1126 J. Phys. Chem.

C 2009, 113, 1126–1136

Electron Injection Efficiency and Diffusion Length in Dye-Sensitized Solar Cells Derived
from Incident Photon Conversion Efficiency Measurements

Piers R. F. Barnes,* Assaf Y. Anderson, Sara E. Koops, James R. Durrant, and

Brian C. O’Regan
Department of Chemistry, Imperial College London, London SW7 2AZ, U.K.
ReceiVed: October 13, 2008; ReVised Manuscript ReceiVed: NoVember 6, 2008

Injection efficiency, ηinj, and diffusion length, L, in dye-sensitized solar cells have been derived from the
spectral response (incident photon to current efficiency, IPCE) of the cells under front side or backside
illumination. Values of L from IPCE are found to be ∼2 times shorter than the values of L derived from the
normal small perturbation transient method. IPCE-derived values of L (2 to more than 40 µm) and ηinj (63-90%)
are found to correlate with the photocurrent (and indirectly with the photovoltage) of the different cells indicating
the extent to which each factor limits the cell efficiency. IPCE spectra varied with light intensity, so that
diffusion lengths derived from both methods show similar trends, e.g., L from IPCE is found to increase 3
times when the light intensity is increased 10 times up to approximately 0.1 sun where L tends to plateau or
peak. The values for ηinj derived from the spectral response are shown to be in quantitative agreement with
those determined from picosecond transient emission spectroscopy. To illustrate the utility of this method, L
and ηinj were measured on cells with and without the TiCl4 chemical bath treatment. The results show that
the increase in photocurrent after the TiCl4 treatment is due to around a 2-fold increase in L despite a 3-fold
reduction in the electron diffusion coefficient. The increased L can be explained by a factor of 10 decrease
in electron recombination rate.

1. Introduction The value of ηinj is dependent upon the electronic coupling

of the dye to the TiO2 and energy of the dye’s lowest unoccupied
A key issue for the progress of dye-sensitized solar cells
(DSSCs) is the development of characterization tools relating molecular orbital (LUMO) relative to the band of electron-
device performance to underlying processes. In DSSCs opti- accepting orbitals in the TiO2. The injection process can be
mized for maximum photocurrent, the incident photon to influenced by the presence of the surrounding electrolyte6 and
electron conversion efficiency (IPCE, also known as external may also be influenced by the wavelength of the exciting photon
quantum efficiency, action spectrum, or spectral response) can if significant injection occurs before thermal relaxation in the
approach 100% after reflection losses.1-3 However, in practice dye.7 Electron injection in dye-sensitized cells is often assumed
for most cells using new materials or cells optimized for to be a fast process occurring on the femtosecond time scale;
maximum efficiency, the maximum IPCEs are significantly however recent studies suggest that slow injection from the
lower with correspondingly less than ideal photocurrent. Quali- triplet state may dominate photocurrent generation in working
tatively, the decrease in photocurrent can often be ascribed to cells sensitized with the widely used ruthenium bipyridyl dye,
one or more loss routes; however, reliable measurements N719.6,8,9
allowing a quantitative explanation for the lost photocurrent have The efficiency of electron collection from within the TiO2
not been available. With mushrooming numbers of new materi- will be dependent on competition between recombination and
als, dyes, and electrolytes being tested in dye-sensitized cells, transport of electrons to the anode. For most liquid electrolyte
methods to characterize the limitations of these new components dye cells, charge transport is dominated by the diffusion of
are increasingly important. electrons since the dimensions and high dielectric constant of
The IPCE, denoted as η in this paper, can be factorized into the TiO2 nanoparticles do not allow a significant electric field
three components enabling the origin of photocurrent loss to in the semiconducting component and the injected electrons are
be examined.4,5 The factors are the efficiency of photon screened by the high charge density of the electrolyte.10
harvesting by dye molecules, ηLH, the efficiency of charge Recombination of electrons before collection may occur either
injection from the excited dye molecules to the TiO2 substrate, with iodine or triiodide in the electrolyte or with photogenerated
ηinj, and the efficiency of charge collection from the TiO2 matrix dye cations which have not been regenerated by the electrolyte.
to the external circuit, ηcol, such that4 The competing processes of recombination and diffusion can
jsc(λ) conveniently be described by the parameter diffusion length,
η(λ) ) ) ηLH(λ)ηinj(λ)ηcol(λ) (1) L, which is the average distance (normal to the plane of the
qφ(λ)
where jsc is the short circuit current, q is the elementary charge, film) an injected electron can travel through the cell before
and φ is the photon flux density at wavelength λ incident on recombination. If L > d (where d is the TiO2 film thickness),
the cell. The ηLH is primarily dependent on the absorption only a small fraction of injected charge will be lost before it is
spectrum and loading of the dye on the TiO2. collected, making long diffusion lengths desirable.
There has been a recent resurgence in interest in using the
* Corresponding author, [email protected]. parameter L to assess the collection efficiency with dye-
10.1021/jp809046j CCC: $40.75  2009 American Chemical Society
Published on Web 12/23/2008
Injection and Diffusion Length in DSSCs from IPCE J. Phys. Chem. C, Vol. 113, No. 3, 2009 1127

sensitized cells.4,11-21 However many groups have erroneously geometry was 1 × 1 cm, and the TiO2 films were heated using
derived L from small perturbation frequency domain or transient a hot plate up to 500 °C for 30 min the heating procedure
measurements as a function of light intensity by taking the described elsewhere.3 A TiCl4 treatment was applied to most
square root of the product of the “effective” diffusion coefficient films by soaking in a 50 mM TiCl4 solution for 30 min at 70
measured at short circuit and “effective” electron lifetime at °C, followed by a water rinse and heating at 450 °C for half an
open circuit at the same light intensity. Under open circuit hour.22 The treatment is typically thought to deposit a layer of
conditions the concentration of electrons in the cell is higher TiO2 1-2 nm thick on the surface of the porous film.22 The
compared to short circuit. This seemed to lead to underestima- resulting films were relatively transparent; their thicknesses d
tion of the diffusion length. This problem can be avoided by were measured using a profilometer and porosity estimated from
finding the appropriate electron lifetime corresponding to the mass measurements. Dye was applied to the samples from a
quantity of charge in the cell under the short circuit conditions 0.3 mM solution of N719 ruthenium-based dye supplied by
used to find the diffusion coefficient.22 A similar approach can Dyesol in 1:1 mixture of tert-butyl alcohol (TBA) and aceto-
also be used to find appropriate diffusion coefficient values at nitrile (ACN). Films were dyed for 6 days to ensure complete
open circuit by extrapolation from short circuit measurements.16 surface coverage; the dyed films were then rinsed in TBA:ACN
These methods yield higher values of L than the previous solution prior to cell construction to help remove aggregated
underestimations which often suggest electron collection does dye molecules from the TiO2 surface. Transparent counter
not limit photocurrents. However, in addition there are a number electrodes were prepared by chemically depositing platinum
of further complications involved in finding L by transient from 0.05 M hexachloroplatanic acid in 2-propanol on to a
methods which will be discussed in sections 2.9, 3.3, and 3.5. second slide of conducting glass. The cells were sealed using
Knowledge of ηinj and L in principle enables the factors Surlyn 1472 polymer (DuPont Ltd.) at 120 °C. Various
limiting cell photocurrent to be determined. This paper describes electrolyte compositions and additives were used to modulate
a method to derive both ηinj and L (and consequently ηcol) from the performance of the cells. The series of cells with different
IPCE measurements based on an approach originally applied film thicknesses (made using the Dyesol paste) had an electrolyte
to porous TiO2 films by Sodergren et al.23 which was first applied composed of a methoxypropionitrile solvent with 0.8 M
to dye-sensitized cells by Lindstrom et al.24 and most recently 1-propyl-3-methylimidazolium iodide, 0.3 M benzimidazole, 0.1
by Halme et al.4 In the present study, cells with different film M iodine, and 0.05 M guanidinum thiocyanate. The remaining
thicknesses, electrolyte compositions, and TiO2 preparations cells (a series with 4 and 13 µm TiO2 films) contained 0.6 M
were fabricated. These factors all cause changes in the photo- tetrabutylammonium iodide, and 0.1 M iodine in 3-methox-
current and photovoltage. The parameters ηinj and L were derived ypropionitrile. A number of different combinations of additive
for each cell by fitting a model to IPCE and optical data. The concentrations were used in this electrolyte to give cells of
key to this analysis is the relationship between η(λ) and L so differing performance: 0, 0.1, 0.2, and 0.5 M of tert-butyl
that if sufficient optical measurements are made, η(λ) can be pyridine (TBP), with and without 0.1 M lithium iodide.
described by a function where the only unknowns are ηinj and 2.2. Current-Voltage Measurements. The current-voltage
L. Moreover, the ratio of η(λ) measured from the back and front characteristics of the cells were determined by illuminating using
sides (in the case of a transparent cell) depends only on L (see a 150 W xenon lamp (Sciencetech model SS150Wsolar simula-
section 2.8). tor), equipped with an IR filter (water filter) and an AM 1.5
The focus of this paper is to demonstrate that this technique filter (Sciencetech). Beam intensity was calibrated using a silicon
yields useful values which are directly related to the photocurrent photodiode with a spectral response modified to approximately
(and indirectly to the photovoltage) of a wide range of cells match the absorption profile of the N719 dye. Current and
(section 3.2). Additionally, the technique is verified by com- voltage were measured and controlled using a Keithley 2400
parison with an independent method for determining ηinj based source meter.
on transient photoemission studies (section 3.4) and the values 2.3. IPCE Measurement. Incident photon conversion ef-
of L are compared to values determined from small perturbation ficiency measurements were made using a 100 W tungsten
transient photovoltage and photocurrent measurements (section halogen lamp (Bentham IL1 with Bentham 605 stabilized current
3.5). To illustrate the application of the technique, the role of power supply) coupled to a monochromator with computer-
the TiCl4 chemical bath treatment22,25 in increasing cell photo- controlled stepper motor (Bentham M300, 300 mm focal length,
current is examined in more depth as for example in section slit width 3.7 nm, 1800 lines/m grating). The photon flux of
3.5. The derived values of ηinj and L are shown to be useful for light incident on the samples was calibrated using a UV-
predicting the performance of cells, and those wishing to see enhanced silicon photodiode. A 590 nm long-pass glass filter
only the application of this technique can skip to the table in was inserted into the beam at illumination wavelengths longer
section 3.5. than 620 nm to remove light from second-order diffraction.
Photocurrent was measured using a Keithley 2400 source meter;
2. Method measurement duration for a given wavelength was sufficient to
2.1. Cell Fabrication. Cells were fabricated as described ensure the current had stabilized (up to around 5 s under low
previously22,25 with the following details. Transparent conductive or zero bias light conditions). Measurements were typically made
F-doped SnO2 glass (Pilkington LOF Tec15) was used as the at 10 or 20 nm wavelength intervals between 400 and 850 nm.
photoanode and cathode substrate material. TiO2 films were The cells were illuminated from either the TiO2 photoanode
deposited by doctor blading, using Scotch tape spacers to vary (substrate-electrode, SE) side or the Pt cathode (electrolyte-
film thickness, from a paste of particles either supplied by electrode, EE) side.
Dyesol (20 nm anatase particles suspended 20% by weight as IPCE measurements with additional bias light used an array
a colloid in an acidic solution pH <2, made into a paste by of eight white light emitting diodes (Luxeon star/O, white
adding hydroxypropyl cellulose, poly(ethylene glycol), and LXHL_NWE8) driven by a TTL QL564 dc power supply to
HNO3) or synthesized from Ti-isopropoxide following the nitric generate a constant background photocurrent in the cell. Bias
acid/acetic acid route followed by autoclaving.26 The cell illumination was from the same direction as the monochromatic
1128 J. Phys. Chem. C, Vol. 113, No. 3, 2009 Barnes et al.

source. The monochromatic beam was chopped using a computer-

controlled shutter and the difference in resulting stabilized
photocurrents used to calculate the IPCE. Averaging of up to
four shutter cycles was employed at higher bias light intensities
to reduce noise.
2.4. Time-Correlated Single-Photon Counting (TCSPC)
Measurements. The technique for collecting TCSPC measure-
ments is described in detail elsewhere.9,27 A JobinYvon TBX
Flurocube system used a laser pulse (wavelength 467 nm) to
excite dye molecules within the cells at a repetition rate of 1
MHz and average intensity of 80 µW cm-2. The resulting
photoluminescence intensity was measured as a function of time
after the pulse with an instrumental response time of 250 ps at
full width half-maximum. The photoluminescence relaxation
could be well fit by the convolution of the instrument response
with a stretched exponential;9,27 this function is related to the
rate at which excited dye states nonradiatively inject electrons
into the TiO2, quenching the photoluminescent relaxation of the
dye to its ground state. Deconvolving the fitted function allows
the half-time and integrated luminescence to be determined.
Comparison was made between the signals emitted from the
cells described above and control cells which were identical in
all respects except that the TiO2 film had been replaced by a
nanocrystalline ZrO2 film. Since excited dye states are unable
to inject charge into ZrO2, there is no equivalent route for
photoluminescent quenching. Thus analysis of the TiO2 and the
ZrO2 cell relaxation functions allows the electron injection yield
from the dye to the TiO2 to be determined (equivalent to ηinj).9
2.5. Photoelectrical Transient and Charge Extraction
Measurements. Photocurrent and photovoltage transients were
generally taken with the use of a pump pulse generated by an
array of 1 W red light emitting diodes (LEDs) controlled by a
fast solid-state switch.22 White bias light was supplied by an
array of white 1 W LEDs as described in section 2.3, the light
intensity was modulated by varying the current supplied to the Figure 1. (a) The absorption coefficient (R) of N719 dye adsorbed to
TiO2, scattering coefficient (s) of the TiO2 and the attenuation of light
LEDs and calibrated using a silicon cell. The pulse was incident due to iodine absorption and platinum with EE side illumination where
on the SE side of the cell and pulse lengths of 10-100 µs were d ) 13 µm. (b) The excited dye state generation rate profile (G) as a
used with rise and fall times of e1 µs. Pulse intensity was function of position in the cell for SE and EE AM1.5 illumination32
adjusted to maintain the amplitude of the transient photovoltage (eqs 4 and 5) at λ ) 535 nm (bandwidth 10 nm, φ ) 4.2 × 1015 cm-2
below 5 mV. The magnitudes of the photocurrent transient tails s-1) and 700 nm (bandwidth 10 nm, φ ) 4.5 × 1015 cm-2 s-1). (c)
were less than 5% of the bias photocurrent at 1 “sun” bias Corresponding calculated conduction band electron concentration
profiles assuming D0 ) 0.45 cm2 s-1,33 τ0 ) 5 × 10-7 s (L ) 4.7 µm),
illumination. Transients were measured using National Instru- ηinj ) 1, and R ) 0 (calculated using eqs A1 and A2).
ments USB-6211 multifunction data acquisition box interfaced
to a PC. No signal averaging was required, the photovoltage charge extraction current. This assumption appears valid in the
decays were single exponential, and the time constant, τn, was majority of cases where τj , τn for a given charge concentration.
determined by fitting. The photocurrent transients at short circuit Thus electron concentration, n, in the cell as a function of light
have close to single exponential tails (after sufficient time) where intensity, photovoltage, and photocurrent could be determined.
the time constant, τj, was also determined by fitting. The value 2.6. Optical Measurements. Optical transmission measure-
of τj can be approximately related to the chemical diffusion ments were made using a Shimadzu UV-vis dual beam
coefficient of electrons in the TiO2 according to Dn ≈ d2/(2.47τj). spectrophotometer (UV-1601). Measurements of the conducting
The geometrical factor of 2.47 was determined by solving the glass, platinized layer, and electrolyte were made to determine
one-dimensional continuity equation (1) for the flux of charge the reflective and absorption losses in the cell not attributable
diffusing out from one side of a slab at fixed initial concentration to the dye/TiO2 system. These losses are denoted by R for the
(due to red light pulse) with no subsequent generation and reflectance from the glass, (1 - TPt) for the absorption due to
recombination. The factor is similar to the factors of 2.35, 2.77, the Pt and (1 - TI) for the absorption in the electrolyte between
and 2.4-2.5 used in other work analyzing the time constants the TiO2 and the Pt; see Figure 1a. The fraction of light
from IMPS and stepped transient measurements21,28,29 and is transmitted by complete cells with (Tdye/TiO2) and without (TTiO2)
appropriate for our photocurrent transient perturbation method dye and of film thickness d was used to estimate the absorption
in cases where recombination during transport is negligible. coefficient of the porous dyed film according to
The quantity of charge stored within the cell under either 1 Tdye⁄TiO2(λ)
short circuit or open circuit conditions was determined by R(λ) ) - ln (2)
d TTiO2(λ)
integrating the total current extracted from the cell immediately
following switching the light off.30 This assumes that the This assumes an eponential variation of light intensity with
recombination current in the cell is insignificant relative to the position and that Tdye/TiO2/TTiO2 ≈ [Tdye/TiO2/(1 - Rdye/TiO2)]total
Injection and Diffusion Length in DSSCs from IPCE J. Phys. Chem. C, Vol. 113, No. 3, 2009 1129

where total refers to all direct specular and diffuse reflected or corresponding to the dye absorption peak, but there is very little
transmitted light (i.e., the dye molecules are assumed not to variation for a weakly absorbed red light at λ ) 700 nm.
scatter a significant fraction of light). Similar values for R were To describe the generation profile within a more strongly
achieved measuring TiO2 films on FTO in acetonitrile covered scattering film, a more sophisticated approach using optical
with a glass coverslip rather than complete cells and had the measurements with an integrating sphere5 and employing a
advantage of reducing the uncertainty introduced by iodine at Kubelka-Munk four flux model to include the effects of
short wavelengths. An example of a measured absorption scattering would be appropriate31 but was not found necessary
coefficient made for one of the films is shown in Figure 1a. here.
Note that the absorption coefficient is related to the molar 2.8. Determination of the Steady-State Diffusion Length
extinction coefficient of the dye, ε, by R ) εc ln(10) where c and Injection Efficiency. L and ηinj can be determined from
is the concentration of dye attached within the TiO2 film. The IPCE and optical measurements. The approach assumes that
same approach was used to determine the absorption coefficient only electrons in the conduction band of the TiO2 can diffuse
of the iodine in the electrolyte, filling the TiO2 pores, RI(λ), by or recombine. The photocurrent generated by the cell can be
measuring transmission with and without iodine in the electro- determined by considering the electron continuity equation for
lyte and estimating the porosity of the films at 70% from mass the concentration of conduction band electrons nc at steady-
and thickness measurements (50% porosity after TiCl4 treat- state conditions
ment). The influence of scattering was assessed by also
measuring the transmission of a cell containing no TiO2 layer ∂2nc(x) nc(x) - nc0
(Tblank), the scattering coefficient is estimated by D0 - + ηinjG(x) ) 0 (6)
∂x2 τ0

1 TTiO2(λ) D0 and τ0 are the electron diffusion coefficient and recombina-

s(λ) ) - ln (3) tion lifetime in the conduction band and are assumed indepen-
d Tblank(λ)
dent of position, x, in the film. The injection efficiency, ηinj is
This represents an upper limit to the coefficient since it assumes also assumed to be constant for a given cell, and nc0 is the dark
no sub-band gap absorption in the TiO2 and intrinsically includes conduction electron concentration. Expression 6 can be solved
the component of light that has been specularly reflected. analytically to give nc(x) at short circuit for cells of thickness d
Scattering contributes substantially for wavelengths <450 nm with the boundary conditions nc(0) ) nc0 and dn(d)/dx ) 0 and
where more than 50% of the attenuation of the specular beam taking G(x) from eqs 4 for SE side illumination or 5 for EE
was due to scattering. Approximately 25% of the specular beam side illumination. Solutions for ncSE(x) and ncEE(x) are given by
attenuation (µm-1) at the dye absorption peak at 535 nm was eqs A1 and A2 in the Appendix. Examples of calculated ncSE(x)
due to scattering in the thicker cells as shown in Figure 1a. and ncEE(x) profiles are shown in Figure 1c in the case of SE
While this clearly introduces uncertainty into the following and EE side illumination for two different wavelengths. At λ
analysis when calculating the dye excited state generation profile ) 700 nm the profiles are very similar, such that the resulting
in the cell, we found that the observed IPCE data can still be flux of electrons collected at x ) 0 µm will be almost
fit successfully using a simple nonscattering approximation by independent of illumination direction. However for λ ) 535
ignoring measurements at λ <450-480 nm. nm the concentration profiles are very different and there is a
The values of R were found to be relatively independent of substantially higher concentration gradient at x ) 0 for the case
film thickness as can be seen in Figure 2. This suggests relatively of SE side illumination reflecting the greater flux of collected
homogeneous dye loading within the films, implying minimal charge. This is because the majority of charge generated by SE
influence from dye aggregation in the outer layers of TiO2 illumination is close to the collection electrode. In contrast most
despite the long dyeing time. The observation of homogeneity of the charge generated with EE illumination must diffuse across
simplifies the following optical analysis. The independence of the thickness of the cell before collection, allowing more time
R with d also implies that the assumptions required for eq 2 for recombination.
are sufficient for the following analysis at the level of uncertainty The IPCE is given by η ) [D0 dnc(0)/dx]/φ and, if R, RI, and
quoted in the results. d are known, η becomes a function of ηinj and L by making the
2.7. Light Harvesting Profiles. To infer information about substitution L ) (D0 τ0)1/2.23 See eqs A3 and A4 in the Appendix
the generation and transport of charge within the cell, it is for the expressions for IPCE under SE and EE side illumination,
necessary to estimate the light absorption profile throughout the ηSE and ηEE, respectively. Thus analysis of optical and IPCE
dyed TiO2 film. If φ is the flux of incident photons on the cell, data for a cell yields L and ηinj. Furthermore a fit to the ratio
the generation rate of excited dye states as a function of position ηEE/ηSE (eq A5), gives L alone since the ηinj terms cancel.
x in the film is approximated by The fit to ηEE/ηSE is particularly useful because it is
independent of calibration errors in the IPCE measurement (ηEE/
GSE(x) ) (1 - R)Rφe-(R+R1)x (4) ηSE can be normalized to unity at long wavelengths if the loss
due to Pt and iodine have not been well characterized as long
for substrate-electrode side illumination (SE) where x ) 0 is
as second-order diffracted light from the monochromator is
the position of the FTO substrate-TiO2 interface, and
removed). In this work eq A5 was first fit to measurements of
GEE(x) ) (1 - R)TPtTIRφe-(R+RI) d-x
( )
(5) ηEE/ηSE for a cell to yield L using a nonlinear least-squares
method; this value of L was then used to fit eqs A3 and A4 to
for the electrolyte-electrode side illumination of the cell (EE) ηEE and ηSE, respectively, to give ηinj, such that only single
where d is the thickness of the TiO2 film. Expressions for ηLH parameter fits were performed. Examples of data and corre-
are given by eqs A6 and A7 in the Appendix. An example sponding fits can be seen in Figure 3. Fits to data for λ <500
excited-state generation profiles for SE and EE side illumination nm were sometimes unreliable due to uncertainty in the
can be seen in Figure 1b for a film thickness of d ) 13 µm. determination of G(x) at these wavelengths due to scatter in
They indicate that the generation rate of excited dye states varies the TiO2 and iodine absorption; in these cases the fitting range
strongly across the film for incident light at λ ) 535 nm was reduced to avoid these data. The average confidence interval
1130 J. Phys. Chem. C, Vol. 113, No. 3, 2009 Barnes et al.

film under short circuit conditions (Figure 1c shows an example

of the variation of nc at short circuit which will be related to
variation in n). Consequently in this work it is useful to consider
the diffusion length values presented as a weighted mean of
the localized property defined by Ln ) (Dn τn)1/2 which can vary
as a function of position.
2.9. Determining L from Transient Measurements. The
transient perturbation measurements described above can be used
as an alternative independent method to determine L. The most
widely adopted description for DSSCs is the multiple-trapping
model with the quasi-static approximation. The assumptions of
this model lead to the corollary that L ) (Dnτn)1/2 ) (D0 τ0)1/2
Figure 2. Measured absorption coefficient at λ ) 530 and 700 nm ) constant such that L must be independent of electron
and diffusion length plotted against cell thickness. concentration and position if the approximation holds34 giving
us the standard definition of L. Thus L determined from the
transient measurements should be consistent with L derived from
conduction band electron diffusivity and lifetime, D0 and τ0.
To find L the appropriate τn must be determined for the
concentration of electrons in the cell n. A charge extraction
measurement at short circuit yields n allowing the appropriate
value of τn to be found (by interpolation or fitting) from the
photovoltage transient versus Voc relationship, such that


d2τn(n)
L) (7)
2.47τj(n)
where τn and τj are the photovoltage and photocurrent transient
decay times measured at matched electron concentrations, n.
We note that determining L using this approach has various
limitations. As described in the previous section on determining
L from IPCE measurements, L could vary with n due to nonideal
recombination or transport processes. In addition to this there
may be an intrinsic problem in determining Dn from short circuit
measurements of τn, since even in an ideal cell Dn is dependent
on n and is measured at short circuit, it represents a mean or
Figure 3. (a) Examples of fits to the IPCE ratio for two cells with characteristic value across a range of electron concentrations,
(open circles) and without (open triangles) the TiCl4 treatment (d ) and thus it may not be appropriate for a homogeneous
13 µm, electrolyte containing 0.1 M LiI but 0 M TBP). The values of concentration of charge in the cell expected at open circuit. Also
L lie within (7% and (4% of the indicated values, respectively, at under conditions where L < d, recombination of electrons during
the 95% confidence level. (b) Examples of fits to the SE and EE side transport will contribute to the measured value of τj. Furthermore
illumination IPCE for the cell without the TiCl4 treatment; the τj could also be influenced by the profile of electrons generated
uncertainty on ηinj is (10%.
by the incident light pulse. These effects could lead to
misleadingly large derived values of L. These issues will not
around the fitted values of L for measurements in this study be considered further quantitatively in this work but form the
was (8% (at the 95% confidence level). Note that when L > subject of a forthcoming study.
d, the fit becomes increasingly uncertain, limiting the meaningful
3. Results and Discussion
determination of L to cases where L is less than approximately
2d. The average confidence interval for ηinj was (5% of the The models for ηEE/ηSE, ηSE, and ηEE (eqs A5, A3, and A4)
fitted value; however, we estimate a measurement uncertainty could typically be fit well to the measured IPCE data collected
for this value at around (10%, due to the absolute nature of for this study except in some cases at short wavelengths, where
the photon accounting required. calculation of G(x) could be inaccurate (see Figure 3). The good
The approach described above assumes that D0 and τ0 are fits imply that a dependence of ηinj on λ is not required to explain
constant such that recombination flux scales linearly with nc the shape of the action spectra. Figure 3a shows the fits deriving
and is independent of trapped charge (n - nc) so that L is L from ηEE/ηSE for two cells with and without the TiCl4
constant regardless of nc (and thus n). As we show later, this treatment. These values of L in Figure 3a clearly show that the
assumption has only limited validity for the DSSCs studied TiCl4 treatment has a significant effect on the diffusion length;
herein. As such, the values of L obtained should be regarded as this will be discussed in more detail in the section 3.5.
approximate or a “characteristic” L referring to a global property 3.1. Dependence of L on Cell Thickness. Figure 2 shows
for the whole cell. This is because the values of L derived by the diffusion length determined from ηEE/ηSE for similar cells
fitting eqs A5, A6, or A7 incorporate any inadequacies inherent of differing TiO2 thicknesses, d; despite some scatter, there is
to assumptions and model given by eq 6. The continuity equation no trend in L with varying d as expected. In contrast Halme et
may require an additional term to account for variation of L al.4 showed a strong dependence between L and d; however, as
with n or D0, and τ0 may not be constant but may vary with n. they discussed, their cells were made from pressed TiO2 and
If L varies as a function of n, then it can no longer be considered the dependence was likely to be due to variation in the
a global property since n varies across the thickness of the TiO2 mechanical properties of the film with differing thicknesses.
Injection and Diffusion Length in DSSCs from IPCE J. Phys. Chem. C, Vol. 113, No. 3, 2009 1131

Figure 5. Dependence of L on illumination intensity for two different

cells, both with the TiCl4 treatment: closed circles, cells made with
TiO2 synthesized via the titanium isopropoxide route, d ) 13 µm,
electrolyte with 0 M LiI and 0 M TBP; open circles, cell made using
Dyesol TiO2 paste, d ) 16 µm.

available for charge separation. This leads to faster injection

kinetics as is shown in detail elsewhere.9
Figure 4. (a) L determined from ηEE/ηSE measurements (no bias light)
3.3. Dependence of L on Light Intensity. Values of L
for different cells (d ) 13 µm) as a function of the cell photocurrent
(SE side illumination). (b) Fitted ηinj values as a function of open circuit determined from ηEE/ηSE using a bias light indicate a clear
photovoltage. The inset shows injection efficiency as function of dependence on intensity, examples for two different cells are
conduction band edge position relative to the conduction band of the shown in Figure 5. Except for light intensities <0.01 sun, neither
TiCl4 treated cell with 0 M TBP and 0 M LiI. The cells with no TiCl4 cell (d ) 13 and 16 µm) is significantly limited by charge
treatment have not been corrected for any shift in the conduction band collection (for SE side illumination) since L > d. However, the
position caused by the treatment (estimated previously22 to be ∆Ec ≈ variation of L with intensity is interesting as the widely adopted
-80 mV).
multiple trapping model with the quasi-static approximation34
predicts L should be independent of electron concentration, and
3.2. Dependence of L and ηinj on Cell Performance. Fits thus light intensity, if recombination and transport only occurs
like those in Figure 3 were used to determine L and ηinj for a via conduction band electrons.
number of cells of similar thickness (d ) 13 µm) with differing This variation of L with incident flux and thus electron
concentrations of electrolyte additives (LiI and TBP) and concentration implies that either D0 and τ0 could vary with nc,
differing TiO2 preparations (with and without the TiCl4 treat- for example, Peter and Wijayantha35 suggested that the intensity
ment). The relationship between L and ηinj and cell performance dependence of L on nc could be related to the order of
under the solar simulator is shown in Figure 4. The cells all recombination kinetics. Alternatively the change in L observed
have a very similar ηLH; for d ) 13 µm typically 20% of the here could indicate an additional recombination route: a deep
photons in the AM1.5 spectrum (λ < 4000 nm) were harvested surface state mediated recombination process which becomes
by the dye molecules in the films. Thus the variation in jsc can saturated by relatively low light levels. This would imply an
be attributed to changes in ηinj and ηcol. additional recombination term is required in the continuity
A strong correlation can be seen between L and jsc for equation (6) and its solutions might not yield a global value of
different cells of the same thickness (d ) 13 µm) in Figure 4a. L with the conventional meaning as discussed in section 2.8.
Since ηcol (given by eq A10 for SE side illumination) becomes Apparently the additional loss process would no longer dominate
sensitive to L when L/d < 1, it is clear from Figure 4a that recombination at higher light levels where eq 6 would be
charge collection efficiency is the dominant factor influencing sufficient to describe the system. Similar suggestions have been
photocurrent in this series of cells. Thus a nontrivial fraction made based on the nonlinear response of photocurrent at low
of injected charge can be lost to recombination during transport light levels.36 Alternatively, although the mechanism is not clear,
even under short circuit conditions. This observation is of the observed increase in L could imply that the increase in Dn
particular importance for cells illuminated from the EE side achieved by filling traps using the bias light is not fully
where transport losses will be greater than those for SE side compensated by a corresponding reduction in τn.37 The variation
illumination. The variation in ηinj also contributed, to a lesser in L with intensity and hence n could also be related to nonideal
extent, to the observed differences in jsc. variation of Dn with n which can be observed in previous work
Figure 4b indicates a weak negative correlation between the involving photovoltage risetime measurements.38
injection efficiency and open circuit photovoltage, Voc. The inset The diffusion length in the 13 µm cell in Figure 5 peaks at
shows the ηinj as a function of ∆EC, the conduction band (or ∼0.2 suns and begins to fall at higher light intensities, perhaps
mobility band) position relative to that of the cell containing 0 indicating a limitation of transport or field build up in the
M TBP and 0 M Li with the TiCl4 treatment. ∆EC is estimated electrolyte phase at higher currents. This behavior was not
from charge extraction data assuming a similar distribution of apparent in all cells examined but is consistent with a previous
trapping states below the conduction band in all the cells.22 It study of IPCE as a function of light intensity.36 The IPCE data
is apparent that as the conduction band drops the ηinj improves, plotted against intensity presented by Halme et al.4 shows a
presumably because of improved coupling between the orbitals similar effect. They proposed that it may be explained by
of the dye LUMO and TiO2 conducting states, with the drop in enhanced electron recombination due to accumulation of I3- at
conduction band corresponding to an increase in the free energy the photoelectrode at higher light intensities.
1132 J. Phys. Chem. C, Vol. 113, No. 3, 2009 Barnes et al.

of ηinj < 100% are reasonable and consistent with the “slow”
injection times observed in TCSPC data.
3.5. The Influence of the TiCl4 Treatment and Compari-
son with L Derived from Transient Perturbation Methods.
To illustrate the usefulness of the method, we compared cells
prepared with and without the TiCl4 treatment. During electrode
fabrication, the treatment adds a thin layer of TiO2 to the surface
of the TiO2 nanoparticles prior to dyeing. The treatment often
leads to substantial improvements in the photocurrent generated
by cells, particularly when the cells perform poorly prior to the
treatment. There have been a number of studies examining the
influence of TiCl4; however, some doubt still remains regarding
the nature of the changes it causes.22,25 The fits shown in Figure
3a are on measurements of cells without and with the treatment;
the fitted diffusion lengths of 8 and 22 µm correspond to a
difference in jscAM1.5 photocurrents of 8.6 and 11.4 mA cm-2,
respectively. This indicates that the addition of a TiO2 layer to
the film has indeed increased the photocurrent and this is
correlated with the observed increase in diffusion length. In the
rest of this section we concentrate on a more extreme example
where jsc in the solar simulator at “one sun” increased from 5.2
to 8.9 mA cm-2 with the treatment; this pair of cells had no LiI
and no TBP in the electrolyte (see Table 1). Additional data
confirming the trends described below are shown in Figure S1
Figure 6. (a) Examples of photoluminescence decay curves measured in the Supporting Information. We now explore the mechanism
using time correlated single photon counting on a 4 µm TiO2 cell and for the increase in jsc in more detail using small perturbation
a 4 µm ZnO2 cell with electrolyte containing 0.5 M TBP and 0.1 M transient measurements and charge extraction data in addition
LiI. The injection yield derived from fitting emission rate functions is to analysis of the IPCEs of the cells. This also enables a
shown. (b) Photoluminescent injection yield (ηinj TCSPC) plotted against
injection efficiency inferred from IPCE measurements (ηinj IPCE). comparison of L determined by the two techniques.
The effective diffusion coefficient, Dn, at short circuit was
determined from measurements of τj, at different light intensities.
Any change in L resulting from the minor changes in the The mean concentration of electrons in the cells, n, was also
mean value of n that might be expected due variation in film determined, at each light intensity, using the short circuit charge
thickness are masked by the experimental noise in Figure 2. extraction technique. Examples of Dn plotted as a function of n
We also note that the cells examined in this study did not show (corrected for porosity) are shown in Figure 7a for a cell
significant changes in ηinj with varying bias light. prepared with and without the TiCl4 treatment to the TiO2. The
3.4. Comparison with Injection Efficiencies Determined values of Dn can be seen to increase with increasing electron
from Time Correlated Single Photon Counting. The injection concentration within the cell. This trend is expected and can be
efficiencies observed in Figure 4b varied from approximately explained by the multiple trapping model of the cell.34 As the
60 to 90% depending on the device’s electrolyte composition concentration of charge in the cell increases with increasing
and TiO2 and thus relative conduction band position (Figure light intensity and photocurrent, the proportion of mobile charge
4b). These values are substantially below the 100% often in the conduction band (or mobility band) increases relative to
assumed in analyses of DSSC device performance. Since ηinj the concentration of localized immobile charge in traps. This
measurements obtained by using IPCE are prone to errors leads to the increase in Dn with n measured.
introduced at the photon accounting step of the analysis Somewhat surprisingly, Figure 7a shows that for a given
(determining ηLH), it is useful to compare the derived values of charge concentration the addition of the TiCl4 treatment reduces
ηinj with measurements from an independent method. the value of Dn by a factor of approximately 3. This is in direct
Figure 6a shows examples of the transient photoemission contrast to a number of studies which have concluded the
decays of TiO2 and ZrO2 based cells determined by TCSPC. treatment increases the effective diffusion coefficient due to a
The injection half-time for the cell shown was found to be ∼200 supposed increase in the neck diameter between neighboring
ps, typical of the cells examined.9 The difference in area under particles (e.g., refs 26 and,39) although there has been little
the fitted decay functions gives an estimate of the injection yield satisfying evidence to support this hypothesis. The multiple
which is equivalent to injection efficiency under steady-state trapping model34 predicts that Dn is approximately proportional
conditions. A detailed analysis of injection dynamics and the to NL-γ where NL is the total concentration of trap states and γ
relationship with device performance and composition will be ≈ 3.4 in these cells and is related to the distribution of traps.
presented elsewhere.9 Here we focus here on a comparison ηinj Thus an increase in the mean density of traps by a factor of 1.4
from IPCE and TCSPC data. is required to account for the observed reduction in Dn. On this
A plot of injection efficiency determined from TCSPC basis, if we assume the mean concentration of trapping states
(ηinj TCSPC) against that determined using the IPCE method as a function of energy below the conduction band is 1.4 times
(ηinj IPCE) is shown in Figure 6b for cells with varying electrolyte higher after the TiCl4 treatment, then charge extraction measure-
composition. It indicates an acceptable correlation between the ments of the cells indicated that the conduction had been reduced
two entirely independent measurements. This agreement be- by ∼80 mV by the treatment. This is consistent with previous
tween the techniques helps to validate both methods for observations where the treatment appeared to shift the conduc-
determining injection efficiency. We can conclude that estimates tion band down by 80 mV in the case where NL was assumed
Injection and Diffusion Length in DSSCs from IPCE J. Phys. Chem. C, Vol. 113, No. 3, 2009 1133

TABLE 1: Measured (jsc AM1.5) and Calculated (jsc calc) Photocurrents for Two Cells (d ) 13.5 µm) Prepared with and without
the TiCl4 Treatment under “Front” and “Back” Illumination (SE and EE side)a
cell, j(sc AM1.5)
illumination [measd] LIPCE Ltrans ηLH ηcol ηcol jsc calc jsc calc
direction (mA cm-2) (µm) (µm) [AM1.5] ηinj [LIPCE] [Ltrans] [LIPCE] (mA cm-2) [Ltrans] (mA cm-2)
no TiCl4, SE 5.2 8.3 20 0.18 0.63 0.67 0.91 5.0 7.1
no TiCl4, EE 3.8 8.3 20 0.16 0.63 0.49 0.85 3.4 5.9
with TiCl4, SE 8.9 28 55 0.20 0.69 0.95 0.99 8.9 9.3
with TiCl4, EE 7.8 28 55 0.18 0.69 0.91 0.98 7.7 8.3
a
The values of L used to calculate the photocurrent with eq 8 (or 9) were determined either by analysis of IPCE data (LIPCE) or from transient
perturbation measurements (Ltrans). ηLH is the light harvesting efficiency weighted for the integrated photon flux of the AM1.5 spectrum (λ <
4000 nm). ηcol is the corresponding electron collection efficiency weighted over all incident wavelengths and is shown for both values of L.
The photocurrent, jsc calc, calculated using LIPCE analysis agrees well with the photocurrent measured using the solar simulator, jsc AM1.5.

cell. It has been well documented previously that τn decreases

as the concentration of electrons in the cell increases with
photovoltage.40 This can be explained by a greater fraction of
charge residing in the conduction band rather than trapped states
as the quasi-Fermi level rises. Hence proportionally less trapped
charge is available to buffer the relaxation of a small perturbation
in conducting electron concentration as the total concentration
of electrons increases.34
Figure 7b indicates that the TiCl4 treatment increases τn by
an order of magnitude or more for short circuit values of n, in
general agreement with previous observation.22 An increased
concentration of trapping states would be expected to increase
τn: for example if NL increased by 1.4 times then τn would be
expected to increase by a factor of about 3 since we expect τn
proportional to NLγ,34 this is not sufficient to explain the observed
increase in τn. Thus it seems likely that an additional process is
also involved. There are a number of possible mechanisms that
might explain the change. Charge transfer to the particle surface
could be slowed preferentially by an inhomogeneous distribution
of the additional trapping states (perhaps concentrated at the
interface between the original particles and new layer) or a
locally raised band edge might block recombination. Alterna-
tively, it is plausible that a proportion of recombination occurs
between electrons and iodine or triiodide bound to oxygen atoms
on the TiO2 surface. If the semiamorphous/nanocrystalline TiO2
coating formed by the treatment reduces iodine binding, perhaps
due to poorer charge transfer from Ti to O, then an increased
electron lifetime would be expected.
Diffusion lengths derived from these transient measurements
(Ltrans ) [Dn τn]1/2, see eq 7) as a function bias light flux (where
Figure 7. Diffusion coefficient, lifetime, and diffusion length measure- the corresponding value of n was determined by charge
ments for two cells with (squares) and without (circles) the TiCl4
treatment (d ) 13 µm, electrolyte containing 0.5 M TBP and 0 M LiI).
extraction) are shown in Figure 7c. As the photocurrent (and
(a) “Effective” electron diffusion coefficient as a function of the mean consequently n) increases, there is a corresponding increase in
electron concentration measured at short circuit. (b) Photovoltage Ltrans, similar to the general trend observed in other samples,
transient lifetimes as a function of the electron concentration in the e.g., Figure 5. The diffusion lengths with the TiCl4 treatment
cell. (c) Diffusion lengths derived from the IPCE measurements (LIPCE) are more than double those of nontreated values, so it is clear
and from the transient measurements in (a) and (b) (Ltrans, calculated that the reduction in the Dn resulting from the TiCl4 treatment
for matched n) as a function of light intensity.
is more than compensated for by the increased τn and leads to
the improved charge collection.
constant with and without the treatment since no change in Dn Comparison of Ltrans with L values derived using the IPCE
was observed.22 Thus it appears that the change in Dn can be method (labeled as LIPCE in Figure 7c) with the same samples
explained by a change in the concentration trapping states measured above indicates similar trends: LIPCE increases with
introduced by the treatment (which also appears to reduce the photocurrent and is at least three times greater for the TiCl4
conduction band position). If the treatment were to reduce the treated samples (the difference in ηinj was not significant).
mean trap state concentration instead of increasing it, then Dn However there is a substantial difference in the absolute
would be expected to increase; this was not seen with the cells magnitudes of the measurements; Ltrans values are approximately
in this study but could occur with other TiO2 preparations. 2-2.5 times greater than LIPCE for equivalent light intensities.
The photvoltage decay time constant, τn, is shown in Figure A similar difference between diffusion lengths measured using
7b as a function of n at open circuit; τn is related to the inverse the two different techniques was seen for most other cells
of the pseudo-first-order recombination rate constant within the examined in this study.
1134 J. Phys. Chem. C, Vol. 113, No. 3, 2009 Barnes et al.

Table 1 shows the measured photocurrent for two cells with The evidence presented in Table 1 and Figure 7, and the above
and without the TiCl4 treatment. Measured values are compared discussion suggests that absolute values of L derived from IPCE
to photocurrents calculated using the parameters L and ηinj measurement may be a more reliable indicator of cell behavior
according to than L derived from small perturbation transient or frequency
domain techniques which give τn and Dn, unless appropriate
jsc ) q ∫ φAM1.5(λ)ηSE(d, R, R1, λ, ηinj, L) dλ (8)
corrections are applied. The reason for the discrepancy is likely to
be related to the determination of Dn under short circuit conditions
particularly where substantial recombination occurs during trans-
jsc ) qΦηLHηinjηcol (9)
port; this question is currently under investigation.
where φAM1.5 is the photon flux of the AM1.5 spectrum and 4. Conclusion
ηSE is given by (A3). Equation 8 may be written as eq 9 using
We have demonstrated that meaningful values of injection
eq 1 and eqs A8, A9, and A12. Φ is the integrated photon flux
efficiency and diffusion length can be derived from careful IPCE
for the AM1.5 spectrum such that the “solar current density”
and optical measurements on dye-sensitized cells. The parameters
(for λ < 4000 nm, eq A9) is q Φ ) 68.3 mA cm-2, ηLH is the
correlated with device performance: increasing the diffusion length
mean light harvesting efficiency weighted by the intensity of
and injection efficiency led to higher short circuit current. Injection
the AM1.5 spectrum as a function of wavelength given by eq
efficiency was negatively correlated with open circuit voltage due
A10, and ηLH is the corresponding mean weighted collection
to higher conduction band positions limiting injection but increasing
efficiency for all injected electrons given by eq A12.
photovoltage. The diffusion length showed a clear dependence on
It is interesting to note that while the Ltrans values are
light intensity, implying the possibility of a saturatable surface-
substantially different for cells with and without the TiCl4
trap-mediated recombination route in the cells examined. This
treatment (Figure 7c), they cannot explain the difference in
observation has more general implications for IPCE measurements
photocurrents observed since in both cases Ltrans > d ) 13 µm
on DSSCs since the measured action spectra may not accurately
(see Table 1). A similar problem can be seen in previous work
reflect the IPCE of the cell under normal operating conditions where
where, using the transient data presented in ref 22, L ≈ 13 µm
light intensities are higher. This problem can be rectified by the
at 1 sun for the cell without TiCl4sthis accounts for only ∼2%
addition of a bias light.
collection losses in that cell (d ) 3 µm, ηLH ≈ 0.98). By
The IPCE derived injection efficiency values were validated
comparison the LIPCE values measured here can predict the
using an independent technique based on the transient photo-
measured photocurrents well (see Table 1); note that this is
emission measurements of complete cells. Comparison of the
expected since integrating a fit to the IPCE (with bias light)
IPCE diffusion lengths with values derived from small perturba-
convolved with the incident flux spectrum should yield jsc if
tion techniques indicated similar trends but a difference in the
the fit is good. It is worth emphasizing that the photocurrent
magnitude of the values. The transient diffusion length was
losses due to collection under EE side illumination are sub-
typically a factor of 2-2.5 times higher than that derived from
stantially larger than for SE side illumination even in the cell
the IPCE data. The difference is likely to be related to the
treated with TiCl4.
measurement of Dn at short circuit and suggests caution may
The previous study from this group observed either no change be required when interpreting transient-derived diffusion lengths.
or a decrease in transport time, τj at approximately the same n The diffusion length of electrons was shown to increase in
for TiCl4 treated devices relative to those with no treatment.22 TiO2 films treated with TiCl4; this change was attributed
Charge extraction measurements from that study suggested that primarily to an increase in the effective electron lifetime despite
the conduction band had been shifted down by ∼80 mV by the a decrease in the diffusion coefficient.
treatment. We might expect that this would lead to an increase One of the significant conclusions drawn from this work is
in ηinj as described in section 3.2 due to an increase the orbital the observation that the electron diffusion length can have a
overlap between the excited dye state and the population of substantial influence on device photocurrent at short circuit even
isoenergetic conduction band states. Analysis of the IPCE data when illuminating the TiO2 side of the cell. The architecture of
for the cells in the present study confirms a modest improvement DSSCs now being manufactured on metal substrates involves
in injection efficiency (Table 1), although this is insufficient to illumination through the counter electrode. In this configuration
explain the majority of the observed increase in jsc. Thus in the collection losses will be greater for a given L. Thus in this
contrast to the previous paper22 we find that recombination case, knowledge of L becomes important for optimizing cell
during transport does significantly reduce the photocurrent of materials and film thicknesses effectively.
untreated cells observed here (Table 1). However if we assume More generally it is hoped that the technique may prove useful
the approximate empirical relationship observed here LIPCE ≈ for assessing the limitations of new materials and devices,
Ltrans/2.5 then L ≈ 5 µm and ηLH ≈ 0.90 for the thin (d ) 3 particularly given the widespread availability of IPCE measure-
µm) cell discussed in ref 22. This would account for ap- ment facilities. In the future, the technique could also be
proximately half the loss in photocurrent relative to the TiCl4 extended to measure diffusion lengths at working potentials in
treated cell, the rest being attributable to drops in ηinj and ηLH; the cell by the application of a voltage bias during measurement.
this interpretation seems reasonable in light of the 20-fold
reduction in recombination rate that was observed.22 We can Acknowledgment. This work is supported by the EPSRC
conclude that the TiCl4 increases photocurrent by reducing the Materials for Energy (no. EP/E035175/1) and SUPERGEN
rate of recombination during electron collection, there is also a “Excitonic Solar Cells” programmes and the European Union
contribution from increased injection efficiency resulting from Robust DSC project (no. 212792). We thank Dyesol for the
lowering of the conduction band edge position relative to the provision of some of the materials used in this work and Hans
dye LUMO. For thick cells, a reduced rate of recombination of Desilvestro for his comments. We are grateful to Janne Halme
electrons during transport (increase in L) is the dominant effect and Laurie Peter for helpful discussions on this approach as
of TiCl4. well as the reviewers for their insightful comments.
Injection and Diffusion Length in DSSCs from IPCE J. Phys. Chem. C, Vol. 113, No. 3, 2009 1135

Appendix
The solution to eq 6 for conduction band electron concentration in the case of SE side illumination at short circuit as described
in the main text is given by

(1 - R)RτI0ηinje- x⁄L -(d+x)(R+RI)

{
( )

ncSE ) ed⁄L[e x⁄L +(d+x)(R+RI) + L(R + RI)ex(R+RI)] + e2x⁄L[e(d+x)(R+RI) - L(R + RI)e x⁄L +x(R+RI)] +
( ) ( )

(e2d⁄L + 1)[L2(R + RI)2 - 1]

e x⁄L [1 + e2d⁄L]
( )
[ nc0(L2(R + RI)2 - 1)e(d+x)(R+RI)
(1 - R)RτI0ηinj ]}
- ed(R+RI) (A1)

For EE side illumination the conduction band electron concentration is given by

(1 - R)TPtTIRτI0ηinje- x⁄L -d(R+RI)

{
( )

ncEE ) e2x⁄L[1 + L(R + RI)ex(R+RI)] + [ex(R+RI+ x⁄L ) - e2d⁄L + L(R + RI)ed(R+RI+ 1⁄L )] +
( ) ( )

(e2d⁄L
+ 1)[L (R + RI) - 1]
2 2

ex⁄L [ nc0(L2(R + RI)2 - 1)(1 + e2d⁄L)ed(R+RI)

(1 - R)(1 - APt)(1 - Ai)RτI0ηinj
(
]}
- ex(R+RI)+ 2d⁄L (A2)
)

The expression for the SE side IPCE is given by

(1 - R)LηinjRe-d(R+R1)[(L(R + R1) - 1)ed(R+R1+ 2⁄L ) + (L(R + R1) + 1)ed(R+R1) - 2L(R + R1)ed⁄L]

( )

nSE ) (A3)
(e2d⁄L + 1)[L2(R + R1)2 - 1]
The expression for the EE side IPCE is given by

(1 - R)TPtT1LηinjRe-d(R+RI)[(L(R + RI) + 1)e2d⁄L - (2L(R + RI)ed(R+RI+ 1⁄L ) + L(R + RI) - 1]

( )

nSE ) (A4)
(e2d⁄L + 1)[L2(R + RI)2 - 1]
The expression describing the ratio of the EE to SE side IPCEs is then given by

TPtTI[(L(R + RI) + 1)e2d⁄L - 2L(R + RI)ed(R+RI+ 1⁄L ) + L(R + RI) - 1]

( )
ηEE
) (A5)
ηSE (L(R + R ) - 1)ed(R+RI+(2⁄L)) + (L + (R + R ) + 1)ed(R+RI) - 2L(R + R )ed⁄L
I I I

In the case where RI ) 0, R ) 0, TPt ) 1, and TI ) 1, expressions A3, A4, and A5 become equivalent to expressions A11, A12, and
B1 in ref 4.
The light harvesting efficiency is given for SE and EE side illumination by integrating eqs 4 and 5 across the thickness of the
TiO2 to determine the fraction of light absorbed

(1 - R)R(1 - e-(R+RI)d)
ηLHSE(λ) ) (A6)
R + RI
(1 - R)TPtTIR(1 - e-(R+RI)d)
ηLHSE(λ) ) (A7)
R + RI
The total light harvesting efficiency, over all wavelengths, of a cell illuminated by a photon flux spectrum, φ(λ), is given by the
integrated convolution of the flux spectrum with the light harvesting efficiency

ηLH )
1
Φ
∫ φ(λ)ηLH(λ) dλ (A8)

where equations A6 or A7 are substituted into the ηLH(λ) for the required illumination direction and φ(λ) can be integrated to give
the total photon flux incident on the cell

Φ) ∫ φ(λ) dλ (A9)
The collection efficiency in the cell for SE and EE side illumination at a given λ is given by rearranging eq 1
ηSE(λ)
ηcolSE(λ) ) (A10)
ηLHSE(λ)ηinj(λ)
ηEE(λ)
ηcolEE(λ) ) (A11)
ηLHEE(λ)ηinj(λ)
The total collection efficiency for all electrons injected into a cell under AM1.5 illumination is given by
1136 J. Phys. Chem. C, Vol. 113, No. 3, 2009 Barnes et al.

ηcol )
1
ΦηLHηinj
∫ φ(λ)η(λ) dλ (A12)

and substituting eq A3 or A4 into the η(λ) term as required.

Supporting Information Available: Transient and IPCE (19) Qu, P.; Meyer, G. J. Langmuir 2001, 17, 6720.
derived L data for an additional pair of cells prepared with and (20) Xia, J. B.; Masaki, N.; Jiang, K. J.; Yanagida, S. J. Photochem.
Photobiol., A 2007, 188, 120.
without the TiCl4 treatment, confirming the trends shown in (21) Jennings, J. R.; Ghicov, A.; Peter, L. M.; Schmuki, P.; Walker,
Figure 7. This material is available free of charge via the Internet A. B. J. Am. Chem. Soc. 2008, 130, 13364.
at http://pubs.acs.org. (22) O’Regan, B. C.; Durrant, J. R.; Sommeling, P. M.; Bakker, N. J.
J. Phys. Chem. C 2007, 111, 14001.
References and Notes (23) Sodergren, S.; Hagfeldt, A.; Olsson, J.; Lindquist, S. E. J. Phys.
Chem. 1994, 98, 5552.
(1) Nazeeruddin, M. K.; Kay, A.; Rodicio, I.; Humphrybaker, R.; (24) Lindstrom, H.; Rensmo, H.; Sodergren, S.; Solbrand, A.; Lindquist,
Muller, E.; Liska, P.; Vlachopoulos, N.; Gratzel, M. J. Am. Chem. Soc. S. E. J. Phys. Chem. 1996, 100, 3084.
1993, 115, 6382. (25) Sommeling, P. M.; O’Regan, B. C.; Haswell, R. R.; Smit, H. J. P.;
(2) Chiba, Y.; Islam, A.; Watanabe, Y.; Komiya, R.; Koide, N.; Han, Bakker, N. J.; Smits, J. J. T.; Kroon, J. M.; van Roosmalen, J. A. M. J.
L. Y. Jpn. J. Appl. Phys. 2006, 45, L638. Phys. Chem. B 2006, 110, 19191.
(3) Ito, S.; Murakami, T. N.; Comte, P.; Liska, P.; Gratzel, C.; (26) Barbe, C. J.; Arendse, F.; Comte, P.; Jirousek, M.; Lenzmann, F.;
Nazeeruddin, M. K.; Gratzel, M. Thin Solid Films 2008, 516, 4613. Shklover, V.; Gratzel, M. J. Am. Ceram. Soc. 1997, 80, 3157.
(4) Halme, J.; Boschloo, G.; Hagfeldt, A.; Lund, P. J. Phys. Chem. C (27) Koops, S. E.; Durrant, J. R. Inorg. Chim. Acta 2008, 361, 663.
2008, 112, 5623. (28) van de Lagemaat, J.; Frank, A. J. J. Phys. Chem. B 2001, 105,
(5) Tachibana, Y.; Hara, K.; Sayama, K.; Arakawa, H. Chem. Mater. 11194.
2002, 14, 2527. (29) Nakade, S.; Kanzaki, T.; Wada, Y.; Yanagida, S. Langmuir 2005,
(6) Haque, S. A.; Palomares, E.; Cho, B. M.; Green, A. N. M.; Hirata, 21, 10803.
N.; Klug, D. R.; Durrant, J. R. J. Am. Chem. Soc. 2005, 127, 3456.
(30) Duffy, N. W.; Peter, L. M.; Rajapakse, R. M. G.; Wijayantha,
(7) Moser, J. E.; Gratzel, M. Chimia 1998, 52, 160.
K. G. U. Electrochem. Commun. 2000, 2, 658.
(8) Durrant, J. R.; Haque, S. A.; Palomares, E. Coord. Chem. ReV.
(31) Murphy, A. B. Appl. Opt. 2007, 46, 3133.
2004, 248, 1247.
(9) Koops, S. E.; O’Regan, B. C.; Barnes, P. R. F.; Durrant, J. R. J. Am. (32) Murphy, A. B.; Barnes, P. R. F.; Randeniya, L. K.; Plumb, I. C.;
Chem. Soc., Submitted. Grey, I. E.; Horne, M. D.; Glasscock, J. A. Int. J. Hydrogen Energy 2006,
(10) O’Regan, B.; Moser, J.; Anderson, M.; Gratzel, M. J. Phys. Chem. 31, 1999.
1990, 94, 8720. (33) Forro, L.; Chauvet, O.; Emin, D.; Zuppiroli, L.; Berger, H.; Levy,
(11) Ahn, K. S.; Kang, M. S.; Lee, J. W.; Kang, Y. S. J. Appl. Phys. F. J. Appl. Phys. 1994, 75, 633.
2007, 101. (34) Bisquert, J.; Vikhrenko, V. S. J. Phys. Chem. B 2004, 108, 2313.
(12) Fukai, Y.; Kondo, Y.; Mori, S.; Suzuki, E. Electrochem. Commun. (35) Peter, L. M.; Wijayantha, K. G. U. Electrochem. Commun. 1999,
2007, 9, 1439. 1, 576.
(13) Guo, L.; Liang, L. Y.; Chen, C.; Wang, M. T.; Kong, M. G.; Wang, (36) Trupke, T.; Wurfel, P.; Uhlendorf, I. J. Phys. Chem. B 2000, 104,
K. J. Acta Phys. Sin. 2007, 56, 4270. 11484.
(14) Hamann, T. W.; Martinson, A. B. F.; Elam, J. W.; Pellin, M. J.; (37) Boschloo, G. K.; Goossens, A. J. Phys. Chem. 1996, 100, 19489.
Hupp, J. T. J. Phys. Chem. C 2008, 112, 10303. (38) O’Regan, B. C.; Bakker, K.; Kroeze, J.; Smit, H.; Sommeling, P.;
(15) Heimer, T. A.; Heilweil, E. J.; Bignozzi, C. A.; Meyer, G. J. J. Durrant, J. R. J. Phys. Chem. B 2006, 110, 17155.
Phys. Chem. A 2000, 104, 4256. (39) Zeng, L. Y.; Dai, S. Y.; Wang, K. J.; Pan, X.; Shi, C. W.; Guo, L.
(16) Jennings, J. R.; Peter, L. M. J. Phys. Chem. C 2007, 111, 16100. Chin. Phys. Lett. 2004, 21, 1835.
(17) Kang, M. S.; Ahn, K. S.; Lee, J. W.; Kang, Y. S. J. Photochem. (40) Schlichthorl, G.; Huang, S. Y.; Sprague, J.; Frank, A. J. J. Phys.
Photobiol., A 2008, 195, 198. Chem. B 1997, 101, 8141.
(18) Madhwani, S.; Ameta, R.; Vardia, J.; Punjabi, P. B.; Sharma, V. K.
Energy Sources, Part A 2007, 29, 721. JP809046J