Steric control of the donor/acceptor interface: Implications in organic photovoltaic charge generation

Article abstract of DOI:10.1021/ja203235z

The performance of organic photovoltaic (OPV) devices is currently limited by modest short-circuit current densities. Approaches toward improving this output parameter may provide new avenues to advance OPV technologies and the basic science of charge tra

Full text of DOI:10.1021/ja203235z

ARTICLE
pubs.acs.org/JACS
Steric Control of the Donor/Acceptor Interface: Implications in
Organic Photovoltaic Charge Generation
Thomas W. Holcombe,† Joseph E. Norton,# Jonathan Rivnay,^ Claire H. Woo,‡,§ Ludwig Goris,^,r
Claudia Piliego,§ Gianmarco Griffini,§,O Alan Sellinger,^ Jean-Luc Brꢀedas,# Alberto Salleo,^ and
||
Jean M. J. Frꢀechet*,†,‡,§,
†Department of Chemistry, and ‡Department of Chemical Engineering, University of California Berkeley,
Berkeley, California 94720-1460, United States
§Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States
King Abdullah University of Science and Technology, Thuwal, Saudi Arabia 23955-6900
^Department of Materials Science and Engineering, Stanford University, Stanford, California 94305, United States
#Center for Organic Photonics and Electronics and School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta,
Georgia 30332, United States
rInstitute for Materials Research (IMO), Hasselt University, Diepenbeek, Belgium
ODepartment of Chemisty, Politecnico di Milano, 20133 Milan, Italy
S Supporting Information
b
ABSTRACT: The performance of organic photovoltaic (OPV) devices is
currently limited by modest short-circuit current densities. Approaches toward
improving this output parameter may provide new avenues to advance OPV
technologies and the basic science of charge transfer in organic semiconductors.
This work highlights how steric control of the charge separation interface can be
effectively tuned in OPV devices. By introducing an octylphenyl substituent onto
the investigated polymer backbones, the thermally relaxed charge-transfer state, and potentially excited charge-transfer states, can be
raised in energy. This decreases the barrier to charge separation and results in increased photocurrent generation. This finding is of
particular significance for nonfullerene OPVs, which have many potential advantages such as tunable energy levels and spectral
breadth, but are prone to poor exciton separation efficiencies. Computational, spectroscopic, and synthetic methods were combined
to develop a structureꢀproperty relationship that correlates polymer substituents with charge-transfer state energies and, ultimately,
device efficiencies.
’ INTRODUCTION
constant of ca. 39a versus fullerenes with a dielectric constant of
ca. 4.9b These properties of fullerenes generally facilitate charge
separation and the generation of free carriers.
State-of-the-art solution processable organic photovoltaic
(OPV) devices generally rely on fullerene derivatives as both
the electron acceptor and the electron transporter.1 Fullerene:
polymer blends, termed bulk heterojunctions (BHJs), hold
record efficiencies around 8%.2 Although these devices have
provided exceptional growth for the field of OPVs and have
demonstrated rapid performance improvement over the past two
decades, alternative n-type materials3 and device architectures4
could lead to “break-through” technological and basic science
advances. Currently, the best nonfullerene OPV device efficiencies
hover around 2%.5 To move beyond fullerene-based OPVs, a
greater understanding of charge generation in organic photo-
voltaics is critical.
Fullerenes provide several potential advantages over polymers
and nonfullerene small molecules in photovoltaic applications:
they possess high molecular symmetry,6 are strongly polarizable,
and present triply degenerate LUMO levels.7 Conjugated poly-
mers and planar small molecules are less symmetric, often have
well-defined charge-transport axes,8 and are generally not as
highly polarizable overall; conjugated polymers have a dielectric
Because OPVs require a donor/acceptor interface to separate
the photoexcited state (Frenkel-type excitons),10 it is important
to understand the thermodynamics of charge separation at this
interface.1d,11 The relative free energy of charge separation
(ΔGCSrel) for several donor materials combined with a fullerene
acceptor has previously been estimated by the abbreviated Weller
rel
equation ΔGCS = Es ꢀ |(HOMOdonor ꢀ LUMOacceptor)|,
where the difference between the singlet excited state energy
(Es) and the relative band offsets provided good agreement with
measured short-circuit current (Jsc).12 Although values for
rel
ΔGCS calculated from this equation correlated with the ob-
served Jsc for several devices,12,13 other factors such as active layer
absorption breadth, optical density, morphology, as well as
charge-carrier mobility and electrode choice are all known to
critically affect Jsc in addition to ΔGcs. A brief description of how
Received: April 8, 2011
Published: June 20, 2011
r
2011 American Chemical Society
12106
dx.doi.org/10.1021/ja203235z J. Am. Chem. Soc. 2011, 133, 12106–12114
|

Journal of the American Chemical Society
ARTICLE
morphology can specifically impact charge separation is pre-
sented in the Supporting Information, and it is discussed with
reference to the current investigation. Notably, the abbreviated
Weller equation does not include the lattice polarization energy
or Coulomb attraction terms, as these are not easily measured.14
Toward expanding our understanding of charge generation in
OPVs, we must explore factors beyond the thermodynamics of
charge separation as estimated from bulk electronic properties.
For instance, charge generation depends not only on the donor
and acceptor material state energies, but also on the specificmolecular
environment at the donor/acceptor (D/A) interface and on the
kinetics of exciton separation/recombination.15,11 Akin to a chemical
reaction, exciton separation to yield free charges can proceed via more
than one mechanism. In some cases, no “reaction intermediates” are
observed, whereas in other cases there is a spectroscopically obser-
vable “geminate pair” or charge transfer (CT) state. Probing the
parameters that control the mechanism of charge generation, parti-
cularly for nonfullerene devices, is of great importance to the field of
OPVs. Studies show that this electronꢀhole (eꢀh) pair is sensitive to
applied electric field and, intriguingly, to hydrostatic pressure: an
externally applied field during device operation is known to increase
the current extracted from the device; when the bias is applied
opposite (“reverse”) to the voltage generated under illumination, free
carriers are quickly removed from the active layer, and the dipolar
geminate pair is driven to separate.16 External pressure on the system
is believed to have the opposite effect on the geminate pair, decreasing
the intermolecular distance at the D/A interface, and leading to
increased radiative recombination of the CT state with a lower energy,
implying a more stable, deeply trapped intermediate.17
Our work toward understanding charge generation started
from a structural point of view, and we drew inspiration from
studies reported by Granstrom et al. in 1998.18 In that publica-
tion, poly[3-(4-n-octyl)-phenylthiophene] (POPT) was shown
to produce the most photocurrent in any OPV device at the time,
a notable achievement with the common electron acceptor material
poly[2-methoxy-5-(20-ethylhexyloxy)-1,4-(1-cyanovinylene)phen-
ylene] (CNPPV). Motivated by that research, we reported
studies in which (POPT) outperformed poly(3-hexylthiophene)
(P3HT) in both bilayer devices with CNPPV5b and BHJ
devices with the newer acceptor material 4,7-bis(2-(1-(2-
ethylhexyl)-4,5-dicyanoimidazol-2-yl)vinyl)benzo[c][1,2,5]-
thiadiazole (EV-BT).5c Our first report focused on the incon-
sistency between expected and realized performance values for
OPV devices with the acceptor CNPPV, specifically the almost
double short-circuit current density (Jsc) for POPT devices
despite reduced optical density as compared to P3HT; however,
very little was understood at that time about why better
performance was achieved with POPT instead of P3HT. In our
second report, we utilized EV-BT to make progress toward
elucidating the physical properties that governed the OPV
performance parameters of these nonfullerene devices. For
example, reverse bias analysis suggested a tighter binding of the
geminate pair at the P3HT:EV-BT interface; that is, a lower-
energy CT state provided a deeper energetic well (a trapped
intermediate) for partially separated charges. We suspected that
the octylphenyl content of POPT played a critical role at the D/A
interface, potentially facilitating geminate pair separation.
Figure 1. (a) Structures of P3HT and POPT, (b) the structures of four
different acceptors (two polymers and two small molecules) that are
tested in a head-to-head comparison between P3HT and POPT. (c) JꢀV
curves for the devices corresponding to the acceptor components in (a).
the CT states were necessary to draw a fitting conclusion. Herein, we
utilize a combination of computational and spectroscopic methods, as
well as tailored synthesis and extensive device engineering, to under-
stand how modifying thiophene substitution from alkyl to octylphe-
nyl on two otherwise identical backbones, polythiophene and
polyquarterthiophene, leads to a greater understanding of the effects
that side group interactions at D/A interfaces have on charge
generation. Structural control of the D/A interface may prove to
be a powerful tool for tuning charge separation dynamics, and we
To shed light on how using an alkylphenyl side group enhances Jsc
as compared to a simple alkyl side group, we investigated analogous
material combinations with different substitution and acceptor
materials. Beyond correlating structure to performance on the basis
of multiple device comparisons, more direct methods to investigate
12107
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
Table 1. PV Output Characteristics of POPT versus P3HT
Devices, and Maximum Efficiencies for Optimized Device
Systemsa
of individually optimized devices in all four comparisons, POPT
consistently outperforms P3HT (Table 1 and Figure 1b). While the
Voc and FF of the POPT and P3HT devices are comparable in most
cases, the Jsc values of POPT devices are at least twice those of P3HT
devices, leading to the higher overall efficiencies of POPT devices. It
should be noted that the P3HT/N2200 device results are consistent
with two recent reports that demonstrated N2200 in a BHJ device
with P3HT yields ∼0.2% efficiency.21 Additionally, the kink observed
for the P3HT/N2200 device may be due to similar effects discussed
in prior work.16b The effect of morphology on Jsc, generally and in the
cases of POPT/CNPPV and P3HT/CNPPV devices, is discussed in
the Supporting Information.
To generalize the effect of interfacial steric interactions on charge
generation, we expanded the scope of this study beyond POPT and
P3HT to another polymer backbone, polyquarterthiophene. We
synthesized poly(3,3-di(4-n-octyl)phenylquaterthiophene) PQT-
OP and compared it to poly(3,3-didodecylquaterthiophene)
PQT-DD (Table 2 and Figure 2). Independently optimized devices
with CNPPV, EV-BT, and PDI were consistently found to perform
nearly twice as well with PQT-OP than PQT-DD, due largely to an
increase in Jsc. The Voc values for PQT-based devices with the same
device active layer
Jsc [mA/cm2]
Voc [V]
FF
PCE [%]
POPT/CNPPV
P3HT/CNPPV
POPT:EV-BT
P3HT:EV-BT
POPT/N2200
P3HT/N2200
POPT:PDI
ꢀ5.44
ꢀ2.63
ꢀ5.70
ꢀ2.81
ꢀ2.50
ꢀ0.80
ꢀ5.70
ꢀ1.70
1.06
1.08
0.62
0.77
0.52
0.46
0.24
0.57
0.35
0.33
0.40
0.51
0.47
0.46
0.37
0.41
2.00
0.93
1.41
1.11
0.61
0.17
0.51
0.39
P3HT:PDI
a Symbol “/” indicates a bilayer device, while symbol “:” indicates a BHJ
device. Devices were optimized first on the basis of thickness (solvent
choice and solution concentration) and then on the basis of annealing
conditions (various temperatures and times).
provide a seminal example of how steric effects can improve charge
separation in organic photovoltaics.
’ RESULTS
Table 2. PV Output Characteristics of PQT-OP versus PQT-
DD Devicesa
Both P3HT and POPT were synthesized via the GRIM
polymerization method.5b The structure of POPT consists of
phenyl groups covalently bound to the polythiophene backbone
as part of the solubilizing substituent. This functionality increases
the ionization potential (deepens the HOMO level) to ꢀ5.5 eV
from ꢀ5.2 eV as compared to P3HT. Additionally, the optical
properties are shifted toward a broader spectral response while
maintaining similar charge-transport properties.5b The energetic
changes result in an excited state that is lower in energy (and the
electron affinity is also more exothermic) in POPT as compared
to P3HT; thus, POPT is thermodynamically less likely to
undergo exciton separation with a given acceptor, as compared
to P3HT (when a normal Marcus regime can be invoked).
Contrary to thermodynamic expectations, however, POPT
yielded more efficient charge separation as a donor material in
PV cells with polymeric and small molecule acceptors, such as
CNPPV and EV-BT, respectively.5b,c In particular, the consider-
ably and recurrently higher Jsc prompted an in-depth investiga-
tion combining device fabrication, theoretical modeling, and
advanced spectroscopy to gain insight into these systems.
The following results exploit observed differences in perfor-
mance caused by the presence of phenyl substituents to better
understand the charge separation process. By expanding our
data set beyond the two systems already reported,5b,5c we aim to
probe the universality of this design strategy for improving
charge generation in nonfullerene OPVs.
device active layer
Jsc [mA/cm2]
Voc [V]
FF
PCE [%]
PQT-OP/CNPPV
PQT-DD/CNPPV
PQT-OP/EV-BT
PQT-DD/EV-BT
PQT-OP:PDI
ꢀ2.43
ꢀ1.51
ꢀ2.68
ꢀ1.48
ꢀ3.33
ꢀ2.18
1.18
1.20
0.95
0.98
0.63
0.66
0.39
0.38
0.48
0.43
0.42
0.34
1.12
0.69
1.22
0.62
0.88
0.49
PQT-DD:PDI
a Reported are maximum efficiencies for individually optimized device
systems. A “/” indicates a bilayer device, while a “:” indicates a BHJ device.
Four acceptors were utilized with POPT and P3HT in head-
to-head comparisons: CNPPV, EV-BT, N-(1-hexylheptyl)-N0-(1-
ethylpropyl)perylene-3,4,9,10-tetracarboxylic diimide (PDI),19 and
poly{[N,N0-bis(2-octyldodecyl)-naphthalene-1,4,5,8-bis(dicar-
boximide)-2,6-diyl]-alt-5,50-(2,20-bithiophene) (Polyera ActiveInk
N2200),20 Figure 1a (see the Supporting Information for device
fabrication details). In most cases, both bilayer and bulk heterojunc-
tion devices were compared, provided that an orthogonal solvent
system was found to allow the fabrication of bilayers; here, we report
the device architecture that demonstrated the higher efficiencies for
each acceptor material. Polymerꢀpolymer solar cells performed
better in the bilayer device architecture, whereas polymerꢀsmall
molecule solar cells were better in the BHJ architecture. On the basis
Figure 2. PQT polymer structures and a comparison of individually
optimized devices. Devices were optimized first on the basis of thickness
(solvent choice and solution concentration) and then on annealing
conditions (various temperatures and times).
12108
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
acceptor material were greater than for the polythiophene-based;
however, there were no significant Voc differences between the
phenyl and alkyl PQT derivatives. This lends credence to the
hypothesis that interfacial interactions could play a role that rivals
the importance of the materials state energies.22,23 The CT state
energy, whatever its physical structure, has already been strongly
correlated with Voc.22 It is worth noting that PQT-OP provides PDI-
based devices with the highest performance to date. These data
supported our hypothesis that the effect of this substituent could be
generalized to other systems, as this is the same trend that was
observed for POPT as compared to P3HT.
Because the highest performing devices utilized POPT and
P3HT in combination with CNPPV (Table 1, Figure 1) as the
component materials, these systems were characterized in
more detail to understand how their structural properties influence
interfacial interactions and, ultimately, charge generation. The
component materials were first analyzed using a computational
description of their molecular geometries. Modeling at the
Density Functional Theory (DFT) B3LYP/6-31G(d,p) level of
theory provided optimized geometries of the neutral ground
states for (isolated) hexamers of the relevant species (Figure 3).
Two POPT conformations were explored: the first structure
allows the phenyl rings to participate in conjugation with the
thiophene backbone (Figure 3b, POPT-unconstrained), and the
second structure forces the phenyl rings to twist perpendicular to
the backbone (Figure 3c, POPT-perp). POPT-perp minimizes
conjugation between the pendant phenyl ring of the side group
and the thiophene ring of the polymer backbone but maximizes
conjugation along the backbone (see Supporting Information,
Figure S1). The neutral ground-state geometries were also
calculated for P3HT (Figure 3a) and CNPPV (Figure 3d), where
the alkyl chains were modeled as methyl groups. The calculations
show that the backbone of POPT is strictly planar only when the
phenyl rings are forced out of plane with respect to the backbone,
minimizing steric or electronic interactions between the thio-
phene and phenyl groups.
Vertical transition energies of the polymers can be qualita-
tively described from those of the oligomers by a Kuhn-type
dependence on 1/N where N is the number of double bonds
along the shortest path connecting the terminal carbon atoms of
the molecular backbone.25 The electronic structures for oligo-
mers of increasing length were calculated, and a Kuhn fit of the
data was used to extrapolate the S0 f S1 transition energies of the
extended polymers. The plots for the two POPT structures,
P3HT, and CNPPV are presented in Figure 4. The best agree-
ment between theory and experiment, that is, where the optical
bandgap (Egopt) for POPT equals 1.8 eV, occurs when the
polymer backbone is planar, suggesting that the phenyl groups
of polymer side chains prefer to orient perpendicular to the
backbone in thin films. The results for P3HT and CNPPV are
also in good agreement with experiment.
These experimental results are in contradiction with predic-
tions based on a simple comparison of the donor polymer state
energies. The larger ionization potential (lower HOMO level) of
POPT as compared to P3HT (ꢀ5.5 vs ꢀ5.2 eV) in combination
with a smaller bandgap should thermodynamically result in a
lower Jsc based on the abbreviated Weller equation. However,
octylphenyl devices produce significantly increased Jsc values as
compared to devices utilizing the alkyl analogues. Morphological
and light absorption parameters were ruled out in previous
studies as the dominant factor in this kind of comparison,
through a careful examination of device output parameters
during optimization and analysis (such as reverse-bias analysis),
as well as characterization of the films by AFM.5b,c Examination
of the PQT polymers provided similar results. PQT-OP has a
slightly larger ionization potential (IP) and a similar optical gap
as compared to PQT-DD; PQT-OP and PQT-DD have IPs of
ꢀ5.4 versus ꢀ5.3 eV, respectively, and optical gaps of approxi-
mately 1.9ꢀ2.0 eV (with absorption onsets of 640 and 620 nm).
Again, devices using the octylphenyl-containing donor polymers
consistently produce a substantially greater Jsc. These data clearly
confirm that the material state energies and optical properties are
not the only factors affecting the charge generation efficiencies in
these systems. More importantly, we hypothesize that the
molecular interactions at the D/A interface are a determining
factor in these devices. Modeling of the D/A interface has
recently predicted that the molecular configurations23 and
environment at this interface are critical in the charge-generation
process, and here we aim to correlate theory with a benchmark
physical test system.15,24
Figure 3. B3LYP/6-31G(d,p)-optimized neutral ground-state structures of the hexamers of (a) P3HT, (b) POPT-unconstrained, (c) POPT-perp, and
(d) CNPPV shown from the top-view (top) and side-view (bottom).
12109
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
packing distances arise from the two major conformations for
the phenyl rings relative to the POPT backbone: the πꢀπ
packing distance of 3.8 Å correlates to the phenyl ring oriented
parallel with the backbone, while the πꢀπ stacking distance of
5.1 Å correlates to a POPT-perp orientation where the phenyl
ring is twisted perpendicular to the backbone and causes an
increase in separation between adjacent polymers. It should also
be noted that the peak broadening observed in the GIXS pattern
may be an indication that the phenyl ring can adopt varying
degrees of rotation between the parallel and perpendicular
conformations. The packing parameters of POPT and several
phenyl-substituted polythiophenes have been studied in-depth
elsewhere.26 GIXS data of PQT-OP also evidence two dominant
conformations for the phenyl ring, resulting in πꢀπ spacings of
5.1 and 3.9 Å (Supporting Information, Figure S2). The relative
scattering intensity of the two πꢀπ spacings in PQT-OP is
reversed from that of POPT. This reversal in PQT-OP can be
attributed to the lack of substituents on two of the four
thiophenes in the polymer repeat unit, thereby favoring the
tighter πꢀπ spacing at 3.9 Å. Detailed GIXS data for P3HT27
and PQT-DD28 have been analyzed previously, and backbone
spacings of 3.8 and 4.2 Å were reported, respectively. CNPPV
derivatives are known to be relatively amorphous; however, weak
diffraction signals between 4 and 5 Å have been observed.29 X-ray
scattering is limited to the investigation of regular periodicity
present in a bulk material, and it is not appropriate for the study
of polymer blend or bilayer interfaces. However, as the D/A
interface in these material systems was our primary focus, we
turned to computational analysis to develop a model interface for
the charge separation event.
Model dimer configurations were constructed from best-fit
planes of polythiophene/CNPPV separated at distances (R)
between 4 and 5 Å in 0.2 Å increments (Figure 5d). To construct
CT states from these dimers, charges were constrained to each
molecule using the constrained density functional theory (C-
DFT) method implemented in NWChem Version 4.6.30
A
conductor polarizable continuum model (CPCM) with ε = 4
was used to approximate polarization effects expected in organic
solid-state systems. Given the limitations of the theoretical
approach, we are mainly interested in the relative CT-state
energies, which are plotted in Figure 5c. The model dimer
configuration of POPT-perp is predicted to have the highest
CT state energy followed by the P3HT and then the POPT-
unconstrained configurations. PQT-OP model calculations re-
quire many more nuclei at the interface, which is beyond the
scope of the present work.
To verify our calculations of the CT state energies in these
D/A systems, we used spectroscopic techniques to experimentally
observe their CT states. Sensitive photocurrent measurements,
via Fourier transform photocurrent spectroscopy (FTPS),22,24b
can extract the weak sub-bandgap external quantum efficiency,
and photothermal deflection spectroscopy (PDS) can detect sub-
bandgap absorption. These tools have previously been used to
investigate charge-transfer states.31 A recent and very significant
FTPS study suggests that the CT state, sometimes called a CT
exciton, is very efficiently split into free charge carriers at room
temperaturein P3HT:PCBM and MDMO-PPV:PCBM devices.32
Spectral evidence and device studies of various D/A systems
suggest that these CT states determine the Voc of the PV cell
and act as an intermediate in the generation and recombination of
free charge carriers.22,24b Consequently, spectroscopic techniques
rooted in sub-bandgap absorption are considered a good indicator
Figure 4. Vertical S0 f S1 transition energies of (a) P3HT and (b)
POPT-unconstrained and POPT-perp where the phenyl group is con-
strained to be perpendicular to the polymer backbone (c) CNPPV. N is
the number of double bonds along the backbone.
The presence of two dominant conformations of POPT is
supported empirically by two-dimensional grazing incidence
X-ray scattering (2D GIXS) measurements. Figure 5a shows
the 2D GIXS pattern and the in-plane line scan of the POPT
sample, while Figure 5b illustrates the schematic of solid-state
packing for POPT. The presence of two peaks at 3.8 and 5.1 Å
suggests that there are two different πꢀπ packing distances in
the POPT thin film. Importantly, these two different πꢀπ
12110
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
Figure 5. (a) An X-ray line scan taken parallel to the substrate surface showing peaks at d spacing equal to 28.6, 5.1, and 3.8 Å corresponding to the “a”
distance and two different “c” distances, respectively, taken from the 2D GIXS pattern of POPT on Si substrate (inset). (b) Schematic of the polymer
packing relative to the substrate, with corresponding labels to the peaks indicated in (a). (c) CT-state energies for the D/A systems illustrated in (d),
estimated at the C-DFT B3LYP/6-31G(d,p) level. (d) Physical representation of dimers of POPT and P3HT with a single repeat unit of CNPPV, both
superimposed and side-by-side.
of the presence of such CT states and of the maximum Voc that can
be expected with a given D/A combination.
to the intrinsic limitations of the methodologies, the difference in
energy between theory and experiment may be due in part to
the fact that the physical size of the CT state (e.g., the extent that
the CT exciton is delocalized) could be larger than what was
considered in the calculations.) To verify that the energies of
these CT states remain unchanged with film morphology
and film thickness, POPT:CNPPV (1:1) films were compared
as both drop cast and spun cast from 1,2-dichlorobenzene
(Supporting Information, Figure S3). While the CT state peak
positions do not change, the relative intensity of the sub-bandgap
absorption to the UVꢀvis absorption is enhanced in the spun
cast film, likely a result of finer scale phase segregation that leads
to greater D/A interfacial surface area and increased relative sub-
bandgap absorption.
PDS measurements were also performed to probe the CT
state energies of the PQT-based polymers blended with CNPPV.
The PQT-DD:CNPPV (1:1 wt/wt) shows little nonadditive sub-
bandgap absorption, while PQT-OP presents two sub-bandgap
absorption peaks at 1.25 and 1.56 eV. The higher energy CT state
peak is significantly less intense for PQT-OP as compared to
POPT (Supporting Information, Figure S4); the reason for this is
discussed below. It should be noted that our first attempts to
obtain a clean PDS signal from PQT polymers were difficult until
we discovered that residual palladium from the cross-coupling
polymerization led to an erroneous mirage effect and dramati-
cally increased background signal (Figure S5).
Here, PDS spectra were obtained by detecting the mirage effect in
a transparent, inert medium (Fluorinert) with a probe HeNe laser
beam. Nonradiative heating associated with absorption of a mono-
chromatic pump beam causes the mirage effect to occur. PDS was
used in this investigation to support our hypothesis that molecular
orientation of the phenyl groups affects the CT state energy. PDS
measurements were performed on drop cast and spun cast films of
POPT and P3HT blended with CNPPV. We underline here that
bilayer films do not provide enough interfacial surface area to produce
good signal-to-noise ratios; in addition, the molecular level interface is
not expected to change upon going from the bilayer to BHJ
morphology, vide infra, and see extended discussion in the Supporting
Information. Figure 6 shows the PDS spectra of the homopolymers
and the polymer blends under investigation; spectra are scaled to
absolute values of the absorption coefficient by matching the signal
near the absorption edge to that from the UVꢀvis spectra of the
same films.
Blends of both P3HT and POPT with CNPPV produce
nonadditive absorptions that are attributed to the presence of
CT states at the D/A interfaces. For P3HT:CNPPV (1:1 wt/wt),
a CT state absorption is present at 1.26 eV. For POPT, however,
there are two sub-bandgap peaks attributed to CT states, one at
1.17 eV and one at 1.50 eV, possibly indicating two distinct
interfacial configurations. These peak maxima are extracted by
fitting an exponential for the band edge and Gaussian curves for
the CT peaks in the sub-bandgap regions. The results of the
calculations presented in Figure 5c are qualitatively in agreement
with the observed PDS absorption peaks, in that the P3HT:
CNPPV blend has a CT state energy that resides between the
two POPT:CNPPV CT state energies. (We note that, in addition
’ DISCUSSION
A thermodynamic driving force for charge generation, that is,
exciton dissociation leading to charge separation, is present at the
interface between the donor/acceptor (D/A) materials in an
12111
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
Figure 7. A diagram of possible electron flow pathways at the D/A
interface, relative to potential energy (adapted from ref 11). CT state
energies (ground-state solid black, excited-state dashed black) are shown
in relation to the D/A singlet excited state (S1), triplet state (T1), and
ground state (S0). Competing energetic pathways and rates are also
depicted: vibrational relaxation of the CT state (kVR), intersystem
crossing of the CT state to the donor triplet state (kIC), recombination
of the CT state to the ground state (kRec), and finally charge separation
(kCS). In addition to thermodynamics considerations, the kinetics of
these processes will determine the charge separation behavior for each
photovoltaic system.
Figure 6. Absorption spectra of P3HT, POPT, and blends with CN-
PPV. Thick solid lines are UVꢀvis absorption spectra of the homo-
polymers; symbols are the PDS absorption spectra of drop cast films.
The arrows indicate the sub-bandgap features attributed to CT states in
the blend systems. P3HT:CNPPV peak maximum at 1.26 eV, while
POPT:CNPPV possesses two peaks at 1.17 and 1.50 eV.
see Figure 7. If kCS > kVR, then the electron is expected to readily
escape the Coulomb potential and proceed to the CS state. If kVR
> kCS, then relaxation to the CT1 state leads to a more tightly
bound (lower energy) intermediate. The electron can still escape
from this state;32 however, other processes start to compete with
charge separation: if either the donor or the acceptor material
possesses a triplet level (T1) below the CT1 state, intersystem
crossing leads to long-lived metastable triplets. Also, the CT1
state for some systems can radiatively or vibrationally decay to
the ground state S0.16c,35 For these reasons, the kinetics of CS
must be considered when parsing the charge-generation process.
In this work, OPV devices comparing POPT to P3HT and
PQT-OP to PQT-DD were fabricated and analyzed. POPT and
PQT-OP possess phenyl groups covalently bound to the poly-
mer backbone as part of the solubilizing substituents. This
functionality decreases the thermodynamic driving force for
charge separation (vide supra), but both POPT and PQT-OP
produced remarkably higher Jsc relative to their alkyl analogs. All
relevant PV characteristics are summarized in Tables 1 and 2.
Further, X-ray scattering data evidenced that both POPT and
PQT-OP may adopt planar-with- and perpendicular-to-the-back-
bone conformations for the pendant phenyl rings. CT state
energies for model dimer configurations were calculated and are
plotted in Figure 5c: POPT-perp is predicted to have the highest
CT state energy, while the CT state energy for P3HT lies
between those of the POPT-perp and POPT-unconstrained
conformations. Finally, experimental spectroscopic evidence of
charge-transfer states at the interface with the acceptor CNPPV,
gathered via PDS for all four donor polymers, is consistent with
the relative values predicted by the model dimer calculations.
With an out-of-plane twist of the phenyl rings, the separation
distance between POPT and the acceptor molecule would likely
increase as steric repulsion from the phenyl rings hinders back-
boneꢀbackbone interaction. PDS data confirm the presence of
two distinct features in the sub-bandgap regime, which could be a
direct result of these two dominant conformations at the D/A
interface, as it has been explained in the previous section. Because
these conformationally dependent states are both involved as
intermediates in the charge-generation process, the correspond-
ing geminate pairs would overcome different energetic barriers to
split into free charges. We postulate that a twisted phenyl ring
OPV active layer. Photon absorption by either the donor or the
acceptor materials produces the opportunity for charge-carrier
generation: in the case of “donor” excitation, the system
decreases in potential energy from the singlet excited state (Es)
by transferring an electron from donor to acceptor, and in the
case of “acceptor” excitation, by transferring a hole from acceptor
to donor. For simplicity, the process is generally discussed from
the viewpoint of an excited donor material. A general diagram
depicting charge separation is presented in Figure 7.
Although a thermodynamic driving force helps to generate
free charges, the immediate physical separation of the electron
and hole does not necessarily lead directly to free charges. The
low dielectric constant of the active layer can produce a Coulomb
trap for a partially separated exciton at the D/A interface. This
state is usually referred to as a charge-transfer (CT) state. The
CT state may either recombine to the initial ground states of the
donor and acceptor materials or undergo further separation into
free charges. It is broadly debated whether an intermediate CT
state is requisite to charge separation,11,32 and it is more recently
debated whether this separation/generation can occur from
lowest-lying CT1 states.32 It was also recently reported that a
modest thermodynamic driving force of 0.1 eV leads to reasonable
quantum yields of photocurrent with fullerene-based devices,
based on commonly measured material properties, which supports
our previous finding with POPT/CNPPV.34 The contention
surrounding the lowest-lying CT states is that they can be bound
(vs the charge-separated states) by more energy than thermally
available from kBT (where kB is the Boltzmann constant); thus, it
is postulated that excess energy released during partial exciton
dissociation could create higher-lying (excited) CT states (CT*),
which are more likely to escape the Coulomb trap.33c However,
strictly discussing the thermodynamics of charge separation
ignores the important kinetic considerations of this process.
The efficiency of the charge separation (CS) process indeed
depends on kinetic factors. Given that there is greater potential
energy stored in the singlet exciton than a charge transfer exciton,
and the possibility that sub-bandgap absorption produces excited
CT* states that may relax down to CT1, two rates are of critical
importance: the rate of charge separation (kCS) and the rate of
vibrational relaxation of an excited CT state down to CT1 (kVR);
12112
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
substantially greater photocurrents, and overall power conversion
efficiencies, than the alkyl analogues.
’ CONCLUSIONS
We have utilized computational modeling, PDS spectroscopy
and tailored synthetic design to probe the importance of steric
interactions at the donor/acceptor interface in nonfullerene
OPV devices. By introducing the octylphenyl substituent onto
the investigated polymer backbones, the thermally relaxed
charge-transfer state, and potentially excited charge-transfer
states, are likely raised in energy, as evidenced by PDS. This
decreases the barrier to charge separation, and assuming we have
controlled properly for changes in morphology by thoroughly
optimizing each materials/device system, it could be the source
of increased photocurrent generation. The design principle was
shown to be general across two polythiophene backbones and
with four different acceptors, two polymers and two small
molecules. Additionally, the lower energy PDS onset for
POPT-based devices with CNPPV (1.17 eV) versus the onset
for PQT-OP with CNPPV (1.26 eV) is reflected in the Voc of
these devices. The combined data from POPT and PQT-OP
devices and their materials analyses suggest that controlling the
steric interaction at the D/A interface could be a general design
principle toward improving charge generation in nonfullerene
OPVs.
Figure 8. (a) Cartoon of how steric interactions can lead to an increase
in backbone spacing, a decrease in the Coulomb binding force, and
destabilization of the geminate pair. (b) Schematic of how the change
depicted in cartoon (a) leads to a different energy landscape with
increased charge separation probability in POPT, as the CT state is
considered an intermediate trapped in an energetic well.
conformation of POPT (POPT-perp) is beneficial for charge
generation, as an intermediate with increased potential energy is
more likely to fully separate into free charges (Figure 8). FTPS
measurements could lead to a quantitative description of the
quantum yields for these two states, and this is the focus of fu-
ture work.
This study also generated two additional significant and
supportive findings. PDI-based acceptors have garnered much
attention as alternative n-type materials to replace fullerenes;3a
here, we produce the highest efficiency devices with this acceptor
to date, despite tremendous efforts with alternative approaches
toward higher efficiency.3a,13a,19 This is just another indication
that control over the interfacial geometry at the molecular level
can lead to much improved device performance, as a comple-
mentary tool to morphology and state energy control. Addition-
ally, the photovoltaic performance with the high mobility n-type
polymer ActiveInk N2200 demonstrates that POPT outperforms
P3HT both in our laboratories and as compared to two very
recent reports.21
’ ASSOCIATED CONTENT
S
Supporting Information. Experimental, device fabrica-
b
tion, and characterization details. This material is available free of
charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
frechet@berkeley.edu
’ ACKNOWLEDGMENT
This work was supported by the Center for Advanced Mole-
cular Photovoltaics (Award No. KUS-C1-015-21), supported by
King Abdullah University of Science and Technology (KAUST),
Combining all of the data, analysis, and literature context, we
have synthesized and proposed a general design principle for
improved charge separation in nonfullerene OPVs: tuning the
D/A interfacial interaction through steric control to facilitate
photocurrent generation. Regardless of whether charge separation
happens from a relaxed CT1 state or an excited CT state,
increasing the steric bulk at the D/A interface likely decreases
the Coulomb binding strength exerted on the geminate pair. We
postulate that the phenyl ring pendant to POPT and PQT-OP
provides an almost ideal interaction distance between the charge
carrying components of the D/A interface, and this leads to two of
the best nonfullerene devices to date. The higher energy of the
intermediate CT state, with a lower activation barrier to free carrier
generation, improves photocurrent generation and provides the
key to the observed phenomenon (Figure 8). This effect was not
limited to one donor polymer or one acceptor material, but rather
it was general for two donors and four acceptors (four polymers
and two small molecules), for a total of seven material combina-
tions. All of these material combinations yielded optimized devices
with the phenyl containing polymeric substituents producing
and the U.S. Department of Energy under Contract No. DE-
AC02-05CH11231 (synthesis and some device characterization
work). T.W.H., C.H.W., and J.R. thank the National Science
Foundation for graduate research fellowships. We gratefully
acknowledge Polyera Inc. and Paul Armstrong for providing
the Active Ink N2200 and PDI, respectively, used in this study.
Paul Armstrong and Yoshi Miyamoto are thanked for assistance
with device optimization. We also thank David Kavulak and
Barry Thompson for helpful discussions.
’ REFERENCES
(1) (a) Boudreault, P. T.; Najari, A.; Leclerc, M. Chem. Mater. 2011,
23, 456–459. (b) He, Y. J.; Chen, H. Y.; Hou, J. H.; Li, Y. F. J. Am. Chem.
Soc. 2010, 132, 1377–1382. (c) Piliego, C.; Holcombe, T. W.; Douglas,
J. D.; Woo, C. H.; Beaujuge, P. M.; Frechet, J. M. J. J. Am. Chem. Soc.
2010, 132, 7595–7597. (d) Thompson, B. C.; Frechet, J. M. J. Angew.
Chem., Int. Ed. 2008, 47, 58–77.
(2) Chen, H. Y.; Hou, J. H.; Zhang, S. Q.; Liang, Y. Y.; Yang, G. W.;
Yang, Y.; Yu, L. P.; Wu, Y.; Li, G. Nat. Photonics 2009, 3, 649–653.
12113
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114

Journal of the American Chemical Society
ARTICLE
(3) (a) Anthony, J. E. Chem. Mater. 2010, 23, 583–590. (b) Brunetti,
F. G.; Gong, X.; Tong, M.; Heeger, A. J.; Wudl, F. Angew. Chem., Int. Ed.
2010, 49, 532–536. (c) Shin, R. Y. C.; Kietzke, T.; Sudhakar, S.;
Dodabalapur, A.; Chen, Z. K.; Sellinger, A. Chem. Mater. 2007,
19, 1892–1894.
2008, 7, 483–489. (b) Vandewal, K.; Tvingstedt, K.; Gadisa, A.; Inganas,
O.; Manca, J. V. Phys. Rev. B 2010, 81, 1252041–8.
(25) Gierschner, J.; Cornil, J.; Egelhaaf, H. J. Adv. Mater. 2007,
19, 173–191.
(26) Fell, H. J.; Samuelsen, E. J.; Mardalen, J.; Andersson, M. R.
Synth. Met. 1995, 69, 283–284.
(4) Lee, M. R.; Eckert, R. D.; Forberich, K.; Dennler, G.; Brabec,
C. J.; Gaudiana, R. A. Science 2009, 324, 232–235.
(27) Verploegen, E.; Mondal, R.; Bettinger, C. J.; Sok, S.; Toney,
M. F.; Bao, Z. A. Adv. Funct. Mater. 2010, 20, 3519–3529.
(28) Thompson, B. C.; Kim, B. J.; Kavulak, D. F.; Sivula, K.; Mauldin,
C.; Frechet, J. M. J. Macromolecules 2007, 40, 7425–7428.
(29) (a) Chen, S. A.; Chang, E. C. Macromolecules 1998, 31, 4899–
4907. (b) Chen, S. H.; Su, C. H.; Su, A. C.; Chen, S. A. J. Phys.
Chem. B 2004, 108, 8855–8861.
(30) (a) Wu, Q.; Van Voorhis, T. J. Chem. Theory Comput. 2006,
2, 765–774. (b) Wu, Q.; Van Voorhis, T. J. Phys. Chem. A 2006,
110, 9212–9218. (c) Wu, Q.; Van Voorhis, T. Phys. Rev. A 2005,
72, 0245021–4. (d) Valiev, M.; Bylaska, E. J.; Govind, N.; Kowalski,
K.; Straatsma, T. P.; Van Dam, H. J. J.; Wang, D.; Nieplocha, J.; Apra, E.;
Windus, T. L.; de Jong, W. Comput. Phys. Commun. 2010, 181, 1477–
1489.
(31) (a) Goris, L.;Haenen, K.;Nesladek, M.;Wagner, P.;Vanderzande,
D.; De Schepper, L.; D’Haen, J.; Lutsen, L.; Manca, J. V. J. Mater. Sci. 2005,
40, 1413–1418. (b) Goris, L.; Poruba, A.; Hod’akova, L.; Vanecek, M.;
Haenen, K.; Nesladek, M.; Wagner, P.; Vanderzande, D.; De Schepper, L.;
Manca, J. V. Appl. Phys. Lett. 2006, 88, 0521131–3.
(32) Lee, J.; Vandewal, K.; Yost, S. R.; Bahlke, M. E.; Goris, L.; Baldo,
M. A.; Manca, J. V.; Van Voorhis, T. J. Am. Chem. Soc. 2010, 132,
11878–11880.
(5) (a) McNeill, C. R.; Abrusci, A.; Zaumseil, J.; Wilson, R.;
McKiernan, M. J.; Burroughes, J. H.; Halls, J. J. M.; Greenham, N. C.;
Friend, R. H. Appl. Phys. Lett. 2007, 90, 1935061–3. (b) Holcombe,
T. W.; Woo, C. H.; Kavulak, D. F. J.; Thompson, B. C.; Frechet, J. M. J.
J. Am. Chem. Soc. 2009, 131, 14160–14161. (c) Woo, C. H.; Holcombe,
T. W.; Unruh, D. A.; Sellinger, A.; Frechet, J. M. J. Chem. Mater. 2010,
22, 1673–1679. (d) Zhou, E.; Cong, J.; Wei, Q.; Tajima, K.; Yang, C.;
Hashimoto, K. Angew. Chem., Int. Ed. 2011, 50, 2799–2803.
(6) The fullerene structure is a truncated icosohedron.
(7) Kanai, Y.; Grossman, J. C. Nano Lett. 2007, 7, 1967–1972 and
references therein.
(8) W€oll, C. Physical and Chemical Aspects of Organic Electronics:
From Fundamentals to Functioning Devices; Wiley-VCH: Weinheim,
2008.
(9) (a) Knipper, M.; Parisi, J.; Coakley, K.; Waldauf, C.; Brabec, C. J.;
Dyakonov, V. Z. Naturforsch., A 2007, 62, 490–494. (b) Mihailetchi,
V. D.; Koster, L. J. A.; Blom, P. W. M.; Melzer, C.; de Boer, B.; van
Duren, J. K. J.; Janssen, R. A. J. Adv. Funct. Mater. 2005, 15, 795–801.
(10) (a) Sariciftci, N. S.; Smilowitz, L.; Heeger, A. J.; Wudl, F. Science
1992, 258, 1474–1476. (b) Halls, J. J. M.; Walsh, C. A.; Greenham,
N. C.; Marseglia, E. A.; Friend, R. H.; Moratti, S. C.; Holmes, A. B.
Nature 1995, 376, 498–500.
(11) Bredas, J. L.; Norton, J. E.; Cornil, J.; Coropceanu, V. Acc. Chem.
Res. 2009, 42, 1691–1699.
(12) Ohkita, H.; Cook, S.; Astuti, Y.; Duffy, W.; Tierney, S.; Zhang,
W.; Heeney, M.; McCulloch, I.; Nelson, J.; Bradley, D. D. C.; Durrant,
J. R. J. Am. Chem. Soc. 2008, 130, 3030–3042.
(13) (a) Shoaee, S.; Clarke, T. M.; Huang, C.; Barlow, S.; Marder,
S. R.; Heeney, M.; McCulloch, I.; Durrant, J. R. J. Am. Chem. Soc. 2010,
132, 12919–12926. (b) Clarke, T. M.; Ballantyne, A. M.; Nelson, J.;
Bradley, D. D. C.; Durrant, J. R. Adv. Funct. Mater. 2008, 18, 4029–4035.
(14) Morteani, A. C.; Sreearunothai, P.; Herz, L. M.; Friend, R. H.;
Silva, C. Phys. Rev. Lett. 2004, 92, 2474021–4.
(33) (a) Clarke, T. M.; Durrant, J. R. Chem. Rev. 2010,
110, 6736–6767. (b) Deibel, C.; Strobel, T.; Dyakonov, V. Adv. Mater.
2010, 22, 4097–4111. (c) Zhu, X. Y.; Yang, Q.; Muntwiler, M. Acc. Chem.
Res. 2009, 42, 1779–1787.
(34) Zhou, Y.; Tvingstedt, K.; Zhang, F. L.; Du, C. X.; Ni, W. X.;
Andersson, M. R.; Inganas, O. Adv. Funct. Mater. 2009, 19, 3293–3299.
(35) Loi, M. A.; Toffanin, S.; Muccini, M.; Forster, M.; Scherf, U.;
Scharber, M. Adv. Funct. Mater. 2007, 17, 2111–2116.
(15) Beljonne, D.; Cornil, J.; Muccioli, L.; Zannoni, C.; Bredas, J. L.;
Castet, F. Chem. Mater. 2011, 23, 591–609.
(16) (a) Marsh, R. A.; Hodgkiss, J. M.; Friend, R. H. Adv. Mater.
2010, 22, 3672–3676. (b) Ooi, Z. E.; Tam, T. L.; Sellinger, A.; deMello,
J. C. Energy Environ. Sci. 2008, 1, 300–309. (c) Veldman, D.; Ipek, O.;
Meskers, S. C. J.; Sweelssen, J.; Koetse, M. M.; Veenstra, S. C.; Kroon,
J. M.; van Bavel, S. S.; Loos, J.; Janssen, R. A. J. J. Am. Chem. Soc. 2008,
130, 7721–7735.
(17) Schmidtke, J. P.; Friend, R. H.; Silva, C. Phys. Rev. Lett. 2008,
100, 1574011–4.
(18) Granstrom, M.; Petritsch, K.; Arias, A. C.; Lux, A.; Andersson,
M. R.; Friend, R. H. Nature 1998, 395, 257–260.
(19) Rajaram, S.; Armstrong, P. B.; Kim, B. J.; Frechet, J. M. J. Chem.
Mater. 2009, 21, 1775–1777.
(20) Yan, H.; Chen, Z. H.; Zheng, Y.; Newman, C.; Quinn, J. R.;
Dotz, F.; Kastler, M.; Facchetti, A. Nature 2009, 457, 679–686.
(21) (a) Fabiano, S.; Chen, Z.; Vahedi, S.; Facchetti, A.; Pignataro, B.;
Loi, M. A. J. Mater. Chem. 2011, ASAP. (b) Moore, J. R.; Albert-Seifried,
S.; Rao, A.; Massip, S.; Watts, B.; Morgan, D. J.; Friend, R. H.; McNeill,
C. R.; Sirringhaus, H. Adv. Energy Mater. 2011, 1, 230–240.
(22) Vandewal, K.; Tvingstedt, K.; Gadisa, A.; Inganas, O.; Manca,
J. V. Nat. Mater. 2009, 8, 904–909.
(23) (a) Yi, Y. P.; Coropceanu, V.; Bredas, J. L. J. Am. Chem. Soc.
2009, 131, 15777–15783. (b) Yi, Y.; Coropceanu, V.; Bredas, J. L.
J. Mater. Chem. 2011, 21, 1479–1486.
(24) (a) Huang, Y. S.; Westenhoff, S.; Avilov, I.; Sreearunothai, P.;
Hodgkiss, J. M.; Deleener, C.; Friend, R. H.; Beljonne, D. Nat. Mater.
12114
dx.doi.org/10.1021/ja203235z |J. Am. Chem. Soc. 2011, 133, 12106–12114