Synergistic Steric Hindrance and Chlorination Enable Efficient Binary Organic Solar Cells with Low Energy Loss

Abstract

We report a series of small-molecule acceptors Z1–Z4 that reconcile low non-radiative energy loss with efficient charge transport by combining a nitrogen-containing core with out-of-plane steric-hindrance/chlorination-regulated packing. The dibenzo[b,f]azepine unit enables low non-radiative energy losses below 0.20 eV, while progressive backbone unlocking and targeted chlorination reshape intermolecular packing by suppressing unfavorable core-centered aggregation and promoting multidimensional electronic coupling. Single-crystal analysis, GIWAXS, and electronic-structure calculations reveal enhanced packing coherence and stronger coupling pathways from Z1 to Z4. Transient absorption spectroscopy and kinetic analysis further show that improved exciton delocalization accelerates interfacial exciton dissociation and suppresses monomolecular recombination. Photo-induced force microscopy confirms more continuous acceptor-rich fibrils in Z3- and Z4-based blends. Consequently, the optimized Z4-based binary device achieves a PCE of 20.28% with a high open-circuit voltage of 0.95 V, demonstrating an effective molecular strategy for low energy-loss high-performance organic solar cells.

Structure-property design principles are necessary for improving organic solar cells’ performance. Here, the authors report small-molecule acceptors with low non-radiative energy loss and efficient charge transport using a nitrogen-containing core with out-of-plane steric-hindrance/chlorination-regulated packing.

Introduction

The development of high-performance organic solar cells (OSCs) has been propelled by the rational design of non-fullerene small-molecule acceptors (SMAs), in which fine-tuning the molecular backbone provides precise control over optoelectronic properties and solid-state packing^1–10. Among the structural motifs, the A-DA₁D-A architecture has proven particularly successful, where the electron-deficient central unit A₁ plays a decisive role in dictating intermolecular interactions and charge-transport characteristics^11–27. Subtle variations in the A₁ core not only modulate frontier orbital energies but also strongly influence exciton behavior, including delocalization, diffusion, and recombination losses^28–38. In particular, simultaneously suppressing non-radiative energy loss and maintaining efficient charge transport remains a critical challenge, because molecular motifs that enhance luminescence or reduce non-radiative recombination do not always generate favorable intermolecular packing networks for charge transport.

A central challenge in advancing OSC performance lies in understanding and regulating exciton delocalization^39–43. In typical organic semiconductors, strong Coulombic binding confines excitons within a single chromophore, limiting their diffusion length and increasing the probability of non-radiative decay before charge separation^44–48. In contrast, spatially delocalized excitons benefit from reduced binding energy, enhanced mobility, and extended lifetimes, all of which facilitate efficient charge transfer at donor–acceptor interfaces and improve power conversion efficiency (PCE) and fill factor (FF)^39–48. Recent studies demonstrate that exciton delocalization can be enhanced through both processing and aggregation engineering. Hao and co-workers showed that a sonication-induced J-aggregation strategy strengthens intermolecular coupling in L8-BO, extending the exciton diffusion length from ~30 to ~37 nm and boosting device efficiency from 17.5% to 19.4%⁴⁸. Likewise, strain engineering in crystalline phthalocyanine films softens low-frequency vibrations, stabilizes delocalized exciton states, and prolongs exciton lifetimes⁴⁰. At the solid-state packing level, Ginger and co-workers demonstrated that promoting face-on π–π stacking in Y6-type acceptors stabilizes delocalized excitons with enhanced charge-transfer character, which not only suppresses monomolecular recombination but also enables more efficient exciton-mediated hole transfer from the acceptor to the donor, thereby boosting the internal quantum efficiency from ~84% to ~97%⁴³. Even donor polymers such as D18 exhibit similar behavior, where radiative delocalized excitons correlate with higher photoluminescence quantum yield and improved charge collection⁴⁵. These studies collectively show that the way donor and acceptor molecules arrange in the solid state plays a decisive role in shaping exciton delocalization and photovoltaic behavior.

Notably, steric hindrance has emerged as a viable structural lever for optimizing Y6 derivatives, because steric effects introduced on the backbone and/or side chains can twist the three-dimensional packing of molecular units and allow adjacent chromophores to slide into alternative stacking registries, thereby reshaping orbital overlap and electronic coupling⁴. Against this backdrop, achieving precise control over exciton delocalization remains difficult, as it is governed by a subtle interplay among electronic coupling, intermolecular packing, and the relative contributions of LE and CT characters. More importantly, excessive molecular planarity or overly strong core-centered aggregation may improve certain aspects of electronic communication but can also lead to unfavorable packing, unbalanced charge transport, or enhanced recombination. Therefore, an ideal molecular design should not simply maximize backbone planarity, but should regulate the packing geometry in a way that preserves low energy loss while constructing efficient electronic-coupling pathways. While approaches such as backbone planarization, heteroatom incorporation, and halogenation have shown promise, a clear structure–property design principle directly linking molecular structure to exciton delocalization and device efficiency is still lacking.

To translate these excitonic insights into molecular design, tailoring the A₁ electron-deficient core offers a powerful and molecularly precise strategy for regulating electronic coupling, solid-state packing, and non-radiative energy loss in A-DA₁D-A acceptors. We previously demonstrated that quinoxaline and its derivatives, when employed as A₁ units, hold great potential for achieving low energy loss and high efficiency^28,31,34. Moreover, we showed that incorporating sp³-hybridized nitrogen atoms can enhance electroluminescence efficiency and suppress non-radiative recombination⁴⁹. Guided by these insights, we rationally integrated quinoxaline with a nitrogen-containing dibenzo[b,f]azepine (DBA) unit to construct a series of A-DA₁D-A acceptors. This design is expected to inherit the low-energy-loss and luminescence-favorable characteristics of the nitrogen-containing unit, while allowing the packing geometry of the acceptor core to be further regulated through backbone steric effects. By using a fully ring-locked, semi-locked, and unlocked (ring-open) strategy, we intentionally introduced out-of-plane backbone steric hindrance at the donor core, thereby modulating the A₁-core geometry and its associated intermolecular contacts. Using this strategy, we obtained Z1, Z2, and Z3, respectively, and further introduced chlorination on the Z3 backbone to afford Z4 (Fig. 1a). This systematic modulation of the A₁-core geometry provides a precise means to tune intermolecular electronic coupling and exciton delocalization in the resulting acceptors. As the structural constraints are gradually relaxed and chlorination is incorporated, the packing motifs evolve toward more coherent and multidimensional π–π networks, leading to progressively stronger coupling between adjacent chromophores. These changes promote the formation of more spatially extended exciton wavefunctions with reduced monomolecular recombination rates and longer diffusion lengths, as revealed by transient absorption measurements. Importantly, the Z-series acceptors demonstrate that low non-radiative energy loss and efficient charge transport can be reconciled through cooperative molecular design. The DBA unit enables low non-radiative losses below 0.20 eV, while out-of-plane steric hindrance and chlorination in Z4 suppress unfavorable core-centered aggregation and promote more effective intermolecular coupling. Consequently, the optimized Z4-based binary device achieves a PCE of 20.28% with a high V_OC of 0.95 V, together with a certified PCE of 19.99%. Our results reveal a coherent link between molecular structure, electronic coupling, exciton delocalization, non-radiative energy loss, and charge transport across the Z1–Z4 series, underscoring the DBA-modified quinoxaline core as a versatile platform for developing high-performance, low-energy loss non-fullerene acceptors.

Fig. 1. Molecular structures and synthetic routes.

A Molecular design of Z1, Z2, Z3, and Z4; B The synthetic routes of Z1, Z2, Z3, and Z4.

Results

Molecular design of backbone unlocked acceptors

The four SMAs Z1–Z4 were synthesized according to the procedures illustrated in Fig. 1b, the Supplementary Methods, Supplementary Figs. 1–12. The syntheses of Z1 and Z2 followed similar procedures, beginning with the preparation of diketone intermediates 1 and 2. Commercially available dibenz[b,f]azepine first underwent a selective Buchwald–Hartwig amination with 2,6-dichloro-1-iodobenzene (or 1-chloro-2-iodobenzene), followed by a high-temperature Pd-catalyzed direct arylation to achieve cyclization, forming two (or one) five-membered rings. Subsequent oxidation with benzeneseleninic anhydride (BSA) afforded the corresponding diketones 1 and 2. In contrast, the synthesis of Z3 and Z4 employed the reverse order—oxidation followed by Buchwald–Hartwig amination—while the remaining steps were identical to those reported previously. NMR spectroscopy, mass spectrometry, and single-crystal X-ray diffraction analyses were used to verify the chemical structures of four SMAs (Z1 – Z4). All compounds demonstrated excellent solubility in chloroform and chlorobenzene.

Steric hindrance and chlorination regulate intermolecular packing

Figure 2a displays the single-crystal structures of Z1–Z4 (CCDC: 2514642-2514645) together with their key dihedral angles and intramolecular S···O interactions. In all four molecules, the torsion between the thieno[3,2-b]thiophene (TT) bridge and the terminal units is less than 5°, and the central N–C–C–N dihedral angles are small and comparable, indicating a highly planar backbone. The intramolecular S···O contact distances fall within 2.62–2.69 Å. Notably, this unlock strategy significantly affects the dihedral angle between the quinoxaline core and the upper benzene ring of the core, which is 0° for Z1, 31° for Z2, 77° for Z3, and 71° for Z4 (Supplementary Fig. 13), introducing pronounced out-of-plane backbone steric hindrance at the core. This out-of-plane distortion effectively suppresses excessive π–π stacking between the central cores and enforces alternative packing geometries.

Fig. 2. Single-crystal structures and intermolecular packing modes.

A The single-molecule crystallographic structures of Z1, Z2, Z3, and Z4; Intermolecular packing modes, intermolecular dimeric stacking, and calculated electronic coupling of B Z1, C Z2, D Z3, and E Z4.

In terms of crystal-packing behavior, these four acceptors exhibit distinctly different three-dimensional transport networks (Fig. 2b–e). Z1 shows dominant end-group/core-arm (E/C) stacking with strong π-overlap (3.42 Å), together with a secondary end-group/end-group (E/E) motif (3.37 Å), reflecting limited steric constraint. In Z2, the partially twisted core introduces moderate out-of-plane steric effects, leading to reduced core overlap and partial molecular slipping, accompanied by the emergence of end-group/TT-unit (E/B) and core-arm/core-arm (C/C) interactions (3.37 and 3.36 Å). In contrast, Z3 exhibits a markedly slipped packing configuration, in which the out-of-plane backbone steric hindrance group protrudes toward the void between the quinoxaline core and the TT unit of neighboring molecules, completely suppressing the E/C motif while enabling additional end-group-involved stacking. For Z4, chlorination on the steric hindrance group further reinforces this packing behavior: In addition to conventional B/B, E/B, and E/E stacking, the chlorinated Z4 shows the shortest E/E π–π distance of 3.22 Å, together with pronounced Cl-based noncovalent interactions, including Cl···S = 3.55 Å and Cl···H = 3.03 Å. These cooperative out-of-plane steric and noncovalent interactions stabilize molecular orientations and strengthen three-dimensional electronic coupling pathways, thereby facilitating more efficient and anisotropic charge transport.

Enhanced electronic coupling promotes exciton delocalization

Furthermore, theoretical calculations of the electronic coupling constants |J| for the four acceptors under different stacking motifs reveal a general increase from Z1 to Z4²⁸. In Z1, both packing modes give similarly small |J| values (≈4 meV), suggesting that its end-group-dominated arrangements fail to establish effective transport pathways. In Z2, additional E/B and C/C modes emerge; however, their coupling strengths differ drastically (44.8 meV vs. 0.03 meV), so the overall transport dimensionality remains poorly connected. For Z3, both the E/B′ and E/E modes exhibit markedly enhanced couplings (≈ 40 meV), suggesting better-defined three-dimensional transport channels. Z4 finally achieves the strongest and most multidimensional couplings, with |J| as high as 79 meV in the E/E motif, which is expected to promote exciton delocalization.

Backbone unlocking and chlorination contribute differently to the packing evolution of the Z-series acceptors. From Z1 to Z3, the increasing dihedral angle between the quinoxaline core and the upper benzene ring indicates that backbone unlocking mainly introduces out-of-plane steric hindrance, which suppresses unfavorable core-centered stacking and promotes slipped, end-group-involved packing. In contrast, the improvement from Z3 to Z4 is not primarily due to further steric distortion, as Z4 has a comparable dihedral angle to Z3. Instead, chlorination reinforces the unlocked packing geometry through Cl-mediated noncovalent interactions, stabilizing molecular orientation and enhancing multidimensional electronic coupling. Thus, Z4 benefits from the cooperative effect of backbone-unlocking-induced steric regulation and chlorine-assisted packing stabilization.

Such progressive strengthening of intermolecular electronic coupling should manifest directly in the thin-film molecular packing^50,51. This trend is indeed corroborated by the grazing-incidence wide-angle X-ray scattering (GIWAXS) results, which provide quantitative insights into both the in-plane (IP) lamellar organization and the out-of-plane (OOP) π–π stacking characteristics of Z1–Z4 films (Fig. 3a, b). In the IP direction, all four acceptors exhibit pronounced lamellar peaks at q ≈ 0.32–0.37 Å⁻¹. From Z1/Z2 to Z3 and particularly Z4, the lamellar peak gradually shifts to higher q (from 0.33 to 0.37 Å⁻¹), corresponding to a contracted lamellar spacing (from 19.65 to 16.97 Å). Meanwhile, the coherence length (Lc) increases significantly (from ≈ 70 Å to 94 Å). In the OOP direction, the π–π stacking peaks appear at q ≈ 1.65–1.73 Å⁻¹ with d-spacings of 3.80–3.63 Å. Compared with Z1 and Z2, Z3 and Z4 show both reduced π–π distances and enhanced coherence lengths (from ≈ 22–26 Å to 35–38 Å). This reflects a clear enhancement in vertical π–π stacking order, which promotes stronger intermolecular interactions and better electron transport (Supplementary Fig. 14).

Fig. 3. Molecular packing and photophysical properties.

A 2D GIWAXS patterns, B in-plane and out-of-plane line-cut profiles from 2D GIWAXS data of neat Z1-Z4 films; C Normalized TA decay kinetics of neat Z1–Z4 films probed at 830 nm; Normalized absorption spectra of Z1-Z4 in D dilute CF solution and E solid state.

To quantitatively assess the recombination behavior of delocalized excitons in the four acceptor films (Fig. S14), we describe the carrier-density decay using the standard rate equation,

$$\frac{-dn}{dt}=an+\frac{1}{2}{bn}^2,$$ 1

where $a$ (s⁻¹) and b (cm³ s⁻¹) represent the monomolecular and bimolecular recombination rate respectively, and n denotes the exciton population as a function of time t⁵². The corresponding time-dependent solution is obtained through numerical integration of this expression:

$$n\left(t\right)=\frac{n_0{e}^{-at}}{1+\frac{b}{2a}{n}_0\left[1-e^{-at}\right]},$$ 2

which enables simultaneous extraction of both $a$ and b from the transient absorption kinetics⁴³.

Using Eqs. (1) and (2), we reveal distinct recombination signatures for Z1-Z4 that correlate strongly with their exciton delocalization characteristics from transient absorption spectra (Fig. 3c, Supplementary Fig. 15 and Supplementary Table 1). The monomolecular coefficient $a$ reflects the first-order recombination of individual excitonic or CT-like states. A smaller generally corresponds to slower monomolecular recombination and a more persistent excited-state population, which is more favorable for exciton delocalization and dissociation⁴³. Since the fitted $a$ values of Z1–Z4 are close to each other (within 1-3 × 10⁹s⁻¹), we avoid overinterpreting their subtle differences. Instead, we directly compared the normalized TA decay kinetics probed at 830 nm. As shown in Fig. 3c, Z4 shows the slowest decay, corresponding to the longest effective decay time and the smallest effective monomolecular recombination rate. In contrast, Z1 decays more rapidly, indicating faster excited-state recombination and a stronger tendency toward localized exciton recombination. This behavior is unfavorable for exciton delocalization and subsequent exciton/charge dissociation. Together with the enhanced intermolecular electronic coupling resolved from structural analysis, these results support more favorable exciton/charge dynamics in Z4 and less favorable exciton-delocalization behavior in Z1.

Furthermore, the exciton diffusion lengths of neat Z1–Z4 films were further extracted from pump-fluence-dependent TA decay measurements (Supplementary Figs. 15 and 16). The decay kinetics were fitted by considering both monomolecular decay and bimolecular exciton–exciton annihilation (EEA) processes, and the exciton diffusion length was calculated according to $L_D=\sqrt{D\mathrm{\tau }}$, where D is the exciton diffusion coefficient and τ is the exciton lifetime^52,53. To evaluate the reliability of the fitting, multiple pump energies were used for each acceptor film. As summarized in Table S1, the extracted L_D values show good reproducibility, giving average values of 29.2 ± 0.6 nm, 32.5 ± 0.8 nm, 37.9 ± 0.2 nm, and 40.2 ± 0.6 nm for Z1, Z2, Z3, and Z4, respectively. The small standard deviations support the reliability of the fitting analysis. The progressively increased L_D from Z1 to Z4 indicates more efficient exciton diffusion in Z4, consistent with its slower TA decay, enhanced intermolecular electronic coupling, and improved photovoltaic performance. These parameters provide complementary evidence for exciton delocalization. Stronger electronic coupling, reduced monomolecular recombination, and increased exciton diffusion length from 29.2 to 40.2 nm collectively indicate more effective intermolecular wavefunction overlap and longer-range exciton migration from Z1 to Z4⁵⁴.

We systematically investigated the optical properties of Z1–Z4 in solution and solid films, and evaluated their electrochemical characteristics from thin-film cyclic voltammetry measurements. In solution (Fig. 3d and Table 1), all four acceptors display nearly identical absorption profiles with λ_max^sol clustered at 741–745 nm, indicating that peripheral substitutions exert only a minor influence on the intrinsic electronic structure of the conjugated backbone. In the solid state (Fig. 3e), all four acceptors display markedly redshifted absorption features (783–809 nm) accompanied by clearer vibronic structures, indicative of strengthened π–π interactions and increased molecular order. Among them, Z3 and Z4 exhibit the largest shifts, consistent with their tighter intermolecular packing. To further evaluate the light-harvesting capability, the absorption coefficients of the neat acceptor films were calculated from non-normalized spectra with comparable film thicknesses (Supplementary Fig. 17). The absorption coefficients of Z1, Z2, Z3, and Z4 are 0.96, 0.98, 1.02, and 1.05 × 10⁵cm⁻¹, respectively, indicating that backbone unlocking and chlorination do not reduce the intrinsic film absorption intensity. The slightly increased absorption coefficient from Z1 to Z4 is consistent with the improved molecular ordering and stronger intermolecular interactions in Z3- and Z4-based films. Although Z2–Z4 possess different backbone conformations and packing motifs, their neat-film absorption spectra show nearly identical peak positions. To rationalize this observation, we calculated the frontier orbital and electron–hole distributions of representative Z2–Z4 dimers. As shown in Supplementary Figs. 18-20, Z2–Z4 exhibit similar orbital and electron–hole distributions, with the HOMO mainly located on the dibenzo[b,f]azepine unit and the LUMO primarily localized on the terminal acceptor units. These results indicate that their primary optical-transition characteristics are largely preserved, thereby explaining the similar thin-film absorption profiles.

Table 1.

Optical and electrochemical parameters of Z1, Z2, Z3 and Z4

Acceptor	λmaxsol (nm)	λmaxfilm (nm)	λonsetfilm (nm)	Egopta (eV)	LUMOb (eV)	HOMOb (eV)	EgCV (eV)	LUMOc (eV)	HOMOc (eV)	Electron coupling (meV)
Z1	741	783	881	1.41	−3.91	−5.63	1.72	−3.69	−5.71	4.3/4.3
Z2	744	801	889	1.39	−3.94	−5.71	1.77	−3.66	−5.68	3.4/44.8 /0.03
Z3	745	809	892	1.39	−3.99	−5.64	1.65	−3.67	−5.70	3.1/42.1/38.5
Z4	742	806	888	1.40	−3.98	−5.69	1.69	−3.68	−5.73	20.3/39.2/79.0/0.8

^aCalculated from E_g^opt = 1240/λ_onset; ^bCalculated from CV curves; ^cObtained from DFT calculations.

Cyclic voltammetry measurements on thin films (Supplementary Fig. 21) show well-defined redox processes for all acceptors, yielding LUMO levels of −3.91 to −3.98 eV and HOMO levels of −5.63 to −5.71 eV. The close similarity of these energy levels indicates that neither the ring-opening modification nor the peripheral chlorination perturbs the intrinsic electronic structure significantly, while still allowing subtle tuning of the intermolecular interactions.

Theoretical calculations show a gradual increase in molecular dipole moment from Z1 (1.48 D) to Z4 (3.48 D), indicating enhanced intramolecular charge polarization across the series. The HOMO/LUMO distributions reveal that Z1 features relatively localized orbital densities near the central core (Supplementary Fig. 22), while Z2–Z4 display progressively more extended distributions along the π-conjugated backbone. Although the degree of delocalization varies, the overall trend suggests strengthened intramolecular electronic coupling from Z1 to Z4. Meanwhile, the HOMO ( − 5.68 to −5.73 eV) and LUMO ( − 3.66 to −3.69 eV) energies remain essentially unchanged, indicating that the substitutions do not substantially perturb the overall frontier-orbital energetics. In contrast, the frontier-orbital distributions exhibit noticeable differences in spatial delocalization, suggesting that these structural modifications mainly regulate the wavefunction distribution and intermolecular electronic coupling rather than simply shifting the frontier-orbital energy levels.

Device physics and energy loss analysis

To assess how ring-opening and chlorination modifications influence the photovoltaic behavior of the four acceptors, we fabricated devices using the conventional architecture ITO/2-PACZ/D18:Acceptor/PNDIT-F3N/Ag. The complete fabrication procedures are provided in the Supplementary Information. Figure 4a displays the J–V characteristics of the corresponding champion devices, and the key photovoltaic parameters are summarized in Table 2.

Fig. 4. Photovoltaic performance and device physics of D18:Z1–D18:Z4 devices.

a J–V curves, b EQE curves, c Charge extraction time constants extracted from TPC measurements under different light intensities, d Charge-carrier lifetimes extracted from TPV measurements under different bias light intensities, e Impedance spectra, and f Light intensity dependence of J_SC for D18:Z1, D18:Z2, D18:Z3, and D18:Z4 devices.

Table 2.

Device data of D18:acceptors-based devices. Values in parentheses are averages based on at least 10 independent devices

Devices	VOC (V)	JSC (mA cm−2)	FF (%)	PCE (%)
Z1	0.971 (0.962 ± 0.009)	23.45 (23.02 ± 0.33)/22.01a	66.27 (65.71 ± 0.56)	15.09 (14.54 ± 0.55)
Z2	0.962 (0.951 ± 0.011)	23.48 (23.12 ± 0.36)/22.45a	69.15 (68.73 ± 0.42)	15.62 (15.11 ± 0.51)
Z3	0.947 (0.933 ± 0.014)	25.52 (25.40 ± 0.12)/24.40a	72.58 (72.32 ± 0.26)	17.53 (17.14 ± 0.23)
Z4	0.950 (0.945 ± 0.005)	27.02 (26.75 ± 0.23)/25.82a	79.01 (78.52 ± 0.52)	20.28 (19.86 ± 0.24)
Z4b	0.937	27.10	78.76	19.99

^aThe calculated J_SC from EQE curves; ^bThe certified device data for the D18:Z4 device.

Z1 and Z2 deliver the open-circuit voltage (V_OC) values of 0.971 and 0.962 V, respectively, consistent with their similar electronic structures. As the backbone becomes more flexible (Z3) or chlorinated (Z4), V_OC slightly decreases to ~ 0.95 V. In contrast, the short-circuit current density (J_SC) shows a monotonic increase from Z1 to Z4, rising from 23.45 to 27.02 mA cm⁻², in agreement with the progressively enhanced external quantum efficiency (EQE) responses, particularly in the 500–800 nm region (Fig. 4b). The FF increases substantially across the series, rising from 66.27% for Z1 to nearly 80% for Z4, which reflects more efficient charge transport and suppressed recombination in the Z3- and Z4-based devices. As a result, the D18:Z4 blend delivers a PCE of 20.28%, notably higher than the 15.09% achieved with D18:Z1. To verify the reliability of the photovoltaic performance, the optimized device was independently certified by the South China National Center of Metrology-Guangdong Institute of Metrology, giving a certified PCE of 19.99% (Supplementary Fig. 23). Furthermore, we summarized representative binary OSCs with PCEs over 19% and V_OC values over 0.90 V in Supplementary Table 2. This comparison highlights that achieving both a PCE exceeding 20% and a V_OC of approximately 0.95 V remains uncommon in binary OSCs, underscoring the effectiveness of the present molecular design strategy.

The thermal and photo-stability of the Z-series devices were further evaluated and compared with that of the previously reported pure quinoxaline-based D18:AQx-8⁵³ device (Supplementary Fig. 24). For thermal aging, the encapsulated devices were stored in a dark oven at 65 °C under N₂, and their J–V characteristics were periodically measured under AM 1.5 G illumination. After 240 h of thermal aging, the D18:Z4 device retained approximately 84% of its initial PCE, outperforming D18:Z1, D18:Z2, D18:Z3, and D18:AQx-8. For photo-stability, the devices were tested under continuous 100 mW cm⁻² LED illumination in ambient atmosphere. The D18:Z4 device retained about 79% of its initial PCE after approximately 330 h, higher than those of D18:Z1, D18:Z2, D18:Z3, and D18:AQx-8. These results indicate that the out-of-plane steric-hindrance/chlorination strategy not only improves exciton delocalization and charge transport but also contributes to enhanced device stability.

To further examine the generality of this molecular-design strategy, we fabricated devices by replacing D18 with another representative donor polymer, PM6. As summarized in Supplementary Table 3 and Supplementary Fig. 25, the PM6:Z1–PM6:Z4 devices show a performance trend consistent with that of the D18-based devices, with PM6:Z4 again delivering the highest efficiency among the four acceptors. Although the absolute photovoltaic parameters differ between the two donor systems, the consistent performance evolution from Z1 to Z4 supports the broader applicability of out-of-plane backbone steric modulation in acceptor design.

To evaluate the charge extraction and recombination behaviors under different operating conditions, intensity-dependent transient photocurrent (TPC) and transient photovoltage (TPV) measurements were carried out for the D18:Z1–D18:Z4 devices. As shown in Fig. 4c and Supplementary Fig. 26, the TPC extraction time slightly increases with increasing light intensity for all devices. Among them, D18:Z4 exhibits the shortest extraction time across the whole investigated intensity range, suggesting more efficient charge extraction and collection. In comparison, D18:Z1 shows the longest extraction time, consistent with its relatively inferior charge-transport properties. The TPV measurements were performed under open-circuit conditions with different steady-state bias light intensities. The photovoltage decay traces were fitted with a single-exponential function to extract the charge-carrier lifetime. The carrier lifetime decreases with increasing light intensity for all devices (Fig. 4d and Supplementary Fig. 27), reflecting enhanced recombination at higher carrier densities. Notably, the D18:Z4 device shows the longest carrier lifetime over most of the measured intensity range, especially under low-to-medium bias illumination, indicating suppressed recombination. Therefore, the intensity-dependent TPC and TPV results further confirm that the superior photovoltaic performance of D18:Z4 is associated with faster charge extraction and reduced recombination loss.

Electrochemical impedance spectroscopy (EIS) experiments were performed to investigate the charge transport and recombination resistance characteristics in the D18:Z1–Z4 devices. As shown in Fig. 4e and Supplementary Table 4, the series resistance (R_s) decreased from 26.94 Ω in D18:Z1 to 15.43 Ω in D18:Z4, indicating a reduction in ohmic losses in the latter blend. The interfacial capacitances (C₁ and C₂) steadily increased from Z1 to Z4, indicating a gradual increase in the density of accumulated/moving charge at the active layer interface. Simultaneously, the recombination-related series resistances (R₁ and R₂) decreased significantly (from 138.3/3204 Ω in D18:Z1 to 74.25/1249 Ω in D18:Z4), indicating a gradual enhancement in charge transfer and a reduction in charge accumulation during operation. These impedance trends are consistent with the more efficient carrier extraction and lower resistive losses in Z3 and Z4-based devices. To gain further insight into the recombination pathways, we analyzed how V_OC responds to variations in light intensity (P_light). As expected, V_OC scales linearly with ln P_light, reflecting the recombination-governed nature of the photovoltage. Analysis of the V_OC–light intensity slopes in Fig. 4f reveals only minor variations among the four devices, with slope (n) values spanning 1.10–1.15 kT/q. These values are only slightly higher than 1kT/q and far below 2kT/q, suggesting that trap-assisted recombination is present but not dominant in these devices. The D18:Z4 device shows a slightly lower slope of 1.118 kT/q, implying a modestly reduced trap-assisted recombination contribution, consistent with its improved charge extraction and longer carrier lifetime. To further examine whether trap-state density contributes to the FF enhancement, deep-level transient spectroscopy (DLTS) measurements were performed for the D18:Z1–D18:Z4 devices (Supplementary Fig. 28). The extracted trap-state densities are 2.067 × 10¹⁵, 1.904 × 10¹⁵, 1.857 × 10¹⁵, and 1.758 × 10¹⁵cm⁻³ for D18:Z1, D18:Z2, D18:Z3, and D18:Z4, respectively. The gradual decrease in trap density from Z1 to Z4 indicates that the Z4-based blend contains fewer electrically active trap states, which is consistent with the reduced trap-assisted recombination and improved carrier lifetime.

To examine the energetic characteristics of Z-series SMAs, we evaluated the energy-loss components of the Z1–Z4 devices (Supplementary Fig. 29 and Supplementary Table 5). The non-radiative energy loss (ΔE₃) was determined from the electroluminescence quantum efficiency according to ΔE₃ = kTln(EQE_EL)^28,31,34. Although Z4 shows the highest PCE among the Z-series acceptors, its ΔE₃ is relatively larger than those of Z1–Z3, indicating that the improved efficiency of Z4 does not originate from the lowest non-radiative voltage loss.). Although Z4 shows the highest PCE among the Z-series acceptors, its ΔE₃ is relatively larger than those of Z1–Z3, indicating that the improved efficiency of Z4 does not originate from the lowest non-radiative voltage loss. To understand this behavior, we further calculated the reorganization energies associated with the excited-state and charged-state transitions. As summarized in Supplementary Table 6, the λ values from S₁ → S₀ of Z1–Z4 are very close, ranging from 0.125 to 0.126 eV, suggesting comparable excited-state structural relaxation. In contrast, the charge-state-related reorganization energies show a clearer variation: λ_S0→anion gradually increases from 0.064 eV for Z1 to 0.068 eV for Z4, while λ_anion→S0 increases from 0.061 eV to 0.064 eV. These results indicate that Z4 undergoes slightly stronger geometry relaxation and electron–vibration coupling during charge-state formation and recombination, which provides a reasonable explanation for its relatively larger ΔE₃ among the Z-series devices²¹.

Nevertheless, these values remain lower than those of L8-BO, which shows λ_S0→anion and λ_anion→S0 values of 0.071 and 0.066 eV, respectively. This suggests that Z4 has weaker charge-state-related geometry relaxation and electron–vibration coupling than benchmark L8-BO acceptor, thereby helping suppress non-radiative recombination loss (Supplementary Fig. 30). Consistently, the Z4-based device exhibits a higher V_OC than the L8-BO-based device (Supplementary Fig. 31 and Supplementary Table 7), while maintaining comparable J_SC and FF, leading to a higher PCE. Therefore, compared with L8-BO, the performance advantage of Z4 mainly arises from reduced voltage loss associated with lower non-radiative recombination, together with efficient charge transport and extraction. Future molecular design should further reduce ΔE₃ by suppressing charge-state reorganization and electron–vibration coupling while preserving the favorable packing features of Z4.

Probing the blend film morphology and charge generation dynamics

To probe the morphological evolution during film formation, we performed in-situ absorption measurements for the D18-based blends with Z1, Z2, Z3 and Z4 (Fig. 5a, b). Across all blends, the D18 absorption remains essentially unchanged, whereas the acceptor 0–0 transition shifts with time and serves as a sensitive indicator of the crystallization process. In the Z1 blend, this transition undergoes a slow redshift and reaches a steady position at around 630 ms, followed by a more gradual relaxation over roughly another 490 ms, consistent with relatively modest crystallization kinetics. The Z2 and Z3 blends reorganize more quickly: both show a pronounced crystallization event at approximately 910 ms and 840 ms, respectively, and the corresponding peaks settle within about 210 ms. The Z4 blend evolves the fastest; its main crystallization step is completed by around 770 ms, and the subsequent relaxation finishes within about 140 ms. This accelerated evolution aligns well with the tighter and more coherent packing observed for Z4 in the solid state.

Fig. 5. Morphological evolution and charge transport in D18:Z1–D18:Z4 blends.

A 2D contour maps, B Position of 0–0 peaks of the in-situ UV–vis absorption for donor and acceptor, C 2D GIWAXS patterns, D in-plane and out-of-plane line-cut profiles from 2D GIWAXS data of D18:Z1, D18:Z2, D18:Z3, and D18:Z4 blend films; E J-V characteristics in dark for hole-only and electron-only devices based on for D18:Z1, D18:Z2, D18:Z3, and D18:Z4 blends.

Consistent with the in-situ absorption results, GIWAXS of the D18:Z1–Z4 blend films reveals that the overall packing motif is preserved while the crystalline coherence is progressively enhanced along the series (Fig. 5c, d). In the IP direction, all blends show a similar lamellar peak at q ≈ 0.29–0.30 Å⁻¹ (d ≈ 21 Å), indicating that the long-period spacing of the donor–acceptor stacks is essentially unchanged from Z1 to Z4. However, the IP peaks gradually sharpen, with the full width at half maximum (FWHM) decreasing from 0.08 Å⁻¹ (Z1) to 0.06 Å⁻¹ (Z3/Z4), leading to an increase of the corresponding coherence length Lc from ~71 Å to ~94 Å. A more pronounced evolution is observed for the OOP π–π stacking peaks at q ≈ 1.67–1.70 Å⁻¹ (d ≈ 3.69–3.76 Å): while the π–π distance changes only slightly, the FWHM continuously narrows from 0.29 Å⁻¹ (Z1) to 0.11 Å⁻¹ (Z4), and Lc increases from ~19 Å to ~51 Å. These results indicate that, although the average packing distances remain nearly constant, the blend films based on Z3 and especially Z4 develop markedly longer-range ordered π–π stacks and higher crystallinity, in line with their faster crystallization kinetics and improved exciton delocalization. In line with the enhanced crystallinity revealed by GIWAXS, the blend films also exhibit steadily improved charge transport (Fig. 5e and Supplementary Table 8). From Z1 to Z4, hole (μ_h) mobility increases from 1.30 × 10⁻⁴ to 1.96 × 10⁻⁴cm²V⁻¹ s⁻¹, and electron (μ_e) mobility rises from 1.50 × 10⁻⁴ to 10⁻⁴cm²V⁻¹ s⁻¹, confirming that the more ordered D18:Z4 blend support more efficient carrier transport.

Photo-induced force microscopy (PiFM) was employed to visualize the nanoscale morphology of the D18:Z1–Z4 blend films, by selectively probing D18 at 1045 cm⁻¹ and the Z-series acceptors at 1289 cm⁻¹ (Fig. 6a–c and Supplementary Figs. 32–33). The fibril-diameter histograms show that the average width of the acceptor-rich fibers increases progressively from 13.5 nm (D18:Z1) to 15.2 nm (D18:Z2), 17.6 nm (D18:Z3), and 20.7 nm (D18:Z4). At the same time, the fibril distributions for the Z3- and Z4-based blends become more centered and well-defined, indicating more continuous and interconnected acceptor networks. This evolution toward thicker and better-organized fibers is consistent with the enhanced crystallinity and longer π–π coherence lengths revealed by GIWAXS, and provides a straightforward structural basis for the improved charge mobilities observed in the Z3 and Z4 devices.

Fig. 6. Nanoscale morphology and excited state dynamics of D18:Z1–D18:Z4 blend films.

A PiFM images at the wavenumber of 1289 cm⁻¹ correspond to the ZX acceptor; B The combined PiFM images, C The ZX fiber width, D fs-TA spectra at different time delays in D18:ZX blends using 780 nm pump laser.

To gain further insight into the excited-state dynamics of the D18:Z1–Z4 blends, we analyzed their femtosecond transient absorption (fs-TA) responses (Fig. 6d and Supplementary Figs. 34 and 35). Upon selective excitation of the acceptor at 780 nm, all blends show an immediate ground-state bleach in the Z-series absorption region together with the appearance of D18 bleach at shorter wavelengths, evidencing rapid exciton formation in the acceptor domains followed by efficient hole transfer to the donor. The temporal evolution of the bleach signals was fitted with a bi-exponential function, in which the fast component (τ₁) is associated with interfacial exciton dissociation and charge transfer, while the slower component (τ₂) reflects subsequent exciton/charge migration and decay within the bulk.

For D18:Z1, relatively sluggish dynamics are obtained (τ₁ = 4.33 ± 0.30 ps, τ₂ = 62.34 ± 3.30 ps), indicating slower exciton dissociation and longer-lived populations in the less ordered blend. Introducing Z2–Z4 markedly accelerates both processes: τ₁ decreases to 1.51 ± 0.19 ps, 1.80 ± 0.18 ps and 1.57 ± 0.21 ps for D18:Z2, D18:Z3 and D18:Z4, respectively, while τ₂ is shortened from 25.29 ± 1.38 ps to 18.12 ± 0.53 ps and 14.34 ± 0.57 ps. The Z3 and Z4 blends thus exhibit the most favorable kinetics, with more efficient interfacial charge generation and substantially faster population evolution, which is consistent with their more coherent molecular packing, enhanced exciton delocalization, and superior photovoltaic performance discussed above.

To further examine whether the observed TA dynamics are affected by the selected excitation wavelength, we additionally performed excitation-wavelength-dependent TA measurements using 750 nm and 820 nm pump excitation. As shown in Supplementary Figs. 36–38, similar kinetic trends are observed under both excitation wavelengths, and effective hole-transfer processes can still be identified in the D18:Z1–D18:Z4 blend films. In particular, the Z4-based blend consistently shows faster hole-transfer dynamics than the other blends, indicating that the observed trend is not caused by a specific pump wavelength. These additional results confirm the robustness of the TA analysis and further support the conclusion that Z4 facilitates more efficient hole transfer in the blend film. Additionally, pump-fluence-dependent TA measurements were performed on the neat acceptor films under 780 nm excitation to determine the appropriate excitation condition. The decay kinetics remain nearly unchanged below 0.56 μJ cm⁻² (Supplementary Fig. 16), indicating that the measurements are within the linear response regime and are not significantly affected by exciton–exciton annihilation or other high-fluence-induced nonlinear effects. Therefore, all TA measurements were conducted at pump fluences close to or below 0.56 μJ cm⁻². TA measurements were performed on donor-rich blend films with a D:A ratio of 3:1 to exclude the dominant influence of blend morphology. The increased donor content dilutes the acceptor-rich domains and shortens the exciton diffusion distance to the donor/acceptor interface. The slower hole-transfer component becomes faster for all blends (Supplementary Figs. 39 and 40), consistent with reduced morphology-limited exciton diffusion. Nevertheless, the Z1-based blend still shows relatively less efficient hole-transfer dynamics, supporting that its unfavorable recombination behavior mainly originates from weaker exciton delocalization and intermolecular coupling rather than from the original blend morphology alone.

Discussion

We constructed a quinoxaline-based acceptor series (Z1–Z4) by systematically varying backbone conformation and introducing chlorine in a controlled manner. This progression—from fully ring-locked to semi-locked and ultimately unlocked frameworks—introduces out-of-plane backbone steric hindrance on the backbone, providing a practical means to regulate electronic coupling and the extent of exciton delocalization. The nitrogen-containing dibenzo[b,f]azepine unit enables low non-radiative energy losses below 0.20 eV for all Z-series devices, whereas the out-of-plane steric hindrance and chlorination in Z4 suppress unfavorable core-centered aggregation and promote more effective intermolecular coupling. The associated changes in backbone planarity and conjugation continuity are reflected in both single-crystal and GIWAXS analyses, which show a steady improvement in π–π stacking and an increase in electronic coupling, reaching up to 79 meV for Z4. Transient absorption spectroscopy and PiFM measurements further reveal that moving from Z1 to Z4 leads to a general reduction in monomolecular recombination rate, accompanied by longer exciton diffusion distances. The Z4-based blends also display larger and more continuous acceptor-rich fibrillar domains. This molecular design therefore provides a practical strategy to overcome the trade-off between high luminescence efficiency and efficient charge transport, leading to a high PCE of 20.28% with a V_OC of 0.95 V. Overall, this work demonstrates that combining a DPA unit with out-of-plane steric-hindrance-regulated packing is an effective strategy to reconcile low non-radiative energy loss with efficient charge transport in high-performance organic solar cells.

Methods

Device fabrication and testing

The OSCs with the structure of ITO/2PACZ/D18:Acceptors/PNDIT-F3N/Ag were fabricated. The indium tin oxide (ITO) glass substrates were sequentially cleaned with acetone, detergent, deionized water, and ethanol. Then, the ITO substrate was treated with UV ozone for 15 min. The 2PACZ solution, prepared at a concentration of 0.3 mg/mL in methanol, was spin-coated onto an ITO substrate and subsequently annealed at 100 °C for 5 minutes. For D18:Acceptor solution, the total concentration was 11 mg/ml (weight ratio, w/w, 1:1.2), dissolved in chloroform at 50°C. Herein, 4 mg/ml 1,3,5-trichlorobenzene was added into active layer solution as an additive. These active layers were spin-coated in 2PACZ layer with 3000 rpm, then, the annealing was carried out for 5 min at 90°C. For above devices, a methanol with 0.5% vol acetic acid blend solution of PNDIT-F3N at a concentration of 0.5 mg mL⁻¹ was spin-coated onto the active layer at 2000 rpm for 30 s. Finally, around 100 nm of Ag was evaporated under 1×10⁻⁴Pa through a shadow mask. Then, the encapsulation was carried out.

The current density-voltage (J-V) curves of all encapsulated devices were measured using a Keithley 2400 Source Meter under AM 1.5 G (100 mW cm⁻²) using an Enlitech solar simulator. The light intensity was calibrated using a standard Si diode with a KG5 filter to bring the spectral mismatch to unity. An optical microscope (Olympus BX51) was used to define the device area (4.01 mm²). EQEs were measured using an Enlitech QE-S EQE system equipped with a standard Si diode. Monochromatic light was generated from a Enlitech 300 W lamp source. FTPS-EQE was measured using an integrated system (PECT-600, Enlitech), where the photocurrent was amplified and modulated by a lock-in instrument. EL-EQE measurements were performed by applying external voltage/current sources through the devices (REPS-Pro, Enlitech). All of the devices were prepared for EL-EQE measurements according to the optimal device fabrication conditions. EL-EQE measurements were carried out from 0 to 1.8 V.

General information

All solvents and reagents were used as received from commercial sources and used without further purification unless otherwise specified. All reactions were heated by a metal sand bath (WATTCAS, LAB-500, https://www.wattcas.com). ¹H NMR (400 MHz or 500 MHz) and ¹³C NMR (100 MHz or 125 MHz) spectra were measured on a MERCURYVX300 spectrometers. Mass spectra were recorded on a Shimadzu spectrometer. High-resolution mass spectrometry was carried out on Thermo Scientific TM Q-Exactive.

Electrochemical characterizations

Cyclic voltammetry measurements were carried out on a CHI voltammetric analyzer at room temperature. Tetrabutylammonium hexafluorophosphate (n-Bu₄NPF₆, 0.1 M) was used as the supporting electrolyte. The conventional three-electrode configuration consists of a platinum working electrode with a 2 mm diameter, a platinum wire counter electrode, and an Ag/AgCl wire reference electrode. Cyclic voltammograms were obtained at a scan rate of 100 mV/s.

Transient photovoltage (TPV) and transient photocurrent (TPC) measurements

TPV and TPC measurements were performed using a Paios setup from FLUXiM AG. A high-power white light-emitting diode (LED) with an operating voltage of 12 V was used as the light source, corresponding to a total irradiance of approximately 475 Wm⁻² (defined as 100% output intensity). For TPV measurements, the device was kept under open-circuit conditions. A series of steady-state bias light intensities were applied to vary the open-circuit voltage (V_OC). A small light perturbation (pulse intensity ranging from 10% to 100% of the bias light intensity) was then superimposed, and the subsequent photovoltage decay was recorded to directly monitor the recombination of non-geminate charge carriers. The photovoltage decay kinetics followed a single-exponential function: δV=Aexp(−t/τ), where A is the initial photovoltage perturbation amplitude, t is the time, and τ is the charge carrier lifetime. For TPC measurements, the device was held under short-circuit conditions, and the light intensity of the same white LED was varied from 20% to 100% of its maximum output.

Deep level transient spectroscopy (DLTS) measurement

DLTS was performed using a current-based simulation method. The transient current response was analyzed by applying a negative pressure −3 V to the device in the dark. The transient current reached its maximum response at t = 0.1 s regardless of the bias voltage, with the peak gradually rising at 0.1 s as the bias voltage increased. In addition to the displacement current generated by the circuit, there is also the current emitted by the discrete energy trap. The following equation was used to estimate the defect distributions in organic semiconductors. Considering the dependence of carrier transport position in the device, the current peak in the first 0.1 μs is mainly caused by the displacement current, so the fitting of 0.1 ~ 10 μs at room temperature dominated.

Mobility measurements

The electron mobility and hole mobility were evaluated using the space charge current limiting (SCLC) method. The device architecture for the electron-only device was ITO/2PACZ/D18:acceptors/PNDIT-F3N/Ag, and the device architecture for the hole-only device was ITO/PEDOT:PSS/D18:acceptors/MoO_x/Ag. The charge carrier mobility was determined by fitting the dark current to a model of single-carrier SCLC according to the following equation: J = 9ε₀ε_rμV²/8d³, where J is the current density, d is the film thickness of the active layer (~100 nm), μ is the charge carrier mobility, ε_r is the relative dielectric constant of the transport medium, and ε₀ is the permittivity of free space. V used in the equation is defined as follows: V = V_app –V_bi, where V_app is the applied voltage, V_bi is the built-in voltage. The carrier mobility is calculated from the slope of the J ~ V² curve.

PiFM measurements

Photo-induced force microscopy (PiFM) results were acquired using a VistaScope microscope from Molecular Vista, inc. All PiFM experiments were excited by a pulsed quantum cascade laser (Block Engineering) with a gap-free narrowband tunable wavenumber of 760-1950 cm⁻¹. The spectral linewidth is ~2 cm⁻¹ with a wavenumber resolution of 0.5 cm⁻¹. The PiFM experiment here was operated at the sideband excitation with the laser-frequency modulated at f_m = f₁-f₀, where f₀ is the first mechanical eigenmode resonance of the cantilever that is used for PiF signal detection, while f₁ denotes the second ones recorded for the AFM topography of the sample. The probe is an Pt-coated tip with a resonant frequency of ~350 kHz (PPP-NCHPt-MB, Nanosensors).

Transient absorption spectroscopy

Transient absorption spectroscopy (TAS) was recorded with Agilent Technologies Cary 60 UV−visible spectrophotometer and ultrafast TA spectrometer (Harpia-TA, Light Conversion), respectively. A Yb:KGW laser (1030 nm, 54 kHz, Light Conversion) is split into two fundamental light beams in ultrafast transient measurement. One of the light beams is transferred to the optical parametric amplifier to generate a high-intense specific wavelength for the excitation (pump) beam. At the same time, the other is focused on 5 mm sapphire to generate a low intense continuum light ranging from 600–1020 nm, employed as a probe beam. A delay stage monitors the time delays between the pump and probe light beams. The pump and probe rays are spatially overlapped at the donor-acceptor networks, and the transmitted probe light is collected using a charge-coupled device. An angle of 54.7°, called the magic angle, is set between the polarization of the pump and probe light beams for all TA experiments. The TA experiment is done five times on several spots of the film, and on measurement completion, no sample degradation is detected. To improve the reliability of the TA analysis, the kinetic traces under 780 nm excitation were fitted using a biexponential function: ΔA(t)= A₁ exp(-t/τ₁) + A₂ exp(-t/τ₂) + y₀, where A₁ and A₂ are the amplitudes of the fast and slow components, respectively, τ₁ and τ₂ are the corresponding time constants, and y₀ is the baseline offset.

Reporting summary

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

Supplementary information

Supplementary Information ^{(4.9MB, pdf)}

Reporting Summary ^{(1.8MB, pdf)}

Transparent Peer Review file ^{(3.6MB, pdf)}

Source data

Source Data ^{(253.3KB, xlsx)}

Author contributions

Z.L. and J.Z. conceived the idea and wrote the manuscript. J.Z. synthesized the Z-series materials. K.S. and Y.W. characterized the small area devices under the supervision of G.Z. J.Z. performed the GIWAXS characterizations and analysis. Z.C. conducted theoretical calculations and analysis. G.R. conducted the TA experiments under the supervision of W.Z., K.S., and L.Y. conducted the in-situ absorption spectroscopy characterization and analysis. The certified device was characterized by L.Z. under the supervision of F.L. The manuscript was approved by all authors. G.R., C.Y., and Z.L. provided important insights during the project design and supervision.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC, Nos. 22309119 and 22475133), the Guangdong Basic and Applied Basic Research Foundation (2026B1515020049), the Shenzhen Science and Technology Program (ZDSYS20210623091813040), and the Shenzhen University 2035 Program for Excellent Research (Grant No. 2024C007).

Data availability

The experiment data generated in this study are provided in the Supplementary Information/Source Data file. Crystallographic data for the structures reported in this study have been deposited at the Cambridge Crystallographic Data Centre (CCDC) under deposition numbers CCDC 2514642 (Z4), 2514643 (Z2), 2514644 (Z1), and 2514645 (Z3). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures/. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76217-9.