Self-assembling fullerenes for improved bulk-heterojunction photovoltaic devices

Full text of DOI:10.1021/ja807627u

Published on Web 12/02/2008
Self-Assembling Fullerenes for Improved Bulk-Heterojunction Photovoltaic
Devices
Robert D. Kennedy, Alexander L. Ayzner, Darcy D. Wanger, Christopher T Day, Merissa Halim,
Saeed I. Khan, Sarah H. Tolbert,* Benjamin J. Schwartz,* and Yves Rubin*
Department of Chemistry and Biochemistry, UniVersity of California, Los Angeles, California 90095-1569
Received September 26, 2008; E-mail: tolbert@chem.ucla.edu; schwartz@chem.ucla.edu; rubin@chem.ucla.edu
Scheme 1
The recent dramatic increase in the cost of fossil fuels has
generated a strong resurgence of interest in renewable solar energy
sources. Although highly efficient, inorganic photovoltaic devices
are costly. Advances in organic “bulk heterojunction” (BHJ)
photovoltaic cells have opened prospects for the development of
less expensive, easily produced alternatives.1 To date, the efficien-
cies of organic BHJ cells have reached 5-6%, still short of the ca.
10% required for commercialization.2
The most efficient BHJ devices have been based on conjugated
semiconducting polymers blended with a fullerene derivative that
serves as an electron acceptor to split excitons created when the
polymer absorbs light. These two components must be mixed
together on a length scale that is shorter than the exciton diffusion
length, typically ca. 10 nm, to ensure that all excitons are harvested.1
It is also necessary that the polymer and fullerene components form
a bicontinuous interpenetrating network for efficient carrier transport
of the separated charges. Creating an optimal BHJ is a delicate
balancing act: having the components blended too well prevents
the formation of separate polymer and fullerene networks, but too
much phase segregation leads to the formation of unconnected
islands that can act as carrier traps. Indeed, low carrier mobilities
in BHJ networks remain the major obstacle for improving device
efficiency.1 Thus, the key to improving BHJ devices is controlling
the nanometer-scale morphology of the interpenetrating organic
networks.
Several groups have presented methods for controlling the nm-
scale morphology of polymer/fullerene blends based on macroscopic
device-processing conditions. These include thermal annealing,3
solvent-vapor treatments,4 and the use of solvent additives that
preferentially dissolve the fullerene.5 Here, we propose a molecular
approach to the control of both the length-scale of polymer/fullerene
phase segregation and the electron mobility in the fullerene network.
Our method is based on replacing conventional fullerene electron
acceptors, such as [6,6]-phenyl-C61-butyric acid methyl ester
(PCBM), with self-assembling fullerene derivatives that may form
one-dimensional (1-D) wire-like domains within the active layer.
Specifically, we examine a pair of fullerene derivatives that are
chemically nearly identical (1a and 1b) but whose molecular shapes
either promote or hinder the formation of 1-D stacks. We find that
even with identical processing conditions, the nonstacking fullerene
1b gives poor performance in solar cells based on blends with
regioregular poly(3-hexylthiophene-2,5-diyl) (P3HT) but that the
stacking fullerene 1a leads to enhanced photovoltaic efficiency.
The new fullerene adducts 1a and 1b are related to those
synthesized by Nakamura and co-workers, who have developed a
highly regioselective and efficient penta-addition of aryl copper
reagents to C60 that produces molecules with “shuttlecock”-like
shapes (Scheme 1).6 An intriguing feature of these molecules is
that some of them tend to form columnar motifs in the solid state,7
in which the curved surface of one fullerene nestles inside the cavity
produced by the five arms of the adjacent molecule. In a future
study of the crystal structures of 20 of these adducts, we will show
that modification of the substituent arms allows control over their
propensity for stacking. Here we focus on BHJ solar cells prepared
from 1a and 1b, with the goal of understanding how molecular
control over fullerene-fullerene interactions might translate into a
more favorable BHJ morphology and thus improve electron mobility
and photovoltaic device performance.
The molecular design of fullerene shuttlecock adducts for use
in BHJ solar cells is challenging for a number of reasons: the
fullerene molecules must have a propensity to self-assemble into
desirable structures and display high solubility in solvents used for
device fabrication, two properties that are often mutually exclusive.
To determine the optimal geometry of a shuttlecock that is likely
to stack efficiently with short fullerene-fullerene contact distances,
we performed a series of molecular modeling calculations (AM1,
Spartan 04). These calculations indicated that the introduction of
the large tert-butyl group at the para-position of each of the phenyl
substituents produces a deep, well-shaped cavity that should
promote self-assembly. We thus synthesized compound 1a and
found that the addition of the para tert-butyl groups makes 1a an
efficient self-assembling 1-D “stacker” that also is suitably soluble
in o-dichlorobenzene (ODCB).
Following the general procedure of Nakamura et al.,6,8 1a
was obtained as a bright red solid and purified by column
chromatography on silica gel. Figure 1 shows the X-ray structure
of 1a · (C5H12)3 obtained by slow diffusion of n-pentane into a
saturated solution of 1a in 1-chloronaphthalene.9 In this structure,
the fullerenes form 1-D stacks that are aligned antiparallel and
separated by sheaths of n-pentane molecules. The distance
between fullerene centroids within each column is 10.83 Å, and
the shortest distances between fullerene centroids in adjacent
columns are 14.85 and 15.86 Å.
We also examined the crystal structures of 1a crystallized from
ODCB, CS2, and C6H5Cl and found 1a is a “universal stacker”:
crystals of 1a display a packing motif characterized by the assembly
of 1-D fullerene stacks that is solvent-independent.8 For the four
solvents we examined, it appears that the overall packing structure
is dictated by the self-assembly of the 1-D stacks and that the
solvent molecules simply fill in the voids.
9
17290 J. AM. CHEM. SOC. 2008, 130, 17290–17292
10.1021/ja807627u CCC: $40.75
2008 American Chemical Society

C O M M U N I C A T I O N S
16.55 Å (Figure 2e). The fullerenes within each layer are roughly
hexagonally packed, and each fullerene has five near-neighbors (C60
centroid-centroid distances e 10.193 Å). Nearly identical packing
is observed for 1b·C6H5CH3 (slow evaporation of PhCH3).10
However, the crystal structure of 1b·(C6H5Cl)1.875 is very different,
witheachfullerenehavingthreenear-neighbors(C60 centroid-centroid
distances e 10.318 Å), forming a layered, puckered, honeycomb
network (Figure 2c).10 These diverse structures make it clear that
1b does not tend to form 1-D stacks, so we shall use it as a model
nonstacking shuttlecock.
Figure 1. Crystal structure of 1a·(C5H12)3. (a) Antiparallel a-axis stacks,
viewed along the (a) bc plane and (b) a-axis. Hydrogens are removed for
clarity; n-pentane solvent molecules are shown in space-filling mode.
Clearly, the packing of 1a contrasts greatly with that of 1b, in
which subtle changes in the solvent produced entirely different
structural motifs. The difference can be rationalized on the basis
of molecular shape: the deeper cavity of 1a resulting from para-
substitution promotes efficient 1-D stacking, while the meta-
substituted phenyl arms of 1b produce only a shallow cavity.
An ideal BHJ solar cell should have two networks that are both
continuous and fully interconnected throughout the thickness of
the active layer. For polymer/fullerene-based BHJ solar cells, a
fullerene derivative with a propensity to form 1-D stacks may form
a more ideal network than one that makes isotropic aggregates that
could more easily lead to island formation. To test this idea, we
fabricated BHJ solar cells from blends of either 1a or 1b with
P3HT.8 Although we do not necessarily expect these blends to
contain fullerene aggregates with structures identical to those shown
in Figures 1 and 2, it seems likely that the robust stacking and
nonstacking molecular motifs found in the single crystals also will
be present in the devices.
Figure 2. Crystal structure of 1b·CHCl3. (a) ac layer, viewed along the
b axis, with disordered CHCl3 solvent shown in space-filling representation.
(b) Single b-axis chain, viewed along the a axis. Carbon atoms with close
contacts (3.412-3.551 Å) are highlighted. (c) Crystal structure of
1b·(C6H5Cl)1.875. Single bc layer, viewed along the a axis, with selected
solvent molecules shown in space-filling representation. (d) Crystal structure
of 1b·C6H4Cl2. Single bc layer, viewed along the a axis, with the solvent
removed for clarity. (e) bc Layers for 1b·C6H4Cl2 viewed along the b axis,
with solvent molecules shown in space-filling representation. Hydrogen
atoms are removed for clarity in all structures above.
Figure 3. (a) I-V curves for as-cast and thermally annealed BHJ solar
cells made from a 1:0.35 blend of 1a and P3HT. Annealing of these cells
also occurs rapidly, with significantly improved device efficiencies compared
to the data in (b). (b) I-V curves for as-cast and thermally annealed BHJ
solar cells made from a 1:0.45 blend of 1b and P3HT. Annealing of these
cells is complete in 5 min, but the overall device performance is low.
Because the reduction of the local symmetry of a group tends to
increase overall molecular solubility, we also investigated the 3-tolyl
system 1b: the solubility of this molecule in ODCB is ca. 10 mg/
mL, which is high enough for the ready fabrication of devices by
spin-coating from a concentrated solution of this fullerene mixed
with P3HT. Crystals suitable for single-crystal X-ray diffraction
were grown from five different solvent systems, resulting in very
different extended crystal structures.10 Figure 2a,b shows the
packing structure of compound 1b·CHCl3 grown from a concen-
trated solution in a CHCl3/CS2 mixture. The chloroform molecules
reside just within the cavity formed by the five meta-tolyl groups,
and the fullerene-solvent units form a close-packed 3-D diamond-
like network in which each fullerene has four near-neighbors
arranged in an approximately tetrahedral geometry. The fullerene
centroid-centroid distances are 10.436 (edge-vertex close contacts,
Figure 2a) and 9.894 Å (hexagon-hexagon, Figure 2b), respec-
tively. An identical motif is observed in the crystal structure of
1b·CS2, grown by slow diffusion of n-pentane into a CS2 solution
of 1b. In contrast, 1b·1,2-C6H4Cl2 (Figure 2d,e) forms a layered
structure.10 Although the cocrystallized ODCB molecules reside
just within the cavity in a similar fashion to 1b ·CHCl3, alternating
layers of fullerene and solvent are observed with a periodicity of
A comparison between the current density-voltage (I-V) charac-
teristics of as-cast P3HT:1a and P3HT:1b devices11 under simulated
AM1.5 solar irradiation8 is shown in Figure 3a,b. The I-V curves
show that both the short-circuit current density (Isc) and open-circuit
voltage (Voc) are larger for the devices fabricated with 1a. The larger
value of Isc for 1a is consistent with the idea of this stacking shuttlecock
producing a more interconnected network compared to 1b. The
difference in Voc is less straightforward to interpret. The value of Voc
is thought to depend only on the difference of the donor HOMO and
acceptor LUMO energies,12 but these devices use the same donor and
the LUMO levels of 1a and 1b are nearly identical. Thus, the difference
must be related to the nm-scale morphology of the fullerene network.
One explanation is that the different fullerene networks have different
densities of charge-carrier traps, which contribute to the space-charge
buildup and thus affect the built-in potential in the device. Alternatively,
the different fullerene aggregation geometries may affect the self-
organization and/or crystallinity of the P3HT: a change in the
9
J. AM. CHEM. SOC. VOL. 130, NO. 51, 2008 17291

C O M M U N I C A T I O N S
crystallinity or average conjugation length could be reflected in the
spatially averaged HOMO of the polymer, thereby directly affecting
Voc.
shuttlecock molecules. PCBM has unusually short fullerene C-C
contact distances, as observed in the crystal structures of
PCBM·PhCl (g2.902 Å) and PCBM·ODCB (g3.171 Å); this
high degree of contact should greatly favor electron mobility.16
Even with the high propensity of 1a to form columnar stacks, the
closest fullerene-fullerene contacts in the crystal structures are
relatively long (g4.035 Å, 1a·(C5H12)3). We attribute this both to
the repulsive interaction between meta-aryl hydrogens of the
shuttlecock “feathers” with the neighboring fullerene ball, which
prevents closer approach of molecules along the stacks, and to the
interstack steric crowding by the “feathers” that keep individual
columns well separated. The shuttlecock system, however, allows
for incredible chemical diversity within a single molecular motif.
We are thus currently exploring crystal structures and BHJ device
behavior of a large range of penta-substituted fullerene derivatives
with the goal of improving BHJ solar cell efficiency by simulta-
neously optimizing the propensity for fullerenes to stack, the
fullerene-fullerene contact distances, and the degree of phase
segregation with conjugated polymers.
Since thermal annealing frequently improves the efficiency of
polymer-based BHJ solar cells, we chose to anneal our solar cells
based on blends of P3HT with 1a or 1b under conditions similar
to those used to improve the efficiency of P3HT:PCBM solar cells.3
As with P3HT:PCBM devices, annealing our devices based on
1a and 1b decreased Voc but increased Isc, leading to an overall
increase in efficiency (Figure 3). The reduction in Voc is generally
assigned to reduced space-charge buildup while the increase in Isc
is related to improved network formation. We find that the
annealing-induced increase in Isc, which is the most direct indicator
of the underlying network structure, is less than half as large for
1b compared to 1a. Since the electronic structures of these two
fullerenes are essentially identical, the larger improvement in
photovoltaic performance upon annealing devices based on tert-
butyl system 1a must be associated with the increased self-
organization of this fullerene derivative.
Thermal annealing is a kinetically slow process that requires the
diffusive motion of molecules. Indeed, P3HT:PCBM devices show
a steady improvement in performance with thermal annealing for
ca. 20 min, after which time the performance plateaus.3 This large
improvement does not occur for devices that have been subjected
to “solvent annealing” by casting from slow-evaporating solvents
such as ODCB.4 In contrast, we find that thermal annealing has a
pronounced effect on shuttlecock-based devices cast out of ODCB.
Moreover, this thermally induced improvement occurs quickly: for
devices based on both 1a and 1b, no further improvement in the
photovoltaic efficiency is produced past 5 min of annealing (Figure
3a,b). We also find that device performance is relatively insensitive
to the amount of 1a but depends critically on the concentration of
PCBM.3 This indicates that there are qualitative differences in both
the degree and nature of the phase segregation for shuttlecock-
based devices relative to photovoltaic devices based on the
conventional PCBM.
The reason that thermal annealing improves the efficiency of
polymer-based BHJ solar cells is still the subject of investigation:
it is known that annealing increases the crystallinity of P3HT, thus
improving the hole mobility in the device,13 but the effects of
annealing on the fullerene network are less clear. Annealing is
known to promote phase segregation, which could either enhance
electron mobility due to better fullerene-fullerene contacts14 or
decrease electron mobility as phase segregation produces islands
within the fullerene network.11,15 The strong propensity of 1a to
form 1-D stacks regardless of solvent conditions (Figure 1) suggests
that this molecule may be less likely to form islands or other
unconnected structures upon thermal annealing. In contrast, fullerenes
like 1b that tend to aggregate into more isotropic structures may
promote the same type of undesirable phase segregation that is
observed with traditional fullerene derivatives such as PCBM.
Moreover, it is possible that the packing of the P3HT chains is
strongly dependent on the nanometer-scale morphology of the
fullerene derivative, so that the propensity of 1a to stack in a
columnar motif could be conducive to improved polymer π-π
interactions and thus lead to higher hole mobility.
Acknowledgment. This work was supported by research and
instrumentation grants ((NSF-CHE-0527015 and ONR-N00014-04-
1-0410; Y.R./B.S./S.T.), NSF-CHE-0617052 (Y.R.), NSF-CHE-
9871332 (X-ray), and NSF-CHE-9974928 (NMR)).
Supporting Information Available: Experimental procedures,
spectra and CIF files for 1a·(C5H12)3, 1b·C6H5CH3, 1b·(C6H5Cl)1.875
,
1b·1,2-C6H4Cl2, 1b·CHCl3, 1b·CS2. This material is available free
of charge via the Internet at http://pubs.acs.org.
References
(1) Hoppe, H.; Saricifci, S. J. Mater. Chem. 2006, 16, 45–61.
(2) Li, G.; Shrotriya, V.; Yao, Y.; Huang, J.; Yang, Y. J. Mater. Chem. 2007,
17, 3126–3140.
(3) Ma, W.; Yang, C.; Gong, X.; Lee, K.; Heeger, A. AdV. Funct. Mater. 2005,
15, 1617–1622.
(4) Li, G.; Yao, Y.; Yang, H.; Shrotriya, V.; Yang, G.; Yang, Y. AdV. Funct.
Mater. 2007, 17, 1636–1644.
(5) Lee, J. K.; Ma, W. L.; Brabec, C.; Yuen, J.; Moon, J. S.; Kim, J. Y.; Lee,
K.; Bazan, G.; Heeger, A. J. Am. Chem. Soc. 2008, 130, 3619–3623.
(6) (a) Sawamura, M.; Iikura, H.; Nakamura, E. J. Am. Chem. Soc. 1996, 118,
12850–12851. (b) Sawamura, M.; Iikura, H.; Ohama, T.; Hackler, U. E.;
Nakamura, E. J. Organomet. Chem. 2000, 599, 32–36. (c) Matsuo, Y.;
Nakamura, E. Chem. ReV. 2008, 108, 3016–3028.
(7) (a) Sawamura, M.; Kawai, K.; Matsuo, Y.; Kanie, K.; Kato, T.; Nakamura, E.
Nature 2002, 419, 703–705. (b) Zhong, Y.-W.; Matsuo, Y.; Nakamura, E.
Org. Lett. 2006, 8, 1463–1466. (c) Zhong, Y.-W.; Matsuo, Y.; Nakamura,
E. J. Am. Chem. Soc. 2007, 129, 3052–3053. (d) Matsuo, Y.; Tahara, K.;
Morita, K.; Matsuo, K.; Nakamura, E. Angew. Chem., Int. Ed. 2007, 46,
2844–2847.
(8) See Supporting Information.
(9) X-ray data for 1a · (C5H12)3: Red platelet (0.3 × 0.2 × 0.2 mm3) from
1-chloronaphthalene/C5H12. Space group: Pnma; a ) 21.661(3) Å; b )
22.721(4) Å; c ) 17.128(3) Å; V ) 8430(2) Å3; Z ) 4; T ) 100(2) K;
Nref(unique) ) 11 521; R1 ) 0.076; Rw ) 0.196.
(10) X-ray data for 1b ·CHCl3: red platelet (0.4 × 0.1 × 0.1 mm3) from CHCl3/
CS2. Space group: Pnma; a ) 18.298(5) Å; b ) 17.955(5) Å; c ) 17.352(5)
Å; V ) 5701(3) Å3; Z ) 4; T ) 100(2) K; Nref(unique) ) 7277; R1
)
0.077; Rw ) 0.200. 1b · CS2: Red prism (0.2 × 0.2 × 0.05 mm3) from
CS2/C5H12. Space group: Pnma; a ) 18.311(5) Å; b ) 17.969(4) Å; c )
17.344(5) Å; V ) 5707(3) Å3; Z ) 4; T ) 100(2) K; Nref(unique) ) 7858;
R1 ) 0.063; Rw ) 0.171. For 1b · 1,2-C6H4Cl2, 1b · C6H5CH3, and
1b · (C6H5Cl)1.875, see Supporting Information.
(11) Ayzner, A. L.; Wanger, D. D.; Tassone, C. J.; Tolbert, S. H.; Schwartz,
B. J. J. Phys. Chem. C, in press (ASAP available at, DOI: 10.1021/
jp8076497).
(12) Branec, C.; Cravino, A.; Meissner, D.; Sariciftci, S.; Fromherz, T.; Rispens,
M.; Sanchez, L.; Hummelen, Y. AdV. Funct. Mater. 2001, 11, 374–380.
(13) Erb, T.; Zhokhavets, U.; Gobsch, G.; Raleva, S.; Stuhn, B.; Schilinsky, P.;
Waldauf, C.; Brabec, C. AdV. Funct. Mater. 2005, 15, 1193–1196.
(14) Dante, M.; Peet, J.; Nguyen, T.-Q. J. Phys. Chem. C 2008, 112, 7241–
7249.
We close by noting that the efficiency of the photovoltaic devices
based on 1a reaches up to ca. 1.5%, lower than the 5-6% achieved
in devices based on PCBM.1,2 Although our processing conditions
(e.g., polymer-fullerene ratio, solvent, spin-coating conditions, etc.)
will benefit from further optimization, we believe that a greater
parameter to improve lies in the optimization of contacts between
the fullerene moieties within the solid-state packing of the
(15) Warman, J.; de Haas, M. P.; Anthopoulos, T. D.; de Leeuw, D. M. AdV.
Mater. 2006, 18, 2294–2298.
(16) Rispens, M. T.; Meetsma, A.; Rittberger, R.; Brabec, C. J.; Sariciftci, N. S.;
Hummelen, J. C. Chem. Commun. 2003, 2116–2118.
JA807627U
9
17292 J. AM. CHEM. SOC. VOL. 130, NO. 51, 2008