Large increases in photocurrents and solar conversion efficiencies by UV illumination of dye sensitized solar cells

Article abstract of DOI:10.1021/jp011612o

Ultraviolet (UV) treatment of dye sensitized solar cells (DSSCs) containing tetra-n-butylammonium iodide electrolyte increases photocurrents dramatically. The effect remains after cessation of UV illumination. Depending upon the photosensitizing dye, the increase in photocurrent can be as much as 2 orders of magnitude. The photocurrent increase more than compensates for slight decreases in photovoltage and fill factor, resulting in overall solar conversion efficiency increases up to 45 ?? for some dyes. The primary effect of the UV treatment appears to be a positive shift in the conduction band of the nanocrystalline titanium dioxide, which promotes electron injection from the dye. The dyes and the solar cells are both found to be stable to this treatment. This effect offers the ability to tune the properties of the semiconductor to match the requirements of a specific dye, thus providing a versatile analytical tool for characterizing DSSCs while also enabling the use of new classes of sensitizing dyes. This letter describes the UV effect and summarizes the results of its application to a number of perylene-based sensitizers and two ruthenium bipyridyl sensitizers.

Full text of DOI:10.1021/jp011612o

7602
J. Phys. Chem. B 2001, 105, 7602-7605
Large Increases in Photocurrents and Solar Conversion Efficiencies by UV Illumination of
Dye Sensitized Solar Cells
Suzanne Ferrere* and Brian A. Gregg*
Basic Sciences Center, National Renewable Energy Laboratory, 1617 Cole BouleVard, Golden, Colorado 80401
ReceiVed: April 27, 2001; In Final Form: June 17, 2001
Ultraviolet (UV) treatment of dye sensitized solar cells (DSSCs) containing tetra-n-butylammonium iodide
electrolyte increases photocurrents dramatically. The effect remains after cessation of UV illumination.
Depending upon the photosensitizing dye, the increase in photocurrent can be as much as 2 orders of magnitude.
The photocurrent increase more than compensates for slight decreases in photovoltage and fill factor, resulting
in overall solar conversion efficiency increases up to 45× for some dyes. The primary effect of the UV
treatment appears to be a positive shift in the conduction band of the nanocrystalline titanium dioxide, which
promotes electron injection from the dye. The dyes and the solar cells are both found to be stable to this
treatment. This effect offers the ability to tune the properties of the semiconductor to match the requirements
of a specific dye, thus providing a versatile analytical tool for characterizing DSSCs while also enabling the
use of new classes of sensitizing dyes. This letter describes the UV effect and summarizes the results of its
application to a number of perylene-based sensitizers and two ruthenium bipyridyl sensitizers.
Introduction
We present here a method by which the solar conversion
efficiency of DSSC devices can be markedly improved by
exposure to ultraviolet (UV) light. The effect is most dramatic
for cells containing photosensitizers with very poor initial
quantum yields. For the perylene sensitizers included here, the
short circuit photocurrent improvements range from 3- to more
than 100-fold. Accompanied by some loss in photovoltage and
fill factor, the improvements in overall conversion efficiencies
range from 2-fold to 45-fold. The UV exposure appears to shift
the conduction band edge positively, which increases the yield
of photoinjection. Although there are previous reports of band
edge shift brought about by UV illumination,10,11 this is the first
demonstration that it can markedly improve device performance.
However, the band edge shift cannot explain all of our
observations, which suggest that other effects, such as the
formation of a surface dipole which opposes recombination and
promotes injection, also contribute to the improved device
performance.
A recent analysis found both the solar conversion efficiency
and production cost of the dye sensitized solar cell (DSSC) to
be competitive with amorphous silicon.1 In theory, there are
many possibilities for dye sensitized configurations, that is, dyes,
semiconductors, and electrolytes could be mixed and matched
according to individual properties such as energetics, chemical
structure or functionality, and kinetic behavior. In reality, only
a very specific combination (titanium dioxide (TiO2) semi-
conductor, a ruthenium based photosensitizing dye (N3), and
the I-/I3 electrolyte) gives an efficiency that is competitive
-
with traditional photovoltaics.2,3 The excellent performance of
this system relies on a well-tuned interplay between components.
The ruthenium based dyes are excellent energetic matches to
the TiO2 conduction band and bind to the semiconductor surface
with strong orbital overlap, resulting in a nearly quantitative
charge injection yield.4 Similarly, iodide (I-) has favorable
energetics and kinetics for regeneration of the ruthenium dye,
whereas triiodide (I3-) displays uniquely slow kinetics for
recombination with conduction band electrons at the TiO2
interface but fast kinetics for reduction at the platinum counter
electrode.5,6 Thus, each component has a very specialized fit
within the system.
Experimental Section
The nanocrystalline titanium dioxide films were prepared as
previously reported from TiO2 colloids made with acetic acid.12
Film thicknesses were measured on an Alpha Step Profiler
(Tencor) and were in the range 6-7 µm. Sandwich-like solar
cells were assembled from a dyed TiO2 film and a platinized
conducting SnO2 counter electrode as previously described.2 A
3-methoxyproprionitrile solution containing 0.5 M tetra-n-
butylammonium iodide (TBAI), 0.05 M iodine (I2), and 0.2 M
4-tert-butylpyridine was used as electrolyte. For UV treatments,
assembled cells were illuminated with the entire spectrum of a
Xenon arc lamp (PTI, model A 1010) which was passed through
a water filter. The intensity was approximately 190 mW/cm2,
and current-time traces were recorded using a program written
in LabView. Current-voltage (J-V) measurements were made
at white light intensity approximating AM 1.5 (100 mW/cm2)
solar emission for the wavelength region between 400 and 800
nm, obtained by filtering and adjusting the same Xenon arc
Unfortunately, there are currently many more dyes than there
are semiconductors to pair them with in a dye sensitized
configuration. Nanocrystalline TiO2 remains the most commonly
employed semiconductor because its conduction band energetics
are in a region accessible to many types of dyes, and it has
useful surface chemistry and excellent materials properties.
However, many of the dyes heretofore investigated on TiO2
show much lower efficiencies than N3.4,7-9 The poorer
performance of other dyes can often be related to insufficient
energetic matching of their excited states with the TiO2
conduction band.8,9 Thus, a strategy by which the semiconductor
properties could be adjusted in order to improve the semi-
conductor’s match to the sensitizing dye would be beneficial.
10.1021/jp011612o CCC: $20.00 © 2001 American Chemical Society
Published on Web 07/19/2001

Letters
J. Phys. Chem. B, Vol. 105, No. 32, 2001 7603
TABLE 1: Values of Cell Parameters Measured before and
after UV Treatmenta
Figure 1. Current versus time of UV illumination for nanocrystalline
TiO2 cells at short circuit in the presence of 0.5 M TBAI, 0.05 M I2,
and 0.2 M 4-tert-butylpyridine in 3-methoxyproprionitrile with a
platinized SnO2 counter electrode. Bare TiO2 is shown in the lower
curve and TiO2 sensitized by [Ru(2,2′-bipyridine)(4,4′-dicarboxy-2,2′-
bipyridine)] is shown in the upper curve.
a The results are averages of two dye-sensitized TiO2 cells prepared
from each sensitizer. Same conditions as for Figure 1.
source used for UV illumination. The data were collected and
solar conversion efficiencies were calculated using a program
written in LabView. The dye [Ru(4,4′-dicarboxy-2,2′-bipyridine)2-
(NCS)2] (“N3”) was purchased from Solaronix and was adsorbed
to the TiO2 films out of ethanol. The dye [Ru(2,2′-bipyridine)2-
(4,4′-dicarboxy-2,2′-bipyridine)][PF6]2 was prepared analogously
to previously reported RuL2L complexes13 and was adsorbed
out of acetonitrile. Perylene-3,4-dicarboxylic acid anhydride-
N-(dodecyl)-9,10-carboximide (P4 in Table 1) was prepared by
the reaction of monopotassium 3,4:9,10-perylene tetracarbox-
ylate with dodecylamine.14 It was adsorbed to TiO2 out of
1-methyl-2-pyrrolidinone. The dye 9-dioctylamino-perylene-3,4-
dicarboxylic acid anhydride (P2 in Table 1) was prepared by
first reacting 9-bromo-N-(2,5-di-tert-butylphenyl)perylene-3,4-
dicarboximide15 with dioctylamine to yield 9-dioctylamino-N-
(2,5-di-tert-butylphenyl)perylene-3,4-dicarboximide; the tert-
butylphenyl group was then removed by treatment in base to
yield the product.15 The dye 9-dioctylamino-N-(acetic acid)-
perylene-3,4-dicarboximide (P3 in Table 1) was prepared by
the reaction of 9-dioctylamino-perylene-3,4-dicarboxylic acid
anhydride with glycine. The dye 9-(dimethylamino)-perylene-
3,4-dicarboxylic acid anhydride (P1 in Table 1) was obtained
by hydrolyzing the tert-butylphenyl group from 9-(dimethyl-
amino)-N-(2,5-di-tert-butylphenyl)perylene-3,4-dicarboximide,15
which had been obtained as a byproduct in the reaction of
9-bromo-N-(2,5-di-tert-butylphenyl)perylene-3,4-dicarbox-
imide with diethyl iminodiacetate. Finally, 9-aminomethyl car-
boxylic acid-N-(2,5-di-tert-butylphenyl)perylene-3,4-dicarbox-
imide (P5 in Table 1) was prepared by the reaction of 9-amino-
N-(2,5-di-tert-butylphenyl)perylene-3,4-dicarboximide15 with
bromoacetic acid. The preceding four perylene derivatives were
adsorbed to TiO2 out of chloroform. All of the perylene dyes
are novel sensitizers, and full synthetic details and sensitization
behavior will be provided in a future publication.
Figure 2. Current density-voltage measurement under visible il-
lumination for a TiO2 cell sensitized by 9-dioctylamino-N-(acetic acid)-
perylene-3,4-dicarboximide (P3), before (lower curve) and after (upper
curve) UV treatment. Same conditions as for Figure 1. η is the solar
conversion efficiency.
Results and Discussion
When a dye sensitized nanocrystalline TiO2 solar cell
containing TBAI and I2 electrolyte is illuminated at short circuit
with white light between 350 and 800 nm, the photocurrent
increases steadily over a period of approximately 20 min (Figure
1). Figure 1 also indicates the effect for an unsensitized TiO2
cell (lower trace). If the UV portion (λ < 400 nm) of the light
is blocked during this illumination, the photocurrent is constant
and stable. When current-voltage measurements (under visible
illumination only) are made after UV treatment, there are
significant improvements in short circuit photocurrent (Jsc),
slight decreases in open circuit photovoltage (Voc), and decreases
in the fill factor, as compared to initial values of these
parameters. Figure 2 shows representative before and after
current-voltage measurements for a film sensitized with
9-dioctylamino-N-(acetic acid)perylene-3,4-dicarboximide (P3
in Table 1). The solar conversion efficiency of the device
increases by 25× with UV treatment. The accompanying
incident photon-to-current efficiency (IPCE) curves are shown
in Figure 3 for the same sensitizer. The change in quantum yield
is remarkable: initially the IPCE is less than 1% and after

7604 J. Phys. Chem. B, Vol. 105, No. 32, 2001
Letters
that under UV illumination the flatband potential of a nano-
porous TiO2 film shifted positively.10,11 Hagfeldt and co-workers
attributed the effect to an unpinning of the conduction band
brought about by the accumulation of holes trapped in surface
states. Furthermore, they confirmed a positive band shift by
demonstrating that the photocurrent response for a dye with a
very positive excited-state potential could be increased by
background UV illumination of the cell.11
Light harvesting by the dye, excited-state electron injection,
electron transport through the film, dye regeneration by I-, and
recombination of injected electrons all contribute to the photo-
current in a DSSC; increases in Jsc must be related to changes
in one or more of these processes. The dye spectra are stable
and thus there are no changes in the light absorption capabilities
of the dyes. If the photocurrent were initially limited by a
collection or transport problem in the TiO2, this would have
been a common problem among all of the devices; yet some
devices work quite well to start. Finally, none of the dyes should
be energetically limited by regeneration from I-, and if they
were, the initial trends among them would be different. If there
were an inherent kinetic barrier to regeneration of the perylenes
by I-, it too would be a common limitation. Thus, there is no
evidence to suggest that the regeneration process is altered by
the UV treatment. The recombination process with I3- is already
so slow that it is not current limiting in DSSCs,5 so any UV-
induced change in this rate could not increase the current. The
most likely explanation for our results is that excited-state
electron injection is improved, although a decrease in recom-
bination rate between the injected electron and the oxidized dye
may also play a role. A positive shift in the conduction band of
TiO2 is expected to increase the injection rate of all dyes,
because of an increased driving force for injection and a higher
density of conduction band states accessible to the dye excited
state. The effect should be most pronounced for dyes that
originally have little or no driving force for injection.7 This
explanation is supported by the trend that dyes with more
positive excited states (P3 and P4) show the greatest relative
increases in Jsc.
Figure 3. Incident photon-to-current efficiency versus wavelength, at
short circuit, for the same device measured in Figure 2.
treatment it increases to more than 25%. Table 1 summarizes
before and after Jsc, Voc, and solar conversion efficiencies (η)
for a variety of sensitizing dyes. In each case, the improvement
in photocurrent significantly outweighs the decreases in other
parameters, resulting in overall solar conversion efficiency
improvements ranging from 13% (for N3) to 4400% (for P4).
The trend among the dyes is that the lower the photocurrent
and IPCE are to begin with, the greater the improvement under
UV illumination. All of the dyes listed in Table 1 were stable
to UV treatments in the cells. Many batches of TiO2 films were
investigated, and the effect was present in all samples.
When a UV treated cell remains assembled, but all illumina-
tion is discontinued, the cell can maintain most of the improved
performance for hours (provided that it does not dry out). If a
cell is disassembled, rinsed, and reassembled, most of its
improved performance is lost; however, the increased efficiency
can be fully regenerated by another UV treatment. The UV effect
depends on electrolyte conditions. If the 4-tert-butylpyridine
additive is eliminated, the Jsc increase is greater but the Voc
loss is also greater, resulting in a smaller increase in overall
efficiency. Cells containing KI exhibit changes in photocurrent
similar to TBAI-containing cells during UV illumination,
whereas cells containing LiI exhibit only slight increases (5-
10%) in current. As a measure of possible band edge motion
during the evolution of the UV effect, J-V curves were recorded
(with 400-800 nm illumination) under conditions where Jsc was
kept constant by adjusting the incident light intensity. Voc
decreased monotonically with UV illumination time, indicating
a positive shift in the conduction band edge.16
The presence of the effect on bare TiO2 and the stability of
the dyes indicate that the UV light is acting upon the TiO2 and
not the adsorbed dye. Titanium dioxide’s activity under band
gap (UV) excitation, where electrons and highly oxidizing holes
are generated, is an active area of photoelectrochemistry and
photocatalysis.17 The presence of holes must be essential to the
effect we observe, since it is the only difference between UV
illumination and normal operation (visible illumination) of a
DSSC, where only electrons are present in the TiO2. Others
have reported UV-induced changes in dye sensitized solar cells
and presented hole-driven mechanisms to explain them.11,18
O’Regan and Schwartz found an efficiency enhancement after
UV illumination of a solid-state DSSC which employed CuSCN
as hole conductor. They reasoned that the TiO2 valence band
holes oxidized (SCN)-, leaving behind a polymeric thiocyanate
species with better kinetics than CuSCN for regeneration of the
oxidized dye.18 Both Hagfeldt et al. and Zaban et al. observed
The positive conduction band shift is most likely due to a
buildup of positive charge in the TiO2 particles. It could arise
from surface adsorption of positively charged speciess
electrolyte components that have been oxidized by valence band
holes, for example. If any species does adsorb to the surface, it
is not strongly bound, as even cursory rinsing after the UV
treatment returns the cell to its initial level of performance. The
possibility that adsorbed oxygen or water is mediating such a
surface reaction was discounted, since the UV effect also occurs
when oxygen and water are excluded from the cells by using
rigorously dried materials and making all measurements in a
glovebox. Alternatively, the particles may accumulate positive
charge from holes trapped in surface states, a mechanism
suggested by Hagfeldt and co-workers.11 A substantial amount
of charge can be stored in surface states of these films because
the nanocrystalline particles are dominated by their surface
properties. Clearly, the mechanism warrants further study.
It is important to note the almost complete absence of the
effect when Li+ is present. Lithium ions are known to both
adsorb to the surface and intercalate into TiO2 films, whereas a
large and bulky cation like TBA+ is not intercalated, nor can it
closely approach the dye-covered surface.19-21 Thus, Li+ ions
fix the band edges to more positive potentials than TBA+;10
this is manifest in the higher short circuit photocurrents and
lower open circuit photovoltages typically observed for Li+-
versus TBA+-containing cells. Here, the UV treatment effects

Letters
J. Phys. Chem. B, Vol. 105, No. 32, 2001 7605
(2) Nazeeruddin, M. K.; Kay, A.; Rodicio, I.; Humphry-Baker, R.;
Mu¨ller, E.; Liska, P.; Vlachopoulos, N.; Gra¨tzel, M. J. Am. Chem. Soc.
1993, 115, 6382-6390.
(3) Nazeeruddin, M. K.; Pe´chy, P.; Gra¨tzel, M. Chem. Commun. 1997,
1705-1706.
(4) Hagfeldt, A.; Gra¨tzel, M. Acc. Chem. Res. 2000, 33, 269-277.
(5) Gregg, B. A.; Pichot, F.; Ferrere, S.; Fields, C. L. J. Phys. Chem.
B 2001, 105, 1422-1429.
(6) Salafsky, J. S.; Lubberhuizen, W. H.; van Faassen, E.; Schropp,
R. E. I. J. Phys. Chem. B 1998, 102, 766-769.
(7) Kay, A.; Gra¨tzel, M. J. Phys. Chem. 1993, 97, 6272-6277.
(8) Gholamkhass, B.; Koike, K.; Negishi, N.; Hori, H.; Takeuchi, K.
Inorg. Chem. 2001, 40, 756-765.
this positive shift to a controllable extent, allowing it to be
optimized for each dye, and in many cases cells retain a higher
photovoltage than for a Li+-containing cell. Thus, cells prepared
with LiI using the sensitizers presented here always start out
with higher efficiencies than the TBAI-containing cells but, after
the UV treatment, the TBAI cells often have higher overall
efficiencies. The fact that some cells have both higher Voc and
higher Jsc after UV illumination than cells with the same dye in
LiI solution suggests that the positive band edge shift is not the
only factor leading to the increased efficiency. Other possible
factors include a decrease in recombination rate between injected
electrons and the oxidized dye, and the formation of a dipole
at the TiO2 solution interface that enhances injection and slows
recombination.
(9) Ferrere, S. Chem. Mater. 2000, 12, 1083-1089.
(10) Zaban, A.; Meier, A.; Gregg, B. A. J. Phys. Chem. B 1997, 101,
7985-7990.
Conclusion
(11) Hagfeldt, A.; Bjo¨rkste´n, U.; Gra¨tzel, M. J. Phys. Chem. 1996, 100,
8045-8048.
We have demonstrated that by illuminating a DSSC with UV
light, large increases in photocurrent and solar cell efficiencies
can occur. The primary mechanism appears to be a positive
movement of the conduction band, which promotes electron
injection. Thus, the UV treatment can be used to “adjust” the
band edges for a photosensitizing dye with a poor injection yield.
Further investigations are under way to understand this very
interesting phenomenon.
(12) Zaban, A.; Ferrere, S.; Sprague, J.; Gregg, B. A. J. Phys. Chem. B
1997, 101, 55-57.
(13) Zaban, A.; Ferrere, S.; Gregg, B. A. J. Phys. Chem. B 1998, 102,
452-460.
(14) Ferrere, S.; Zaban, A.; Gregg, B. A. J. Phys. Chem. B 1997, 101,
4490-4493.
(15) Feiler, L.; Langhals, H.; Polborn, K. Liebigs. Ann. 1995, 1229-
1244.
(16) Cahen, D.; Hodes, G.; Gra¨tzel, M.; Guillemoles, J. F.; Riess, I. J.
Phys. Chem. B 2000, 104, 2053-2059.
(17) Kung, H. K. Transition Metal Oxides: Surface Chemistry and
Catalysis; Elsevier: New York, 1989; Vol. 45.
Acknowledgment. We thank the Office of Science, Division
of Basic Energy Sciences, Chemical Sciences Division of the
U.S. Department of Energy and the Director’s Discretionary
Research and Development program at NREL for funding this
research.
(18) O’Regan, B.; Schwartz, D. T. Chem. Mater. 1998, 10, 1501-1509.
(19) Lunell, S.; Stashans, A.; Ojama¨e, L.; Lindstro¨m, H.; Hagfeldt, A.
J. Am. Chem. Soc. 1997, 119, 7374-7380.
(20) Pelet, S.; Moser, J.-E.; Gra¨tzel, M. J. Phys. Chem. B 2000, 104,
1791-1795.
References and Notes
(1) Schmela, M. In PHOTON International, 2000; Solar Verlag
GmbH: Germany; pp 14-16.
(21) Kelly, C. A.; Farzad, F.; Thompson, D. W.; Stipkala, J. M.; Meyer,
G. J. Langmuir 1999, 15, 7047-7054.