17278 J. Phys. Chem.

C 2009, 113, 17278–17290

Electron Lifetime in Dye-Sensitized Solar Cells: Theory and Interpretation of

Measurements

Juan Bisquert,* Francisco Fabregat-Santiago, Iván Mora-Seró, Germà Garcia-Belmonte, and

Sixto Giménez
PhotoVoltaic and Optoelectronic DeVices Group, Departament de Fı́sica, UniVersitat Jaume I, 12071 Castelló, Spain
ReceiVed: April 23, 2009; ReVised Manuscript ReceiVed: June 30, 2009

The electron lifetime τn in dye-sensitized solar cells (DSC) is a central quantity to determine the recombination
dynamics in the solar cell. It can be measured by several methods: impedance spectroscopy, IMVS, stepped
time transients, and open-circuit voltage decays. The paper aims at a better understanding of this fundamental
parameter. We summarize the main models that describe the lifetime dependence on bias voltage or carrier
density, and find that there are two complementary approaches to clarify the structure of the lifetime. The
first is to treat the lifetime as a product of the chemical capacitance and recombination resistance. This approach
is important because the resistance largely determines steady state operation characteristics of the solar cell
close to open-circuit voltage. The second approach is based on a kinetic model that describes in detail the
different processes governing the decay of the carrier population in a measurement of τn. The lifetime is
composed of a trapping factor and a free electron lifetime. Since the diffusion coefficient contains the reciprocal
of the trapping factor, it is found that a product (diffusion coefficient) × (lifetime) reveals the shape of the
free electron lifetime, which contains the essential information on kinetics of electron transfer at the surface
as a function of the position of the Fermi level. A model based on an exponential distribution of surface
states provides a good description of the voltage and temperature dependence of free electron lifetime and
diffusion lengths in high performance DSCs.

1. Introduction potential, recombination flow matches the photocurrent, and this

equilibrium determines the open circuit voltage Voc of the solar
Considerable research efforts have been devoted to the study
cell.
of mesoscopic dye-sensitized solar cells (DSC) since the seminal
demonstration1 of their feasibility as a low cost photovoltaic In a DSC or QDSC, transport of separated electron and hole
device. Charge generation on a DSC occurs on a molecular carriers in their respective media is intercepted by interfacial
sensitizer grafted on a nanostructured wide bandgap semicon- charge transfer. The understanding and material control of the
ductor. Upon excitation of the sensitizer, fast electron transfer kinetic barriers that impede recombination are key factors for
occurs to the electron transport material. The oxidized dye is improving the performance of these solar cells. Optical transient
regenerated from a hole transport material or redox electrolyte. spectroscopies have provided important insights in back electron
The two transport media are interpenetrated on a fine scale, transfer7 and dye regeneration dynamics.8 However, reasonably
allowing for a large internal interface for carrier generation. DSC efficient devices often contain a complex combination of
devices with ruthenium sensitizers and liquid electrolytes have materials and surface treatments. Pushing performance up
achieved demonstrated efficiencies over 11% at 1 full sun, and requires the in situ determination of the recombination flow in
remarkable stability has been obtained using ionic liquids. working devices.
Alternative configurations of nanostructured sensitized solar cells Due to the complex morphology of these devices, direct
are raising a lot of interest. Quantum dot sensitized solar cell measure of the recombination current does not provide sufficient
(QDSCs) performance is increasing rapidly,2,3 and the use of information on the underlying mechanisms, and special methods
ordered semiconductor architectures such as nanotubes4 could have been devised based on a small perturbation approach. The
improve the rate of electron transport and enhance diffusion solar cell is set in determined conditions of steady state by bias
lengths in DSC.5
voltage, illumination, temperature, etc., that remain constant in
The operation of solar cells is based on the competition of time. A small variation of a stimulus signal (harmonically
two basic processes.6 On the one hand, photoexcitation and modulated, or stepped in time) induces a small variation of the
charge separation (i.e., electron injection in TiO2 and dye response signal. Here “small” has a precise meaning: the
regeneration in a DSC7,8) creates a photocurrent, which is a flow response amplitude is linear to that of the signal. By these means,
of carriers toward the outer contacts. On the other hand, the quotient of both (the “transfer function”) is independent of
recombination is an opposite flow that internally annihilates the
the amplitudes, and thus is a unique property of the steady state
moving carriers.9 Recombination normally increases with the
conditions. In particular, the voltage-to-current transfer function,
carrier density and depends (usually in exponential function)
usually termed impedance spectroscopy (IS), has emerged as a
on the bias voltage V between the outer contacts. At a certain
major tool for DSC characterization.10-16 Measurements with
* To whom correspondence should be addressed. E-mail: IS technique provide a full picture of the processes determining
[email protected]. the operation of DSCs, and this allows for a deeper study of
10.1021/jp9037649 CCC: $40.75  XXXX American Chemical Society
Published on Web 08/18/2009
Feature Article J. Phys. Chem. C, Vol. 113, No. 40, 2009 17279

Juan Bisquert (BSc. Physics 1985, Ph.D. Physics 1992) is professor of Germà Garcia-Belmonte received his Ph.D. degree at Universidad
Applied Physics at Universitat Jaume I, Spain. His recent research activity Nacional de Educación a Distancia, 1996. Recently he follows research in
is focused on hybrid and organic photovoltaic devices, in particular dye- various topics within the field of organic electronics and photovoltaics as
sensitized solar cells. electronic mechanisms in organic lightemitting diodes and plastic and thin-
film solar cells. Device physics using impedance spectroscopy (including
modeling and measuring) is his main subject.

Francisco Fabregat-Santiago (BsC 1995, Ph.D. Physics 2001) is an

associate professor at Universitat Jaume I, Spain. He is an expert in electro-
optical characterization of devices and particularly known by his studies Sixto Giménez is curently Ramón y Cajal researcher at the Department of
on the electrical characteristics of nanocolloids, nanorods, nanotubes, dye Physics of the Universitat Jaume I de Castellón. His research is mainly
solar cells and electrochromic materials using impedance spectroscopy. focused on the manufacturing and electrochemical and optoelectronic
characterization of QDSC.

be obtained by IS14,19 and by small amplitude step time transient

decays.20-22 A widely used Voc decay method23,24 is based on
the reciprocal time derivative of the open-circuit voltage decay,
normalized to the thermal voltage

τn ) - ( )
kBT dVoc
q dt
-1
(1)

Here kB is Boltzmann’s constant, T is the temperature, and q

is the elementary charge. Another important quantity that gives
us insight in the recombination properties of DSCs is the
recombination resistance11,13
Iván Mora-Seró (M.Sc. Physics 1997, Ph.D. Physics 2004) is a researcher

( )
at Universitat Jaume I, Spain since 2002. He has worked on crystal growth
-1
and characterization of nanostructured devices, making both experimental 1 ∂jrec
and theoretical research. Nowadays he is focused on the study of charge Rrec ) (2)
generation and separation in quantum dots and the application in QDSCs.
A ∂V

specific aspects, avoiding the contamination of side effects in where jrec is the recombination current and A is the cell area.
these complex devices. The dominant method to obtain Rrec is IS.10-14
It has become popular to characterize recombination in a DSC Essential to good device performance is that the carrier
by means of a characteristic time constant that is usually called extraction time should be shorter than the recombination time.
the electron lifetime, τn. This can be measured by intensity Therefore many papers report the lifetime τn in comparison to
modulated photovoltage spectroscopy (IMVS),17,18 which is a the characteristic transport time τd, that relates to the electron
voltage-to-light transfer function. The electron lifetime can also diffusion coefficient, Dn, and active film thickness, L, as
17280 J. Phys. Chem. C, Vol. 113, No. 40, 2009 Bisquert et al.

L2 dn
τd ) (3) ) -Un(n) (7)
Dn dt

We identify a small portion of the population much smaller

A similar procedure is to compare the diffusion length than the average value, ∆n < <nj, and this decays by the law

Ln ) √Dnτn (4) d∆n

dt
)-
∂Ujn
( )
∂n nj
∆n (8)

with the film thickness. Since the diffusion length is the average
distance that a carrier diffuses before disppearing by recombina- Equation 8 is linear with ∆n and provides always an
tion, Ln > L indicates good collection efficiency. exponential decay that defines a lifetime in terms of the
It is usually thought that the electron lifetime in DSC reflects recombination rate Un(n) as
the basic kinetics of the recombination of electrons. In fact, we

( )jn
would ideally like to reduce the measured lifetime to a set of -1
∂U
microscopic parameters describing a capture probability, such τn ) (9)
as the formula for bulk homogeneous semiconductors ∂n nj

In the particular case that recombination rate is linear in

1
τn ) (5) electron density
VthσNt
Un ) krn (10)
where Vth is the thermal velocity of electrons, σ is the capture
cross section and Nt is the density of recombination states. the lifetime is a constant τ0 ) kr-1. However, in general eq 10
However, in hybrid solar cells made of a mixture of nanoscale is not the case, and τn in eq 9 depends on the steady state via
phases, as in semiconductors containing numerous defect the carrier density. Moreover, often it is necessary to expand
levels,25 the situation is generally quite complex. In fact τn may the kinetic eq 7 with trapping terms, in addition to the
represent a combination of processes that depend not only on recombination rate, see eq 20 below. Therefore the general
charge transfer kinetics but on several of the cell constituents. definition is
Despite hundreds of measurements of τn having been reported
in DSCs, our general understanding of the physical meaning
∆n
must be rated as limited, as this is connected to the complexity τn ) - (11)
of recombination in DSC. The standard redox carrier in DSC (d∆n/dt)
is iodine/iodide. The overall recombination reaction
The measurement of lifetime in a DSC basically consists of
a perturbation of the Fermi level that induces the recombination
I3- + 2e-(TiO2) f 3I- (6) by charge transfer. It means that the definition (11) applies to
those measurements in which the excess carriers ∆n are not
extracted at the contacts. In ref 23, it was shown that all the
consists of a multiple-step mechanism, very probably involving
methods mentioned in the Introduction to measure the lifetime
the species I2,9 and one of the steps will be rate determining.
in DSC provide the quantity in eq 11.
The oxidized species in the electrolyte may therefore be I2, and/
We mention a different approach, which is to define an
or I3-, and essential aspects of electron transfer are not yet
apparent lifetime from eq 7 as27
known in detail.26
Aiming at progress toward a microscopic understanding of
recombination by charge transfer in DSC devices, in this paper n n
τapp ) - )- (12)
we try to clarify the structure of the electron lifetime in DSC. (dn/dt) Un(n)
Our main strategy is to outline a general model for τn that
contains all of the main factors that intervene in the measurement It should be emphasized that eq 12 uses total density and
of τn. We also suggest to discern in the lifetime the two recombination rate instead of the small perturbation. However,
components that can be separately measured by IS: a resistance when Un(n) is nonlinear eq 7 does not give a characteristic time.
and a capacitance. And we take an additional step which is to In fact it is well-known that nonexponential decay contains a
identity the voltage dependence of a newly defined “free electron distribution of lifetimes.28 Thus τapp in eq 12 cannot be measured
lifetime”, that is directly related to the relatively small, but in general and must be obtained after some data treatment.
important, variations of the diffusion length that have been Equation 12 is commonly used in silicon solar cells, but there
reported. We will discuss a variety of representative measure- one can separately measure at steady state the minority carrier
ments to support the discussion. density by the photovoltage (or by several contactless methods,
such as photoconductance29) and the recombination rate (equal
2. Definition of the Lifetime
to photogeneration), so that τapp can be measured under some
The basic definition of the electron lifetime is quite simple conditions, especially at low injection levels, and theory exists
for a homogeneous system.23 We identify a given electronic to describe Un(n), even in the presence of defect levels or surface
state populated with an ensemble of electrons with volume recombination.30 This is very far from being the case in DSC.
density n. If the recombination rate per unit volume is Un, then An established physical picture for Un(n) in DSC is not available
the decay of a population is governed by the equation (as the following discussion will show), and even interpretation
Feature Article J. Phys. Chem. C, Vol. 113, No. 40, 2009 17281

It is useful to underline that the simple model of Figure 1b

can describe the essential operation of crystalline Si solar cells
as well as DSC. The reason is that both kinds of solar cells are
predominantly controlled by the variation of a single electronic
carrier, and no space-charge effects that introduce strong carrier
homogeneities or large electrical fields for transport need to be
considered.10,34 Although this basic model is used as a common
starting point, we will also highlight the main differences
between the two types of solar cells, that are mainly related to
strong energy disorder in a DSC.
It is not difficult to calculate the response of the basic model
of Figure 1b in IS, and the result35 is a parallel connection of
the recombination resistance Rrec and the chemical capacitance
Cµ as shown in Figure 1c. The lower value of Cµ, inherent in
an ideal model, is the chemical capacitance of the absorber.
However, in a DSC the dominant contribution to Cµ is the
electron transport material, the nanostructured TiO2 as indicated
in Figure 1a.36 The density of electrons in the conduction band
relates to the voltage as

nc ) n0e(EFn-EF0)/kBT ) n0eqVF/kBT (13)

Assuming the recombination process of electrons in the

semiconductor conduction band is given by eq 10, we have that
jrec ) qLUncb ) qLkrnc and thus
Figure 1. (a) Scheme of a DSC device. (b) Basic energy diagram

( ) [ ]
indicating voltage injection of electrons (1) and holes (2) and -1 k BT qVF
1 ∂jrec
recombination (3) processes in a DSC. Also shown are the transport Rrec ) ) nc-1 ) Rcb
0 exp -
energy levels and Fermi levels of electrons and holes, the potential V A ∂VF 2
ALq kr kBT
applied between the contacs, and the potential VF associated with (14)
separation of Fermi levels. (c) The basic equivalent circuit for ac
electrical perturbation. (d) Characteristic impedance spectra measured
on a liquid electrolyte DSC showing the RC arc, and a small Warburg The chemical capacitance for conduction band electrons is36
(diffusion) feature at high frequency.

of how n is distributed in the different kinds of electronic states

is not exempt of problems. We thus restrict the discussion to
Cµcb ) ALq
∂nc
∂VF
)
q2ALnc
kBT
) Cµ0
cb
exp
qVF
kBT [ ] (15)

quantities that can be measured in DSC. The small perturbation

lifetime, that always can be measured in itself, is therefore called Therefore the lifetime is
here simply “the lifetime”, as most authors have done so far.
The need to measure the small perturbation lifetime was
progressively recognized in amorphous silicon31 and crystalline 1
τ0 ) RrecCµcb ) (16)
silicon solar cells.32,33 kr

3. Structure of the Electron Lifetime and Simple Model

This characteristic time corresponds to the angular frequency
A useful theoretical approach to the electron lifetime, with at the top of the arc in Figure 1d, ωrec ) τ0-1 and an equivalent
important practical applications as discussed later on, is to relate result is obtained from IMVS. Taking a Laplace transform of
τn to equivalent circuit elements that can be separately measured the frequency response function, it is indeed shown that such
by IS. As introduction to this approach we use the basic model frequency corresponds to the time constant for exponential
of a solar cell described in Figure 1b, consisting on a light decay.23
absorber of thickness L and area A, where the Fermi levels of Equation 16 effectively shows the separation of lifetime in
electrons (EFn) and holes (EFp) are separately equilibrated to two main components. The rationale for this is that Rrec contains
selective contacts.6 An external bias voltage V can inject carriers both a density term, which is precisely the chemical capacitance,
and induce recombination, as shown in the figure, and cor- and a kinetic constant, so the product in eq 16 leaves only the
respondingly, promotion of carriers to the high energy level by latter.
photons produces a photovoltage. We point out in Figure 1b In Figure 2 we show the data of 16% efficient monocrystalline
the voltage VF ) (EFn - EF0)/q associated to the splitting of p-type silicon solar cell where such reciprocal dependence is
Fermi levels. In a solar cell VF is normally less than V due to realized.34 The chemical capacitance, associated to minority
series resistances and other elements that can be identified by carrier accumulation, obtained at forward bias in excess of VF
IS. We assume that the Fermi level of majority carrier holes, ) 0.5 V, and the recombination resistance, compensate to
remains at dark equilibrium level, EFp ) EF0. This is usually provide a nearly constant lifetime at different illumination levels,
the case in liquid electrolyte DSC where the redox energy Eredox Figure 2d (data values at VF < 0.4 V are discussed later on).
of I-/I3-, is not significantly modified during operation. Contactless methods to determine minority carrier lifetime have
17282 J. Phys. Chem. C, Vol. 113, No. 40, 2009 Bisquert et al.

Figure 3. Schematic representation of the steps involved in the

recombination between the electrons in TiO2 nanoparticles and the
oxidized species in the electrolyte (or hole conductor). EF0 is the position
of the Fermi level in the dark, equilibrated with the redox potential
Eredox of the acceptor species in solution, EFn is the Fermi level of
electrons, and Ec is the conduction band energy. The following processes
are indicated: Electron transport in the transport level; trapping and
release in an exponential density of localized states in the bulk of
semiconductor nanoparticles; capture and release by surface states;
electron transfer through conduction band with rate Ucbn ; electron transfer
through surface state with rate Uss n . On the right side we show the
fluctuating energy levels of oxidized species in solution according to
the Marcus-Gerischer model. λ is the reorganization energy of the
acceptor species in the ionic or hole transport material, with an effective
density of states D. Eox is the most probable energy level for the oxidized
state of the acceptor species.

νel of charge transfer at energy level E is given39,40 by the Marcus

model, outlined in the right-hand side of Figure 3

νel(E) ) 2k0cox


kBT
4πλ
exp - [
(E - Eredox - λ)2
4λkBT ] (17)

where k0 is a time constant for tunnelling, which is dependent

on the distance of the acceptor to the surface,41 and λ is the
reorganization energy.

4. Factors Governing a Variable Lifetime in DSC

In the presence of impurity levels, lifetimes in silicon become
highly variable42-44 and even more so in amorphous inorganic
semiconductors.25 The characteristic result in DSC consists of
a strong exponential decrease of τn with increasing carrier
density or bias voltage in the solar cell, as shown in Figure 4c,
and further discussed later.18,21,24
In order to provide a suitable framework of the interpretation
Figure 2. Experimental data for a Si monocrystalline solar cell. (a)
Current density-voltage characteristic, under dark and at different
of the lifetime, we consider in more detail a number of kinetic
illumination intensities. (b) Cell capacitance, indicating the regions of processes that are indicated in Figure 3.24,45,46 Free (conduction
potential where the depletion layer capacitance and the chemical band) electrons in TiO2 nanoparticles undergo trapping-detrap-
capacitance are separately observed. (c) Minority carrier recombination ping events in bandgap states. Such states can be directly
resistance. (d) Electron lifetime. measured by capacitance techniques.47 The results usually imply
an exponential distribution of localized states as described by
been used for many years in inorganic solar cells.29 The results the density of states (DOS)
are in good agreement with the IS method.33,37,38
In the denominator of eq 5 we have given the microscopic NL
expression of the kinetic rate constant, kr ) VthσNt, in a bulk g(E) ) exp[(E - Ec)/kBT0] (18)
kBT0
semiconductor. In a DSC, the corresponding expression is the
probability of a conduction band electron to be captured by the
oxidized species in the electrolyte with concentration cox (or Here Ec is the energy of the lower edge of the conduction
holes, in solid DSC). On fundamental grounds, the probability band, NL is the total density, and T0 is a parameter with
Feature Article J. Phys. Chem. C, Vol. 113, No. 40, 2009 17283

temperature units that determines the depth of the distribution,

which can also be expressed as a coefficient R ) T/T0. The
measured chemical capacitance is given by

Cµtraps ) q2g(EFn) ) Ctraps

0 exp[qVF /kBT0b] (19)

where the prefactor C0traps is independent of voltage. Most of

the traps may be located in the bulk of nanoparticles.
It should be emphasized that recombination is an interfacial
charge transfer event at the surface between the semiconductor
and the ionic/hole carrier. Since a distance for electron tunneling
should be of the order of 1 nm, which is usually much less
than the typical size of nanoparticles in a DSC, it is useful to
distinguish between bulk traps, with parameter T0b, and surface
states, with a parameter T0s, including in the latter class all the
electronic states that participate in charge transfer.
In any transient measurement for determination of lifetime,
a change of the Fermi level implies the change of occupation
of bulk traps. The kinetic eq 7 must be extended as

∂nc ∂nL
) -Un(nc) - (20)
∂t ∂t

where the last term describes the variation of free carrier density
by trapping (which causes an increase of the localized electrons
with density nL) and release processes. In eq 20, Un groups all
possible channels for recombination of free carriers by charge
transfer through different interfacial levels, as outlined in Figure
3. Taking into account the transfer from conduction band and
from a variety of surface states (at distinct energy levels labeled
E(i)), we may write for the rate of recombination of electrons
in the conduction band, a sum of the transference rates as
follows48

Un ) Ucb
n + ∑ Uss(i)
n (21)
E(i)

Equation 21 contains the assumption that the electron

transference occurs independently at the surface states in
different energy levels. In the case of of I-/I3- redox couple it
is not established that charge transfer is a one electron process,26
and we cannot rule out that different surface channels are
interdependent for reaction (6). On other hand, for solid state
DSC using OMeTAD hole conductor,14 recombination is a
process involving one electron and hole, very probably through
a variety of channels as in eq 21.49 In general it is very likely

TABLE 1: Properties of the Measured Dye-Sensitized Solar

Cellsa
sample L5 L9 L10 IL6 OMeTAD 4
Figure 4. Parameters result from IS in the dark, as a function of n-TiO2 layer (µm) 8.1 6.8 12 6.8 1.8
potential, in several DSCs, with the following characteristics: L5: 4.6% scatter layer (µm) 0 4 2 4 0
efficient, dye N3, electrolyte H10; L9: 9.6% efficient, dye K19, dye N3 K19 N719 K19 Z907
electrolyte Z325; IL6: 6.3% efficient, dye K19, ionic liquid Z594; electrolyte H10 Z325 Z300 Z594 solid
OMeTAD4: 4% efficient, dye Z907, solid hole conductor, see Table area (cm2) 0.48 0.28 0.28 0.28 0.128
1. (a) Chemical capacitance per unit volume. (b) Recombination porosity 70% 68% 68% 68% 68%
resistance per volume. (c) Electron lifetime obtained from IS data Voc (V) 0.58 0.76 0.80 0.71 0.86
(points) and Voc decays (lines). (d) Electron conductivity. For comparing Jsc (mA cm-2) 12.3 17 17.5 14.0 9.1
the results of the different samples, the values of potentials are referred FF 0.66 0.72 0.73 0.71 0.51
to the same distance between EFn and EC. Using the electron conductiv- efficiency (%) 4.6 9.6 10 6.3 4.0
ity, L9 is taken as reference sample to determine the shifts of voltage
obtaining: ∆VL5 ) 20 mV, ∆VIL6 ) 39 mV, and ∆VOMeTAD4 ) 226 a
Please refer to the original publication for the full names of the
mV. coded dyes and electrolytes.13,14,62
17284 J. Phys. Chem. C, Vol. 113, No. 40, 2009 Bisquert et al.

that, rather than a single energy level, a distribution of charge Therefore, trapping factors compensate and disappear in the
transfer states exists at the interface in DSC, and we adopt the diffusion length, eq 4
reasonable assumption in eq 21. This approach is supported by
steady state measurements,50 and important additional implica-
tions for interpretation of experimental measurements of lifetime Ln ) √D0τf (27)
are discussed below.
For the calculation of the lifetime from eq 20 we deal first Since the diffusion coefficient of the free electrons, D0, is
with the trapping term, assuming in quasistatic approximation believed to be constant,55 variation of Ln should be attributed
that trapped and free electrons remain in equilibrium, so that merely to the free electron lifetime τf. In experiments, the
we have46,51 electron diffusion length is found to moderately vary with
electron density.22,56-58 Such variation should be proportional
∂nL ∂nL ∂nc to τf1/2 according to the model of eq 27. The observed variations
) (22) of Ln are further discussed in section 8.
∂t ∂nc ∂t
5. RC Structure of the Variable Lifetime in Standard
Inserting eq 22 in 20, we obtain for a small perturbation DSC
We have pointed out in section 3 that the lifetime, τn,
∂∆nc 1 ∂Un measured in liquid electrolyte DSC, is related to the RC arc of
)-
( )
∆n (23) IS shown in Figure 1d. This was shown in eq 15 for a
∂t ∂nL ∂nc c
1+ preliminary simple model with a unique electronic level, and
∂nc we next discuss the structure of the lifetime quite generally,
including both the bulk trapping effect and the presence of
Equation 23 strongly suggests that we define the lifetime of several parallel channels for transference at the surface. We first
free carriers, applying eq 9, as give the general expressions of the model and thereafter we
suggest a specific trap distribution which allows us to describe

( ) ( )
-1 the voltage dependence of the electron lifetime and diffusion
∂Un -1 ∂Uss(i)
τf )
∂nc
) τ0 -1
+ ∑
n
∂nc
(24) length in DSC.
Calculation of the recombination resistance, eq 2, with the
E(i)
general charge transfer expression in eq 21, provides the result
Here, τ0 ) νel(Ec)-1 is the lifetime of free electrons by direct
transference through conduction band, the same as that given
in eq 16. Additional terms in eq 24 correspond to the
transference through the bandgap surface states as indicated in
Rrec-1 ) qLA ( ∂Ucb
n
∂nc
+ ∑
E(i)
n
)
∂Uss(i)
∂nc ∂VF
∂nc
(28)

Figure 3.
From eqs 23 and 24, using the general definition of eq 11, and from eqs 15 and 24 we obtain
we obtain the expression for the measured lifetime
τf ) RrecCµcb
( )
(29)
∂nL
τn ) 1 + τ (25)
∂nc f
Note that eq 29 contains the chemical capacitance of
conduction band electrons. However, as shown in Figure 4, the
In eq 25, the term in parentheses corresponds to the delay measured chemical capacitance is usually dominated by bulk
by traps at the Fermi level when we attempt to measure τn. More traps, eq 19. Therefore the lifetime from IS is
precisely, by detailed balance ∂nL/∂nc ) Cµtraps/Cµcb equals the
quotient of detrapping and trapping rates, which is a proportion
of the time an electron spends in traps with respect of the time τn ) RrecCµtraps (30)
in the conduction band. This term does not correspond to the
time survival of a free electron, which is given by τf. In general, We should also point out that eq 30 underlies all of the
τn, measured by the methods indicated in the Introduction, equivalent experimental methods for the measurements of
should be called a response time of the recombination process, lifetime in DSC.
but for simplicity we may still denote τn the “lifetime”, once Combining eqs 29 and 30 we find
the content of this quantity is properly understood. It has been
long recognized that trapping levels produce a considerable
delay in the response time with respect to the recombination Cµtraps
τn ) τf (31)
time τf.42,52 For a continuous distribution of traps, increasing Cµcb
the Fermi level progressively occupies deep traps, and a large
variation of the response time results.53
There is strong evidence that the trapping factor in eq 25 This is the same result as eq 25 (provided that ∂nL/∂nc . 1),
actually occurs in DSC. This is because the same factor applies showing the decomposition of τn in trapping and free carrier
in the measurement of the (chemical) diffusion coefficient46,54 lifetime terms. As mentioned above, this approach has received
ample attention in the literature,58 due to the general emphasis

( )
∂nL -1 on the calculation of the diffusion length. Equation 31 shows
Dn ) 1 + D0 (26) that the two approaches used in this paper to model the lifetime,
∂nc the kinetic model and RC approach, give the same result.
Feature Article J. Phys. Chem. C, Vol. 113, No. 40, 2009 17285

We are interested to have access to fundamental kinetic For an interpretation of the lifetime, leading toward the
parameters of recombination. The previous analysis shows that extraction of the true kinetic constants that govern recombination
Rrec contains a capacitance, and τn contains trapping terms. The in a DSC, we have two basic approaches.
closest we get to the recombination microscopic parameters, The first approach is given by eq 25 that states that, to a
using the small perturbation methods, is the free electron lifetime large extent, the bias variation of the lifetime in a DSC is an
τf as defined in eq 24. This parameter usually cannot be extracted unavoidable consequence of the disorder in the material. The
from experimental data using eq 29. But we note that eq 26 fact that there is an exponential distribution of traps in the
can be expressed55 nanostructured semiconductor, introduces an exponential de-
pendence in τn, via the trapping prefactor in eq 25. Besides, the
crucial parameter governing recombination by interfacial charge
Cµcb transfer, τf, may also show a potential dependence on its own,
Dn ) D0 (32) as we explain in detail below.
Cµtraps
At this point it is useful to revise the comparably simpler
results obtained in the silicion solar cell, shown in Figure 2.
The main difference between Figure 2 and Figure 3 is the slope
Therefore, combining eqs 31 and 32, we can write
of the chemical capacitances. For the silicon solar cell the slope
is q/kBT, which indicates that electron density increase is
Dnτn due to carriers in the conduction band. For the DSC, Figure 3a,
τf ) (33) the slope is much less steep, it is q/kBT0, or equivalently Rq/
D0 kBT, indicating that the charging with electrons occurs in the
exponential distribution of bandgap states. So for the silicon
solar cell, the recombination resistance is given by Rrec )
We can thus obtain the free electron lifetime as the product
(Cµcb)-1τ0 and the product RrecC provides a constant lifetime.
of experimental quantities: the diffusion coefficient and the
This analysis leads us to the second approach to discern the
measured lifetime. The denominator of eq 33 is a constant which
properties of τn in a DSC, which is via eq 30. We have already
is estimated as D0 ) 0.4 cm2 s-1.59 It has long been recognized
noted the slope of the traps capacitance, Ctraps
µ , observed in Figure
that the mobility-lifetime product relates to free carrier recom-
4a. As for the recombination resistance, it also shows an
bination time in amorphous silicon solar cells.60,61
exponential dependence of the form
6. Characteristic Experimental Results for DSC and

[ ]
Preliminary Interpretation qVF
Let us discuss the characteristic experimental behavior of τn
Rrec ) R0′ exp -β (34)
kBT
and Rrec in DSCs. Data sets representative of a variety of DSCs
with different types of electrolytes and hole conductor (including
two cells with volatile electrolyte with high and lower efficiency, Since the β parameter is typically in the range 0.5-0.7, we
one with ionic liquid and another one with the solid hole must conclude that recombination flux in a DSC cannot be
conductor OMeTAD) are shown in Figure 4 (Table 1).13,14,62 simply proportional to the free electron density, which is the
First of all we remark the exponential decrease of τn on bias case described above in eq 14, and a more complex process is
voltage, as observed in Figure 4c, which is a characteristic result involved. A specific model based on eq 28 will be described
that is very well established and routinely reported in papers.40,58,63 later on.
Furthermore good agreement is often found between the values We should remark in passing that eq 34 is an empirical
of electron lifetime determined by independent methods.63,64 This approximation that works well in restricted domains of bias
is confirmed in Figure 4c, that shows great coherence of the voltage. The observed dependence of recombination resistance
results obtained by IS and Voc decays. The consistency of the on bias may contain additional features, such as a curvature,
outcome of different techniques clearly indicates that τn is an and also a valley, i.e., a minimum of resistance, at low
experimentally well-defined quantity that calls for a more potential,65 as discussed later.
complete understanding, since it contains important information In any case, it clearly appears that in the DSC the two
on recombination at steady state. quantities in the product RrecC, depart from the ideal behavior
In general in DSCs with a resonable performance, the described in eqs 14 and 15. These quantities have different
photocurrent is basically determined by the absorption and slopes on voltage, that are determined by constants R and β.
injection properties of the dye and other light management As a consequence, and in contrast to the crystalline silicon solar
aspects of the solar cell. Near the maximum power point and cells, the lifetime τn in the DSC is not a constant and obtains
close to open-circuit conditions, is where the electron dynamics an exponential dependence on the bias, as shown in Figure 4c.
properties of the device have a more acute influence, via the Such variation can be attributed to the rather different features
recombination resistance, as mentioned before, Figure 4b. of the DSC in comparison with the bulk crystalline semiconduc-
Therefore a combined understanding of τn and Rrec, in the region tor. The preparation of DSC uses low temperature chemical
of voltage around Voc, is an important key to obtain DSC with routes, which provides a lower grade, and more disordered,
higher voltage and fill factor. Here we must highlight that the semiconductor. Further, the DSC is characterized by a hugue
steady state current density-potential curve of a DSC, can be internal interface where the recombination by charge transfer
entirely reconstructed from the knowledge of resistances occurs. At this point it may be useful to recall that a milestone
obtained by IS (by an integration procedure), provided that short- in the development of crystalline silicon photovoltaics was, and
circuit photocurrent is known.62 The most important piece of still is, the control of the main recombination interface, which
information to construct the JV curve and investigate the is the rear contact. In a DSC the recombination interface
performance of the solar cell is the data of recombination occupies by construction the whole active layer, and it is
resistance as a function of voltage. expedient to master the properties of charge transfer events in
17286 J. Phys. Chem. C, Vol. 113, No. 40, 2009 Bisquert et al.

the path toward higher efficiencies. It should be added that the 1 T

photovoltage in a DSC can double that in Si solar cell, implying
β) + Rs ) 0.5 + (39)
2 T0s
a higher driving force for recombination.

7. Lifetimes and Recombination Resistance for an If the bias voltage VF is considerably less than λ/q, eq 38
Exponential Distribution of Surface States simplifies to eq 34. The justification to assume an exponential
distrution of surface states instead of another type of distribution,
The main method of analysis of small perturbation parameters
is that eq 34 is in agreement with the experimental results, as
(recombination resistance, capacitance, lifetime, diffusion length)
shown in Figure 4c. Therefore, by the model in eq 37 we obtain
is to observe their dependence with the voltage as in Figures 2
a microscopic description of the β parameter that is determined
and 4. To calculate such dependencies from the general
in measurements of recombination in a DSC.
expressions defined previously, we assume a specific model
In eq 39 we observe that β parameter has two different
which has two main elements, as indicated in Figure 4:24 (1) an
components and these can be traced to eq 37. The recombination
exponential distribution of bulk trap states, with parameter T0b
resistance decreases exponentially with the voltage by two
(and Rb ) T/T0b), and (2) an exponential distribution of surface
reasons. The first is that the density of states participating in
states gss, with parameter T0s (Rs ) T/T0s).
charge transfer, gss, increases when the Fermi level raises. The
The reasons for distinguising bulk traps and surface states
second is the increase of the charge transfer probability, νel,
have been commented above. We have already mentioned that
with EFn. Eventually, νel may decrease at higher voltage,
the chemical capacitance provides a direct measure of the DOS,
provided that Marcus inverted region is reached. In this situation
but it is a spatial average in which the more abundant states
eq 34 is not valid and we should use a more complete expression
will dominate. Therefore in principle Ctraps
µ in eq 19 corresponds
such as eq 38.
to bulk traps located in the interior of the metal oxide
Importantly, the two components of the recombination
nanoparticles, and a different approach is required to determine
resistance have a different behavior with temperature, since gss
the crucial parameter involved in interfacial charge transfer, Rs
should depend very weakly on T, while νel is thermally activated
) T/T0s. However, if there is evidence that interfacial traps are
as indicated in eq 17. This observation forms the basis for a
more abundant, or that the distribution is the same, the model
method to obtain Rs ) T/T0s that is discussed later. Temperature
can be simplified at any stage by writing simply Rs ) Rb ) R.
of the DSC appears as a critical variable in order to discern the
The probability of charge transfer is described by eq 17.
microscopic components of the recombination parameter β.
Therefore, we have that the rate of charge transfer in an interval
Implicit in the previous modeling of Rrec is the assumption
∆E at energy E is given by
that surface states are occupied according to the bulk Fermi
level. In general, the occupation of surface states, which loose
Uss(i) ) gss(Ei)f(Ei, EFn)νel(Ei)∆E (35) charge by interfacial charge transfer, can be much less than a
n
bulk bandgap state at the same energy level.65 However, if the
transfer rate is not too large, as expected in good quality DSC,
where f(E,EFn) is the Fermi-Dirac function. The calculation of a common equilibrium can be assumed. In quantitative terms,
Rrec in eq 28 with a sum (integral) over the surface state levels a demarcation level can be defined, above which level surface
is states occupation is governed by EFn. An estimation of the
demarcation level in a DSC indicates that it lies quite deep,

( )
0.30 eV above the dark Fermi level.48 Therefore the equilibrium
df(Ei, EFn) -1
Rss
rec
-1
) qAL ∑ gss(Ei) dVF
νel(Ei)∆E assumption in the model of eq 37 is well supported.
Having obtained a complete description of the recombination
E(i)
(36) resistance dependence on voltage in microsocopic terms, we
now use the model to derive also the potential dependence of
the different quantities that can be measured with the small
The sum (or integration) in eq 36 gives the result13 perturbation methods.
Using eqs 19, 30, and 39, we obtain an expression of the
-1 potential dependence of the lifetime
Rss
rec ) q2ALgss(EFn)νel(EFn) (37)

Equation 37 states that the reciprocal charge-transfer resis-

tance is proportional to the product of the density of surface
τn ) R0′Ctraps
0 [
exp -γn
qVF
kBT ] (40)

states at the Fermi level, and the probability of electron transfer

from such states. This result occurs because the resistance is a where
differential quantity corresponding to the current gained by a
small step of voltage. A small displacement of the Fermi level
fills the surface states precisely at the Fermi level, hence the 1
γn ) β - Rb ) + Rs - Rb (41)
resistance detects only those states. 2
Rearrangement of eq 37 leads to13
Note that γn ) 0.5 if Rs ≈ Rb. On another hand, from eq 29,

[( ) ]
qVF qVF the free carrier lifetime can be stated as
Rss
rec ) R0′ exp - β - (38)
4λ kBT

with R0′ a voltage-independent prefactor and

τf ) R0′Ccb
0 exp γf[ ] qVF
kBT
(42)
Feature Article J. Phys. Chem. C, Vol. 113, No. 40, 2009 17287

where

1
γf ) 1 - β ) - Rs (43)
2

and from eq 27, the voltage dependence of the diffusion length

is

Ln ) B exp [ ]
γf qVF
2 kBT
(44)

8. Discussion and Applications

The fundamental problem of understanding recombination
mechanisms in a DSC is the interpretation of the exponent β of
recombination resistance. The model suggested above, by a
combination of exponential distribution of surface states, and
Marcus probabilities of transfer, explains the main observed
trends in the measurement, as already discussed, and provides
two additional predictions.13 First, log Rrec(VF) is not perfectly
straight but shows a curvature over a sufficient wide bias range,
eq 37, and second, β depends on temperature as indicated in eq
39. At the present time, a quantitative confirmation of this model
is not generally supported, since the data at different temper-
atures in high performance DSCs are scarce.13 However, both
characteristics (the curvature, and temperature dependence of
β) will be probably confirmed in general.
In the results compared in Figure 4 for DSCs with different
types of electrolytes and dyes, it is observed that the higher
efficiency reached by some of the cells is explained by the
combined effect of both a higher charge transfer resistance and
the higher charge injection from the dye, as suggested by the
short circuit current in Table 1. In particular the solid hole
conductor OMeTAD cell displays the lowest charge transfer
resistance, which reduces considerably the performance although
this type of DSC attains higher photovoltage, Figure 4d.14,66
An interpretation of the electron lifetime in DSCs requires
an understanding of the exponent γn in eq 40. In eq 41, we
note that γn arises from the β exponent of charge transfer
resistance and Rb for bulk traps. Equation 34 is strongly
supported by recombination data,9 and the chemical capacitance
of traps is also a well established result;47 therefore, γn ) β -
Rb is a reliable result.
It is also interesting is to address observed properties of the
diffusion length in DSCs. We note first that the diffusion length
is not a transient quantity, but a steady state parameter, and
consequently can be expressed in terms of the IS resistances as
Ln ) L(Rrec/Rt)1/2, being Rt the transport resistance.10 Therefore,
bulk traps do not intervene in the diffusion length.
According to the model, Ln is proportional to the square root
of the free carrier lifetime. Diffusion length in DSC usually Figure 5. Parameters result from IS in the dark in several DSCs: a
10% efficient DSC (L10, dye N719, electrolyte Z300) at several
increases slightly at increasing bias voltage.56,58,67 Characteristic
temperatures and in a 4.6% efficient DSC (L5, dye N3, electrolyte H10).
voltage dependence of Ln both for high efficiency (10%)13 and (a) Electron diffusion length (Ln )
Dnτn) as a function of Fermi level
regular efficiency (5%)11 DSCs with liquid electrolyte are shown potential. (b) The free electron lifetime τf, calculated from diffusion
in Figure 5a. A similar temperature variation of Ln has been coefficient-lifetime product, indicating a best fit to a straight line that
reported recently in a 11.4% efficient DSC.16 Since characteristic corresponds to eq 43. (c) The coefficient γf, defined in eq 42, as a
β values are about 0.7, the coefficient governing the variation function of temperature.
of Ln, following the above theory, (1 - β)/2, is indeed very
small, and this provides preliminary support of the model. In tance techniques measure T0b which usually ranges between 800
addition, according to eq 43 we have γf ) 0.5 - Rs with Rs ) and 1200 K. Nonetheless it should be remarked that near room
T/T0s. Therefore the exact value of T0s and the temperature of temperature we expect that γf ) 0.5 - Rs is positive but very
the solar cell become critical issues. Unfortunately T0s cannot close to 0, so that an increase of the temperature may invert
be independently determined. As mentioned before, the capaci- the tendency of Ln, from increasing to decreasing with the
17288 J. Phys. Chem. C, Vol. 113, No. 40, 2009 Bisquert et al.

voltage. In fact for OMeTAD DSC, the diffusion length is

observed to increase with voltage in the dark, while a contrary
tendency is observed under illumination: it decreases with the
potential,14,59,66,68 and we may tentatively attribute this to heating
of the solar cell.
Let us focus our attention on the new quantity that has been
introduced above, the free carrier lifetime τf. As mentioned
before, this is an important quantity that contains essential
information on the recombination kinetics. In Figure 5b we plot
the diffusion coefficient-lifetime products for the 10% and 5%
efficiency DSCs as a function of voltage. From the fits to straight
lines we derive the γf exponent that is plotted in Figure 5c.
First we observe that for the 10% efficient DSC, γf shows a
linear decrease with the temperature in good agreement with
the model prediction in eq 43. A second and important result is
that γf value decreases considerably for the less efficient solar
cell, at room temperature. Using the model results we obtain
that the parameter of the surface state distribution is T0s ) 1770
K for the 10% efficient DSC while it is T0s ) 750 K for the 5%
efficient DSC. It means that the surface states in the more
efficient DSC are much more concentrated near the conduction
band edge, while in the less efficient cell the charge transfer
states extend deeper in the bandgap. Therefore obtaining a large
T0s, which concentrates the recombination at higher voltages,
could be an important reason why high performance DSC
provide a large photovoltage. This interpretation illustrates the
significance of the control of recombination parameters for
improving the actual performance of DSCs, but must regarded
as speculative until wider sets of data become available.
Nonetheless we emphasize the importance of analyzing the free
carrier lifetime under variation of temperature.
An important conclusion of these considerations is that
exponents governing electron lifetime and free carrier lifetime
variation with bias voltage, are very sensitive to temperature.
Temperature has seldom been controlled in the measurements
of recombination,13,16,69 and even in these cases, it is not the
ambient temperature that matters, but the internal temperature
of the solar cell. In order to stabilize the measurement at
illumination, DSC is irradiated during minutes or more and
becomes internally heated. It may be difficult to control this
problem in a device that is double glass sealed, but improve-
ments of experimental methods are clearly necessary to progress
in the understanding and control of recombination.
In the model of an exponential distribution of surface states,
the recombination resistance varies continuously with the bias
voltage, since more and more surface states become available
for charge transfer as the Fermi level raises, eq 37. Therefore,
one should recognize that the resistance in eq 34 does not
provide a manifest evidence of the actual presence of the surface
states. The situation is different for a monoenergetic, deep
surface state.48,50 In this case, the resistance first decreases, when
the Fermi level approaches the surface state level Ess, but when Figure 6. Experimental data of QDSC (polysulfide aqueous electrolyte)
EFn > Ess, the rate of trapping decreases and the resistance formed by mesoscopic TiO2 sensitized with colloidal CdSe, using
increases.65 Figure 6 shows very direct evidence for charge cysteine as a QD linker molecule, and coated with ZnS. (a) Cyclic
transfer through a surface state in a QDSC with aqueous voltammetry at different scan rates. (b) Current density-voltage
polysulfide electrolyte. The valley in Rrec is accompanied by a characteristic, under dark and under full 1 sun illumination. (c) Charge
transfer resistance and chemical capacitance. (d) Electron lifetime.
peak of the capacitance Cµ, Figure 6c, but the latter peak due
to surface states is damped by the background capacitance of
the exponential distribution of the bulk states, eq 19. As a result, However, it must be taken into account that τn in Figure 6c
the lifetime in Figure 6d displays a valley at low voltage that is results from a stationary measurement, while the current-potential
undoubtedly associated to the surface state. This effect causes data are obtained at a certain velocity of voltage scan. Kinetic
an S shape in the current-potential (j-V) curve as shown in effects progressively displace the response of the surface state
Figure 6b. It is also observed that the depression in the j-V to higher voltage, as shown by voltammetry data in the dark in
curve appears at higher voltage than in the case of the lifetime. Figure 6a.
Feature Article J. Phys. Chem. C, Vol. 113, No. 40, 2009 17289

We have argued that the response time for recombination, 9. Conclusions

generally called the “lifetime”, always is composed of a
The electron lifetime in dye-sensitized solar cells can be
recombination resistance and a capacitance, and this is very
measured by several methods that usually give coherent results.
directly revealed by IS measurement, as illustrated in Figure
We have highlighted the approach to the lifetime suggested by
1d. However, the time constant of the recombination arc of IS
impedance spectroscopy, that shows that τn is a product of a
must be written, in general, as follows
chemical capacitance and recombination resistance. The model-
ing of lifetime in DSC shows two main components: a trapping
τn ) RrecC (45) factor and the free electron lifetime, τf. The latter quantity, τf,
is shown to reflect the basic kinetics of recombination of free
electrons, and depends on the bias when charge transfer at the
Here we write C for the capacitance, which may not be metal oxide/electrolyte interface is governed by surface states.
chemical in origin. The nature of the capacitance is a very The diffusion coefficient contains the reciprocal of the trapping
important issue in order to properly interpret τn as a lifetime. factor, therefore from a product Dnτn the potential dependence
The point is very familiar in silicon solar cells. In certain of free electron lifetime can be visualized. The temperature of
conditions, the transient measurement produces a charging or the solar cell exerts an important influence on the coefficients
discharging of the depletion layer, and then the measured that describe the voltage dependence of the lifetime. A model
“lifetime” has no relationship to recombination time at all.70-72 based on the transference through an exponential distribution
This is clearly illustrated in Figure 2d: The values of τn at VF < of surface states is found to provide a good description of the
0.4 V correspond to depletion capacitance in Figure 2c and observed exponential dependence of the recombination resis-
cannot be associated to the minority carrier lifetime. tance with the voltage, with a parameter β. The model also
This effect may be found also in some classes of nanostruc- accounts for the temperature variations of the free electron
tured hybrid solar cells. For example it has been shown that lifetime and electron diffusion length in high performance dye
ZnO nanowire arrays prepared by electrodeposition show a very solar cells.
high level of doping with a depletion region in the surface.15,73,74
Acknowledgment. The work was supported by Ministerio
IS of DSCs prepared with such arrays will provide the
de Ciencia e Innovación of Spain under projects HOPE
impedance pattern of Figure 1d, but the time constant of the
CSD2007-00007 and MAT2007-62982 and Generalitat Valen-
recombination arc cannot be interpreted as a recombination
ciana under project PROMETEO/2009/058.
lifetime, since it is simply a product as indicated in eq 45, where
C is a surface depletion capacitance.
Another valuable aspect of the R-C approach to the electron References and Notes
lifetime is the control of experimental artifacts. In fact, very (1) O’ Regan, B.; Grätzel, M. Nature 1991, 353, 737.
often the recombination arc does not singly dominate the (2) Shen, Q.; Kobayashi, J.; Diguna, L. J.; Toyoda, T. J. Appl. Phys.
impedance spectra, a significance portion of the capacitance may 2008, 103, 084304.
be due to counterelectrode or electrolyte diffusion contribu- (3) Kamat, P. V. J. Phys. Chem. C 2008, 112, 18737.
(4) Shankar, K.; Bandara, J.; Paulose, M.; Wietasch, H.; Varghese,
tions.62 Time transient decays will capture such capacitances O. K.; Mor, G. K.; LaTempa, T. J.; Thelakkat, M.; Grimes, C. A. Nano
rendering the measurement of recombination time devoid of any Lett. 2008, 8, 1654.
value. While the IS measurement allows to separately recognize (5) Martinson, A. B. F.; Goes, M. S.; Fabregat-Santiago, F.; Bisquert,
J.; Pellin, M. J.; Hupp, J. T. J. Phys. Chem. A 2009.
the recombination arc from which τn may be safely extracted. (6) Bisquert, J.; Cahen, D.; Rühle, S.; Hodes, G.; Zaban, A. J. Phys.
It should be emphasized that in determination of electron Chem. B 2004, 108, 8106.
lifetime in DSCs it is also important to check that the (7) Tachibana, Y.; Moser, J. E.; Grätzel, M.; Klug, D. R.; Durrant,
recombination resistance is in fact associated with interfacial J. R. J. Phys. Chem. 1996, 100, 200556.
(8) Clifford, J. N.; Palomares, E.; Nazeeruddin, M. K.; Grätzel, M.;
charge transfer at the nanostructured metal oxide, as in some Durrant, J. R. J. Phys. Chem. C 2007, 111, 6561.
cases recombination flux is dominated by charge transfer at the (9) Huang, S. Y.; Schilchthörl, G.; Nozik, A. J.; Grätzel, M.; Frank,
exposed substrate.75 It is generally accepted that this effect A. J. J. Phys. Chem. B 1997, 101, 2576.
(10) Bisquert, J. J. Phys. Chem. B 2002, 106, 325.
occurs at low voltage, and it can be neglected close to open- (11) Fabregat-Santiago, F.; Bisquert, J.; Garcia-Belmonte, G.; Boschloo,
circuit voltage. G.; Hagfeldt, A. Sol. Energy Mater. Sol. Cells 2005, 87, 117.
Finally, we comment briefly on the determination of carrier (12) Wang, Q.; Moser, J.-E.; Grätzel, M. J. Phys. Chem. B 2005, 109,
lifetimes in organic bulk heterojunction (BHJ) devices. Several 14945.
(13) Wang, Q.; Ito, S.; Grätzel, M.; Fabregat-Santiago, F.; Mora-Seró,
techniques have been applied to this end,76 including modulated I.; Bisquert, J.; Bessho, T.; Imai, H. J. Phys. Chem. B 2006, 110, 19406.
photoinducedabsorption,77 transientabsorption,78 photo-CELIV,79,80 (14) Fabregat-Santiago, F.; Bisquert, J.; Cevey, L.; Chen, P.; Wang, M.;
double-injection currents,81 and time-of-flight methods.82 How- Zakeeruddin, S. M.; Grätzel, M. J. Am. Chem. Soc. 2009, 131, 558.
(15) He, C.; Zheng, Z.; Tang, H.; Zhao, L.; Lu, F. J. Phys. Chem. C
ever many of these methods often depart from steady state 2009, 113, 10322.
operation of the solar cell. Some recent papers have determined (16) Cao, Y.; Bai, Y.; Yu, Q.; Cheng, Y.; Liu, S.; Shui, D.; Gao, F.;
the lifetime by small-amplitude perturbation of a steady state, Wang, P. J. Phys. Chem. C 2009, 113 ASAP.
using transient photovoltage83 and IS.84 In these works, it was (17) Schlichthörl, G.; Huang, S. Y.; Sprague, J.; Frank, A. J. J. Phys.
Chem. B 1997, 101, 8141.
observed that when the Fermi level of electrons increases, then (18) Okermann, T.; Zhang, D.; Yoshida, T.; Minoura, H. J. Phys. Chem.
(i) the lifetime decreases and (ii) the capacitance increases.83,84 B 2004, 108, 2227.
However, as generally remarked above, for a proper interpreta- (19) Fabregat-Santiago, F.; Garcı́a-Cañadas, J.; Palomares, E.; Clifford,
J. N.; Haque, S. A.; Durrant, J. R.; Garcia-Belmonte, G.; Bisquert, J. J. Appl.
tion of the lifetime in BHJ, it is necessary to simultaneously Phys. 2004, 96, 6903.
measure the chemical capacitance, which provides access to the (20) O’Regan, B. C.; Scully, S.; Mayer, A. C.; Palomares, E.; Durrant,
DOS of the electron transporting phase,47,84 over a sufficiently J. R. J. Phys. Chem. B 2005, 109, 4616.
wide range of variation of the Fermi level. More work in this (21) Nakade, S.; Kanzaki, T.; Wada, Y.; Yanagida, S. Langmuir 2005,
21, 10803.
direction is needed to clarify the interpretation of variations of (22) Nissfolk, J.; Fredin, K.; Hagfeldt, A.; Boschloo, G. J. Phys. Chem.
the lifetime in BHJs. B 2006, 110, 17715.
17290 J. Phys. Chem. C, Vol. 113, No. 40, 2009 Bisquert et al.

(23) Zaban, A.; Greenshtein, M.; Bisquert, J. ChemPhysChem 2003, 4, (55) Bisquert, J. Phys. Chem. Chem. Phys. 2008, 10, 3175.
859. (56) Fisher, A. C.; Peter, L. M.; Ponomarev, E. A.; Walker, A. B.;
(24) Bisquert, J.; Zaban, A.; Greenshtein, M.; Mora-Seró, I. J. Am. Chem. Wijayantha, K. G. U. J. Phys. Chem. B 2000, 104, 949.
Soc. 2004, 126, 13550. (57) Krüger, J.; Plass, R.; Grätzel, M.; Cameron, P. J.; Peter, L. M. J.
(25) Shen, D. S.; Conde, J. P.; Chu, V.; Wagner, S. J. Appl. Phys. 1994, Phys. Chem. B 2003, 107, 7536.
75, 7349. (58) Peter, L. M.; Dunn, H. K. J. Phys. Chem. C 2009, 113, 4726.
(26) Peter, L. M. J. Phys. Chem. C 2007, 111, 6601. (59) Jennings, J. R.; Peter, L. M. J. Phys. Chem. C 2007, 111, 16100.
(27) Hornbeck, J. A.; Haynes, J. R. Phys. ReV. 1955, 97, 311. (60) Street, R. A. Appl. Phys. Lett. 1982, 41, 1060.
(28) Alcala, J. R.; Gratton, E.; Prendergast, F. G. Biophys. J. 1987, 51, (61) Crandall, R. S.; Balberg, I. Appl. Phys. Lett. 1991, 58, 508.
597. (62) Fabregat-Santiago, F.; Bisquert, J.; Palomares, E.; Otero, L.; Kuang,
(29) Sinton, R. A.; Cuevas, A. Appl. Phys. Lett. 1996, 69, 2510. D.; Zakeeruddin, S. M.; Grätzel, M. J. Phys. Chem. C 2007, 111, 6550.
(30) Macdonald, D.; Kerr, M.; Cuevas, A. Appl. Phys. Lett. 1999, 75, (63) Miettunen, K.; Halme, J.; Toivola, M.; Lund, P. J. Phys. Chem. C
1571. 2008, 112, 4011.
(31) Ritter, D.; Zeldov, E.; Weiser, K. Phys. ReV. B 1988, 38, 8296. (64) Miettunen, K.; Halme, J.; Lund, P. J. Phys. Chem. C 2009, 113
(32) Brendel, R. Appl. Phys. A: Mater. Sci. Process. 1995, 60, 523. ASAP.
(33) Schmidt, J. IEEE Trans. Electron DeVices 1999, 46, 2018. (65) Mora-Seró, I.; Bisquert, J. Nano Lett. 2003, 3, 945.
(34) Mora-Seró, I.; Garcia-Belmonte, G.; Boix, P. P.; Vázquez, M. A.; (66) Wang, M.; Chen, P.; Humphry-Baker, R.; Zakeeruddin, S. M.;
Bisquert, J. Energy EnViron. Sci. 2009, 2, 678. Grätzel, M. ChemPhysChem 2009, 10, 290.
(35) Mora-Seró, I.; Bisquert, J.; Fabregat-Santiago, F.; Garcia-Belmonte, (67) Barnes, P. R. F.; Anderson, A. Y.; Koops, S.; Durrant, J. R.;
G.; Zoppi, G.; Durose, K.; Proskuryakov, Y. Y.; Oja, I.; Belaidi, A.; Dittrich, O’Regan, B. J. Phys. Chem. C 2009, 113, 1126.
T.; Tena-Zaera, R.; Katty, A.; Lévy-Clement, C.; Barrioz, V.; Irvine, S. J. C. (68) Snaith, H. J.; Grätzel, M. AdV. Mater. 2007, 19, 3643.
Nano Lett. 2006, 6, 640. (69) O’Regan, B.; Durrant, J. R. J. Phys. Chem. B 2006, 110, 8544.
(36) Bisquert, J. Phys. Chem. Chem. Phys. 2003, 5, 5360. (70) Castañer, L.; Vilamajo, E.; Llaberia, J.; Garrido, J. J. Phys. D: Appl.
(37) Garcia-Belmonte, G.; Garcı́a-Cañadas, J.; Mora-Seró, I.; Bisquert, Phys. 1981, 14, 1867.
J.; Voz, C.; Puigdollers, J.; Alcubilla, R. Thin Solid Films 2006, 514, 254. (71) Bail, M.; Schulz, M.; Brendel, R. Appl. Phys. Lett. 2003, 82, 757.
(38) Mora-Seró, I.; Luo, Y.; Garcia-Belmonte, G.; Bisquert, J.; Muñoz, (72) Cuevas, A.; Recart, F. J. Appl. Phys. 2005, 98, 074507.
D.; Voz, C.; Puigdollers, J.; Alcubilla, R. Sol. Energy Mater. Sol. Cells
(73) Mora-Seró, I.; Fabregat-Santiago, F.; Denier, B.; Bisquert, J.; Tena-
2008, 92, 505.
Zaera, R.; Elias, J.; Lévy-Clement, C. Appl. Phys. Lett. 2006, 89, 203117.
(39) Lyon, L. A.; Hupp, J. T. J. Phys. Chem. B 1999, 103, 4623.
(40) Miyashita, M.; Sunahara, K.; Nishikawa, T.; Uemura, Y.; Koumura, (74) Tena-Zaera, R.; Elias, J.; Leı̀vy-Cleı̀ment, C.; Bekeny, C.; Voss,
N.; Hara, K.; Mori, A.; Abe, T.; Suzuki, E.; Mori, S. J. Am. Chem. Soc. T.; Mora-Seroı̀, I.; Bisquert, J. J. Phys. Chem. C 2008, 112, 16318.
2008, 130, 17874. (75) Cameron, P. J.; Peter, L. M. J. Phys. Chem. B 2005, 109, 7392.
(41) Clifford, J. N.; Palomares, E.; Nazeeruddin, M. K.; Grätzel, M.; (76) Pivrikas, A.; Sariciftci, N. S.; Juska, G.; Österbacka, R. Prog.
Nelson, J.; Li, X.; Long, N. J.; Durrant, J. R. J. Am. Chem. Soc. 2004, 126, PhotoVoltaics: Res. Appl. 2007, 15, 677.
5225. (77) Arndt, C.; Zhokhavets, U.; Mohr, M.; Gobsch, G.; Al-Ibrahim, M.;
(42) Fan, H. Y. Phys. ReV. 1953, 92, 1424. Sensfuss, S. Synth. Met. 2004, 147, 257.
(43) Hornbeck, J. A.; Haynes, J. R. Phys. ReV. 1955, 97, 311. (78) Shuttle, C. G.; O’Regan, B.; Ballantyne, A. M.; Nelson, J.; Bradley,
(44) McIntosh, K. R.; Paudyal, B. B.; Macdonald, D. H. J. Appl. Phys. D. D. C.; Durrant, J. R. Phys. ReV. B 2008, 78, 113201.
2008, 104, 084503. (79) Mozer, A. J.; Dennler, G.; Sariciftci, N. S.; Westerling, M.; Pivrikas,
(45) Bisquert, J.; Garcia-Belmonte, G.; Pitarch, A. ChemPhysChem 2003, A.; Österbacka, R.; Juska, G. Phys. ReV. B 2005, 72, 035217.
4, 287. (80) Dennler, G.; Mozer, A. J.; Juska, G.; Pivrikas, A.; Österbacka, R.;
(46) Bisquert, J.; Vikhrenko, V. S. J. Phys. Chem. B 2004, 108, 2313. Fuchsbauer, A.; Saricifti, N. S. Org. Electron. 2006, 7, 229.
(47) Bisquert, J.; Fabregat-Santiago, F.; Mora-Seró, I.; Garcia-Belmonte, (81) Juska, G.; Sliauzys, G.; Genevicius, K.; Arlauskas, K.; Pivrikas,
G.; Barea, E. M.; Palomares, E. Inorg. Chim. Acta 2008, 361, 684. A.; Scharber, M.; Dennler, G.; Sariciftci, N. S.; Österbacka, R. Phys. ReV.
(48) Bisquert, J.; Zaban, A.; Salvador, P. J. Phys. Chem. B 2002, 106, B 2006, 74, 115314.
8774. (82) Pivrikas, A.; Juska, G.; Mozer, A. J.; Scharber, M.; Arlauskas, K.;
(49) Bisquert, J.; Palomares, E.; Qiñones, C. A. J. Phys. Chem. B 2006, Sariciftci, N. S.; Stubb, H.; Österbacka, R. Phys. ReV. Lett. 2005, 94, 176806.
110, 11284. (83) Shuttle, C. G.; O’Regan, B.; Ballantyne, A. M.; Nelson, J.; Bradley,
(50) Salvador, P.; González-Hidalgo, M.; Zaban, A.; Bisquert, J. J. Phys. D. D. C.; de Mello, J.; Durrant, J. R. Appl. Phys. Lett. 2008, 92, 093311.
Chem. B 2005, 109, 15915. (84) Garcia-Belmonte, G.; Munar, A.; Barea, E. M.; Bisquert, J.; Ugarte,
(51) Bisquert, J. Phys. ReV. B 2008, 77, 235203. I.; Pacios, R. Org. Electron. 2008, 9, 847.
(52) Haynes, J. R.; Hornbeck, J. A. Phys. ReV. 1953, 90, 152. (85) Garcı́a-Cañadas, J.; Fabregat-Santiago, F.; Bolink, H.; Palomares,
(53) Rose, A. Concepts in PhotoconductiVity and Allied Problems; E.; Garcia-Belmonte, G.; Bisquert, J. Synth. Met. 2006, 156, 944.
Interscience: New York, 1963.
(54) Bisquert, J. J. Phys. Chem. B 2004, 108, 2323. JP9037649