All the chemicals were analytical grade and used as received without further purification. BiOI powders were synthesized according to the previous report (Xiang et al., 2016). Briefly, Bi (NO₃)₃·5H₂O (AR, 99.0%) was added slowly into an ethylene glycol (EG) solution containing KI (AR, 99.0%) with the Bi/I molar ratio of 1. The mixture was stirred for 0.5 h at room temperature in air, and then poured into a 50 ml Teflon-lined stainless autoclave until 80% of the autoclave volume was filled. The autoclave was allowed to be heated at 160°C for 12 h under autogenous pressure, and was then cooled to room temperature in air. The resulting precipitates were collected and washed with ethanol and deionized water and dried at 60°C in air for 12 h.

Bi₅O₇I/g-C₃N₄ heterojunctioned composites were synthesized by a calcination method. BiOI and urea powers were mixed and ground for 0.5 h with agate grinding bowl. The resulting solid was placed in a crucible covered with an alumina lid and was then heated to 500°C in a muffle furnace for 2 h with a heating rate of 15°C/min. After cooling naturally to room temperature, the pale yellow resultant was washed for three times with deionized water and ethyl alcohol, respectively, and dried at 60°C in air for 12 h. The product was labeled as x% BiOI/g-C₃N₄ where x% is the mass percentage of BiOI to urea. Sample 8% BiOI/g-C₃N₄ was calcined and washed again by the same process. The final product was labeled as 8% Bi₅O₇I/g-C₃N₄. For comparison on photocatalytic performance, pure Bi₅O₇I or g-C₃N₄ samples were synthesized under the same condition without addition of urea or BiOI.

X-ray diffraction (XRD) patterns of the samples were carried out on a diffractometer XRD-6100 (Shimadzu, Japan) operating at 40 KV and 30 mA with Cu K _α radiation (λ = 0.154,056 nm). Morphology of the samples was observed using a field-emission scanning electron microscope (FE-SEM) SU-1510 (Hitachi, Japan) and a high-resolution transmission electron microscope (HRTEM) Tecnai G2 F20 (FEI Company, United States). Reflection spectra of the samples were measured on a UV-Vis-NIR spectrophotometer UV-3600 (Shimadzu, Japan) in the range of 260–750 nm and BaSO₄ was employed as a reflectance. Brunauer–Emmett–Teller (BET) specific surface areas of the samples were analyzed by Autosorb-IQ-C specific surface analyzer (Kanta Instruments, United States). Chemical oxygen demand (COD) was evaluated by COD Max II Chemical Oxygen Demand Velocity Tester (Hach Company, United States). A CHI 660E electrochemical workstation and a three-electrode system including a reference electrode (RE), a counter electrode (CE) and a working electrode (WE) were used in photocurrent tests. The counter and reference electrodes were a piece of Pt mesh and a saturated calomel electrode, respectively. The working voltage was a positive bias of 1 V, the light source was a 20 W LED lamp, and the illumination interval was 20 s.

Degradation of RhB and TCH under visible light was used to evaluate the photocatalytic activities of all as-prepared samples. A 300 W xenon lamp equipped with a 420 nm cutoff filter severed as a visible light source was placed 15 cm above the water bath. Water circulation system was used to keep reactant suspension at 20°C avoiding the influence of photo-thermal effect. In a typical experiment, a suspension was prepared by mixing 0.1 g photocatalyst into 100 ml of 8 mgL⁻¹ RhB aqueous solution and 100 ml of 20 mgL⁻¹ TCH aqueous solution, respectively. Before illumination, the suspension was vigorously stirred in the dark for 1 h to reach adsorption-desorption equilibrium. At given irradiation time intervals of 15 min, 3 ml of suspensions were collected and centrifuged at 5,000 rpm for 3 min to remove the catalysts. The absorbance of the solution was analyzed at 554 nm for RhB and 357 nm for TCH by UV-Vis spectrophotometer to monitor photocatalytic degradation of RhB and TCH, respectively.

XRD was performed to investigate the phase structures of as-prepared samples and the results of g-C₃N₄, BiOI, Bi₅O₇I, BiOI/g-C₃N₄ and Bi₅O₇I/g-C₃N₄ are shown in Figure 1. The strongest peak of pure g-C₃N₄ (JCPDS No.87-1,526) at 27.75° is assigned to its (002) crystal plane (Yang et al., 2013; Dong et al., 2015; Ho et al., 2015). The characteristic diffraction peaks at 2θ = 29.64° and 31.65° for pure BiOI, correspond to the (102) and (110) crystal planes of tetragonal BiOI, respectively (JCPDS, No.10-0445). The diffraction peaks at 28.06°, 31.08°, 33.02°, and 33.40° observed on pure Bi₅O₇I are indexed to crystal planes (312), (004), (204) and (020) of orthorhombic Bi₅O₇I, respectively, according to the XRD standard card (JCPDS No.40-0548), which suggested the formation of Bi₅O₇I after calcination for twice at 500°C. The XRD pattern of 8% BiOI/g-C₃N₄ presents characteristic diffraction peaks corresponding to BiOI and g-C₃N₄, indicating formation of composite including BiOI and g-C₃N₄ after the first calcination at 500°C. All the peaks in the XRD pattern of 8% Bi₅O₇I/g-C₃N₄ were indexed to the characteristic diffraction of crystal planes of orthorhombic Bi₅O₇I, suggesting that BiOI has been transformed into Bi₅O₇I after the second calcination at 500°C. No obvious characteristic diffraction peaks were identified for g-C₃N₄ in the sample, which can be ascribed to low crystallinity of g-C₃N₄ and overlap of the main peaks with Bi₅O₇I at 28.06° and g-C₃N₄ at 27.75°. In addition, no impurity peaks are detected in these patterns, revealing that the composites are composed of g-C₃N₄ and Bi₅O₇I with a high purity.

Figure 2 shows typical FE-SEM images of as-prepared photocatalysts. Microspherical structures with diameter of about 2.5 μm, composed of nanosheets, were observed in BiOI sample, as shown in Figure 2A. Figure 2B shows the aggregated particles of Bi₅O₇I photocatalyst which was formed by calcination of BiOI in air. Figures 2D,E show the aggregation of micrometer sized irregular particles of BiOI/g-C₃N₄ and Bi₅O₇I/g-C₃N₄ composites, respectively. In the specific surface analysis testing, the specific surface areas of Bi₅O₇I, g-C₃N₄ and 8% Bi₅O₇I/g-C₃N₄ were determined to be 17.6, 14.2 and 16.5 m²g⁻¹, respectively which indicated that specific surface area had a trivial effect on photocatalytic performance.

Sample 8% Bi₅O₇I/g-C₃N₄ was further characterized by TEM, as shown in Figure 3. Nanoparticles with diameter of several nanometers were distributed in the composite, suggesting that the as-synthesized Bi₅O₇I/g-C₃N₄ composite was possibly mixtures of two different components. In the HRTEM image (Figure 3B), the lattice fringes of 0.338, 0.249 and 0.316 nm coincide with the interplanar spacing of plane (002) of g-C₃N₄ and planes (114) and (312) of Bi₅O₇I, respectively, which is consistent with the results of XRD analysis. Furthermore, Bi₅O₇I nanospheres were distributed on the surface of g-C₃N₄, which implied a superior absorption of visible light. Based on the results of XRD, SEM, and TEM, the composites were successfully synthesized and formation of the interfaces between Bi₅O₇I and g-C₃N₄ confirmed a p-n junction structure in the composite.

Figure 4 exhibits the UV-Vis diffuse absorbance spectra of as-prepared photocatalysts. Apparently, BiOI absorbs light most effectively with an absorption edge at about 700 nm. Comparatively, the pure g-C₃N₄ sample exhibits a shorter absorption edge, which is located at about 450 nm. The light absorption threshold of samples extended to longer wavelength with the introduction of BiOI and Bi₅O₇I, which indicated that the visible light adsorption of as-prepared composites was contributed by the formation of BiOI and Bi₅O₇I. The band gap energies of as-prepared composites can be calculated using classical Tauc’s approach a h v = A ( hv − E g ) n / 2 (1) where α, h, υ, A and E _g, are absorption coefficient, Planck’s constant, light frequency, proportionality constant and band gap energy, respectively (Huang et al., 2014). Among them, values of n are 4 for indirect transition and 1 for direct transition, respectively (Xu et al., 2015). Since both Bi₅O₇I and g-C₃N₄ are indirect transition semiconductors (Yang et al., 2014; Liu et al., 2015), values of n are 4. Hence, values of E _g for Bi₅O₇I and g-C₃N₄ can be estimated to be 2.59 and 2.71 eV respectively. These E _g values are in accordance with those in previous reports (Chen et al., 2015; Bai et al., 2016).

The photocatalytic activities of as-synthesized samples were first characterized by degradation of RhB under visible light illumination (≥420 nm), as shown in Figure 5A. C was the concentration of RhB during the reaction and C ₀ was the initial concentration of RhB. As for pure g-C₃N₄ and BiOI, only 21 and 39% of RhB were photodegraded after visible-light irradiation for 120 min, respectively. Meanwhile, Bi₅O₇I achieved superior photocatalytic activity over BiOI. After formation of BiOI/g-C₃N₄ and Bi₅O₇I/g-C₃N₄, the composites show a significant increase in RhB degradation performance compared with the pure BiOI, Bi₅O₇I and g-C₃N₄. The degradation of RhB was fitted for first-order kinetics with a Langmuir-Hinshelwood model (Sun et al., 2014): ln (C ₀ /C) = kt-ln (C ₀ /C ₁), where k is the reaction rate constant, which were presented in Figure 5B. The reaction rate constants of pure g-C₃N₄, pure BiOI and pure Bi₅O₇I were estimated to be 0.00370 min⁻¹, 0.00820 min⁻¹, 0.01456 min⁻¹, respectively. With formation of composites, the k values increased to 0.02095 min⁻¹, 0.02843 min⁻¹ 0.02028 min⁻¹ and 0.03497 min⁻¹ for 6% BiOI/g-C₃N₄, 8% BiOI/g-C₃N₄, 10% BiOI/g-C₃N₄ and 8% Bi₅O₇I/g-C₃N₄, respectively. The optimal degradation rate constant achieved by 8% Bi₅O₇I/g-C₃N₄ was approximately 9.45 times higher than that of pure g-C₃N₄, 4.26 times of pure BiOI and 2.40 times of pure Bi₅O₇I, respectively.

The photostability of 8% Bi₅O₇I/g-C₃N₄ was characterized through a five-cycle photocatalytic experiment, as shown in Figure 5C. No apparent deactivation was detected in five successive degradation reactions under visible light irradiation, suggesting the possibility for practical application in treatment of water pollution.

To investigate active species involved in the photocatalytic reaction, BQ, IPA and EDTA were used as scavengers to trap O₂ ⁻, ·OH and h⁺, respectively. The effects of different scavengers on the removal efficiency of RhB were shown in Figure 5D. The results indicated that the removal percentage of RhB was significantly decreased to approximate 57, 38, 31% after addition of EDTA, BQ and IPA, respectively. These results revealed that all the reactive species of h⁺, ·OH, and O₂ ⁻ participated in the photocatalytic degradation processes and all of them played important roles in the photocatalytic degradation reaction of RhB.

Figure 6 shows the photocatalytic activity of as-prepared samples in degradation of TCH under visible light illumination (≥420 nm). Similarly, 8% Bi₅O₇I/g-C₃N₄ achieved the superior photocatalytic activity. Photocatalytic activity descended in the queue as 8% Bi₅O₇I/g-C₃N₄> 8% BiOI/g-C₃N₄ > 10% BiOI/g-C₃N₄ > 6% BiOI/g-C₃N₄ > Bi₅O₇I > g-C₃N₄ > BiOI. Stability experiments indicated that sample 8% Bi₅O₇I/g-C₃N₄ showed a stable efficient photocatalytic performance in photodegradation of TCH. Scavengers experiments illustrated that h⁺ played an important role in the photocatalytic degradation reaction of TCH.

To further confirm complete photodegradation of RhB and TCH, COD of the solutions were measured before and after degradation tests. The COD values of the solutions before degradation of RhB and TCH were evaluated to be 9.5 mg/L and 32.0 mg/L respectively. Those after degradation of RhB and TCH were 1.2 mg/L and 2.8 mg/L, respectively, both of which were significantly lower than the WHO guideline value of 10 mg/L. It suggested very rare organic species in the solutions.

Testing on transient photocurrent response were performed to study the transfer and separation efficiency of photogenerated charge carriers under visible light irradiation. Stronger photocurrent intensity usually indicates higher separation efficiency of holes and electrons. Figure 7 illustrates the transient photocurrent response curves of pure BiOI, Bi₅O₇I, g-C₃N₄ and Bi₅O₇I/g-C₃N₄ composite. Once the light source was turned on and off, the sample electrode exhibited a rapid photocurrent change. Furthermore, the Bi₅O₇I/g-C₃N₄ composite showed a strongest photocurrent response. It suggests that more efficient separation efficiency of electrons and holes occurred in Bi₅O₇I/g-C₃N₄ heterojunction, thus an enhanced photocatalytic activity was achieved by Bi₅O₇I/g-C₃N₄ composite photocatalyst.

Figure 8 shows the proposed photocatalytic mechanism of Bi₅O₇I/g-C₃N₄. The valence band edge potential and the conduction band edge potential of Bi₅O₇I and g-C₃N₄ were calculated using electronegativity with the following empirical equations (Li et al., 2015b) E VB   =   X   +   0 . 5 E g   −   E e (2) E CB   =   E VB   −   E g (3) X   =   [ x ( A ) a x ( B ) b x ( C ) c ] 1 / ( a + b + c ) (4) where E _VB is the valence band edge potential, E _CB is the conduction band edge potential, x is the electronegativity of chemical elements (a, b, and c are the atomic number of compounds), E _e is the energy of free electrons on the hydrogen scale (about 4.5 eV), E _g is the band-gap energy of semiconductor and X is the electronegativity of the semiconductor. The X value for Bi₅O₇I is about 6.22 eV. For g-C₃N₄, the X value is 4.73 eV (An et al., 2018). Therefore, the valence of E _VB and E _CB of Bi₅O₇I and g-C₃N₄ were calculated to be 3.02, 0.43 eV and 1.59, −1.13 eV, respectively. The Fermi energy (E _f) level of Bi₅O₇I as a p-type semiconductor is closed to the valance band, while that of g-C₃N₄ as a n-type semiconductor is closed to the valence band. After contact was made between Bi₅O₇I and g-C₃N₄, electrons and holes migrated, i.e., electrons in g-C₃N₄ diffused to Bi₅O₇I and holes in Bi₅O₇I diffused to g-C₃N₄, which led to the formation of an electric field (Wang et al., 2014). The diffusion process continued until the Fermi levels of Bi₅O₇I and g-C₃N₄ reached an equilibrium state. Simultaneously, the valence and conduction bands of heterojunctioned Bi₅O₇I and g-C₃N₄ were bent to render a balanced state.

After visible light illumination, the induced holes from the valance band of g-C₃N₄ transferred to the valance band of Bi₅O₇I and the photoexcited electrons transferred from the conduction band of Bi₅O₇I to the conduction band of g-C₃N₄ driven by built-in electric field. It allows more effective separation and longer lifetime of photoinduced electron-hole pairs. Electrons gathered on the surface of g-C₃N₄ reacted with O₂ in the water to form •O₂ ⁻ and holes gathered on the surface of Bi₅O₇I reacted with OH⁻ in the water to form •OH (Lin et al., 2016), leading to the degradation of RhB. Simultaneously, holes on the valance band of Bi₅O₇I possesses a higher oxidation capability on RhB and TCH. The formation of Bi₅O₇I/g-C₃N₄ heterojunction not only expands the photoresponse region, but also suppresses the recombination of photogenerated electron-hole pairs and improves the photocatalytic activity.