Regio-regular alternating diketopyrrolopyrrole-based D1–A–D2–A terpolymers for the enhanced performance of polymer solar cells

We designed and synthesized regio-regular alternating diketopyrrolopyrrole (DPP)-based D1–A–D2–A terpolymers (PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT) using a primary donor (D1) [3,3′-difluoro-2,2′-bithiophene (F2T2)] and a secondary donor (D2) [2,2′-bithiophene (T2), (E)-1,2-di(thiophen-2-yl)ethene (TVT), or dithieno[3,2-b:2′,3′-d]thiophene (DTT)]. A PDPP2DT-F2T2 D–A polymer was synthesized as well to compare optical, electronic, and photovoltaic properties. The absorption peaks of the terpolymers (PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT) were longer (λmax = 801–810 nm) than the peak of the PDPP2DT-F2T2 polymer (λmax = 799 nm), which is associated with the high-lying HOMO levels of the terpolymers (−5.08 to −5.13 eV) compared with the level of the PDPP2DT-F2T2 polymer (−5.38 eV). The photovoltaic properties of these DPP-based polymers were investigated under simulated AM 1.5G sunlight (100 mW cm−2) with a conventional structure (ITO/PEDOT:PSS/polymer:PC71BM/Al). The open-circuit voltages (Voc) of photovoltaic devices containing the terpolymers were slightly lower (0.68–0.70 V) than the Voc of the device containing the PDPP2DT-F2T2 polymer (0.79 V). The short-circuit current (Jsc) of the PDPPF2T2DPP-DTT device was significantly improved (14.14 mA cm−2) compared with that of the PDPP2DT-F2T2 device (8.29 mA cm−2). As a result, the power conversion efficiency (PCE) of the PDPPF2T2DPP-DTT device (6.35%) was increased by 33% compared with that of the simple D–A-type PDPP2DT-F2T2 device (4.78%). The highest Jsc and PCE values (the PDPPF2T2DPP-DTT device) were attributed to an optimal nanoscopically mixed morphology and strong interchain packing with a high face-on orientation in the blend film state. The study demonstrated that our strategy of using multiple donors in a regio-regular alternating fashion could fine-tune the optical, electronic, and morphological properties of D–A-type polymers, enhancing the performance of polymer solar cells.


Introduction
Polymer solar cells (PSCs) have been investigated as a future energy source because of advantages such as the low-cost of source materials and production and the potential for fabricating lightweight, exible products. [1][2][3] Improving the photovoltaic properties of PSCs requires the development of appropriate donor-acceptor-type conjugated polymers consisting of an electron-donating unit (donor or D) and an electronaccepting unit (acceptor or A). D-A polymers have been synthesized and demonstrated the desired energy level, energy bandgap, and light absorption properties. 4 In the blend lm where D-A polymers and n-type molecules are combined to form a photoactive layer, they are required to exhibit high crystallinity, good carrier transport, and optimal morphology; 5 PSC devices made using these photoactive materials would have a high open-circuit voltage (V oc ), short-circuit current (J sc ), llfactor (FF), and power conversion efficiency (PCE). However, because of the synthetic difficulty in developing new donor and acceptor units, combining known D or A moiety derivatives is a common approach, which could limit the ne-tuning of the optoelectronic properties of conjugated polymers for highperformance PSCs. To overcome these limitations, the use of D 1 -A-D 2 -A terpolymers has recently emerged as an effective strategy for constructing high-performance conjugated polymers. 6,7 D 1 -A-D 2 -A terpolymers are composed of three monomers with one acceptor unit and two different donor units in a conjugated polymer backbone. The use of the D 1 -A-D 2 -A structure allows the production of a number of conjugated copolymers by varying the D and A monomer combination in the polymerization process. Moreover, this rational polymer structural design would enhance physical properties such as energy level, energy bandgap, and light absorption properties. In particular, the regio-regular connection of D and A units seems to be preferable for high photovoltaic performance. 8, 9 2,5-Dihydropyrrolo [4,3-c]pyrrolo-1,4-dione (DPP) moiety, rst reported in 1974 by Farnum, has been utilized as an acceptor unit for organic eld-effect transistors (OFETs) and organic photovoltaics (OPVs). [10][11][12][13][14][15] Aromatic substituents (such as phenyl or thienyl groups) to the DPP unit are oen used to strongly modulate electrical and optical properties. As the DPP unit is strongly electron-withdrawing and forms a planar conjugated backbone, DPP-based polymers have high charge carrier mobilities and small bandgaps. [16][17][18][19][20] In addition, uorine atoms have been used in recently developed high-performance D-A polymers. 21 The strong electronegativity of the uorine atoms can enhance the oxidative stability of the D-A polymers by lowering the highest occupied molecular orbital (HOMO) energy level. More importantly, uorine atoms substituted in the conjugated backbone can induce strong interchain interactions through the polar C-F bond dipole and intramolecular interactions with close-lying protons or sulfur atoms in neighboring conjugated units (i.e., through-space H/F and S/F interactions). These interactions can enhance the orbital overlap between aromatic units, promoting backbone planarity and interchain stacking. [22][23][24] We have recently investigated the optical and electrical properties of a DPP-based D-A copolymer (PDPP2DT-F2T2) synthesized using DPP and 3,3 0 -diuoro-2,2 0bithiophene (F2T2). A comparison with PDPP2DT-T2 synthesized using DPP and 2,2 0 -bithiophene (T2) units revealed that the F2T2 donor moiety could facilitate a planar polymer backbone conformation and enhance crystallinity and carrier transport. 25 In this study, we synthesized three D 1 -A-D 2 -A-type DPPbased terpolymers, in which the D 1 unit was xed with the F2T2 donor, and D 2 units were varied using T2, (E)-1,2di(thiophen-2-yl)ethene (TVT), and dithieno[3,2-b:2 0 ,3 0 -d]thiophene (DTT). The variation of the second donor moiety enabled the ne-tuning of electronic structures, backbone conformation, and the resulting photovoltaic performance. The structural change caused signicant changes in physical properties and photovoltaic performances. In comparison with the PDPP2DT-F2T2 polymer, the PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT polymers had high-lying HOMO levels and lower bandgaps. Among PSC devices containing polymer: [6,6]-phenyl-C 71 butyric acid methyl ester (PC 71 BM) blend lms, the performance of the device containing the PDPPF2T2DPP-DTT:PC 71 BM blend lm was the best. When diphenyl ether (DPE) was used as a processing additive, photovoltaic performances were further improved. The D 1 -A-D 2 -A-type PDPPF2T2DPP-DTT device exhibited a PCE of 6.35%, which was a great improvement compared with the PCE of the simple D-A-structured PDPP2DT-F2T2:PC 71 BM device. Here, we highlight the advantages of using D 1 -A-D 2 -A polymers (PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT) and explain the origin of the difference in photovoltaic property among the PSCs in detail.
Synthesis of poly(3- 1302 mmol), tetrakis(triphenyl-phosphine)palladium(0) (Pd(PPh 3 ) 4 ) (0.006 g, 4 mol%) were added to a ame dried one-neck round bottomed ask. Degassed DMF (1.1 mL) and toluene (5.5 mL) were added to the ask and the solution was heated initially to 60 C with stirring. The reaction temperature was raised to 90 C gradually at a rate of 1 C/5 min. Aer 6 min, the reaction solution became a gel state. The reaction temperature was cooled down to 60 C and chloroform (5 mL) was added was added to the reaction mixture. Aer vigorous stirring for 1 h, the reaction mixture was transferred to diethylammonium diethyldithiocarbamate (1 M) aqueous solution (50 mL) in 250 mL round bottomed ask using hot chloroform (30 mL) and stirred at 50 C for another 1 h. The polymer solution was extracted with chloroform (30 mL Â 3). The collected organic layer was then washed brine and deionized water. Aer removing organic solvent under reduced pressure, the crude polymer was dissolved in chloroform (21 mL) and then was precipitated in methanol (300 mL). The collected polymer was further puried by Soxhlet extraction using methanol, acetone, hexane, cyclohexane, dichloromethane, and chloroform. The chloroform fraction was precipitated in methanol, ltered, and dried under vacuum to yield PDPPF2T2DPP-DTT polymer (0.2819 g, 92.6% yield). Gelpermeation chromatography (GPC) (o-dichlorobenzene, 80 C)

Characterizations and measurements
To identify the molecular structures of all the synthesized products, 1 H-NMR, 13 C-NMR and 19 F-NMR spectra were taken on Bruker Avance III 300MHz. Average molecular weights and polydispersity index (PDI) of synthesized polymers were determined by an Agilent GPC system (GPC 1200 system) at 80 C. o-Dichlorobenzene was an eluent, and polystyrene standards were used for molecular weight calibration. UV-visible absorption spectra were taken on Agilent 8453 UV-Vis spectrophotometer. Differential scanning calorimetry (DSC) data were obtained on DSC Q2000 differential scanning calorimeter from TA instruments. Heating and cooling temperature was scanned at a rate of 10 C min À1 . Thermogravimetric analysis (TGA) data were obtained on TGA Q50 under N 2 atmosphere at a temperature scan rate of 25 C min À1 . Cyclic voltammetry (CV) was performed using a CorrTest instruments and 0.1 M tetrabutylammonium hexauorophosphate (TBAPF 6 ) solution in anhydrous acetonitrile was prepared as an electrolyte. Pt wires were used as the counter electrode and working electrode. Synthesized polymers were coated on the Pt wire working electrode; Ag wire was used as a reference electrode. Ferrocene (Fc) was used as the internal standard; ferrocene oxidation (Fc/Fc + ) potential was assumed to be À4.8 eV. The voltage sweep rate was 50 mV s À1 . HOMO levels were determined from the onsets of the anodic curves. Density functional theory (DFT) calculations was conducted as follows: HOMO and LUMO energies were calculated by quantum calculation as implemented in Gaussian 09 package. Geometry optimization, single point calculation and frequency analysis were performed using density functional theory (B3LYP method) with 6-311G(d) basis set. In order to obtain the representative structure of the target molecules, the global minimum structures were searched in two steps. Firstly, the optimized structures of fragment molecules consisting of two different residues were determined by the calculated potential energy surface as a function of dihedral angle between residues. The model molecules for the investigated polymers was built based on the optimized structure of the fragment molecule. Tetramer, dimer, dimer and dimer for PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT and PDPPF2T2DPP-DTT, respectively were chosen and the bulky decyltetradecyl groups were replaced with methyl groups to reduce computational cost. The model molecules were freely optimized and conrmed by frequency analysis. The global minimum structures were used to determine the HOMO and LUMO energies. GIWAXS measurements were accomplished at PLS-II 9A U-SAXS beamline of the Pohang Accelerator Laboratory in Republic of Korea. X-rays coming from the in-vacuum undulator (IVU) were monochromated (E k ¼ 11.065 keV, wavelength l ¼ 1.10994Å) using a Si(111) double crystal monochromator and focused both horizontally and vertically at the sample position (450 (H) Â 60 (V) mm 2 in FWHM@sample position) using K-B type mirrors system. GIWAXS sample stage was equipped with a 7-axis motorized stage for the ne alignment of sample, and the incidence angle of X-ray beam was set to be 0.12 -0.14 for polymer lms and polymer:PC 71 BM blend lms. GIWAXS patterns were recorded with a 2D CCD detector (Rayonix SX165) and X-ray irradiation time was 6-9 s, dependent on the saturation level of the detector. Diffraction angles were calibrated using a sucrose standard (Monoclinic, P2 1 , a ¼ 10.8631Å, b ¼ 8.7044Å, c ¼ 7.7624Å, b ¼ 102.938 ) and the sample-to-detector distance was $231 mm. Samples for GIWAXS measurements were prepared by spin-coating polymers and polymer:PC 71 BM blend solutions with various ratios on top of the PEDOT:PSS coated Si wafer substrates. AFM characterization: an Agilent 5500 scanning probe microscope (SPM) running with a Nanoscope V controller was used to obtain AFM images of polymer:PC 71 BM blend thin lms. AFM images were recorded in high-resolution tapping mode under ambient conditions. Premium silicon cantilevers (TESP-V2) were used with a rotated tip to provide more symmetric representation of features over 200 nm.

Device fabrication
Glass/patterned indium tin oxide (ITO) substrates were cleaned with detergent and ultra-sonicated in deionized water, acetone, and isopropyl alcohol sequentially and dried in an oven for 12 h. The substrates were subjected to UV/ozone treatment for 15 min and then poly(3,4-ethylenedioxythiophene):polystyrene sulfonic acid (PEDOT:PSS, CLEVIOS P VP Al4083) was spin-coated on top of the cleaned ITO substrates through 0.45 mm cellulose acetate syringe lter and were dried at 140 C to remove moisture. The substrates were transferred to globe box lled in N 2 and photoactive solution was spin-cast. These polymers (12 mg mL À1 ) were blended with PC 71 BM with various ratio in chlorobenzene. 100 nm thick Al electrode was evaporated on top of the photoactive layer under high vacuum (<10 À6 torr) through a mask. The active area of the device is 3.67 mm 2 . The photovoltaic characteristics were measured in glove box by using a high quality optical ber to guide light from the solar simulator. The current density-voltage (J-V) characteristics of the PSC were measured using a Keithley 2635A source measurement unit under AM 1.5G illumination at 100 mW cm À2 . The incident photon-to-current efficiency (IPCE) was measured by using a QEX7 from PV measurement Inc. For light intensity dependence measurement, the devices were placed under a solar simulator by using a set of neutral density lters. Neutral density lters have the ability to block a certain amount of the light, which reduces the intensity.

Polymer synthesis and thermal properties
The synthetic routes for all of the DPP-based terpolymers as well as the PDPP2DT-F2T2 polymer are shown in Scheme 1. To synthesize the terpolymers, compound 3, a key monomer, was prepared via three steps: monobromination of DPP2DT, Stille coupling with 2,6-bis(trimethylstannyl)dithieno[3,2-b:2 0 ,3 0 -d] thiophene compound, and dibromination. Polymers were prepared via the Stille polymerization of an acceptor monomer (Br-DPP2DT-Br or Br-DPPF2T2DPP-Br) and a bis(trimethylstannyl) donor monomer (F2T2, T2, TVT, or DTT) using Pd(PPh 3 ) 4 catalyst in a toluene:DMF cosolvent system. The crude polymer solutions were treated with 1 M diethylammonium diethyldithiocarbamate solution to remove the Pd-catalyst, precipitated in methanol, and further puried by the Soxhlet extraction method. The nal fraction was precipitated and dried. The resulting polymers were obtained with a high yield (up to 83%) and were highly soluble in organic solvents like chloroform, toluene, and chlorobenzene (CB). Thermal stability was determined by thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC). TGA results revealed that all terpolymers had a decomposition temperature (T d ) of > 400 C, while DSC data showed no noticeable transitions (see Fig. S12 †).

Optical and electrochemical properties
Optical and electrochemical properties of the synthesized polymers were summarized in the Fig. 1 and Table 1. Fig. 1a and b shows that all of the synthesized polymers had dual absorption regions at 400-500 nm and 700-800 nm, which are attributed to two different p-p transitions. The vibronic peaks in the 700-800 nm region were more pronounced in the lm state ( Fig. 1b) than in solution (Fig. 1c). The maximum absorption peaks in the solution were slightly more red-shied in the lm state (PDPP2DT-F2T2: 791 nm / 799 nm, PDPPF2T2DPP-T2: 791 nm / 801 nm, PDPPF2T2DPP-TVT: 791 nm / 803 nm, PDPPF2T2DPP-DTT: 803 nm / 810 nm). This nding suggests that polymer aggregation was increased in the lm state. In particular, the PDPPF2T2DPP-DTT polymer had not only a red-shied absorption peak but also a pronounced vibronic peak at 810 nm, indicating that the PDPPF2T2DPP-DTT polymer signicantly improved interchain interactions in the lm state. Optical bandgaps (E opt g s) were estimated from the absorption onsets of the polymer lms, which were 1.40, 1.43, 1.42, and 1.41 eV for PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT, respectively.
Cyclic voltammetry (CV) measurements were conducted to investigate the electrochemical properties of the synthesized polymers. Fig. 1c shows the cyclic voltammograms of the polymer lms, which were measured in a 0.1 M tetrabutylammonium hexauorophosphate (TBAPF 6 ) solution of anhydrous acetonitrile. The onsets of oxidation potential for PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT were 0.49, 0.30, 0.28, and 0.33 V, corresponding to À5.38, À5.10, À5.08, and À5.13 eV of HOMO energy levels, respectively. Fluorine atoms in the PDPP2DT-F2T2 effectively lowered the HOMO level of the PDPP2DT-T2 polymer that has the same chemical structure except uorine atoms, 25 whereas electron-donating units (T2, TVT, and DTT moieties) slightly increased the HOMO levels compare to that of Scheme 1 Synthetic routes for the DPP-based regio-regular terpolymers.

Density functional theory (DFT) calculations
Molecular backbone geometries and electronic properties were evaluated using Gaussian 09 program package at the DFT level (B3LYP, 6-311G(d)). The model molecules of the polymers were (DPP2Me-F2T2) 4 , (DPP2MeF2T2DPP2Me-T2) 2 , (DPP2MeF2T2DPP2Me-TVT) 2 , and (DPP2MeF2T2DPP2Me-DTT) 2 , where the long 2-decyltetradecyl groups were superseded by methyl groups. The energy-minimized molecular geometries of the model molecules are shown in Fig. 2a-d. The dihedral angles (DAs) were examined. The DA between two uorothiophene rings in the F2T2 unit was almost 0 through S/F interaction; however, there was some degree of distortion (17)(18)(19) between the F2T2 units and neighboring thiophenes due to steric hindrance in the (DPP2Me-F2T2) 4 structure. When the F2T2 units were replaced with the T2, TVT, and DTT units, the DA values between the F2T2 unit and the neighboring thiophene were decreased to 13-16 for the le half structure or almost 0 for the right half structure, and the planarity was improved in the order of T2, TVT, and DTT. In addition, the replacement of the F2T2 units with T2 units increased the DAs in the bithiophene regions corresponding to the F2T2 units, whereas the DAs between thiophene rings connected to the DPP moiety and T2 units were reduced compared with those of the F2T2-based structure, i.e., (DPP2Me-F2T2) 4 . Accordingly, the overall twists of (DPP2MeF2T2DPP2Me-T2) 2 appeared to be similar to those of (DPP2Me-F2T2) 4 . On the other hand, the additional planar moieties of the TVT and DTT units resulted in more planar structures. In comparison with the UV-visible absorption spectra of the polymers in the solution state (Fig. 1a), the contribution of DA differences was minor because the l peak values were nearly the same. This result is consistent with previous studies reporting that DA values under 30 do not signicantly alter electronic properties. 27 However, the high coplanarity greatly affected interchain interactions in the lm state (Fig. 1b), especially for the (DPP2MeF2T2DPP2Me-DTT) 2 structure.

Photovoltaic characteristics
We investigated the photovoltaic characteristics of PSC devices containing the DPP-based polymers under simulated AM 1.5G sunlight (100 mW cm À2 ) with a conventional structure (ITO/ PEDOT:PSS/polymer:PC 71 BM/Al).
To optimize the polymer:PC 71 BM blend ratio, blend lms were processed using pure CB or CB containing 3 vol% DPE as the solvent. The optimal polymer:PC 71 BM ratios were 1 : 2, 1 : 3, 1 : 3, and 1 : 2 (w/w) for the PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT polymers, respectively. The device current-voltage (J-V) characteristics and IPCE spectra of each optimized device are shown in Fig. 3, and the device photovoltaic parameters are summarized in Table 2. The PDPP2DT-F2T2 device had the highest V oc (0.79 V). This high V oc , one of the highest V oc values among the DPP polymer-based PSCs, may be correlated with the lower-lying HOMO level of the PDPP2DT-F2T2 lm due to the high number of F atom substituents in the conjugated backbone. This can be conrmed by comparison with PSCs containing PDPPF2T2DPP-T2 (F atom substitution in the alternate quarter thiophene bridges) and DT-PDPP4T 14 (no F atom substitution in the quarter thiophene bridges), which had V oc values of 0.70 and 0.64 V, respectively. In comparison with devices containing PDPP2DT-F2T2 and PDPPF2T2DPP-TVT, devices containing PDPPF2T2DPP-T2 and PDPPF2T2DPP-DTT demonstrated a higher performance with a maximum PCE of 3.20 and 3.01%, J sc of 6.32 and 6.59 mA cm À2 , V oc of 0.70 and 0.69 V, and FF of 0.70 and 0.66, respectively.
Processing additives can be used to modify the morphology of the photoactive layer, greatly improving device performance. [28][29][30] Following optimization with various processing additives, DPE was found to be suitable for our DPP polymerbased PSCs. As shown in Fig. 3 and Table 2, the addition of 3 vol% DPE to CB led to a marked enhancement in all device parameters. In particular, the PCEs were signicantly increased from 2.18, 3.20, 1.77, and 3.01% to 4.78, 5.89, 3.20, and 6.35% for devices containing PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT, respectively. High FF values (0.66-0.73) were obtained for all devices, and J sc values were a main factor for determining differences in PCE values. The optimized thicknesses of photoactive layers processed with the DPE additive were 100-170 nm. Among four devices containing DPP polymer-based PSCs processed with the CB:DPE solvent, the PDPPF2T2DPP-DTT device had the best PCE (6.35%) with a J sc of 14.14 mA cm À2 , V oc of 0.68 V, and FF of 0.66. The PDPPF2T2DPP-DTT device had the highest IPCE values in the whole wavelength region (300-900 nm) as well. In

Charge transport characteristics
Vertical charge transport characteristics in the photoactive lms was evaluated using a space charge limited current (SCLC) model. Hole-only (ITO/PEDOT:PSS/polymer:PC 71 BM/Au) and electron-only (uorine-doped tin oxide (FTO)/polymer:PC 71 BM/ Al) devices were fabricated under optimized device fabrication conditions. To ensure accuracy in carrier mobility estimation, the potential loss due to the series resistance of the ITO and the built-in potential were taken into consideration. J-V characteristics (Fig. 4) showed a quadratic dependence on voltage over a range of several volts, which was consistent with the Mott-Gurney equation: 31,32 where 3 0 is the free-space permittivity, 3 r is the dielectric constant of the polymer:PC 71 BM blend lms, m is the mobility, V is the applied voltage and L is the thickness of the photoactive lms. As shown in Table 3, the hole (m h ) mobilities were 1.08 Â 10 À3 , 3.19 Â 10 À3 , 7.03 Â 10 À4 , and 1.74 Â 10 À3 cm 2 V À1 s À1 , and the electron (m e ) mobilities were 2.75 Â 10 À3 , 4.21 Â 10 À3 ,    Table S1 †) were compared with those of the polymer:PC 71 BM blend lms, the hole mobilities in the blend lms were decreased slightly whereas the electron mobilities were improved signicantly because of using PC 71 BM.
To gain insights on the charge recombination and charge extraction of PSCs containing the terpolymers, we examined the light intensity dependence of J-V characteristics under shortcircuit conditions. Fig. 5a shows a log-log plot of J sc as a function of the light intensity. The J-V curves were tted according to the power-law dependence of J sc on the light intensity: where I is the light intensity, and a is the exponent constant for polymer:PC 71 BM-based PSCs. The a values were 0.9698, 0.9893, 0.9661, and 0.9868 for the devices containing the PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT polymers, respectively. The devices containing PDPPF2T2DPP-T2 and PDPPF2T2DPP-DTT had a values closer to 1, indicating little bimolecular recombination under short circuit conditions. The device containing PDPPF2T2DPP-TVT had a relatively lower a value, indicating a relatively larger degree of bimolecular recombination due to a poorer morphology (see below) and lower charge transport. In addition, we examined the dependence of the net photocurrent (J ph ) on the effective voltage (V eff ) (Fig. 5b). J ph is the difference between the current density under illumination and dark conditions of the PSCs (J ph ¼ J L À J D ). V eff is the difference between the compensation voltage (V 0 ) at J ph ¼ 0 and applied bias voltage (V), i.e., V eff ¼ V 0 À V. At a high V eff , the photocurrent was saturated without recombination, demonstrating that all photo-generated charges were collected at the electrodes. The devices containing PDPPF2T2DPP-T2 and PDPPF2T2DPP-DTT had the highest J ph s, whereas those containing PDPP2DT-F2T2 and PDPPF2T2DPP-TVT had a substantially lower J ph s. With increasing V eff , the photocurrent gradually became saturated for the blend lms with PDPP2DT-F2T2, PDPPF2T2DPP-T2, and PDPPF2T2DPP-DTT at V eff ¼ $0.80 V; however, the blend lm with PDPPF2T2DPP-TVT did not exhibit a saturation region over 1 V. The devices containing PDPP2DT-F2T2, PDPPF2T2DPP-T2, and PDPPF2T2DPP-DTT had high J ph /J sat ratios (96%, 99%, and 99%, respectively), where the ratio of J ph to J sat is the product of charge dissociation and collection probabilities. However, the  device containing PDPPF2T2DPP-TVT had a slightly lower J ph / J sat ratio (89%) due to the high bimolecular recombination. These results are in good agreement with the mobility and light intensity-dependent J sc data.

Morphological characteristics
To analyze the improvement in device performance caused by the use of the DPE additive, we examined the polymer:PC 71 BM blend lms by atomic force microscopy (AFM), eld-emission transmission electron microscopy (FE-TEM), and twodimensional grazing incidence wide-angle X-ray scattering (GIWAXS). According to AFM results, the topography and phase images of photoactive lms with and without DPE addition showed a clearly different morphology ( Fig. 6 and S14 † To examine the morphology of polymer:PC 71 BM blend lms processed with DPE additive, FE-TEM images were also taken as shown in Fig. 7. Considerable differences in the morphology were observed; nevertheless, all of the blend lms had interconnected bril structures, which were formed by DPP polymers. 17,28,33 The PDPP2DT-F2T2:PC 71 BM and PDPPF2T2DPP-TVT:PC 71 BM lms contained thick aggregated nanobers ( Fig. 7a and c). Due to the large polymer aggregation, the effective interfacial area between the polymer chains and PC 71 BM molecules for efficient exciton separation would have been relatively reduced, resulting in the moderate PCEs. The  PDPPF2T2DPP-T2:PC 71 BM lm had sparsely distributed very narrow nanobers (Fig. 7b), which is consistent with the lowest carrier mobility and the lowest PCE in this study. In comparison with other blend lms, the PDPPF2T2DPP-DTT:PC 71 BM blend lm had well developed brillary structures with an even distribution (Fig. 7d). This type of nanobrillar crystalline morphology is essential for achieving high device performance because it allows efficient charge transport. It is worth noting that the PDPPF2T2DPP-DTT polymer had brils and at the same time good miscibility with PC 71 BM domains. Considering that the DPP polymers in this study were not signicantly different in terms of their geometries, it appears that a fused ring structure of the DTT moiety might facilitate p-p interactions between the PDPPF2T2DPP-DTT polymer chains and PC 71 BM molecules.
To further investigate the morphologies of the four DPP polymer-based blend lms in terms of the orientation and packing of polymer chains, 2D-GIWAXS images were obtained. Fig. 8 and S15 † show the 2D-GIWAXS images and in-plane and out-of-plane line-cut proles of pristine polymers and polymer:PC 71 BM blend lms without and with DPE addition. The extracted 2D-GIWAXS scattering features are summarized in Table S2. † The patterns of the pristine DPP polymer lms were similar, showing strong lamellar scattering up to (500) in the out-of-plane direction with an edge-on orientation. From the out-of-plane (100) peaks, the estimated lamellar d-spacing values were similar (23.43, 21.95, 21.95, and 22.11Å for PDPP2DT-F2T2, PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT, respectively), as expected based on the chemical structures. All of the pristine DPP polymer lms also demonstrated strong (010) p-p stacking peaks in the in-plane direction with the same d-spacing distance (3.63Å). In case of the polymer:PC 71 BM blend lms without DPE addition, there was no large difference in the packing structure of the PDPP2DT-F2T2, PDPPF2T2DPP-T2, and PDPPF2T2DPP-TVT polymers; the lamellar and p-p stacking peaks were signicantly decreased in the three DPP polymer:PC 71 BM blend lms. In contrast, the PDPPF2T2DPP-DTT:PC 71 BM blend lm  exhibited clear p-p stacking in the out-of-plane direction. Upon the use of the DPE additive, we observed that p-p stacking peaks in the out-of-plane direction reappeared for the PDPP2DT-F2T2:PC 71 BM and PDPPF2T2DPP-T2:PC 71 BM blend lms, and they became more pronounced for the PDPPF2T2DPP-DTT:PC 71 BM blend lm. At the same time, (100) peaks in the in-plane direction were sharply increased for all of the DPP polymer:PC 71 BM blend lms. These changes suggest that the DPE additive could facilitate the ordering and stacking of DPP polymer chains while promoting a face-on orientation. Because the face-on orientation is benecial for charge transport and extraction in the vertical direction, these results clearly supported the high photovoltaic performance of the PDPP2DT-F2T2, PDPPF2T2DPP-T2, and PDPPF2T2DPP-DTT devices compared with that of the PDPPF2T2DPP-T2 device.

Conclusions
We investigated the photovoltaic properties of newly designed DPP-based D 1 -A-D 2 -A-type terpolymers (PDPPF2T2DPP-T2, PDPPF2T2DPP-TVT, and PDPPF2T2DPP-DTT; F2T2 as the D 1 unit and either T2, TVT, or DTT as the D 2 unit) as well as the D-A-type PDPP2DT-F2T2 polymer. Depending on the D 2 unit, electronic levels and bandgaps were nely modulated; the terpolymers exhibited slightly higher lying HOMO energy levels. Measurement of photovoltaic properties revealed that the PSC containing the PDPPF2T2DPP-DTT:PC 71 BM blend lm had the highest PCE (6.35%) with DPE addition. This good photovoltaic performance may be attributed to the face-on crystalline features of the PDPPF2T2DPP-DTT polymer in the polymer:PC 71 BM blend lm, which could facilitate good carrier transport, and the close interaction of the long nanoscale brils of the PDPPF2T2DPP-DTT polymer with the PC 71 BM domains. We emphasize that the PCE of the PDPPF2T2DPP-DTT device was much higher than that of the D-A-type PDPP2DT-F2T2 device (4.78%). As the DTT unit is a stronger donor than the F2T2 unit, the V oc value of the PDPPF2T2DPP-DTT device was slightly lower. However, the DTT unit generated a more ideal morphology and chain orientation; thus, J sc values were markedly improved from 8.29 to 14.14 mA cm À2 . This study highlights the great potential of using D 1 -A-D 2 -A-type conjugated polymers rather than simple D-A-type polymers to tailor electronic properties and morphology for fabricating highperformance PSCs.

Conflicts of interest
There are no conicts to declare.