Enhancing Fullerene-Based Solar Cell Lifetimes by Addition of a Fullerene Dumbbell

Article abstract of DOI:10.1002/anie.201407310

Cost-effective, solution-processable organic photovoltaics (OPV) present an interesting alternative to inorganic silicon-based solar cells. However, one of the major remaining challenges of OPV devices is their lack of long-term operational stability, esp

Full text of DOI:10.1002/anie.201407310

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Angewandte
Chemie
DOI: 10.1002/anie.201407310
Organic Photovoltaics
Enhancing Fullerene-Based Solar Cell Lifetimes by Addition of
a Fullerene Dumbbell**
Bob C. Schroeder,* Zhe Li,* Michael A. Brady, Gregꢀrio Couto Faria, Raja Shahid Ashraf,
Christopher J. Takacs, John S. Cowart, Duc T. Duong, Kar Ho Chiu, Ching-Hong Tan,
Jo¼o T. Cabral, Alberto Salleo, Michael L. Chabinyc, James R. Durrant, and Iain McCulloch
Abstract: Cost-effective, solution-processable organic photo-
voltaics (OPV) present an interesting alternative to inorganic
silicon-based solar cells. However, one of the major remaining
challenges of OPV devices is their lack of long-term opera-
tional stability, especially at elevated temperatures. The syn-
thesis of a fullerene dumbbell and its use as an additive in the
active layer of a PCDTBT:PCBM-based OPV device is
reported. The addition of only 20% of this novel fullerene
not only leads to improved device efficiencies, but more
importantly also to a dramatic increase in morphological
stability under simulated operating conditions. Dynamic
secondary ion mass spectrometry (DSIMS) and TEM are
used, amongst other techniques, to elucidate the origins of the
improved morphological stability.
The morphology of polymer:PCBM BHJs is crucial for
optimal PCE; it is important to achieve the right trade-off
between domain size and finely interconnected network
throughout the bulk of the active layer that allows efficient
charge generation and extraction. The morphology can be
controlled during device fabrication by employing solvent
mixtures to deposit the active layer, adding high boiling
solvent additives or alternatively by post-deposition treat-
ments like solvent vapor or thermal annealing.[4] However
once the active layer is deposited and the desired morphology
achieved, it is essential to “lock” that morphology in order to
maintain high power efficiencies and to avoid device degra-
dation over time. The “locking” of the active layer morphol-
ogy is extremely non-trivial, due in part to the fullerene
derivatives miscibility in the polymer phase, where they
readily diffuse throughout the polymer on short timescales.[5]
Over an extended period of time or under thermal stress, the
fullerene derivatives, especially PC61BM, form micron-scale
crystallites within the polymer phase acting as charge traps
and inhibiting efficient charge transport to the electrodes.[6]
One of the most common approaches to avoid aggrega-
tion of fullerenes into large domains is to chemically crosslink
either or even both the materials in the BHJ after deposi-
tion.[7] Crosslinking, usually triggered by external stimuli (e.g.
heat or light), has been successfully used in lithography for
decades, but has proven rather difficult to apply to organic
semiconductor thin films. In most cases crosslinking the active
layer inadvertently causes a significant performance decrease.
We have recently demonstrated a new approach upon which
T
he power conversion efficiency (PCE) of single-junction
organic solar cells has increased significantly during the last
decade to 9–10%, now approaching the threshold considered
necessary to commercialize the technology.[1] During this
period, the structural diversity of semiconducting donor
polymers for solar cells has increased dramatically.[2] These
materials have helped enable an accelerated development of
bulk heterojunction (BHJ) organic solar cells based on
polymer donor materials and molecular fullerene derivatives.
However, the development of electron-accepting materials
that lead to BHJs with high PCE has been significantly
slower.[3] The most commonly used n-type acceptors to date
remain [6,6]-phenyl-C61-butyric acid methyl ester (PC61BM)
and its slightly larger counterpart PC71BM.
[*] Dr. B. C. Schroeder, Dr. Z. Li, Dr. R. S. Ashraf, K. H. Chiu, C.-H. Tan,
Prof. J. R. Durrant, Prof. I. McCulloch
Department of Chemistry and
Dr. G. C. Faria
S¼o Carlos Physics Institute, University of S¼o Paulo
PO.Box: 369, 13560-970, S¼o Carlos, SP (Brazil)
Centre for Plastic Electronics
[**] The authors would like to acknowledge the Imperial College High
Performance Computing Service for providing the computing
facilities to perform the quantum-mechanical calculations. B.C.S.
gratefully acknowledges funding from the EPSRC knowledge trans-
fer secondment scheme. Financial support from Solvay SA is
grateful acknowledged. This work was in part carried out under the
EPSRC projects EP/F056710/1, EP/I019278/1, and EP/G037515/
1 with support from the Centre for Plastic Electronics at Imperial.
M.L.C. and M.A.B. were partially funded by NSF CHE ICC award
1026664. Thanks to Christine Schroeder-Schanck for designing the
TOC graphic.
Imperial College London, London, SW7 2AZ (UK)
E-mail: b.schroeder10@imperial.ac.uk
zhe.li@imperial.ac.uk
M. A. Brady, Dr. C. J. Takacs, J. S. Cowart, Prof. M. L. Chabinyc
Materials Department
University of California Santa Barbara
Santa Barbara, California 93106 (USA)
Dr. G. C. Faria, Dr. D. T. Duong, Prof. A. Salleo
Department of Materials Science and Engineering
Stanford University
476 Lomita Mall, Stanford, California 94305 (USA)
Supporting information for this article, including synthetic details
and additional graphs (CV, UV/Vis, OPV, DSC, OFET) and images
(microscopy, TEM), is available on the WWW under http://dx.doi.
org/10.1002/anie.201407310.
Dr. J. T. Cabral
Department of Chemical Engineering and
Centre for Plastic Electronics, Imperial College London
London, SW7 2AZ (UK)
Angew. Chem. Int. Ed. 2014, 53, 1 – 7
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1
These are not the final page numbers!

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
.
Angewandte
Communications
the active layer morphology can be stabilized by the photo-
induced [2+2] cycloaddition of the PC61BM derivatives,
which results in covalent bonding between the fullerene
groups.[8] The strategy was found to hold generally for
a variety of polymer–fullerene pairs.[6c,9] Light soaking
significantly reduced the formation of PC61BM crystallites
under thermal stress, and thus the device lifetime subse-
quently improved. Mechanistic studies indicated the enhance-
ment in device thermal stability is primarily associated with
inhibition of the nucleation of nanoscale PC61BM aggrega-
tes.[6c] These observations have motivated us to consider
whether enhancement of device stability could also be
achieved by the direct addition of covalently linked PC61BM
dimers in solution prior to deposition of the active layer. Such
an approach could provide a more structurally controlled and
robust route to inhibit PC61BM aggregation, as well as
enabling direct quantification of the degree of PC61BM
dimerization required to enhance device stability, thus
avoiding limitations associated with thermally induced dis-
sociation observed for photoinduced PC61BM oligomers.[6c]
Our synthetic strategy was to make dumbbell-shaped
dimeric fullerenes where the fullerenes are linked by an alkyl
bridge between the ester functional group on PCBM. In a first
step commercially available PCBM was hydrolyzed using
a previously described literature procedure and the resulting
PCBA was recovered as a black soot.[10] The PCBA was
immediately used in the next reaction step without any
further purification. The use of thionyl chloride to convert the
PCBA into the corresponding acyl chloride was disregarded
because thionyl chloride potentially p-dopes carbon-based
molecules like fullerenes, leading ultimately to poor device
efficiencies.[11] To prevent undesired doping with thionyl
chloride, two equivalents of PCBA were reacted with one
equivalent of ethylene glycol under mild reaction conditions
using Steglich esterification. Isolating the desired (PCB)2C2
dumbbell in sufficient purity from the complex reaction
mixture however proved difficult and gave unsatisfactory
yields (< 10%). As a result of the impossibility to purify the
poorly soluble PCBA starting material, it is likely that the acid
and diol were not present in the required stoichiometric ratio,
leading to the formation of a variety of compounds, difficult
to separate. To circumvent these problems, the esterification
reaction was performed in two steps as shown in Scheme 1.
The crude PCBA was reacted with a large excess of diol
(10 equivalents) to encourage the formation of PCBC2OH.
The highly soluble PCBC2OH intermediate was readily
purified by column chromatography and recovered in good
yields. After a second esterification between PCBC2OH and
PCBA, the (PCB)2C2 dumbbell could be successfully assem-
bled and isolated in reasonable yield and excellent purity as
evidenced by the H-MALDI and NMR spectra shown in
1
Figures S5 and S10 in the Supporting Information. The singlet
associated with the terminal methyl protons in PCBM
appears at 3.67 ppm, whereas the singlet of the bridging
methylene protons in (PCB)2C2 are shifted towards lower
fields and appear at 4.26 ppm. All remaining signals associ-
ated with the aromatic and alkyl protons are identical in both
cases and confirm that our dimerization approach has no
effect on the chemical environment of the fullerenes.
Because our goal was to blend the dimeric fullerenes with
PCBM, we investigated the impact of the dimerization on
their electronic structure. There have been numerous
approaches to synthesize covalently bound fullerene dimers,
however most of these synthetic procedures also modify the
frontier energy levels of the fullerene dumbbells.[12] From this
perspective our controlled dimerization approach is advanta-
geous, allowing the synthesis of structurally well-defined
PCBM dumbbells with similar frontier energy levels to
PCBM, as evidenced by density functional theory (DFT)
calculations (Table S1) and cyclic voltammetry (CV) meas-
urements (Figure S11). In both cases the HOMO was
estimated at À5.94 eV and the LUMO at À3.85 eV, thus
confirming the trend predicted by DFT calculations. The
result might seem intuitive, but ensuring similar energy levels
for both PCBM and the dimer (PCB)2C2 is crucial in order to
achieve high PCE. In case the LUMO energy level of the
dimer would be significantly lower than the one of PCBM, the
added dimer would act as a charge trap, which would be
detrimental for device performance.
To investigate the effect of addition of (PCB)2C2 on the
active layer morphology and device performance, a blend of
PCBM with the carbazole-based PCDTBT polymer was
chosen as model system. PCDTBT is a high-performing
amorphous polymer with a glass-transition temperature (Tg
ꢀ 1068C) in neat films below the temperatures used here for
examination of thermal stability.[13] To determine the mini-
mum weight percentage of (PCB)2C2 dumbbell necessary to
stabilize the PCDTBT:PCBM blend under thermal stress,
blends of PCDTBT:fullerene (1:2) were spin-coated on
silicon substrates from chlorobenzene and thermally
annealed at 1408C for 1 hour. We have studied the effect of
dimer content in the fullerene component, while maintaining
the overall weight ratio of polymer:fullerene constant. From
the optical micrographs presented in Figure S14, it is apparent
that the addition of dimer leads to a more thermally stable
blend. At low dimer loadings (< 1% by mass) the growth of
micron-sized PCBM crystallites in the blend cannot be
inhibited. However, the formation of micron-sized PCBM
Scheme 1. Synthesis of the PCBM dumbbell (PCB)2C2.
2
www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 1 – 7
These are not the final page numbers!

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Angewandte
Chemie
Figure 1. Optical images of a) PCDTBT:PCBM as-cast blend films (scale bar=200 mm); b) PCDTBT:PCBM:(PCB)2C2 (20%) as-cast films (scale
bar=200 mm); c) PCDTBT:PCBM blend films after thermal annealing at 1408C for 1 h (scale bar=200 mm)and d) PCDTBT:PCBM:(PCB)2C2
(20%) blend films after thermal annealing at 1408C for 1 h on SiOx substrates (scale bar=200 mm). The insets in (c,d) highlight the drastic
differences in PCBM crystal formation in the annealed blend films with and without the (PCB)2C2 (scale bar=40 mm). AFM images of
e) PCDTBT:PCBM as-cast blend films (scale bar=5.0 mm); f) PCDTBT:PCBM:(PCB)2C2 (20%) as-cast films (scale bar=5.0 mm);
g) PCDTBT:PCBM blend films after thermal annealing at 858C for 1 h (scale bar=5.0 mm); and h) PCDTBT:PCBM:(PCB)2C2 (20%) blend films
after thermal annealing at 858C for 1 h on PEDOT:PSS substrates (scale bar=5.0 mm).
morphology upon thermal
annealing at 858C, cus-
tomary for standard ther-
mal stress tests, on
PEDOT:PSS substrates is
shown in Figure 1.[15] As
revealed by AFM, as cast
samples are relatively
smooth and featureless,
with
a similar surface
Figure 2. A) Number density of micron-sized PCBM crystallites formed in blend films after thermal annealing at
1408C for 1 h on SiOx/Si substrates, plotted as a function of the (PCB)2C2 dimer loading. B) Comparison of the
J–V characteristics of optimized conventional PCDTBT:PCBM devices and PCDTBT:PCBM:(PCB)2C2 (20%).
C) Degradation of solar cell PCE as a function of annealing time at 858C in nitrogen atmosphere. The error bars
represent the spread in degradation kinetics of three typical devices
roughness for both with
and without (PCB)2C2.
Upon thermal annealing
at 858C for 1 h, the sam-
ples without (PCB)2C2
exhibit densely formed
crystallites at temperatures above the Tg of the pure polymer
is completely impeded upon addition of 20% of (PCB)2C2
(Figure 1), and indeed the number density of the PCBM
crystallites is suppressed by over 50% by adding only 5% of
dimer to the blend (Figure 2A). This low threshold is
consistent with our previous conclusion that the primary
role of the PC61BM oligomers, respectively (PCB)2C2, is to
frustrate and eventually inhibit PC61BM crystal nucleation,
with such nucleation processes often being hindered by low
compositions of additive materials.[6c,14]
The effect of (PCB)2C2 on the nanomorphological behav-
ior upon thermal stress analogous to solar-cell-operating
conditions was also investigated. Our previous studies have
shown that thermal stress at temperatures relevant to device
operation does not lead to micron-size crystallites, but instead
results in the formation of nanometer-sized PCBM features in
PCDTBT:PCBM blend films.[6c] The evolution of the blend
PCBM crystallites of the order of 150 nm in diameter and
30 nm in height. We have previously assigned these nanoscale
features to PCBM aggregates, and correlated their appear-
ance with the degradation of device efficiency under modest
thermal stress.[6c] In clear contrast, the samples with 20%
(PCB)2C2 show drastically suppressed PCBM aggregation
with no measurable change in film surface morphology after
thermal treatment.
The film morphology was further probed by bright-field
transmission microscopy (TEM). At 808C, both blend films,
with and without (PCB)2C2, present very comparable fea-
tures. The non-(PCB)2C2 containing blend contains spherical
aggregates dispersed homogenously throughout the film
(Figures S15 to S17), with diameters ranging from about
150–250 nm. The (PCB)2C2 containing blend showed compa-
rable features, but a much lower density of fullerene nuclei of
approximate diameter of 100–150 nm was found. This modest
Angew. Chem. Int. Ed. 2014, 53, 1 – 7
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3
These are not the final page numbers!

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
.
Angewandte
Communications
difference suggests that the observed spherical aggregates in
the non-(PCB)2C2 containing blend might play a key role in
PCBM crystal nucleation and device degradation even though
the exact mechanism is not yet elucidated. Upon thermal
annealing at 1408C, the formation of fullerene nuclei varying
in size (about 50–100, 100–150, and 150–200 nm) was
observed in the non-(PCB)2C2 containing blend. More
importantly however, we observed the formation of needle-
like structures, similar to the ones previously observed in
optical microscopy image, but much smaller in size (50–
100 nm in length and 10 nm in width). In the annealed blend
film containing 20% of (PCB)2C2 these sheaf-like structures
were not observed, but the morphology was instead charac-
terized by the presence of spherical aggregates, on average
smaller than the ones observed prior to 1408C annealing.
These results are in further agreement with our previously
made observation that a low concentration of (PCB)2C2 in the
fullerene phase is stabilizing the blend nanomorphology and
seems to influence the shape and number density of
0% (100% of PCBM) to 100% (0% of PCBM) while keeping
the total concentration of fullerene in the active layer blend
constant. The initial device J–V characteristics, as a function
of the (PCB)2C2 loading, are shown in Figure S18. From 0 to
20% of (PCB)2C2 loading, modest increases in PCE are
observed because of enhanced FF values. However, further
increase in dimer loading results in decreasing device PCE,
caused by decreases in both Jsc and FF (see Table S2 in the
Supporting Information for detailed device parameters). The
loss in PCE results in part from a higher series resistance,
indicative of a drop in electron mobility in the fullerene phase
with increasing PCBM dimer loading. To confirm this
assumption, organic field-effect transistors (OFET) were
built to measure the electron mobility as a function of
(PCB)2C2 loading. The electron mobility of a neat PCBM film
was found to be around 1 ꢀ 10À3 cm2 VÀ1 sÀ1, which is in good
agreement with previously reported literature values.[16] As
evidenced from Figure 3A, a sharp drop in electron mobility
PCBM aggregates under modest thermal stress.
Knowing that the PCBM dumbbell has a positive
effect on the thermal stability of active layer blend
films, we focused on its effect on the performance of
PCDTBT:PCBM solar cells. A standardized ther-
mal stress of 858C was applied to the devices under
nitrogen atmosphere, with the current–voltage (J–
V) curve recorded using repeated dark thermal
annealing/room-temperature device testing cycles.
For J–V measurements, devices were exposed under
a solar simulator for less than 30 seconds during
each measurement to minimize the effect of PCBM
photo-oligomerization. Figure 2 demonstrates the
Figure 3. A) Electron mobility of PCBM:(PCB)2C2 measured in organic field-effect
transistors as a function of (PCB)2C2 loading. B) DSC first heating thermogram of
PCBM, (PCB)2C2, and PCBM:(PCB)2C2 (20%) after the samples were thermally
impact of a low percentage of PCBM dimers on the annealed at 858C during two hours. Heating rate 108CminÀ1
efficiency and thermal stability of PCDTBT:PCBM
.
devices. Devices containing 20% of (PCB)2C2 show
a modest improvement in efficiency compared to devices
without dimer. This improvement is primarily due to an
enhancement in fill factor (FF) from around 0.56 to 0.62,
whereas both Voc and Jsc are nearly identical in both sets of
devices (Table S2). The cause of improved FF is not obvious
from the AFM images (e) and (f) in Figure 1, but is likely due
to an improvement in the blend morphology induced by the
dimers during film deposition. For the devices fabricated
without (PCB)2C2, a rapid degradation in device performance
was observed upon thermal stress, with a 20% loss in power
conversion efficiency (PCE) within the first 25 min, primarily
due a loss of fill factor, and likely linked to the densely formed
nanoscale PCBM aggregates. Remarkably, devices containing
20% of the (PCB)2C2 show significantly improved thermal
device stability with a reduction in the PCE by 20% occurring
only after 2000 minutes. It is thus apparent that the device
thermal stability of conventional PCDTBT:PCBM devices
can be enhanced by an order of magnitude by using a low
percentage of (PCB)2C2, at least under the thermal stress
conditions and time scale studied herein.
is observed at (PCB)2C2 loadings exceeding 10–20%, which
correlates well with the lower Jsc measured at higher dumbbell
loadings and the resulting drop in PCE. These findings are in
line with previous work by Distler et al., who demonstrated
that photodimerization of PCBM causes a significant reduc-
tion of charge carrier mobility in a polymer–fullerene system,
resulting in a reversible degradation of device efficiency
under constant illumination.[17] The electron mobility meas-
urements in conjunction with the device stability data there-
fore conclusively show that there is an optimal (PCB)2C2
loading around 20% at which both the device efficiency and
the device lifetime can be simultaneously enhanced.
Based on the aforementioned results, the addition of
(PCB)2C2 leads to improved power conversion efficiencies,
mainly because the growth of nanometer-sized PCBM
crystallites in the active layer under thermal stress is
prevented as evidenced by AFM measurements. To further
investigate by what mechanism this crystallite growth is
suppressed, we have carried out interdiffusion measurements.
We hypothesized above that the addition of (PCB)2C2 would
lead to lower mass diffusion coefficient, thus slowing down
and eventually preventing, the growth of PCBM nanocrys-
tallites. The diffusion of (PCB)2C2 from a bulk heterojunction
blend with PCDTBT into a neat PCDTBT thin film was thus
Given that the addition of (PCB)2C2 not only enhances
the long-term stability, but also the device efficiency, we
evaluated the potential to further enhance the performance
by systematically increasing the PCBM dimer loading from
4
www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 1 – 7
These are not the final page numbers!