Blog
About

4
views
0
recommends
+1 Recommend
0 collections
    0
    shares
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      The effect of side-chain substitution and hot processing on diketopyrrolopyrrole-based polymers for organic solar cells†

      Read this article at

      Bookmark
          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

          Abstract

          The effects of cold and hot processing on the performance of polymer–fullerene solar cells are investigated for polymers designed to exhibit temperature-dependent aggregation in solution via second-position branched alkyl side chains.

          Abstract

          The effects of cold and hot processing on the performance of polymer–fullerene solar cells are investigated for diketopyrrolopyrrole (DPP) based polymers that were specifically designed and synthesized to exhibit a strong temperature-dependent aggregation in solution. The polymers, consisting of alternating DPP and oligothiophene units, are substituted with linear and second position branched alkyl side chains. For the polymer–fullerene blends that can be processed at room temperature, hot processing does not enhance the power conversion efficiencies compared to cold processing because the increased solubility at elevated temperatures results in the formation of wider polymer fibres that reduce charge generation. Instead, hot processing seems to be advantageous when cold processing is not possible due to a limited solubility at room temperature. The resulting morphologies are consistent with a nucleation-growth mechanism for polymer fibres during drying of the films.

          Related collections

          Most cited references 19

          • Record: found
          • Abstract: found
          • Article: found
          Is Open Access

          Aggregation and morphology control enables multiple cases of high-efficiency polymer solar cells

          Conventional inorganic solar cells can achieve high efficiencies but are produced through complicated, costly processes. The desirability of lower costs is driving the development of several third-generation solar technologies. Among these, polymer solar cell (PSC)1 2 3 4 5 6 technology is an excellent example of low-cost production because PSCs can be produced using extremely high-throughput roll-to-roll printing methods similar to those used to print newspapers7. PSCs also offer several other advantages: vacuum processing and high-temperature sintering are not needed, and no toxic materials are used in the end product. Most importantly, a tandem cell architecture6 8 9 10 can be easily implemented with PSCs and has proven to improve PSC efficiency by ~40–50% (refs 6, 8). As PSCs are two-component, donor–acceptor material systems, it is generally important to control the morphology of the donor:acceptor blends and to find an optimal materials combination with excellent optical and electronic properties. In the last few years, record-efficiency PSCs were achieved with only three donor polymers (which all belong to a specific polymer family based on fluorinated thieno[3,4-b]thiophene, for example, PTB7) that are, furthermore, constrained to be used with a specific fullerene, PC71BM, to achieve their best performance11 12 13. In general, the morphologies14 15 16 and thus performance of state-of-the-art donor polymers (for example, PTB7 (refs 11, 17) and PBDT-DTNT18) are sensitive to the choice of fullerene and replacing PC71BM with another C60-based or non-PCBM fullerene decreases PSC efficiency to 6-7% (refs 11, 18, 19, 20) The dominant role of PC71BM places serious constraints on PSC material development, because the properties of the polymers must be precisely matched with fixed targets set by PC71BM. As tandem PSCs require two sets of perfectly matching polymer/fullerene materials, the constraint on their development is compounded. It has thus been pointed out that it is crucial to have the flexibility of being able to use different fullerenes and more generally to remove material constraints to achieve tandem PSCs with 15–20% efficiency envisioned by Brebac and colleagues9 10 21. The development of polymer:fullerene material systems that are morphologically insensitive to fullerene choice will remove these material constraints, and greatly accelerate material development for single-junction and tandem PSCs10 22. Another important fundamental issue for the PSC field is how to control the morphology of polymer:fullerene blends to achieve the best PSC performance. There is likely more than one near-optimum PSC morphology. The famous PTB7 family donor polymers enabled one type of the near-optimum PSC morphology, as high external quantum efficiencies (EQEs~80%) have been reported for PTB7-based cells11. However, the PTB7-based PSC materials and devices have certain limitations. Besides the sensitivity of the choice of fullerenes, another important limitation for PTB7 family polymers is that they cannot perform well when relatively thick active layers (~300 nm) are used in the PSC device. Thick-film PSCs are important for the industrial application of PSCs, and thick films should also further increase the absorption strength of the solar cell and thus cell efficiency. The reason why PTB7 does not perform well in thick-film PSCs is partially owing to the relatively low hole transport ability (space charge limited current (SCLC) mobility ~6 × 10−4 cm2 V−1 s−1; ref. 17) related to the low crystallinity of the PTB7 polymer. There has been also evidence that high purity of the polymer domain may be an important factor to achieve efficient thick-film PSCs14 23 24. The PTB7-based materials systems are characterized by relatively impure polymer domains25, which could be a reason why these polymers do not perform well in thick-film PSCs. Clearly, there is a need for new materials systems that explore a different ‘near-optimum’ PSC morphology in order to achieve thick-film PSCs that have comparable or higher efficiencies than state-of-the-art PTB7 materials systems. In the following, we report the achievement of high-performance (efficiencies up to 10.8% and fill factors (FFs) up to 77%) thick-film PSCs based on three different donor polymers and 10 polymer:fullerene combinations, all of which exhibit efficiencies higher than the previous state of the art. In contrast to state-of-the-art PTB7-based materials systems, the high PSC performances in this report are achieved via the formation of an ‘optimum PSC morphology’ that contains highly crystalline, sufficiently pure, yet reasonably small polymer domains. The high polymer crystallinity and thus excellent hole transport ability, combined with sufficiently pure polymer domains, are the main reasons why the PSCs exhibit high FFs and efficiency even when the active layer is 300 nm thick. Importantly, this ipso facto near-perfect morphology is controlled by the temperature-dependent aggregation behaviour of the donor polymers during casting and is insensitive to the choice of fullerenes. Taking advantage of the robust polymer:fullerene morphology enabled by the three donor polymers, many non-traditional fullerenes are also used. Traditional PCBMs, the most dominant fullerenes in PSCs, are out-performed by several other non-traditional fullerenes, clearly indicating the benefits of exploring different fullerenes and the robust morphology formation. Comparative studies on several structurally similar polymers reveal that the 2-octyldodecyl (2OD) alkyl chains sitting on quaterthiophene is the key structural feature that causes the polymers’ highly temperature-dependent aggregation behaviour that allows for the processing of the polymer solutions at elevated temperature, and, more importantly, controlled aggregation and strong crystallization of the polymer during the film cooling and drying process. The branching position and size of the branched alkyl chains are critically important in enabling an optimal aggregation behaviour. With our approach, PSC production is no longer constrained by the use of a single fullerene or by a very thin active layer. Our aggregation and morphology control approach and polymer design rules can be applied to multiple polymer:fullerene materials systems and will allow the PSC community to explore many more polymers and fullerene materials and to optimize their combinations (energy offsets, bandgap and so on) under a well-controlled morphological landscape, which would greatly accelerate the materials and process development towards improved PSCs. Results PSC device performance Among the three donor polymers, we developed that achieved power conversion efficiency>10%, we first focus on poly[(5,6-difluoro-2,1,3-benzothiadiazol-4,7-diyl)-alt-(3,3′′′-di(2-octyldodecyl)-2,2′;5′,2′′;5′′,2′′′-quaterthiophen-5,5′′′-diyl)], PffBT4T-2OD (Fig. 1a). PffBT4T-2OD is a material that enables six cases of high-efficiency (9.6–10.8%), high FF (73–77%) and thick-film (250–300 nm) PSCs (Table 1) when combined with traditional PCBM and many non-traditional fullerenes (Fig. 1b). A typical J–V plot of a PffBT4T-2OD:fullerene PSC is shown in Fig. 1c, with EQE spectra shown in the inset. The benefits of thick-film PSCs are obvious. The thick cell exhibits 10–20% higher EQE values, and the effective absorption bandwidth of a thick PSC can be increased as the result of a ~20 nm red-shift of the ‘leading, low energy edge’ of a PSC’s EQE spectrum. Combined, these account for a ~30% increase in short circuit current (J SC). Taking advantage of PffBT4T-2OD’s excellent aggregation properties (as delineated further below), we synthesized more than a dozen known or new fullerene derivatives (Fig. 1b) to find the best acceptor match for PffBT4T-2OD. All of these fullerenes form similar morphologies with PffBT4T-2OD and can produce PSCs with high efficiencies in the range of 8.6–10.8% (Table 1 and Supplementary Table 1). The best efficiency (10.4%) in the C60 family was achieved by PC61PM (Fig. 1b), and the most commonly used C60-based fullerene, PC61BM, is not the best match for PffBT4T-2OD. Polymer crystallinity and hole mobility Grazing incident wide-angle X-ray diffraction (GIWAXS)26 reveals the molecular packing and orientational texture of pure PffBT4T-2OD and PffBT4T-2OD:fullerene blend films. Both exhibit a high degree of molecular order, as evidenced by strong lamellar (100), (200) and even (300) reflection peaks and, more importantly, a large (010) coherence length (GIWAXS 2D patterns shown in Fig. 2a,b and Supplementary Fig. 1). The (010) coherence length (that is, extent of ordering) of PffBT4T-2OD:PC61PM blend films was calculated using Scherrer analysis27 to be ~8.5 nm, which corresponds to ~24 π-stacked copolymers. In contrast, the (010) coherence length of PTB7:PC61BM, for example, is only ~2 nm (ref. 16). Owing to the high crystallinity and preferential face-on orientation of polymer domains, relatively high SCLC hole mobility of 1.5–3.0 × 10−2 cm2 V−1 s−1 were obtained for various PffBT4T-2OD:fullerene blend films in a hole-only diode device configuration (Supplementary Fig. 2). The importance of mobility for good FF was recently illustrated23. Polymer:fullerene domain size and average domain purity In addition, resonant soft X-ray scattering14 15 25 28 29 30 (R-SoXS; Fig. 2c) and atomic force microscopy (AFM; Supplementary Fig. 3) analysis reveals that the various PffBT4T-2OD:fullerene films all exhibit multi-length scale morphologies with reasonably small median domain sizes of ~30–40 nm, which is similar to previous cases of high-performance polymers16 25. R-SoXS can also reveal the average composition variations, which are indicative of the average purity of the polymer and fullerene regions as well as a possible third phase of polymer-rich domains26 31. An annealing sequence on PffBT4T-2OD:fullerene blends revealed that the non-annealed devices presented here exhibited almost 90% average purity compared with the asymptotic limit (Fig. 3a,b), which corresponds to an unusually low residual concentration of 3.2% fullerene averaged over all PffBT4T-2OD domains in the film as measured by X-ray microscopy (Fig. 3c)25 32 33. In general, PSCs with significantly impure polymer phases exhibit detrimental bimolecular charge recombination when the polymer film is too thick, whereas pure phases can help to minimize recombination23. These morphological data show that PffBT4T-2OD can form a polymer:fullerene morphology containing highly crystalline and sufficiently pure yet reasonably small polymer domains. Note that PTB7-type polymers have been the best donor polymer in PSCs for the past few years. By its very nature of high performance in thin films, PTB7 can form a ‘near-optimum’ PSC morphology characterized by relatively low molecular ordering, relatively low hole mobilities and impure polymer domains25. PffBT4T-2OD exhibits high molecular ordering (‘crystallinity’), high hole mobilities and purer polymer domains, which appears to be a different ipso facto ‘near-optimum’ PSC morphology. Although PTB7 enabled great thin-film PSC performance, PffBT4T-2OD offers high performance even in thick-film PSCs owing to the high mobility of the highly ordered and sufficiently pure polymer domains it forms. Morphology control via temperature-dependent aggregation We attribute PffBT4T-2OD’s excellent performance and robust morphology to its significant temperature-dependent aggregation behaviour that can be exploited during device fabrication. The UV-Vis absorption spectra exhibit a marked red-shift when a low concentration PffBT4T-2OD solution in 1,2-dichlorobenzene (DCB) is lowered from 85 to 25 °C (Fig. 1d). At elevated temperature, PffBT4T-2OD is well dissolved and disaggregated. At progressively lower temperatures, a strong 0-0 transition peak at ~700 nm emerges with significant strength at 25 °C, indicating strong aggregation of the polymer chains in solution at that temperature. Note that the absorption spectrum of the 25 °C solution of PffBT4T-2OD is almost identical to that of the optimized PffBT4T-2OD solid film (Fig. 1d), which is observed to be highly crystalline by GIWAXS. Consequently, devices are always cast from warm solutions (60–80 °C) of PffBT4T-2OD, which then aggregates during the cooling and film-forming process. To understand details of PffBT4T-2OD’s aggregation behaviour during the film-forming process, the critical π–π molecular ordering ((010) coherence length and intensity of the (010) peak) is determined with X-ray diffraction (XRD) for a series of PffBT4T-2OD films spun at different rates. As shown in Fig. 2d and Supplementary Table 2, the π–π ordering decreases markedly with increasing spin rates. As PffBT4T-2OD exhibits a strong yet progressively evolving aggregation property, the extent of PffBT4T-2OD’s aggregation depends upon temperature and concentration changes, the film drying time and the kinetics of aggregation. During a slow spin process (for example, 700 r.p.m.), it takes a relatively long time for the solution and film to dry, during which the temperature of the substrate and the wet film also decreases significantly. When using an ultra-fast rate (for example, 5,000 r.p.m.), however, the solvent evaporates more quickly, which results in kinetically quenched, poorly ordered films, whereas slow spin rates provides PffBT4T-2OD sufficient time to aggregate and to form crystalline polymer domains with large coherence lengths. Importantly, studies for PffBT4T-2OD pure films and PffBT4T-2OD blend films with two different fullerenes yield similar trends (Supplementary Table 2 and Supplementary Fig. 4), demonstrating that the aggregation of PffBT4T-2OD is insensitive to the presence of fullerenes. Not surprisingly, high substrate temperatures were found to have a similar effect to fast spin rates. PffBT4T-2OD:fullerene films prepared with fast spin rates/high substrate temperatures show a decrease in the 0-0 transition peaks and a pronounced shift in the 0-0 transition energy in their UV-Vis absorption spectra, indicative of significant disorder (Fig. 2e). The corresponding hole-only and PSC devices fabricated using high spin rates and high substrate temperature also exhibit markedly decreased hole mobilities (3.1 × 10−3 cm2 V−1 s−1; Supplementary Table 3) and PSC efficiencies (3.6%; Supplementary Table 4). These morphological, spectroscopic and electric data demonstrate that PffBT4T-2OD’s morphology is mainly controlled by the progress of its aggregation during the film-casting process until the film is dry, which locks-in the length scale of the morphology. PffBT4T-2OD’s strong yet well-controllable aggregation property allows for convenient optimization of processing conditions that led to a near-ideal polymer:fullerene morphology that is insensitive to the choices of fullerene. This approach of controlling the extent of polymer aggregation during a warm solution casting process is different from the common processing protocol of PTB7 family polymers that are typically processed at room temperature. Discussion The key structural feature of PffBT4T-2OD that enables its pronounced, yet gradual temperature-dependent aggregation (Fig. 1d) is the second-position branched alkyl chains (2OD) on a quaterthiophene (4T-2OD). To elucidate this aspect, we contrast PffBT4T-2OD with two structurally very similar polymers. These two polymers have the same backbone but their alkyl chains are branched at the first or third side-chain carbon atom (for ease of comparison, these two polymers are named as PffBT4T-1ON and PffBT4T-3OT; Fig. 4a). In contrast to PffBT4T-2OD, PffBT4T-1ON is disaggregated at 85 °C, and more importantly, also disaggregated at 25 °C (Fig. 4b). As a result, PffBT4T-1ON cannot aggregate easily during the film-forming process, leading to films with poor crystallinity (GIWAXS pattern shown in Fig. 4c) and thus PSC devices of only ~0.6% efficiency. PffBT4T-3OT exhibits the other extreme, showing excessive aggregation at both 25 and 85 °C. During our attempt to process PffBT4T-3OT, the PffBT4T-3OT solution quickly becomes a gel (Fig. 4d) even before the start of spin casting. These comparisons indicate that PffBT4T-1ON’s alkyl chains cause too much steric hindrance, which results in poor aggregation and crystallinity. PffBT4T-3OT’s alkyl chains provide too little steric hindrance that makes aggregation of PffBT4T-3OT too strong even at 85 °C and makes it difficult to process. PffBT4T-2OD’s second-position branched alkyl chains offer an optimal tradeoff that allows for controllable aggregation of PffBT4T-2OD during the film-forming process. More discussions on the impact of alkyl chain branching positions are provided in Supplementary Note 1. Several structurally similar donor polymers containing quaterthiophene substituted with second-position branched alkyl chains were reported in the literature34 35 36 37, including a recent report in which a polymer with longer alkyl chains, FBT-Th4(1,4) (named as PffBT4T-2DT for simplicity in this paper, structure shown in Fig. 4a), achieved 7.64% efficiency34. The difference of PffBT4T-2OD and PffBT4T-2DT devices can be understood from the following results. R-SoXS (Fig. 4e) and prior AFM studies34 show that the average domain size of PffBT4T-2DT:fullerene films is too large (~100 nm). The R-SoXS data furthermore show that the average purity of the polymer/polymer-rich domains in an PffBT4T-2DT:fullerene film is ~87% of that in PffBT4T-2OD:fullerene film. Lower purity of average polymer/polymer-rich domains can result in significant recombination for thick-film PSC devices and thus lower performance14 23 24. Regarding molecular ordering, the two-dimensional GIWAXS mapping of PffBT4T-2DT:PC71BM films34 shows that PffBT4T-2DT:fullerene films exhibit weak laminar packing peaks, which are significantly weaker than those of PffBT4T-2OD:fullerene films. The smaller degree of laminar packing of PffBT4T-2DT is consistent with the lower average purity of PffBT4T-2DT polymer domains compared with that of PffBT4T-2OD’s, as more fullerene is expelled by the crystalline polymer domains of PffBT4T-2OD. Lastly, the absorption coefficient of PffBT4T-2DT is lower than that of PffBT4T-2OD owing to longer alkyl chains that does not contribute to light absorption (Supplementary Fig. 5). These studies show that the branching position and the size of the alkyl chains are critically important in obtaining the optimal aggregation properties of PffBT4T-2OD. Insufficient aggregation (for example, PffBT4T-1ON) and unnecessarily long alkyl chains (for example, PffBT4T-2DT) resulted in low crystallinity and/or impure polymer domains. Excessive aggregation (for example, PffBT4T-3OT) makes the processing and aggregation difficult to control. Similar to recent observations32 36 38, the molecular weight of PffBT4T-2OD has a significant impact on its aggregation property and performance. Lower molecular weight (M n=16.6 kDa, M w=29.5 kDa) batches of PffBT4T-2OD exhibit weaker aggregation and thus lower efficiency (7.7%) than the high molecular weight (M n=47.5 kDa, M w=93.7 kDa) PffBT4T-2OD batches (Supplementary Table 5 and Supplementary Fig. 6). The impacts of polymer molecular weight on PSC performances are discussed in details in the Supplementary Note 2. Following the rationale described above, we synthesized two other polymers (poly[(2,1,3-benzothiadiazol-4,7-diyl)-alt-(4′,3′′-difluoro-3,3′′′-di(2-octyldodecyl)-2,2′;5′,2′′;5′′,2′′′-quaterthiophen-5,5′′′-diyl)] (PBTff4T-2OD) and poly[(naphtho[1,2-c:5,6-c′]bis[1,2,5]thiadiazol-5,10-diyl)-alt-(3,3′′′-di(2-octyldodecyl)-2,2′;5′,2′′;5′′,2′′′-quaterthiophen-5,5′′′-diyl)] (PNT4T-2OD); Fig. 1a) with significantly different polymer backbones but with similar arrangements of 2OD alkyl chains. Both PBTff4T-2OD and PNT4T-2OD exhibit significant temperature-dependent aggregation behaviour that leads to processing and morphology control and thus efficiency (including >10% for thick-film PSCs; Table 1, Supplementary Table 1 and Supplementary Fig. 7) comparable to those achieved by PffBT4T-2OD-based PSCs. R-SoXS and AFM studies confirmed that the polymer domain size of these two new polymers are similar to that of PffBT4T-2OD (30–40 nm). XRD characterization of PBTff4T-2OD:fullerene and PNT4T-2OD:fullerene films also showed strong (010) π–π stacking peaks that are similar to those observed for PffBT4T-2OD. Note that PNT4T-2OD also significantly outperforms its analogue polymer with 2-decyltetradecyl (2DT) alkyl chains35, providing another example that supports the critical importance of the size of the alkyl chains. The synthesis, characterization and device performance of PBTff4T-2OD and PNT4T-2OD are described in detail in the Supplementary Information (Supplementary Fig. 7, Table 1 and Supplementary Table 1). Although second-position branched alkyl chains are a well-known structural motif and have been previously used on quaterthiophene-based polymers, previous work did not utilize a polymer with the most suitable alkyl chains nor were warm-casting methods used that optimally harnessed aggregation. They thus failed to reveal the connections between chemical structure, polymer aggregation during warm processing, morphology formation, polymer crystallinity and consequently PSC performance. Our study uncovered a new approach of aggregation and morphology control enabled by a structural feature (2OD alkyl chain) that is seemingly simple and commonly known, yet has surprisingly profound impact on PSC performances. The wide ranging applicability of our morphology control approach is supported by the three polymers and over 10 polymer:fullerene combinations that all yielded similar blend morphology and high-efficiency thick-film PSCs. Furthermore, the aggregation behaviour as observed by UV-Vis might serve as a useful screening tool to identify materials that yield good devices when cast from warm solutions. Note that the chemical structures of the three donor polymers presented in the paper are distinctively different from previous state-of-the-art PTB7 family of polymers12 13 17. The PTB7 family polymers consist of an electron deficient fluorinated thieno[3,4-b]thiophene unit and a benzodithiophene unit with alkoxy, alkylthienyl or alkylthiothienyl substitution groups. The three polymers in this paper consist of an electron deficient unit (either difluorobenzothiadiazole or benzothiadiazole or naphthobisthiadiazole) combined with a quaterthiophene unit with two 2OD alkyl chains sitting on the first and fourth thiophenes. The difference in the chemical structures caused different aggregation properties, based on which different processing protocols are used. While PTB7 family polymers do not exhibit a strong temperature-dependent aggregation property and are often processed from room temperature solutions, the 4T-2OD based polymers are processed from warm solutions to utilize their temperature-dependent aggregation property so that the morphology and extent of molecular ordering can be explicitly controlled during casting. To summarize, we report that exquisite control of aggregation results in high-performance thick-film PSCs for three different donor polymers and 10 polymer:fullerene combinations, all of which yielded efficiencies higher than the previous state of the art (9.5%). The common structural feature of the three donor polymers, the 2OD alkyl chains on quaterthiophene, causes a temperature-dependent aggregation behaviour that allows for the processing of the polymer solutions at moderately elevated temperature, and more importantly, controlled aggregation and strong crystallization of the polymer during the film cooling and drying process. This results in a near-ideal polymer:fullerene morphology (containing highly crystalline, preferentially orientated, yet small polymer domains) that is controlled by polymer aggregation during casting and thus insensitive to the choice of fullerenes. The branching position and size of the branched alkyl chains are critically important in enabling a well-controllable aggregation behaviour. Unnecessarily long alkyl chains (for example, 2DT) cause several detrimental effects including weaker laminar stacking, poorer absorption properties and less pure polymer domains. Our structural design rationales and aggregation and morphology control approach offer a new route to achieve high-performance thick-film PSCs that cannot be obtained from previous state-of-the-art material systems. Given that the field and record performance in the last few years has been mostly dominated by a single system (PTB7 family with PC71BM), the 10 material systems and three polymers based on a single and simple design feature presented here point to a plethora of possible materials combinations that should further improve the performance. Our approach will allow the PSC community to explore many more polymers and fullerene materials and to optimize their combinations (energy offsets, bandgap and so on) under a well-controlled morphological landscape that would greatly accelerate the materials and process development towards improved PSCs. Methods X-ray diffraction XRD data were obtained from a PANanalytical XRD instrument (model name: Empyrean) using the parallel beam mode that is recommended by the instrument manufacturer to characterize thin-film samples. All XRD samples were spin cast on Si substrates to avoid strong scattering background of glass substrates. To rule out the effect of substrate properties on the crystallinity of polymer film samples, we also investigated polymer films on Si/ZnO substrates and found that the polymer films have similar scattering profiles (Supplementary Fig. 8) on these two types of substrates (Si/ZnO and Si). The polymer crystallinity is thus rather insensitive to the surface properties of the substrates. More details of XRD characterizations are provided in Supplementary Note 3. Cyclic voltammetry Cyclic voltammetry was performed in an electrolyte solution of 0.1 M tetrabutylammonium hexafluorophosphate in acetonitrile, both working and counter electrodes were platinum electrode. Ag/AgCl electrode was used as the reference electrode; the Fc/Fc+ redox couple was used as an external standard (Supplementary Fig. 9 and Supplementary Table 6). UV-Vis absorption UV-Vis absorption spectra were acquired on a Gary 50 UV-Vis spectrometer. All film samples were spin cast on ITO/ZnO substrates. Hole-only device The hole mobility were measured using the SCLC method by using a device architecture of ITO/V2O5/PffBT4T-2OD (300 nm)/V2O5/Al by taking current–voltage curves and fitting the results to a space charge limited form, where the SCLC is described by: Where ε 0 is the permittivity of free space, ε r is the dielectric constant of the polymer, μ is the hole mobility, V is the voltage drop across the device and L is the thickness of the polymer. The dielectric constant ε r is assumed to be ~3, which is a typical value for conjugated polymers. GIWAXS characterization GIWAXS measurements were performed at beamline 7.3.3 at the Advanced Light Source (ALS)39. Samples were prepared on Si substrates using identical blend solutions as those used in devices. The 10 keV X-ray beam was incident at a grazing angle of 0.11°–0.15°, which maximized the scattering intensity from the samples. The scattered X-rays were detected using a Dectris Pilatus 1 M photon counting detector. Resonant soft X-ray scattering R-SoXS transmission measurements were performed at beamline 11.0.1.2 at the ALS30. Samples for R-SoXS measurements were prepared on a PSS modified Si substrate under the same conditions as those used for device fabrication, and then transferred by floating in water to a 1.5 × 1.5 mm, 100-nm thick Si3N4 membrane supported by a 5 × 5 mm, 200 μm thick Si frame (Norcada Inc.). Two dimensional scattering patterns were collected on an in-vacuum CCD camera (Princeton Instrument PI-MTE). The beam size at the sample is ~100 μm by 200 μm. The composition variation (or relative domain purity) over the length scales probed can be extracted by integrating scattering profiles to yield the total scattering intensity. The purer the average domains are, the higher the total scattering intensity. Owing to a lack of absolute flux normalization, the absolute composition cannot be obtained by only R-SoXS. In order to get a sense of how pure the domains are, we annealed the PffBT4T-2OD/fullerene blend at 130 °C for different length of time, 0, 10, 20, 40 and 120 min. The unannealed sample exhibits very pure domains, that is, almost 90% of the saturated value. AFM characterization AFM measurements were performed by using a Scanning Probe Microscope-Dimension 3100 in tapping mode. All film samples were spin casted on ITO/ZnO substrates. Photoluminescence quenching measurements Photoluminescence spectra were measured on samples on ITO/ZnO substrates upon excitation of a 671-nm laser beam. The PL quenching efficiency of PffBT4T-2OD was estimated from the ratio of the PL intensity of a PffBT4T-2OD:fullerene film sample to that of the PffBT4T-2OD control sample. (Supplementary Fig. 10) Solar cell fabrication and testing Pre-patterned ITO-coated glass with a sheet resistance of ~15 Ω per square was used as the substrate. It was cleaned by sequential sonications in soap DI water, DI water, acetone and isopropanol for 15 min at each step. After ultraviolet/ozone treatment for 60 min, a ZnO electron transport layer was prepared by spin coating at 5,000 r.p.m. from a ZnO precursor solution (diethyl zinc). Active layer solutions (D/A ratio 1:1.2) were prepared in CB/DCB (1:1 volume ratio) with or without 3% of DIO (polymer concentration: 9 mg ml−1). To completely dissolve the polymer, the active layer solution should be stirred on a hot plate at 110 °C for at least 3 h. Before spin coating, both the polymer solution and ITO substrate are preheated on a hot plate at ~110 °C. Active layers were spin coated from the warm polymer solution on the preheated substrate in a N2 glovebox at 800 r.p.m. to obtain thicknesses of ~300 nm. (The spin casting of high-performance PSC films is described in Supplementary Note 4 in details. The processing of PNT4T-2OD also requires the use of a metal chuck as described in Supplementary Note 4. High-temperature and high spin rate samples are described in Supplementary Note 5). The polymer/fullerene films were then annealed at 80 °C for 5 min before being transferred to the vacuum chamber of a thermal evaporator inside the same glovebox. At a vacuum level of 3 × 10−6 Torr, a thin layer (20 nm) of MoO3 or V2O5 was deposited as the anode interlayer, followed by deposition of 100 nm of Al as the top electrode. All cells were encapsulated using epoxy inside the glovebox. Device J–V characteristics was measured under AM1.5G (100 mW cm−2) using a Newport solar simulator. The light intensity was calibrated using a standard Si diode (with KG5 filter, purchased from PV Measurement) to bring spectral mismatch to unity. J–V characteristics were recorded using a Keithley 236 source meter unit. Typical cells have devices area of ~5.9 mm2, which is defined by a metal mask with an aperture aligned with the device area. EQEs were characterized using a Newport EQE system equipped with a standard Si diode. Monochromatic light was generated from a Newport 300 W lamp source. One of our best cells was sent to an accredited solar cell calibration laboratory (Newport Corporation) for certification, confirming an efficiency of 10.36±0.22%, with V OC=0.7743±0.0077 V, I SC=0.00079±0.00001 A, area=0.0425±0.0001, cm2, FF=72.0±1.5 (Supplementary Fig. 11). Author contributions Y.L. synthesized PffBT4T-2OD; J.Z. designed PNT4T-2OD, synthesized 5,10-Dibromonaphtho[1,2-c:5,6-c']bis[1,2,5]thiadiazole and carried out AFM measurements; Z.L. synthesized PBTff4T-2OD; J.Z., H.L., Y.L., H.H. and Z.L. synthesized fullerenes; C.M., H.H., K.J. and H.Y. fabricated and tested PSC devices; W.M. carried out GIWAXS and R-SoXS measurements and analysis; K.J. carried out XRD analysis; H.L. synthesized PNT4T-2OD; H.A. supervised GIWAXS and R-SoXS work, and helped design experimental protocols; H.Y. conceived and directed the project; H.A. and H.Y. wrote the paper with input from all authors who reviewed the final paper. Additional information How to cite this article: Liu, Y. et al. Aggregation and morphology control enables multiple cases of high-efficiency polymer solar cells. Nat. Commun. 5:5293 doi: 10.1038/ncomms6293 (2014). Supplementary Material Supplementary Information Supplementary Figures 1-11, Supplementary Tables 1-6, Supplementary Notes 1-5, Supplementary Methods and Supplementary References
            Bookmark
            • Record: found
            • Abstract: found
            • Article: found
            Is Open Access

            Competition between recombination and extraction of free charges determines the fill factor of organic solar cells

            Among the parameters that characterize a solar cell and define its power-conversion efficiency, the fill factor is the least well understood, making targeted improvements difficult. Here we quantify the competition between charge extraction and recombination by using a single parameter θ, and we demonstrate that this parameter is directly related to the fill factor of many different bulk-heterojunction solar cells. Our finding is supported by experimental measurements on 15 different donor:acceptor combinations, as well as by drift-diffusion simulations of organic solar cells in which charge-carrier mobilities, recombination rate, light intensity, energy levels and active-layer thickness are all varied over wide ranges to reproduce typical experimental conditions. The results unify the fill factors of several very different donor:acceptor combinations and give insight into why fill factors change so much with thickness, light intensity and materials properties. To achieve fill factors larger than 0.8 requires further improvements in charge transport while reducing recombination.
              Bookmark
              • Record: found
              • Abstract: found
              • Article: found
              Is Open Access

              Determining the optimum morphology in high-performance polymer-fullerene organic photovoltaic cells

              In organic photovoltaic cells (OPVs), the highest power conversion efficiencies are achieved in bulk heterojunction (BHJ) devices in which the active layer is made by blending electron donor and acceptor materials1 2 3 4. Light absorption in each of these materials generates excitons that have to diffuse to an interface with the other material, where there is an offset in energy that can split them into electron–hole pairs. These pairs then have to overcome Coulombic attraction in order to dissociate into free-charge carriers and to be extracted to the corresponding electrodes via percolation pathways. Critical to all of these processes is the morphology of the blend—how the two materials mix together or segregate. It is a pressing challenge in the science of OPV materials and devices to be able to determine the optimum morphology of blends5 6 7 8 9 10 11 12 13 14 15. Recently, single layer OPVs with a record power conversion efficiency of 9.2% were reported16 using a blend of the fullerene derivative [6,6]-phenyl-C71-butyric acid methyl ester (PC71BM) and a polymer with alternating units of thieno[3,4–b]thiophene and benzodithiophene, commonly denoted as PTB7 (chemical structures shown in Fig. 1a) and so consequently this system is of great interest. In this Article, we investigate the nanoscale morphology of such blends, their ultrafast photophysics and performance in devices. We use advanced scanning probe techniques and also exciton diffusion in PC71BM as a nanoscale ruler to determine the domain sizes. We investigate the exciton dissociation dynamics with a femtosecond time resolution, which allows us to get information on the domain composition. Blends were prepared with and without the high boiling point additive diiodooctane (DIO). Addition of DIO improves the device’s internal quantum efficiencies by a factor of 2, which we determine is mainly because of improved charge extraction. We show that blends prepared with DIO have a fibre-like morphology of polymer-rich and fullerene-rich domains. We suggest that a concentration gradient between these domains helps in the dissociation of geminate pairs by charge diffusion and that charge extraction is enhanced with a fibrous morphology. We conclude that a fibrous morphology of polymer-rich and fullerene-rich domains is highly advantageous for OPVs. Results Photovoltaic response Absorption spectra of PTB7 and PC71BM films as well as the 1:1.5 blend of the two are shown in Fig. 1b, along with the photoluminescence (PL) spectra of PC71BM and PTB7. OPVs were prepared with and without the additive DIO as described in Methods and the device characteristics are shown in Fig. 1c,d and current–voltage curves in Supplementary Fig. S1. The electrode structures were not optimized for the highest possible power conversion efficiencies. It should be noted that the efficiency of our device is comparable to the reported value of 7.4% with the same architecture7. The world record 9.2% efficient device was fabricated in an inverted geometry with optimized electrode work-functions that increased the photocurrent. However, the active layer deposition process of the record cell (blend ratio, solvent, use of additive, spin-coating speed) was identical to the deposition parameters we have used to make films. Thus, the morphology and photophysics of our blend films, and the conclusions we deduce, are transferable to devices with the highest reported efficiencies. The external quantum efficiency (EQE) increased by a factor of 2 in devices prepared with DIO and is consistent with previous reports7 8. The internal quantum efficiency (IQE)—corrected for the number of photons absorbed in the active layer—is determined by optical modelling of the device stack and also increases by a factor of 2 in devices prepared with DIO. The IQE is >90% with DIO in the polymer absorption region and is slightly lower in the region of fullerene absorption. Blends prepared without additive We have combined time-resolved PL (TRL) and atomic force microscopy (AFM) to provide complementary information on phase separation of the fullerene and polymer. Previous studies reported large domains of fullerene of 100–200 nm in diameter7 8 10 in blends prepared without additive. With scanning electron microscopy (SEM) we indeed see large domains (Fig. 2a). To extract more information, we made a score through the film to enable the profile of the domains to be imaged. Visible in the same SEM image is a top ‘skin’ layer of material that can be clearly seen to have been pulled back from the domains at the score, enabling us to observe the material beneath. Then we removed 30–40 nm of material by plasma-ashing to take away the top skin and measured the AFM topography (Fig. 2b), which shows the large 100–200 nm diameter domains but with fine structure inside them. In an AFM-phase image (Fig. 2c), the fine structure is better resolved and shows small sphere-like objects of 20–60 nm diameter. We have then employed TRL measurements to learn about the composition of the blend. TRL intensity from PC71BM is proportional to the exciton population in the fullerene, and thus quenching of PC71BM emission by PTB7 is able to tell us about the segregation of molecules. Neat films of PTB7 are fluorescent; however, it is important to note that even with selective excitation of PTB7 in the blend (λ ex=670 nm) no polymer fluorescence was observed, even at timescales as short as 2 ps; thus with 400 nm excitation we look at solely PC71BM emission. Previously, Kazaoui et al. 17 have reported that with high-energy photoexcitation it is possible to excite intermolecular charge transfer states in neat films of C70, directly producing charge pairs from an absorbed photon. To confirm that we were not generating significant amounts of charge pairs directly, we measured the PL excitation spectra and compared it with the absorption spectrum of the film (see Supplementary Fig. S2) and find that at 400 nm, ~90% of the absorbed photons produce excitons, with ~10% producing charge pairs. The discrepancy between absorbed photons and emitted photons only becomes significant below 375 nm. The decay of PL is observed on femtosecond (Fig. 2d) and picosecond–nanosecond (Fig. 2e) timescales. Decay of the PL in the blend is faster than in films of only PC71BM (Fig. 2e), indicating dynamic quenching by the polymer. The ultrafast decay of PC71BM emission in Fig. 2d indicates a 330-fs quenching with 0.8 amplitude of fullerene excitons by PTB7. This can occur by hole transfer to PTB7 or resonant energy transfer (RET) to the polymer. The Förster radius for RET from PC71BM to PTB7 was calculated to be 2.17 nm from the spectral overlap as described in the Methods. From this value we conclude that the fast quenching has to be short ranged in nature (<10 nm) and is consistent with a large number of small spheres of fullerene, presenting a large surface area for quenching by the polymer. The slower part of the TRL decay (Fig. 2e) is mediated by exciton diffusion to an interface with the polymer and can give us the size of the pure fullerene spheres if the exciton diffusion coefficient in PC71BM is known. Measurement of the exciton diffusion coefficient is described in the next section, and gives a value of 1.6 × 10−4 cm2 s−1 in PC71BM. This enables us to describe the dynamics observed in Fig. 2e based on the information obtained in Fig. 2a–c. The most obvious model is spheres of PC71BM surrounded by a matrix of PTB7. We assume that excitons diffuse inside the spheres by a random walk and are instantaneously quenched at the surface with PTB7. The diffusion equation was solved analytically with these boundary conditions following Crank18, and the number of fullerene excitons in a sphere is given by: where a is the radius of the sphere and D the diffusion coefficient which we set to the measured value stated above. Equation 1 was multiplied by the fluorescence intensities in the absence of quencher and convoluted with the instrument response to fit the PL decay. The simulated decay using a pure 150-nm-diameter sphere of PC71BM is displayed in Fig. 2e as the green dashed line and is slower than the measured decay. The best fit with a single sphere model gives a fullerene sphere with a diameter of 60 nm (Fig. 2e, solid blue line), agreeing well with the AFM data (Fig. 2b,c). The good agreement between the directly observed morphology of domains 20–60 nm in diameter with AFM and the deduced fullerene domain size with TRPL indicates that using the decay of PL in combination with the exciton diffusion coefficient in the bulk to model the morphology is valid. To further test this validity, we turn our attention to the ‘pureness’ of the small spheres of PC71BM. We can modify our model by including a small percentage of PTB7 uniformly distributed inside the 60-nm-fullerene spheres (a fuller description is provided in the Methods). We find that even a very small amount (0.2 wt%) of PTB7 mixed into the spheres renders the decay faster than is measured, shown in Fig. 2e as the orange dotted line, giving strong evidence that the small 60 nm spheres are comprised of pure fullerene. To explore this further, we have performed differential scanning calorimetry (DSC) on similar blends of PTB7:PC71BM. With DSC, we can monitor the phase transitions of the fullerene in the blend. Such phase transitions (for example, melt temperature) would be modified if significant amounts of the fullerene were in a mixed phase with the polymer, indicating purity. The melt temperature of the fullerene would be the best feature to observe and was found to be at 315–320 °C (see Supplementary Figs S3,S4), consistent with published values19 20. This fullerene melt cannot be observed in the blend with DSC because of overlapping melt crystallization in PTB7 at this temperature (Supplementary Fig. S3). Information on the fullerene purity can instead be derived from the melt crystallization peak (205 °C)20, where there is no overlapping polymer phase transition. Looking at the neat fullerene and the blend, the same melt crystallization peak is observed (see Supplementary Figs S3,S4), thus supporting that the small spheres are composed of pure fullerene. The overall determined morphology is schematically shown in Fig. 2f, where small pure spheres of fullerene 20–60 nm in diameter in a polymer matrix agglomerate into 150-nm-diameter fullerene-rich domains. These domains sit in a finely mixed phase matrix, and a finely mixed phase skin blankets everything. This morphology is substantially different from those that have been reported using transmission electron microscopy7 10 and X-ray scattering8 11. In a report by Chen et al. 11, a hierarchical morphology was observed, with crystallites of a few nanometres forming heterogeneity with a 75-nm-length scale and creating domains of 175 nm size. The work concludes, however, that the domain purity peaks at only ~45%. We are able to determine that the 60-nm-fullerene domains are 100% pure, as indicated in Fig. 2e. In addition, the work by Chen et al. 11 uses PC61BM, which has a different miscibility with polymers21 22 23 from PC71BM, and thus may produce different morphologies in the blend. The report by Collins et al. 8 suggests pure fullerene domains of 200 nm in size—which they state would limit charge carrier generation. We show that these large domains are not pure and are in fact comprised of smaller pure fullerene spheres 60 nm in size surrounded by polymer. We suggest that the top skin as well as the small-length scales make these structures difficult to detect with previously used techniques. As noted above, two processes can contribute to quenching of fullerene fluorescence in the blend: charge pair formation and RET to PTB7. Even though RET can occur, the fact that we do not observe PTB7 fluorescence in the blend indicates that fast exciton splitting into charge pairs occurs no matter which process takes place. The efficiency of charge pair formation due to light absorbed on the fullerene can be derived from the fluorescence decay using: where τ Neat is the average fluorescence lifetime of the fullerene in an inert blend with PMMA (600 ps) and τ Blend is the average fluorescence lifetime in the photovoltaic blend with PTB7 (67 ps). Hence, for the blend without additive φ CP=0.1+(0.89 × 0.9)=0.9—that is, 90% of absorbed photons generate charge pairs, where the 0.1 is from directly formed charge pairs in the fullerene as discussed earlier. This number contrasts with an IQE of 40% in the fullerene absorption region, suggesting that only 45% of the initially generated charges are extracted. This is an extraction efficiency under short-circuit conditions and takes into account both geminate and non-geminate recombination losses—that is, all losses that occur between charge generation and charge extraction at short-circuit conditions. Photoconductive–AFM of this blend in Supplementary Fig. S5 shows low photocurrent in the regions between the fullerene-rich domains, which indicate poor electron extraction from polymer-rich domains. This is consistent with an IQE of ~50% in the spectral range of polymer absorption without additive, despite total quenching of the polymer fluorescence. We conclude that even though good exciton dissociation is achieved, large fullerene-rich domains limit charge extraction and OPV performance. Exciton diffusion in PC71BM Exciton diffusion in PC71BM is important in its own right, as the fullerene is used in many efficient OPV blends with the polymers PCDTBT24, PSBDTBT25 26 as well as PTB7 (refs 7, 16, 27, 28). To determine the exciton diffusion coefficient, we have dispersed electron donor-functionalized molecules into the PC71BM film and measured the time-resolved luminescence (TRL) of PC71BM to tell us about the degree of PL quenching. Small known concentrations of DPP-NMe2 were added to PC71BM films. The synthesis of DPP-NMe2 is described in the Methods; synthesis route is shown in Fig. 3a; and absorption spectrum shown in Fig. 3b. Increasing concentrations of the quencher caused a reduction in the excited state lifetimes of the blend relative to the pristine film as measured by TRL. In this low-concentration regime, the molecules of DPP-NME2 are assumed to be randomly distributed throughout the film. In the simplest model, these molecules act as quenching sites so that an exciton in the PC71BM will decay if it encounters DPP-NMe2 in the course of its random walk. Mathematically, this can be expressed as the solution to the diffusion equation in spherical coordinates, using the boundary condition that the concentration of the mobile species is held at zero when it gets to within a critical distance, R c, from a quenching site. This relation defines the rate of quenching of the diffusing species and is known as the Smoluchowski equation29. There is spectral overlap between the emission of PC71BM and the absorption of DPP-NMe2 (see Fig. 3b), which means that it is possible for some quenching at a distance away from the quencher via resonance energy transfer, rendering our simply collisional boundary conditions unphysical. It was found, however, that the very low PLQY of PC71BM means that the Förster radius is relatively short and is hence in the regime that has been calculated by Klein et al. 30 to be well approximated by the Smoluchowski equation. Thus, the rate of change of population of excitons can be described according to equation 3: where N(t) is the population of excitons, c q is the concentration of quenching sites in the films, D is the exciton diffusion coefficient, R AD is a distance of closest approach, inside which the exciton is quenched more rapidly than can be resolved via picosecond TRL measurements. k f(t) is the rate at which the excitons undergo fluorescence decay and was calculated from a 2-exponential fit to the PL emission from the pristine PC71BM film. Equation 3 was solved analytically, convoluted with the instrument response and then fitted to the time-resolved fluorescence decays of films with seven different low concentrations of the DPP-NMe2 quencher as shown in Fig. 3c, using D and R AD as the only fitting parameters. This generated a value of 1.5 nm for the distance of closest approach and an exciton diffusion coefficient in PC71BM of 1.6 × 10−4 cm2 s−1. An upper limit on the exciton diffusion coefficient can also be estimated from the characteristic time the exciton takes to undergo the first hopping step in its diffusion. The initial hop can be observed in excitons that move from a higher-energy chromophore to a lower one because the energy migration is exhibited in a spectral shift in the fluorescence to longer wavelength at longer time. Figure 3d shows the time-resolved PL in the spectral region 640–690 nm—defined as the blue side of the steady-state PC71BM PL spectrum, along with dynamics in the spectral region 690–740 nm, around the steady-state PL peak—both wavelength regions are indicated in Fig. 3b. On the blue side, a fast decay with a time constant of 7 ps is observed, while on the PL peak no decay is recorded, and instead a small rise-time of 7 ps is required for a good fit. The differences in dynamics at these two wavelengths indicate that the fast decay is not related to a loss of oscillator strength or electronic reconfiguration, but rather is consistent with spectral relaxation because of exciton hopping. Consequently, a time for the first hop of 7 ps can be deduced. Exciton diffusion is expected to slow down with time because excitons relax to low energy sites and the 7-ps-hopping time can be used to determine the diffusion coefficient’s initial value. By making the simplifying approximation that the PC71BM molecules lie on a cubic lattice, the diffusion coefficient can be determined using29: where R c-c is the centre to centre distance and τ hop is the characteristic hopping time. Using a density of a similar material, PC61BM, reported31 to be 1.3 g cm−3, the distance between chromophores on a cubic lattice is calculated as 1.3 nm. This gives the initial diffusion coefficient of 3.6 × 10−4 cm2 s−1. This is twice higher than the effective long-time value determined by the volume quenching technique. A similar value has been reported32 of D=2 × 10−4 cm2 s−1 for the other fullerene derivative, PC61BM. The one-dimension (1-D) diffusion length (L D ) can be calculated using the relation: Using the values of D and τ measured, LD =3.1 nm. Blends prepared with additive Solvent additives have been shown to greatly enhance device performance in a number of OPV blends7 8 33. Past studies have also shown that DIO increases intermixing of fullerene derivatives with PTB7 and suggested that proximity of the polymer to pure fullerene aggregates are important for efficient charge separation and transport10 11. Here we show that DIO dramatically improves the extraction of photogenerated charges, which can be explained by the formation of narrow elongated fibre-shaped fullerene-rich and polymer-rich domains, with a continuous change in composition observed by photocurrent mapping implying a concentration gradient between the two materials. First, we can obtain information on the degree of mixing between the two materials using TRL. The PL decay of the blend prepared with DIO at 710 nm is shown in Fig. 4a as filled circles and it displays very fast quenching of fullerene fluorescence by the polymer. The fit to a three-exponential function gives a decay time constant of 100 fs for 0.8 fraction of the fullerene, a slower 700 fs component (of fraction 0.15) and a slow component of 100 ps for 0.05 fraction—mis-fits showing the poor fitting that is achieved when any of these values deviate significantly from our best fit are shown in Supplementary Figs S6–S8. While the exponential fitting of the PL decay has no absolute physical meaning, it can tell us directly about the timescale of the dissociation of excitons into charge pairs in the blend, where the very fast decay of exciton emission agrees with the reported charge formation on a sub-picosecond timescale34. This qualitatively suggests that the fullerene molecules are in close proximity to the polymer and that the fullerene excitons do not need to transverse regions of fullerene via exciton diffusion to reach the PTB7, in contrast to the blend without DIO where more fluorescence persists for longer as the fullerene excitons diffuse to the polymer. PTB7 fluorescence in neat films has a peak at 810 nm; therefore, it is possible to distinguish between fullerene and polymer fluorescence spectrally, especially because the PTB7 radiative rate is almost twenty times that of the fullerene (0.44 versus 0.025 ns−1, as described in the Methods). Shown in Fig. 4a is ultrafast emission at 760 nm (open circles), where the PTB7 fluorescence intensity is ~0.65 of its peak intensity, and the dynamics are identical to emission at 710 nm. This indicates that all the emission we detect is from the fullerene and no polymer emission is detected—that is, the PTB7 excitons are dissociated extraordinarily quickly (<100 fs)—and suggests that polymer chains are intimately mixed with fullerene, agreeing with previous morphological studies7 8 9. The charge pair generation efficiency from absorbed photons in the blend with DIO is calculated using equation 2, and experimental values for τ Neat=220 ps and τ Blend=5.2 ps give a charge pair formation efficiency, φ CP=0.98. The IQE in devices prepared with DIO is 80% in the fullerene absorption region and 90% in the polymer absorption region. Since the charge generation is nearly unity, IQE can be taken as a measure of the charge extraction efficiency at short-circuit conditions, including all losses after charge generation in both cases. Intimate mixing will give large interfacial areas for efficient charge formation; however, charge pairs formed at the donor–acceptor interface are Coulombically bound and will recombine geminately35 36 37 38 39 unless they are extracted by continuous networks of polymer and fullerene. Consequently, we must look at the morphology to explain the efficient charge extraction. Measuring the surface topography with AFM of the as spun film (Fig. 4b) and after plasma-ashing that removes ~20 nm of the material (Fig. 4c) shows similar topographies, while a profile of the blend with SEM (Fig. 4d and Supplementary Figs S9–S12) shows no obvious lateral structure, indicating that morphologies of the surface and of the bulk film are not substantially different. This contrasts with morphology in the blend without DIO. Recently it has been shown40 41 42 43 44 that photocurrent mapping with photoconductive–AFM (pc-AFM) gives high resolution and high-contrast morphological information and we have therefore used this technique to study blend morphology. A pc-AFM image of the PTB7:PC71BM blend prepared with DIO and acquired with a tip voltage of +3 V and 670 nm red light excitation is shown in Fig. 4e and a zoomed area in Fig. 4f. Dark current in the same area was at least five times lower than with 670 nm light. Details of the pc-AFM experiments and photocurrent maps with negative tip bias as well as maps at different positive tip bias are given in the Methods and Supplementary Figs S13,S14. Figure 4g shows typical photocurrent–voltage curves of high and low current regions. Without an applied bias, photogenerated electrons cannot flow to the gold tip owing to a Schottky barrier but they can be extracted to the indium tin oxide (ITO), while holes can be extracted to the gold tip, generating a negative photocurrent. With a positive tip bias higher than the open circuit voltage (V OC) of ~0.7 V, electrons overcome the Schottky barrier and are extracted to the tip, whereas holes flow to the ITO and the photocurrent is positive. The absolute value of the negative photocurrent is about 5–10 times lower for the same voltage, which can be explained by a barrier to extract photogenerated holes from the tail of the density of states of the polymer into the gold tip. The 670-nm light preferentially excites the polymer; therefore, high photocurrent is likely to correspond to polymer-rich regions. This is confirmed by the observation that photocurrent maps obtained at negative tip bias correlate with positive tip bias maps (see Supplementary Fig. S13) as well as confirming that the measured quantity is photocurrent. Consequently, lower photocurrent is from fullerene-rich regions of the sample. The first thing to notice in the photocurrent map is that the domains of high and low photocurrent are smaller and more elongated than those without the additive, which indicates that polymer-rich and fullerene-rich domains have a high aspect ratio, with their lengths being longer than their widths. The regions of the blend with the highest and lowest photocurrents are narrow and resemble elongated thin fibre-like strands or swirling ‘brushed fibres’, with the close-up image shown in Fig. 4f indicating that these richest domains are typically 10–50 nm wide and 200–400 nm long. The photocurrent values show a gradual change going from high to low current regions, as indicated in Supplementary Fig. S15. The same regions of high photocurrent are observed in the range of +1.5 to +6 V (see Supplementary Fig. S14), which indicates that photocurrent contrast is not affected by charge extraction; thus, the photocurrent contrast is dependent on the polymer concentration as it is absorbing the 670-nm tip light. Therefore, a continuous change of photocurrent in the pc-AFM map implies that the mixing ratio of the polymer and fullerene also changes gradually. The interdiffusion of the polymer and fullerene to form mixed phases has been observed in a number of OPV blends45 46 47 48 49 and can have an important role in device performance. Previous studies have not reported any sort of fibre-like morphology in these systems, instead observing hierarchical morphology11 little different than without additive when PTB7 is blended with PC61BM. When PC71BM was blended with PTB7, Collins et al. 8 were able to identify that on a length scale larger than ~100 nm that the fullerene purity varied from 52 to 68%, while the average fullerene domain size was measured to be a distribution centred around 30 nm, ranging from 6 to 200 nm. This domain size was unable to determine any spatial organization of the domains, and so our work observing the elongated fibre-like alternating phases of polymer-rich and fullerene-rich domains enables the lengths observed by Collins et al. 8 to be placed in a spatial context. It should be noted that while the pc-AFM image shows the morphology in-plane, and that in a working device charge transport will be perpendicular to the plane (towards the contacts), our understanding is that the concentration gradient of each material that forms between the polymer-rich and fullerene-rich regions helps the dissociation of geminate charge pairs by diffusion as positive/negative charge diffuses preferentially towards the polymer-rich and fullerene-rich domains, respectively. In pc-AFM, the charges are extracted by a strong electric field, whereas in working OPVs the built-in field is much weaker and charge diffusion has an important role in separation of Coulombically bound charge pairs39. Directional charge diffusion driven by a concentration gradient is expected to lead to more dissociation than a random walk. The separated charges are then transported in the fibre-like polymer-rich and fullerene-rich domains with low probability of recombination. The overall morphology is thus beneficial for exciton dissociation into a charge pair, geminate pair dissociation into free charges and charge extraction, all of which are consistent with the observed high IQE of 80–90%. Rather than simply being an agent to ensure that the materials are well mixed, the solvent additive creates an environment where the optimum morphology for a working device is created. Discussion The results presented have enabled significant new nanoscale morphology of the high-performance PTB7:PC71BM blends to be observed for the first time, determining how the morphology of the blend affects the efficiency of charge carrier generation, extraction and the quantum efficiency of OPV devices. We show that blends prepared from solutions without any processing additive form large (100–200 nm) fullerene-rich domains comprised of small 20–60 nm pure fullerene spheres. The exciton dissociation efficiency is high (90%) in these blends but the IQE is only 40–60%, limited by charge extraction. This indicates that large fullerene- and polymer-rich domains give poor charge extraction and high recombination, leading to low quantum efficiencies. By using the additive DIO, narrow and extended fibre-like polymer-rich and fullerene-rich domains are created, with fibres 10’s of nanometres wide and 100’s of nanometres long for both materials. Exciton dissociation in the blends prepared with DIO is very fast, with 95% of exciton decay amplitude occurring on a sub-picosecond timescale. The fibre-like polymer-rich and fullerene-rich domains form part of a continual change in the polymer-fullerene mixing ratio when going from one to another, implying a concentration gradient that is favourable for charge pair dissociation and minimizes geminate recombination, thus giving high extraction efficiency (80%) of photogenerated charges. Consequently, our study of the PTB7:PC71BM system tells us about the optimum morphologies that are required, and in this system created, for efficient OPV devices. Although our study has looked specifically at PTB7:PC71BM blends, parallels and contrasts in the morphology can be drawn with other high-performance OPV blends. Charge transport modelling indicates the importance of continuous phases of the donor and acceptor materials and good connectivity to the appropriate electrodes, whereas the domain size has been found to be less important50. The extended fullerene-rich and polymer-rich domains observed in our work give efficient charge extraction, indicating that they are well connected, despite being only 10–20 nm wide. We note that fibrous morphology has been observed in P3HT- (ref. 51), PCDTBT- (ref. 24) and PCPDTBT (ref. 52)-based blends with PCBM. However, fibres were observed only in the polymer phase. We see that not only the polymer but also the fullerene show this kind of spatial organization, giving bi-continuous networks for both electron and hole transport. Short-range ordering of PTB7 has been explored with grazing-incidence X-ray diffraction10 and indicates that in the blend ~20% of the polymer can be considered crystalline. We are therefore left to consider that bundles of extended PTB7 chains form a template that helps self-organization of fullerene molecules around them, forming fibrous strands of both the donor and acceptor materials, which are advantageous for efficient OPV performance. Owing to the spatial resolution of pc-AFM, it is also possible that both polymer-rich and fullerene-rich domains contain regions of a few nanometres in width of pure material, which are too small for us to observe. The optimum morphology in PTB7:PC71BM is substantially different from previous generations of BHJ blends and suggests that the properties of PTB7 are different from other conjugated polymers used in OPVs and can be exploited in future material development. Methods Sample preparation PTB7 was purchased from 1-Material and had a molecular weight of 92,000 Da with a polydispersity of 2.6. PC71BM of 99% purity was purchased from Solenne. 1,8-DIO was purchased from Fluka and chlorobenzene (HPLC grade) from Sigma Aldrich. Samples of the blend were prepared with parameters as described in Liang et al. 7 and He et al. 16 In short, 10 mg of PTB7 and 15 mg PC71BM were dissolved in 1 ml of chlorobenzene at 50 °C with gentle stirring in a nitrogen-filled glovebox. After letting the solution cool, blends without additive were then spun directly on to clean fused silica disks at 1,000 r.p.m. in the glovebox. For the samples with DIO, 3% by volume of DIO was added to the solution and left to stir for 5 min prior to spin-coating with the same parameters. Time-resolved PL Measurements were performed on two setups; femtosecond data were acquired with upconversion spectroscopy (FOG100 from CDP systems), while picoseconds–nanosecond data were acquired with a synchroscan streak camera (C6860 from Hamamatsu). The excitation in both cases was the frequency doubled output of a Ti:sapphire laser, giving 400 nm 100 fs full-width half-maximum pulses at 80 MHz. Samples for streak camera measurements were held under an active vacuum of ~10−5 mbar. Samples for upconversion measurements were encapsulated by sealing with a fused silica window in a nitrogen-filled glovebox and were rotated during measurements, with no degradation visible over the length of the scan. Device fabrication ITO-coated glass substrates (15 Ω per square) from Xin Yan Technology Ltd were masked and etched in hydrochloric acid (37%) for 20 min. The mask was removed and the substrates cleaned by sonication in deionized water, acetone and isopropanol. The substrates were then dried with nitrogen and treated in an oxygen plasma asher for 5 min. Poly(3, 4-ethylenedioxythiophene:poly(styrenesulfonate) (PEDOT:PSS)(Clevios AI4083) was spin-coated at 4,000 r.p.m. The PEDOT:PSS-coated ITO substrates were annealed on a hotplate at 120 °C for 20 min before being placed in a nitrogen-filled glovebox and PTB7:PC71BM spin-coated on top with the same conditions as noted above. The substrates were then inserted into an evaporator for top electrode deposition. An ~20-nm calcium layer and an ~200-nm aluminium cathode were thermally evaporated at a pressure of 2 × 10−6 mb. Immediately after top electrode deposition, the devices were removed from the evaporator and encapsulated with a UV optical adhesive and a glass coverslip. Devices were then removed from the glovebox, masked and characterized under an illumination intensity of 100 mW cm−2 in air using a K.H. Steuernagel AM1.5G solar simulator and a Keithley 2400 source-measure unit. The illumination intensity was verified with an NREL-calibrated monosilicon detector and a KG-5 filter. EQE measurements were obtained with an incident photon to charge carrier efficiency setup, which consists of an NPL-calibrated photodiode, Keithley 6517A picoammeter and a TMc300 monochromator. SEM and AFM The dual beam system Nova Nanolab (FEI Co.) was employed for SEM imaging. The sample was tilted by 52 degrees and the cross-section of device was imaged near the scratch. AFM measurements were obtained with an SPM Solver Next (NT-MDT) AFM. Topography and phase images were measured in tapping mode with repulsive average tip-sample force by using probes NSG11 (NT-MDT). Photoconductive–AFM Pc-AFM was performed with an SPM Solver P47H (NT-MDT) in a nitrogen-filled glovebox. Conductive gold-coated probes NSC36/Cr-Au and CSC37/Cr-Au (Micromash) with three levers on each probe were used for pc-AFM measurements. All measurements were performed with the longest lever on chip. Photoexcitation to enable photoconductivity measurements was provided by the AFM laser at a wavelength of 670 nm. We confirmed that photocurrent was measured by scanning a region with the 670-nm tip light on to determine the topography and conductivity, and then rescanned the same area with the 670-nm tip light off and measuring only conductivity, using the tip-sample distance derived from the first scan—in this scenario we see current with the tip light on but none on the second pass with the tip light off, as shown in Supplementary Fig. S5 along with a description of the geometry of the AFM tip laser, enabling direct illumination of tip-sample contact area. The sample was grounded and the denoted voltage was applied to the tip. The sample comprised of an ITO-coated glass substrate on which PEDOT:PSS and then PTB7:PC71BM were deposited by spin-coating with the same conditions as noted above. Optical modelling The optical density and absorption simulations were performed using finite-element in the COMSOL package. The refractive indexes of the PTB7:PC71BM blend and of PEDOT used in the simulations were measured by ellipsometry as shown in Supplementary Fig. S16. The refractive indexes of ITO, Calcium and Aluminium were taken from the literature. By deriving the true active layer absorption with optical modelling, all interference effects in the stack are taken into account, and thus the IQE spectra can be deduced directly from the EQE spectra. Synthesis and characterization of DPP-NMe2 Compound A1 53 (600 mg, 0.88 mmol), as shown in Fig. 3a, 10% Pd/C (103 mg, 0.10 mmol), SPhos (40 mg, 0.10 mmol) and K2CO3 (305 mg, 2.20 mmol) were dissolved/suspended in dimethylacetamide (5 ml). The mixture was degassed with N2 for 5 min, and then 4-ethynyl-N,N-dimethylaniline (282 mg, 1.94 mmol) was added to the mixture. The solution was heated at 110 °C overnight. The reaction was cooled to room temperature, diluted with ethyl acetate and washed repeatedly with brine. The organic layer was separated, dried over MgSO4, filtered and the solvent removed under reduced pressure to afford a crude product that was subjected to silica gel column chromatography (petroleum ether/dichloromethane (1:1)) to afford the product as a blue metallic solid (0.32 g, 45%). Mp. 275 °C (Dec.). 1H NMR (400 MHz, CDCl3) δ=0.90 (12H, m), 1.22–1.42 (16H, m), 1.90 (2H, m), 3.02 (12H, s), 4.01 (4H, m), 6.66 (4H, J=9.0 Hz, d), 7.30 (2H, J=4.2 Hz, d), 7.41 (4H, J=9.0 Hz, d), 8.92 (2H, J=4.2 Hz, d). 13C NMR (100 MHz, CDCl3) δ=10.5, 14.0, 23.0, 23.6, 28.3, 30.1, 46.1, 46.0, 40.2, 80.8, 99.8, 108.5, 108.7, 111.8, 129.5, 130.0, 131.8, 132.8, 135.8, 139.4, 150.6, 161.7. MS (FAB/NOBA): m/z 810.5 [M+H]+. Anal. Calcd for C50H58N4O2S2: C, 74.03; H, 7.21; N, 6.91. Found: C, 74.17; H, 7.42; N, 6.88. Pureness of fullerene spheres The ‘pureness’ of the small spheres of PC71BM can also be determined using the time-resolved PL data. This can be modelled by using the product of the normalized populations of the solution to the differential equation (equation 3) and the population in the sphere quenching case (equation 1). The polymer was assumed to be outstretched and an effective radius was calculated from the length of the chain (~140 nm for 92,000 Da MW, assuming equilibrium bond lengths) and a capture radius from the chain approximated as 1 nm (reasonable in comparison to the capture radius calculated from the DPP-NMe2 complex). These parameters were used to approximate the straight polymer as a highly prolate spheroid, which can be used to generate an effective spheroidal capture distance of 14 nm using the method described by Sreearunothai et al. 54 The simulated decay for a 60-nm-diameter sphere of PC71BM with 0.2 wt% of PTB7 homogenously mixed in to it is faster than is observed experimentally (Fig. 2e), suggesting that even a very small amount of PTB7 in the sphere causes significant quenching. Consequently, we can conclude that the spheres have to be pure fullerene, as even a very small amount of PTB7 causes more quenching than is observed. Determination of PC71BM to PTB7 Förster radius To determine the Förster radius, R 0, for RET from the PC71BM donor to the PTB7 acceptor we use equation 6 (ref. 55): where κ is the orientation factor between the donor and acceptor and is taken to be 2/3 for randomly orientated dipoles, η D is the PL quantum yield of the donor (measured to be 0.015 for PC71BM doped in PMMA at 1% concentration), n is the refractive index of the film (taken to be 1.9), N A is Avogadro’s number, f d (λ) is the PC71BM fluorescence spectrum normalized to area 1 and ε A (λ) is the molar absorption coefficient spectrum of PTB7, with values derived from a solution of PTB7. With the values defined above, and the spectral overlap as shown in Fig. 1b we calculate that the Förster radius is 2.17 nm. Determination of radiative rates for PTB7 and PC71BM To determine the radiative rate, k R, of the PL for PTB7 and PC71BM we use equation 7: where η is the PLQY and τ is the natural lifetime. For PTB7, η=0.02 and τ ave=45 ps, thus k R=0.44 ns−1. For PC71BM, η=0.015 and τ ave=600 ps, giving k R=0.025 ns−1. Experimental methodology for DSC DSC was performed with a Netzsch DSC204F1, using hermetically sealed aluminium pans, with an empty pan used as a reference. The scan of PC71BM shown in Supplementary Fig. S4 was measured with a Netzsch STA449 with open air pans. All DSC scans were performed with a scan rate of 10 K min−1 with a first heat to 170 °C, then a cool to 30 °C, followed by a second heat to 380 °C. Blend samples were made by solution casting from a solution of PTB7:PC71BM made with the same concentration and ratio as used for the spin-coated samples (10:15 mg ml−1 dissolved in 1 ml of chlorobenzene at 50 °C with gentle stirring in a nitrogen-filled glovebox). The solution was deposited on a glass microscope slide and left to dry in a nitrogen-filled glovebox for ~3 days to allow the solvent to evaporate. Complete removal of the solvent was assured by cycling the sample under vacuum of ~10−2 mbar. The film thickness was determined with a Veeco Dektak 150 surface profiler, making a scratch on the film and measuring the depth. A film thickness of ~4 μm was measured. The film was then scraped off the glass substrate and deposited into the aluminium pan for the DSC measurements. Author contributions G.J.H. and I.D.W.S. planned the research. G.J.H. performed measurements of time-resolved fluorescence on the blends and spectral diffusion in PC71BM, made the blend samples for morphology measurements and performed the analysis of pc-AFM and TRPL data with A.R. A.J.W. measured the fluorescence quenching of PC71BM by dispersed quencher and carried out the calculations for determining the exciton diffusion coefficient in the fullerene and diffusion-defined morphology in the blends. A.A. measured the AFM, pc-AFM and SEM of the blends. C.T.H. fabricated and characterized the solar cells. E.R.M. measured the optical constants of the materials with ellipsometry and performed the optical modelling. L.A.S. and G.C. synthesized DPP-NMe2 used in the volume quenching experiments. A.R. and I.D.W.S. guided the research. The article was written by G.J.H., A.R. and A.J.W. Additional information How to cite this article: Hedley, G. J. et al. Determining the optimum morphology in high-performance polymer-fullerene organic photovoltaic cells. Nat. Commun. 4:2867 doi: 10.1038/ncomms3867 (2013). Supplementary Material Supplementary Information Supplementary Figures S1-S16
                Bookmark

                Author and article information

                Journal
                J Mater Chem A Mater Energy Sustain
                J Mater Chem A Mater Energy Sustain
                Journal of Materials Chemistry. A, Materials for Energy and Sustainability
                Royal Society of Chemistry
                2050-7488
                2050-7496
                14 July 2017
                8 June 2017
                : 5
                : 26
                : 13748-13756
                Affiliations
                [a ] Molecular Materials and Nanosystems , Institute for Complex Molecular Systems , Eindhoven University of Technology , P.O. Box 513 , 5600 MB Eindhoven , The Netherlands . Email: r.a.j.janssen@ 123456tue.nl
                [b ] Institute for Materials Research (IMO-IMOMEC) , Design & Synthesis of Organic Semiconductors (DSOS) , Hasselt University , Agoralaan, 3590 Diepenbeek , Belgium
                [c ] Dutch Institute for Fundamental Energy Research , De Zaale 20 , 5612 AJ Eindhoven , The Netherlands
                Article
                c7ta01740e
                10.1039/c7ta01740e
                5735362
                This journal is © The Royal Society of Chemistry 2017

                This article is freely available. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence (CC BY 3.0)

                Categories
                Chemistry

                Notes

                †Electronic supplementary information (ESI) available. See DOI: 10.1039/c7ta01740e

                Comments

                Comment on this article