Journal of New Materials for Electrochemical Systems 8, 327-338 (2005)
c J. New. Mat. Electrochem. Systems
Rate-Determining Processes in Photoelectrochromic Devices
Anneke Georg 1∗ , Andreas Georg2 , Urša Opara Krašovec3 , V. Wittwer2
1 Freiburg
Materials Research Centre, Stefan Meier Str. 21, 79104 Freiburg, Germany
Institute for Solar Energy Systems, Heidenhofstr.2, 79100 Freiburg, Germany
Faculty of Electrical Engineering, University of Ljubljana, Tržaška 25, SI-1000 Ljubljana, Slovenia
2 Fraunhofer
3
( Received May 30, 2005 ; received in revised form March 9, 2006 )
Abstract: Photoelectrochromic windows are a combination of a dye solar cell and an electrochromic material, usually WO3 . They change their
transmittance on illumination. We prepared a particularly advantageous configuration, which can be coloured and bleached under illumination,
and bleached in the dark. In this paper we discuss variations of the light intensity, the ion concentrations in the electrolyte and of the solvent. With
this approach, we define the rate-determining processes of the device. The rate of the colouring process with an open circuit is determined by the
electron excitation at the dye by the incident light. The electron transfer from the WO3 to the I−
3 in the electrolyte is the dominating loss reaction
and the phase change of the WO3 during the charge injection is also of importance. Transport processes in the TiO2 or WO3 and loss reactions at
the TiO2 are less significant.
For the bleaching process with an open circuit, the electron transfer from the WO3 to the I−
3 is dominant. Bleaching under short circuit conditions
is limited by the Li+ diffusion in the WO3 .
Key words :
1.
INTRODUCTION
We developed the photoelectrochromic configuration illustrated
in Fig.1, which is a particularly advantageous device. It consists
of several components (Fig.1): a dye-covered nanoporous TiO2
layer, a porous electrochromic layer, such as WO3 , two glass
substrates coated with a transparent conductive oxide (TCO), of
which one is coated with Pt, an iodide/triiodide redox couple
and Li+ ions in an organic solvent. Both the TiO2 and the Pt
layers can be kept thin, so that they are transparent. The pores
of the TiO2 and WO3 layers are filled with the electrolyte.
Photoelectrochromic systems combine electrochromic layers
[1,2] and dye solar cells [3,4]. Electrochromic layers change
their transmittance reversibly when electrons and cations are injected. In photoelectrochromic systems, the dye solar cell provides the energy for the coloration of the electrochromic layer.
Thus, the transmittance of the photoelectrochromic device can
be decreased under illumination and can be increased again with
short circuiting the electrodes when illuminated or in the dark.
An external voltage supply is not required. Applications of
these devices include, for example, switchable sunroofs in cars
or smart windows in buildings.
During illumination (upper part of Fig.1), a dye molecule absorbs a photon of the incident light. Then an electron is rapidly
injected from the excited state of the dye into the conduction
band of the TiO2 and diffuses to the WO3 . Ionised dye molecules are reduced by I− in the electrolyte according to the reaction:
+
2dye+ + 3I− → I−
3 + 2dye. Li ions intercalate into the WO3
and keep the charges balanced. Because of the double injec-
∗ To whom correspondence should be addressed: Anneke Georg, Freiburg
Materials Research Centre, Stefan-Meier-Str. 21, 79104 Freiburg, Germany, phone: ++49 (0) 761 203 4799, fax: ++49 (0)761 2034801, email:
[email protected]
327
328
A. Georg et al./ J. New Mat. Electrochem. Systems 8, 327-338 (2005)
Figure 1: Construction and operating principle of the photoelectrochromic device. The upper part shows the coloration in open
circuit (switch open) and the lower part the bleaching in short
circuit (switch closed).
Figure 2: Transmittance spectra of the photoelectrochromic device in the coloured and bleached states. For a cell with liquid
electrolyte, the visible (solar) transmittances changes from 51%
(35%) to 4.8% (1.5%) in 3 min.
ferences to the alternative photoelectrochromic system and the
advantages of our new system.
tion of electrons and Li+ ions, the WO3 changes its colour from
colourless to blue.
If electrons are allowed to flow via an external circuit from the
WO3 via a TCO layer to the Pt electrode (lower part of Fig.1,
external switch closed), then the Pt catalyses the reverse reac−
tion, i.e. the reduction of I−
3 to I . Lithium cations leave the
WO3 , and the WO3 is bleached fast. This process occurs also
during illumination. If the external switch is open, electrons can
leave the WO3 only by loss reactions. This process is very slow.
If the electrolyte is liquid, the device’s visible (solar) transmittance changes under 1000 W/m2 of illumination (1 sun) from
51% to 5% (35% to 1.5%) with switching times of about 3 minutes (Fig. 2). With a solid electrolyte, a visible transmittance
change from 62% to 1.6% and a solar transmittance change
from 41% to 0.8% are achieved with switching times of 10 min.
The colouring time is independent of the area.
An alternative photoelectrochromic configuration was first published by Bechinger et al. [5]. In this device the colouring and
the bleaching processes are competing, because the bleaching
is possible only via loss reactions. Therefore, either fast colouring and bleaching with a small transmittance change [5] or a
high contrast with slow bleaching is achievable [6], or an external voltage is used for bleaching [7]. In our new device, the
materials can be optimised for colouring and bleaching independently, so it simultaneously allows fast colouring and bleaching
and high contrast [8].
In Ref. 8 we introduced this new device and discussed the dif-
Experiments with different layer configurations of photoelectrochromic devices were reported [9]. From these experiments,
we concluded that the loss reactions of electrons from the TiO2
can be neglected compared to the loss reactions of electrons
from the WO3 .
We developed special WO3 - and TiO2 -layers for the photoelectrochromic system. They were examined in detail in [10]. A
very interesting effect, which is also important for the photoelectrochromic device, is the phase transition that takes place
in WO3 during charge intercalation. The nanoparticles of WO3
have a crystalline core [10] which has a monoclinic structure in
the bleached state and changes during charge intercalation to a
tetragonal and then to a cubic structure [11, 12].
Knowledge of the limiting processes of the photoelectrochromic
device is important for the optimisation of the materials. In
this paper we investigate the different processes that occur in
the system and determine their role in the system. For some
processes, we measured their characteristic kinetic parameters
such as diffusion constants or charge transfer resistances; for
some processes, we demonstrated their importance by varying
parameters such as the light intensity, layer thickness, the electrolyte solvent, concentrations of Li+ , I− , and I−
3 ions.
2.
2.1
EXPERIMENTAL
Sample Preparation
Transparent nanoporous TiO2 and WO3 layers were prepared
using sol-gel techniques described in [9]. A TCO (F:SnO2 )
Rate-Determining Processes in Photoelectrochromic Devices . / J. New Mat. Electrochem. Systems 8, 327-338 (2005)
329
coated glass plate from Pilkington was covered by dip-coating
first with WO3 , then with TiO2 , and sintered in between and afterwards at 450◦ C for 30 min. The thickness of the TiO2 layer
was about 150 nm, and of the WO3 layer about 500 nm. The
size (diameter) of the particles in the WO3 layer is around 20
to 30 nm, in the TiO2 10 nm, as we showed by SEM measurements [9]. The Pt layer was sputtered with a mean thickness
of about 2 nm. The dye (Ru 535 bis-TBA from Solaronix) was
deposited by soaking the TCO/WO3 /TiO2 layers in a solution
of the dye in ethanol. The area of the samples was 5*5 cm2 .
If not described otherwise, the electrolyte was 0.5 M LiI and
0.005 M I2 in propylene carbonate. A silicone film with 0.5 or 1
mm thickness was used as a spacer between the electrodes. The
samples were sealed with acrylic adhesive.
ched on once every minute for a short measurement so that it did
not influence the coloration of the photoelectrochromic device.
2.2
There are four different switching modes of the photoelectrochromic cells (see Fig. 3): Colouring is possible only under
illumination in open circuit. Fast bleaching is possible in short
circuit both under illumination and in the dark. In the dark,
the cell is bleached completely, but under illumination a weak
coloration, which depends on the amount of dye in the TiO2 and
on the concentration of I2 in the electrolyte, remains. In open
circuit in the dark, the cell is also bleached, but very slowly,
because here the bleaching is possible only via loss reactions.
Characterisation
For characterising the kinetics of the colouring and bleaching
processes, we developed a special experimental set-up to measure the transmittance of the samples together with the open
circuit voltage or short circuit current. The set-up consists of
a halogen lamp for the illumination, a Si photodiode to detect
the transmitted light, an optical filter and collimators. It is able
to measure the transmittance in the same way when the device
is activated by illumination as well as when it is not activated.
The halogen lamp simulated sunlight with an intensity of 1000
Wm−2 (1 sun) and a spectrum close to the spectrum of the sun.
The mismatch factor of dye solar cells was taken into account.
The optical filter is transparent only in the range of the spectrum (above 715 nm) where the dye is not sensitive and no electrons are excited, as we demonstrated by spectral response measurements. Thus, the filter is placed between the halogen lamp
and the photoelectrochromic cell for dark measurements and between the cell and the photodiode for the illuminated state. The
measured value of the transmittance is a convolution of the spectra of the halogen lamp, the filter and the photodiode. We calibrated our set-up by measuring the transmittance spectra with
a Perkin-Elmer 330 spectrometer for different values of the optical density. The spectra were convoluted with the visible and
the solar spectra to calculate the visible and solar transmittance,
respectively. We found a simple linear relation between the optical density as measured with our set-up and the visible (solar)
optical density determined with the spectrometer. This is because the coloration efficiency is independent of the degree of
coloration [8]. With this calibration it is possible to calculate
the values of the solar and the visible transmittance from the
values measured with our set-up. More details of this set-up are
described in [11].
Long-duration transmittance measurements of self-bleaching in
the dark under open circuit conditions were made with a lightemitting diode with an intensity maximum at a wavelength of
655 nm without any filters. This light-emitting diode was swit-
The light intensity was varied by using different distances between the cell and the halogen lamp. The light intensity at different distances was determined with the help of a reference
silicon solar cell. For the very low intensities (50 W/m2 and
20 W/m2 ), additional grey optical filters from Schott were used:
Filter NG4 (transmittance 0.1%) and filter NG9 (transmittance
0.01%).
3.
3.1
RESULTS AND DISCUSSION
Switching Modes
Figure 3: Switching modes of the photoelectrochromic device.
In short circuit, some of the electron transfer processes can be
excluded as rate-determining processes for the bleaching because they are very fast: The electron transfer between the two
TCO electrodes via the shunt resistance, between the Pt-TCOelectrode and the electrolyte, and between the WO3 and the
TCO substrate. The resistance of the TCO substrates is 10Ω, the
shunt resistance for current measurement in short circuit is 10Ω,
the short circuit current density in 1 sun is ISC = 0.15 mA/cm2
(Fig. 4). The ohmic voltage drop between the two TCO electrodes for a closed switch is R*ISC = 0.15 mA/cm2 *20Ω =
330
A. Georg et al./ J. New Mat. Electrochem. Systems 8, 327-338 (2005)
3 mV/cm2 and is therefore negligible. The charge transfer resistance of the electron transfer from the Pt electrode to the I−
3
ions in the electrolyte was determined to be RCT = 40Ωcm2
with the help of impedance spectroscopy [13]. The voltage
drop (not ohmic, but following the Butler-Vollmer equation) is
smaller than RCT *ISC = 6mV. The transfer resistance between
the WO3 and TCO is R = 10Ωcm2 for an evaporated WO3 layer
in the bleached state [14], in the coloured state it should be even
smaller.
Figure 4: Colouring in open circuit and bleaching in short circuit with 1 sun illumination for photoelectrochromic cells with
propylene carbonate and acetonitrile as the solvents.
Compared to bleaching in open circuit in the dark, which is
driven by loss reactions, bleaching in short circuit is much faster.
The loss reactions are therefore too slow to play an important
role in short circuit.
3.2
Solar Cell Inside the Photoelectrochromic Cell
Inside the photoelectrochromic cell there is a very thin, transparent dye solar cell which supplies the power for the electrochromic coloration. It has the configuration
glass - TCO – TiO2 with dye – electrolyte – Pt – TCO - glass,
which is the same configuration as the photoelectrochromic cell
without WO3 . In contrast to normal dye solar cells, the TiO2
layer here is very thin (150 nm instead of typically 10 µm) and
has no light-scattering clusters, so that it is transparent. However, the particle size of the TiO2 and the electrolyte are similar,
and the dye is the same as is commonly used in dye solar cells.
The electron generation process in the dye-sensitised TiO2 consists of three sub-processes, which are very fast (femto- to na-
nosecond time-scales [15, 16, 17, 18]): excitation of the dye by
incident light, injection of electrons into TiO2 , reduction of the
dye by I− . Because nearly all electrons excited in the dye are
injected into the TiO2 [17, 19], the electron generation is limited
by the dye content and the light intensity.
It should also be noted that there is almost no Li+ intercalation in the TiO2 layer compared to the WO3 layer. This is confirmed by measurements on dye solar cells [18]. The reason is
the high redox potential of TiO2 with respect to this intercalation (-0.7 V versus NHE [20] to -0.9 V versus NHE [21, 22,
23]), whereas WO3 can be intercalated at 0 V versus NHE for
an intercalation degree x = 0 (x: number of intercalated Li+
per W atom in the WO3 layer) [12, 24]. The consequence is
that the electron density in TiO2 under the conditions of dye solar cells is about 40000 times lower than in electrochromically
coloured WO3 : In TiO2 the electron density is 2*1017 /cm3 [25],
in coloured WO3 it is 8*1021 /cm3 (calculated from the density
of WO3 (7.2 g/cm3 ) and the molar mass (232 g/mol) for 0.4 Li+
ions per W atom). If we take into account the porosity of 50% of
the WO3 [11], the electron density is halved. Because the electron density in WO3 is so much higher than in TiO2 , the electrochemical potential of both layers is dominated by the WO3 .
This can be seen also in the time dependence of the open circuit
voltage of the photoelectrochromic device. Directly after starting the illumination the electrochemical potential in the TiO2 is
increasing very fast (equilibrium value of open circuit voltage is
reached after approximately 0.5 s [26]). As mentioned before,
nearly all electrons are injected into the WO3 . The electrochemical potential in WO3 is increasing (according to equation (5),
see below). However, in the WO3 this takes much longer, VOC
reaches equilibrium after 20 min.
We prepared this dye solar cell, which is included inside the
photoelectrochromic cell, separately. All layers were made in
the same way as for the photoelectrochromic cell. The cell had
an efficiency of 0.054%, an open circuit voltage of 580 mV and
a short circuit current of 0.16 mA/cm2 . (The I-V-curve of this
solar cell can be found in [9].) This shows that even the very low
efficiency of 0.054% is sufficient for the excellent coloration
characteristic of the photoelectrochromic cell.
The short circuit current in 1 sun in equilibrium is always the
same (about 0.15 mA/cm2 ) for the photoelectrochromic device
and for the solar cell, if the TiO2 layers of both are of the same
thickness. We can conclude that the dye content of the WO3
layer does not contribute significantly to the current. This is
confirmed by the observation that after soaking in the dye solution, WO3 layers without TiO2 do not show a significant colouring in illumination.
The diffusion constant of I−
3 in the electrolyte was determined
as described in [13] to be D = 2*10−6 cm2 /s. If diffusion were
the only transport mechanism in the electrolyte, with
Rate-Determining Processes in Photoelectrochromic Devices . / J. New Mat. Electrochem. Systems 8, 327-338 (2005)
2neo DcNA
jlim =
L
(1)
[26], this would give a diffusion-limited current density of jlim
= 0.04 mA/cm2 (c = [I2 ] = [I3 ] = 0.005M, thickness of electrolyte layer L = 1 mm, number of electrons transferred n = 2,
NA : Avogadro’s number, e0 : elementary charge). This value
is much lower than the measured short circuit current of the
cell. This is possible because convection takes place in the cells
with an electrolyte layer of 1 mm thickness, especially when the
cell is illuminated: The halogen lamp which was used for illumination heats the cell up to 37◦ C from one side. With an assumed diffusion layer thickness of about 100 µm [28], we obtain
a diffusion-limited current density of about jlim = 0.2 mA/cm2 ,
which is similar to the short circuit current of the cell. We conclude that diffusion and convection are both limiting processes
in short circuit under illumination.
3.3
Bleaching in Short Circuit in the Dark
In Fig. 5, the optical density for bleaching in short circuit in
the dark is plotted versus the square root of the time. With the
help of this plot, we can examine the role of diffusion processes
more closely. If the bleaching is limited by diffusion processes
(that means that diffusion processes are slower than all other
processes during bleaching), this plot is linear.
331
the interface between WO3 particle and electrolyte. In case of
infinite diffusion, such limits do not exist and the concentration
changings can spread out infinitely.) In the electrolyte, only
the diffusion of I−
3 could be important, because the concentration of Li+ and I− is 100 times higher than the concentration of
I−
3 , and the diffusion constants should be of the same order of
magnitude. According to the literature [29] the diffusion in the
pores of TiO2 is not much slower than in the bulk of the electrolyte, because the interaction between I−
3 and the TiO2 surface
is weak.
From literature we know the diffusion constant of photogenerated electrons in dye-sensitised nanocrystalline TiO2 : D =
10−5 cm2 /s [30 - 33], and the electron density ne = 2*1017 /cm3
[25]. With the layer thickness of about 150 nm and Eq. 1, we
estimate the diffusion-limited current density to be 40 mA/cm2 .
This is much higher than the measured currents. Therefore the
diffusion of electrons in TiO2 is not important in photoelectrochromic devices, in contrast to dye solar cells, where the
TiO2 layer is much thicker and the diffusion of electrons is important [34, 35].
The diffusion constant of electrons in WO3 was measured according to the method described in [36]. The result was D
= 0.02 cm2 /s. The electron density in coloured WO3 is ne =
8*1021 /cm3 . The film thickness is L = 500 nm. This leads to a
diffusion-limited current density of 106 A/cm2 .
This means that diffusion of electrons in WO3 and in TiO2 is so
fast that its effect on the bleaching rate is negligible.
From the electrochromic studies on WO3 it is known that diffusion of Li+ ions in WO3 can dominate the bleaching process [1].
Thus two processes could be dominating: The diffusion of I−
3 in
the electrolyte and diffusion of Li+ out of the WO3 particles.
If the diffusion of I−
3 in the electrolyte is limiting, the optical
density is given by:
√
2nF
OD(t) = √ CE c∞ Dt
π
(2)
Figure 5: Bleaching in short circuit in the dark, optical density
and extracted charge plotted versus the square root of the time.
OD: optical density, F: Faraday constant, n: number of electrons
transferred (here n=2), CE : coloration efficiency, c∞ : concentra−
tion of I−
3 in the electrolyte, DI : diffusion constant of I3 in the
electrolyte, t: time.
Here the curve is linear approximately between 1.9 and 7.8 s0.5
(or 4 and 60s). This is an indication that diffusion processes are
rate-determining in this range. After 60 s, the dominating diffusion very probably changes from infinite to finite diffusion.
(Finite diffusion means that changings of concentration (here
the concentration of the ions) can spread out only up to a certain limit, e.g. when the concentration of Li+ changing reaches
Because OD = CE Q, Q being the intercalated charge per area,
Eq. 2 is equivalent to the Cottrell equation [26], multiplied by
the coloration efficiency. The Cottrell equation is a direct consequence of the solution of Fick’s law of diffusion for a potential
step at a flat electrode. The linear dependence of the optical
density on the charge was proved by plotting the optical density
as a function of the extracted charge for the measurement shown
332
A. Georg et al./ J. New Mat. Electrochem. Systems 8, 327-338 (2005)
in Fig. 5. From the slope of the very linear plot obtained [13],
we determined the coloration efficiency to be CE =100 cm2 /C.
With c∞ = 0.005 mol/l the gradient in Fig. 5 gives a value of
DI = 1.8*10−6 cm2 /s for the diffusion constant of I−
3 in the electrolyte. This would be a realistic value.
If the diffusion of Li+ out of the WO3 particles is rate-determining, we use spherical symmetry instead of the flat electrode,
so the factor 2 is replaced by a factor 6 in Eq. 2 [37]. With the
optical density in equilibrium OD∞ =CE nFc∞ r , we obtain
√
6 O∞
d
Dt
OD(t) = √
π r
(3)
where r is the radius of WO3 particle (15nm).
Eq. 3 applied to the gradient in Fig. 5 leads to a diffusion
constant for Li+ in WO3 of DL = 1.6*10−15 cm2 /s. This is a
realistic order of magnitude for coloured crystalline WO3 .
On this basis, the question about the dominating diffusion process is still open. The answer is given by variation of the solvent
(see below).
3.4
Variation of LiI Concentration
We prepared photoelectrochromic cells with different concentrations of LiI: [LiI]= 0.2 M, 0.5 M and 1 M. The concentration
of I2 was 0.005 M in all cells.
The self-bleaching in open circuit in the dark, which was measured after coloration in 1 sun (Fig. 6), is faster for lower
LiI concentration. We can conclude that the loss reactions are
slower for higher LiI concentration. A possible explanation is
that Li2 CO3 is formed as a layer on the WO3 /TiO2 surface in
contact with the electrolyte (which has propylene carbonate as
solvent) and blocks the loss reactions at the WO3 /electrolyte
interface. A layer of crystals on the WO3 /TiO2 surface was detected by SEM measurements, and the presence of the Li2 CO3
in the electrolyte was confirmed with XRD measurements not
shown here. The colouring in open circuit is shown in Fig. 7.
Several important observations can be made with this figure:
• The coloration velocity (initial rate of change of the optical density) is independent of the LiI concentration. This
leads to the following conclusions:
– The ion transport in the electrolyte is not limiting
the kinetics of the coloration.
– Processes involving Li+ or I− at the interface between TiO2 or WO3 and the electrolyte are not limiting the kinetics of the coloration, because the velocity of the processes at the interfaces depends on
Figure 6: Self bleaching in open circuit in the dark for photoelectrochromic cells with varied concentration of LiI. Optical
density at 655 nm.
ion concentration gradients and on the coverage of
the surface with ions. The coverage again depends
on the ion concentration, if the surface is not already
saturated.
– The diffusion of Li+ in WO3 is not limiting the kinetics of the coloration. The diffusion of Li+ into
the WO3 particles should become faster with higher
concentration gradient, i.e. with higher surface coverage with Li+ .
• The coloration in equilibrium increases with the LiI concentration. This is in accordance with the conclusion from
self-bleaching measurements, that the loss reactions are
low for high LiI concentration.
• The open circuit voltage in equilibrium increases with decreasing LiI concentration. This is initially astonishing,
because the loss reactions are faster at lower LiI concentration. An explanation with the help of the Nernst equation
Rate-Determining Processes in Photoelectrochromic Devices . / J. New Mat. Electrochem. Systems 8, 327-338 (2005)
333
The bleaching in short circuit in the dark (Fig. 8) is faster for
lower LiI concentration. This is in accordance with the hypothesis explained before, that the bleaching is limited by the diffusion of Li+ out of the WO3 particles, because then the bleaching
velocity would be determined by the gradient of concentration
of Li+ on the surface of the WO3 particle. With higher LiI concentration, the surface coverage with Li+ is higher (if not already saturated). Then the Li+ concentration gradient is lower
and the bleaching slower.
Figure 7: Colouring in open circuit with 1 sun illumination for
photoelectrochromic cells with different LiI concentrations.
kB T
∆E =
ln
2eo
− − !
I3 / I3 std
< 10mv
[I − ] / [I − ]std
(4)
−
−
−
([I−
3 ]std ,[I ]std : concentrations of I3 and I in the standard
state, kB : Boltzmann constant, T: temperature)
is not possible, because this effect is too small. (Another consequence of this small ∆E is that in the photoelectrochromic device, the Pt electrode in the presence of the redox couple I− /I−
3
acts as a reference electrode under open circuit conditions.)
As the LiI concentration increases, the optical density in equilibrium, which is proportional to the intercalation degree x, also
increases. Therefore the electron density also increases with LiI
concentration. Considering
E(X) = a + bx −
kB T
x
ln
eo
1−×
(5)
[38], (a,b: constants),
this would mean that the electrochemical potential in the WO3
layer and therefore the open circuit voltage would increase with
the LiI concentration. The experimental result showing the opposite behaviour could be explained by a shift of the constant
“a” in Eq. 5 to positive potentials. The term “a” contains the
energies of reduction, interaction and bonding of Li+ ions and
electrons in WO3 . A similar effect is observed in dye-sensitised
solar cells: Here also the open circuit voltage decreases with increasing LiI concentration. This is explained by a positive shift
of the conduction band in TiO2 [39, 40].
Figure 8: Bleaching in short circuit in the dark for photoelectrochromic cells with different LiI concentrations.
3.5
Phase Transition
In Fig. 7 the optical density increases continuously, while the
voltage increases rapidly during the first minute, reaches a maximum and then decreases slightly. On first inspection, this looks
like a contradiction. The continuous increase of the optical density shows that the intercalation degree x, defined as the number
of injected electrons per W atom in the WO3 layer, is increasing. On the other hand, Eq. 5 postulates that the electrochemical
potential E and the voltage are also increasing continuously.
We found out by x-ray diffraction measurement (XRD), IR spectroscopy and by comparison with the literature results [12,41]
that this maximum in the voltage is due to the phase transition
of the WO3 crystals: During coloration the crystal structure
changes from monoclinic through tetragonal to cubic. If the
electron injection is faster than the phase transition, then such a
maximum in the voltage occurs. For slower electron injection,
e.g. lower light intensity or low dye content, it does not occur.
Also, the abruptly changing gradient of the voltage after some
seconds is an effect of phase transition. More detailed studies
of these phase transitions are published elsewhere [11]. In [11]
we also show the electrochemical potential of the WO3 layer
measured for galvanostatic charging. It should be noted that the
Pt electrode in the presence of the redox couple I− /I−
3 acts as a
334
A. Georg et al./ J. New Mat. Electrochem. Systems 8, 327-338 (2005)
reference electrode under open circuit conditions. The changes
in the potential due to the changes in the ion concentrations in
the Nernst equation (Eq. 4) during colouring are only about 3
mV and can be neglected.
In the measurement of self-bleaching in open circuit in the dark
(Fig. 6), the phase transitions are even more obvious. The open
circuit voltage corresponds here to the equilibrium potential of
the WO3 versus the Pt electrode, because the bleaching via loss
reactions is so slow. The plateau in the voltage occurs during
phase transition, the decrease of the voltage during the extraction of Li+ in the single phase. The optical density decreases
continuously.
3.6
Variation of I2 Concentration
The main loss reaction in dye solar cells is the recombination
−
of electrons from the TiO2 with I−
3 in the electrolyte so that I
is formed [16]. The exchange current density of this process is
10−11 to 10−9 A/cm2 [42] for nanoporous TiO2 films of about
10µm thickness. Transferring this to the photoelectrochromic
device, this corresponds to the recombination of electrons from
the TiO2 and the WO3 with I−
3 . In order to investigate the
influence of I−
3 (I2 ) in the electrolyte on the behaviour of the
photoelectrochromic cell, cells with different concentrations of
I2 ([I2 ]=0M, 0.005M and 0.05M) were constructed. For all of
these cells, the concentration of LiI was constant: [LiI]=0.5M.
The I2 is transformed to I−
3 nearly completely due to the reaction
I − + I2 → I3−
Figure 9: Self bleaching in open circuit in the dark for photoelectrochromic cells with different I2 concentrations. Optical
density at 655 nm.
[43].
The measurement of self-bleaching (Fig. 9) shows that the loss
reactions are faster for higher I2 concentration. That means that,
similarly to dye solar cells, the recombination of electrons from
−
the TiO2 or WO3 with I−
3 in the electrolyte to form I is the
main loss reaction in the photoelectrochromic device. From our
previous work [9], we know that the loss reactions from TiO2
are negligible compared to those from WO3 . The loss reactions at the TCO / electrolyte interface are negligible because of
the very high charge transfer resistance of 25 MΩcm2 [13]. So
we conclude that the main loss reaction is the recombination of
electrons in the WO3 with I−
3 in the electrolyte.
The colouring under illumination in open circuit (Fig. 10) is
faster and deeper for lower I2 concentration. This is due to the
lower loss reactions. The open circuit voltage decreases logarithmically with increasing I2 concentration. This is the same
effect as observed by Huang et. al [29] for dye solar cells. It can
be explained by the higher loss reaction, which induces lower
electron density.
In short circuit under illumination, the cells with the lowest and
the middle I2 concentration do not bleach completely (Fig. 11).
The optical density after bleaching of the cells with the lowest
Figure 10: Colouring in open circuit in 1 sun illumination for
photoelectrochromic cells with different I2 concentrations.
Rate-Determining Processes in Photoelectrochromic Devices . / J. New Mat. Electrochem. Systems 8, 327-338 (2005)
335
I2 concentration is still quite high, the one of the cells with the
middle concentration is low. The reason is the limitation by
the ion transport. A possible application of this effect is that
the degree of bleaching under illumination can be adjusted by
selecting an appropriate I2 concentration.
Figure 12: Bleaching in short circuit in the dark for photoelectrochromic cells with propylene carbonate and acetonitrile as
the solvents.
Figure 11: Colouring in open circuit and bleaching in short circuit with 1 sun illumination for photoelectrochromic cells with
different I2 concentrations.
3.7
Variation of solven
Photoelectrochromic devices with two different solvents were
prepared: Propylene carbonate and acetonitrile. As we know
from measurement of the diffusion-limited current density, the
diffusion constant of I−
3 in acetonitrile is about ten times higher
than in propylene carbonate for the same ion concentrations
[13]. The optical density for the coloration in illumination in
open circuit does not show any significant difference between
the solvents (Fig. 4). This means that diffusion of ions in the
electrolyte does not influence the colouring. The short circuit
currents in illumination are 0.25 mA/cm2 for acetonitrile, and
0.17 mA/cm2 for propylene carbonate. This shows a certain influence of the ion transport, but it is less than the factor of ten
which would be expected if the current were limited by diffusion. Again the influence of convection can be seen.
The bleaching in short circuit in the dark is shown in Fig. 12.
Both optical density and current density are the same for both
solvents. Because of the large difference in the diffusion constants, this is proof that the ion transport has no influence on the
bleaching in short circuit in the dark.
On comparing with the variation of LiI concentration, we can
conclude that the diffusion of Li+ out of the WO3 particles is
limiting the bleaching in short circuit in the dark. Additionally,
it is now clear that the coverage of the WO3 surface with Li+
has not reached saturation.
3.8
Variation of Light Intensity
The coloration under illumination in open circuit for different
values of light intensity can be seen in Fig. 13. From Fig. 13,
the coloration velocity (initial rate of change of the optical density) was determined and is plotted as a function of the light
intensity in Fig. 14. The dependence of the coloration velocity
on the intensity is linear. This is proof that the colouring is determined by the electron generation in the dye-sensitised TiO2
layer.
The development of the open circuit voltage during colouring
shows several interesting results:
For lower light intensity (≤ 0.1 sun), the voltage increases continuously. For higher light intensity (≥0.25 sun), there is, as
described before, an increase of the voltage up to a maximum
value and then a decrease. This is in agreement with the interpretation above: For high intensity, when the electron injection
is faster than the phase transition, such a maximum in the voltage occurs. For lower light intensity, it does not occur.
For lower light intensity, the equilibrium voltage increases with
the intensity, for higher intensity it decreases again. This effect
cannot yet be explained completely.
For dye solar cells, the dependence of the equilibrium open circuit voltage (of the TiO2 layer) on the light intensity is logarithmic. This was shown experimentally and explained theoretically in [29]. This is not the case here, because the electro-
336
A. Georg et al./ J. New Mat. Electrochem. Systems 8, 327-338 (2005)
chemical potential of the WO3 , which changes its crystal structure during electron and Li+ injection, depends differently on
the electron density than the electrochemical potential of the
TiO2 in dye solar cells, where only electrons are injected and
not Li+ ions.
The short circuit current in equilibrium as a function of the light
intensity is plotted in Fig. 15. The dependence is linear for
intensity up to 50% sun, in this range the electron generation in
TiO2 is limiting the current. For 1 sun, it is limited by the ion
transport in the electrolyte.
Figure 13: Colouring in open circuit with different light intensities. 100% corresponds to an intensity of 1 sun (1000W/m2 ).
Figure 15: Short circuit current density in equilibrium for different light intensities. The dotted line is a linear fit of the measurements for intensities lower than 1 sun.
3.9
Variation of TiO2 Thickness
We constructed photoelectrochromic cells with different thicknesses of the TiO2 layer (150 nm and 370 nm). They were prepared by dipping the substrates several times into the solution.
For coloration in open circuit in 1 sun, the voltage and the optical density in equilibrium are higher for thicker TiO2 . The coloration velocity is roughly proportional to the thickness. This
supports the conclusion made after variation of light intensity
that the kinetic of the colouring process is limited by electron
generation.
4.
CONCLUSION
With this comprehensive investigation, a broad understanding
of the processes in the photoelectrochromic device has been
gained. The dominating processes are:
Figure 14: Coloration velocity (initial rate of change of the optical density), determined from fig. 13, as a function of the light
intensity.
For colouring in open circuit under illumination:
• Electron generation in the dye-sensitised TiO2
Rate-Determining Processes in Photoelectrochromic Devices . / J. New Mat. Electrochem. Systems 8, 327-338 (2005)
337
• Recombination of electrons from Lix WO3 with I−
3 in the
electrolyte
[9] A. Hauch, A. Georg, U. Opara Krašovec, B. Orel, J. Electrochem. Soc. 149 (9), H159 (2002).
• Phase transition in the WO3 crystallites (monoclinic =>
tetragonal => cubic)
[10] U. Opara Krašovec, Anneke Georg, Andreas Georg, G.
Dražič, submitted to Journal of Sol-Gel Science and Technology.
For self-bleaching in open circuit in the dark:
• Recombination of electrons from Lix WO3 with I−
3 in the
electrolyte
• Phase transition in the WO3 crystallites (monoclinic =>
tetragonal => cubic)
[11] Anneke Georg, Andreas Georg, U. Opara Krašovec, V.
Wittwer, submitted to the Journal of New Materials for
Electrochemical Systems.
[12] Q. Zhong, J.R. Dahn, K. Colbow, Physical Review B 46,
N◦ 4, 2554 (1992).
[13] A. Hauch, A. Georg, Electrochim. Acta 46, 3457 (2001).
For bleaching in short circuit in the dark:
[14] N. Yoskiike, M. Ayusawa, S. Kondo, J. Electrochem. Soc.
131 (11), 2600 (1984).
• Diffusion of Li+ out of WO3 particles
[15] M.K. Nazeeruddin, A. Kay, I. Rodicio, R. HumphryBaker, E. Müller, P. Liska, V. Vlachopoulos, M. Grätzel,
J. Am. Chem. Soc. 115, 6382 (1993).
For bleaching in short circuit under illumination:
[16] A. Hagfeldt, M. Grätzel, Chem. Rev. 95, 49 (1995).
• Diffusion of I−
3 and convection in the electrolyte
5.
ACKNOWLEDGMENTS
This work was supported financially by the University of Freiburg,
Germany and by the German Ministry of Education and Research BMBF.
REFERENCES
[17] R. Eichberger, F. Willig, Chem. Phys. 141, 159 (1990).
[18] N.W. Duffy, L.M. Peter, R.M.G. Rajapakse, K.G.U.
Wijayantha, Electrochemistry Communications 2, 658
(2000).
[19] M. Grätzel, MRS Bulletin 61 (1993).
[20] H. Lindström, S. Södergren, A. Solbrand, H. Rensmo, J.
Hjelm, A. Hagfeld, S.-E. Lindquist, J. Phys. Chem. B, 101,
7717 (1997).
[1] C.G. Granqvist, “Handbook on Inorganic Electrochromic
Materials”, ELSEVIER SCIENCE B.V., Amsterdam,
Netherlands, 1995.
[21] A. Hagfeldt, N. Vlachopoulos, S. Gilbert, M. Grätzel,
SPIE 2255, 297 (1994).
[2] C.G. Granqvist, Solar Energy Materials and Solar Cells
60, 201 (2000).
[22] R. van de Krol, A. Goossens, J. Schoonman, J. Phys.
Chem. B, 103, 7151 (1999).
[3] B. O’Regan, M. Grätzel, Nature 353, 737 (1991).
[23] M. Ottaviani, S. Panero, S. Morzilli, B. Srosati, M. Lazzari, Solid State Ionics 20, 197 (1986).
[4] K. Kalyanasundaram, M. Grätzel, Coordination Chemistry Reviews 77, 347 (1998).
[5] C. Bechinger, S. Ferrere, A. Zaban, J. Sprague, B. A.
Gregg, Nature 383, 608 (1996).
[6] F. Pichot, S. Ferrere, R. J. Pitts, B. A. Gregg, J. Electrochem. Soc. 146 (11), 4324 (1999).
[7] G. Teowee, T. Gudgel, K. McCarthy, A. Agrawal, P. Allemand, J. Cronin, Electrochim. Acta 44, 3017 (1999).
[8] A. Hauch A. Georg, S. Baumgärtner, U. Opara Krašovec,
B. Orel, B, Electrochim. Acta 46, 2131 (2001).
[24] S. K. Mohapatra, S. Wagner, J. Electrochem. Soc.: Electrochemical Science and Technology 1603 (1978).
[25] S. Kambe, S. Nakade, T. Kitamura, Y. Wada, S. Yanagida,
J. Phys. Chem. B 106, 2967 (2002)
[26] N.W. Duffy, L.M. Peter, R.M.G. Rajapakse, K.G.U. Wijayantha, Electrochemistry Communications 2, 658-662,
(2000)
[27] R. Greef, R. Peat, L.M. Peter, D. Pletcher, J. Robinson, “Instrumental Methods in Electrochemistry”, ELLIS
HORWOOD LIMITED, Chichester, England, 1985.
338
A. Georg et al./ J. New Mat. Electrochem. Systems 8, 327-338 (2005)
[28] K.J. Vetter, “Elektrochemische Kinetik”, SPRINGERVERLAG, Berlin Heidelberg, Germany 1961.
[29] S.Y. Huang, G. Schlichthörl, A.J. Nozik, M. Grätzel, A.J.
Frank, J. Phys. Chem. B 101, 2576 (1997).
[30] B. Zanden, A. Gossens, J. Phys. Chem. B 104, 7171
(2000).
[31] A. Solbrand, H. Lindström, H. Rensmo, A. Hagfeldt, S.
Lindquist, J. Phys. Chem. B 101, 2514 (1997).
[32] L. Dloczik, O. Ileperuma, I. Lauermann, L.M. Peter, E.A.
Ponomarev, G. Redmond, N.J. Shaw, I. Uhlendorf, J.
Phys. Chem. B 101, 10281 (1997).
[33] R. Kern, R. Sastrawan, J. Ferber, R. Stangl, J. Luther, Electrochim. Acta 47, 4213 (2002).
[34] J. Ferber, R. Stangl, J. Luther, Solar Energy Materials and
Solar Cells 53, 29 (1998).
[35] J. Ferber, Joachim Luther, J. Phys. Chem. B, 105, 4895
(2001).
[36] R. S. Crandall, B. W. Faughnan, Applied Physics Letters
26, No. 3, 120 (1975).
[37] J. Crank, “The Mathematics of Diffusion”, Oxford at the
Clarendon Press 1956.
[38] R.S. Crandall, P.J. Wojtowicz, B.W. Faughnan, Solid State
Communications 18, 1409 (1976).
[39] G. Redmond, D. Fitzmaurice, J. Phys. Chem. 97, 1426
(1993).
[40] S. A. Haque, Y. Tachibana, R. L. Willis, J. E. Moser, M.
Grätzel, D. R. Klug, J. R. Durrant, J. Physical Chemistry
B, 104, 538 (2000).
[41] U. Opara Krašovec, A. Šurca Vuk, B. Orel, Electrochim.
Acta 46, 1921 (2001).
[42] G. Smestad, Solar Energy Materials and Solar Cells 32,
273 (1994).
[43] V.A. Macagno, M.C. Giordano, A.J. Arvía, Electrochim.
Acta 14, 335 (1969).