www.advmat.de www.MaterialsViews.com
COMMUNICATION
Dominant Effects of First Monolayer Energetics at Donor/ Acceptor Interfaces on Organic Photovoltaics Seiichiro Izawa, Kyohei Nakano, Kaori Suzuki, Kazuhito Hashimoto, and Keisuke Tajima* Even after 20 years of research, it is still unknown how the charges in organic photovoltaics (OPVs) are separated and collected so efficiently.[1] Charge separation occurs via charge transfer of photogenerated excitons across the donor/acceptor (D/A) interface to form electron/hole pairs. However, Coulomb binding between the charge pair is strong compared with the thermal energy and is effective over a long range due to the low dielectric constants of organic semiconductors. Therefore, the bound charge transfer state (CTS) can be formed at the D/A interface before the free charges form. The CTS can decay to the ground state with a certain probability (geminate recombination), resulting in loss of the photocurrent. The CTS is also formed when free charge pairs encounter each other at the D/A interface and decay to the ground state via non-geminate (bimolecular) recombination, the kinetics of which are strongly related to the fill factor (FF) and open-circuit voltage (VOC).[2] Recently, it was proposed that the higher energy hot (delocalized) CTS may be important for efficient charge separation.[3] Ultrafast spectroscopy studies show that delocalization of the CT state over several nanometers on aggregated fullerene molecules[4] and the excess energy assists free charge formation.[5] In contrast, the excitation of the relaxed CTS can be efficiently separated into free charges without excess energy.[6] It was proposed that a gradient forms in the free-energy landscape near the D/A interface because the disorder of the materials widens the energy gap.[7] The unintentional “cascade” energy landscape at the interface could push the energy of the CTS close to the charge separated state, resulting in highly efficient charge separation.[8] Groves used a kinetic Monte Carlo simulation to predict that thin cascade layers (1 nm) suppressed geminate recombination, whereas thicker cascade layers did not.[9] Although the mechanism for the efficient formation of free charge is still unknown,[10] the aggregation structures or the gradient of the free-energy landscape close to the D/A interface could be important for obtaining efficient OPVs through efficient Dr. S. Izawa, Dr. K. Nakano, K. Suzuki, Dr. K. Tajima RIKEN Center for Emergent Matter Science (CEMS) 2-1 Hirosawa, Wako Saitama 351-0198, Japan E-mail: [email protected] Dr. S. Izawa, Prof. K. Hashimoto Department of Applied Chemistry Graduate School of Engineering The University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan Dr. K. Tajima Precursory Research for Embryonic Science and Technology (PRESTO) Japan Science and Technology Agency 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan
DOI: 10.1002/adma.201500840
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
charge separation and suppressing recombination. The effects of the energy cascade at the D/A interface have been explored with model bilayer systems. However, because the materials were either spin coated successively or evaporated, the interfacial layers were thick and the interfacial structures were mixed on a molecular scale.[11] Thus, the effects of the energy cascade close to the interface at the molecular monolayer level remain unclear. In this work, we combine our contact film transfer (CFT) method[12] with a self-organized surface segregated monolayer (SSM)[13] to fabricate bilayer OPVs with well-defined D/A interfaces.[14–16] To tune the energetics at the surface precisely, a new surface modifier based on fluoroalkyl bis(1-[3-(methoxycarbonyl) propyl]-1-phenyl)-[6.6]C62 (F-bisPCBM) (Figure 1a) was synthesized. Fullerene bisadducts generally have lower electron acceptability (higher-lying lowest unoccupied molecular orbital (LUMO)) than monoadducts because the π-conjugation is broken.[17] We combine two surface modifiers with different LUMO energy levels (F-PCBM and F-bisPCBM) with two fullerene acceptors (PCBM and bisPCBM), and construct four energy landscapes at the D/A interface in the bilayer OPVs (Figure 1b). We investigate the effect of the energy levels of the first monolayer at the D/A interface on the photoelectric conversion. F-PCBM and F-bisPCBM were synthesized via an acid-catalyzed ester exchange reaction[18,19] as described in the Supporting Information. We compared the molecular electronic properties of PCBM, bisPCBM, F-PCBM, and F-bisPCBM by cyclic voltammetry (CV) in solution. The reduction potentials of the fullerene bisadducts (bisPCBM and F-bisPCBM, −1.20 V vs Fc/Fc+) have a reduction potential that is 0.12 V lower than those of the monoadducts (PCBM and F-PCBM, −1.08 V vs Fc/Fc+) (Figure S3, Supporting Information). Therefore, the fullerene bisadducts have higher-lying LUMO energy levels compared with the monoadducts.[17] There is no difference between the fluoroalkyl and methyl ester derivatives, indicating that the fluoroalkyl chain attached to the ester group does not affect the energy level of the fullerene.[13] The identical absorption spectra of the fluoroalkyl and methyl ester derivatives in solution also show that the fluoroalkyl chains have little effect on the molecular orbitals of the fullerene group (Figure S4, Supporting Information). Surface segregation behaviors of F-PCBM and F-bisPCBM in the blended films were investigated by X-ray photoelectron spectroscopy (XPS). The films prepared from different combinations of the surface modifiers and bulk materials (F-PCBM/PCBM, F-bisPCBM/PCBM, F-PCBM/bisPCBM, and F-bisPCBM/bisPCBM) were prepared by spin coating solutions containing different concentrations of F-PCBM and F-bisPCBM (0.2–1.2 g L−1) with a fixed concentration of PCBM and bisPCBM (10 g L−1). The surface F/C atomic ratio was
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
1
www.advmat.de
COMMUNICATION
www.MaterialsViews.com
Figure 1. a) Chemical structures of the surface modifiers and the bulk materials, and schematic images of the surface modification by SSM and the fabrication of bilayer devices by CFT. b) Schematic images of the four different energy landscapes at the D/A interface in the bilayer OPVs.
calculated from the F 1s and non-fluorinated C 1s XPS peak intensities. Figure S5 in the Supporting Information shows the F/C atomic ratio as a function of the concentration of F-PCBM and F-bisPCBM in the solutions. The results are similar to those of our previous studies on SSM formation in F-PCBM/ PCBM; the F/C ratios measured by XPS were higher than those calculated for homogeneously mixed films for all films, indicating the surface segregation of F-PCBM and F-bisPCBM.[19] The clear saturation of F/C ratios was observed at high concentrations of the surface modifiers in all the combinations. The F/C atomic ratios at the saturated conditions were 0.128, 0.111, 0.111, and 0.102 in F-PCBM/PCBM, F-bisPCBM/PCBM, F-PCBM/bisPCBM, and F-bisPCBM/bisPCBM, respectively. The F/C atomic ratios reflect the density of the segregated fluoroalkyl chains at the surface. As previously reported, the surface of F-PCBM/PCBM was almost fully covered by SSM according to the maximum F/C ratios calculated for full coverage conditions.[19] The slightly lower densities in the films of the bisadduct fullerene derivatives compared with that of F-PCBM/PCBM could be attributed to larger steric hindrance and thus a lower packing density of the fullerene bisadducts. The concentration of the surface modifiers close to the saturation point (0.56 × 10−3 M (0.75 g L−1) for F-PCBM and 0.56 × 10−3 M (0.86 g L−1) for F-bisPCBM) were used to investigate the film properties and for device fabrication. XPS depth profiles of the films showed that the F 1s peak of all the films disappeared after surface etching to a depth of about 1.5 nm with an Ar+ ion beam (Figure S6, Supporting Information). This suggests that the fluoroalkyl chains of F-PCBM and F-bisPCBM segregated to the surface and a small amount remained in the bulk of the films.
2
wileyonlinelibrary.com
To elucidate the surface structure further, we also performed angle-resolved XPS. The peak ratios of fluorinated to non-fluorinated carbon were used for measurements. A bilayer model, in which the surface layers consist of fluoroalkyl chains, fitted the experimental data well (details provided in the Supporting Information). This indicated that F-PCBM or F-bisPCBM formed a monolayer at the surface of both PCBM and bisPCBM films with fluoroalkyl chains exposed to the surface. The ionization potentials (IPs) of the films were measured by UV photoelectron spectroscopy (UPS; Figure S8, Supporting Information and Table 1). There were small differences of 0.04 eV in IPs between F-PCBM/PCBM and F-bisPCBM/PCBM and of 0.03 eV between F-PCBM/bisPCBM and F-bisPCBM/ bisPCBM. The oriented fluoroalkyl chain at the surface should form a surface dipole moment; therefore, the IPs of the films should be shifted compared with the pristine films depending on the surface density of the fluoroalkyl chains.[14] The small difference in the IPs between the surface modifiers (F-PCBM Table 1. IPs of the films measured by UPS and reduction potentials (half potentials) versus Fc/Fc+ of PCBM and bisPCBM determined by CV in CH2Cl2.
PCBM
IP [eV]
Ered [V]
5.81
−1.08
F-PCBM/PCBM
5.41
F-bisPCBM/PCBM
5.45
bisPCBM
5.79
F-PCBM/bisPCBM
5.51
F-bisPCBM/bisPCBM
5.54
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
−1.20
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
www.advmat.de www.MaterialsViews.com
COMMUNICATION
Figure 2. J–V curves of the four bilayer OPVs with a,b) P3HT and c,d) PTB7 as the donor layer under AM 1.5, 100 mW cm−2 irradiation.
and F-bisPCBM) could be caused by the difference in surface density of the fluoroalkyl chains suggested by XPS. Figure 1b shows a schematic of the energy levels of the four kinds of acceptor films deduced from the CV and UPS results. The energy levels between the surface monolayer and the bulk are the same in the F-PCBM/PCBM and F-bisPCBM/bisPCBM films. A LUMO energy trap with a depth of 0.12 eV formed at the surface monolayer in the F-PCBM/bisPCBM film, whereas a LUMO energy cascade of 0.12 eV formed at the surface monolayer in the F-bisPCBM/PCBM film. The advantage of this model is that besides the surface energy levels, other properties can be tuned to the same conditions; the optical absorption of the films was identical (Figure S9, Supporting Information), the film thicknesses were similar (about 35 nm), and the planarity of the film surfaces was similar with Ra of 0.16–0.26 nm (Figure S10, Supporting Information). The surface dipole moments caused by the fluoroalkyl chains exist in all four films. UPS showed that the vacuum level shift for different surface modifiers was similar, with differences of 0.03–0.04 eV. This enabled us to compare the device performance of the bilayer OPVs directly. Four bilayer devices with different energy levels at the D/A interface were fabricated via CFT of the same donor polymer layer onto the four fullerene layers (Figure 1a). The structures of the devices were indium tin oxide (ITO)/TiO2/acceptor/ SSM//donor/MoO3/Ag, where // denotes the interface created by film transfer. It has been shown that the D/A interfaces of the bilayer devices fabricated by CFT are flat and sharp and thus the surface structure of films before the transfer are retained.[14] As the result, the surface energy level differences of the fullerene films are preserved and the four bilayer devices with different energy landscapes at the D/A interface (donor//
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
F-PCBM/PCBM: normal (PCBM); donor//F-bisPCBM/PCBM: cascade; donor//F-bisPCBM/bisPCBM: normal (bisPCBM); donor//F-PCBM/bisPCBM: trap) were fabricated (Figure 1b). The J–V characteristics of the four devices with P3HT and PTB7 as the donor layer under simulated solar light irradiation are shown in Figure 2 and the results are summarized in Table S1, Supporting Information. The cascade device with a P3HT donor layer exhibited VOC and FF values that were larger than the normal (PCBM) device, whereas the JSC values were similar. The changes in VOC and FF were +0.09 V and +0.13, respectively. In contrast, the trap device with P3HT as the donor layer exhibited much lower VOC, FF, and JSC values than the normal (bisPCBM) device. The changes in VOC and FF were −0.12 V and −0.18, respectively. The devices with PTB7 as the donor layer exhibited similar behavior (Figure 2c,d): the VOC of the cascade (trap) device was higher (lower) by +0.11 V (−0.15 V) and the FF of the cascade (trap) device was higher (lower) by +0.20 (−0.21). However, the JSC value of the cascade device was smaller than that of the normal devices. The differences in device performance between the four devices can be attributed to the difference in the energy landscape at the D/A interface because the bulk properties, such as the absorption and charge mobility, were the same. In particular, the large difference in FF suggested that the charge recombination was suppressed in the cascade configuration, whereas it was promoted in the trap configuration. The s-shape of J–V curve in Figure 2d could be caused by severe charge recombination at D/A interface in the trap configuration. The molecular dipole moment of the fluoroalkyl chains in SSM layers caused VOC change as previously reported;[14] PTB7//F-PCBM/PCBM and PTB7//F-bisPCBM/ bisPCBM showed lower VOC by 0.18 and 0.06 V, respectively, compared to those of PTB7//PCBM and PTB7//bisPCBM,
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
3
www.advmat.de
COMMUNICATION
www.MaterialsViews.com
respectively (Table S1, Supporting Information). However, the large changes in VOC of the cascade and trap devices were not due to the difference in the dipole moments but the different energy landscapes at the D/A interface because all the four films have the interfacial fluoroalkyl chains with the similar densities and the vacuum level shifts measured by UPS for the different surface modifiers were also similar. The changes in device performance may be caused by the presence of a small amount of surface modifiers below the XPS detection limit in the bulk. To eliminate this possibility, small amounts of PCBM and bisPCBM were intentionally added to the bisPCBM and PCBM layers, respectively, in the bilayer devices containing PTB7 (details provided in the Supporting Information). Figure S11 in the Supporting Information shows that the addition of a small amount of the different acceptors in the bulk of the films induced no change in the performance. Therefore, the energy level difference at the D/A interface, not in the bulk, caused the changes. To explain the changes in JSC, the external quantum efficiency (EQE) spectra of four devices with P3HT and PTB7 as the donor layer are shown in Figure 3. The EQE peak at 600 nm was from the absorption of P3HT in Figure 3a,b, the peak at 650 nm was from the absorption of PTB7 in Figure 3c, d, and the peaks at 350 and 450 nm had a large contribution from PCBM absorption. This was confirmed by optical simulation and EQE calculations for devices, based on the transfer matrix model and an exciton diffusion model (Figure S13, Supporting Information). Figure 3a shows that the cascade device had a larger EQE in the donor absorption region, and a smaller EQE in the acceptor absorption region compared with the normal (PCBM) device. Therefore, the JSC in the cascade device was similar to that of the normal device. In PTB7 devices, the EQE in the donor
absorption region was unchanged and was smaller in the acceptor absorption region for the cascade device, resulting in the smaller JSC under solar irradiation (Figure 3c). The decrease in EQE in the fullerene absorption region may be because of the larger bandgap of F-bisPCBM at the D/A interface than that of PCBM in the bulk (Figure 1b); therefore, the exciton in the bulk PCBM could not reach the F-bisPCBM layer at the D/A interface in the cascade devices since the exciton migration process is energetically unfavorable by 0.12 eV. This can be avoided by creating a cascade for the highest occupied molecular orbital (HOMO) levels as well and matching the optical bandgap for the interface and the bulk. In contrast, the trap device exhibited a smaller EQE value over the entire spectrum compared with the normal (bisPCBM) device (Figure 3b,d). This indicates that the electrons could be trapped in F-PCBM at the D/A interface after charge transfer, decreasing the photocurrent. Next, we discuss the origin of the changes in FF. The change could be related to the geminate recombination from the initial CTS and non-geminate recombination of the once-free charges. The JSC values were measured under white LED light for four types of devices with P3HT as donor layer at temperatures from 20 to −40 °C. The temperature dependence of JSC is expressed by the following Arrhenius equation: ⎛ Δ⎞ JSC = J 0 (PLight ) exp ⎜ − ⎟ ⎝ kT ⎠
(1)
where J0(Plight) is the preexponential factor and Δ is the activation energy.[20] The results are shown in Figure 4 and the activation energies obtained from fitting are summarized in Table S2 in the Supporting Information. JSC decreased sharply when the temperature decreased in the trap device compared
Figure 3. EQE plots of the four bilayer devices with a,b) P3HT and c,d) PTB7 as the donor layer.
4
wileyonlinelibrary.com
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
www.advmat.de www.MaterialsViews.com
COMMUNICATION
Figure 4. Temperature dependence of JSC in the four bilayer OPVs with P3HT as the donor layer under white LED irradiation.
with the normal (bisPCBM) device. Interestingly, JSC of the cascade device was almost independent of the temperature. A nearly linear increase in JSC with light intensity was observed for all the devices, even at low temperatures (Figure S14, Supporting Information). This indicates that non-geminate recombination was a negligible loss mechanism under these conditions, which are close to the short circuit conditions in this temperature range. The difference in the temperature dependence of the charge mobility can be eliminated for the origin of the observed difference, because the bulk materials in both the donor and the acceptor layers are the same. Moreover, the mean drift lengths of the charges were calculated to be much larger than the film thickness (Supporting Information). Therefore, JSC was determined by the generation rate of the charges with geminate recombination as the only loss mechanism. The origin of the activation energy was attributed to the trapping of the Coulomb attraction at the D/A interface. Charge pairs were deeply trapped at the D/A interface in the trap devices, promoting geminate recombination. In contrast, the temperature independent JSC in the cascade devices indicated free charge generation at the D/A interface without thermal activation. These mechanisms of promotion and suppression of the geminate recombination produced the change in FF. This result suggested that the Coulomb binding in the CTS was weak enough to form free charges in the cascade device. To explain the large VOC changes caused by different energy landscapes at the D/A interface, the energy of the CTS of four types of devices with P3HT as the donor layer was measured by electroluminescence (EL) and the temperature dependence of VOC. It has been reported that VOC of bulk heterojunction photovoltaic devices at the low-temperature limit can be directly related to the energy of the CT state (ECT) measured by EL.[21] The EL spectra of the bilayer devices are shown in Figure 5. Because of the detection limit of the photodetector, the peak top of the CTS emission from the normal (PCBM) device was not visible. The other peaks were fitted with a Gaussian function to extract the value of peak top and the reorganization energy, λ. The energy values are summarized in Table S3 in the Supporting Information. The calculated values of peak top of the
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
Figure 5. EL spectra of the four bilayer OPVs with P3HT as the donor layer.
cascade, normal (bisPCBM), and trap devices were 0.90, 0.94, and 0.87 eV, respectively. A shift of about +0.1 eV in the EL spectra was observed for the cascade device compared with the normal (PCBM) device. However, the peak top for the trap device shifted by about −0.1 eV compared with the normal (bisPCBM) device. Shifts in the CTS energy were also observed in the temperature dependence of VOC under white LED light (Figure S15, Supporting Information). ECT extrapolated from the low-temperature limits for cascade, normal (PCBM), normal (bisPCBM), and trap are 0.95, 0.86, 1.06, and 0.94 eV, respectively. The energy changes for the cascade (+0.09 eV) and trap (−0.12 eV) compared with the corresponding normal devices generally agree with those from EL measurements, although direct comparison would be difficult because of the temperature dependence of ECT.[21] These differences in the energy of CTS were mainly attributed to the reason for the change in VOC in the cascade and trap devices. The energy of CTS is expressed by: E CT = E DA −
e2 4πε 0 ε r d
(2)
where EDA is the energy level difference between the HOMO of the donor and LUMO of the acceptor, ε0 is the vacuum
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
5
www.advmat.de www.MaterialsViews.com
COMMUNICATION
dielectric constant, εr is the relative dielectric constant, and d is the electron–hole distance. The energy level of the LUMO at the first layer of the D/A interface in the trap device was lower than that of the normal (bisPCBM) device; thus, the charge recombination occurred via the trap at the D/A interface and decreased the energy of CTS in the trap device. The energy level of the LUMO at the first layer of the D/A interface in the cascade device was higher than that of the normal (PCBM) device. The energy of the electrons in the first layer of the acceptor were raised by a higher LUMO, and the spatially separated charge pairs further away than the first monolayer from the D/A interface experienced a weaker Coulomb attraction.[16] The energy of the CTS was destabilized in both cases, although we could not conclude at this stage which factor made the larger contribution. We also measured the transient photovoltage (TPV) for the four devices with P3HT as the donor layer to trace the nongeminate recombination. The charge lifetimes of the four types of devices were plotted as a function of carrier density measured by TPV and transient photocurrent (TPC) in Figure 6, by using the analytical method reported by Durrant et al. (Supporting Information).[22] All the TPV signals were fitted well with single exponential decays, except for the devices with a cascade energy landscape where two exponential components
with different lifetimes were necessary to fit the data (“fast” and “slow” in Figure 6a). They could be assigned to the recombination of electrons at the first monolayer of F-bisPCBM (fast) and at the second layer of PCBM (slow). At any charge concentration, the lifetimes of the free charges were longer in the cascade device than those in the normal device. The cascade energy level at the interface may have reduced the population of the electrons at the first monolayer under the photoirradiated steady state and the recombination happened through spatially separated CTS, resulting in the change in the kinetics and the suppression of the non-geminate recombination. In contrast, the trap device showed a lifetime similar to that of the normal (bisPCBM) device. This suggests that under open-circuit conditions, the trap at the interface did not promote bimolecular recombination. We conclude that the energetics of the first monolayer at the D/A interfaces controls the energy of the CTS and largely affects the geminate and the non-geminate recombination processes. In particular, the cascade energy landscape weakens the Coulomb binding energy at the interface and suppresses both the recombination processes, resulting in a higher FF and VOC, and activation-free charge generation. The current model system experimentally demonstrates the importance of controlling the molecular structure at the D/A interfaces. This should also apply to the D/A interfaces in bulk heterojunction structures, in which the interfacial structures are currently illdefined and have large variations because they are constructed by mixing. The next major challenge is to develop methods to fabricate cascade energy interfaces precisely at the molecular level in bulk heterojunction structures, preferably through molecular self-organization. This could lead to another breakthrough in high-performance OPVs beyond the limit of simple mixing of the materials.
Experimental Section
Figure 6. Total perturbation lifetime of the four bilayer OPV devices with P3HT as the donor plotted as a function of carrier density measured by TPV and TPC.
6
wileyonlinelibrary.com
We followed our previous paper for the inverted bilayer device fabrication with the optimized film thickness and thermal annealing conditions.[14] An ITO-coated glass substrate was cleaned by sequential ultrasonication in detergent solution, water, acetone, 2-propanol, and water. A TiO2 precursor solution (NDH-510C, Nippon Soda) diluted in ethanol was spin-coated on the ITO substrates at a spinning rate of 3000 rpm for 30 s, dried at 140 °C for 40 min, and calcined at 500 °C for 30 min to form an electron-transporting layer. The thickness of the TiO2 layer was 40 nm. The substrates were cleaned by ultrasonication in acetone and 2-propanol. Chlorobenzene solution containing 10 g L−1 of PCBM (Solenne) or bisPCBM (Frontier carbon) and F-PCBM (0.75 mg), F-bisPCBM (0.86 mg), PCBM (0.51 mg), or bisPCBM (0.62 mg) was spin-coated onto the ITO/TiO2 substrates at a spinning rate of 600 rpm for 60 s. The thickness of all acceptor layers was about 35 nm. The substrates were thermally annealed at 155 °C for 10 min inside a N2-filled glove box. Substrates with the structure glass/ poly(sodium 4-styrenesulfonate) (PSS) (Mw: 4300 g mol−1, Aldrich)/ P3HT(Merck) or PTB7 (1-Material) were prepared by successive spin coating. PSS was a sacrificial layer prepared by spin coating an aqueous solution containing 30 g L−1 of PSS at a spinning rate of 4000 rpm for 30 s on glass substrates that were precleaned and exposed to UV-O3 using the same method as the ITO substrates. Chlorobenzene solution containing 10 g L−1 of P3HT or PTB7 was spin-coated on glass/PSS substrates at a spinning rate of 1000 rpm for 60 s. The thicknesses of the P3HT and PTB7 layer were 50 and 40 nm, respectively. The glass/PSS/
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
www.advmat.de www.MaterialsViews.com
Supporting Information Supporting Information is available from the Wiley Online Library or from the author.
Acknowledgements The authors thank Professor Ohkita and Dr. Tamai of Kyoto University for their advice on EL measurements. This research was supported by JST, PRESTO. S.I. thanks JSPS for financial support. The authors thank Frontier Carbon Co., Ltd. for providing bisPCBM. Received: February 16, 2015 Revised: March 10, 2015 Published online:
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
COMMUNICATION
P3HT or PTB7 substrate was gently placed upside down on the ITO/ TiO2/PCBM substrate, and one drop of water was placed on the edge of the two substrates. Water selectively penetrated and dissolved the PSS layer, allowing the P3HT layer to be transferred onto the PCBM layer. A MoO3 hole-transporting layer (10 nm) and Ag electrodes (100 nm) were deposited by thermal evaporation under high vacuum (≈10−4 Pa) in a vacuum evaporation system (H-2807 system with E-100 load lock, ALS Technology). The J–V characteristics of the devices were measured under simulated solar illumination (AM 1.5, 100 mW cm−2) from a solar simulator based on a 150 W Xe lamp (PEC-L11, Peccell Technologies). The light intensity was calibrated with a standard silicon solar cell (BS520, Bunkoh-Keiki). The active area of the devices was defined by using a 0.12 cm2 photo mask. The EQE of the devices was measured on a Hypermonolight SM-250F system (Bunkoh-Keiki). EL was measured on a spectrofluorometer equipped with an InGaAs detector (Nanolog, Horiba) and a constant DC voltage was applied to the devices by a DC power supply (PAN 110-3A, Kikusui). The J–V characteristics of the devices at different temperatures (r.t. to −40 °C) were measured under white LED light at an operation power of 3.4 W (10 W LED XM-L, Cree) using a source-meter (2400, Keithley). TPV and TPC measurements were carried out at room temperature. The light source for measuring small perturbations of VOC (ΔVOC) was a N2-dye pulse laser (KEC-160, Usho) with an excitation wavelength, repetition rate, and pulse duration of 510 nm, 2.5 Hz, and 0.4 ns, respectively. The intensity of the laser pulse was controlled by a neutral density filter to keep ΔVOC below 5 mV. The bias light source was a white LED (10 W LED XM-L, Cree) with a neutral density filter. The electrical signal was detected with a digital oscilloscope (DS-5632, Iwatsu). For the TPV measurements, a resistance of 1 MΩ was used for the input impedance of the oscilloscope, which kept the sample under opencircuit conditions. For TPC measurements, a resistance of 50 Ω was used parallel to the input impedance of the oscilloscope, and the TPC was calculated using Ohm’s law. The carrier density and lifetime were calculated using the method reported by Durrant and co-workers.[22] To calculate the carrier density, the differential capacitance method was used. Differential capacitance was calculated from Cdiff = ΔQSC/ΔVOC and plotted as a function of VOC. ΔQSC is the perturbation of carrier density calculated by time integration of the photocurrent in the TPC measurements. Total carrier density, n(V), is determined by integrating Cdiff with VOC. The small perturbation lifetime, τΔn, was determined by fitting the TPV signal with single-exponential decays. All the TPV signals were well fitted with this function, except for the cascade devices where two exponential components with different lifetimes were necessary to fit the data (Figure S16, Supporting Information). The slope of n–τΔn gave the order of recombination, λ (dn/dt = −k0nλ+1). λ is taken into account for calculating carrier lifetime τ, which is determined from τ = (λ + 1) × τΔn.
[1] S. Few, J. M. Frost, J. Nelson, Phys. Chem. Chem. Phys. 2015, 17, 2311. [2] G. Lakhwani, A. Rao, R. H. Friend, Annu. Rev. Phys. Chem. 2014, 65, 557. [3] P. K. Nayak, K. L. Narasimhan, D. Cahen, J. Phys. Chem. Lett. 2013, 4, 1707. [4] S. Gelinas, A. Rao, A. Kumar, S. L. Smith, A. W. Chin, J. Clark, T. S. van der Poll, G. C. Bazan, R. H. Friend, Science 2014, 343, 512. [5] a) G. Grancini, M. Maiuri, D. Fazzi, A. Petrozza, H. J. Egelhaaf, D. Brida, G. Cerullo, G. Lanzani, Nat. Mater. 2013, 12, 29; b) A. E. Jailaubekov, A. P. Willard, J. R. Tritsch, W.-L. Chan, N. Sai, R. Gearba, L. G. Kaake, K. J. Williams, K. Leung, P. J. Rossky, X. Y. Zhu, Nat. Mater. 2013, 12, 66; c) A. A. Bakulin, A. Rao, V. G. Pavelyev, P. H. M. van Loosdrecht, M. S. Pshenichnikov, D. Niedzialek, J. Cornil, D. Beljonne, R. H. Friend, Science 2012, 335, 1340. [6] K. Vandewal, S. Albrecht, E. T. Hoke, K. R. Graham, J. Widmer, J. D. Douglas, M. Schubert, W. R. Mateker, J. T. Bloking, G. F. Burkhard, A. Sellinger, J. M. J. Frechet, A. Amassian, M. K. Riede, M. D. McGehee, D. Neher, A. Salleo, Nat. Mater. 2014, 13, 63. [7] a) S. Sweetnam, K. R. Graham, G. O. N. Ndjawa, T. Heumueller, J. A. Bartelt, T. M. Burke, W. Li, W. You, A. Amassian, M. D. McGehee, J. Am. Chem. Soc. 2014, 136, 14078; b) F. C. Jamieson, E. B. Domingo, T. McCarthy-Ward, M. Heeney, N. Stingelin, J. R. Durrant, Chem. Sci. 2012, 3, 485. [8] K. Vandewal, S. Himmelberger, A. Salleo, Macromolecules 2013, 46, 6379. [9] a) C. Groves, Energy Environ. Sci. 2013, 6, 1546; b) M. L. Jones, R. Dyer, N. Clarke, C. Groves, Phys. Chem. Chem. Phys. 2014, 16, 20310. [10] F. Gao, O. Inganas, Phys. Chem. Chem. Phys. 2014, 16, 20291. [11] a) S. Sista, Y. Yao, Y. Yang, M. L. Tang, Z. Bao, Appl. Phys. Lett. 2007, 91, 223508; b) Y. Kinoshita, T. Hasobe, H. Murata, Appl. Phys. Lett. 2007, 91, 083518; c) Z.-K. Tan, K. Johnson, Y. Vaynzof, A. A. Bakulin, L.-L. Chua, P. K. H. Ho, R. H. Friend, Adv. Mater. 2013, 25, 4131. [12] Q. Wei, K. Tajima, K. Hashimoto, ACS Appl. Mater. Interfaces 2009, 1, 1865. [13] Q. Wei, K. Tajima, Y. Tong, S. Ye, K. Hashimoto, J. Am. Chem. Soc. 2009, 131, 17597. [14] A. Tada, Y. Geng, Q. Wei, K. Hashimoto, K. Tajima, Nat. Mater. 2011, 10, 450. [15] a) D. Bilby, J. Amonoo, M. E. Sykes, B. Frieberg, B. Huang, J. Hungerford, M. Shtein, P. Green, J. Kim, Appl. Phys. Lett. 2013, 103, 203902; b) A. Tada, Y. Geng, M. Nakamura, Q. Wei, K. Hashimoto, K. Tajima, Phys. Chem. Chem. Phys. 2012, 14, 3713; c) Y. Zhong, J. Ma, K. Hashimoto, K. Tajima, Adv. Mater. 2013, 25, 1071. [16] Y. Zhong, A. Tada, S. Izawa, K. Hashimoto, K. Tajima, Adv. Energy Mater. 2014, 4, 1301332. [17] M. Lenes, G.-J. A. H. Wetzelaer, F. B. Kooistra, S. C. Veenstra, J. C. Hummelen, P. W. M. Blom, Adv. Mater. 2008, 20, 2116. [18] Y. C. Lai, T. Higashihara, J. C. Hsu, M. Ueda, W. C. Chen, Sol. Energy Mater. Sol. Cells 2012, 97, 164. [19] S. Izawa, K. Hashimoto, K. Tajima, Phys. Chem. Chem. Phys. 2014, 16, 16383. [20] I. Riedel, V. Dyakonov, Phys. Status Solidi A 2004, 201, 1332. [21] K. Vandewal, K. Tvingstedt, A. Gadisa, O. Inganas, J. V. Manca, Phys. Rev. B 2010, 81, 125204. [22] a) B. C. O’Regan, S. Scully, A. C. Mayer, E. Palomares, J. Durrant, J. Phys. Chem. B 2005, 109, 4616; b) C. G. Shuttle, B. O’Regan, A. M. Ballantyne, J. Nelson, D. D. C. Bradley, J. de Mello, J. R. Durrant, Appl. Phys. Lett. 2008, 92, 093311.
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
7
COMMUNICATION
Dominant Effects of First Monolayer Energetics at Donor/ Acceptor Interfaces on Organic Photovoltaics Seiichiro Izawa, Kyohei Nakano, Kaori Suzuki, Kazuhito Hashimoto, and Keisuke Tajima* Even after 20 years of research, it is still unknown how the charges in organic photovoltaics (OPVs) are separated and collected so efficiently.[1] Charge separation occurs via charge transfer of photogenerated excitons across the donor/acceptor (D/A) interface to form electron/hole pairs. However, Coulomb binding between the charge pair is strong compared with the thermal energy and is effective over a long range due to the low dielectric constants of organic semiconductors. Therefore, the bound charge transfer state (CTS) can be formed at the D/A interface before the free charges form. The CTS can decay to the ground state with a certain probability (geminate recombination), resulting in loss of the photocurrent. The CTS is also formed when free charge pairs encounter each other at the D/A interface and decay to the ground state via non-geminate (bimolecular) recombination, the kinetics of which are strongly related to the fill factor (FF) and open-circuit voltage (VOC).[2] Recently, it was proposed that the higher energy hot (delocalized) CTS may be important for efficient charge separation.[3] Ultrafast spectroscopy studies show that delocalization of the CT state over several nanometers on aggregated fullerene molecules[4] and the excess energy assists free charge formation.[5] In contrast, the excitation of the relaxed CTS can be efficiently separated into free charges without excess energy.[6] It was proposed that a gradient forms in the free-energy landscape near the D/A interface because the disorder of the materials widens the energy gap.[7] The unintentional “cascade” energy landscape at the interface could push the energy of the CTS close to the charge separated state, resulting in highly efficient charge separation.[8] Groves used a kinetic Monte Carlo simulation to predict that thin cascade layers (1 nm) suppressed geminate recombination, whereas thicker cascade layers did not.[9] Although the mechanism for the efficient formation of free charge is still unknown,[10] the aggregation structures or the gradient of the free-energy landscape close to the D/A interface could be important for obtaining efficient OPVs through efficient Dr. S. Izawa, Dr. K. Nakano, K. Suzuki, Dr. K. Tajima RIKEN Center for Emergent Matter Science (CEMS) 2-1 Hirosawa, Wako Saitama 351-0198, Japan E-mail: [email protected] Dr. S. Izawa, Prof. K. Hashimoto Department of Applied Chemistry Graduate School of Engineering The University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan Dr. K. Tajima Precursory Research for Embryonic Science and Technology (PRESTO) Japan Science and Technology Agency 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan
DOI: 10.1002/adma.201500840
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
charge separation and suppressing recombination. The effects of the energy cascade at the D/A interface have been explored with model bilayer systems. However, because the materials were either spin coated successively or evaporated, the interfacial layers were thick and the interfacial structures were mixed on a molecular scale.[11] Thus, the effects of the energy cascade close to the interface at the molecular monolayer level remain unclear. In this work, we combine our contact film transfer (CFT) method[12] with a self-organized surface segregated monolayer (SSM)[13] to fabricate bilayer OPVs with well-defined D/A interfaces.[14–16] To tune the energetics at the surface precisely, a new surface modifier based on fluoroalkyl bis(1-[3-(methoxycarbonyl) propyl]-1-phenyl)-[6.6]C62 (F-bisPCBM) (Figure 1a) was synthesized. Fullerene bisadducts generally have lower electron acceptability (higher-lying lowest unoccupied molecular orbital (LUMO)) than monoadducts because the π-conjugation is broken.[17] We combine two surface modifiers with different LUMO energy levels (F-PCBM and F-bisPCBM) with two fullerene acceptors (PCBM and bisPCBM), and construct four energy landscapes at the D/A interface in the bilayer OPVs (Figure 1b). We investigate the effect of the energy levels of the first monolayer at the D/A interface on the photoelectric conversion. F-PCBM and F-bisPCBM were synthesized via an acid-catalyzed ester exchange reaction[18,19] as described in the Supporting Information. We compared the molecular electronic properties of PCBM, bisPCBM, F-PCBM, and F-bisPCBM by cyclic voltammetry (CV) in solution. The reduction potentials of the fullerene bisadducts (bisPCBM and F-bisPCBM, −1.20 V vs Fc/Fc+) have a reduction potential that is 0.12 V lower than those of the monoadducts (PCBM and F-PCBM, −1.08 V vs Fc/Fc+) (Figure S3, Supporting Information). Therefore, the fullerene bisadducts have higher-lying LUMO energy levels compared with the monoadducts.[17] There is no difference between the fluoroalkyl and methyl ester derivatives, indicating that the fluoroalkyl chain attached to the ester group does not affect the energy level of the fullerene.[13] The identical absorption spectra of the fluoroalkyl and methyl ester derivatives in solution also show that the fluoroalkyl chains have little effect on the molecular orbitals of the fullerene group (Figure S4, Supporting Information). Surface segregation behaviors of F-PCBM and F-bisPCBM in the blended films were investigated by X-ray photoelectron spectroscopy (XPS). The films prepared from different combinations of the surface modifiers and bulk materials (F-PCBM/PCBM, F-bisPCBM/PCBM, F-PCBM/bisPCBM, and F-bisPCBM/bisPCBM) were prepared by spin coating solutions containing different concentrations of F-PCBM and F-bisPCBM (0.2–1.2 g L−1) with a fixed concentration of PCBM and bisPCBM (10 g L−1). The surface F/C atomic ratio was
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
1
www.advmat.de
COMMUNICATION
www.MaterialsViews.com
Figure 1. a) Chemical structures of the surface modifiers and the bulk materials, and schematic images of the surface modification by SSM and the fabrication of bilayer devices by CFT. b) Schematic images of the four different energy landscapes at the D/A interface in the bilayer OPVs.
calculated from the F 1s and non-fluorinated C 1s XPS peak intensities. Figure S5 in the Supporting Information shows the F/C atomic ratio as a function of the concentration of F-PCBM and F-bisPCBM in the solutions. The results are similar to those of our previous studies on SSM formation in F-PCBM/ PCBM; the F/C ratios measured by XPS were higher than those calculated for homogeneously mixed films for all films, indicating the surface segregation of F-PCBM and F-bisPCBM.[19] The clear saturation of F/C ratios was observed at high concentrations of the surface modifiers in all the combinations. The F/C atomic ratios at the saturated conditions were 0.128, 0.111, 0.111, and 0.102 in F-PCBM/PCBM, F-bisPCBM/PCBM, F-PCBM/bisPCBM, and F-bisPCBM/bisPCBM, respectively. The F/C atomic ratios reflect the density of the segregated fluoroalkyl chains at the surface. As previously reported, the surface of F-PCBM/PCBM was almost fully covered by SSM according to the maximum F/C ratios calculated for full coverage conditions.[19] The slightly lower densities in the films of the bisadduct fullerene derivatives compared with that of F-PCBM/PCBM could be attributed to larger steric hindrance and thus a lower packing density of the fullerene bisadducts. The concentration of the surface modifiers close to the saturation point (0.56 × 10−3 M (0.75 g L−1) for F-PCBM and 0.56 × 10−3 M (0.86 g L−1) for F-bisPCBM) were used to investigate the film properties and for device fabrication. XPS depth profiles of the films showed that the F 1s peak of all the films disappeared after surface etching to a depth of about 1.5 nm with an Ar+ ion beam (Figure S6, Supporting Information). This suggests that the fluoroalkyl chains of F-PCBM and F-bisPCBM segregated to the surface and a small amount remained in the bulk of the films.
2
wileyonlinelibrary.com
To elucidate the surface structure further, we also performed angle-resolved XPS. The peak ratios of fluorinated to non-fluorinated carbon were used for measurements. A bilayer model, in which the surface layers consist of fluoroalkyl chains, fitted the experimental data well (details provided in the Supporting Information). This indicated that F-PCBM or F-bisPCBM formed a monolayer at the surface of both PCBM and bisPCBM films with fluoroalkyl chains exposed to the surface. The ionization potentials (IPs) of the films were measured by UV photoelectron spectroscopy (UPS; Figure S8, Supporting Information and Table 1). There were small differences of 0.04 eV in IPs between F-PCBM/PCBM and F-bisPCBM/PCBM and of 0.03 eV between F-PCBM/bisPCBM and F-bisPCBM/ bisPCBM. The oriented fluoroalkyl chain at the surface should form a surface dipole moment; therefore, the IPs of the films should be shifted compared with the pristine films depending on the surface density of the fluoroalkyl chains.[14] The small difference in the IPs between the surface modifiers (F-PCBM Table 1. IPs of the films measured by UPS and reduction potentials (half potentials) versus Fc/Fc+ of PCBM and bisPCBM determined by CV in CH2Cl2.
PCBM
IP [eV]
Ered [V]
5.81
−1.08
F-PCBM/PCBM
5.41
F-bisPCBM/PCBM
5.45
bisPCBM
5.79
F-PCBM/bisPCBM
5.51
F-bisPCBM/bisPCBM
5.54
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
−1.20
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
www.advmat.de www.MaterialsViews.com
COMMUNICATION
Figure 2. J–V curves of the four bilayer OPVs with a,b) P3HT and c,d) PTB7 as the donor layer under AM 1.5, 100 mW cm−2 irradiation.
and F-bisPCBM) could be caused by the difference in surface density of the fluoroalkyl chains suggested by XPS. Figure 1b shows a schematic of the energy levels of the four kinds of acceptor films deduced from the CV and UPS results. The energy levels between the surface monolayer and the bulk are the same in the F-PCBM/PCBM and F-bisPCBM/bisPCBM films. A LUMO energy trap with a depth of 0.12 eV formed at the surface monolayer in the F-PCBM/bisPCBM film, whereas a LUMO energy cascade of 0.12 eV formed at the surface monolayer in the F-bisPCBM/PCBM film. The advantage of this model is that besides the surface energy levels, other properties can be tuned to the same conditions; the optical absorption of the films was identical (Figure S9, Supporting Information), the film thicknesses were similar (about 35 nm), and the planarity of the film surfaces was similar with Ra of 0.16–0.26 nm (Figure S10, Supporting Information). The surface dipole moments caused by the fluoroalkyl chains exist in all four films. UPS showed that the vacuum level shift for different surface modifiers was similar, with differences of 0.03–0.04 eV. This enabled us to compare the device performance of the bilayer OPVs directly. Four bilayer devices with different energy levels at the D/A interface were fabricated via CFT of the same donor polymer layer onto the four fullerene layers (Figure 1a). The structures of the devices were indium tin oxide (ITO)/TiO2/acceptor/ SSM//donor/MoO3/Ag, where // denotes the interface created by film transfer. It has been shown that the D/A interfaces of the bilayer devices fabricated by CFT are flat and sharp and thus the surface structure of films before the transfer are retained.[14] As the result, the surface energy level differences of the fullerene films are preserved and the four bilayer devices with different energy landscapes at the D/A interface (donor//
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
F-PCBM/PCBM: normal (PCBM); donor//F-bisPCBM/PCBM: cascade; donor//F-bisPCBM/bisPCBM: normal (bisPCBM); donor//F-PCBM/bisPCBM: trap) were fabricated (Figure 1b). The J–V characteristics of the four devices with P3HT and PTB7 as the donor layer under simulated solar light irradiation are shown in Figure 2 and the results are summarized in Table S1, Supporting Information. The cascade device with a P3HT donor layer exhibited VOC and FF values that were larger than the normal (PCBM) device, whereas the JSC values were similar. The changes in VOC and FF were +0.09 V and +0.13, respectively. In contrast, the trap device with P3HT as the donor layer exhibited much lower VOC, FF, and JSC values than the normal (bisPCBM) device. The changes in VOC and FF were −0.12 V and −0.18, respectively. The devices with PTB7 as the donor layer exhibited similar behavior (Figure 2c,d): the VOC of the cascade (trap) device was higher (lower) by +0.11 V (−0.15 V) and the FF of the cascade (trap) device was higher (lower) by +0.20 (−0.21). However, the JSC value of the cascade device was smaller than that of the normal devices. The differences in device performance between the four devices can be attributed to the difference in the energy landscape at the D/A interface because the bulk properties, such as the absorption and charge mobility, were the same. In particular, the large difference in FF suggested that the charge recombination was suppressed in the cascade configuration, whereas it was promoted in the trap configuration. The s-shape of J–V curve in Figure 2d could be caused by severe charge recombination at D/A interface in the trap configuration. The molecular dipole moment of the fluoroalkyl chains in SSM layers caused VOC change as previously reported;[14] PTB7//F-PCBM/PCBM and PTB7//F-bisPCBM/ bisPCBM showed lower VOC by 0.18 and 0.06 V, respectively, compared to those of PTB7//PCBM and PTB7//bisPCBM,
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
3
www.advmat.de
COMMUNICATION
www.MaterialsViews.com
respectively (Table S1, Supporting Information). However, the large changes in VOC of the cascade and trap devices were not due to the difference in the dipole moments but the different energy landscapes at the D/A interface because all the four films have the interfacial fluoroalkyl chains with the similar densities and the vacuum level shifts measured by UPS for the different surface modifiers were also similar. The changes in device performance may be caused by the presence of a small amount of surface modifiers below the XPS detection limit in the bulk. To eliminate this possibility, small amounts of PCBM and bisPCBM were intentionally added to the bisPCBM and PCBM layers, respectively, in the bilayer devices containing PTB7 (details provided in the Supporting Information). Figure S11 in the Supporting Information shows that the addition of a small amount of the different acceptors in the bulk of the films induced no change in the performance. Therefore, the energy level difference at the D/A interface, not in the bulk, caused the changes. To explain the changes in JSC, the external quantum efficiency (EQE) spectra of four devices with P3HT and PTB7 as the donor layer are shown in Figure 3. The EQE peak at 600 nm was from the absorption of P3HT in Figure 3a,b, the peak at 650 nm was from the absorption of PTB7 in Figure 3c, d, and the peaks at 350 and 450 nm had a large contribution from PCBM absorption. This was confirmed by optical simulation and EQE calculations for devices, based on the transfer matrix model and an exciton diffusion model (Figure S13, Supporting Information). Figure 3a shows that the cascade device had a larger EQE in the donor absorption region, and a smaller EQE in the acceptor absorption region compared with the normal (PCBM) device. Therefore, the JSC in the cascade device was similar to that of the normal device. In PTB7 devices, the EQE in the donor
absorption region was unchanged and was smaller in the acceptor absorption region for the cascade device, resulting in the smaller JSC under solar irradiation (Figure 3c). The decrease in EQE in the fullerene absorption region may be because of the larger bandgap of F-bisPCBM at the D/A interface than that of PCBM in the bulk (Figure 1b); therefore, the exciton in the bulk PCBM could not reach the F-bisPCBM layer at the D/A interface in the cascade devices since the exciton migration process is energetically unfavorable by 0.12 eV. This can be avoided by creating a cascade for the highest occupied molecular orbital (HOMO) levels as well and matching the optical bandgap for the interface and the bulk. In contrast, the trap device exhibited a smaller EQE value over the entire spectrum compared with the normal (bisPCBM) device (Figure 3b,d). This indicates that the electrons could be trapped in F-PCBM at the D/A interface after charge transfer, decreasing the photocurrent. Next, we discuss the origin of the changes in FF. The change could be related to the geminate recombination from the initial CTS and non-geminate recombination of the once-free charges. The JSC values were measured under white LED light for four types of devices with P3HT as donor layer at temperatures from 20 to −40 °C. The temperature dependence of JSC is expressed by the following Arrhenius equation: ⎛ Δ⎞ JSC = J 0 (PLight ) exp ⎜ − ⎟ ⎝ kT ⎠
(1)
where J0(Plight) is the preexponential factor and Δ is the activation energy.[20] The results are shown in Figure 4 and the activation energies obtained from fitting are summarized in Table S2 in the Supporting Information. JSC decreased sharply when the temperature decreased in the trap device compared
Figure 3. EQE plots of the four bilayer devices with a,b) P3HT and c,d) PTB7 as the donor layer.
4
wileyonlinelibrary.com
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
www.advmat.de www.MaterialsViews.com
COMMUNICATION
Figure 4. Temperature dependence of JSC in the four bilayer OPVs with P3HT as the donor layer under white LED irradiation.
with the normal (bisPCBM) device. Interestingly, JSC of the cascade device was almost independent of the temperature. A nearly linear increase in JSC with light intensity was observed for all the devices, even at low temperatures (Figure S14, Supporting Information). This indicates that non-geminate recombination was a negligible loss mechanism under these conditions, which are close to the short circuit conditions in this temperature range. The difference in the temperature dependence of the charge mobility can be eliminated for the origin of the observed difference, because the bulk materials in both the donor and the acceptor layers are the same. Moreover, the mean drift lengths of the charges were calculated to be much larger than the film thickness (Supporting Information). Therefore, JSC was determined by the generation rate of the charges with geminate recombination as the only loss mechanism. The origin of the activation energy was attributed to the trapping of the Coulomb attraction at the D/A interface. Charge pairs were deeply trapped at the D/A interface in the trap devices, promoting geminate recombination. In contrast, the temperature independent JSC in the cascade devices indicated free charge generation at the D/A interface without thermal activation. These mechanisms of promotion and suppression of the geminate recombination produced the change in FF. This result suggested that the Coulomb binding in the CTS was weak enough to form free charges in the cascade device. To explain the large VOC changes caused by different energy landscapes at the D/A interface, the energy of the CTS of four types of devices with P3HT as the donor layer was measured by electroluminescence (EL) and the temperature dependence of VOC. It has been reported that VOC of bulk heterojunction photovoltaic devices at the low-temperature limit can be directly related to the energy of the CT state (ECT) measured by EL.[21] The EL spectra of the bilayer devices are shown in Figure 5. Because of the detection limit of the photodetector, the peak top of the CTS emission from the normal (PCBM) device was not visible. The other peaks were fitted with a Gaussian function to extract the value of peak top and the reorganization energy, λ. The energy values are summarized in Table S3 in the Supporting Information. The calculated values of peak top of the
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
Figure 5. EL spectra of the four bilayer OPVs with P3HT as the donor layer.
cascade, normal (bisPCBM), and trap devices were 0.90, 0.94, and 0.87 eV, respectively. A shift of about +0.1 eV in the EL spectra was observed for the cascade device compared with the normal (PCBM) device. However, the peak top for the trap device shifted by about −0.1 eV compared with the normal (bisPCBM) device. Shifts in the CTS energy were also observed in the temperature dependence of VOC under white LED light (Figure S15, Supporting Information). ECT extrapolated from the low-temperature limits for cascade, normal (PCBM), normal (bisPCBM), and trap are 0.95, 0.86, 1.06, and 0.94 eV, respectively. The energy changes for the cascade (+0.09 eV) and trap (−0.12 eV) compared with the corresponding normal devices generally agree with those from EL measurements, although direct comparison would be difficult because of the temperature dependence of ECT.[21] These differences in the energy of CTS were mainly attributed to the reason for the change in VOC in the cascade and trap devices. The energy of CTS is expressed by: E CT = E DA −
e2 4πε 0 ε r d
(2)
where EDA is the energy level difference between the HOMO of the donor and LUMO of the acceptor, ε0 is the vacuum
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
5
www.advmat.de www.MaterialsViews.com
COMMUNICATION
dielectric constant, εr is the relative dielectric constant, and d is the electron–hole distance. The energy level of the LUMO at the first layer of the D/A interface in the trap device was lower than that of the normal (bisPCBM) device; thus, the charge recombination occurred via the trap at the D/A interface and decreased the energy of CTS in the trap device. The energy level of the LUMO at the first layer of the D/A interface in the cascade device was higher than that of the normal (PCBM) device. The energy of the electrons in the first layer of the acceptor were raised by a higher LUMO, and the spatially separated charge pairs further away than the first monolayer from the D/A interface experienced a weaker Coulomb attraction.[16] The energy of the CTS was destabilized in both cases, although we could not conclude at this stage which factor made the larger contribution. We also measured the transient photovoltage (TPV) for the four devices with P3HT as the donor layer to trace the nongeminate recombination. The charge lifetimes of the four types of devices were plotted as a function of carrier density measured by TPV and transient photocurrent (TPC) in Figure 6, by using the analytical method reported by Durrant et al. (Supporting Information).[22] All the TPV signals were fitted well with single exponential decays, except for the devices with a cascade energy landscape where two exponential components
with different lifetimes were necessary to fit the data (“fast” and “slow” in Figure 6a). They could be assigned to the recombination of electrons at the first monolayer of F-bisPCBM (fast) and at the second layer of PCBM (slow). At any charge concentration, the lifetimes of the free charges were longer in the cascade device than those in the normal device. The cascade energy level at the interface may have reduced the population of the electrons at the first monolayer under the photoirradiated steady state and the recombination happened through spatially separated CTS, resulting in the change in the kinetics and the suppression of the non-geminate recombination. In contrast, the trap device showed a lifetime similar to that of the normal (bisPCBM) device. This suggests that under open-circuit conditions, the trap at the interface did not promote bimolecular recombination. We conclude that the energetics of the first monolayer at the D/A interfaces controls the energy of the CTS and largely affects the geminate and the non-geminate recombination processes. In particular, the cascade energy landscape weakens the Coulomb binding energy at the interface and suppresses both the recombination processes, resulting in a higher FF and VOC, and activation-free charge generation. The current model system experimentally demonstrates the importance of controlling the molecular structure at the D/A interfaces. This should also apply to the D/A interfaces in bulk heterojunction structures, in which the interfacial structures are currently illdefined and have large variations because they are constructed by mixing. The next major challenge is to develop methods to fabricate cascade energy interfaces precisely at the molecular level in bulk heterojunction structures, preferably through molecular self-organization. This could lead to another breakthrough in high-performance OPVs beyond the limit of simple mixing of the materials.
Experimental Section
Figure 6. Total perturbation lifetime of the four bilayer OPV devices with P3HT as the donor plotted as a function of carrier density measured by TPV and TPC.
6
wileyonlinelibrary.com
We followed our previous paper for the inverted bilayer device fabrication with the optimized film thickness and thermal annealing conditions.[14] An ITO-coated glass substrate was cleaned by sequential ultrasonication in detergent solution, water, acetone, 2-propanol, and water. A TiO2 precursor solution (NDH-510C, Nippon Soda) diluted in ethanol was spin-coated on the ITO substrates at a spinning rate of 3000 rpm for 30 s, dried at 140 °C for 40 min, and calcined at 500 °C for 30 min to form an electron-transporting layer. The thickness of the TiO2 layer was 40 nm. The substrates were cleaned by ultrasonication in acetone and 2-propanol. Chlorobenzene solution containing 10 g L−1 of PCBM (Solenne) or bisPCBM (Frontier carbon) and F-PCBM (0.75 mg), F-bisPCBM (0.86 mg), PCBM (0.51 mg), or bisPCBM (0.62 mg) was spin-coated onto the ITO/TiO2 substrates at a spinning rate of 600 rpm for 60 s. The thickness of all acceptor layers was about 35 nm. The substrates were thermally annealed at 155 °C for 10 min inside a N2-filled glove box. Substrates with the structure glass/ poly(sodium 4-styrenesulfonate) (PSS) (Mw: 4300 g mol−1, Aldrich)/ P3HT(Merck) or PTB7 (1-Material) were prepared by successive spin coating. PSS was a sacrificial layer prepared by spin coating an aqueous solution containing 30 g L−1 of PSS at a spinning rate of 4000 rpm for 30 s on glass substrates that were precleaned and exposed to UV-O3 using the same method as the ITO substrates. Chlorobenzene solution containing 10 g L−1 of P3HT or PTB7 was spin-coated on glass/PSS substrates at a spinning rate of 1000 rpm for 60 s. The thicknesses of the P3HT and PTB7 layer were 50 and 40 nm, respectively. The glass/PSS/
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
www.advmat.de www.MaterialsViews.com
Supporting Information Supporting Information is available from the Wiley Online Library or from the author.
Acknowledgements The authors thank Professor Ohkita and Dr. Tamai of Kyoto University for their advice on EL measurements. This research was supported by JST, PRESTO. S.I. thanks JSPS for financial support. The authors thank Frontier Carbon Co., Ltd. for providing bisPCBM. Received: February 16, 2015 Revised: March 10, 2015 Published online:
Adv. Mater. 2015, DOI: 10.1002/adma.201500840
COMMUNICATION
P3HT or PTB7 substrate was gently placed upside down on the ITO/ TiO2/PCBM substrate, and one drop of water was placed on the edge of the two substrates. Water selectively penetrated and dissolved the PSS layer, allowing the P3HT layer to be transferred onto the PCBM layer. A MoO3 hole-transporting layer (10 nm) and Ag electrodes (100 nm) were deposited by thermal evaporation under high vacuum (≈10−4 Pa) in a vacuum evaporation system (H-2807 system with E-100 load lock, ALS Technology). The J–V characteristics of the devices were measured under simulated solar illumination (AM 1.5, 100 mW cm−2) from a solar simulator based on a 150 W Xe lamp (PEC-L11, Peccell Technologies). The light intensity was calibrated with a standard silicon solar cell (BS520, Bunkoh-Keiki). The active area of the devices was defined by using a 0.12 cm2 photo mask. The EQE of the devices was measured on a Hypermonolight SM-250F system (Bunkoh-Keiki). EL was measured on a spectrofluorometer equipped with an InGaAs detector (Nanolog, Horiba) and a constant DC voltage was applied to the devices by a DC power supply (PAN 110-3A, Kikusui). The J–V characteristics of the devices at different temperatures (r.t. to −40 °C) were measured under white LED light at an operation power of 3.4 W (10 W LED XM-L, Cree) using a source-meter (2400, Keithley). TPV and TPC measurements were carried out at room temperature. The light source for measuring small perturbations of VOC (ΔVOC) was a N2-dye pulse laser (KEC-160, Usho) with an excitation wavelength, repetition rate, and pulse duration of 510 nm, 2.5 Hz, and 0.4 ns, respectively. The intensity of the laser pulse was controlled by a neutral density filter to keep ΔVOC below 5 mV. The bias light source was a white LED (10 W LED XM-L, Cree) with a neutral density filter. The electrical signal was detected with a digital oscilloscope (DS-5632, Iwatsu). For the TPV measurements, a resistance of 1 MΩ was used for the input impedance of the oscilloscope, which kept the sample under opencircuit conditions. For TPC measurements, a resistance of 50 Ω was used parallel to the input impedance of the oscilloscope, and the TPC was calculated using Ohm’s law. The carrier density and lifetime were calculated using the method reported by Durrant and co-workers.[22] To calculate the carrier density, the differential capacitance method was used. Differential capacitance was calculated from Cdiff = ΔQSC/ΔVOC and plotted as a function of VOC. ΔQSC is the perturbation of carrier density calculated by time integration of the photocurrent in the TPC measurements. Total carrier density, n(V), is determined by integrating Cdiff with VOC. The small perturbation lifetime, τΔn, was determined by fitting the TPV signal with single-exponential decays. All the TPV signals were well fitted with this function, except for the cascade devices where two exponential components with different lifetimes were necessary to fit the data (Figure S16, Supporting Information). The slope of n–τΔn gave the order of recombination, λ (dn/dt = −k0nλ+1). λ is taken into account for calculating carrier lifetime τ, which is determined from τ = (λ + 1) × τΔn.
[1] S. Few, J. M. Frost, J. Nelson, Phys. Chem. Chem. Phys. 2015, 17, 2311. [2] G. Lakhwani, A. Rao, R. H. Friend, Annu. Rev. Phys. Chem. 2014, 65, 557. [3] P. K. Nayak, K. L. Narasimhan, D. Cahen, J. Phys. Chem. Lett. 2013, 4, 1707. [4] S. Gelinas, A. Rao, A. Kumar, S. L. Smith, A. W. Chin, J. Clark, T. S. van der Poll, G. C. Bazan, R. H. Friend, Science 2014, 343, 512. [5] a) G. Grancini, M. Maiuri, D. Fazzi, A. Petrozza, H. J. Egelhaaf, D. Brida, G. Cerullo, G. Lanzani, Nat. Mater. 2013, 12, 29; b) A. E. Jailaubekov, A. P. Willard, J. R. Tritsch, W.-L. Chan, N. Sai, R. Gearba, L. G. Kaake, K. J. Williams, K. Leung, P. J. Rossky, X. Y. Zhu, Nat. Mater. 2013, 12, 66; c) A. A. Bakulin, A. Rao, V. G. Pavelyev, P. H. M. van Loosdrecht, M. S. Pshenichnikov, D. Niedzialek, J. Cornil, D. Beljonne, R. H. Friend, Science 2012, 335, 1340. [6] K. Vandewal, S. Albrecht, E. T. Hoke, K. R. Graham, J. Widmer, J. D. Douglas, M. Schubert, W. R. Mateker, J. T. Bloking, G. F. Burkhard, A. Sellinger, J. M. J. Frechet, A. Amassian, M. K. Riede, M. D. McGehee, D. Neher, A. Salleo, Nat. Mater. 2014, 13, 63. [7] a) S. Sweetnam, K. R. Graham, G. O. N. Ndjawa, T. Heumueller, J. A. Bartelt, T. M. Burke, W. Li, W. You, A. Amassian, M. D. McGehee, J. Am. Chem. Soc. 2014, 136, 14078; b) F. C. Jamieson, E. B. Domingo, T. McCarthy-Ward, M. Heeney, N. Stingelin, J. R. Durrant, Chem. Sci. 2012, 3, 485. [8] K. Vandewal, S. Himmelberger, A. Salleo, Macromolecules 2013, 46, 6379. [9] a) C. Groves, Energy Environ. Sci. 2013, 6, 1546; b) M. L. Jones, R. Dyer, N. Clarke, C. Groves, Phys. Chem. Chem. Phys. 2014, 16, 20310. [10] F. Gao, O. Inganas, Phys. Chem. Chem. Phys. 2014, 16, 20291. [11] a) S. Sista, Y. Yao, Y. Yang, M. L. Tang, Z. Bao, Appl. Phys. Lett. 2007, 91, 223508; b) Y. Kinoshita, T. Hasobe, H. Murata, Appl. Phys. Lett. 2007, 91, 083518; c) Z.-K. Tan, K. Johnson, Y. Vaynzof, A. A. Bakulin, L.-L. Chua, P. K. H. Ho, R. H. Friend, Adv. Mater. 2013, 25, 4131. [12] Q. Wei, K. Tajima, K. Hashimoto, ACS Appl. Mater. Interfaces 2009, 1, 1865. [13] Q. Wei, K. Tajima, Y. Tong, S. Ye, K. Hashimoto, J. Am. Chem. Soc. 2009, 131, 17597. [14] A. Tada, Y. Geng, Q. Wei, K. Hashimoto, K. Tajima, Nat. Mater. 2011, 10, 450. [15] a) D. Bilby, J. Amonoo, M. E. Sykes, B. Frieberg, B. Huang, J. Hungerford, M. Shtein, P. Green, J. Kim, Appl. Phys. Lett. 2013, 103, 203902; b) A. Tada, Y. Geng, M. Nakamura, Q. Wei, K. Hashimoto, K. Tajima, Phys. Chem. Chem. Phys. 2012, 14, 3713; c) Y. Zhong, J. Ma, K. Hashimoto, K. Tajima, Adv. Mater. 2013, 25, 1071. [16] Y. Zhong, A. Tada, S. Izawa, K. Hashimoto, K. Tajima, Adv. Energy Mater. 2014, 4, 1301332. [17] M. Lenes, G.-J. A. H. Wetzelaer, F. B. Kooistra, S. C. Veenstra, J. C. Hummelen, P. W. M. Blom, Adv. Mater. 2008, 20, 2116. [18] Y. C. Lai, T. Higashihara, J. C. Hsu, M. Ueda, W. C. Chen, Sol. Energy Mater. Sol. Cells 2012, 97, 164. [19] S. Izawa, K. Hashimoto, K. Tajima, Phys. Chem. Chem. Phys. 2014, 16, 16383. [20] I. Riedel, V. Dyakonov, Phys. Status Solidi A 2004, 201, 1332. [21] K. Vandewal, K. Tvingstedt, A. Gadisa, O. Inganas, J. V. Manca, Phys. Rev. B 2010, 81, 125204. [22] a) B. C. O’Regan, S. Scully, A. C. Mayer, E. Palomares, J. Durrant, J. Phys. Chem. B 2005, 109, 4616; b) C. G. Shuttle, B. O’Regan, A. M. Ballantyne, J. Nelson, D. D. C. Bradley, J. de Mello, J. R. Durrant, Appl. Phys. Lett. 2008, 92, 093311.
© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
wileyonlinelibrary.com
7
