Incomplete Exciton Harvesting from
Fullerenes in Bulk Heterojunction Solar
George F. Burkhard,†Eric T. Hoke,†Shawn R. Scully,‡and Michael D. McGehee*,‡
Department of Applied Physics and Department of Materials Science and Engineering,
Stanford UniVersity, Stanford, California 94305
Received July 9, 2009; Revised Manuscript Received September 13, 2009
We investigate the internal quantum efficiencies (IQEs) of high efficiency poly-3-hexylthiophene:[6,6]-phenyl-C61-butyric acid methyl ester
(P3HT:PCBM) solar cells andfindthemtobe lower at wavelengths where the PCBMabsorbs. Because the excitondiffusionlengthinPCBM
istoosmall, excitonsgeneratedinPCBMdecaybeforereachingthedonor-acceptor interface. Thisresult hasimplicationsfor most stateof
the art organic solar cells, since all of the most efficient devices use fullerenes as electron acceptors.
Since their inception, organic photovoltaics (OPVs) have
steadily improved in performance. OPVs generate power
through three major processes: exciton generation (absorp-
tion), exciton harvesting (the process of excitons migrating
to the donor/acceptor interface and being split into their
constituent charges), and charge transport.1,2A typical device
consists of a charge generating active layer sandwiched
between hole extracting and electron extracting electrodes.
The active layer consists of an electron-donating material in
contact with an electron-accepting material. Excitons, bound
electron-hole pairs, are generated when light is absorbed
in one of the materials. If an exciton is sufficiently close to
the donor/acceptor interface, the exciton is split into its
constituent charges, leaving an electron in the acceptor and
a hole in the donor. Today’s best OPVs are made with active
layers using a bulk heterojunction structure obtained by
blending a polymer donor with a fullerene acceptor.3-5In
bulk heterojunction solar cells, the donor and acceptor are
naturally nanostructured due to phase segregation of the
polymer and fullerene. The morphology of the nanostructure
is somewhat tunable through thermal and solvent annealing.
Annealing typically increases the size of the domains in the
blend, which increases the distance excitons need to travel
to dissociate at the heterojunction interface. The increase in
domain size also affects charge carrier mobilities and
therefore the recombination mechanisms in the devices.6
Optimized devices have power conversion efficiencies of
5-6%.4,5,7Pushing these efficiencies higher requires detailed
analysis of the losses in these devices. A fraction of the
excitons in most pure materials decay radiatively, so exciton
harvesting is usually evaluated by observing photolumines-
cence quenching. In C60 fullerene systems like poly-3-
hexylthiophene:[6,6]-phenyl-C61-butyric acid methyl ester
(P3HT:PCBM), the PCBM emission is extremely weak and
its emission spectrum overlaps that of the polymer, so this
technique can only effectively probe exciton quenching in
the polymer phase. Most analyses assume the overall exciton
harvesting efficiency to be very close to 100%.8-10In this
paper we show that for the highest efficiency P3HT:PCBM
cells, the internal quantum efficiency is lower at wavelengths
where the PCBM absorbs. We find that the exciton harvesting
in the fullerene phase is less than 50% efficient. While
absorption in the fullerene is weak compared to the polymer,
recovering this loss would increase the photocurrent by
External quantum efficiency (EQE), the ratio of charges
extracted from a device to the number of incident photons
is an important benchmark of solar cell performance. Figure
1a shows the EQE and absorption of a typical high
performance (power conversion efficiency >4%) P3HT:
PCBM solar cell used in this study and is consistent with
EQE spectra of high efficiency cells published in the
literature.4,6,11Internal quantum efficiency, the ratio of
charges extracted from a device to the number of photons
absorbed by the active layer, provides a useful way to isolate
electronic loss mechanisms from light coupling and parasitic
absorption losses in a solar cell. The top curve in Figure 1b
shows a typical IQE spectrum for the same high efficiency
P3HT:PCBM devices. We found that the IQE curves were
far from flat; the IQE ranges from 50 to 75%, with lower
IQE at shorter wavelengths. Because more of the short-
* Corresponding author, email@example.com.
†Department of Applied Physics.
‡Department of Materials Science and Engineering.
Vol. 9, No. 12
10.1021/nl902205n CCC: $40.75
Published on Web 10/07/2009
2009 American Chemical Society
wavelength absorption occurs in the PCBM, the low IQE in
this region suggests that not all excitons generated in the
PCBM are harvested.
The IQE can be factored into three distinct parts: exciton
diffusion, charge transfer, and charge collection.
Each of these terms can have wavelength dependence.
Exciton diffusion and charge transfer are processes that
involve excitons in either the donor or the acceptor phase
and therefore might have different efficiencies depending on
the properties of the phase in question. Such differences
would result in wavelength dependence of the exciton
diffusion and charge transfer efficiencies since the absorption
contribution and thus exciton generation contribution of each
of the phases changes with wavelength. Effects that generate
multiple excitons from a single photon could also result in
wavelength-dependent exciton diffusion and charge transfer
efficiencies; however these effects have not been observed
in polymer-fullerene blend systems. Charge collection
encompasses all of the transport processes involved in
moving electrons and holes to their respective electrodes and
includes geminate and bimolecular recombination. The
charge collection process begins after excitons are split at
the heterojunction interface and is therefore insensitive to
the exciton’s origin; the charge collection process always
begins with an electron in the acceptor and a hole in the
donor. The charge collection efficiency, however, can vary
with position in the device due to differences in distances
the charges need to travel to be extracted, variations in
morphology, or interactions with electrodes. Optical interfer-
ence effects cause exciton generation profiles for different
wavelengths of light to have maxima at different locations
in the device. When combined, these two effects can lead to
wavelength dependence of the charge collection efficiency.
In a device where the exciton diffusion and charge transfer
efficiencies are equal in both materials, and the charge
transport efficiency does not change much throughout its
thickness, the IQE should be independent of excitation
To allow for different exciton diffusion and charge
efficiencies for the donor and acceptor materials, we modeled
the IQE with the equation
where AbsDand AbsAare the contributions to the absorption
spectrum and ηDand ηAare the exciton harvesting efficien-
cies ηEDηCTof the donor and acceptor, respectively. Note
that the numerator corresponds to the external quantum
efficiency and the denominator corresponds to the total
absorption in the active layer. AbsDand AbsAwere deter-
mined by measuring the total reflectance of the device and
deriving from this the absorption of the active layer using a
transfer matrix optical model.1,12The absorption due to either
component in the blend was then calculated by multiplying
the active layer absorption by the ratio of the k value (the
imaginary part of the complex index of refraction) in question
to the total k at the wavelength of interest
Further details on these measurements are provided at the
end of this paper and in the Supporting Information.
The top curves in Figure 1b show the fit between the
modeled and experimental IQE. Figure 1a shows the
experimentally measured EQE and absorption data used to
generate the IQE curve. The best fit of eq 2 to the IQE curve
in Figure 1b is obtained by taking ηCC) 79 +1/-4%, ηD)
Figure 1. (a) Experimentally measured EQE and absorption of a
P3HT:PCBM cell cast from 1,2-dichlorobenzene (solvent and
thermally annealed). The absorption in the active layer was extracted
from the total absorption using a transfer matrix optical model. The
contributions of the P3HT and PCBM to the active layer absorption
were determined by multiplying the active layer absorption by the
ratio of each component’s imaginary index of refraction to the total
imaginary index of refraction of the blend. (b) Experimentally
measured IQE curves of P3HT:PCBM cells cast from 1,2-
dichlorobenzene (solvent and thermally annealed), chlorobenzene
(as cast), and chloroform (as cast) as well as modeled IQE for the
ηCC(x)(ηDAbsD(λ) + ηAAbsA(λ))
AbsD(λ) + AbsA(λ)
Nano Lett., Vol. 9, No. 12, 2009
95 +5/-2%, and ηA) 41 +5/-1%. These are standard
errors and represent the extreme values the fit parameters
could take at the 50% confidence level assuming a normal
distribution of error. The fit value for exciton harvesting in
the donor, ηD) 95%, is consistent with photoluminescence
measurements which show 95% quenching of the emissive
The observation that only 41% of excitons in PCBM are
harvested indicates either that the diffusion length is smaller
than the PCBM domain size or that there is some other
excitation decay pathway. If the former is true, then by
reducing the domain size we should be able to recover all
of the excitons lost in the PCBM. Modeling the system with
ηA) ηD) 95% suggests that if all of the PCBM excitons
were harvested, we would see an increase in the photocurrent
To probe the dependence of exciton harvesting on domain
size, we created blends cast from lower boiling point solvents
(chloroform and chlorobenzene) without annealing to ensure
that the domains were as small as possible. As seen in the
two lower curves in Figure 1b, the IQEs of these devices
are independent of wavelength, indicating that both the
PCBM and P3HT have high exciton harvesting efficiencies.
Of course the IQE is also low, indicating that while shrinking
the domains improved exciton harvesting, it dramatically
decreased the charge transport efficiency, making the device
less efficient overall.
We considered the possibility that the wavelength depen-
dence of the IQE could be due to variations in ηCCfrom
optical interference effects. If this were the case, we would
expect to see IQE minima at different wavelengths for
devices of different thicknesses. However, we observed that
the minimum IQE is always in the blue end of the spectrum
(Figure 2). This implies that the charge collection has
negligible dependence on excitation wavelength. We also
considered the possibility that singlets generated in the
PCBM might be lost via mechanisms other than internal
conversion, such as energy transfer to polarons13-15or
intersystem crossing to the triplet state.16It is possible that
triplets have lower charge separation efficiency due to their
lower energy; however, the fact that we were able to recover
these excitons by making the PCBM domains smaller points
to exciton diffusion rather than charge separation as the
reason for the reduced IQE. Additionally, the IQE curves
are independent of excitation intensity up to one sun (data
not shown), which discounts energy transfer to polarons since
this recombination pathway depends on the carrier density.
The exciton diffusion lengths in PCBM have not yet been
thoroughly studied; however, Cook et al. have performed
measurements that suggest a value as small as 5 nm.17We
have not measured the exciton diffusion length because most
methods for doing so detect photoluminescence quenching
and PCBM is a very weak emitter. Furthermore any
technique that analyzed thin films of pure PCBM might not
reveal the exciton diffusion length for PCBM in a bulk
heterojunction due to differences in morphology. PCBM
domain sizes vary and are typically 10-100 nm after
annealing for 5 min at 100 °C18and up to tens of micrometers
after annealing at higher temperatures for longer periods of
time.19It is therefore not surprising that the domains might
be significantly larger than the exciton diffusion length.
Incomplete exciton harvesting from fullerenes might help
explain some effects seen by others. Moule ´ et al. observed
a “reduced generation zone” (RGZ) in the active layer near
(PEDOT:PSS) interface.20Several studies have shown verti-
cal phase segregation of the active layer,3,21where larger
fullerene domains lie near this interface. This could be due
to the more polar nature of PCBM compared with P3HT22-25
or because PCBM is more soluble than the polymer in the
casting solvent, which evaporates from the top surface first.
These studies used multiple characterization techniques
including ellipsometry, near-edge X-ray absorption fine
structure spectroscopy, dynamic secondary ion mass spec-
troscopy, and energy compatibility arguments. Our data are
consistent with the hypothesis that the observed RGZ is due
to excitonic losses in the fullerene due to the larger domain
size near the active layer/PEDOT boundary. Having less
polymer at this interface would also weaken absorption in
this area, reducing generation; however this would not
explain the wavelength dependence of the IQE we observe.
Figure 2 shows IQE curves for devices made in the same
manner as the high-efficiency cells but with active layers of
varying thickness. We observed that while all of the cells
had similar IQEs at longer wavelengths, thinner cells
generally had higher IQEs at shorter wavelengths than thicker
cells. This is consistent with a vertical phase segregation
model where the largest PCBM domains appear close to the
PEDOT interface. TEM tomography has suggested that
vertical phase segregation occurs with the opposite orienta-
tion (PCBM accumulating near the metal electrode).26A
thorough discussion of this is beyond the scope of this paper;
however, we note that TEM methods can only differentiate
PCBM from crystalline P3HT. Thus, the TEM data could
be interpreted to mean that there is more polymer at the
PEDOT interface or that the P3HT is more crystalline in
this region. While the general effect of vertical phase
Figure 2. IQE curves for devices of varying active layer thickness.
These devices were processed in the same manner as the high-
efficiency device shown as the top curve in Figure 1a.
Nano Lett., Vol. 9, No. 12, 20094039
segregation would be that thicker films will show a larger
fraction of oversized PCBM domains, this effect might be
especially important for the thick films used to make high
efficiency solar cells. Optical interference modeling, shown
in Figure 3, confirms that for these cells (optimized to 220
nm active layer thickness), the highest excitation rates occur
near this interface. It should be noted that solvent and thermal
annealing of the blend results in larger PCBM domains
throughout the film, so we would predict poorer exciton
harvesting in the PCBM phase regardless of vertical phase
Park et al. recently published work on a 6.1% efficient
cell using poly[N-900-hepta-decanyl-2,7-carbazole-alt-5,5-
(40,70-di-2-thienyl-20,10,30-benzothiadiazole) (PCDTBT) as
a donor material that does not require long periods of solvent
or thermal annealing to achieve good device performance5
and does not suffer from an exciton harvesting problem in
the fullerene, as evidenced by their flat, near-100% IQE.
Their device uses a very thin (80 nm) active layer. Because
charge carriers have less far to travel, mobility requirements
are less stringent and annealing is not required for good
device performance. Because they do not anneal their films,
the donor and acceptor are probably more intimately mixed.
If this is true, it might explain why their exciton harvesting
efficiency is very close to 100% for both the polymer and
the fullerene. Their high, flat IQE shows that the exciton
harvesting problem is not an insurmountable one and that
better design rules are enough to make higher efficiency
organic solar cells.
We used IQE measurements as a tool to investigate exciton
harvesting efficiencies in P3HT:PCBM bulk heterojunction
solar cells and found that in the best performing cells with
high electron and hole mobilities, there is incomplete
harvesting of excitons in the fullerene phase. The exciton
diffusion length in the fullerene is generally shorter than the
domain size, and approximately 60% of excitons generated
in the fullerene phase decay before being harvested. We are
currently investigating the same effect in devices made with
PC70BM, which absorbs more strongly in the solar spectrum.
Preliminary results indicate that these devices have a similar
exciton harvesting issue; however quantitative analysis is still
underway. Our findings have implications for most bulk
heterojunction solar cells since the vast majority use PCBM
as an electron acceptor. Novel geometries that use strongly
absorbing, thin active layers may bypass this issue by using
blends with smaller domains since having high charge carrier
mobilities is less important in a thinner device. It may also
be possible to solve this problem in more standard devices
using novel nanostructures or new acceptors.
Devices were made with the structure indium tin oxide
(ITO)/PEDOT:PSS/P3HT:PCBM/Ca/Al with the following
thicknesses (in nm): 110/35/220/7/200. ITO substrates were
purchased from Sorizon Technologies, PEDOT:PSS from
Baytron, P3HT from Rieke, PCBM from NanoC, and metals
from K. J. Lesker. Substrates were cleaned in an ultrasonic
bath with Extran 300, rinsed in deionized water, and then
cleaned in acetone and 2-propanol followed by 20 min of
UV-ozone treatment. PEDOT:PSS was spin-coated and the
substrates were annealed at 140 °C for 10 min. They were
then transferred to a nitrogen glovebox, where they remained
for the duration of the fabrication process as well as for all
characterizations performed. P3HT:PCBM (1:1 ratio by
weight) was cast from 1,2-dichlorobenzene, chlorobenzene,
or chloroform. The devices cast from dichlorobenzene were
allowed to slow-dry overnight and were thermally annealed
at 110 °C for 10 min. Calcium and aluminum metal
electrodes were deposited in a thermal evaporator.
IQE was calculated using an external quantum efficiency
(EQE) measurement as well as a reflection-mode absorption
measurement. EQE was taken at short circuit using mono-
chromated white light from a tungsten lamp. EQE was
calculated by comparing the photocurrent action spectrum
of the device to that of a NIST traceable calibration
photodiode. The device absorption spectrum was measured
in reflection mode inside of an integrating sphere (to capture
scattered light) using the same monochromated white light
source and calibrated photodiode. Parasitic absorption in the
ITO, PEDOT, and metals was calculated using a transfer
matrix formalism1,12to evaluate the coherent superposition
of light waves at each interface which is described in the
Supporting Information. The active layer absorption was then
calculated by subtracting the modeled parasitic absorption
from the experimentally measured total absorption. Indices
of refraction for the various materials were either taken from
literature2or measured using a combination of spectroscopic
ellipsometry and absorption/reflection measurements.
Acknowledgment. This work was supported by the Center
for Advanced Molecular Photovoltaics (Award No KUS-
C1-015-21), made by King Abdullah University of Science
and Technology (KAUST) and partially by the Global
Climate Energy Project (GCEP). E.T.H. is supported by the
National Science Foundation GRFP and the Fannie and John
Figure 3. Exciton generation rate in the active layer vs position in
the device for an optimized (220 nm active layer thickness) under
AM1.5G illumination. The left side of the plot (0 nm) represents
the interface with the PEDOT; the right boundary (220 nm)
represents the boundary with the reflective metal electrode.
Nano Lett., Vol. 9, No. 12, 2009
Supporting Information Available: Calculating AbsDand Download full-text
AbsAand a typical comparison of calculated total reflectance
vs experimentally measured reflectance. This material is
available free of charge via the Internet at http://pubs.acs.org.
(1) Peumans, P.; Yakimov, A.; Forrest, S. J. Appl. Phys. 2003, 93, 3693–
(2) Kotlarski, J. D.; Blom, P. W. M.; Koster, L. J. A.; Lenes, M.; Slooff,
L. H. J. Appl. Phys. 2008, 103, 084502.
(3) Li, G.; Shrotriya, V.; Huang, J. S.; Yao, Y.; Moriarty, T.; Emery, K.;
Yang, Y. Nat. Mater. 2005, 4 (11), 864–868.
(4) Peet, J.; Kim, J. Y.; Coates, N. E.; Ma, W. L.; Moses, D.; Heeger,
A. J.; Bazan, G. C. Nat. Mater. 2007, 6 (7), 497–500.
(5) Park, S. H.; Roy, A.; Beaupre, S.; Cho, S.; Coates, N.; Moon, J. S.;
Moses, D.; Leclerc, M.; Lee, K.; Heeger, A. J. Nat. Photonics 2009,
3 (5), 297–302.
(6) Li, G.; Yao, Y.; Yang, H.; Shrotriya, V.; Yang, G.; Yang, Y. AdV.
Funct. Mater. 2007, 17, 1636–1644.
(7) Kim, Y.; Cook, S.; Tuladhar, S. M.; Choulis, S. A.; Nelson, J.; Durrant,
J. R.; Bradley, D. D. C.; Giles, M.; McCulloch, I.; Ha, C.-S.; Ree, M.
Nat. Mater. 2006, 5, 197–203.
(8) Mayer, A. C.; Scully, S. R.; Hardin, B. E.; Rowell, M. W.; McGehee,
M. D. Mater. Today 2007, 28–33.
(9) Yang, X.; Loos, J. Macromolecules 2007, 40 (5), 1353–1362.
(10) Dennler, G.; Scharber, M. C.; Brabec, C. J. AdV. Mater. 2009, 21,
(11) Dennler, G.; Forberich, K.; Scharber, M. C.; Brabec, C. J.; Tomis ˇ, I.;
Hingerl, K.; Fromherz, T. J. Appl. Phys. 2007, 102, 054516.
(12) Pettersson, L. A. A.; Roman, L. S.; Inganas, O. J. Appl. Phys. 1999,
86 (1), 487–496.
(13) Scully, S. R.; McGehee, M. D. J. Appl. Phys. 2006, 100 (3), 034907.
(14) Liu, Y. X.; Summers, M. A.; Scully, S. R.; McGehee, M. D. J. Appl.
Phys. 2006, 99, 093521.
(15) Hukka, T. I.; Toivonen, T.; Hennebicq, E.; Bredas, J. L.; Janssen,
R. A. J.; Beljonne, D. AdV. Mater. 2006, 18 (10), 1301–1306.
(16) Cook, S.; Durrant, J. R. Chem. Phys. Lett. 2007, 445 (4-6), 276–
(17) Cook, S.; Furube, A.; Katoh, R.; Han, L. Chem. Phys. Lett. 2009,
(18) Vanlaeke, P.; Swinnen, A.; Haeldermans, I.; Vanhoyland, G.; Aernouts,
T.; Cheyns, D.; Deibel, C.; D’Haen, J.; Heremans, P.; Poortmans, J.;
Manca, J. V. Sol. Energy Mater. Sol. Cells 2006, (90), 2150–2158.
(19) Savenije, T. J.; Kroeze, J. E.; Yang, X. N.; Loos, J. AdV. Funct. Mater.
2005, 15 (8), 1260–1266.
(20) Moule ´, A. J.; Meerholz, K. Appl. Phys. B: Laser Opt. 2009, 92, 209–
(21) Yang, X. N.; Loos, J.; Veenstra, S. C.; Verhees, W. J. H.; Wienk,
M. M.; Kroon, J. M.; Michels, M. A. J.; Janssen, R. A. J. Nano Lett.
2005, 5 (4), 579–583.
(22) Campoy-Quiles, M.; Ferenczi, T.; Agostinelli, T.; Etchegoin, P. G.;
Kim, Y.; Anthopoulos, T. D.; Stavrinou, P. N.; Bradley, D. D. C.;
Nelson, J. Nat. Mater. 2008, 7, 158–164.
(23) Germack, D. S.; Chan, C. K.; Hamadani, B. H.; Richter, L. J.; Fischer,
D. A.; Gundlach, D. J.; DeLongchamp, D. M. Appl. Phys. Lett. 2009,
(24) Bjo ¨rstro ¨m, C. M.; Bernasik, A.; Rysz, J.; Budkowski, A.; Nilsson, S.;
Svensson, M.; Andersson, M. R.; Magnusson, K. O.; Moons, E. J.
Phys.: Condens. Matter 2005, 17, L529–L534.
(25) Barrau, S.; Andersson, V.; Zhang, F.; Masich, S.; Bijleveld, J.;
Andersson, M. R.; Ingana ¨s, O. Macromolecules 2009, 42 (13), 4646–
(26) van Bavel, S. S.; Sourty, E.; With, G. d.; Loos, J. Nano Lett. 2009, 9
Nano Lett., Vol. 9, No. 12, 20094041