Understanding the correlation and balance between the miscibility and optoelectronic properties of polymer–fullerene solar cells

Institute of Materials for Electronics Friedrich-Alexander University Erlangen Erlangen, Germany. E-mail: chaohong.zh brabec@fau.de Institute for Problems of Chemical Physics Prospect 1, Chernogolovka, 142432, Russia Moscow Institute of Physics and Technolo Dolgoprudny, 141700, Russia Moscow State University, Faculty of Fundam GSP-1, 1-51 Leninskie Gory, Moscow, 11999 The Institute for Advanced Studies, Wuhan Institut de Sciences des Matériaux de Mullho 68051 Mulhouse, France Skolkovo Institute of Science and Technology Moscow, 143026, Russian Federation. E-ma Bavarian Center for Applied Energy Researc Erlangen, Germany † Electronic supplementary informa 10.1039/c7ta03505e Cite this: J. Mater. Chem. A, 2017, 5, 17570

In this study, we in-depth investigate the loss mechanism of polymer solar cells based on the amorphous dominated fullerene acceptor PyF5 as well as on the semi-crystalline fullerene acceptor FAP1.Increased ll factors are observed for PyF5-and FAP1-based solar cells with higher fullerene loadings, while the ll factor of PC 61 BM based devices is quite saturated aer certain fullerene loadings.Space-charge-limited current (SCLC) measurements are performed to study the charge carrier transport properties of pristine fullerenes and blends with various fullerene contents.The electron mobilities of these three pristine fullerenes are quite comparable but distinctly differing electron mobilities are found in BHJ composites.We further investigate the morphological evolution of the blend lms by means of in situ temperaturedependent grazing-incidence wide-angle X-ray scattering (GIWAXS).Combining our ndings and understanding, insights into the balance between miscibility and photovoltaic performance are given in a conclusive picture.Th has a strong absorption area between 500 and 800 nm.The photoluminescence (PL) peak of PTB7-Th appears at $830 nm.When blended with PC 61 BM (donor (D) : acceptor (A) ¼ 1.5 : 1), PyF5 (D : A ¼ 1 : 1 or 1 : 3), or FAP1 (D : A ¼ 1 : 1 or 1 : 3), the PL emission of PTB7-Th is signicantly quenched by over 98% for all blends.These phenomena suggest that the exciton dissociation/charge separation process is highly efficient for PTB7-Th:PyF5 and PTB7-Th:FAP1 with either high or low fullerene loadings.We further fabricate solar cells consisting of PTB7-Th as the donor, PC 61 BM, PyF5 or FAP1 as the acceptor with various D : A ratios.The solar cell architecture is shown in Fig. 1c.

Photovoltaic properties of amorphous and crystalline fullerene based solar cells
The optimized photovoltaic performance of PyF5 and FAP1 based devices is slightly lower than that of PC 61 BM based devices, owing to the slightly lower ll factor and J SC .The photovoltaic characteristics of PTB7-Th:fullerene with different donor : acceptor ratios are summarized in Fig. 2 and Table 1.It is important to point out that PC 61 BM instead of PC 71 BM was employed for reference devices to ensure rational comparison with the two C 60 -based fullerene derivatives.As observed in Fig. 2, when the acceptor loading is low (D : A ¼ 1 : 0.8), J SC , ll factor and power conversion efficiency (PCE) are much lower than the optimized device performance for each corresponding system.With increasing the fullerene loadings, the photovoltaic parameters of the three systems have a trend to increase.In PC 61 BM-based BHJ systems, PCE reaches a summit at a D : A ratio of 1 : 1.5.Further increasing the ratio beyond 1 : 1.5 only moderately increases the ll factor.In the case of PTB7-Th:PyF5 and PTB7-Th:FAP1, PCE peaks at a D : A ratio of 1 : 2. Further increasing the fullerene ratio slowly enhances the ll factor beyond 60% at a D : A ratio of 1 : 3. It is noticed that, at the D : A ratio of 1 : 3, when the ll factor of PyF5 and FAP1 based devices is above 60%, J SC and PCE are almost the same for all three systems; similar phenomena are observed for pDDP5T-2:PyF5 and pDDP5T-2:FAP1 solar cells (Fig. S1 and Table S1 †).

Carrier dynamics along with structural evolution
To nd out the correlation between the photovoltaic parameters, especially the FF, and the fullerene loadings, space-chargelimited current (SCLC) measurements were performed to study the charge carrier transport properties of fullerene:polymer blends for various fullerene loadings.The electron-only devices were constructed with the architecture of ITO/Al/active layer/Ca/ Ag for pristine lms and ITO/ZnO/active layer/Ca/Ag for blend lms.The electron mobility was estimated by tting the current-voltage curves according to the SCLC modied Mott- Gurney model.As shown in Fig. 3, Table S2 and Fig. S2, † pristine PyF5 and FAP1 demonstrate a high electron mobility of 4.0 Â 10 À3 cm 2 V À1 s À1 and 4.4 Â 10 À3 cm 2 V À1 s À1 , respectively, which are very similar to that determined for PC 61 BM (4.5 Â 10 À3 cm 2 V À1 s À1 ), suggesting that the electron transporting properties of the functional fullerenes are very comparable to the ones of PC 61 BM.However, the electron mobilities of PC 61 BM-based blends remain high ($10 À4 cm 2 V À1 s À1 ) in composites with up to 50 wt% polymer loading.Such high electron mobility in blends is essential to guarantee a high ll factor and a correspondingly high efficiency.In stark contrast, the electron mobility of PTB7-Th : PyF5 (1 : 1) and PTB7-Th : FAP1 (1 : 1) BHJ systems was signicantly reduced to the level of 10 À7 -10 À6 cm 2 V À1 s À1 .This is associated with the rather low FF observed for such low fullerene loadings (Table 1).When more PyF5 or FAP1 was added to the blends, the electron mobility was improved by two orders of magnitude to 8.8 Â 10 À5 cm 2 V À1 s À1 for PTB7-Th : PyF5 (1 : 3) and 9.3 Â 10 À5 cm 2 V À1 s À1 for PTB7-Th : FAP1 (1 : 3).This is in good agreement with the comparable J SC and PCE for all three systems.
We note that for PTB7-Th:PyF5 and PTB7-Th:FAP1 blends signicantly higher fullerene loadings are necessary to reach comparably high electron mobilities as for PTB7-Th:PC 61 BM blends.Surprisingly, we interpret these ndings to indicate that the crystallinity of a fullerene acceptor is not the sole material property dominating the electron mobility of a BHJ composite.Our studies demonstrate that the amorphous-dominated fullerene acceptor PyF5 has the same mobility as the semicrystalline fullerene acceptor FAP1.When blended with PTB7-Th, at low fullerene loadings, the electron mobility drops dramatically for both fullerene composites.Upon increasing the fullerene loading, the electron mobility increases for both the amorphous and the crystalline fullerene.This observation leads us to the conclusion that crystallinity per se is an important material parameter but probably not the decisive parameter guaranteeing good solar cell performance.This trend is best documented by Fig. 3, where the electron motilities of PTB7-Th:fullerene as a function of fullerene content are presented.The much distinct electron mobility delivered by the polymer:fullerene blends can be mainly attributed to the different microscale BHJ morphology.
In situ temperature-resolved GIWAXS was employed to acquire a deeper insight into the polymer:fullerene morphology.According to the literature, pristine PC 61 BM and PyF5 lms mainly reveal amorphous ngerprints while the FAP1 thin lm exhibits highly crystalline reections with narrow peaks.In Fig. S3, † the presence of an in-plane (100) peak and an out-of-plane (010) peak (polymer p-p stacking) indicates that the polymer chains preferentially form face-on orientation in pristine PTB7-Th.With S as axes, the position of the (100) and (010) peaks of polymer is at 0.048 and 0.255 ÅÀ1 corresponding to 0.30 and 1.60 ÅÀ1 with Q as axes (Q ¼ 2pS), respectively, which is in good agreement with the literature. 54Temperatureresolved GIWAXS measurements were performed in the temperature range from 25 to 150 C. At each temperature (70 C and 150 C, Fig. S4 †) the annealing process was performed for 5 minutes before the measurement.
Fig. 4 depicts the 2D GIWAXS patterns, the corresponding inplane and out-of-plane cuts, and the azimuthal distribution of the PTB7-Th (100) peak of three BHJ blends.PTB7-Th:PC 61 BM blends present the reections of the crystal phase of the  polymer and amorphous halo of PC 61 BM.Annealing leads to an increase of the crystalline polymer (100) and (010) peak intensity in both, in-plane and out-of-plane, directions as well as enhanced PC 61 BM aggregation (the distinction between the polymer p-p stacking and the halo of PC 61 BM aggregates can be found in Fig. S5 †).For PTB7-Th:PyF5 and PTB7-Th:FAP1 blends, aer annealing, the polymer out-of-plane (100) peak increases while the polymer in-plane (100) peak as well as the maximum of the azimuthal distribution of the (100) peak slightly decreases.From previous studies, it was found that the photoluminescence of PTB7-Th:PCBM blends signicantly increases aer thermal annealing, with a strong rise of the singlet emission of PTB7-Th, indicating a strong separation of the donor and acceptor phases.For both PTB7-Th:PyF5 and PTB7-Th:FAP1 systems, the PL spectra remain almost unchanged during thermal annealing, indicating that PCBM and PTB7-Th have a strong tendency to phase separate on a large scale while both PyF5 and FAP1 maintain their good intermixing. 53In summary, all three PTB7-Th:fullerene blends re-organize in the nano-scale during this short-time annealing; polymer packing and fullerene aggregates were observed to increase in PC 61 BM blends, which indicates a tendency to phase separate, while no such tendency was observed for PyF5-and FAP1-blends.

Morphological evolution of polymer-fullerene BHJ blends
It is essential to explore why the same polymer matrix (PTB7-Th or pDDP5T-2) can show such a distinctly different microstructure evolution for the one fullerene (PC 61 BM) over the other fullerenes (PyF5 and FAP1).Following this analysis we will identify the miscibility of the single components as one essential parameter inuencing photovoltaic performance.
In principle, there are three phases in OPV BHJ blends: pristine donor and acceptor domains, and a mixed domain; [55][56][57][58] the mixed domain should be as inter-mixed as possible to facilitate charge generation and dissociation; the pristine domains should be as pure and crystalline as possible in order to reduce recombination and lead to better charge transportation.Previous studies discussed even more complete pictures of the BHJ microstructure involving 5 or even more phases; 59-63 however, we will limit the discussion here to one amorphous mixed phase which allows us to simplify the introduction of the miscibility concept without limiting its generality.Previous research indicates that a certain degree of miscibility between the donor and acceptor is necessary to provide a sufficiently mixed polymer-fullerene domain which is important for charge generation and dissociation.Nevertheless, if the polymer donor and the fullerene acceptor have too ne inter-mixing, full percolation of the fullerene phase gets hampered and the electron mobility starts to break down, resulting in reduced ms products and worsened ll factors.][66][67] The rather low miscibility of PTB7-Th and PCBM typically does result in an unfavorable thin lm morphology.Advanced processing recipes relying on additives provide a path to microstructure optimization at rather low fullerene loading; 29 nevertheless, this optimized microstructure turns out to be metastable. 53The microstructural instability makes these composites vulnerable against external stress (e.g.9][70] We previously demonstrated that PyF5 and FAP1 are better miscible with PTB7-Th than PCBM, leading to superior thermal stability of the corresponding solar cells. 53In the current study, we nd that the PL signal of PTB7-Th is efficiently quenched by adding PCBM, PyF5 or FAP1 (Fig. 1b), indicating equally efficient charge dissociation; however the electron mobilities of PyF5and FAP1-based blends at a D : A ratio of 1 : 1 are two orders of magnitude lower than that of the PCBM-based blend, indicating decient charge transport.At a D : A ratio of 1 : 3, the electron mobilities as well as the photovoltaic performance of PyF5-and FAP1-based composites advance to a performance level comparable to that of the PCBM-based composite at a mixing ratio of 1 : 3. From the absorption spectra of thin lms (Fig. S6 †), the low polymer content in PTB7-Th : PyF5 (1 : 3) and PTB7-Th : FAP1 (1 : 3) lms results in reduced absorption and J SC , compared to the optimized PTB7-Th : PCBM (1 : 1.5) blend with the same layer thickness.[73][74] Balance between miscibility and optoelectronic properties 5][76] It is suggested that polymers with higher crystallinity require lower PCBM loadings to form effective percolation pathways.However, this hypothesis cannot fully explain the optimized PCBM loadings reported for multiple organic solar cells, such as MDMO-PPV:PCBM, PTB7-Th:PCBM or PffBT4T-2OD:PCBM.Both MDMO-PPV and PTB7-Th are dominantly amorphous polymers, but PTB7-Th requires far less PCBM than MDMO-PPV to attain optimized performance.The required fullerene loading for PTB7-Th is very similar to that of the highly crystalline/aggregated PffBT4T-2OD.Thus, we suggest considering that the D : A ratio in BHJ blends does not solely depend on the crystalline nature of the polymer donor but may be governed by other factors.This thought provoking impulse is further motivated by the data from this work where both, PyF5and FAP1-based, systems require more fullerene loadings than the PCBM-based system independent of the crystalline nature of fullerene acceptors.
It has been reported that the dimensionless Flory-Huggins interaction parameter (x 12 ) can be used for evaluating the miscibility between polymers and small molecules. 77,78As already described in previous publications, the interaction parameters can be determined as follows: 77,78 Fig. 5 Schematic of the morphological evolution of polymer-fullerene with (a) poor miscibility and (b) good miscibility.
This journal is © The Royal Society of Chemistry 2017 where v 0 is the lattice site volume and is dened by the smallest unit; d 1 and d 2 are the Hildebrand solubility parameters of the fullerenes and the polymer, respectively; R is the ideal gas constant (8.314 cm 3 MPa K À1 mol À1 ) and T is the temperature (298 K).
The Hildebrand parameters as determined by Hansen solubility parameters (HSP) are summarized in Table S3 † for a selection of relevant polymer donors and fullerene acceptors.In addition to taking HSP of polymers from the literature, we further determined the HSP for a number of polymers.According to eqn (1), the interaction parameter for various polymer-fullerene combinations was calculated and is summarized in Table S4.† The site volume v 0 in eqn (1) must be specied whenever discussing the interaction parameter as it is dened in terms of energy per site. 30When v 0 is xed in eqn (1), the value of (d 1 À d 2 ) 2 is proportional to the interaction parameter x 12 which is directly correlated with the polymerfullerene miscibility.The optimized acceptor : donor ratios of the corresponding polymer-fullerene solar cells were taken from the literature 3,7,11,13,30,[79][80][81][82][83] (Table S4 †).
Fig. 6 summarizes the acceptor : donor ratios for optimized devices as a function of the relative polymer-fullerene miscibility.Most surprisingly we nd a well expressed relation between the acceptor : donor ratio and the miscibility.Two separate regions are observed in Fig. 6.Systems with a relatively low miscibility, below the threshold, and a signicant tendency to phase separate perform best at low fullerene ratios of 1 : 1 or 2 : 1 (acceptor : donor), more or less independent of their crystallinity.On the other hand, composites with high miscibility, above the threshold, show a more expressed dependency on the interaction parameter and exhibit the best performance at high fullerene loadings of 3 : 1 to 5 : 1 (acceptor : donor).Although the miscibility of PC 61 BM and PC 71 BM was found to be different when blended with the same polymer, the difference is still below the threshold of miscibility as shown in Fig. 6.In this case, the optimized acceptor : donor ratio is expected to be very similar for a certain polymer.Fig. 6 is in good agreement with the observation that polymers having better miscibility with fullerenes require higher fullerene loadings to attain optimized photovoltaic performance, quite independent of their crystallinity.The correlation between the miscibility and optimized acceptor : donor ratio is more clear from Fig. S8, † where the acceptor : donor ratio is plotted logarithmically.
We want to accentuate a nal important aspect of miscibility.Polymer-fullerene BHJ composites with high miscibility are promising opportunities for enhanced thermal stability.
On the other hand, high fullerene loadings do result in weak absorption and reduced photo-current generation at equal layer thickness.This is a strong argument to further foster the development of highly absorbing, non-fullerene acceptors, [84][85][86] which have the potential to overcome this limitation.

Conclusion
In summary, we in-depth investigated the carrier transport loss mechanism for PTB7-Th:PCBM, PTB7-Th:PyF5 and PTB7-Th:FAP1 BHJ composites as a function of their microstructure and especially their miscibility.An obvious correlation between the microstructural compatibility of the polymer-fullerene components with the electron transport properties and the ultimate photovoltaic performance of BHJ solar cells was observed.Surprisingly, this interaction is independent of the crystalline nature of the fullerenes but rather depends on the miscibility of the polymer donor with the fullerene acceptor.We further found that the miscibility between the donor and acceptor dominates the optimized acceptor loadings in polymer-fullerene BHJ systems.BHJ composites with good polymer-fullerene miscibility require higher fullerene loadings than composites with a tendency to phase separate.It has to be carefully considered for future design of the donor and acceptor BHJ composites that the miscibility and the optoelectronic properties need to be well balanced in order to maximize the photovoltaic performance of organic BHJ solar cells.

Materials
The ZnO-nanoparticle dispersion in isopropyl alcohol was received from Nanograde.pDPP5T-2 (batch no.GKS1-001) was provided by BASF.PTB7-Th and PC 61 BM (99%) were purchased from One Material and Solenne BV, respectively.PyF5 and FAP1 were synthesized according to the literature. 48,87vice fabrication All solar cells were fabricated in an inverted structure of ITO/ZnO/ active layer/MoO x /Ag.The pre-structured ITO coated glass substrates were subsequently cleaned in toluene, acetone and isopropyl alcohol for 10 min each.Then 30 nm ZnO (Nanograde, N10) was doctor bladed on the ITO substrate and annealed at 85 C in ambient air.The PTB7-Th based active layer from a chlorobenzene : 1,8-diiodooctane (100 : 3) solution was spin coated on top of ZnO in a nitrogen atmosphere.The pDPP5T-2 based active layer from a chloroform : 1,2-dichlorobenzene (90 : 10) solution was bladed under ambient conditions.10 nm MoO x and 100 nm silver were deposited subsequently under 6 Â 10 À6 Torr by thermal evaporation through a shadow mask to form an active area of 10.4 mm 2 .For SCLC devices, a pristine fullerene layer was spin coated under an ambient atmosphere.The BHJ active layer was processed the same as solar cells.15 nm calcium and 100 nm silver were deposited subsequently under 6 Â 10 À6 Torr by thermal evaporation through a shadow mask to form an active area of 10.4 mm 2 .

General device characterization
The current-voltage characteristics of the solar cells were measured under AM 1.5G irradiation on an OrielSollA Solar simulator (100 mW cm À2 ).The light source was calibrated by using a silicon reference cell.All cells were tested in ambient air.
Film absorption was characterized with a UV-vis-NIR spectrometer Lambda 950 from PerkinElmer.

PL
Thin lm photoluminescence measurements were conducted under excitation from a 375 nm diode laser.The spectra were recorded with a Si-CCD attached to an iHR320 monochromator (Horiba).

Fig. 1a summarizes
Fig.1asummarizes the chemical structures of the polymer donor, PTB7-Th and three fullerene acceptors studied in this work.Fig.1bdepicts the normalized absorption of pristine PTB7-Th, PC 61 BM, Py5 and FAP1.The absorption region of the three fullerenes is mainly located at around 320 nm while PTB7-

Fig. 2
Fig. 2 (a) Current density-voltage (J-V) characteristics of PTB7-Th:fullerenes organic solar cells; (b) PCE and (c) fill factor of the corresponding organic solar cells as a function of D : A ratios.

Fig. 3
Fig. 3 SCLC electron motilities of PTB7-Th:fullerene as a function of fullerene content.The dashed lines are a guide to the eye.

.
To present the 2D GIWAXS patterns in coordinates S xy and S z , the original rectangular images were deformed aer recalculation in reciprocal space.White stripes on the images are the gaps between photoelectronic chips.The 1D proles represent the original distribution of intensity as a function of the wave vector modulus.

Table 1
Photovoltaic properties of PTB7-Th:fullerene organic solar cells