Gating hole back-transfer with molecular spacers programs rapid micropollutant decontamination

Abstract

Precise control of oxygen activation to generate reactive oxygen species (ROS) for on-demand micropollutant removal remains challenging. Herein, we develop a bottom-up molecular approach to systematically tailor dithienyl spacers (alkene/alkyl/benzene), from model fragments, amorphous polymers to high crystalline covalent-organic frameworks (COFs). We observe that the ordered assembly of benzene-bridged bithiophene units facilitates long-range electron transport, thereby suppressing the spatial recombination of photogenerated electron–hole pairs. These effects synergically reduce the oxygen adsorption barrier, efficiently generating superoxide radicals as the dominant ROS, with subsequent singlet-oxygen formation. Therefore, the removal of acetaminophen in water via the optimized Tapt-BDD COF catalyst achieves a reaction rate constant of 0.80 min⁻¹. The degradation rates are enhanced by 63–95 fold over fragments (local repeating unit structure of the COF) and 5–8 fold over polymers (disordered COFs). Additionally, the proposed system demonstrates efficient retention in complex wastewater treatment scenarios, and under continuous operation. This work establishes a strategy of spacer-directed supermolecule preorganization for environmental catalysis.

This work demonstrates that molecularly designed COF spacers enhance charge transport and oxygen activation, enabling efficient reactive oxygen species generation and rapid micropollutant removal from water.

Introduction

Oxygen activation is regarded as a key step in the generation of reactive oxygen species (ROS) in photosynthesis and energy–environment chemistry¹. Given the pivotal role of ROS in disinfection and micropollutant removal, solar-driven oxygen activation aligned with global carbon-neutrality goals has emerged as an imperative in sustainable environmental chemistry. Accordingly, solar-driven catalysis, which directly harnesses renewable solar energy with high energy-conversion efficiency, is widely regarded as a promising strategy for organic micropollutant removal^2,3. Nevertheless, most existing systems rely on semiconductor photocatalysis and remain fundamentally limited by short charge-carrier diffusion lengths and rapid energy dissipation at charge-transfer interfaces^4–7. A prevailing yet often unscrutinized assumption in photocatalyst design is that maximizing structural rigidity through C≡C linkages inherently guarantees optimal performance. This rigidity-centric paradigm, however, overlooks the subtle interplay between molecular conformation, electronic delocalization, and interfacial charge dynamics.

Early advances in the controlled motion of molecular machines and the modular assembly of “molecular LEGO” systems established the importance of precise spatial control at the nanoscale^8–11 (Fig. 1a). Transcending these paradigms requires moving beyond dynamic motion or structural modularity toward predictable macroscopic function. Here, we dismantle the rigidity–performance dogma by integrating molecular spacer engineering with hierarchical assembly (Fig. 1b). In recent years, studies on charge transport modulation in COFs have demonstrated that precise engineering of molecular structures and stacking configurations enables the synergistic optimization of electrical and photophysical properties across multiple length scales, spanning from the molecular level to the extended framework architecture^12,13. Zhu et al. introduced conjugated enyne chains into the interlayer of COFs to provide additional charge-carrier pathways through direct π-electron delocalization, resulting in a conductivity of 2.8 × 10⁻⁴ S cm⁻¹ for TAPFY-COF, nearly one order of magnitude higher than that of the non-conjugated counterparts¹⁴. Zhou et al. incorporated pyrene (PY, electron donor) and nickel bipyridine (BIP, acceptor) units to decouple interlayer electron coupling and boost photogenerated electron transfer toward the single-Ni catalytic center¹⁵. In PY-CN-BIP, the C^δ+ center of the imine bond adjacent to the PY lowers the HOMO energy level, while the N^δ− site near the BIP elevates the LUMO energy level. Müllen et al. found that saturated linkers (e.g., cyclohexane) in 2D COFs interrupt the intralayer conjugation and interlayer interactions, resulting in the suppression of photoinduced charge transfer along with photoluminescence enhancement through the aggregation-induced emission effect of tetraphenylethene units¹⁶. Meanwhile, both the carbonyl groups in the basal plane¹⁷ and proton-mediated topological interlayer shift¹⁸ are demonstrated to efficiently regulate intralayer charge transport, exciton binding energy, and interlayer exciton dissociation.

Fig. 1. Molecular-to-macroscopic design evolution.

a The importance of precise spatial control at the nanoscale^8–11 (molecular shuttle, molecular motor, chelation-induced self-assembly, non-covalent molecular LEGO). b The strategy of molecular spacing engineering combined with hierarchical assembly. Model compounds, polymers, and COFs were constructed from bithiophene–triazine units with different spacer groups. c Experimental, refined, and simulated PXRD patterns of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF.

However, the complex periodic network introduces multiple variables, such as interlayer stacking and pore heterogeneity¹⁹ complicating the precise isolation of individual structural factors influencing charge transport. In contrast, low-dimensional models such as well-defined molecular fragments and polymers feature simpler charge transport pathways, effectively decoupling macroscopic aggregation effects and enabling systematic evaluation of how different spacer units modulate electronic behavior. Building on this foundation, three types of molecular fragments with distinct electronic structures and conformational characteristics were rationally designed and synthesized, then polymerized and ultimately assembled into highly ordered, periodic COFs. This approach facilitates a hierarchical structural evolution from molecular “dots” to polymeric “lines” and finally to extended framework “planes,” establishing a coherent progression across multiple length scales. The key factors governing the enhancement of rigid π-conjugated frameworks in COFs are identified by strategically regulating the spatial sequence of donor-acceptor (D-A) segments.

Inspired by the coupling bridges involved in the energy cascade of natural photosynthesis, we designed a series of spacer units ranging from flexible alkyl chains and partially conjugated alkene linkers to rigid aromatic rings to systematically regulate the degree of π conjugation, structural stability, and the effective length of electronic transport pathways within the donor segments (Fig. 1c). Compared with rigid alkene and flexible alkyl units, benzene spacers in organic molecules exhibit favorable performance associated with preorganized π-conjugation. The strategy can maximize π-delocalization, lowering exciton binding energies and amplifying surface charge density versus flexible analogs. This restructuring establishes long-range charge highways, boosting oxygen activation efficiency via single-electron transfer. Meanwhile, spacer chemistry in the donor unit regulates the degree of hole back transfer for the transformation of superoxide radicals (·O₂^-) and subsequent singlet oxygen (¹O₂), culminating in ultrafast micropollutant destruction. The optimized catalyst (Tapt-BDD COF) confers universal resilience across broad acidity/alkalinity, high salinity, and complex matrices. Validated in continuous-flow reactors, this work highlights the potential importance of spacer-directed molecular preorganization over rigidity in the design of next-generation environmental catalysts.

Results

Spacer-engineered molecular evolution

To elucidate how spacer units govern charge transport, we implemented a rational bottom-up design spanning dithienyl–triazine molecular fragments (Fig. 2a–c), linear polymers (Supplementary Figs. 1–3), and crystalline COFs. Systematic photoelectronic characterization via transient photocurrent response and electrochemical impedance spectroscopy (Fig. 2d, e and Supplementary Figs. 4-5) demonstrated that benzene spacers enhance photocurrent density and reduce charge-transfer resistance compared with alkene or flexible alkyl linkers in both dithienyl–triazine molecular fragments and macromolecule polymers.

Fig. 2. Comprehensive Characterization of fragments, polymers, and COFs.

a–c Solid state ¹³C CP/MAS NMR spectra of Tapt-TD fragment, Tapt-BD fragment, and Tapt-BDD fragment. d, e Transient photocurrents spectra of molecular fragments and polymers. f The natural transition orbital distributions of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF, the white, tan, blue, and yellow spheres refer to hydrogen, carbon, nitrogen, and sulfur, respectively. (NTO: Natural transition orbitals; %: contribution rate). g PXRD pattern of three COFs. h FTIR patterns of three COFs. i N₂ adsorption-desorption isotherms of Tapt-BDD COF (The inserted image is the pore distribution of Tapt-BDD COF). j, HRTEM images of Tapt-BDD COF. k The Kubelka−Munk-transformed reflectance spectra of three COFs.

Electronic structure analysis of dithienyl–triazine fragments established thiophene units as electron donors and the electron-deficient Tapt moieties as acceptors (Supplementary Fig. 6). Crucially, benzene-spaced Tapt-BDD fragments showed an upward shift in sulfur orbital energy of benzodithiophene moieties near the Fermi level (Supplementary Fig. 7). This shift arises from enhanced π-conjugation and stronger orbital coupling between the benzene spacer and adjacent thiophene units, which increases the delocalization of donor-centered orbitals and thereby strengthens their electron-donating capability upon photoexcitation. As a direct consequence, spacer-induced electronic modulation facilitates more efficient donor–acceptor charge transfer. Spatial charge distribution mapping under visible light irradiation (Supplementary Fig. 8) demonstrates that benzene spacers govern carrier separation. Electron accumulation (green) and depletion (blue) regions exhibit maximal spatial displacement in Tapt-BDD unit. Analysis of the singly excited electronic states reveals a reduction in electron density on the benzodithiophene units and an increase in the triazine moieties in Tapt-BDD fragments, confirming a distinct long-range electron transfer from benzodithiophene to triazine units along imine linkages (Fig. 2f). Quantitative metrics via the t-index and D-index (Supplementary Fig. 8) analyze this molecular design advantage via space chemistry. Compared with Tapt-TD fragment (t-index = −0.721 Å, D-index = 3.093 Å) and Tapt-BD fragment (t-index = −1.993 Å, D-index = 2.202 Å), Tapt-BDD fragment exhibits markedly greater charge separation distance and centroid displacement (t-index = 1.717 Å, D-index = 5.177 Å). These findings substantiate that benzene spacer engineering constructs extended charge-transfer channels, enabling efficient exciton dissociation and carrier migration.

Subsequently, these fragments were polymerized to yield macromolecular polymers. As shown in Fig. 2e, the photocurrent density of these polymers is higher than that of corresponding molecular fragments, which is ascribed to the extended conjugation and intermolecular connectivity for enhancing exciton diffusion and charge transport. Building upon this foundation of molecular design hierarchy, we engineered spacer-optimized COFs where molecular preorganization dictates macroscopic order.

Molecular preorganized assembly validation

The periodic ordered structure of these building units was verified by Powder X-ray diffraction (PXRD) patterns. Three COFs share the isostructural hcb topologies with A-A stacking (Supplementary Tables 1–3). Specially, all of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF exhibit high crystallinity with corresponding (100) plane angles of 2.54°, 2.33°, and 2.21°, respectively (Fig. 2g). The full width at half maximum (FWHM) values of the (100) crystalline planes for Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF were calculated to be 0.49°, 0.55°, and 0.31°, corresponding to the microcrystalline sizes of 16.2 nm, 14.5 nm, and 25.6 nm, respectively. According to the Scherrer equation^20,21, Tapt-BDD COF exhibits sharp reflections and narrow FWHM, indicative of maximal crystallite size, a highly ordered domain and the highest π-conjugation degree along the x-axis. Additionally, the broad diffraction peaks observed near 25° for the three COFs correspond to the (001) plane due to the π–π stacking of 2D layers (Supplementary Fig. 9). Among them, Tapt-BDD COF displays the smallest interlayer spacing, suggesting longitudinally superior π-conjugation along the y-axis. The degree of π-conjugation is evidenced by bond length alternation (BLA), defined as the average difference in bond lengths between adjacent carbon–carbon bonds along the conjugated backbone²². A lower BLA value typically reflects a greater contribution from quinoidal resonance structure²³. Notably, the BLA value of 0.0277 for Tapt-BDD COF is substantially lower than those of Tapt-TD COF (0.0302) and Tapt-BD COF (0.0424), indicating a more delocalized and strongly conjugated framework. These enhanced structural features, including larger microcrystalline domains, tighter π–π stacking, and greater conjugation, are expected to facilitate efficient charge transport and exciton dissociation, thereby contributing directly to the superior photocatalytic performance of Tapt-BDD COF.

Spectroscopic analyses verified framework integrity. Fourier-transform infrared (FT-IR) spectroscopy identified C=N stretching vibration at 1575 cm⁻¹ (Fig. 2h). High-resolution X-ray photoelectron spectroscopy (XPS) resolved characteristic C 1s bonding states (Supplementary Figs. 10 and 11), attributing to C=C, C=N, and C-N at 284.8 eV, 286.8 eV, and 288.6 eV, respectively. The chemical structure of three COFs with typical imine linkages at ~154 ppm was also confirmed by ¹³C solid-state nuclear magnetic resonance (NMR) spectroscopy (Supplementary Fig. 12). Scanning electron microscopy (SEM) revealed that both Tapt-BDD COF and Tapt-TD COF exhibit roughened particulate surface, whereas Tapt-BD COF displays comparatively smoother morphology (Supplementary Fig. 13). Nitrogen adsorption-desorption measurements conducted at 77 K revealed that Tapt-BDD COF, Tapt-BD COF and Tapt-TD COF possess Brunauer-Emmett-Teller (BET) surface area of 1281, 31 and 191 m² g⁻¹ (Fig. 2i and Supplementary Fig. 14), respectively. This dramatic enhancement stems from spacer-suppressed interlayer disorder²⁴, corroborated by High-resolution transmission electron microscopy (HRTEM) images. An apparent lattice spacing of 2.74, 3.44, and 2.71 nm was observed for Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF, corresponding to the (100) plane (Fig. 2j and Supplementary Fig. 15). Moreover, all COFs maintained structural integrity to 500 °C (Supplementary Fig. 16). Overall, the spacer not only modulates the electronic structure of the framework but also leads to variations in crystallite size, stacking mode, and porosity. These features act in concert, providing a basis for the subsequent photocatalytic processes.

Optoelectronic characterization revealed spacer-modulated band architectures. Both optical properties and redox potential of three COFs are determined by UV-visible diffused reflectance spectra (UV-vis DRS) and Mott-Schottky equation. The absorption edge (Fig. 2k) of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF were closed to 594, 597, and 541 nm, respectively. The corresponding band gaps were calculated to be 2.23, 2.16, and 2.37 eV, respectively, from Tauc plots estimation (Supplementary Fig. 17). Mott-Schottky analysis (Supplementary Fig. 18) revealed flat-band potentials of −1.15, −1.06, and −0.13 eV relative to the normal hydrogen electrode (NHE), with the corresponding conduction band edges calculated at −0.6 V, −0.51 V, and −0.48 V, respectively. This optimized alignment (Supplementary Fig. 19) creates substantial thermodynamic permissible for O₂ activation in photo-induced micropollutants elimination from wastewater.

Ultrafast micropollutant removal via spacer-directed active species modulation

To validate our spacer-directed molecular strategy, we selected acetaminophen (ACE), a persistent pharmaceutical pollutant with high water solubility and low octanol-water partition coefficient (K_ow)—as a model contaminant, commonly found in pharmaceutical manufacturing, hospital effluents, landfill leachate, rural sewage systems, and surface water. Implementing a bottom-up design hierarchy across three structural tiers with various spacers, explicit structure-property relationships were established. As shown in Supplementary Figs. 20 and 21, all photocatalysts exhibit effective ACE removal under visible light irradiation governed by synergistic interplay of hierarchy and spacer chemistry. Dual-scaling laws were demonstrated: i) structural complexity amplifies performance. Model compounds underperform polymers, which are surpassed by COFs, with the latter exhibiting degradation rates enhanced by ~63–95-fold over fragments and 5–8-fold over polymers; ii) spacer chemistry in hierarchical assemblies dictates efficacy with the following order of benzene spacer > alkene spacer > flexible alkyl linker. This cascading enhancement confirms that catalytic potency is molecularly encoded in spacer-engineered electronic frameworks, with polymerization and crystalline assembly progressively magnifying their photochemical potential.

Typically, only 10%, 13%, and 6% of ACE was adsorbed by Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF under dark condition, respectively (Supplementary Fig. 22), which is consistent with weak physisorption energies (Supplementary Fig. 23). Upon visible light irradiation, the removal efficiency of ACE reached to 83%, 65%, and 96% within 5 minutes for Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF, respectively (Fig. 3a and Supplementary Fig. 24). The superior catalytic performance corresponds directly to the stronger and sharper PXRD reflections and the enhanced imine C=N stretching band observed in the FTIR spectrum. The overall degradation trend is broadly consistent with the UV–Vis diffuse reflectance spectra, indicating that light absorption is the fundamental prerequisite for driving the catalytic reaction (Supplementary Fig. 25). The first-order rate constant of 0.80 min^-1 for Tapt-BDD COF exceeded Tapt-BD COF by 4.2-fold (0.19 min^-1) and Tapt-TD COF by 2.4-fold (0.33 min^-1), setting a global benchmark (Fig. 3b, Supplementary Fig. 25 and Supplementary Table 4). Although enhanced hydrophilicity of Tapt-BDD COF (Fig. 3c) facilitated superior aqueous dispersion and pollutant-catalyst contact for optimizing mass transfer^25,26, this performance of the hierarchy directly reflects molecularly engineered electronic structures controlled by spacer genome of catalyst.

Fig. 3. Photocatalytic activity and mechanism of COFs.

a Removal of ACE by Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF. Conditions: ACE solution (60 mL, 5 mg L^-1), catalyst (12 mg), 300 W Xe lamp, λ  >  420 nm. Multiple contemporary measurements from the same experimental unit. Error bars depict the standard deviations (n = 2). The data are presented as average ± SD. b Comparison of Tapt-BDD COF and other reported advanced catalysts for photocatalytic degradation of ACE (Supplementary Table 4). c Contact angles of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF. d–f Contribution of active species to ACE removal in Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF (The relative contribution was estimated using indirect scavenger experiments; the presence of other ROS cannot be entirely excluded). g Yield of superoxide radicals. h Yield of singlet oxygen. i O₂-TPD curves of three COFs. j Reaction energy barriers of the ACE removal pathway. The white, gray, blue, and red spheres refer to hydrogen, carbon, nitrogen, and oxygen, respectively.

The initiator of ACE elimination in the presence of COFs was investigated by electron spin resonance (ESR) spectroscopy and scavenger experiments (Supplementary Figs. 27–30). In the presence of isopropanol (IPA, ·OH scavenger), the ACE removal by Tapt-BDD COF was almost unaffected (Supplementary Figs. 27 and 28). A slight decline trend was observed upon the introduction of triethanolamine and furfuryl alcohol, suggesting a suboptimal role of photogenerated holes (h⁺) and singlet oxygen (¹O₂). Notably, ACE removal was nearly halted when p-benzoquinone (·O₂^- quencher) was added, indicating the decisive role of ·O₂^- induced by molecular O₂ activation. The presence of h⁺, ¹O₂ and ·O₂^- in both Tapt-TD COF and Tapt-BDD COF was also verified by ESR signals (Supplementary Fig. S29). Once depletion of dissolved oxygen via argon (Ar) purging, a marked decline of ACE removal was observed, substantiating O₂ activation as essential (Supplementary Fig. 30). These demonstrated that the superoxide with 100% selectivity is generated predominantly through electron transfer between adsorbed O₂ and catalyst. The contribution of ROS in Tapt-BDD COF reached 80.4%, which is 1.5- and 1.1-fold higher than those in Tapt-TD COF (53.3%) and Tapt-BD COF (72.3%), respectively (Fig. 3d–f). Correspondingly, the overall oxygen activation of Tapt-BDD COF significantly exceeds both Tapt-TD COF and Tapt-BD COF. Relative to Tapt-BD and Tapt-TD COFs, Tapt-BDD COF generated ·O₂^- at concentrations approximately 1.5 and 1.8 times greater, respectively (Fig. 3g). Using 1,3-diphenylisobenzofuran as a probe, Tapt-BDD COF produced ¹O₂ with a concentration of 37.8 μmol L^-1, showing increases of about 1.4- and 1.3-fold over Tapt-TD and Tapt-BD COFs, respectively (Fig. 3h). The yield of ¹O₂ increased concomitantly with the elevated production of ·O₂^- and the enhanced consumption of photogenerated holes, indicating that ¹O₂ formation proceeds primarily through the reaction between ·O₂^- and holes—a pathway governed predominantly by hole availability. Although Tapt-BD COF generated much more bulk ROS, this apparent discrepancy with its lower degradation activity compared to Tapt-TD COF is accounted for by differences in the effective ROS availability (I_effect) at the catalytic surface (Supplementary Fig. 31, Supplementary Note. 1). Integrating excellent oxidative activation and strong interaction of ACE and Tapt-BDD COF, more apparent removal is exhibited.

Theory calculations demonstrated the strongest O₂ adsorption onto triazine region (acceptor unit) of Tapt-BDD COF (-0.093 eV), followed by Tapt-TD COF (-0.084 eV) and Tapt-BD COF (0.308 eV) (Supplementary Fig. 32). This adsorption affinity trend was corroborated by oxygen programmed temperature rise desorption (O₂-TPD) test (Fig. 3i). Upon O₂ adsorption, all COFs exhibit an increase in the summed Mulliken charge of the carbon atoms in triazine ring (Supplementary Fig. 33), indicative of electron depletion at adsorption sites. This redistribution of electron density reflects a net electron transfer from the framework to the adsorbed O₂ for activation. Triazine carbon segments in Tapt-BDD COF underwent minimal charge increase upon O₂ adsorption, indicating efficient electron transfer to adsorbed O₂—a direct consequence of benzene spacer-enhanced electronic delocalization. These findings underscore the pivotal role of O₂ activation in removing ACE with low energy consumption.

The spacer-enhanced reactive species execute precision molecular cleavage of ACE through regioselective attacks. Electrostatic potential mapping (Supplementary Fig. 34) identifies electron-rich carbonyl regions vulnerable to nucleophilic ·O₂^- and electron-deficient amine sites susceptible to electrophilic oxidants (¹O₂/h⁺). Frontier orbital analysis (Supplementary Fig. 35) confirms the benzene ring region located at HOMO facilitates electrophilic oxidation while the carbonyl fragments at LUMO favors nucleophilic cleavage—enabling spacer-engineered radicals to selectively target molecular vulnerabilities. High-resolution liquid chromatography–mass spectrometry (LC–MS) and Gibbs free energy analysis validated multiple degradation pathways. It is revealed that, under complex oxidative environment, ACE molecules undergo sequential degradation via coupling, diazotization, hydrolysis, and oxidative ring-opening reactions, leading to the formation of various transformation products (Fig. 3j, Supplementary Figs. 36–37, Supplementary Tables 6–8 and Supplementary Note 2). Finally, Tapt-BDD COF achieves 35% mineralization within 10 minutes, confirming deep oxidation capacity despite transient dimer accumulation via radical coupling of single-electron ACE²⁷.

Computational toxicity profiling shows progressive detoxification of intermediates versus parent ACE (Supplementary Fig. 38), following thorough deep oxidative treatment. Wheat germination assays reveal that the germination was unaffected (Supplementary Fig. 39a), but seedling growth was initially inhibited. Chlorophyll content increased by approximately 45% (Supplementary Fig. 39b), likely due to beneficial nutrients for intermediates. Within the first 5-min reaction, root and shoot lengths decreased by approximately 32% and 15%, respectively, indicating transient growth inhibition (Supplementary Fig. 39c). After 7-min reaction, a synergy of recovery and enhancement was shown, because deep oxidation products with reduced toxicity promote wheat growth, with root and shoot lengths increasing to 30.4 cm and 8.7 cm. This biphasic response confirms the ultimate biocompatibility of degradation by-products, fulfilling the environmental safety imperative of molecular design. Simultaneously, a life cycle assessment (LCA) was conducted, which demonstrates that the environmental burden of the COFs system is predominantly dictated by electricity and solvent consumption during synthesis, whereas the contribution from the framework building units is negligible (Supplementary Figs. 40–42, Table 5, Note 3).

Engineering Spatiotemporally Synergistic Charge Dynamics in COFs via Molecular Spacers

Now that all-possible pathways with corresponding products during the ACE elimination have been identified, the question of how the chemical reaction mechanisms are created in spacer-controlled-COFs catalytic system remains. Temperature-dependent Photoluminescence (PL) spectra (Fig. 4a–c) reveal Tapt-BDD COF achieves the lowest exciton binding energy (35.1 meV versus 48.8 meV and 59.6 meV), confirming benzene spacers maximize exciton dissociation efficiency. Logically, upon undergoing charge transfer in three COFs^28,29, both intrinsic electric fields and surface potentials form, as evidenced by in situ atomic force microscopy–Kelvin probe force microscopy (AFM–KPFM) measurements. The surface potential of Tapt-BDD COF reaches to 263 mV in dark, markedly higher than that of Tapt-TD COF (55 mV) and Tapt-BD COF (90 mV) (Fig. 4d–f). The electrochemical impedance of COFs follows the order of Tapt-BDD COF <Tapt-TD COF <Tapt-BD COF (Supplementary Fig. 43), confirming that Tapt-BDD COF has electrochemical impedance the minimal resistance and favorable conductivity. The surface charge density of Tapt-BDD COF (Fig. 4g, Supplementary Fig. 44 and Supplementary Note. 4) accumulates 0.17 μC cm^-2, which is 8.3 and 10.4 times higher than that of Tapt-TD COF and Tapt-BD COF, respectively. According to the Kanada model^30,31, the built-in electric field strength of Tapt-BDD COF is enhanced by factors of 8.82 and 7.79 compared to the other two COFs (Fig. 4h). These results collectively verify that the introduction of rigid π-conjugated benzene spacer amplify electrostatic driving forces for directional photogenerated charge migration toward the surface/interface.

Fig. 4. Photophysical properties and charge dynamics of COFs.

a–c Temperature-depended PL spectra (insert) obtained with an excitation wavelength of 350 nm for Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF, along with derived exciton binding energy for COFs. d–f The KPFM image of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF under water-soaked conditions in dark. g Transient photocurrents of three COFs. h Internal electric field intensities. The inserted image is the electrostatic potential of Tapt-BDD COF. The white, grey, blue, and yellow spheres refer to hydrogen, carbon, nitrogen, and sulfur, respectively. (ELF: Electron Localization Function) i Surface potential change of Tapt-BDD COF after visible-light illumination. j–l In-situ FTIR spectra of Tapt-BDD COF under O₂ atmosphere in light and dark. *O₂^- signal and the vibration of the O-O bond. The vibration of the K, C-S-C bond. l The vibration of the triazine and imine bonds.

Upon visible-light excitation, the surface potential of Tapt-BDD COF increases to 396 mV, substantially exceeding those of Tapt-TD (93 mV) and Tapt-BD COF (184 mV) (Fig. 4i, Supplementary Fig. 45). This trend is in agreement with TD-DFT simulations of photoexcited charge transfer. Electron localization function (ELF) analysis (Supplementary Fig. 46) visualizes that the incorporation of benzene spacer in Tapt-BDD COF enhances electron delocalization and facilitates efficient migration of free electrons from thiophene donors to triazine acceptors³². This phenomenon not only improves the spatial distribution of excited-state charges but also establishes molecular highways for long-range charge transfer. In contrast, flexible alkyl-spaced Tapt-BD COF shows localized electron density around heteroatoms, where conformational dynamics (e.g., twisting, rotation) disrupt orbital coupling and impede carrier mobility³³.

In-situ FTIR spectroscopy elucidates the dynamic evolution of active sites during light-driven O₂ activation (Fig. 4j-l). The characteristic O-O stretching vibration at ~893 cm^-1 progressively intensifies under illumination, indicating photoinduced adsorption and stepwise activation of O₂ on the catalyst surface. Concurrently, the band at 1142 cm^-1, assigned to ^*O₂^-, markedly increases with irradiation time, reflecting a continuous single-electron reduction mediated by photoexcited electrons. The enhanced and red-shifted C-S-C vibration at 951 cm^-1 indicates pronounced electronic polarization and structural rearrangement within the thiophene units, providing direct evidence of localized charge response during photoactivation. Notably, the intensified stretching vibration of the imine (C=N) bond at 1519 cm^-1—linking thiophene donor and triazine acceptor—is markedly enhanced. Correlated with the electron–hole distribution in the excited state (Supplementary Fig. 8), this observation indicates that excitons efficiently migrate to the vicinity of the imine linkage and then undergo dissociation within the conjugated framework, generating spatially separated electrons and holes. The holes predominantly localize near the imine bonds. Meanwhile, a subtle shift of the triazine-associated band from 1363 cm^-1 to 1373 cm^-1 can be attributed to interfacial charge transfer between adsorbed O₂ and triazine rings³⁴, confirming the pivotal role of triazine fragments as an active site in oxygen activation.

To verify the long-range charge transport channels enabled by spacer-controlled COFs, femtosecond time-resolved transient absorption (fs-TA) spectroscopy was performed (Fig. 5a–c). As shown in Fig. 5d–f, a positive signal around 560 nm can be assigned to excited-state absorption (ESA). After 27 fs, a negative signal emerges, corresponding to ground-state bleaching (GSB), consistent with the strong absorption of the three COFs in this region (Supplementary Fig. 17). Notably, the GSB peak undergoes a pronounced blue shift (Fig. 5d–f), which is attributed to interfacial charge transfer³⁵. The resultant microscopic electric field induces a marked shift of local transition energies³⁶, providing direct evidence for the occurrence of photoinduced hole back transfer. Importantly, the amplitude of GSB bleaching in Tapt-BDD COF (0.0168) is markedly smaller than that in Tapt-TD (0.0226) and Tapt-BD (0.0215). Given that the transient absorption amplitude scales proportionally with the density of photoexcited carriers³⁷, this reduced decay indicates that, within the same timescale, Tapt-BDD COF facilitates faster and more efficient hole transport, thereby promoting more complete spatial separation of electrons and holes. This experimental observation is fully corroborated by our TD-DFT calculations, which reveal that the energy-level alignment and orbital distributions in Tapt-BDD COF are more conducive to forming efficient hole transport pathways. The convergence of experiment and theory firmly establishes the superiority of Tapt-BDD COF in suppressing charge recombination and enhancing charge transport efficiency.

Fig. 5. Charge separation mechanism and stability of COFs.

a–c 2D mapping fs-TA spectra of Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF. d–f Time-resolved absorption (TA) spectra of Tapt-TD COF (d), Tapt-BD COF (e), and Tapt-BDD COF (f) in the range of timescales. g Corresponding decay kinetics curves of three COFs (Supplementary Table 9). h Schematic of electron transitions (τ: lifetime). (GSB: ground-state bleaching; SE: stimulated emission; ESA: excited-state absorption; ΔA: ΔAbsorbance.) i Oxygen activation mechanism in Tapt-BDD COF. The white, pink, green, and red spheres refer to hydrogen, carbon, nitrogen, and oxygen, respectively. j Removal of four kinds of typical micropollutants. The white, grey, blue, red, and green spheres refer to hydrogen, carbon, nitrogen, oxygen, and chlorine, respectively. (TC: tetracycline; DCP: 2,4-dichlorophenol; 4CP: p-chlorophenol; MOP: m-hydroxyphenol isomers) k Long-term operation for ACE (Acetaminophen) Removal based on a continuously flowing reactor. Multiple contemporary measurements from the same experimental unit. Error bars depict the standard deviations (n = 2). The data are presented as average ± SD.

In addition, a stimulated emission (SE) feature was observed around 620 nm (Supplementary Fig. 47), consistent with the steady-state photoluminescence spectrum (Fig. 4a–c). Tapt-BDD COF also exhibits nearly synchronous rise, plateau, and decay dynamics at 463, 514, and 655 nm. Kinetic fitting at these wavelengths yields a fast rise constant of 3.48 ps, which is attribute to vibrational relaxation of photoexcited electrons and holes toward the LUMO minimum and HOMO maximum, respectively (Fig. 5g, h, Supplementary Fig. 48 and Supplementary Table 9). An additional ESA signal at 514 nm may originate from hole transfer from the electron-accepting Tapt units to the electron-donating BDD units. The kinetic analysis further resolves two characteristic time constants of 0.66–3.48 ps and 350.22–389.18 ps. The faster component (τ₁) corresponds to rapid charge transfer, involving exciton annihilation or dissociation assisted by transfer and trapping processes. Once excitons or polarons dissociate, the resulting charge-separated states display extended lifetimes. The slower component (τ₂) is attributed to carrier recombination (Supplementary Fig. 49). Strikingly, the lifetime of photogenerated carriers (τ₂) in Tapt-BDD COF is significantly prolonged compared with Tapt-TD COF and Tapt-BD COF, highlighting its pronounced advantage in maintaining charge separation and suppressing recombination. This pronounced lifetime extension is due to the fact that the benzene spacer facilitates the spatial electron–hole separation and weakens the electronic coupling, thereby suppressing hole back transfer. Integrating ¹O₂ generation and spatiotemporal charge dynamics, the spacer structure modulates hole back-transfer dynamics, which reduces the non-productive consumption of holes and enables their increased availability for singlet oxygen generation via reaction with ·O₂^-. Notably, wavelength-dependent ACE degradation experiments (Supplementary Fig. 26) further validate this kinetic advantage: although light absorption decreases at longer wavelengths, Tapt-BDD COF maintains high degradation efficiency, indicating that the catalytic performance is more directly governed by carrier lifetimes and interfacial charge transfer efficiency rather than simply by the number of absorbed photons.

The spatiotemporally cooperative optimization of charge dynamics enables synergistic oxygen activation (Fig. 5i). Under photoexcitation, the electron-rich thiophene donors induce exciton migration toward the vicinity of the imine bonds, where dissociation occurs under the influence of the built-in electric field. The polarized nature of the imine bond facilitates this process: the partially positive carbon (δ⁺) attracts the negatively charged electron from the exciton, promoting electron delocalization toward the triazine acceptor, while the partially negative nitrogen (δ⁻) localizes the positive hole near the imine linkage, thereby enhancing spatial charge separation. On the one hand, the long-range electron flux from thiophene to triazine units drives single-electron O₂ reduction at the triazine sites, producing ·O₂^- via the radical pathway. On the other hand, the residual photogenerated holes controlled by benzene spacer engage with the ·O₂^-, driving ¹O₂ generation through a coupled electron-transfer mechanism. This spatially separated dual-pathway on the molecular scale markedly engages in the successive oxidation of micropollutants at the interface.

Molecularly engineered catalyst excels in real environments

Benzene-spaced Tapt-BDD COF demonstrates exceptional environmental adaptability under optimized dosage and varying water substrates (Supplementary Fig. 50). Although exposure to natural sunlight accelerates the fluctuations of water temperature, near-complete ACE removal maintained across a temperature range of 5-45°C, indicating a thermal resilience (Supplementary Fig. 51). As illustrated in Supplementary Figs. 52–53, when the pH value in wastewater changes from 3 to 11 and the ionic strength increases to 50 mM, Tapt-BDD COF also preserved 100% efficiency toward ACE removal, exhibiting strong resilience against variations in acid and alkaline wastewater and the coexisting anions. A co-contaminant synergy was observed. In the presence of humic acid (HA) weakly enhanced the rate of ACE destruction from 1 to 5 mg/L HA (Supplementary Fig. 54), likely resulting from the photochemical promotion³⁸. In terms of broad-spectrum micropollutant applicability (Fig. 5j, Supplementary Table 10), Tapt-BDD COF is capable of achieving over 93% removal efficiency for four representative compounds—tetracycline (TC), 2,4-dichlorophenol (DCP), p-chlorophenol (4CP), and m-hydroxyphenol isomers (MOP) within just 10 minutes. Although high removal efficiencies are highlighted, the reaction kinetic rates vary due to the influence of intrinsic molecular structure. Even after five consecutive cycles, the catalyst sustained a removal efficiency of exceeding 96% within 10 minutes (Supplementary Fig. 55) despite partial framework alteration (Supplementary Figs. 56–57). Importantly, post-reaction LC–MS analysis revealed that no framework-derived species or organic ligands were detected in the treated solution (Supplementary Fig. 58), thereby excluding the possibility of secondary contamination during repeated operation.

Testing across diverse water sources confirmed universal applicability via various scenarios including tap water (TW), lake water (LW), landfill leachate (LL), and pharmaceutical wastewater (PW). Consistent performance of ACE removal in four kinds of water bodies is observed (Supplementary Fig. 59). In contrast, the ACE decomposition rate in pharmaceutical wastewater decreased significantly, attributed to the competitive quenching induced by high organics load. To transform molecularly engineered COFs from batch-mode catalysts into sustainable water remediation platforms, a one-pot and continuous flow membrane interception and retention device was developed (Supplementary Fig. 60 and Supplementary Note 4). This device operated over 80% removal efficiency continuously and stably for 12 h (Fig. 5k), validating practical implementation potential.

Discussion

This work deciphers how spacer chemistry governs long-range charge transport through a bottom-up molecular design paradigm—spanning model fragments, polymers, and crystalline COFs. Within the investigated dithienyl–triazine spacer systems, benzene spacers exhibit the most effective charge delocalization, reducing exciton binding energies by 28 % and 41% and enhancing surface charge density by 16.1-fold and 20.8-fold compared with their alkene and alkyl counterparts. This electronic restructuring suppresses hole back-transfer near the imine linkage fragment and facilitates efficient oxygen activation at the triazine site, leading to ultrafast pollutant destruction with a record reactive rate. The scalability of this approach is inherently dependent on the synthetic feasibility of the spacers, emphasizing the need for efficient, high-yield strategies for both spacer synthesis and COFs assembly. Unique spacer-directed charge highways synchronize the transformation of oxygen activation to ROS pathways: i) single-electron transfer for ·O₂^- formation through sustained electron flux; ii) the interaction of ·O₂^- with holes for ¹O₂ generation through prolonged carrier lifetime. By correlating molecular preorganization with macroscopic performance in continuous-flow reactors, this study establishes a mechanistic framework for understanding charge-transfer regulation within dithienyl–triazine-based porous catalytic systems. However, the synthetic complexity and associated costs of these COFs currently hinder their direct deployment in practical wastewater treatment. Future efforts should therefore prioritize developing scalable, green synthetic routes to enhance cost-effectiveness.

Methods

Materials

4,4’,4”-(1,3,5-triazine-2,4,6-triyl)trianiline (98%), Thieno[3,2-b]thiophene-2,5-dicarbaldehyde (97%), [2,2’-Bithiophene]-5,5’-dicarboxaldehyde (98%), and Benzo[1,2-b:4,5-b’] dithiophene-2,6-dicarboxaldehyde (98%) were supplied by Yanshen Technology Co., Ltd (Jilin Chinese). All other reagents and solvents were bought from Sinopharm Chemical Reagent Co., Ltd.

Synthesis of Tapt-TD COF

4,4’,4”-(1,3,5-triazine-2,4,6-triyl) trianiline (21.3 mg, 0.06 mmol) and Thieno[3,2-b] thiophene-2,5-dicarbaldehyde (17.6 mg, 0.09 mmol) were dissolved in 1.5 mL of o-DCB/n-butanol (V/V = 1:2). The mixture was sonicated for 10 min, followed by the addition of 0.1 mL of 6 M acetic acid and further sonication for 3 min. After three freeze–pump–thaw cycles, the tube was sealed and heated at 120 °C for 3 days. The obtained solid was washed with THF and methanol three times and dried under vacuum at 80 °C for 12 h to afford Tapt-TD COF.

Synthesis of Tapt-BD COF

4,4’,4”-(1,3,5-triazine-2,4,6-triyl) trianiline(21.3 mg, 0.06 mmol) and [2,2’-Bithiophene]-5,5’-dicarboxaldehyde (20 mg, 0.09 mmol) were dissolved in 1 mL of o-DCB/n-butanol (V/V = 3:7). The mixture was sonicated for 10 min, followed by the addition of 0.1 mL of 6 M acetic acid and further sonication for 3 min. After three freeze–pump–thaw cycles, the tube was sealed and heated at 120 °C for 3 days. The resulting solid was washed with THF and methanol three times and dried under vacuum at 80 °C for 12 h to afford Tapt-BD COF.

Synthesis of Tapt-BDD COF

4,4’,4”-(1,3,5-triazine-2,4,6-triyl) trianiline(17.7 mg, 0.05 mmol) and Benzo[1,2-b:4,5-b’] dithiophene-2,6-dicarboxaldehyde (18.5 mg, 0.075 mmol) were dissolved in 1 mL of o-DCB/n-butanol (V/V = 3:7). The mixture was sonicated for 10 min, followed by the addition of 0.1 mL of 6 M acetic acid and further sonication for 3 min. After three freeze–pump–thaw cycles, the tube was sealed and heated at 120 °C for 3 days. The resulting solid was washed with THF and methanol three times and dried under vacuum at 80 °C for 12 h to yield Tapt-BDD COF.

Sample characterization

Zeiss Sigma 300 XRD patterns were obtained by a Bruker D8 advance powder diffractometer using Ni filtered Cu Kα radiation (5°/min). Scanning electron microscopy (SEM, 300 Sigma, Zeiss., Germany) and field emission transmission electron microscopy (FE-TEM, JEOL JEM-2100F, Japan) were utilized to obtain the morphology and microstructure of samples. The chemical compositions and characteristic functional groups of COFs were analyzed using X-ray photoelectron spectroscopy (XPS, Scientific K-Alpha spectrometer, Thermo Fisher Scientific) and Fourier transform infrared (FT-IR, Thermo Scientific Nicolet iS10 spectrometer). The Brunauer-Emmett-Teller (BET) specific surface area was investigated on THS-114 (Micromeritics Instrument Ltd., USA). The UV-visible diffused reflectance spectra (UV-vis DRS) were recorded using PerkinElmer Lambda 750S. Photoluminescence (PL) spectroscopy was undertaken using a photoluminescence Spectrometer (FLS1000, Edinburgh). The static contact angle was recorded by optical contact angle meter (Dataphysics OCA20., Germany). Solid- state ¹³C NMR (ssNMR) spectra were acquired on Bruker AVANCE III spectrometry (400 MHz). Both the electrochemical impedance spectroscopy (EIS) and photocurrent response were measured on the electrochemical workstation (CHI760E, China). The electrochemical workstation was a standard three-electrode system, in which the catalyst-coated indium tin oxide (ITO) was the working electrode, the Pt wire was the counter electrode, and the saturated calomel electrode was the reference electrode. Atomic force microscopy (AFM) and Kelvin probe force microscopy (KPFM) measurements were both completed at SPM-9700 surface probe system (Bruker, MultiMode8, Germany). Before the test, the probe is calibrated with the iron plate, which was the base for the test. The light source for in-situ KPFM is a 420 nm visible light.

Catalytic performance and kinetic measurements

The experimental procedures of photocatalytic degradation were carried out according to the literature³⁹. Time-dependent photocatalytic decomposition of acetaminophen (ACE) on Tapt-TD COF, Tapt-BD COF, and Tapt-BDD COF at a specific initial concentration (200 mg/L). 12mg catalyst and 60mL ACE solutions were evenly dispersed in a 100mL beaker and stirred for 60 min in the dark to reach the adsorption-desorption equilibrium, and then irradiated with a Xenon lamp source (PLS-SXE300, Beijing Perfect Light), and a cutoff filter was used to achieve visible-light irradiation (λ>420nm). Samples were taken at 1, 3, 5, 7, and 10 minutes of the 10-minute reaction time, and solid-liquid separation was performed using a 0.22 μm filter.

First-order kinetic model was utilized as follows:

$$\ln C/C_0=kt$$ 1

where C₀ is the initial ACE concentration.

Theoretical calculation

Theoretical calculations were performed using the Gaussian 16 program (Revision C.01). All geometrical optimizations were conducted at the B3LYP-D3(BJ)/SVP level of theory. Based on these B3LYP-D3(BJ)/SVP optimized geometries, subsequent calculations for single-point energies, frontier molecular orbit (FMO), electrostatic potentials (ESP), and Fukui indices were performed at the B3LYP-D3(BJ)/TZVP level.

The Fukui indices, derived from Natural Population Analysis (NPA) charges and representing electrophilic attack (f ^-) and nucleophilic attack (f ⁺), were employed alongside the ESP to predict the regioselectivity of radical attack on the ACE molecule. The M06-2X/TZVP level of theory was utilized exclusively for the calculation of Gibbs free energies. Electrostatic potential molecular orbitals were analyzed and visualized by Multiwfn⁴⁰ and VMD⁴¹.

To evaluate the adsorption properties of three COFs materials for O₂/ACE, a study was conducted Vienna Ab-initio Simulation Package (VASP)⁴² using the Density Functional Theory (DFT) method⁴³. The van der Waals (vdW) modified DFT-D3⁴⁴ method and the generalized gradient approximation (GGA)⁴⁵ of Perdew-Burke-Ernzerhof (PBE) is used to optimize the structure and calculate the electronic properties. The kinetic energy cut-off energy of the plane wave basis group is set to 400 eV. The energy convergence criterion for the iterative solution of the Kohn-Sham equation is 1×10^-5eV, and the atomic force convergence threshold during the structural optimization process is set at 0.02 eV/ A⁴⁶. The electron occupation state adopts the Gaussian broadening technique, with the broadening width σ = 0.03 eV. The Brillouin zone integral adopts the Monkhorst-Pack scheme, a gamma-centred 3 × 3 × 1 Monkhorst-Pack grid⁴⁷.The adsorption energy of O₂/ACE on COF was calculated with the following equation:

$${\mathrm{E}}_{\mathrm{a}}={\mathrm{E}}_{\left(\mathrm{COF}+{\mathrm{O}}_{\mathrm{2}}/\mathrm{ACE}\right)}-{\mathrm{E}}_{\mathrm{COF}}-{\mathrm{E}}_{{\mathrm{O}}_2/\mathrm{ACE}}$$ 2

Phytotoxicity of ACE degradation byproducts

After 7 days of seed culture, the germination rate was calculated by observing the germination number of wheat seeds (New Ji Mei 22 Triticum aestivum L). The shoot length and root length of wheat seeds were measured by the ImageJ software.

Reporting summary

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

Author contributions

W.Z. and C.Q. performed experiments, analyzed data and writing this paper; W.Z. and Y.W. came up with the original idea, performed photocatalytic experiments, and processed data; H.W. helped with the calculation part; Z.F, A.S., and Z.S. conducted the DFT calculations on Fukui index; K.Y. assisted in the toxicology experiments; S.D. helped with the data interpretations; Y.Y. and X.Y. supervised the project; W.Z., C.Q., H.W., and Y.W. wrote the manuscript; all authors commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

Open access funding provided by Hunan Agricultural University.

Data availability

The data that support the findings of this study are available within the article and its Supplementary Information. 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/s41467-026-76398-3.