Design of organic dyes and cobalt polypyridine redox mediators for high-efficiency dye-sensitized solar cells

Article abstract of DOI:10.1021/ja1088869

Dye-sensitized solar cells (DSCs) with cobalt-based mediators with efficiencies surpassing the record for DSCs with iodide-free electrolytes were developed by selecting a suitable combination of a cobalt polypyridine complex and an organic sensitizer. The effect of the steric properties of two triphenylamine-based organic sensitizers and a series of cobalt polypyridine redox mediators on the overall device performance in DSCs as well as on transport and recombination processes in these devices was compared. The recombination and mass-transport limitations that, previously, have been found to limit the performance of these mediators were avoided by matching the properties of the dye and the cobalt redox mediator. Organic dyes with higher extinction coefficients than the standard ruthenium sensitizers were employed in DSCs in combination with outer-sphere redox mediators, enabling thinner TiO₂ films to be used. Recombination was reduced further by introducing insulating butoxyl chains on the dye rather than on the cobalt redox mediator, enabling redox couples with higher diffusion coefficients and more suitable redox potential to be used, simultaneously improving the photocurrent and photovoltage of the device. Optimization of DSCs sensitized with a triphenylamine-based organic dye in combination with tris(2,2′-bipyridyl) cobalt(II/III) yielded solar cells with overall conversion efficiencies of 6.7% and open-circuit potentials of more than 0.9 V under 1000 W m^-2 AM1.5 G illumination. Excellent performance was also found under low light intensity indoor conditions.

Full text of DOI:10.1021/ja1088869

Published on Web 11/03/2010
Design of Organic Dyes and Cobalt Polypyridine Redox
Mediators for High-Efficiency Dye-Sensitized Solar Cells
Sandra M. Feldt,† Elizabeth A. Gibson,† Erik Gabrielsson,‡ Licheng Sun,‡
Gerrit Boschloo,*,† and Anders Hagfeldt†
Department of Physical and Analytical Chemistry, Uppsala UniVersity,
Box 259, 751 05 Uppsala, Sweden and Organic Chemistry, Center of Molecular DeVices,
Chemical Science and Engineering, Royal Institute of Technology, 100 44 Stockholm, Sweden
Received October 1, 2010; E-mail: gerrit.boschloo@fki.uu.se
Abstract: Dye-sensitized solar cells (DSCs) with cobalt-based mediators with efficiencies surpassing the
record for DSCs with iodide-free electrolytes were developed by selecting a suitable combination of a cobalt
polypyridine complex and an organic sensitizer. The effect of the steric properties of two triphenylamine-
based organic sensitizers and a series of cobalt polypyridine redox mediators on the overall device
performance in DSCs as well as on transport and recombination processes in these devices was compared.
The recombination and mass-transport limitations that, previously, have been found to limit the performance
of these mediators were avoided by matching the properties of the dye and the cobalt redox mediator.
Organic dyes with higher extinction coefficients than the standard ruthenium sensitizers were employed in
DSCs in combination with outer-sphere redox mediators, enabling thinner TiO2 films to be used.
Recombination was reduced further by introducing insulating butoxyl chains on the dye rather than on the
cobalt redox mediator, enabling redox couples with higher diffusion coefficients and more suitable redox
potential to be used, simultaneously improving the photocurrent and photovoltage of the device. Optimization
of DSCs sensitized with a triphenylamine-based organic dye in combination with tris(2,2′-bipyridyl)cobalt(II/
III) yielded solar cells with overall conversion efficiencies of 6.7% and open-circuit potentials of more than
0.9 V under 1000 W m-2 AM1.5 G illumination. Excellent performance was also found under low light
intensity indoor conditions.
Introduction
transparent conductive oxide substrate, which limits the per-
formance of large area DSCs and modules, is hindered. Several
Dye-sensitized solar cells (DSCs) have received great interest
as cost-effective alternatives to silicon-based photovoltaic
devices. In the DSC, light is absorbed by a dye molecule, which
is anchored to a mesoporous wide band gap semiconductor,
normally TiO2. Upon light absorption the photoexcited dye
injects an electron into the conduction band of the semiconductor
and its resulting oxidized state is regenerated by a redox
mediator in the surrounding electrolyte.1-3 Thus far, the best
conversion efficiencies in DSCs, about 11%,4 have been
obtained with ruthenium-based dyes together with an iodide/
triiodide redox couple.
redox mediators have so far been examined as alternative redox
couples in DSCs6-13 in order to avoid the problems associated
with the iodide/triiodide system. The best alternative redox
mediator to date is a recently reported disulfide/thiolate redox
couple that yielded an efficiency of 6.4% under full sunlight.14
One-electron outer-sphere redox couples, such as cobalt
complexes, are interesting alternative redox mediators since they
show weak visible light absorption and are less aggressive
For large-scale manufacturing of dye-sensitized solar cells
(DSCs) it is, however, of interest to find an alternative redox
mediator to the iodide/triiodide redox couple. The iodide/
triiodide system is limited by its relatively high overpotential
for dye regeneration5 and by competitive light absorption by
the triiodide.2,6 Moreover, the scale up of DSCs and the module
stability is hindered by the corrosiveness of the iodide/triiodide
redox couple toward most metals and sealing materials. As a
result, methods such as the use of current collecting metal grids
over the substrate to reduce the high series resistance of the
(4) Chiba, Y.; Islam, A.; Watanabe, Y.; Komiya, R.; Koide, N.; Han, L.
Jpn. J. Appl. Phys., Part 2 2006, 45, L638.
(5) Boschloo, G.; Hagfeldt, A. Acc. Chem. Res. 2009, 42, 1819.
(6) Nusbaumer, H.; Zakeeruddin, S. M.; Moser, J.-E.; Gra¨tzel, M.
Chem.sEur. J. 2003, 9, 3756.
(7) Gregg, B. A.; Pichot, F.; Ferrere, S.; Fields, C. L. J. Phys. Chem. B
2001, 105, 1422.
(8) Nusbaumer, H.; Moser, J.-E.; Zakeeruddin, S. M.; Nazeeruddin, M. K.;
Gratzel, M. J. Phys. Chem. B 2001, 105, 10461.
(9) Sapp, S. A.; Elliott, C. M.; Contado, C.; Caramori, S.; Bignozzi, C. A.
J. Am. Chem. Soc. 2002, 124, 11215.
(10) Oskam, G.; Bergeron, B. V.; Meyer, G. J.; Searson, P. C. J. Phys.
Chem. B 2001, 105, 6867.
(11) Hattori, S.; Wada, Y.; Yanagida, S.; Fukuzumi, S. J. Am. Chem. Soc.
2005, 127, 9648.
† Uppsala University.
‡ Royal Institute of Technology.
(12) Zhang, Z.; Chen, P.; Murakami, T. N.; Zakeeruddin, S. M.; Gra¨tzel,
M. AdV. Funct. Mater. 2008, 18, 341.
(1) O’Regan, B.; Gra¨tzel, M. Nature 1991, 353, 737.
(2) Hagfeldt, A.; Gra¨tzel, M. Chem. ReV. 1995, 95, 49.
(3) Hagfeldt, A.; Gra¨tzel, M. Acc. Chem. Res. 2000, 33, 269.
(13) Teng, C.; Yang, X.; Yuan, C.; Li, C.; Chen, R.; Tian, H.; Li, S.;
Hagfeldt, A.; Sun, L. Org. Lett. 2009, 11, 5542.
9
16714 J. AM. CHEM. SOC. 2010, 132, 16714–16724
10.1021/ja1088869  2010 American Chemical Society

High-Efficiency Dye-Sensitized Solar Cells
A R T I C L E S
Chart 1. Chemical Structure of D29 and D35
toward metallic conductors than iodine.6,9,15 Unfortunately,
recombination of electrons in the TiO2 with one-electron redox
mediators is much faster.7,16,17 It was, however, anticipated that
recombination of cobalt(III) should be relatively slow due to a
change in spin upon reduction to cobalt(II).18 Nevertheless,
previous investigations have suggested that the performance of
cobalt redox mediators in DSCs is limited by rapid recombina-
tion from electrons in the TiO2 conduction band to the cobalt(III)
species and furthermore by slow dye regeneration and mass
transport problems in the mesoporous TiO2 electrode.17,19 Much
research has recently been focused on passivating the TiO2
surface using, for example, atomic layer deposition (ALD) in
order to diminish the fast recombination processes occurring
when cobalt redox mediators are used.17,20,21
and D35 (Chart 1),25 were compared to investigate the effect
of bulky alkoxyl substituents on the sensitizer in devices
employing cobalt polypyridine redox mediators. In the case of
D29 the electron-donating substituent was p-N,N-dimethyla-
nilinyl and for D35 o,p-dibutoxyphenyl. In a previous study it
was shown that inclusion of bulky alkoxy electron-donating
substituents efficiently suppresses recombination in iodide/
triiodide-based DSCs.26 These organic sensitizers have high
extinction coefficients (about 30 000 M-1 cm-1 at their absorp-
tion maximum)27 compared to ruthenium-based dyes (about
14 000 M-1 cm-1 for N719). Dyes with high extinction
coefficients are advantageous in DSCs that use ionic liquids or
solid-state hole conductors, since they allow for good light
harvesting efficiencies using thinner TiO2 films. A reduction in
film thickness will decrease the charge recombination rate and
the transport time, and can improve surface wetting and pore
filling of the electrode by the electrolyte or hole conductor, thus
improving the charge collection efficiency.28,29 In liquid DSCs
incorporating an electrolyte with diffusion-limiting properties,
it is of great importance to build thin-film DSCs in order to
avoid mass transport problems of redox species through the
mesoporous semiconductor structure.30
Here, we will report on highly efficient DSCs based on TiO2,
organic dyes, and cobalt redox couples. A strategy of tuning
the steric properties of both the sensitizer and the redox mediator
and replacing standard ruthenium sensitizers with organic dyes
with high extinction coefficients, to enable thinner mesoporous
TiO2 films to be used, has been employed to overcome mass
transport and recombination limitations associated with cobalt
polypyridine redox electrolytes. Triphenylamine (TPA)-based
organic dyes have been widely reported as sensitizers in DSCs
in recent years, due to the favorable electron-donating properties
of the triphenylamine unit and their potential for low-cost
production.22-24 Here, two TPA-based organic sensitizers, D29
The steric bulk of a selection of bipyridyl and phenanthroline
cobalt redox shuttles was varied by introducing different alkyl
substituents on the polypyridine ligands. An energy diagram of
a DSC sensitized with D29 and D35, employing the different
cobalt redox shuttles compared here, and the chemical structure
of the complexes are shown in Figure 1a and 1b, respectively.
The redox species were cobalt(III/II) tris(2,2′- bipyridine),
[Co(bpy)3]3+/2+, cobalt(III/II) tris(4,4′-dimethyl-2,2′- bipyridine),
[Co(dmb)3]3+/2+, cobalt(III/II) tris(4,4′-ditert-butyl-2,2′-bipyri-
dine), [Co(dtb)3]3+/2+, and cobalt(III/II) tris(1,10-phenanthroline),
(14) Wang, M.; Chamberland, N.; Breau, L.; Moser, J.-E.; Humphry-Baker,
R.; Marsan, B.; Zakeeruddin, S. M.; Gra¨tzel, M. Nat. Chem. 2010, 2,
385.
(15) Nusbaumer, H.; Moser, J.-E.; Zakeeruddin, S. M.; Nazeeruddin, M. K.;
Gratzel, M. J. Phys. Chem. C 2001, 105, 10461.
[Co(phen)3]3+/2+
.
(16) Feldt, S. M.; Cappel, U. B.; Johansson, E. M. J.; Boschloo, G.;
Hagfeldt, A. J. Phys. Chem. C 2010, 114, 10551.
(17) Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C 2009, 113, 14040.
(18) Cameron, P. J.; Peter, L. M.; Zakeeruddin, S. M.; Gra¨tzel, M. Coord.
Chem. ReV. 2004, 248, 1447.
(25) Hagberg, D. P.; Jiang, X.; Gabrielsson, E.; Linder, M.; Marinado, T.;
Brinck, T.; Hagfeldt, A.; Sun, L. J. Mater. Chem. 2009, 19, 7232.
(26) Jiang, X.; Marinado, T.; Gabrielsson, E.; Hagberg, D. P.; Sun, L.;
Hagfeldt, A. J. Phys. Chem. C 2010, 114, 2799.
(19) Nelson, J. J.; Amick, T. J.; Elliott, C. M. J. Phys. Chem. C 2008, 112,
18255.
(20) Ondersma, J. W.; Hamann, T. W. J. Phys. Chem. C 2009, 114, 638.
(21) DeVries, J. W.; Pellin, T. W.; Hupp, J. T. Langmuir 2010, 26, 9082.
(22) Kitamura, T.; Ikeda, M.; Shigaki, K.; Inoue, T.; Anderson, N. A.; Ai,
X.; Lian, T.; Yanagida, S. Chem. Mater. 2004, 16, 1806.
(23) Hagberg, D. P.; Edvinsson, T.; Marinado, T.; Boschloo, G.; Hagfeldt,
A.; Sun, L. Chem. Commun. 2006, 2245.
(27) In ethanol: D29: εmax(485 nm) ) 27.9 × 103 M-1 cm-1; D35: εmax(463
nm) ) 28.6 × 103 M-1 cm-1. Note that erroneous values are reported
in refs 25 and 26.
(28) Kuang, D.; Ito, S.; Wenger, B.; Klein, C.; Moser, J.-E.; Humphry-
Baker, R.; Zakeeruddin, S. M.; Gra¨tzel, M. J. Am. Chem. Soc. 2006,
128, 4146.
(24) Hagberg, D. P.; Marinado, T.; Karlsson, K. M.; Nonomura, K.; Qin,
P.; Boschloo, G.; Brinck, T.; Hagfeldt, A.; Sun, L. J. Org. Chem.
2007, 72, 9550.
(29) Wang, M.; Liu, J.; Cevey-Ha, N.-L.; Moon, S.-J.; Liska, P.; Humphry-
Baker, R.; Moser, J.-E.; Gra¨tzel, C.; Wang, P.; Zakeeruddin, S. M.;
Gra¨tzel, M. Nano Today 2010, 5, 169.
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16715

A R T I C L E S
Feldt et al.
Figure 1. (a) Schematic energy diagram for a nanostructured TiO2 electrode sensitized with D35 or D29 employing [Co(bpy)3]2+/3+-, [Co(dmb)3]2+/3+-,
[Co(dtb)3]2+/3+-, and [Co(phen)3]2+/3+-based electrolytes. Formal potentials indicated are vs NHE. (b) Chemical structure of the different cobalt polypyridyl
redox couples employed herein.
After the sintering, when the temperature cooled to about 90 °C,
the electrodes were immersed in a dye bath containing 0.2 mM
D35 in ethanol or 0.2 mM D29 in acetonitrile and left overnight.
The films were then rinsed in ethanol to remove excess dye. Solar
cells were assembled, using a 30 µm thick thermoplastic Surlyn
frame, with a platinized counter electrode (TEC8), which was
prepared by depositing 10 µL cm-2, 4.8 mM H2PtCl6 solution in
ethanol to the glass substrate followed by heating in air at 400 °C
for 30 min. An electrolyte solution was then introduced through
two holes predrilled in the counter electrode, and the cell was sealed
with thermoplastic Surlyn covers and a glass coverslip. Unless
otherwise noted, the electrolyte consisted of 0.22 M Co(L)3(PF6)2,
0.033 M Co(L)3(PF6)3, 0.1 M LiClO4, and 0.2 M 4-tert-butylpy-
ridine (TBP) in acetonitrile, where L ascribes the different poly-
pyridine ligands. For comparison, DSCs were prepared using 0.6
M tetrabutylammonium iodide, 0.1 M LiI, 0.05 M I2, and 0.2 M
TBP in acetonitrile.
Solar Cell Characterization. Current-voltage (I-V) charac-
teristics were measured using a Keithley 2400 source/meter and a
Newport solar simulator (model 91160) giving light with AM 1.5
G spectral distribution, which was calibrated using a certified
reference solar cell (Fraunhofer ISE) to an intensity 1000 W m-2
or with the help of a neutral density filter to 100 W m-2. A black
mask with an aperture slightly larger (0.6 × 0.6 cm2) than the active
area of the solar cell (0.5 × 0.5 cm2) was applied on top of the cell
to avoid significant additional contribution from light falling on
the device outside the active area. Photovoltaic measurements at
low light intensities were performed using a compact fluorescent
lamp (Osram Dulux Value 8W/827, warm white) as a light source
at an illuminance of 250 lx. Incident photon to current conversion
efficiency (IPCE) spectra were recorded using a computer-controlled
setup consisting of a xenon light source (Spectral Products ASB-
XE-175), a monochromator (Spectral Products CM110), and a
potentiostat (EG&G PAR 273), calibrated using a certified reference
solar cell (Fraunhofer ISE). Electron lifetime, transport times, and
extracted charge measurements were performed using a white LED
(Luxeon Star 1W) as the light source. Voltage and current traces
were recorded with a 16-bit resolution digital acquisition board
(National Instruments) in combination with a current amplifier
(Stanford Research Systems SR570) and a custom-made system
using electromagnetic switches. Transport and lifetimes were
determined by monitoring photocurrent and photovoltage transients
at different light intensities upon applying a small square wave
modulation to the base light intensity. The photocurrent and
photovoltaic responses were fitted using first-order kinetics to obtain
time constants.31 Extracted charge measurements were performed
by illuminating the cell for 10 s under open-circuit conditions and
then turning the lamp off to let the voltage to decay to a voltage V.
By tuning the steric bulk of the dye and the redox mediator,
the recombination and mass transport problems associated with
the electrolyte were minimized and highly efficient DSCs were
obtained. Solar cell efficiencies recorded in this study surpass
the highest values reported so far for devices containing iodine-
free electrolytes, reaching 6.7% overall conversion efficiency
under 1000 W m-2 AM1.5G illumination and 7.8% at 100 W
m-2
.
Experimental Section
All chemicals were purchased from Sigma Aldrich unless
otherwise noted. D35 and D29 were prepared according to the
published procedure.25
Synthesis of Cobalt Complexes. The cobalt complexes
[Co(bpy)3(PF6)2]2+, [Co(dmb)3(PF6)2]2+, [Co(dtb)3(PF6)2]2+, and
[Co(phen)3(PF6)2]2+ were synthesized as described elsewhere.9,17
Briefly, 1 equiv of CoCl2 ·6H2O and 3.3 equiv of the polypyridine
ligand were dissolved in a minimal amount of methanol (Merck),
and the solution was stirred at reflux for 2 h. An excess of
ammonium hexafluorophosphate was then added to the solution to
precipitate the compound that was filtered, washed with methanol
and ethanol, dried under vacuum, and used without further
purification.
Oxidation of the cobalt(II) complexes was performed by adding
a slight excess of NOBF4 to an acetonitrile solution of the complex
and then by removing the acetonitrile solution by rotary evaporation.
The complex was then redissolved in acetonitrile, and a large
amount of NH4PF6 was added to the solution. The final product
was precipitated with diethyl ether, filtered, dried under vacuum,
and used without further purification.
Solar Cell Preparation. Fluorine-doped tin oxide (FTO) glass
substrates (Pilkington, TEC15) were cleaned in an ultrasonic bath
overnight using (in order) ethanol, water, and ethanol. The
conducting glass substrates were pretreated by immersion in a 40
mM aqueous TiCl4 solution at 70 °C for 30 min and then washed
with water. Mesoporous TiO2 films of different thicknesses were
prepared with an area of 0.25 cm2 by repeatedly screen printing
colloidal TiO2 paste (Dyesol DSL 30 NRD-T) and drying at 300
°C between deposition steps. After sintering, the thickness of the
films was measured with a profilometer (Veeco Dektak 3) and was
5.6 µm, unless otherwise noted. Only in the optimization study for
some cells a light-scattering TiO2 layer (PST-400C, JGC Catalysts
and Chemical Ltd., gratefully received from JGC Catalysts and
Chemical Ltd.) was deposited on top of the mesoporous TiO2 film.
The electrodes were gradually heated in an oven (Nabertherm
Controller P320) in an air atmosphere. The temperature gradient
program had four levels at 180 (10 min), 320 (10 min), 390 (10
min), and 500 °C (60 min). After sintering, the electrodes were,
once again, immersed in aqueous TiCl4 and rinsed as above. A
second and final heating step, at 500 °C for 60 min, was performed.
(30) Ito, S.; Zakeeruddin, S.; Humphry-Baker, R.; Liska, P.; Charvet, R.;
Comte, P.; Nazeeruddin, M.; Pe´chy, P.; Takata, M.; Miura, H.; Uchida,
S.; Gra¨tzel, M. AdV. Mater. 2006, 18, 1202.
9
16716 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010

High-Efficiency Dye-Sensitized Solar Cells
A R T I C L E S
The cell was then short circuited, and the current was measured
under 0.1 s and integrated to obtain Qoc (V).
the steric bulk of the cobalt complex. As no antireflection coating
was applied on these solar cells, about 10% of the incoming
light is reflected. This implies that 100% of absorbed photons
are converted into electrons. In the spectral range between 360
and 460 nm the IPCE was higher for DSCs employing cobalt
redox mediators than for those employing iodide/triiodide. This
is due to competitive light absorption by triiodide. This effect
is even more marked when the cell is illuminated from the
counter electrode side rather than through the working electrode
(see Figure S1 in the Supporting Information).
Photoinduced Absorption Spectroscopy (PIA). PIA measure-
ments were performed using the same setup as described else-
where.32 A square wave modulated blue LED (Luxeon Star 1 W,
Royal Blue, 460 nm) used for excitation was superimposed on a
white probe light provided by a 20 W tungsten-halogen lamp. The
transmitted probe light was focused onto a monochromator (Acton
Research Corp. SP-150) and detected using a UV-enhanced Si
photodiode, connected to a lock-in amplifier (Stanford Research
Systems model SR830) via a current amplifier (Stanford Research
Systems model SR570). The intensity of the probe light was
approximately 100 W m-2, and the intensity of the excitation LED
was approximately 80 W m-2. The modulation frequency of the
LED was 9.33 Hz.
Electrochemical Measurements. Cyclic voltammetry was per-
formed on a CH Instruments 660 potentiostat using a three-electrode
setup. For the diffusion measurements a 20 µm diameter platinum
microelectrode was used as the working electrode, glassy carbon
as the counter electrode, and a Ag/AgNO3 (10 mM AgNO3, 0.1 M
TBAPF6 in acetonitrile) reference electrode. The reference electrode
was calibrated versus ferrocene in the same supporting electrolyte.
The redox electrolyte solution was 15 mM Co(L)3(PF6)2 and 0.1
M TBAPF6 in acetonitrile. The scan rate was 10 mV s-1. Charge-
transfer resistances at the counter electrode were determined by
impedance spectroscopy, using a two-electrode setup consisting of
two platinized FTO electrodes (prepared as for the DSCs) in a
sandwich device using a 30 µm thick thermoplastic frame (inner
dimensions 7 mm × 7 mm), infiltrated with redox electrolyte as
used in the DSCs. Measurements were performed on an Iviumstat
XR potentiostat at a bias of 0 V and an ac frequency range between
100 kHz and 0.1 Hz.33
The current-voltage curves for this series of DSCs are shown
in Figure 2c and 2d. The photovoltaic performance of cobalt
polypyridine redox mediators was superior to that of the iodide/
triiodide redox shuttle for thin-film DSCs sensitized with
triphenylamine-based organic sensitizers. The best efficiency
at 1000 W m-2 illumination for the DSCs sensitized with D29
of 2.6% was obtained using a [Co(dmb)3]3+/2-based electrolyte.
At this intensity a lower efficiency was obtained when [Co-
(dtb)3]3+/2 was used, which is, most likely, a result of slow mass
transport of this redox mediator. Despite the enhanced spectral
response of D29 in the red region of the electromagnetic
spectrum (see Figure S2 in the Supporting Information) the
efficiency was higher for DSCs sensitized with D35. The higher
efficiency obtained using D35 is due to the decreased recom-
bination rate between electrons in TiO2 and Co(III) species in
the electrolyte, as follows from electron lifetime studies (see
next section) and I-V curves recorded in the dark (see Figure
S3 in the Supporting Information). The bulky butoxyl groups
on D35 molecules prevent close contact between the oxidized
species in the electrolyte and the TiO2 surface.
Results and Discussion
Among the different redox mediators employed, [Co-
(bpy)3]3+/2+ in combination with D35 gave the highest efficiency
of 4.9%. By comparison, a modest efficiency of 1.3% was
obtained using [Co(bpy)3]3+/2+ in combination with ruthenium-
based dyes in Klahr and Hamann’s study, which was explained
by fast recombination, even though insulating alumina coatings
were applied on the TiO2 surface.17 The open-circuit voltage
(Voc) for DSCs sensitized with D35 is remarkably high as a
result of the insulating effect of the alkoxyl groups and the more
positive potential of Co-based redox mediators. The high Voc
for the D35-based DSCs employing [Co(bpy)3]3+/2+ (0.88 V)
is in stark contrast to the results obtained by other groups, where
the use of one-electron redox couples with a more positive redox
potential than the iodide/triiodide redox shuttle was found to
lead to high dark current densities and lower photovoltages.7,10,34
The lower photocurrent observed using the iodide/triiodide redox
couple illustrates one of the advantages of using cobalt redox
couples in DSCs and can partly be explained by the loss in
photocurrent due to light adsorption by triiodide. In agreement
with earlier reports, the fill factor was rather poor using
[Co(dtb)3]3+/2+, which has been attributed to the mass transport
problems associated with these species that lead to increased
series resistance in the device.17,19 The fill factor was better for
DSCs based on D29 compared to D35. This could be a result
of the lower resistive losses due to the lower current densities
obtained with D29 or, possibly, due to the steric bulk of D35
that slightly hinders the mass transport of the cobalt bipyridyl
complexes through the mesoporous TiO2 film. The effect of
mass transport limitations in cobalt mediator-based DSCs will
be discussed in a later section.
Effect of the Steric Properties of the Dye and Redox
Shuttle on the Photovoltaic Performance. Incident-photon-to-
current conversion efficiency (IPCE) spectra were recorded for
DSCs sensitized with the organic dyes D29 and D35, containing
[Co(bpy)3]3+/2+, [Co(dmb)3]3+/2+, [Co(dtb)3]3+/2+, or the iodide/
triiodide redox shuttle, see Figure 2a and 2b. The IPCE for the
D29-based DSCs increased with the steric bulk of the cobalt
complexes, and a maximum IPCE of about 70% was found for
DSCs employing [Co(dtb)3]3+/2+. A similar trend was found by
Hamann and co-workers for DSCs sensitized by the Ru
complexes Ru(dcby)2(NCS)2 and Ru(dcbp)(bpy)2(PF6)2 with the
same series of redox mediators.17,20 The good performance of
[Co(dtb)3]3+/2+ in the IPCE measurements is attributed to the
bulky tert-butyl substituents that slow down the recombination
rate.17,20 Although Nelson et al. showed that mass transport of
[Co(dtb)3]3+/2+ in the pores of the TiO2 film can be slow and
limit the overall photovoltaic performance,19 this should not limit
the photocurrent in IPCE measurements, because low mono-
chromatic light intensities (1-5 W m-2) are used. The lower
IPCE values obtained for the devices containing the less bulky
cobalt complexes used in conjugation with D29 could be a result
of the smaller driving force for the regeneration process, i.e.,
the reduction of oxidized dye by the Co(II) complex, and/or
faster recombination kinetics between injected electrons in TiO2
and the Co(III) complex in the electrolyte. The different aspects
will be discussed in detail in the next sections.
The IPCE for DSCs sensitized with D35 was about 90% for
each of the different redox mediators employed, regardless of
(31) Boschloo, G.; Haggman, L.; Hagfeldt, A. J. Phys. Chem. B 2006, 110,
13144.
(32) Boschloo, G.; Hagfeldt, A. Inorg. Chim. Acta 2008, 361, 729.
(33) Hauch, A.; Georg, A. Electrochim. Acta 2001, 46, 3457.
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16717

A R T I C L E S
Feldt et al.
Figure 2. Spectra of incident photon to current efficiency (IPCE) for DSCs sensitized with (a) D29 and (b) D35. Current density versus applied potential
curves under AM1.5G illumination for DSCs sensitized with (c) D29 and (d) D35. Electron lifetime as a function of extracted charge under open-circuit
conditions for DSCs sensitized with (e) D29 and (f) D35. Every fifth data point is marked with a symbol in the IPCE spectra and every 20th data point in
the I-V plots.
insulating blocking layers to the TiO2 surface.20,21,35,36 While
Effect of the Steric Properties of the Dye and Redox
this had been shown to reduce the charge recombination and
increase the photovoltage obtained, the electron injection kinetics
may be affected, leading to a reduction in photocurrent.
The current-voltage curves for the DSCs sensitized with D35
Shuttle on the Charge Recombination Reaction. Losses due to
charge recombination between the electrons in the semiconduc-
tor and the oxidized species in the redox electrolyte limit the
DSC efficiency by lowering the charge-collection efficiency and,
therefore, the photocurrent. In addition, fast recombination will
lead to lower concentrations of electrons in the conduction band
and correspondingly lower photovoltages. For many one-
electron, outer-sphere redox couples, faster recombination and
shorter electron lifetimes have been measured compared to
devices incorporating iodide/triiodide, and this is thought to be
the main limitation to these alternative redox systems7,11,12,16,34
To avoid this problem, a number of groups have applied thin,
employing [Co(bpy)3]3+/2+-, [Co(dmb)3]3+/2+-, or [Co(phen)3]3+/2+
-
based electrolytes are shown in Figure 3. The Voc increased with the
Nernst potential of the redox shuttles and the highest Voc was
obtained using [Co(phen)3]3+/2+. DeVries et al.21 found that for
(34) Li, T. C.; Spokoyny, A. M.; She, C.; Farha, O. K.; Mirkin, C. A.;
Marks, T. J.; Hupp, J. T. J. Am. Chem. Soc. 2010, 132, 4580.
(35) Hamann, T. W.; Farha, O. K.; Hupp, J. T. J. Phys. Chem. C 2008,
112, 19756.
9
16718 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010

High-Efficiency Dye-Sensitized Solar Cells
A R T I C L E S
dependent on the radius of the ions, and therefore, [Co-
(dtb)3]3+/2+ should have a smaller reorganization energy than
[Co(dmb)3]3+/2+. However, the electronic coupling of the donor
and acceptor has a larger effect on the rate of electron transfer.
Since this is exponentially distance dependent, the rate of
electron transfer (k) depends on the spatial separation of the
donor and acceptor (R) according to eq 137
k ) k0 exp[-ꢀ(R - R0)]
(1)
where R0 is the van der Waals separation and k0 and ꢀ are
constants. A possible (smaller) distance dependence of the
reorganization energy is ignored in this equation.
The increased steric bulk of [Co(dtb)3]3+/2+ decreases the
electronic coupling compared to [Co(dmb)3]3+/2+ and reduces
the rate of recombination accordingly. The trend of increasing
electron lifetime as the steric bulk of the redox mediator was
increased is consistent with the corresponding increase in IPCE.
This is in agreement with previous results from Klahr and
Hamann17 where the bulky groups on [Co(dtb)3]3+/2+ were found
to act as insulating spacers, hindering the recombination
processes.
For D35, however, the insulating effect of the alkoxy chains
on the dye itself efficiently hinders recombination and the
electron lifetimes were increased significantly for all of the redox
couples investigated. D35 blocks the dark current for each of
the redox couples more efficiently than D29, consistent with
the trend in electron lifetime for the dyes. The differences in
electron lifetime between the cobalt complexes with different
substituents were less pronounced with D35 than with D29.
Since the electron transfer rate is exponentially dependent on
the distance, at longer distances from the semiconductor surface
the steric effects of the bipyridine substituents have less effect.
It appears that D35 prevents the close approach of the cobalt
complexes to the TiO2 surface to such extent that there is little
difference in the recombination rate between [Co(dmb)3]3+/2+
and [Co(dtb)3]3+/2+. Instead, the driving force for recombination
appears to be more important at this distance, and accordingly,
the device containing [Co(bpy)3]3+/2+ has a shorter lifetime than
the DSCs with the other redox couples. While the electron
lifetime was slightly shorter when [Co(bpy)3]3+/2+ was em-
ployed, the photovoltaic performance was still good when used
in combination with D35 and fast recombination kinetics did
not appear to be limiting the efficiency.
When the electron lifetimes in DSCs with either the iodide/
triiodide or the cobalt-based mediators were compared, it was found
that the former gave the longest lifetimes, with one exception: for
D29 lifetimes were very similar for I-/I3- and [Co(dtb)3]3+/2+. As
was noted in our previous work, the electron lifetimes in I-/I3-
-based DSCs were shorter for D29 than for D35, which was mainly
attributed to the blocking effect of the steric groups on D35.28 On
substituting D35 for D29, the electron lifetimes with the I-/I3-
mediator increased significantly more than with the [Co(dtb)3]3+/
2+ redox mediator (Figure 2e and 2f). Specific chemical interactions
between triiodide (or iodine) with the dye molecules are likely38
and probably play a role in the relatively short lifetime in D29-
sensitized solar cells. Similar chemical interactions are not likely
to occur between the Co complexes and the dye molecules. One
potential disadvantage of the investigated cobalt-based mediators
is their positive charge. Nelson et al.19 observed electrostatic
Figure 3. Current density versus applied potential curves under AM1.5G
illumination for DSCs sensitized with D35. Every 20th data point is marked
with a symbol.
a series of cobalt complexes with phenanthroline derivatives
the Voc increased as the formal potential of the redox couple
became more positive. In contrast, the quasi-Fermi level in the
TiO2 at open circuit became less negative with more positive
redox potential, because of faster interception of photoinjected
electrons by Co(III) as the driving force for recombination
increased, indicating Marcus normal-region behavior. Our results
here are in agreement with this explanation; however, we
propose that the recombination is not only dependent on the
driving force for recombination but also on the steric properties
of the redox mediator.
The recombination kinetics were investigated using a small
amplitude light-modulation technique to study the effects of the
steric properties of the dyes and the redox mediators on the electron
lifetime. The electron lifetime as a function of extracted charge
for the DSCs sensitized with D29 and D35, employing the three
different cobalt bipyridyl redox mediators or iodide/triiodide, is
shown in Figure 2e and 2f, respectively. As anticipated, for the
devices incorporating D29 as the sensitizer, the electron lifetime
was dependent on the substituents around the periphery of the cobalt
complexes. According to the Marcus-Gerischer model, electron
transfer rates are dependent on the rate of diffusion of the reactants
(we will discuss this in a later section), the driving force for the
reaction, the reorganization energy, and the electronic coupling.
Cobalt complexes are known to have relatively slow electron
transfer kinetics compared to other transition metal complexes such
as iron(II/III) on account of the large internal reorganization energy
when going from d7 (high spin) to d6 (low spin). Both iron(II) (d6)
and iron(III) (d5) are low-spin, octahedral complexes and, therefore,
fast redox shuttles which perform poorly as redox electrolytes in
complete devices, in the absence of insulating layers on the TiO2
surface, on account of the substantially shorter electron lifetime.16,35
The redox potentials of the cobalt bipyridine complexes
employed in this study are shown in Figure 1 and are 0.56 V
vs NHE for [Co(bpy)3]3+/2+ and 0.43 V vs NHE for [Co-
(dmb)3]3+/2+ and [Co(dtb)3]3+/2+ 9,17 Assuming that recombina-
.
tion occurs between electrons in the conduction band of the
TiO2 and cobalt(III) species in the electrolyte, the driving force
for recombination will be proportional to the potential difference
between the electrons in the TiO2 conduction band and the redox
potential of the redox couple in solution. Therefore, in terms
of driving force, the recombination is expected to be fastest for
[Co(bpy)3]3+/2+; the electron lifetime data (Figure 2e) is in
agreement with this. The redox potential of [Co(dmb)3]3+/2+ and
[Co(dtb)3]3+/2+ is similar, but the electron lifetimes for the two
devices were very different. Solvent reorganization energy is
(36) Palomares, E.; Clifford, J. N.; Haque, S. A.; Lutz, T.; Durrant, J. R.
J. Am. Chem. Soc. 2002, 125, 475.
(37) Closs, G. L.; Miller, J. R. Science 1988, 240, 440.
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16719

A R T I C L E S
Feldt et al.
Figure 4. Light intensity dependence of (a) the open-circuit voltage and (b) the current density under short-circuit conditions for DSCs sensitized with D35.
adsorption of cationic complexes, including [Co(dtb)3]3+/2+, on the
surface of TiO2. Electrostatic interactions would likely speed up
the interfacial charge transfer,39 giving rise to the shorter lifetimes
observed compared to the anionic triiodide/iodide redox electrolyte.
The best overall solar cell efficiency was obtained with the
combination of D35 and [Co(bpy)3]3+/2+, despite the fact that
electron lifetimes were about 1 order of magnitude lower than
Figure S4 in the Supporting Information). The slope increased
up to 131 mV/decade for DSCs with electrolyte containing
[Co(bpy)3]3+/2+ and 0.1 M TBAPF6. The increase of the slope
is correlated with the decrease of steric bulk of the cobalt
complexes and the decrease of electron lifetime. This effect was
not found when Li+ ions were present in the electrolyte, possibly
because charge compensation of the electrons in TiO2 occurs
mainly by Li+ ions in this case. Nakade et al.43 showed that
recombination was slowed down by addition of Li+ in DSCs
employing a cobalt redox mediator, and they attributed the effect
to a decrease in local concentration of Co(III) caused by
electrostatic repulsion due to adsorbed Li cations. When Li+
ions are replaced by bulky TBA+ ions, it is likely that the
positively charged Co complexes are more involved in local
charge compensation of electron in TiO2, which will affect the
recombination kinetics.
Mass Transport Limitations in Cobalt-Based DSCs. The
effect of slow mass transport in DSCs employing cobalt
bipyridyl redox mediators was investigated by monitoring
photocurrent transients using a large modulation (on/off) of the
incident light.19 Using a white LED, with the output set to match
the photocurrent obtained under 1000 W m-2 AM1.5G il-
lumination, photocurrent transients were recorded for D29- and
D35-sensitized DSCs with the different cobalt bipyridyl redox
mediators, see Figure 5. A maximum in photocurrent was
observed when the light is switched on, followed by a decrease
in photocurrent, reaching a constant value after a few seconds,
similar to the observations by Nelson et al.19 They found that
the steady-state current increased with increasing concentration
of Co(III) in the electrolyte and suggested that diffusion of
Co(III) to the counter electrode limits the cell performance.19
We found that cell performance and, in particular, the fill factor
increase with the increasing concentration of Co(III) (see Figure
S5 in the Supporting Information), which agrees with the
conclusion that slow mass transport of Co(III) to the counter
electrode limits the cell performance. The ratio of initial peak
current to steady state current in the photocurrent transients
-
that of the D35-I-/I3 combination at the same electron
concentration in the TiO2 electrode. This can be attributed to
the ∼0.2 V more positive redox potential of the cobalt mediator.
Nonlinearity of the Charge Recombination Kinetics. The light
intensity dependence on the open-circuit potential for DSCs
sensitized with D35 employing the three different cobalt
bipyridyl redox mediators and iodide/triiodide is shown in Figure
4a. The slope in the Voc versus light intensity plot was about 76
mV/decade for each of the redox mediators. If recombination
occurs solely via the conduction band of the TiO2 and the
reaction is first order in the concentration of conduction band
electrons, a linear increase of the conduction band electron
concentration and the recombination rate with light intensity is
predicted, resulting in a slope of 59 mV/decade at 298 K in the
Voc versus log(light intensity) plot. The observed deviation,
implying nonlinear recombination kinetics, can be due to
electron transfer via surface states, effects of the transfer
coefficient for the electron transfer process, and/or shifts in the
conduction band under illumination.40 A recombination reaction
mediated by a distribution of surface states at the TiO2/
electrolyte interface appears to be the most likely cause.41
Jennings and Wang showed recently for N719-sensitized DSCs
with iodide/triiodide-based electrolyte that the “ideality” and
the linearity of the semilogarithmic plots of the photovoltage
versus incident photon flux increase with increasing Li+
concentration.42 For the Co-mediator-based DSCs investigated
here, the slope in the Voc vs light intensity plots was also affected
by the choice of cation in the supporting electrolyte. The slope
for DSCs containing 0.1 M TBAPF6 in the electrolyte was higher
than for DSCs containing 0.1 M LiClO4 and was found to
depend on the substituents on the cobalt redox mediators (see
decreases in the order [Co(dtb)3]3+/2+ > [Co(dmb)3]3+/2+
>
[Co(bpy)3]3+/2+, indicating that mass transport is retarded by
the steric bulk of the redox mediator. For the combination D35
and [Co(bpy)3]3+/2+ the decrease in photocurrent was negligible
(<1.5%). At lower light intensities, mass transport is no longer
a problem and no current spike is observed (data not shown).
The initial peak current for DSCs sensitized with D29 increases
with the steric bulk of the cobalt complexes but is about the
(38) O’Regan, B. C.; Walley, K.; Juozapavicius, M.; Anderson, A.; Matar,
F.; Ghaddar, T.; Zakeeruddin, S. M.; Klein, C. d.; Durrant, J. R. J. Am.
Chem. Soc. 2009, 131, 3541.
(39) Gaal, D. A.; McGarrah, J. E.; Liu, F.; Cook, J. E.; Hupp, J. T.
Photochem. Photobiol. Sci. 2004, 3, 240.
(40) Villanueva-Cab, J.; Wang, H.; Oskam, G.; Peter, L. M. J. Phys. Chem.
Lett. 2010, 1, 748.
(41) Salvador, P.; Hidalgo, M. G.; Zaban, A.; Bisquert, J. J. Phys. Chem.
B 2005, 109, 15915.
(42) Jennings, J. R.; Wang, Q. J. Phys. Chem. C 2009, 114, 1715.
9
16720 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010

High-Efficiency Dye-Sensitized Solar Cells
A R T I C L E S
Figure 5. Plots of current transients measured at an illumination intensity of 1000 W m-2 for DSCs sensitized with (a) D29 and (b) D35.
same for DSCs sensitized with D35, suggesting that the current
depends not only on the slower mass transport of the bulky
redox species in the case of D29 but also on the recombination
of photoinjected electrons with the oxidized form of the redox
mediator, as discussed above.
the redox mediator and the dye, since the size of the electron
donor and acceptor should affect the reorganization energy and
the electronic coupling.44 While no time constants for the
regeneration were measured in this work, Nusbaumer et al.
compared the regeneration of ruthenium-based sensitizers and
a cobalt-based redox mediator, [Co(dbbip)2](ClO4)2.5 This redox
couple was found to regenerate N719 and the hydrophobic dye
Z907 on the microsecond time scale with high efficiency
(ꢁreg(N719) ) 0.98 and ꢁreg(Z907) ) 0.91), suggesting that
neutral, sterically hindered dyes are regenerated more slowly
than smaller, negatively charged dyes. Cazzanti et al. reported
the lifetime of Z907 in the presence of [Co(dtb)3]3+/2+ to be
about 390 ns.44 This was substantially slower than regeneration
of the same dye by ferrocene, which does not suffer from the
large reorganization energy requirements that apply to cobalt
redox couples. By comparison, triiodide/iodide regenerates N719
and Z907 in the microsecond range.45,46
The regeneration of the dyes by the cobalt complexes was
investigated with photoinduced absorption spectroscopy, performed
under conditions similar to the operational conditions of DSCs (ca.
100 W m-2). The PIA spectra of D29- and D35-sensitized TiO2
in the presence and absence of the three different cobalt bipyridyl
redox mediators or iodide/triiodide are shown in Figure 6a and
6b, respectively. In the absence of a redox mediator, a bleach was
found at 520 nm for D35 and 540 nm for D29, respectively, and
an increase in absorbance in the near infrared. The latter can be
attributed to the presence of oxidized dye and to a lesser extent
electrons in TiO2 (in agreement with the results from measurements
by spectroelectrochemistry shown in Figure S8 in the Supporting
Information). The bleach is attributed to superposition of the bleach
corresponding to the ground-state dye on the formation of the
oxidized dye together with a Stark shift caused by the photoinjected
electrons that give a local change in the electrical field across the
dye molecules.47,48 When the electrolyte was added to the sensitized
film, the features from the oxidized dye in the near infrared
disappeared, indicating that the redox mediators had regenerated
the dye, and a broad absorption due to accumulated electrons in
the TiO2 was seen. A bleach at about 520 nm persisted, caused by
The diffusion coefficients of the cobalt bipyridyl redox mediators
were determined from the diffusion-limiting current measured by
slow scan cyclic voltammetry to 5.2 × 10-6 cm2 s-1 for
[Co(dtb)3]2+, 7.7 × 10-6 cm2 s-1 for [Co(dmb)3]2+, and 9.1 × 10-6
cm2 s-1 for [Co(bpy)3]2+ in acetonitrile. The diffusion-limiting
current for [Co(bpy)3]2+ is shown in Figure S6 in the Supporting
Information. The diffusion is somewhat slower for the cobalt
bipyridyl redox mediators than for triiodide, which has a diffu-
sion coefficient of 18 × 10-6 cm2 s-1.19 The diffusion of [Co-
(bpy)3]3+/2+ seems, however, to be fast enough in a 6 µm thick
DSC, as the current spike in the photocurrent transient was
negligible. The short-circuit current is also shown in Figure 4b to
be linear with light intensity for [Co(bpy)3]3+/2+ but deviates from
linearity at higher light intensity for [Co(dmb)3]3+/2+ and [Co-
(dtb)3]3+/2+ due to slower mass transport of the bigger cobalt
complexes. The high efficiency for the DSCs sensitized with
D35 and using [Co(bpy)3]3+/2+ can, therefore, be explained by
the efficient blocking of the recombination by the bulky alkoxy
substituents on the dye. By adding steric bulk to the dye, rather
than the cobalt complex, mass transport problems are avoided.
In the cyclic voltammetry experiments, the electrochemical
reaction at the platinum electrode was increasingly less reversible
for the cobalt complexes in the following order [Co(bpy)3]3+/2+
<
[Co(dmb)3]3+/2+< [Co(dtb)3]3+/2+. Cobalt complexes containing
alkyl substituents in the 4 and 4′ or equivalent positions have been
shown to be surface sensitive with respect to the roughness and
material of the electrode, and quasi-irreversible behavior on
platinized electrodes has been reported elsewhere.9 The charge-
transfer resistance at the platinized counter electrode for the different
cobalt bipyridyl complexes was compared using impedance spec-
troscopy (see Figure S7 in the Supporting Information). The lowest
charge-transfer resistance at the counter electrode was measured
for [Co(bpy)3]3+/2+ and demonstrates the advantage of employing
the less bulky redox mediator in DSCs. Initial results comparing
alternative materials as counter electrodes are promising and suggest
that alternative, cheap counter electrode materials will become
available.
(43) Nakade, S.; Makimoto, Y.; Kubo, W.; Kitamura, T.; Wada, Y.;
Yanagida, S. J. Phys. Chem. B 2005, 109, 3488.
(44) Cazzanti, S.; Caramori, S.; Argazzi, R.; Elliott, C. M.; Bignozzi, C. A.
J. Am. Chem. Soc. 2006, 128, 9996.
(45) Clifford, J. N.; Palomares, E.; Nazeeruddin, M. K.; Gra¨tzel, M.;
Durrant, J. R. J. Phys. Chem. C 2007, 111, 6561.
Effect of the Steric Properties of the Dye and Redox
Shuttle on the Dye Regeneration Reaction. As for the recom-
bination reaction, the rate of electron transfer between the dye
and the redox couple is expected to depend on the structure of
(46) Pelet, S.; Moser, J.-E.; Gratzel, M. J. Phys. Chem. B 2000, 104, 1791.
(47) Ardo, S.; Sun, Y.; Staniszewski, A.; Castellano, F. N.; Meyer, G. J.
J. Am. Chem. Soc. 2010, 132, 6696.
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16721

A R T I C L E S
Feldt et al.
Figure 6. PIA spectra of TiO2 sensitized with (a) D29 and (b) D35. The electrolyte was 0.1 M Co(L)3(ClO4)2, 0.01 M Co(L)3(ClO4)3, 0.1 M LiClO4, and
0.2 M TBP in acetonitrile. Every 10th data point is marked with a symbol.
Figure 7. (a) Transport time as a function of short-circuit current density and (b) electron lifetime as a function of open-circuit voltage for [Co-
(bpy)3]2+/3+-based DSCs sensitized with D35 using a different thickness of TiO2.
the Stark effect. The electroabsorption spectra of D35 and D29
have a shape similar to the first derivative of the dye absorption
spectrum (shown in Figure S9 in the Supporting Information) and
demonstrate that the observed bleaches are caused by a change in
the absorbance spectra of the dye by a local electric field.47,48 The
absence of any trace of oxidized dye in the steady-state spectra in
the presence of electrolyte suggests that all the dye has been
efficiently regenerated for all the redox couples investigated.
Co(II) in the pores of TiO2. This can inhibit efficient dye
regeneration and increase undesirable recombination processes.
Effect of Film Thickness. The results discussed above refer to
DSCs with a mesoporous TiO2 film thickness of 5.6 µm. Figure 7
shows the effect of varying film thickness on the electron trans-
port and lifetime in DSCs sensitized with D35 using a [Co-
(bpy)3]3+/2+-based electrolyte. The electron lifetime decreased and
the transport time increased, i.e., the electron transport became
slower, with increasing thickness of the TiO2. Electron transport
is suggested to be driven by diffusion and has been shown
elsewhere to require a longer time in DSCs with thicker films.49
The charge-collection efficiency of photoinjected electrons is higher
in DSCs with thinner films since the distance required for electron
diffusion to the charge collector is smaller, resulting in shorter
transport times and lower electron densities in the film, which limits
recombination losses. The decrease in electron lifetime, measured
at open-circuit conditions, with film thickness may be explained
as follows: At a given VOC, the electron concentration in the TiO2
is constant, even though the strong light absorption by the organic
dyes leads to charge injection mostly close to the substrate.
Regeneration of the dye leads to generation of Co(III) species close
to the substrate, from where it diffuses toward the bulk electrolyte.
Under steady-state conditions, the rather slow diffusion of the cobalt
complexes in the pores of the TiO2 leads to a local increase of
The results of the PIA measurements and the high IPCE
values for the DSCs sensitized with D35 with each redox
mediator, despite structural differences, suggest that the regen-
eration efficiency is high under the conditions used in the
experiments. The photocurrent in our devices was, furthermore,
not significantly affected by the redox potential of the cobalt
complexes, in agreement with observations by DeVries et al.,21
suggesting that the driving force for regeneration is sufficient
for the cobalt-mediator-based DSCs employed here. The lower
IPCE value for DSCs sensitized with D29 was, therefore, not
attributed to a lower driving force for regeneration but rather
to fast recombination kinetics as discussed in the electron
lifetime section above. However, in the PIA measurements we
used thinner films (about 2 µm) than those applied in real
devices. This means that we cannot rule out that dyes may be
located deep in the pores in very thick films, which may not be
accessible to larger redox species. Slow mass transport of the
cobalt mediator can lead to depletion of Co(III) at the counter
electrode but also to accumulation of Co(III) and a decrease of
(48) Cappel, U. B.; Feldt, S. M.; Scho¨neboom, J.; Hagfeldt, A.; Boschloo,
G. J. Am. Chem. Soc. 2010, 132, 9096.
9
16722 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010

High-Efficiency Dye-Sensitized Solar Cells
A R T I C L E S
Figure 8. Plots of current transients measured at an illumination intensity
of 1000 W m-2 for [Co(bpy)3]2+/3+-based DSCs sensitized with D35 using
a different thickness of TiO2.
their concentration, which will depend on the film thickness. The
result will be a smaller electron lifetime for thicker TiO2 films.
Figure 8 shows the effect of varying the TiO2 film thickness on
the photocurrent transients for the DSCs sensitized with D35,
employing [Co(bpy)3]3+/2+ in the electrolyte. The photocurrent
transients were normalized with respect to the steady-state current
in order to illustrate the increase in the current spike with increasing
film thickness as the light is switched on. In DSCs with thicker
TiO2 films, the current is clearly limited by slow mass transport
through the mesoporous TiO2 film. The cell performance was not
substantially affected by varying the thickness of the Surlyn spacer,
suggesting that it is the diffusion through the mesoporous TiO2
network rather than in the bulk electrolyte that limits the perfor-
mance of DSCs employing cobalt polypyridine redox mediators.
The diffusion of [Co(dtb)3]n+ through mesoporous titania films was
determined by Nelson et al.19 to be about 1 order of magnitude
slower than for I3-. Organic dyes with high extinction coefficients
can, however, be used to build thin-film DSCs in order to overcome
mass transport and recombination limitations related to cobalt-based
redox mediators.
Figure 9. Current density versus applied potential curves under (a) 1000
W m-2 AM1.5G and (b) 250 Lux illumination for DSCs sensitized with
D35. The electrolytes were 0.5 M Co(bpy)3(PF6)2, 0.1 M Co(bpy)3(PF6)3,
0.5 M TBP, and 0.1 M LiClO4 in acetonitrile or 0.6 M 1-butyl-3-
methylimidazoliumiodide, 0.03 M I2, 0.5 M TBP, and 0.1 M guanidinium
thiocyanate in acetonitrile, respectively.
Table 1. Current-Voltage Characteristics of DSCs Sensitized with
D35 Using [Co(bpy)3]2+/3+ and Iodide/Triiodide-Based Electrolytes
Measured Using a 6 mm × 6 mm Mask under Different
Illumination Conditions
Optimized Photovoltaic Performance. The current-voltage
characteristics were optimized for the DSCs sensitized with D35
employing the [Co(bpy)3]3+/2+ or the iodide/triiodide electrolyte,
and the best efficiencies obtained are shown in Figure 9 and Table
1. Although the mass transport problems were shown to be avoided
in the thin-film DSCs employing [Co(bpy)3]3+/2+, the best efficiency
was still obtained for thicker films because of the increased light-
harvesting efficiency, i.e., two layers of TiO2 and one light-
scattering layer giving a total thickness of 15 µm (12 + 3 µm).
The best efficiencies obtained in this study measured at 1000 W
m-2 AM1.5G illumination are shown in Figure 9a and were 6.7%
for [Co(bpy)3]3+/2+ and 5.5% for iodide/triiodide. The efficiency
reported above (and for all the devices mentioned so far) was
measured using a 6 mm ×6 mm mask, following the recom-
mendation by Ito et al.50 Since certified solar cell efficiencies for
DSCs are quoted for a defined aperture size,51 we also measured
the most efficient cells with a mask that was slightly smaller than
the film area (4 mm × 4 mm). In this case, the efficiency found
was 6.3% for [Co(bpy)3]3+/2+ and 5.4% for iodide/triiodide.
The efficiency of DSCs sensitized with D35 employing cobalt
complexes is superior to that obtained by other groups working
max power
(µW cm-2
)
light intensity
Voc(V)
Jsc (mA cm-2
)
FF
η (%)
Co(bpy)3(PF6)2+/3+
a
1000 W m-2
0.92
0.85
0.70
10.7
1.12
0.68
0.76
0.80
6.7
7.2
a
100 W m-2
250 lxb
18.5 × 10-3
10.7
iodide/triiodide
a
1000 W m-2
0.91
0.86
0.70
9.38
0.65
0.75
0.79
5.5
5.7
a
100 W m-2
0.89
250 lxb
12.6 × 10-3
7.2
a AM1.5G spectral distribution. b Fluorescent light (Osram Dulux
Value 8W/827).
with cobalt-based DSCs using ruthenium-based sensitizers.
Control devices with the ruthenium-based sensitizers Z907 and
N719 were performed, see Figure S10 in the Supporting
Information. The photovoltaic performance of Z907 in combina-
tion with the coadsorber 1-decylphosphonic acid (DPA) was
much better than that of N719 and illustrates the importance of
introducing alkoxyl groups on the dye that prevents the TiO2
surface from Co(III) species in the electrolyte in cobalt-based
DSCs. The structural properties of D35 seem, however, to
prevent recombination more efficiently than Z907 in cobalt-
based DSCs and a higher efficiency and Voc was obtained with
this organic dye. It was further noted that the efficiency increased
when the solar cells were warmed under the solar simulator
(49) Nissfolk, J.; Fredin, K.; Simiyu, J.; Ha¨ggman, L.; Hagfeldt, A.;
Boschloo, G. J. Electroanal. Chem. 2010, 646, 91.
(50) Ito, S.; Nazeeruddin, M. K.; Liska, P.; Comte, P.; Charvet, R.; Pe´chy,
P.; Jirousek, M.; Kay, A.; Zakeeruddin, S. M.; Gra¨tzel, M. Prog.
PhotoVoltaics Res. Appl. 2006, 14, 589.
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16723

A R T I C L E S
Feldt et al.
since the diffusion of the cobalt complexes is faster at higher
temperatures. Liberatore et al.52 found that the standard rate
electron coefficient for [Co(dtb)3]n+ in methoxypropionitrile
increased more than six times and the ionic diffusion coefficient
more than three times at 80 °C.
thicker film of TiO2 due to enhanced light-harvesting efficiency,
and it is therefore a need of organic dyes with even higher
extinction coefficients and more red-shifted absorption spectra.
One advantage of using one-electron redox couples in DSCs is
that the driving force for regeneration is expected to be lower
than that for the iodide/triiodide redox shuttle, and the absorption
of the dye can therefore be broadened by tuning the HOMO
level of the dye. The current was, in addition, higher in the
DSCs employing cobalt redox mediators since the cobalt
complexes absorb less light than iodide/triiodide.
The Voc and electron lifetime are approximately the same for
the optimized DSCs employing iodide/triiodide or cobalt-based
redox mediators when a large amount of TBP is present in the
electrolyte (see Figure S12 in the Supporting Information). The
Voc and solar cell efficiency increase much upon addition of
TBP to both types of redox electrolyte. For iodide/triiodide-
based electrolyte this has been explained by a negative shift in
the conduction band as a result of a decreased amount of
adsorbed Li cations at the TiO2 surface as well as an increase
of electron lifetime.43,53,51 A similar mechanism is expected for
the cobalt-based redox mediators. In agreement with findings
from Nakade et al.,43,53 we found that TBP increases the electron
lifetime in cobalt mediator-based DSCs.
The DSCs employing cobalt redox mediators are also promising
for indoor applications as the efficiency and the open-circuit voltage
remain high even at low light intensities. The photovoltage for
DSCs sensitized with D35 employing [Co(bpy)3]3+/2+ at an indoor
light illumination of 250 lx was about 700 mV, as shown in the
current-voltage characteristics in Figure 9b. For comparison,
Burnside et al.54 reported 539 mV per cell at 250 lx for ruthenium-
based sensitizer (N719) based DSC modules employing the iodide/
triiodide redox couple, while a commercial amorphous silicon
module gave 677 mV/element. At 100 W m-2 AM1.5G illumina-
tion we obtained 7.8% efficiency for DSCs with cobalt redox
mediators using a TiO2 film thickness of 22 µm and a lower
concentration of cobalt mediator in the electrolyte (see Figure S11
in the Supporting Information).
Cobalt polypyridine complexes are promising alternative redox
mediators to the iodide/triiodide redox couple for large-scale
manufacturing of DSCs, since they are less aggressive toward metal
contacts and sealing materials. DSCs sensitized with organic dyes
employing cobalt polypyridine redox mediators are promising for
indoor applications since the voltage and efficiency remain high
under low light intensities. Mass transport limitations are also
efficiently avoided at lower light intensities, and the stability of
the devices can be extended by using a more viscous electrolyte
solvent. The Voc of the devices is high (>900 mV) because of the
more positive redox potential of the complexes and can easily be
tuned, as the large amount of commercially available ligands makes
it possible to adjust the redox potential by changing the coordination
sphere of the complexes.
Most importantly, we demonstrated that one-electron outer-
sphere complexes may be employed in n-type DSCs without the
need for surface-insulating layers. By choosing an appropriate dye,
the recombination, which is otherwise fast with organometallic
complexes, can be prevented and the lifetime of the electrons can
be increased to time scales comparable with the standard iodide/
triiodide electrolyte. This in turn removes an extra step in the
preparation of the working electrodes, which would increase the
manufacturing cost if such devices were to be produced commercially.
Conclusion
Efficient dye-sensitized solar cells using cobalt-based media-
tors can be obtained by matching the steric bulk of tripheny-
lamine-based organic sensitizers and cobalt polypyridine redox
mediators. Careful tuning of their properties minimizes the
recombination between electrons in TiO2 and Co(III) species
in the electrolyte solution and avoids the mass transport
limitations of the redox mediator. The organic sensitizer D35,
equipped with bulky alkoxyl groups, efficiently suppresses
recombination, allowing the use of cobalt redox mediators with
smaller steric bulk. By building thin-film DSCs sensitized with
dyes with high extinction coefficients, previously encountered
problems associated with cobalt redox mediators, such as fast
recombination and mass transport limitations, can efficiently be
avoided. The best efficiency for DSCs, sensitized with D35 and
employing a [Co(bpy)3]3+/2+-based electrolyte of 6.7% at full
sunlight (1000 W m-2 AM1.5G illumination) is superior to the
highest efficiency obtained so far with iodide-free electrolytes.
The best efficiency was, however, achieved for DSCs with a
Acknowledgment. The Swedish Energy Agency, the Swedish
Research Council (VR) and the Knut and Alice Wallenberg
Foundation are acknowledged for financial support.
Supporting Information Available: IPCE spectra from CE
illumination for D29 and D35; absorption spectra of D29 and
D35; dark current-voltage characteristics of D29 and D35; light
intensity dependence of the open-circuit voltage for D35-based
DSCs using a supporting electrolyte of TBAPF6 or LiClO4;
current-voltage characteristics of D35-based DSCs with a
different amount of oxidized [Co(bpy)3]3+ in the electrolyte;
diffusion-limiting current for [Co(bpy)3]2+/3+ determined by
cyclic voltammetry; impedance spectra of the different cobalt
bipyridyl redox mediators employed herein; spectroelectro-
chemistry measurements of D29 and D35; electroabsorption
spectra of D29 and D35; current-voltage characteristics com-
paring D35 and ruthenium sensitizers in cobalt-based DSCs;
current-voltage characteristics for D35-based DSCs employing
[Co(bpy)3]2+/3+ at 100 W m-2 AM1.5G illumination; electron
lifetime as a function of extracted charge plot for the optimized
DSCs sensitized with D35 employing [Co(bpy)3]2+/3+- or iodide/
triiodide-based electrolytes. This material is available free of
charge via the Internet at http://pubs.acs.org.
(51) Green, M. A.; Emery, K.; Hishikawa, Y.; Warta, W. Prog. Photo-
Voltaics Res. Appl. 2010, 18, 144.
(52) Liberatore, M.; Petrocco, A.; Caprioli, F.; La Mesa, C.; Decker, F.;
Bignozzi, C. A. Electrochim. Acta 2010, 55, 4025.
(53) Nakade, S.; Kanzaki, T.; Kubo, W.; Kitamura, T.; Wada, Y.; Yanagida,
S. J. Phys. Chem. B 2005, 109, 3480.
(54) Burnside, S.; Winkel, S.; Brooks, K.; Shklover, V.; Gra¨tzel, M.; Hinsch,
A.; Kinderman, R.; Bradbury, C.; Hagfeldt, A.; Pettersson, H. J. Mater.
Sci.-Mater. Electron. 2000, 11, 355.
JA1088869
9
16724 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010