Introduction

With the advancement of artificial intelligence and the rapidly growing demand for technology-integrated lifestyles, flexible electronic devices are now becoming indispensable players in the future market1,2. Organic solar cells (OSCs) have been regarded as promising candidates for wearable power sources due to their advantages of lightweight, flexible, and thin3,4,5,6,7,8. Benefiting from the advancements in photovoltaic materials and device technologies, the power conversion efficiencies (PCEs) of state-of-the-art rigid and flexible OSCs have surpassed 20% and 19%, respectively9,10,11,12,13,14,15,16,17,18.

Balancing power conversion efficiency and mechanical robustness in flexible electronics remains challenging. Here, the authors introduced 1,3,5-tris(bromomethyl)benzene into polymer donors for intramolecular crosslinking, which improved the material’s flexibility while achieving an efficiency of 20.48% in ternary organic solar cells.

In flexible devices, long-chain polymer donors (PDs) typically form intertwined and interspersed networks within the active layer, allowing the localized deformation of chain segments under mechanical stress without causing misalignment of molecular chains and disruption of molecular arrangement. This structural adaptability enhances the flexural endurance of OSCs more effectively than small-molecule acceptors19,20. However, high-performance PDs like PM6 and D18 typically feature rigid linear structures with conjugated backbones, limiting their applicability in deformable (flexible and intrinsically stretchable (IS)) OSCs, particularly in IS devices21,22,23. While more flexible and stretchable PDs, such as PNTB6-Cl, have been developed for these applications, their inferior photovoltaic performance makes them unsuitable for practical daily use24,25,26. This performance-flexibility dichotomy underscores the critical need for molecular design engineering of PDs that simultaneously optimize photovoltaic performance and mechanical robustness for next-generation OSCs. Extensive efforts have been devoted, including blending thermoplastic elastomers (e.g., styrene block copolymers, polyurethane, polymethylmethacrylate) into the active layer or incorporating flexible moieties (e.g., oligothiophene or flexible chains) into the PDs structure22,23,27,28,29,30. Nevertheless, introducing non-photoactive components or reducing conjugation typically disrupts light-harvesting capabilities and charge transport pathways in OSCs, which detrimentally impact critical device parameters31,32,33,34. Therefore, developing novel PDs that overcome the trade-off between PCE and flexibility remains a critical challenge for achieving high-efficiency flexible- and IS-OSCs.

In this study, we develop star-branched polymer donors (SPDs) through the ternary copolymerization strategy by introducing 1,3,5-tris (bromomethyl) benzene as a multi-reactive-site unit to covalently crosslink PM6 polymer chains into an intramolecular meshing architecture. This structural innovation significantly mitigates aggregation in conventional linear polymer donors while simultaneously enhancing their flexibility and stretchability, enabling their effective integration into flexible- and IS-OSCs. The SPDs (named S1, S2 and S3) with systematically tuned branching ratios are synthesized and paired with L8-BO, yielding remarkable PCEs of 18.50%, 19.51%, and 19.07% in rigid devices, respectively. Notably, the optimal efficiency achieved by S2:L8-BO OSCs is comparable to that of PM6:L8-BO devices (19.43%), and a PCE of 20.48% has been further achieved in S2:L8-BO:ZY-8F ternary OSCs. When S2 is incorporated into deformable OSCs, the resulting flexible- and IS-OSCs deliver PCEs of 18.39% and 15.40%, which are among the highest reported for binary devices. Moreover, the intramolecular meshing configuration enhances PM6 backbone entanglement density, effectively restricting chain slippage under stress, thereby synergistically improving macroscopic tensile strain resistance. Therefore, all SPDs-based flexible OSCs could retain over 80% of their initial PCEs after 6,000 bending cycles, and S2- and S3-based IS-OSCs exhibit the best tensile stress resistance and cyclic stretching endurance. Notably, both deformable OSCs based on donor S2 exhibit relatively high PCE values and mechanical robustness.

Results and discussion

Materials and photovoltaic performance of rigid devices

The synthetic pathways for SPDs and PM6 are schematically depicted in Fig. 1 and Supplementary Fig. 1. All required raw materials, catalysts, and solvents for post-treatment were commercially available, with detailed synthesis procedures provided in the Methods section. Both PM6 and SPDs were synthesized through Stille coupling reactions using the same monomers M1, M2, and varying contents (molar ratios with M2) of 1,3,5-tris (bromomethyl) benzene (M3) as crosslinker. In details, (ⅰ) methyltin-containing parts (monomers, oligomers, or polymeric intermediates) underwent sequential coupling with M3, initiating the hierarchical assembly of star-branched architectures; (ⅱ) the three radiating arms from each crosslinker underwent progressive elongation through iterative monomer incorporation; (ⅲ) the M3 on the branch chain will act as the next active center, continuing the elongation and propagation of the polymer chain. This dynamic process culminated in the formation of an intramolecular three-dimensional entanglement network through covalent interconnection between adjacent crosslinkers. Notably, the topological density of this molecular-scale entanglement network exhibited tunability by strategically adjusting the crosslinker-to-monomer stoichiometry. Molecular weight characterization data (GPC traces) and 1H NMR for the three SPDs are presented in Supplementary Figs. 2–9. The chemical structures of PM6, SPDs and acceptor L8-BO are shown in Supplementary Fig. 10. The more detailed schematic diagram for PDs and the three-dimensional network structure diagrams of SPDs are shown in the Supplementary Figs. 11 and 12.

Fig. 1: Synthetic pathways for PM6 and SPDs.
Fig. 1: Synthetic pathways for PM6 and SPDs.
Full size image

Blue and green squares represent the two polymeric monomers of PM6, red circles represent cross-linking centers, and orange squares represent the chain segments of PM6. (SPDs: Star-branched Polymer Donors).

The normalized UV-vis absorption profiles of PDs in neat films are presented in Fig. 2a. Three SPDs displayed (0-0) transition maxima at 616, 618, and 615 nm in neat films, exhibiting a consistent blue-shift relative to PM6 (622 nm). Furthermore, the absorption intensity ratio of (I0-1/I0-0) relative to SPDs showed varying degrees of enhancement compared to PM6, suggesting suppressed aggregation in the SPDs35. The similar trends can be seen in chloroform (CF) (Fig. 2b) and o-xylene solutions (Supplementary Fig. 13), indicating effective suppression of donor pre-aggregation in both solutions. This phenomenon is related to that the introduction of cross-linking centers (M3) suppresses the aggregation process between PM6 backbones, which is advantageous for material solubility and large-scale processing with green solvents36. The UV-vis absorption profiles of blend films are depicted in Fig. 2c. The characteristic (0-1) transition of donor materials in all blends shows significant suppression after incorporating L8-BO, which is related to that the strategic introduction of cross-linking centers primarily regulates molecular assembly dynamics during the film-forming process, while showing little effect on the molecular arrangement within blends. Moreover, the optimized S2:L8-BO blend demonstrates enhanced photon capture capability across the visible spectrum, which benefits the achievement of higher short-circuit current density (Jsc) in photovoltaic devices.

Fig. 2: Optical and crystalline characterization of materials.
Fig. 2: Optical and crystalline characterization of materials.
Full size image

UV-vis absorption spectra in neat films (a), CF solutions (b) and blend films (c). d Energy level alignment of PDs and L8-BO. e The d-spacing and CCL in OOP direction. f 2D GIWAXS pattern of PDs. g Stress-strain curves of the neat films.

Optical bandgaps (\({E}_{{\mbox{g}}}^{{\mbox{opt}}}\)) derive from absorption onsets in films obtained by CF solution, yielding values of 1.52 eV, 1.50 eV, 1.49 eV, and 1.51 eV for PM6, S1, S2, and S3, respectively. The corresponding electrochemical bandgaps probed by cyclic voltammetry (CV) (Fig. 2d and Supplementary Fig. 14) of SPDs were: 1.85 eV (S1), 1.88 eV (S2) and 1.90 eV (S3), with LUMO (the Lowest unoccupied molecular orbital)/HOMO (the Highest occupied molecular orbital) levels of −3.55/−5.40 eV (S1), −3.58/−5.46 eV (S2), and −3.55/−5.45 eV (S3). It can be observed that the introduction of cross-linking centers leads to band gap narrowing in the donors. Then Grazing-incidence wide-angle X-ray scattering (GIWAXS) was used to probe molecular packing and crystallinity, with detailed parameters exhibited in Fig. 2e, Supplementary Fig. 15 and Supplementary Table 1. As shown in Fig. 2f, all neat films of PDs display strong π-π stacking signals at out-of-plane (OOP) direction. The (010) diffraction peaks for PDs all appear at 1.65 Å−1, corresponding to calculated d-spacings of 3.81 Å. The crystal coherence length (CCL) values, determined from peak full-width-at-half-maximum (FWHM), are 21.15 Å, 21.97 Å, and 21.37 Å for S1, S2, and S3, respectively. These results demonstrate that the optimal OOP π-π stacking in S2, comparable to that of PM6 (CCL = 21.67 Å), facilitates efficient charge transport and extraction. Subsequently, to investigate the effect of star-branched structure on the donor material’s flexibility, we conducted FOW (Film-on-Water) tests on PM6 and SPDs. The corresponding stress-strain curves are shown in Fig. 2g. All SPDs demonstrated superior elongation at break compared to PM6 (fracture strain = 10.46%), whereas S3 exhibited the highest fracture strain of 19.16%, followed by S2 (15.33%) and S1 (12.70%), which originated from their intramolecular entanglement network, fully demonstrating the effectiveness of this strategy37.

To systematically investigate the impact of crosslinker incorporation on photovoltaic performance, all devices with the architecture of ITO/2PACz/active layer/PNDIT-F3N/Ag were fabricated and characterized under AM1.5G illumination. Detailed device fabrication protocols are provided in the Supplementary Information. Table 1 summarizes the photovoltaic parameters under optimized conditions, with the corresponding current density-voltage (J-V) curves presented in Fig. 3a. The PM6:L8-BO based device achieves a PCE of 19.43% with an open-circuit voltage (Voc) of 0.900 V, Jsc of 26.68 mA cm−2, and fill factor (FF) of 80.9%. Among the SPDs-based OSCs, the S2:L8-BO device demonstrates optimal performance, delivering a PCE of 19.51% (Voc = 0.902 V, Jsc = 26.70 mA cm−2, FF = 81.0%), whereas the devices based on S1 and S3 exhibit significantly reduced PCEs compared to PM6, likely attributed to the unsatisfactory molecular aggregation and stacking38,39. Furthermore, studies on S2 with different molecular weights were conducted to investigate the effect of molecular weight on device performance (Supplementary Fig. 16 and Supplementary Table 2). It can be observed that the material with a molecular weight of 25.9 kDa (similar to the molecular weight of the PM6 used) achieved better performance. The corresponding box chart in Fig. 3b shows the statistical histograms of PCEs obtained from 20 independent devices. Additionally, the thermal stability and photostability of binary devices have also been tested, with the results shown in Supplementary Fig. 17.

Fig. 3: Photovoltaic performance and physical characteristics.
Fig. 3: Photovoltaic performance and physical characteristics.
Full size image

a J-V characteristics based on rigid OSCs. b The box chart depicts the statistical histograms of the PCEs obtained from 20 devices. c EQE spectra (solid lines) and integrated current densities (dashed lines) of rigid OSCs. d Jph versus effective bias (Veff) characteristics of devices. e Voc and f Jsc versus light intensity characteristics of devices. g, h Space-charge-limited current (SCLC) characteristic curves of devices. i The electron (μe) and hole (μh) mobilities of devices.

Table 1 Photovoltaic parameters of the rigid OSCs
Full size table

External quantum efficiency (EQE) spectra in Fig. 3c reveals broad spectral responses across 350-1000 nm for all devices, indicating effective sunlight utilization. Integrated current densities derived from EQE curves show consistency (<3% deviation) with Jsc values obtained from J-V measurements, confirming measurement reliability. When incorporating guest acceptor ZY-8F (Supplementary Fig. 10) as a third component in S2:L8-BO blend, a significantly enhanced Voc of 0.921 V and a champion PCE of 20.48% are achieved in ternary OSCs (Supplementary Fig. 18 and Supplementary Table 3). Detailed preparation details of ternary devices and the optimization process for the content of the third component are given in the Supplementary Information (Supplementary Fig. 19 and Supplementary Table 4). Moreover, benefiting from the perfect solubility of SPDs in o-xylene, ternary devices yield an efficiency of 20.26% (Supplementary Fig. 20 and Supplementary Table 3), which is comparable to that of CF-processed devices (20.48%) and facilitates green large-area preparation of OSCs and modules.

To evaluate exciton dissociation dynamics in these devices, we analyzed the photocurrent density (Jph) versus effective voltage (Veff) characteristics, as shown in Fig. 3d. The exciton dissociation probability (Pdiss) under short-circuit conditions, defined as Jsc/Jsat (where Jsat is the saturation photocurrent density), was determined to be 99.1%, 98.2%, 99.3%, and 98.7% for PM6-, S1-, S2- and S3-based devices, respectively. Charge recombination mechanisms were further probed through light intensity (Plight) dependence of Voc. The relationship between Voc and Plight follows the formula Voc ∝ nln(Plight), where a slope (n) closer to kT/q indicates dominant bimolecular recombination, while a slope > kT/q suggests trap-assisted recombination. As shown in Fig. 3e, experimentally derived slopes are 1.16 kT/q (PM6:L8-BO), 1.46 kT/q (S1:L8-BO), 1.06 kT/q (S2:L8-BO), and 1.34 kT/q (S3:L8-BO). The significantly increased slopes in S1 and S3 devices suggest enhanced trap-assisted recombination due to the introduction of inappropriate amounts of cross-linking centers. The Jsc and Plight relationship follows Jsc ∝ (Plight)α with α approaching 1 indicating suppressed bimolecular recombination. The fitted exponents (Fig. 3f) are α = 0.989 (PM6:L8-BO), 0.970 (S1:L8-BO), 0.996 (S2:L8-BO), and 0.987 (S3:L8-BO). Notably, the S2-based device exhibits near-unity α, demonstrating perfect charge collection efficiency with the lowest bimolecular recombination among all devices.

To elucidate charge transport characteristics in devices, the hole and electron mobilities (μh and μe) of PDs:L8-BO blend films were quantitatively evaluated using the space-charge-limited current (SCLC) method. As detailed in Fig. 3g–i and Supplementary Table 5, the PM6:L8-BO, S1:L8-BO, S2:L8-BO, and S3:L8-BO devices exhibit μh/μe values of 3.91 × 10−4/3.19 × 10−4 cm2 V−1 s−1, 2.01 × 10−4/1.35 × 10-4 cm2 V−1 s−1, 5.59 × 10−4/4.72 × 10−4 cm2 V−1 s−1, and 3.13 × 10−4/2.17 × 10−4 cm2 V−1 s−1, respectively. The corresponding μh/μe ratios are calculated as 1.22, 1.49, 1.18, and 1.44. Clearly, the S2 devices demonstrate both enhanced and balanced charge transport, which is beneficial for OSCs to obtain higher Jsc and FF40,41.

Film formation kinetics and morphological characterization

To establish multiscale structure-property correlations in devices, we performed hierarchical structural characterization through synergetic atomic force microscopy (AFM) and photoinduced force microscopy (PiFM) for nanoscale morphological mapping, in situ UV-vis absorption for molecular stacking process and GIWAXS for molecular-scale crystallographic analysis. AFM height and phase images of neat films and blend films are shown in Supplementary Figs. 21 and 22. With the increasing number of cross-linking centers, the fibers in both neat and blend films show a tendency to become progressively thinner but more developed, which is associated with the effective suppression of donor chain aggregation and facilitates charge dissociation and transport.

Then we conducted PiFM measurements on the blend films. PiFM, a technique combining AFM with infrared (IR) spectroscopy, enables selective chemical mapping by targeting material-specific absorption bands. The characteristic absorption peaks of the donor materials were observed at 1650 cm-1, while L8-BO exhibited a distinct peak at 1270 cm−1. Figure 4a presents the PiFM images of PM6:L8-BO and SPDs:L8-BO blend films, all of which display clear phase-separation phenomena. Notably, compared to PM6, the SPDs-based blend films exhibited pronounced network-like structures with isolated “island” domains. Crosslinker-dependent morphological evolution was clearly observed: PM6 blends featured short, thick fibrils, whereas increasing crosslinker content progressively transformed these into fine, entangled nanofibrils.

Fig. 4: In-situ absorption spectroscopy and thin-film morphology characterizations.
Fig. 4: In-situ absorption spectroscopy and thin-film morphology characterizations.
Full size image

a PiFM images of blend phases of blend films. b In-situ 2D UV-vis absorption spectra of PDs:L8-BO blend films. c The time evolution of the peak location of PDs and L8-BO in the blends during spin-coating process. d 2D GIWAXS patterns and e the corresponding OOP and IP line cuts of blend films.

Then in situ UV-vis absorption measurements were employed to study the evolution of active layer morphology in blend films during the solvent evaporation process. As shown in Fig. 4b, the main absorption peaks of the blend films are located around 625 nm and 790 nm, corresponding to PDs and L8-BO, respectively. Throughout the solvent evaporation process, the evolution of both absorption characteristic peaks can be categorized into three stages, reflecting the film condition: the initial solvent evaporation in wet film state (Stage 1), the solute precipitation process after the solution concentration reaches solubility limit (Stage 2), and the stable dry film state (Stage 3)11. The primary focus is on stage 2, the duration of which typically reflects the level of molecular aggregation of the donor/acceptor42. For the stage 2 of donors, the (0-0) transition of PM6 shifts from solution to film within 0.6 s, while that of S1 within 1.0 s, of S2 within 1.1 s and of S3 within 1.5 s (Fig. 4c). Cross-linking centers disrupt PM6 interchain aggregation, leading to prolonged aggregation. For acceptor peak evolution, the stage 2 can typically be further subdivided into a slow nucleation stage (blue region) and a fast crystal growth stage (brown region). It can be seen that as the content of cross-linking centers increases, the duration of acceptor nucleation is progressively longer, but the crystallization time is rapidly shortened. These results demonstrate that the introduction of cross-linking centers suppress the donor crystallization, while regulating the acceptor crystallization process in a two-fold manner: suppressing nucleation and facilitating crystal growth. More crystallization information can be obtained from the GIWAXS measurements, and the relevant 2D patterns and line profiles are collected in Fig. 4d and 4e. All blends maintained nearly identical π-π stacking distances (qz ≈ 1.70 Å−1, d-spacing ≈ 3.70 Å in OOP direction), indicating preserved local molecular packing. Interestingly, the establishment of cross-linked networks triggered marked variations in CCL parameters (Supplementary Table 6). Comparative analysis revealed that S2:L8-BO blend exhibited the highest CCL value (29.92 Å) among the SPDs series, exceeding those of S1 (26.74 Å) and S3 (28.17 Å). This enhanced CCL parameter in S2 suggests a higher degree of crystalline long-range ordering, closely paralleling the structural characteristics observed in PM6 films (CCL = 29.36 Å). This crystallinity preservation in S2:L8-BO blend correlates with its perfect exciton dissociation efficiency (99.3%) and balanced charge transport (μh/μe = 1.18), collectively explaining its optimal device performance among SPDs-based devices43,44.

Photovoltaic and deformable performance of flexible- and IS- devices

To investigate the practicality and usefulness of star-branched polymer donors in deformable OSCs, flexible and stretchable devices were constructed based on the conventional device architecture of PEN/ITO/PEDOT:PSS/Active layer/PNDIT-F3N/Ag and Strem-AT45/PEDOT:PSS/Active layer/PNDIT-F3N/EGaIn, respectively. Detailed fabrication procedures are shown in the Supplementary Information. Table 2 and Fig. 5a exhibit the photovoltaic parameters and the corresponding current density-voltage (J-V) curves of flexible devices. Clearly, the PM6:L8-BO-based device achieves a PCE of 17.70%, while the S2-based device exhibits PCE of 18.39%. Mechanical endurance testing under extreme bending conditions (radius = 2 mm) demonstrates the flexibility retention. All SPDs-based devices maintain > 80% initial PCE after 6000 bending cycles, while PM6 counterparts suffer the rapid performance degradation beyond 2000 cycles (Fig. 5b). The results demonstrate that by meshing PM6, their photovoltaic performance in rigid devices can be well inherited to flexible devices while simultaneously enhancing device flexibility.

Fig. 5: Photovoltaic and deformable performance of flexible- and IS- devices.
Fig. 5: Photovoltaic and deformable performance of flexible- and IS- devices.
Full size image

a J-V characteristics based on flexible OSCs. b PCE degradation of the flexible OSCs versus bending cycles at a radius of 2 mm. c J-V characteristics based on IS-OSCs. d Stress-strain curves of the IS-OSC films. e Normalized PCEs of IS-OSCs as a function of a strain and f cyclic tensile times at 10% strain based on PDs-based blend films. g Schematic diagram of PDs under mechanical deformation. (The green circular blocks represent the acceptor L8-BO, while the purple wireframe structures denote the polymer donors).

Table 2 Photovoltaic parameters of the flexible- and IS-OSCs
Full size table

For IS-OSCs (Photographs of the actual devices are shown in the Supplementary Figs. 23 and 24), the S2-based device achieves a remarkable PCE of 15.40% outperforming the PM6-based (14.21%) and other SPDs-based (13.93% for S1, 14.52% for S3) devices, as shown in Fig. 5c and Table 2. Notably, the 15.40% efficiency is one of the highest reported PCEs for IS-OSCs. Additional FOW tests on blend films were performed (stress-strain curves in Fig. 5d). PM6:L8-BO demonstrates a fracture strain of 2.6%, consistent with previously reported values in the literatures23,46,47. With a meshing structure, all SPDs-based blend films exhibit superior fracture strain values compared to the former. Specifically, S1:L8-BO shows a fracture strain of 6.1%, S2:L8-BO demonstrates 7.4%, and S3:L8-BO achieves the highest fracture strain of 8.1%. Notably, the S2- and S3-based blend films exhibit about threefold enhancement in fracture strain compared to that of PM6:L8-BO, demonstrating that the meshing strategy can effectively enhance the stretchability of blend films. Notably, while this networked approach can effectively enhance the flexibility and stretchability of PDs, the introduction of cross-linking centers disrupts the backbone packing of PDs. As exemplified in this study, although the S3-based blend film achieves the best fracture strain, the excessive content of cross-linking agent disrupts its crystalline stacking and results in a significant PCE drop. In contrast, the S2-based device achieves an optimal balance between fracture strain and PCE, as demonstrated by the strain stability (Fig. 5e) and fatigue resistance (Fig. 5f) tests. As shown in Fig. 5e, the PCE of the intrinsically stretchable OSC devices under varying tensile strains was further investigated. Upon reaching 15% strain, PM6- and S1-based stretchable devices exhibited severe PCE degradation to ≈40% of initial efficiency, while S2 and S3 counterparts retained > 90% efficiency retention. Notably, the PM6 and S1 devices experienced mechanical failure with further straining, whereas S2 and S3 configurations withstood strains up to 35% while maintaining operational integrity. Furthermore, cyclic tensile testing under 10% strain revealed stark durability contrasts: PM6- and S1-based devices exhibited progressive efficiency degradation to ≈65% of initial PCE after 50 cycles, whereas S2 and S3 counterparts maintained > 80% retention. Moreover, the S2 and S3 configurations demonstrated better cyclic stability, sustaining > 60% PCE retention through 200 deformation cycles. The improved stretchability of SPDs is likely attributed to their stronger intramolecular entanglement networks. As illustrated in Fig. 5g, mechanical deformation (stretching or bending) caused disentanglement of polymer chains in linear PM6, resulting in crack propagation and even material rupture at structural weak points, while the covalently stabilized network structure in star-branched donors resisted chain slippage, enhancing mechanical robustness.

Discussion

In this work, three SPDs (S1, S2, S3), with star-branched meshing architectures, were synthesized by incorporating varying contents of 1,3,5-tris (bromomethyl) benzene during the polymerization of PM6. The flexibility and tensile robustness of SPDs are significantly enhanced due to the unique deformable networked structure, which effectively restricts the movement and misalignment of molecular chains. Therefore, the SPDs achieve an obvious improvement in intrinsic stretchability compared to pristine PM6, successfully enhancing mechanical compliance. In flexible OSCs, all SPDs-based devices demonstrate perfect bending durability, retaining > 80% initial efficiency after 6000 bending cycles. Moreover, in IS-OSCs, S2- and S3-based devices withstood twice the tensile stress and cyclic stretching tolerance compared to PM6 counterparts. Besides better deformation capability, S2-based rigid, flexible, and stretchable OSCs all exhibit better photovoltaic performance compared to PM6-based devices. In addition, the introduction of cross-linking centers effectively suppresses the aggregation of donor materials in eco-friendly non-halogenated solvents, and the o-XY-processed OSCs thus achieve a PCE value of 20.26%, which is comparable to that of CF-devices (20.48%). This study presents network engineering as an effective approach to make high-performance donor materials suitable for flexible and stretchable OSCs and for non-halogenated solvent processing, which is essential for large-area industrial fabrication of wearable organic photovoltaics.

Methods

Materials

PM6, with Mn = 20 kDa, was purchased from Solarmer Materials (Beijing, China). The reactants for donor synthesis were purchased from Leyan (Shanghai, China), Hyperchemical (Zhejiang, China), Energy Chemical, and Bidepharm. The methanol, n-hexane, and dichloromethane used in the polymer post-processing steps were procured from Shanghai Titan Scientific. Chloroform (CF) was purchased from Sigma Aldrich. 4-BDBTP were purchased from Macklin (Shanghai, China). The silver for electrodes were purchased from ZhongNuo Advanced Material (Beijing) Technology Co., Ltd.

Synthesis of compound S-X: M1 (Cat No.1685468, Leyan, Shanghai, China), M2, M3, Pd2(dba)3 (1.2 mg), P(o-tol)3 (2.4 mg) and toluene (5 mL) were added into 10 mL two-necked round bottom flask. The mixture was heated at 110 °C for 1 h. After cooling to room temperature, the mixture was poured into methanol (20 mL), and precipitation occurred. Then the resulting mixture was filtered and subjected to the sequential Soxhlet extraction with methanol, hexane, and dichloromethane as the solvents to remove impurities and low molecular weight fractions. The residue was extracted with chloroform as the final solvent, and the chloroform solution was concentrated and precipitated into 100 mL methanol to get the product as a golden-dark solid.

Synthesis of compound S1 (poly((1-(5-(4,8-bis(5-(2-ethylhexyl)−4-fluorothiophen-2-yl)benzo[1,2-b:4,5-b’]dithiophen-2-yl)thiophen-2-yl)−5,7-bis(2-ethylhexyl)−3-(thiophen-2-yl)−4H,8H-benzo[1,2-c:4,5-c’]dithiophene-4,8-dione)-co-(2-(3,5-dimethylbenzyl)−4,8-bis(5-(2-ethylhexyl)−4-fluorothiophen-2-yl)benzo[1,2-b:4,5-b’] dithiophene))(mole ratio of 20:1)): M1 (65.7 mg, 0.0700 mmol), M2 (50.0 mg, 0.0650 mmol), M3 (1.2 mg, 0.0033 mmol). The final product S1 was obtained as a golden-dark solid (100.5 mg, 86.0%). Molecular weight: Mn = 26 kDa, PDI = 1.50.

Synthesis of compound S2 (poly((1-(5-(4,8-bis(5-(2-ethylhexyl)−4-fluorothiophen-2-yl)benzo[1,2-b:4,5-b’]dithiophen-2-yl)thiophen-2-yl)−5,7-bis(2-ethylhexyl)−3-(thiophen-2-yl)−4H,8H-benzo[1,2-c:4,5-c’]dithiophene-4,8-dione)-co-(2-(3,5-dimethylbenzyl)−4,8-bis(5-(2-ethylhexyl)−4-fluorothiophen-2-yl)benzo[1,2-b:4,5-b’] dithiophene))(mole ratio of 10:1)): M1 (70.3 mg, 0.0750 mmol), M2 (50.0 mg, 0.0650 mmol), M3 (2.3 mg, 0.0065 mmol). The final product S2 was obtained as a golden-dark solid (102.3 mg, 83.4%). Three batches of S2 with different molecular weights were obtained by setting the polymerization time to 0.5 h, 1 h, and 2 h, respectively. Molecular weight: Mn = 6.5 kDa, PDI = 1.58; Mn = 25.9 kDa, PDI = 1.76; Mn = 103.6 kDa, PDI = 1.98.

Synthesis of compound S3 (poly((1-(5-(4,8-bis(5-(2-ethylhexyl)−4-fluorothiophen-2-yl)benzo[1,2-b:4,5-b’]dithiophen-2-yl)thiophen-2-yl)−5,7-bis(2-ethylhexyl)−3-(thiophen-2-yl)−4H,8H-benzo[1,2-c:4,5-c’]dithiophene-4,8-dione)-co-(2-(3,5-dimethylbenzyl)−4,8-bis(5-(2-ethylhexyl)−4-fluorothiophen-2-yl)benzo[1,2-b:4,5-b’] dithiophene))(mole ratio of 5:1)): M1 (79.5 mg, 0.0850 mmol), M2 (50.0 mg, 0.0650 mmol), M3 (4.6 mg, 0.013 mmol). The final product S3 was obtained as a golden-dark solid (104.8 mg, 78.1%). Molecular weight: Mn = 20.4 kDa, PDI = 1.85.

Optical characterization

UV-vis absorption spectra were obtained using a Shimadzu (model UV-3700) UV-vis spectrophotometer. CV measurements were made on a CHI660 potentiostat/galvanostat electrochemical workstation at a scan rate of 20 Mv/s. Electrochemical measurements were carried out at room temperature in an acetonitrile solution of 0.1 mol L−1 Bu4NPF6 using ferrocene (−4.3 eV) as standard reference. All cyclic voltammetry (CV) curves were measured by films cast from CHCl3 solution on the glassy carbon electrode with Pt wire as the counter electrode and Ag/AgCl as reference electrode. SCLC measurements were adopted to examine charge mobilities of the blend by using hole-only devices with a structure of ITO/PEDOT:PSS/active layer/MoO3/Ag, electron-only devices with a structure of ITO/ZnO/active layer/PNDIT-F3N/Ag under dark conditions. The mobilities were determined by fitting the corresponding J-V characteristics according to the Mott-Gurneylaw: J = (9/8)εrε0µ(V2/L3), in which J is the current density, εr is the dielectric permittivity of the transport medium, ε0 is the vacuum permittivity of free space, L is the thickness of the active layer, and µ is the mobility. V = Vapp - Vbi - Vr, where Vapp is the applied voltage, Vbi is the offset voltage, and Vr is the voltage loss on series resistance.

Morphology characterizations

GIWAXS measurements were carried out at the PLS-II 9 A U-SAXS beamline of the Pohang Accelerator Laboratory in Korea. AFM measurements were performed by Dimension Icon AFM (Bruker) in a tapping mode. PiFM is a multimodal Atomic Force Microscopy (AFM) technique that integrates the high resolution of AFM with infrared spectroscopy to acquire topographical and molecular information with sub-10 nanometer resolution. All measurements were conducted using Molecular Vista’s Visascope microscope and Block Engineering’s Laser Tune QCL, with an adjustable range from 775 to 1885 cm-1 and a spectral linewidth of 2 cm−1. PiFM images were created with a resolution of 256 × 256 pixels, line speed of 0.89 lines/s, and a size of 1 µm2, using wavenumbers 1650 cm−1 and 1270 cm−1 to emphasize the copolymer donor and acceptor. In-situ absorption measurements were detected by the Spectrum Microvision DU-200. The donors and acceptors were dissolved in o-xylene solvent for testing. The detection range of the instrument is 186-1026 nm with a spectral resolution of 0.01 nm and a time resolution of 1 ms.

Flexibility and tensile testing

The stress-strain curves were obtained using a commercial FOW instrument (designed by Yangzhou Super-Flex-Tech Company, No. ZRA-01, China). In the FOW test, the film was spin-coated onto a pre-ultrasonically cleaned glass substrate (2 × 2 cm), then cut into rectangles and transferred to the water surface. The film was positioned under a PDMS fixture, and the water level was raised to allow the film to adhere to the PDMS. The PDMS was stretched at a speed of 1 mm/min with an acceleration of 3 mm/min, and data were acquired via a sensor.

Device photostability characterizations

The photostability data of the devices were measured by using the Maximum Power Point (MPP) tracking mode (YH-VMPP-IV-16). The decay curves of these encapsulated devices were measured under a continuous LED light source (380–810 nm, one solar intensity) in ambient air conditions (the average humidity was 20%, the average tested temperature was 25 °C). The device encapsulation procedure is shown as follows: Firstly, a layer of UV-curable adhesive was evenly smeared on the electrode surface of the device in the N2 atmosphere. Then, a piece of a coverslip was put on the top of the adhesive. Finally, the device was illuminated under a UV light source (365 nm) for 3 minutes.

Device fabrication and measurement

Rigid organic solar cells (rigid-OSCs) were fabricated with a device structure of ITO/2PACz/PDs:acceptors/PNDIT-F3N/Ag. The ITO-coated substrates were cleaned by detergent, deionized water, acetone, and isopropyl alcohol for 15 min each in an ultrasonic cleaning machine sequentially. The cleaned ITO-coated substrates were dried in an oven at 100 °C overnight. Before use, the ITO-coated substrates were pretreated by a plasma cleaner for 2 min under a vacuum condition of 10 Pa. 2PACz was adequately dissolved in ethanol at a concentration of 0.3 mg/ml, after which the solution was spin-coated on top of the ITO substrates at 3000 rpm for 30 s and then thermal annealed at 100 °C for 10 min under nitrogen conditions. The active layers were generated by spin-coating the active layer solution on top of the 2PACz layer for 30 s at a rate of 2500 rpm with an optimal thickness of 110 nm under a nitrogen atmosphere. The D:A weight ratios of the CF solution were 1:1. The optimal weight ratios of DBCl to the total mass of PM6 acceptor were 165%. The total concentration of the CF solution was 12.5 mg/ml. The CF solutions were stirred at normal temperature (25 °C) for 4 hours before use. The active layer was spin-coated onto the 2PACz layer, followed by cooling. The corresponding blends were generated by spin-coating the solutions on the 2PACz layer. Then, the active layers were all thermally annealed at 80 °C for 7 min. PNDIT-F3N was dissolved in methanol at a concentration of 0.5 mg/mL, with glacial acetic acid added to facilitate its dissolution. Subsequently, a PNDIT-F3N layer with a concentration of 1.2 mg/ml in mixed solvent (methanol: acetic acid is 500:1 by volume) was spin-coated onto the active layer at a rate of 4200 rpm for 30 s. Finally, 120 nm Ag electrode were deposited under a vacuum condition of 4 × 10−4 Pa. The detailed fabrication protocols for binary and ternary devices are identical, except for the types and proportions of acceptor materials.

Device performance was measured by using a 510 Air Mass 1.5 Global (AM1.5 G) solar simulator (SS-F5-3A, Enlitech) with an irradiation intensity of 100 mW cm−2, which was determined using a calibrated silicon solar cell (SRC2020, Enlitech). The J-V curves were measured from −0.5 to 1 V with a scan step of 50 mV and a dwell time of 10 ms, along the forward scan direction, using a Keithley 2400 Source Measure Unit. EQE spectra were obtained by using a QE-R3011 Solar Cell EQE measurement system (Enlitech).

The detailed fabrication protocols for other types of devices are provided in the Supplementary Information.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.