Amorphous Ternary Charge-Cascade Molecules for Bulk Heterojunction Photovoltaics.

Ternary bulk heterojunctions with cascade-type energy-level configurations are of significant interest for further improving the power conversion efficiency (PCE) of organic solar cells. However, controlling the self-assembly in solution-processed ternary blends remains a key challenge. Herein, we leverage the ability to control the crystallinity of molecular semiconductors via a spiro linker to demonstrate a simple strategy suggested to drive the self-assembly of an ideal charge-cascade morphology. Spirobifluorene (SF) derivatives with optimized energy levels from diketopyrrolopyrrole (DPP) or perylenediimide (PDI) components, coded as SF-(DPP)4 and SF-(PDI)4, are synthesized and investigated for application as ternary components in the host blend of poly(3-hexylthiophene-2,5-diyl):[6,6]phenyl-C61-butyric acid methyl ester (P3HT:PCBM). Differential scanning calorimetry and X-ray/electron diffraction studies suggest that at low loadings (up to 5 wt %) the ternary component does not perturb crystallization of the donor:acceptor host blend. In photovoltaic devices, up to 36% improvement in the PCE (from 2.5% to 3.5%) is found when 1 wt % of either SF-(DPP)4 or SF-(PDI)4 is added, and this is attributed to an increase in the fill factor and open-circuit voltage, while at higher loadings, the PCE decreased because of a lower short-circuit current density. A comparison of the quantum efficiency measurements [where light absorption of SF-(DPP)4 was found to give up to 95% internal conversion] suggests that improvement due to enhanced light absorption or to better exciton harvesting via resonance energy transfer is unlikely. These data, together with the crystallinity results, support the inference that the SF compounds are excluded to the donor:acceptor interface by crystallization of the host blend. This conclusion is further supported by impedance spectroscopy and a longer measured charge-carrier lifetime in the ternary blend.

Introduction

Organic photovoltaics (OPVs) have received tremendous attention as a potential low-cost large-area solar energy conversion technology, now routinely reaching over 10% solar power conversion efficiency (PCE) in laboratory-sized devices.[1,2] Further development toward commercialization requires careful optimization of the optoelectronic properties as well as morphology of the photoactive donor:acceptor bulk heterojunction (BHJ). In this respect, the use of a ternary BHJ (i.e., including a third active component in the donor:acceptor BHJ) has recently been established as a powerful approach to engineering high-efficiency OPVs.[3−6] Blending multiple donor molecules to extend the absorption spectrum[7−10] and using additives to control the morphology[11] or doping,[12] as well as a cascade strategy,[13−17] exemplify the advantages of ternary systems. Charge-cascade ternaries, in particular, represent an elegant and versatile way to mitigate bimolecular recombination in OPVs. This approach consists of including a tertiary molecule with intermediate energy levels ideally at the interface between the donor and acceptor domains so that free photogenerated holes and electrons accumulate on the donor and acceptor materials, respectively, and are spatially separated by the cascade molecule. Several reports even show the possibility of enlarging the exciton harvesting distance through long-range energy transfer from the donor to the ternary material (a mechanism known as energy cascade), provided that the donor has a larger band gap.[9,18,19] These two considerations lead to well-known design rules for the energetics of cascade molecules: highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energy levels must be between those of the donor and acceptor while retaining a narrow absorption band gap. Molecules matching these properties have been successfully used in planar junction devices with the ternary compound selectively placed at the interface, reaching over 8% PCE.[18] In BHJ OPVs, a PCE over 11% has recently been reported.[20]

Despite these encouraging results, the relative improvement in the performance when applying the ternary cascade approach to the BHJ remains inferior to the planar junction configuration, in part, because of the inherent morphological complexity of the BHJ active layer. Indeed, controlling the self-assembly of a ternary BHJ into the ideal cascade morphology—especially driving the cascade compound to the donor:acceptor interface—remains a key challenge. Concrete efforts to prepare ternary blends with one component at the interface of the other two have been reported by Honda et al. using a silicon phthalocyanine derivative in a poly(3-hexylthiophene-2,5-diyl):[6,6]phenyl-C61-butyric acid methyl ester (P3HT:PCBM) host blend.[21] The self-assembly of that system was later ascribed to surface energy differences between the three materials,[22] which limits the chemical nature of the cascade material. Moreover, engineering surface energies in the BHJ blend remains nontrivial. Alternatively, Gu et al. have recently reported the formation of a multilength-scale ternary morphology attributed to the crystallization of P3HT, allowing the formation of two distinct donor phases working as subcells in parallel.[23] Even though this system does not form an ideal cascade structure, the use of the host blend crystallization as a driving force for the self-assembly is a promising and simple strategy to engineer ternary morphology. Inspired by recent work employing a spiro linker[24−26] to control the 3D self-assembly and crystallinity in materials for organic electronics, herein we demonstrate that controlling the crystallinity can be used for the efficient self-assembly of cascade BHJ toward the formation of an ideal charge-cascade structure. Specifically, using a noncrystalline “amorphous” cascade molecule as the ternary component and tuning the crystallinity of the electron donor and acceptor phases, thereby driving the ternary molecule to the interface, lead to a maximum 36% improvement in the PCE of a P3HT:PCBM solar cell.

Results and Discussion

In order to demonstrate the self-assembly of an ideal cascade morphology using an amorphous ternary additive, two molecules (Figure ) were synthesized via the Suzuki coupling reaction between functionalized conjugated aryl groups and a spiro core. The spiro core is chosen to discourage long-range π–π stacking in the material, while a perylenediimide derivative, commonly coded as PDI, and a diketopyrrolopyrrole derivative, coded as DPP, are used as aryl groups, in order to tune the energy levels of the molecule. The PDI aryl group yields the molecule coded here as SF-(PDI)4, which was recently reported as an electron acceptor for OPVs,[25,26] and, alternatively, coupling DPP yields a novel molecule coded as SF-(DPP)4. See the Supporting Information for full synthetic information and basic characterization. The electrochemical and optical properties of both molecules were measured by cyclic voltammetry and ultraviolet–visible (UV–vis) spectroscopy (details are shown in Figure S1) in order to position the HOMO and LUMO levels (summarized in Figure c together with the host blend components). The optoelectronic properties of SF-(DPP)4 and SF-(PDI)4 are similar to those reported for parent DPP(TBFu)2 and PDI derivatives,[27,28] supporting the notion that the band gap and energy levels of the overall spiro molecule can be effectively engineered by varying the aryl groups. It can be seen that the P3HT:PCBM system accommodates well both spiro derivatives as potential charge-cascade molecules given their intermediate HOMO and LUMO, respectively, energy levels. Therefore, this classic system was chosen as the host blend to test our new ternary materials.

Figure 1

Chemical structure of the ternary small molecules used in this work: (a) SF-(DPP)4 with R1 = 2-ethylhexyl; (b) SF-(PDI)4 with R2 = 2-hexyldecyl. (c) Energy levels and band gap of the active molecules used in this work.

DSC measurements of pure SF-(DPP)4 and SF-(PDI)4 did not exhibit any indication of crystallinity. In the thermograms of the pure materials (Figure S2a), no particular melting or crystallization peaks are seen, strongly suggesting that both compounds lack crystalline ordering in the solid state. This is supported by grazing-incidence X-ray diffraction (GIXRD) data measured on spin-coated films (Figure S2b), where no out-of-plane diffraction peak can be seen for either molecule.

In order to examine the mixing behavior of the spiro derivatives in the presence of P3HT and PCBM, DSC measurements of pure P3HT or PCBM as well as blends with 10 or 25 wt % of either ternary component were performed. The (second) heating scans are shown in Figure a, with the solid and broken lines respectively representing the addition of SF-(DPP)4 and SF-(PDI)4. It can be seen that pure P3HT (purple trace) exhibits an endothermic melting transition at ca. 234 °C. This peak remains present after the addition of either spiro additive up to 25 wt %. A sharp endothermic (melting) transition is also seen for pure PCBM (brown trace) at ca. 282 °C. Similar to the P3HT, the addition of SF-(PDI)4 does not significantly affect the melting transition of PCBM. However, upon blending with SF-(DPP)4, the PCBM melting transition strongly decreases in intensity and melting temperature (down to 266 °C for 25 wt %). Interestingly, a small exothermic transition appears at ca. 182 °C, which likely originates from the crystallization of a kinetically frustrated mixture created during cooling of the melt.

Figure 2

(a) DSC second heating curves of neat P3HT and PCBM with varying amounts of SF-(DPP)4 (solid lines) and SF-(PDI)4 (dotted lines). Scale bars represent 0.2 mW g–1. (b) Relative crystallinity (based on melting enthalpies) of PCBM (dots) and P3HT (triangles) as a function of the additive content.

The melting enthalpy extracted from the DSC data allows estimation of the relative crystallinity of the donor or acceptor component in blends with the ternary component (Figure b).[29] In the case of P3HT, the relative crystallinity only slightly decreases upon the addition of both additives and never drops below 95%, supporting the notion that SF-(DPP)4 and SF-(PDI)4 do not disrupt the crystallinity of P3HT and are excluded from the pristine crystalline P3HT domains. In contrast, PCBM shows a significant decrease in the relative crystallinity, with about 38% loss when 25 wt % SF-(DPP)4 additive is included. This suggests that the additive mixes with PCBM and limits its crystallization to some extent. Nonetheless, we note that at 1 wt % of the additive the relative PCBM crystallinity remains 94% (see also Figure S3 for the heating scan of this condition). On the other hand, the PCBM phase exhibits a relative crystallinity higher than 88% at 25 wt % SF-(PDI)4 content, suggesting that this additive mixes less with PCBM.

While the melting transition behavior gives an indication of the mixing of the additives with the host donor and acceptor in the solid state, the DSC measurement conditions are significantly different from the conditions that occur during solution processing of the BHJ in a thin film. Thus, further investigation over the crystallinity of P3HT or PCBM blended with one of the additives, SF-(DPP)4, in thin films spin-casted from dichlorobenzene and annealed at 140 °C for 10 min was performed using GIXRD (out-of-plane) and electron diffraction, respectively, to assess the crystallinity of P3HT and PCBM (Figures S4 and S5). The full width at half-maximum (fwhm; Figure S6) of the diffraction peaks is well-known to be inversely proportional to the crystallite size and quality in paracrystalline materials.[30,31] Pristine P3HT is found to have a fwhm of ca. 0.55 nm–1 corresponding to a crystal coherence length (CCL) of 11.4 nm. A further addition of SF-(DPP)4 does not significantly affect the fwhm for P3HT because the blend of 25 wt % SF-(DPP)4 still exhibits a fwhm of ca. 0.56 nm–1. In the case of pristine PCBM, a much larger fwhm of 3.52 nm–1 is calculated, as expected from its lower crystallinity. Upon the addition of 1, 10, and 25 wt % SF-(DPP)4, the fwhm increases only slightly to ca. 3.71, 3.81, and 4.18 nm–1, respectively, which is in good agreement with the aforementioned DSC results and supports the view that the SF-(DPP)4 additive does not form a cocrystal with P3HT or PCBM, nor does it significantly affect the crystallinity of the pure donor and acceptor components. We note that no meaningful CCL can be extracted for fwhm data from the electron diffraction of PCBM.[32]

The application of ternary additives in BHJ OPV devices was next investigated in optimized P3HT:PCBM host blend BHJs. We note that the PCE observed at the optimized 6:4 ratio (Figure S7) under the best conditions without any ternary additive (ca. 2.5%) is slightly lower than the average PCE reported of 3.0%[33] because of the relatively large device area used (0.16 cm2) and the fact that PC71BM and calcium (as a cathode interlayer) were not employed in this study. We also established that, upon the addition of 5 wt % SF-(DPP)4 to the host blend, no change in the optimal P3HT:PCBM blending ratio was found (Figure S8). Atomic force microscopy (AFM) height images of the film surface, shown in Figure S9, further confirm that the BHJ morphology remains largely unaffected by the addition of SF-(DPP)4.

The effect of varying the content of SF-(DPP)4 or SF-(PDI)4 on the performance was then investigated by fabricating devices containing 0, 1, 2.5, 5, and 10 wt % of the ternary additive. The device figures of merit (averaged over at least four different devices) are plotted on Figure as a function of the additive concentration, and the corresponding J–V curves can be found in Figure S10. It can be seen that the addition of 1 wt % SF-(DPP)4 results in an increased PCE of up to 3.2% compared to the 2.5% of the control blend (about 28% improvement). This increase can be mainly attributed to enhancement in the fill factor (FF; 14%), which improves from 0.43 to 0.49 upon SF-(DPP)4 addition. The open-circuit voltage (VOC) also increases by 20 mV. Taken together, the VOC and FF improvements suggest a reduced recombination within the BHJ. Partially due to the extended absorption spectrum of SF-(DPP)4 compared to the host blend, a slight improvement in the short-circuit current (JSC) is also observed. A further addition of SF-(DPP)4 seems to deteriorate the device performance as both JSC and FF decrease. In contrast, VOC continues to improve gradually from 0.52 to 0.67 V upon the addition of SF-(DPP)4 up to 10 wt %. A similar enhancement in VOC for ternary devices has been attributed to the additional density of states[34] as well as alloy formation;[35,36] however, these studies have only been demonstrated on miscible ternary blends, which is not the case of the SF-(DPP)4 reported here (as demonstrated by the DSC results). Rather, the improvement in VOC in our case could result from either a doping effect or mitigated recombination at the open circuit and will be further discussed hereafter.

Figure 3

Photovoltaic device performance with added amorphous cascade molecules. The average figures of merit (PCE, JSC, VOC, and FF) are shown as a function of the SF-(DPP)4 and SF-(PDI)4 content in P3HT:PCBM BHJ OPVs.

The SF-(PDI)4 additive exhibits a performance trend very similar to that of SF-(DPP)4, with a 36% improvement in the PCE at 1 wt % (due to increases in VOC and FF), reaching an efficiency of ca. 3.5%. Because SF-(PDI)4 also improves JSC [to a lesser extent than SF-(DPP)4], this suggests that the extended absorption spectrum is not the main origin for device improvement. At a higher content than 1 wt % SF-(PDI)4, except for the enhancement in VOC, all of the figures of merits are seen to decrease, similar to SF-(DPP)4. Taken together, the higher PCE for the SF-(PDI)4 additive at 1 wt % compared to SF-(DPP)4 at the same concentration supports the conclusion that the contribution from extended absorption as well as from a potential energy cascade mechanism in the SF-(DPP)4 system is negligible. This is plausible given the intrinsically low optimum concentration of the additive and the already optimized morphology. In stark contrast, the addition of 9,9′-spirobifluorene (the unfunctionalized spiro-core molecule, with Eg = 3.8 eV) results in a loss in the PCE at 1 wt %, further supporting the charge-cascade mechanism (Figure S11).

In order to gain further insight into the working mechanism of the amorphous charge-cascade additives, we next focused in detail on the effect of the of the SF-(DPP)4 additive by estimating the internal quantum efficiency (IQE; Figure a) from the external quantum efficiency and absorption data detailed in Figure S12. The host blend device exhibits IQE values close to 60% (square markers in Figure a) in line with previous reports.[37] Adding 1 wt % SF-(DPP)4 gives higher IQE values over the entire absorption range, correlating well with the observed enhancement in JSC (circle markers in Figure a). Strikingly, the photons absorbed by SF-(DPP)4 at 685 nm reach an IQE of 95%, suggesting that essentially all excitons generated within SF-(DPP)4 molecules are efficiently dissociated into free charge carriers. This strongly supports the argument that SF-(DPP)4 successfully self-assembles at the interface between P3HT and PCBM molecules, either between crystalline domains or within amorphous domains, leading to an ideal cascade morphology. A further addition of SF-(DPP)4 leads to an overall decrease in the IQE, accounting for the loss in JSC. It is worth noting that the IQE at 685 nm decreases significantly more than the IQE within the P3HT absorption region, which translates into a loss in the exciton splitting efficiency within the SF-(DPP)4 phase. This loss could be attributed to SF-(DPP)4 domains growing larger than the exciton diffusion length. However, one might also consider that the disrupted crystallinity of PCBM at high SF-(DPP)4 content, as seen from the DSC data, could account for a PCBM amorphous-rich region at the interface surrounding the SF-(DPP)4 molecules, which leads to hole-trapping-state formation.

Figure 4

(a) IQE of the host blend and containing 1, 2.5, and 10 wt % SF-(DPP)4. (b) Recombination lifetime obtained from fitting the IS response (dots, inset) at open circuit under 1 sun for increasing content of SF-(DPP)4 (0, 1, 2.5, 5, and 10 wt %). Corresponding data as a function of the SF-(PDI)4 content are shown in Figure S15.

Charge-carrier lifetimes were extracted from impedance spectroscopy (IS) data under 1 sun illumination at open circuit because this technique is well-known to provide useful insight into recombination losses.[38] The impedance response of the devices (see the Nyquist plots in Figure b, inset) always exhibited a single semicircle at low frequency, which was fit using a simple constant-phase element (CPE) parallel to the resistance (Rrec). The latter, Rrec, which can be associated with the electrical resistance through recombination channels, is found to increase by more than 1 order of magnitude with the addition of SF-(DPP)4 (Figure S13). This supports the view that the recombination rate decreases in the presence of the amorphous charge-cascade additive at the interface of P3HT and PCBM. The trend in the recombination lifetime, shown in Figure b as a function of the additive concentration, indicates a longer free charge-carrier lifetime with increasing SF-(DPP)4 and confirms that recombination is mitigated in the presence of the ternary additive.

Interestingly, lifetimes are especially longer at high loadings (e.g., 5 and 10 wt %), despite their poor J−V performances. Actually, 5 and 10 wt % SF-(DPP)4 devices both exhibit a sublinear dependence of JSC as a function of the incident illumination intensity (Figure S14), which is typically characteristic of high recombination.[39] However, the aforementioned considerable increase in the recombination lifetime at higher loadings contrasts with this possibility. Rather, this sublinear behavior and loss in performance could reasonably be caused by the formation of a space-charge layer[40] or, less likely at open circuit, poor transport. Under the assumption that the amorphous additive is excluded from both the crystalline donor and acceptor phases at all loading concentrations, upon an increase in the loading and as long as both crystalline donor and acceptor phases remain, there reasonably must exist a loading limit above which a continuous SF-(DPP)4 layer would isolate some crystalline donor or acceptor domains. This would break down the percolation pathways and favor the formation of a space-charge layer within the BHJ (because isolated donor or acceptor domains will accumulate free charges upon light illumination), which will deteriorate the transport. While the morphological verification of this condition is complicated by the optimal donor:acceptor domain size present in P3HT:PCBM BHJs (ca. 20 nm) and the similar chemical nature of the components, the experimentally demonstrated decrease in the OPV device performance at higher loadings reasonably demonstrates this fundamental drawback of the amorphous ternary cascade approach. Despite this, the increased performance of the control BHJ upon optimized loading of the amorphous ternary charge-cascade molecules, observed in this work, gives promise for this approach to enhance the performance of high-performance OPVs based on semicrystalline materials.

Conclusions

In conclusion, we have combined the aspects of the exclusion of an amorphous ternary additive from a crystalline donor:acceptor blend for BHJ OPVs with the charge-cascade approach to establish a proof-of-concept. By employing DPP and PDI chromophores with the ideal energy levels to enhance the electron and hole separation in a P3HT:PCBM BHJ OPV and controlling the chromophore mixing with P3HT and PCBM by using a spiro linking strategy, we demonstrated the successful exclusion of ternary additives from the crystalline host blend using DSC and crystal domain size results. At optimized ternary loadings (i.e., 1 wt %), an improvement of up to 36% relative to the host blend (from a PCE of 2.5% to 3.5%) was found with the SF-(PDI)4 additive, while a similar improvement (28%) was also observed with the SF-(DPP)4 additive. The change in the device performance metrics upon additive inclusion, together with the different UV–vis absorption properties of the two additive molecules and the quantum efficiency measures [which showed an impressive 95% IQE of SF-(DPP)4 in optimized blends], indicated that the improvement was not due to increased light absorption or an increase in the exciton harvesting distance through energy transfer. Further investigation by IS supported the view that the improvement was due to a lower recombination within the active layer at optimized conditions, while at high additive loading (i.e., 5–10 wt %), a decrease in the performance was reasonably attributed to the formation of a space-charge layer likely caused by the isolation of crystalline P3HT or PCBM domains by the ternary compound. The promising performance as well as the simplicity of the strategy demonstrated herein could likely be extended to high efficiency paracrystalline blends to further increase the PCE without requiring a complex optimization, assuming that the segregation and interfacial properties are similar to the P3HT:PCBM system.

Materials and Methods

Solar Cell Fabrication and Testing

Solar cells were fabricated on a glass substrate patterned with 300 nm of ITO. A 40 nm layer of PEDOT:PSS (Ossilla M121 Al 4083) was first spin-coated at 3000 rpm for 1 min prior to annealing at 130 °C in air. The BHJ active layer was then spin-cast at 800 rpm from a blend solution of P3HT (Sigma-Aldrich, 99.995%, RR, MW 54–75 kDa) and PC61BM (Ossila M111, 99.0%) at a 6:4 weight ratio in dichlorobenzene (stirred overnight at 80 °C) at a total solid concentration of 30 mg mL–1, resulting in a thin film with a measured thickness of ca. 140 nm. An 80-nm-thick aluminum cathode was deposited (area 16 mm2) by thermal evaporation (Kurt J. Lesker Mini-SPECTROS). Prior to testing, active devices were annealed at 140 °C for 15 min. Electronic characterization was performed under simulated AM1.5G irradiation from a 300 W xenon arc lamp set to 100 mW cm–2 with a calibrated silicon photodiode (ThorLabs). Current–voltage curves were obtained with a Keithley 2400 source measuring unit. Incident photon-to-charge-carrier efficiency measurement was performed using a xenon arc lamp light source and a monochromator (OBB). A calibrated photodiode was used to measure the incident number of photons at each wavelength. IS was carried out using a SP-300 potentiostat at frequencies ranging from 1 MHz up to 1 Hz. Device fabrication was performed under an argon atmosphere, and testing was performed under a nitrogen atmosphere.

Surface Characterization

AFM was performed under ambient conditions with a Cypher S from Asylum research. A cantilever from Olympus (AC240TS) was used.

Thermal Characterization

DSC was performed using a PerkinElmer DSC8000 calorimeter calibrated with indium and zinc at a scanning rate of 10 °C min–1. Samples were prepared by drop-casting from a precursor solution in chlorobenzene and slow evaporation of the solvent at 80 °C under an argon atmosphere. The enthalpies were calculated by integration over the phase transition.

Optical Characterization

Absorption spectra were recorded with a Shimadzu UV–vis–near-IR UV-3600 spectrophotometer. In order to calculate the IQE, the total absorbance was measured using an integration sphere.

Crystal Characterization

GIXRD was measured with a D8 Discovery (Bruker) diffractometer using Cu Kα radiation and a nickel β filter with a scan rate of 0.05° min–1 and a step width of 0.01°. Transmission electron microscopy (TEM) analysis was performed using a FEI Tecnai Osiris microscope.