《1. Introduction》

1. Introduction

Hydrogen peroxide (H2O2) is one of the most valuable chemicals in pharmaceutical, chemical, food, and environmental industries [1–3]. It is a potential oxidant for carbonaceous fuels in electricity generation [4]. The traditional large-scale production of H2O2 relies on the energy-intensive anthraquinone method [5]. The sustainable in-situ production of H2O2 can significantly reduce the production, storage, and transportation costs with minimized safety risks [6]. Recently, sunlight-driven H2O2 generation using suitable photocatalysts has become a research hotspot [7,8]. Various photocatalysts including metal oxides (TiO2 and BiVO4) [9,10], metal–organic materials (MOMs) [11], metal-free polymers [12], and graphitic carbon nitride (g-C3N4) have been studied for the photocatalytic H2O2 production [13,14]. Among those semiconductors, g-C3N4 is considered a promising photocatalyst owing to the merits of suitable bandgap structure, good thermal and chemical stability, low cost with facile production process [15–18]. However, pristine g-C3N4 only offers a limited visible-light absorption range. It suffers from rapid charge recombination and low charge carrier mobility with a poor selectivity towards the twoelectron (2e–) reaction pathway. Therefore, further enhancement of the photocatalytic activity of g-C3N4 in H2O2 production becomes a hot research topic in recent years.

Introducing defects (nitrogen and carbon vacancies) in g-C3N4 was reported to be effective in enhancing photocatalytic H2O2 production [19,20]. Specifically, N vacancies engineering in g-C3N4 can not only adjust the bandgap structures for harnessing more visible light but also increase the surface energy to promote active sites for oxygen reduction [21]. For instance, with KOH-assisted polymerization of the nitrogen-rich precursors, nitrogen defects and the cyano group were introduced into the g-C3N4 to control the band structure. As the electron-withdrawing groups, nitrogen defects, and cyano groups could improve the charge transportation to delivery charges to the surface terminal sites [22,23]. In addition to KOH, other K or Na salts have also been used to modify g-C3N4 to enhance the production of H2O2 via two specific mechanisms, that is, modifying the pre-synthesized g-C3N4 or participating in the molten polymerization synthesis of g-C3N4 [24–28]. The advantages of the alkali-assisted modification method can be summarized as follows: ① The generated N vacancies inhibit the charge recombination; ② the coexistence of cyano groups and N vacancies affects the bandgap of g-C3N4; ③ introduction of alkali metal dopants broadens the light absorption range; and ④ the incorporation of multiple hetero-elements (O, S, P) into the polymeric CN matrix enhances the charge transfer properties. Although this method can introduce N vacancies in g-C3N4, the formation and identification of nitrogen defects have not received enough attention. Moreover, different types of N vacancies produced in this method may exhibit different photocatalytic performances. Recently, g-C3N4 containing two types of N vacancies (NHx and N2C) has been reported, where NHx promotes the photoexcited charge separation and N2C is responsible for the oxygen activation [29]. N3C vacancy is another type of N vacancy that has been reported in photocatalytic NOx reduction and H2 production [30,31], which has yet been studied in photocatalytic H2O2 production.

Herein, we designed a one-step sodium persulfate (PDS)- assisted molten polymerization method to produce g-C3N4 with N3C vacancies. The effects of N vacancies and defects in g-C3N4 in photocatalysis were characterized by investigating the light absorption performance, charge transfer properties, and the selective two-electron pathway. Importantly, the process and mechanism of PDS-induced formation of N3C vacancy during the polymerization process were also elucidated. The H2O2 evolution study showed that g-C3N4 with the highest amount of N3C offered the best H2O2 evolution activity, which is 4.5 times higher than that of the pristine g-C3N4. To the best of our knowledge, this is the first time to identify the role of the N3C vacancy in g-C3N4 for promoting the photocatalytic production of H2O2. The reported sodium persulfate-assisted synthetic method provides a novel strategy to customize N3C vacancies in g-C3N4 and has significant implications for developing photocatalysts for other applications. 

《2. Experimental》

2. Experimental

《2.1. Chemicals》

2.1. Chemicals

Sodium persulfate (Na2S2O8, 99.0%) and p-Benzoquinone (p-BQ, 99.0%) were purchased from Aladdin Chemistry Co., China. Melamine, ethanol (analytical reagent (AR), 95.0%), and sodium sulfate (Na2SO4) were supplied by Macklin Biochemical Co., Ltd. (China). Ethanol (99.5%) and potassium iodide (KI) were obtained from Sinopharm Chemical Reagent Co., Ltd. (China). Methanol (MeOH, high performance liquid chromatography (HPLC)) and isopropanol (IPA, HPLC) were bought from Anpel Laboratory Technologies Inc., China. Ultrapure water (18.2 MΩ·cm–1 ) produced from Milli-Q System was used in all experiments. 

《2.2. Preparation of the catalysts 》

2.2. Preparation of the catalysts 

Before the synthesis, 10 g melamine was mixed with 0.33 mol Na2S2O8 followed by grounding for 30 min to form a uniform precursor. This mixture was heated to 550 °C in an Ar atmosphere at a heating rate of 5 °C·min–1 and kept for 2 h. The obtained solid was ground and washed with ultrapure water before being dried at 60 °C. The obtained sample was named sodium persulfate doped g-C3N4 (SPCN). The sodium sulfate doped g-C3N4 (SSCN) was prepared by the same method with 10 g melamine mixed with 0.33 mol of sodium sulfate (Na2SO4). The pristine g-C3N4 (CN) was prepared by the same method with only melamine as the precursor. 

《2.3. Characterizations》

2.3. Characterizations

X-ray diffraction (XRD) patterns were collected on a Bruker D8- Advanced X diffractometer (Germany) with Cu–Kα radiation. Fourier transform infrared spectroscopy (FTIR) spectra were acquired on a Nicolet 380 spectrometer (USA). The microstructure of the samples was characterized using a scanning electron microscope (SEM; Hitachi S-4800, Japan). X-ray photoelectron spectroscopy (XPS) study was conducted by a Thermo Scientific K-Alpha spectroscope (USA). Thermogravimetry analysis (TGA) was measured on an FEI Themis Z (USA), and the sample was heated at a rate of 10 °C·min–1 up to 550 °C under Ar atmosphere (30 mL·min–1 ). Ultraviolet–visible diffuse reflectance spectra (UV–vis DRS) were collected on a UV-2450 spectrophotometer (Shimadzu, Japan) equipped with an integrating sphere with BaSO4 as the reference. The O2 temperature-programmed desorption (TPD) measurement was performed on a Thermo Scientific ESCALAB XI+. Electron paramagnetic resonance (EPR; Bruker A300-10/12) was performed to detect the nitrogen vacancy. The photoluminescence (PL) emission spectra were obtained on a Hitachi F-7000 fluorescence spectrophotometer at an excitation wavelength of 355 nm. The zeta potentials of the prepared samples were measured by a Zetasizer Nano ZS90 analyzer (China). The setup of a photoelectrochemical reactor is shown in Text S1 in Appendix A. 

《2.4. Photocatalytic H2O2 evolution》

2.4. Photocatalytic H2O2 evolution

H2O2 evolution tests were conducted in a photocatalytic reactor. Typically, 25 mg photocatalyst was dispersed into 50 mL of isopropanol/H2O mixture (10/90, v/v). The suspension was stirred and ultrasonicated for 5 min. A 300 W white-light light emitting diode (LED; CEL-PE300-4A; AuLight, China) with a 420 nm cutoff filter was used as the light source with a light intensity of 100 mW·cm–2 (at the upper surface of the solution). During the reaction, the solution was stirred continuously without pumping in O2. The solution was sampled every 10 min, which was filtered using a 0.22 μm Millipore filter. The concentration of H2O2 was determined by iodometric titration with a UV–vis spectrophotometer (see details in Fig. S1 in Appendix A) [32].

《3. Results and discussion》

3. Results and discussion

《3.1. Structure and chemical composition of the catalyst》

3.1. Structure and chemical composition of the catalyst

As shown in Fig. S2 in Appendix A, SSCN and SPCN have similar morphologies to pristine CN. However, porous structure and rough surface were observed due to the etching by Na2SO4/Na2S2O8 eutectic salts during the thermal treatment [27]. The XRD patterns (Fig. 1(a)) of all prepared carbon nitride samples show two characteristic peaks at 13.2° and 27.0°, which are assigned to the (100) and (002) crystal planes of g-C3N4, respectively. Compared with pristine CN, both peaks were weakened with the addition of Na2SO4 or Na2S2O8. This weakening in XRD suggests the reduction of ordering in the tri-s-thiazine units (100) and the interlayer (002) stacking [33]. There are no other peaks from sodium or sulfur were observed in the XRD patterns due to the extremely low content of sodium and sulfur. However, weak Na 1s signals were identified in the XPS survey spectra from SSCN and SPCN (Fig. S3 in Appendix A). The surface element contents of the different samples are listed in Table S1. The result indicates that no sulfur species were left in the polymeric CN products. As shown in the FTIR spectra (Fig. 1(b)), multiple peaks between 900 and 1800 cm–1 were assigned to the stretching of the melon structural unit (N–C=N) in the aromatic heterocycles. With the addition of sodium salts in the synthesis, the intensities of the breathing mode of the tri-s-triazine at 810 cm–1 and the stretching of –NHx groups as the broad peaks between 3000–3500 cm–1 were decreased. Meanwhile, a new peak at 2177 cm–1 corresponding to cyano groups emerged, which was formed by the deprotonation of the terminal –C–NH2. 

《Fig. 1》

Fig. 1. Chemical composition of as-synthesized photocatalysts. (a) XRD patterns, (b) FTIR spectra, (c) XPS N1s spectra, and (d) XPS C1s spectra of CN, SSCN, and SPCN. The schematic illustrations of N groups (blue) in (e) pristine g-C3N4, and (f) the N vacancies (red) in modified g-C3N4.

The N/C and O/C atomic ratios were measured (Table S1) to further investigate the effects of Na2S2O8 modification on g-C3N4. Compared with pristine CN, the O/C ratios for both SSCN and SPCN are both increased, accompanied by the decrease in the N/C ratio. Similar O/C ratios are detected for SSCN and SPCN, indicating that sodium persulfate does not introduce more oxygens into the CN framework than sodium sulfate. Considering that the formation of cyano groups does not affect the N contents, other forms of defects, such as N vacancies, must be produced with the addition of Na2S2O8 or Na2S2O8. SPCN has the lowest N/C ratio, which is expected to have the highest density of N defects [22]. To confirm the origin of N defects, the high-resolution XPS spectra of N1s and C1s are analyzed (Figs. 1(c) and (d)). Fig. 1(c) shows that the N1s region is formed by three characteristic peaks including the sp2 hybridized aromatic nitrogen (C–N=C, N2C) at 398.8 eV, the tertiary nitrogen (N–(C)3, N3C) at 400.3 eV, and the uncondensed amino groups N–Hx (x = 1, 2) at 401.1 eV [34]. It is noticed that the decrease of the N content in the modified CN manifests the generation of N vacancies as shown in Figs. 1(e) and (f). As listed in Table 1, the relative compositions of N3C and NHx were decreased along with the increase of N2C for the SPCN sample, which reveals that more N3C vacancies were produced rather than the N2C vacancies [29]. The content of N3C in SPCN is 10.14%, which is lower than that of CN (14.10%) and Na2SO4 (14.08%), indicating that the persulfate induces a large amount of N3C vacancies in SPCN. 

As shown in Fig. 1(d), the C1s region of the pristine CN was deconvoluted into two characteristic peaks of C–C bonds at 286.4 eV and sp2 carbon (N=C–N) at 288.1 eV [35]. A new peak corresponding to the or C–NHx appears at 286.4 eV for the modified CN [28]. The intensity of the 286.4 eV peak in SPCN is higher than that in SSCN. Based on the XPS results of N1s, the content of NHx is similar in SSCN and SPCN. Therefore, the extra peak intensity at 286.4 eV is associated with the higher content in SPCN than in SSCN, providing additional evidence for a higher amount of cyano groups shown in the FTIR spectra in Fig. 1(b). 

《Table 1》

Table 1 The contents of N and C coordination groups in the catalysts estimated by XPS results.

《3.2. Formation of N3C defects 》

3.2. Formation of N3C defects 

To study the phase transformation during the eutectic crystallization of persulfate and melamine, thermogravimetry-derivate thermogravimetric (TG/DTG) measurement was performed to simulate the thermal environment of the synthesis process. As shown in the inset of Fig. 2(a), there is no weight loss before 210 °C, although the decomposition temperature of persulfate is 180 °C. The improved thermal stability may be due to the synergistic effect of electrostatic and hydrogen bonding interaction between the O–O (Na2SO3–O–O–SO3) bond and –NH2 groups of melamine [36,37]. The rapid weight loss peak appeared at the temperature between 210 and 318 °C, associated with the deamination, sublimation, and thermal condensation of melamine [38]. Among them, melamine doped by persulfate showed the least weight loss due to the aforementioned synergistic stabilization effect inhibiting the release of NH3 and the degree of condensation. Meanwhile, melamine doped by sodium sulfate showed less weight loss than pure melamine. Additionally, eutectic Na salts could improve the mass transport of the reactants and achieve a more uniformly distributed dopant in carbon nitride [24]. Notably, there is another weight loss peak at a higher temperature of 412 °C only in the persulfate doped melamine sample, implying a further polymerization to generate small molecule fragments containing oxygen and nitrogen. Compared with the typical phase formation from melamine to g-C3N4, this process forms more cyano and nitrogen vacancies. 

《Fig. 2》

Fig. 2. The formation of N3C vacancies during the polymerization process of SPCN. (a) TG/DTG curves of the prepared catalysts. (b) Solid-state 13C NMR spectra of SPCN. (c) EPR spectra of the prepared catalysts. (d) The speculated formation route of N3C vacancies during the persulfate-assisted molten polymerization process of CN. ppm: parts per million.

Solid-state 13C nuclear magnetic resonance (NMR) spectroscopy of SPCN showed two dominant peaks at 156.4 and 164.5 parts per million (ppm), which corresponds to the chemical shifts of C2N– NHx (peak 2) and C3N (peak 1) in the heptazine units, respectively (Fig. 2(b)). This phenomenon indicates that the PDS-assisted eutectic polymerization process does not alter the main heptazine core structure [26]. Notably, a broad peak at 115.8 ppm assigned to the aromatic pyridine carbon species (peak 3) in the CN framework was observed, implying the formation of N3C vacancies [39]. Solidstate EPR spectra were collected to investigate the electronic band structure of nitrogen vacancies. As shown in Fig. 2(c), the Lorentz signal with g = 2.003 was recorded for all three samples, which is assigned to the unpaired electrons of the sp2 hybridized carbon atoms on the π-conjugated planes of g-C3N4. The higher intensity in the EPR signal represents more unpaired electrons, associated with the higher population of nitrogen vacancies introduced by the thermal eutectic persulfate treatment [40]. Combining the EPR the XPS results, N3C vacancies were mainly introduced in SPCN. In this regard, a higher number of unpaired electrons can facilitate the photocatalytic H2O2 evolution reaction [41].

The zeta-potential analysis was used to further understand the surface properties of the catalysts. As shown in Fig. S4 in Appendix A, CN exhibits the zero-point charge (pHZPC) at pH 3. In contrast, the SSCN and the SPCN surfaces were negatively charged over the entire pH range from 2 to 9. This result indicates the electron-withdrawing ability of SPCN was enhanced due to the strong electronegativity of the N atoms in the and/or unsaturated C atoms surrounding the N3C defects [35,42]. SPCN showed a slightly positive zeta potential than SSCN at the same pH. Based on the element component analysis (Table S1), it can be inferred that the presence of positively charged Na+ ions partially counteracted the negative charges on SPCN [43]. The more negatively charged surface of the photocatalyst can enhance the adsorption of O2, which could subsequently react with H+ to produce H2O2 [44]. 

The schematic illustration of the persulfate-induced surface modification during the synthesis of SPCN is shown in Fig. 2(d). At the first step of polymerization, the released N component was oxidized by the low eutectic salt sodium sulfite salt. In the second step, the oxygen species in the g-C3N4 framework were released to form sodium sulfate. As a consequence, this process generates more defects in SPCN compared to that in SSCN. As shown in Figs. 1(e) and (f), the potential reaction sites (green parts) on the pure CN can be converted into N defects (red parts) following the molten salt eutectic reaction. By summarizing the above observations, the presence of persulfate during the synthesis of g-C3N4 produced more active sites in the form of N3C vacancies and cyano groups for photocatalytic H2O2 evolution. 

《3.3. Band structure and optical properties》

3.3. Band structure and optical properties

The light absorption properties are critical to photocatalytic H2O2 production performance. The UV–vis DRS absorption spectra in Fig. 3(a) show that the catalysts had good absorption in the visible region. Compared with CN, the adsorption edges of both the SPCN and the SSCN are redshifted, indicating improved visible light absorption. The broad increase of the absorption in the long wavelength by SSCN and SPCN indicates the more homogeneous distribution of N defects [22]. SPCN shows the highest absorption in the visible region, which is associated with its highest density of defects. Bandgap energy (Eg) of the samples was calculated using the Kubelka–Munk function (see details in Text S2 in Appendix A) based on the UV–vis DRS data [45]. The results show that the light-harvesting abilities were enhanced from the SSCN and the SPCN in comparison with the pristine CN due to the high population of N defects and cyano groups introduced during the thermal polymerization of melamine with the presence of eutectic sodium salt. 

To identify the band edges of the catalysts, the work function and the valence band (VB) of the catalyst were determined by ultraviolet photoelectron spectroscopy (UPS) (Fig. 3(b) and Fig. S5 in Appendix A). The cut-off binding energy (EC) of all samples was determined as 16.40 eV. Then, the work function (U) of the catalysts was calculated as 4.82 eV by EHe–I – EC, where EHe–I (He–I photoelectron energy) is the photon energy (21.22 eV) [46]. The energy of VB (EVB) of SPCN was determined to be 6.33 eV (vs vacuum) by Φ + EVB/HOMO (HOMO: the highest occupied molecular orbital). With the Eg from the DRS results, the position of the conduction band (CB) of SPCN was calculated to be 3.68 eV by EVB – Eg. Finally, the EVB and ECB of CN, SSCN, and SPCN were converted to the electrochemical potentials with respect to reversible hydrogen electrode (RHE) (–4.44 eV vs vacuum) [47]. The band alignments of the catalysts are shown in Fig. 3(c).   

《Fig. 3》

Fig. 3. Optical properties and electronic band structure of as-synthesized photocatalysts. (a) UV–vis DRS of the catalysts. Inset shows the corresponding plots of transformed Kubelka–Munk function ((αhv) 1/2, where α is the absorption coefficient, h is Planck constant and v is the optical frequency) vs photon energy (hv). (b) The UPS spectrum of SPCN. (c) The band structure alignments, (d) PL spectra, (e) EIS Nyquist plots, and (f) transient photocurrent responses of the catalysts.

As shown in the photoluminescence (PL) spectra (Fig. 3(d)), the major peak around 465 nm is ascribed to the band–band recombination of photogenerated carriers. The fluorescence intensity of this peak is in the order of CN > SSCN > SPCN, which implies that the charge separation ability was improved with the sodium salts modification. It is also found that this peak was redshifted for the SPCN sample, associated with the reduction of the bandgap energy, which is consistent with the DRS results [44,48]. Moreover, as shown in Fig. 3(e), the decreased electrochemical impedance spectroscopy (EIS) slop of SPCN also reflects the improved charge migration rate and the reduced electric transferring resistance, which benefits the photoelectric activity [49]. Considering the above facts, this enhancement is attributed to the improved charge-trapping offered by the defect sites of cyano groups and N vacancies.

To further investigate the influence of N vacancies on photocatalytic performance, the transient photocurrent responses were obtained, which represented the efficiency of using the photogenerated electrons and holes in the redox reactions. Considering the aforementioned results of light-harvesting capability and charge separation efficiency, the order of photoelectric activity is as follows: SPCN > SSCN > CN. The stable photocurrent density of SPCN (plateau: 0.315 μA·cm–2 ) under visible light irradiation is 1.56- and 3.12-fold higher than that of SSCN and CN, respectively (Fig. 3(f)). Therefore, the introduction of N3C vacancies is beneficial to the charge separation of g-C3N4. Due to the presence of extra electrons in the nitrogen-deficient sites on the catalyst, it is easier to trap the photoinduced holes, which leads to the reduced spatial separation of the photoinduced carriers [50] (details of electrochemical and photoelectrochemical tests are shown in Text S3 in Appendix A).

《3.4. Photocatalytic activity for H2O2 generation 》

3.4. Photocatalytic activity for H2O2 generation 

Fig. 4(a) shows that the production of H2O2 increased linearly with time by different catalysts under visible light irradiation ( ≥ 420 nm). SPCN exhibits the highest photocatalytic activity (323.6 μmol·L–1 ·h–1 ·g–1 ) among the three catalysts. Comparatively, the production rate of SPCN was calculated to be 161.8 μmol·L–1 ·h–1 ·g–1 which is about 4.5-fold higher than that of CN. Compared with the performance of other CN-based photocatalysts in the literature (Table S2 in Appendix A), SPCN shows an outstanding photocatalytic H2O2 generation rate. The produced H2O2 can also be effectively activated by carbon nitride under visible light irradiation to generate hydroxyl radicals (·OH) [21]. Hence, the overall H2O2 production rate represents the dynamic competition between the rates of formation (kf, μmol·L–1 ·min–1 ) and decomposition (kd, min–1 ). The rate constants of kf (zero-order kinetics) and kd (first-order kinetics) were calculated by fitting the H2O2 evolution data shown in Fig. S6(a) in Appendix A into the formula (t represents the irradiation accumulation time) : 

As shown in Fig. S6(b) in Appendix A, the kf and kd of SPCN were 3.05 μmol·L–1 ·min–1 and 0.0043 min–1 , respectively. Interestingly, SPCN exhibited the lowest kd value, implying that the decomposition of H2O2 was hindered. This is possibly due to the improvement in the band edge positions of SPCN. The photoexcited electrons in the conduction band were not negative enough to cause the fast decomposition of the generated H2O2, described in Eq. (2).

From Fig. 4(b), the effects of oxygen on H2O2 production were studied. The H2O2 yield was suppressed from 161.8 μmol·L–1 ·h–1 in the air to 52.7 μmol·L–1 ·h–1 in the argon environment. In contrast, the production of H2O2 increases to 194.7 μmol·L–1 ·h–1 with pure O2 gas purging. Therefore, the H2O2 production rate was significantly enhanced under the saturated O2 condition, suggesting that O2 plays a critical positive role in the reaction. 

《Fig. 4》

Fig. 4. Photocatalytic activity of H2O2 production. (a) The photocatalytic performance in H2O2 production with different photocatalysts. (b) Effect of dissolved oxygen on the photocatalytic production of H2O2. (c) Long-term photocatalytic H2O2 production under visible light and dark conditions. Working condition: 0.5 g·L–1 catalyst suspension, 50 ml IPA/H2O mixture (10/90, v/v), pH 3.0, ≥ 420 nm.

The impact of electron donors on photocatalytic H2O2 production was also investigated. As shown in Fig. S7(a) in Appendix A, the H2O2 production performance in the presence of electron donors was in the order of IPA (161.8 μmol·L–1 ·h–1 ) > MeOH (122.0 μmol·L–1 ·h–1 ) > EtOH (103.1 μmol·L–1 ·h–1 ) > deionized (DI) water (10.0 μmol·L–1 ·h–1 ). The MeOH and EtOH are typical sacrificial agents, which show similar negative effects on H2O2 production. In contrast, IPA is more conducive to selective H2O2 production than the two aliphatic alcohols. The electron donor can boost H2O2production, which is in agreement with the results in previous studies [35,51]. In the case of DI water, the water oxidation reaction by photogenerated holes is slow, which leads to the recombination of photogenerated charges. Herein, IPA, as the electron donor, can promote photocatalytic H2O2 production by consuming the photogenerated holes. Besides, the yield of H2O2 decreases as the pH value increases from 5 to 7 (Fig. S7(b) in Appendix A). This is because the reduced proton concentration in the solution slows down the H2O2 production while increasing the H2O2 decomposition rate [52]. The stability and cycle performance of photocatalyst is essential for practical application. As shown in Fig. S8 in Appendix A, SPCN maintains 89.4 % of the initial H2O2 production performance after four cycles of reaction, confirming the good stability of SPCN for continuous photoproduction of H2O2. We further studied the H2O2 production under continuous irradiation for 7 h (Fig. 4(c)), and the SPCN showed a high accumulation of produced H2O2. The generated H2O2 is stable and its concentration has remained under dark conditions for more than 30 hours, which makes it a promising method of in-situ H2O2 generation for practical applications. 

《3.5. Reaction pathway of H2O2 generation》

3.5. Reaction pathway of H2O2 generation

To study the reaction pathway of H2O2 evolution, the catalytic activities of CN, SSCN, and SPCN for oxygen reduction reaction (ORR) were studied using electrochemical voltammetry. Fig. S9 in Appendix A shows the linear sweep voltammetry (LSV) curves of the catalysts on a rotating disk electrode (RDE). The higher current densities were achieved from the SSCN and SPCN compared with that of CN at the same rotating speed, which suggests a higher cathodic ORR performance [4]. Based on the voltammetry results, the corresponding Koutecky–Levich (K–L) plots are obtained (Fig. 5(a)). The electron transfer numbers (n) of CN, SSCN, and SPCN, extracted from the slopes via the K–L analysis, are 1.41, 1.54, and 1.58, respectively. These values indicated that the oxygen reduction reactions on the catalysts have a mixed one- and twoelectron pathways [53]. Among them, SPCN shows the highest n value, indicating the highest selectivity toward two-electron transfer for the H2O2 evolution. The one- and two-electron reactions can be specifically expressed as:

《Fig. 5》

Fig. 5. Reaction pathway of photocatalytic reduction of O2 to H2O2. (a) K–L plots of CN, SSCN, and SPCN were obtained from RDE results (J: the tested current density; ω: rotate speed; rpm: rotation per minute). (b) EPR signals of DMPO–·O2– with SPCN under visible light irradiation. (c) Schematic diagram of efficient visible-light-driven H2O2 conversion by SPCN with N3C vacancies.

One-electron reaction:

Two-electron reaction:

The H2O2 evolution reaction involves oxygen, proton, and photoexcited electron, as described in Reaction (4) [2]. For the oneelectron oxygen reduction, O2 took an electron to produce ·O2–. Then, the ·O2– combines another electron and two protons to convert to H2O2. To identify this photocatalytic O2 reduction process, p-BQ and AgNO3 were used as the chemical quenching agents to scavenge O2– and e–, respectively [12]. Fig. S10 in Appendix A shows that when p-BQ or AgNO3 were added into the photocatalytic reaction, the generation of H2O2 was completely suppressed.

The characteristic EPR signals of 5,5-dimethyl-1-pyrroline N-oxide (DMPO)– ·O2– were collected from SPCN as photocatalyst under visible light irradiation, shown in Fig. 5(b). The results suggest that O2 was reduced to ·O2– via photoexcited electrons and ·O2– is the indispensable intermediate for photocatalytic H2O2 generation [54]. On the other hand, SPCN and SSCN have higher negative zeta potentials compared with CN, suggesting the higher adsorption capacity of H+ which also promotes the generation of H2O2. Although SPCN exhibited a slightly positive zeta potential than SSCN at the same pH, the electron transfer number (n) of SPCN was higher than that of SSCN, indicating that N3C vacancies in SPCN dominated the contribution to the higher activity for H2O2 generation. Moreover, the oxygen adsorption capacity of the catalysts was evaluated by O2-TPD experiments (Fig. S11 in Appendix A). The physisorption of O2 is important in the initial stage of ORR. The TPD results show that SPCN has the highest physical O2 adsorption capacity (< 200 °C) [55], which is because O2 is more likely to be adsorbed on the C sites near the N3C vacancies [31]. 

Based on the above analysis, we propose a reaction pathway for the production of H2O2 by SPCN in Fig. 5(c). At first, O2 molecules were adsorbed onto the C sites around N3c vacancies. With the illumination of visible light, SPCN generates electron-hole pairs. Due to the synergistic effect between the N3c vacancies and cyano group, SPCN offered excellent charge separation ability, which enables the efficient reduction of O2 to ·O2–. Finally, the ·O2– took another electron and reacted with the adsorbed H+ to form H2O2. 

《4. Conclusions》

4. Conclusions

In this study, we design a modified polymeric carbon nitride with rich N3C vacancies by sodium persulfate eutectic polymerization with melamine. The formation mechanism of nitrogen vacancies and other defects during the co-polymerization process is investigated, in which the N3C vacancies are formed by the persulfate-assisted molten polymerization process. The results show that SPCN with a high amount of N3C vacancies exhibits enhanced visible light absorption with accelerated photoexcited charge carrier separation compared with the unmodified CN. The SPCN shows excellent photocatalytic kinetic performance with high kf (3.05 μmol·L–1 ·h–1 ·g–1 ) and low kd (0.0043 min–1 ) and delivers a 4.5-fold enhancement in H2O2 production (323.6 μmol·L–1 ·h–1 ·g–1 ) compared to that of pristine CN. This work reveals insights into the high photocatalytic activity of N3C vacancies in photocatalytic H2O2 generation and offers a new perspective for the development of defect engineering in carbon nitride towards various photocatalytic applications.

《Acknowledgments》

Acknowledgments

This work was supported by the National Key Research and Development Program of China (2019YFC1905400).

《Compliance with ethics guidelines》

Compliance with ethics guidelines

Wei Miao, Yijie Wang, Ying Liu, Hehe Qin, Chengcheng Chu, and Shun Mao declare that they have no conflict of interest or financial conflicts to disclose.

《Appendix A. Supplementary data》

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2021.12.016.