Introduction

In OSCs, the light-to-electric energy conversion is realized by the combination of the carrier formation and transfer processes within the active layer (Fig. S1). In the former process, the photo irradiation creates a donor exciton in the donor region and the donor exciton migrates to the donor (D)/acceptor (A) interface. Finally, the exciton separates into the electron and hole at the D/A interface. In most cases, the electron and hole are weakly bound to each other around the interface. In the latter process, the carriers transfer to the collector electrode and are collected as photocurrent. This is in a sharp contrast with an inorganic solar cell (ISC), in which the photo-irradiation directly creates carriers within the active layer.

The BHJ active layer of OSC consists of phase-separated nano-size (several tens of nm) domains of the donor polymer and acceptor fullerene1,2,3,4,5. The nano-size structure is essential for the efficient carrier formation process, because the length of exciton migration is ~3 nm. The STXM around the carbon K-edge is a powerful tool to clarify the molecular mixing as well as the domain structure in the BHJ active layer6,7, because it can distinguish the fullerene carbon from the polymer carbon. For example, Collins et al.6 revealed the fullerene mixing in the polymer matrix of PTB7/PC71BM blend film. Due to the low spatial resolution (∼several tens of nm) of STXM, however, the domain size had to be enlarged by chemical admixture6 or thermal annealing at higher temperature7. Recently, several experiments revealed sub-structures within the large domains. By means of atomic force microscopy (AFM) coupled with plasma-ashing technique, Hedley et al.8 investigated sub-structure inside the domain (100–200 nm) of PTB7/PC71BM blend film prepared without additive and found that the domain consists of a large number of small fullerene spheres (20–60 nm). By means of energy-filtered electron transmission microscopy (EFTEM), Kesave et al.9 reported fiber-like structure, ~10 nm wide and ~100 nm long, in PGeBTBT/PC71BM blend film.

On the other hands, the femtosecond time-resolved spectroscopy is a powerful tool to reveal the carrier formation process within the active layer10,11,12,13,14, because the spectroscopy monitors the relative numbers of the photo-created exciton and carrier in the time domain. Significantly, the spectroscopy decouples the carrier formation and transfer processes, because the former process completes within several tens of ps. Actually, the time-resolved spectroscopy revealed that the exciton-to-carrier conversion process in PTB7/PC71BM blend film completes within ~0.3 ps10.

In order to clarify the interrelation between molecular mixing and the photovoltaic properties of BHJ-type OSCs, we selected a liquid-crystalline semiconducting polymer, F8T2, as the donor polymer, because the domain structure of the blend film with fullerene derivatives remains large (several hundreds of nm) and independent of Tan15,16. Yasuda et al.16 systematically investigated the photovoltaic properties of the OSCs based on films of F8T2/PC71BM (33 : 67 wt. %) blend annealed at various temperature (Tan): the PCE systematically decreases from the optimal value ( = 2.28%) at Tan = 80°C to 0.81% at 240°C. Yonezawa et al.17 investigated the charge formation dynamics of F8T2/PC71BM blend film by means of the femtosecond time-resolved spectroscopy. Here, we systematically investigated Tan-dependence of the photovoltaic properties, i.e., short-circuit current (Jsc), open-circuit voltage (Voc), fill factor (FF), power conversion efficiency (PCE), internal quantum efficiency (ΦIQ), charge formation efficiency (ΦCF), domain size (L) and fullerene mixing (ΦFullerene,) in the polymer matrix in the OSCs based on films of F8T2/PC71BM blend. ΦCF is defined by nformed/nphoton, where nformed and nphoton are the numbers of the carriers formed at the D/A interface (include weakly bound state) and the absorbed photons, respectively. Absolute value of nformed was estimated by combination of the time-resolved and electrochemical spectroscopies. ΦCF is the same as the exciton quenching efficiency, if all the quenched excitons are converted to carriers. We found that ΦCF, ΦIQ and PCE are dominantly determined by ΦFullerene, indicating an essential role of the fullerene mixing in the polymer matrix on the carrier formation and transfer processes.

Results

Photovoltaic properties

First of all, let us survey the device parameter, i.e., Jsc, Voc, FF, PCE and ΦIQ against Tan (Table I). We fabricated OSCs based on films of F8T2/PC71BM (33 : 67 wt%) blend annealed for 10 min at Tan (Fig. S2). We measured current (J) – voltage (V) curve (Fig. S3) and incident photon-to-current conversion efficiency (IPCE) spectra (Fig. S4) .The magnitudes of Jsc and FF decrease with increase in Tan, while Voc remains nearly independent of Tan. As a result, PCE ( = Jsc × Voc × FF/I0, where I0 is the power density of the incident light) deceases with increase in Tan. The Tan-dependence of Jsc and FF is ascribed to several compounded factors, e.g., the domain size, carrier recombination process at the D/A interface and connectivity among the domains. We confirmed that the domain size (L) of the active layered is nearly independent of Tan (vide infra).

Table 1 Device parameters, i.e., Jsc, Voc, FF, PCE and ΦIQ of OSCs based on films of F8T2/PC71BM (33 : 67 wt%) blend annealed for 10 min at Tan. The error bars of Jsc, Voc, FF, PCE and ΦIQ were estimated from standard deviations of more than five OSC devices
Full size table

Domain structure as investigated by STXM image

Figure 1 shows STXM images of the F8T2/PC71BM blend films annealed at various Tan probed at 284.4 eV. The photon energy (284.4 eV) was at the π*-resonance absorption of the fullerene framework6,18. Therefore, the bright regions correspond to the fullerene-rich domains, while the dark regions the polymer-rich domains. We performed two-dimensional Fourier transformation to evaluate the length scale (L) of the fullerene domain. We regarded the local maxima of the Fourier component as L (Fig. S5). We found that L (~270 nm) is nearly independent of Tan.

Figure 1
figure 1

STXM image of F8T2/PC71BM blend films at 284.4 eV.

The bright regions correspond to the PC71BM-rich domains, while the dark regions the F8T2-rich domains.

Full size image

Fullerene mixing as investigated by STXM spectroscopy

To determine the fullerene mixing, we measured the carbon K-edge absorption spectra (φexp) at every 40 nm within the 2 μm × 2 μm image, i.e., 50 × 50 spectra19. We should be careful for evaluation of the molecular mixing since the STXM spectra is average along the depth direction. We investigated cross-sectional Plasmon loss image (Fig. S6) of the blend film annealed at 80°C. We confirmed that the polymer matrix passes completely through to the other side. That is, the fullerene mixing of the polymer matrix is accurately evaluated by the STXM spectroscopy. Unfortunately, we cannot accurately evaluated the molecular mixing of the fullerene domain, because it overlaps with the polymer domain in the depth direction.

Upper panel of Fig. 2 shows the averaged carbon K-edge absorption spectra of the polymer matrix against Tan. We observed extra bands at both sides of the main peak at 285 eV, as indicated by downward arrows. The bands are ascribed to the 1st and 3rd peaks of PC71BM (see the lower panel of Fig. 2). Their intensities gradually increases with decreases in Tan, indicating that the fullerene mixing increases with decreases in Tan. The magnitudes of ΦFullerene were evaluated by least-squares fitting of the φexp spectra with the linear combination of the F8T2 (φD) and PC71BM (φA) spectra, φcal = CDφD + CAφA. In the lower panel of Fig. 2, we show an example of the least-squares fitting. The linear combination (black thin curve) well reproduces the overall features of the φexp spectra. This indicates that the charge-transfer-type absorption at the D/A interface has negligible effects on the φexp spectra. In the spectral analysis, we select ten φexp spectra at every Tan at the central position of the polymer matrix to avoid the artificial mixing of the materials. The averages and standard deviations of CD and CA were evaluated. The ΦFullerene values were calculated by CA/(CD + CA) and are listed in Table II.

Table 2 Internal quantum efficiency (ΦIQ), carrier formation efficiency (ΦCF), carrier transfer efficiency (ΦCT = ΦIQ/ΦCF) and fullerene mixing (ΦFullerene) within the polymer matrix of OSCs based on films of F8T2/PC71BM (33 : 67 wt%) blend annealed for 10 min at Tan. The error bars of ΦCF were roughly evaluated from the signal/noise ratio of the femtosecond time-resolved spectra. ΦFullerene was evaluated by least-squares fitting of the observed spectra (φexp) with the linear combination of the F8T2 (φD) and PC71BM (φA) spectra, φcal = CDφD + CAφA. The averages and standard deviations of CD and CA were evaluated from ten φexp spectra at every Tan. The ΦFullerene values were calculated by CA/(CD + CA)
Full size table
Figure 2
figure 2

Carbon K-edge absorption spectra of the polymer matrix.

Upper panel shows the averaged absorption spectra against Tan. Lower panel shows an example of the spectral decomposition, which was performed by least-squares fitting of the observed spectra (φexp) with the linear combination of the F8T2 (φD) and PC71BM (φA) spectra, φcal = CDφD + CAφA.

Full size image

Carrier formation efficiency

We evaluate the absolute value of ΦCF by combination of the femtosecond time-resolved and electrochemical spectroscopies20. Solid curve in Fig. 3 is the differential absorption (ΔODEC) spectrum of electrochemically oxidized F8T2 neat film. The observed absorption at 1.8 eV is ascribed to the donor carrier. We investigated the spectral intensity (I1.8eV) at 1.8 eV against the hole-doping level (n) and determined the coefficient (αcarrier = 4.1 × 10−3 nm2) between I1.8eV and n (see Fig. S7). Circles in Fig. 3 is the differential absorption (ΔOD) spectra at 10 ps of the F8T2/PC71BM blend film. A sharp photoinduced absorption (PIA) is observed at 1.7 eV, whose profile is similar to that of the ΔODEC spectrum. We confirmed that the spectral profile is unchanged after 10 ps. In addition, the decay time of the PIA is very long ( = 300 ps). These observations indicate that the PIA is due to the donor carriers. We evaluated the coefficient (αphoton) between the spectral intensity (I1.7eV) at 1.7 eV and the number (nphoton) of the absorbed photons, with considering the reflection and transmission losses. The ΦCF values were calculated by αphoton/αcarrier and are listed in Table II.

Figure 3
figure 3

Differential absorption (ΔOD) spectra at 10 ps of F8T2/PC71BM blend film and differential absorption (ΔODEC) spectrum of electrochemically oxidized F8T2 neat film.

In ΔOD, the excitation photon energy and pulse energy are 3.1 eV and 27 μJ/cm2, respectively. In ΔODEC, the hole-doing level (n) is 0.55 nm−2.

Full size image

Correlation between parameters

We summarize in Fig. 4 the interrelation among ΦCF, ΦFullerene, L, ΦIQ and PCE in OSCs against Tan. We note that ΦCF has no relation with any losses after the carrier formation, e.g., the carrier recombination at the D/A interface or carrier trapping. In this sense, ΦCF is easier to interpret than the other efficiencies such as ΦIQ and PCE. ΦCF systematically decreases from 0.78 to 0.48 with increase in Tan. In the 2nd and 3rd panels, we plotted ΦFullerene and L. The Tan value has no effect on the domain size (L ~ 270 nm), but seems to suppress ΦFullerene in the polymer matrix. The Tan-dependence of ΦFullerene, however, is unclear due to rather large error bars. The error bars come from the bad signal/noise ratio in the featureless spectra. We carefully investigated Tan-dependence of the spectral profile around the fullerene peaks (284–287 eV) in the averaged carbon K-edge absorption spectra. We found that the relative intensities (I284.4eV) of the fullerene peak at 284.4 eV systematically decreases with increase in Tan (Fig. S8). This observation indicates that the fullerene mixing in the polymer matrix decreases with increase in Tan. Thus observed Tan-dependence of the fullerene mixing is reasonable, because the thermal annealing at higher Tan accelerates the phase-separation into more pure domains. Our analysis revealed that the fullerene mixing in the polymer matrix is advantageous for the efficient carrier formation. The decrease in ΦCF with Tan is responsible for the suppressed ΦIQ and PCE (bottom panel of Fig. 4).

Figure 4
figure 4

Interrelation among ΦCF, ΦFullerene, L, ΦIQ and PCE in OSCs against Tan.

Solid straight lines are the results of the lease-squares fittings. Error bars of ΦIQ and PCE are within the symbol size.

Full size image

Discussion

To investigate the morphology within the polymer matrix, we investigated cross-sectional TEM-S mapping image of the blend film annealed at 80°C (Fig. S9). The mapping clarifies the distribution of the F8T2 polymer in ∼ nm resolution. The mapping suggests that the F8T2 polymer matrix consists of the polymer clusters of several nm and the fullerene. Such a sub-structure well explains why the F8T2/PC71BM OSC shows high ΦIQ ( = 0.35 at 40°C) even though its domain size (L ~ 270 nm) is too large for exciton to reach the domain boundaries. According to this scenario, our observation, i. e., ΦCF, ΦIQ and PCE decreases as the fullerene mixing in the polymer matrix decreases, is interpreted as follows. With increase in Tan, the number (size) of the polymer clusters decreases (increases) within the polymer matrix, because the thermal annealing at higher Tan accelerates the phase-separation into more pure domains in every scale. As a result, the average fullerene concentration (ΦFullerene) within the polymer matrix decreases with Tan. The decrease in number and the increase in size of the polymer clusters lead to less donor exciton reaching to the D/A interface, to cause the suppressed ΦCF, ΦIQ and PCE. The fullerene mixing is a two-edged blade, because the sub-structure interface also function as recombination point for the photo-generated carriers21. Here, we define the carrier transfer efficiency (ΦCT) as ncollected/nformed, where ncollected is the number of the carriers collected as photocurrent (Fig. S1). Then, ΦCT is evaluated by ΦIQ/ΦCT (see Table II). We found that ΦCT decreases with decrease in Tan. The suppressed ΦCT is ascribed to the enhanced carrier recombination process at the sub-structure interface.

Summary

In summary, we systematically investigated the interrelations between photovoltaic properties, ΦCF and ΦFullerene in OSCs based on films of F8T2/PC71BM blend annealed at various Tan. We found that ΦCF, ΦIQ and PCE are dominantly determined by ΦFullerene, not by the size scale (L) of the domain. The TEM-S mapping image suggests that F8T2 polymer matrix consist of the polymer clusters of several nm and the fullerene. We interpreted the observation in terms of the polymer clusters within the polymer matrix: the decrease in number and the increase in size of polymer clusters lead to less donor exciton reaching to the D/A interface, to causes of the suppressed ΦCF, ΦIQ and PCE. Even though the stability of the large-scale morphology against Tan is specific to the F8T2/PC71BM combination, the annealing effects on the sub-structure are considered to be general to the polymer/fullerene blend film. Thus, complementary study of STXM, which probes quantitative molecular mixing in several tens of nm scale and TEM-S mapping, which probes molecular distribution in ∼ nm resolution, is effective to comprehend the photovoltaic properties of OSCs.

Method

Synthesis and characterization of the blend film

F8T2 was purchased from American Dye Source. The weight average molecular weight Mw, number average molecular weight Mn and polydispersity Mw/Mn were estimated to be 45,000, 13,000 and 3.4, respectively. PC71BM (purity 99%) was purchased from Solenne.

For the STXM measurements, F8T2/PC71BM blend films were transferred to a Si3N4 membrane. A bilayer film [poly(sodium 4-styrenesulfonate) (PSS)/blend film] was prepared by successive spin-coating of an aqueous solution of PSS and an o-dichlorobenzene (o-DCB) solution of F8T2/PC71BM (33 : 67 wt %). The thicknesses of the as-grown films were 71 nm. The films were annealed for 10 min at Tan = 40, 80, 110, 150, 190 and 240°C in a N2 glove box. Then, the bilayer film was cut into 1 × 1 mm2 pieces and the substrate was immersed for several minutes in distilled water to etch away the PSS film. Thus, we obtained small F8T2/PC71BM films floating on the distilled water. A piece of the floating film was scooped up with the Si3N4 membrane (50 nm in thickness and 500 × 500 μm2 in area).

For the time-resolved spectroscopy, F8T2/PC71BM blend films were prepared by spin-coating of an o-DCB solution of F8T2/PC71BM (33 : 67 wt %) on quartz substrates. The thicknesses of the as-grown films were 60–70 nm. The films were annealed for 10 min at Tan = 40, 80, 110, 150, 190 and 240°C in a N2 glove box. The atomic force microscope (AFM) image of the blend film annealed below 190°C revealed a periodic nanostructure of 300 nm in diameter (Fig. S2). The blend film annealed at 240°C is known to show macro-scale phase-separation into pure-F8T2 and pure-PC71BM domains.

Fabrication and characterization of the OSC

The OSCs were fabricated with a structure of indium tin oxide (ITO)/poly(3,4-ethylenedioxythiophene) (PEDOT):PSS (40 nm)/blend film/LiF (1 nm)/Al (80 nm). The patterned ITO (conductivity: 10 Ω/square) glass was pre-cleaned in an ultrasonic bath of acetone and ethanol and then treated in an ultraviolet-ozone chamber. A thin layer (40 nm) of PEDOT:PSS was spin-coated onto the ITO and dried in air at 110°C for 10 min on a hot plate. The substrate was then transferred to an N2 glove box and dried again at 110°C for 10 min on a hot plate. An o-DCB solution of F8T2:PC71BM (33 : 67 wt %) was subsequently spin-coated onto the PEDOT:PSS surface to form the active layer. The resultant substrates were then annealed at Tan = 40, 80, 110, 150, 190 and 240°C for 10 min in a N2 glove box. Finally, LiF (1 nm) and Al (80 nm) were deposited onto the active layer by conventional thermal evaporation at a chamber pressure lower than 5 × 10−4 Pa. The active area of the OSCs is 2 × 2 mm2. The J - V curves (see Fig. S3) were measured using a voltage - current source/monitor under AM 1.5 solar-simulated light irradiation of 100 mW/cm2 (Bunkou-keiki, OTENTO-SUN III). The IPCE spectra (see Fig. S4) was measured using a SM-250 system (Bunkou-keiki). The internal quantum efficiencies (ΦIQ) at 400 nm were estimated with considering the reflection loss.

STXM spectroscopy and analysis

The STXM measurement was performed using the compact STXM installed at the BL-13A beamline of the Photon Factory (PF), KEK. The details of the compact STXM are described in the literature19. The spatial resolution was 30–40 nm. The carbon K-edge absorption spectra (φexp) were measured at every 40 nm within the 2 μm × 2 μm image, i.e., 50 × 50 spectra. The molecular mixing was evaluated by least-squares fitting of the observed spectra (φexp) with the linear combination of the F8T2 (φD) and PC71BM (φA) spectra, φcal = CDφD + CAφA. We regard the absorption spectra of the F8T2 and PC71BM domains in the F8T2/PC71BM blend film annealed at 240°C as φD and φA, respectively. The coefficients, CA and CD, are determined so that the evaluation function, , becomes the minimum. The background constant component was subtracted so that φcal becomes zero at 280 eV. In the spectral analysis, we select the ten φexp spectra at the central position of the polymer matrix to avoid the artificial mixing of the materials. The averages and standard deviations of CD and CA were evaluated. The volume fractions of fullerene (ΦFullerene) were calculated by CA/(CD + CA).

Femtosecond time-resolved spectroscopy

The time-resolved spectroscopy was performed in a pump-probe configuration. In order to reduce the irradiation damage, the blend films were placed in N2 atmosphere. The pump pulse at 400 nm was generated as the second harmonics of a regenerative amplified Ti: sapphire laser in a β-BaB2O4 (BBO) crystal. The pulse width, repetition rate and pulse energy were 100 fs, 1000 Hz and 27 μJ/cm2 respectively. The frequency of the pump pulse was decreased by half (500 Hz) to provide “pump-on” and “pump-off” conditions. A white probe pulse (500–900 nm), generated by self-phase modulation in a sapphire plate was focused on the sample with the pump pulse. The spot sizes of the pump and probe pulses were 4.0 and 2.0 mm in diameter, respectively. The differential absorption (ΔOD) spectrum is expressed as −log(Ion/Ioff), where Ion and Ioff are the transmission spectra under the pump-on and pump-off conditions, respectively.

Electrochemical spectroscopy

The electrochemical spectroscopy was carried out in an optical two-pole cell with a pair of quartz windows. The electrochemical hole-doping was performed against Li metal in propylene carbonate (PC) solution containing 1 mol/L LiClO4. The F8T2 neat film was spin-coated on an ITO glass substrate from o-DCB solution and was dried in a N2 glove box. The thicknesses was 67 nm. The active area of the film was 2.25 cm2 and the reduction current was 100 nA. The voltage in the hole-doping process were 3.8 V vs. Li. The differential absorption (ΔODEC) spectrum of electrochemically oxidized film is expressed as −log(Idoped/Inon), where Idoped and Inon are the transmission spectra of the hole-doped and non-doped films, respectively.

The charge formation efficiency (ΦCF) was determined by combination of the time-resolved and electrochemical spectroscopies20. The former spectroscopy tells us the coefficient (αphoton) between ΔOD and nphoton, while the latter spectroscopy tells us the coefficient (αcarrier) between ΔODEC and n. Then, the ΦCF value is calculated by αphoton/αcarrier.