Hexavalent chromium (Cr(Ⅵ)), one of the most common heavy metal environmental contaminants, is widely found in wastewater from the dye production, electroplating, and leather processing industries [1, 2]. Demonstrated to be highly toxic, Cr(Ⅵ) has a strong carcinogenic effect on the human body. Methods for dealing with Cr(Ⅵ)-containing wastewater economically and effectively have attracted widespread attention. The most common solution at present is to convert Cr(Ⅵ) into lower toxicity Cr(Ⅲ), which is then removed by precipitation under neutral or alkaline conditions. Yet repeated pollution with Cr(Ⅵ) is a big challenge caused by chromium-containing mud, which is difficult to handle and make use of [3-5].

Recently, the photocatalytic reduction of Cr(Ⅵ) using solar energy as the driving force in the reaction has become a research focus in the field by virtue of the low cost, high efficiency, and lack of collateral contamination. The semiconductor TiO₂ is the most extensively studied compound for this end, due to its excellent photochemical stability, strong reducibility, superhydrophilicity, and its non-toxic nature [6-8]. Despite this, TiO₂ has the inherent drawback of having low solar energy utilization efficiency due to its wide bandgap (3.2 eV) in photocatalytic reactions, which severely limits the practical application of TiO₂ as a photocatalyst. Accordingly, the development of new visible light driven catalytic materials to remove Cr(Ⅵ) effectively is highly desirable [9].

As a kind of non-metallic organic polymer n-type semiconductor, g-C₃N₄ with its unique electrical, optical, structural and physicochemical properties, and suitable forbidden band width (2.7 eV), conduction band (CB, –1.2 eV) and valence band (VB, 1.5 eV) locations also has the advantages of being a common raw material, with low cost, a simple synthesis reaction, and a visible light response. It has been applied in photocatalysis in many aspects owing to its more flexibility in structural modulation and stronger acid and alkali resistance compared with traditional metal oxide catalysts [10-15]. Theoretically, the reduction potential of Cr(Ⅵ)/Cr(Ⅲ) is 1.33 eV. Thus, purification of highly toxic chromium-containing waste water could be achieved by photo-generated electrons. However, pure g-C₃N₄ itself has the shortages of small specific surface area, narrow visible light response range, high electron-hole pair recombination rate, and low photon quantum efficiency, which greatly restrict the practical applications of g-C₃N₄ [16-20].

Chemical element doping for the introduction of new elements into the structure of g-C₃N₄ by chemical or physical methods has been demonstrated to be the simplest and most effective method to modify the electronic structure of g-C₃N₄. Doping of g-C₃N₄ mainly includes metal elements, non-metal elements and self-doping [21-26]. Non-metallic element doping has proven to be an effective and commonly used method to enhance the catalytic performance of g-C₃N₄ [27-29]. The photocatalytic activity and reaction efficiency of g-C₃N₄ could also be enhanced by doping with halogen using carbon nitride precursors and ammonium halides as additives. These include precursors such as ammonium fluoride (NH₄F) [30], ammonium chloride (NH₄Cl) [31], ammonium bromine [32], and ammonium iodide (NH₄I) [33]. Inspired by these research and findings, we thought that valence electrons with lower electronegativity in iodine atoms were more likely to move and then interact with g-C₃N₄ in possession of π electron system. This extended conjugation system should make the light absorption range of g-C₃N₄ redshift and enhance the electron mobility. However, the large radius of the iodine atom leads to instability in the carbon nitride triazine ring structure.

In this work, we successfully obtained a Br-doped g-C₃N₄ photocatalyst via a simple thermal polymerization method with melamine as the precursor and ammonium bromide as the dopant. X-ray powder diffraction (XRD), Fourier-transform infrared spectroscopy (FTIR), X-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM), transmission electron microscopy (TEM), and density functional theory (DFT) were employed to characterize the resulting structure and investigate the synthesis scheme of the materials. Then, photocatalytic reduction of hexavalent chromium was used to evaluate its photocatalytic performance and stability under visible light. The mechanism of the improved photocatalytic effect by afforded bromide doping is discussed at length and was studied by photocurrent analysis, UV diffuse reflectance spectroscopy (UV-DRS), and photoluminescence (PL). These results show that the CN–BrX samples retain the basic framework of g-C₃N₄, and the Br element are uniformly distributed in g-C₃N₄ structure by replacing the N atoms in carbon-nitrogen heterocycles. The introduction of Br element can obviously reduce the band gap, widen the range of visible light response, accelerate the separation efficiency of photogenerated carriers and increase its specific surface area, so that the modified samples have higher photocatalytic redox activity. In addition, photo-generated e-, ^·OH and H₂O₂ play a certain role during the process of Cr(Ⅵ) photocatalytic reduction.

Melamine (99.0%) and NH₄Br (99.0%) obtained, respectively, from Sinopharm Chemical Reagent Co., Ltd and Shanghai Macklin Biochemical Co., Ltd were of analytical grade and used without further treatment. Deionized water used during the experimental process was generated via an ultra-pure purification system.

The g-C₃N₄ (CN) was prepared by heating 4.0 g of melamine in an uncovered crucible to 520 ℃ at a heating rate of 10 ℃/min and for a total incubation time of 4 h. The obtained product was collected and subsequently ground into a powder.

Br-doped g-C₃N₄ was synthesized by heating a mixture of melamine and ammonium bromide in a closed system. Typically, different weights of NH₄Br powder (15.55, 31.10, 62.20, and 93.30 g) and 4.0 g melamine were dissolved in 40 mL deionized water. After stirring for 24 h, the mixture was dried at 90 ℃ to evaporate the water, followed by a conventional thermal polymerization procedure sharing identical conditions to the synthesis of g-C₃N₄ described above. The product was then washed three times with water to remove residual impurities and is referred to as CN–BrX (X = 5, 10, 20, 30), in which X stands for the mole ratio of NH₄Br to melamine.

The crystalline structure of g-C₃N₄ and the various CN–BrX photocatalysts were analyzed using XRD, recorded on a Bruker D8 diffractometer using Cu K_α radiation (λ = 1.5418 Å). Crystallite sizes of the catalysts were calculated using the Scherrer equation. The Fourier-transform infrared (FTIR) spectra were recorded with a Vertex70 infrared spectrophotometer in the frequency range of 4000–400 cm⁻¹ to get information about chemical bonds or functional groups in molecules. In order to analyze the chemical composition, the valence state of atoms, and the level structure in the surface, XPS (Thermo scientific escalab250Xi) were recorded with a monochromatic Mg K_α X-ray source (hν = 1253.6 eV), calibrated by carbon deposit C (1 s) binding energy (BE) at 284.6 eV. UV–vis DRS was measured on a Solid Cary 5000 spectrophotometer and BaSO₄ was used as a reference material. PL were obtained using an FS5 spectrometer with an excitation wavelength of 355 nm, and the fluorescent lifetime was also obtained. The specific surface areas were deduced from N₂ adsorption-desorption isotherms at –196 ℃ (ASAP 2460, Micromeritics USA) on samples preheated to 200 ℃ for 4 h in a vacuum. S_BET values were calculated using the Brunauer-Emmett-Teller (BET) equation. To characterize the detailed morphology and surface structure of the samples, SEM (Verios 460) with an accelerating voltage of 2.00 kV and TEM with an accelerating voltage of 200 kV were used.

To investigate the crystal structure of Br-doped CN, DFT calculations were carried out using the program package CASTEP. For the total and partial density of state (DOS) k-points were adopted in a 2 × 2 × 1 grid, and the energy cut off chosen was 550 eV.

Photoelectrochemical tests were carried out on a CHI 660 workstation with a three-electrode system, using a Pt plate (1 × 1 cm²) as the counter electrode, a saturated calomel electrode as the reference electrode, and a photocatalyst-coated electrode as the working electrode. Typically, the working electrode was prepared as follows. A total of 3 mg of catalyst powder was dispersed in 0.15 mL of ethanol, followed by ultrasonication for 30 min. Then the above solution was dropped onto IFO (1 × 1.5 cm²) conductive glass and dried at room temperature. Finally, a transient photocurrent was achieved in a 0.1 mol/L Na₂SO₄ aqueous electrolyte to study the surface charge transfer efficiency and carrier density. A 300 W Xe lamp with a filter (λ > 420 nm) served as the simulated solar light source.

The photocatalytic activities of obtained samples were evaluated by Cr(Ⅵ) reduction under visible light illumination. A 300 W Xe lamp with a filter (λ > 420 nm) served as the simulated solar light source, and the irradiation distance is 30 cm.

Prior to irradiation, 50 mg of the sample was placed in a beaker containing 50 mL of 20 mg/L potassium dichromate (K₂Cr₂O₇) solution acidulated by concentrated sulfuric acid and kept stirring in the dark for 1 h to reach an adsorption-desorption equilibrium. During the photoreaction process, 3 mL of suspension was withdrawn and centrifuged at 9000 rpm to remove the photocatalyst at certain time intervals for subsequent analysis. Finally, the concentration of Cr(Ⅵ) was analyzed using a UV−vis spectrophotometer, measuring the characteristic absorption peak at 350 nm.

The active species involved in the photocatalytic reduction of potassium dichromate was investigated by adding excess trapping agents to the solution. These generally included potassium persulfate (K₂S₂O₈), which captures e-, ammonium oxalate ((NH₄)₂C₂O₄), which captures H⁺, tert-butanol (TBA), which captures ·OH, and argon was used to reduce the production of superoxide anions (·O_2-).

Cycling experiments were conducted to explore the stability of prepared samples. Photocatalysts after the reaction were recovered by centrifugation and washed thoroughly with deionized water in order to completely remove residual materials. Subsequently, the recycled products were added into fresh Cr(Ⅵ) aqueous solution (K₂Cr₂O₇: 20 mg/L) for the next photocatalytic process.

The production of hydrogen peroxide under visible light irradiation with a 300 W Xenon lamp (λ > 420 nm) serving as the visible light source was analyzed by studying the resulting PL spectra from a photocatalytic reaction. Typically, 0.1 g of photocatalyst sample was dispersed in 100 mL of deionized water in a container with ice water around. Prior to visible light irradiation, the mixture was stirred in the dark for 10 min to reach an adsorption-desorption equilibrium among the photocatalyst, dissolved oxygen, and water. In the irradiation process, 3 mL of the suspension was taken out every 10 mins, and then filtered to remove the photocatalyst particles followed by a fluorescence measurement. Generally, 100 uL of fluorescence reagent (potassium hydrogen phthalate 8.2 g/L, p-hydroxyphenylacetic acid 270 mg/L, and horseradish peroxidase 100 mg/L) was added to the filtrate along with 1.0 mL of 1.0 mol/L NaOH after 10 min. Finally, the concentration of H₂O₂ was detected using a PL spectrophotometer with a characteristic emission peak of 409 nm and an excitation peak of 315 nm.

The obtained samples were characterized by XRD and FTIR to investigate the crystal structure, chemical bonds, and functional groups. It can be seen in Fig. 1a that all CN–BrX samples except for CN–Br30 exhibited a similar layered structure with two diffraction peaks at 2 θ values of about 12.9° and 27.3°, which are well indexed into the tetragonal phase of CN. The (100) peak near 12.9° corresponded to the in-plane structural packing motif of aromatic segments in the CN crystal, which is a characteristic peak of a triazine species [34]. The lattice spacing calculated to be 0.688 nm was related to the hole-to-hole distance between nitrogen pores of the 3-s-triazine units. Another pronounced peak at 27.4°, which we ascribed to the (002) peak, represented the stacking of conjugated hexatomic rings of CN with an interlayer distance of 0.339 nm [35]. For comparison, the XRD patterns of CN–Br20m were also analyzed as shown in Fig. 1a. Three pronounced diffraction peaks at 2 θ values of 22.0°, 31.7°, and 38.0° were observed for CN–Br20m, which we ascribed to characteristic reflections of NH₄Br. We failed to observe any peaks forNH₄Br in our prepared samples, demonstrating the complete washing away of raw materials. When the doping ratio of NH₄Br increased to 30, the diffraction peaks at 12.9° and 27.3° for CN–Br30 were somewhat offset and weakened along with the occurrence of new peaks, which are not attributable to the characteristic peak of NH₄Br and may have been due to defects arising from incomplete polymerization, indicating that the in-plane structure was destroyed or changed by large amounts of NH₄Br.

FTIR spectra of all the CN–BrX samples except for CN–Br30 (Fig. 1b) displayed the similar molecular skeletal vibration modes, confirming that the basic structure of CN was preserved after modification with Br. The sharp peak at 812 cm^‒1 was assigned to the bending vibration characteristic of a triazine unit, while the peak at 1200–1700 cm^‒1 was related to the stretching vibration characteristic peaks of C=N and C–N bonds in a carbon-nitrogen heterocycle structure. The region at 3000–3500 cm^‒1 corresponded to the stretching vibration of –NH₂ groups in the edge or a damaged position, or the O–H bond from water molecules adsorbed on the surface [36-38]. However, when the doping ratio of NH₄Br increased to 30, the absorption peak of CN-Br30 at 1200–1700 cm^‒1 suffered from some attenuation and had deviations compared to bare CN (Fig. 1c), which was consistent with the assumption that the structure of CN was changed based on the XRD results.

To further study the morphology of the prepared CN and CN–Br20 semiconductors, SEM and TEM (Fig. 2) analyses were carried out. We found that SEM and TEM images (Fig. 2a and 2c) of CN exhibited a typical layered structure with no obvious surface damage or holes. Nevertheless, as shown in Fig. 2b and 2d, the introduction of Br had great influence on the morphology and structure of CN. CN–Br20 showed a curl-like porous structure with regular pores uniformly spread on the slice. Additionally, elemental mapping images of CN–Br20 were generated, which evidence that C, N, and Br atoms were detected and homogeneously distributed over the entire structure (Fig. 2e–2h). These results further confirmed the Br doping of the modified CN.

To rule out the existence of Br in the prepared samples, XPS was performed to test the chemical environment of C and N upon Br-doping. As shown in Fig. 3a, an obvious extra Br peak could be detected in CN–Br20 compared to bare CN, and the high revolution of Br 3 d spectra (Fig. 3b) could be deconvoluted into two peaks at 67.6 and 68.5 eV, which are ascribed to Br 3 d_5/2 and Br 3 d_3/2, respectively, attesting to the successful introduction of Br into the g-C₃N₄ structure. The high resolution of C 1 s spectra (Fig. 3c) of CN was divided into two peaks at 284.6 and 287.9 eV. The narrow peak at 284.6 eV was assigned to the sp²-hybridized carbon (C–C), which was determined to be standard carbon. The broad peak at 287.9 eV was ascribed to sp²-bonded carbon in the heterocycle (N=C–N) of an aromatic carbon nitride, which was related to the major skeleton carbon in the triazine-based heterocycle. In CN–Br20, extra peaks located at 286.0 and 290.3 eV, which corresponded to C–NH₂ and C–Br bonds, could be observed. The former one could be attributed to the restrained polymerization rate of CN, due to the formation of NH₃ from the decomposition of NH₄Br in the calcining process. When excess amounts of NH₄Br were added to co-condensation with melamine (which involved the release of ammonia gas during the process), the produced NH₃ should impede the intrinsic deamination kinetics, and thus result in more –NH₂ bonded with C atoms. The latter may be due to the replacement of N atoms with Br, indicating that Br-doped graphitic carbon has been formed and introduced into the CN matrix. As for N 1 s, the spectra (Fig. 3d) of two samples were separated into three peaks at 398.1, 399.4, and 400.5 eV, corresponding to the sp² hybridized aromatic N bonded with carbon atoms of C–N=C, the tertiary nitrogen in the form of N–(C)₃ groups and the surface uncondensed amino groups bound to hydrogen (C–NH₂), respectively. Combined with the high-resolution spectra of C 1 s, it can be demonstrated that CN–Br20 still retained the basic triazine ring structure of carbon nitride. Additionally, it was clear that the N1s band underwent an obvious binding-energy shift, implying that its chemical environment had been affected. Moreover, the peak area of C–N=C decreased significantly, and that of C–NH₂ was increased compared to pure carbon nitride, which was in good agreement with the C 1 s spectra and further gave evidence that Br had replaced the bonded N atoms in the form of C–N=C [39, 40].

To further analyze the presence of Br, CN-Br20 subjected to deep etching X-ray photoelectron spectroscopy (XPS). Fig. 4a shows the high-resolution energy spectrum of Br at the sample surface, etched at 30, 90, and 120 nm. As the etching depth increased, the peak strength remained consistent, indicating that Br existed in the internal structure of CN. Element analysis tests (EA) of the CN and CN–Br20 samples were performed to verify its structural composition (Fig. 4b). The results showed that the C/N ratio increased from 0.49 to 0.65, which attested to the previously stated reasoning that Br had replaced the bonded N. A possible formation mechanism is illustrated in Fig. 5.

The photocatalytic performance of CN and CN-BrX (X = 5, 10, 20, 30, 20 m) samples was analyzed using a time course of Cr(Ⅵ) (7 mg/L) reduction under visible light (λ > 420 nm) at room temperature. Before light irradiation, the solution was kept in the dark for 1 h to ensure the establishment of an adsorption-desorption equilibrium between the photocatalyst and Cr(Ⅵ). It was clear from Fig. 6a that the adsorption removal ratio of Cr(Ⅵ) was negligible (about 5%) for the different catalyst systems. As presented in Fig. 6b, 25.1%, 35.9%, 55.2%, 61.6%, 35.2%, and 18.7% of Cr(Ⅵ) were removed over CN, CN–Br5, CN–Br10, CN–Br20, CN–Br30, and CN–Br20m, respectively. The concentration change of Cr(Ⅵ) was found to follow zero-order kinetics and the photoreduction rate of Cr(Ⅵ) for CN, CN–Br5, CN–Br10, CN–Br20, and CN–Br30 was calculated to be 0.082, 0.119, 0.119, 0.155, and 0.1 mg L^‒1 h^‒1, respectively, normalized with respect to the specific surface area (Fig. 6c and 6d). CN-BrX samples could dramatically improve the photoreduction activity, and CN–Br20 showed the best photocatalytic performance, as high as 0.155 mg L^‒1 h^‒1, displaying a twice enhancement compared to unmodified CN using the same illuminating conditions. Furthermore, the conclusion can be drawn that doping with Br in an internal structure was the primary cause for the improvement in photoreduction ability of Cr(Ⅵ) on the side of CN–Br20m.

In addition to testing the photocatalytic reduction of Cr(Ⅵ), the production of H₂O₂ in air using the prepared samples (Fig. 6e and 6f) was also tested. We found that all the modified samples exhibit strengthened activity, and CN–Br20 possessed the highest yield of H₂O₂, which was 10 times more active than CN, consistent with the above photoreduction results. Consequently, we concluded that the photoreduction capabilities of CN were greatly improved by Br modification.

Density functional theory (DFT) calculations of pure and Br-doped CN were conducted to investigate the electronic properties of Br-doped gC₃N₄. The structure diagrams of pure CN and doping system are shown in Fig. 7a and 7d, where a bonded N atom was substituted by a foreign Br atom. Calculation results of the energy band structure for CN (Fig. 7b) demonstrated that the top of the VB and the bottom of the CB were both at point G, indicating that CN is a direct band-gap semiconductor, which is consistent with previous reports. We found that a new energy band passing through the Fermi level was generated (Fig. 7e), close to the conduction band, implying that the substitution of a Br atom at an N site was an N-type doping. In addition, the top of the VB and the bottom of CB in the doped system were also seated at point G, demonstrating that modified CN was still a direct band gap semiconductor. Simultaneously, the energy density of Br-doped CN was much higher than that of pristine CN, illustrating that Br atoms may have reduced the symmetry of CN and caused electron delocalization, thus improving the photocatalytic activity. The value of E_g for CN was 0.81 eV and that of Br-doped CN was clearly narrowed to 0.46 eV (Fig. 7c and 7f), in addition to the decrease of both the CB and VB positions, which may have been due to the doped Br atom via influencing the distribution of C and N atoms in the lattice. The calculation results of energy band structure are consistent with the enhancement of photocatalytic activity under UV–vis light irradiation.

Photocatalytic activity is often related to the photo-absorption, photoexcitation and specific surface area of prepared samples. UV–vis DRS was utilized to monitor the alterations of the optical properties of CN and CN–BrX. It is important to note that the light responsive range of CN was expanded to cover the entire visible-light region due to Br modification, and the tail peak extending near the IR region was gradually enhanced with increasing NH₄Br concentration, which is in good accordance with the sample color varying from pale yellow to gray (Fig. 8a and 8b). Consistent with this, CN–BrX should absorb more visible light than bulk CN, thereby generating more charges, which was expected to be beneficial for the enhancement of photocatalytic activity under visible light irradiation. Additionally, the CN–BrX samples may have advantages due to the exposure of active sites as they have its looser structures than pure CN. The intrinsic band edges of CN–Br5, CN–Br10, and CN–Br20 showed a slight redshift compared to CN, with the CN band gap of 2.80 eV decreasing to 2.79 and 2.74 eV. While a blue shift was observed for CN–Br30, the band gap increased to 2.86 eV (Fig. 8c).

To further investigate the effect of the Br-doping on the position shifting of the CB and the VB of the synthesized samples, the VB X-ray photoelectron spectroscopy (VB XPS) was used to analyze the energy band structure. As shown in Fig. 8d, the VB of bare CN and CN–BrX (X = 5, 10, 20, 30) catalysts are calculated to be 1.53, 1.55, 1.68, 1.74, and 1.33 eV. Accordingly, the CB positions were derived to be -1.27, -1.24, -1.06, -1.00, and -1.53 eV based on their band gaps. This result indicated that Br element doping could lead to the down shift of CB and VB levels except in the case of CN–Br30, consistent with the DFT calculations. According to the above experimental results and mechanism calculation, a schematic of the band gap energy level of CN and a series of CN–BrX (X = 5, 10, 20, 30) are depicted in Fig. 8e.

After photoexcitation, the photogenerated electrons may undergo two ways: (1) they could be transferred to the surface of the photocatalyst for subsequent chemical reactions, or (2) they could recombine with photogenerated holes. Photoluminescence (PL) measurements were used to investigate the separation and recombination of photogenerated electrons and holes excited at 355 nm. The strong emission peak of CN at 465 nm was derived from the direct electron and hole recombination of the band transition. The PL intensity of CN–BrX was found to be greatly lowered in comparison to that of CN (Fig. 9a), implying that the photogenerated charge carrier recombination could be effectively suppressed by modification of Br. Moreover, apart from the intrinsic peak of CN at 465 nm, a new band at a shorter wavelength appeared in the PL spectrum of CN–Br30 samples, which may have been due to the recombination of electrons from a lower impurity level. Time-resolved PL spectra were captured to confirm the above results. As shown in Fig. 9b, the fluorescent lifetime of CN–Br20 was obviously shorter than that of pristine CN, indicating that the doped Br could serve as a channel to efficiently transport the charge carriers to active sites and thus promote photochemical reactions.

Photoelectrochemical measurements were carried out in a typical three-electrode cell to analyze the CN and CN–BrX samples, and the results are displayed in Fig. 9c. CN–BrX samples exhibited a higher photocurrent intensity than pure CN, indicating the efficient utilization of solar-light resources after the modification. Distinctly, CN–Br20 exhibited the highest intensity among all the samples, which strongly illustrated that the mobility of the photoexcited charge carriers was greatly improved after modification. Furthermore, the laws governing our photocurrent response results agreed with the Cr(Ⅵ) photoreduction results.

N₂ adsorption-desorption isotherms of the prepared CN and CN–BrX materials are shown in Fig. 9d. All samples exhibited a type Ⅳ sorption isotherm with a pronounced hysteresis loop based on IUPAC classification, suggesting the existence of a porous structure. It was noteworthy that the calculated BET surface area of CN–BrX (except for the sample CN–Br30) was higher than that of CN, which may have been caused by NH₃ gas released during the calcining process. Materials with a larger specific surface area could provide more active sites on the surface and promote the adsorption and transportation capabilities of catalytic reactants, thereby leading to improved photocatalytic activity. However, the BET surface area decreased to 6.932 m²/g when the amount of NH₄Br was increased to 30, which could be due to the fact that the texture of the sample was changed.

Control experiments with different scavengers individually trapping the corresponding active species were employed to explore the possible mechanism of Cr(Ⅵ) reduction. The scavengers used in this study were potassium persulfate (K₂S₂O₈) for e-, ammonium oxalate (AO) for h⁺, and tertiary butanol (TBA) for ·OH. The oxygen dissolved in the reaction solution may have helped to generate numerous oxidizing species, improving the oxidation reaction and suppressing the reduction activity. High purity Ar (> 99.99%) was used to purge the reaction system to get rid of the dissolved oxygen as much as possible before experiments began, and Ar was used throughout the photocatalytic process.

As shown in Fig. 10, the photoreduction efficiency of Cr(Ⅵ) increased in both the CN and CN–Br20 systems with the addition of AO. AO can serve as a hole scavenger, which can promote the separation of photogenerated electron-holes and exhibit a promoting effect on the reduction of Cr(Ⅵ). Meanwhile, the trapping of e^‒ by K₂S₂O₈ conversely restrained the removal rate of Cr(Ⅵ), implying that e^‒ was the main active species in the photocatalytic process, while h⁺ appeared to be negligible in the reaction. Moreover, when TBA was added to capture hydroxyl free radicals, the removal efficiency of Cr(Ⅵ) was significantly improved, and the effect of pure CN was much higher than that of Br-doped CN. This was because the Cr(Ⅲ) in solution could be oxidized to Cr(Ⅵ) by hydroxyl free radicals, which are a kind of reactive oxygen species with strong oxidizing abilities. Thus, the reduction efficiency of Cr(Ⅵ) was greatly improved when it was removed. As shown in Fig. 10d, the hydroxyl free radicals produced in the CN–Br20 system were much higher than those produced in the bare CN system, while the inhibitory effect on the removal rate of Cr(Ⅵ) was exactly the opposite, which may have been due to the larger amounts of H₂O₂ generated by CN–Br20. Since H₂O₂ can show a strong reducibility to transfer Cr(Ⅵ) to Cr(Ⅲ) when reacting with K₂Cr₂O₇ under acid condition. Nevertheless, when the reaction was operated under an Ar atmosphere, only a slight decrease in the Cr(Ⅵ) reduction activity could be observed, being attributed to the decreased electron transport of the O₂/·O₂⁻ mediated reduction in the absence of dissolved oxygen, which may have generated H₂O₂ to reduce Cr(Ⅵ) to Cr(Ⅲ) and ·OH to oxidize Cr(Ⅲ) to Cr(Ⅵ) at the same time. In summary, the active species that played an important role in the Cr(Ⅵ) reduction of both the CN and CN–Br20 systems was photogenerated e-, ·OH and H₂O₂, but ·OH can inhibit the reduction of Cr(Ⅵ).

Based on the above experimental results, a probable mechanism for the photocatalytic reduction of Cr(Ⅵ) over CN–Br20 was proposed. CN–Br20 has a porous structure, which means a higher specific surface area and numerous active sites for the reaction process. The visible light response region of a photocatalyst can be expanded by the doping of Br, thus promoting the solar energy utilization efficiency. The separation efficiency of photogenerated carriers can be greatly improved due to Br and its strong electronegativity. CN–Br20 could be motivated to generate electrons and holes under visible light irradiation. Subsequently, the photogenerated electrons in the conductor band could react with dissolved oxygen to produce ·O₂⁻ and reduce Cr(Ⅵ) to Cr(Ⅲ). Then the reaction between ·O₂⁻ and H⁺ could generate H₂O₂ as well as ·OH. According to the analysis mentioned above, the generation of the photogenerated electron-holes, radicals and reaction process can be described by the following equations:

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Considering the practical application of photocatalysts, it is essential to investigate the repeatability and stability of a photocatalyst. As an example, recycling experiments of CN–Br20 were conducted. In these experiments, the recovered catalyst was washed with deionized water and then dried at 60 ℃. Results in Fig. 11 revealed that the photocatalytic efficiency for the reduction of Cr(Ⅵ) could be maintained at approximately 50% even after five photocatalytic cycles, indicating the CN–Br20 was stable during these experiments. However, the removal efficiency of Cr(Ⅵ) dropped from 61.6% to 49.5% in the second cycle, which may have been due to the occupation of some active sites by Cr(Ⅵ) adsorbed on the material during the first reaction. To further confirm the physical and chemical stability of CN–Br20 after the photocatalytic reaction, the fresh and used photocatalysts were then subjected to characterizations by FT-IR and XRD, respectively. As shown in Fig. 12, no significant change could be observed in the structure of CN–Br20 after five successive cycles, further verifying the high stability and durability of the prepared CN–Br20.

A series of Br doped g-C₃N₄ composites were successfully synthesized via a thermo-induced copolymerization strategy by introducing large amount of NH₄Br into melamine. The experimental characterizations showed that Br was intercalated by replacing the lattice N atoms to form C–Br bonds. The doping of bromine into g-C₃N₄ could modulate the morphology and photocatalytic performance of g-C₃N₄, without destroying the major construction and architecture of the g-C₃N₄ polymer. Compared with bulk g-C₃N₄, the Br-doped g-C₃N₄ exhibited dramatically improved photocatalytic activity in Cr(Ⅵ) reduction under visible light irradiation. The photoreduction of Cr(Ⅵ) on a Br-incorporated g-C₃N₄ photocatalyst (CN–Br20) with high stability reached 0.155 mg L-¹ h-¹, which was almost 2.0 times higher than that of bare g-C₃N₄. We confirmed that the extended optical absorption, narrowed energy gap, high transfer rate of electron-hole pairs, and enlarged specific surface area attributed to the introduced Br atoms were responsible for the remarkably enhanced photoactivity, as demonstrated by both experimental and theoretical results. This work demonstrates a novel method for the synthesis of metal-free photocatalysts with high performance through a simple, low-cost, and environmentally friendly strategy.