Introduction

Conversion of solar energy into chemical energy, resulting in high value-added chemicals such as hydrogen (H2) and hydrogen peroxide (H2O2) from sustainable and economical resources like water (H2O) and oxygen (O2), is of immense importance to tackle environmental concerns. Hydrogen peroxide is a versatile green oxidant extensively used in bleaching, water purification, medical disinfection and organic synthesis1,2. In addition, the use of H2O2 in fuel cells and as a non-polluting source of energy for vehicles and rockets has been demonstrated. The use of H2O2 is particularly advantageous because it releases only water and oxygen as byproducts, posing minimal environmental threat3,4,5. Although the output power potential of the H2O2 fuel cell is slightly lower than that of the H2 fuel cell, it is considered a favorable green energy source due to its ease of transport and storage under ambient conditions6,7. The conventional anthraquinone oxidation method used for producing H2O2 industrially has significant disadvantages, including multiple energy-intensive steps for hydrogenation and tedious extraction and purification of H2O28. Additionally, it generates a considerable amount of hazardous chemical waste, posing a substantial threat to the environment9. In this context, photocatalytic strategies for H2O2 production are considered to be the cleanest, greenest and most environmentally friendly, with zero-carbon emissions10,11.

Over the years, significant efforts have been devoted towards the development of photocatalysts for H2O2 generation, including various materials such as inorganic semiconductors (e.g., metal oxides and metal sulfides)12,13,14, graphitic carbon nitride (g-C3N4)15,16, porous and conjugated polymers17,18,19, resins20, metal organic frameworks (MOFs)21,22,23, and covalent organic frameworks (COFs)24,25,26. However, challenges such as poor H2O2 generation efficiencies have been reported owing to the factors such as lack of crystallinity, wide band gap, low porosity or incompatible redox selectivity. Considering the challenges, among the various materials, COFs have emerged as a promising class of crystalline porous organic materials for photocatalytic H2O2 generation due to their ability to offer tunable functionalities, efficient light harvesting in the visible region, narrow band gap, extended π-conjugation, efficient charge separation and high photostability27,28,29,30. Furthermore, the incorporation of triazine31, heptazine32, benzothiazole33, bipyridine34, pyrene35, and porphyrin36 moieties into the COF backbone has been found to be highly beneficial for significantly accelerating the H2O2 generation. Many recent reports demonstrated the applications of COFs for efficient and durable photocatalytic H2O2 generation in the presence of sacrificial electron donors and an adequate supply of oxygen—crucial factors for achieving satisfactory H2O2 activity by enhancing the oxidation half-reaction35,37.

In ongoing research, COFs bearing donor–acceptor (D-A) units have demonstrated efficient photocatalytic H2O2 generation due to improved electron-hole separation and transport, along with a lower recombination rate38,39,40. Excellent charge separation, multi-exciton photogeneration and mass transfer have been observed in COFs containing the benzotrithiophene motif, e.g., benzo[1,2-b:3,4-b’:5,6-b”] trithiophene-2,5,8-tricarbaldehyde (BTT) as an electron donor41. To date, the majority of COFs with imine and vinylene linkages have been found to be active for photocatalytic H2O2 generation42,43. However, these COFs often lack sufficient active sites for effective water oxidation and oxygen reduction. In this context, it is expected that polar hydrazone linkages in COFs can provide desirable sites for docking both water and oxygen, thus promoting simultaneous water oxidation and oxygen reduction to efficiently generate H2O242,43,44,45. Recently, the effect of linkages in donor-acceptor COFs on the photocatalytic H2O2 generation from water using solar light has been demonstrated46, where was observed that compared to imine and vinylene linkages, hydrazone linkages provide efficient charge transfer and mass transport to catalytic sites, with a very high H2O2 generation rate. However, no studies have reported a comparative investigation of photocatalytic H2O2 generation by tuning the density of hydrazone bonds along the COF backbone to improve photocatalytic performance.

In order to investigate the effect of the covalent bonds present in the COF backbone on photocatalytic H2O2 generation, we report here a series of COFs with varying numbers of imine and hydrazone linkages: BTT-DAB with imine linkages, BTT-H1 with one imine and one hydrazone linkage, and BTT-H2 with two hydrazone linkages (Fig. 1). Our work involving controlled experiments, supplemented by density functional theory (DFT) calculations, highlights the crucial role of the hydrophilic and polar hydrazone linkages in activating H2O and lowering the energy barrier of the •O2- and •OOH intermediates in H2O2 generation. Our study demonstrates that the greater water affinity of the hydrazone linkage and its enhanced interaction with oxygen at the catalytic sites synergistically increases the rate of H2O2 generation.

Fig. 1: Synthesis and crystallinity of COF photocatalysts.
figure 1

a Scheme of synthesis of imine-linked BTT-DAB, partial imine and hydrazone-linked BTT-H1 and hydrazone-linked BTT-H2 COFs. Comparison of simulated and experimentally obtained PXRD patterns of b BTT-DAB, c BTT-H1, and d BTT-H2 showing the formation of crystalline COFs.

Full size image

Results and discussion

The careful selection of linker units and linkages is essential for photocatalytic applications as they act as connectors or bridges for the photogeneration of excitons and charge transport through the backbone. To facilitate the photocatalytic H2O2 generation, we designed and synthesized a set of three COFs with variable numbers of imine and hydrazone linkages. To achieve the synthesis of such COFs, we carried out the Schiff base reaction of electron-rich benzo[1,2-b:3,4-b’:5,6-b”] trithiophene-2,5,8-tricarbaldehyde (BTT) with electron-deficient 4-diaminobenzene (DAB), 4-aminobenzohydrazide (H1) and terephthalic dihydrazide (H2) to yield BTT-DAB, BTT-H1, and BTT-H2 COFs, respectively, under solvothermal conditions (Fig. 1a). These COFs were found to be insoluble in commonly used organic solvents, including dimethylsulfoxide, N,N-dimethylformamide, tetrahydrofuran, methanol and acetone.

The long-range periodicity of the BTT-based donor-acceptor COFs was confirmed using powder X-ray diffraction (PXRD) analysis (Cu Kα radiation; λ = 1.5418 Å). The PXRD analyses showed strong reflections at 3.51 and 6.15 2θ degrees for BTT-DAB, 3.06 and 5.39 2θ degrees for BTT-H1, and 2.79 and 4.94 2θ degrees for BTT-H2, corresponding to the (100) and (110) facets of a primitive hexagonal lattice structure (Fig. 1b–d). The presence of additional reflections at 9.35 2θ degrees for BTT-DAB, 8.30 2θ degrees for BTT-H1 and 7.28 2θ degrees for BTT-H2 can be assigned to 210 planes, indicating the formation of COFs with a high degree of crystallinity. In addition, the (001) diffraction planes of the three BTT COFs show peaks around 26 2θ degrees, which can be attributed to the weak long-range order of the π-π interaction with an interlayer stacking distance along the c direction. According to the theoretical modeling and Pawley refinement, BTT-DAB, BTT-H1, and BTT-H2 adopt a similar eclipsed AA model within the P6/m (No. 175), P3 (No. 143), and P6/m (No.175) space groups, respectively (Supplementary Fig. 1 and Tables 1–3). After geometry optimization, the PXRD patterns were calculated and compared with the corresponding experimental patterns. The structural models of BTT-DAB, BTT-H1, and BTT-H2 were formulated by creating anticipated two-dimensional hexagonal layers in AA stacking mode (eclipsed) following the hcb topology, which resulted in reasonably low residual values and a satisfactory level of differentiation in the profiles. On the other hand, AB staggered model (staggered) can be eliminated because of the difference between the simulated and experimental PXRD patterns (Supplementary Fig. 2).

The structural integrity of the imine, imine/hydrazone, and hydrazone-linked COFs was confirmed by solid-state 13C cross-polarization magic angle spinning nuclear magnetic resonance spectroscopy (13C CP-MAS NMR) analyses (Fig. 2a). The lower field signal at 157 ppm corresponds to the imine linkage (C=N) of BTT-DAB COF, while the peaks at 156 and 164 ppm were observed due to the presence of the imine (C=N) and the carbonyl (C=O) functionalities of the hydrazone linkages of the BTT-H1 COF. The signature peak at 164 ppm for the hydrazone-linked BTT-H2 COF corresponds to the retention of the asymmetric vibration of the carbonyl (C=O) group in the hydrazone linkage. The structure and construction of the linkages in the COF backbone were further verified by Fourier transform infrared (FTIR) spectroscopy analysis. The FT-IR spectrum of BTT-DAB shows the imine (C=N) stretching at 1600 cm−1, while the C=O stretching of BTT disappears from 1660 cm−1, demonstrating the formation of imine linkages through the Schiff base condensation reaction (Fig. 2b). The retention of the shoulder peaks at ~3312 cm−1 corresponding to the N–H of the secondary amine and the strong peak at 1645 cm−1 corresponding to the C=O of the hydrazone linkages together indicate the presence of secondary amine and carbonyl of the hydrazone linkages in BTT-H1 and BTT-H2 COFs (Fig. 2b, c and Supplementary Fig. 3).

Fig. 2: Characterization of COF-based photocatalysts.
figure 2

a Solid-state 13C CP-MAS NMR analyses of BTT-DAB, BTT-H1 and BTT-H2 COFs showing the presence of signature peaks. b FT-IR analyses of BTT-DAB COF showing the formation of imine COFs. c FT-IR analyses of BTT-H1 and BTT-H2 COFs showing the formation of imine and hydrazone linkages. Nitrogen sorption isotherms of d BTT-DAB, e BTT-H1 and f BTT-H2 COFs showing the formation of porous materials.

Full size image

The porosity of the synthesized donor-acceptor COFs was determined by using N2 sorption analyses performed at 77 K. The N2 sorption isotherm for imine-linked BTT-DAB exhibited a type-I isotherm characterized by distinct steps, whereas BTT-H1 and BTT-H2 showed type-II isotherms. The specific surface area calculated using the Brunauer-Emmett-Teller (BET) model showed the surface areas of 2469, 190, and 95 for BTT-DAB, BTT-H1, and BTT-H2 COFs, respectively (Fig. 2d–f). The partially and fully hydrazone-linked COFs exhibit a lower surface area than analogous imine COFs due to their flexible and free rotation in the framework, inefficient packing and defect formation, which is consistent with similar findings reported previously47,48,49. The alteration of synthetic conditions by modifying the solvent combinations and temperature resulted in the materials with moderate porosity (Supplementary Fig. 4). The pore size distribution profile also suggests that BTT-DAB, BTT-H1 and BTT-H2 are mesoporous in nature, with the pore size increasing from BTT-DAB to BTT-H2, in agreement with the theoretical prediction (Supplementary Fig. 5).

Further, we evaluated the water adsorption properties of COFs, where we observed that the hydrazone-linked BTT-H2 COF exhibited a strong water uptake capacity (more than 3 times) compared to the imine-linked BTT-DAB COF at a lower pressure (Supplementary Fig. 6), indicating the strong influence of water docking sites present in abundance in hydrazone-linked COF for better interaction with water. However, at high-pressure ranges (P/P0 < 0.4), BTT-DAB showed a sharp uptake of water via the capillary effect due to its high surface area50. It is well known that low pressure adsorption is dominated by chemical interactions, whereas high pressure adsorption is dominated by porosity due to the different interactions between the adsorbate and the adsorbent51,52. Field emission scanning electron microscopy (FESEM) analyses showed that BTT-DAB, BTT-H1, and BTT-H2 COFs bear rod-shaped morphologies generated from assembled small crystallites together (Supplementary Fig. 7), which were further supported by high-resolution transmission electron microscopy (HR-TEM) analyses (Supplementary Fig. 8). Thermogravimetric analyses (TGA) of the COFs were performed to analyze the thermal stability of the COFs, where the thermal stability up to 300 °C was observed (Supplementary Fig. 9).

To analyze the optical properties of the imine and hydrazone-linked COFs, the absorbance of the COF samples was screened using solid-state ultraviolet-visible (UV-vis) spectroscopy analyses (Fig. 3a). In the UV-vis diffuse reflectance analyses, the absorption edges of BTT-DAB, BTT-H1, and BTT-H2 were observed in the range of 450–550 nm, whereas the absorbance tail was found to be extended beyond 750 nm, showing the ability to absorb the light in the visible region. The imine-linked BTT-DAB COF showed an optical band gap of 2.26 eV, whereas imine/hydrazone-linked BTT-H1 COF showed an optical band gap of 2.33 eV and hydrazone-linked BTT-H2 COF showed the band gap of 2.47 eV, as confirmed from the Tauc plots (Fig. 3a, Inset).

Fig. 3: Photophysical characterization and photocatalytic H2O2 generation performance.
figure 3

a Solid-state UV-vis absorption spectra of BTT-DAB, BTT-H1 and BTT-H2. Inset shows a Tauc plot for the determination of band gaps of the COFs. b Experimentally derived energy band alignments of BTT-DAB, BTT-H1 and BTT-H2 COFs compared to oxygen reduction and oxygen evolution potentials (vs. NHE at pH 6.8) along with other reactions. c UPS analyses of BTT-DAB and BTT-H2 COFs. d The photocurrent responses of BTT-DAB and BTT-H2. e EIS Nyquist plots of BTT-DAB and BTT-H2 COFs. f Photoluminescence spectra of BTT-DAB and BTT-H2 COFs. g Time-dependent photocatalytic H2O2 generation of BTT-DAB, BTT-H1 and BTT-H2 COF in pure water under visible light (λ = 467 nm). Error bars indicate the error in the measurement. h Photocatalytic H2O2 activity of BTT-DAB, BTT-H1 and BTT-H2 COF in the presence of different sacrificial electron donors. (Inset: A two-phase system of water-benzyl alcohol). i Recyclability analyses for BTT-DAB and BTT-H2 for over five consecutive cycles (2 h each) for photocatalytic H2O2 generation.

Full size image

Further, the band positions of BTT-DAB, BTT-H1 and BTT-H2 were determined by combining the optical band gaps determined using Tauc plots and reduction potentials from Mott-Schottky plots (Fig. 3b and Supplementary Fig. 10). The conduction bands (CB) of BTT-DAB, BTT-H1 and BTT-H2 were estimated to be −0.52, −0.61, and −0.99 eV [vs. normal hydrogen electrode (NHE)], respectively. The BTT-H2 COF exhibited the maximum negative value of the conduction band potential of −0.99 eV (vs. NHE, pH = 6.8), indicating the superior photoreduction ability of the hydrazone sites in the COF matrix. It is a very crucial property as it assists in reducing molecular oxygen to the superoxide radical (•O2-, dashed pink line, −0.35 V), thereby promoting the oxygen reduction reaction (ORR) to generate H2O253,54. It is, therefore, could be assumed that the generation rate of superoxide radical anions may increase as the density of hydrazone linkages increases. Furthermore, UPS analyses executed to evaluate the energy level of the valence band maximum precisely showed the valence band position at 1.71 eV and 1.51 eV (vs. NHE) for BTT-DAB and BTT-H2, respectively, which were also found to be in agreement with the Mott-Schottky analyses (Fig. 3c). The valence band levels of the BTT-DAB and BTT-H2 COFs were found to be higher than the thermodynamic potential for the direct 2e− water oxidation reaction (WOR) (1.36 V vs. NHE, at pH = 6.8), indicating a thermodynamically feasible overall photocatalytic H2O2 generation. In addition, the positive valence band values of the as-synthesized COFs indicate the favorable four-electron (4e−) water oxidation (dashed black line, +0.83 V).

To analyze the charge carrier generation upon visible light irradiation, the photocurrent density measurements were performed. These analyses proved that the hydrazone linkages benefit the ultrafast transport of charge carriers due to the accumulation of p-orbital population density of electrons near the valence band edge (Fig. 3d). The semicircle radius in the Nyquist plot obtained from the electrochemical impedance spectroscopy (EIS) analyses suggested that the hydrazone-linked BTT-H2 COF has a reduced resistance, and an enhanced charge migration ability as compared to the imine-linked BTT-DAB COF, which are ideal properties for a high-performance photocatalyst (Fig. 3e). Furthermore, photoluminescence (PL) measurements showed the lower peak intensity for BTT-H2 as compared to the BTT-DAB COF, suggesting the possibility of lower electron-hole recombination in the hydrazone-linked COFs (Fig. 3f). Form these photophysical analyses, it is clear that hydrazone-linked COFs possess enhanced charge carrier mobility assisted by efficient charge separation and reduced recombination of electron-hole in comparison with the imine-linked COFs.

Photosynthesis of hydrogen peroxide (H2O2)

Considering the crystallinity, favorable optical band gap and band positions, the imine and hydrazone-linked COFs were used for photocatalytic H2O2 generation using abundant natural air, water and sunlight under ambient conditions. The photocatalytic generation of H2O2 was carried out using pure water and atmospheric oxygen, which is a simple, economical, and sustainable approach. In typical photocatalytic experiments, the imine and hydrazone-based COFs were dispersed in ultrapure water, followed by irradiation with a 40 W blue LED (λ = 467 nm). The reaction progress was monitored over time intervals of 15, 30, 60, 90, and 120 min, each constituting a separate reaction to precisely evaluate the photocatalytic H2O2 activity as a function of time (Fig. 3g). Firstly, the formation of H2O2 was detected using a peroxide test kit and further quantified by UV-Vis spectroscopy using the titanium(IV) oxsulphate sulfuric acid solution55. The time-varying experiment showed that the photocatalytic rate followed a linear variation with time when scanned over 120 min. The BTT-H2 COF with both-sided hydrazone linkages showed H2O2 generation as high as 1359 μmol g−1 h−1, which was superior to the imine/hydrazone-based BTT-H1 COF (1096 μmol g−1 h−1), while the corresponding H2O2 generation for the both-sided imine-linked BTT-DAB was found to be 854 μmol g−1 h−1 (Fig. 3g). We believe that the greater photoreduction potential and photocurrent density of the hydrazone-based BTT-H2 COF account for their superior photocatalytic performance (Fig. 3b, d). This can also be attributed to the improved interaction of water and better oxygen uptake tendency of the hydrazone linkage compared to the imine linkage56.

When pure water was used under atmospheric conditions (air), the H2O2 generation rate of the BTT-H2 COF (1359 μmol h−1 g−1; 40 W Xenon lamp; 2 h) was found to exceed many benchmark heterogeneous photocatalysts, including g-C3N4 and other reported COFs, by a factor of several due to the increasing density of hydrazone bonds in the COF backbone (Supplementary Table 5). In exceptional cases, high values for H2O2 generation were achieved either by using a sacrificial agent26, bubbling of pure O257, or utilizing high intensity lights58 (Supplementary Table 5). The photocatalytic H2O2 generation using BTT-DAB, BTT-H1 and BTT-H2 COF was also screened using different sacrificial electron donors such as ethyl alcohol (EtOH) and benzyl alcohol (BA) (Fig. 3h). A slight increase in H2O2 yield was observed when a 10% aqueous solution of EtOH was used. However, in the presence of 10% aqueous BA, a significant increase in the yield of H2O2 was observed, which was approximately 15 times higher. This significant improvement can be attributed to the phase separation that occurs in the presence of BA, where the COF remains in the organic phase while the H2O2 remains in the aqueous phase and is protected from degradation due to interference with the COF photocatalyst (Fig. 3h inset). As the hydrazone-linked BTT-H2 COF bears a VB edge located at 1.51 V (obtained from UPS analysis), higher than the potential required for oxidation of benzyl alcohol to benzaldehyde (1.98 V vs. NHE), the formation of benzaldehyde from benzyl alcohol is inhibited during the photosynthesis of H2O2. By taking advantage of the phase separation, the implementation of a biphasic water-benzyl alcohol system promotes the unhindered generation of H2O2. The H2O2 produced dissolves in the aqueous phase (pure water), while the COF photocatalyst remains in the organic phase (BA), which can then be rejuvenated for H2O2 synthesis59. This strategy could be promising for the continuous photosynthesis of H2O2 and may provide a bridge from the laboratory to the industry.

The recyclability of the BTT-DAB and BTT-H2 COF photocatalysts was evaluated over five consecutive runs, which showed better recyclability with a drop of ~11% in the photocatalytic activity of the hydrazone-linked BTT-H2 COF compared to the imine-linked BTT-DAB COF that exhibited a drop of ~26% (Fig. 3i). Additionally, the long-term photocatalytic experiment for H2O2 generation was performed over a period of 12 h. The amount of H2O2 produced by the hydrazone-linked BTT-H2 COF increased steadily over 8 h, which however drops down to some extent after 8 h of irradiation due to the equilibrium between H2O2 formation and decomposition that is commonly observed in H2O2 photosynthesis (Supplementary Fig. 12)60,61. Fascinatingly, when the BTT-H2 COF was recovered after 8 h of photocatalysis and used for a fresh photocatalytic reaction, the H2O2 activity of 1198 µmol g−1 h−1 was observed, suggesting that the hydrazone-linked BTT-H2 COF could be reused with considerable activity. As confirmed from the PXRD, FT-IR, porosity and morphological analyses before and after the photocatalytic H2O2 generation reaction, the crystallinity, structural integrity, surface area and morphology of BTT-H2 were found to be retained, confirming the durability of BTT-H2 for H2O2 generation (Supplementary Fig. 11).

In the photocatalytic H2O2 generation process, the H2O2 produced can inevitably react with electrons (e-) or holes (h+) and can further decompose into hydroxyl radicals (•OH) or superoxide radicals (•O2-), leading to the generation of water or O2, respectively. Therefore, the dynamics of H2O2 formation and decomposition were studied using hydrazone-linked BTT-H2 COF, as both phenomena are expected to occur simultaneously during the process of photocatalysis (Supplementary Fig. 13). In these analyses, it was observed that the formation of H2O2 follows the zero-order kinetics, whereas the decomposition of H2O2 follows the first-order kinetics62,63. Hence, the overall H2O2 generation can be stated as follows:

$$[{{{H}}}_{2}{{{O}}}_{2}]={K}_{{{f}}}/{K}_{{{d}}}\times \left\{1\,-{exp}(-{K}_{{{d}}}\times t)\right\}$$
(1)

where t is the reaction time and [H2O2] is the concentration of produced H2O2, and Kf and Kd are the formation and decomposition rate constants, respectively. The linear fitted data determines the Kf to be 421 mM h−1, and the Kd values obtained at 298 K and 313 K were 0.03 h−1 and 0.05 h−1, respectively. Therefore, the amount of H2O2 increased for a certain period, after which the amount of H2O2 decreased (in the long-term photocatalytic reaction) due to the equilibrium between H2O2 formation and decomposition.

The photocatalytic H2O2 generation experiments suggest that the photocatalytic activity of H2O2 generation can be increased by increasing the density of hydrazone linkages in the COF framework. To validate the impact of hydrazone linkages on photocatalytic H2O2 generation, we targeted the synthesis of hydrazone-linked COF with the maximum density of hydrazone linkages. To achieve the synthesis of such hydrazone-linked COF (BTT-H3), the condensation reaction of Benzo[1,2-b:3,4-b’:5,6-b”] trithiophene-2,5,8-tricarbaldehyde (BTT) and benzene-1,3,5-tricarbohydrazide (H3) was carried out under solvothermal conditions (Fig. 4a). Solid-state 13C CP-MAS NMR analyses showed the disappearance of the carbonyl -C=O chemical shift at 190 ppm for the BTT precursor and the appearance of a peak at 164 ppm for the carbonyl groups of the hydrazone linkage, demonstrating the formation of the desired BTT-H3 COF (Fig. 4b). The PXRD, FT-IR and porosity analyses further confirmed the formation of moderately crystalline and porous BTT-H3 COF (Supplementary Fig. 14). Although BTT-H3 COF showed relatively low crystallinity compared to BTT-H1 and BTT-H2, it exhibited a lower optical band gap aided by a more negative reduction potential of −1.2 V and a higher density of catalytic sites for water oxidation and oxygen reduction reactions (Supplementary Fig. 14h). As expected, based on the desired properties such as high density of catalytic sites, low optical band gap, large negative reduction potential, BTT-H3 COF showed the highest photocatalytic H2O2 generation rate with a rate as high as 1588 μmol g−1 h−1 over 2 h compared to the BTT-H2 COF which has a relatively lower density of hydrazone linkages (Fig. 4c). The fluorescence lifetimes of the high-density hydrazone-linked BTT-H3 COF (1.83 ns) were higher than those of the relatively low-density hydrazone-linked BTT-H2 COF (1.52 ns) or the imine-linked BTT-DAB COF (1.60 ns), indicating that BTT-H3 COF can suppress e−-h+ recombination efficiently (Fig. 4g and Supplementary Fig. 15).

Fig. 4: Synthesis, characterization and photocatalytic H2O2 generation performance of BTT-H3 COF.
figure 4

a Scheme of synthesis of BTT-H3 COF with highly dense hydrazone linkages. b Solid-state 13C CP-MAS NMR analyses showing the formation of BTT-H3 COF. c Time-dependent photocatalytic H2O2 generation of BTT-H3 COF in pure water under visible light (λ = 467 nm). Error bars indicate the error in the measurement. d Photocatalytic H2O2 generation using BTT-H3 with variable light intensity in pure water showing the best performance at 467 nm. Error bars indicate the error in the measurement. e Photocatalytic H2O2 generation using BTT-H3 under different atmospheres (Air, O2, N2 and Ar), in pure water under visible light (λ = 467 nm). Error bars indicate the error in the measurement. f Photocatalytic H2O2 activity of BTT-H3 in the presence of different scavenging agents, under visible light (λ = 467 nm). Error bars indicate the error in the measurement. g Fluorescence lifetime decay profile of BTT-H2 and BTT-H3 COFs. h Photograph of COF dispersed water exposed to sunlight showing the generation of significant amounts of H2O2.

Full size image

Furthermore, the photocatalytic H2O2 activity of the BTT-H3 COF evaluated using different light intensities showed better performance at an intensity of 467 nm (Fig. 4d). The apparent quantum yield (AQY) was found to be 17.7% observed at 467 nm (Supplementary Fig. 16), and the solar-to-chemical energy conversion (SCC) was 2.02%, which is higher than the solar-to-biomass conversion efficiency in plants (~0.1%)64. It is fascinating to note, when the BTT-H3 COF dispersed water was exposed to sunlight for 1 h, a significant amount of H2O2 (550 µmol g−1 h−1) was also generated (Fig. 4h). This provides evidence that the readily available renewable resource of sunlight is sufficient for producing a significant amount of H2O2. In addition, we find that BTT-H3 COF can efficiently generate H2O2 in a wider range of available real-world water samples, including RO water, tap water, river water and seawater (Supplementary Fig. 17). Surprisingly, the H2O2 yield from seawater was as high as 1380 µmol g−1 h−1, showing the utility of COFs for practical applications.

Mechanism of photocatalytic H2O2 generation

Understanding the mechanism of photocatalytic H2O2 generation using COFs as photocatalysts is crucial for optimizing their efficiency and stability. Understanding the reaction pathways and active sites within COFs allows targeted modifications to enhance their photocatalytic performance, which is essential for sustainable hydrogen peroxide production and its applications in green chemistry and environmental remediation. In order to analyze the role of oxygen in the photocatalytic H2O2 generation, the H2O2 photosynthesis experiment was carried out using oxygen-saturated water. As a result, it was observed that the rate of H2O2 generation using BTT-H3 COF was increased up to 1770 μmol g−1 h−1 (Fig. 4e). However, when the water was purged with nitrogen (N2), a significant drop in the H2O2 yield was observed, suggesting the necessity of oxygen during the H2O2 generation reaction. The significant drop in H2O2 generation under argon (Ar) purging also emphasized the crucial involvement of oxygen in the photocatalytic oxygen reduction reaction.

In general, in the first step of photocatalytic H2O2 generation, the molecular oxygen accepts electrons generated from the photocatalyst under light irradiation to form the superoxide radical anion (•O2-) through the oxygen reduction reaction (ORR) pathway. Further, in the stepwise ORR route, the •OOH intermediate is formed by the reaction of •O2- with H+ ion. Finally, *OOH intermediate reacts with an electron followed by an H+ ion to generate H2O2. To validate this mechanism, we introduced a scavenger molecule, p-benzoquinone (p-BQ; 10 mM aqueous solution), which is selective for scavenging the superoxide radical anion, into the reaction mixture, which resulted in a significant reduction in the rate of H2O2 generation (Fig. 4f)41,65. The photoreduction ability of the acceptor sites of the hydrazone linkage in the BTT-H3 COF mediates this process. Additionally, the poor H2O2 generation in the presence of the electron scavenger AgNO3 (10 mM aqueous solution) confirmed the involvement of photogenerated one-electron reduction in the H2O2 generation (Fig. 4g). Furthermore, the addition of tertiary butanol, a scavenger of hydroxyl radicals (•OH), hardly affected the H2O2 production rate, suggesting that the •OH radical does not contribute to the H2O2 generation reaction66,67. Similar characteristics were also observed for imine-linked BTT-DAB, and hydrazone-linked BTT-H1 and BTT-H2 COFs (Supplementary Fig. 18).

Rotating ring-disk electrode (RRDE) experiments were conducted to determine the number of electron transfers involved in the ORR during the photosynthesis of H2O2. The average number of electron transfers in the ORR using BTT-H2 and BTT-H3 COFs was found to be 2.64 and 2.28, respectively, according to the RRDE analyses (Supplementary Fig. 19), confirming the reduction of O2 to H2O2 taking place via a 2e− ORR process. In addition, the RRDE experiments were also used to investigate the WOR (2e− WOR and 4e− WOR) pathways for the hydrazone-linked COFs, which suggested that both the hydrazone-linked BTT-H2 and BTT-H3 COFs generate O2 via the 4e− WOR process (Supplementary Fig. 20), consistent with reported COF-based photocatalysts38,58. It is worth mentioning that in the WOR pathway towards H2O2 generation, the generated O2 molecules can participate in the production of H2O2 via the 4e−–2e− cascaded process (H2O → O2 → H2O2), which could be more thermodynamically favorable and also can avoid the catalyst poisoning. Electron paramagnetic resonance (EPR) was assessed to further clarify the role of the O2•- intermediate involved in ORR. The detection of the signal of DMPO-O2•- in the EPR spectra confirmed the generation of O2•- by hydrazone-linked COFs under visible-light irradiation (Supplementary Fig. 21). We believe that the highly conjugated hydrophilic π-columnar arrays in the hydrazone-enriched BTT-H3 COF provide ample catalytic sites for both the water oxidation reaction and the oxygen reduction reaction—demonstrating the efficient photocatalytic H2O2 generation.

Computational studies

To unravel the interaction of water molecules with the COF backbone and the influence of linkers at the microscopic level, the density functional theory (DFT) and molecular dynamics simulations were carried out. In the first step of the analysis, DFT was applied to derive the optimal geometries of the COFs. The calculations converged to stable geometries for both the monolayer and stacked bulk phases (Supplementary Fig. 22). Computed local potentials for hydrazone-linked BTT-H3 COF, which have the pores inlaid with arrays of proton acceptors/donors, showed a high degree of charge separation as compared to imine-linked BTT-DAB as the hydrazone linkages are electron deficient and the benzotrithiophene (BTT) units are electron rich (Supplementary Fig. 23a, b). The computational study thus confirms the presence of intramolecular charge transfer (IRCT) channels in hydrazone-linked COFs, which assist in producing free charges and facilitating exciton dissociation and charge separation. The binding of water molecules at the hydrophilic knots formed by hydrazone linkages promotes the water oxidation reaction alongside the oxygen reduction reaction, influencing the interaction with dissolved oxygen throughout the catalytic framework46. In addition, the importance of the BTT unit for photocatalytic H2O2 generation was compared with an analogous C3 symmetric molecule with sulfur atoms substituted by carbon. However, the presence of sp3 carbon atom in 4,7-dihydro-1H-cyclopenta[e]-as-indacene-2,5,8-tricarbaldehyde molecule induces non-planarity (out-of-plane hydrogen atoms) in such structures due to the interlayer stacking for the formation of COFs (Supplementary Fig. 24). Therefore, it is expected that the photocatalytic activity of 4,7-dihydro-1H-cyclopenta[e]-as-indacene-2,5,8-tricarbaldehyde molecule with carbon atom should be much lower compared to BTT with sulfur atom.

The high electron affinity of the hydrazone indicates its proton-donating potential to act as an efficient catalytic center for oxygen reduction/water oxidation. Therefore, to understand the interaction of water and dioxygen with this catalytic center, classical molecular dynamics (MD) simulations were carried out by placing the COF fragment in an explicit water environment68,69. Peaks at hydrogen bonding distances (~2–3 Å) indicated bound water; the higher intensity of which indicates stronger binding, i.e., the presence of an immobile water layer. In imine-linked BTT-DAB COF, the C=N (imine) binding sites are present, which bind to water with notably less strength compared to the hydrazone linkages present in BTT-H3 COF (Fig. 5a).

Fig. 5: Computational studies of the photocatalytic H2O2 generation.
figure 5

Radial distribution (obtained from classical molecular dynamics simulation) of water with respect to: a N center of BTT-DAB and b O centers of BTT-H3. c Band structure and projected density of states (PDOS) for BTT-H3. d Molecular mechanism of the generation of H2O2 from oxygen and water at the hydrazone center of the BTT-H3 COFs. e Electronic energy profile of the hydrogen peroxide generation reaction path. f Spin densities (semitransparent light green blobs) of the radical species. g Illustration of the overall photocatalytic mechanism of H2O2 generation via oxygen reduction reaction and water oxidation reaction.

Full size image

On the one hand, in BTT-H3 COF, the carbonyl-oxygen forms a highly robust hydrogen bond with water (Fig. 5b). On the other hand, the -NH group of the hydrazone linkages acts as a hydrogen bond donor, forming hydrogen bonds with water, and the C=N- of the hydrazone linkages acts as an acceptor of hydrogen bonding (Supplementary Fig. 23c). These theoretical findings were further validated through contact angle measurements with water. In contact angle measurements for BTT-H3, the strong interaction (hydrophilicity) of water with hydrazone linkages was observed as compared to imine-linked BTT-DAB COF, which is advantageous for photocatalytic H2O2 generation from water (Supplementary Fig. 23d). The calculated density of states (DOS) showed a higher density of p-orbital electrons near the valence band edge for the BTT-H3 COF, which renders the high electron-donating capability in highly dense hydrazone-linked COFs (Fig. 5c). High p-orbital electron density enhances transport by creating delocalized states, which leads to the enhanced light absorption capabilities in the UV-visible range and makes them better at harvesting solar energy to drive photocatalytic reactions. The extended π-system promotes delocalization, lowering the chances of electron-hole recombination, which is detrimental to photocatalysis. The enhanced p-orbital electron density in hydrazone-linked BTT H3 COF enables improved mobility of electrons and holes as compared to imine-linked BTT-DAB COF (Supplementary Fig. 25). We thus conjecture that the amalgamation of these promising features in a single COF-based photocatalyst, namely BTT-H3, could make it possible to perform efficient photosynthesis of H2O2 with abundant water and air utilizing natural sunlight.

Despite the inherent limitations of computational studies, especially when applied to idealized systems that may not capture all details of the experimental setup, the computational study provides invaluable insights into experimental observations. Having understood the molecular mechanism of the generation of H2O2 from oxygen and water at the hydrazone center of the BTT-H3 COF in the overall photocatalytic H2O2 generation, we proceeded to study the energy profile of the potential mechanistic pathways for the H2O2 generation reaction. For this purpose, we used DFT calculations70, employing the MD inputs on the distribution of water and oxygen (reactants) around the catalytic sites of the BTT-H3 COF. In the following, we consider a few possibilities and judge their feasibility. In the first possibility, we consider the reaction to begin with water splitting and the uptake of H• or proton-coupled electrons by the COFs upon photoexcitation, generating one H2O2 molecule. In the subsequent step, upon relaxation, the H• can either combine to form H2 gas or react with dissolved oxygen to form another H2O2 molecule. The electronic energy profile of this reaction pathway is presented in Supplementary Fig. 26, which involves a high energy barrier. The computed H2O/H2O2 potential, exceeding 3 eV, is higher than the theoretical band gap energy of the COF, making it thermodynamically unfavorable. The excited-state acid-base properties of the COF may rationalize this observation. The ground state charge separation in the COF and the accumulation of electron density in the hydrazone moiety are shown in Supplementary Fig. 23a. The higher electron affinity suggests that hydrazone linkages are more likely to release a proton than accept one. Analysis of the frontier molecular orbital (HOMO) reveals significant electron density over the -NH- bond in the ground state (Supplementary Fig. 23e, f). In contrast, the transition orbital shows that the electron density shifts away from the -NH- group, weakening the N–H bond in the excited state. At the same time, the electron density on the -C=O and -C=N groups increases, making these functional groups less likely to accept protons in the excited state.

The increased acidity of the hydrazone linkages of BTT-H3 COF in the excited state implies an alternative possible pathway for H2O2 generation. This pathway initiates with excited-state electron-coupled proton transfer or hydrogen atom transfer from the COF to dioxygen, directly producing hydroperoxide radical (•OOH) or through a superoxide (•O2-) intermediate. This •OOH radical then interacts with another hydrogen atom from the adjacent hydrazone moiety, producing one molecule of H2O2 (Fig. 5d). Upon photoexcitation, one proton is transferred from hydrazone linkages of BTT-H3 COF to dissolved molecular oxygen, producing a radical (I → II). Then, another proton is transferred from the adjacent hydrazine moiety to •OOH, producing one H2O2 (II → III). Thus, two proton vacancies are generated in the BTT-H3 COF (III). BTT-H3 COF readily uptakes protons from the hydrogen-bonded water, generating another molecule of H2O2 (III → V) and regenerating the BTT-H3 COF photocatalyst. The computed reduction potential is 2.01 eV, which is compatible with the band gap energy of the COF, rendering it thermodynamically favorable (Fig. 5e). This process creates two hydrogen vacancies in the COF, driving water oxidation and the production of another H2O2 molecule, thus completing the catalytic cycle. The free energies of the radical intermediates along the reaction pathway were discovered to be lower than those of the ionic intermediates (in step II, •OOH formation is energetically more favorable than HOO–), indicating a radical mediated reaction, a conclusion supported by experimental evidence. Additionally, computed spin densities localized on the radical intermediates further confirm the presence of superoxide anion radical (•O2-) species, suggesting a stepwise oxygen reduction pathway (Fig. 5f). Hence a stepwise one-electron transfer process—a more favored pathway for photocatalytic H2O2 generation—is a less energy-demanding reaction (Fig. 5g).

In summary, we synthesized a series of hydrazone-linked COFs such as, BTT-H1, BTT-H2, and BTT-H3 with varying hydrazone linkages and investigated their efficacy in photocatalytic H2O2 generation from pure water under air. Our findings underscore the critical role of hydrazone linkages in COFs for efficient H2O2 photosynthesis, surpassing the performance of imine-linked BTT-DAB COF due to enhanced water interaction and oxygen uptake tendencies for hydrazone COFs. Notably, BTT-H3 COF exhibited exceptional photocatalytic H2O2 generation (1588 µmol g−1 h−1) without external sacrificial electron donors. Furthermore, it demonstrated significant H2O2 production (550 µmol g−1) under sunlight exposure for 1 h, showcasing its potential for renewable resource utilization. Using molecular simulations, we have demonstrated that the π-columnar arrays in BTT-H3 COF provided abundant catalytic sites for water oxidation and oxygen reduction reactions. The DFT calculations and detailed MD simulations elucidated the charge distribution on hydrazone moieties and their affinity for water and oxygen molecules, supporting our experimental findings. Possible reaction pathways have been suggested. We believe that our research contributes to the development of high-performance COF-based photocatalysts for sustainable H2O2 generation from water and atmospheric oxygen, driven by natural solar light, offering promising prospects in materials science and renewable energy.

Methods

Preparation of BTT-DAB COF

Benzo[1,2-b:3,4-b’:5,6-b”] trithiophene-2,5,8-tricarbaldehyde (BTT, 66 mg, 0.2 mmol) and 4-diaminobenzene (DAB, 33 mg, 0.3 mmol) were taken in a 10 mL Pyrex tube, and dissolved in ortho-dichlorobenzene (o-DCB, 3 mL) and n-butanol (n-BuOH, 3 mL) along with the addition of 6 M acetic (AcOH, 0.5 mL) acid as catalyst. The reaction mixture in the Pyrex tube is sonicated for 10 min. The highly homogeneous reaction mixture is made to undergo three freeze-pump-thaw cycles. The sealed tubes containing the organic linker monomers dispersed in solvents are now heated at 120 °C for 3 days. After cooling to room temperature, the dark, orange-colored powder was filtered using acetone, methanol, tetrahydrofuran, and hexane.

Preparation of BTT-H1 COF

Benzo[1,2-b:3,4-b’:5,6-b”] trithiophene-2,5,8-tricarbaldehyde (BTT, 50 mg, 0.15 mmol) and 4-aminobenzohydrazide (H1, 33 mg, 0.22 mmol) were taken in a 10 mL Pyrex tube, and dissolved in DMSO (1.5 mL) and DMAC (0.5 mL) along with the addition of 6 M acetic acid (0.7 mL) as catalyst. The reaction mixture in the Pyrex tube is sonicated for 10 min. The highly homogeneous reaction mixture is made to undergo three freeze-pump-thaw cycles. The sealed tubes containing the organic linker monomers dispersed in solvents are now heated at 150 °C for 3 days. After cooling to room temperature, the dark, yellow-colored powder was filtered using acetone, methanol, tetrahydrofuran, and hexane.

Preparation of BTT-H2 COF

Benzo[1,2-b:3,4-b’:5,6-b”] trithiophene-2,5,8-tricarbaldehyde (BTT, 50 mg, 0.15 mmol) and terephthalic dihydrazide (H2, 45 mg, 0.22 mmol) were taken in a 10 mL Pyrex tube, and dissolved in DMSO (0.5 mL) and DMAC (1.5 mL) along with the addition of 6 M acetic acid (0.7 ml). The reaction mixture in the Pyrex tube is sonicated for 10 min. The sealed tubes containing the organic linker monomers dispersed in solvents are now heated at 120 °C for 3 days. After cooling to room temperature, the light, yellow-colored powder was filtered using acetone, methanol, tetrahydrofuran, and hexane.

Procedure of the photosynthesis of H2O2

The experimental procedure includes taking an optimum dosage of the COF photocatalyst (5 mg) dispersed in 10 mL pure water by ultrasonication for 15 min. Excessive use of the catalyst is avoided since it may decompose the H2O2 generated. For the oxygen, nitrogen and argon atmosphere, the solution was generally bubbled continuously with pure oxygen gas, nitrogen and argon, respectively. For controlled experiments, different sacrificial agents (e.g., ethanol, isopropyl alcohol and benzyl alcohol) and scavengers (para-benzoquinone, tert-butanol, AgNO3) were used. The reaction mixture was allowed to stir in the dark for 30 min to reach the absorption-desorption equilibrium. The system was irradiated with a xenon lamp source (40 W) by maintaining the reaction temperature around 25 °C. After sampling at a given time interval, the solution is withdrawn with a syringe, centrifuged, and then filtered (using a 0.22 μm filter) to remove the COF photocatalyst further. The formation of H2O2 was detected by a peroxide test kit and was quantified by UV-Vis spectroscopy with the Titanium (IV) oxsulfate-sulfuric acid solution.

Detection of hydrogen peroxide

The aqueous Ti reagent solution was prepared by diluting Titanium (IV) oxysulfate-sulfuric acid hydrate (TiOSO4.xH2O + H2SO4) in sulfuric acid. The colorless acidic solution of aqueous titanium oxysulfate reacts with H2O2, resulting in the formation of a yellow peroxotitanium complex [Ti(O2)OH(H2O)3]+(aq), the absorbance of which was measured at 412 nm. The higher the intensity of coloration, the greater the concentration of H2O2 generated, according to the Lambert-Beer Law. The reactions for the detection of H2O2 through titanium sulfate colorimetry are as follows.

$${{{TiO}}}{{{{SO}}}}_{4}(s)+5{{{H}}}_{2}{{O}}\to {\left[{{{Ti}}}{\left({{{OH}}}\right)}_{3}{\left({{{H}}}_{2}{{O}}\right)}_{3}\right]}^{+}({aq})+{{{{HS}}}{{{O}}}_{4}}^{-}({aq})$$
(2)
$${[{{{Ti}}}{({{{OH}}})}_{3}{({{{H}}}_{2}{{O}})}_{3}]}^{+}({aq})+{{{H}}}_{2}{{{O}}}_{2}({aq})\to {[{{{Ti}}}({{{O}}}_{2}){{{OH}}}{({{{H}}}_{2}{{O}})}_{3}]}^{+}({aq})+{2{{H}}}_{2}{{O}}$$
(3)

The concentration of the produced H2O2 can be calculated from the calibration curve established by the absorbance intensity and the known concentration of the H2O2 solution. A plot of absorbance versus concentration was made into a linear fit graph, obtaining a certain slope and intercept value. The slope and intercept values were used to calculate the concentration of H2O2 generated.

Recyclability experiment of COF-based photocatalyst

After 2 h of photocatalytic reaction, the COF photocatalyst was centrifuged, washed with deionized water, and dried. Then, the resulting photocatalyst was reused for the next photocatalytic reaction. The same cycle was repeated over five reactions using the recycled COF photocatalyst each time over 2 h.

Kinetics of H2O2 formation and decomposition

The dynamics of photocatalytic H2O2 formation and decomposition were studied by dispersing hydrazone-linked BTT-H2 COF (15 mg) in a 1 mM aqueous solution of H2O2 (50 mL) under continuous O2 and N2 bubbling using a Xenon lamp (40 W).