Unraveling the efficiency-limiting morphological issues of the perylene diimide-based non-fullerene organic solar cells

Herein we report a comparative morphological analysis of the perylene diimide (PDI)- and fullerene-based organic solar cells (OSCs) to identify the factors responsible for low performance of PDI-based devices. A PDI derivative, bis-PDI, and a fullerene derivative, PC70BM, are mixed with an efficient polymer donor, PffBT4T-2OD. The large disparity in power conversion efficiencies (PCEs) of OSCs composed of PffBT4T-2OD:bis-PDI (PCE = 5.18%) and PffBT4T-2OD:PC70BM (PCE = 10.19%) observed are attributed to differences in the nanostructural motif of bulk heterojunction (BHJ) morphologies of these blend systems. The X-ray scattering and surface energy characterizations revealed that the structurally dissimilar bis-PDI and PC70BM molecules determine the variation in blend film morphologies, and in particular, the molecular packing features of the donor PffBT4T-2OD polymer. In addition, high-resolution transmission electron microscopy (HRTEM) images explore the BHJ morphologies and presence of longer polymer fibrils in PffBT4T-2OD:bis-PDI system, justifying the unbalanced charge transport and high hole mobility. The low performance of PffBT4T-2OD:bis-PDI devices was further investigated by studying charge carrier recombination dynamics by using light-intensity-dependent and transient photovoltage (TPV) experiments. Furthermore, the temperature-dependent experiments showed the photovoltaic properties, including charge recombination losses, are strongly affected by energetic disorder present in bis-PDI-based system.


Results and Discussion
The chemical structures and energy levels (HOMO and LUMO) of PffBT4T-2OD, PC 70 BM, and bis-PDI are shown in Fig. 1a,b. PffBT4T-2OD is a crystalline polymeric donor with a broad absorption spectrum in visible region, and OSC devices reported with fullerene acceptors exhibited PCE of 10.8%, one of the highest PCE values recorded for single OSC 25 . The bis-PDI small molecule was synthesized by a modified procedure as described in the supporting information, and used as an acceptor material because of its strong light absorption in visible region (420-600 nm), good electron mobility, and favorable energy level matching with PffBT4T-2OD donor polymer. Figure 1c presents the UV-vis absorption spectra of the pristine (bis-PDI, PC 70 BM and PffBT4T-2OD) and blend (PffBT4T-2OD:PC 70 BM and PffBT4T-2OD:bis-PDI) films, where both blend films exhibit good light harvesting in the visible region (400-780 nm). In addition, PffBT4T-2OD:bis-PDI blend film shows a weak absorption near-UV region (300-420 nm), this difference might be attributed to weak absorption of bis-PDIs in this region. Solar cell device performance. BHJ OSCs were fabricated with device structure of glass/indium tin oxide (ITO)/zinc oxide (ZnO)/donor:acceptor/vanadium oxide (V 2 O 5 )/Ag, and current density-voltage (J-V) characteristics of the optimized PffBT4T-2OD:PC 70 BM and PffBT4T-2OD:bis-PDI OSCs are shown in Fig. 2a. The optimal device performance of the PffBT4T-2OD:bis-PDI system was achieved using a 1:1.5 (wt:wt) ratio and of the PffBT4T-2OD:PC 70 BM system using a 1:1.2 (wt:wt) ratio (Table 1 and S1) 25  of 0.85 V, a J SC of 11.22 mA cm −2 , and a FF of 52.6% with DIO (w/DIO). The high value of V OC for PffBT4T-2OD:bis-PDI OSC is assigned to the high LUMO energy level of the bis-PDI acceptor 26 .
The external quantum efficiency (EQE) spectra of the PffBT4T-2OD:PC 70 BM and PffBT4T-2OD:bis-PDI devices were recorded over a broad range 300-800 nm (Fig. 2b), which is corresponding well to the absorption spectra of the respective films ( Fig. 1c and Figure S2a). The maximum EQE values of the PffBT4T-2OD:bis-PDI (50.1%) and PffBT4T-2OD:bis-PDI w/DIO (58.3%) devices are significantly lower than those of the PffBT4T-2OD:PC 70 BM (69.8%) and PffBT4T-2OD:PC 70 BM w/DIO (79.6%) devices, measured at 455 nm. The large difference in EQE values between bis-PDI-and PC 70 BM-based devices is well matched to the J SC values. The J SC values calculated from the EQE spectra are in a good agreement with experimental J SC in the direct J-V measurements (Table S2).
The OSC devices were further set to characterize the effective exciton dissociation into free charge carriers in bis-PDI-and PC 70 BM-based devices. For this, we have studied the relationship between the photocurrent (J ph ) and the effective voltage (V eff ) ( Figure S2b), J ph = J L − J D , where J L is the current density under illumination and J D is the current density under dark conditions 27 . At a high reverse biased voltage, J ph values are saturated in all solar cells. The saturation current density (J sat ) and charge dissociation probability (P diss = J ph /J sat ) were determined for all devices. The calculated P diss values under short-circuit condition are 84.6% for PffBT4T-2OD:PC 70 BM, 73.6% for PffBT4T-2OD:bis-PDI, 91.8% for PffBT4T-2OD:PC 70 BM w/DIO, and 74.1% for PffBT4T-2OD:bis-PDI w/ DIO, respectively. The higher values of P diss obtained from the PffBT4T-2OD:PC 70 BM compared to PffBT4T-2OD:bis-PDI suggest an efficient dissociation of excitons into free charge carriers and better collection of free charge carriers at electrodes 27 . Film morphology and photoluminescence quenching. The morphologies of the PffBT4T-2OD:PC 70 BM and PffBT4T-2OD:bis-PDI blend films were characterized by HRTEM (Fig. 3a,b,d,e). All the TEM images of PffBT4T-2OD:PC 70 BM and PffBT4T-2OD:bis-PDI films have showed a nanofibrillar structures kind of BHJ morphology. These fibrils arise from the highly crystalline features of the PffBT4T-2OD polymer 28 , which was also confirmed in GIWAXS experiments (discussed later in this section). In Fig. 3, TEM images depicted that components in PffBT4T-2OD:bis-PDI film are closely packed, and are contained longer fibrils when compare to the PffBT4T-2OD:PC 70 BM blend film where fibrils appear smaller and farther apart. We assume that these differences arose due to the distinct structure and aggregation properties of PC 70 BM and bis-PDI molecules. The slightly coarser morphology in PffBT4T-2OD:bis-PDI blend film reflects in a smaller PL quenching efficiency of polymer PL spectrum (Fig. 3c,f) 29 . The higher value of PL quenching for PffBT4T-2OD:PC 70 BM (93.2%) indicates a better exciton dissociation at the donor/acceptor interface than in the PffBT4T-2OD:bis-PDI (79.9%). The addition of the DIO additive further improved the PL quenching efficiency significantly to 95.6% for PffBT4T-2OD:PC 70 BM and 83.1% for PffBT4T-2OD:bis-PDI due to the effect of DIO on the organization of the PffBT4T-2OD polymer 30 . We have also investigated the PL quenching efficiency of bis-PDI (excited at 532 nm), revealing almost complete quenching in the PffBT4T-2OD:bis-PDI blend films ( Figure S3). The complete quenching of bis-PDI excimers implies that bis-PDI is relatively close to the polymer or bis-PDI domain sizes lying well within PffBT4T the excimer diffusion length, so bis-PDI excimers could reach a heterojunction before emitting 31,32 . The AFM images of these two systems are showed in Figure S4. The surface roughness values of the PffBT4T-2OD:PC 70 BM blend system decreased w/DIO treatment from 1.8 nm to 1.1 nm, whereas surface roughness of the PffBT4T-2OD: bis-PDI film increased from 1.9 nm to 2.8 nm.
The differences in blend morphologies of two systems might have arisen due to the different surface energies of the components. Therefore, the surface energies of the pristine films were evaluated by measuring the contact angles formed between the film and either of two different liquids (deionized (DI) water and ethylene glycol (EG)) ( Figure S5). Later, contact angles were used in the Owens and Wendt geometric mean equation to calculate the surface energies (Table S3) 33 . The relatively large difference in surface energies of the PffBT4T-2OD (γ = 29.1 mN m −1 ) and the bis-PDI (γ = 41.1 mN m −1 ) is likely to result in a poor miscibility in the blend film, producing a slightly larger phase-separated film morphology. Whereas in the PffBT4T-2OD:PC 70 BM blend film, the spherical shape fullerene molecule and small difference in surface energies of the components enabled the PC 70 BM (γ = 31.6 mN m −1 ) molecule to easily penetrate and diffuse into the polymer region, which may have hindered strong aggregation of polymer 34,35 . However, the bis-PDI molecule could not penetrate into the polymer region because of large surface energy difference, and allow polymer to crystallize largely, as can be seen in TEM images. The high surface energy of the bis-PDI film may be because of more nitrogen and oxygen atoms in molecular structure compared to PC 70 BM 36 .
The crystalline properties of the PC 70 BM and bis-PDI acceptors in the presence of the PffBT4T-2OD polymer were investigated using GIWAXS experiments ( Figures S6, S7). The information extracted from the in-plane and out-of-plane of GIWAXS images are presented in Fig. 4 and summarized in Table S4. In the 2D-XRD images, both blend films exhibit distinctive diffraction peaks, as evidenced by the strong lamellar stacking with (100), (200), and (300) diffraction peaks, as well as the distinct (010) diffraction peak present in the out-of-plane direction, which indicated the presence of a preferential face-on orientation of PffBT4T-2OD polymer domains. In Figure S6a, three isotropic diffraction rings at q = 0.63 Å −1 , 1.28 Å −1 , and 1.90 Å −1 represent the amorphous behavior of the PC 70 BM with randomly oriented molecules 37 . On the other hand in Figure S6b, the pristine bis-PDI film showed a stronger (100) diffraction peak in the out-of-plane profile and a peak near 1.5 Å −1 , which corresponds to the π-π stacking interactions, is more distinctive in the in-plane direction. These features of the in-plane and out-of-plane profiles of the bis-PDI film suggested that the preferred molecular orientation was weakly edge-on. In Figures S7a,c, the broad peaks at q = 1.3 Å −1 are the characteristic of PC 70 BM aggregation 38 and a weak peak near q = 1.42 Å −1 represents bis-PDI aggregation. To understand the impact of two structurally different small molecules (bis-PDI and PC 70 BM) on the polymer chain packing, we have analyzed the coherence length for polymer peak (200) in both the blend systems. The coherence lengths for the (200) diffraction peak in the in-plane direction and the (010) diffraction peak in the out-of-plane direction were calculated for all blend systems using the Scherrer equation 39 . The coherence length in the PffBT4T-2OD:bis-PDI (15.6 nm) was much larger than that in the PffBT4T-2OD:PC 70 PM (11.8 nm), suggesting that the PffBT4T-2OD polymer formed larger crystals when mixed with bis-PDI than with PC 70 BM. Moreover, the PffBT4T-2OD:bis-PDI films prepared with DIO formed larger π-π stacked crystals (5.0 nm) oriented in the out-of-plane direction than were formed in the PffBT4T-2OD:PC 70 PM blend film (4.3 nm). Here, we speculate that the planar molecular structure of bis-PDI would form π-π interactions with PffBT4T-2OD, thereby favoring the crystallization of polymer molecules with a face-on orientation, in contrast with the PffBT4T-2OD:PC 70 BM blend film.
Charge carrier recombination dynamics. In order to gain insight into the charge carrier recombination dynamics in both type of OSCs, we varied light intensity during the measurement of J-V characteristics and the results are shown in Fig. 5a. A plot of J SC versus the light intensity (I) on a log-log scale provides a good comparison of the non-geminated bimolecular recombination losses at 0 V, where I was varied over the range 17-100 mW cm −2 40 . The experimental data points showed in Fig. 5a were fitted to a power law 41 ∝   that J SC for these devices is limited by bimolecular recombination losses 40 , however such deviation in α could also be assign to space charge limited effects 42,43 or to inefficient charge extraction caused by an improper choice of charge carrier-collecting electrodes 44,45 . The difference in charge carrier recombination dynamics of the fullerene and bis-PDI-based OSCs is quantified by measuring the TPV and transient photocurrent (TPC). The charge carrier lifetime (τ) and carrier density (n) were estimated from the TPV and TPC data as described earlier 46,47 . The carrier density versus carrier lifetime is plotted in Fig. 5b and fitted to a power law equation τ τ = λ − n n ( ) 0 , where τ 0 is the intercept at n = 0, and λ is the magnitude of the slope, which is related to the order of the non-geminated recombination (λ + 1) 48 . For similar values of n = 1.15 × 10 18 cm −3 , the charge carrier lifetime in PffBT4T-2OD:PC 70 BM is double to that in the PffBT4T-2OD:bis-PDI device, indicating a much slower recombination rate in the fullerene-based OSCs. Further, the carrier lifetime of the devices increased with the use of the additive, consistent with the improved device performance. The slope λ extracted 1.26 for PffBT4T-2OD:PC 70 BM, 1.02 for PffBT4T-2OD:PC 70 BM w/DIO, 2.05 for PffBT4T-2OD:bis-PDI, and 1.63 for PffBT4T-2OD:bis-PDI w/DIO suggesting that the bis-PDI-based devices severely affected by the non-geminated trap-assisted recombination, whereas the fullerene-based devices are nearly independent of trap-assisted recombination 48 . We expect unbalanced charge carrier transport and largely phase-separated morphology in PffBT4T-2OD:bis-PDI system are the main reasons for having high trap-assisted recombination. To understand the role of the charge carrier transport and the morphology of the blend systems on recombination dynamics, we have performed temperature-dependent SCLC mobility and J-V measurements under different sunlight intensities.
Temperature-dependent measurements. Temperature-dependent SCLC charge carrier mobilities were determined from dark J-V characteristics measured at various temperatures between 150-310 K. Figure 6a Table S5. The energetic disorder parameter σ e for electron transport in PffBT4T-2OD:PC 70 BM (74.6 meV) is smaller than that in PffBT4T-2OD:bis-PDI (81.5 meV), supporting a higher electron mobility in the PffBT4T-2OD:PC 70 BM blend system. Also, the large difference in σ h values for PffBT4T-2OD:PC 70 BM (62.0 meV) and PffBT4T-2OD:bis-PDI (38.8 meV) explains the difference in hole mobility values of the two systems. It seems evident that the values of σ h strongly depends on the PffBT4T-2OD polymer domains in two blend systems. The large difference in device performance might arise from the differences in the energetic disorder and, as a result, slower charge carriers start accumulating in the photoactive layer and increase the recombination losses 50 . To understand the effect of energetic disorder directly on the non-geminated bimolecular and trap-assisted recombinations, we measured intensity-dependent J SC and V OC at various low temperatures 150 K, 200 K, 250 K and 300 K (Figures S9 and S10). In supporting information, the intensity-dependent J SC data was fitted with power law and the α values were extracted in similar method as described above. The exponent α decreased from 0.99 to 0.94 (PffBT4T-2OD:PC 70 BM), 1 to 0.98 (PffBT4T-2OD:PC 70 BM w/DIO), 0.91 to 0.79 (PffBT4T-2OD:bis-PDI), and 0.96 to 0.86 (PffBT4T-2OD:bis-PDI w/DIO) as the measuring temperature decreased (Fig. 6b). The decrease in α could be explained in terms of slow charge extraction at a reduced mobility, which increase the availability of free charge carriers to recombine 50 . The faster rate of α decay with decreasing temperature suggests a higher-order bimolecular recombination losses in the PffBT4T-2OD:bis-PDI system. The intensity-dependent V OC was fitted to equation V OC ∝ β(k B T/e)ln(I) 41 , where T is the absolute temperature, and e is the electron charge ( Figure S10). The extent of trap assisted recombination can be estimated by extracting the β values 40 . β = 1 implies no trap-assisted recombination, whereas β > 1 indicates that trap-assisted recombination is dominating mechanism. The higher values of β, and the faster rate of increase in β values with decreasing temperature obtained for the PffBT4T-2OD:bis-PDI devices suggest that the devices are strongly limited by the trap-assisted recombination in energetically localized states. The increase in β with decreasing temperature is expected because of the stronger occupation of deeply localized states (traps), as the charge carriers no longer had access to the thermal energy needed to escape from the trap states 51 .

Conclusions
In conclusion, we have performed a comprehensive comparative analysis for elucidating the key morphological issues that are responsible for low performance of bis-PDI-based OSCs. We observed that the morphological differences in blend films, that determined the device performance, are arisen due to the distinct miscibility of PC 70 BM and bis-PDI molecules with PffBT4T-2OD polymer. GIWAX, TEM and AFM studies revealed the strongly face-on oriented PffBT4T-2OD polymer forms a well-mixed interconnected BHJ morphology with amorphous PC 70 BM instead of weakly edge-on oriented bis-PDI molecule. In addition, the large coherence length in the PffBT4T-2OD:bis-PDI suggested that bis-PDI molecule favoring the crystallization of PffBT4T-2OD polymer molecules. However, the lower values of J SC and FF in PffBT4T-2OD:bis-PDI are primarily attributed to the weak absorption in near-UV region, inefficient charge generation, low electron mobility and unbalanced charge carrier transport. The relevance of unbalanced charge transports in PffBT4T-2OD:bis-PDI devices was supported by their large differences in the energetic disorder associated with electron and hole transports. Further, low-temperature light-intensity dependent study on devices confirmed the dominance of trap-assisted recombination in PffBT4T-2OD:bis-PDI devices.

Materials.
PffBT4T-2OD and PC 70 BM were purchased from Solarmer Energy Inc., and were used as received.
The bis-PDI small molecule was synthesized according to the published procedure 52 . A ZnO precursor solution was prepared by dissolving 1 g zinc acetate dihydrate (99.999% trace metals basis, Aldrich) in 10 mL 2-methoxyethanol (99.8%, anhydrous, Aldrich) with 0.28 g ethanolamine (≥99.5%, Aldrich) as a surfactant, followed by stirring overnight under ambient conditions 53 . Characterization of the Materials. The bis-PDI and other intermediate were characterized by 1 H NMR, 13 C NMR, and mass spectrometry (MALDI-TOF). The LUMO levels of bis-PDI and PC 70 BM were calculated from cyclic voltammetry measurements. The UV-Vis absorption and PL spectra of the spin-coated films were recorded using a Perkin Elmer, Lambda 1050 spectrometer with a Horiba Jobin Yvon NanoLog spectrofluorimeter, respectively.

Preparation of the solar cells.
The inverted structure of the OSC devices was prepared with the structure: stack glass/ITO (110 nm)/ZnO (40 nm)/donor:acceptor/V 2 O 5 (2 nm)/Ag (100 nm). ITO-coated glass substrates were cleaned by sequential sonications in detergent, distilled water, acetone, and isopropyl alcohol for 15 minutes at each step. After UV/ozone treatment for 30 minutes, a ZnO electron transport layer 7 was prepared by spin-coating at 4000 rpm and then baking at 120 °C for 30 minutes on the hot plate under ambient conditions. The photoactive layer solutions were prepared in CB (polymer concentration: 9 mg mL −1 ) and were heated on a hotplate at 110 °C for 8 hours with stirring. The photoactive layers were spin-coated from the warm solution onto the prepared glass/ITO/ZnO substrate in a N 2 -filled glovebox. The films were then annealed at 100 °C for 5 minutes. The electrodes were deposited by transferring the samples into a vacuum chamber. V 2 O 5 (2 nm)/Ag (100 nm) was then thermally deposited on top of the photoactive layer with the help of a shadow mask having a 0.0555 cm 2 device area.
Characterization of the solar cells. The electrical characteristics were measured using a Keithley 4200 unit under 100 mW cm −2 AM1.5 solar illumination in a N 2 -filled glove box. The light was generated using an Oriel 1-kW solar simulator referenced using a Reference Cell PVM 132 calibrated at the US National Renewable Energy Laboratory. The external quantum efficiency was measured using a photomodulation spectroscopy setup (Merlin, Oriel) with monochromatic light from a Xenon lamp. The power density of the monochromatic light was calibrated using a Si photodiode certified by the National Institute for Standards and Technology. Light intensity-dependent J-V curves were measured using the same solar simulator setup by varying the intensity from 0.17-1 sun.
Preparation of the control devices and their characterization. The hole-and electron-only control devices were prepared with the device structure ITO/PEDOT:PSS/donor:acceptor/Au and ITO/ZnO/donor:acceptor/Ca/Al, respectively. The electrical characteristics were measured using the Keithley 4200 in a N 2 -filled