A novel 2,7-linked carbazole based "double cable" polymer with pendant perylene diimide functional groups: Preparation, spectroscopy and photovoltaic properties

Article abstract of DOI:10.1039/c0jm02673e

The preparation and chemical analysis of a 'double-cable' conjugated polymer comprising a backbone of alternating 2,7-linked carbazole repeat units with covalently attached perylene diimide (PDI) substituents and dithienyl repeat units is presented. The b

Full text of DOI:10.1039/c0jm02673e

View Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/materials | Journal of Materials Chemistry
A novel 2,7-linked carbazole based ‘‘double cable’’ polymer with pendant
perylene diimide functional groups: preparation, spectroscopy and
photovoltaic properties†
a
David K. Mohamad, Achim Fischereder, Hunan Yi, Ashley J. Cadby, David G. Lidzey and Ahmed Iraqi
b
b
a
a
b
*
*
Received 14th August 2010, Accepted 19th October 2010
DOI: 10.1039/c0jm02673e
The preparation and chemical analysis of a ‘double-cable’ conjugated polymer comprising a backbone
of alternating 2,7-linked carbazole repeat units with covalently attached perylene diimide (PDI)
substituents and dithienyl repeat units is presented. The backbone of the new ‘‘double-cable’’ polymer
P1 acts as an electron donor while the pendant PDI substituents act as electron acceptors. In order to
investigate the influence of the PDI moieties on the polymer backbone as well as to elucidate their
photophysical interaction, three reference-compounds were also prepared and analysed: a polymer
with the same backbone as that of P1 but without PDI substituents (P2) and two PDI derivatives with
branched alkyl side chains (PDI 2 with 12-tricosanyl substituents and PDI 1 with 3-pentyl substituents).
We find that polymer P1 shows pronounced electron transfer from the polymer backbone to the PDI
chromophore units covalently bound to it, resulting in highly efficient quenching of excitons and strong
suppression of fluorescence in solutions and in thin films. The existence of long-lived polaronic states
resulting from exciton dissociation on P1 is confirmed using steady state photo-induced absorption
spectroscopy. Despite the improved exciton quenching yield shown by P1 over a blend of P2/PDI,
photovoltaic devices fabricated from P1 have a low external quantum efficiency (EQE) of around
0.43% at 410 nm; a value that is smaller than that from a conventional BHJ device of P2/PDI which is
found to have a peak external quantum efficiency of 3.7% at 463 nm. We tentatively ascribe the reduced
EQE of the double-cable polymer to geminate recombination of charge-carriers as a result of poor
charge transport and a complete lack of donor acceptor phase separation.
Double-cable polymers are generally composed of covalently
linked acceptors to polymer donors, with two ‘cables’ existing on
a single molecule permitting charge carriers of both signs to be
extracted using a single material system. Most studies have so-far
targeted polymer donors that are covalently linked to fullerene
acceptors.3 The covalent attachment of fullerenes to conjugated
polymer donors remains a challenging problem; while many such
systems have been described in the literature, there remain issues
associated with the processability of these materials as well as
their percentage loading with fullerenes. Alternative acceptor
systems to fullerenes such as anthraquinone have been explored
in this area and it has been shown that covalently bound
anthraquinone to conjugated polymers displayed photoinduced
electron transfer from the polymer backbone to the covalently
attached anthraquinone moieties.4 Perylene diimides show
strong n-type character and attractive properties such as long
exciton diffusion ranges of up to 2.5 mm in vacuum deposited
films5 and high electron carrier mobilities.6 Consequently, they
have been used as electron acceptors in bulk heterojunction
devices with conjugated polymers7a–e showing external quantum
efficiencies (EQEs) as high as 16% and also with discotic liquid
crystals7f where the EQE was 34% at 490 nm (the perylene
absorption peak). However, in view of the difficulty in directing
optimal phase separation, these systems have relatively limited
power conversion efficiencies. Covalent attachment of perylene
diimides to conjugated polymers such as poly(p-phenylene
vinylene)s8 and alternating poly(fluorene-alt-phenylene)9 have
1. Introduction
There is currently a concerted effort aimed at developing new
materials for use in photovoltaic (PV) devices. Organic PVs have
been targeted as a potentially promising technology due to the
inherently low costs associated with manufacturing and
production, and research is actively evolving in this area.1 Bulk
heterojunction PV cells comprising conjugated polymer donors
and fullerene acceptors have attracted most interest.2 The control
of the morphology of blends in these cells is an important
component of generating systems in which there is efficient
separation of excitons at the donor–acceptor interface and
charge collection at electrodes. Indeed, donor and acceptor
components should be intimately mixed so that a donor–
acceptor interface is within a distance less than the exciton
diffusion length whilst maintaining a continuous percolation
path to extract charge carriers from the device.
To this end the concept of double-cable polymers was
developed.
aDepartment of Physics and Astronomy, University of Sheffield, Sheffield,
S3 7RH, UK. E-mail: d.g.lidzey@sheffield.ac.uk; Fax: +44 (0)114 222
3555
bDepartment of Chemistry, University of Sheffield, Sheffield, S3 7HF, UK.
E-mail: a.iraqi@sheffield.ac.uk; Fax: +44 (0)114 222 9303
† Electronic Supplementary Information (ESI) available: Material
preparation and analysis, PA dynamics. See DOI: 10.1039/c0jm02673e/
This journal is ª The Royal Society of Chemistry 2011
J. Mater. Chem., 2011, 21, 851–862 | 851

View Online
also been described but to our knowledge, no photovoltaic
devices have been reported on these materials.
2.2. Fabrication of OPV devices
Photovoltaic devices were fabricated using the basic architecture
ITO(50 nm)/PEDOT:PSS(30 nm)/Active Layer(100–150 nm)/
Al(100 nm) where the active layer was a thin-film of the double-
cable polymer P1 or a blend of the model polymer P2 and
acceptor moieties in a variety of blend ratios. For benchmarking
purposes, we also explored blends of polymer P2 with the
fullerene acceptor PCBM to assess the relative suitability of the
polymer as an electron donor and charge transporting material in
a comparative BHJ device. To fabricate OPV devices, a thin-film
of poly(3,4-ethylenedioxythiophene) poly(styrenesulfonate) was
first coated on the ITO patterned substrates. The active layer
solutions and thin films were prepared in an inert N2 atmosphere
maintained in a glove box. Solutions were prepared at 20 g/l
concentration in a variety of solvents (vide infra) then heated
In this work, we report the preparation of alternating poly-
(carbazole-alt-dithienyl) polymers with and without covalently
attached perylene diimide (PDI) units. Owing to the relatively
deep LUMO level of the PDI moieties, they act as strong elec-
tron-acceptors coupled to a carbazole-based backbone-polymer
(donor). We present a range of studies on the photophysical
properties of these materials as pristine materials as well as in
blends with free perylene diimide chromophores. Through
functional control of active layer morphology, we show that the
quenching of excitons on these materials is particularly efficient –
a conclusion supported by steady-state photoinduced absorption
measurements. But, devices based on this particular double-cable
polymer have a lower efficiency than those from blends of
a polymer with a similar backbone with free PDI chromophores.
We tentatively propose that the efficiency of the double-cable
polymer devices we fabricate is limited both by geminate charge
recombination, and poor charge extraction.
ꢁ
overnight at 60 C. To ensure that each material was dissolved,
they were passed through a 0.45 mm PTFE syringe mounted
filter. All films were spin cast in a glove box prior to the depo-
sition of the Al cathode. Devices were encapsulated in nitrogen
using a UV curable epoxy that was used to stick a glass slide to
the device surface. Devices were then thermally annealed by
2. Experimental methods
ꢁ
heating to 120 C for 10 min.
Experimental details of the monomer and polymer synthesis are
provided in the supplementary information. Characterization
data are also presented.
AFM images were obtained with a Veeco Dimension 3100
atomic force microscope in tapping mode using TAP300AL
cantilevers supplied by Budget Sensors Ltd. Thickness
measurements were taken with a Veeco Dektak 150 profilometer.
2.1. Spectroscopy of solutions and thin-films
2.3. Material synthesis and characterization
Samples of polymer thin films for UV-visible absorption, pho-
toluminescence and photoinduced absorption (PA) spectroscopy
measurements were prepared by either spin-coating or dip
coating quartz substrates from polymer solutions in chloroform
(HPLC grade) or dichloromethane (HPLC grade). Typical thin
films had a thickness of ꢀ 100 to 150 nm unless otherwise
specified. Photoluminescence and Absorption spectra were
obtained using Hitachi F-4500 Fluorescence Spectrophotometer
or a JY Fluoromax 4 Fluorimeter. Solution photoluminescence
(PL) measurements were carried out using a quartz fluorescence
cuvette (light path length ¼ 10 mm). Absorption and PL
measurements were carried out at ambient temperature in air.
PA spectroscopy measurements were performed by measuring
the transmission difference caused by photoexcited states of
a 100 W tungsten lamp focussed to a ꢀ 5 mm diameter spot onto
the sample, mounted onto the cold finger of an Oxford-Instru-
ments DN type cryostat cooled to 77 K. Films were optically
pumped using the 488nm line from an Ar+ laser that was chopped
using either an opto-acoustic modulator for modulation resolved
experiments or a mechanical chopper for the wavelength and
intensity dependent measurements. The transmitted light was
imaged into a JY SPEX monochromator using a silicon photo-
diode to detect light from 2.2 eV to 1 eV and thereafter with an
InSb detector to 0.5 eV. Photodiode signals were amplified by an
EG&G current voltage amplifier which separates the DC and AC
signal components and provides some preamplification. A
Stanford Research Systems SR830 Lockin-amplifier was used to
collect the phase information from the AC component. Zero
phase is defined as the laser modulation. Both the in-phase and
out-phase signals were recorded.
Synthesis of 2 was performed via a condensation reaction
(Scheme 1) of perylene-3,4,9,10-tetracarboxylic dianhydride with
1-undecyldodecylamine in imidazole. Partial hydrolysis of bis-
alkylated 2 with potassium hydroxide in tert-butanol followed by
treatment with hydrochloric acid provided 3 in moderate yields.
Reaction of 3 with 6-aminohexanol in the presence of zinc
acetate in N,N-dimethylacetamide afforded 4. Compound 5 was
obtained via tosylation of 4 using p-toluenesulfonyl chloride in
the presence of triethylamine and catalytic amounts of trime-
thylamine hydrochloride. The compound was purified by
recrystallisation of the crude product from diethyl ether and
methanol. Carbazole compound 6 was obtained in moderate
yields on reacting 5 with 2,7-dibromo-3,6-dimethyl-9H-carba-
zole and potassium hydroxide. The NMR spectra of 6 are
concentration dependant and while intermolecular interactions
affecting perylene diimide units are well documented in the
literature, the chemical shifts of signals from the carbazole units
on 6 do also change significantly with concentration. At
a concentration of 9 mg/ml in CDCl3, the carbazole moiety on 6
1
can be easily identified from the H-NMR spectrum, (Fig. S1,
ESI†) with two singlet peaks with an integral intensity of two
protons at 7.788 ppm and 7.516 ppm (aromatic protons) and one
singlet peak with an integral intensity of six protons at 2.583 ppm
(protons of the methyl groups attached to the carbazole) while
the aromatic protons of the perylene ring are observed as one
multiplet at 8.602 ppm. At a higher concentration of 6 in CDCl3,
(37 mg/ml), the signals of the carbazole moieties on the
compound are shifted upfield to 7.630 and 7.391 ppm for the
aromatic protons and at 2.450 ppm for the methyl substituents
852 | J. Mater. Chem., 2011, 21, 851–862
This journal is ª The Royal Society of Chemistry 2011

View Online
Scheme 2 (i) Pd2(dba)3, P(o-totyl)3, toluene, 95 ꢁC.
temperature to some extent even though this portion of the
polymer was not extracted by warm chloroform during the
Soxhlet extraction of the first fraction of the polymer. We believe
that this is due to aggregation of polymer chains during the
Soxhlet extraction process.
Polymer P2 was prepared so as to provide a basis of
comparison to that of the backbone of polymer P1. Stille cross
coupling of 5,50-bis(tri-n-butylstannyl)-2,20-bithiophene and 2,7-
dibromo-3,6-dimethyl-9-(2-hexyldecyl)-9H-carbazole provided
P2 (Scheme 2) which has a weight average molecular weight Mw
of 59,800 Da and a polydispersity PD of 2.2.
Scheme 1 (i) NH2(CH(C2H5)2), Zn(OAc)2, pyridine, 80 ꢁC for 4 h then
reflux overnight; (ii) NH2(CH(C11H23)2), imidazole, 140 ꢁC; (iii) (a)
KOH, t-BuOH, reflux, (b) HCl; (iv) NH2(C6H12)OH, Zn(OAc)2,
DMAC, 110 ꢁC then 160 ꢁC; (v) p-CH3C6H4SO2Cl, NEt3, NH(CH3)3Cl,
0–5 ꢁC; (vi) 2,7-dibromo-3,6-dimethyl-9H-carbazole, KOH, DMSO/
THF, RT.
1H-NMR spectra of P1 in deuterated chloroform provides
broad signals especially for the protons in the aromatic region.
1H-NMR studies of P1 in deuterated 1,1,2,2-tetrachloroethane at
room temperature provided better spectra but still broad signals
in the aromatic region, however, upon heating to 100 ꢁC, the
signals were better defined with a multiplet between 8.54–
8.08 ppm corresponding to four protons and four broad singlets
corresponding to two protons each at respectively 7.790, 7.399,
7.192 and 7.039 (Fig. S2, ESI†). We believe that this behaviour is
due to aggregation of pendant PDI units on P1. Similar effects
have been previously observed for donor–acceptor molecules
such as those incorporating zinc tetraphenyl-porphyrin with PDI
substituents.10 Broadening of the peaks in the1H-NMR spectra of
P1 is persistent even at high dilution which could point to
intramolecular aggregation of the pending PDI units on polymer
chains and that aggregates are reduced at higher temperature.
However, intermolecular aggregates associating different poly-
mer chains could not be ruled out given the changes observed in
the solubility of P1 during the separation of its second fraction.
Thermal gravimetric analysis of P1 and P2 indicated a good
thermal stability for both polymers with an onset of the first
degradation ꢁat 399 ꢁC for P1 and at 428 ꢁC for P2. The residual
mass at 800 C for polymer P1 (59.3%) corresponds well to the
sum of the bare polymer backbone and the aromatic core
structure of its PDI-units. Polymer P2 shows a mass loss after the
first thermal degradation of 29.6%, which is consistent with the
loss of the alkyl chains. Differential scanning calorimetry studies
on P1 did not reveal any clear glass transition while studies on
polymer P2 indicate a glass transition around 95 ꢁC which is
on the carbazole rings. The signals of the aromatic protons of the
perylene ring are split into a multiplet and three doublets at
respectively 8.593, 8.4817, 8.416 and 8.335 ppm. The concen-
tration related changes point to the fact that these must be due to
intermolecular interactions. We believe that these changes are
due to donor/acceptor intermolecular interactions since the
changes in chemical shifts affect both the carbazole donor
moieties and the PDI acceptor moieties on monomer 6.
The preparation of polymer P1 is outlined in Scheme 2. The
polymer was obtained in good yields upon reaction of 5,50-bis-
(tri-n-butylstannyl)-2,20-bithiophene with 6 and using a Stille
type cross coupling condensation polymerization reaction. The
Stille coupling was chosen in preference to the Suzuki coupling
for this polymerization reaction in order to avoid any side
reactions as a consequence of alkaline conditions that might
involve partial hydrolysis of the perylene diimide substituents on
P1 during its preparation. P1 was purified by Soxhlet extraction
with methanol, acetone, toluene then chloroform and the chlo-
roform extracts afforded the first fraction of the polymer which
has a weight average molecular weight Mw of 28,800 Da and
a polydispersity (PD) of 2.2. The residues left in the thimble after
extraction with chloroform were refluxed with chloroform and
afforded a second fraction of the polymer after precipitation in
methanol with a weight average molecular weight Mw of 46,900
Da and a polydispersity PD of 1.8. It is notable that the second
fraction of the polymer is soluble in chloroform at room
This journal is ª The Royal Society of Chemistry 2011
J. Mater. Chem., 2011, 21, 851–862 | 853

View Online
higher than that of the equivalent polymer without methyl
groups on carbazole repeat units reported in the literature.11
The absorption spectra of P2 in toluene solutions and as a thin
film are shown in Fig. 1(a) and display absorption maxima at
394 nm in toluene solutions and at 419 nm as a thin film. The
optical band gap of P2, as determined from the onset of its
absorption band in the solid state, is 2.48 eV. This is slightly
higher than that reported for the equivalent polymer without
methyl groups on carbazole repeat units reported in the literature
(2.36 eV).11 This difference can be explained by the additional
3,6-dimethyl groups on P2 which induce a higher degree of
twisting between the carbazole and the bithiophene repeat units
along the polymer chains, resulting in a reduction of the elec-
tronic conjugation and a small increase in the band gap.
3. Results and discussion
3.1. Absorption and PL emission in solution and thin-film
The absorption spectra of perylene diimides in solutions gener-
ally display three distinct vibronic transitions; (0,0); (0,1); (0,2)
and one shoulder indicating a fourth transition, (0,4). These are
separated by 200 meV corresponding to the C]C stretching
mode. This is indeed observed for PDI 2 as shown in Fig. 1(a).
The ratio between the (0,0) transition and the (0,1) transition is
an indication of the degree of self-assembly of perylene diimides
and can be calculated as 1.66 in both THF and toluene solution
for PDI 2 (only absorption in toluene is shown in Fig. 1(a)). This
suggests that the PDI-units are not affected by aggregation at the
concentrations investigated.9a However, in films of PDI 2,
aggregation becomes predominant, resulting in a red-shift of the
peak maxima and an increase of the (0,1) vibronic transition
relative to the (0,0) transition leading to a lower (0,0)/(0,1) ratio.
These spectral changes derive from a p–p stacking of the PDI
chromophores as observed for similar perylene diimides with
only hydrogen substituents at the core. This stacking has the
form of a face-to-face assembly (H-type aggregation).12 Due to
this aggregation the onset of absorption in thin films is shifted to
higher wavelengths, resulting in an optical band gap of 2.16 eV
for PDI 2 as determined from the onset of its absorption band in
the solid state.
The absorption spectra of the P1 double-cable-polymer in
THF and toluene solutions and as thin films are shown in
Fig. 1(b). Table 1 summarises the measured absorption maxima
and associated extinction coefficients. It can be seen from the
figure that the absorption spectrum of P1 in toluene consists to
some extent a superimposition of the absorption spectra of PD1
2 and P2 (compare with Fig. 1(a)). We assign the strong
absorption band at 392 nm to absorption from the carbazole-
bithiophene backbone of the polymer while the bands at 494 nm
and 527 nm arise from the absorption of the PDI substituents on
P1. However, the absorption bands of the PDI substituents on
P1 bear more of a resemblance to those of compound 2 in a thin
film rather than in solution, indicating a likely aggregation of the
PDI chromophores of P1 in solutions of THF and toluene. It is
tempting to assume that this stacking occurs mainly within the
polymer chain itself rather than between PDI units of different
chains as a result of intramolecular aggregation. Aggregation of
PDI units on a single molecule is distinctly possible; as such
moieties of P1 are covalently bound to the polymer backbone
which leads to high local concentrations and in turn to aggre-
gation. However, we cannot exclude intermolecular interactions
of PDI units from different polymer chains at this stage.
The comparison of the absorption spectra of P1 in solutions
and as thin film (Fig. 1(b)), shows that the absorption band of the
carbazole-bithiophene backbone of the polymer displays a red-
shift (15 nm) in the solid state compared to that in solution
indicating better electronic delocalisation and a more planar
polymer backbone in films of P1. In contrast, the absorption
bands corresponding to the PDI-units show only minor changes
in the position of their absorption bands in THF, toluene and in
films. It appears however that the degree of aggregation of PDI
substituents on P1 depends on the solvent used, as the (0,0)/(0,1)
absorption band ratio changes in different solvents. These ratios
in THF, toluene and CHCl3 are respectively 0.76, 0.82 and 1.04
while that in thin films is 1.02. Interestingly the PDI moieties on
Table 1 Absorption maxima and extinction coefficients of P1 in solu-
tions and as thin films
THF solution
Toluene solution
Thin film
3/Mꢂ1 labs max 3/Mꢂ1 labs max
labs max
nm
/
/
/
Transition
cmꢂ1 cmꢂ1
nm nm
Fig. 1 (a) UV/Vis absorption spectra of polymer P2 in toluene (dashed
line), P2 as a film (solid line), PDI 2 in toluene (dash dotted line) and PDI
2 as a film (dotted line). (b) UV/Vis absorption spectra of polymer P1 in
THF (dotted line), toluene (dashed line), chloroform (solid line) and as
a film (dash dotted line).
PDI-unit (0,0)
PDI-unit (0,1)
Polymer backbone 392
527
494
17700 530
22800 495
39400 393
18200 532
22000 495
35600 407
854 | J. Mater. Chem., 2011, 21, 851–862
This journal is ª The Royal Society of Chemistry 2011

View Online
P1 appear to be more aggregated in solutions of toluene and
THF than in the solid state. Despite such apparent changes in the
PDI absorption, it is clear that the absorption of P1 is an
approximate superposition of that of P2 and PDI suggesting that
there is no ground state charge transfer in the double-cable-
polymer system.
backbone of P1 to its PDI substituents leads to the formation of
long-lived metastable charge states.
3.2. Spectroscopy of thin-film blends of P1 and PDI
In Fig. 2(b), we plot PL emission spectra from thin films of
polymer P1, P2, PDI 1 dispersed in P2 (at 1 : 1 by weight) and for
comparison PDI 1 dispersed in the electronically inert polymer
poly(methyl methacrylate) (PMMA) at the same concentration.
Notably, the vibronic structure evident in the solution spectra of
PDI 1 (shown in Fig. 2(a)) is absent in the thin film. Instead we
observe a broad emission band that peaks at 624 nm. Such
emission is relatively red-shifted and broadened compared to
that from PDI 1 in solution phase (see Fig. 2(a)), and suggests
that intermolecular interactions may well be present. Indeed,
previous studies have identified excimer emission in perylene-
based materials.13
The emission spectra of P1, P2 and PDI 1 in dilute solutions
(10ꢂ6 M) are shown in Fig. 2(a). They are corrected for relative
absorbance and excitation fluence, and thus can be used as an
approximate measure for fluorescence quantum yield. We find
that PDI compound 1 emits intense fluorescence (as is the case
for a large number of such compounds) with polymer P2 emit-
ting relatively modest fluorescence in comparison. Upon excita-
tion at 480 nm, we find that emission from polymer P1 closely
resembles that of PDI 1, however its emission intensity is over 5
orders of magnitude weaker indicating almost complete exciton
quenching and concomitant suppression of PL. A very similar
result is also obtained when P1 is excited using shorter wave-
length radiation that is principally absorbed by the backbone
polymer. This is consistent with the results published for a PDI
functionalised alternating fluorene-phenylene polymer,9a which
also showed residual fluorescence from both the PFP and PDI
moieties. We can understand such quenching on the basis of
rapid charge transfer from the backbone of P1 to its PDI
substituents. We show in section 3.5 that charge transfer from the
The data in Fig. 2(b) is again corrected for relative absorbance
and excitation fluence, and thus can be used as an approximate
measure for fluorescence quantum yield. As was found in solu-
tion measurements (see Fig. 2(a)), it is apparent that the emission
from the P1 double-cable polymer is strongly quenched
compared to either P2 or PDI 1. Critically however, we find that
the emission from the P1 double-cable polymer is weaker than
that emitted by the P2/PDI 1 blend. As we argue section 4, this
result is consistent with phase-separation and crystallization in
the polymer/perylene blend. It can be seen that the P1 polymer
PL emission spectrum resembles that of the PDI monomer,
having characteristic sharp vibronic peaks at 526 and 567 nm
(although these are blue-shifted by 10 nm compared to PDI in
solution, see Fig. 2(a)). Interestingly, we observe a broad emis-
sion band in the as cast P2/PDI 1 blend centred at 736 nm. We
believe that this emission may result from exciplex states; our
assignment is based on the fact that this feature appears only in
the emission of the donor–acceptor systems and does not
correspond to the fluorescence of either moiety. It is also only
accessible from the excited state. Similar exciplex emission has
also been observed in perylene/pthalocyanine thin-film blends.14
Note that the emission from such red-shifted exciplex-like states
is strongly quenched in P2/PDI 1 blends on annealing and thus
we assume, in this case, that they play only a minor role in device
physics. The contribution of excimer and exciplex-like states in
the P1 polymer emission is an area of ongoing research.
In Fig. 3(a), we plot the absorption coefficient of P1 and a 1 : 1
P2/PDI 1 blend. We find that the absorbance of P1 can be
approximately described by a linear superposition of PDI 1 and
P2 absorption spectra confirming that there is no ground state
charge transfer and little electronic hybridization between the
donor and acceptor. The PDI 1 and P2 blend have however an
absorption tail that extends to red-wavelengths which is most
probably accounted for by increased optical scattering within the
blend thin films resulting from phase-separation (vide infra).
In Fig. 3(b) we plot the emission from a 1 : 4 P2/PDI 1 blend
film excited at 490 nm before and after heat treatment on the
same intensity scale. It is apparent that the heat treated film
shows enhanced fluorescence which suggests reduced exciton
quenching due to coarsened phase-separation. This, as we later
demonstrate, is associated with improved OPV device perfor-
mance.
Fig. 2 (a) PL spectra of 10ꢂ6 mol dmꢂ3 solutions in toluene of polymer
P1 excited at 480 nm (dotted line), P2 excited at 360 nm (dashed line) and
PDI 1 excited at 480 nm (solid line). (b) Thin film PL spectra as cast from
chloroform solution: P1 (dotted line) excited at 480 nm, P2 (dashed line)
excited at 410 nm scaled by 1/100, PMMA/PDI 1 (1 : 1 b.w.) blend (solid
line) excited at 480 nm, P2/PDI 1 (1 : 1 b.w.) blend (dash-dot line) excited
at 480 nm. All spectra are corrected for sample absorption and excitation
fluence.
This journal is ª The Royal Society of Chemistry 2011
J. Mater. Chem., 2011, 21, 851–862 | 855

View Online
polymer. It also displays a reduction wave at ꢂ2.42 V with an
associated oxidation wave at ꢂ2.37 V that corresponds to the
reduction of the polymer backbone of the polymer. The PDI
moieties on P1 can be clearly identified on its cyclic voltammo-
gram by the reduction waves at ꢂ1.29 V and ꢂ1.47 V and two
associated oxidation waves that show up as one broad peak
at ꢂ1.15 V.
From the onset of oxidation (+0.58 V) and that of the reduc-
tion (ꢂ2.32 V) of the polymer backbone of P1, the ionisation
potential, IP, and the electron affinity, EA, for the polymer
backbone can be derived as ꢂ 5.38 eV and ꢂ 2.48 eV respectively
(on the basis that ferrocene/ferrrocenium is 4.8 eV below the
vacuum level9). Therefore the electrochemical band gap can be
calculated as 2.9 eV. It is not possible to calculate the optical
band gap of polymer P1 directly from its absorption spectrum in
films since the edge of the absorption from the polymer backbone
is masked by the absorption of the PDI substituents, however, we
can safely say that the band gap of the backbone of the polymer
should be slightly higher than that of polymer P2 since the
absorption maxima for the backbone of the polymers are
respectively at 407 nm for P1 and 419 nm for P2. The optical
band gap of P1 should be around 2.6 eV (optical band gap of P2
is 2.48 eV). This value of 2.6 eV is lower than that deduced from
the electrochemical measurements (2.9 eV), however, it is not
uncommon to observe slight variations between the optical and
electrochemical band gaps. The LUMO level of the PDI moieties
attached to P1 as determined from their onset of reduction
(ꢂ1.06 V) in Fig. 4(a) is ꢂ3.74 eV. On the basis of these
measurements, we present the energy band diagram for P1 in
Fig. 5 and compare it with the known HOMO and LUMO levels
of PDI. It is clear that efficient chatge transfer is expected in P1
since PDI has a large electron affinity with its LUMO level lying
some 1.26eV deeper than the LUMO level of P1.
Fig. 3 (a) Absorption of P1 film (dotted line) and 1 : 1 P2/PDI 1 blend
(dot dashed line). (b) PL from a 1 : 4 P2/PDI 1 blend film before and after
thermal treatment (solid and dashed lines respectively) excited at 490 nm.
3.3. Electrochemical studies
Cyclic voltammetry measurements on drop-cast P1 and P2
polymer films were conducted in acetonitrile with tetrabuty-
lammonium hexafluorophosphate as an electrolyte. The cyclic
voltammograms of these materials are shown in Fig. 4.
The cyclic voltammogram of P2 is very much similar to that of
P1 (Fig. 4(b) apart from the absence of the features resulting
from the PDI units. The ionisation potential, IP, and the electron
affinity, EA, for polymer P2 can be derived as ꢂ5.38 eV and
ꢂ2.34 eV respectively.
The cyclic voltammogram of P1 (Fig. 4(a)) displays an
oxidation wave at +0.67 V and an associated reduction wave at
+0.49 V that corresponds to the oxidation of the backbone of the
Fig. 4 CV curves of thin films of (a) polymer P1 and (b) polymer P2 on
platinum disc electrodes (area 0.031 cm2) at a scan rate of 100 mV sꢂ1 in
acetonitrile/tetrabutylammonium hexafluorophosphate (0.1 mol dmꢂ3).
Fig. 5 Energy level diagram of the polymer backbone and PDI moieties
on polymer P1.
856 | J. Mater. Chem., 2011, 21, 851–862
This journal is ª The Royal Society of Chemistry 2011

View Online
3.4. Thin-film morphology
self-organization in all films, even those cast from a single
material. For example, characteristic elongated crystals of PDI
are seen in Fig. 6(b) and also in a matrix of P2 as shown in
Fig. 6(d) and (e).7f,15,16 Thermally assisted diffusion of PDI leads
to the growth of PDI crystals and the formation of smaller
surface crystallites as is seen in Fig. 6(f). A clear comparison is
made in Fig. 6(g) where we plot the deviation of the height from
the mean for the 1 : 4 P2/PDI 1 blend before and after thermal
treatment (left and right panels respectively). We find that the as
spun film shows a bimodal height distribution, and is fitted by
two Gaussians. The primary peak is offset from the mean by
ꢂ2.3 nm with a FWHM of 6.6 nm and a secondary peak centred
around +14.1 nm from the mean which we assign to the elon-
gated crystals appearing on the surface. After thermal treatment,
the height distribution broadens (FWHM increases to 9.4 nm)
and loses its bimodal character. We tentatively attribute this to
the breakup of the large PDI crystals and an increase in sub-
micron length-scale roughness owing to further PDI crystal-
lisation and associated P2 enrichment.
We have used tapping mode Atomic Force Microscopy (AFM)
to study thin-films of P1, PDI 1, P2 and blends of P2 and PDI 1.
The results of our studies are shown in Fig. 6, where we plot
AFM images of thin-films of thermally annealed (120 ꢁC for
10 min) P1 (part (a)), as spun PDI 1 (part (b)), as spun P2 (part
(c)), a thermally annealed 4 : 1 P2/PDI 1 blend (part (d)), an as
spun 1 : 4 P2/PDI 1 blend (part (e)) and a heat treated 1 : 4 P2/
PDI 1 (part (f)). It is evident that there is a degree of
The coarse morphology resulting from crystallization that is
evident in PDI 1 and the P2/PDI 1 blend thin-films is suppressed
in films of the P1 double-cable polymer as seen in Fig. 6(a).
Furthermore, thin films of P1 have an RMS roughness of
0.64 nm in contrast to a value of 18.9 nm determined from the
thermally annealed thin films of the 1 : 4 P2/PDI 1 blend. This
relative flatness in the double-cable polymer is indicative of a lack
of coarse phase-separation that results from the covalent link-
ages between PDI substituents and the backbone of P1 that
prohibit their demixing and demonstrates the potential of the
double-cable concept to affect film morphology.
3.5. Photoinduced absorption spectroscopy
We have used photoinduced absorption (PA) spectroscopy to
study the various excited charged states of P1 double-cable
polymer and its constituent components. The PA signatures of
P1, a 4 : 1 P2/PDI 1 blend, P2 and a 1 : 4 blend of P2 with PCBM
are shown in Fig. 7(a), (b), (c) and (d) respectively. In parts (a)
and (b), we detect prominent features that arise from the PDIꢂ
radical anion which are located at 1.28, 1.54 and 1.72 eV in the
P2/PDI blend and in the P1 double-cable polymer.8,17 We can
gain further insight into these spectra on the basis of the PA
spectrum of the 1 : 4 P2/PCBM blend (shown in Fig. 7(d)). Here,
correlated P2+ bands are seen at 0.56 and 1.83 eV that we assign
to the P1 and P2 transitions respectively. A feature at 1.25 eV is
ꢂ 18
also observed and we assign this to the C60
.
On the basis of
measurements on the 1 : 4 P2/PCBM blend, we are able to
identify the low energy P1 transition of P2+ radical cation which
is evident in the PA spectra of both P1 and the 4 : 1 P2/PDI 1
blend (see parts (a) and (b)). Note that the high energy P2 tran-
sition of the P2+ radical is likely to be superimposed on the PDIꢂ
transition that is located at around 1.8 eV.
Further support for the polaron assignment made here derives
from the FBC model that predicts the intragap polaron levels.19
In the model, if the p–p* energy gap is equivalent to 2D and u is
half the energy separation of the two polaron levels, then the
following ratio u/D $ 1/O2 is expected. From the absorption
spectrum of P2 shown in Fig. 1(a) we determine a p–p* energy
gap of 2.48 eV. From the PA spectrum shown in Fig. 7(d), we
Fig. 6 Tapping mode AFM height images. Part (a): P1 after thermal
treatment 10 min at 120 ꢁC, part (b): pristine PDI 1, part (c): pristine P2,
part (d): heat treated 4 : 1 P2/PDI 1 blend, part (e): pristine 1 : 4 P2/PDI 1
blend, part (f): heat treated 1 : 4 P2/PDI 1 and part (g): height distribu-
tion of image (e) and (f) in the left and right panel respectively with fits.
Note that all images have the same lateral scale but the height scaling
varies between images.
This journal is ª The Royal Society of Chemistry 2011
J. Mater. Chem., 2011, 21, 851–862 | 857

View Online
carbazole.20 Note that without access to LESR or PADMR spin
resolved measurements, any excited state assignment must still be
considered tentative at this stage.
We can relate the PA signal to the density of photoexcited
states using ꢂDT/T z nds as described by Lanzani et al.21 Here,
n is the density of excited states, s is their absorption cross-
section and d is the thickness of sample. It is more appropriate to
use the penetration depth of the pump laser, or 1/aL, as this is the
effective thickness in which most of the laser is absorbed. As the
absorption coefficient of all the films studied are roughly
comparable, we can relate DT/T measured for the different
samples to the polaron yield. Comparing the peak values of DT/T
in Fig. 7 parts (a), (b) and (d), it appears that that there is a larger
photoexcitation density in P1 in comparison with both the P2/
PDI and P2/PCBM blends.
Using our measured parameters, we determine a singlet-
exciton to polaron yield of hp ¼ 6% (see ESI†). Although this
value is smaller than expected from the efficient exciton
quenching that we observe in P1, we associate this yield with
a long-lived (s ¼ 0.56 ms – see ESI†) population of trapped
charge carriers that we discuss later. We note that this value is
relatively large in comparison with other studies by List et al.22
where a yield of 0.08% was measured in a fluorescent conjugated
ladder-type polymer, however in such materials the direct
dissociation of excitons into polarons is anticipated to be a rela-
tively inefficient process.
3.6. Device fabrication and characterization
We have made a series of OPV devices based on P1 double-cable
polymer together with devices based on a number of different
weight blends (4 : 1, 1 : 1 and 1 : 4) of P2 and PDI 1 donor and
acceptor materials. As a bench-mark, devices were also fabri-
cated from P2 in a 1 : 4 blend with the fullerene acceptor [6,6]-
phenyl-C61-butyric acid methyl ester (PCBM). The effect of
casting solvent and the use of thermal annealing protocols were
in all cases explored in order to optimize device efficiency (see
Table 2 for details). The performance metrics for the various
devices studied (JSC, VOC, FF, PCE and peak EQE) are listed in
Table 2. In all cases, we report the metrics of the highest efficiency
devices fabricated, along with the conditions (casting-solvent,
film-thickness and thermal-annealing protocol) used to create
such a device. It can be seen that devices based on the double-
cable polymer P1 are less efficient than those of the optimised
1 : 4, 1 : 2 and 1 : 1 P2/PDI 1 blends. Specifically, we find that the
1 : 4 P2/PDI 1 blend has a maximum power conversion efficiency
(PCE) of 0.15% compared with devices based on double-cable
polymer P1 that have a PCE of 0.011%. A clear trend also
emerges in the efficiency of the blend devices as a function of
relative blend composition, with devices having increased PDI
concentrations having improved efficiencies.
Fig. 7 Steady state PA spectra. Part (a) shows the PA spectrum from P1
(f ¼ 135 Hz), part (b): a P2/PDI 1 4 : 1 blend (f ¼ 400 Hz), part (c):
unblended P2 (f ¼ 135 Hz) and part (d): P2 (f ¼ 135 Hz) blended with
PCBM in a 1 : 1 ratio. All samples are held at 77 K, pumped at 488nm
with a photon density of 1.2 ꢃ 1017 cmꢂ2 sꢂ1
.
deduce that the P2 intragap polaron levels are spaced in energy
by 1.83 eV from which we calculate a ratio of u/D ¼ 1.36 and the
FBC criterion is satisfied. Note that bipolaron levels would be
found much deeper in the energy gap.
The relative superiority of the blend films compared with the
double-cable material is also reflected in the external quantum
efficiency (EQE) spectra presented in Fig. 8. Specifically, parts (a)
and (b) are measurements made on fully optimized devices
composed of a 1 : 4 P2/PDI 1 blend before and after thermal
treatment respectively. Part (c) shows an EQE spectrum of an
optimized device based on P1. It can be seen that the P1 device
reaches a maximum EQE of 0.43% at 405 nm, whilst the
We have also performed PA on a pure film of P2 as shown in
Fig. 7(c). Here, the spectrum is dominated by a single feature at
1.64 eV that we assign to the triplet–triplet exciton transition (T1)
of P2. This assignment is made on the basis that it appears
without an accompanying transition. Some support for this
particular assignment is found in the literature since this feature
bears a resemblance to the triplet transition in polyvinyl
858 | J. Mater. Chem., 2011, 21, 851–862
This journal is ª The Royal Society of Chemistry 2011

View Online
Table 2 Performance metrics from the P1 and P2/PDI 1 blend devices
Material
Preparation
Thickness/nm
JSC/mA cmꢂ2
VOC/V
FF (%)
PCE (%)
EQE (%)
P1
Spun from DCB, 10 min 120C
Chloroform cast 10 min 120 ꢁC
Chloroform cast 10 min 120 ꢁC
Chloroform cast 10 min 120 ꢁC
Chloroform cast 10 min 120 ꢁC
Spun from chloroform, no treatment
Slow spun from TCB, 10 min 120 ꢁC.
53
156
52
57
66
133
0.070
0.71
1.02
0.72
0.71
0.62
0.60
0.40
0.58
0.63
21
30
24
24
27
25
32
0.011
0.15
0.15
0.43
3.7
5.0
3.1
1 : 4 P2/PDI 1
1 : 2 P2/PDI 1
1 : 1 P2/PDI 1
4 : 1 P2/PDI 1
P2
0.67
0.094
5.5 ꢃ 10ꢂ3
9.1 ꢃ 10ꢂ4
7.34
5.5 ꢃ 10ꢂ3
1.3 ꢃ 10ꢂ4
1.49
0.038
—
48
a
1 : 4 P2/PCBM
Not determined
a
Beyond instrument sensitivity.
intermolecular interaction resulting from crystallization in the
thermally annealed blend.
In general, we find that optimum performance from P1
double-cable polymer based devices results from using thin-films
cast from a slow-drying solvent (dichlorobenzene). In contrast,
optimum performance from P2/PDI 1 blend-based devices is
achieved when a much more volatile solvent (chloroform) is used
to cast thin films. This observation suggests that allowing the P1
double-cable polymer to undergo self-organization enhances the
disassociation and extraction of charge, whereas this is not the
case for the blend. Rather, it appears that a thermal annealing
step in conjunction with a fast drying solvent encourages the
formation of PDI crystals which we presume are dispersed in
a matrix of P2. It is possible that the use of slow-drying solvents
to cast blend films results in the formation of domains that are so
large that exciton dissociation is suppressed resulting from the
low density of interfaces between donor and acceptor compo-
nents.
As anticipated, the lowest efficiency devices (1.3 ꢃ 10ꢂ4
%
PCE) were produced from a ‘pristine’ film of polymer P2.
Clearly, the lack of a donor–acceptor couple in this co-polymer
system results in a very low degree of charge separation and thus
very poor device efficiency. When polymer P2 is blended with
PCBM and cast from TCB, devices created had a PCE of 1.5%.
On the basis of the model by Scharber et al.,23 we can expect
a maximum power efficiency of PCBM/P2 blends of ꢀ2%, sug-
gesting that our devices are relatively well optimized.24 In
contrast to what is observed in P1 or the P2/PDI 1 blends, it
appears therefore that the devices based on a blend of P2 and
PCBM are significantly more efficient than those utilizing PDI,
as an acceptor material. As we argue below, we believe that P2/
PDI 1 and P1 devices are dominated by geminate or non-gemi-
nate recombination; an effect that is presumably suppressed in
the P2/PCBM devices.
Fig. 8 Comparison of EQE spectra and active layer absorbance. Part
(a): the most efficient 1 : 4 P2/PDI 1 device prior to annealing, part (b):
the same device after annealing and part (c): the most efficient P1 device.
In Fig. 9(a) and (b), we plot current–voltage (J–V) spectra
from a P1 and a 1 : 4 P2/PDI 1 OPV device respectively recorded
under illuminated and dark conditions plotted on a semilog
scale. The inset shows the same J–V traces over a ꢂ1 V to +1.5 V
range on a linear scale. It can be seen that the P1 based device
exhibits an approximately linear J–V trace up to VOC with low fill
factors indicative of high series and low shunt resistances25 that
most likely arises from poor charge transport. An improved Fill
Factor is evident in the 1 : 4 P2/PDI 1 device. We can quantify
charge transport in the devices we have studied by estimating the
active region charge carrier mobility by fitting to the quadratic,
thermally annealed 1 : 4 P2/PDI 1 blend has a maximum EQE of
3.7% at 462 nm. Thus the optimized blend device is at least six
times more efficient than the cable polymer based device. Inter-
estingly, the spectral shape of the absorbance and the EQE
spectra of the annealed and un-annealed blend devices are
different, with (0,0) electronic transition in the annealed device
being relatively suppressed compared to the (0,1) transition. As
discussed above, this observation suggests increased
This journal is ª The Royal Society of Chemistry 2011
J. Mater. Chem., 2011, 21, 851–862 | 859

View Online
estimates are subject to an uncertainty of ca. 20% arising from
errors in measuring device thickness and fitting, however it
appears that carriers in the P2/PDI 1 blends have a mobility that
is approximately 10 times larger than that in the P1 device.
As summarized in Table 2, devices based on polymer P1 are
characterized by an open circuit voltage (VOC) of 0.72 V. The
VOC of the P2/PDI 1 blend ranges from 0.4 V to 0.71 V as the
blend composition is varied between 4 : 1 and 1 : 4 P2/PDI
1 respectively. Empirically, Scharber et al.23 have shown that
VOC ¼ (1/e)(LUMOA - HOMOD)-0.3 V where e is the elementary
charge. Using the same reasoning, we should expect a value of
VOC ¼ 1.34 V in devices based on double-cable polymer P1. This
is however not observed in practice, with VOC in all devices being
much lower than ideality. Even in state of the art P3HT/PCBM
devices, there is some voltage loss (ca. 0.3 V) owing to the non-
ideal nature of the current–voltage characteristic.27 Because VOC
is the voltage at which the injected current compensates the
photocurrent, any dark-injection will act to reduce VOC from the
idealized limit of Vbi which here is anticipated to be 1.64 V
determined from the difference between acceptor LUMO and
donor HOMO levels. The most likely explanation for the
discrepancy is that photocurrent is largely field driven with the
reduced charge carrier mobility responsible for a reduced VOC
.
An insight into loss mechanisms in the devices studied can be
gained by studying the dependence of the short-circuit current
(JSC) against incident flux as presented in Fig. 9(c). We find that
for the double-cable polymer based device, JSC is to a good
approximation linear with optical power (fitted by an exponent
of 0.92). This suggests that the dominant charge-carrier recom-
bination mechanism within P1 based devices is likely to be via
geminate recombination of charge carriers rather than a bimo-
lecular process involving non-geminate recombination. Bimo-
lecular recombination would instead have a sub-linear JSC vs.
incident flux following a JSC ꢀ P3/4 law.28 In the blend devices
however, the situation is more complicated, with JSC being fitted
by an exponent that varies between 0.87 in a 4 : 1 P2/PDI 1 blend
to 0.96 in a 1 : 4 P2/PDI 1 blend. This result suggests that in
devices with a low PDI fraction, recombination is predominantly
bi-molecular. As the PDI concentration increases however, the
device efficiency improves with recombination becoming signif-
icantly more monomolecular in character. This finding is in
qualitative accord with those of Howard et al.,29 who also
identified bimolecular recombination as a dominant loss mech-
anism in blends of PDI and a conjugated polymer, where the PDI
and polymer had relative mass fractions of 3 : 2 respectively.
Fig. 9 Current–voltage spectra under simulated AM1.5G irradiation at
1 Sun and in the dark with space charge limited current fits, part (a): the
most efficient P1 device after annealing, part (b): the most efficient 1 : 4
P2/PDI 1 device after annealing. The same data is plotted on a linear scale
inset. Part (c): JSC vs. 488 nm laser illumination intensity with fits for P1
(closed squares), 1 : 4 P2/PDI 1 (open squares) and 4 : 1 P2/PDI 1 (open
circles) devices.
space charge limited region of the dark J–V curves presented in
Fig. 9(a) and (b). Specifically at high applied-bias, we fit the
following space-charge limited expression26 that includes a Pool–
Frenkel mobility term
4. Discussion
!
rffiffiffiffi
9
8
V
V2
L3
We find that photovoltaic devices based on a double-cable
polymer-type material in which donor and acceptor moieties are
covalently linked are less efficient than those of an optimised
blend of donor and acceptor materials. It is interesting to explore
the reasons for the reduced efficiency of the double-cable poly-
mer. We found that the J–V traces of the P1-based devices are
almost linear, (symmetric around VOC) suggesting that charge
transport is poor. This conclusion is supported by the fact that
the short circuit currents and fill-factors measured are low
resulting from poor charge transport and resistive losses within
the device.25
JeðhÞ
¼
303rmeðhÞexp 0:891geðhÞ
(1)
L
Using a value of 3 for 3r, together with the device thickness (L)
listed in Table 2, we obtain values of charge carrier mobility of
4.0 ꢃ 10ꢂ7 cm2 Vꢂ1 sꢂ1 and 3.8 ꢃ 10ꢂ6 cm2 Vꢂ1 sꢂ1 for the P1 and
1 : 4 P2/PDI 1 devices respectively. The fits used to derive these
values are also shown in Fig. 9(a) and (b) along with the exper-
imentally determined J–V spectra. Note that should transport be
imbalanced between donor and acceptor, the values calculated
here are representative of highest mobility carrier. These
860 | J. Mater. Chem., 2011, 21, 851–862
This journal is ª The Royal Society of Chemistry 2011

View Online
It is clear however that excitons are quenched efficiently in P1,
as seen from the suppresion of PL evidenced in Fig. 2(b).
However, the low values of EQE and JSC measured demonstrate
that charge is not being efficiently collected. On the basis of
previous work,30,31 we assume that the long-lived polarons
detected in PA are associated with charge stored in deep traps
whereas photocurrent measurements on devices sample
a different population of mobile polarons. This suggests there-
fore that there is a large population of charge carriers that
undergo recombination on a timescale that is faster than the
resolution of our PA measurement. Our measurements of
photocurrent vs. optical pump-power on P1 devices suggest that
this recombination mainly occurs via geminate recombination.
Given the lack of significant observable structure in AFM
height images of P1, we conclude that the donor and acceptor
components are intimately mixed at least at a length-scale below
the resolution of our microscope (ꢀ10 nm corresponding to the
tip radius), suggesting a lack of the mesoscale order of donor and
acceptor domains essential for a bicontinuous charge trans-
porting network. It is likely that the resultant morphology in the
double-cable polymer also prevents complete exciton separation
into free carriers and results in rapid geminate recombination.
Indeed work on polymer-polymer blends32 has argued that the
spatially-confined geometry present in an intimately mixed
polymer blend suppresses charge-separation and results in
geminate recombination. It is also possible that the relatively
high ‘steady-state’ trapped-charge population detected in P1 also
directly results from charges stabilized on PDI molecules that are
effectively ‘‘isolated’’ from the polymer backbone by the satu-
rated linker group.
bimolecular recombination. This apparently switches over to
monomolecular recombination in more efficient devices in which
PDI is in excess. Bimolecular recombination has previously been
observed in PDI/conjugated polymer based devices.29 We there-
fore suggest that there is a competition between geminate and
non-geminate recombination dynamics in P2/PDI blends. At
high PDI fractions, separate percolation pathways exist for
electrons and holes, formed as a result of phase-separation.33
This spatial-separation of electron and hole currents, together
with a general increase in carrier mobility in the blends results in
a reduced bimolecular recombination rate, with a consequence
being that monomolecular recombination effectively dominates.
However as the PDI fraction is reduced, self-organization of the
PDI is suppressed, resulting in a lack of percolation paths for
electrons and a general reduction in the bulk mobility of charge
carriers. The consequence of this is that the bimolecular recom-
bination rate apparently increases and ‘overtakes’ the mono-
molecular recombination channel. Clearly, time-resolved
spectroscopy would be useful in confirming this hypothesis.
5. Conclusion
We have synthesized and characterized a double-cable polymer
composed of perylene diimide acceptor groups that are cova-
lently linked to 2,7-linked carbazole bi-thiophene polymer
chains. Excitons created on this polymer following optical exci-
tation are very efficiently dissociated as fluorescence from the
polymer is strongly quenched both in solution and in thin film.
Photoinduced absorption measurements indicate that polarons
are generated on the double-cable polymer with high yields.
Photovoltaic devices utilizing the double-cable polymer generate
a symbatic photoresponse, demonstrating charge transport
arising from photoexcitations. However such devices have low
peak quantum efficiency resulting from poor charge transport
and resistive losses within the device. Comparative studies on
devices based on thin-film blends of perylene diimide moieties
and a carbazole bi-thiophene polymer without PDI substituents
had a higher efficiency than those based on the double-cable
polymer. AFM studies on thin films of the double-cable polymer
suggest that phase-separation is almost completely suppressed,
whereas a coarser grained structure is observed in the thin-film
blend. Herein lies the explanation for reduced device efficiency of
the double-cable polymer; such fine-scale phase separation
results in geminate recombination since the absence of phase
separated domains prevent the separation of electron-hole pairs
into free carriers. This latter effect is consistent with the emer-
gence of charge-traps that are evidenced by a large and long-lived
polaron population (see Fig. S3, ESI†). Optimized blend devices
based on an excess of PDI acceptor molecules however are
characterized by a reduced trapped polaron population and
superior operational efficiency. We propose that this results from
the formation of improved charge extraction pathways resulting
from phase-separation and crystallization of the PDI acceptor
molecules.
AFM measurements on P2/PDI 1 blends suggest that phase-
separation occurs over significantly larger length-scales than in
the double-cable polymer - an effect that results in a reduced
exciton quenching efficiency. On thermal annealing, we find that
the PL intensity emitted by the blends increases, a result
consistent with the increased phase-separation, domain nucle-
ation and crystallization of the PDI (as evidenced by AFM and
modified device absorbance). Despite the reduced exciton
quenching efficiency in annealed P2/PDI 1 blends, it appears that
a coarsening of phases in an excess of PDI is necessary for effi-
cient device operation. However, in P1 devices, neither of these
criteria is met since the weight concentrations of donor and
acceptor are roughly equal and phase separation is arrested by
their covalent-linkage. Consequently, we see an increased EQE in
P2/PDI 1 in comparison with the double-cable polymer based
devices as shown in Fig. 8. Although exciton quenching is very
efficient in P1, it is likely that device efficiency is limited by either
poor extraction of carriers or incomplete separation of gemin-
ately bound electron-hole pairs. The observation of a large
polaron yield with monomolecular recombination kinetics would
suggest that it is the latter. By contrast, we find a reduced polaron
yield in a 4 : 1 P2/PDI 1 blend, illustrating that fewer polarons
become trapped – presumably as a result of an improved nano-
scale-morphology.
We have found that the recombination kinetics of the P2/PDI
1 blend apparently change as the composition of the material is
varied. In blend materials that make the worst performing
devices (having a low PDI fraction), the photocurrent depen-
dence against pump-power is sub-linear, suggestive of
Acknowledgements
This work was funded by the Engineering and Physical
Sciences Research Council (EPSRC) under grant EP/F056397:
This journal is ª The Royal Society of Chemistry 2011
J. Mater. Chem., 2011, 21, 851–862 | 861

View Online
ꢀ
R. Esquembre, R. Gomez, R. Blanco and J. L. Segura, J. Phys.
Chem. C, 2008, 112, 16668.
‘‘High-efficiency Block Copolymer Solar Cells: A Scalable
Prototype for Low Cost Energy Generation’’.
10 M. J. Ahrens, R. F. Kelley, Z. E. X. Dance and M. R. Wasielewski,
Phys. Chem. Chem. Phys., 2007, 9, 1469.
11 S. Wakim, N. Blouin, E. Gingras, Y. Tao and M. Leclerc, Macromol.
Rapid Commun., 2007, 28, 1798.
Notes and references
1 (a) C. Brabec, Sol. Energy Mater. Sol. Cells, 2004, 83, 273; (b)
€
S. Gunes, H. Neugebauer and N. Sariciftci, Chem. Rev., 2007, 107,
1324.
12 T. van der Boom, R. T. Hayes, Y. Zhao, P. J. Bushard, E. A. Weiss
and M. R. Wasielewski, J. Am. Chem. Soc., 2002, 124, 9582.
13 J. Ferguson, J. Chem. Phys., 1966, 44, 2677–2683.
2 (a) G. Dennler, M. Scharber and C. Brabec, Adv. Mater., 2009, 21,
1323; (b) G. Li, J. Huang, Y. Tao, T. Moriarty, K. Emery and
Y. Yang, Nat. Mater., 2005, 4, 864; (c) S. Park, A. Roy,
~
14 R. Aroca, T. Del Cano and J. de Saja, Chem. Mater., 2003, 15, 38–45.
15 M. Mamada, J. Nishida, S. Tokito and Y. Yamashita, Chem.
Commun., 2009, 2177–2179.
ꢀ
S. Beaupre, S. Cho, N. Coates, J. Moon, D. Moses, M. Leclerc,
16 S. Han, T. Ha, C. Kim and K. Kim, Langmuir, 1998, 14, 6113–6120.
17 J. Salbeck, J. Electroanal. Chem., 1992, 340, 169–195.
18 K. Lee, R. Janssen, N. Sariciftci and A. Heeger, Phys. Rev. B., 1994,
49, 5871–5784.
19 K. Fesser, R. Bishop and D. Campbell, Phys. Rev. B, 1983, 106, 5136–
5190.
20 (a) D. Comoretto, I. Moggio, C. Cuniberti and G. Dellepiani, Synth.
Met., 1998, 94, 229–234; (b) G. Dellepiani, C. Cuniberti,
D. Comoretto and G. Musso, Phys. Rev. B: Condens. Matter, 1993,
48, 7850–7856.
21 V. Wohlgennant, Ehrenfreund and Z. V Vardeny., Spectroscopy of
Long-Lived Photoexcitations in p-Conjugated Systems; Lanzani, G.,
Ed.; Wiley-VCH, Weinheim 2006, (ch. 5).
K. Lee and A. Heeger, Nat. Photonics, 2009, 3, 297; (d) H. Yi,
R. G. Johnson, A. Iraqi, D. Mohamad, R. Royce and
D. G. Lidzey, Macromol. Rapid Commun., 2008, 29, 1804; (e)
M. Campoy-Quiles, T. Ferenczi, T. Agostinelli, P. Etchegoin,
Y. Kim, T. Anthopoulos, P. Stavrinou, D. Bradley and J. Nelson,
Nat. Mater., 2008, 7, 158.
3 A. Cravino and N. S. Sariciftci, J. Mater. Chem., 2002, 12, 1931.
4 (a) S. Luzzati, M. Scharber, M. Catellani, N.-O. Lupsac,
F. Giacalone, J. L. Segura, N. Martin, H. Neugebauer and
N. S. Sariciftci, Synth. Met., 2003, 139, 731; (b) S. Luzzati,
M. Scharber, M. Catellani, F. Giacalone, J. L. Segura, N. Martin,
H. Neugebauer and N. S. Sariciftci, J. Phys. Chem. B, 2006, 110,
5351; (c) M. Catellani, S. Luzzati, N.-O. Lupsac, R. Mendichi,
R. Consonni, A. Famulari, S. V. Meille, F. Giacalone, J. L. Segura
22 E. List, C. Kim, A. Naik, U. Scherf, G. Leising, W. Graupner and
J. Shinar, Phys. Rev. B: Condens. Matter Mater. Phys., 2001, 64,
155204.
ꢀ
and N. Martin, J. Mater. Chem., 2004, 14, 67; (d) R. Gomez,
R. Blanco, D. Veldman, J. L. Segura and R. A. J. Janssen, J. Phys.
Chem. B, 2008, 112, 4953.
5 (a) B. A. Gregg, J. Sprague and M. W. Peterson, J. Phys. Chem. B,
1997, 101, 5362; (b) A. J. Makinen, A. R. Melnyk, S. Schoemann,
R. L. Headrick and Y. Gao, Phys. Rev. B: Condens. Matter Mater.
Phys., 1999, 60, 14683.
6 C. W. Struijk, A. B. Sieval, J. E. J. Dakhorst, M. van Dijk, P. Kimkes,
R. B. M. Koehorst, H. Donker, T. J. Schaafsma, S. J. Picken,
A. M. van de Craars, J. M. Warman, H. Zuilhof and
E. J. R. Sudholter, J. Am. Chem. Soc., 2000, 122, 11057.
€
23 M. Scharber, D. Muhlbacher, M. Koppe, P. Denk, C. Waldauf,
A. Heeger and C. Brabec, Adv. Mater., 2006, 18, 789–794.
24 Our estimate is based on the HOMO and LUMO levels of PCBM
being ꢂ6.0 and ꢂ4.2 eV respectively and the HOMO and LUMO
levels of P2 being 5.38 and 2.9 eV respectively. Note, the LUMO
level of P2 is calculated by subtracting the optical energy gap of
2.40 eV (determined from the onset of absorption) from the HOMO
level measured using cyclic voltametry. We also assume EQE and
fill-factors of 65%.
25 C. J. Brabec, S. E. Shaheen, C. Winder, N. S. Sariciftci and P. Denk,
Appl. Phys. Lett., 2002, 80, 1288–1290.
€
7 (a) J. Li, F. Dierschke, J. Wu, A. C. Grimsdale and K. Mullen,
J. Mater. Chem., 2006, 16, 96; (b) P. E. Keivanidis, I. A. Howard
and R. H. Friend, Adv. Funct. Mater., 2008, 18, 3189; (c)
P. K. Keivanidis, N. C. Greenham, H. Sirringhaus, R. H. Friend,
J. C. Blakesley, R. Speller, M. Campoy-Quilles, T. Angostinelli,
D. D. C. Bradley and J. Nelson, Appl. Phys. Lett., 2008, 92,
023304; (d) J. J. Dittmer, E. A. Marseglia and R. H. Friend, Adv.
Mater., 2000, 12, 1270; (e) J. J. Dittmer, K. Petritsch,
E. A. Marseglia, R. H. Friend, H. Rost and A. B. Holmes, Synth.
26 P. N. Murgatroyd, J. Phys. D: Appl. Phys., 1970, 3, 151–156.
27 L. Koster, V. Mihailetchi and P. Blom, Appl. Phys. Lett., 2005, 86,
123509.
28 P. Blom, V. Mihailetchi, L. Koster and D. Markov, Adv. Mater.,
2007, 19, 1551–1566.
29 I. A. Howard, F. Laquai, P. E. Keivanidis, R. H. Friend and
N. C. Greenham, J. Phys. Chem. C, 2009, 113, 21225–21232.
30 A. Dhoot, G. Ginger, D. Beljonne, Z. Shuai and N. Greenham, Chem.
Phys. Lett., 2002, 360, 195–201.
31 T. Offermans, S. Meskers and R. Janssen, Org. Electron., 2006, 7,
213–221.
€
Met., 1999, 102, 879; (f) L. Schmidt-Mende, A. Fechtenkotter,
K. Mullen, E. Moons, R. H. Friend and J. D. MacKenzie, Science,
€
2001, 293, 1119.
8 E. E. Neuteboom, P. A. van Hal and R. A. J. Janssen, Chem.–Eur. J.,
2004, 10, 3907.
ꢀ
9 (a) R. Gomez, D. Veldman, R. Blanco, C. Seoane, J. L. Segura and
R. A. J. Janssen, Macromolecules, 2007, 40, 2760; (b) F. Montilla,
32 C. R. McNeill, S. Westenhoff, C. Groves, R. H. Friend and
N. C. Greenham, J. Phys. Chem. C, 2007, 111, 19153–19160.
33 R. A. Marsh, C. R. McNeill, A. Abrusci, A. R. Campbell and
R. H. Friend, Nano Lett., 2008, 8, 1393–1398.
862 | J. Mater. Chem., 2011, 21, 851–862
This journal is ª The Royal Society of Chemistry 2011