out
of a solution in ethanol (0.3 mM), and the porous layer was
-
-
filled with electrolyte containing I /I3 as redox couple with
-
c(I3 ) ) 15 mM. A reflective platinum back electrode was
attached with a distance of 6 µm from the porous layer.
Characterization. Transmission electron microscopy (TEM)
and energy-dispersive X-ray (EDX) analysis were performed
using a TECNAI G2 F20 transmission electron microscope from
FEI. X-ray diffraction studies exploited the KR line of a copper
anode with a wavelength λ ) 1.54 Å. The lattice spacing d
was then calculated via λ/d ) 2 sin θ/2. External quantum
efficiency measurements were performed at constant light
Figure 4. Change in lattice spacing of mixed oxides with respect to
TiO as obtained from the evaluation of the XRD data shown in the
2
inset (Zr content in mixed oxide of inset was 20%). For the linear fit,
only the data of the lowest five values of Zr content were taken into
account.
2
intensity of 2 mW/cm . Current-voltage characteristics were
for mixed oxide with 20% Zr content in Figure 4, inset. This
behavior clearly indicates that most of the Zr atoms are evenly
distributed in the whole nanocrystal. Otherwise, for example,
in the case of a core-shell formation with TiO2 as core and
ZrO2 as shell material, no such change in the peak position of
the TiO2 anatase should be observed. For a quantitative analysis,
the maximum of the single peaks is determined by fitting
Lorentz-shaped curves to the data, and the respective change
in lattice constant is depicted in Figure 4. One observes a linear
relationship between the increasing lattice constant and the
content of Zr atoms in the mixed oxide, in good agreement with
taken under illumination with white light from a sulfur lamp,
2
and the intensity was 100 mW/cm .
Results and Discussion
Pure TiO2 and mixed oxide was synthesized by means of
thermal hydrolysis starting from either the pure TiO2 precursor,
i
that is, titanium isopropoxide, Ti( PrO)4, or a mixture of Ti-
i
i
(
PrO)4 and zirconium isopropoxide, Zr( PrO)4, respectively. The
ratio of the precursors was found to determine the atomic
composition on the nanoparticles’ scale. This is documented in
Figure 2, where a typical transmission electron micrograph of
mixed oxide particles is depicted together with the atomic ratio
of Zr and Ti in 12 single particles as determined by means of
energy-dispersive X-ray analysis (EDX). A good match between
the initial ratio of the precursors and the ratio of the metal ion
in the particles is observed. The reproduction of the ratio of the
precursors in the mixed oxide particles was also confirmed on
the macroscopic scale by means of X-ray photoelectron
spectroscopy (XPS).
8
,9
Vegard’s law. Only at 20% Zr content, the measured lattice
spacing deviates from this linear behavior. This is in accordance
with the observation that, starting from 11% of Zr content, addi-
tional features appear in the XRD diffractogram of the mixed
oxides around 2θ = 30°. Most likely, they origin from smaller
segregations with a majority of ZrO2 as a comparison with data
1
0,11
of pure ZrO2 suggests.
As a consequence, the actual Zr
content in the Ti-rich phase of the mixed oxide is smaller than
20%, leading to a lower increase of the lattice spacing.
Furthermore, X-ray diffraction (XRD) measurements of the
mixed oxide material show that for lower Zr content the Zr
atoms are distributed homogeneously in the nanoparticles. As
shown in Figure 3, the anatase structure of TiO2 prevails in the
case of the mixed oxides. A more detailed analysis of the XRD
data from Figure 3 shows that the peak maxima shift toward
smaller angles when the Zr content is increased. As an example,
the peak at 2θ = 25° is shown both for pure TiO2 as well as
The increased Zr content does not only account for the
increased lattice constant but also for a larger band gap in
accordance with the scenario depicted in Figure 1. Experimental
evidence is obtained from UV-vis spectroscopy of both TiO2
and mixed oxide layers. For the latter one, the onset of absorp-
tion starts at lower wavelengths. For example, two spectra are
shown in the inset of Figure 5, both for pure TiO2 and for a
mixed oxide with x ) 0.2. From the different threshold for

Metal Oxides for Dye-Sensitized Solar Cells
J. Phys. Chem. B, Vol. 110, No. 43, 2006 21901
Figure 7. External quantum efficiency EQE as a function of irradiating
2
wavelength for TCPP-Pd colored layers of TiO and mixed oxide (MO,
x ) 20%). Inset: Schematics of the energy levels (not to scale).
Figure 5. Open circuit voltage VOC of DSSC as a function of Zr content
in the mixed oxide and linear fit to the data; each value represents the
averaged results of three cells from a single, representative data set,
and error bars are calculated from the standard deviation. Inset: Total
absorption of 2-µm-thick films as measured by means of an integrating
a change in band positions rather than a changed recombination
rate. A reduced recombination rate causes a different slope of
VOC(T), whereas a change in band positions leads to a change
in VOC(T ) 0 K) but the temperature dependence stays
2 2
sphere for both TiO and Ti0.8Zr0.2O .
13,14
unchanged.
The latter behavior is even better observed when
extrapolating the data to T ) 0 K, as shown in the inset of
Figure 6.
Second, external quantum efficiency (EQE) measurements
are shown in Figure 7. The shift of the conduction band mini-
mum does not only change the open circuit voltage but also
influences the injection efficiency from the excited state of the
dye molecules to the conduction band of the semiconductor
because the energy difference is the main driving force for this
1
5
mechanism. One therefore observes in general a lower short
circuit current density the higher the conduction band minimum
1
6
is located. This is also observed for the mixed oxides of this
contribution. Additionally, external quantum efficiency mea-
surements with porous layers stained with porphyrin dye
molecules (5,10,15,20-tetrakis(4-carboxyphenyl)-porphyrin-
Pd(II), short: TCPP-Pd) directly image the shift of the conduc-
tion band edge. The EQE spectra of the TCPP-Pd colored TiO2
layers in Figure 7 show two maxima around 530 and 400 nm
due to the absorption of the first two electronic excitations of
Figure 6. Open circuit voltage VOC as a function of temperature T for
cells both made from TiO and made from Ti0.8Zr0.2 (two samples
each). The linear fits show a similar slope for all cells. Inset: Zoom-
out of the same data.
2
O
2
absorption, an increase of the band gap by approximately 0.2 eV
was derived. If one assumes a similar shift of both the valence
17
the TCPP-Pd, that is, the Q-band and the S-band, respectively.
For TiO2, the ratio between the two maxima is about 1.5, in
5
band edge and the conduction band edge, then both should
17
agreement with previous investigations. For the mixed oxide,
move by about 0.1 eV, which sets the upper limit for the change
in VOC. Indeed, an increase of VOC by up to 80 mV for DSSC
made from mixed oxide when compared to pure TiO2 was
observed (Figures 5 and 6). A closer look at Figure 5 reveals
that, at low Zr content, the effect of the band widening on VOC
is not so pronounced; up to a Zr content of nominally 5.7%,
not much change of VOC is observed. However, at this point it
should be noted that VOC is not only dependent on the energy
difference between conduction band minimum and redox
however, one observes not only an in general lower intensity
but the ratio between the two EQE maxima is increased by a
factor of 5. As illustrated in the inset of Figure 7, this clearly
indicates that the conduction band edge of the mixed oxide is
shifted toward higher values: such a shift affects the injection
from the lower excited state (Q-band) much more than injection
from the higher excited state (S-band).
With respect to power conversion efficiency, we observe
improvements only for the DSSC with mixed oxides of low Zr
content, x ) 1% (Figure 8). With higher Zr content and therefore
higher conduction band minimum, we observe a drop in short
circuit current density due to the reduced driving force for elec-
tron injection into the semiconductor, which cannot be com-
pensated by the higher VOC. The photovoltaic parameters taken
from the best cells’ data set compare as follows: for the best
-
-
potential of the redox couple I /I3 . Recombination losses
between electrons in the conduction band and tri-iodide of the
electrolyte can additionally decrease VOC.1
2-14
A changed
surface morphology and additional recombination channels may
therefore (over)compensate for the effect of band-gap widening
in the case of lower Zr content in the mixed oxides.
2
To further strengthen our argumentation on the relationship
between Zr content, band position, and open circuit voltage,
two additional experiments were performed. First, the depen-
dence of VOC on temperature T was measured as displayed in
Figure 6. The data for the cells with mixed oxide are in general
higher but show the same linear dependence on temperature as
the cells with pure TiO2. This behavior is indeed indicative of
cell based on pure TiO2, we measured JSC ) 15.6 mA/cm , VOC
) 700 mV, and fill factor FF ) 0.64, which results in a power
conversion efficiency of 7.0%, whereas for the best cell based
2
on mixed oxide we measured JSC ) 16.5 mA/cm , VOC ) 715
mV, FF ) 0.69 and a resulting power conversion efficiency of
8.1%. Both porous layers were measured to be 8 µm thick. In
addition to the increased VOC due to the shift of the conduction

21902 J. Phys. Chem. B, Vol. 110, No. 43, 2006
D u¨ rr et al.
semiconductors with a wider band gap. The so far widely
unemployed longer wavelength region of the solar spectrum
becomes then more easily accessible when a semiconductor with
lower CB minimum and a dye absorbing in the longer wave-
length region is chosen in a further compartment.
Acknowledgment. We would like to thank Michael Steiert
for the XRD measurements, as well as A. Bamedi, G. Fuhrmann,
and M. Obermaier for fruitful discussions and experimental help.
Note Added after Print Publication. Due to a production
error, the corresponding author’s e-mail address was incorrect
in the original published version of this article (Web
Figure 8. J-V characteristics of DSSC with porous layers consisting
10/10/2006; print Volume 110, No. 43, pp 21899-21902). The
both of pure TiO
2
and of mixed oxide with 1% Zr content. Active area
2
corrected electronic version was reposted on November 22, 2006
and an Addition and Correction appears in the December 28,
was 0.24 cm .
2
006 issue (Vol. 110, No. 51).
gap of the mixed oxide, an increase in JSC is observed, which
is attributed to a slight increase of internal surface area from
References and Notes
2
2
8
6 m /g for the pure TiO2 material to 97 m /g for the mixed
(
1) Fujishima, A.; Honda, K. Nature 1972, 238, 37.
oxide, in agreement with ref 7. The higher FF might be also
associated with the changed morphology of the nanoporous film.
Differences in the diffusion pathways through the nanoporous
network can lead to a reduced resistivity of the film and hence
an increased FF.18
(2) Linsebigler, A. L.; Lu, G.; Yates, J. T. Chem. ReV. 1995, 95, 735.
(3) Fujishima, A.; Rao, T. N.; Tryk, D. A. J. Photochem. Photobiol.,
C 2000, 1, 1.
4) Gr a¨ tzel, M. Nature 2001, 414, 338.
In summary, we could show that the use of mixed Ti-Zr
oxides in DSSC leads to an increasing open circuit with
increasing Zr content x. It was traced back to a shift of the
conduction band edge toward higher energies. Improved power
conversion efficiency was found for small content of Zr. Toward
higher values of x, the reduced electron injection due to the
higher conduction band causes a drop in power conversion
efficiency. Nonetheless, the presented results have some impor-
tance with respect to further improvements of DSSC. Current
dye molecules are optimized for the combination with TiO2;
however, the successful demonstration of mixed oxides in DSSC
might trigger new dye synthesis with respect to these new
materials. Especially with respect to tandem solar cell concepts,
108, 20204.
(7) Kitiyanan, A.; Ngamsinlapasathian, S.; Pavasupree, S.; Yoshikawa,
S. J. Solid State Chem. 2005, 178, 1044.
(8) Vegard, L. Z. Phys. 1921, 5, 17.
(9) Denton, A. R.; Ashcroft, N. W. Phys. ReV. A 1991, 43, 3161.
10) Noh, H.-J.; Seo, D.-S.; Kim, H.; Lee, J.-K. Mater. Lett. 2003, 57,
(
2
425.
(11) Manr ´ı quez, M. E.; L o´ pez, T.; G o´ mez, R.; Navarrete, J. J. Mol.
Catal. A 2004, 220, 229.
(12) Nazeeruddin, M. K.; et al. J. Am. Chem. Soc. 1993, 115, 6382.
(
13) Kron, G.; Egerter, T.; Werner, J. H.; Rau, U. J. Phys. Chem. B
003, 107, 3556.
14) D u¨ rr, M.; Yasuda, A.; Nelles, G. Appl. Phys. Lett. 2006, 89, 061110.
2
2
(
(15) Wang, Z.-S.; Yamaguchi, T.; Sugihara, H.; Arakawa, H. Langmuir
005, 21, 4272.
(
(
16) Moser; et al. Z. Phys. Chem. 1999, 212, 85.
17) D u¨ rr, M.; et al. Proc. 19th EPSEC 2004, 1, 21.
1
9,20
which aim toward a better use of the full sun spectrum,
one
(18) D u¨ rr, M.; Schmid, A.; Obermaier, M.; Rosselli, S.; Yasuda, A.;
Nelles, G. Nat. Mater. 2005, 4, 607.
19) D u¨ rr, M.; Bamedi, A.; Yasuda, A.; Nelles, G. Appl. Phys. Lett.
004.
20) Shevaleevskiy, O.; Larina, L.; Lim, K.-S. Proc. 15th PVSEC 2005,
2, 754.
can easily think of an optimization of one or both of the
compartments with the help of band-gap engineered semicon-
ductors. For example, the compartment that uses dyes absorbing
in the shorter wavelength region might be combined with
(
2
(