Introduction

In heterogeneous catalysis, the nature and arrangement of active sites critically dictate catalyst performance by governing the recognition, activation, and transformation of reactants1,2,3. Conventional catalyst design typically relies on single-component active sites, such as metal nanoparticles or oxide surfaces, which are highly effective for isolated chemical processes or simple reactions4,5. However, in complex redox reactions, where reactants undergo multiple, sequential oxidation-reduction steps, single active sites always fail to accommodate the diverse mechanistic requirements6. Moreover, their catalytic efficiency is frequently limited by inherent geometric and electronic constraints7,8. In contrast, many natural systems, including photosynthesis and biodegradation, exploit the cooperative interplay of multiple active sites to achieve efficient, multi-step conversions9,10. Such system provides compelling inspiration for the development of artificial catalysts that harness synergistic interactions with multiple active centers.

The design of artificial catalysts featuring multiple active sites remains a significant challenge due to the minute spatial scale and challenging operability. Conventional strategies, such as heterostructure engineering and metal loading11,12, seek synergy by combining metals, oxides, or organic ligands through simple mixing or bonding. Although these approaches have modestly enhanced catalytic performance, they always suffer from uneven active site distribution and limited control over synergistic interactions13. As catalytic reactions occur at the nanoscale interface between catalyst and reactants, precise control of active sites at the atomic and molecular levels offers a route to achieve the cascade and control of complex redox reactions, driving functional catalytic materials from macroscopic composites toward microscopic manipulation.

Molecular junction catalyst, assembled by molecular structure composed of multiple functional groups, such as oxidative and reductive groups, represents an ideal platform for designing artificial catalysts with multiple active sites14,15. With exceptional design flexibility, it enables precise modulation of catalytic activity, selectivity, and efficiency through tailored adjustments to the redox center and linker chemistry, and their modular architecture allows for optimization across a wide range of reactions, enhancing performance metrics such as turnover frequency and reaction rates16,17,18. Specifically, designing redox centers at the molecular level enables controlled regulation of active sites, particularly in constructing their spatial distribution (Fig. 1a). This molecular junction catalyst facilitates highly efficient coordination of oxidation and reduction reactions at the molecular level, avoids interference between active sites, and enhances the selectivity and efficiency of complex reactions.

Fig. 1: Concept of F substituent coordination inducing oxidation-reduction molecular organic junction to promote carrier separation.
Fig. 1: Concept of F substituent coordination inducing oxidation-reduction molecular organic junction to promote carrier separation.
Full size image

a Separated oxidation-reduction centers promoting catalytic reaction with carrier separation at the dual active site. D and A: Donor and acceptor molecule. D+ and A−: Donor cation and acceptor anion. b The proposed structures of oxidation-reduction sites. c Gradual increase in the number of F substitutions to tune the oxidation-reduction sites of CTFs.

Inspired by these, here we modulated the substituent coordination number of covalent triazine framework (CTF), a representative molecular organic junction, to manipulate the redox properties of adjacent triazine rings, thereby promoting a typical heterogeneous catalytic reaction—hydrogen peroxide (H2O2) interfacial synthesis via oxygen reduction reaction (ORR) and water oxidation reaction (WOR) at independent oxidation and reduction centers. In particular, CTFs and fluorine-substituted CTFs (CTF-TF) were synthesized via a low-temperature approach based on our previous reports19. By precisely modulating the number of different precursors of triazine, benzene, and F-substituted benzene, CTF-1, CTF-TF-0.5, CTF-TF were obtained, and the coordination number of F-substituted benzene linkers to affect the redox nature of the triazine ring in CTFs, thereby presenting separable oxidation-reduction assignment in the conjugated skeleton of CTFs. Based on theoretical calculations, we found that the triazine connected a benzene and two F-substituted benzene exhibits as reduction site, while the triazine connected two benzene and a F-substituted benzene exhibits as oxidation site (Fig. 1b). With a gradual increase in the number of F substitutions in CTFs, only CTF-TF-0.5 exhibits a clear separation of oxidation-reduction assignment, relative to CTF-1 and CTF-TF (Fig. 1c). Under external forces, the photogenerated-holes allow to transfer to the oxidation site and photogenerated-electrons to the reduction site via benzene ring bridges. Then, O2 and H2O are adsorbed at the N active sites of triazine connected a benzene and two F-substituted benzene and triazine connected two benzene and a F-substituted benzene in CTF-TF-0.5 for H2O2 synthesis at the gas-liquid-solid triphasic interface. This separable oxidation-reduction assignment in the molecular junction catalyst enables the cascade and control of complex redox reactions at the molecular level, and provides opportunities for designing functional artificial catalysts with multiple active sites.

Results

Spatial distribution of oxidation-reduction sites in CTFs

We began to explore the spatial distribution of oxidation-reduction sites of all CTFs by using density functional theory (DFT) simulations. For catalytic reactions, the energy band positions of highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) determine the oxidation and reduction capacities, and their spatial distribution determines the distribution of oxidation-reduction site on the catalyst molecular20. As shown in Fig. 2a, CTF-TF-0.5 exhibited a clear separated spatial distribution of HOMO and LUMO. Specially, its HOMO is localized at the triazine connected two benzene and a F-substituted benzene (N1), while LUMO is localized at the triazine connected a benzene and two F-substituted benzene (N2). In contrast, both CTF-1 and CTF-TF showed irregular and overlapping spatial distributions of LUMO and HOMO. This difference in HOMO-LUMO spatial distribution indicated that CTF-TF-0.5 showed a clear spatial distribution of oxidation-reduction assignment due to the modulation of F-substituted number, which is further confirmed through electron-hole distribution simulation with time-dependent DFT calculations (Fig. S1).

Fig. 2: Theoretical simulation of oxidation-reduction sites in CTFs with F substituent coordination.
Fig. 2: Theoretical simulation of oxidation-reduction sites in CTFs with F substituent coordination.
Full size image

a HOMO and LUMO distributions. b ΔG diagram of CTFs in the ORR and WOR reaction. The insets show the transition state in the ORR and WOR, and the white, red, cyan, brown, and gray colors represent hydrogen, oxygen, fluorine, carbon, and nitrogen.

To further verify the feasibility of H2O2 photosynthesis in CTFs with different oxidation and reduction sites, the projected density of states (PDOS) was carried out in Fig. S2. Compared to CTF-1 and CTF-TF, the PDOS of the VB and CB edge (HOMO and LUMO) in CTF-TF-0.5 simultaneously present more nitrogen orbital contributions, suggesting that the VB and CB in CTF-TF-0.5 mainly originate from N on the triazine ring and trace amounts of C. Furthermore, to distinctly distinguish the contributions of N1 and N2 on the triazine ring for HOMO and LUMO, a single CTF-TF-0.5 single unit was selected for PDOS analysis. As shown in Fig. S3, the N1 2p orbital contributed the more HOMO, while the N2 2p orbital contributed the more LUMO. This also help confirm that N1 and N2 in CTF-TF-0.5 serve as oxidation and reduction sites, respectively. Furthermore, we also investigated the thermodynamics of the WOR and ORR to produce H2O2 at all sites on CTF-1, CTF-TF-0.5, and CTF-TF by analyzing Gibbs free energy (ΔG) and intermediate configurations (Figs. 2b and S4–S11). The results revealed that CTF-TF-0.5_N2 exhibited the lowest free energy for ORR to H2O2, accompanied by a sequential transition of reactive oxygen intermediates (*O2− to *OOH), followed by proton coupling to form H2O2. Meanwhile, CTF-TF-0.5_N1 showed the lowest free energy for WOR to H2O2, with a reactive oxygen intermediate (*OH), which then two *OH combines to produce H2O2. These confirmed that the N1 site of CTF-TF-0.5 serves as the oxidation center for WOR, while the N2 site acts as the reduction center for ORR. Moreover, CTF-TF-0.5 facilitates more efficient bulk charge separation of photogenerated carriers into distinct molecular units, enhancing H2O2 overall photosynthesis compared to CTF-1 and CTF-TF.

Structural characterization

CTF-1, CTF-TF-0.5, and CTF-TF were synthesized via direct cyclotrimerization of aromatic aldehydes using NH4I (Fig. S12)19. The resulting nanosheet-like morphologies were confirmed by scanning electron microscopy and transmission electron microscopy (TEM) (Figs. S13 and S14). The X-ray diffraction pattern of CTF-1 revealed two diffraction peaks at 7.5° and 18.2°, corresponding to the (100) and (210) planes, respectively, indicating a moderate degree of crystallinity, while that of CTF-TF-0.5 and CTF-TF exhibited no obvious diffraction peaks, showing that both of them are amorphous (Fig. S15)19. The pore structures and surface areas of CTFs were characterized by N2 adsorption-desorption measurements (Fig. S16). All samples displayed significant nitrogen adsorption in the low-pressure region (P/P0 < 0.1), and the Brunauer–Emmett–Teller surface areas of CTF-1, CTF-TF-0.5, and CTF-TF were determined to be 301, 426, and 385 m2 g−1, respectively.

The chemical environment and structural integrity were explored using solid-state NMR (13C, 1H, and 19F) spectroscopy. In the CP-MAS 13C-NMR spectrum (Fig. 3a), the CTF-1 demonstrates the obvious carbon signals from the triazine units at 170.4 ppm (a) and the phenyl carbons (b and c) at 138.8 ppm and 128.7 ppm. In the CTF-TF, the carbon signal of triazine rings shifts to 159 ppm because of the introduction of fluorine atoms (a’), and the carbon signal of fluorinated benzene rings are located at 145.1 ppm (c’). Moreover, the carbon atoms connected with triazine rings display a peak at 119 ppm (b’). Note that the CTF-TF-0.5 simultaneously displayed the carbon signals (a and a’) from the triazine units without and with fluorine atoms and the phenyl carbons (b and c), and the carbon signal (c’) of fluorinated benzene rings, and the carbon atoms connected with triazine rings (b’), which indicated that CTF-TF-0.5 contains both benzene ring and fluorine-substituted benzene ring19. The solid-state 1H NMR spectra of CTF-1 and CTF-TF-0.5 exhibit a signal at approximately 7.5 ppm, originating from H atom on the benzene ring (Fig. S17a), while the solid-state 19F NMR spectra of CTF-TF-0.5 and CTF-TF exhibit a signal at approximately −144 ppm, originating from F atom on fluorine-substituted benzene ring (Fig. S17b).

Fig. 3: Structural characterization of CTFs.
Fig. 3: Structural characterization of CTFs.
Full size image

a Solid-state 13C NMR spectra and structural model representation. b FT-IR spectra. c Solid EPR spectra. d–f Phase hysteresis loop (blue line) and amplitude butterfly loop (red line) of CTF-1 (d), CTF-TF-0.5 (e), and CTF-TF (f).

The chemical bonding characteristics of CTF-1, CTF-TF-0.5, and CTF-TF were investigated using Fourier transform infrared spectroscopy (FT-IR) (Fig. 3b). All samples exhibited two characteristic vibration signals associated with triazine units, appearing in the range of 1300–1600 cm−1. For CTF-1, sharp peaks at 1356 and 1510 cm−1 corresponded to triazine unit vibrations, while the peak at 813 cm−1 is attributed to benzene rings. Similarly, CTF-TF-0.5 and CTF-TF displayed peaks at about 1356 and 1515 cm−1, indicative of triazine units, along with an additional peak at 983 cm−1, assigned to the stretching vibrations of C−F bonds21, confirming the formation of CTF-TF and CTF-TF-0.5. Meanwhile, as the number of F substituents increases, the peak at 813 cm−1 gradually decreases, indicating that the benzene ring is progressively substituted by F as the amount of F substitution increases. These results revealed the successful synthesis of CTF-TF-0.5 with our previous report19, which was further corroborated by simulated FT-IR spectra with using Gauss (Figs. 18 and S19) and X-ray photoelectron spectroscopy analyses (Figs. S20–23 and Table S1).

The optical absorption properties of the samples were investigated using UV–visible diffuse reflectance spectroscopy. All samples exhibited an absorption edge around 600 nm (Fig. S24). As the F content increased, the yellow color of CTF gradually deepened, and the light absorption range broadened. This enhancement is attributed to extended electron delocalization and increased n–π* electronic transitions22,23. Electron paramagnetic resonance (EPR) spectroscopy was employed to probe the presence of unpaired electrons (Fig. 3c). CTF-TF-0.5 exhibited the strongest single Lorentzian signal at g = 2.005, indicative of delocalized electrons on the triazine ring24,25. This is likely due to the enhanced asymmetry introduced by the F-substituted benzene in CTF-TF-0.5, which strengthens the electronic delocalization on the triazine ring.

Charge behaviors promoted by F substituent coordination

By tuning the number of F substitutions on CTFs, distinct spatial distributions of oxidation-reduction sites were achieved to enhance the piezoelectric effect. To further explore the modulation of surface charges, Kelvin probe force microscopy was employed to map the surface piezoelectric potential distribution of all samples (Figs. S25 and S26). The contact potential difference (VCPD) of CTF-TF-0.5 ranged from −51 to −14 mV, significantly lower than that of CTF-1 (3.07 to 17.2 mV) and CTF-TF (−18 to 0.82 mV). Additionally, VCPD values were calculated using the following equation26:

$${{{{\bf{V}}}}}_{{CPD}}=\frac{{{{{\boldsymbol{\varphi }}}}}_{{tip}}-{{{{\boldsymbol{\varphi }}}}}_{{sample}}}{{{{\bf{e}}}}}$$
(1)

where φtip and φsample represent the work functions of the probe and the sample, respectively, and e represents the electron charge. Through this relationship, we found that CTF-TF-0.5 exhibited a higher work function than CTF-1 and CTF-TF, which indicated that electrons don’t escape easily without light27. Moreover, under illumination, CTF-TF-0.5 demonstrated an average surface potential enhancement of approximately 57 mV, while those of CTF-1 and CTF-TF showed almost no change under light. This enhancement is attributed to the spatially separated oxidation-reduction sites of CTF-TF-0.5, which allow for photogenerated electrons to escape from the surface, significantly boosting surface potential.

The piezoelectric properties of all samples were further evaluated using piezoresponse force microscopy. The phase-voltage hysteresis loop for all CTFs (blue line, Fig. 3d–f) displayed a clear hysteresis with a 180° phase shift at 5 V, indicating distinct local polarization switching behavior28. Moreover, the amplitude-voltage butterfly loop for CTF-TF-0.5 exhibited a remarkable amplitude of approximately 4.74 nm, significantly surpassing the 0.495 nm of CTF-1 and 0.733 nm of CTF-TF (red line), demonstrating the superior piezoelectric performance of CTF-TF-0.5. These results strongly confirmed that the spatial distribution of oxidation-reduction assignment in CTFs plays a crucial role in enhancing the local piezoelectric properties of CTF-TF-0.5.

Charge migration behavior with separated oxidation-reduction assignment

As previously noted, CTF-TF-0.5 exhibits separated redox assignments resulting from the modulation of F substitution on the benzene ring, which significantly promotes directed bulk electron-hole separation within the catalyst. Moreover, structural reorganization of piezoelectric materials under external force leads to an enhanced local dipole moment, further facilitating carrier migration29. To probe the structural changes under high pressure, in-situ FT-IR spectroscopy was employed on all CTFs (Fig. 4a–c). As pressure increased, the characteristic peaks of the triazine ring exhibited a gradual blue shift. At approximately 15 GPa, the triazine ring peaks in CTF-1 shifted from 1357 cm−1 and 1519 cm−1 to 1420 cm−1 and 1567 cm−1, respectively, while the benzene ring peaks at 811 cm−1 remained nearly unchanged. Upon incorporating F substitution, the triazine ring peaks in CTF-TF-0.5 shifted from 1356 cm−1 and 1504 cm−1 to 1421 cm−1 and 1566 cm−1, whereas CTF-TF exhibited a shift from 1501 cm−1 to 1554 cm−1. The C-F peaks of both CTF-TF-0.5 and CTF-TF also displayed a slight blue shift under pressure (Fig. 4b). The calculated blue-shift rates for the triazine rings (triazine-1 and triazine-2) were 3.26 and 3.34 cm−1 GPa−1 in CTF-1, 3.73 and 3.68 cm−1 GPa−1 in CTF-TF-0.5, and 3.38 cm−1 GPa−1 (triazine-2) in CTF-TF. For C-F, the blue-shift rates were 1.21 cm−1 GPa−1 in CTF-TF-0.5 and 1.29 cm−1 GPa−1 in CTF-TF (Fig. 4c). The more pronounced blue-shift rates observed in CTF-TF suggest shorter bond length changes during structural reorganization under pressure, which serve to enhance the local dipole moment (Fig. 4d)30. These variations in blue-shift magnitudes underscore the distinct mechanical responsiveness of the samples and demonstrate that the separated oxidation-reduction sites in CTF-TF-0.5 significantly enhanced its piezoelectric properties under pressure.

Fig. 4: In-situ high pressure characterization.
Fig. 4: In-situ high pressure characterization.
Full size image

a–c In-situ high pressure FT-IR pseudocolor plot (a), spectra (b), and the corresponding peaks fitting (c). d Schematic representation of electron conduction and changes in functional groups of in CTFs under pressure. e, f In-situ high pressure UV‒Vis absorption spectra (e) and bandgap variation with pressure (f). g, h In-situ high pressure PL pseudocolor plot (g) and the corresponding PL peak position (blue line) and PL peak intensity (red line) as a function of pressure (h) at an excitation wavelength of 355 nm.

Compared to CTF-1 and CTF-TF, the enhanced performance of CTF-TF-0.5 for H2O2 production suggests that rapid electron migration occurs during the photocatalytic process. To explore the charge migration behavior, the fundamental energy band structures of all samples were determined using Tauc plots and Mott-Schottky plots (Figs. S27–29). Additionally, in-situ pressure-dependent UV‒vis absorption spectra were measured to examine the force-induced shifts in the absorption edges of all CTFs. As shown in Figs. 4e and S30–32, with increasing pressure from 0 to 15 GPa, the absorption edges of all samples exhibited significant redshifts, with the color of the samples gradually transitioning from yellow to brown. The bandgap of CTF-1 decreased from 2.20 to 1.73 eV, CTF-TF-0.5 from 2.16 to 1.84 eV, and CTF-TF from 2.12 to 1.74 eV (Figs. S33 and S34). Upon releasing the pressure to 0 GPa, the bandgaps of CTF-1, CTF-TF-0.5, and CTF-TF increased to 2.18, 2.15, and 2.03 eV, respectively (Fig. 4f). These observations indicated that pressure application effectively modulates the local energy bands, enhancing the migration and transport of photocatalytic charge carriers.

To understand the charge carrier behavior in CTF-TF-0.5 with separated oxidation-reduction site, the solid-state photoluminescence (PL) spectra were analyzed. As shown in Fig. S35, CTF-TF-0.5 exhibited a lower PL intensity than other samples with a quenching efficiency of 88.8% relative to CTF-1 and that of 87.8% relative to CTF-TF. These results suggested that the separated oxidation-reduction site in CTF-TF-0.5 effectively suppress photogenerated charge recombination. The peak position of PL emission peak displayed slight red-shifted, perhaps due to the narrowing of energy band after adding F substitution. The charge carrier migration properties were further explored through electrochemical techniques, including transient photocurrent, electrochemical impedance spectroscopy, and linear sweep voltammetry (Figs. S36–S38).

To investigate the charge migration process in photoexcited CTFs under an external force, in-situ pressure-dependent PL spectroscopy was conducted. With pressure increased from ambient to 15 GPa, all CTFs exhibited a color change from yellowish to colorless (Figs. S39–41). As shown in Figs. 4g, h and S42, the PL peak positions of all samples exhibited a noticeable redshift with increasing pressure, while the PL intensity of CTF-TF-0.5 was significantly lower than those of CTF-1 and CTF-TF. Notably, the decrease rate of intensity in CTF-TF-0.5 was more pronounced than in the other samples. Upon pressure release to 0 GPa, the peak position reverted, though the peak intensity remained diminished, suggesting that structural damage under applied pressure weakened the material’s luminescence31. The exciton binding energy (Eb), determined from temperature-dependent PL spectra, and the emission peak intensities from steady-state PL spectra followed the order: CTF-1 > CTF-TF > CTF-TF-0.5 (Figs. S43–S45). This indicated that exciton dissociation is most efficient in CTF-TF-0.5, followed by CTF-TF and CTF-1. These all findings confirm that external pressure enhances charge separation efficiency in photoexcited CTFs, along with CTF-TF-0.5 exhibiting the highest charge separation efficiency. Therefore, this separated oxidation-reduction assignment could improve efficiently charge migration behavior.

Heterogeneous interfacial H2O2 photosynthesis pathway via separated oxidation-reduction assignment

To explore the role of the separated oxidation-reduction site in H2O2 photosynthesis, CTF-1, CTF-TF-0.5, and CTF-TF were tested in a reactor under both ultrasonic (Us) force and simulated sunlight (Fig. 5a). As shown in Figs. 5b, S46, and S47, the photocatalytic performance of CTF-1, CTF-TF-0.5, and CTF-TF in various scenarios after 60 min followed the order: CTF-TF-0.5/Us/light > CTF-1/Us/light > CTF-0.5/light > CTF-TF/Us/light > CTF-TF/light > CTF-TF-0.5/Us > CTF-1/light >CTF-1/Us > CTF-TF/Us. Notably, the H2O2 yield of CTF-TF-0.5 showed only a slight increase relative to CTF-TF and CTF-1 under Us or light alone. However, when both Us and light are applied simultaneously, the H2O2 yield of CTF-TF-0.5 (3.28 mmol g−1 h−1) was 2.43 times higher than with light alone, 4.75 times higher than with Us alone, and 2.13 and 2.69 times higher than the yields from CTF-1 and CTF-TF under Us and light, respectively. In addition, the mass transfer of reactants to the surface of photocatalyst is also an important factor affecting the reaction efficiency, which will be studied in our subsequent research. Note that the H2O2 yield is governed by both the formation rate (kf) and the decomposition rate (kd), and the reaction kinetics can be described by the equation [H2O2] = (kf/kd) {1-exp(-kd t)}, where kf and kd are derived from zero-order and first-order kinetics (Figs. S48 and S49)32. These analyses demonstrated that CTF-TF-0.5/Us/light exhibits the highest kf, suggesting that this condition leads to the highest H2O2 yield compared to other conditions (Fig. S50). The apparent quantum yield (AQY) of CTF-TF-0.5 was calculated at specific wavelengths and shown to approximately match the UV–Vis spectrum (Fig. S51), which demonstrated the excellent AQY at 420 nm compared to the reported catalysts (Table S2). Special attention needs to be paid to the fact that the current yield shows a high value among most of the reported COF-based photocatalysts and piezocatalysts and is even higher than that of most piezo-photocatalysts (Table S3).

Fig. 5: Heterogeneous interfacial H2O2 photosynthesis and its reaction mechanism.
Fig. 5: Heterogeneous interfacial H2O2 photosynthesis and its reaction mechanism.
Full size image

a Schematic diagram of triphase H2O2 photosynthesis with Us. b Time profiles of CTFs for H2O2 photosynthesis in various scenarios. Experimental conditions: catalyst (0.125 g L−1) under Us and light, T = 25 °C, water. c Comparison of H2O2 synthesis using CTF-TF-0.5 in triphase and diphase reaction system at 60 min. d Stability of CTF-TF-0.5 over 10 h. e H2O2 synthesis in real water samples using CTF-TF-0.5. f, g In-situ EPR spectra and concentration and lifetime of different free radicals during ORR (f) and WOR (g) process via CTF-TF-0.5/Us/Light. h Proposed reaction mechanism. Error bars indicate the standard deviation from three measurements.

It is worth noting that CTF-TF-0.5 floated on the surface of water to react during the photosynthesis, which was very favorable for direct sunlight irradiation and direct mass transfer of O2 from the gas to catalyst without passing through water (Fig. 5a). To identify the role of gas-solid-liquid triphase in H2O2 production, the photocatalytic performance of CTF-TF-0.5 in three-phase and two-phase systems were investigated (Figs. 5c, S52, and S53). By using cyclohexane, a liquid with a density lower than that of CTFs and water, to covering CTF-TF-0.5 to prevent transfer mass of O2 from air to catalyst, the H2O2 photosynthesis performance of diphase rapidly decreased by about 64.5% relative to that of triphase. Water contact angles were further used to explore the interface state of triphasic system using CTF-1, CTF-TF-0.5, and CTF-TF (Fig. S54). With the increase of F content, contact angel of CTF-TF-0.5 and CTF-TF increased by ~20°, relative to that of CTF-1. This suggests that the increase of F content significantly enhance the hydrophobicity of CTFs, which is more favorable to the formation of three-phase interfaces. These indicated that the floating CTF-TF-0.5 via separated oxidation-reduction assignment effectively enhanced H2O2 photosynthesis.

Widespread practical application of piezo-photocatalysts requires several factors, including the solar-to-chemical conversion (SCC) efficiency, stability, and suitability in various media. The photocatalytic H2O2 production using the AM-1.5 G filter with Us was carried out and the SCC efficiency of CTF-TF-0.5 with mechanical force was also calculated to 0.16% (Fig. S55 and Table S4), surpassing the typical photosynthetic efficiency of plants of 0.10%33, which demonstrated the excellent SCC compared to the reported catalysts (Table S5). Moreover, H2O2 production over CTF-TF-0.5 for a wide pH range (pH = 1 ~ 11) were also explored (Fig. S56), and the optimum pH is 3. CTF-TF-0.5 exhibited good stability over 6 h, and then it gradually decreased and eventually stabilized at 10 h with about 24 mmol g−1 (Figs. 5d and S57). Furthermore, the H2O2 production rate in various water samples was examined. As shown in Fig. 5e, the H2O2 yield from CTF-TF-0.5 consistently exceeded 1.6 mmol g−1 h−1 across the different real water samples, despite slightly lower performance than pure water due to the impact of a large number of ions or organics in the real water on H2O2 production. These results underscored the practical potential of CTF-TF-0.5 as an effective catalyst for H2O2 photosynthesis, making it a promising candidate for environmental application.

Heterogeneous interfacial H2O2 photosynthesis pathway via CTF-TF-0.5

The intrinsic electronic structure of CTFs plays a pivotal role in determining catalytic activity and provides insights into the microscopic processes at the catalytic active sites. To elucidate the localized electron distribution profiles, the electron localization function was computed. As shown in Fig. S58, the electron cloud distribution in CTF-1 and CTF-TF exhibited symmetric spatial patterns, while CTF-TF-0.5 displayed an asymmetric distribution, attributed to its separated oxidation-reduction sites. Upon applying pressure, all CTFs showed a similar trend of slightly increased electron cloud density (Fig. S58). The strong charge interactions between adsorbed O2 and nitrogen atoms in the CTFs were further explored through charge density difference (Figs. S59 and S60). In the case of F-substituted CTFs, electron accumulation on O atoms and depletion at the N edges became more pronounced under applied pressure, confirming the activating influence of F substitution. Notably, CTF-TF-0.5 and CTF-TF exhibited higher electron transfer and adsorption energies under pressure. These findings highlighted that the separated oxidation-reduction sites in CTF-TF-0.5 enhance electron activity and O2 activation, thereby boosting catalytic performance, and O2 adsorption capacity of CTF-TF-0.5 was demonstrated through O2 temperature-programmed desorption (Fig. S61).

As demonstrated in Fig. 2b, CTF-TF-0.5 exhibits a higher propensity for H2O2 production via both ORR and WOR compared to CTF-1 and CTF-TF. To further elucidate the reaction pathways, H2O2 production was investigated under varying O2/N2 atmospheres. As shown in Fig. S62, the H2O2 production rate was significantly higher in an O2 atmosphere than in air, while production in N2 was markedly lower, remaining at approximately 1.3 mmol g−1. The outstanding 2e− ORR of CTF-TF-0.5 for H2O2 production with high selectivity (~85%) was further evaluated by rotating ring-disk electrode analysis (Fig. S63)34. Various sacrificial agents experiment revealed that *O2− predominantly contribute to H2O2 production, whereas *OH play a lesser role (Fig. S64). The generation of *O2− and *OH was further corroborated by EPR spectroscopy (Figs. S65 and S66). Isotopic labeling experiments provided additional insights into the reaction pathways (Table S6). In a system with H216O and 18O2, the 18O signal in the produced H2O2 under light irradiation was significantly higher than that without light, confirming the contribution of ORR35. Similarly, in a system with H218O and 16O2, the presence of 18O signals in the produced H2O2 under light further confirmed the contribution of WOR. These results provide compelling evidence that CTF-TF-0.5 achieves efficient H2O2 production through a dual-pathway mechanism leveraging both ORR and WOR in our catalytic system36,37. The H2O2 production via WOR pathway was also been proved through isotope experiments using Gas chromatography-mass spectrometry (Fig. S67).

To investigate the intermediates involved in photocatalytic H2O2 production using CTF-TF-0.5 under Us, in-situ EPR experiments were performed (Figs. S68 and 5f, g) to understand the dynamic reaction processes. DMPO was employed to capture the in-situ generated *O2−, *OOH, and *OH intermediates. For the ORR intermediates (Fig. 5f), the *O2− signal emerged at 72 s under Us and light, increasing steadily until approximately 216 s, with a peak concentration of 1.63 × 10−4 M. Subsequently, *O2– was rapidly converted to *OOH by protonation and continues to grow to 376 s, with optimum concentration of 3.07 × 10−4 M, and it also maintained to 600 s. For the WOR intermediates (Fig. 5g), the *OH signal appeared at 123 s, progressively increasing and stabilizing within the range of 1.52 × 10−5 to 3.24 × 10−5 M by 600 s. These findings indicated that H2O2 production occurs through a combination of an indirect 2e− ORR and a 1e− WOR process. The full EPR spectra with a sweep time of 4 s per sample were provided in Figs. S69 and S70 for detailed analysis. Figure S71 illustrates the simulated EPR spectra of *O2–,*OOH, and *OH.

According to the findings of this study, we proposed a mechanism for separated oxidation-reduction site in F-substituted CTFs to promote H2O2 overall photosynthesis (Fig. 5h). Initially, due to F substitutions in CTFs, a separated oxidation-reduction site is formed in the CTFs-TF-0.5 plane. Under Us/light conditions, this separated oxidation-reduction center allows the photogenerated electrons and holes to undergo directional migration. Then, the O2 is easily adsorbed on the surface of N2 atoms and the adequate electrons localized at N2 reduces O2 to undergo ORR through an indirect 2e− transfer pathway with intermediates of *O2− and *OOH to form H2O2. Meanwhile, the positive holes localized at N1 oxidized H2O to form *OH, which subsequently combined to produce H2O2 through 1e− WOR.

Evaluation of the environmental applications of heterogeneous interfacial H2O2 photosynthesis system

It is well known that arsenic (As)-containing wastewater from acid mine drainage and metallurgical industries remain a huge challenge for water pollution control due to its biotoxicity and bioaccumulation38. Here, we used CTF-TF-0.5 with separated oxidation-reduction sites for acidic wastewater containing As (III) purification. As shown in Figs. 6a and S72a, CTF-TF-0.5 showed the highest oxidation capacity of As (III) (63.5%) relative to CTF-1 (26.8%) and CTF-TF (40.6%), while the concentration of As (V) of all CTFs gradually increased. During the degradation, the production of H2O2 also shows a positive correlation with its oxidation rate of As (III) (Fig. S72b, c). Interestingly, the adsorption of As (V) in CTF-TF-0.5 was significantly higher than those of CTF-1 and CTF-TF, which was attributed to the excellent adsorption capacity of CTF-0.5 for As (V) (Fig. S72d). Furthermore, the As (III) oxidation of CTF-TF-0.5 was extended to a long time reaction (Fig. 6b). The concentration of As (III) was oxidized by about 95.7% within 3 h, while the concentration of As (V) was about 806 mg L−1. During 3 h reaction, it also exhibits a continuously increasing H2O2 concentration, reaching 244.5 μM at 3 h (Fig. S73). The effect of different pH on As oxidation was also evaluated in the Fig. 6c, the concentration of As (III) was completely oxidized within 90 min at pH = 3 and displayed a highest oxidation rate of 0.0527 min−1 than those at pH = 7 and pH = 10 (Fig. S74a). During 3 h reaction, it also exhibits a continuously increasing H2O2 concentration at pH = 3, reaching 471.5 μM at 3 h, which was significantly higher than that at pH = 7 and pH = 10 (Fig. S74b, c). The adsorption of As (V) in CTF-TF-0.5 at pH = 3 was significantly higher than those at pH = 7 and pH = 10, which was due to the adsorption of As (V) could be promoted under acid condition (Fig. S74d). The effect of initial concentration of As (III) oxidation and the in-situ generation of H2O2 was also explored (Figs. S75 and S76). Furthermore, scavengers were added to the initial solution to investigate the reaction mechanism of photocatalytic As (III) oxidation with Us. TBA for *OH have no effect on As (III) oxidation, while p-BQ for *O2− have some effect for that. Note that catalase for H2O2 mainly contribute to As (III) oxidation, due to that the addition of catalase resulted in its near-total loss of oxidation capacity (Fig. S77). These results demonstrate that As (III) oxidation via piezo-photocatalysis was directly utilized for the oxidation of As (III), without conversion into other reactive species.

Fig. 6: Evaluation of environmental applications for mine wastewater treatment.
Fig. 6: Evaluation of environmental applications for mine wastewater treatment.
Full size image

a, b Time profiles of all CTFs in 1 h (a) and CTF-TF-0.5 (b) in 3 h for heterogeneous interfacial As (III) oxidation. c The effect of different pH. d As-containing mine wastewater treatment. The inset shows the photographs of wastewater during the degradation process. Error bars indicate the standard deviation from three measurements.

It is worth that the pH of acidic As-containing wastewater from mine drainage always ranges from 2.0 to 3.539. Thus, our proposed catalytic system was also used to treat real As-containing acidic mine wastewater with pH = 2.8 (Fig. 6d). First, CTF-TF-0.5 did not exhibit significant adsorption under natural conditions for 20 min. With the reaction beginning, about 34 μg L−1 As (III) was completely oxidized within 3 h, and about 6 μg L−1 of As (V) at initial solution and As (V) converted via As (III) was completely adsorbed within 30 min. These results fully demonstrated the application potential of CTF-TF-0.5 with separated oxidation-reduction site in real As-containing acidic mine wastewater. It is also expected to be a key component for mine wastewater treatment in the future (Fig. S78).

Discussion

In summary, we synthesized CTFs and fluorine-substituted CTFs via a low temperature approach. This process established a separable oxidation-reduction site in the conjugated skeleton of CTF-TF-0.5, by modulating the coordination number of F-substituted benzene linkers to influence the redox characteristics of the triazine rings. CTF-TF-0.5 floats on the water surface and reacts at the gas-solid-liquid triple-phase interface, facilitating rapid oxygen transfer to the catalyst surface, and it achieves a high H2O2 photosynthesis rate of 4664 µmol g−1 h−1 at gas-solid-liquid triphasic interface. The separable oxidation-reduction sites accelerate redox reactions in the separated space, with photogenerated electrons localized at the reduction sites (N2, triazine connected a benzene and two F-substituted benzene) and photogenerated holes localized at the oxidation sites (N1, triazine connected two benzene and a F-substituted benzene). Under external force, O2 is efficiently adsorbed at the N2 site, where localized electrons reduce it via an indirect 2e− transfer ORR pathway to form H2O2. Simultaneously, holes at the N1 site oxidize H2O through a 1e− WOR to produce H2O2, enabling highly efficient overall H2O2 photosynthesis. Additionally, the unique system demonstrated significant potential for As-containing acid mine wastewater treatment. This approach highlights a promising strategy for manipulating complex chemical reactions using molecular junction catalysts with multiple active sites and offers novel insights into advanced catalyst design at molecular and atomic levels.

Methods

The material, instruments, other experiments and characterizations are discussed in Supplementary Information (Texts S1–S8).

H2O2 photosynthesis at a triphasic interface

Five milligram catalyst was added in 40 mL pure water and the suspension was dispersed by ultrasonication. The reactor was maintained at 25 ± 0.5 °C using a cooling circulating water system and irradiated with full-spectrum light from a 300 W Xe lamp (PLS-SXE300D, Beijing Perfectlight Technology Co., Ltd) at an intensity of 150 mW cm−2. Simultaneously, the suspension was subjected to react using an ultrasonic cleaner (40 kHz, 100 W, Jielimei, Kunshan, China). For light-only control experiments, the reactor was placed under the Xe lamp with stirring. The concentration of H2O2 was determined using the KI colorimetric method40. One milliliter of freshly prepared KI reagent A (0.4 M KI, 0.05 M NaOH, 1.6 × 10−4 M (NH4)6Mo7O24⋅4H2O) and 1 mL reagent B (0.1 M KHC8H4O4) were mixed with 1 mL the above samples. The absorbance of the above mixture was measured at 350 nm by a UV‒Vis spectrophotometer (JASCO V-770, Japan).

In-situ high-pressure FT-IR, UV–Vis absorption, and PL spectra

High-pressure experiments were conducted using a symmetric diamond anvil cell equipped with ultra-low fluorescence diamond anvils featuring a culet diameter of 400 µm. A T301 stainless steel gasket was pre-indented to a thickness of 45 µm and laser-drilled to create a 150 µm diameter sample chamber, which housed both the sample and a ruby ball for pressure calibration. The pressure was determined using the ruby fluorescence technique by monitoring the shift of the R1 line of ruby fluorescence. Silicon oil was employed as the pressure-transmitting medium to ensure hydrostatic conditions41. All high-pressure experiments were performed at room temperature.

High-pressure PL measurements were conducted using a custom-built optical measurement system. The pressure-dependent PL spectra were obtained with a semiconductor laser at an excitation wavelength of 355 nm. To ensure consistency, parameters such as laser intensity and luminous flux were fixed during each high-pressure PL experiment, minimizing potential variations in PL intensity for CTFs. Absorption spectra were recorded in the exciton absorption band region using a deuterium-halogen light source. Both high-pressure PL and absorption spectra of CTFs were acquired using an Ocean Optics QE65000 optical fiber spectrometer. For in situ high-pressure infrared (IR) experiments, KBr was employed as the pressure-transmitting medium. IR spectroscopy of TriPE was performed using a Nicolet iN10 microscope spectrometer (Thermo Fisher Scientific, USA) equipped with a liquid nitrogen-cooled detector41.

Arsenic containing acid mine wastewater treatment

Before the start of the experiment, As (III) standard solution (1000 mg L−1) was diluted stepwise into 1 mg L−1 As (III) solution. 1 mg CTFs was added to 40 mL As (III) solution (1 mg L−1) to form the reaction system. A mechanical stirrer was used for 20 min to stabilize the adsorption, then the reactor was kept at 25 ± 0.5 °C with cooling circulating water and was illuminated using the Xe lamp with a light intensity of 150 mW cm−2, and simultaneously subjected to ultrasonication by the ultrasonic cleaner without stirring. The pH was adjusted with 1 M HCl and NaOH to perform pH experiments. For acidic As-containing mining wastewater treatment, 40 mL wastewater was filtered to conduct the same experimental procedure. The concentration of As (III) and As (V) was determined by HPLC-AFS. With the instrument beginning, it was step-by-step cleaned with 30% methanol-water and the configured mobile phase (1.98 g (NH4)2HPO4 was dissolved in 1 L pure water, and then formic acid and ammonia was used to adjust the pH to 6 with ultrasound for 30 min) for 20 min. In addition, 3% potassium borohydride-5‰ sodium hydroxide solution was prepared as the reducing agent, and the gas was opened for 30 min. 5% HCl was used as the carrier solution.