The electron donor-acceptor (D-A) interface is an essential component for realizing efficient exciton dissociation and charge generation in organic photovoltaic cells (OPVs). It can also however enable rapid charge recombination due to the close spatial proximity of electrons and holes. To frustrate recombination losses, attempts have been made to separate charge carriers by introducing an insulating blocking interlayer at the D-A interface. It is challenging to realize increased efficiency using this approach as the relative similarity of interlayer optical and transport energy gaps may also frustrate exciton harvesting and charge generation. To overcome this trade-off, the interlayer must block charge carriers while continuing to permit exciton migration to the dissociating interface. In this work, we demonstrate this configuration in archetypical copper phthalocyanine (CuPc)-C60 planar OPVs containing a rubrene interlayer to frustrate charge recombination. Critically, the similarity in triplet exciton energy levels between rubrene and CuPc allows the interlayer to be permeable to excitons. Devices containing an interlayer show a reduction in the charge transfer state binding energy and non-geminate recombination rate with increasing interlayer thickness. For thin interlayers, geminate recombination is also suppressed. Thus, devices containing an exciton permeable interlayer show a simultaneous increase in open-circuit voltage, short-circuit current, and power efficiency.

With power conversion efficiencies (ηP) exceeding 14%, organic photovoltaic cells (OPVs) continue to demonstrate their potential as a low-cost photovoltaic technology.1,2 State-of-the-art devices are frequently limited by a low open-circuit voltage (VOC) relative to the optical gap of the active materials. To maximize VOC, a variety of studies have sought to better elucidate its physical origin.3–6 Previous work has demonstrated that VOC in a donor-acceptor (D-A) type OPV depends not only on the energy of the relaxed charge transfer (CT) state at the D-A interface, which sets the theoretical maximum for VOC, but also on the recombination of free charge carriers (non-geminate recombination).7,8 In many systems, the current density due to non-geminate recombination (JNGR) is often the primary factor limiting VOC and hence must be suppressed to reduce energy losses.7,9

Molecular design is frequently used to reduce JNGR and improve VOC, as an increased interface energy level offset (EDA) between the highest occupied molecular orbital (HOMO) of the donor and the lowest unoccupied molecular orbital (LUMO) of the acceptor increases the effective energy gap via the CT state energy (ECT) and reduces the dark current [Fig. 1(a)].10–13 Despite these efforts, there remains room for further improvement, even for energetically optimized D-A systems. ECT is generally lower than EDA due to the Coulombic binding energy of relaxed CT states.14 Consequently, efforts have also exploited variations in device architecture to engineer ECT and reduce JNGR. For example, an increase in the CT state separation has been shown to reduce the binding energy and non-geminate recombination.3,15,16 Inserting a wide energy gap interlayer (i.e., deeper HOMO than the donor and shallower LUMO than the acceptor) between the donor and acceptor materials can spatially separate the electron and hole that comprise the CT state, leading to an increase in the maximum achievable VOC.3,15,16 Unfortunately, increasing the thickness of the insulating interlayer [Fig. 1(b)] can also lead to a drastic reduction in short-circuit current density (JSC) by reducing the efficiency of exciton dissociation by charge transfer.3,16

FIG. 1.

Energy diagram for donor-acceptor heterojunction OPVs with (a) an archetypical planar heterojunction, (b) a wide energy gap interlayer between the donor and acceptor materials which increases CT state energy while also frustrating exciton diffusion to the dissociating interface, and (c) a triplet exciton permeable wide energy gap interlayer that allows exciton dissociation at the interlayer-acceptor interface. (d) Device architecture of interest with a rubrene interlayer between the CuPc donor and the C60 acceptor. An exciton blocking layer of bathocuproine (BCP) is placed between C60 and the Al cathode. The CuPc-rubrene-C60 planar cell corresponds to the situation described in (c).

FIG. 1.

Energy diagram for donor-acceptor heterojunction OPVs with (a) an archetypical planar heterojunction, (b) a wide energy gap interlayer between the donor and acceptor materials which increases CT state energy while also frustrating exciton diffusion to the dissociating interface, and (c) a triplet exciton permeable wide energy gap interlayer that allows exciton dissociation at the interlayer-acceptor interface. (d) Device architecture of interest with a rubrene interlayer between the CuPc donor and the C60 acceptor. An exciton blocking layer of bathocuproine (BCP) is placed between C60 and the Al cathode. The CuPc-rubrene-C60 planar cell corresponds to the situation described in (c).

Close modal

To simultaneously increase the spatial extent of the CT state and maintain efficient exciton harvesting, we demonstrate a device architecture using an interlayer with a HOMO level between those of the donor and the acceptor and a LUMO level that is shallower than that of both the donor and the acceptor [Fig. 1(c)]. Despite the similarity of this architecture to energy cascade OPVs, the photoconversion process is different. Even with an interlayer, the current device has only a single dissociating interface (between the interlayer and the acceptor), contrasting energy cascade OPVs that have multiple dissociating interfaces.15 As such, only photogenerated holes are transported within the interlayer. To avoid significant exciton loss from the donor layer, the interlayer must be permeable to donor excitons. Figure 1(d) shows the device architecture and molecular structures of the electron donor and interlayer used in this work. Here, triplet exciton forming metal-phthalocyanines are used as donors with a rubrene interlayer. For the archetypical donor copper phthalocyanine (CuPc), the triplet yield under optical excitation is near unity due to its rapid intersystem crossing.17 The triplet level (T1) of rubrene (1.14 eV) is similar to CuPc (1.16 eV), permitting exciton transport and dissociation.18,19 Here, we elucidate the impact of the interlayer on the interfacial morphology, device performance, CT state binding energy, and recombination losses. Detailed experimental procedures are described in the supplementary material.

Figure 2(a) shows the operating parameters for CuPc-rubrene-C60 planar devices in Fig. 1(d) as a function of interlayer thickness (0–3 nm) under simulated AM1.5G (100 mW/cm2) illumination. VOC increases with the interlayer thickness while JSC shows an improvement of 20% for a nominally 1-nm-thick layer of rubrene before rolling-off with the increased thickness. This increase in VOC is similar to the results of previous work employing a wide energy gap material as a blocking interlayer for CuPc-C60 planar devices.3 In this prior work [case in Fig. 1(b)], a steep reduction in JSC was observed with no initial improvement. In contrast, here the JSC, fill factor (FF), and resulting ηP are maximized when a 1-nm-thick rubrene interlayer is included at the interface. Optimized rubrene interlayer devices show a 50% enhancement in ηP.

FIG. 2.

(a) Open-circuit voltage (VOC), short-circuit current density (JSC), fill factor (FF), and power conversion efficiency (ηP) versus rubrene thickness for the device in Fig. 1(d). (b) Voltage dependence (open-circuit) of the excess carrier lifetime (τΔn) as a function of interlayer thickness.

FIG. 2.

(a) Open-circuit voltage (VOC), short-circuit current density (JSC), fill factor (FF), and power conversion efficiency (ηP) versus rubrene thickness for the device in Fig. 1(d). (b) Voltage dependence (open-circuit) of the excess carrier lifetime (τΔn) as a function of interlayer thickness.

Close modal

The role of the interlayer in suppressing non-geminate recombination was examined by probing the charge carrier lifetime using transient photovoltage (TPV).20 In this measurement, devices are held at open-circuit under constant background illumination (blue light-emitting diode (LED), λpeak = 455 nm) and then excited with an additional short light pulse (20 μs, green LED, λpeak = 530 nm, intensity: 16.5 mW cm−2) to induce excess charge carriers.20,21 Since at open-circuit recombination within the device is the only pathway to consume charge carriers, the TPV decay back to steady-state reflects the excess charge carrier lifetime. Figure 2(b) shows the small perturbation carrier lifetime (τΔn) as a function of steady-state voltage and interlayer thickness for planar devices on pre-patterned indium-tin-oxide (ITO)-coated glass substrates (voltage perturbation ∼5 mV). For the same voltage, the measured τΔn increases with the interlayer thickness. Recent work has suggested that τΔn measured by TPV might reflect both the transit of charge carriers from the electrodes and eventual recombination at the D-A interface.22 Here, since only the D-A interface is modified, charge carrier transit from the electrodes should be unaffected. Thus, the observed increase in τΔn confirms that non-geminate recombination can be slowed by spatially separating oppositely charged photogenerated carriers.

Additional insight into the role of the interlayer can be realized by considering the dark current density (JDark)-voltage characteristics for the devices in Fig. 1(d) (Fig. S3). JDark decreases with the interlayer thickness for the “diode” dominant region and the effective interlayer coverage can be estimated to be 70% (1 nm), 89% (2 nm), and 96% (3 nm) from this region (Fig. S4). To identify the origin of enhanced VOC in interlayer devices, JDark in Fig. S3 is fit with a Shockley model (see Sec. 3 of the supplementary material).23 Increasing the thickness leads to increased shunt resistance and a reduction in the saturation current density (Table S1). These reflect a reduction in current leakage between the electrodes and reduced non-geminate recombination at the D-A interface, respectively. This dark current analysis is consistent with the carrier lifetime results of Fig. 2(b).

To understand the impact of the rubrene interlayer on the CT state binding energy, ECT was extracted as a function of interlayer thickness via the temperature dependence of VOC as follows:3 

qVOCECTnkBTlnJ00JGenVOC.
(1)

Here, q is the fundamental charge, n is the diode ideality factor, kB is the Boltzmann constant, T is the temperature, JGen is the generation flux of free charge carriers, namely the sum of the current density under illumination and JNGR,20 and J00 is a temperature independent prefactor which determines the reverse saturation current density J0 [Eq. (S1)]. In practice, the observed temperature-dependence of VOC often deviates from a linear trend and plateaus at low temperature due to the limited quasi-Fermi level offset at finite illumination intensity.6,24 Thus, the theoretical maximum achievable VOC and ECT are often approximated as the extrapolated VOC at 0 K (known as V0).24,25 Figure S6 shows the measured VOC as a function of temperature (190–294 K) for CuPc-C60 planar devices as a function of rubrene interlayer thickness. A linear dependence of VOC on temperature is observed for all devices. The extracted value of V0 (≈ ECT/q) increases with the rubrene interlayer thickness (from 0.79 to 0.84, 0.87, and 0.90 V). As the V0 of a standalone rubrene-C60 bilayer device (1.19 V) is larger than that of a CuPc-C60 bilayer device (0.80 V), one possible explanation for the increased value of ECT is the presence of a different relaxed CT state formed at the interlayer-acceptor interface (Fig. S7). To examine this further, the measured ECT is fit with a simple electrostatic model assuming that the relaxed CT state continues to reside on molecules of CuPc and C60 with ECT depending on the interlayer thickness (d) as follows:

ECT=EDAEba0a0+dεiεawithEb=q24πε0εaa0.
(2)

Here, Eb and a0 are the binding energy and the separation distance of the CT state in the absence of an interlayer, ε0 is the permittivity of vacuum, and εa and εi are the relative permittivity of the active material and interlayer, respectively.3 To reduce fitting variables, the values of εa and εi are estimated from the near-infrared refractive indices of the D-A (1:1) mixture (εa = 3.8) and rubrene (εi = 2.9). The value of EDA (1.00 eV) is taken from the literature, which is measured directly from CuPc/C60 thin films deposited on ITO using photoelectron spectroscopy.26 The measured ECT in Fig. S6 can be well fit with this simple model [Fig. 3(a)]. An initial separation distance of a0 = (1.9 ± 0.3) nm is obtained, corresponding to a binding energy Eb = (0.20 ± 0.03) eV, similar to previous results (error bars are determined from the 95% confidence interval of the fit).3 This suggests that the relaxed CT state resides on molecules of CuPc and C60 up to an interlayer thickness of 3 nm. Thus, the binding energy of the CuPc-C60 CT state can be reduced by spatially separating the electron and hole, leading to an improved ECT and larger VOC under practical operating conditions.

FIG. 3.

(a) Modeling the increased ECT of a CuPc-C60 CT state with the incorporation of a rubrene interlayer by fitting the interlayer thickness dependence of ECT. (b) Normalized external quantum efficiency (ηEQE) of metal phthalocyanine (PbPc, CuPc, PtPc)-C60 planar OPVs (structure: 15 nm donor/x nm rubrene/35 nm C60/10 nm BCP/100 nm Al) as a function of interlayer thickness. The ηEQE is normalized to 500 nm (C60 absorption dominant) to isolate the impact of the interlayer on exciton harvesting. The donor component is shown as solid lines. (c) Donor diffusion efficiency (normalized to bilayer case) as a function of donor triplet energy and interlayer thickness.

FIG. 3.

(a) Modeling the increased ECT of a CuPc-C60 CT state with the incorporation of a rubrene interlayer by fitting the interlayer thickness dependence of ECT. (b) Normalized external quantum efficiency (ηEQE) of metal phthalocyanine (PbPc, CuPc, PtPc)-C60 planar OPVs (structure: 15 nm donor/x nm rubrene/35 nm C60/10 nm BCP/100 nm Al) as a function of interlayer thickness. The ηEQE is normalized to 500 nm (C60 absorption dominant) to isolate the impact of the interlayer on exciton harvesting. The donor component is shown as solid lines. (c) Donor diffusion efficiency (normalized to bilayer case) as a function of donor triplet energy and interlayer thickness.

Close modal

In order for the interlayer-based device in Fig. 1(d) to exhibit high JSC, the rubrene interlayer must transport donor excitons to the dissociating interlayer-acceptor interface. To verify the transport of donor excitons via the triplet level of the interlayer, the relative change in the exciton diffusion efficiency (ηDiff) is investigated as a function of rubrene thickness in three metal phthalocyanine (MPc)-C60 systems. Devices based on donor layers of platinum phthalocyanine (PtPc) and lead phthalocyanine (PbPc), with triplet energies of 1.28 eV and 1.02 eV, respectively, are compared with those based on CuPc (1.16 eV).19,27,28 As the triplet excitons in all three metal phthalocyanine (MPc) donors cannot be dissociated at the MPc-rubrene interface (Fig. S8), a more exothermic exciton transfer from the donor to interlayer will frustrate back transfer and improve exciton harvesting. Figure 3(b) shows the external quantum efficiency (ηEQE) normalized to its value at a wavelength of λ = 500 nm (C60 absorption dominant) for MPc-rubrene-C60 planar devices. For all MPcs, the donor contribution to the ηEQE falls with the interlayer thickness. The ηEQE can be thought of as the product ηEQE = ηA ηDiff ηCC, with ηA, ηDiff, and ηCC defined as the absorption, exciton diffusion, and charge collection efficiencies, respectively. Here, the exciton dissociation efficiency at the D-A interface is taken as unity, while ηCC, the collection yield of CT states, is considered as a wavelength independent property.29 As exciton diffusion in C60 is unaffected by the presence of the interlayer, the normalization of ηEQE removes any impact of the interlayer on ηCC. Thus, the normalized ηEQE is equivalent to the product ηAηDiff. Optical transfer matrix simulations of ηA show no change for either active material with the presence of the interlayer (Fig. S9). The observed reduction in normalized ηEQE for the donor is therefore attributed to a decrease in ηDiff. To better compare the reduction in ηDiff induced by the interlayer in MPc devices, we plot the normalized ηDiff (100% for bilayer case) as a function of triplet energy and interlayer thickness for the MPc absorption region [Fig. 3(c)]. The normalized donor ηDiff is observed to increase with donor material triplet energy. With a 3-nm-thick rubrene layer, the normalized ηDiff for PtPc devices is 1.5 and 3 times higher than that for CuPc and PbPc, respectively, likely reflecting reduced back transfer to the MPc layer. The exciton permeable interlayer thus provides a promising strategy to improve VOC while maintaining efficient exciton harvesting.

Since the inclusion of the interlayer does not improve ηDiff and ηA, an increase in ηCC must be responsible for the initial increase in JSC. Figure 4(a) shows the ηEQE spectra for the device in Fig. 1(d) and a rubrene-C60 bilayer device for comparison. Since the simulated ηA and ηDiff of C60 are almost unchanged with rubrene thickness (Fig. S10), the relative increase in ηCC can be readily isolated from the C60 component of the ηEQE.29 For C60 absorption at λ ∼ 450 nm, there is an increase of ∼50% in ηEQE when a 1-nm-thick layer of rubrene is inserted at the D-A interface. Relative to the 1 nm case, the ηEQE at this wavelength decreases with the increasing interlayer thickness, while remaining larger than the case of control CuPc-C60 bilayer devices. This suggests that the interlayer suppresses charge recombination at short-circuit, leading to improved ηCC. As the dependence of JSC on light intensity is linear for all of the devices studied (Fig. S11), geminate recombination (first-order) serves as the dominant recombination mechanism at short-circuit.30 Accordingly, the CT state separation efficiency (ηCS), the dissociation yield of relaxed CT states, can be approximated as ηCC.

FIG. 4.

(a) ηEQE for the devices in Fig. 1(d) and a rubrene-C60 bilayer OPV (structure: 15 nm rubrene/35 nm C60/10 nm BCP/100 nm Al). (b) Charge separation efficiency (ηCS) of CuPc devices in (a) as a function of rubrene (RUB) interlayer thickness. The green dashed line is the ηCS of the rubrene bilayer device in (a). The exciton diffusion length (LD) of C60 is taken as 30 nm to extract ηCS. The ηCS of CuPc devices extracted using C60LD = 25 and 35 nm is also shown as the red dashed line and red dotted line, respectively.

FIG. 4.

(a) ηEQE for the devices in Fig. 1(d) and a rubrene-C60 bilayer OPV (structure: 15 nm rubrene/35 nm C60/10 nm BCP/100 nm Al). (b) Charge separation efficiency (ηCS) of CuPc devices in (a) as a function of rubrene (RUB) interlayer thickness. The green dashed line is the ηCS of the rubrene bilayer device in (a). The exciton diffusion length (LD) of C60 is taken as 30 nm to extract ηCS. The ηCS of CuPc devices extracted using C60LD = 25 and 35 nm is also shown as the red dashed line and red dotted line, respectively.

Close modal

To better understand the increased efficiency of CT state separation in interlayer devices, the ηCS (≈ ηCC) is extracted as a function of interlayer thickness by taking the ratio of ηEQE spectra to ηAηDiff products (LD = 30 nm for C60) [Fig. 4(b)].31,32 Due to the uncertainty of the LD in C60, we also plot the results using LD = 25 and 35 nm for C60. To confirm that the ηDiff of C60 can be accurately simulated with the assumption of the flat D-A interface, the surface topography of rubrene-covered CuPc is measured by atomic force microscopy (Fig. S12), and the root-mean square (RMS) roughness of the D-A interface is nearly unchanged with the inclusion of a thin rubrene layer. As such, the increased C60ηEQE is from improved ηCS rather than a significant variation in the interface morphology. As shown in Fig. 4(b), the ηCS of CuPc devices with an interlayer is in all cases larger than that of rubrene-C60 bilayer OPVs, reaching 100% for 1-nm-thick interlayer when taking LD = 25 nm for C60. This suggests that the observed enhancement in ηCS is not simply due to the formation of a more easily dissociated rubrene-C60 CT state. Instead, CT state separation here is likely a two-step process, with the rubrene-C60 CT state first relaxing to the CuPc-C60 CT state, followed by the dissociation of the CuPc-C60 CT state into free carriers. The roll-off in ηCS above an interlayer thickness of 1 nm suggests that a potential transport barrier must be overcome for the transition from the rubrene-C60 to CuPc-C60 CT state. Previously, Groves predicted a similar roll-off in ηCS when increasing the interlayer thickness in an energy cascade heterojunction using a Marcus model and a kinetic Monte Carlo simulation, in agreement with our observations.33 With the increasing rubrene thickness, relaxed CT states are more likely to be rubrene-C60 CT states, with the resulting devices resembling a rubrene-C60 OPV. For a thin interlayer, the two-step exothermic charge transfer leads to larger CT states frustrating geminate recombination.

While in this work the exciton permeable interlayer is applied with a triplet exciton-forming host, we envision broader applications for this concept in singlet fission devices or potentially in ternary bulk heterojunctions (BHJs). Despite a reduction in recombination losses, the realization of high efficiency in triplet-based OPVs will require the detrimental energy loss incurred during singlet-triplet conversion to be overcome. This can be realized by reducing the singlet-triplet energy difference through molecular design or by generating multiple low energy triplet excitons from one high energy singlet exciton through singlet fission. The former permits the realization of a higher VOC, while the latter seeks to offset the reduction in VOC with a larger photocurrent. In singlet fission materials, OPVs often utilize a planar device architecture as singlet fission is an intermolecular process.34 Typically, singlet fission materials are paired with small triplet energy materials for long wavelength absorption.34,35 For instance, Jadhav et al. have used CuPc to accept and transport the triplet excitons of tetracene to a dissociating CuPc-C60 interface.36 The use of an exciton permeable interlayer would allow for improved VOC and efficiency in these systems. An area of potential analogy is in morphology-controlled ternary bulk heterojunction (BHJ) OPVs.37 Formation of the third BHJ component located at the D-A interface of the other two components has been previously reported and remains an active research area in OPVs.38,39 Our interlayer work may provide additional insight into the operation of these devices as the third component may act as a discontinuous interlayer to optimize recombination.

In summary, we present a device architecture that exploits a triplet exciton permeable interlayer at the D-A interface in an OPV to spatially separate oppositely charged carriers while maintaining efficient exciton harvesting. By comparing three MPc donors with varying triplet levels, we find that a larger donor triplet level facilitates exciton migration through the interlayer. For CuPc-C60 planar devices, the interlayer is found to reduce the saturation current density and increase the τΔn and ECT, leading to suppressed non-geminate recombination and an increased VOC. The thin interlayer also creates an energy level cascade for hole transfer that results in the formation of relaxed CT states with a reduced binding energy. This suppresses geminate recombination and leads to an increase in short-circuit current density. Overall, a 50% increase in efficiency is realized in optimized interlayer devices. The demonstrated exciton permeable interlayer design could have further application in singlet fission devices and ternary mixture systems when the morphology of active materials can be accurately controlled.

See supplementary material for detailed experimental methods and discussions for charge extraction measurements, interlayer coverage, dark current density fitting, CT state energy measurement, metal phthalocyanine-C60 OPVs, optical simulations, intensity dependence of short-circuit current density, and morphology of the donor-acceptor interface.

This work was supported by the National Science Foundation (NSF) Electronics, Photonics and Magnetic Devices under ECCS-1509121 and NSF Solid State and Materials Chemistry under DMR-1708177. The authors acknowledge the research group of Professor C. D. Frisbie for use of their atomic force microscope.

1.
X.
Che
,
Y.
Li
,
Y.
Qu
, and
S. R.
Forrest
,
Nat. Energy
3
,
422
(
2018
).
2.
S.
Zhang
,
Y.
Qin
,
J.
Zhu
, and
J.
Hou
,
Adv. Mater.
30
,
1800868
(
2018
).
3.
Y.
Zou
and
R. J.
Holmes
,
ACS Appl. Mater. Interfaces
7
,
18306
(
2015
).
4.
M. D.
Perez
,
C.
Borek
,
S. R.
Forrest
, and
M. E.
Thompson
,
J. Am. Chem. Soc.
131
,
9281
(
2009
).
5.
K.
Vandewal
,
K.
Tvingstedt
,
A.
Gadisa
,
O.
Inganas
, and
J. V.
Manca
,
Nat. Mater.
8
,
904
(
2009
).
6.
B. P.
Rand
,
D. P.
Burk
, and
S. R.
Forrest
,
Phys. Rev. B
75
,
115327
(
2007
).
7.
D.
Credgington
,
R.
Hamilton
,
P.
Atienzar
,
J.
Nelson
, and
J. R.
Durrant
,
Adv. Funct. Mater.
21
,
2744
(
2011
).
8.
T. M.
Burke
,
S.
Sweetnam
,
K.
Vandewal
, and
M. D.
McGehee
,
Adv. Energy Mater.
5
,
1500123
(
2015
).
9.
A.
Maurano
,
R.
Hamilton
,
C. G.
Shuttle
,
A. M.
Ballantyne
,
J.
Nelson
,
B.
O'Regan
,
W.
Zhang
,
I.
McCulloch
,
H.
Azimi
,
M.
Morana
,
C. J.
Brabec
, and
J. R.
Durrant
,
Adv. Mater.
22
,
4987
(
2010
).
10.
K. L.
Mutolo
,
E. I.
Mayo
,
B. P.
Rand
,
S. R.
Forrest
, and
M. E.
Thompson
,
J. Am. Chem. Soc.
128
,
8108
(
2006
).
11.
Y.
He
,
H.-Y.
Chen
,
J.
Hou
, and
Y.
Li
,
J. Am. Chem. Soc.
132
,
1377
(
2010
).
12.
I. H.
Campbell
and
B. K.
Crone
,
Appl. Phys. Lett.
101
,
023301
(
2012
).
13.
A.
Kumar
,
G.
Pace
,
A. A.
Bakulin
,
J.
Fang
,
P. K. H.
Ho
,
W. T. S.
Huck
,
R. H.
Friend
, and
N. C.
Greenham
,
Energy Environ. Sci.
6
,
1589
(
2013
).
14.
B.
Kippelen
and
J.-L.
Brédas
,
Energy Environ. Sci.
2
,
251
(
2009
).
15.
T. D.
Heidel
,
D.
Hochbaum
,
J. M.
Sussman
,
V.
Singh
,
M. E.
Bahlke
,
I.
Hiromi
,
J.
Lee
, and
M. A.
Baldo
,
J. Appl. Phys.
109
,
104502
(
2011
).
16.
Y.
Zhong
,
A.
Tada
,
S.
Izawa
,
K.
Hashimoto
, and
K.
Tajima
,
Adv. Energy Mater.
4
,
1301332
(
2014
).
17.
B. W.
Caplins
,
T. K.
Mullenbach
,
R. J.
Holmes
, and
D. A.
Blank
,
Phys. Chem. Chem. Phys.
18
,
11454
(
2016
).
18.
19.
P. S.
Vincett
,
J. Chem. Phys.
55
,
4131
(
1971
).
20.
T.
Zhang
and
R. J.
Holmes
,
J. Mater. Chem. C
5
,
11885
(
2017
).
21.
T. K.
Mullenbach
,
I. J.
Curtin
,
T.
Zhang
, and
R. J.
Holmes
,
Nat. Commun.
8
,
14215
(
2017
).
22.
D.
Kiermasch
,
A.
Baumann
,
M.
Fischer
,
V.
Dyakonov
, and
K.
Tvingstedt
,
Energy Environ. Sci.
11
,
629
(
2018
).
23.
R. H.
Bube
and
A. L.
Fahrenbruch
,
Adv. Electron. Phys.
56
,
163
(
1981
).
24.
J.
Widmer
,
M.
Tietze
,
K.
Leo
, and
M.
Riede
,
Adv. Funct. Mater.
23
,
5814
(
2013
).
25.
S. R.
Cowan
,
A.
Roy
, and
A. J.
Heeger
,
Phys. Rev. B
82
,
245207
(
2010
).
26.
S.
Cho
,
L.
Piper
,
A.
DeMasi
,
A.
Preston
,
K.
Smith
,
K.
Chauhan
,
P.
Sullivan
,
R.
Hatton
, and
T. S.
Jones
,
J. Phys. Chem. C
114
,
1928
(
2010
).
27.
K.
Kaneto
,
K.
Yoshino
, and
Y.
Inuishi
,
J. Phys. Soc. Jpn.
37
,
1297
(
1974
).
28.
N.
Minami
,
J. Chem. Soc., Faraday Trans. 2
78
,
1871
(
1982
).
29.
R.
Pandey
and
R. J.
Holmes
,
Appl. Phys. Lett.
100
,
083303
(
2012
).
30.
C. M.
Proctor
,
M.
Kuik
, and
T.-Q.
Nguyen
,
Prog. Polym. Sci.
38
,
1941
(
2013
).
31.
L. A. A.
Pettersson
,
L. S.
Roman
, and
O.
Inganas
,
J. Appl. Phys.
86
,
487
(
1999
).
32.
D.
Qin
,
P.
Gu
,
R. S.
Dhar
,
S. G.
Razavipour
, and
D.
Ban
,
Phys. Status Solidi A
208
,
1967
(
2011
).
33.
C.
Groves
,
Energy Environ. Sci.
6
,
1546
(
2013
).
34.
J.
Xia
,
S. N.
Sanders
,
W.
Cheng
,
J. Z.
Low
,
J.
Liu
,
L. M.
Campos
, and
T.
Sun
,
Adv. Mater.
29
,
1601652
(
2017
).
35.
B.
Ehrler
,
M. W.
Wilson
,
A.
Rao
,
R. H.
Friend
, and
N. C.
Greenham
,
Nano Lett.
12
,
1053
(
2012
).
36.
P. J.
Jadhav
,
A.
Mohanty
,
J.
Sussman
,
J.
Lee
, and
M. A.
Baldo
,
Nano Lett.
11
,
1495
(
2011
).
37.
Q.
An
,
F.
Zhang
,
J.
Zhang
,
W.
Tang
,
Z.
Deng
, and
B.
Hu
,
Energy Environ. Sci.
9
,
281
(
2016
).
38.
J.-S.
Huang
,
T.
Goh
,
X.
Li
,
M. Y.
Sfeir
,
E. A.
Bielinski
,
S.
Tomasulo
,
M. L.
Lee
,
N.
Hazari
, and
A. D.
Taylor
,
Nat. Photonics
7
,
479
(
2013
).
39.
S.
Honda
,
H.
Ohkita
,
H.
Benten
, and
S.
Ito
,
Adv. Energy Mater.
1
,
588
(
2011
).

Supplementary Material