n-Type Semiconducting Naphthalene Diimide-Perylene Diimide Copolymers: Controlling Crystallinity, Blend Morphology, and Compatibility Toward High-Performance All-Polymer Solar Cells

Article abstract of DOI:10.1021/ja513260w

Knowledge of the critical factors that determine compatibility, blend morphology, and performance of bulk heterojunction (BHJ) solar cells composed of an electron-accepting polymer and an electron-donating polymer remains limited. To test the idea that bulk crystallinity is such a critical factor, we have designed a series of new semiconducting naphthalene diimide (NDI)-selenophene/perylene diimide (PDI)-selenophene random copolymers, xPDI (10PDI, 30PDI, 50PDI), whose crystallinity varies with composition, and investigated them as electron acceptors in BHJ solar cells. Pairing of the reference crystalline (crystalline domain size L_c = 10.22 nm) NDI-selenophene copolymer (PNDIS-HD) with crystalline (L_c = 9.15 nm) benzodithiophene-thieno[3,4-b]thiophene copolymer (PBDTTT-CT) donor yields incompatible blends, whose BHJ solar cells have a power conversion efficiency (PCE) of 1.4%. However, pairing of the new 30PDI with optimal crystallinity (L_c = 5.11 nm) as acceptor with the same PBDTTT-CT donor yields compatible blends and all-polymer solar cells with enhanced performance (PCE = 6.3%, J_sc = 18.6 mA/cm², external quantum efficiency = 91%). These photovoltaic parameters observed in 30PDI:PBDTTT-CT devices are the best so far for all-polymer solar cells, while the short-circuit current (J_sc) and external quantum efficiency are even higher than reported values for [70]-fullerene:PBDTTT-CT solar cells. The morphology and bulk carrier mobilities of the polymer/polymer blends varied substantially with crystallinity of the acceptor polymer component and thus with the NDI/PDI copolymer composition. These results demonstrate that the crystallinity of a polymer component and thus compatibility, blend morphology, and efficiency of polymer/polymer blend solar cells can be controlled by molecular design. (Figure Presented).

Full text of DOI:10.1021/ja513260w

Article
pubs.acs.org/JACS
n‑Type Semiconducting Naphthalene Diimide-Perylene Diimide
Copolymers: Controlling Crystallinity, Blend Morphology, and
Compatibility Toward High-Performance All-Polymer Solar Cells
Ye-Jin Hwang, Taeshik Earmme, Brett A. E. Courtright, Frank N. Eberle, and Samson A. Jenekhe*
Department of Chemical Engineering and Department of Chemistry, University of Washington, Seattle, Washington 98195-1750,
United States
S
* Supporting Information
ABSTRACT: Knowledge of the critical factors that determine
compatibility, blend morphology, and performance of bulk
heterojunction (BHJ) solar cells composed of an electron-
accepting polymer and an electron-donating polymer remains
limited. To test the idea that bulk crystallinity is such a critical
factor, we have designed a series of new semiconducting
naphthalene diimide (NDI)-selenophene/perylene diimide
(PDI)-selenophene random copolymers, xPDI (10PDI,
30PDI, 50PDI), whose crystallinity varies with composition,
and investigated them as electron acceptors in BHJ solar cells.
Pairing of the reference crystalline (crystalline domain size Lc
= 10.22 nm) NDI-selenophene copolymer (PNDIS-HD) with
crystalline (Lc = 9.15 nm) benzodithiophene-thieno[3,4-b]thiophene copolymer (PBDTTT-CT) donor yields incompatible
blends, whose BHJ solar cells have a power conversion efficiency (PCE) of 1.4%. However, pairing of the new 30PDI with
optimal crystallinity (Lc = 5.11 nm) as acceptor with the same PBDTTT-CT donor yields compatible blends and all-polymer
solar cells with enhanced performance (PCE = 6.3%, Jsc = 18.6 mA/cm2, external quantum efficiency = 91%). These photovoltaic
parameters observed in 30PDI:PBDTTT-CT devices are the best so far for all-polymer solar cells, while the short-circuit current
(Jsc) and external quantum efficiency are even higher than reported values for [70]-fullerene:PBDTTT-CT solar cells. The
morphology and bulk carrier mobilities of the polymer/polymer blends varied substantially with crystallinity of the acceptor
polymer component and thus with the NDI/PDI copolymer composition. These results demonstrate that the crystallinity of a
polymer component and thus compatibility, blend morphology, and efficiency of polymer/polymer blend solar cells can be
controlled by molecular design.
blend but rather denotes a blend with enhanced or useful new
properties compared to the components.9 Furthermore, unlike
universally applicable fullerene derivatives such as phenyl-C71-
butyric acid methyl ester (PC71BM) and phenyl-C61-butyric
acid methyl ester (PC61BM), current electron acceptor
polymers are rarely compatible with multiple donor polymers
in BHJ solar cells.3−6 An important exemption to the universal
compatibility of fullerene acceptors with multiple donor
polymers (i.e., gives rise to high photovoltaic efficiency) is
the indene-C60-bisadduct, 1′,1″,4′,4″-tetrahydro-di[1,4]-
methanonaphthaleno[1,2:2′,3′,56,60:2″,3″][5,6]fullerene-C60
(ICBA). ICBA is an amorphous material which is found to be
only compatible with a very few semicrystalline polymers,
notably poly(3-hexylthiophene) (P3HT)10 and poly[(4,4′-
bis(2-ethylhexyl)dithieno[3,2-b:2′,3′-d]silole)-2,6-diyl-alt-(2,5-
bis(3-(2-ethylhexyl)thiophen-2-yl)thiazolo[5,4-d]thiazole)]
(PSEHTT).11 Studies of amorphous ICBA and crystalline
fullerene acceptors (PC61BM and PC71BM) with various donor
INTRODUCTION
■
All-polymer solar cells,1 composed of an electron-donating
polymer and an electron-accepting polymer components, are of
growing interest because of their potential advantages over the
more widely studied donor polymer/fullerene acceptor
systems.1,2 Although significant progress has been made in
investigating and developing all-polymer solar cells, their
performance in terms of the power conversion efficiency
(PCE) of single-junction devices remains limited to about
5%,3−6 while that of polymer/fullerene systems is over 9%.7
One of the major challenges to further progress in developing
highly efficient all-polymer solar cells is that besides having a
suitable offset in the frontier molecular orbital (highest
occupied molecular orbital/lowest unoccupied molecular
orbital, HOMO/LUMO) energy levels of the polymer/polymer
donor/acceptor pair, the other key factors that govern
compatibility and blend morphology of the blend pair toward
achieving efficient bulk heterojunction (BHJ) solar cells are not
known or clear.3−6,8 The longstanding useage of the term
compatibility of a blend or compatible blend adapted here does
not imply a specific scale of miscibility or microstructure in the
Received: December 30, 2014
© XXXX American Chemical Society
A
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
polymers have led to the hypothesis that material crystallinity
controls polymer/fullerene compatibility and blend micro-
structure and thus charge photogeneration efficiency and
transport in such BHJ devices.12 There are as yet no analogous
or other studies of the role of crystallinity of the acceptor (or
donor) polymer component on the compatibility, blend
morphology, and performance of polymer/polymer BHJ solar
cells.
In this paper we show that an incompatible polymer/
polymer blend system with poor photovoltaic efficiency can be
rendered compatible with over 4-fold enhancement in PCE by
controlling the bulk crystallinity of the acceptor polymer while
holding the donor polymer constant. Our initial studies of the
reference all-polymer blend system, composed of a known
crystalline electron-accepting NDI-selenophene copolymer
PNDIS-HD6 and a known crystalline benzodithiophene-
thieno[3,4-b]thiophene copolymer (PBDTTT-CT)13 donor
(Chart 1) showed that the extensively optimized PNDIS-
RESULTS AND DISCUSSIONS
Synthesis, Absorption Spectra, and Electronic Struc-
■
ture. The new n-type semiconducting, random, NDI-
selenophene/PDI-selenophene copolymers denoted as xPDI,
where x is the molar percentage of PDI-selenophene segments,
were synthesized by Stille coupling copolymerization of three
monomers, 4,9-dibromo-2,7-bis(2-hexyldecyl)benzo[lmn][3,8]-
phenanthroline-1,3,6,8-tetraone (4,9-dibromo-NDI), N,N′-bis-
(2-ethylhexyl)-1,7-dibromo-3,4,9,10-perylene diimide (1,7-di-
bromo-PDI), and 2,5-bis(trimethylstannyl)-selenophene
(Scheme 1). The 4,9-dibromo-NDI and selenophene mono-
Scheme 1. Synthesis of n-Type Semiconducting NDI/PDI
Random Copolymers, xPDI
Chart 1. Molecular (a) and Electronic (b) Structure of
PNDIS-HD, xPDI, and PDBTTT-CT
mers were synthesized and purified by following our previous
reports,6 whereas the 1,7-dibromo-PDI monomer was obtained
commercially (SunaTech Inc.) and was used without further
purification. Three different compositions of xPDI, i.e. 10PDI,
30PDI, and 50PDI, were synthesized and investigated. In
addition, the benchmark NDI-selenophene copolymer, PNDIS-
HD,6 as well as PDI-selenophene copolymer was also
synthesized for the purpose of comparison. PNDIS-HD,
10PDI, 30PDI, and 50PDI showed good solubility (>30 mg/
mL) in common organic solvents, including chloroform,
chlorobenzene, and dichlorobenzene, at room temperature.
However, PPDIS precipitated out during the polymerization
and showed very limited solubility due to the short alkyl side
chains, 2-ethylhexyl, and thus PPDIS was not further studied.
The molecular structures of the monomers and copolymers
HD:PBDTTT-CT blend solar cells had PCEs of only 1.2−
1.4%, and thus it constitutes an incompatible blend despite the
very favorable HOMO/LUMO energy offset of the pair (Chart
1). To test the hypothesis that polymer crystallinity is a critical
factor that controls the compatibility, blend morphology, and
performance of polymer/polymer blend solar cells, we designed
a series of new n-type semiconducting NDI-selenophene/PDI-
selenophene random copolymers (Chart 1), xPDI (10PDI,
30PDI, 50PDI), in which the bulk crystallinity (e.g., mean
crystalline domain size Lc determined by X-ray diffraction
(XRD) analysis) decreased with increasing amount of PDI
moiety substituted for NDI moiety in the reference PNDIS-
HD. Pairing of the new 30PDI copolymer acceptor, which has a
substantially decreased and optimal crystallinity, with
PBDTTT-CT donor is found to give compatible blends and
all-polymer solar cells with enhanced performance (PCE =
6.3%, Jsc = 18.6 mA/cm2, external quantum efficiency (EQE) =
91%). Charge transport in neat films of the reference PNDIS-
HD and the new NDI/PDI copolymers (10PDI, 30PDI,
50PDI) was investigated by using organic field-effect transistors
(OFETs), and bulk charge transport in the polymer/polymer
BHJ active layers was studied by the space-charge-limited
current (SCLC) method. The surface and bulk morphologies of
the PNDIS-HD/PBDTTT-CT and xPDI/PBDTTT-CT blend
systems were investigated by atomic force microscopy (AFM)
and transmission electron microscopy (TEM) imaging,
respectively.
1
were confirmed by H NMR spectra (Figures S1−6). The
molecular weight and polydispersity index (Mw/Mn) of the
polymers were measured by gel permeation chromatography
(GPC) against polystyrene standards in o-dichlorobenzene at
130 °C. The reference polymer, PNDIS-HD, had a weight-
average molecular weight (Mw) of 42.6 kDa with Mw/Mn of 1.5.
The Mw of xPDI was comparable with PNDIS-HD sample and
was in the range of 32.1−57.2 kDa with a Mw/Mn of 1.3−1.4
(Table 1). Initially, GPC was performed in chlorobenzene at a
lower temperature (60 °C), and xPDI showed significantly
higher Mw in the range of 269.1−430.8 kDa with high Mw/Mn
in the range of 2.6−3.5, indicating aggregation of the samples
(Table S1). The p-type polymer PBDTTT-CT had Mn > 20
kDa with Mw/Mn ∼ 3. Thermogravimetric analysis (TGA) of all
the n-type polymers showed good thermal stability with onset
decomposition temperature (Td) of over 400 °C under
nitrogen flow (Figure S7).
The optical absorption of the new NDI/PDI copolymers,
xPDI, was characterized by UV−vis absorption spectroscopy of
dilute (10−6 M) chloroform solutions (Figure 1a) and of thin
films (95−110 nm) on glass substrates (Figure 1b). The optical
absorption spectra of the reference PNDIS-HD acceptor and
donor polymer, PBDTTT-CT, were also obtained for the
purpose of comparison. In solution, xPDI materials showed two
distinct absorption bands at 300−400 nm and 500−700 nm
B
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 1. Molecular Weight, Thin-Film Optical Absorption, Electronic Structure, and Field-Effect Electron Mobility of PNDIS-
HD and xPDIs
polymer
Mw (kDa)
Mw/ Mn
λmax (nm)
Eogpt (eV)
LUMO (eV)
μe (cm2/(V s))
Vth (V)
Ion/Ioff
PNDIS-HD
10PDI
42.6
57.2
52.5
32.1
1.5
1.3
1.4
1.4
348, 622
352, 615
349, 578
347, 544
1.65
1.70
1.77
1.77
3.84
3.84
3.89
3.89
9.2 × 10−2
5.5 × 10−2
7.1 × 10−3
4.0 × 10−3
10
10
5
103
103
103
103
30PDI
50PDI
5
consequence of the incorporation of PDI moieties into the n-
type semiconducting copolymers, xPDI, is that the absorption
coefficient is slightly lowered while the visible region absorption
maximum λmax progressively decreases from 615 nm (αmax = 2.6
× 104 cm−1) in 10PDI to 578 nm (αmax = 2.3 × 104 cm−1) in
30PDI and finally to 544 nm (αmax = 2.4 × 104 cm−1) in 50PDI.
This trend is largely explained by the progressive decrease in
planarity of the polymer backbone due to the increasing steric
hindrance from the large PDI moiety. We note that the
absorption maximum of 50PDI is already comparable to those
of known PDI-containing alternating conjugated copolymers.4
The measured α values for PNDIS-HD and xPDI are
comparable with other reported NDI and PDI-based acceptor
polymers.3,4 The optical band gap (Eogpt), determined from the
onset absorption band edge of the acceptor polymers, increased
from 1.65 eV in the reference PNDIS-HD to 1.70 eV in 10PDI
and to 1.77 eV in 30PDI and 50PDI. Our measured thin-film
absorption spectrum of PBDTTT-CT has an absorption
maximum at 706 nm and Egopt of 1.55 eV, which are in good
agreement with previous report of λmax of 692 nm and Eogpt of
1.58 eV.13 Although the absorption coefficient of PBDTTT-CT
is not previously reported, our measured αmax of 8.3 × 104 cm−1
in PBDTTT-CT is in good agreement with αmax values (7 ×
104 − 9 × 104 cm−1) of other benzodithiophene-thieno[3,4-
b]thiophene copolymers (PTB7 and PTB7-Th).3b,f
Figure 1. UV−vis absorption spectra of xPDI, PNDIS-HD, and
PBDTTT-CT in dilute chloroform solution (a) and as thin films on
glass substrates (b).
The electronic structures of the acceptor (xPDI, PNDIS-
HD) and donor (PBDTTT-CT) polymers were investigated by
using cyclic voltammetry (CV) of thin films. The ferrocene/
ferrocenium (Fc/Fc+) reference was used as an internal
standard, which was assigned an absolute energy of −4.8 eV
vs vacuum level.16 Based on the onset reduction potentials of
the polymers obtained from the cyclic voltammograms (Figure
which can be assigned to π−π* or n−π* transition of the NDI
and PDI chromophores and an intramolecular charge transfer
(ICT) band, respectively. These absorption features are
typically observed in NDI-based conjugated copolymers.6,14
10PDI has absorption maxima (λmax) at 570 nm (maximum
molar extinction coefficient, εmax = 1.48 × 104 M−1 cm−1) and
at 346 nm (εmax = 2.46 × 104 M−1 cm−1) which are very similar
to those of PNDIS-HD (λmax = 568 and 344 nm). By increasing
the amount of PDI moiety in the polymer backbone, however,
the λmax of the ICT band is blue-shifted to 555 nm (εmax = 1.59
× 104 M−1 cm−1) in 30PDI and to 539 nm (εmax = 1.25 × 104
M−1 cm−1) in 50PDI. The enhanced absorption near 500 nm in
30PDI and 50PDI originates from the PDI-selenophene moiety
and can be explained by the fact that PDI copolymers typically
have absorption maximum in the 500−560 nm range.4 Our
measured solution absorption spectrum of PBDTTT-CT has
two peaks at 645 and 693 nm (εmax = 4.1 × 104 M−1 cm−1) with
much higher molar extinction coefficient compared to the
acceptor polymers. However, we note that the molar extinction
coefficient of the new acceptor polymers, xPDI, is more than an
order of magnitude higher than fullerene acceptors in the
visible region.15
S8), the LUMO energy levels were determined (LUMO =
onset
red
−(eE
(V vs Fc/Fc+) + 4.8 eV)). In this way, we obtained
LUMO energy levels of −3.84 eV for PNDIS-HD and 10PDI
and −3.89 eV for 30PDI and 50PDI (Chart 1). The very similar
LUMO energy levels of these polymers is to be expected from
the similar electron-accepting strengths of the NDI and PDI
moieties. Oxidation wave was not observed up to 2 V (vs SCE)
in CV experiments on the acceptor polymers. Therefore, the
HOMO energy levels were estimated in two ways: (i) by
subtracting the Eogpt from the above CV-determined LUMO
energy levels and (ii) considering that the HOMO energy level
of −5.95 eV was experimentally observed for NDI-biseleno-
phene copolymer,14a the real HOMO energy levels of each
acceptor polymer (PNDIS-HD, 10PDI, 30PDI, 50PDI) should
thus be lower-lying than −5.95 eV, since selenophene is a
weaker electron-donating moiety than biselenophene. We
consider the HOMO energy levels (−5.95 eV) from the latter
estimate to be more reliable than those from using Eogpt. The
HOMO/LUMO energy levels of PBDTTT-CT were also
measured by using the onset oxidation/reduction potentials of
CV scans, giving −5.25/−3.19 eV. These HOMO/LUMO
energy levels of PBDTTT-CT provide sufficient energy offsets
The thin-film absorption spectra are very similar to the
solution spectra, but the longer wavelength (ICT) bands are
red-shifted due to the increased conjugation and intermolecular
interactions in the solid state. The reference PNDIS-HD has a
visible region λmax at 622 nm (αmax = 4.0 × 104 cm−1) which is
54 nm red-shifted compared to that in solution. An important
C
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(>0.3 eV) with each of the acceptor polymers (PNDIS-HD,
10PDI, 30PDI, 50PDI) for efficient electron transfer17 and hole
transfer18 essential for photovoltaic devices.
mixed in other conjugated polymer backbones.3d,20 The
progressive increase of the π−π stacking distance with
increasing amount of PDI moiety incorporated into the
copolymer chain can be understood as a consequence of lattice
distortion and decrease in polymer backbone planarity as PDI
replaces some NDI units.3d The average crystalline domain size
Lc decreased slightly from 10.22 nm in the reference PNDIS-
HD to 9.47 nm in 10PDI and much more dramatically to 5.11
and 3.62 nm in 30PDI and 50PDI, respectively (Figure 2b). We
conclude that random copolymerization has enabled the tuning
of the bulk crystallinity of NDI/PDI-based n-type semi-
conducting polymers.
Crystallinity of Neat Polymer Films. We used wide-angle
XRD analysis of thermally annealed (175 °C, 10 min) drop-
casted films on glass substrates to characterize the molecular
packing structure and bulk crystallinity of the new n-type
copolymers (10PDI, 30PDI, and 50PDI), the reference
acceptor polymer (PNDIS-HD), and the donor polymer
(PBDTTT-CT) as neat films. Such solution cast films have
randomly oriented crystallites which facilitate analysis and
interpretation of the XRD patterns shown in Figures 2a and S9.
Table 2. XRD Data and the Mean Crystalline Domain Size
(Lc) of Thermally Annealed Neat Polymers
sample
(100) (°) d100 (Å) (010) (°) d010 (Å) Lc (nm)
PBDTTT-CT
PNDIS-HD
10PDI
4.50
4.10
4.26
4.76
5.08
19.61
21.53
20.72
18.54
17.37
22.42
22.54
21.88
21.24
20.38
3.96
3.94
4.06
4.18
4.35
9.15
10.22
9.47
5.11
3.62
30PDI
50PDI
XRD patterns of the donor polymer PBDTTT-CT film
showed a lamellar (100) peak at 2θ = 4.50° and a π−π stacking
(010) peak at 2θ = 22.42° with corresponding d-spacings of
19.61 and 3.96 Å, respectively (Figure 2a). These results are in
good agreement with the previously reported XRD patterns for
this polymer.13,21 The average crystalline domain size Lc of
PBDTTT-CT was found to be 9.15 nm, which is comparable
with those of PNDIS-HD and 10PDI but is much larger
compared to the new random copolymers 30PDI and 50PDI.
We also performed XRD analysis of the film-aged (25 °C, 96
h) electron-accepting polymer (PNDIS-HD, xPDIs) films to
understand how film-forming conditions affect the bulk
crystallinity and molecular packing. The XRD patterns are
shown in Figure S10, and the data are summarized in Table
S2.The lamellar (100) peaks of PNDIS-HD and xPDIs were
found with 2θ of 4.48°, 4.88°, 5.89°, and 6.02° and d-spacing of
19.70, 18.09, 14.99, and 14.66 Å in PNDIS-HD, 10PDI, 30PDI,
and 50PDI films, respectively. The π−π stacking distance in
these room-temperature-aged neat films, d010, was 3.82, 3.84,
3.93, and 4.00 Å in PNDIS-HD, 10PDI, 30PDI, and 50PDI
films, respectively. The trend of progressive decrease of the
lamellar packing distance d100 with increasing amount of PDI
moiety is observed similar to what was seen in the thermally
annealed films. Also, similar to the observed trend in the
thermally annealed films, the π-stacking distance d010 increased
with increasing amount of the PDI moiety in the copolymer.
However, the d100 and d010 spacings in the room-temperature-
aged films are decreased by 9−19% and 3−8%, respectively,
compared to the corresponding thermally annealed films. The
average crystalline domain size Lc of these room temperature-
aged films decreased from 5.22 nm in PNDIS-HD and 3.42 nm
in 30PDI to 1.60 nm in 50PDI (Table S2). Thus, the observed
variation of bulk crystallinity with copolymer composition is
maintained under the different film-forming conditions.
Figure 2. (a) XRD patterns of solution-casted neat films of PNDIS-
HD, xPDIs, and PBDTTT-CT on glass substrates annealed at 175 °C
for 10 min. (b) Dependence of the mean crystalline domain size (Lc)
and π-stacking distance on random copolymer composition.
XRD patterns of the reference PNDIS-HD showed an intense
lamellar (100) peak at 2θ = 4.1° and a π−π stacking (010) peak
at 2θ = 22.54°, giving a lamellar packing distance of 21.53 Å
and a π−π stacking distance of 3.94 Å. The mean size of
crystalline domains (Lc) was also calculated from the Scherrer
equation19 by using the full-width-at-half-maximum (fwhm)
value of the lamellar (100) peak. The PNDIS-HD film has a
large Lc of 10.22 nm, implying a highly crystalline film. A similar
XRD analysis of the xPDI copolymers shows that the molecular
packing and crystallinity of drop cast neat films of 10PDI,
30PDI, and 50PDI are significantly different from those of the
reference PNDIS-HD. The lamellar (100) peaks of xPDIs are
found with 2θ of 4.26°, 4.76°, and 5.08° and d-spacings of 20.72
Å, 18.54 Å, and 17.37 Å in 10PDI, 30PDI, and 50PDI films,
respectively (Figure 2a). This represents a large progressive
decrease of the lamellar d-spacing with increasing amount of
the PDI moiety in the chain. In contrast, the π−π stacking
distance is increased linearly from 3.94 Å in PNDIS-HD to
4.06, 4.18, and 4.35 Å in 10PDI, 30PDI, and 50PDI films,
respectively (Figure 2b). The observed progressive decrease of
lamellar packing distance with increasing amount of PDI
moiety results from the shorter ethylhexyl side chains on the
PDI moieties and is in good agreement with similar
observations when side chains of different sizes are randomly
Field-Effect Electron Mobility. The intrinsic electron
mobility of the n-type semiconducting polymers was
investigated by fabricating and testing field-effect transistors
with bottom-gate and top-contact geometry.14 The electrical
parameters are summarized in Table 1, and the output and
transfer characteristics are presented in Figure S13. All the new
D
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
polymers (10PDI, 30PDI, and 50PDI) along with the reference
PNDIS-HD showed unipolar n-channel characteristics and no
p-channel charge transport as expected from their low-lying
LUMO and HOMO energy levels (Chart 1). PNDIS-HD neat
films showed the highest saturation region field-effect electron
mobility of 0.092 cm2/(V s) with a threshold voltage (Vth) of
10 V and on/off ratio of 103; this electron mobility is
comparable with the previously reported values for PNDIS-
HD6 and other NDI copolymers.14,22 The electron mobility
slightly decreased to 0.055 cm2/(V s) in 10PDI films, whereas it
decreased substantially to 7.1 × 10−3 cm2/(V s) and 4.0 × 10−3
cm2/(V s) for 30PDI and 50PDI films, respectively. The
observed trend of decreasing electron mobility from the highest
value in PNDIS-HD to the lowest value in 50PDI is in excellent
agreement with the XRD data which showed a progressive
lowering of the crystallinity and increasing of the π−π stacking
distance as NDI moieties are progressively substituted with PDI
moieties in the reference PNDIS-HD (Figure 2b). We also note
that the observed electron mobility of 4.0 × 10−3 − 0.055 cm2/
(V s) in neat films of the NDI/PDI random copolymers is
comparable to those of other acceptor polymers used in all-
polymer solar cells.3−5
All-Polymer BHJ Solar Cells. The photovoltaic properties
of all-polymer solar cells based on xPDI:PBDTTT-CT (1:1 w/
w) and PNDIS-HD:PBDTTT-CT (1:1 w/w) blend active
layers were investigated by fabricating and evaluating diodes
with the inverted structure (ITO/ZnO/polymer blend/MoO3/
Ag), and they were tested under 100 mW/cm2 AM 1.5G solar
illumination in ambient air. The polymer/polymer blend active
layer was prepared under optimized conditions of spin coating
from a chlorobenzene solution containing 3 vol % 1,8-
diiodooctane (DIO) and aging of the blend film at room-
tempaerature for 96 h to facilitate slow solvent evaporation and
self-organization in an argon-filled glovebox.
photovoltaic parameters, including the short-circuit current
density (Jsc), the open-circuit voltage (Voc), fill factor (FF), and
PCE, are summarized in Table 3.
The reference PNDIS-HD:PBDTTT-CT solar cells gave a
maximum PCE of 1.23% with Jsc of 5.39 mA/cm2, Voc of 0.69,
and FF of 0.35. This poor photovoltaic performance of the
PNDIS-HD/PBDTTT-CT blend system means that it is
incompatible. The incompatibility of this reference polymer/
polymer BHJ active layer is not obvious in light of the following
facts. The HOMO/LUMO energy level offsets between
PNDIS-HD acceptor and PBDTTT-CT donor (Chart 1) are
excellent for enabling efficient electron/hole transfer processes
and charge separation.17,18 PNDIS-HD is known to have a
relatively high electron mobility (Table 1) and prior reports
have showed that its blends with other donor polymers (e.g.,
PSEHTT) can create high efficiency all-polymer photovoltaic
devices (PCE = 3−5%).6 By virtue of its broad optical
absorption in the 500−800 nm region and small optical
bandgap (Eogpt = 1.55 eV), the PBDTTT-CT donor has
previously been used to produce high performance (PCE >
7%) polymer/[70]PCBM devices.13 We conclude that
incompatibility of PNDIS-HD/PBDTTT-CT blend system
implies that these factors are not sufficient to a priori predict
compatibility of a polymer/polymer blend pair for BHJ solar
cells.
Compared to the reference PNDIS-HD:PBDTTT-CT blend
system, all the new xPDI acceptors (10PDI, 30PDI, 50PDI),
respectively, paired with the same PBDTTT-CT donor gave
rise to enhanced photovoltaic performance of the BHJ devices.
In the case of 10PDI:PBDTTT-CT blends, the performance of
the BHJ solar cells (PCE = 2.8%, Jsc = 6.94 mA/cm2, Voc = 0.78
V, and FF = 0.51) is significantly improved even though the
change from PNDIS-HD to 10PDI is relatively minor. A
dramatic enhancement in photovoltaic performance is observed
in 30PDI:PBDTTT-CT blend solar cells with Voc of 0.77 V, FF
of 0.43, Jsc of 15.55 mA/cm2, and thus a high PCE of 5.10%.
This observed efficiency in 30PDI blends represents a 4.15-fold
enhancement of PCE compared to the reference PNDIS-HD
blend devices. It is also to be noted that the observed
photocurrent in 30PDI:PBDTTT-CT devices is comparable to
those seen in the best fullerene-based PC71BM:PBDTTT-CT
solar cells13 (15.5−17.5 mA/cm2). Although the performance
The current density−voltage (J−V) curves and the EQE
spectra of the devices are given in Figure 3a,b, respectively. The
of 50PDI:PBDTTT-CT BHJ solar cells (PCE = 2.66%, Jsc
=
9.68 mA/cm2, Voc = 0.73 V, and FF = 0.38) is still significantly
higher compared to the reference PNDIS-HD devices,
however, it is clear that further increase in the amount of
PDI moieties in the NDI/PDI copolymers results in decrease in
performance relative to the 30PDI acceptor. The dependence
of the PCE on the NDI/PDI copolymer composition is shown
in Figure 4, revealing that the 30 mol % PDI is the optimum.
The EQE spectra of the best xPDI:PBDTTT-CT and
PNDIS-HD:PBDTTT-CT blend devices are shown in Figure
3b. The photoresponse of each of the blend photodiodes starts
at 800 nm, which corresponds to the onset absorption of the
donor polymer, and covers the entire visible region down to the
UV at ∼300 nm. The EQE spectra peak at about 600−700 nm,
where both the p-type and n-type polymers strongly absorb.
The observed EQE spectra indicate that the n-type polymer in
each blend contributes to light harvesting and photocurrent
generation. In the case of the 30PDI:PBDTTT-CT blend
devices, the EQE peaks at 73%. We note that the photocurrent
density calculated from the EQE spectra (Figure 3b) are in
Figure 3. Current density−voltage (J−V) curves (a) and EQE spectra
(b) of PNDIS-HD:PBDTTT-CT (1:1 w/w) and xPDI:PBDTTT-CT
(1:1w/w) blend solar cells with film-aged (25 °C, 96 h) active layer.
E
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 3. Photovoltaic Properties of PNDIS-HD:PBDTTT-CT and xPDI:PBDTTT-CT Blend Solar Cells
a
blend
Jsc (mA/cm2)
Voc (V)
FF
PCEmax (%)
RS (Ω cm2) RSH (Ω cm2)
PNDIS-HD: PBDTTT-CT
10PDI: PBDTTT-CT
30PDI: PBDTTT-CT
50PDI: PBDTTT-CT
PNDIS-HD: PBDTTT-CT
5.39 (5.19 0.20)
6.94 (6.55 0.33)
15.55 (15.26 0.38)
9.68 (9.44 0.25)
4.16 (4.13 0.03)
18.55 (18.22 0.28)
0.69 (0.66 0.02)
0.78 (0.77 0.01)
0.77 (0.76 0.005)
0.73 (0.73 0.003)
0.81 (0.80 0.01)
0.35 (0.35 0.002)
0.51 (0.51 0.005)
0.43 (0.42 0.01)
0.38 (0.38 0.003)
0.40 (0.39 0.01)
0.45 (0.43 0.01)
1.23 (1.19 0.05)
2.80 (2.56 0.16)
5.10 (4.90 0.17)
2.66 (2.58 0.09)
1.36 (1.31 0.05)
6.29 (6.17 0.10)
85.0
19.7
14.0
21.9
18.2
11.3
290.7
616.1
219.2
199.2
211.5
160.6
b
b
30PDI: PBDTTT-CT
0.79 (0.78 0.002)
a
b
The photovoltaic properties were averaged over 10 devices. ZnO layer modified by a spin-coated polyethylenimine (PEI) interfacial layer.
cm2) is observed with Voc of 0.79 V and FF of 0.45, leading to a
PCE of 6.29%. The EQE spectrum of the 30PDI:PBDTTT-CT
photodiode (Figure 5b) has a maximum of 91% at 650 nm. The
calculated Jsc from the EQE spectrum is 17.47 mA/cm2, which
is in good agreement (5.8% mismatch) with that from the J−V
measurement (18.55 mA/cm2). The best performance of the
reference PNDIS-HD:PBDTTT-CT blend devices with PEI-
modified ZnO layer (PCE = 1.36%, Jsc = 4.16 mA/cm2, Voc
=
0.82 V, and FF = 0.40) is only slightly improved relative to
devices without the PEI interlayer. Thus, the photovoltaic
efficiency of the 30PDI blend devices is 4.6-fold enhanced
compared to the reference PNDIS-HD blend cells. We note
that the photocurrent (18.55 mA/cm2), PCE (6.29%), and
EQE (91%) seen in the present 30PDI:PBDTTT-CT BHJ
solar cells are the best observed to date in all-polymer solar
Figure 4. Dependence of PCE and the mean crystalline domain size
(Lc) on random copolymer composition. Half-filled purple circles are
the PCE of the devices with PEI interfacial layer modified ZnO layer.
cells.3−6 Furthermore, the short-circuit current density (Jsc
=
good agreement with the Jsc values measured directly from the
J−V curves in Figure 3a.
18.55 mA/cm2) and EQE (91%) are also the highest measured
for the donor polymer PBDTTT-CT, including
PC71BM:PBDTTT-CT devices.13
The 30PDI:PBDTTT-CT blend and the reference PNDIS-
HD:PBDTTT-CT blend devices were further optimized by
using a polyethylenimine (PEI) surface modifier on the top of
the zinc oxide (ZnO) layer which is known to lower the
cathode work-function and thus can provide better electron
injection and collection.23 The J−V curve and EQE spectrum of
the 30PDI:PBDTTT-CT blend devices are shown in Figure 5,
and a summary of the photovoltaic parameters is given in Table
3. A large increase in short-circuit current density (18.55 mA/
The observed high PCE (6.29%) of the 30PDI:PBDTTT-CT
BHJ solar cells implies that this polymer/polymer blend system
is definitely compatible, whereas the reference PNDIS-
HD:PBDTTT-CT blend is not. Given the far inferior
photovoltaic properties of the 10PDI:PBDTTT-CT and
50PDI:PBDTTT-CT blends, compared to the 30PDI blends,
we conclude that they are also incompatible like the reference
PNDIS-HD blends. The dependence of the crystallinity (i.e.,
average crystalline domain size Lc) on copolymer composition
shown in Figure 4b suggests that there is an optimum bulk
crystallinity of the acceptor polymer (Lc = 5.11 nm for 30PDI)
that facilitates its compatibility in blends with a given donor
polymer (PBDTTT-CT). The fact that the crystallinity in
50PDI (Lc = 3.62 nm), which pairs with PBDTTT-CT to
produce incompatible blends, is much smaller than in 30PDI,
which partners with the same donor polymer to produce
compatible blends, highlights our use of random copolymeriza-
tion as a facile means to discover the optimum bulk crystallinity
of the acceptor polymer essential to compatibility. Although
these results demonstrate that bulk crystallinity of the acceptor
polymer paired with a given donor polymer (PBDTTT-CT) is
a critical factor that dictates compatibility and photovoltaic
efficiency, the detailed mechanism of how an optimal
crystallinity endows blend compatibility and enhances perform-
ance of photovoltaic devices is not yet clear, and its elucidation
would require much future studies on various blend pairs.
Nevertheless, we believe that the optimal crystallinity exerts its
influence through the bulk morphology and bulk charge
transport of the polymer/polymer blend systems as will be
further discussed in subsequent sections.
Figure 5. Current density−voltage (J−V) curve (a) and EQE
spectrum (b) of 30PDI:PBDTTT-CT (1:1 w/w) blend solar cells
with film-aged (25 °C, 96 h) active layer and a PEI interfacial layer
modified ZnO layer.
Finally, we compare the observed best photovoltaic proper-
ties of 30PDI:PBDTTT-CT blends (PCE = 6.29%, Jsc = 18.55
mA/cm2, Voc = 0.79 V, and FF = 0.45) with those reported for
F
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
the best PC71BM:PBDTTT-CT blends.13 As already pointed
out above, the measured short-circuit current density in the
present all-polymer photodiodes is significantly higher than the
maximum Jsc (15.5−17.7 mA/cm2) observed in
PC71BM:PBDTTT-CT,13 while the Voc is also slightly higher.
However, much higher power conversion efficiencies (6.91−
7.59%) and FFs (58.7−59.5%) have been reported for the
PC71BM devices.13 The low FF (45%) is the main limitation of
the efficiency of 30PDI:PBDTTT-CT BHJ solar cells. Indeed, if
FFs comparable to the PC71BM device values (e.g., 59.5%)
could be obtained in the 30PDI:PBDTTT-CT blend devices,
PCEs exceeding 8% and thus comparable to the corresponding
fullerene-based solar cells would be achieved. The observed
much higher maximum EQE (91%) for 30PDI:PBDTTT-CT
blend devices compared to 66% for PC71BM:PBDTTT-CT13
suggests feasibility of such high PCEs if the low FF due to
recombination losses could be addressed by further device
optimization, including exploration of methods of facilitating
vertical phase segregation.24
Morphology of BHJ Solar Cells. The surface and bulk
morphologies of the xPDI:PBDTTT-CT (1:1 w/w) blend films
as well as the reference PNDIS-HD:PBDTTT-CT (1:1 w/w)
blend films prepared in identical ways as for the solar cells (film
aged at room temperature for 96 h in an argon-filled glovebox)
were investigated by AFM and bright-field transmission
electron microscopy (BF-TEM) imaging, respectively. AFM
height and phase images (1 μm × 1 μm) are shown in Figure 6.
The reference PNDIS-HD:PBDTTT-CT blend films showed
the largest phase-separated domains (∼200 nm) with a
roughness of 2.25 nm, indicating a strong tendency of the
two components to phase separate into large domains. A
dramatic change in the blend surface morphology is observed
by increasing the amorphous PDI-selenophene component and
thus decreasing the crystallinity of the acceptor polymer paired
with PBDTTT-CT. First, the surface roughness decreased from
2.25 nm in PNDIS-HD blends to 1.61 and 0.62 nm in 10PDI
and 30PDI blends, respectively. Second, the phase-separated
domain sizes observed in xPDI:PBDTTT-CT blends decreased
significanlty compared to the PNDIS-HD:PBDTTT-CT
blends. Although crystallinity (Lc = 9.47 nm) of 10PDI is
only slightly lower than that of PNDIS-HD (Lc = 10.22 nm),
the observed phase-separated domain sizes (∼100 nm) in the
10PDI:PBDTTT-CT blend morphology is decreased by half,
indicating that small differences in the crystallinity of the
component polymers can have a large impact on the surface
morphology of the blends. In the case of 30PDI:PBDTTT-CT
blends, a uniform phase-separated morphology with domains of
about 20 nm in size was observed, and domain size was
estimated from 200 × 200 nm2 phase images (Figure S11).
However, a clear phase-separated microstructure with distinct
domains was not observed in the 50PDI:PBDTTT-CT blends;
a surface roughness of 0.33 nm was determined from the AFM
height image of this blend. The phase-separated microstructure
with ∼20 nm distinct domain sizes observed in the surface
morphology of 30PDI:PBDTTT-CT blends (Figures 6c,g and
S11) is consistent with the observed high-performance
photovoltaic devices and our above earlier conclusion that
this blend system is compatible. We note that based on the
previously discussed photovoltaic properties, we concluded that
the 50PDI:PBDTTT-CT blend system is incompatible; the
AFM images of the surface morphology of this blend do not
reveal a phase-separated microstructure with distinct domains,
implying that its incompatibility may originate elsewhere.
Overall, these results of AFM imaging demonstrate that the
surface morphology of polymer/polymer blend solar cells is
controlled by the crystallinity of the components, which in turn
is controlled by the NDI/PDI random copolymer composition.
The BF-TEM images of the bulk morphologies of similarly
prepared polymer/polymer blends are shown in Figures 7 and
S12. The overall trend of phase-separated domains in the
blends decreasing in size progressively from PNDIS-HD to
50PDI blends is seen in the BF-TEM results similar to the
trend seen in the AFM images. In the PNDIS-HD:PBDTTT-
CT blend film, isolated domains with sizes of about 200 nm are
observed (Figure 7a). Such isolated domains could act as a
charge trap, impede charge transport, and act as recombination
centers and thus may explain the incompatibility and the poor
photovoltaic properties of PNDIS-HD:PBDTTT-CT blends. In
general, the xPDI:PBDTTT-CT blends showed much smaller
and better interconnected phases compared to the PNDIS-HD
blends. The 10PDI blend has significantly decreased phase-
separated domain sizes (∼100 nm), while the 30PDI blend has
phases separated on the scale of 20 nm. In the
50PDI:PBDTTT-CT blends, however, a clear contrast between
two phases is not observed (Figure 7d). The observed well
interconnected nanoscale domain in the bulk morphology of
30PDI:PBDTTT-CT blend as seen in the BF-TEM image
(Figure 7c) is consistent with compatibility of this blend system
and its excellent photovoltaic properties. From these results, we
conclude that the bulk morphology of polymer/polymer blend
solar cells can be controlled by tuning the crystallinity of one of
the polymer components.
Figure 6. AFM height (a−d) and phase (e−h) images (1 μm × 1 μm)
of PNDIS-HD:PBDTTT-CT (1:1 w/w) and xPDI:PBDTTT-CT (1:1
w/w) blend solar cells.
G
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 7. BF-TEM images of PNDIS-HD:PBDTTT-CT (1:1 w/w)
and xPDI:PBDTTT-CT (1:1 w/w) blend films.
Bulk Charge Transport in BHJ Blend Films. We
investigated the bulk charge carrier mobilities in the
polymer/polymer blend films, which were prepared similarly
to the photovoltaic devices, by using SCLC measurement. The
electron mobility was measured in an ITO/ZnO/active layer/
LiF/Al device structure, and the hole mobility was measured in
an ITO/PEDOT:PSS/active layer/Au device structure. The
current−voltage curves and SCLC fittings of the data are shown
in Figure 8 and the bulk carrier mobilities are summarized in
Tables S3 and S4.
Figure 8. Current−voltage curves and SCLC fittings of PNDIS-
HD:PBDTTT-CT and xPDI:PBDTTT-CT blend films. Hole-only
SCLC devices (a−d): ITO/PEDOT:PSS/blend/Au and electron-only
SCLC devices (e−h): ITO/ZnO/blend/LiF/Al.
The observed bulk hole and electron mobilities in the
polymer/polymer blends (Tables S3 and S4) were found to
vary substantially with crystallinity of the acceptor polymer
component and thus with the copolymer composition. The
reference PNDIS-HD:PBDTTT-CT blend systems have low
bulk hole and electron mobilities of 7.5 × 10−4 and 7.2 × 10−5
cm2/(V s), respectively, and this result can be understood from
the observed blend morphology in which large phase-separated
isolated domains are not interconnected. These low and
unbalanced bulk charge carrier mobilities (μh/μe = 2.6) of
PNDIS-HD:PBDTTT-CT blends can explain the poor
performance of the solar cells.
Compared to the reference PNDIS-HD blends, the
xPDI:PBDTTT-CT blends showed significantly enhanced
bulk charge carrier mobilities. The hole mobility of xPDI/
PBDTTT-CT blends varied from 6.0 × 10−4 cm2/(V s) in
50PDI to 1.7 × 10−3 cm2/(V s) in 10PDI and to 2.6 × 10−3
cm2/(V s) in 30PDI. The hole mobility in 30PDI blends is
enhanced by a factor of 3.3 compared to the hole mobility in
the reference PNDIS-HD blends. Clearly, the bulk morphology
in 30PDI:PBDTTT-CT is more favorable to hole transport in
this blend than in either PNDIS-HD or 10PDI or 50PDI blends
even though the donor polymer is identical in all the blends.
The bulk electron mobility varied from 1.8 × 10−4 cm2/(V s) in
50PDI blends to 8.5 × 10−4 cm2/(V s) in 10PDI blends and to
1.0 × 10−3 cm2/(V s) in 30PDI blends. It is to be noted that
electron mobility in 30PDI blends has increased by a factor of
13.8 compared to the electron mobility in the reference
PNDIS-HD blends. We point out that both electron and hole
mobilities are highest and the carrier asymmetry (μh/μe = 2.6)
lowest in the 30PDI:PBDTTT-CT blends, which are highly
beneficial to high photocurrent and conversion efficiency in the
BHJ solar cells. These results can be understood in terms of the
previously discussed bulk morphology as imaged by BF-TEM
(Figure 7), which revealed a phase-separated, interconnected
microstructure with small (∼20 nm) domains in
30PDI:PBDTTT-CT blends.
It is very instructive that the trends observed in the field-
effect electron mobility of neat films (Table 1) of the acceptor
polymers (PNDIS-HD, 10PDI, 30PDI, and 50PDI), in which
μe decreased monotonically from PNDIS-HD to 50PDI, are not
observed in the bulk electron transport in the blends. This
means that the electron mobility of neat films per se could not
be a useful guide in selecting components that would form
compatible blends in all-polymer solar cells. However, the
present results demonstrate that the bulk crystallinity,
quantified here in terms of Lc, could be used as an important
criterion in selecting compatible donor/acceptor pairs in
polymer/polymer blend solar cells.
CONCLUSIONS
■
Our study has investigated the problem of how to control the
crystallinity of the acceptor polymer component of polymer/
polymer blend solar cells and its impact on blend compatibility,
blend morphology, and performance of all-polymer solar cells.
H
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
mL) was refluxed under magnetic bar stirring for 72 h under argon and
cooled down to room temperature. The polymerization mixture was
poured into 200 mL methanol and 5 mL hydrochloric acid solution
and stirred overnight. The polymer precipitated out as a solid and was
filtered using a filter paper. The polymer was purified by Soxhlet
extraction with methanol, acetone, and hexane, sequentially. 10PDI
(260 mg; yield = 98%), GPC: Mw = 57.2 kDa, Mn = 44.0 kDa, Mw/Mn
= 1.3; TGA: Td = 400 °C.
30PDI: NDI (250 mg, 0.286 mmol), PDI (94.83 mg, 0.123 mmol),
and 2,5-bis(trimethylstannyl)selenophene (186.8 mg, 0.409 mmol)
with Pd2(dba)3 (7.5 mg, 2 mol %) and P(o-tolyl)3 (9.96 mg, 8 mol %)
in 16 mL of chlorobenzene. 30PDI (329 mg; yield =97.6%), GPC: Mw
= 52.5 kDa, Mn = 37.5 kDa, Mw/Mn = 1.4; TGA: Td = 420 °C.
50PDI: NDI (169.5 mg, 0.194 mmol), PDI (150 mg, 0.194 mmol),
and 2,5-bis(trimethylstannyl)selenophene (177.2 mg, 0.388 mmol)
with Pd2(dba)3 (7.1 mg, 2 mol %) and P(o-tolyl)3 (9.45 mg, 8 mol %)
in 15 mL of chlorobenzene. 50PDI (305 mg; yield =98.4%), GPC: Mw
= 32.1 kDa, Mn = 22.9 kDa, Mw/Mn = 1.4; TGA: Td = 415 °C.
Characterization. The structure and physical properties of the
new polymers were investigated by 1H NMR, GPC analysis and TGA.
1H NMR spectra at 300 MHz were recorded on a Bruker-AF300
spectrometer to verify the molecular structure, and the molecular
weight was measured using a Waters 1515 GPC with a refractive index
detector against polystyrene standards in o-dichlorobenzene at 130 °C.
TGA thermograms of the polymers were acquired on a TA
Instruments Q50 TGA at a heating rate of 20 °C per minute under
nitrogen gas flow.
Electrochemical properties of the polymers were investigated by
CV. CV experiments were done on an EG&G Princeton Applied
Research potentiostat/galvanostat (model 273A) in an electrolyte
solution of 0.1 M tetrabutylammonium hexafluorophosphate
(Bu4NPF6) in acetonitrile at a scan rate of 40 m/(V s). Platinum
wires were used as counter and working electrodes, and Ag/Ag+ (Ag in
0.1 M AgNO3 solution, Bioanalytical System, Inc.) was used as a
reference electrode. Ferrocene/ferrocenium was used as an internal
standard, and the reference potential was converted to the saturated
calomel electrode (SCE) scale. Each sample for CV was prepared by
dip-coating the copolymer solutions in chloroform onto Pt wires.
Optical absorption spectra of the polymers were measured on a
PerkinElmer model Lambda 900 UV−vis/near-IR spectrophotometer.
Solution- and solid-state absorption spectra were obtained from dilute
(10−6 M) polymer solutions in chloroform and as thin films on glass
substrates, respectively. Thin films (95−110 nm) were spin coated
from 20 mg/mL solutions in chlorobenzene.
We have found that the crystallinity (e.g., Lc) of new n-type
semiconducting NDI/PDI-selenophene random copolymers
xPDI (x = 10, 30, 50 mol % PDI), synthesized by Stille
copolymerization, varies with the copolymer composition,
which thus provides a synthetic means of controlling the
crystallinity of a polymer component of BHJ solar cells. Blends
of the reference crystalline (Lc = 10.22 nm) acceptor polymer
PNDIS-HD with crystalline (Lc = 9.47 nm) donor polymer
PBDTTT-CT were found to be incompatible with poor
photovoltaic properties (PCE = 1.4%). However, similar blends
of the new NDI/PDI copolymer (30PDI) acceptor with
optimal crystallinity (Lc = 5.11 nm) and the same PBDTTT-
CT were found to be compatible with substantially enhanced
photovoltaic properties (PCE = 6.3%, Jsc = 18.6 mA/cm2, and
EQE = 91%), which are the highest to date for all-polymer solar
cells. Indeed, the observed Jsc and EQE in 30PDI:PBDTTT-CT
blend solar cells are higher than reported values for
PC71BM:PBDTTT-CT devices.
AFM and BF-TEM imaging of the surface and bulk
morphologies of the various polymer/polymer blends found
that the blend microstructure varied substantially with the
crystallinity of the acceptor polymer component. Both surface
and bulk morphologies of the reference PNDIS-HD:PBDTTT-
CT blends revealed a phase-separated microstructure with large
(∼200 nm) isolated domains, whereas the 30PDI:PBDTTT-
CT blends have a phase-separated microstructure with small
(∼20 nm) interconnected domains. The bulk hole (μh) and
electron (μe) mobilities and carrier asymmetry (μh/μe) in the
polymer/polymer blends varied dramatically with crystallinity
of the acceptor polymer component, reaching their optimal
values at the optimum crystallinity found in 30PDI. The results
of this study demonstrate that the bulk crystallinity of a
polymer component in polymer/polymer blend solar cells is a
critical factor that determines blend compatibility, blend
morphology, and photovoltaic properties; furthermore, this
crystallinity can be controlled by molecular design. The bulk
crystallinity quantified here in terms of the average crystalline
domain size Lc is a material property that can be used as an
important criterion for selecting donor/acceptor pairs in
polymer/polymer blend solar cells.
XRD patterns were obtained from Bruker F8 power XRD with a Cu
Kα radiation as the X-ray source, and the solid samples were prepared
by drop-casting of highly concentrated polymer solutions (30 mg/mL)
in chloroform onto glass substrates and followed by annealing on a hot
plate at 175 °C for 10 min or film-aging in an argon filled glovebox at
room temperature for 96 h. The mean size of the crystalline domains
(Lc) of the polymers was calculated from the lamellar peaks using the
Scherrer equation, Lc = Kλ/β cos θ, where K is shape factor (0.9), λ is
X-ray wavelength (1.54 Å), and β is the fwhm in radian. The peak
center and the fwhm were obtained by fitting the lamellar peak using
Gaussian function in Origin software.
AFM and TEM Imaging. AFM characterization of surface
morphology was done on the active layers of the actual BHJ solar
cells by using a Veeco Dimension 3100 scanning probe microscope
(SPM) system. BF-TEM images were measured on an FEI Tecnai G2
F20 TEM at 200 kV accelerating voltage and acquired with a CCD
camera and recorded with Gatan Digital Micrograph software with
proper exposure time (0.1 s). The TEM images were slightly
defocused to enhance the phase contrast. The sample films were
spin-casted on top of ITO/PEDOT:PSS substrates and peeled off by
putting the samples in water. A peeled-off film was deposited on a
TEM grid (Electron Microscopy Sciences, Inc.) and dried overnight in
a vacuum oven.
EXPERIMENTAL SECTION
■
Materials. 4,9-Dibromo-2,7-bis(2-hexyldecyl)benzo[lmn][3,8]-
phenanthroline-1,3,6,8-tetraone (NDI) and 2,5-bis(trimethylstannyl)-
selenophene were synthesized according to the known literature
procedures.6 N,N′-bis(2-ethylhexyl)-1,7-dibromo-3,4,9,10-perylene dii-
mide (PDI) was purchased from Sunatech Inc., and the donor polymer
PBDTTT-CT (Mn > 20 kDa, Mw/Mn ∼ 3) was purchased from
Solarmer Energy, Inc., and both were used as received. The ZnO
precursor solution was prepared by dissolving 1 g of zinc acetate
dihydrate (99.999% trace metals basis, Aldrich) in 10 mL of 2-
methoxyethanol (99.8%, anhydrous, Aldrich) with 0.28 g of ethanol-
amine (≥99.5%, Aldrich) as a surfactant and stirring overnight under
ambient conditions.
Poly{([N,N′-bis(2-hexyldecyl)-naphthalene-1,4,5,8-bis-
(dicarboximide)-2,6-diyl]-alt-5,5′-selenophene)-ran-([N,N′-bis(2-eth-
ylhexyl)-1,7-dibromo-3,4,9,10-perylene diimide]-alt-5,5′-seleno-
phene))} (10PDI, 30PDI, and 50PDI). In the case of 10PDI, NDI
(250 mg, 0.286 mmol), PDI (22.42 mg, 0.0318 mmol), and 2,5-
bis(trimethylstannyl)selenophene (145.12 mg, 0.318 mmol) with
Pd2(dba)3 (5.82 mg, 2 mol %) and P(o-tolyl)3 (7.74 mg, 8 mol %)
were added into a 100 mL three-neck round-bottom flask. The flask
equipped with a condenser was then degassed and filled with argon
three times. Afterward, 12.6 mL of chlorobenzene was added and
degassed one time. The highly concentrated reaction mixture (34 mg/
Fabrication and Characterization of Field-Effect Transistors.
A heavily n-doped silicon substrate with a 200 nm thermally grown
silicon oxide layer acted as a gate electrode and insulating layer,
I
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
respectively (Ci = 17 nF/cm2). The substrate was cleaned by
ultrasonication in acetone and isopropyl alcohol for 30 min each
and dried under a flow of nitrogen gas. The substrate was further
cleaned using ozone-plasma for 5 min. A dilute chloroform solution
(0.1 M) of octyltrichlorosilane (OTS8) was spin-coated onto the
silicon dioxide layer at 3000 rpm in air for 10 s. The substrates were
then annealed for 10 min at 100 °C, washed with toluene, then
annealed for 10 min at 150 °C. The polymer semiconductors were
spin coated onto the OTS8 treated substrate from a solution in
chloroform (8 mg/mL) at 2000 rpm for 60 s in an argon-filled
glovebox. The films were annealed at 175 °C for 10 min, and then
source/drain electrodes were deposited via thermal evaporation of
silver (100 nm) that defined a channel width (W) of 1000 μm and
length (L) of 100 μm. Current−voltage characteristics of the
completed transistors were measured under nitrogen atmosphere
using a Signatone probe station and a semiconductor parameter
analyzer. The saturation region field-effect mobility (μ) and threshold
film-aged inside the glovebox at room temperature for 96 h (4 days)
followed by thermal vacuum deposition of LiF (1 nm)/Al (100 nm) or
Au (40 nm).
ASSOCIATED CONTENT
* Supporting Information
■
S
Synthesis of monomers and polymers, NMR spectra, AFM
images of surface morphology, output and transfer curves for
field-effect transistors, and SCLC device characteristics. This
material is available free of charge via the Internet at http://
pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
*jenekhe@u.washington.edu
■
1/2
voltage (Vth) were calculated from plots of Ids vs Vgs in a forward
Notes
scan with Vds at 80 V by using the saturation-region transistor
The authors declare no competing financial interest.
equation: Ids= (μWCi)(Vgs − Vth)2/(2L).
Fabrication and Characterization of All-Polymer Solar Cells.
ITO glass substrates were cleaned sequentially in ultrasonic baths with
acetone and isopropyl alcohol for 20 min, dried using nitrogen gas, and
stored in a vacuum oven. The ITO glass substrate was O2 plasma
treated for 90 s right before coating the ZnO layer. The ZnO precursor
solution was spin-coated onto the ITO glass at 5000 rpm for 40 s,
annealed at 250 °C on a hot plate in air for 1 h to make 20−30 nm
thick ZnO layer, and the glass/ITO/ZnO substrate was transferred
into an argon-filled glovebox. A 1.0 vol % ethanolamine or 0.05 wt %
PEI in 2-methoxyethanol solution was spin-coated onto the ZnO layer
and dried at 110 °C on a hot plate for 10 min right before use. Each
active layer (PNDIS-HD:PBDTTT-CT or xPDI:PBDTTT-CT blend
(1:1 w/w) solution (25 mg/mL) in chlorobenzene with 3 vol % DIO
solvent additive was spin-coated at 1000 rpm for 20 s. After spin-
coating, the wet film was film-aged inside the glovebox at room
temperature for 96 h (4 days) followed by thermal vacuum deposition
of MoO3 (7.5 nm) and Ag anode (100 nm). All the active layers have
thicknesses of 90 10 nm. Five pixels, each with an active area of 4
mm2, were fabricated per ITO substrate. The photovoltaic cells were
tested under AM 1.5G solar illumination at 100 mW/cm2 in ambient
condition using a Solar Simulator (model 16S, Solar Light Co.,
Philadelphia, PA) with a 200 W xenon lamp power supply (Model
XPS 200, Solar Light Co., Philadelphia, PA) calibrated by NREL
certified Si photodiode (Model 1787−04, Hamamatsu Photonics K.K.,
Japan) and a HP4155A semiconductor parameter analyzer (Yokogawa
Hewlett-Packard, Japan). After the J−V measurement, the EQE was
measured by using a solar cell quantum efficiency measurement system
(Model QEX10, PV Measurements, Inc., Boulder, CO) with a 2 mm2
(2 × 1 mm) size masked incident light source and TF Mini Super
measurement apparatus for multiple devices in a single substrate. The
EQE system was calibrated with a Si photodiode before each
measurement.
ACKNOWLEDGMENTS
■
This work was supported by the NSF (DMR-1409687) and in
part by the Office of Naval Research (ONR) (N00014−11−1−
0317).
REFERENCES
■
(1) (a) 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.
(b) Jenekhe, S. A.; Yi, S. J. Appl. Phys. Lett. 2000, 77, 2635. (c) Alam,
M. M.; Jenekhe, S. A. Chem. Mater. 2004, 16, 4647. (d) McNeill, C. R.
Energy Environ. Sci. 2012, 5, 5653.
(2) (a) Ahmed, E.; Ren, G.; Kim, F. S.; Hollenbeck, E. C.; Jenekhe, S.
A. Chem. Mater. 2011, 23, 4563. (b) Li, H.; Earmme, T.; Ren, G.;
Saeki, A.; Yoshikawa, S.; Murari, N. M.; Subramaniyan, S.; Crane, M.
J.; Seki, S.; Jenekhe, S. A. J. Am. Chem. Soc. 2014, 136, 14589. (c) Guo,
X.; Facchetti, A.; Marks, T. J. Chem. Rev. 2014, 114, 8943.
(3) (a) Schubert, M.; Dolfen, D.; Frisch, J.; Roland, S.; Steyrleuthner,
R.; Stiller, B.; Chen, Z.; Scherf, U.; Koch, N.; Facchetti, A.; Neher, D.
Adv. Energy Mater. 2012, 2, 369. (b) Zhou, E.; Cong, J.; Hashimoto,
K.; Tajima, K. Adv. Mater. 2013, 25, 6991. (c) Mori, D.; Benten, H.;
Okada, I.; Ohkita, H.; Ito, S. Energy Environ. Sci. 2014, 7, 2939.
(d) Hwang, Y. J.; Earmme, T.; Subramaniyan, S.; Jenekhe, S. A. Chem.
Commun. 2014, 50, 10801. (e) Mori, D.; Benten, H.; Okada, I.;
Ohkita, H.; Ito, S. Adv. Energy Mater. 2014, 4, 1301006. (f) Kang, H.;
Kim, K. H.; Choi, J.; Lee, C.; Kim, B. J. ACS Macro Lett. 2014, 3, 1009.
(g) Zhou, N.; Lin, H.; Lou, S. J.; Yu, X.; Guo, P.; Manley, E. F.; Loser,
S.; Hartnett, P.; Huang, H.; Wasielewski, M. R.; Chen, L. X.; Chang, R.
P. H.; Facchetti, A.; Marks, T. J. Adv. Energy Mater. 2014, 4, 1300785.
(h) Mu, C.; Liu, P.; Ma, W.; Jiang, K.; Zhao, J.; Zhang, K.; Chen, Z.;
Wei, Z.; Yi, Y.; Wang, J.; Yang, S.; Huang, F.; Facchetti, A.; Ade, H.;
Yan, H. Adv. Mater. 2014, 26, 7224.
(4) (a) Zhan, X.; Tan, Z.; Domercq, B.; An, Z.; Zhang, X.; Barlow, S.;
Li, Y.; Zhu, D.; Kippelen, B.; Marder, S. R. J. Am. Chem. Soc. 2007,
129, 7246. (b) Zhou, E.; Cong, J.; Wei, Q.; Tajima, K.; Yang, C.;
Hashimoto, K. Angew. Chem., Int. Ed. 2011, 50, 2799. (c) Zhou, Y.;
Yan, Q.; Zheng, Y. Q.; Wang, J. Y.; Zhao, D.; Pei, J. J. Mater. Chem. A
2013, 1, 6609. (d) Zhou, Y.; Kurosawa, T.; Ma, W.; Guo, Y.; Fang, L.;
Vandewal, K.; Diao, Y.; Wang, C.; Yan, Q.; Reinspach, J.; Mei, J.;
Appleton, A. L.; Koleilat, G. I.; Gao, Y.; Mannsfeld, S. C. B.; Salleo, A.;
Ade, H.; Zhao, D.; Bao, Z. Adv. Mater. 2014, 26, 3767. (e) Kozycz, L.
M.; Gao, D.; Tilley, A. J.; Seferos, D. S. J. Polym. Sci., Part A: Polym.
Chem. 2014, 52, 3337.
SCLC Measurement. Current−voltage (J−V) characteristics of the
SCLC devices were measured by using a HP4155A semiconductor
parameter analyzer (Yokogawa Hewlett-Packard, Tokyo). The carrier
mobility was extracted by fitting the J−V curves in the near quadratic
region according to the modified Mott−Gurney equation:25
⎛
⎞
⎟
9
8
V2
L3
V
L ⎠
J = εε0μ
exp 0.89β
⎜
⎝
where J is the current density, ε0 is the permittivity of free space, ε is
the relative permittivity, μ is the zero-field mobility, V is the applied
voltage, L is the thickness of active layer, and β is the field-activation
factor. The SCLC device structures for electron-only and hole-only
measurements were ITO/ZnO/active-layer/LiF/Al and ITO/PE-
DOT:PSS/active-layer/Au, respectively. Each active layer (PNDIS-
HD:PBDTTT-CT or xPDI:PBDTTT-CT blend (1:1 w/w)) solution
(25 mg/mL) in chlorobenzene with 3 vol % DIO solvent additive was
spin-coated at 1000 rpm for 20 s. After spin-coating, the wet film was
(5) (a) Kietzke, T.; Horhold, H. H.; Neher, D. Chem. Mater. 2005,
17, 6532. (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.
(c) Mori, D.; Benten, H.; Ohkita, H.; Ito, S.; Miyake, K. ACS Appl.
Mater. Interfaces 2012, 4, 3325. (d) Guo, C.; Lin, Y. H.; Witman, M.
D.; Smith, K. A.; Wang, C.; Hexemer, A.; Strzalka, J.; Gomez, E. D.;
J
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Verduzco, R. Nano Lett. 2013, 13, 2957. (e) Li, W.; Christian Roelofs,
W. S.; Turbiez, M.; Wienk, M. M.; Janssen, R. A. J. Adv. Mater. 2014,
26, 3304. (f) Jung, I. H.; Lo, W. Y.; Jang, J.; Chen, W.; Zhao, D.;
Landry, E. S.; Lu, L.; Talapin, D. V.; Yu, L. Chem. Mater. 2014, 26,
3450.
(6) (a) Earmme, T.; Hwang, Y. J.; Murari, N. M.; Subramaniyan, S.;
Jenekhe, S. A. J. Am. Chem. Soc. 2013, 135, 14960. (b) Earmme, T.;
Hwang, Y. J.; Subramaniyan, S.; Jenekhe, S. A. Adv. Mater. 2014, 26,
6080.
(7) (a) Small, C. E.; Chen, S.; Subbiah, J.; Amb, C. M.; Tsang, S. W.;
Lai, T. H.; Reynolds, J. R.; So, F. Nat. Photonics 2012, 6, 115. (b) He,
Z.; Zhong, C.; Su, S.; Wu, H.; Cao, Y. Nat. Photonics 2012, 6, 591.
(c) Liao, S. H.; Jhuo, H. J.; Cheng, Y. S.; Chen, S. A. Adv. Mater. 2013,
25, 4766. (d) Cabanetos, C.; El Labban, A.; Bartelt, J. A.; Douglas, J.
D.; Mateker, W. R.; Frechet, J. M. J.; McGehee, M. D.; Beaujuge, P. M.
J. Am. Chem. Soc. 2013, 135, 4656. (e) Chen, J. D.; Cui, C.; Li, Y. Q.;
Zhou, L.; Ou, Q. D.; Li, C.; Li, Y.; Tang, J. X. Adv. Mater. 2015, 27,
1132. (f) Xu, T.; Yu, L. Mater. Today 2014, 17, 11.
(8) (a) Anthony, J. E.; Facchetti, A.; Heeney, M.; Marder, S. R.;
Zhan, X. Adv. Mater. 2010, 22, 3876. (b) Facchetti, A. Mater. Today
2013, 16, 123.
(9) (a) Utracki, L. A. Polymer Alloys and Blends; Hanser Publishers:
New York, 1990. (b) Alam, M. M.; Tonzola, C. J.; Jenekhe, S. A.
Macromolecules 2003, 36, 6577.
(10) Zhao, G.; He, Y.; Li, Y. Adv. Mater. 2010, 22, 4355.
(11) (a) Subramaniyan, S.; Xin, H.; Kim, F. S.; Shoaee, S.; Durrant, J.
R.; Jenekhe, S. A. Adv. Energy Mater. 2011, 1, 854. (b) Xin, H.;
Subramaniyan, S.; Kwon, T.-W.; Shoaee, S.; Durrant, J. R.; Jenekhe, S.
A. Chem. Mater. 2012, 24, 1995.
(12) (a) Shoaee, S.; Subramaniyan, S.; Xin, H.; Keiderling, C.;
Tuladhar, P. S.; Jamieson, F.; Jenekhe, S. A.; Durrant, J. R. Adv. Funct.
Mater. 2013, 23, 3286. (b) Jamieson, F. C.; Domingo, E. B.;
McCarthy-Ward, T.; Heeney, M.; Stingelin, N.; Durrant, J. R. Chem.
Sci. 2012, 3, 485.
(13) (a) Huo, L.; Zhang, S.; Guo, X.; Xu, F.; Li, Y.; Hou, J. Angew.
Chem., Int. Ed. 2011, 50, 9697. (b) Fan, X.; Fang, G.; Cheng, F.; Qin,
P.; Huang, H.; Li, Y. J. Phys. D: Appl. Phys. 2013, 46, 305106.
(14) (a) Hwang, Y. J.; Ren, G.; Murari, N. M.; Jenekhe, S. A.
Macromolecules 2012, 45, 9056. (b) Hwang, Y. J.; Murari, N. M.;
Jenekhe, S. A. Polym. Chem. 2013, 4, 3187.
(15) Kronholm, D.; Hummelen, J. C. Mater. Matters 2007, 2, 16.
(16) Pavlishchuk, V. V.; Addison, A. W. Inorg. Chim. Acta 2000, 298,
97.
(17) Heeger, A. J. Adv. Mater. 2014, 26, 10.
(18) Ren, G.; Schlenker, C. W.; Ahmed, E.; Subramaniyan, S.; Olthof,
S.; Kahn, A.; Ginger, D. S.; Jenekhe, S. A. Adv. Funct. Mater. 2013, 23,
1238.
(19) Patterson, A. L. Phys. Rev. 1939, 56, 978.
(20) (a) Wu, P. T.; Ren, G.; Jenekhe, S. A. Macromolecules 2010, 43,
3306. (b) Burkhart, B.; Khlyabich, P. P.; Thompson, B. C. ACS Macro
Lett. 2012, 1, 660.
(21) Yan, H.; Zhu, L.; Li, D.; Zhang, Y.; Yi, Y.; Yang, Y.; Wei, Z.;
Bredas, J. L. J. Phys. Chem. C 2014, 118, 29473.
(22) Guo, X.; Kim, F. S.; Seger, M. J.; Jenekhe, S. A.; Watson, M. D.
Chem. Mater. 2012, 24, 1434.
(23) Zhou, Y.; Fuentes-Hernandez, C.; Shim, J.; Meyer, J.; Giordano,
A. J.; Li, H.; Winget, P.; Papadopoulos, T.; Cheun, H.; Kim, J.; Fenoll,
M.; Dindar, A.; Haske, W.; Najafabadi, E.; Khan, T. M.; Sojoudi, H.;
Barlow, S.; Graham, S.; Bredas, J. L.; Marder, S. R.; Kahn, A.; Kippelen,
B. Science 2012, 336, 327.
(24) (a) Kim, Y.; Cook, S.; Tuladhar, S. M.; Choulis, S. A.; Nelson, J.;
Durrant, J. R.; Bradley, D. D. C.; Giles, M.; McCulloch, I.; Ha, C. S.;
Ree, M. Nat. Mater. 2006, 5, 197. (b) Murari, N. M.; Crane, M. J.;
Earmme, T.; Hwang, Y. J.; Jenekhe, S. A. Appl. Phys. Lett. 2014, 104,
223906.
(25) Mott, N. F.; Gurney, R. W. Electronic Processes in Ionic Crystals;
Oxford University Press: London, 1940.
K
DOI: 10.1021/ja513260w
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX