COMMUNICATIONS

DOI: 10.1002/adma.200501717

Design Rules for Donors in Bulk-Heterojunction Solar

Cells—Towards 10 % Energy-Conversion Efficiency**
By Markus C. Scharber,* David Mühlbacher, Markus Koppe, Patrick Denk, Christoph Waldauf,
Alan J. Heeger, and Christoph J. Brabec

There has been an intensive search for cost-effective photo- e.g., in the field of portable electronics. One major obstacle
voltaics since the development of the first solar cells in the for the commercialization of bulk-heterojunction solar cells is
1950s.[1–3] Among all alternative technologies to silicon-based the relatively small device efficiencies that have been demon-
pn-junction solar cells, organic solar cells could lead the most strated up to now.[5] The best energy-conversion efficiencies
significant cost reduction.[4] The field of organic photovoltaics published for small-area devices approach 5 %.[9–11] A de-
(OPVs) comprises organic/inorganic nanostructures like dye- tailed analysis of state-of-the-art bulk-heterojunction solar
sensitized solar cells, multilayers of small organic molecules, cells[8] reveals that the efficiency is limited by the low open-
and phase-separated mixtures of organic materials (the bulk- circuit voltage (Voc) delivered by these devices under illumi-
heterojunction solar cell). A review of several OPV technolo- nation. Typically, organic semiconductors with a bandgap of
gies has been presented recently.[5] Light absorption in organic about 2 eV are applied as photoactive materials, but the ob-
solar cells leads to the generation of excited, bound electron– served open-circuit voltages are only in the range of 0.5–1 V.
hole pairs (often called excitons). To achieve substantial ener- There has long been a controversy about the origin of the Voc
gy-conversion efficiencies, these excited electron–hole pairs in conjugated polymer–fullerene solar cells. Following the
need to be dissociated into free charge carriers with a high classical thin-film solar-cell concept, the metal–insulator–met-
yield. Excitons can be dissociated at interfaces of materials al (MIM) model was applied to bulk-heterojunction devices.
with different electron affinities or by electric fields, or the In the MIM picture, Voc is simply equal to the work-function
dissociation can be trap or impurity assisted. Blending conju- difference of the two metal electrodes. The model had to be
gated polymers with high-electron-affinity molecules like C60 modified after the observation of the strong influence of the
(as in the bulk-heterojunction solar cell) has proven to be an reduction potential of the fullerene on the open-circuit volt-
efficient way for rapid exciton dissociation. Conjugated poly- age by introducing the concept of Fermi-level pinning.[12] It
mer–C60 interpenetrating networks exhibit ultrafast charge has also been shown that the Voc of polymer–fullerene solar
transfer (∼40 fs).[6,7] As there is no competing decay process cells is affected by the morphology of the active layer[13] and
of the optically excited electron–hole pair located on the poly- that it can be influenced by the electrochemical potential of
mer in this time regime, an optimized mixture with C60 con- the cathode (poly(3,4-ethylenedioxythiophene):poly(styrene
verts absorbed photons to electrons with an efficiency close to sulfonate), PEDOT:PSS).[14] However, even an extended
100 %.[8] The associated bicontinuous interpenetrating net- MIM model was found to be insufficient for a complete expla-
work enables efficient collection of the separated charges at nation of the Voc of bulk-heterojunction solar cells. Recently,
the electrodes. The bulk-heterojunction solar cell has at- Gadisa et al.[15] studied the correlation of electrochemical
tracted a lot of attention because of its potential to be a true properties of six similar polythiophene derivatives and the
low-cost photovoltaic technology. A simple coating or printing open-circuit voltage of polymer–fullerene solar cells using
process would enable roll-to-roll manufacturing of flexible, [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) as an ac-
low-weight PV modules, which should permit cost-efficient ceptor and the same polythiophenes derivatives as donors.
production and the development of products for new markets, The authors attributed the variation of the photovoltage to
variation of the oxidation potential of the conjugated poly-
mer. Concluding the summary of the recent literature, there is
– no general consistent understanding of the origin of the open-
[*] Dr. M. C. Scharber, D. Mühlbacher, M. Koppe, P. Denk,
C. Waldauf, Dr. C. J. Brabec circuit voltage of the bulk-heterojunction solar cell. In this
Konarka Austria work, the relation between the energy levels of the donor–ac-
Altenbergerstrasse 69, A-4040 Linz (Austria) ceptor blend and the open-circuit voltage of 26 different bulk-
E-mail: [email protected]
heterojunction devices is studied. A simple relation between
Prof. A. J. Heeger
Department of Materials Science the energy level of the highest occupied molecular orbital
Broida Hall 6125 (HOMO) of the polymer and the Voc is derived, which is used
University of California at Santa Barbara to estimate the maximum efficiency of bulk-heterojunction
Santa Barbara, CA 3106-5090 (USA)
solar cells. Based on the model, the ideal material parameters
[**] This work was partially supported by the Commission of the Eur-
opean Community (Project MOLYCELL Contract No. 502783) and for a conjugated polymer–PCBM device are determined. The
the BMBF project EKOS: Förderkennzeichen 03N2023. results presented here can be used as a guideline for the selec-

Adv. Mater. 2006, 18, 789–794 © 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 789
COMMUNICATIONS

tion and synthesis of new active materials for bulk-heterojunc- As a practical limit to the efficiency, the external quantum
tion solar cells. efficiency (EQE) of the solar cell for photon energies equal to
In Figure 1 the open-circuit voltage of different bulk-het- or larger than the bandgap energy of the donor and the FF
erojunction solar cells is plotted versus the oxidation potential are set to 65 %. Similar values have been found by several
groups for optimized devices in the absence of recombination
losses.[8,18] In addition, any contribution to the short-circuit
HOMO Position [ eV ] current from photons absorbed by the fullerene is neglected.
-4.6 -4.8 -5.0 -5.2 -5.4 -5.6 -5.8 Under these assumptions, the efficiency calculation requires
1200 1200
only an integration of the Air Mass 1.5 (AM1.5) spectrum
1000 1000 times the constant EQE for the Jsc and Equation 1 for Voc.
The result is shown in Figure 2 as a contour plot where the
Measured Voc [ mV ]

800 800
x- and y-axes are the bandgap and the LUMO level of the
600 600 donor, respectively, and the contour lines indicate constant
power-conversion efficiencies. The straight lines in Figure 2
400 400 define lines of constant donor HOMO levels of –5.7 and
200 200
–4.8 eV. Calculations were performed for LUMO levels of the
donor in the range of –3 to –4 eV assuming that an energy
0 0 difference DE of 0.3 eV between the LUMO of the donor and
-200 0 200 400 600 800 1000 1200
the LUMO of the acceptor is sufficient for efficient charge
Onset of Oxidation [ mV ]
separation.[19] Figure 2 suggests that the energy-conversion
Figure 1. Open-circuit voltage (Voc) of different bulk-heterojunction solar
efficiency of a bulk-heterojunction solar cell should be much
cells plotted versus the oxidation potential/HOMO position of the donor more sensitive to changes of the donor LUMO level com-
polymer used in each individual device. The straight line represents a lin- pared to variations of the donor bandgap. For example, for a
ear fit with a slope of 1. LUMO level of –3.6 eV the calculated efficiency at a bandgap
of 2.1 eV is ∼5.8 %. Upon decreasing the bandgap, the effi-
of the conjugated polymers used in these devices. The conju- ciency increases to ∼6.7 % at a bandgap of 1.8 eV and gradu-
gated polymers corresponding to open squares are listed in ally decreases to ∼5.8 % at a gap of 1.45 eV. A variation of the
Table 1. In addition, data points corresponding to polymer donor bandgap of 0.65 eV leads to a variation of only 1 % in
materials proprietary to Konarka (filled triangles) are added. the device efficiency. In contrast, a 0.65 eV variation of the
The line representing a linear fit to the measured points has a polymer LUMO level results in efficiency changes between
slope equal to 1 and intersects the x-axis at –140 mV, which 3.5 and 8 % depending on the donor bandgap. For energy-
corresponds to a HOMO level of ∼4.6 eV. Figure 1 demon- conversion efficiencies exceeding 10 %, the donor polymer
strates that for a total of 26 different bulk-heterojunction must have a bandgap < 1.74 eV and a LUMO level < –3.92 eV,
solar cells, a linear relation between Voc and the conjugated assuming the FF and the average EQE remain equal to 0.65.
polymer oxidation potential is found; the open-circuit voltage For a 3.92 eV LUMO value, the calculated efficiency value is
of a conjugated polymer–PCBM solar cell can be estimated almost constant upon decreasing the donor bandgap down to
by 1.3 eV. The results presented here demonstrate the impor-
tance of the donor LUMO level in bulk-heterojunction solar
Voc = (1/e)(EDonorHOMO – EPCBMLUMO) – 0.3 V (1) cells. Besides a reduction of the bandgap, new donor materials
must be designed to optimize the LUMO as this parameter
where e is the elementary charge and using –4.3 eV for the dominantly drives the solar-cell efficiency. It is important to
PCBM lowest occupied molecular orbital (LUMO) energy. note that an optimized open-circuit voltage is a prerequisite
The value of 0.3 V in Equation 1 is an empirical factor which to achieve certain device efficiencies; however, it is not suffi-
will be discussed below. It is interesting to note that cient. In addition, the charge-carrier mobility of electrons and
PEDOT:PSS with a Fermi level of –5.0 eV can contact poly- holes in the donor acceptor blend must be high enough to
mers with a wide range of HOMO levels. Although there may allow efficient charge extraction and FF of 0.65. The relation
be an optimized doping level of the PEDOT:PSS layer for between charge-carrier mobility and FF can be deduced from
every conjugated polymer,[14] even polyfluorene copolymers a model which has been reported recently.[20] The main
with a HOMO level of ∼–5.7 eV can be used as donors in effi- feature of this extended pn-junction model is that the photo-
cient bulk-heterojunction solar cells when PEDOT:PSS is current is dominantly field-driven. For devices with an active-
applied as the anode material.[16,17] Based on this finding, the layer thickness of several hundred nanometers, carrier mobili-
limiting efficiency of a bulk-heterojunction solar cell (given ties of ∼10–3 cm2 V–1 s–1 are required to prevent losses in the
by the open-circuit voltage (Voc) times the short-circuit cur- photocurrent in the absence of second-order recombination.
rent density (Jsc) times the electrical fill factor (FF) divided Equally important is the successful description of the light-in-
by the incident-light intensity) can be predicted solely as a tensity dependence of the Voc of bulk-heterojunction solar
function of the bandgap and the LUMO level of the donor. cells. Recently, Koster et al. suggested a complicated model

790 www.advmat.de © 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 789–794
 COMMUNICATIONS
Table 1. Structures of different donor polymers used in bulk-heterojunction solar cells.

Structure Name HOMO

[eV]
H33C16 C16H 33
Poly-[2,6-(4- –4.82
dihexadecylmethylene-
cyclopentadithiophene)]
* *
S S n

C10H23

O O
Poly-[(1′dodecyl)-3,4- –4.83
ethylenoxythiophene]
* *
S n
C8 H17

O
N

O Poly-[(3,4- –4.84
O
ethylenedioxythiophene)-N-2′-
ethyl-4,5-dicarboxylic-imide-
* *
S S
n
benzo[c]thiophene)]

C4 H9

Poly(3-butylthiophene) –5.05
* *
S n

C6H13

Poly(3-hexylthiophene) –5.1
* *
S n

H17C8 C8H17

Poly[2,5-(7,7-dioctyl)- –5.15
*
cyclopentadithiophene]
*
S S n

S
Poly[2,7-9,9-dioctylfluorene)-alt- –5.25
* S
C 8H 17 n
*
2,6-cyclopentadithiophene]
H 17C 8

H3C

CH3

H3C

O
Poly[(2-methyloxy-5-[3,7- –5.26
* dimethyloctyloxy])paraphenylene
*
n vinylene]
O

CH3

S
N
S
N N
N
R

R S Poly[N-dodecyl-2,5-bis(2′- –5.28
S
N thienyl)pyrrole-2,1,3-
n

C12H25 benzothiadiazole]

O2N C6H13
Poly(3-hexyl-4-nitroxythiophene) –5.45

* *
S

H17C8 C8H17
N
S
N
Poly[2,7-(9,9-dioctylfluorene)- –5.75
S S
alt-5,5′′-(4′,7′-di-2-thienyl-
*
*
2′,1′,3′-benzothiadiazole]
n

Adv. Mater. 2006, 18, 789–794 © 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.advmat.de 791
COMMUNICATIONS

Power Conversion Efficiency [ % ]

-4 .0
0
8.00 10.00
1.00
9.00 2.00
3.00
-3 .8 4.00
LUMO Level Donor [ eV ]

5.00
7.00 6.00

5.00 7.00
-3 .6 8.00
9.00
HOMO -5.7 eV 6.00 10.00
11.00

-3 .4 1.00
4.00
LUMO

∆E LUMO -4.3 eV
-3 .2 Eg
2.00
3.00
HOMO -4.8 eV HOMO
2.00
-3 .0 HOMO -6.0 eV

3. 0 2.7 2. 4 2 .1 1. 8 1.5 1. 2 Donor PCBM

Band Gap Donor [ eV ]

Figure 2. Contour plot showing the calculated energy-conversion efficiency (contour lines and colors) versus the bandgap and the LUMO level of the
donor polymer according to the model described above. Straight lines starting at 2.7 eV and 1.8 eV indicate HOMO levels of –5.7 eV and –4.8 eV, re-
spectively. A schematic energy diagram of a donor PCBM system with the bandgap energy (Eg) and the energy difference (DE) is also shown.

based on a MIM concept. The experimental findings pre- the theoretical maximum value for the VBI in bulk-hetero-
sented here are in contradiction to the classical MIM picture, junction systems. As discussed above and shown in Figure 1,
i.e., for more than 26 different material combinations, no in- deviations between the theoretical maximum VBI and the Voc
fluence of the contact work function on the Voc is observed. on the order of 0.3 V are found. In Figure 3, the current–vol-
These conceptual problems can be overcome by using the tage curve of a poly(3-hexylthiophene) (P3HT)/PCBM bulk-
extended pn-junction model with a field-dependent photocur- heterojunction solar cell acquired in the dark (solid line) and
rent.[21] Underlying the basic physics behind the model is the
fact that all photovoltaic cells work essentially the same way:
absorbed light is transferred into photogenerated charge car-
0.015
riers, which are separated by a junction later. The junction
guarantees that separated charge carriers do not recombine 0.010
before reaching the contacts. In contrast to classical solar-cell
materials like Si or Ge, however, the quantum efficiency for 0.005 2
slope = µτ/d
current [ A ]

charge generation in pristine organic semiconductors is low, ∆V~220 mV

typically in the range of 1 %. It is therefore a necessity to 0.000
blend semiconductors with appropriate electron affinities and
work functions[19] to increase the quantum efficiency for -0.005 Voc VBI
charge generation. This so-called donor–acceptor principle is
successfully applied in liquid-electrolyte and solid-state dye- -0.010
sensitized solar cells as well as in polymer/fullerene solar cells.
A consequence of this approach is the presence of a charge- -0.015
0.0 0.2 0.4 0.6 0.8 1.0 1.2
generating junction that is identical to the charge-separating voltage [ V ]
junction. In other words, the energy-level differences between
the polymer and the fullerene form the junction for the photo- Figure 3. Current–voltage curve of (P3HT)/PCBM solar cell measured in
the dark (solid line) and field-driven photocurrent (dotted line). The
voltaic action. The Voc and the built-in potential VBI are deter- built-in potential VBI is given by ELUMO(PCBM) minus EHOMO(P3HT)
mined by the difference between the HOMO of the donor (–4.3 V – (–5.1 V) = 0.8 V). l: charge-carrier mobility; s: charge-carrier
and the LUMO of the acceptor. It is trivial to note that this is lifetime; d: active-layer thickness.

792 www.advmat.de © 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 789–794
 COMMUNICATIONS
the idealized field-driven photocurrent (dotted line) are Experimental
plotted. The superposition of both curves gives the current–
voltage curve under illumination.[20] The open-circuit voltage In this study, commercially available compounds and polymers
of the solar cell is defined as the voltage which compensates which were prepared for Konarka to be tested in solar cells by compa-
the current flow through the external circuit (indicated by the nies, universities, or other research institutions were investigated. The
chemical structures, which are public and not proprietary to Konarka,
vertical line in Fig. 3). Figure 3 shows that a reduction of Voc are listed in Table 1. The electron-acceptor material PCBM was pur-
is caused by the dark current–voltage curve of the diode, chased from NanoC.
which is determined by the ideality factor n and the reverse Solar cells were prepared as small area (0.25 cm2) sandwich devices.
dark current I0 of the diode. This loss has been found to be The hole conductor PEDOT:PSS (H. C. Starck) was doctor-bladed
from an aqueous solution on a pre-cleaned, patterned indium tin ox-
typically on the order of 200 mV. In addition, a smaller part of ide (ITO)/glass substrate. The blend polymer/PCBM solution was
the Voc loss (∼ 100 mV or less) originates from the fact that doctor-bladed on top of the PEDOT:PSS. The devices were finalized
the photocurrent in bulk-heterojunction devices is dominantly by the deposition of a 1 nm thick LiF layer and a ∼100 nm aluminum
field-driven. The open-circuit voltage depends on the slope layer as top electrodes. LiF and aluminum were thermally evaporated
at a pressure <10–5 mbar (1 mbar 100 Pa). The geometry of the elec-
(= ls/d2) of the field-driven current around VBI, where l is the trode was defined by a shadow mask. Device preparation was per-
charge-carrier mobility, s is the charge-carrier lifetime, and d formed under ambient conditions, except for the evaporation of the
is the active-layer thickness. A steeper slope moves the Voc top electrode. Devices were characterized in a dry-argon atmosphere.
closer to VBI, which is typically observed for thin-film devices Current–voltage characteristics were measured using a Keithley
2400 SMU while the solar cells were illuminated by an ORIEL Solar
where the electric field is maximized across the active layer.
Simulator at 80 mW cm–2 (AM1.5D) white light. The electrochemical
The highest energy-conversion efficiencies derived above properties of all polymers were studied by cyclic voltammetry (CV).
are in the range of 10 % for single-bulk-heterojunction solar Electrochemical experiments were carried out on thin films at room
cells. This number appears achievable, although the donor temperature in a glove box under argon atmosphere using a comput-
er-controlled Jaissle 1002 T-NC potentiostat. The supporting electro-
material comprising all the required material properties has
lyte was tetrabutylammonium hexafluorophosphate (TBAPF6, 98 %,
probably not yet been prepared. To go beyond 10 % efficien- Aldrich, ∼ 0.1 M) in anhydrous acetonitrile (Aldrich). Platinum foils
cy, one could try to increase the external quantum efficiency were used as the working electrode (WE) as well as the counter elec-
of the device by improved light management and/or improv- trode (CE). Polymer films were deposited on the WE by drop-casting.
As a reference electrode (RE), a silver wire coated with AgCl was
ing FF by optimizing the electrical design. This would shift the
used. After each measurement, the RE was calibrated with ferrocene
efficiencies in Figure 2 to higher values, but would not change (oxidation potential E0 = 400 mV versus normal hydrogen electrode
the shape of the contour plot. Upon optimizing the open-cir- (NHE)), and the potential axis was corrected to NHE according to
cuit voltage of a bulk-heterojunction solar cell, the maximum the difference between E0 (ferrocene) and the measured E1/2 (ferro-
efficiency in Figure 2 would move to smaller bandgaps. This cene). The oxidation onset potentials were determined as the position
where the oxidation current starts to differ from the baseline. To eval-
may be achieved by reducing the losses resulting from a non- uate the accuracy of our measurement procedure, several polymers
perfect diode, as discussed above. In addition, upon reducing were also characterized by the more precise method, electrochemical
the required difference between the LUMO levels of the voltage spectroscopy (EVS) [22]. An excellent agreement between
donor and the acceptor (e.g., by developing materials with CV and EVS measurements was found, and therefore only oxidation
potentials extracted from CV measurements are used in this study.
very small exciton-binding energies), the energy-conversion Measured onset potentials were corrected to NHE and recalculated
efficiency of the bulk-heterojunction solar cell would ap- to electron volts versus vacuum level using a potential value of
proach the classical single-bandgap solar cell that has been –4.75 eV for NHE [23,24]. The LUMO level of PCBM was also deter-
discussed under various assumptions before.[3] mined by a CV measurement. The onset of the first reduction of
PCBM was found to start at a potential of –4.3 eV with respect to
In conclusion, it has been demonstrated that the open-cir- vacuum.
cuit voltage of bulk-heterojunction devices using PCBM as
acceptor is determined by the HOMO level of the donor and Received: August 18, 2005
Final version: December 29, 2005
the LUMO level of the acceptor molecule. Together with ear- Published online: February 22, 2006
lier investigations on the role of the LUMO level of fullerene
acceptors, this work confirms that the MIM model is not –
applicable to bulk-heterojunction devices. The observed [1] D. M. Chapin, C. S. Fuller, G. L. Pearson, J. Appl. Phys. 1954, 25,
676.
deviation of Voc from VBI (HOMO donor minus LUMO
[2] Clean Electricity from Photovoltaics (Eds: M. D. Archer, R. Hill),
acceptor) is related to the working principle of the bulk-het- Series on Photoconversion of Solar Energy, Vol. 1, Imperial College
erojunction solar cell and can be minimized by optimizing Press, London 2001.
materials, the active-layer thickness, and the charge-carrier [3] A. Goetzberger, C. Hebling, Sol. Energy Mater. Sol. Cells 2000, 62, 1.
mobility. Based on the presented findings, a relation between [4] C. J. Brabec, J. Hauch, P. Schilinsky, C. Waldauf, MRS Bull. 2005, 30,
energy-conversion efficiency of a bulk-heterojunction solar 50.
[5] Special Issue on “Organic Photovoltaics”, MRS Bull. 2005, 30, 1.
cell, bandgap, and the LUMO level of the donor was derived.
[6] N. S. Sariciftci, L. Smilowitz, A. J. Heeger, F. Wudl, Science 1992,
In addition, the model derived in this work can be used as a 258, 1474.
guideline for the material selection and material development [7] C. J. Brabec, G. Zerza, N. S. Sariciftci, G. Cerullo, S. DeSilvestri,
for bulk-heterojunction solar cells. S. Luzatti, J. C. Hummelen, Chem. Phys. Lett. 2001, 340, 232.

Adv. Mater. 2006, 18, 789–794 © 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.advmat.de 793
COMMUNICATIONS

[8] P. Schlinsky, C. Waldauf, C. J. Brabec, Appl. Phys. Lett. 2002, 81, [17] Q. Zhou, Q. Hou, L. Zheng, X. Deng, G. Yu, Y. Cao, Appl. Phys.
3885. Lett. 2004, 84, 1653.
[9] J. Xue, S. Uchida, B. P. Rand, S. R. Forrest, Appl. Phys. Lett. 2004, [18] V. D. Mihailetchi, L. J. A. Koster, J. C. Hummelen, P. W. M. Blom,
85, 5757. Phys. Rev. Lett. 2004, 93, 216 601.
[10] L. Schmidt-Mende, U. Bach, R. Humphry-Baker, T. Horizchi, [19] J. L. Brédas, D. Beljonne, V. Coropceanu, J. Cornil, Chem. Rev.
H. Miura, S. Ito, S. Uchida, M. Grätzel, Adv. Mater. 2005, 17, 813. 2004, 104, 4917.
[11] W. Ma, C. Yang, X. Gong, K. Lee, A. J. Heeger, Adv. Mater. 2005, [20] P. Schilinsky, C. Waldauf, J. Hauch, C. J. Brabec, J. Appl. Phys. 2004,
17, 1617. 95, 2816.
[12] C. J. Brabec, A. Cravino, D. Meissner, N. S. Sariciftci, T. Fromherz, [21] L. J. A. Koster, V. D. Mihailetchi, R. Ramaker, P. W. M. Blom,
M. T. Rispens, L. Sanchez, J. C. Hummelen, Adv. Funct. Mater. 2001, Appl. Phys. Lett. 2005, 86, 123 509.
11, 374. [22] D. Mühlbacher, Diplom Thesis, Johannes Kepler Universität Linz
[13] L. Liu, Y. Shi, Y. Yang, Adv. Funct. Mater. 2001, 11, 420. 2002.
[14] H. Frohne, S. E. Shaheen, C. J. Brabec, D. C. Müller, N. S. Sariciftci, [23] R. J. Gomer, G. Tryson, J. Chem. Phys. 1977, 66, 4413.
K. Meerholz, ChemPhysChem 2002, 3, 795. [24] R. Kötz, H. Neff, K. A. Müller, J. Electroanal. Chem. Interfacial
[15] A. Gadisa, M. Svensson, M. R. Andersson, O. Inganäs, Appl. Phys. Electrochem. 1986, 215, 331.
Lett. 2004, 84, 1609.
[16] M. Svensson, F. Zhang, S. C. Veenstra, W. J. H. Vehees, J. C. Hum-
melen, J. M. Kroon, O. Inganäs, M. R. Andersson, Adv. Mater. 2003,
15, 988.

______________________

794 www.advmat.de © 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 789–794