J. Phys. Chem.

C 2007, 111, 1035-1041 1035

Comparison of Dye-Sensitized ZnO and TiO2 Solar Cells: Studies of Charge Transport and
Carrier Lifetime

Marı́a Quintana,† Tomas Edvinsson,‡ Anders Hagfeldt,‡ and Gerrit Boschloo*,‡

Facultad de Ciencias, UniVersidad Nacional de Ingenierı́a, P. O. Box 31-139, AVenida Tupac Amaru 210,
Lima, Perú, and Center of Molecular DeVices, Department of Chemistry, Royal Institute of Technology (KTH),
Teknikringen 30, 100 44 Stockholm, Sweden
ReceiVed: September 12, 2006; In Final Form: NoVember 6, 2006

Nanocrystalline particles of ZnO and TiO2 of approximately equal size (∼15 nm) were used to prepare
mesoporous electrodes for dye-sensitized solar cells. Electron transport in the solar cells was studied using
intensity-modulated photocurrent spectroscopy and revealed very similar results for ZnO and TiO2. Apparent
activation energies for electron transport in nanostructured ZnO of e0.1 eV were calculated from the
temperature dependence of transport times under short-circuit conditions. The lifetime of electrons in the
nanostructured semiconductors was evaluated from open-circuit voltage decay and intensity-modulated
photovoltage spectroscopy. Significantly longer lifetimes were obtained with ZnO. Despite the reduced
recombination, ZnO-based solar cells performed worse than TiO2 cells, which was attributed to a lower electron
injection efficiency from excited dye molecules and/or a lower dye regeneration efficiency. The internal
voltage in the nanostructured ZnO film under short-circuit conditions was about 0.23 V lower than the open-
circuit potential at the same light intensity. Results may be explained using a multiple trapping model, but as
electrons are usually only shallowly trapped in ZnO, an alternative view is presented. If there is significant
doping of the ZnO, resulting band bending in the nanocrystals will form energy barriers for electron transport
and recombination that can explain the observed properties.

Introduction more flexibility in synthesis and morphologies than TiO2. The

chemical stability of ZnO is, however, less than that of TiO2,
Dye-sensitized solar cells based on nanocrystalline mesopo- which was found to be problematic in the dye adsorption
rous metal oxide films have attracted much attention in recent procedure.6
years.1,2 They offer the prospect of low-cost photovoltaic energy
The solar-to-electrical energy conversion efficiencies of dye-
conversion. Promising solar to electrical energy conversion
sensitized ZnO solar cells are so far significantly lower than
efficiencies of more than 10% have been achieved,2 and good
those reported for TiO2. The highest reported values are 5% at
progress has been made on long-term stability.3 The working
1/10 sun17 and 4.1% at 1 sun.18 Several reports suggest that
mechanism of dye-sensitized solar cells differs completely from
dye adsorption is the main problem in dye-sensitized ZnO solar
conventional p-n junction solar cells,2 but, after more than 15
cells.6,19 Cells with high dye loading tend to be inefficient,
years of research, is still not completely resolved. Research has
whereas cells with lower dye loading show good quantum
largely focused on nanostructured TiO2 (anatase) as the metal
efficiencies. These problems are mainly related to the high
oxide to which the dye is bound. Good results have, however,
acidity of the carboxylic acid binding groups of the dyes that
also been obtained using other n-type metal oxides, such as
can lead to dissolution of ZnO and precipitation of dye-Zn2+
ZnO,4-6 Nb2O5,7 and SnO2.8
complexes, leading to a poor overall electron injection efficiency
ZnO is an attractive material for nanoscale optoelectronic of the dye.
devices, because it is a wide band gap semiconductor with good
In comparison to TiO2, few electron transport studies have
carrier mobility and can be doped both n-type and p-type.9
been done on dye-sensitized nanostructured ZnO solar cells.20-22
Electron mobility is much higher in ZnO than in TiO2, while
In this study we compare the electron transport, accumulation,
the conduction band edge of both materials is located at
and recombination properties of dye-sensitized solar cells based
approximately the same level. One would therefore expect
on nanostructured electrodes of ZnO and TiO2. There is a
nanostructured ZnO to be a good candidate as electron acceptor
remarkable similarity in the transport properties of the two
and transport material in dye-sensitized solar cells. A large range
materials, despite the significant difference in the mobility and
of fabrication procedures is available for ZnO nanostructures,
the effective mass of electrons in the conduction band of the
such as sol-gel processes,4,10 chemical bath deposition,11-13
pure, single-crystalline materials.
electrodeposition,14,15 and vapor-phase processes.16 Different
morphologies such as spherical particles,4,10 rods,13 wires,16 and
hollow tubes12 can be prepared with relative ease. ZnO shows Experimental Procedures
Preparation of Nanostructured ZnO and TiO2 Films. ZnO
* To whom correspondence should be addressed. Phone: +46 87908178. colloids were prepared in ethanol by addition of tetramethyl-
Fax: +46 87908207. E-mail: gerrit@kth.se.
† Universidad Nacional de Ingenierı́a. ammonium hydroxide (25% in methanol) to a suspension of
‡ KTH. zinc acetate in ethanol.19 The resulting ZnO sol was refluxed at
10.1021/jp065948f CCC: $37.00 © 2007 American Chemical Society
Published on Web 12/14/2006
1036 J. Phys. Chem. C, Vol. 111, No. 2, 2007 Quintana et al.

80 °C for 30 min. The suspension was left to cool down and

settle overnight, decanted, washed with ethanol, and finally
concentrated until visibly viscous. Transparent nanostructured
ZnO electrodes were obtained by depositing the paste onto
conducting glass substrates (TEC8, Pilkington) by doctor
blading, followed by heating in a hot-air stream at 380 °C for
30 min. The porosity of the resulting films was 50%. Transpar-
ent nanostructured TiO2 electrodes were prepared from HNO3-
stabilized TiO2 colloids autoclaved for 15 h at 200 °C.1
Electrodes were heated at 450 °C for 30 min. Film thickness
(3-4 µm) was determined by profilometry.
Dye Sensitization and Cell Assembly. The concentration
of the dye bath and the adsorption time strongly influence the
efficiency of dye-sensitized ZnO solar cells.6 In this study a
rapid sensitization method was used.23 A 8 µL volume of a
20 mM N719 solution in dimethyl sulfoxide (DMSO) was
applied to the ZnO or TiO2 electrode (at 50 °C) and left for 1
min (N719 corresponds to (TBA)2 cis-Ru(Hdcbpy)2(NCS)2). The
dyed electrode was rinsed with ethanol, dried, and assembled
with a platinized conducting glass counter electrode using a
50 µm thick thermoplastic frame (Surlyn 1702). The electrolyte
composition was as follows: 0.5 M LiI; 50 mM I2 in
3-methoxyproprionitrile with either 0.1 M 1-methylbenzimida-
zole (electrolyte 1) or 0.5 M 4-tert-butylpyridine (electrolyte
2) as an additive.
Characterization Methods. UV-vis spectra were recorded
using a Hewlett-Packard 8453 diode array spectrometer. The
setups for recording incident photon to current efficiency (IPCE)
spectra and I-V curves under simulated sunlight have been
described elsewhere.19,24 Intensity-modulated photocurrent and
photovoltage spectroscopy (IMPS and IMVS, respectively) and
charge extraction measurements were performed using a 10 mW
diode laser (Coherent Lablaser, λ ) 635 nm) or a light-emitting
diode (Lumiled Luxeon Star 1W, λmax ) 640 nm) as the light
source as described previously.25 Red light was chosen to obtain
a relative uniform light absorption in the film. Measurements
were carried out at room temperature, except for the tempera-
ture-dependent IMPS experiments, which were performed in
the range of 6-60 °C using a Peltier element.

Results and Discussion

Film Characterization. Transmission electron microscopy
(TEM) revealed that colloidal ZnO solution consisted of
crystalline particles with an average size of about 15 nm; see
Figure 1a. The colloidal TiO2 solution used in this study
contained slightly smaller nanoparticles with an average size
of 10 nm found by TEM. The X-ray diffraction spectrum of a
sintered nanostructured ZnO film is shown in Figure 1b and
shows peaks characteristic of wurtzite. The grain size, calculated
from the peak broadening using the Scherrer formula, was
16 nm. The TiO2 films consist of 14-nm-size anatase crystals
with a trace of brookite (data not shown). The results suggest
that significant crystal growth occurs in the TiO2 films during
heat treatment, but not in ZnO films. The resulting nanostruc-
tured ZnO and TiO2 films were fully transparent.
Cyclic Voltammetry. The nanostructured ZnO and TiO2
electrodes were characterized using cyclic voltammetry in
aqueous electrolyte. The experiment was performed on bare
(non-sensitized) electrodes that were scanned toward negative
potential and back. Typical results are shown in Figure 2. The
current onset for electron accumulation in ZnO is at slightly
more positive potentials than that of TiO2, in agreement with
similar experiments by Willis et al.26 As the conduction band
edges of both materials are expected to be located at similar
Figure 1. (a) Transmission electron microscopic picture of colloidal
ZnO. The bar corresponds to 100 nm. (b) X-ray diffractogram of a
ZnO film sintered at 380 °C.
potentials, this might indicate the presence of more deep traps
in ZnO. At more negative potentials (-0.65 to -1 V), however,
currents were much smaller for the ZnO than for the TiO2
electrode. The relatively good reversibility of the voltammo-
grams suggests that the current is mainly due to charging or
discharging of the nanostructured electrode by electrons. On
the basis of the difference in effective density of states of the
two semiconductor materials, higher electron concentrations and
therefore larger currents can be expected for TiO2.27 Calculated
electron densities in the semiconductors are also shown in Figure
2. The electron concentration increases approximately expo-
nentially with more negative applied potential for TiO2, n ∝ n0
exp(-6.2Vappl), whereas a clearly different behavior is observed
for ZnO. The charging of the electrodes can be explained in
terms of filling of conducting band states and/or trap states and
will be discussed further below in the section on Traps in ZnO.
Solar Cell Characterization. A rapid method for dye
adsorption was used to sensitize the films. This method was
chosen to prevent sensitization problems in the case of ZnO,
which can slightly dissolve in acidic environment, resulting in
a precipitate of Zn2+ and the dye.6 The optimum adsorption
Dye-Sensitized ZnO and TiO2 Solar Cells
Figure 2. Cyclic voltammograms of nanostructured ZnO and TiO2
electrodes at a scan rate of 10 mV s-1. The electrolyte was a deaerated
aqueous solution of 0.2 M KCl with 0.01 M phosphate buffer, pH 6.7.
The thicknesses of the ZnO and TiO2 films were 1.9 and 1.5 µm,
respectively. Below and on the right-hand axis the calculated electron
densities in ZnO and TiO2 are shown, assuming 50% porosity and
calculated from the scan toward negative potentials. A correction for
the capacitance of the substrate/electrolyte interface (∼28 µF cm-2)
has been made.
Figure 3. Absorption spectra of N719 dye adsorbed on ZnO and TiO2
films (thickness 4 µm). The spectra are corrected for ZnO, TiO2, and
substrate absorption.
time was 1 min for ZnO solar cells. UV-vis absorption spectra
of sensitized ZnO and TiO2 films are shown in Figure 3. Slightly
more of the N719 dye was adsorbed by TiO2. Importantly, the
absorption maximum of the adsorbed dye on ZnO was not blue-
shifted. This gives evidence that there is no formation of the
Zn2+-dye complex.6
Current-voltage characteristics of the ZnO and TiO2 solar
cells in simulated sunlight are shown in Figure 4a. TiO2 solar
cells give higher power conversion efficiencies than correspond-
ing ZnO solar cells, as they give higher photocurrents, open-
circuit potentials, and fill factors. The overall efficiency was
found to be rather low for all cells (<3%), as the nanostructured
films were relatively thin. Also, the rapid dye adsorption
procedure may be a cause for low efficiency, although good
results with this method have been reported by Nazeeruddin et
al. for TiO2-based solar cells.23
IPCE spectra of N719-sensitized ZnO and TiO2 solar cells
are shown in Figure 4b. The photocurrent generation efficiencies
are higher for TiO2 than for ZnO, being 50% and 37% at the
maximum at 520 nm for TiO2 and ZnO, respectively. The
incident photon to current conversion efficiency can be described
as the product of four terms:
IPCE ) (LHE)ΦinjΦregηc (1)
J. Phys. Chem. C, Vol. 111, No. 2, 2007 1037
Figure 4. Solar cell characteristics of solar cells based on 4 µm thick
nanostructured ZnO and TiO2 films with rapid dye adsorption (1 min
in 20 mM N719 in DMSO). (a) I-V curves under illumination
corresponding to 100 W/m2 with AM 1.5G spectral distribution (1/10
sun). (b) IPCE spectra. (c) APCE spectra.
where LHE is the light-harvesting efficiency, defined as LHE
) 1 - 10-A with A is the sample absorbance, Φinj is the quantum
yield for electron injection, Φreg is the quantum yield for
regeneration of the oxidized dye by iodide, and ηc is the
efficiency of collecting injected electrons at the conducting
substrate. The absorbed photon to current efficiency (APCE),
shown in Figure 4c, was calculated by dividing the IPCE by
the LHE. In most of the visible light range APCE values for
ZnO and TiO2 solar cells are rather constant at about 40% and
60%, respectively. As will be discussed later, losses during
charge collection appear to be very low. The rather low quantum
efficiencies must therefore be attributed to poor electron
injection efficiency and/or dye regeneration efficiency. The rapid
sensitization method is likely to give dye aggregates that are
less efficient in both electron injection28 and dye regeneration
by iodide.
Electron Transport Studies. Several time and frequency-
domain measurement techniques were used to further character-
ize the nanostructured ZnO and TiO2 solar cells. The electron
transport through the nanostructured metal oxides was studied
using intensity-modulated photocurrent spectroscopy under
short-circuit conditions,29-31 while the electron lifetime was
probed using intensity-modulated photovoltage spectroscopy
under open-circuit conditions.31-33 Figure 5a shows the IMPS
and IMVS time constants as function of light intensity. The
1038 J. Phys. Chem. C, Vol. 111, No. 2, 2007
Figure 5. (a) IMPS and IMVS time constants as a function of light
intensity (λ ) 640 nm) for N719-sensitized ZnO and TiO2 solar cells
with electrolyte 1. Drawn lines are power-law fits. (b) Extracted charge
at short-circuit conditions (QSC) as a function of light intensity for the
same solar cells. (c) Temperature dependence of the electron transport
(τIMPS) in a N719-sensitized ZnO solar cell with electrolyte 2. The short-
circuit current densities and the apparent activation energies are
indicated in the figure.
general trend for all time constants is that they decrease with
increasing light intensity. The IMVS time constant, which
corresponds to the electron lifetime τe, is larger in ZnO than in
TiO2 solar cells. As τe is much larger than τIMPS, τIMPS can be
interpreted as the electron transport time in the investigated solar
cells.33 Electron transport times in nanostructured ZnO were
similar to those in TiO2, and they change in the same way as a
function of the incident light intensity. It is noted that the
electron mobility in single crystals is at least 1 order of
magnitude larger in ZnO34 than in TiO2 (anatase).35 The
similarity of the electron transport times in the nanostructured
metal oxide solar cells suggests that they are to a large part
determined by the morphology of the film. The films consist
of porous networks of nanocrystals, having a high percentage
of surface atoms and many grain boundaries. It was demon-
strated recently by Galoppini et al. that very rapid transport
(<30 µs) occurred in DSSCs based on ZnO nanorods.22 In the
single-crystalline nanorods electron transport will not be limited
by grain boundaries.
The amount of charge present in the nanostructured metal
oxide films was determined by photocurrent transients under
short-circuit conditions. The current after switching off the light
was measured over 1 s and integrated numerically to obtain
Quintana et al.
Figure 6. (a) Open-circuit voltage decay transients of dye-sensitized
nanostructured ZnO and TiO2 solar cells with electrolyte 1. (b)
Calculated electron lifetime (eq 2) vs open-circuit potential.
QSC. It is noted that a fraction of the electrons can recombine
before they are extracted. The fact that τe is much larger than
τIMPS suggests, however, that most charge will be extracted.
Furthermore, not all electrons may be extracted during the
integration period. This can result in a systematic underestimate
of the charge. The extracted charge increases with light intensity
and is quite similar for the ZnO and TiO2 solar cells at higher
light intensities; see Figure 5b. At lower intensities, more charge
was extracted for the ZnO cell. At the highest light intensity,
the charge in the ZnO electrode was about 10-5 C cm-2. As
the ZnO film is 4 µm thick and 50% porous, this corresponds
to an electron concentration in the ZnO of 3 × 1017 cm-3, or
on average about 0.6 electron per nanocrystal assuming spherical
particles with a diameter of 16 nm.
Results of temperature-dependent IMPS measurements on
dye-sensitized nanostructured ZnO under short-circuit conditions
are shown in Figure 5c. Short-circuit current densities were
practically independent of the temperature. IMPS time constants
decreased with increasing temperature, indicating that the charge
transport process is thermally activated. Apparent activation
energies (Ea), determined using the Arrhenius equation, were
rather small, 0.07-0.10 eV, and similar to those measured for
dye-sensitized TiO2 under similar conditions.25
Electron Recombination and Potential-Charge Relation.
Figure 6a shows the decay of the open-circuit potential that
follows after switching off the light for ZnO and TiO2 solar
cells. The decay is significantly slower in the ZnO cell, in
agreement with the larger IMVS time constants found for ZnO.
The electron lifetime can be calculated from the voltage
transients using eq 2:36
τe ) -
e ( )
kT dVOC
dt
-1
(2)
where k is the Boltzmann constant, T is the absolute temperature,
and e is the positive elementary charge. The calculated electron
lifetimes are shown in Figure 6b as function of the open-circuit
potential.37 The lifetime increases approximately exponentially
with decreasing voltage. For the TiO2 cell there appears to be
Dye-Sensitized ZnO and TiO2 Solar Cells
Figure 7. Extracted charge as function of open-circuit potential for a
dye-sensitized nanostructured ZnO cell with electrolyte 2. The drawn
line is an exponential fit.
a change in recombination mechanism when the voltage
decreases. This may be related to the relative importance of
recombination via the conducting glass substrate (SnO2:F) that
may be dominant over the recombination via the nanostructured
TiO2 at lower potentials.38 In the ZnO solar cell, however, no
such effect is seen. This could be a consequence of a more
complete coverage of the conducting glass in case of the
nanostructured ZnO electrode due to the absence of polymer in
the preparation solution. Additionally, some dissolved zinc ions
may have been present in the ZnO colloidal solution, which
could have resulted in a dense ZnO underlayer on the conducting
glass substrate.39
The relation between potential and charge in the nanostruc-
tured ZnO solar cell was investigated using a charge extraction
technique.40 The solar cell was illuminated under open-circuit
conditions, the potential was left to decay in the dark for a fixed
time, and finally the charge was extracted by discharging over
a 100 ohm resistor. Results are shown in Figure 7. At an open-
circuit potential of 0.56 V the electron concentration per ZnO
nanocrystal is about 5. The charge increases roughly exponen-
tially with potential: Q ) Q0 exp(V/m), with Q0 ) 5.5 × 10-6
C cm-2 and m ) 0.19 V. It is noted that a better fit is obtained
using a power-law equation. Part of the extracted charge is
related to discharge of double-layer capacitance of the conduct-
ing glass/electrolyte interface, but its contribution is relatively
small at higher potentials. The shape of electron density vs
potential curve of the ZnO solar cell is similar to that of the bare
ZnO electrode (Figure 2), but absolute values were larger by a
factor ∼3 for the latter. The difference may be explained by
differences in the electrolytes and in the ZnO batch that was used.
To get insight into the electrochemical potential present in
the illuminated dye-sensitized nanostructured ZnO film under
short-circuit conditions, the following experiment was per-
formed: the solar cell was illuminated for some time under
short-circuit conditions, then the light was switched off and the
cell switched to open circuit simultaneously.25 The voltage that
developed gives a good indication of the electrochemical
potential (the quasi Fermi level) under short-circuit conditions;
see Figure 8a. We will refer to the maximum voltage that
develops as the “short-circuit voltage” VSC. Figure 8b shows
VSC and VOC as a function of light intensity. The difference is
constant, ∼230 mV. Similar results have been obtained for dye-
sensitized nanostructured TiO2.25 Using transient photovoltage
as a tool to study recombination, O’Regan and Lenzmann
concluded that the quasi Fermi level inside dye-sensitized
nanostructured TiO2 was lowered by 240 mV when changing
from open-circuit to short-circuit conditions under 80 mW cm-2
white-light illumination.41
The electron lifetime in ZnO cells under short-circuit condi-
tions can be estimated using VSC and the lifetime-potential
J. Phys. Chem. C, Vol. 111, No. 2, 2007 1039
Figure 8. (a) Voltage transients of a dye-sensitized ZnO solar with
electrolyte 2 recorded in the dark at open circuit after illumination (0.2-
24 mW cm-2) under short-circuit conditions. (b) Open-circuit potential
(VOC) and internal potential under short-circuit conditions (VSC) in a
dye-sensitized ZnO solar cell as a function of light intensity.
relation shown in Figure 6b. Electron lifetimes under short-
circuit conditions appear to be very long, more than 10 s. This
means that no significant recombination takes place during
charge transport in the nanostructured ZnO. The low IPCE
values for dye-sensitized ZnO solar cells must therefore be
attributed to poor electron injection efficiency.
Comparison with Exponential Trapping Model. The
electron transport properties in nanostructured TiO2 are fre-
quently explained using trapping/detrapping models. A large
part of the electrons is assumed to be trapped in localized states
below the conduction band edge. Trapped electrons can be
thermally excited to the conduction band and move freely before
they are trapped again. The multiple trapping/detrapping process
increases the transport time for the electrons strongly, and it
can explain the light-intensity dependence of the transport by
the energy distribution of the traps. Specifically, an exponential
increase of the trap state density NT(E) toward the conduction
band edge (eq 3) fits the experimental data well, as it results in
power-law dependencies of transport time and charge with light
intensity.31,42
NT(E) ) NT0 exp (E - EF0
m ) (3)
In eq 3 NT0 is the trap state density at EF0, which is the Fermi
level of the TiO2 in the dark, and m is the slope of the trap
distribution. Considering the similarities in the results obtained
with TiO2 and ZnO, we now apply this model to nanostructured
ZnO. From the potential-charge relation (Figure 6) it follows
that m ) 0.19 eV and NT0 ) 1.7 × 1017 cm-3. Assuming the
conduction band edge to be located about 0.9 eV above the
redox energy of the electrolyte, a total density of traps in the
ZnO of about 2 × 1019 cm-3, or 40 traps per ZnO particle, is
calculated. This appears to be a reasonable result as it is well-
known that as-prepared ZnO can contain a significant amount
of defects. Following the approach outlined by van de Lagemaat
and Frank,42 the slope m can also be calculated from the light-
1040 J. Phys. Chem. C, Vol. 111, No. 2, 2007
intensity dependence of the short-circuit charge and the IMPS
time constant, yielding in our case for ZnO only a reasonable
value in the case of QSC, m ) 0.049 eV.
The quasi Fermi level (EF) in the mesoporous ZnO film under
illumination at short-circuit conditions was determined using
the VSC (voltage extraction) measurement.25 Recently, Lobato
et al. measured EF in a nanostructured TiO2 solar cell directly
by addition of an extra electrode on top the porous TiO2 film.43
Their results are in excellent agreement with our voltage
extraction method. In our experiment with ZnO a clear increase
of VSC with light intensity is found, from 0.2 to 0.4 V with
increasing light intensity. Considering the trap model described
above and the estimated value of ECB of 0.9 eV (vs Eredox), one
might expect to find activation energies for electron transport
of about 0.6 eV. The apparent activation energies that are
calculated from the temperature-dependent IMPS measurements
under short-circuit conditions are, however, much smaller, about
0.1 eV. This apparent discrepancy was explained recently by
Peter et al. using the quasi-static multiple trapping model.44 The
anomalously low apparent activation energies are a consequence
of the boundary conditions imposed by the short-circuit condi-
tion and the quasi-static relationship between changes in the
densities of free and trapped electrons. As the short-circuit
current is almost independent of temperature, EF is forced to
move downward with increasing temperature, so that the
decrease in transport time is much less than one might expect.
Traps in ZnO. The presence of an exponential distribution
of trap states in nanostructured ZnO may be questioned.
Evidence from spectroelectrochemistry45,46 and laser spectros-
copy47 suggests that electrons in nanostructured ZnO electrodes
are located in the conduction band and/or shallow traps.
Specifically, upon electron accumulation a bleach of the
excitonic absorption band of ZnO as well as free-electron
absorption is observed. The presence of localized electronic
states in the band gap of ZnO is, however, well documented.9
Intrinsic defects, such as oxygen vacancies and interstitial Zn
atoms, will affect the electronic properties of ZnO. As-prepared
ZnO tends to show significant n-type conductivity, which is
mainly caused by interstitial Zn atoms that act as shallow donors,
with an energy level about 0.03 eV below the conduction band
edge.48 Also hydrogen can act as a shallow donor in ZnO.49,50
While being very important for the conductivity of ZnO, such
shallow donors (traps) do not directly explain the observed slow
and light-intensity-dependent electron transport in nanostructured
ZnO.
If the donor density in nanostructured ZnO films is suf-
ficiently high, band bending within the individual nanocrystal
can occur and may determine the electronic properties of the
film. If we assume that the donor density is equal to the
estimated total density of traps from the charge extraction
measurements (∼2 × 1019 cm-3), which implies that all traps
are positively charged impurities, an estimation of the maximum
potential drop within a single ZnO nanocrystal (φsc) can be
calculated using eq 4:
eND 2
φsc ) r (4)
60
where is  the relative dielectric constant of ZnO ( ) 8 0 9),
is the permittivity of vacuum, and r is the radius of the particle.
For the fully depleted particle φsc is calculated to be ∼0.5 V. A
significant band bending may therefore be present in the ZnO
particles. This is a possible cause of the very slow recombination
of electrons in ZnO with triiodide in the electrolyte. In
Quintana et al.
comparable TiO2 particles band bending would be much less
because of the higher dielectric constant in anatase ( ≈ 50).
As a result of band bending between adjacent ZnO particles,
energy barriers would be formed at grain boundaries. The
thermally activated crossing of these barriers by the electrons
would result in slow transport in the nanostructured ZnO films.
Upon electron accumulation, band bending and consequently
barrier height would decrease and electron transport would
become faster. Such a barrier model can, at least qualitatively,
explain light-intensity-dependent electron transport in doped
nanostructured films.
In a barrier model, electron transport is determined by the
rate of electron hopping from ZnO particle to particle. From
the transport time (0.57 ms at the highest light intensity in Figure
5a) an effective electron diffusion coefficient in the nanostruc-
tured ZnO of 1.2 × 10-4 cm2 s-1 is calculated using Deff )
w2/(2.35τIMPS),51 where w is the film thickness. The time τj for
a single electron jump across the barrier from one ZnO
nanocrystal to the next can be calculated using the Einstein-
Smoluchowski equation Deff ) d2/2τj, where d is jump distance,
which is equal to the nanocrystal diameter. It can thus be
calculated that τj ) 11 ns, and the electron makes on average
5.3 × 104 jumps before reaching the conducting substrate. The
thermionic emission model can be used to describe electron
transfer in semiconductors across a barrier:52
∆E
j ) AT2 exp - ( kT ) (5)
In this expression j is the current density, ∆E is the barrier
height, and A is the Richardson constant for thermionic emission
(A ) 120(m*/me) A cm-2). Using the effective electron mass
in ZnO (m*) of 0.3me, and assuming that the contact area
between the nanocrystals is 50 nm2 (the area of circle with radius
4 nm), it can be calculated that a jump time of 11 ns is obtained
with a barrier height of 0.3 eV, which seems to be a very
reasonable value considering that the band bending in a fully
depleted ZnO nanoparticle was estimated to be 0.5 eV.
The rather triangular shape of the voltammogram for nano-
structured ZnO electrode (Figure 2) may be attributed to a nearly
linearly increasing density of trap states toward the conduction
band of ZnO. Alternatively, it may be attributed to significant
n-type doping of the ZnO in combination with band bending.
Nanostructured films composed of highly doped spherical
semiconductor particles will show similarly shaped voltammo-
grams, as was demonstrated by model calculations, as well as
by experiments with highly doped nanostructured SnO2:Sb
electrodes, by Boschloo and Fitzmaurice.53 We can estimate that
the doping density in the ZnO needs to be about 5 × 1019 cm-3
to allow for sufficient potential drop in the ZnO particles (0.8 V),
so that voltammograms similar to that shown in Figure 2 can
be obtained. This doping density seems, however, rather high
for unintentionally doped ZnO material. At this point, however,
we cannot exclude the possibility that the electron transport and
recombination properties of nanostructured ZnO electrodes are
caused by a significant doping density that leads to band bending
and energy barriers at grain boundaries.
Summary
Electron transport in dye-sensitized solar cells prepared from
approximately equally sized nanocrystalline ZnO and TiO2
particles was very similar in terms of transport times and light-
intensity dependence, but the electron lifetime was significantly
higher in ZnO than in TiO2. The performance of ZnO-based
solar cells was less than that for TiO2-based cells, despite the
Dye-Sensitized ZnO and TiO2 Solar Cells
reduced recombination rate. This was attributed to a lower
electron injection efficiency from excited dye molecules into
the ZnO conduction band and/or a lower dye regeneration
efficiency. The apparent activation energy for electron transport
in nanostructured ZnO under short-circuit conditions was
∼0.1 eV, nearly independent of light intensity. The internal
voltage in the nanostructured ZnO under short-circuit conditions
was about 0.23 V lower than VOC. Results may be explained
using a multiple trapping model. An alternative view was
presented: If there is significant doping of the ZnO, band
bending in the nanocrystals will result in energy barriers for
electron transport and recombination that can explain the
observed properties.
Acknowledgment. This work was financially supported by
the International Science Program at Uppsala University and
the Swedish Energy Agency. T.E. thanks BASF A.G. for
financial support. We thank Göran Karlsson, Department of
Physical Chemistry, Uppsala University, for transmission
electron microscopy.
References and Notes
(1) O’Regan, B.; Grätzel, M. Nature (London) 1991, 353, 737.
(2) Hagfeldt, A.; Grätzel, M. Acc. Chem. Res. 2000, 33, 269.
(3) Wang, P.; Zakeeruddin, S. M.; Moser, J. E.; Nazeeruddin, M. K.;
Sekiguchi, T.; Grätzel, M. Nat. Mater. 2003, 2, 402.
(4) Redmond, G.; Fitzmaurice, D.; Grätzel, M. Chem. Mater. 1994, 6,
686.
(5) Rensmo, H.; Keis, K.; Lindström, H.; Södergren, S.; Solbrand, A.;
Hagfeldt, A.; Lindquist, S. E.; Wang, L. N.; Muhammed, M. J. Phys. Chem.
B 1997, 101, 2598.
(6) Keis, K.; Lindgren, J.; Lindquist, S.-E.; Hagfeldt, A. Langmuir 2000,
16, 4688.
(7) Sayama, K.; Sugihara, H.; Arakawa, H. Chem. Mater. 1998, 10,
3825.
(8) Ferrere, S.; Zaban, A.; Gregg, B. A. J. Phys. Chem. B 1997, 101,
4490.
(9) Ozgur, U.; Alivov, Y. I.; Liu, C.; Teke, A.; Reshchikov, M. A.;
Dogan, S.; Avrutin, V.; Cho, S. J.; Morkoc, H. J. Appl. Phys. 2005, 98,
041301.
(10) Hoyer, P.; Weller, H. J. Phys. Chem. 1995, 99, 14096.
(11) O’Brien, P.; Saeeda, T.; Knowles, J. J. Mater. Chem. 1996, 6, 1135.
(12) Vayssieres, L.; Keis, K.; Hagfeldt, A.; Lindquist, S. E. Chem. Mater.
2001, 13, 4395.
(13) Peterson, R. B.; Fields, C. L.; Gregg, B. A. Langmuir 2004, 20,
5114.
(14) Yoshida, T.; Tochimoto, M.; Schlettwein, D.; Wöhrle, D.; Sugiura,
T.; Minoura, H. Chem. Mater. 1999, 11, 2657.
(15) O’Regan, B.; Sklover, V.; Grätzel, M. J. Electrochem. Soc. 2001,
148, C498.
(16) Huang, M. H.; Wu, Y.; Feick, H.; Tran, N.; Weber, E.; Yang, P.
AdV. Mater. 2001, 13, 113.
(17) Keis, K.; Magnusson, E.; Lindström, H.; Lindquist, S.-E.; Hagfeldt,
A. Sol. Energy Mater. Sol. Cells 2002, 73, 51.
(18) Kakiuchi, K.; Hosono, E.; Fujihara, S. J. Photochem. Photobiol.,
A: Chem. 2005, 179, 81.
(19) Bauer, C.; Boschloo, G.; Mukhtar, E.; Hagfeldt, A. J. Phys. Chem.
B 2001, 105, 5585.
J. Phys. Chem. C, Vol. 111, No. 2, 2007 1041
(20) Oekermann, T.; Yoshida, T.; Minoura, H.; Wijayantha, K. G. U.;
Peter, L. M. J. Phys. Chem. B 2004, 108, 8364.
(21) Oekermann, T.; Yoshida, T.; Boeckler, C.; Caro, J.; Minoura, H.
J. Phys. Chem. B 2005, 109, 12560.
(22) Galoppini, E.; Rochford, J.; Chen, H.; Saraf, G.; Lu, Y.; Hagfeldt,
A.; Boschloo, G. J. Phys. Chem. B 2006, 110, 16159.
(23) Nazeeruddin, M. K.; Splivallo, R.; Liska, P.; Comte, P.; Grätzel,
M. Chem. Commun. 2003, 1456.
(24) Lindström, H.; Magnusson, E.; Holmberg, A.; Södergren, S.;
Lindquist, S.-E.; Hagfeldt, A. Sol. Energy Mater. Sol. Cells 2002, 73, 91.
(25) Boschloo, G.; Hagfeldt, A. J. Phys. Chem. B 2005, 109, 12093.
(26) Willis, R. L.; Olson, C.; O’Regan, B.; Lutz, T.; Nelson, J.; Durrant,
J. R. J. Phys. Chem. B 2002, 106, 7650.
(27) From the effective mass of conduction band electrons of 0.3me (ref
9)and 1me [Tang, H.; Prasad, K.; Sanjines, R.; Schmid, P. E.; Levy, F. J.
Appl. Phys. 1994, 75, 2042], effective densities of conduction band states
of 4 × 1018 and 2.5 × 1019 cm-3 are calculated for ZnO and TiO2 (anatase),
respectively.
(28) Wenger, B.; Grätzel, M.; Moser, J.-E. J. Am. Chem. Soc. 2005,
127, 12150.
(29) Cao, F.; Oskam, G.; Searson, P. C. J. Phys. Chem. 1996, 100, 17021.
(30) Dloczik, L.; Ileperuma, O.; Lauermann, I.; Peter, L. M.; Ponomarev,
E. A.; Redmond, G.; Shaw, N. J.; Uhlendorf, I. J. Phys. Chem. B 1997,
101, 10281.
(31) Fisher, A. C.; Peter, L. M.; Ponomarev, E. A.; Walker, A. B.;
Wijayantha, K. G. U. J. Phys. Chem. B 2000, 104, 949.
(32) Schlichthörl, G.; Huang, S. Y.; Sprague, J.; Frank, A. J. J. Phys.
Chem. B 1997, 101, 8139.
(33) Schlichthörl, G.; Park, N. G.; Frank, A. J. J. Phys. Chem. B 1999,
103, 782.
(34) Seager, C. H.; Myers, S. M. J. Appl. Phys. 2003, 94, 2888.
(35) Forro, L.; Chauvet, O.; Emin, D.; Zuppiroli, L.; Berger, H.; Lévy,
F. J. Appl. Phys. 1994, 75, 633.
(36) Zaban, A.; Greenshtein, M.; Bisquert, J. ChemPhysChem 2003, 4,
859.
(37) There is some difference in the lifetimes of Figures 6b and 5a; this
can probably be attributed to the fact that the data in the two figures stem
from two different series of cells and measurements, separated in time by
about a year.
(38) Cameron, P. J.; Peter, L. M. J. Phys. Chem. B 2005, 109, 7392.
(39) We thank one reviewer for suggesting this possibility.
(40) Duffy, N. W.; Peter, L. M.; Rajapakse, R. M. G.; Wijayantha, K.
G. U. Electrochem. Commun. 2000, 2, 658.
(41) O’Regan, B.; Lenzmann, F. J. Phys. Chem. B 2004, 108, 4342.
(42) van de Lagemaat, J.; Frank, A. J. J. Phys. Chem. B 2000, 104,
4292.
(43) Lobato, K.; Peter, L. M.; Wurfel, U. J. Phys. Chem. B 2006, 110,
16201.
(44) Peter, L. M.; Walker, A. B.; Boschloo, G.; Hagfeldt, A. J. Phys.
Chem. B 2006, 110, 13694.
(45) Redmond, G.; O’Keeffe, A.; Burgess, C.; MacHale, C.; Fitzmaurice,
D. J. Phys. Chem. 1993, 97, 11081.
(46) Bauer, C.; Boschloo, G.; Mukhtar, E.; Hagfeldt, A. Chem. Phys.
Lett. 2004, 387, 176.
(47) Katoh, R.; Furube, A.; Hara, K.; Murata, S.; Sugihara, H.; Arakawa,
H.; Tachiya, M. J. Phys. Chem. B 2002, 106, 12957.
(48) Look, D. C.; Hemsky, J. W.; Sizelove, J. R. Phys. ReV. Lett. 1999,
82, 2552.
(49) Van de Walle, C. G. Phys. ReV. Lett. 2000, 85, 1012.
(50) Jokela, S. J.; McCluskey, M. D. Phys. ReV. B 2005, 72, 113201.
(51) van de Lagemaat, J.; Frank, A. J. J. Phys. Chem. B 2001, 105,
11194.
(52) Sze, S. M. Physics of Semiconductor DeVices; Wiley: New York,
2006.
(53) Boschloo, G.; Fitzmaurice, D. J. Phys. Chem. B 1999, 103, 3093.