Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: A steady-state numerical model of dye-sensitized solar cell is based on continuity and transport equa-
Received 12 January 2011 tions for electrons, iodide and triiodide ions. The cell model consists of an active layer, where photo-
Received in revised form voltaic effect including diffusion of electrons in mesoporous TiO2 and ions in electrolyte takes place, and
7 June 2011
a bulk electrolyte layer, where only ions diffuse. Exponential distribution of trap states in TiO2 and
Accepted 13 June 2011
Gaussian distributions of energy levels in the electrolyte within active layer are included in modeling of
Available online 21 June 2011
the recombination dynamics, according to Shockley-Read-Hall statistics and Marcus-Gerischer electron
transfer theory. Recombinations at the front contact and a voltage drop at the platinum covered back
Keywords:
Dye-sensitized solar cell
contact are included in the model. Simulation results are compared with the measured currentevoltage
Electron recombinations characteristics at different light intensities. In particular, light intensity dependence of open circuit
Numerical model voltage is studied over 4 decades. Optimization of cell efficiency regarding active layer and electrolyte
layer thickness is carried out. Simulation results show that best efficiency is achieved when electrolyte
layer thickness is minimized as much as possible and that active layer thickness is traded off with respect
to recombination rates and/or diffusion limited current determined with the selection of the electrolyte.
Ó 2011 Elsevier B.V. All rights reserved.
1. Introduction reaction model [10] and applied it to anodic half cell of DSSC [11].
Chen et al. [12] modeled the effects of partial shading of DSSC on its
Dye-sensitized solar cell (DSSC) based on dyed mesoporous performance. Villanueva-cab et al. [13,14] presented models based
semiconductor surrounded with electrolyte or other hole con- on continuity equation for electrons and considered trapping
ducting media is a promising device for efficient, low-cost, stable effects as well as different recombination dynamics. Recently
and sustainable solar energy conversion. Current record efficiencies Barnes et al. developed a time dependant numerical model of
for laboratory scale DSSC are above 12% [1]. To further increase a complete DSSC [15]. In our previous work [16] we focused on
efficiency, the parameters that influence cell performance need to optical modeling of DSSC.
be optimized. Numerical simulations based on models that Our model includes both the active layer and the bulk electro-
describe physical and chemical processes that govern DSSC oper- lyte layer in the DSSC. Effects of transparent conductive oxide (TCO)
ation are useful in finding the optimal cell design. at the front contact and platinum at the back contact are considered
Södergren et al. [2] developed analytical model based on when currentevoltage characteristics are calculated. Recombina-
diffusion of electrons in TiO2. Papageorgiou et al. [3] included tions are assumed to be affected by an exponential distribution of
diffusion of electrolyte species and effects of porosity of meso- trap states in the band gap of TiO2. Shockley-Read-Hall (SRH)
porous film. Ferber et al. presented numerical models treating statistics [17] are applied to determine trapping and detrapping
electrolyte permeated dye covered TiO2 as quasi-homogeneous rates and Marcus-Gerischer (MG) theory [18e20] is considered
medium [4] as well as a 2D model with simplified TiO2 structure when determining electron transfer rate between the semi-
[5]. Effects of monoenergetic trap states in the band gap were conductor and ions in the electrolyte.
considered by Usami and Ozaki [6]. Nelson and Chandler [7] used
continuous-time random walk simulations to describe electron 2. Implemented models
transport. Lee et al. [8] developed closed-form model with detailed
description of interface reactions. Layer of bulk electrolyte was 2.1. Basic device operation and device structure
studied by Oda et al. [9]. Penny et al. developed detailed interfacial
First step of energy conversion in DSSC is incident light
absorption by a monolayer of dye molecules covering mesoporous
* Corresponding author. Tel.: þ386 1 4768 276; fax: þ386 1 4264 630. semiconductor, usually TiO2. After absorption dye molecules inject
E-mail address: miha.filipic@fe.uni-lj.si (M. Filipi
c). electrons from their excited states into TiO2 conduction band. In
1567-1739/$ e see front matter Ó 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cap.2011.06.011
M. Filipic et al. / Current Applied Physics 12 (2012) 238e246 239
nA;e ð0Þ
EF ð0Þ ¼ ECB þ kTln ; (9)
NCB
where NCB is the effective concentration of conduction band states
in TiO2, ECB is the edge of TiO2 conduction band, k is the Boltzmann
constant, and T is temperature. Using Eqs. (7e9) we can express
electron concentration at the front contact in terms of applied
voltage.
At the end of active layer (x ¼ D) electrons cannot leave the
mesoporous TiO2 and their current density is zero
Fig. 1. DSSC structure with active layer (0 < x D) and bulk electrolyte layer
(D < x L). jA;e ðDÞ ¼ 0: (10)
240 M. Filipic et al. / Current Applied Physics 12 (2012) 238e246
At the front contact some of the electrons that already left TiO2
recombine with triiodide ions through the TCO. Taking into account
reaction (1) the recombination current jTCO (see Section 2.5)
determines the value of triiodide and iodide current densities at the
front contact
3
jA;I ð0Þ ¼ j ; (11)
2 TCO
1
jA;I3 ð0Þ ¼ j : (12)
2 TCO
Densities of iodide and triiodide ions in the active layer are
dependent on porosity p of TiO2, and jump discontinuously at the
border of active layer and bulk electrolyte, where TiO2 does not
occupy any volume (i ¼ I, I3 )
1 1 izational energy.
nA;I3 þ nA;I dx þ nB;I3 þ nB;I dx
3 3
0 D
1 1 In order to obtain the difference GeRe we need to know at what
¼ n0;I3 þ n0;I pD þ n0;I3 þ n0;I ðL DÞ; (15)
3 3 rates the transitions shown in Fig. 2 occur. Rate of transition (a) is
determined by the optical generation rate GL of conduction band
ZD ZL electrons and is calculated with optical simulator SunShine [16,24].
1 1 1 Trapping and detrapping rates (b and c) of conduction band elec-
nA;e þ nA;I dx þ nB;I dx
2 3 3 trons are calculated according to Shockley-Read-Hall recombina-
0 D
tion statistics [17]. Electron transfers rates between the
1 1 1 semiconductor and ions in the electrolyte (d, e and f) are modeled
¼ n þ p n0;I D þ n0;I ðL DÞ; (16)
2 0;e 3 3 according to Marcus-Gerischer theory (MG) [18e20].
where n0,i are equilibrium concentrations of involved particles Electrons in the conduction band can be trapped by an expo-
(i ¼ e, I, I3 ). nential distribution of localized states in the band gap, from where
they can be thermally released back to the conduction band, or they
can recombine with the triiodide ions in the electrolyte. Expo-
nential distribution of trap states is assumed, although generally
2.4. Generation and recombination rates any distribution of trap states can be included:
where kox(E) includes Gaussian distribution of energy levels of Besides active and bulk electrolyte layer the front and back
oxidized ions and ket determines the rate of electron transfer. contacts also affect the cell operation. Back contact must allow for
fast electron transfer to triiodide ions so they can be reduced to
kox ðEÞ ¼ 2kT$ket $Dox ðEÞ: (25) iodide ions that are able to regenerate dye molecules. If this elec-
tron transfer is not fast enough, the voltage drop at this contact
Injection rate from iodide ions into TiO2 band gap states is becomes significant. To increase the rate of electron transfer the
back contact is covered by a layer of catalyst, usually platinum.
dGMG;e ðEÞ ¼ kred ðEÞgðEÞð1 f ðEÞÞdE; (26)
Since the TiO2 structure is mesoporous, the electrolyte pene-
where kred(E) includes Gaussian distribution of energy levels of trates the pores and reaches the front TCO. Electrons that have
reduced ions already reached the front contact can recombine with triiodide
electrons present at the TCO layer [25]. TCO layer is a highly doped
kred ðEÞ ¼ 2kT$ket $Dred ðEÞ: (27) semiconductor and acts like a metal, therefore we can apply But-
lereVolmer equation that describes current density jk dependence
At steady-state the density of trapped electrons in the band gap
on applied voltage Uk at a metal/electrolyte interface
does not change, and the same amount of electrons that enters
(dRSRH,e, dGMG,e) the trap states must leave (dGSRH,e, dRMG,e) them
ð1 bk ÞUk bU
jk ¼ ak jk;0 exp q exp q k k ; (36)
df kT kT
gðEÞ ¼ dRSRH;e dGSRH;e dRMG;e þ dGMG;e ¼ 0: (28)
dt where jk,0 is the exchange current density, bk is symmetry factor
By inserting Eqs. (18, 19, 24 and 26) into Eq. (28) and solving for f and ak determines the effective area of the contact surface reached
we get a steady-state occupation function by the electrolyte. Index k represents different contacts, namely
k ¼ TCO at the front contact and k ¼ Pt at the back contact.
ktrap nA;e þ kred ðEÞ At the back contact UPt equals the voltage applied to the whole
f ðEÞ ¼ : (29)
ktrap nA;e þ n1 ðEÞ þ kox ðEÞ þ kred ðEÞ cell, and surface of platinum contact is assumed to be flat, therefore
aPt ¼ 1. At the front contact the UTCO is determined by the difference
Eqs. (18, 19, 24 and 26) give expressions for electron transitions between the quasi-Fermi level of TCO and the redox potential in
at a fixed energy level. In order to get total contributions from all electrolyte at the front contact. Quasi-Fermi level of TCO is assumed
energy levels these equations have to be integrated through the to be the same as in the TiO2 at the front contact
band gap
qUTCO ¼ EF ð0Þ Eredox ð0Þ: (37)
ZECB
Eredox(0) is calculated using Eq. (8) but with concentrations
RSRH;e ¼ ktrap nA;e gðEÞð1 f ðEÞÞdE; (30) present at the front contact which results in a small value for UTCO
EVB even when the cell is in short circuit. Only parts of the TCO that are
not covered by the TiO2 are in contact with the electrolyte, there-
fore aTCO < 1. In our case we assume that aTCO is the same as TiO2
ZECB porosity (aTCO ¼ p).
In order to get good agreement with measured currentevoltage
GSRH;e ¼ krelease gðEÞf ðEÞdE; (31)
characteristics, resistances of the TCO layers along with other
EVB possible shunting effects are included as the series Rs and parallel
242 M. Filipic et al. / Current Applied Physics 12 (2012) 238e246
Experimental section in Ref. [16]) for 1, 0.5, 0.25 and 0.1 sun (AM1.5)
Measurement shows three distinct regions of UOC(GL) with
light intensity in Fig. 4. Besides the varied parameter cell is
different slopes. In order to get good agreement of measurement
measured under standard test conditions (STC: 25 C, AM1.5,
and simulation results in the lowest light intensity region (light
100 mW/cm2). Same input parameters listed in Table 1 are used for
intensity < 2 103 sun) a single level (monoenergetic) trap states
all light intensities.
had to be included in TiO2 band gap at energy level
Emono ¼ 4.35 eV (0.15 eV above Eredox 0 ) with concentration of
17 3
3.2. Light intensity dependence of open circuit voltage Nmono ¼ 1.5 10 cm . This deep level of trap states influences
mainly the shape of UOC(GL) in the lowest light intensity region,
Fig. 5 shows the light intensity (GL) dependence of open circuit while the UOC(GL) shape is not affected in higher light intensity
voltage (UOC). The UOC(GL) characteristic was measured using regions since the trap states become filled. In the second low
a white light emitting diode (LED Bridgelux BXRA-C0603) at cell intensity region (2 103 sun < light intensity < 2 102 sun)
temperature of 25 C. Lower light intensities were obtained with UOC(GL) is determined by recombinations through exponential
neutral density filters, by increasing the distance between LED and distribution of traps and exchange with ions in electrolyte. Slope in
measured DSSC and/or by decreasing the current supply through this region is affected mainly by trap distribution parameter a and
the LED. At lower light intensities special care was taken to elimi- electrolyte reorganizational energy l. Trap concentration Nt and
nate stray light. The UOC(GL) measurements were performed in the recombination rate constants ket and ktrap determine where regions
range of four orders of magnitude, starting at 1 sun and reduced with different slopes start and end but do not influence the slope.
down to 104 sun (from 100 down to 0.01 mW/cm2) (note: the Reorganizational energy l also influences recombination factor for
100 mW/cm2 AM1.5 spectrum in the wavelength range between conduction band electrons (Eqs. (34) and (25)). Increase of l
300 and 800 nm corresponds to approximately 1.8 1013 photons/ increases Eox and widens the Gaussian distribution of triiodide ions
s/cm2). and consequently increases recombinations through conduction
Table 1 59 mV/decade [26] which does not agree with the measured slope
Input parameters of the model. of 92 mV/decade. A better fit is obtained if we include the effect of
Parameter Symbol Value recombinations through TCO layer (full line in Fig. 5) with
Electron transfer rate ket 5 1018 cm3/s jTCO,0 ¼ 104 mA/cm2 and bTCO ¼ 0.65 (see Eq. (36)), but this would
Trapping rate ktrap 4 1011 cm3/s suggest that UOC(GL) in the third region is determined mainly by
Reorganizational energy l 0.3 eV recombinations through uncovered TCO at the front contact and
Electron diffusion constant De 3.8 104 cm2/s
not by conduction band recombinations. Another possibility of
Iodide diffusion constant DA;I ; DB;I 2.5 107 cm2/s
Triiodide diffusion constant DA;I3 ; DB;I3 2.5 107 cm2/s influencing the slope in the third region is to include nonlinear
Initial iodide concentration n0I 2.3 1021 cm3 recombination from TiO2 conduction band [27,28]
Initial triiodide concentration n0I 1.2 1020 cm3
3
g
Conduction band ECB 3.6 eV RCB;e ¼ nA;e knonlin ; (39)
Standard redox potential 0
Eredox 4.5 eV
Valence band EVB 6.8 eV
where knonlin is recombination constant for the case of nonlinear
Conduction band effective density of states NCB 3.3 1020 cm3
Total trap concentration Nt 5 1018 cm3
recombinations and g is the exponent that influences the slope of
Trap distribution parameter a 0.24 UOC(GL) [26]. With Eq. (39) good fit in the third region can be
Monoenergetic trap concentration Nmono 1.5 1017 cm3 achieved (not shown in Fig. 5; g ¼ 0.65,
Energy level of monoenergetic trap states Emono 4.35 eV knonlin ¼ 1.5 106 cm3(1g) s1), but it does not give us insight in
Porosity p 0.5
the recombination mechanism.
Series resistance RS 19.6 U
Parallel resistance RP 3.6 kU In the inset of Fig. 5 the measured efficiency of the cell with
Cell area A 0.56 cm2 varying light intensity is shown. Maximum efficiency is measured
Active layer thickness D 13 mm at 0.25 sun light intensity. At light intensities above 0.25 sun the
Electrolyte layer thickness LD 9 mm efficiency decreases due to serial resistance losses and a decrease in
Front contact TCO exchange current density jTCO,0 104 mA/cm2
Back contact Pt exchange current density jPt,0 80 mA/cm2
fill factor. Below 0.25 sun light intensity a decrease of UOC causes
TCO symmetry parameter bTCO 0.65 a fall in the efficiency of the cell.
Pt symmetry parameter bPt 0.5
3.3. Electron densities
band. Increasing l therefore increases the slope of UOC(GL) in the
second region and shifts the start of the third region to lower light Fig. 6 shows a profile of electron density in the active layer at
intensities. short and open circuit for different illumination intensities. At short
In the normal range of light intensities during daylight outdoors circuit the electrons are effectively collected at the front contact,
or indoors (third region e light intensity > 2 102 sun) UOC(GL) is and their concentration is decreasing toward the contact. At open
determined by recombinations through conduction band of TiO2 circuit only recombination currents flow inside the cell. Electrons
(dashed line in Fig. 5). Since the iodide and triiodide ion concen- recombine with triiodide ions in the electrolyte, or enter the front
trations in active layer are higher than electron concentrations the TCO layer and recombine from there. Since at open circuit no
variation of Eredox with light intensity is negligible and does not current flows out of the cell, the current of electrons is compen-
affect the UOC(GL) (variation of Eredox with light intensity is dis- sated by current of triiodide and iodide ions in the electrolyte. At
cussed in the last paragraph of this chapter). For conduction band open circuit UTCO is higher than at short circuit and recombination
recombinations in the third region the slope for constant Eredox is current jTCO cannot be neglected. This recombination current
density causes the electron concentration to start decreasing near
the front contact. In our case (jTCO,0 ¼ 104 mA/cm2 and bTCO ¼ 0.65)
it is a rather small decrease as can be seen in Fig. 6 for open circuit
conditions in the region 0 < x < 1 mm.
Fig. 5. Simulated and measured light intensity dependence of UOC. In the light
intensity region >2 102 sun, simulations with and without front contact recom- Fig. 6. Short circuit and open circuit electron densities under different illumination
binations show different slopes. The inset shows the measured efficiency of the cell conditions. In the region <1 mm a decrease of open circuit concentrations is observed,
with varying light intensity. caused by front TCO/electrolyte contact recombinations.
244 M. Filipic et al. / Current Applied Physics 12 (2012) 238e246
Fig. 7. Occupation function for different ratios of ket/ktrap for fixed electron and ion Fig. 9. Marcus-Gerischer recombination and generation rates between trap states in
concentrations corresponding to open circuit conditions. TiO2 and electrolyte for different values of ket and ktrap.
M. Filipic et al. / Current Applied Physics 12 (2012) 238e246 245
time because the occupancy of trap states is low and also not
df
influenced much by ket. For medium values of ktrap dk s0 and slope
et
Fig. 12. Dependence of STC efficiency on active layer thickness for bulk electrolyte
Fig. 10. Effects of recombinations between front contact TCO and electrolyte on cur- layers (L e D) of 0, 2, 9, 15 and 30 mm. Crosses mark the points of highest efficiency.
rentevoltage characteristic for various front contact TCO exchange current densities jTCO,0. Efficiency of measured cell is shown for reference.
246 M. Filipic et al. / Current Applied Physics 12 (2012) 238e246
Plots in Fig. 12 clearly show that highest efficiency can be obtained [4] J. Ferber, R. Stangl, J. Luther, An electrical model of the dye-sensitized solar
by minimizing the electrolyte layer thickness. For 0, 2 and 9 mm cell, Sol. Energy Mater. Sol. Cells 53 (1998) 29e54.
[5] J. Ferber, J. Luther, Modeling of photovoltage and photocurrent in dye-
thicknesses the diffusion limited current is higher than current sensitized Titanium Dioxide solar cells, J. Phys. Chem. B. 105 (2001)
generated by light and does not influence cell efficiency. For 15 mm 4895e4903.
thicknesses diffusion limited current is lower and limits the photo- [6] A. Usami, H. Ozaki, Computer simulations of charge transport in dye-
sensitized nanocrystalline photovoltaic cells, J. Phys. Chem. B. 105 (2001)
generated current when active layer is thick enough to generate 4577e4583.
currents comparable to the diffusion limited current. This effect is [7] J. Nelson, R.E. Chandler, Random walk models of charge transfer and transport
even more pronounced for 30 mm electrolyte layer thickness where in dye sensitized systems, Coord. Chem. Rev. 248 (2004) 1181e1194.
[8] J.-J. Lee, G.M. Coia, N.S. Lewis, Current density versus potential characteristics
diffusion limited current is lowest. For 15 and 30 mm electrolyte of dye-sensitized nanostructured semiconductor photoelectrodes. 1. Analyt-
layer thickness current becomes diffusion limited for active layers ical expressions, J. Phys. Chem. B. 108 (2004) 5269e5281.
thicker than 20 and 10 mm, respectively. Simulation results there- [9] T. Oda, S. Tanaka, S. Hayase, Differences in characteristics of dye-sensitized
solar cells containing acetonitrile and ionic liquid-based electrolytes studied
fore show that optimum active layer thickness is a trade-off using a novel model, Sol. Energy Mater. Sol. Cells 90 (2006) 2696e2709.
between light harvesting on one side and recombination rates at [10] M. Penny, T. Farrell, C. Please, A mathematical model for interfacial charge
the uncovered TiO2 surface and/or diffusion limited current deter- transfer at the semiconductoredyeeelectrolyte interface of a dye-sensitised
solar cell, Sol. Energy Mater. Sol. Cells 92 (2008) 11e23.
mined with the diffusion properties of ions in the actual electrolyte [11] M. Penny, T. Farrell, G. Will, A mathematical model for the anodic half cell of
on the other side. Our cell is not far from optimal active layer a dye-sensitised solar cell, Sol. Energy Mater. Sol. Cells 92 (2008) 24e37.
thickness for given 9 mm electrolyte layer. According to the results [12] S. Chen, J. Weng, Y. Huang, C. Zhang, L. Hu, F. Kong, L. Wang, S. Dai, Numerical
model analysis of the shaded dye-sensitized solar cell module, J. Phys. D: Appl.
of the simulation, increasing active layer thickness of our DSSC to
Phys. 43 (2010) 305102e305108.
18 mm and decreasing bulk electrolyte layer to 2 mm would further [13] J. Villanueva, J. Anta, E. Guille cn, G. Oskam, Numerical simulation of the cur-
increase the cell efficiency for an absolute value of 0.3%. rentvoltage Curve in dye-sensitized solar cells, J. Phys. Chem. C 113 (2009)
19722e19731.
[14] J. Villanueva-cab, G. Oskam, J. Anta, A simple numerical model for the charge
4. Conclusion transport and recombination properties of dye-sensitized solar cells:
a comparison of transport-limited and transfer-limited recombination, Sol.
The model of the DSSC based on continuity and transport Energy Mater. Sol. Cells 94 (2010) 45e50.
[15] P.R.F. Barnes, A.Y. Anderson, J.R. Durrant, B.C. O’Regan, Simulation and
equations in the active as well as the bulk electrolyte layer including measurement of complete dye sensitised solar cells: including the influence of
series resistance of contacting layers is presented. Trapping/ trapping, electrolyte, oxidised dyes and light intensity on steady state and
detrapping processes and electron exchange between semi- transient device behaviour, Phys. Chem. Chem. Phys. 13 (2011) 5798e5816.
[16] M. Topi
c, A. Campa, M. Filipi
c, M. Berginc, U.O. Krasovec, F. Smole, Optical and
conductor and electrolyte is considered when determining recom- electrical modelling and characterization of dye-sensitized solar cells, Curr.
bination rates. Generation and recombination rates are calculated Appl. Phys. 10 (2010) S425eS430.
and influence of trapping and detrapping probabilities on recom- [17] W. Shockley, W. Read, Statistics of the recombinations of holes and electrons,
Phys. Rev. 87 (1952) 835e842.
bination rates are evaluated. Simulated currentevoltage character- [18] R.A. Marcus, On the theory of electron-transfer reactions. VI. Unified treat-
istics and light intensity dependence of open circuit voltage show ment for homogeneous and electrode reactions, J. Chem. Phys. 43 (1965)
good agreement with measured data. Recombinations of electrons 679e701.
[19] J. Bisquert, A. Zaban, P. Salvador, Analysis of the mechanisms of electron
through the front TCO contact are analyzed, and their effect on the
recombination in Nanoporous TiO 2 dye-sensitized solar cells. Nonequilib-
currentevoltage characteristic and electron concentration in the rium steady-state statistics and interfacial electron transfer via surface states,
cell is shown. As an application of the presented model optimization J. Phys. Chem. B. 106 (2002) 8774e8782.
of cell efficiency with varying active layer thickness is carried out [20] N. Sato, Electrochemistry at Metal and Semiconductor Electrodes, first ed.
Elsevier, Amsterdam, 1998.
and optimal thickness is found to be 18 mm for the measured cell. [21] A. Hagfeldt, M. Grätzel, Light-induced redox reactions in nanocrystalline
Simulations show that optimal active layer thickness and efficiency systems, Chem. Rev. 95 (1995) 49e68.
of the cell increase for thinner electrolyte layers. [22] G. Boschloo, A. Hagfeldt, Characteristics of the iodide/triiodide redox mediator
in dye-sensitized solar cells, Acc. Chem. Res. 42 (2009) 1819e1826.
[23] J. Rowley, G.J. Meyer, Reduction of I2/I3 by Titanium Dioxide, J. Phys. Chem. C
Acknowledgments 113 (2009) 18444e18447.
[24] J. Krc, F. Smole, M. Topi&ccaron, Analysis of light scattering in amorphous Si:H
solar cells by a one-dimensional semi-coherent optical model, Prog. Photo-
The authors are grateful for the funding by the Slovenian voltaics Res. Appl. 11 (2003) 15e26.
Research Agency (P2-0197). The authors would like to thank the [25] P.J. Cameron, L.M. Peter, How does back-reaction at the conducting glass
DSSC team of LPVO for providing the DSSC samples. substrate influence the dynamic photovoltage response of nanocrystalline
dye-sensitized solar cells? J. Phys. Chem. B 109 (2005) 7392e7398.
[26] P. Salvador, M.G. Hidalgo, A. Zaban, J. Bisquert, Illumination intensity
References dependence of the photovoltage in nanostructured TiO2 dye-sensitized solar
cells, J. Phys. Chem. B. 109 (2005) 15915e15926.
[1] M. Grätzel, Optimizing Photon Harvesting and Carrier Collection in Meso- [27] P.R.F. Barnes, B.C. O’Regan, Electron recombination kinetics and the analysis of
scopic Solar Energy Conversion Systems Presentation at MRS Fall Meeting, collection efficiency and diffusion length measurements in dye sensitized
Boston, MA (2009). solar cells, J. Phys. Chem. C 114 (2010) 19134e19140.
[2] S. Soedergren, A. Hagfeldt, J. Olsson, S.-E. Lindquist, Theoretical models for the [28] J. Villanueva-Cab, H. Wang, G. Oskam, L.M. Peter, Electron diffusion and back
action spectrum and the currentevoltage characteristics of microporous reaction in dye-sensitized solar cells: the effect of nonlinear recombination
semiconductor films in photoelectrochemical cells, J. Phys. Chem. 98 (1994) kinetics, J. Phys. Chem. Lett. 1 (2010) 748e751.
5552e5556. [29] M. Berginc, U. Opara-Krasovec, M. Jankovec, M. Topic, The effect of temper-
[3] N. Papageorgiou, P. Liska, A. Kay, M. Grätzel, Mediator transport in multilayer ature on the performance of dye-sensitized solar cells based on a propyl-
nanocrystalline photoelectrochemical cell configurations, J. Electrochem. Soc. methyl-imidazolium iodide electrolyte, Sol. Energy Mater. Sol. Cells 91
146 (1999) 898e907. (2007) 821e828.