High-efficiency all-polymer ternary blends enable exceptional thermal stability in organic photovoltaics

Abstract

Organic photovoltaics (OPVs) hold great promise due to their flexibility, lightweight nature, and compatibility with solution processing. However, achieving both high efficiency and thermal stability remains a significant challenge. In this study, the polymer donor PBQx-TF was blended with the high-crystallinity polymer D18, followed by the sequential deposition of the polymer acceptor PY-IT. Molecular packing differences between PBQx-TF and D18 films affected PY-IT layer deposition, impacting device charge transport. Ternary blending optimized morphology, balancing electron/hole mobility and reducing recombination. This, combined with optimized energy level alignment and minimized energy losses, enabled the ternary all-polymer OPV to achieve a power conversion efficiency (PCE) of 16.07%, surpassing the PCEs of the corresponding binary devices, which yielded 15.26% (PBQx-TF : PY-IT) and 14.39% (D18 : PY-IT), respectively. The optimized ternary OPV devices exhibited remarkable thermal stability, retaining 80% of their initial PCE following 1500 hours of sustained thermal stress at 120 °C, with structural integrity maintained across all device layers. This performance represents one of the highest levels of thermal stability reported for OPVs, underscoring the critical contribution of D18 as a third component in enhancing both the morphological stability and overall PCE. The ternary strategy-revealed correlation between thermal stability and morphology is crucial for the future development of high-performance and highly stable OPVs, potentially accelerating their commercialization.

---

Introduction

Organic photovoltaic (OPV) cells have emerged as promising candidates for renewable energy technologies due to their inherent advantages, including lightweight nature, low manufacturing costs, environmental friendliness, and short energy payback time.^1–3 In the past decade, rapid advancements in non-fullerene acceptors (NFAs) and device engineering have pushed the power conversion efficiency (PCE) meeting commercialization standards.^3–6 Various strategies, such as ternary blends, self-assembled interlayers, and pseudo-bulk heterojunctions, have been employed to enhance PCEs, even exceeding 20%.^7–12 Nevertheless, beyond achieving high PCE, device stability remains a critical challenge for the successful commercialization of OPVs.¹³

Under typical operating conditions, prolonged exposure to light, moisture, oxygen, and heat can induce performance degradation.^1,14–17 While advanced encapsulation techniques can effectively mitigate the detrimental effects of light and moisture/oxygen ingress, the inherently non-thermodynamically stable nature of the active layer may still result in blend morphology evolution under elevated temperatures, ultimately compromising device performance.¹⁷ To address this issue, various strategies have been explored to enhance morphological stability, including the incorporation of crosslinking agents within the active layer and the development of conjugated block-copolymer architectures.^18–20 Although these approaches improve stability, they often impose trade-offs that can impact the overall PCE of the device. Intrinsic entanglement between the polymeric donor and acceptor offers inherent advantages in terms of high stability.^21,22 For example, Ma and Yuan et al. observed that all-polymer (AP) OPV devices exhibited significantly superior thermal stability at 80 °C compared to their counterparts incorporating the small molecule acceptor ITIC or the polymer acceptor N2200.²³ Bao et al. showed that reducing polymer acceptor molecular weight with defined oligomers improves active layer order and crystallinity, enhancing device performance and achieving 84.9% PCE retention after 932 hours at 80 °C.²⁴ Recently, the ternary strategy has emerged as a viable route to simultaneously enhance the PCE and stability of OPV devices, and this approach holds promise for the development of AP OPVs. Min et al. demonstrated that ternary AP OPV devices, incorporating PYT into PM6 : PY-2F blends, achieved a 17.2% PCE and enhanced photo-thermal stability, surpassing the 15% PCE of the binary PM6 : PY-2F system.²⁵ Kim et al. utilized P1 as an interfacial compatibilizer in PBQx-TF : PY-IT blends, which led to enhanced PCE and thermal stability, with 90% PCE retention after 96 hours at 125 °C.²⁶ Hou et al. achieved an 18.2% peformance and enhanced stability by incorporating high molecular weight PBDB-TF into PBQx-TF : PY-IT-based OPVs.²⁷ Kim et al. demonstrated 18.53% PCE and 88% PCE retention after 500 hours at 120 °C using vinylene-linked polymer acceptors in PM6-based ternary blends.²⁸ Li et al. optimized PBQx-TCl aggregation and ternary blend morphology through PY-IT/PY-IV ratio adjustments, resulting in 18.81% PCE and 55% PCE retention after 150 hours at 85 °C. Peng et al. reported the development of a high-performance terpolymer PBZ-10, which, when paired with PY-IT in an AP binary OPV, achieved a PCE of 19.06%.²⁹ Subsequently, the introduction of PBQx-TF to create an AP ternary OPV led to a breakthrough, exceeding 20% in PCE. Despite their promise, AP ternary OPVs remain less explored than small molecule acceptor-based OPVs.^22,30,31 These advancements pave the way for highly stable AP OPV.

This study investigated OPV devices utilizing poly[(2,6-(4,8-bis(5-(2-ethylhexyl)-4-fluoro)thiophen-2-yl)-benzo[1,2-b : 4,5-b′]dithiophene)-alt-5,5′-(6,9-bis(4-(2-butyloctyl)thiophen-2-yl)dithieno[3,2-f : 2′,3′-h]quinoxaline)] (PBQx-TF), poly[(2,6-(4,8-bis(5-(2-ethylhexyl)-4-fluoro)thiophen-2-yl)-benzo[1,2-b : 4,5-b′]dithiophene)-alt-5,5′-(5,8-bis(4-(2-butyloctyl)thiophen-2-yl)dithieno[3′,2′ : 3,4; 2′′,3′′ : 5,6]benzo[1,2-c][1,2,5]thiadiazole)] (D18), and poly[[12,13-bis(2-octyldodecyl)-12,13-dihydro-3,9-diundecylbisthieno[2′′,3′′ : 4′,5′]thieno[2′,3′ : 4,5]pyrrolo[3,2-e : 2′,3′-g][2,1,3]benzothiadiazole-2,10-diyl]methylidyne[1-(dicyanomethylene)-1,3-dihydro-3-oxo-2H-inden-yl-2-ylidene]-2,5-thiophenediyl[1-(dicyanomethylene)-1,3-dihydro-3-oxo-2H-inden-yl-2-ylidene]methylidyne] (PY-IT) to elucidate factors governing device performance. We observed that distinct molecular packing variations in PBQx-TF and D18 neat films significantly impacted PY-IT layer deposition, subsequently influencing charge transport and recombination dynamics. Ternary blending effectively modulated the active layer morphology, achieving balanced electron and hole mobility and reducing charge recombination. Leveraging the established benefits of ternary strategies, including energy level optimization and energy loss minimization, the all-polymer ternary OPV device achieved a PCE of 16.07%, exceeding the PCE values of the binary reference systems, which yielded 15.26% (PBQx-TF : PY-IT) and 14.39% (D18 : PY-IT), respectively. Notably, our ternary device exhibited exceptional thermal stability, retaining 80% of its initial PCE after 1500 hours of thermal stress at 120 °C, with all device layers intact. This result demonstrates one of the highest thermal stabilities reported for all-polymer ternary OPVs, highlighting the significant contribution of D18 as a third component in enhancing both morphological stability and PCE.

Results and discussion

Fig. 1a shown the chemical structures of PBQx-TF, D18 and PY-IT. D18 achieves excellent PCEs due to its ithieno[3,2-b : 3′,4′-e][1,2-c][1,2,5]thiadiazole (DTBT) unit, which enhances crystallinity and hole mobility.³² PBQx-TF, with its wider bandgap and dithieno[3,2-f  2′,3′-h]quinoxaline (DTQx) moiety, is a potential PM6 replacement for improved high-energy photon harvesting.³³ PY-IT, a widely utilized polymer acceptor, was employed in this study to fabricate all-polymer OPVs in combination with PBQx-TF or D18. Fig. 1b illustrates the energy levels of PBQx-TF, D18, and PY-IT, showing their HOMO/LUMO values to be −5.36/−3.33 eV, −5.47/−3.53 eV, and −5.66/−3.70 eV, respectively.^34,35 D18, with its deeper HOMO level, was anticipated to yield a higher open-circuit voltage (V_OC) in the corresponding binary device compared to its PBQx-TF counterpart. The chemical structural similarity between D18 and PBQx-TF suggests the potential formation of a donor alloy phase. Consequently, the ternary OPV, incorporating D18 into the PBQx-TF binary blend, was expected to demonstrate an enhanced V_OC. Surface potential analysis via Kelvin probe revealed work function or energy level modulations in pristine PBQx-TF and D18 films, as well as their 9 : 1 blend. The measured surface potential values were −385 mV, −281 mV, and −329 mV, respectively. Incorporation of a small amount of D18 into PBQx-TF resulted in a significant increase in the surface potential of the PBQx-TF : D18 blend film, implying energy level modification. Fig. 1c presents the absorption spectra of the neat films. PBQx-TF and D18 were dissolved in chlorobenzene (CB) and deposited onto a glass substrate via spin-coating, whereas PY-IT was dissolved in chloroform (CF). Previous studies have investigated the dominant aggregation behavior (H- or J-aggregation) in DTQx-based polymer blend films by analyzing the relative intensities of the 0–1 and 0–0 vibrational peaks.³⁶ The D18 neat film exhibited a higher I_0–0/I_0–1 ratio than the PBQx-TF neat film, suggesting a dominant J-aggregation that favors hole transport. Upon incorporating 10 wt% D18 into PBQx-TF (i.e. PBQx-TF : D18), a slight increase in the I_0–0/I_0–1 ratio was observed, indicating that the aggregation of PBQx-TF can be modulated through blending with D18. The contact angles of PM6, PTQ10, and PY-IT were measured using water and diiodomethane as probe solvents (Fig. S1†). Based on these measurements, the surface energies, Flory–Huggins interaction parameters (χ), and wetting coefficients (ω) were calculated and are summarized in Table S1.† ³⁷ The surface energies of PBQx-TF, D18, and PY-IT were determined to be 37.64, 35.67, and 46.10 mJ m⁻², respectively. PY-IT exhibited a significantly higher surface energy compared to PBQx-TF and D18, resulting in Flory–Huggins interaction parameters (χ) of 0.146 and 0.227, respectively, when paired with these materials. These χ values indicate a strong tendency for phase separation in the active layer. In contrast, the similar surface energies of PBQx-TF and D18 led to a considerably lower χ value of 0.009, suggesting minimal phase separation between them. To predict the location of a third component within the active layer, the wetting coefficient (ω) was calculated. An ω value greater than 1 suggests preferential localization within the donor domain, a value less than −1 suggests localization within the acceptor domain, and a value between −1 and 1 suggests localization at the donor/acceptor interface. D18 displayed an ω value exceeding 1 when considered in the context of a PBQx-TF : PY-IT blend, indicating a tendency to reside within the PBQx-TF domain. This, combined with their structural similarity, suggests that PBQx-TF and D18 are likely to form an alloy phase. To investigate exciton dissociation, we recorded the photoluminescence (PL) spectra (Fig. S2†). Under 530 nm excitation, PBQx-TF and D18 exhibited maximum emission peaks at 686 nm and 684 nm, respectively. The PBQx-TF : D18 (9 : 1) blend film showed an emission peak comparable to that of the PBQx-TF neat film. Upon incorporating PY-IT, all blend films displayed effective quenching behavior, demonstrating quenching efficiencies above 95%, suggesting that excitons from the donor were efficiently dissociated.

Fig. 1 (a) Chemical structure, (b) energy level of PBQx-TF, D18, PY-IT. (c) UV-vis absorption spectra of PBQx-TF, D18, PY-IT, and (d) donor blend film.

Inverted organic photovoltaic (OPV) devices, employing an ITO/ZnO/active layer/MoO₃/Ag architecture (Fig. 2a), were fabricated. While inverted architectures may exhibit slightly lower power conversion efficiencies (PCEs) compared to normal structures, they offer enhanced device stability due to the inherent robustness of the oxide transport layers. To further optimize thermal resilience, the active layers were prepared using a sequential deposition (SD) method, a technique known to improve OPV thermal stability.^38,39 Specifically, the active layer was formed by pre-depositing PBQx-TF, D18 neat films, or PBQx-TF : D18 blend films, followed by the deposition of PY-IT. A comprehensive description of the fabrication procedure can be found in the ESI.† The J–V curve of OPV was recorded under AM1.5 G illumination (100 mW cm⁻²). Optimization of the device performance was achieved through fine-tuning the donor and total film thicknesses (Tables S2 and S3†). Given the structural similarity between PBQx-TF and D18, the binary devices exhibited nearly identical optimized fabrication parameters. The average OPV performance was determined from at least five individual devices. The J–V curves and corresponding parameters are presented in Fig. 2b and Table 1. The PBQx-TF binary OPV achieved an average PCE of 15.03 ± 0.16%, characterized by a short-circuit current density (J_SC) of 21.82 ± 0.38 mA cm⁻², an V_OC of 0.92 ± 0.01 V, and a fill factor (FF) of 74.3 ± 0.1%. In comparison, the D18 binary OPV demonstrated an average PCE of 14.21 ± 0.23%, with a J_SC of 20.63 ± 0.43 mA cm⁻², a V_OC of 0.97 ± 0.01 V, and an FF of 71.5 ± 0.2%. The D18 binary device, possessing a deeper HOMO level, exhibited a higher V_OC than the PBQx-TF binary device, as anticipated. Nevertheless, the reduced J_SC and FF values resulted in a lower PCE for the D18 binary OPV relative to the PBQx-TF binary OPV. The ternary device was then optimized by adjusting the D18 component content, as detailed in Table S2.† The optimal device, incorporating 10 wt% D18, exhibited the higher PCE to be 15.95 ± 0.07%, with a J_SC of 22.36 ± 0.43 mA cm⁻², a V_OC of 0.95 ± 0.01 V, and an FF of 75.0 ± 0.1%. A maximum PCE of 16.07% was achieved for the ternary OPV, due to simultaneous improvements in J_SC, V_OC and FF relative to the PBQx-TF binary OPV. Fig. 2c shown the external quantum efficiencies (EQE) spectra of binary and ternary devices. The integrated current density values (EQE-J_SC), derived from the EQE spectra, were 20.4, 19.5, and 20.9 mA cm⁻² for the PBQx-TF binary, D18 binary, and ternary OPVs, respectively. Well agreement was observed between these values and the J_SC values derived from the J–V curves. The minor deviations can be attributed to several factors, including variations in measurement conditions between the solar simulator and the EQE system, and differences in their respective light sources.⁴⁰ EQE spectra of the two binary OPVs revealed comparable EQE in the donor response region, but a notable difference in the PY-IT response region. Specifically, the D18 binary device showed a reduced EQE in the PY-IT absorption range, indicating limited charge extraction from PY-IT within the active layer. Conversely, the PBQx-TF binary device maintained a higher EQE in this region, explaining its superior JSC compared to the D18 binary device. Furthermore, the ternary OPV exhibited an enhanced EQE compared to the PBQx-TF binary OPV, particularly within the donor response region. This suggests that the incorporation of a small amount of D18 effectively modulated the molecular packing of PBQx-TF, leading to improved charge extraction.

Fig. 2 (a) OPV device structure. (b) J–V and (c) EQE curves of binary and ternary OPVs.

Table 1 Photovoltaic device parameters of binary and ternary OPVs

Active layer	J SC (mA cm^−2)	V OC (V)	FF (%)	PCE (%)	Best PCE (%)
PBQx-TF/PY-IT	21.82 (±0.38)	0.92 (±0.01)	74.3 (±0.1)	15.03 (±0.16)	15.26
PBQx-TF : D18/PY-IT	22.36 (±0.43)	0.95 (±0.01)	75.0 (±0.1)	15.95 (±0.07)	16.07
D18/PY-IT	20.63 (±0.43)	0.97 (±0.01)	71.5 (±0.2)	14.21 (±0.23)	14.39

---

Atomic force microscopy (AFM) and grazing-incidence wide-angle X-ray scattering (GIWAXS) were employed to gather morphological characterization of the blend films, thereby elucidating the impact of blend morphology variations on device performance. Topographical and phase images of the PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT blend films are shown in Fig. S3.† Topographical analysis revealed root-mean-square (RMS) roughness values of 5.7 nm, 2.8 nm, and 3.0 nm for the PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT films, respectively. The PBQx-TF/PY-IT film exhibited the highest surface roughness, potentially leading to interfacial defects during the top transport layer deposition and consequently, impaired charge collection.⁴¹ Notably, incorporating D18 into the ternary blend significantly reduced the roughness to a level comparable to the D18 binary film, suggesting improved interfacial charge collection. Phase images of all three films displayed a network morphology with distinct fiber-like domains, conducive to efficient charge transport. The domain size, quantified by measuring the fiber-like domain diameter, was approximately 14 nm for all films. This consistent domain size correlates with the observed efficient PL quenching, indicating a well-mixed morphology across all compositions. Phase images revealed a higher contrast between bright and dark regions for the D18/PY-IT film relative to the PBQx-TF/PY-IT film. This increased contrast implies a greater disparity in mechanical properties, possibly arising from variations in molecular packing. The ternary blend exhibited an intermediate contrast, suggesting that D18 effectively modulated the molecular packing or phase separation characteristics.

Two-dimensional (2D) GIWAXS patterns of the neat, donor blend, binary blend, and ternary blend films are shown in Fig. 3a. One-dimensional (1D) GIWAXS profiles were subsequently extracted from the 2D patterns along both the out-of-plane (OP, solid line) and in-plane (IP, dashed line) directions, as shown in Fig. 3b. Distinct diffraction peaks were observed in the D18 neat film at 1.731 Å⁻¹ (OP) and 0.317 Å⁻¹ (IP), which correlate to the (010) and (100) planes with corresponding distances of 3.629 Å and 19.802 Å, respectively.⁴² Due to the similar molecular chemical structure, PBQx-TF exhibited a diffraction pattern similar to that of D18, with the (010) and (100) planes located at 1.701 Å⁻¹ (OP) and 0.313 Å⁻¹ (IP), respectively, corresponding to interplanar distances of 3.694 Å and 20.057 Å. To further compare the packing differences between PBQx-TF and D18, we calculated the crystalline coherence lengths of π–π stacking (L_c-ππ), yielding values of 21.0 and 22.9 Å, respectively. Upon the introduction of a small fraction of D18 into the PBQx-TF film, the resulting PBQx-TF : D18 donor blend film exhibited diffraction patterns resembling that of the PBQx-TF neat film, with the (010) and (100) reflections observed at 1.702 Å⁻¹ (OP) and 0.313 Å⁻¹ (IP), corresponding to interplanar distances of 3.692 Å and 20.057 Å. The L_c-ππ of the PBQx-TF : D18 blend film was measured to be 21.3 Å, a slight increase compared to the PBQx-TF neat film, indicating improved molecular packing. Furthermore, the (100) signal was observed in both OP and IP directions for neat and donor blend films. We calculated the area ratio of the OP (100) to the IP (100) signals, which were 4.52 and 5.70 for PBQx-TF and D18 neat films, respectively. A higher area ratio suggests that the crystalline domains are mainly distributed along the OP direction, which is more favorable for charge transport in the vertical direction. Upon blending a small amount of D18 with PBQx-TF, the area ratio of the donor blend film increased to 4.66, indicating that the donor blend film becomes more conducive to vertical charge transport, likely contributing to the enhanced device performance. Subsequently, PY-IT was deposited onto these donor films via SD to form the complete active layer. During this process, PY-IT is expected to infiltrate downwards into the donor layer through solvent wetting, thereby forming a bulk heterojunction (BHJ). Phase images, which are presumably indicative of variations in domain density or mechanical stiffness, revealed a higher contrast between bright and dark regions for the D18/PY-IT film than the PBQx-TF/PY-IT film. This increased contrast implies a greater disparity in mechanical properties, possibly arising from variations in molecular packing. Both binary blend films exhibited distinct π–π stacking signals, with corresponding location/distances/L_c-ππ values of 1.711 Å⁻¹/3.672 Å/25.3 Å for the PBQx-TF/PY-IT blend film and 1.734 Å⁻¹/3.624 Å/26.8 Å for the D18/PY-IT blend film. Although D18, due to its relatively larger χ parameter indicating lower miscibility with PY-IT, promoted more distinct microstructural packing—reflected by a reduced π–π stacking distance and an increased L_c-ππ in the D18/PY-IT blend film—its overall diffraction intensity remained significantly lower than that of the PBQx-TF/PY-IT blend film. This suggests that the lower crystallinity or higher miscibility (i.e., lower χ parameter) of PBQx-TF may enhance the molecular intermixing with PY-IT during the spin-drying process, thereby promoting the formation of a greater number of smaller crystalline domains, which is favorable for charge transport, as schematically depicted in the upper panel of Fig. 3c. In contrast, D18 appears to induce the formation of only a few larger crystalline domains of PY-IT (the middle panel of Fig. 3c), consequently limiting charge extraction in the device and resulting in a lower EQE at longer wavelengths. The ternary blend film exhibited location/distances/L_c-ππ values of 1.712 Å⁻¹/3.670 Å/26.0 Å, demonstrating an improved L_c-ππ while maintaining the strong diffraction intensity of the PBQx-TF : PY-IT blend. This suggests that the controlled introduction of D18 into the PBQx-TF film can optimize molecular packing and induce the growth of larger crystalline domains without significantly impeding PY-IT infiltration (the bottom panel of Fig. 3c). This is expected to suppress unwanted charge recombination pathways, ultimately leading to the highest FF and PCE observed for the ternary device.

Fig. 3 (a) 2D-GIWAXS and (b) 1D-GIWAXS profiles of neat and blend films. (c) Schematic diagram illustrating active layer formation during the SD process.

However, alterations in composition or blend morphology can directly affect the efficiency of charge dissociation and extraction. When free holes and electrons are not effectively dissociated or extracted, they undergo recombination, leading to energy dissipation through pathways like radiation and heat, which is termed energy loss (E_loss). E_loss analysis, encompassing Fourier transform photocurrent spectroscopy–EQE (FTPS-EQE) and electroluminescence (EL) measurements, is widely utilized as an effective tool for elucidating V_OC variations in OPVs, as illustrated in Fig. 4a–d. E_loss was quantified as the difference between the V_OC and the device bandgap energy (E_gap) using the relationship E_loss = E_gap – qV_OC, where E_gap was determined from the intersection of the FTPS-EQE and EL spectra. Given the comparable E_gap of the three devices, the E_loss exhibited a positive correlation with the V_OC. The E_loss values for the PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT-based OPVs were determined to be 0.560, 0.510, and 0.527 eV, respectively. The E_loss in the OPVs can be further dissected into three fundamental contributions: the energy loss incurred during charge generation (ΔE_CT), the inherent energy loss stemming from radiative recombination (ΔE_rad), and the energy loss arising from non-radiative recombination (ΔE_non-rad). The ΔE_CT was quantified as the difference between the E_gap and the charge transfer energy (E_CT), where E_CT was determined from the intersection of the fitting curves obtained from FTPS-EQE and EL spectra. Typically, the E_CT is expected a positive correlation with the energy difference between the donor's HOMO and acceptor's LUMO. Herein, the E_CT values for the PBQx-TF and D18 binary OPVs were determined to be 1.425 and 1.435 eV, respectively, which is consistent with the expected trend based on their energy level variations. The E_CT value for the ternary OPV was determined to be 1.447 eV, which was higher than those of both binary OPVs. This suggests that in addition to the typical alloy-phase-modulated energy level, the device performance is also influenced by variations in the donor/acceptor (D/A) interfacial geometry.⁴³ However, effective experimental techniques for directly observing the D/A interfacial geometry within BHJ are currently lacking, thus limiting further exploration in this study. Owing to its highest E_CT, the ternary OPV exhibited the smallest ΔE_CT of 0.030 eV, in comparison to the PBQx-TF and D18 binary OPVs which displayed ΔE_CT values of 0.055 and 0.045 eV, respectively. ΔE_rad inherently accounts for the unavoidable energy loss dictated by radiative recombination pathways, resulting in similar values for the three OPVs. Lastly, ΔE_non-rad, representing the energy loss through non-radiative recombination, was quantitatively assessed using the equation ΔE_non-rad = −k_BT/q ln(EQE_EL), where k_B, T, and q are the Boltzmann constant, absolute temperature and elementary charge, respectively. The detailed parameters are summarized in Table S4.† The ΔE_non-rad value for the PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT-based OPVs were determined to be 0.228, 0.187, and 0.223 eV, respectively. In comparison to the two binary OPVs, the PBQx-TF/PY-IT-based OPV displayed a higher ΔE_non-rad, which can be ascribed to the lower molecular packing of PBQx-TF and the induction of numerous small PY-IT crystallites, potentially leading to unwanted charge recombination during carrier transport. However, the introduction of an optimized amount of D18 resulted in a slight improvement of these factors, leading to a lower ΔE_non-rad in the ternary OPV. In summary, the lower E_loss observed in the ternary OPV can be attributed to its higher E_CT and the slightly improved ΔE_non-rad, consequently resulting in a higher V_OC than the PBQx-TF binary OPV, demonstrating the benefits of ternary blending for V_OC enhancement.

Fig. 4 The normalized and fitted FTPS-EQE and EL curves of (a) PBQx-TF/PY-IT-, (b) PBQx-TF : D18/PY-IT-, and (c) D18/PY-IT-based OPVs. (d) The EQE_EL spectra of the binary and ternary OPVs. Plots of (e) J_SC- light intensity, (f) V_OC- light intensity, (g) The histogram of the hole and electron mobilities of devices. (h) Photo-CELIV, and (i) TPC of the binary and ternary OPVs.

To gain deeper insights into the charge transport and recombination, the J–V characteristics of both binary and ternary OPVs were recorded under a range of light intensities (P_light). The resulting J_SC–P_light and V_OC–P_light relationships are presented in Fig. 4e–f. To analyze the degree of bimolecular recombination, the J_SC−P_light curves were fitted using the equation J_SC ∝ P_light^α, where α is the fitting parameter. A deviation of α from 1 signifies an increase in bimolecular recombination. The values of α for PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT-based OPVs were 0.998, 0.983, and 1.000, respectively. The PBQx-TF binary OPV exhibited a higher α value than the D18 binary, suggesting that the blend morphology of the PBQx-TF binary is less prone to bimolecular recombination. Upon the introduction of D18 as a third component, bimolecular recombination in the ternary OPV was almost negligible. To analyze the degree of trap-assisted recombination, the V_OC–P_light curves were fitted using the equation V_OC = (nk_BT/q) ln(P_light), where n is the fitting parameter. The values of n for PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT-based OPVs were 1.35, 1.25, and 1.32, respectively. The D18 binary exhibited the lowest n value, indicating that an active layer composed of larger crystallites can effectively suppress trap-assisted recombination in the device, a finding consistent with the lowest ΔE_non-rad observed for the D18 binary OPV. Furthermore, the ternary blend film formed by incorporating D18 also displayed slightly larger crystallites than the PBQx-TF binary, leading to an improved suppression of trap-assisted recombination, as evidenced by a reduction in the n value. However, the degree of molecular packing within the active layer materials typically directly influences electron and hole mobility, consequently affecting charge carrier balance and overall device performance. Here, to quantify these transport properties, we fabricated hole-only and electron-only devices and employed the space-charge-limited current (SCLC) method to determine the hole mobilities (μ_h) and electron mobilities (μ_e) of the three active layer compositions. Fig. S4† shows the experimental current density–voltage (J–V) characteristics for the hole-only and electron-only devices, and the calculated mobilities are summarized in Fig. 4g. The μ_h of the PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT-based devices were determined to be 3.84 × 10⁻⁴, 7.00 × 10⁻⁴, and 4.57 × 10⁻⁴ cm² V⁻¹ s⁻¹, respectively, while the corresponding μ_e were found to be 7.06 × 10⁻⁴, 4.65 × 10⁻⁴, and 6.66 × 10⁻⁴ cm² V⁻¹ s⁻¹, respectively. The PBQx-TF binary device exhibited a lower μ_h and a higher μ_e compared to the D18 binary device. This can be attributed to the intrinsically less efficient molecular packing of PBQx-TF and its ability to induce effective stacking of PY-IT, respectively. Upon the introduction of D18, the ternary device demonstrated an improved μ_h while maintaining excellent μ_e, resulting in a more balanced μ_e/μ_h ratio of 0.69, which is closer to the ideal value of 1 for efficient charge collection, compared to the ratios of 0.54 for the PBQx-TF binary and 1.51 for the D18 binary. Subsequently, we employed the photoinduced charge extraction by linearly increasing voltage (Photo-CELIV) method to directly probe the OPV devices and obtain the corresponding carrier mobility, which is a combined result of electron and hole mobilities, as shown in Fig. 4h. The carrier mobilities of the PBQx-TF/PY-IT, D18/PY-IT, and PBQx-TF : D18/PY-IT-based OPVs were determined to be 4.15 × 10⁻⁵, 4.08 × 10⁻⁵, and 5.72 × 10⁻⁵ cm² V⁻¹ s⁻¹, respectively. The ternary OPV exhibited the highest carrier mobility, attributed to its more balanced electron and hole mobilities, a finding consistent with the SCLC results. This high carrier mobility is crucial for enabling faster and more efficient charge extraction in the device, a characteristic that can be further investigated and confirmed through transient photocurrent (TPC) measurements. Fig. 4i illustrates the photocurrent density decay transients of the binary and ternary OPV, as acquired via TPC measurement. These decay curves were analyzed by fitting them with a mono-exponential decay function, yielding charge extraction times of 0.741 μs for the PBQx-TF/PY-IT-based OPV, 0.714 μs for the D18/PY-IT-based OPV, and 0.695 μs for the PBQx-TF : D18/PY-IT-based OPV, respectively. As anticipated due to its balanced charge transport, the ternary OPV exhibited the shortest charge extraction time. All-in-all, despite their similar chemical structures, PBQx-TF and D18 exhibit significant differences in molecular packing and miscibility, profoundly impacting the SD process and the resulting active layer morphology. The weak molecular packing and high miscibility of PBQx-TF facilitate PY-IT infiltration and crystallization, leading to higher μ_e at the expense of μ_h. Conversely, while D18 demonstrates good molecular packing, its high crystallinity and low miscibility limit PY-IT infiltration and crystallization, resulting in high μ_h but low μ_e. By strategically introducing D18 as a third component into the PBQx-TF binary system, the molecular packing within the donor domain can be further improved while maintaining good PY-IT infiltration. This enables the resulting ternary devices to achieve more balanced charge carrier mobilities and suppressed unwanted charge recombination, leading to enhancements in J_SC and FF. Combined with a high E_CT and low E_loss, the ternary devices ultimately exhibit the highest PCE, highlighting the effectiveness of ternary blending for optimizing both charge transport and energy loss in OPV.

Finally, unencapsulated binary and ternary OPVs (complete device including top transport layer and anode) were stored in a dark, Ar-filled glovebox, and thermal stability tests were conducted at 120 °C. This accelerated aging test aimed to better observe the morphological evolution across the different blend systems. To enable a fair comparison despite different starting performances, the performance of each device was normalized to its initial value. Average performance changes were obtained through statistical analysis of several independent devices to ensure reproducibility, and the evolution of key device parameters over time is summarized in Fig. 5a and S5.† Despite exhibiting a lower initial PCE, the D18 binary OPV demonstrated outstanding device stability, retaining over 90% of its initial PCE after 500 hours of heating and still maintaining over 84% after 1500 hours. In contrast, the PBQx-TF binary OPV, which possessed a relatively high initial PCE, showed inferior stability, with its PCE decreasing to 80% of its initial value (T₈₀) at approximately 1080 hours, primarily due to a significant reduction in J_SC and FF. While the stability of the ternary OPV did not surpass that of the D18 binary OPV, it was superior to the PBQx-TF binary, with a T₈₀ exceeding 1500 hours, attributed to a less decrease in J_SC and FF, highlighting the beneficial role of ternary blending in enhancing long-term stability.

Fig. 5 (a) Normalized PCE of binary and ternary OPVs under thermal aging process at 120 °C. (b) Plot of I_t/I₀ – heating time and I_0–0/I_0–1 – heating time of binary and ternary blend films. (c) The AFM topographic and phase images of binary and ternary blend films aged 696 h. (d) Arrhenius plot with respective linear fitting for ternary OPVs.

To understand the reason of the difference in thermal stability, those active layers were subjected to UV-vis absorption spectroscopy and AFM to monitor changes in their optical properties and surface morphology at different heating durations, as shown in Fig. S3 and S6.† In the UV-vis spectra, all three blend films exhibited similar absorption profiles in the long-wavelength region, with only minor intensity variations and a slight red-shift, indicating that prolonged heating had a limited impact on PY-IT. In contrast, the short-wavelength absorption, primarily contributed by the polymer donors, showed the significant differences, particularly in the 0–0 and 0–1 vibrational peaks. During the initial heating stages, a decrease in the intensity of the 0–0 and 0–1 peaks was observed for all blend films, suggesting the disruption or reorganization of some metastable aggregates. As the heating time was extended, the changes in absorption intensity became less pronounced, indicating that the microstructure tended towards stabilization. The relative change in absorption intensity was quantified by the ratio I_t/I₀, where I_t and I₀ are the maximum absorption intensities at a given heating time and initially, respectively, as presented in Fig. 5b. The D18 binary blend film tended towards stabilization after 400 hours of heating, whereas the PBQx-TF binary blend film still exhibited a noticeable decrease after the same duration, indicating ongoing morphological instability. However, the introduction of a small amount of D18 effectively mitigated this issue, leading to a significantly improved reduction in the I_t/I₀ ratio for the ternary blend film after 400 hours, suggesting an improvement of morphological stability. Beyond the decrease in absorption intensity, the heating process also induced changes in the I_0–0/I_0–1 ratio, with the values at different heating times summarized in Fig. 5b. Among the three blend film, the D18 binary blend film exhibited the highest I_0–0/I_0–1 ratio at all heating time, indicating that D18 maintained robust J-aggregation. The PBQx-TF binary blend film showed a decrease in the I_0–0/I_0–1 ratio, suggesting that the aggregation of PBQx-TF was affected by heating and implying the morphological instability. In contrast, the ternary blend film maintained a consistent I_0–0/I_0–1 ratio throughout the heating process, indicating this factor was largely unaffected by heating time, further supporting the superior morphological stability of the ternary blend film. Fig. 5c and S3† illustrate the evolution of topographical and phase images at different heating durations. Between 0 and 360 hours of heating, all blend films exhibited similar surface morphologies, with distinct fiber-like domains still observable in the phase image. However, as the heating approached 700 hours, the fiber-like domains became less pronounced in certain regions, a phenomenon particularly evident in the PBQx-TF binary, as indicated by the yellow circles. This suggests that prolonged heating induces macroscopic morphological changes, thereby limiting device performance. In contrast, under the same heating conditions, the ternary blend film maintained well-defined fiber-like domains, which is crucial for efficient charge transport and collection, ultimately contributing to the superior thermal stability of the ternary OPV compared to the PBQx-TF binary OPV.

Table S5† presents a compilation of AP OPV reports demonstrating superior thermal stability. It is noteworthy that in a few of the studies investigating thermal stability above 100 °C, the heating process is applied to incomplete devices comprising the active layer, with subsequent deposition of transport layers and electrodes prior to performance evaluation. Furthermore, most studies have only reported thermal stability changes within 200 hours, suggesting a persistent lack of long-term thermal stability monitoring at high temperatures. In contrast, our study employed complete OPV devices for thermal stability assessment, allowing for continuous monitoring of device performance evolution until T₈₀. This rigorous approach provides stronger evidence for the inherent high stability of AP OPVs. At the same temperature (i.e. 120 °C), our devices exhibit a T₈₀ value comparable to that of the best ternary OPV devices reported by Kim et al., where their T₈₀ was extrapolated from stability curves.²⁸ This underscores the excellent thermal stability of our ternary OPV devices. Furthermore, we also monitored the thermal stability of the ternary OPVs at 100 °C and estimated their T₈₀ to be 2544 hours by fitting the linear decay portion in the later heating stage. Utilizing the thermal stability data obtained at 100 °C and 120 °C, we determined the activation energy (E_a) for the ternary OPVs. Subsequently, the Arrhenius equation was employed to predict the T₈₀ lifetimes of the ternary OPV at various operational temperatures,⁴⁴ with the results summarized in Fig. 5d. The extrapolated T₈₀ values for the ternary devices at 85 °C, 65 °C, and 25 °C were 3890, 7158, and 32570 hours, respectively, indicating promising long-term stability even at moderate operating temperatures. Moreover, we recorded the device stability under continuous illumination (6500 k LED; 2.23 mW cm⁻², 6400 lx) in Ar-filled glovebox, as shown in Fig. S7.† With over 83% of their initial PCE maintained after 200 hours, AP-OPV devices clearly demonstrate excellent thermal stability and photostability. These findings highlight their increased potential for commercial applications. This study demonstrates that incorporating a highly crystalline third polymer effectively enhances both PCE and thermal stability by tuning active layer morphology. This strategy may be extended to other ternary OPVs, particularly those using benzodithiophene (BDT)-based donors with reduced crystallinity for miscibility, where adding a structurally similar crystalline donor could boost performance.

Conclusion

This study investigates the fabrication of highly stable AP inverted OPV devices employing a ternary active layer composed of PBQx-TF, D18, and PY-IT, utilizing a SD method. We conducted a systematic comparative analysis of device performance and thermal stability between binary and ternary OPVs. Despite their structural similarities, PBQx-TF and D18 exhibited distinct molecular packing characteristics, which significantly impacted PY-IT deposition and the resulting active layer morphology. This led to imbalanced charge carrier mobilities and limited PCEs in the binary OPVs. However, the incorporation of a small fraction of D18 into the PBQx-TF film optimized donor blend film packing without impeding PY-IT infiltration, enabling the ternary OPV to achieve balanced charge transport and suppressed charge recombination. Leveraging the established advantages of ternary strategies, including energy level tuning and energy loss reduction, the all-polymer ternary OPV achieved a PCE of 16.07%, exceeding the binary OPV PCEs of 15.26% (PBQx-TF) and 14.39% (D18). Furthermore, the ternary OPV demonstrated superior thermal stability compared to the PBQx-TF binary OPV, exhibiting a T₈₀ of 1530 hours at 120 °C. This enhanced stability is attributed to D18-mediated optimization of donor blend film packing, which mitigates performance degradation resulting from morphological instability.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

The authors acknowledge the financial support from the National Science and Technology Council in Taiwan (NSTC 112-2113-M-131-001; NSTC 111-2221-E-131 -021 -MY3).

References

1. Y. C. Tseng, L. H. Yeh, H. L. Yip and C. C. Chueh, Small, 2025, e2412507.
2. Z. Fu, J. W. Qiao, F. Z. Cui, W. Q. Zhang, L. H. Wang, P. Lu, H. Yin, X. Y. Du, W. Qin and X. T. Hao, Adv. Mater., 2024, 36, e2313532.
3. C. Yan, S. Barlow, Z. Wang, H. Yan, A. K. Y. Jen, S. R. Marder and X. Zhan, Nat. Rev. Mater., 2018, 3, 18003.
4. D. Luo, W. Jang, D. D. Babu, M. S. Kim, D. H. Wang and A. K. K. Kyaw, J. Mater. Chem. A, 2022, 10, 3255–3295.
5. K. H. Huang, B. H. Jiang, H. C. Lu, Y. J. Xue, C. F. Lu, Y. Y. Chang, C. L. Huang, S. Y. Chien, C. P. Chen and Y. J. Cheng, Adv. Sci., 2025, 12, e2413045.
6. Y. B. Wang, C. L. Tsai, Y. J. Xue, B. H. Jiang, H. C. Lu, J. C. Hong, Y. C. Huang, K. H. Huang, S. Y. Chien, C. P. Chen and Y. J. Cheng, Chem. Sci., 2025, 16, 3259–3274.
7. H. Zhou, L. Zhang, H. Tian, Y. Ni, Y. Xie, S. Y. Jeong, T. Huang, H. Y. Woo, J. Zhang, X. Zhu, L. Lu, X. Ma and F. Zhang, Small, 2025, 21, e2410581.
8. H. Tian, H. Zhou, L. Zhang, W. Xu, R. Gong, Y. Ni, S. Y. Jeong, X. Zhu, H. Y. Woo, X. Ma, L. Lu and F. Zhang, Adv. Funct. Mater., 2025, 2422867.
9. Y. Ni, J. Li, M. Zhang, H. Tian, H. Zhou, L. Zhang, W. Xu, G. Fu, X. Lü, S. Y. Jeong, W.-Y. Wong, H. Y. Woo, X. Ma and F. Zhang, Chem. Eng. J., 2025, 507, 160359.
10. N. Zhang, K. Jiang, F. R. Lin, Y. An, G. Du, T. Xia, A. K. Y. Jen and H.-L. Yip, Mater. Today Energy, 2024, 42, 101548.
11. Y. Jiang, S. Sun, R. Xu, F. Liu, X. Miao, G. Ran, K. Liu, Y. Yi, W. Zhang and X. Zhu, Nat. Energy, 2024, 9, 975–986.
12. Y.-Y. Yu, J.-H. Xu, B.-H. Jiang, Y.-C. Hsu and C.-P. Chen, ACS Appl. Energy Mater., 2025, 8, 1776–1782.
13. K. N. Zhang, X. Y. Du, L. Yan, Y. J. Pu, K. Tajima, X. Wang and X. T. Hao, Small Methods, 2024, 8, e2300397.
14. W. Li, D. Liu and T. Wang, Adv. Funct. Mater., 2021, 31, 2104552.
15. J. Xiao, N. Li, Q. Yin, Y. Min and H. L. Yip, Adv. Opt. Mater., 2024, 12, 2302202.
16. G. Du, W. Jiang, N. Zhang, Y. Wang, M. Liu, T. Lei, Y. An, L. Ke, C. Ge, F. R. Lin, A. K. Y. Jen and H.-L. Yip, Energy Environ. Sci., 2025, 18, 799–806.
17. L. Duan and A. Uddin, Adv. Sci., 2020, 7, 1903259.
18. B. Li, X. Yang, S. Li and J. Yuan, Energy Environ. Sci., 2023, 16, 723–744.
19. B. Liu, H. Sun, J. W. Lee, Z. Jiang, J. Qiao, J. Wang, J. Yang, K. Feng, Q. Liao, M. An, B. Li, D. Han, B. Xu, H. Lian, L. Niu, B. J. Kim and X. Guo, Nat. Commun., 2023, 14, 967.
20. S. Yoon, B. Reyes-Suárez, S. T. Pham, H. Vezin, Y. A. Tobon, M. Lee, S. Mugiraneza, B. M. Kim, M. Y. T. Oide, S. Yoo, S. Lee, S. H. Wang, S. M. Collins, C. M. Bates, Y. Park, B. Kim, G. N. Manjunatha Reddy and T.-Q. Nguyen, ACS Energy Lett., 2025, 10, 541–551.
21. J. Yi, G. Zhang, H. Yu and H. Yan, Nat. Rev. Mater., 2023, 9, 46–62.
22. K. Zhou, K. Xian and L. Ye, InfoMat, 2022, 4, e12270.
23. Y. Zhang, Y. Xu, M. J. Ford, F. Li, J. Sun, X. Ling, Y. Wang, J. Gu, J. Yuan and W. Ma, Adv. Energy Mater., 2018, 8, 1800029.
24. J. Wang, C. Sun, Y. Li, F. Bi, H. Jiang, C. Yang, X. Bao and J. Chu, Nat. Commun., 2025, 16, 1784.
25. R. Sun, W. Wang, H. Yu, Z. Chen, X. Xia, H. Shen, J. Guo, M. Shi, Y. Zheng, Y. Wu, W. Yang, T. Wang, Q. Wu, Y. Yang, X. Lu, J. Xia, C. J. Brabec, H. Yan, Y. Li and J. Min, Joule, 2021, 5, 1548–1565.
26. G. U. Kim, C. Choi, D. Jeong, D. J. Kim, T. N. L. Phan, S. Song, J. Park, T. S. Kim, Y. H. Kim and B. J. Kim, Adv. Energy Mater., 2023, 13, 2302125.
27. L. Ma, Y. Cui, J. Zhang, K. Xian, Z. Chen, K. Zhou, T. Zhang, W. Wang, H. Yao, S. Zhang, X. Hao, L. Ye and J. Hou, Adv. Mater., 2023, 35, e2208926.
28. J.-W. Lee, J. Kim, T. H.-Q. Nguyen, D. C. Lee, Z. Tan, J. Park, T. N.-L. Phan, S. Cho and B. J. Kim, Nano Energy, 2024, 122, 109338.
29. W. Qiu, C. Liao, Y. Li, M. Deng, Y. Duan, X. Xu and Q. Peng, Adv. Funct. Mater., 2025, 2503009.
30. H. Yin, C. Yan, H. Hu, J. K. W. Ho, X. Zhan, G. Li and S. K. So, Mater. Sci. Eng., R, 2020, 140, 100542.
31. P. Cong, Z. Wang, Y. Geng, Y. Meng, C. Meng, L. Chen, A. Tang and E. Zhou, Nano Energy, 2023, 105, 108017.
32. Q. Liu, Y. Jiang, K. Jin, J. Qin, J. Xu, W. Li, J. Xiong, J. Liu, Z. Xiao, K. Sun, S. Yang, X. Zhang and L. Ding, Sci. Bull., 2020, 65, 272–275.
33. Y. Cui, Y. Xu, H. Yao, P. Bi, L. Hong, J. Zhang, Y. Zu, T. Zhang, J. Qin, J. Ren, Z. Chen, C. He, X. Hao, Z. Wei and J. Hou, Adv. Mater., 2021, 33, e2102420.
34. T. Zhang, Y. Xu, H. Yao, J. Zhang, P. Bi, Z. Chen, J. Wang, Y. Cui, L. Ma, K. Xian, Z. Li, X. Hao, Z. Wei and J. Hou, Energy Environ. Sci., 2023, 16, 1581–1589.
35. Y. Cheng, Q. Mao, C. Zhou, X. Huang, J. Liu, J. Deng, Z. Sun, S. Jeong, Y. Cho, Y. Zhang, B. Huang, F. Wu, C. Yang and L. Chen, Angew Chem. Int. Ed. Engl., 2023, 62, e202308267.
36. R. Ma, H. Li, T. A. Dela Pena, X. Xie, P. W. Fong, Q. Wei, C. Yan, J. Wu, P. Cheng, M. Li and G. Li, Adv. Mater., 2024, 36, e2304632.
37. C.-Y. Lin, B.-H. Jiang, P.-J. Weng, Y. Hsuan Lin, Y.-W. Su, H.-S. Shih, Z.-E. Shi, Y.-R. Lin, J. Vailassery, S.-S. Sun, C.-P. Chen and Y. J. Chang, Chem. Eng. J., 2024, 494, 153183.
38. J. Chen, J. Cao, L. Liu, L. Xie, H. Zhou, J. Zhang, K. Zhang, M. Xiao and F. Huang, Adv. Funct. Mater., 2022, 32, 2200629.
39. T. Kumari, I. Vyalih, M. Á. León Luna, H. Ahmed, M. Ahmad, R. Atajanov, E. Jayaraman, S. Manikandan, B. Paci, A. Di Carlo, J. W. Andreasen, V. Turkovic and M. Madsen, Cell Rep. Phys. Sci., 2024, 5, 102027.
40. M. Saliba and L. Etgar, ACS Energy Lett., 2020, 5, 2886–2888.
41. B.-H. Jiang, P.-H. Chan, Y.-W. Su, H.-L. Hsu, R.-J. Jeng and C.-P. Chen, Org. Electron., 2020, 87, 105944.
42. Z. Wang, Z. Peng, Z. Xiao, D. Seyitliyev, K. Gundogdu, L. Ding and H. Ade, Adv. Mater., 2020, 32, e2005386.
43. Y. L. Lin, M. A. Fusella and B. P. Rand, Adv. Energy Mater., 2018, 8, 1702816.
44. G.-L. Chen, S.-H. Wang, K.-W. Tseng, C.-I. Huang and L. Wang, J. Mater. Chem. A, 2025, 13, 2668–2676.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta02541a

‡ B.-H. Jiang and Z.-R. Huang contributed equally to this work.

---