Introduction

Recent advances in the area of solution-processed electronic devices, such as organic field-effect transistors (OFETs) and organic photovoltaics (OPVs), have been achieved due to considerable research on π-conjugated polymers that are used as semiconducting materials1,2,3,4,5,6,7,8. The performance of these devices is highly dependent on the structural order of π-conjugated polymers in the thin film. π-Conjugated polymers with high backbone coplanarity show strong intermolecular interactions and are able to self-organize to form a crystalline structure9,10. With the long-range order of the π–π stacked conjugated backbones as well as that along the backbone in the crystalline phase, the polymers can provide high charge transport, improving device performance. However, similar to other polymer materials, π-conjugated polymers also contain amorphous phases11,12, where the backbones are assumed to be less coplanar and more disordered than those in the crystalline phases; specifically, long-range order is absent. Therefore, the amorphous phase is generally thought to be detrimental to charge transport, limiting the device performance, although there are some exceptions3.

In bulk heterojunction OPVs, the relationship between polymer order and device performance is slightly more complex. Since charge separation occurs at the polymer donor and acceptor heterojunction, the donor–acceptor mixed phase plays a critical role in determining the photovoltaic performance13,14,15. However, the mixed phase, where the materials are thought to be amorphous, limits the charge transport. If the polymer order can be controlled or a high polymer order can be obtained in the mixed phase, the photovoltaic performance will be greatly improved.

Here, we synthesized two π-conjugated polymers based on benzodithiophene (BDT) and thiazolothiazole (TzTz) fused rings, named PSTz1 and PSTz2 (Fig. 1a), which possess branched alkyl (2-ethylhexyl) and tripropylsilyl groups on the BDT moiety, respectively. Interestingly, these two polymers showed highly contrasting changes in the backbone order when the environment changed from a solution to a neat film and to the polymer/acceptor blend film; here, fullerene [6,6]-phenyl-C71-butyric acid methyl ester (PC71BM) and nonfullerene IT-4F were used as the acceptor (Fig. 1b). More specifically, PSTz1 had a higher backbone order in the aggregated state, such as the polymer neat film and the polymer/PC71BM blend film, than in the dispersed and/or isolated state, such as the solution and the polymer/IT-4F blend film; however, PSTz2 had a higher backbone order in the isolated state than in the aggregated state. Importantly, due to the significantly high backbone order in the IT-4F blend film that was nearly amorphous as a whole, PSTz2 showed much higher photovoltaic performance than PSTz1.

Fig. 1: Chemical structures of the materials.
figure 1

a PSTz1 and PSTz2 and b fullerene acceptor PC71BM and nonfullerene acceptor IT-4F.

Full size image

Results and discussion

Polymer properties

Both polymers were synthesized by copolymerization of the corresponding distannylated BDT monomer and the dibrominated TzTz monomer (Supplementary Figs. 1–3). The polymers were soluble in warm chlorinated solvents, such as chlorobenzene (CB) and o-dichlorobenzene (DCB). The number-average molecular weight (Mn) and dispersity (Ð) were 45,000 and 2.2 for PSTz1 and 70,700 and 2.1 for PSTz2, respectively (Supplementary Table 1). The highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) energy levels determined by cyclic voltammetry measurements (Supplementary Fig. 4) were –5.37 and –3.35 eV for PSTz1 and –5.49 and –3.48 eV for PSTz2, respectively; these results were consistent with computation (Supplementary Fig. 5). The polymers showed glass transition peaks at 270 °C (PSTz1) and 305 °C (PSTz2), as observed by differential scanning calorimetry (DSC) (Supplementary Fig. 6).

Photovoltaic properties

We investigated the photovoltaic properties of the polymers by fabricating inverted cells (ITO/ZnO/polymer:acceptor/MoOx/Ag), where the polymer to acceptor (D/A) weight ratio was 1:2 and 1:1.5 (w/w) for the cells based on PC71BM and IT-4F, respectively. Figure 2a and c show the current density (J)–voltage (V) curves, and Fig. 2b and d depict the external quantum efficiency (EQE) spectra of the cells based on PC71BM and IT-4F, respectively. Table 1 summarizes the photovoltaic parameters. The short-circuit current density (JSC) values for all the cells were consistent with those calculated from the EQE characteristics. In the PC71BM cells, both polymers showed similar power conversion efficiencies (PCEs) of approximately 6% (5.7% for PSTz1 and 5.9% for PSTz2) and similar JSC and fill factor (FF) values. In stark contrast, in the IT-4F cells, PSTz2 exhibited a PCE of 13.0% with a JSC of 18.6 mA cm−2 and an FF of 0.76; all these results were significantly higher than those of PSTz1 (PCE = 8.9%, JSC = 14.4 mA cm−2, and FF = 0.70). In addition, in both IT-4F and PC71BM cells, PSTz2 consistently gave a higher open-circuit voltage (VOC) than PSTz1, which was ascribed to the deeper HOMO energy level for PSTz2 than for PSTz1. Notably, PSTz2 with an Mn of 45,500, which was similar to PSTz1, showed a slightly lower PCE of ca. 12% but consistently had higher performance than that of PSTz1 (Supplementary Fig. 7).

Fig. 2: Organic photovoltaic characteristics of PSTz1 and PSTz2.
figure 2

a, c Current density–voltage curves and b, d external quantum efficiency (EQE) spectra of photovoltaic cells. a, b Polymer/PC71BM cells and c, d polymer/IT-4F cells.

Full size image
Table 1 Photovoltaic parameters of cells based on PSTz1 and PSTz2.
Full size table

Polymer backbone order in solution and neat film

We examined the polymer backbone order by UV–vis absorption spectroscopy. Both PSTz1 and PSTz2 showed absorption spectra spanning the 400 to 650 nm wavelength range in both solution and neat films, in which vibronic structures with two absorption maxima assignable to the 0–0 and 0–1 transition bands were observed (Fig. 3a, e). The spectrum of PSTz1 in the neat film was slightly redshifted by approximately 10 nm relative to the spectrum of PSTz1 in the solution, and the 0–0 band intensified; the intensity ratio of the 0–0 band with respect to the 0–1 band (I0–0/I0–1) was 1.03 and 1.12 in the solution and the neat film, respectively (Fig. 3a). This result indicated that the PSTz1 polymer backbone became more coplanar and/or ordered in the solid state than in the solution state, likely as a result of self-organization, which is often the case for π-conjugated polymers. Interestingly, on the other hand, PSTz2 showed an extremely intense 0–0 band with I0–0/I0–1 of 1.47 in the solution, which was significantly weakened in the neat film (I0–0/I0–1 = 1.13) (Fig. 3e). The change in the spectrum between the solution and the neat film sharply contrasted that in PSTz1 despite having an identical backbone. As a result, the absorption spectra of PSTz1 and PSTz2 in the neat film were very similar even though the spectra in the solution were quite distinct. This unconventional behavior in PSTz2 indicated that the backbone was highly coplanar and rigid and thus highly ordered when the polymer chains were likely isolated; however, when the polymer chains were aggregated, they became less ordered. We speculated that the three-dimensional bulky tripropylsilyl groups in PSTz2 could potentially hinder interaction among the backbones, resulting in the high order in the solution. Conversely, the bulky tripropylsilyl groups could sterically diminish the backbone order of adjacent polymer chains when the polymers were closely packed in the aggregates (thin film). Notably, although the backbone order of PSTz2 was reduced in the film, it was still similar to that of PSTz1.

Fig. 3: Investigation of polymer order by UV‒vis absorption spectroscopy and GIXD measurement.
figure 3

UV−vis absorption spectra of a–c PSTz1 and e–g PSTz2. a,e Absorption spectra in chlorobenzene (CB) solution, neat film, and polystyrene (PS) blend film where the polymer to PS weight ratio is 1:5. b,f Absorption spectra in solutions of CB and methanol with different CB:methanol ratios. c,g Temperature-dependent absorption spectra in CB solution measured at 20–100 °C. Two-dimensional grazing-incidence wide-angle X-ray diffraction (2D GIXD) patterns of d PSTz1 and h PSTz2 in the neat films.

Full size image

To confirm this deduction, we measured the absorption spectra of PSTz1 and PSTz2 in the polymer/polystyrene (PS) blend film, in which the polymer to PS weight ratio was 1:5 (Fig. 3a for PSTz1 and Fig. 3e for PSTz2). Blending with PS diluted the polymer chains in the film and hence eliminated intermolecular interactions, creating the so-called pseudo-solution state. Interestingly, as expected, the 0–0 band weakened for PSTz1 (I0–0/I0–1 = 1.03) but intensified for PSTz2 (I0–0/I0–1 = 1.26) in the polymer/PS blend film relative to that in the neat film. We also measured the absorption spectra of PSTz1 and PSTz2 in CB solution by adding methanol (Fig. 3b for PSTz1 and Fig. 3f for PSTz2) since the addition of this poor solvent enabled the polymer chains to form aggregates. As the ratio of methanol increased, the 0–0 band intensified for PSTz1 but weakened for PSTz2. These results clearly indicated that the PSTz2 backbone was more ordered in the isolated state than in the aggregates, and the PSTz1 backbone was more ordered in the aggregates than in the isolated state.

In addition, we conducted temperature-dependent absorption measurements in a CB solution (Fig. 3c for PSTz1 and Fig. 3g for PSTz2). Although the 0–0 band weakened as the temperature increased from 20 to 100 °C for both polymers, the spectra of PSTz1 showed a more significant blueshift and became featureless, whereas those of PSTz2 retained a vibronic structure. Interestingly, the spectrum of PSTz1 at 20 °C was similar to that of PSTz2 at 100 °C; thus, these polymers were similarly coplanar and ordered under these different conditions. Furthermore, the PSTz2 backbone was much more rigid than the PSTz1 backbone.

Polymer packing structure and charge transport in neat film

The packing order of polymers in the neat film was studied by two-dimensional grazing incident wide-angle X-ray diffraction (2D GIXD) measurements (Fig. 3d, h). Both PSTz1 and PSTz2 had a certain degree of intermolecular order with a face-on backbone orientation, as the diffractions for the lamellar structure and the π–π stacking structure were observed along the qxy axis and the quasi-qz axis, respectively. PSTz2 had a wider d-spacing for π–π stacking (dπ) than PSTz1: dπ was 3.79 Å for PSTz1 and 3.98 Å for PSTz2 (Supplementary Fig. 8). The difference in the dπ is ascribed to the difference in the side chain; the tripropylsilyl group in PSTz2 was bulkier than the 2-ethylhexyl group in PSTz1, thus reducing interchain interaction. Nevertheless, both polymers had a similar coherence length (LC) of approximately 13 Å for the π–π stacking order.

We examined the bulk charge transport property of the polymers by fabricating OFET and hole-only devices. In the OFET device, the charge carriers were transported along the substrate plane, and the polymers showed similar transistor properties with similar hole mobilities (µFETs) of 8.8 × 10−3 cm2 V−1 s−1 for PSTz1 and 5.5 × 10−3 cm2 V−1 s−1 for PSTz2 (Supplementary Fig. 9 and Supplementary Table 2). In the hole-only device, the charge carriers were transported in the direction normal to the substrate plane, and the polymers also showed similar hole mobilities (µSCLCs), which were evaluated on the basis of the space-charge limited current (SCLC) model: 7.0 × 10−4 cm2 V−1 s−1 for PSTz1 and 6.0 × 10−4 cm2 V−1 s−1 for PSTz2 (Supplementary Fig. 10a). These results were consistent with the fact that the polymer packing structure in the neat film was similar for these polymers.

Polymer backbone order in blend film

Herein, we discussed the polymer backbone order in the polymer/acceptor blend film. We initially measured the UV–vis absorption spectra of the polymers in polymer/PC71BM (Fig. 4a) and polymer/IT-4F (Fig. 4c) blend films with polymer toacceptor weight ratios of 1:2 and 1:1.5, respectively, which were the same as those for the corresponding photovoltaic cells. In the polymer/PC71BM blend film, the absorption spectra of PSTz1 and PSTz2 were almost identical; thus, the backbone order was similar in these two polymers, as is the case in the polymer neat film. However, the I0–0/I0–1 values were 1.01 and 0.99 for PSTz1 and PSTz2, respectively, which were smaller than those in the neat film. In contrast, the absorption spectra of PSTz1 and PSTz2 in the IT-4F blend film were quite distinct. The PSTz2/IT-4F blend film provided an intense spectrum, particularly in the polymer absorption region, and a significantly large I0–0/I0–1 of 1.32 compared with the PSTz1/IT-4F blend film (I0–0/I0–1 = 1.13). This result clearly indicated that the backbone order was significantly higher for PSTz2 than for PSTz1 when blended with IT-4F. The trend observed for the polymer/IT-4F blend film was similar to that observed for the solution and the polymer/PS blend film, where the polymer chains were assumed to be mostly dispersed and/or isolated.

Fig. 4: Investigation of polymer order in blend films by UV–vis absorption spectroscopy and GIXD measurement.
figure 4

UV–vis absorption spectra of polymers in a polymer/PC71BM and c polymer/IT-4F blend films. 2D GIXD patterns of b polymer/PC71BM and d polymer/IT-4F blend films.

Full size image

Polymer packing structure in blend film

We also showed the 2D GIXD patterns of the polymer/PC71BM (Fig. 4b) and polymer/IT-4F (Fig. 4d) blend films (see Supplementary Fig. 11 for the cross-sectional profiles). In both PC71BM and IT-4F blend films, PSTz1 and PSTz2 showed extremely weak diffraction in the wide qz region. For the acceptor materials, PC71BM showed a typical diffuse ring diffraction pattern, and IT-4F did not show any diffractions, consistent with their neat films (Supplementary Fig. 12). In the PC71BM blend films, both dπ and LC could not be estimated since the diffraction of the polymer π–π stacking overlapped with the diffraction of PC71BM. Nevertheless, the π–π stacking diffraction was very weak. In contrast, in the IT-4F blend films, dπ and LC were estimated to be 3.70 Å and ca. 7 Å for PSTz1 and 3.82 Å and ca. 8 Å for PSTz2, respectively. These results indicated that both polymers had a similar amorphous-like nature in the two blend films12. This result was probably due to the sterically hindered π–π interaction of the polymer chains, which originated in the largely twisted alkylthiophene or tripropylsilylthiophene substituents with respect to the BDT unit. In these conjugated polymers with sterically hindered π–π interactions, the π–π stacking could potentially be further weakened when another material was blended. In addition, we conducted atomic force microscopy (AFM) and transmission electron microscopy (TEM) for the blended films (Supplementary Figs. 13 and 14). In both measurements, we observed a slight difference in the texture between the PC71BM and IT-4F; however, we did not observe any significant difference between PSTz1 and PSTz2.

In accordance with the similar packing structure, similar charge carrier mobilities were observed for PSTz1 and PSTz2 blend films when the same acceptor was used (Supplementary Fig. 10b, 10c). We further conducted transient photovoltage (TPV) and photocurrent (TPC) measurements to study the bimolecular recombination dynamics. By analyzing the charge carrier lifetime and bimolecular recombination rate constant as a function of charge carrier density (Supplementary Fig. 15a and 15b, Supplementary Table 3), the bimolecular recombination reduction factor was found to be in a similar range of 0.01–0.02 for all the blend films (Supplementary Fig. 15c, Supplementary Table 3), indicating that bimolecular recombination was similarly suppressed for these cells.

Morphological behavior of polymers

An overall picture of the change in polymer backbone order as well as morphology in different states is shown in Fig. 5 (a: solution, b: neat film, c: PC71BM blend film, d: IT-4F blend film) with consideration of the results observed in the absorption and GIXD measurements. In the case of PSTz1, because I0–0/I0–1 for the absorption spectrum of this polymer in the neat film was larger than that in the solution, which was typical of π-conjugated polymers, the backbones were more ordered in the neat film than in the solution. Thus, the size of the coplanar segments in the single polymer chain became larger in the neat film than in the solution. In the case of PSTz2, because I0–0/I0–1 for the absorption spectrum of this polymer in the solution was significantly larger than that in the neat film, the coplanar segments were larger in the solution than in the neat film; this result sharply contrasted that of PSTz1. Nevertheless, the backbone order of the two polymers appeared to be similar in the neat film since the absorption spectra were similar. In addition, the π–π stacking order of the two polymers was similar, as evidenced by the similar absorption spectra and GIXD patterns.

Fig. 5: Schematic illustrations of the backbone structures and morphology of the polymers.
figure 5

PSTz1 (upper) and PSTz2 (lower) in a solution, b neat film, c PC71BM blend film, and d IT-4F blend film.

Full size image

In the blend films, the phase separation was apparently quite distinct by the acceptor material. In the PC71BM blend film (Fig. 5c), PC71BM was found to have formed aggregates with some order according to the GIXD patterns; thus, it was assumed that phase separation occurred to a certain extent. On the other hand, in the IT-4F blend film (Fig. 5d), the polymer chains of PSTz1 and PSTz2 were likely to be dispersed based on resemblance of the absorption spectra of the polymers in the polymer/IT-4F blend film to those in the solution. This deduction was supported by the difference in the surface energy. Here, we measured the contact angles of the neat films of the polymers and the acceptors and determined the surface energy according to the Owens–Wendt methods (Supplementary Fig. 16, Supplementary Table 4)16. As a result, the surface energies were 23–24 mN m−1 for the polymers, 32.8 mN m−1 for PC71BM, and 28.1 mN m−1 for IT-4F. Since the difference in the surface energy between the polymers and IT-4F was markedly smaller than that between the polymers and PC71BM, the polymers were likely to be more miscible with IT-4F than PC71BM.

According to the absorption spectra, the backbone order in the PC71BM blend film was also similar in the two polymers, although it appeared that the backbones were less ordered than those in the neat film (Fig. 5c). Additionally, the polymers had a negligibly low π–π stacking order and were thus mostly amorphous. Notably, in the IT-4F blend film, the PSTz2 backbone possessed much larger coplanar segments than the PSTz1 backbone (Fig. 5d). In addition, however, there was almost negligible π–π stacking order for both polymers.

Therefore, structural order of PSTz1 and PSTz2 was similar in the aggregated state, such as in the neat film and the PC71BM blend film, whereas it was quite different in the dispersed/isolated state, such as in the solution and the IT-4F blend film. Interestingly, different from most π-conjugated polymers, the PSTz2 backbones were significantly more ordered in the dispersed/isolated state with negligible interchain interactions than in the aggregated state. Notably, this difference in the backbone order in the blend films correlated well with the photovoltaic performance.

Exciton and charge delocalization and local charge transport

We performed electroabsorption (EA) spectroscopy measurements of the polymer neat films and polymer/acceptor blend films. As reported previously17, the excited-state polarizability derived from the EA spectra is a good measure of exciton delocalization, which is sensitively dependent on the backbone coplanarity/order. Furthermore, the backbone coplanarity/order is strongly correlated with the π-conjugation along the backbone; Δp can also be a good measure of charge delocalization along the backbone. The change in the optical absorption coefficient Δα is given by Eq. 118.

$$\Delta \alpha \,=\,\left[\frac{1}{2}\Delta p{\alpha }^{{\prime} }+\,\frac{1}{6}{\left(\Delta \mu \right)}^{2}\alpha ^{\prime\prime} \right]{F}^{2}$$
(1)

where Δp is the average excited-state polarizability change; Δμ is the average permanent dipole moment change of the material upon photoexcitation from the ground state to the excited state; α′ and α″ are the first and second derivatives of the optical absorption coefficient; and F is the electric field. Thus, we can estimate the exciton delocalization as well as the charge delocalization by analyzing the EA spectra using Eq. 1.

The EA spectra of the polymer neat and polymer/acceptor blend films were carefully measured in transmission mode to eliminate optical interference effects that could have an impact on the spectral shape18. Fig. 6a shows the EA signals and the fitting curves of the polymer neat films in the polymer absorption range. For both polymers, the fitting curves effectively matched the EA signals, particularly in the low-energy region, in which PSTz2 consistently produced large signals. Notably, the extracted Δp for PSTz2 (302 Å3) was larger than that for PSTz1 (204 Å3) (Table 2), indicating that the exciton, and thus the charge, was more highly delocalized in PSTz2 than in PSTz1. This also indicated that the backbone of PSTz2 had longer effective conjugation than that of PSTz1, even though the UV–vis absorption spectra of the polymers in the neat films did not show a significant difference. This result could originate from the higher π–π stacking order, as evidenced by the GIXD study. Furthermore, this difference was also apparent in the local charge carrier transport studied by time-resolved microwave conductivity (TRMC) measurements (Fig. 6d). In the TRMC measurements, the charge transport property is expressed as φΣµ, where φ is the charge carrier generation efficiency and ∑μ is the charge carrier mobility19. The maximum φΣµ for PSTz2 was 6.8 × 10−5  cm2 V−1 s−1, which was slightly higher than that for PSTz1 (5.6 × 10−5 cm2 V−1 s−1) (Table 2).

Fig. 6: Exciton and charge delocalization and local charge transport.
figure 6

a–c Electroabsorption spectra and d–f laser-flash TRMC kinetic traces. a, d polymer neat films, b, e polymer/PC71BM blend films, and c, f polymer/IT-4F blend films.

Full size image
Table 2 Exciton/charge delocalization evaluated by EA measurements and charge carrier transport properties evaluated by TRMC measurements.
Full size table

In the polymer/PC71BM blend films (Fig. 6b), PSTz2 also showed a larger EA signal than PSTz1, resulting in Δp values of 279 and 346 Å3 for PSTz1 and PSTz2, respectively, and the difference in Δp was similar to that in the neat film (Table 2). The trend of the EA spectra of these polymers in the neat film and the PC71BM blend film agreed well with that of the UV–vis absorption spectra. Accordingly, the TRMC signal was slightly higher in the PSTz2/PC71BM blend (3.0 × 10−4 cm2 V−1 s−1) than in the PSTz1/PC71BM blend (2.3 × 10−4 cm2 V−1 s−1), as in the case of the neat films (Fig. 6e, Table 2). However, JSC and FF, which should correlate with the charge carrier mobility, were similar for these polymers. Thus, these differences in exciton delocalization and charge carrier mobility were not highly significant as to result in large differences in photovoltaic performance.

In contrast, in the polymer/IT-4F blend films, the differences in these properties between the two polymers were much more significant. Figure 6c shows the EA spectra of the blend films in the polymer absorption region. Clearly, the difference in the signal intensity in the polymer absorption region between PSTz1 and PSTz2 in the IT-4F blend films was much larger than that in the neat films and the PC71BM blend films. In fact, the Δp values were 205 and 605 Å3 for PSTz1 and PSTz2 (Table 2), respectively, suggesting that the exciton/charge delocalization was more significant in PSTz2 than in PSTz1 when blended with IT-4F. This result was consistent with the fact that I0–0/I0–1 for the UV–vis absorption spectrum of PSTz2 in the IT-4F blend film was much larger than that of PSTz1 in the IT-4F blend film. Interestingly, the Δp values in the IT-4F absorption region were also larger in the PSTz2/IT-4F blend film (662 Å3) than in the PSTz1/IT-4F blend film (250 Å3) (Supplementary Fig. 17), even though the UV–vis absorption spectra in the IT-4F absorption region did not show a significant difference between these blend films (Supplementary Fig. 14). This potentially occurred because the local molecular packing and/or the connectivity of the molecular packing of IT-4F was more developed in the PSTz2-based blend film than in the PSTz1-based blend film. The PSTz2/IT-4F blend film consistently produced a φΣµ of 2.2 × 10−3 cm2 V−1 s−1, which was almost one order of magnitude higher than that of the PSTz1/IT-4F blend film (5.7 × 10−4 cm2 V−1 s−1) (Fig. 6f, Table 2). These results agreed well with the fact that PSTz2 demonstrated markedly higher photovoltaic performance, in particular, JSC and FF, than PSTz1 in the cells using IT-4F as the acceptor. More specifically, since this charge delocalization and charge transport occurred in the shorter range along the backbone, the higher photovoltaic performance in the PSTz2 system could be ascribed to the more efficient charge separation process at the polymer/IT-4F interface.

Conclusion

In this work, we synthesized two π-conjugated polymers PSTz1 and PSTz2, which had the same backbone but different side chains, namely, branched alkyl (2-ethylhexyl) groups and tripropylsilyl groups, respectively. Interestingly, these two polymers showed highly contrasting aggregation behaviors. The PSTz2 backbone was significantly more ordered in the dispersed or amorphous state than in the aggregated state, and the PSTz1 backbone was more ordered in the aggregated state than in the dispersed or amorphous state. With the unique aggregation behavior, PSTz2 showed highly delocalized excitons and charges presumably along the backbone in the amorphous film but not in the film with π–π stacking order. These characteristics were likely more beneficial to charge delocalization and charge transport in relatively short ranges than in long ranges. Accordingly, PSTz2 showed higher performance than PSTz1 in the OPV cell based on IT-4F, where the active layer was amorphous. Thus, this result was most likely caused by a more efficient charge separation process in the PSTz2 system.

We demonstrated that significantly high polymer order could be achieved in the mixed phase with an amorphous nature by using a π-conjugated polymer having a rigid and coplanar backbone with a three-dimensional bulky side chain. Our results provide a direction for more precise control of structural order in bulk heterojunction films where both crystalline pure-material phases and amorphous-like mixed phases coexist, which are crucial for achieving even higher performance in OPVs.

Methods

Polymer synthesis

Polymerization was carried out using a Biotage Initiator microwave reactor. The molecular weights of the polymers were evaluated by a high-temperature GPC (TOSOH HLC-8121GPC/HT) using DCB as the eluent at 140 °C and calibrated with polystyrene standard. NMR spectra of the polymers were measured in deuterated DCB with a Varian System 500 (1H, 500 MHz) (Supplementary Figs. 2, 3).

PSTz1

2,6-Bis(trimethyltin)-4,8-bis(5-(2-ethylhexyl)thiophen-2-yl)benzodithiophene (1) (90.4 mg, 0.10 mmol), which was purchased from Ossila, 2,5-bis(5-bromo-3-(2-butyloctyl)thiophen-2-yl)thiazolo[5,4-d]thiazole (3) (80.0 mg, 0.10 mmol), and Pd(PPh3)4 (1.0 mg, 0.002 mmol) were mixed with toluene (5 mL) in a reaction tube. Herein, 3 was synthesized via the cyclization using 3-(2-butyloctyl)thiophene-2-carbaldehyde and dithiooxamide, and the subsequent bromination using N-bromosuccinimide20. The tube was purged with argon and sealed, which was then heated at 150 °C for 20 min using the microwave reactor. The reaction mixture was added to methanol (100 mL) containing hydrochloric acid (5 mL), which was stirred for 5 h. The resulting precipitate was collected and washed with methanol, hexane, and chloroform to eliminate low molecular weight fractions. The residue was dissolved in CB and reprecipitated in methanol (100 mL), yielding the polymer as dark red solids (yield = 84%). Anal. calcd for C70H94N2S8: C, 68.90; H, 7.75, N, 2.25. Found: C, 68.96; H, 8.00, N, 2.52.

PSTz2

PSTz2 was synthesized by the same procedure as PSTz1 using bis(trimethylstannyl)-4,8-bis(5-(tripropylsilyl)thiophen-2-yl)benzo[1,2-b:4,5-b′]dithiophene (2), which was purchased from Ossila, and 3 as the comonomers (yield = 87%). Anal. calcd for C72H102N2S8Si2: C, 68.91; H, 7.77, N, 2.30. Found: C, 68.95; H, 8.01, N, 2.52.

OPV fabrication and measurements

ITO substrates were precleaned with a detergent, deionized water, acetone, and isopropanol at room temperature in an ultrasonication bath, and with boiled isopropanol each for 10 min, and then were baked at 120 °C for 10 min. The substrates were further cleaned by UV/ozone treatment at room temperature for 20 min. The ZnO layer was deposited by spin coating a ZnO nanoparticle ink (purchased from Sigma–Aldrich) at 1200 rpm for 30 s. The photoactive layer was deposited in a glove box (Korea Kiyon, KK-011AS-EXTRA). Herein, PC71BM and IT-4F were purchased from Solenne BV and 1-Material, respectively. For the polymer/PC71BM blend, a blend solution in CB (5 g L−1 based on the polymer) heated at 100 °C was spin-coated at 800 rpm for 20 s. For the polymer/IT-4F blend, a blend solution in CB (4 g L−1 based on the polymer) heated at 100 °C was spin-coated at 800 rpm for 20 s. MoOx (7.5 nm) and Ag (100 nm) layers were then deposited sequentially by thermal evaporation through a shadow mask in a vacuum evaporator (ALS Technology, E-100J), where the photoactive area was 0.126 cm2. The J–V measurement of the cells were carried out using a Keithley 2400 source–measure unit in a glove box under 1 sun (AM1.5 G) conditions. The light intensity of the solar simulator (SAN-EI Electric, XES-40S1) was calibrated with a reference PV cell (KONICA MINOLTA AK-100 certified by the National Institute of Advanced Industrial Science and Technology, Japan). The EQE spectra were measured using a spectral response measuring system (SOMA OPTICS, S-9241). More than 10 different substrates (four photoactive areas each) were prepared and measured.

UV–vis absorption measurements

UV–vis absorption spectroscopy was performed with a Shimadzu UV-3600 Plus spectrometer. Polymer neat films were prepared by spin coating 5 g L−1 chlorobenzene solution at 800 rpm for 20 s, in which the solution was kept heated at 100 °C. Polymer/PS (1:5 w/w) blend films were prepared by spin coating the chlorobenzene solution of the blend (5 g L−1 based on the polymer) at 800 rpm for 20 s. Polymer/PC71BM and polymer/IT-4F blend films were prepared as described for the solar cell fabrication.

Grazing incidence X-ray diffraction (GIXD) measurements

GIXD measurements were performed with a HUBER multiaxis diffractometer installed in beamline BL46XU at SPring-8. The X-ray beam from the undulator was monochromatized by a double-crystal Si(111) monochromator. The X-ray energy was 12.39 keV (λ = 1 Å), and the X-ray beam size was 40 µm (height) × 300 µm (width) at the sample position. The diffraction from the samples was detected by a two-dimensional (2D) X-ray photon counting pixel detector (PILATUS 300 K). The X-ray beam incidence angle was set to 0.12°, and the camera length (sample-to-detector distance) was 174 mm. The polymer neat films were prepared on the ZnO-coated ITO glass substrate under similar conditions to the blend films for OPV fabrication. For the blend films, films on the OPV cells were directly used for the measurements. Measurements were performed in air at room temperature. The exposure time was 1 s, and no irradiation damage was observed on the samples.

Time-resolved microwave conductivity (TRMC) measurements

A resonant cavity was used to obtain a high degree of sensitivity in the conductivity measurement. The resonant frequency and the microwave power were set ca. 9.1 GHz and 3 mW, respectively, such that the electric field of the microwave was sufficiently small to not disturb the motion of charge carriers. The third harmonic generation (THG; 355 nm) of a Nd:YAG laser (Continuum Inc., Surelite II, 5–8 ns pulse duration, 10 Hz) was used as the excitation source. The incident photon intensity was 9.1 × 1015 photons cm−2 pulse−1. The photoconductivity transient Δσ was converted into the product of the quantum efficiency φ and the sum of charge carrier mobilities Σμ by φΣμ = Δσ(eI0Flight)−1, where e and Flight are the unit charge of a single electron and the correction (or filling) factor, respectively.

Electroabsorption (EA) measurements

A 450-W Xe lamp (Newport, 6279NS) was used as the light source. White light was passed through a monochromator (Cornerstone, 130) to be used as probe light. The probe light passing through a sample film sandwiched with transparent electrodes was finally detected with a p-i-n photodiode, such as Si or InGaAs, depending on the measuring wavelength. The signal from the photodiode was preamplified (Stanford Research Systems, SR570) and then sent to a lock-in amplifier (Stanford Research Systems, SR830) and a digital oscilloscope (Tektronix, TDS3052B). A sinusoidal modulated AC signal with a peak-to-peak voltage (Vp-p) of 0.6 to 0.8 V and a frequency of 1 kHz was applied on the sample film with a negative DC bias voltage (Voff) of −2 to −1 V by using a function generator (NF, WF1973). The EA signal was recorded as −ΔT/T, where ΔT is the transmittance AC signal with the electric field and T is the transmittance DC signal without it. All measurements were performed at room temperature.