Introduction

Hydrogen peroxide (H2O2), an eco-friendly oxidizing agent and energy carrier, is a vital and versatile chemical widely utilized in chemical synthesis and environmental remediation1,2,3. The global H2O2 market is projected to grow at a compound annual growth rate (CAGR) of 4.6%, reaching 5.7 million tons by 20284. Currently, the anthraquinone oxidation process conducted at high temperatures dominates industrial H2O2 production, accounting for over 95% of global output5. An alternative approach involves the direct reaction of H2 and O2 catalyzed by noble metal catalysts6. However, both methods are burdened by high costs, significant energy consumption, and pollutant emissions7,8. Moreover, the highly concentrated H2O2 produced poses considerable safety risks during storage and transportation9, although end-use scenarios often require only dilute solutions (e.g., <0.1 wt% for water treatment)10. Therefore, it is urgent to develop green and sustainable technologies for on-site, on-demand production of ready-to-use H2O2 solutions.

Solar-driven oxygen reduction reaction (ORR) is emerging as a promising technique to produce clean H2O2 solution from O2 and water11,12,13. In the photocatalytic ORR process, H2O2 can be produced via a one-step two-electron (2e–) route (O2 + 2e– + 2H+ → H2O2) or a sequential two-step 2e– one (O2 + e– → O2•− + e– + 2H+ → H2O2)14,15. Since O2 adsorption is the key step of photocatalytic ORR, the affinity and initial configurations of O2 adsorption are crucial for H2O2 generation16. For instance, Yeager-type O2 adsorption benefits efficient and selective H2O2 photosynthesis via the direct 2e– route17. Once O2 is adsorbed, it is transformed into H2O2 via proton-coupled electron transfer (PCET)18. To enhance the efficiency and selectivity of H2O2 production, it is essential to achieve a triple synergy in providing O2, protons, and electrons. Recent studies, such as the one conducted by Han et al., demonstrate that engineered electric fields and hydrophilic surfaces can facilitate the transfer of charges and protons, resulting in enhanced H2O2 generation19. Despite these advancements, simultaneously optimizing all three key factors within a single catalyst remains a challenge, leading to unsatisfactory efficiency in solar-driven O2-to-H2O2 conversion.

Covalent organic frameworks (COFs), as metal-free porous materials with customizable active sites and good visible light absorption, are burgeoning as top candidates for H2O2 photosynthesis20,21,22,23. Since the seminal work reported in 2020 by Thomas and Van Der24, various strategies such as constructing heterojunctions, donor-acceptor (D-A) engineering, and heteroatom doping have been employed to enhance COFs’ performance in H2O2 photosynthesis25,26. Nevertheless, nearly all reported COFs feature neutral skeletons, which weakens the interaction with guest molecules like O2 and hinders synergistic optimization of photocatalytic ORR. Zwitterionic COFs, incorporating equimolar and spatially arranged cationic and anionic groups27, have recently emerged as promising candidates for enhancing charge transfer and guest molecule interaction, offering the potential for a single material to perform multiple functions28,29. Although zwitterionic COFs have shown great promise in gas separation and proton conduction30, their application in H2O2 photosynthesis remains largely unexplored, likely due to the challenges in the design of suitable charge groups.

Here, a squaric acid (SQ)-based zwitterionic COF (STT COF) was designed to demonstrate the triple synergy in the photocatalytic ORR process. The resulting STT COF achieves an H2O2 yield of 14,356.5 μmol g–1 h–1 in pure water, with an apparent quantum yield (AQY) of 40.0% at 420 nm, outperforming its charge-neutral PTT COF counterpart and the state-of-the-art systems ever reported. To address issues related to oxygen supply and catalyst recovery, a continuous flow membrane reactor incorporating STT COF and gas diffusion layers was further developed, enabling sunlight-driven generation of 5 L of H2O2 (400 μM) without external aeration. The produced H2O2 can be directly used for the in-situ degradation of emerging organic pollutants via Fenton reactions, demonstrating great promise for practical applications. Experiments and theoretical calculations jointly elucidate the triple synergy mechanism in the zwitterionic structure derived from SQ units, which promotes photogenerated charge transfer, ensures a continuous supply of protons, and facilitates O2 adsorption in the Yeager-type configuration. These findings underscore how the structural design and variation of zwitterionic COFs synergistically promote the ORR processes, paving a feasible avenue for the practical application of H2O2 photosynthesis.

Results

Synthesis and structural characterization of COFs

Squaric acid (SQ) and 4,4’,4”-(1,3,5-triazine-2,4,6-triyl) trianiline (TAPT) were elaborately selected as the electron donor and acceptor, respectively, to construct the target zwitterionic COF (STT COF), as illustrated in Fig. 1a. For comparative investigation, SQ was substituted by terephthalaldehyde (PDA) to prepare charge-neutral COF (PTT COF) with an identical electron acceptor (TAPT). According to the powder X-ray diffraction (PXRD) patterns (Fig. 1b), the STT COF exhibits diffraction peaks at ca. 3.4°, 6.32°, and 27.1°, corresponding to the (100), (110), and (001) facets, respectively. Similarly, the PTT COF shows corresponding peaks at around 2.87°, 5.1°, and 25.9°, respectively (Supplementary Fig. 1). The experimental PXRD data of these COFs closely match the simulated AA stacking model, with Pawley fitting analysis yielding R factors of less than 5% for both unweighted-profile (Rp) and weighted-profile (Rwp).

Fig. 1: Synthesis and structural characterization of COFs.
Fig. 1: Synthesis and structural characterization of COFs.
Full size image

a Schematic illustration of the synthesis process of PTT COF and STT COF. b The experimental PXRD data and stimulated patterns of STT COF. c 13C ssNMR spectra of PTT COF and STT COF. d XPS spectra of O 1s of PTT COF and STT COF. e TEM image of STT COF. f Ultraviolet–visible absorption spectra from diffuse reflectance of the resultant COFs. g Band structure diagrams for PTT COF and STT COF.

The chemical states of the resultant COFs were then systematically investigated. Solid-state 13C nuclear magnetic resonance (13C ssNMR) spectra (Fig. 1c) display two peaks at ca. 130 and 170 ppm, which can be attributed to the hydrogen-bonded aromatic carbon and the carbon atoms of the triazine units, respectively31. The peak at 152 ppm on PTT COF can be assigned to the carbon atom of the imine (C = N) linkage, while the peak at 118 ppm on STT COF refers to the C-N bond32. Besides, two peaks at 150 and 175 ppm are observable at STT COF, corresponding to the C atoms of SQ units. Fourier transform infrared (FTIR) spectra (Supplementary Fig. 2) show the disappearance of amino group peaks (3352–3432 cm–1) and the emergence of C = N bonds (1595 cm–1) after the Schiff base reaction. For STT COF, the enol peak of SQ (2420 cm–1) vanishes, accompanied by a new signal of C-N-C bond (1388 cm–1), whereas the aldehyde peak of PTT COF (2680–2840 cm–1) disappears. According to the X-ray photoelectron spectroscopy (XPS) survey spectra (Supplementary Fig. 3a), C, N, and O are the primary elements of these two COFs. Both C 1s spectra exhibit two distinctive peaks at 284.7 eV and 286.4 eV, corresponding to the sp2 carbon and C = N bonds, respectively. Besides, peaks at 285.6 and 287.1 eV in the C 1s spectra of STT COF belong to C-N/C-O and C = O bonds, respectively (Supplementary Fig. 3b)33. As for N 1s spectra, two peaks at ca. 398.8 and 400.4 eV can be detected (Supplementary Fig. 3c), which are attributed to the C-N = C and hybridized -C = N-, respectively34. Noticeable differences in their O 1s spectra (Fig. 1d), in which distinct signals for O = C and O-C are observable on STT COF, but only small signals ascribed to the adsorbed O2 on PTT COF. The obtained results clearly confirm the successful fabrication of the targeted COFs.

The microstructural features of these two COFs were characterized by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). As shown in SEM images (Supplementary Fig. 4a–d), both PTT COF and STT COF are composed of irregular and uniform nanorods, respectively. Their morphology was further validated by TEM images (Fig. 1e and Supplementary Fig. 5). According to N2 adsorption-desorption isotherms (Supplementary Fig. 6a, b), the Brunauer–Emmett–Teller (BET) surface areas of PTT COF and STT COF were calculated to be 1017.1 m2 g–1 and 217.9 m2 g–1, respectively, along with the average pore size of ca. 3.0 and 2.5 nm (Supplementary Fig. 6c, d). As shown in Fig. 1f, the ultraviolet–visible (UV–vis) absorption spectra from diffuse reflectance reveal that STT COF exhibits a broader absorption in the visible light range than PTT COF, with their bandgaps of 1.97 eV and 2.66 eV, respectively (Supplementary Fig. 7). Based on the Mott–Schottky plots (Supplementary Fig. 8a, b), the conduction band (CB) edges of PTT COF and STT COF are calculated to be −0.80 and −0.46 V (vs. NHE), respectively (Fig. 1g), which are thermodynamically favorable for ORR (O2 + 2H+ + 2e− → H2O2, ENHE = 0.68 V).

H2O2 photosynthesis and its practical application

The H2O2 photosynthesis performance of these two COFs was evaluated in pure water. As shown in Fig. 2a, under visible light irradiation (Xe light, 420 nm < λ < 780 nm), PTT COF delivers an almost steady H2O2 generation during the 1-h operation with a moderate rate of 1503.4 μmol h−1 g−1, while STT COF exerts a 6.7-fold enhancement (10,004.2 μmol h−1 g−1) under the identical conditions. Despite comparable light adsorption nature at 420 nm, STT COF still shows about sevenfold higher H2O2 yield than PTT COF (Supplementary Fig. 9), signifying that light absorption is not the key factor for differentiating their performance. All the monomers are almost inactive for H2O2 photosynthesis (Supplementary Fig. 10a), highlighting the crucial role of COF structure. As shown in Supplementary Fig. 10b, the slightly increased H2O2 content suggests that H2O2 decomposition is unlikely to account for their different photocatalytic activity. The optimal dosage of STT COF was found to be 0.25 g L−1 (Supplementary Fig. 10c), as both lower and higher concentrations may hinder effective O2 and light utilization35. Furthermore, STT COF exhibits high compatibility with water matrices and can efficiently produce H2O2 from a range of water sources, including tap water, seawater, and Pearl River water (Supplementary Fig. 11). Under UV–vis light irradiation (320 nm < λ  < 780 nm), STT COF exerts a high H2O2 yield of 14356.5 μmol g−1 h−1 with an apparent quantum yield (AQY) of 40.0% at 420 nm (Fig. 2b), which is 7.9 times higher than that of PTT COF (Supplementary Fig. 12) and surpasses other reported systems in pure water (Supplementary Table 1), demonstrating its notable meliority in this arena. Of note, the produced H2O2 can be utilized directly to eliminate contaminants (e.g., methylene blue and bisphenol A) via Fenton reactions (Fig. 2d and Supplementary Fig. 13). Cycling experiments and corresponding characterizations (e.g., PXRD and FTIR analyses, Supplementary Figs. 14 and 15) jointly validate the good stability of STT COF in this system.

Fig. 2: Photocatalytic performance of H2O2 production and practical application of the as-prepared COFs.
Fig. 2: Photocatalytic performance of H2O2 production and practical application of the as-prepared COFs.
Full size image

a Photocatalytic activity of PTT COF and STT COF for H2O2 generation in pure water without any sacrificial agents. Conditions: water (20 mL), catalyst (5 mg), visible light, 420 nm < λ < 780 nm. b Comparison of photocatalytic H2O2 concentration and yield from STT COF under 300 W Xe lamp at different wavenumbers (visible light, 420 nm < λ < 780 nm; UV–vis light, 320 nm < λ < 780 nm). c Comparison of H2O2 production activities of COFs under various conditions over 1 h, including standard, air, dark, N2, N2 + KBrO3 (10 mM), 10% IPA and p-BQ (0.01 mM). d Degradation of methyl blue via Fenton reaction with H2O2 produced by STT COF. e Labelled diagram of the continuous flow membrane reactor. f Experiment on the effectiveness of gas diffusion layers under UV–vis light irradiation. g A photograph of the continuous flow membrane reactor under natural sunlight irradiation. h The performance of a continuous flow reactor under natural sunlight. The error bars in the figures represent the standard deviations from triplicate tests.

A series of experiments was then conducted to investigate the reaction pathways. As demonstrated in Fig. 2c, almost no H2O2 could be detected under dark conditions, in an N2-saturated solution or N2-saturated KBrO3 (electron scavenger) solution, while the addition of isopropanol (IPA, hole scavenger) boosted H2O2 production. These results suggest that H2O2 comes from the photocatalysis-driven ORR process rather than the hole-driven water oxidation reaction (WOR). As revealed in Supplementary Fig. 17, the H2O2 production in both COF systems remains nearly unaffected by p-benzoquinone (0.01–0.05 mM), suggesting a direct one-step 2e⁻ ORR pathway17. Based on the results of rotating ring-disk electrode (RRDE) measurements, STT COF exhibits a lower onset potential, higher H2O2 current, and an electron transfer number closer to 2 than PTT COF (Supplementary Fig. 18), which benefits the 2e− ORR pathway. In the N2-saturated KBrO3 solution, the concentration of dissolved O2 increases as the reaction proceeds (Supplementary Fig. 19), suggesting that the photogenerated holes oxidize H2O to yield O2. As illustrated in electron paramagnetic resonance (EPR) spectra (Supplementary Figs. 19 and 20), no typical signal for radicals (such as O2·− and ·OH) can be detected in these two COF systems, further excluding the generation of radicals, which aligns closely with quenching experiment results. 18O2 isotope experiments were then conducted to confirm the reaction pathways of STT COF. In the STT COF system filled with 18O2, the 16O2 content remained negligible after 1 h (Supplementary Fig. 21a), corroborating the good airtightness of the reaction system. After 6 h of light illumination, the 16O2 level in the STT COF system escalates because of O2 being produced by 4e– WOR. As depicted in Supplementary Fig. 21b, both 18O2 and 16O2 were detected after H2O2 decomposition, with 18O2 as the predominant species (78.7%), further validating the origin of H2O2 from the ORR process.

The practical implementation of H2O2 photosynthesis is now hindered by the quick consumption of the dissolved O2 in water and the challenge of catalyst recovery36. Considering the boosted performance of STT COF, a continuous flow membrane reactor with carbon cloth equipped with gas diffusion layers (GDL) and the sample layer on different side (Fig. 2e) was constructed to further evaluate its potential for practical application. Under UV–vis light irradiation (320 nm < λ < 780 nm) for 2 h, the O2 concentration in the GDL-free reactor dropped (ca. 33%), but it remained almost unchanged in the GDL-based system (Fig. 2f), demonstrating the efficient O2 permeability of GDL. In a laboratory-grade reactor (d = 10 cm, h = 2.5 cm, Supplementary Fig. 22a, b), 1.5 L of H2O2 solution with a concentration of ca. 250 μM can be produced during a 14-h operation with a flow rate of 2 mL min−1 (Supplementary Fig. 22c). Then, the volume of the reactor was further expanded (40 × 40 × 2 cm) and its performance was evaluated under natural sunlight irradiation, as displayed in Fig. 2g. On August 8th, 2024 with sunny weather, the light intensity at the campus of Guangdong University of Technology was between 65 and 100 mW cm–2 (Fig. 2h). During a reaction period from 11 AM to 4 PM, H2O2 could be continuously produced with a flow of 15 mL min−1, and 5 L of H2O2 solution (ca. 400 μM) was finally harvested, which ranks among the top of the documented systems under similar conditions (Supplementary Table 2), highlighting the great practical potential of both STT COF and the continuous flow membrane reactor.

Identification of active sites

The effects of SQ units were examined by evaluating the H2O2 generation of both COFs under varying pH levels. As shown in Fig. 3a, the activity of PTT COF declines as pH elevates, while the performance of STT COF is virtually unaffected by pH changes. Given that H2O2 photosynthesis is a PCET-mediated ORR process, the aforementioned disparity demonstrates that including SQ units considerably enhances the proton utilization capacity of STT COF37. To further evaluate the proton utilization capacity of SQ units, the pH variations of these two COF systems were tested. During the 1-h photocatalytic H2O2 production process, the pH of the STT COF-based system increases from 8.16 to 9.17 (Supplementary Fig. 23a), whereas it remains mostly unchanged in the PTT COF-based system. This discrepancy may be attributed to the enhanced ability of STT COF to extract protons from water during the photocatalytic process. Notably, in the dark, the pH of the PTT COF-based system stays mostly stable whereas it gradually increases in the STT COF-based one (Supplementary Fig. 23b). The pH variation of the former can be caused by the negatively charged O atom at the C-O− site in SQ units absorbing protons from water by spontaneous hydrogenation (STT COF → STT COF*), which also results in a structural change (from C-O− to C-OH), as schematically illustrated in Fig. 3c. Since the D-O bond in D2O is more thermodynamically stable than the H-O bond in H2O38, it may be harder for COFs to break the D-O bond and extract D⁺ from D2O. Therefore, photocatalytic hydrogen peroxide production tests using these two COFs were carried out in H2O and D2O, respectively. As revealed in Fig. 3b, the activity of PTT COF in D2O decreased by 24% compared to H2O, while STT COF delivers almost unchanged performance. Besides, the slightly smaller water contact angle and more negative Zeta potential of STT COF (Supplementary Fig. 24) also reflect facilitated proton uptake by STT COF. Therefore, these findings jointly demonstrate that the SQ units promote proton utilization, which benefits H2O2 photosynthesis.

Fig. 3: Proton utilization ability and the spontaneous hydrogenation process.
Fig. 3: Proton utilization ability and the spontaneous hydrogenation process.
Full size image

a Photocatalytic H2O2 evolution by STT COF and PTT COF under different pH. b Photocatalytic activity of STT COF and PTT COF for H2O2 generation in H2O and D2O. c Schematic diagram of the spontaneous hydrogenation process of STT COF. d In-situ DRIFT spectra of PTT COF and e STT COF under dark conditions with continuous water vapor. The error bars in the figures represent the standard deviations from triplicate tests.

To further substantiate the above hypothesis, in-situ diffuse reflectance infrared Fourier transform spectroscopy (in-situ DRIFT) was performed. In dark with continuous water vapor aerated conditions (Fig. 3d), peaks from C = C stretching modes of benzene ring (1492 and 1523 cm–1) and benzene ring (1573 cm–1) could be observed in PTT COF39,40, and their intensity gradually increased with time, implying that water is inclined to be adsorbed on the benzene ring in PTT COF. The intensifying peaks of C-O (1414 and 1441 cm–1) and O-H (1354 and 1384 cm–1) may originate from the adsorbed water41,42. As for STT COF (Fig. 3e), the peaks at 1259 cm–1 and 1558 cm–1 are attributed to the C = C-O and C = NH+ vibrations43, respectively, while the peak at 1441 cm–1 may come from the vibration of SQ units or adsorbed water. The emerging prominent peak at 1402 cm–1 can be ascribed to the C-OH vibration44, further validating the spontaneous hydrogenation process of STT COF (STT COF → STT COF*), which is accompanied by the structural transformation from C-O– to C-OH.

Investigation of exciton and oxygen adsorption behaviors

To disclose the disparities in their H2O2 photosynthesis performances, the exciton dissociation and subsequent electron transfer behaviors of these two COFs were thoroughly investigated. Temperature-dependent photoluminescence (TD-PL) spectra were employed to analyze the exciton binding energy (Eb), a crucial parameter that characterizes the Coulomb interaction of excitons and illustrates the energy needed to ionize excitons into free charge carriers45. Arrhenius fitting of temperature-dependent integrated PL intensities yields Eb of 83.58 meV for PTT COF and 73.14 meV for STT COF (Fig. 4a, b). The decreased Eb (12%) indicates better exciton dissociation in STT COF46. Besides, the intensity of steady-state photoluminescence (PL) spectra of STT COF is much lower than that of PTT COF (Fig. 4c), with an increased average carrier lifetime from 0.787 ns to 1.550 ns (Supplementary Fig. 25), further signifying that the SQ units can significantly suppress the recombination of photogenerated electron-hole (e−-h+) pairs. STT COF exerts a higher photocurrent density than PTT COF (Fig. 4d), indicating more efficient charge separation47. The lower charge transference resistance of STT COF obtained by electrochemical impedance spectroscopy (Fig. 4e) implies a lower charge transfer barrier, and the higher surface potential of STT COF determined by Kelvin probe force microscopy (KPFM) test (Fig. 4f) validates its better separation of photogenerated e−-h+ pairs.

Fig. 4: Exciton dissociation and electron utilization investigation.
Fig. 4: Exciton dissociation and electron utilization investigation.
Full size image

a STT COF and b PTT COF (inset: the relationship between PL intensity and temperature, fitted using the Arrhenius equation). c Photoluminescence emission spectra of PTT COF and STT COF. d Transient photocurrent under visible light irradiation. e Electrochemical impedance spectra (EIS) under visible light irradiation. f Images of KPFM surface potentials (left) of PTT COF and STT COF. Contact potential difference (CPD) (right) of two COFs measured by KPFM. g 2D pseudocolor mapping TA spectra of STT COF. h TA spectra signals of STT COF. i TA kinetics plots and typical fitting curves of PTT COF and STT COF probed at 485 nm probe wavelength.

Then, femtosecond transient absorption (fs-TA) spectroscopy was carried out to investigate exciton dynamics and carrier transfer mechanisms. As displayed in the two-dimensional (2D) pseudocolor mapping TA spectra and TA spectra signals (Fig. 4g, h and Supplementary Fig. 26), both PTT COF and STT COF exhibit broad positive absorption between 440 nm and 800 nm, corresponding to excited state absorption signals from electronic transitions between the singlet state (S1) and higher energy excited states (Sn)48. Fig. 4i depicts the TA kinetics of these two COFs at a detection wavelength of 485 nm. Based on the decay curves of the TA signals fitted by a double exponential function (with τ1 attributed to Sn → S1 and τ2 corresponding to Sn→ ground state (S0))49, the average lifetime (τave) of PTT COF and STT COF is calculated to be 200.9 ps and 576.6 ps, respectively. The longer τave indicates that STT COF features an enhanced charge separation capability with a reduced recombination rate50. Therefore, introducing SQ units into COF promotes exciton dissociation and then enhances charge separation and transfer efficiency.

In-situ DRIFT was used to identify the real-time intermediates and pathways in H2O2 photosynthesis. Under O2-saturated conditions with water vapor and visible light irradiation, the benzene ring in PTT COF exhibits characteristic C = C stretching vibrations (Supplementary Fig. 27a), with increasing intensity over time, indicating its role as the catalytic active site. On the other hand, the C = NH⁺ vibration peak in STT COF (Supplementary Fig. 27b) shows a continuously augmented intensity, indicating that SQ units actively participate in ORR. Besides, distinct peaks attributed to the 1,4-endoperoxide intermediate (910 cm−1) and the -O-O- bond (ca. 1115–1295 cm−1) can be gradually observed in both systems51, implying their Yeager-type O2 adsorption configuration, which favors H2O2 photosynthesis via a direct one-step 2e− pathway. Besides, the stronger -O-O- signal in STT COF indicates its enhanced O2 adsorption affinity, which is further validated by the results of O2 temperature-programmed desorption (O2-TPD) (Supplementary Fig. 28).

Theoretical calculations

Time-dependent density functional theory (TD-DFT) and Density functional theory (DFT) calculations were performed to gain an in-depth understanding of the H2O2 photosynthesis process over these two COFs (atomic coordinates provided in the Supplementary Data 1 file). Based on the minimum periodic structural units, the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbitals (LUMO) for these two COFs were calculated. As displayed in Fig. 5a, the HOMO and LUMO of PTT COF overlap distinctly, while STT COF exhibits a pronounced spatial separation between HOMO and LUMO, which is conducive to more efficient charge separation upon photoexcitation52. Then, the charge density difference between the ground and excited states is analyzed to give insights into their different electron transfer characteristics. The charge transfer distance (D) denotes the distance between the centers of electron density accumulation and depletion zones formed by electronic excitation, reflecting the separation efficiency of e−-h+ pairs and the spatial separation of orbitals in this D-A system53. As depicted at the top of Fig. 5b, the TAPT unit exhibits its electron-accepting nature, while the SQ and PDA units feature electron-donating properties, signifying clear D-A interactions in these two COFs. The direction and magnitude of charge transfer can be determined through charge density difference and Bader charge calculation, with the corresponding results visualized through the centroids of the electron and hole clouds. As shown in Fig. 5b, STT COF exhibits a larger D (0.361 Å) and transferred charge (0.151 e) than PTT COF (0.056 Å, 0.077 e). The larger D implies the enhanced spatial separation between electron and hole centroids, as well as a stronger D-A interaction, which facilitates charge migration. In STT COF, the SQ units serve as stronger donors, promoting charge migration toward the TAPT acceptors54. Therefore, these calculation results suggest that the SQ units in STT COF facilitate the separation and transfer of photogenerated charges, aligning well with experimental observations.

Fig. 5: TD-DFT calculations and mechanism analysis.
Fig. 5: TD-DFT calculations and mechanism analysis.
Full size image

a Calculated HOMO-LUMO distribution of PTT COF and STT COF. b Above: calculation of the charge density difference between the ground and excited states of PTT COF and STT COF. c the adsorption energy of O2 on the optimum sites of STT COF and STT COF*. d Free-energy diagrams for the reduction of O2 to H2O2 on PTT COF and STT COF*. e The mechanism of STT COF for photocatalytic H2O2 formation.

The adsorption of O2 molecules on the surface of catalysts is the initial step of the ORR process, which has been then simulated based on the model fragments (Supplementary Fig. 29). As shown in Supplementary Fig. 30, the benzene ring on TAPT units features an enhanced O2 adsorption energy (−0.099 eV) than the other sites in PTT COF, corresponding to a Yeager-type adsorption configuration, which is consistent with the results of in-situ DRIFT. As for STT COF (Supplementary Fig. 31), the adsorption energies of O2 on all possible adsorption sites are positive, and the potential adsorption site (C = NH+) features a Pauling-type O2 adsorption configuration. Neither the O2 adsorption energies nor configurations match the results of H2O2 photosynthesis performance, O2-TPD, and in-situ DRIFT. Such a discrepancy motivated further theoretical and experimental investigations. Since the spontaneous sites formed during the reaction may serve as active centers18, the O2 adsorption process on the STT COF* (STT COF after the spontaneous hydrogenation) was further investigated. As displayed in Fig. 5c, after converting C-O− to C-OH, STT COF* offers a pair of adjacent H sites (from C = NH+ and C-OH groups, respectively) for O2 adsorption. The adsorbed O2 with Yeager-type configuration rapidly acquires H+ from the C-OH site and transforms into *OOH, which can serve as a metastable O2-adsorption configuration due to the presence of a hydrogen bond between STT COF and the *OOH, as illustrated in Supplementary Video 1. Such an O2 adsorption configuration features an increased adsorption energy (−2.02 eV), which can be supported by the aforementioned experimental results. Therefore, the C = NH+ and C-OH sites in STT COF* instead of the sites in pristine STT COF serve as the actual adsorption centers, implying that the spontaneous hydrogenation of STT COF is capable of boosting O2 uptakes.

The ORR process over these two COFs was then simulated to understand the reaction pathways from a thermodynamic perspective. PTT COF follows the conventional one-step 2e⁻ pathway for H2O2 production. As shown in Fig. 5d, the conversion of adsorbed O2 into intermediates (*O2 → *HOOH) serves as the rate-determining step, with an energy barrier (ΔG) of 0.56 eV. As for STT COF*, the adsorbed O2 (i.e., *OOH) undergoes a thermodynamically favorable process to form *HOOH, which is eventually desorbed to give free H2O2 with a small barrier (ΔG) of 0.14 eV (*HOOH → H2O2(l)). During the whole reaction process, the lower ΔG in STT COF* favors H2O2 generation, suggesting the importance of the spontaneous hydrogenation in STT COF for photocatalytic production of H2O2.

Based on the above experimental and theoretical results, plausible mechanisms for photocatalytic H2O2 production over these two COFs are proposed. STT COF with enhanced D-A configuration can improve the separation of photogenerated e−-h+ pairs. Moreover, as illustrated in Fig. 5e, the SQ units in STT COF undergo a spontaneous hydrogenation under aqueous conditions, converting the pristine C-O– to C-OH (STT COF → STT COF*), providing a pair of adjacent H sites to facilitate O2 adsorption in Yeager-type configuration with lower O2 adsorption energy. Then the adsorbed O2 quickly captures the H+ from the spontaneous C-OH site to form *OOH. Subsequently, the *OOH further gains electrons and a proton via a one-step 2e− to produce *HOOH. After generating H2O2 via *HOOH desorption, the original STT COF state is regenerated from STT COF*and rolls for the next photocatalytic cycle. On the other hand, as for PTT COF, the adsorbed O2 on the benzene ring on TAPT unites with a Yeager-type configuration and then obtains protons and electrons for H2O2 production (Supplementary Fig. 32).

Discussion

In summary, we present a facile triple synergy for boosting H2O2 photosynthesis by an elaborately designed zwitterionic STT COF. Systematic investigations demonstrate that STT COF not only features a well-designed D-A configuration capable of improving the separation of photogenerated e−-h+ pairs, but also undergoes spontaneous hydrogenation to facilitate proton utilization and O2 adsorption in Yeager-type configuration. As a result, a high H2O2 yield (14,356.5 μmol g–1 h–1) with an AQY of 40.0% (at 420 nm) has been achieved over STT COF in pure water, outperforming its charge-neutral counterpart and other documented systems. A sunlight-driven continuous flow membrane reactor integrated with STT COF and gas diffusion layers achieves scalable production of 5 L of H2O2 (400 μM) within 7 h, demonstrating strong practical applicability. These findings underscore the significance of triple synergistic optimization of each successive reaction step of ORR, paving the avenue for efficient H2O2 photosynthesis.

Methods

Materials

4,4’,4”-(1,3,5-Triazine-2,4,6-triyl)trianiline (TAPT, 98%), squaric acid (SQ, 98%) were obtained from Jilin Chinese Academy of Sciences-Yanshen Technology Co., Ltd. Terephthalaldehyde (PDA, 98%), N, N-diethyl-1,4-phenylene-diamine sulfate (DPD, 96%), peroxidase (POD, horseradish, 99%), p-benzoquinone (p-BQ, 99%), potassium bromate (98%), isopropyl alcohol (IPA, AR), n-butyl alcohol (spectral pure), methylbenzene (AR), 1,2-dichlorobenzene (spectral pure), mesitylene (AR), sodium dihydrogen phosphate anhydrous (99%), potassium phosphate dibasic (99%), ferrous sulfate heptahydrate (99%), methyl blue (95%), bisphenol A (BPA, 99%), tetrahydrofuran (AR), N, N-dimethylformamide (DMF, AR), n-hexane (AR), methanol (MeOH, AR), acetic acid (AR), HCl (99%) and NaOH (98%) were purchased from Aladdin Co., Ltd., Shanghai, China. All chemicals were used without further purification. The carbon cloth (W1S1011) used for the continuous-flow membrane reactor was purchased from CeTech. The Ultrapure water (UP) with the resistivity of 18.3 MΩ cm−1 was used in this work.

Synthesis of COFs

The synthesis of STT COF

Squaric acid (25.1 mg, 0.22 mmol) and 4,4’,4”-(1,3,5-triazine-2,4,6-triyl) trianiline (39.7 mg, 0.112 mmol) were mixed in a 10 mL pressure tube, then n-butyl alcohol (1.5 mL), methylbenzene (0.75 mL), and a suitable-sized magnetic stir bar were added. After 15 min of sonication, the mixture was degassed through three freeze-pump-thaw cycles and sealed under vacuum, then heated at 120 °C for 72 h with a rotation speed of 200 rpm. The obtained orange-colored sample was washed with DMF, tetrahydrofuran, water, and methyl alcohol. Finally, the products were evacuated at 60 °C under vacuum. The yield of STT COF is calculated to be ca. 89%.

The fabrication of PTT COF

PTT COF was synthesized according to a reported procedure with minor modifications55. Typically, 4,4’,4”-(1,3,5-triazine-2,4,6-triyl) trianiline (21.6 mg, 0.063 mmol), terephthalaldehyde (12.7 mg, 0.095 mmol), 1,2-dichlorobenzene (0.5 mL), n-butyl alcohol (0.5 mL), and acetic acid (0.1 mL, 6 M) were added to a 10 mL pressure tube. The mixture was sonicated for 15 min, followed by degassing through three freeze-pump-thaw cycles, vacuum sealing, and heating at 120 °C for 72 h. The obtained yellow sample was washed with tetrahydrofuran, DMF, and n-hexane. Finally, the products were evacuated at 60 °C in a vacuum drying oven overnight. The yield of PTT COF is ca. 92%.

Photocatalytic H2O2 production and other reactions

Photocatalytic H2O2 production

A flask was charged with the photocatalyst 5 mg, 20 mL of H2O, and sealed with a septum. After thorough ultrasonication, the obtained suspension was degassed by O2 bubbling for 20 min. Then the acquired solution was stirred in the dark for 30 min to establish equilibrium between absorption and desorption. A 300 W Xe lamp (PLSSXE300D/300DUV, Beijing Perfectlight, China) was used for the photocatalytic reaction. The entire reaction system was maintained at 25 °C using a cooling water system. At each sampling time, a syringe sampler was used to collect samples from the suspension, which were then immediately filtered through a 0.22 μm needle filter to remove the catalyst, resulting in a hydrogen peroxide solution ready for measurement.

Photocatalytic H2O2 decomposition

The evaluation of photocatalytic consumption of H2O2 involved dispersing 5 mg of COF-based catalysts in a 20 mL aqueous solution with 0.2 mM H2O2 under N2 conditions.

Cycling experiments

To assess the photocatalytic stability of the catalyst, several parallel tests were conducted under identical conditions. After each cycle, the catalysts were separated and washed multiple times with methanol and water. The samples were then dried and reused in subsequent activity tests. This process was repeated for multiple cycles to evaluate the catalyst’s performance over time.

Continuous flow membrane experiment

Gas diffusion layers (GDL) composed of carbon black and polytetrafluoroethylene (PTFE) were coated on one side of carbon cloth, which provides a high surface area and porosity to ensure efficient oxygen permeability56. Then, a certain number of STT COF is coated on the other side of the carbon cloth.

Lab-setup reactor was applied with 100 mg STT COF, which was thoroughly dispersed within ethyl alcohol after 15 min of ultrasonication. Then the mixture was evenly coated on the carbon cloth (a circular area with a diameter of 10 cm) and left to naturally air-dry; it was fixed in the reactor. The reactor was charged with 200 mL of deionized water at the beginning of the experiment. The concentration of the produced H2O2 solutions was tested at specific time intervals. This experiment was conducted under UV–vis light irradiation.

A continuous flow membrane experiment under natural sunlight was conducted based on the above procedure, while the total amount of STT COF used was 4g, and the size of the carbon cloth was 40 cm × 40 cm. The pre-charged volume of deionized water is 1.6 L.

pH variation experiments

A quartz reactor was used to add 50 mg of COF and 20 mL of deionized water. After ultrasonication, the mixture was placed in a dark condition for 4 h. The experiment under light irradiation was carried on based on the usual route of photocatalytic H2O2 production in the main text.

Isotopic labeling experiment

STT COF (30 mg) was suspended in H2O (10 mL) in a 20 mL glass vial. The suspension was well-dispersed by sonication for 5 min, then bubbled with 18O2 for 10 min before sealing the vial. The reaction was irradiated for 6 h using a 300 W Xe lamp (320 nm < λ < 780 nm). The post-reaction photocatalytic hydrogen peroxide solution was transferred to a new vial, which was pre-vacuumed and charged with MnO2. After the thorough decomposition of hydrogen peroxide, the evolved oxygen gas was analyzed through GC-MS (Agilent 8860-5977B).

Photocatalytic O2 evolution

5 mg of catalyst was charged in 20 mL of 0.01 M KBrO3 solution, then 20 mg of La2O3 was added to maintain the pH57. The experiment was conducted in an N2 atmosphere under UV–vis light irradiation.

Fenton degradation

First of all, 10 mg L–1 methyl blue containing 0.01 M FeSO4 with pH = 3 was prepared. The configured methyl blue solution and the produced H2O2 solution (photocatalytic conditions: 20 mg of STT COF was placed in 20 mL of deionized water, exposed to O2 for 20 min, and then reacted under UV–vis light for 2 h) were added in equal proportions, and the color change of methyl blue was observed.

As for bisphenol A (BPA) degradation, the procedure is similar to the above, except that the concentration of BPA is 5 mg L–1, and the concentration of the pollutant is tested under liquid chromatography at a specific period.

Detection of H2O2 concentration

DPD colorimetry was used to detect the concentration of H2O2. The sodium phosphate buffer was prepared by slowly adding a 0.5 M sodium dihydrogen phosphate solution to a 0.5 M monosodium phosphate solution, monitoring the mixture until the pH reached 6. Additionally, 0.1g of DPD was dissolved in 10 mL of 0.05 M H2SO4. The other solution was prepared by adding 5 mg of POD to 5 mL of deionized water58. To determine the H2O2 concentration, first fill a 10 mL colorimetric tube with 3 mL of phosphate buffer. Next, add a specific quantity of sample so that the absorbance falls within the standard curve. Next, sequentially add 50 μL of DPD solution and 50 μL of POD solution. Finally, use deionized water to bring the total volume to 10 mL and mix thoroughly. After stirring for 50 s, the absorbance at 551 nm was measured using a UV–vis spectrophotometer from Agilent Technologies. Dilute the 30% H2O2 stock solution to a concentration of 0–200 μM, and use the colorimetric method to obtain the corresponding H2O2 concentration-absorbance standard curve. The linear relationship between the obtained H2O2 concentration and absorbance is shown in the Supplementary Fig. 33.

Characterization methods

The X-ray diffraction (XRD) pattern of samples was recorded on a Rigaku Smart Lab diffractometer with Cu Kα radiation (λ = 1.5418 Å) at room temperature. X-ray photoelectron spectroscopy (XPS) experiments were carried out using Thermo Fisher’s ESCALAB 250Xi. Fourier transform infrared spectroscopy (FTIR) spectra were collected on a Nicolet iN10 (Thermo Fisher). Solid-state NMR spectra were obtained at the Bruker Avance III 400 MHz. The morphology of the sample was characterized on a field emission scanning electron microscope (FESEM, Zeiss Sigma 300, Gemini) and a transmission electron microscope (TEM, Tecnai G2-20, USA). Thermogravimetric analysis (TGA) was carried out on NETZSCH STA 449C by heating samples from 70 to 700 °C in a dynamic nitrogen atmosphere with a heating rate of 10 °C min–1. The oxygen temperature-programmed desorption (O2-TPD) test for the catalyst was conducted on an AutoChem II 2920. The photoluminescence (PL) tests were recorded on a F-4600 FL spectrophotometer at emission scan mode. In-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was carried out in Thermo IS 50. Temperature-dependent PL spectra (TD-PL) were explored on the Shimadzu RF-6000 spectrometer. EPR spectra were recorded on a Bruker EMX spectrometer. The pore size, pore volume, and surface area of samples were obtained by the Brunauer–Emmett–Teller (BET) instrument (ASAP 2460, Micromeritics). The concentration of bisphenol A was tested by High-Performance Liquid Chromatography (HPLC, Waters, USA). Concentrations of dissolved O2 during the O2 evolution experiment were tested by a portable oxygen dissolving meter (HACH, HQ 30 d). The Kelvin probe force microscopy (KPFM) (SPM-9700, Shimadzu, Japan) was used to test the surface potential of the samples. Ultraviolet–visible spectroscopy (UV–Vis-DRS) was performed by PerkinElmer Lambda 950. Electrochemical measurements were carried out on a CHI760e (Chenhua, Shanghai) electrochemical workstation in a three-electrode cell system. TA spectroscopy was performed using a commercial TA spectrometer (Helios, Ultrafast System, USA), with excitation provided by a femtosecond-pulsed amplifier (Astrella, 80 fs, 1 kHz, Coherent) and frequency-doubled by a Coherent Libra regenerative amplifier (800 nm, 80 fs, 1 kHz). Isotopic labeling experiments were analyzed through GC-MS (Agilent 8860-5977B).

Photoelectrochemical measurements and calculations of AQY

Photoelectrochemical characterizations

10 mg of photocatalyst was added to 1 mL of ethanol. After 1 h of sonication, 20 μL of the mixture was applied onto the FTO conductive glass and dried to create a functioning electrode. The Ag/AgCl electrode served as the reference electrode, while the Pt electrode acted as the counter electrode. A 0.1 M Na2SO4 solution, freshly prepared before each experiment, was used as the electrolyte and analyzed using the electrochemical workstation (CHI760e, Shanghai Chenhua Instrument Co., Ltd.). A 300 W Xe lamp (420 nm < λ < 780 nm, PLSSXE300D/300DUV, Beijing Perfectlight, China) was used as the light source, and the transient photocurrent was measured over time during the on/off light cycle. The electrochemical impedance spectrum was evaluated across a frequency range from 1000 kHz to 0.01 Hz.

RRDE measurement

The RRDE experiments were carried out using a CHI760e electrochemical workstation in a standard three-electrode configuration. A RRDE electrode (glassy carbon disk with a platinum ring, disk area = 0.2475 cm2) was employed as the working electrode, with an Ag/AgCl electrode as the reference electrode and a graphite rod as the counter electrode. The electrolyte was a 0.1 M phosphate buffer solution (pH = 7), prepared fresh and used immediately for each experiment. The catalyst ink was prepared by dispersing 5 mg of the COF sample in 2 mL of isopropanol containing 50 μL of naphthol, followed by sonication for 30 min to obtain a homogeneous suspension. A fixed volume of the catalyst ink (10 μL) was drop-cast onto the surface of the RRDE and dried under ambient conditions. Before measurement, the electrolyte was saturated with O2 by purging for at least 30 min. Linear sweep voltammetry (LSV) was employed to obtain the ORR polarization curves at a scan rate of 10 mV·s–1, with measurements taken at 1600 rpm. All measurements were conducted at room temperature. The number of transferred electrons (n) was calculated from the disk current (Id) and ring current (Ir) using Eq. (1):

$${{\rm{n}}}=4\times \frac{{{{\rm{I}}}}_{{{\rm{d}}}}}{{{{\rm{I}}}}_{{{\rm{d}}}}+{{{\rm{I}}}}_{{{\rm{d}}}}/{{\rm{N}}}}$$
(1)

where N is the Pt ring’s collecting efficiency (N = 0.37), and Id and Ir are the disk and ring currents (μA), respectively.

In-situ DRIFT measurements

For the tests in dark conditions, the samples were filled into an in-situ IR cell, and H2O vapor was introduced into the cell. Before the measurement, the samples were degassed at 393 K for 6-h and then and then purged with Ar for 1 h at 353 K. The baseline was obtained after the samples had cooled down. The tests under photocatalytic conditions were started with Ar purging for 1 h at room temperature to remove the impurities. Next, moist oxygen (a mixture of oxygen and water vapor) was introduced in the dark, and the container was sealed to stop ventilation. The adsorption process lasted for 60 min, during which data were recorded at 30 min and 60 min until oxygen adsorption equilibrium was achieved. The reaction was then initiated by turning on the 300 W Xe lamp, and data were recorded at regular intervals.

Exciton dissociation energy calculation

Steady-state photoluminescence fluorescence (PL) spectra were recorded at different temperatures, and the Eb of the sample was calculated using the Arrhenius equation59:

$${{\rm{I}}}({{\rm{T}}})=\frac{{{{\rm{I}}}}_{0}}{1+{{\rm{A}}} \, {\mathrm{exp}}\left(\frac{-{{{\rm{E}}}}_{{{\rm{b}}}}}{{{{\rm{k}}}}_{{{\rm{b}}}}{{\rm{T}}}}\right)}$$
(2)

where I0 represents the PL intensity at 0 K; T stands for temperature; kb refers to Boltzmann’s constant.

AQY measurement

The AQY of STT COF was tested in pure deionized water (30 mL) with a photocatalyst (20 mg) in a baker. Then, irradiated under Xe lamp (PLSSXE300D/300DUV, Beijing Perfectlight, China) at the wavelength of 420 nm for 30 min with magnetic stirring and continuous O2 aeration. The active area of the reactor was approximately 7 cm2, and the light intensity was controlled to be 5 W m–2. The AQY was calculated as follows:

$${{\rm{AQY}}}=\frac{({{\rm{number}}}\; {{\rm{of}}}{{{\rm{H}}}}_{2}{{{\rm{O}}}}_{2}{{\rm{production}}})\times 2}{{{\rm{number}}} \; {{\rm{of}}} \; {{\rm{incident}}} \; {{\rm{photons}}}}\times 100\%$$
(3)

The number of incident photons is:

$${{{\rm{N}}}}_{{{\rm{incident}}}}=\frac{{{\rm{Pt}}}}{{{\rm{hv}}}}=\frac{{{\rm{Pt}}}{{\rm{\lambda }}}}{{{\rm{hc}}}}=\frac{{{\rm{ISt}}}{{\rm{\lambda }}}}{{{\rm{hc}}}}$$
(4)

where I is the light power intensity (W m–2), S is the irradiation area (m2), t is the reaction time (s), λ is the wavelength length (m) of the monochromatic light, h is Planck’s constant (6.63 × 10–34 m2 kg s–1), and c is the speed of light in free space (3.0 × 108 m s–1).

DFT calculations

Spin-polarized DFT calculations were performed utilizing the VASP 5.4.460, employing the projector-augmented wave (PAW)61 methodology. The Perdew−Burke−Ernzerhof (PBE) exchange-correlation functional62, within the generalized gradient approximation (GGA) framework, was adopted. A plane-wave kinetic energy cutoff of 450 eV and a Gaussian smearing width of 0.1 eV were implemented throughout the calculations. All atomic geometries and lattice vectors were optimized via the conjugate-gradient algorithm, with stringent convergence criteria set at 10–5 eV for total energy and 0.02 eV/Å for interatomic forces. The Brillouin zone was accomplished using a Γ-centered Monkhorst–Pack grid of 2 × 2 × 1 grid for structural relaxation and a denser 4 × 4 × 1 grid for self-consistent field (SCF) calculations. To avoid periodic artifacts, a 15 Å vacuum layer was introduced. Dispersion corrections were accounted for via the DFT-D3 methodology to accurately describe van der Waals interactions63. Charge redistribution was analyzed using:

$$\Delta \rho={\rho }_{{{\rm{tot}}}}-({\rho }_{{{\rm{part}}}1}+{\rho }_{{{\rm{part}}}2})$$
(5)

where ρtot is the total charge density, while ρpart1 and ρpart2 represent the charge densities of the correspond to isolated components. Visualization was performed with VESTA64, and vibrational frequencies were obtained using VASPKIT65.

The Gibbs free energy (ΔG) was evaluated as:

$$\Delta {{\rm{G}}}=\Delta {{\rm{E}}}+{\Delta {{\rm E}}}_{{{\rm{ZPE}}}}{{\rm{\hbox{--}}}}{{\rm{T}}}\Delta {{\rm{S}}}$$
(6)

with ΔE from DFT, ΔΕZPE from zero-point energy, and ΔS from vibrational analysis at 298.15 K (T). Entropies of gaseous H2 and H2O2 were taken from the NIST database with standard conditions.

Optical excitations were studied using time-dependent DFT (TD-DFT) in Gaussian. Ground states were optimized at the B3LYP/6-311G(d) level, confirmed by frequency analysis. Excited states were calculated at CAM-B3LYP/6-311G(d) based on optimized geometries. Excited-state electron distributions were analyzed with Multiwfn (http://sobereva.com/multiwfn) and visualized in VESTA66,67. Solvent effects were included via the conductor-like polarizable continuum model (C-PCM)68, and Grimme’s D3 correction was applied for dispersion63.