Available online at www.sciencedirect.

com

Thin Solid Films 516 (2008) 7144 – 7148

www.elsevier.com/locate/tsf

Charge separation in excitonic and bipolar solar cells — A detailed

balance approach
Thomas Kirchartz a , Uwe Rau b,⁎
a
Institut für Physikalische Elektronik, Universität Stuttgart, 70569 Stuttgart, Germany
b
IEF5-Photovoltaik, Forschungszentrum Jülich, 52425 Jülich, Germany
Available online 15 December 2007

Abstract

A generalized solar cell model for excitonic and classical, bipolar solar cells is developed that describes the combined transport and interaction
of electrons, holes and excitons. Both, conventional inorganic solar cells as well as organic solar cells, where excitons play a dominant role for
energy transport, turn out to be special cases of this model. Due to the inclusion of photon recycling effects, the approach is compatible with the
principle of detailed balance and the Shockley-Queisser limit. We show how varying the interaction between excitons and charge carriers as well
as varying the respective mobilities of the different species changes the operation mode of the solar cell path between excitonic and bipolar.
© 2008 Published by Elsevier B.V.

Keywords: Organic solar cells; Exciton transport; Mobility effects; Charge separation

1. Introduction similar as in Refs. [7,8]. At the surfaces of the absorber, we

The introduction of organic materials as photovoltaic
absorbers in organic [1] or dye-sensitized [2] solar cells has
raised questions whether or not these new types of devices can
be described with the help of theories that initially have been
developed for inorganic solid-state type devices usually
provided with a pn-homo- or heterojuncion. In addition, the
nano-scale mixture of different organic materials [3] has not
only boosted the efficiencies of organic solar cells but also led to
an apparent new type of solar cell, the bulk-heterojunction.
Recently, it has been proposed to divide all solar cells into two
classes, namely the classical inorganic solar cells and the
‘excitonic’ solar cells, embracing essentially all devices that use
organic absorber materials [4,5].
The present paper proposes a model that allows us to
describe virtually all single-absorber solar cells by an identical
approach that will be published in more detail elsewhere [6].
For the bulk of the absorber, we use a set of differential
equations for excitons, electrons, and holes that is coupled by
the dissociation of excitons into an electron/hole pair and the
recombination of this pair into an exciton. This approach is
allow for cross-dissociation of the exciton into an electron in the
absorber and a hole in the contact and vice versa.
We also include in our approach the optical interaction (photon
recycling) between the different regions in the device. As noted by
Martì et al. [9], such a feature is necessary to make a microsopic
transport theory consistent with the global detailed balance
approach of Shockley and Queisser (SQ) [10]. Because the SQ-
theory only accounts for the balance of photons entering or
leaving the device photon recycling is imminent. To check the
compatibility with the detailed balance approach, we verified with
our calculations that in the limit of high mobilities for excitons,
electrons and holes our model delivers the maximum efficiencies
given by the SQ-theory as will be shown elsewhere [6].
The present paper concentrates on the carrier collection
properties of the solar cells under short circuit conditions. We
especially show that by changing the dissociation rate of
excitons we can continuously tune the model from the excitonic
range considered in Refs. [4,5] towards a situation where the
solar cell works practically like a classical bipolar device.
2. Model

⁎ Corresponding author. For simplicity, we restrict the photovoltaic action to the

E-mail address: u.rau@fz-juelich.de (U. Rau). photogeneration of excitons and neglect that of free electron/
0040-6090/$ - see front matter © 2008 Published by Elsevier B.V.
doi:10.1016/j.tsf.2007.12.084
 T. Kirchartz, U. Rau / Thin Solid Films 516 (2008) 7144–7148
hole pairs. The radiative lifetime τrx of excitons is defined
similar to that of charge carriers in a semiconductor [11] via
R
c̃=srx ¼ ax ð E Þ/vol
bb ð E ÞdE ð1Þ
where c̃ denotes the concentration of excitons in thermal
equilibrium, αx(E) is the absorption coefficient resulting from
the generation of excitons, E the photon energy, and /vol bb ð E Þ
is the black body radiation per area into the full spherical
angle 4π. Note at this point that, due to the finite cell
thickness, a recombination event in the volume of the
photovoltaic absorber does not necessarily lead to an
emission of a photon from the surface of the solar cell.
Instead the possibility of re-absorption of this photon by
generation of another exciton has to be considered even in
equilibrium [9]. In the present work, we use the computation
scheme of Ref. [12] to involve this photon recycling effect
into our calculations.
Once generated, the excitons can dissociate into electrons
and holes in the bulk of the absorber with a dissociation lifetime
τdx. The dissociation of an exciton into an electron/hole pair and
the recombination of this pair into an exciton are also
interlinked by detailed balance and we have
Rx ñ p̃ ¼ c̃=sdx ð2Þ
where Rx is the recombination rate and ñ, p̃ denote the
equilibrium concentrations of electrons and holes. Similarly,
at the front or the back surface we allow for dissociation of
an exciton into an electron in the respective junction and a
hole that remains in the bulk of the absorber, as well as the
complementary process, dissociation into a hole in the
junction and an electron in the bulk. The rates for this type
of charge separation that is thought to be the dominant one in
excitonic solar cells are determined by the detailed balance
relations
Hn ñj p̃b ¼ Sxnc̃ ð3Þ
and
Hp ñb p̃j ¼ Sxpc̃ ð4Þ
where Hn/p are the cross recombination rates of electrons/
holes in the junction with holes/electrons in the bulk of the
absorber. The equilibrium concentrations of electrons and
holes in the bulk/junction are denoted by ñb/j, pj̃ /b. The
dissociation velocities Sxn/Sxp describe the probabilities that a
free exciton dissociates into an electron (hole) in the contact
and a hole (electron) in the bulk. An asymmetry in the values
of Sxn/Sxp causes the selectivity of a contact for the
dissociation of excitons much like an asymmetry of Sn/Sp
causes the selectivity for free carriers.
The collection of free electrons and holes at the interface is
determined by Sn⁎ñj = Snñb and Sp⁎pj̃ = Spp̃b where the rate
constants Sn/p, Sn⁎/p again have the dimension of a (collection
or injection) velocity. For the following computations, we set
ñ j = ñ b, p̃ j = p̃ b, and, consequently, Sn/p⁎ = Sn/p.
7145
Fig. 1. Band diagrams for a pin-type (a) and a pn-type solar cell (b) with the
conduction and valence band energy EC and EV as well as the equilibrium
Fermi-energy EF. The photovoltaic active absorber extends from coordinate
x = 0 (electron extracting contact) to x = d (hole extracting contact). In case of the
pin-type solar cell (a) the built-in voltage Vb drops across the undoped absorber,
whereas for pn-type device (b) Vb is accommodated within a space charge
region situated completely outside of the absorber. Schematically shown are the
ways how photogenerated excitons dissociate into electron/hole pairs in the bulk
of the absorber or at the interface to the electron extracting contact for both types
of devices.
The motion of excitons in the bulk of the absorber is
described by the continuity (diffusion) equation
d2 c c
Dx c¼ þ  Rx np  gx ð5Þ
dx2 sdx srx
where Dx is the diffusion constant for excitons and gx their
photogeneration rate. The continuity (drift-diffusion) equation
for electrons reads
d2 d c
Dn n  FAn n ¼ Rx np  ð6Þ
dx 2 dx sdx
and we have an analogous equation for the concentration p of
free holes with the ‘−‘ sign in front of the electrical field F
replaced by a ‘+’. The diffusion constants Dx/n/p are linked to the
mobilities μx/n/p by Einstein's relation Dx/n/p = kTμx/n/p/q where
kT/q denotes the thermal voltage.
Note that in a pin type structure the electrical field F
results from the difference between the built-in potential Vb
(cf. Fig. 1a) and the applied external voltage V according to
F = (Vb − V)/d where d is the thickness of the intrinsic
(undoped) absorber layer as sketched in Fig. 1a. Whereas in
this pin solar cell the voltage dependence of the current
voltage characteristics is exclusively determined by the
electrical field F present in the intrinsic layer, no such field
is present in a pn structure as sketched in Fig. 1b. The voltage
dependence of this type of device is determined by the change
of the minority carrier concentration nj = nj̃ exp(qV/kT ) that is
available in the injecting contact.
3. Results
In order to focus on the charge separation mechanism in
excitonic and bipolar solar cells, the following computations are
7146 T. Kirchartz, U. Rau / Thin Solid Films 516 (2008) 7144–7148
restricted to the short circuit situation of pin-type and pn-type
solar cells. In both type of devices, the necessary selectivity of
the contacts [13] is warranted by the asymmetry of the
dissociation velocities Sxn/xp of excitons as well as by the
asymmetry of the collection velocities Sn/p of the free carriers.
In all following situations, we assume Sxp = Sp = 0 at x = 0 and
Sxn = Sn = 0 at x = d (cf. Fig. 1). That means that free holes are
only extracted at the back contact and perfectly reflected by the
front contact. Accordingly, free electrons are collected only at
the front contact and reflected from the back contact. The
boundary conditions for the excitons are designed in an
analogous way: at the front contact excitons can only dissociate
into an electron in the contact and a hole in the bulk (finite Sxn)
whereas exciton dissociation into an electron in the bulk and a
hole in the contact is prohibited by Sxp = 0. The role of electrons
and holes is interchanged at the back contact (finite Sxp and
Sxn = 0).
In addition to the selectivity of the contacts, the built-in
field F (resulting from the built-in voltage Vb = 1.3 V, in the
following) in case of the pin-device provides additional help
for the separation of free carriers, whereas the excitons are
unaffected by the field. In case of the pn-device the built-in
voltage Vb is located completely outside the photovoltaic
absorber such that we have no field driven charge separation
under short circuit conditions. We may look at this situation
as the limiting case of a single-sided pn-junction under
neglect of the space charge region. The assumed absorption
coefficient α(E) is that of ZnPc (taken from Fig. 4 in Ref.
[14]) leading to a maximum short circuit current density
max
JSC = 20.5 mAcm− 2 for an absorber thickness of d = 300 nm
and under illumination with an AM1.5G solar spectrum. We
have further assumed a radiative exciton lifetime τrx = 200 μs
implying c̃ = 4.4 × 10− 3 cm− 3 because of Eq. (1). We note
here that radiative recombination of excitons is the only loss
mechanism considered throughout this paper.
In first step, we show that our model is able to describe a
continuous transition from a situation where transport is
basically arranged by excitons and a situation where transport
is exclusively supported by electrons and holes. We achieve this
transition by a variation of the dissociation lifetime τdx from
10− 4 s to 10− 9 s as shown in Fig. 2a for the pin and in Fig. 2b for
the pn-device. Here, we have assumed that the recombination
rate Rx is linked to τdx via Eq. (2), i.e. the concentration c̃ as
well as the product ñp̃ = 106 cm− 6 are kept constant. The current
density Jx as displayed in Fig. 2a and b is defined at the
coordinate x = 0 by Jx = Sxnx(0) − Hnnjp(0). Similarly, Jn is
defined by Jn = Snn(0) − Sn⁎nj such that Jn corresponds to the
net number of free electrons collected by the junction at x = 0
whereas Jx counts the number of excitons that are separated into
electrons (in the junction) and holes (in the bulk) at that
interface. Because of our assumption that excitons only
dissociate at the light exposed surface of the absorber (x = 0),
the hole current density Jp (analyzed at x = d) corresponds to the
sum Jp = Jx + Jn.
As can be seen from Fig. 2a, for dissociation lifetimes
τdx b 10− 8 s, the device acts like a classical pin-solar cell,
where the charge separation basically takes place within the
Fig. 2. Exciton, electron, and hole contribution Jx, Jn, and Jp to the short circuit
current density for different values of the bulk dissociation lifetime τdx. For the
for a pin-type (a) and a pn-type solar cell (b) the exciton contribution decreases
with decreasing τdx, i.e. with increasing coupling of the excitonic to the
electron/hole system, whereas the electron contribution increases at the same
time. The hole contribution Jp analysed at coordinate x = d (cf. Fig. 1)
corresponds to the total short circuit current density JSC and is unaffected by
the changes of τdx. The assumed mobilities are μx = μn = μp = 103 cm2 V− 1 s− 1 in
(a) and μx = 10− 2 cm2 V− 1 s− 1, μn = μp = 103 cm2 V− 1 s− 1 in (b).
bulk of the absorber and with the help of the built-in field.
This is because in this situation the photogenerated excitons
dissociate faster than they diffuse to the surface and also
faster than they recombine with their radiative lifetime τrx. In
contrast, for τdx N 10− 5 s, the total current of the device is
carried by excitons towards the front surface where the short
circuit current is generated via interfacial charge separation.
This limit corresponds to the situation of an excitonic solar
cell as described in Refs. [4,5]. The transition between the
excitonic and the bipolar operation mode occurs at approxi-
mately at τdx ≈ 2 × 10− 7 s for both the pin-type (Fig. 2a) and
the pn-type device (Fig. 2b). In the first case, the transition
regime extends approximately over 3 orders of magnitude in
τdx, whereas in the latter one the transition is pronouncedly
more smeared out extending over about 6 orders of
magnitude.
Within our model, this transition between excitonic and
bipolar operation not only depends on τdx and τrx but also on the
respective mobilities μx and μn/p of the excitons and the charge
carriers as well as on the interface dissociation/collection
velocities Sxn and Sn. For the situation illustrated in Fig. 2a with
μx = μn = μp = 103 cm2 V− 1 s− 1, we find that Jx and Jn are
unaffected by the choice of Sxn and Sn as long as Sxn,
Sn N 105 cms− 1.
For the pn-case we observe a similar transition from the
excitonic to the bipolar operation mode only if the mobility μx is
chosen considerably smaller than μn and μp. This is due to the
fact that in such a device the number of excitons always exceeds
the number of minority carriers (electrons). For the same reason
the collection velocity Sn must be chosen considerable larger
than the dissociation velocity Sxn. In Fig. 2b we have therefore
used μx = 10− 2 cm2 V− 1 s− 1, μn=μp = 103 cm2 V− 1 s− 1,
 T. Kirchartz, U. Rau / Thin Solid Films 516 (2008) 7144–7148
Sn = 1010 cms− 1 and Sxn = 106 cms− 1. We note that in all cases
considered in Fig. 2a and b, the short circuit current JSC
(corresponding to the hole current Jp at x = d) is as large as the
total number of absorbed photons per unit time. Thus, in all
situations we have a full photocurrent collection, regardless
whether the current is carried by electrons or by excitons.
Ongoing work [6] demonstrates that in all these situations the
solar cell efficiency corresponds to the SQ limit [10] and that the
results are also compatible with the fundamental reciprocity of
carrier collection and electroluminescent emission of solar cells
[15,16]. Thus, the SQ theory correctly describes the limiting
efficiency for bipolar as well as for excitonic solar cells.
In the next step, we investigate the effect of small exciton
and electron mobilities μx and μn on the current collection. Fig.
3a and b show the exciton, electron, and hole currents as a
function of the exciton mobility μx where we have chosen a
dissociation lifetime τdx = 100 μs corresponding to the excitonic
limit (cf. Fig. 2a and b, for the pin-and the pn-case,
respectively). Obviously, a decreasing μx hampers the current
collection by excitonic diffusion leading to a decrease of Jx
starting at approximately μx ≈ 2 × 10− 3 cm2 V− 1 s− 1 for the pin
case (Fig. 3a) as well as for the pn-case (Fig. 3b). As soon as Jx
starts to decline in the pin-or the pn-type cell, we observe an
increase of Jn. Obviously, part of the excitonic current is taken
over by the electrons and the devices change their character
from excitonic to bipolar in terms of our definition above.
However, the overall currents JSC = Jp = Jn + Jx are smaller in the
bipolar than in the excitonic limit of Fig. 3a and b. This is
because of the fact that the dissociation lifetime τdx = 100 μs is
chosen comparable to the radiative recombination lifetime
τrx = 200 μs. Therefore, a part of the photogenerated excitons
Fig. 3. Exciton, electron, and hole contribution Jx, Jn, and Jp to the short circuit
current density for different values of the exciton mobility μx with an assumed
dissociation lifetime τdx = 10− 4 s corresponding to the excitonic limit of Fig. 2.
For the for a pin-type (a) and a pn-type solar cell (b) the exciton contribution
decreases with decreasing μx, whereas the electron contribution increases at the
same time but not to a level such that JSC = Jp = Jx + Jn remains constant. With
decreasing μx the device changes its character from excitonic to bipolar.
Eventhough radiative recombination is the only loss mechanism, the device does
not deliver the maximum possible short circuit current in the limit of low
exciton mobilities μx. The assumed electron and hole mobilities are μn = μp =
103 cm2 V− 1 s− 1 in (a) and (b).
7147
Fig. 4. Exciton, electron, and hole currents Jx, Jn, and Jp for different values of
the electron mobility μn with an assumed dissociation lifetime τdx = 10− 9 s for
the pin-and τdx = 10− 11 s for the pn-configuration. These dissociation lifetimes
correspond to the bipolar limit of Fig. 2. In both the pin-type (a) and the pn-type
solar cell (b), the electron contribution decreases with decreasing μn and the
exciton contribution increases at the same time fully compensating the loss in Jn.
Thus, the device changes its character from bipolar to excitonic without any
losses in JSC. The transition from excitonic to bipolar occurs in a range around
μn ≈ 2 × 10− 2 cm2 V− 1 s− 1 for the pin-type cell (a) and μn ≈ 5 cm2 V− 1 s− 1 for
the pin-cell (b).
just recombines before they dissociate into electron/hole pairs
that are quickly separated in both devices due to the assumed
high mobilities μn = μp = 103 cm2 V− 1 s− 1. Despite of the fact
that radiative recombination is the only loss mechanism
assumed in the present calculations, neither the pin-nor the
pn-type device achieves full carrier collection. Consequently,
both devices fall short of the SQ-limit if the exciton mobility
μx is below a critical value and if the dissociation lifetime is
longer or comparable with the recombination lifetime.
Fig. 4a and b show the influence of the electron mobility μn
on the charge carrier collection properties of the pin-and pn-
solar cell. Here, we have chosen very short dissociation
lifetimes (τdx = 1 ns for the pin configuration and τdx = 10 ps
for the pn configuration) corresponding to the bipolar limit in
Fig. 2a and b. As expected, for both the pin-type (4a) and the pn-
type solar cell (4b), the electron contribution decreases with
decreasing μn. However, the exciton contribution increases at
the same time and fully compensates the loss in Jn. Thus, the
device changes its character from bipolar to excitonic without
any losses in JSC. The transition from excitonic to bipolar
occurs in a range around μn ≈ 2 × 10− 2 cm2 V− 1 s− 1 for the pin-
type cell (a) and μn ≈ 5 cm2 V− 1 s− 1 for the pin-cell (b). This
example shows that for the present case of two competing
transport channels, a decreasing mobility of one type of particle
does not necessarily lead to short circuit currents below the
SQ-limit.
Fig. 5 investigates the combined effect of simultaneously
lowering both electron and hole mobility μn, μp for the case
of a pin device. In the chosen configuration with interface
dissociation being only allowed at the front surface, the total
current at the backside is carried by holes. Consequently, low
7148 T. Kirchartz, U. Rau / Thin Solid Films 516 (2008) 7144–7148
Fig. 5. Exciton, electron, and hole currents Jx, Jn, and Jp for different values of
the electron and hole mobilities μn = μp (pin configuration, μx = 103 cm2 V− 1 s− 1
for all data points). The different dissociation lifetimes correspond to the bipolar
(τdx = 10− 9 s, open symbols) and the excitonic limit (τdx = 10− 4 s, full symbols) of
Fig. 2a. For τdx = 10− 9 s, the electron current Jn decreases and the excitonic
current Jx is not capable of compensating this loss. For τdx = 10− 4 s, the excitonic
current starts to decrease already at μp ≈ 2 × 102 cm2 V− 1 s− 1. Although μn is
decreasing parallel to μp, Jn partly compensates the loss in Jx, before at
μn/p b 10− 1 cm2 V− 1 s− 1 all contributions to the short circuit current vanish
simultaneously.
hole mobilities always lead to a decline of the total current,
though the mobility of excitons μx is kept as high as
μx = 103 cm2 V− 1 s− 1.
The intermediate region, between totally quenched short
circuit current and a sufficiently high mobility to collect all
created electron/hole pairs, shows different effects depending
on the dissociation lifetime. The open symbols in Fig. 5
represent the case of the low dissociation lifetime, τdx = 1 ns
(bipolar case in Fig. 2a) where for high mobilities, the short
circuit current at x = 0 is carried by electrons. Lowering the
values of μn/p leads to an almost synchronous decline of Jn and
Jp starting at μn/p ≈ 3 × 10− 1 cm2 V− 1 s− 1 . The exciton
component Jx plays no role in this situation. In contrast, for
τdx = 100 μs (full symbols), the exciton current Jx dominates at
μn/p N 2 × 102 cm2 V− 1 s− 1. Lowering μn/p below this value
decreases the contribution Jx whereas Jn increases. This is due
to the fact that exciton dissociation at the light exposed surface
(resulting from a high exciton current Jx at x = 0) enforces a hole
current through the entire thickness of the device. In contrast,
dissociation of the excitons into electrons and holes in the bulk
(leading to a high electron current Jn at x = 0) allows for a
shorter way for the holes to the back contact. However, a further
decrease of the mobilities μn/p b 10− 1 cm2 V− 1 s− 1 of the
electrons and holes leads to a total decline of all contributions to
the short circuit current.
4. Conclusions
The present paper has introduced a detailed balance model
for solar cells that includes transport and interaction of excitons,
electrons and holes. The model is designed to be compatible
with the SQ efficiency limit [10] as well as with the fundamental
reciprocity of carrier collection and electroluminescent emis-
sion of solar cells [15,16]. The model is able to describe a
continuous transition between an excitonic and a bipolar
operation mode of pin-as well as pn-type solar cells. In the
excitonic limit, the short circuit current of the cells exclusively
results from the dissociation of excitons at the photovoltaic
active interface. In the bipolar limit, the dissociation of excitons
(for simplicity, assumed to be the only photogenerated species
in the cell) takes place already in the bulk of the photovoltaic
absorber. Photovoltaic charge separation, in this limit, results
with the help of the built-in field and/or by the selectivity of the
contacts. Thus, the distinction between excitonic and classical
solar cells [4,5], though useful for the analysis of practical
devices, turns into limiting cases of the present general theory.
The paper also demonstrates that it is not possible to construct a
device operating in the excitonic limit that does not rely also on
a sufficient mobility of at least one type of the charged particles.
References
[1] C.W. Tang, Appl. Phys. Lett. 48 (1986) 183.
[2] B. O'Regan, M. Grätzel, Nature 353 (1991) 737.
[3] N.S. Sariciftci, L. Smilowitz, A.J. Heeger, F. Wudl, Science 258 (1992)
1474.
[4] B.A. Gregg, M.C. Hanna, J. Appl. Phys. 93 (2003) 3605.
[5] B.A. Gregg, J. Phys. Chem., B 107 (2003) 4688.
[6] T. Kirchartz, U. Rau, (unpublished).
[7] R. Corkish, D.S.P. Chan, M.A. Green, J. Appl. Phys. 79 (1996) 195.
[8] M. Burgelman, B. Minnaert, Thin Solid Films 511 (2006) 214.
[9] A. Martì, J.L. Balenzategui, R.F. Reyna, J. Appl. Phys. 82 (1997) 4067.
[10] W. Shockley, H.J. Queisser, J. Appl. Phys. 32 (1961) 510.
[11] W. van Roosbroeck, W. Shockley, Phys. Rev. 94 (1954) 1558.
[12] J. Mattheis, J.H. Werner, U. Rau, Phys. Rev. B 77 (2008) 085203.
[13] P. Würfel, Physica E 14 (2002) 18.
[14] J. Rostalski, D. Meissner, Sol. Energy Mater. Sol. Cells 63 (2000) 37.
[15] U. Rau, Phys. Rev. B 76 (2007) 085303.
[16] T. Kirchartz, U. Rau, M. Kurth, J. Mattheis, J.H. Werner, Thin Solid Films
515 (2007) 6238.