Surface defect engineering and MOF derivatives regulate the electron transport pathway of polymeric carbon nitride for an efficient photocatalytic 2e⁻ oxygen reduction reaction to form H₂O₂

Min Liu^a, Siyuan Di^b, Hailan Qin^a, Pin Chen^a, Yunkang Liu^a, Huan Liu^a, Qiuyue Zhang^a, Zihan Li^a and Shukui Zhu*^a
aState Key Laboratory of Geomicrobiology and Environmental Changes, China University of Geosciences, Wuhan 430074, China. E-mail: shukuizhu@126.com
^bEngineering Research Center of Ministry of Education for Clean Production of Textile Printing and Dyeing, Wuhan Textile University, Wuhan 430200, China

First published on 22nd August 2025

---

---

1. Introduction

Hydrogen peroxide (H₂O₂), as an environmentally friendly strong oxidant and disinfectant,¹ is widely used in sterilization,² wastewater treatment,³ textile industry,⁴ and chemical synthesis.⁵ Traditionally, H₂O₂ was predominantly produced via the anthraquinone oxidation (AO) method,⁶ which involved complex technology, high energy consumption,⁷ and significant pollution.⁸ Therefore, it was necessary to develop a new synthesis method for H₂O₂ from the perspectives of energy savings, cost efficiency, and sustainable development. In recent years, the photocatalytic synthesis of H₂O₂ through oxidation–reduction reactions has drawn great attention and is considered the most promising method for replacing the traditional AO process.⁹ In this method, only air and water are used as reactants, and H₂O₂ can be continuously produced on the catalyst surface under sunlight exposure. It has been reported that photocatalysts produce H₂O₂ through the oxygen reduction reaction (ORR) and water oxidation reaction (WOR).¹⁰ Compared with the WOR, the ORR occurs more easily and requires a less positive valence band position of the catalyst. However, the ORR is often limited by the transport efficiency of photogenerated carriers as a high concentration of free electrons is crucial for efficient reaction. Thus, it is imperative to develop simple and readily available photocatalysts with adjustable photo-generated carrier transport efficiency to address this challenge.

Graphite-phase carbon nitride (g-C₃N₄) is an inorganic non-metallic semiconductor with a two-dimensional layered nanostructure,¹¹ offering advantages such as abundant raw materials,¹² good chemical stability,¹³ and an appropriate band gap.¹⁴ However, the performance of pristine g-C₃N₄ does not meet the requirements for practical applications.¹⁵ Due to its thick and stacked structure, the path for photo-generated electron migration is long, which is highly unfavourable for interlayer electron transport and results in a high recombination rate of photo-generated carriers.¹⁶ Therefore, researchers have focused on modifying the properties of pristine g-C₃N₄ to improve its photocatalytic performance using methods such as morphology control,¹⁷ element doping,¹⁸ energy-band engineering,¹⁹ and defect modification.²⁰ Among these approaches, element doping and defect introduction are the most widely used techniques to enhance the photocatalytic activity of pristine g-C₃N₄.^21,22 For example, doping g-C₃N₄ with alkali metals modifies the band gap or electronic structure, broadens its light absorption range, and promotes charge migration.²³ Defects, such as N and O vacancies, were also reported as reactive sites and electron traps in photocatalytic reactions, increasing the electron concentration in the system.^24,25 However, for g-C₃N₄ as a photocatalyst for H₂O₂ production, the above methods did not effectively address the issue of a low H₂O₂ yield.

In photocatalytic H₂O₂ production, the main factors that can influence the yield include the surface oxygen adsorption capacity of the catalyst, the migration rate of unpaired electrons in the catalyst, the positions of the valence and conduction bands, and the decomposition rate of H₂O₂. The low H₂O₂ yield of g-C₃N₄ is primarily due to the following reasons: (1) small specific surface area, which limits the adsorption of sufficient oxygen, (2) few active sites, which do not provide enough platforms for H₂O₂ production, (3) as a layered structure, the unpaired electrons between the layers cannot migrate to the catalyst surface in time to participate in the reaction, and (4) the high H₂O₂ decomposition rate on the catalyst surface. Thus, when the material is doped with elements and defects are introduced simultaneously, theoretically, the light absorption capacity of the material can be enhanced, electron migration can be promoted, and more active sites and electron traps can be generated.²⁶ As a result, more photoinduced electrons and holes are generated due to the enhanced light absorption,²⁷ more free electrons can be captured because of the improved electron migration, and consequently, more H₂O₂ is generated due to the increased number of active sites.²⁸

Based on the above-mentioned background, herein, through the regulation of KCl and 2D Zn-MOF-NH₂, a novel K-doped g-C₃N₄ was successfully prepared by a molten salt method for preparing H₂O₂. The structure and photoelectric properties of the catalyst were tested by several characterizations, the preparation conditions of the catalyst were evaluated, and several key parameters influencing catalytic efficiency were optimized, including the catalyst dosage, initial pH, coexisting ions, and dissolved organic acid. The potential reaction mechanism of H₂O₂ production was discussed with the aid of density functional theory (DFT) calculations. Moreover, the contaminated and authentic water matrices had little effect on the production of H₂O₂, thus reflecting the feasibility of the catalyst for producing H₂O₂ in natural water.

2. Experimental

2.1 Materials and sources

Dicyandiamide (C₂H₄N₄, AR), zinc nitrate hexahydrate (Zn(NO₃)₂‧6H₂O, AR), potassium chloride (KCl, AR), N,N-dimethylformamide (DMF, AR), ethanol, potassium iodide (KI, AR), potassium biphthalate (C₈H₅O₄K, AR), benzyl alcohol (C₇H₈O, AR), tert-butanol (TBA, AR), p-benzoquinone (PBQ, AR) and potassium dichromate (K₂Cr₂O₇, AR) were supplied by Sinopharm Chemical Reagent Co., Ltd. 2-Aminoterephthalic acid (BDC-NH₂, >98%) and triethylamine (C₆H₁₅N, AR) were supplied by Shanghai Macklin Biochemical Co., Ltd.

2.2 Synthesis of Zn-MOF-NH₂

Zn(NO₃)₂‧6H₂O (743.8 mg) and BDC-NH₂ (460 mg) were dissolved in 150 mL of DMF. Ultrapure water (9 mL), ethanol (9 mL), and triethylamine (3.5 mL) were added into the above mixture under ultrasound and then stirred for 12 h at room temperature. The 2D Zn-MOF-NH₂ nanosheets were collected by centrifugation, washed with DMF and ethanol three times, and dried in an oven at 60 °C.

2.3 Synthesis of K-C₃N₄-like

Dicyandiamide (2 g) and KCl (3 g) were put into an agate mortar and ground for 10 min, then transferred to a porcelain boat with a cover and 2 mL of ultrapure water was added. The porcelain boat was placed into a tube furnace and heated to 550 °C for 3 hours (at a ramping rate of 10 °C min⁻¹). After naturally cooling to room temperature, the obtained products were ground to a power and washed with ultrapure water until neutral, and then dried in an oven at 60 °C. The sample was denoted as K-C₃N₄-like to indicate that the material has a structure like g-C₃N₄ while containing K.

2.4 Synthesis of K-CZ-X

Dicyandiamide (2 g), KCl (3 g) and 2D Zn-MOF-NH₂ (X mg, X = 1, 2, and 3) were put into an agate mortar and ground for 10 min, then transferred to a porcelain boat with a cover and 2 mL of ultrapure water was added. The porcelain boat was placed into a tube furnace and heated at 550 °C for 3 hours (at a ramping rate of 10 °C min⁻¹). After naturally cooling to room temperature, the obtained products were ground to a power and washed with ultrapure water until neutral, and then dried in an oven at 60 °C. The samples were denoted as K-CZ-1, K-CZ-2, and K-CZ-3, respectively.

2.5 Synthesis of g-C₃N₄

Dicyandiamide (2 g) was put into an agate mortar and ground for 10 min, then transferred to a porcelain boat with a cover and 2 mL of ultrapure water was added. The porcelain boat was placed into a tube furnace and heated at 550 °C for 3 hours (at a ramping rate of 10 °C min⁻¹). After naturally cooling to room temperature, the obtained products were ground to power and washed with ultrapure water until neutral, and then dried in an oven at 60 °C.

2.6 Photocatalytic reaction

Photocatalytic H₂O₂ production was conducted with a home-made LED light (50 W, λ > 420 nm), and the temperature was controlled at 25 °C by a circulation cooling device. The photocatalyst (25 mg) was added to 47 mL of ultrapure water, then, 3 mL of benzyl alcohol (BA) was added as an electron donor. During the photocatalytic experiment, 1 mL of the mixture solution was taken every 10 minutes until the photocatalytic reaction ended (60 minutes). After certification, the concentration of H₂O₂ was measured by the iodometric method. In addition, 25 mg of the as-prepared photocatalyst was dispersed in 50 mL of 0.5 mM H₂O₂ solution to investigate the decomposition behaviour of H₂O₂ under the home-made LED light.

To identify the reactive species that participated in the reaction, K₂Cr₂O₇, p-benzoquinone (PBQ), and tert-butanol (TBA) were used as scavengers for electrons (e⁻), superoxide radicals (˙O₂⁻), and hydroxyl radicals (˙OH), respectively, while maintaining the other reaction conditions.

3. Results and discussion

3.1 Structural characterization of K-CZ-2

Fig. 1a schematically illustrates the synthesis procedure of the K-CZ-X catalyst. In brief, K-CZ-2 was synthesized via a two-step method: first, 2D Zn-MOF-NH₂ was synthesized by stirring at room temperature²⁹ and then it was mixed and calcined with dicyandiamide, KCl, and 2D Zn-MOF-NH₂ to obtain K-CZ-X through the molten salt method. The morphology of all samples was characterized by scanning electron microscopy (SEM, Fig. 1b–d and S1a). Compared with the massive structure of K-C₃N₄-like (Fig. 1b), K-CZ-2 had smaller lamellae and fewer layers (Fig. 1c), which is beneficial for mass transfer between the lamellae and the transport of photogenerated carriers. The EDS element mapping and summed spectrum of K-CZ-2 revealed that the Zn content was zero and was presumably below the detection limit, rendering it undetectable. But O was present, suggesting that K-CZ-2 might not have contained ZnO (Fig. 1d and S1c). Moreover, transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) images of K-CZ-2 clearly showed that it had an ultra-thin lamellar structure and no Zn–O lattice stripes (Fig. 1e, f and S1b), further confirming that K-CZ-2 did not contain ZnO but was a polycrystalline nanomaterial (Fig. 1g) whose lamellar structure was regulated by 2D Zn-MOF-NH₂.

The crystal structures of the samples were recorded using powder X-ray diffraction (XRD) measurements (Fig. 2a). The diffraction pattern of g-C₃N₄ exhibited two typical peaks at 13° (100) and 27.3° (002), reflecting the in-plane ordering of the tri-s-triazine motifs and the interlayer stacking of the aromatic systems, respectively.^30,31 Notably, the peak at 13° disappeared, and the peak at 27.3° shifted to the right for K-C₃N₄-like and K-CZ-2 after the addition of KCl. This change could be attributed to the disruption of the in-plane long-range order and the smaller spacing between the layers caused by the introduction of K⁺ and N vacancies.^32–34 This modification was highly beneficial for the migration of inner electrons to the surface layer and helped inhibit recombination of the photo-generated carriers.³⁵ Meanwhile, the molecular structural changes of K-C₃N₄-like and K-CZ-2 were characterized using Fourier-transform infrared spectroscopy (FTIR, Fig. 2b). The peaks centered at 809, 914, 1114–1700, and 3000–3650 cm⁻¹ were observed in both K-C₃N₄-like and K-CZ-2, which were attributed to tri-s-triazine ring modes, deformation of N–H, stretching of C–O–C, aromatic carbon nitride heterocycles, and N–H stretching vibrations,^36,37 respectively. Notably, compared with K-C₃N₄-like, the peak intensity of K-CZ-2 at 2178 cm⁻¹ was stronger, corresponding to the stretching vibration of the cyan group (–CN), generated from the deprotonation of –C–NH₂.³⁸ These observations revealed that 2D Zn-MOF-NH₂ not only changed the morphological structure of K-CZ-2 but also introduced more cyan groups (a type of nitrogen defect) into the building blocks of K-CZ-2. The cyan group can adjust the band structure of K-CZ-2 to enhance its activity.³⁹ It also served as an oxygen adsorption site, where local charge polarization facilitated O₂ adsorption and protonation.³⁹ All the samples exhibited type III adsorption isotherms and H3 type hysteresis loops (Fig. S2a and b), indicating that the samples consisted of slit pores caused by stacked nanosheets. Additionally, K-CZ-2 showed a smaller CA (20.8°) and higher surface energy (68.454 mN m⁻¹, Fig. S3a and b), which was beneficial for mass transfer during photocatalytic H₂O₂ generation.⁴⁰

X-ray photoelectron spectroscopy (XPS) was conducted to analyse the surface chemical composition (Fig. 2c–e and S4a–c); the C 1s peak at 284.8 eV was used for calibration. The C 1s spectra of the samples could be deconvoluted into three peaks at 284.8, 286.45, and 288.28 eV, corresponding to carbon impurities (C–C), the C–NH_x group on the edge of aromatic units, and sp²-hybridized C (NC–N) in the plane of the aromatic ring (Fig. 2c), respectively.^41,42 Interestingly, compared with g-C₃N₄, the C–NH_x peak of K-C₃N₄-like and K-CZ-2 was more intense, which was attributed to the introduction of the cyan group, as the binding energy of –CN was similar to that of C–NH_x.⁴³ For N 1s (Fig. 2d), the spectrum could also be deconvoluted into four peaks. The peak located at 398.8 eV corresponded to the pyridinic N atom (CN–C, N₂C) in the plane of the aromatic ring, and the binding energy peak at 399.6 eV was attributed to the pyrrolic N atom in N–3C (N₃C).^44,45

The peak at 401.3 eV was assigned to the –NH_x group, and the peak located at 404.7 eV could be attributed to the charge effect generated by π-excitation.^46,47 Moreover, elemental K had binding energies located at 292.1 and 294.8 eV, corresponding to the K 2p_3/2 and K 2p_1/2 peaks (Fig. 2e), respectively. These peaks were also observed for K-C₃N₄-like and K-CZ-2, suggesting the presence of K species in the form of K–N and K–C bonds.⁴⁸

Compared with K-C₃N₄-like, the binding energies of K–C and K–N in K-CZ-2 increased, indicating that K–C and K–N acted as “electron bridges” to transport electrons from the inner layer to the surface layer.⁴⁸ The O 1s spectra (Fig S4b) for ZnO could be divided into two peaks at 530.7 and 532.1 eV, corresponding to lattice oxygen and oxygen vacancies, respectively.⁴⁹ In contrast, the O 1s spectrum of K-CZ-2 showed only one peak at 532.1 eV, corresponding to O–H,⁵⁰ indicating that K-CZ-2 did not contain a Zn–O bond. Furthermore, the Zn 2p spectra of ZnO (Fig. S4c) could be deconvoluted into two peaks at 1022.3 and 1045.4 eV, corresponding to Zn 2p_3/2 and Zn 2p_1/2, respectively.⁵⁰ The Zn content in K-CZ-2 was too low and unevenly distributed, which might have resulted in too little Zn in the analysed region to be detected. Electron paramagnetic resonance (EPR) spectroscopy is a powerful technique for directly probing unpaired electrons. As shown in Fig. 2f and S5, all samples presented Lorentzian lines with a g-value of 2.003, corresponding to the delocalized electrons of the π-conjugated aromatic ring.⁵¹ As shown in Fig. 2f, after regulation with Zn-MOF-NH₂, the vacancy intensity of K-CZ-2 significantly increased compared to that of K-C₃N₄-like. As a metal oxide, ZnO contains more O vacancies. According to the XPS data (Tab. S1), the O atom content of K-C₃N₄-like and K-CZ-2 remained unchanged, while the C/N ratios were 0.85 and 0.87, respectively; both are higher than the ideal C/N ratio of 0.75 for g-C₃N₄. Thus, both K-C₃N₄-like and K-CZ-2 contained nitrogen vacancies,⁵² and the ESR results indicated that the nitrogen vacancy intensity of K-CZ-2 was much higher than that of K-C₃N₄-like. It is widely accepted that N vacancies are both active sites and dangling bonds of local electrons. More N vacancies can improve the photocatalytic performance while inhibiting electron–hole recombination.

According to the XRD and XPS analyses, K was in the interlayer and formed K–C and K–N bonds. So, we analyzed the differential charge density, electron localization function (ELF), and Mulliken charge of g-C₃N₄ and K-CZ-2 (Fig. 2g–i and S6, S7). Compared with the independent layers of g-C₃N₄, in K-CZ-2, electron depletion near K and the outer C/N layers, and accumulation in the inner layers indicate asymmetric charge redistribution (Fig. 2g). Although the differential charge density only reflects spatial redistribution relative to isolated atoms, this pattern suggests that the presence of K modulates the electronic environment between the layers. ELF is a dimensionless quantity that takes values between 0 and 1, where ELF = 1 corresponds to perfect localization and ELF = 0.5 corresponds to a uniform electron gas.⁵³ The ELF values of g-C₃N₄ and K-CZ-2 are displayed in Fig. 2h. The ELF values of C–N and CN in the triazine ring of g-C₃N₄ (1) and K-CZ-2 (1) were greater than 0.5 and close to 1, indicating that they had highly localized charge density. This further suggests that these sites might act as potential electron donors. The large electronegativity difference between K (0.82) and C (2.55) or N (3.04) indicates that the ionic character dominated the bonding; however, the ELF values suggest the presence of partial covalent character (g-C₃N₄ (2) and K-CZ-2 (2)). In other words, K atoms could chemically bond with atoms in the two adjacent layers to form charge-transport channels, bridging the layers and promoting the flow of electrons in the inner layers, thus reducing electron localization. It could be observed that the ELF values of the K atom (K-CZ-2 (3)) and N connected to K on the surface layer of K-CZ-2 increased significantly after K was added, indicating that electrons migrated between K and N. This further verified that K and N formed an electronic bridge, which is consistent with the conclusion drawn from the XPS data of K-CZ-2. As shown in Fig. 2i, compared to g-C₃N₄, K-CZ-2 exhibited a decrease in the Mulliken charge of the C atoms and an increase in that of the N atoms. In addition, the doped K atoms had a Mulliken charge of +1.857e. It is well known that a positive Mulliken charge indicates electron loss by an atom, whereas a negative Mulliken charge indicates electron loss. The decrease in the charge of C was likely due to the intercalation of K⁺, which induced a redistribution of the interlayer charges, leading to changes in the polarity of the C–N bonds and thus affecting the charge distribution. Meanwhile, the increase in the charge of N occurred because the electrons lost by K were transferred to the conjugated π-system of g-C₃N₄, particularly to the more electronegative N atoms, enhancing their electron-accepting ability. These results suggest that across layers, electrons migrated from the inner-layer atoms to the K dopant, and then from the K atoms to the surface N atoms via K–N bonding, thereby forming an efficient electron-transport channel. This unique electronic structure contributed to the transfer and separation of charge carriers.

3.2 Band structure and carrier separation characteristics of K-CZ-2

The band structure of the photocatalysts was further studied using UV-visible diffuse reflection spectroscopy (DRS) and the Mott–Schottky technique. DRS was utilized to determine the visible-light absorption capabilities of all the samples. As shown in Fig. 3a, because ZnO was a black material, it had a very wide light absorption range, while the light absorption ranges of K-C₃N₄-like and K-CZ-2 were similar. The bandgaps of all the samples (Fig. 3b) were calculated based on the Tauc plots derived from the transformed Kubelka–Munk function. Furthermore, the work function and valence band maximum of K-CZ-2 were determined to be 3.2 eV and 1.45 eV, respectively, by UPS measurements (Fig. 3c). Combined with the DRS data, the conduction band was calculated to be −1.14 eV. In addition, the Mott–Schottky curves (Fig. S8) of all samples indicated n-type semiconductor behaviour, and the conduction band position of K-CZ-2 was found to be −1.18 eV, which was in good agreement with the UPS results. The band structures of the samples were calculated based on the bandgaps (Fig. 3d). Additionally, the carrier concentration of the samples was estimated from the slope of the Mott–Schottky curves.^54–56 As shown in Fig. 3e, K-CZ-2 had the highest carrier concentration, mainly because K-CZ-2 formed an electron bridge that promoted the rapid migration of photogenerated electrons and reduced their recombination rate with holes. The N vacancies within K-CZ-2 acted as local free electrons associated with dangling bonds, thus increasing the carrier concentration to the highest level.

To explore the separation and transfer behaviours of the photoinduced charge carriers, photoluminescence (PL), photocurrent spectra (I–t), and electrochemical impedance spectra (EIS) were acquired. As shown in Fig. 3f, the PL spectral intensity of K-CZ-2 was the lowest, indicating that the recombination rate of photogenerated carriers was the lowest. This is because the photogenerated electrons in K-CZ-2 were captured and anchored by N vacancies during the transition from the valence band to the conduction band, preventing them from recombining with the holes. As a result, the photocurrent density of K-CZ-2 is the highest (Fig. 3g), and as shown in Fig. 3h, K-CZ-2 exhibits the smallest semicircle radius, indicating the lowest charge transfer resistance, which suggests more efficient interfacial electron transfer during the photocatalytic process. This result aligns well with the PL and I–t data.

3.3 Photocatalytic H₂O₂ production activity and selectivity of K-CZ-2

Upon obtaining the structural information and physicochemical properties, we evaluated the photocatalytic performance of the synthesized sample in a self-designed reactor with 6 vol% benzyl alcohol solution under light irradiation (LED light, 50 W, λ > 420 nm). The H₂O₂ yield was detected using the iodometric method. Based on these measurements, we optimized the amount of 2D Zn-MOF-NH₂ in the composite photocatalyst (Fig. 4a), with and without a sacrificial agent (Fig. S9a), and with different amounts of sacrificial agent (Fig. S9b), and altered the catalyst dosage to test H₂O₂ synthesis (Fig. S9c and d). Compared with K-C₃N₄-like and ZnO, all the K-CZ-X samples exhibited an enhanced H₂O₂ yield rate. As shown in Fig. 4a, the H₂O₂ yield of g-C₃N₄ was only 184 μM (0.36 mmol g⁻¹ h⁻¹), whereas the H₂O₂ generation rate of K-C₃N₄-like increased to 2610.9 μM (5.2 mmol g⁻¹ h⁻¹), indicating that the introduction of K⁺ shortened the interlayer spacing and promoted the rapid transport of photogenerated electrons to the catalyst surface to participate in the photocatalytic reaction. Remarkably, after further decoration with 2D Zn-MOF-NH₂, K-CZ-2 showed a dramatic increase in the H₂O₂ activity, reaching 3878.5 μM (7.8 mmol g⁻¹ h⁻¹), which was 1.5 times higher than that of K-C₃N₄-like. The photocatalytic H₂O₂ activity of ZnO (25.8 μM, 51.6 μmol g⁻¹ h⁻¹) was significantly lower than that of K-CZ-2, indicating a synergistic effect between K⁺ and the N vacancies. K–C and K–N acted as “electron bridges” to transport electrons from the inner layer to the surface layer, while N vacancies served as active sites and local electron traps, effectively promoting the photocatalytic H₂O₂ performance compared to that of K-CZ-2 and recently reported CN-based catalysts for H₂O₂ production. The standardized output of H₂O₂ demonstrated that K-CZ-2 is an excellent photocatalyst for H₂O₂ production (Fig. 4b and Table S2). Notably, the final production of H₂O₂ resulted from the in situ generation and decomposition rate during the actual reaction process. To evaluate the photon utilization efficiency, the apparent quantum yield (AQY) of K-CZ-2 was measured. As shown in Fig. 4c, the AQY of K-CZ-2 exhibited a wavelength dependence, which was in accordance with its absorption spectrum. The maximum AQY of K-CZ-2 at 420 nm reached 3.08% and the solar-to-chemical energy conversion (SCC) efficiency was measured to be 0.63%. The zero-order kinetics of H₂O₂ generation (Fig. 4a) and the first-order kinetics of H₂O₂ decomposition (Fig. S10) were calculated. As shown in Fig. 4d, compared with K-C₃N₄-like, ZnO, K-CZ-1, and K-CZ-3, K-CZ-2 exhibited lower first-order kinetics and the highest zero-order kinetics, indicating that the synergistic effect between K⁺ and the N vacancies more effectively promoted the photocatalytic formation of H₂O₂ while inhibiting its decomposition.

To clarify the formation mechanism and pathway of H₂O₂ during the reaction process, systematic active species capture experiments and rotating disk electrode (RDE) measurements were conducted on K-CZ-2; K₂Cr₂O₇, p-benzoquinone (PBQ), and tert-butanol (TBA), were used as scavengers for electrons (e⁻), superoxide radicals (˙O₂⁻), and hydroxyl radicals (˙OH), respectively. As shown in Fig. 4e and S11a, b, the H₂O₂ yield decreased significantly when K₂Cr₂O₇ and PBQ were introduced into the reaction system, suggesting that H₂O₂ production mainly originated from the ORR. The addition of TBA had little effect on the H₂O₂ yield (Fig. S11c), indicating that ˙OH did not play a significant role in H₂O₂ formation. Moreover, K-CZ-2 exhibited stronger O₂-TPD signals, revealing that the decoration of 2D Zn-MOF-NH₂ dramatically increased the O₂-adsorption capacity (Fig. 4f). Correspondingly, characteristic signals of DMPO-˙O₂⁻, TEMPO-e⁻, and DMPO-˙OH were detected in the EPR spectra of K-CZ-2 (Fig. 4g). Illustrating that H₂O₂ was likely produced by a sequential two-step ORR. Moreover, the number of electrons transferred (n), as a key index for estimating the selectivity of the ORR process, was determined by electrochemical RDE measurements (Fig. 4h and S12). The average number of transferred electrons was 2.2, further indicating that K-CZ-2 had a high selectivity for 2e⁻ ORR in H₂O₂ generation. Combined with the electrochemical analysis and photocatalytic performance measurements, it was further confirmed that the decoration of 2D Zn-MOF-NH₂ effectively promoted the selectivity for the 2e⁻ ORR, which in turn enhanced the H₂O₂ generation activity.

3.4 Photocatalytic ORR mechanism

Based on the above analysis of the experiments and characterization, the mechanism for the high selectivity of H₂O₂ generation by K-CZ-2 photocatalysis is shown in Fig. 5. Firstly, K-CZ-2 has a smaller and thinner lamellar structure and better hydrophilicity, which enhanced the light absorption of the catalyst and promoted mass transfer during the reaction. Secondly, the coordination bond formed by K and C/N between the layers acted as an electron bridge, creating an electron transport channel between the layers and transferring electrons from the inner layer to the surface layer. This greatly reduced the recombination rate of photogenerated carriers, ensuring a sufficient electron supply in the reaction system. In addition, the abundant N vacancies served as active sites for the reaction while also acting as local electrons traps, providing a robust platform for photocatalytic H₂O₂ production. Thirdly, K-CZ-2 had a suitable band structure, with a VB potential of −1.18 eV and a CB potential of +1.41 eV, which fully met the redox potential required for the O₂ →˙O₂⁻ → H₂O₂ process. As a result, O₂ adsorbed on the surface of K-CZ-2 was first reduced to ˙O₂⁻ by electrons, and then reacted with protons to generate H₂O₂.

3.5 Application potential

The results of the real-time pH change experiment showed that pH decreased as the experiment progressed (Fig. S13a), mainly because BA and holes reacted to generate benzaldehyde and H⁺, which increased the H⁺ concentration in the reaction system and caused the pH to decrease. In the experiments with different initial pH values (Fig. 6a), the catalyst showed good H₂O₂ production performance in the range of pH 5–11, indicating that the catalyst had a wide pH application range. K-CZ-2 produced more H₂O₂ in different cation and anion (2 mM) solutions (Fig. S13b and 6b), but the H₂O₂ yield in H₂PO₄⁻ solution was greatly reduced. Overall, K-CZ-2 was suitable for most solutions containing common anions and cations for H₂O₂ production. In addition, K-CZ-2 also produced a high concentration of H₂O₂ in citric acid and humic acid solutions of different concentrations (Fig. 6c and d), demonstrating that K-CZ-2 was also suitable for producing H₂O₂ in dissolved organic acids. From the above, it can be seen that K-CZ-2 can be used in some solutions or wastewater containing specific substances. Furthermore, K-CZ-2 produced a large amount of H₂O₂ in all four pollutants (Fig. 6e), making it possible to remove pollutants from the solution using the H₂O₂ produced by the catalyst itself. K-CZ-2 could still produce H₂O₂ in different authentic water matrices (Fig. 6f), which means that K-CZ-2 can generate H₂O₂ in the field by utilizing natural water.

To explore the stability and durability of K-CZ-2, five-cycle experiments were conducted (Fig. 6g). The experimental results showed that the yield decreased by about 16% in the second cycle, but remained almost unchanged from the second to the fifth cycle, indicating that the catalyst exhibited good durability. Additionally, the XRD and FT-IR spectra of the catalyst after five cycles were analysed (Fig. 6h and i). The results showed that the XRD and FT-IR spectra did not change, indicating that the catalyst maintained good stability.

4. Conclusions

In summary, we presented a two-step synthesis strategy for introducing alkali metal dopants and surface defect sites (N vacancies) into graphite-phase carbon nitride (g-C₃N₄) for effective photocatalytic H₂O₂ production. The well-designed K-CZ-2 achieved an excellent photocatalytic H₂O₂ yield of 7.8 mmol g⁻¹ h⁻¹ in 6 vol% BA solution, surpassing that of the most reported CN-based photocatalysts. The introduction of alkali metal dopants successfully shortened the interlayer spacing and formed charge delivery channels that bridged the layers, promoting electron transport. Further introduction of N vacancies captured abundant electrons transmitted from the inner layer to the surface through electronic bridges and served as active reaction sites. Thus, their synergy facilitated optimized carrier separation and surface reactions, which were essential for the elevated H₂O₂ activity of K-CZ-2. Moreover, K-CZ-2 demonstrated outstanding H₂O₂ production performance across a wide pH range, in the presence of common anions and cations, dissolved organic acids, pollutant solutions, and authentic water matrices, indicating its strong practical application potential. Our work provides new insights for simultaneously achieving metal dopant and vacancy introduction, and expands the application of modified g-C₃N₄ in the field of H₂O₂ production via the 2e⁻ ORR.

Author contributions

L. M.: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, writing-original draft; D. S.: conceptualization, formal analysis, funding acquisition, visualization; Q. H.: data curation, formal analysis; C. P.: data curation, formal analysis; L. Y.: data curation. L. H.: data curation; Z. Q.: data curation; L. Z.: data curation; Z. S.: formal analysis, funding acquisition, project administration, resources, supervision, writing-review & editing.

Conflicts of interest

There are no conflicts to declare.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article and/or its SI. The data that support the findings of this study are available from the corresponding author upon reasonable request. Supplementary information is available. Detailed information regarding the characterization analysis, DFT calculations, supporting figures (S1–S13), supporting tables (S1 and S2), and corresponding references. See DOI: https://doi.org/10.1039/d5ta04269k.

Acknowledgements

The study was supported by grants from the National Key Research and Development Program of China (Grants No. 2023YFC3706505 and 2023YFC3707701), State Key Laboratory of Geomicrobiology and Environmental Changes, China University of Geosciences (No. GKZ25Y665), the National Natural Science Foundation of China (Grant No. 42407117).

References

1. X. Y. Shen, Z. M. Wang, H. Z. Guo, Z. D. Lei, Z. Liu and L. Wang, Small, 2023, 19(43), 2303156 CrossRef CAS PubMed .
2. Y. Li, J. X. Chen, Y. X. Ji, Z. L. Zhao, W. J. Cui, X. H. Sang, Y. Cheng, B. Yang, Z. J. Li, Q. H. Zhang, L. C. Lei and Z. H. Wen, Angew. Chem., Int. Ed., 2023, 135(34), e202306491 CrossRef .
3. Y. J. Zhang, J. S. Tao, Y. Hu, G. X. Huang, Y. Pan, W. W. Li, J. J. Chen and H. Q. Yu, Nat. Water, 2024, 2, 770–781 CrossRef CAS .
4. L. Y. Li, X. M. Lv, Y. Y. Xue, H. B. Shao, G. F. Zheng and Q. Han, Angew. Chem., Int. Ed., 2024, 63(15), e202320218 CrossRef CAS PubMed .
5. R. Y. Liu, Y. Z. Chen, H. D. Yu, M. Položij, Y. Y. Guo, T. C. Sum, T. Heine and D. L. Jiang, Nat. Catal., 2024, 7, 195–206 CrossRef CAS .
6. M. H. Liu, S. Yang, S. J. Liu, Q. Y. Miao, X. B. Yang, X. W. Li, Q. Xu and G. F. Zeng, Small, 2022, 18(50), 2204757 CrossRef CAS PubMed .
7. W. T. Li, B. Han, Y. H. Liu, J. Y. Xu, H. R. He, G. G. Wang, J. S. Li, Y. X. Zhai, X. L. Zhu and Y. F. Zhu, Angew. Chem., Int. Ed., 2024, 64, e202421356 CrossRef PubMed .
8. J. X. Su, L. Jiang, B. B. Xiao, Z. X. Liu, H. Wang, Y. F. Zhu, J. Wang and X. F. Zhu, Small, 2024, 20(24), 2310317 CrossRef CAS PubMed .
9. S. Fukuzumi, Y. M. Lee and W. Nam, Chem.–Eur. J., 2018, 24, 5016 CrossRef CAS PubMed .
10. Y. Yang, Q. Y. Guo, Q. W. Li, L. P. Guo, H. Q. Chu, L. J. Liao, X. P. Wang, Z. Z. Li and W. Zhou, Adv. Funct. Mater., 2024, 34, 2400612 CrossRef CAS .
11. J. Xiao, X. L. Liu, L. Pan, C. X. Shi, X. W. Zhang and J. J. Zou, ACS Catal., 2020, 10(20), 12256–12283 CrossRef CAS .
12. N. Gou, W. Y. Yang, S. Gao and Q. Li, J. Hazard. Mater., 2023, 447, 130795 CrossRef CAS PubMed .
13. L. Li, H. H. Liu, C. Cheng, X. Y. Dai, F. Chen, J. Q. Ning, W. T. Wang and Y. Hu, ACS Catal., 2024, 14, 10204–10213 CrossRef CAS .
14. J. Y. An, W. J. Jiang, F. W. Zhuang, Y. H. Ma, S. Zhan and F. Zhou, ACS Catal., 2024, 14(22), 17014–17024 CrossRef CAS .
15. M. Wu, X. He, B. H. Jing, T. Wang, C. Y. Wang, Y. L. Qin, Z. M. Ao, S. B. Wang and T. C. An, J. Hazard. Mater., 2020, 384(15), 121323 CrossRef CAS PubMed .
16. M. M. Ma, J. Z. Li, X. G. Zhu, K. Liu, K. G. Huang, G. D. Yuan, S. Z. Yue, Z. J. Wang and S. C. Qu, Carbon Energy, 2024, 6(3), e447 CrossRef CAS .
17. C. F. Gao, Y. Sun, S. S. Yu, L. F. Liu, C. Y. Liu, Y. H. Li, H. B. Wang and X. B. Chang, Chem. Eng. J., 2024, 500, 156944 CrossRef CAS .
18. S. Wu, H. T. Yu, S. Chen and X. Quan, ACS Catal., 2020, 10(24), 14380–14389 CrossRef CAS .
19. C. Y. Feng, X. L. Ouyang, Y. C. Deng, J. J. Wang and L. Tang, J. Hazard. Mater., 2023, 441, 129845 CrossRef CAS PubMed .
20. D. M. Zhao, C. L. Dong, B. Wang, C. Chen, Y. C. Huang, Z. D. Diao, S. Z. Li, L. J. Guo and S. H. Shen, Adv. Mater., 2019, 31(43), 1903545 CrossRef CAS PubMed .
21. H. R. Miao, W. Q. Zhang, T. Wang, Z. M. Yang and C. C. Kong, Prog. Nat. Sci. Mater. Int., 2023, 33(4), 407–424 CrossRef CAS .
22. J. K. Lin, W. J. Tian, H. Y. Zhang, H. Q. Sun and S. B. Wang, Acc. Chem. Res., 2024, 57(16), 2303–2315 CrossRef CAS PubMed .
23. M. Liu, H. L. Qin, H. M. Xu and Y. Zheng, Sep. Purif. Technol., 2025, 355, 129785 CrossRef CAS .
24. W. Miao, Y. J. Wang, Y. Liu, H. H. Qin, C. C. Chu and S. Mao, Engineering, 2023, 25, 214–221 CrossRef CAS .
25. F. Bi, Q. J. Meng, Y. L. Zhang, X. L. Weng and Z. B. Wu, ACS Appl. Mater. Interfaces, 2023, 15(26), 31409–31420 CrossRef CAS PubMed .
26. H. Yuan, H. R. Sun, Y. X. Shi, J. X. Wang, A. Bian, Y. Y. Hu, F. Guo, W. L. Shi, X. Du and Z. H. Kang, Chem. Eng. J., 2023, 472, 144654 CrossRef CAS .
27. P. F. Wu, C. L. Zhou, Y. P. Li, M. H. Zhang, P. X. Tao, Q. L. Liu and W. Q. Cui, Appl. Surf. Sci., 2021, 540(2), 148362 CrossRef CAS .
28. G. Chen, Y. P. Zhu, Y. R. Ying, Y. D. Yao, Z. W. Hu, D. Zu, Z. Z. Lin, C. W. Pao, Y. C. Zhang, L. Li, Y. Zhu and H. T. Huang, Matter, 2024, 7(6), 2265–2277 CrossRef CAS .
29. J. Lu, S. Wang, Y. Zhao, K. Ge, J. B. Wang, H. Cui, Y. Yang and Y. F. Yang, Catal. Commun., 2023, 175, 106613 CrossRef CAS .
30. D. D. Liu, L. P. Jiang, D. Q. Chen, Z. K. Hao, B. W. Deng, Y. Y. Sun, X. Liu, B. Y. Jia, L. M. Chen and H. T. Liu, ACS Catal., 2024, 14(7), 5326–5343 CrossRef CAS .
31. W. W. Wang, S. J. Song, P. Wang, M. He, Z. Fang, X. L. Yuan, H. Li, C. Y. Li, X. Wang, Y. C. Wei, W. Y. Song, H. Xu and Z. X. Li, ACS Catal., 2023, 13(7), 4597–4610 CrossRef CAS .
32. X. Ma and H. F. Cheng, Sep. Purif. Technol., 2024, 330, 125260 CrossRef CAS .
33. T. Xiong, W. L. Cen, Y. X. Zhang and F. Dong, ACS Catal., 2016, 6(4), 2462–2472 CrossRef CAS .
34. G. G. Zhang, G. S. Li, Z. A. Lan, L. H. Lin, A. Savateev, T. Heil, S. Zafeiratos, X. C. Wang and M. Antonietti, Angew. Chem., Int. Ed., 2017, 56(43), 13445–13449 CrossRef CAS PubMed .
35. S. Z. Hu, F. Y. Li, Z. P. Fan, F. Wang, Y. F. Zhao and Z. B. Lv, Dalton Trans., 2015, 44(3), 1084–1092 RSC .
36. W. S. Hou, Y. X. Li, S. X. Ouyang, H. Y. Chen, J. H. Ye, X. P. Han and Y. D. Deng, Chem. Commun., 2019, 55(88), 13279–13282 RSC .
37. S. Y. Zhong, Y. Y. Xiang, Y. Y. Dai, Y. Y. Wang, W. F. Su, S. M. Li and J. Li, J. Membr. Sci., 2024, 708, 123058 CrossRef CAS .
38. C. Y. Feng, L. Tang, Y. C. Deng, J. J. Wang, J. Luo, Y. N. Liu, X. L. Ouyang, H. R. Yang, J. F. Yu and J. J. Wang, Adv. Funct. Mater., 2020, 30(39), 2001922 CrossRef CAS .
39. L. Chen, C. Chen, Z. Yang, S. Li, C. H. Chu and B. L. Chen, Adv. Funct. Mater., 2021, 31(46), 2105731 CrossRef CAS .
40. R. R. Deng and Q. B. Zhang, Adv. Energy Mater., 2024, 14(35), 2401444 CrossRef CAS .
41. L. Zhang, F. X. Mao, L. R. Zheng, H. F. Wang, X. H. Yang and H. G. Yang, ACS Catal., 2018, 8(12), 1103–11041 Search PubMed .
42. L. H. Lin, H. H. Ou, Y. F. Zhang and X. C. Wang, ACS Catal., 2016, 6(6), 3921–3931 CrossRef CAS .
43. Y. J. Cui, Z. X. Ding, X. Z. Fu and X. C. Wang, Angew. Chem., Int. Ed., 2012, 51(47), 11814–11818 CrossRef CAS PubMed .
44. Q. H. Liang, Z. Li, X. L. Yu, Z. H. Huang, F. Y. Kang and Q. H. Yang, Adv. Mater., 2015, 27(31), 4634–4639 CrossRef CAS PubMed .
45. T. Y. Ma, S. Dai, M. Jaroniec and S. Z. Qiao, Angew. Chem., Int. Ed., 2014, 126(28), 7281–7285 CrossRef PubMed .
46. Z. Z. Lin and X. C. Wang, Angew. Chem., Int. Ed., 2013, 52(6), 1735–1738 CrossRef CAS PubMed .
47. K. H. Park, B. H. Kim, S. H. Song, J. Y. Kwon, B. S. Kong, K. Kang and S. Jeon, Nano Lett., 2012, 12(6), 2871–2876 CrossRef CAS PubMed .
48. L. X. Su, D. Gong, N. Yao, Y. B. Li, Z. Li and W. Luo, Adv. Funct. Mater., 2021, 31(49), 2106156 CrossRef CAS .
49. H. G. Li, B. Yu and H. Z. Yu, Adv. Funct. Mater., 2024, 34(37), 2402128 CrossRef CAS .
50. H. Sun and S. J. Park, J. Environ. Sci., 2022, 118, 57–66 CrossRef CAS PubMed .
51. X. Zhang, P. J. Ma, C. Wang, L. Y. Gan, X. J. Chen, P. Zhang, Y. Wang, H. Li, L. H. Wang, X. Y. Zhou and K. Zheng, Energy. Environ. Sci., 2022, 15(2), 830–842 RSC .
52. W. K. Wang, H. M. Zhang, S. B. Zhang, Y. Y. Liu, G. Z. Wang, C. H. Sun and H. J. Zhao, Angew. Chem., Int. Ed., 2019, 58(46), 16644–16650 CrossRef CAS PubMed .
53. A. D. Becke and K. E. Edgecombe, J. Chem. Phys., 1990, 92(9), 5397–5403 CrossRef CAS .
54. L. N. Su, P. F. Wang, X. L. Ma, J. H. Wang and S. Zhan, Angew. Chem., Int. Ed., 2021, 133(39), 21431–21436 CrossRef .
55. H. J. Yu, F. Chen, X. W. Li, H. W. Huang, Q. Y. Zhang, S. Q. Su, K. Y. Wang, E. Y. Mao, B. Mei, G. Mul, T. Y. Ma and Y. H. Zhang, Nat. Commun., 2021, 12, 4594 CrossRef CAS PubMed .
56. Z. J. Zhang, T. Tsuchimochi, T. Ina, Y. Kumabe, S. Muto, K. Ohara, H. Yamada, S. L. Ten-No and T. Tachika-wa, Nat. Commun., 2022, 13, 1499 CrossRef CAS PubMed .

This journal is © The Royal Society of Chemistry 2025