1 Introduction

The emission of various non-biodegradable and persistent pollutants in industrial processes and household wastewater pose a serious threat to the ecological environment and human health [1,2,3,4]. Researchers have adopted various methods to solve the above problems. Among them, homogeneous Fenton reaction or heterogeneous sulfate radicals based advanced oxidation processes (AOPs) have been widely applied in eliminating the organic pollutants through the efficient generation of reactive oxygen species (e.g., \(\cdot {\text{OH}}\), \(\cdot {\text{SO}}_{4}^{ - }\) and singlet oxygen (\(^{1} {\text{O}}_{2}\)), etc.) [5, 6]. However, the degradation efficiency is usually limited by the following aspects: (1) The sluggish regeneration (Fe3+ to Fe2+) and narrow pH adaptability limit the practical applications of homogeneous Fenton reaction. Previous studies show that the oxyhydroxides is formed by co-precipitation of Fe3+ and Fe2+ at pH > 3 [7]; (2) Additional hydrogen peroxide (H2O2) or persulfates (peroxymonosulfate (PMS) and peroxydisulfate (PDS)) are required [8, 9]. The H2O2 solution is a typical hazardous chemical, and persulfates have strong oxidation property. They all may pose a risk of secondary environmental pollution; (3) The utilization and activation efficiency of H2O2 and persulfates are usually low [10, 11]. Nowadays, many efforts focus on improving the degradation efficiency by improving Fenton reaction, which can be realized by adding cocatalysts (such as carbon materials, non-metallic elements) or using external techniques (including UV, ultrasound or heat) [12,13,14,15,16]. However, some problems still exist, such as the toxicity, instability, high cost and complex preparation process. Therefore, an efficient, affordable and environmentally friendly treatment approach is urgently desired.

Graphite carbon nitride (g-C3N4), as a metal-free polymeric photocatalyst, has earned increasing attention because of its non-toxicity, suitable electronic band gap (ca. 2.7 eV) and reliable stability [17,18,19]. However, the limited visible light absorption and rapid complexation efficiency of photogenerated electron–hole pairs significantly restrict the applications of g-C3N4 [20]. Constructing heterostructures is an effective strategy, in which g-C3N4 was coupled by other semiconductors with suitable band structure [21]. The heterojunction structures can effectively enhance the separation efficiency of photogenerated charge carriers, broaden the visible-light absorption range and improve the photocatalytic activity. Besides, for practical applications, the separation of catalyst without loss during the recycling process is also an important hotspot to be considered [22].

Zinc ferrite (ZnFe2O4) has attracted widespread attention in photocatalysis due to its broad visible spectrum, narrow band gap (ca. 1.9 eV), magnetic recyclability and photochemical stability [23, 24]. Doped zinc ferrite and zinc ferrite-based heterojunctions have been constructed for photocatalytic applications [25,26,27,28]. Coupling g-C3N4 nanosheets with magnetic ZnFe2O4 nanoparticles can overcome several drawbacks of g-C3N4 monomer, such as low recycle ability, fast photogenic charge carriers recombination, poor redox ability and so on [29, 30]. Researchers have reported that the ZnFe2O4/g-C3N4 heterojunction shows high photocatalytic degradation activity under visible light irradiation, during which H2O2 was added to induce photo-Fenton reaction [31, 32]. The addition of ZnFe2O4 can accelerate the electron–hole separation and migration efficiency. However, a careful study and analysis is still required to answer these questions: (1) whether additive H2O2 is prerequisite during the photocatalytic process; (2) the relationship between H2O2 concentration and catalytic efficiency; (3) the generation path of free radicals; (4) the mechanism of improved photocatalytic performance.

Herein, magnetic ZnFe2O4/g-C3N4 heterojunctions with different ZnFe2O4 contents were constructed as reusable visible-light photocatalysts. The g-C3N4 component promotes the spontaneous production of H2O2, inducing photo-self-Fenton reaction for boosting the photocatalytic performance, that can overcome the shortcomings of traditional Fenton technology during which additional H2O2 is needed. Furthermore, the generation path of free radicals and photo-self-Fenton enhanced photocatalytic mechanism are systematically studied. This work not only provide a method for the design of magnetic g-C3N4 based heterojunctions as reusable visible-light photocatalysts, but also gives a new sight to enhance the photocatalytic efficiency through photo-self-Fenton reaction.

2 Experiments

2.1 Synthesis of ZnFe2O4/g-C3N4 heterojunctions

g-C3N4 nanosheets and ZnFe2O4 nanoparticles were prepared by calcination process and solvothermal method, respectively. The detailed experiments are shown in the supporting information. ZnFe2O4/g-C3N4 heterojunctions were synthesized by a reflux route. In a typical process, 200 mg g-C3N4 was dispersed into 60 mL EG by ultrasonic treatment for 1 h. Then 20 mg ZnFe2O4 nanoparticles were dissolved in the g-C3N4-containing solution, followed by ultrasonic treatment for another 1 h. Finally, the mixture solution was refluxed at 200 °C for 2 h under magnetic stirring. The obtained ZnFe2O4/g-C3N4 heterojunctions were centrifuged at 10,000 rpm, and washed with water and ethanol for several times. The products were dried at 60 °C in air. By changing the ZnFe2O4 content to be 5, 10, 20, 30 and 40 mg, the samples were named as ZFO/CN-X (X = 5, 10, 20, 30 and 40). The materials and characterizations can be seen in the supporting information.

2.2 Photocatalytic activity evaluation

The photocatalytic activity of ZnFe2O4/g-C3N4 heterojunctions was evaluated by photodegradation of methylene blue (MB) under visible light irradiation (300 W Xenon lamp with a cut-off filter of 420 nm) in a photocatalytic reactor. 50 mg of the ZnFe2O4/g-C3N4 heterojunctions were dispersed in 50 mL of MB (20 ppm) water solution. Prior to the light irradiation, the suspension of mixtures were agitated in dark for 30 min. The suspension solutions were exposed under visible light irradiation for 180 min. 3 mL aliquots of the reaction mixture suspensions were taken at certain intervals, and subjected to centrifugation to remove the ZnFe2O4/g-C3N4 heterojunction photocatalysts from the filtrate. The clear solution containing MB were analyzed by UV–vis spectrophotometer at an absorbance of 664 nm.

2.3 Detection experiment of H2O2

The concentration of H2O2 produced during photocatalysis was determined by iodometry. The detection experiment of H2O2 is consistent with the photocatalytic experiment (Sect. 2.2) without MB. 50 mL water solution containing 50 mg ZnFe2O4/g-C3N4 heterojunctions were exposed under visible light irradiation for 180 min. At certain intervals, 3 mL aliquots were taken, and centrifuged to remove the ZnFe2O4/g-C3N4 heterojunction photocatalysts. Then, 1 mL potassium hydrogen phthalate (C8H5KO4) aqueous solution (0.1 M) and 1 mL potassium iodide (KI) aqueous solution (0.4 M) were added to the above 3 mL solution, and kept for 30 min. C8H5KO4 and KI are used to provide hydrogen ion (H+) and iodide ion (\({\text{I}}^{ - }\)), respectively. Under acidic condition, the H2O2 molecules react with \({\text{I}}^{ - }\), generating \({\text{I}}_{{3}}^{ - }\) (\({\text{H}}_{2} {\text{O}}_{2} + 3{\text{I}}^{ - } + 3{\text{H}}^{ + } \to {\text{I}}_{3}^{ - } + 2{\text{H}}_{2} {\text{O}}\)). The amount of \({\text{I}}_{{3}}^{ - }\) was measured by UV–vis spectrometer at its characteristic absorbance peak (350 nm), and then the amount of H2O2 generated in the reaction was obtained.

3 Results and discussion

3.1 Characterization of as-prepared photocatalysts

Figure 1a schematically illustrates the fabrication process of ZnFe2O4/g-C3N4 (ZFO/CN) heterojunctions. g-C3N4 nanosheets and ZnFe2O4 nanoparticles were prepared by calcination process and solvothermal method, respectively. Then ZFO/CN heterojunctions were obtained by a refluxed route. Figure 1b shows the XRD patterns of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN heterojunctions with different amount of ZnFe2O4 nanoparticles. The bare g-C3N4 nanosheets show two characteristic peaks at about 13.01° and 27.45°, belonging to the (100) and (002) crystalline planes, which were related to the in-plane structure of tri-s-triazine units and the interlayer stacking reflection of conjugated aromatic segments, respectively [33, 34]. For the ZnFe2O4 nanoparticles, the diffraction peaks at 18.18°, 29.91°, 35.23°, 42.82°, 53.11°, 56.61° and 62.15° are attributed to the (111), (220), (222), (400), (422), (511) and (440) planes of cubic ZnFe2O4 (JCPDS No. 82-1049). The characteristic peaks of both ZnFe2O4 and g-C3N4 can be observed in the ZFO/CN heterojunctions without other impurity phases. This indicates that ZFO/CN heterojunctions were successfully synthesized.

Fig. 1
figure 1

a Schematic illustration of the preparation of ZnFe2O4/g-C3N4 heterojunctions; b XRD patterns and c FTIR spectra of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZnFe2O4/g-C3N4 heterojunction photocatalysts

Full size image

Figure 1c shows the typical FTIR spectra of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20. For the pure g-C3N4 nanosheets, the broad absorption bands from 3000 to 3400 cm−1 is attributed to the –NH stretching vibration modes and –OH of surface adsorbed hydroxyl species. A series of characteristic peaks located in the region of 1200–1600 cm−1 are due to the typical stretching vibration of C–N and C=N bonds (heterocycle) in the tri-s-triazine units of g-C3N4. The characteristic peak at 805 cm−1 is observed for pristine g-C3N4, which is related to the bending vibration of tri-s-triazine units [35,36,37]. The stretching vibrations of Zn–O and Fe–O modes for ZnFe2O4 are observed at 547 and 423 cm−1 [24, 38]. The band positioned at 1636 cm−1 signifies the O–H bending vibration of surface adsorbed hydroxyl group. No significant differences are observed between g-C3N4 and ZFO/CN-20, suggesting that the addition of ZnFe2O4 didn’t change the chemical structure of g-C3N4.

TEM image (Fig. 2a) and AFM image (Fig. 2b) of g-C3N4 suggest the two-dimensional sheet-like structure with micro size. The thickness is measured to be about 3.3 nm (Inset of Fig. 2b). The interlayer spacing of g-C3N4 is 0.36 nm [39], indicating that the g-C3N4 nanosheets are about nine molecular layers in thickness. TEM image of ZnFe2O4 nanoparticles (Fig. S1) reveals that the particles are well-distributed with an average diameter of approximately 8.4 nm. Figure S2 shows the SEM image of g-C3N4 sheets and ZnFe2O4/g-C3N4 heterojunction photocatalysts with different amounts of ZnFe2O4. Obviously, the ZnFe2O4 nanoparticles are uniformly deposited on the surface of g-C3N4 sheets, which increases by increasing the ZnFe2O4 contents. We consider that the EG solvent with multiple hydroxyl functional groups plays an important role during the formation of heterojunctions. During the reflux process, ZnFe2O4 nanoparticles were surrounded by EG molecules with multiple hydroxyl groups, which can anchor on the surface of g-C3N4 nanosheets. Besides, EG is relatively viscous, which can improve the dispersity of ZnFe2O4 nanoparticles and g-C3N4 nanosheets. The good contact recombination is conducive to constructing heterojunction, which is beneficial to improve the photocatalytic performance of monomer g-C3N4. The typical TEM image (Fig. 2c) and STEM image (Fig. 2e) of ZFO/CN-20 further suggest that tiny ZnFe2O4 nanoparticles are uniformly anchored on the g-C3N4 nanosheets. HRTEM image (Fig. 2d) shows that the lattice spacing distances are 0.243 and 0.253 nm, which are ascribed to the (222) and (311) planes of ZnFe2O4. EDS spectrum of ZFO/CN-20 (Fig. 2f) revealed the existence of C, N, Zn, Fe and O elements in the heterostructure without other element, indicating that the composites are of high purity.

Fig. 2
figure 2

a TEM image and b AFM image of g-C3N4 nanosheets; c TEM image, d HRTEM image, e STEM image and f EDS spectrum of ZFO/CN-20 heterojunction

Full size image

XPS technique was performed to study the surface functional groups and electronic environment of elements in the samples, as shown in Fig. 3. The survey XPS spectra of g-C3N4, ZnFe2O4 and ZFO/CN-20 (Fig. 3a) indicate that C, N, Zn, Fe and O as existed in the ZFO/CN, which is accordant with the XRD data and EDS analysis. The C 1s spectrum of g-C3N4 nanosheets (Down in Fig. 3b) can be fitted into two peaks at about 284.8 and 288.2 eV, which are attributed to C–C/C=C and N–C=N in g-C3N4 lattice, respectively [40]. In comparison, an additional peak at 285.6 eV can be found in ZFO/CN-20, corresponding to the C=O groups. The presence of C=O groups suggest that ZnFe2O4 nanoparticles are chemically bonded on the surface of g-C3N4 nanosheets in the heterojunction. N 1 s spectra of both g-C3N4 and ZFO/CN-20 (Fig. 3c) can be fitted by four peaks, about 398.7 and 399.3 eV for the C=N–C involved in triazine rings and the tri-coordinated N-(C)3, about 400.8 eV for amino groups (C–N–H), and about 404.6 eV for the typical π-excitation in the N–C heterocycles [39]. The Zn 2p spectra of ZnFe2O4 nanoparticles and ZFO/CN-20 (Fig. 3d) show two peaks at about 1021.2 and 1044.4 eV, associated to the Zn 2p3/2 and Zn 2p1/2 of Zn2+, respectively. Figure 3e shows the Fe 2p spectra. The Fe 2p3/2 peak can be fitted into two peaks at about 710.6 and 712.8 eV, which are attributed to the Fe3+ in octahedral and tetrahedral sites, respectively. The peak located at about 724.85 eV is attributed to Fe 2p1/2 of spinel ZnFe2O4. The other peaks at about 718.5 and 732.0 eV are identified as the shake-up satellite structure. Compared to bare ZnFe2O4, the satellite peaks of Fe 2p in ZFO/CN-20 composite are slightly shift to lower binding energy, which suggests that defects may exist in the heterojunction. In the O 1s spectra (Fig. 3f), the broad curve exhibit two fitting peaks located at about 529.6 and 532.0 eV, ascribed to the lattice oxygen (denoted as Fe–O and Zn–O) and surface absorbed oxygen species (such as H2O), respectively [40,41,42]. The peak parameters of g-C3N4, ZnFe2O4 and ZFO/CN-20 are summarized in Table S2. Compared with binding energies of monomers (ZnFe2O4 or g-C3N4), the binding energies of C 1s and N 1s of ZFO/CN-20 slightly shift toward higher values, while the binding energies of Zn 2p and Fe 2p shift toward lower values. The shift to higher binding energy suggests the elements lose electrons, while the shift to lower binding energy suggests the elements get electrons [27, 43]. The opposite shift indicate that the electron transfer from g-C3N4 nanosheets to ZnFe2O4 nanoparticles at the interface of ZFO/CN heterojunction, which is beneficial for the electron migration during photocatalytic process.

Fig. 3
figure 3

a Survey XPS spectra of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20; b C 1s and c N 1s spectra of g-C3N4 and ZFO/CN-20; d Zn 2p, e Fe 2p and f O 1s spectra of ZnFe2O4 and ZFO/CN-20

Full size image

Specific surface area is also an important factor that can affect the photocatalytic activity. Figure S3 shows the N2 adsorption–desorption isotherms and pore size distribution curves of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20. The isotherms of both bare g-C3N4 nanosheets and ZFO/CN-20 show the similar shape with classical type IV, indicating the mesoporous structure [44]. The BET specific surface area, pore diameter and pore volume of the three samples are summarized in Table S1. ZFO/CN-20 reveals the highest specific surface area, about 1.7 times larger than g-C3N4 nanosheets. The enhanced specific surface area of ZFO/CN-20 is probably due to that the anchoring tiny ZnFe2O4 nanoparticles can avert the aggregation of g-C3N4 nanosheets. Large specific surface areas could provide abundant active reaction sites and facilitate the adsorption of pollutant molecules [44, 45].

3.2 Optical and photoelectrochemical properties

The photocatalytic activity is affected by the optical absorption performance. Figure 4a displays the UV–vis absorption spectra of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN heterojunctions. The bare g-C3N4 nanosheets show a strong absorption just at wavelength shorter than around 460 nm. Pure ZnFe2O4 nanoparticles show significant absorption both in the ultraviolet and visible regions. Upon the formation of heterojunctions, The ZFO/CN heterojunctions possess increased optical absorption especially in the visible region, and the absorption edge shows red-shift compared to the pure g-C3N4 nanosheets. Generally, the band gap energy (Eg) of semiconductor can be obtained by the Kubelka–Munk method [46]:

$$\left( {\alpha hv} \right)^{n} = A\left( {hv - E_{{\text{g}}} } \right)$$
(1)

where \(\alpha\), \(h\), \(v\) and \(A\) represent the absorption coefficient, Planck constant, light frequency and constant, respectively. The optical band gaps of g-C3N4 (n = 0.5 for indirect transition) and ZnFe2O4 (n = 2 for direct transition) were determined by a Tauc plot with a linear extrapolation, as shown in Fig. 4b. The Eg of g-C3N4 nanosheets and ZnFe2O4 nanoparticles are calculated to be 2.36 and 1.80 eV, respectively. Based on the Eg values, the valence band energy (EVB) and conduction band energy (ECB) of g-C3N4 nanosheets and ZnFe2O4 nanoparticles can be calculated by the following equations [47]:

$$E_{\text{CB}} = X - E_{\text{e}} - 0.5E_{\text{g}}$$
(2)
$$E_{{{\text{VB}}}} = E_{{{\text{CB}}}} + E_{{\text{g}}}$$
(3)

where X is the absolute electronegativity of semiconductor and Ee is the energy of free electrons on hydrogen scale (about 4.5 eV) [48]. The ECB and EVB of g-C3N4 nanosheets are calculated to be − 0.95 and 1.41 eV, respectively. The ECB and EVB of ZnFe2O4 are − 0.35 and 1.45 eV, respectively.

Fig. 4
figure 4

a UV–vis absorption spectra of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN heterojunctions; b Kubelka–Munk plots of g-C3N4 nanosheets and ZnFe2O4 nanoparticles; c PL spectra of g-C3N4 nanosheets and ZFO/CN; d Transient photocurrent response and e EIS Nyquist spectra of g-C3N4 nanosheets and ZFO/CN-20; f M–H curves of g-C3N4 nanosheets, ZnFe2O4 nanosheets and ZFO/CN-20. Inset shows the photo of magnetic separation tests of ZFO/CN-20

Full size image

The separation and transfer ability of photogenerated electron–hole pairs is a crucial factor for photocatalysis. Photoluminescence (PL) spectra are applied to examine the photo-induced charge carrier excitation, migration, transfer and recombination processes of semiconductor photocatalysts. Figure 4c shows the PL spectra of g-C3N4 nanosheets and ZFO/CN heterojunctions. An obvious emission peak can be seen at approximately 460 nm, which is related to the band gap recombination of photogenerated carriers for g-C3N4 nanosheets. The lower PL intensity represents higher electron–hole separation efficiency, resulting in higher photocatalytic activity [49, 50]. Compared to the pure g-C3N4 nanosheets, ZFO/CN (except ZFO/CN-5) show lower PL intensity, revealing efficient suppression of the recombination between photogenerated electron–hole pairs. The PL spectrum of ZFO/CN-20 possesses the lowest intensity, indicating the excellent carriers separation and photocatalytic activity [51, 52]. The efficient interfacial charge migration and separation can be further investigated through the photoelectrochemical properties. Figure 4d shows the transient photocurrent (PC) response of g-C3N4 nanosheets and ZFO/CN-20. ZFO/CN-20 shows the highest photocurrent density. The larger photocurrent response reveals the higher electrons and holes separation efficiency. Therefore, the photoexcited carriers in ZFO/CN-20 are separated effectively and the photoexcited charge carriers have an improved lifetime during the photocatalytic degradation reaction. Electrochemical impedance spectroscopy (EIS) was performed to provide the evidence of the faster transportation of photoexcited electrons. Generally speaking, a higher mobility and separation of photoexcited electron–hole pairs would result in a smaller arc radius. As seen in Fig. 4e, g-C3N4 nanosheets display an arc radius value than ZFO/CN-20. The PC and EIS results indicate that the combination of ZnFe2O4 nanoparticles could greatly inhibit the recombination of electron–hole pairs in g-C3N4 nanosheets, and efficiently facilitate the separation of photogenerated charges at the interface between ZnFe2O4 nanoparticles and g-C3N4 nanosheets. These results were consistent with the PL analysis, which further validates that the introduction of ZnFe2O4 nanoparticles can enhance the photocatalytic performance.

Figure 4f reveals the M–H curves of g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20 measured at room temperature. The corresponding zoomed M–H curves are displayed in Fig. S4. Obviously, the M–H curve of g-C3N4 nanosheets shows a linear shape, displaying nonmagnetic characteristic. The M–H curves of ZnFe2O4 and ZFO/CN-20 reveal S-like shape, and the coercivity (Hc) values are only about 13 and 15 Oe, respectively. These results suggest that ZnFe2O4 and ZFO/CN-20 show magnetic behavior at room temperature. The magnetization saturation (Ms) value of ZFO/CN-20 is 0.85 emu/g, smaller than that of ZnFe2O4 nanoparticles (14.2 emu/g). This is owing to the non-magnetic g-C3N4. Inset of Fig. 4f shows the photograph of magnetic separation tests of ZFO/CN-20. The transparent solution indicates that the magnetic ZFO/CN heterojunctions can be easily collected by an applied magnet. These findings suggest that ZFO/CN heterojunctions have an excellent magnetic property, which brings convenience for recycling in wastewater management.

3.3 Photocatalytic performance

Methylene Blue (MB) is a typical cationic organic pollutant, which is widely found in various industrial and domestic wastewater. Besides, MB is generally used to evaluate the photocatalytic performance of semiconductor photocatalyst. Photocatalytic properties of the samples were evaluated by degradation of MB solution (20 ppm) under visible-light (λ > 420 nm) irradiation. Figure 5a shows the UV–vis absorption spectra of MB solution after photodegradation by ZFO/CN-20 at different time. The absorption intensity of characteristic peak for MB (around 664 nm) decreased gradually by increasing the irradiation time, indicating that MB was gradually photodegraded. Figure 5b shows the comparison of photocatalytic activities curves for different photocatalysts. The ZFO/CN heterojunctions show greatly enhanced synergetic photocatalytic activities compared to the ZnFe2O4 and g-C3N4 monomers. The photocatalytic activities of ZFO/CN heterojunctions are sensitive to the content of ZnFe2O4 nanoparticles, which increase first and then decrease by increasing the amount of ZnFe2O4. Especially, ZFO/CN-20 shows the highest and fastest photocatalytic activity, in which 96.27% of MB was degraded in 3 h. The enhanced photocatalytic capacity of ZFO/CN heterojunctions is possibly due to that moderate amount of ZnFe2O4 nanoparticles could induce abundant heterostructure interfaces and endow higher separation efficiency of photo-induced electron–hole pairs. Continuously increasing the amount of ZnFe2O4 nanoparticles may result in the aggregation of ZnFe2O4 on the surface of g-C3N4 nanosheets, reducing the amount of active sites and heterostructure interfaces.

Fig. 5
figure 5

Degradation of MB solution by g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN heterojunctions: a UV–vis absorption spectra of MB solution by ZFO/CN-20 at different time; b Time profiles of the photocatalytic degradation of MB by different samples, error bars represent one standard deviation; c Pseudo-first-order kinetics fitted curves of different photocatalysts; d Recycling tests of ZFO/CN-20

Full size image

Pseudo-first-order kinetics model was used to estimate the reaction kinetics of photocatalysts [46]:

$$- \ln (C_{t} /C_{{0}} ) = kt$$
(4)

where C0 is the initial concentration, Ct is the remaining concentration at different irradiation time, t is the irradiation time, and k (min−1) is the apparent first-order rate constant. Figure 5c shows the pseudo-first-order kinetics fitted curves of different photocatalysts. Compared to g-C3N4 and ZnFe2O4 monomers, the ZFO/CN heterojunctions show enhanced degradation rates due to the synergistic effect, which increases first and then decreases by increasing the content of ZnFe2O4 nanoparticles. Among them, ZFO/CN-20 exhibits the highest degradation rate of 0.019 min−1, which is nearly 2.4 and 6.3 times higher than that of g-C3N4 (0.008 min−1) and ZnFe2O4 (0.003 min−1), respectively. These results suggest that ZFO/CN-20 shows significantly enhanced photocatalytic performance for MB degradation, which is attributed to the unique heterojunction structure.

The reusability is another crucial factor for the practical applications of catalysts. Recycling experiments of optimal ZFO/CN-20 is shown in Fig. 5d. The degradation rate of ZFO/CN-20 has no apparent deactivation after four repeated cycles, suggesting good chemical stability and reusability. 88.09% of MB was still degraded after four cycles. The slight decline after cycles is probably due to the photocatalyst loss during the collection. To further confirm the stability, the ZFO/CN-20 heterojunctions after four cycles were characterized by XRD, SEM and XPS techniques, as shown in Fig. S5. After four cycling, XRD patterns (Fig. S5a) of ZFO/CN-20 remains the phase without discrepancy, implying that ZFO/CN-20 shows great structure stability during the photocatalytic process. Besides, the XPS spectra (Fig. S5c-h) confirm that the chemical composition and electric states of all the elements in ZFO/CN-20 are similar to the fresh samples, further suggests the stability. The SEM image (Fig. S5b) suggests that the morphology of the recycled photocatalyst is a little different from the fresh ZFO/CN-20 (Fig. S2d). This is mainly due to that photo-induced electron–hole migration may slightly damage the microstructure of photocatalyst during the multiple photocatalytic recycling, which can reasonably explain the slightly loss of 8.2% for ZFO/CN-20 photocatalyst after four cycles. The characterizations of recycled photocatalyst (Fig. S5) suggest that the recycling can gently change the morphology, but has little influence on the phase, chemical composition and electric states of ZFO/CN heterojunction.

The universal suitability of ZFO/CN heterojunctions was further studied by photodegradation of rhodamine B (RhB, another typical cationic organic pollutant) and methyl orange (MO, a typical anionic organic pollutant), as shown in Fig. S6. Compared to the g-C3N4 monomer, ZFO/CN-20 shows improved photocatalytic efficiency. The photodegradation efficiencies of RhB and MO by ZFO/CN-20 are 92.4 and 73.17%, respectively, much higher than those values by g-C3N4 monomer (81.22% for RhB and 25.76% for MO), as shown in Fig. S6a-d. The kinetic constants of ZFO/CN-20 for degradation of RhB and MO are 0.015 and 0.007 min−1, respectively, which are about 1.7 and 3.5 times higher than the g-C3N4 monomer (0.009 min−1 for RhB, 0.002 min−1 for MO), as displayed in Fig. S6e and f. Table S3 compares the photocatalytic performance between ZnFe2O4/g-C3N4 and other g-C3N4 based materials. Obviously, the ZnFe2O4/g-C3N4 heterostructures show excellent photocatalytic performance at a relatively short time and large pollutant concentration. These results suggest that ZFO/CN heterojunctions show high-efficient photocatalytic degradation performance to different kinds of organic pollutants.

3.4 Mechanism analysis

The light-harvesting capacity, photoelectrochemical properties and PL properties of g-C3N4, ZnFe2O4, and ZnFe2O4/g-C3N4 were studied to deeply understand the photocatalytic mechanism of ZFO/CN. The photo-induced reactive species, including photogenerated holes (\({h}^{{ + }}\)), \({e}^{ - }\), hydroxyl radicals (\(\cdot {\text{OH}}\)) and superoxide radicals (\(\cdot {\text{O}}_{2}^{ - }\)) are involved in the photocatalytic degradation process. It is necessary to study the roles through the trapping experiment of the photo-induced reactive species. Four trapping agents, including EDTA-2Na, AgNO3, BQ and IPA were employed as the scavengers of \({h}^{{ + }}\), \({e}^{ - }\), \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\), respectively [23, 37]. The reactive species capture experiments were carried out by g-C3N4 and ZFO/CN-20. During all the trapping experiments, the amount of trapping agents was kept to be 1 mM. As shown in Fig. 6a and b, the photocatalytic efficiencies of MB by ZFO/CN-20 were decreased from original 96.27% (without trapping agent) to 82.25% (with IPA), 85.61% (with BQ) and 55.76% (with AgNO3), respectively. For g-C3N4 monomer (Fig. S7), the photocatalytic efficiencies of MB were also reduced from original 76.42% (without trapping agent) to 71.29% (with IPA), 45.65% (with BQ) and 49.04% (with AgNO3), respectively. The capture experiments show that the \(\cdot {\text{OH}}\), \(\cdot {\text{O}}_{{2}}^{ - }\) and \({e}^{ - }\) species improve the photocatalytic activity for MB degradation. Among them, photogenerated electrons (\({e}^{ - }\)) play a crucial role in the photocatalytic system of g-C3N4-based materials, while \(\cdot {\text{OH}}\) and \(\cdot {\text{O}}_{{2}}^{ - }\) are involved in the photocatalytic process. The addition of EDTA-2Na can greatly improve the photocatalytic efficiency of g-C3N4 monomer, in which the photocatalytic efficiency is boosted from 76.42 to 98.98% (Fig. S7). According to the mechanism of semiconductor photocatalytic reaction, \({e}^{ - }\) can reduce O2 to \(\cdot {\text{O}}_{{2}}^{ - }\) by a reduction reaction, but there is a serious recombination phenomenon between \({e}^{ - }\) and \({h}^{{ + }}\) [53]. EDTA-2Na, as the scavenger of \({h}^{{ + }}\), plays a role in improving the separation efficiency of photogenerated electron–hole pairs. When EDTA-2Na was added, the main reason for the excessive increase of photocatalytic efficiencies is due to the improved separation efficiency (\({e}^{ - }\) and \({h}^{{ + }}\)), resulting in the increased quantity of \(\cdot {\text{O}}_{{2}}^{ - }\) involved in the degradation of MB. For ZFO/CN-20 (Fig. 6a), the degradation efficiencies of MB reach 96.21% in 90 min and 96.99% in 180 min after introducing EDTA-2Na, while the degradation efficiencies are 74.46% in 90 min and 96.27% in 180 min without EDTA-2Na. The introducing EDTA-2Na has little influence on the final photocatalytic efficiency, but greatly boosts the photocatalytic rate. This suggests that ZnFe2O4 nanoparticles can improve the separation efficiency of photogenerated electron–hole pairs in ZFO/CN heterojunctions and avoid the rapid recombination in g-C3N4 monomer. The capture experiment results of g-C3N4 and ZFO/CN-20 indicated that ZFO/CN heterojunction behaves higher separation efficiency of photogenerated electron–hole pairs and faster photocatalytic rate than g-C3N4 monomer.

Fig. 6
figure 6

a, b Capture experiments of photodegradation of MB solution by ZFO/CN-20 after adding different trapping agents, error bars represent one standard deviation; ESR spectra of c DMPO-\(\cdot{\text{O}}_{2}^{ - }\) and d DMPO-\(\cdot {\text{OH}}\) for g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20 under dark condition and visible light irradiation

Full size image

To determine if \(\cdot {\text{OH}}\) and \(\cdot {\text{O}}_{{2}}^{ - }\) radicals are generated during the photocatalytic process, ESR technique is carried out under dark condition and visible light irradiation for comparison, as shown in Fig. 6c and d. During the detection, 5,5-Dimethyl-l-pyrroline N-oxide (DMPO) was used to trap \(\cdot {\text{OH}}\) in aqueous solution and \(\cdot {\text{O}}_{{2}}^{ - }\) in methanol solution, which can produce the adducts of DMPO-\(\cdot {\text{OH}}\) and DMPO-\(\cdot {\text{O}}_{{2}}^{ - }\), respectively [39]. Under the darkness environment, no ESR signals are detected in the DMPO-containing aqueous solution or methanol solution with g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20. When the solutions were irradiated by visible light, the characteristic peaks of DMPO-\(\cdot {\text{O}}_{{2}}^{ - }\) adduct (Fig. 6c) and DMPO-\(\cdot {\text{OH}}\) adduct (Fig. 6d) can be detected in the solutions with g-C3N4 nanosheets, ZnFe2O4 nanoparticles and ZFO/CN-20. These results demonstrate that active species of \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\) radicals are definitely generated by photogenerated electrons under visible light irradiation.

As described in Fig. 6, the active species trapping experiment and ESR analysis demonstrates that \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\) radicals are produced in the photocatalytic reaction system. However, the calculated CB potentials of g-C3N4 (− 0.95 eV vs. Normal Hydrogen Electrode (NHE)) and ZnFe2O4 (− 0.35 eV vs. NHE) are more negative than the standard redox potentials of O2/\(\cdot {\text{O}}_{{2}}^{ - }\) (− 0.33 eV vs. NHE). Besides, the standard redox potentials of H2O/\(\cdot {\text{OH}}\) (2.72 eV vs. NHE) and \({\text{OH}}^{ - }\)/\(\cdot {\text{OH}}\) (1.99 eV vs. NHE) are more positive than the VB potentials of g-C3N4 (1.41 eV vs. NHE) and ZnFe2O4 (1.45 eV vs. NHE). This implies that the photoexcited holes in the VB of g-C3N4 and ZnFe2O4 cannot directly oxidize the adsorbed H2O molecules or surface hydroxyl group to form \(\cdot {\text{OH}}\) [54, 55]. Therefore, an indirect way for the formation of \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\) radicals is proposed. The photogenerated electrons on the CB of g-C3N4 and ZnFe2O4 react with the resolved O2, forming \(\cdot {\text{O}}_{{2}}^{ - }\) radical species. Then the photogenerated electrons would react with \(\cdot {\text{O}}_{{2}}^{ - }\) and \({\text{H}}^{ + }\) to produce H2O2, which could be further reduced to \(\cdot {\text{OH}}\) radicals. The process is speculated as follows [56, 57]:

$${\text{O}}_{2} + {e}^{ - } \to \, \cdot {\text{O}}_{{2}}^{ - }$$
(5)
$${e}^{ - } + \, \cdot {\text{O}}_{{2}}^{ - } + 2 {\text{H}}^{ + } \to {\text{H}}_{2} {\text{O}}_{2}$$
(6)
$${e}^{ - } + {\text{H}}_{{2}} {\text{O}}_{2} \to \cdot {\text{OH}} + {\text{OH}}^{ - }$$
(7)

As described above, if the active species of ⋅O2− and \(\cdot {\text{OH}}\) radicals are generated according to the Eqs. (5–7), H2O2 should be also spontaneously produced during the photocatalytic process. In order to verify this hypothesis, we further carried out the H2O2 detection experiments under UV–vis illumination. Figure S8a shows the UV–vis absorption spectra of aqueous solution with different contents of H2O2. Obviously, the absorption intensity of characteristic peak for H2O2 (around 350 nm) increases by increasing the H2O2 content. Figure 7a displays the fitted curve of the relationship between H2O2 contents and absorbency in UV–vis region, which is obtained from Fig. S8a. The H2O2 contents and absorbency shows typical linear relationship. Figure S8b-d reveals the UV–vis absorption spectra of aqueous solution containing g-C3N4, ZnFe2O4 and ZFO/CN-20 photocatalysts, which were irradiated at different time. Under ultra-voilet illumination, the characteristic absorption peak for H2O2 can be obviously observed in the aqueous solutions containing g-C3N4, ZnFe2O4 and ZFO/CN-20 photocatalysts. Besides, the absorption intensity of characteristic peak for H2O2 increased gradually as the irradiation time increases. These phenomena demonstrate that a certain amount of H2O2 is spontaneously generated during the photocatalytic process, verifying the about hypothesis. Figure 7b shows the concentration of spontaneously produced H2O2 species by g-C3N4, ZnFe2O4 and ZFO/CN-20 at different illumination time, which is simulated according Figs. 7a and S8. Obviously, the g-C3N4 monomer shows the highest yield of H2O2 (327.8 μM). However, the photocatalytic performance is lower than ZFO/CN-20, as described before. This is mainly due to the high electron–hole recombination rate in g-C3N4 nanosheets. The photogenerated electrons in g-C3N4 nanosheets rapidly recombine with the holes, and cannot adequately react with H2O2, resulting in the accumulation of H2O2. Meanwhile, \(\cdot {\text{OH}}\) radicals were slowly generated by g-C3N4 monomer upon photoexcitation and the increasing concentration of H2O2 leads to the production of a new radical (\(\cdot {\text{OOH}}\)), which shows a considerably lower oxidation potential than that of \(\cdot {\text{OH}}\) radicals [31, 58]. On the contrary, although ZFO/CN-20 shows a lower yield of H2O2 (116.9 μM), but the separation of photogenerated electron–hole pairs is effectively promoted. The spontaneously produced H2O2 species can adequately react with the photogenerated electrons, resulting in abundant active species of \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\) radicals for photocatalysis. The H2O2 yield of ZnFe2O4 monomer is only 7.2 μM, which can be negligible (Fig. 7b). g-C3N4 and ZnFe2O4 can continuously supply H2O2 and iron ions, respectively. Self-Fenton reaction wouldn’t happen in the ZnFe2O4 monomer.

Fig. 7
figure 7

a Fitted curve of the relationship between H2O2 contents and absorbency in UV–vis region; b The concentration of spontaneously produced H2O2 species by g-C3N4, ZnFe2O4 and ZFO/CN-20 at different illumination time

Full size image

Based on the above investigations, the photocatalytic mechanism of ZFO/CN heterojunctions is proposed, as depicted in Fig. 8. Under visible-light irradiation, the g-C3N4 and ZnFe2O4 components generate electron–hole pairs. The photogenerated electrons generated from VB migrate to CB, leaving holes in the VB. According to the results of XPS characterization, we consider that the photogenerated electrons produced in g-C3N4 nanosheets transferred to the CB of ZnFe2O4 nanoparticles, and the holes produced in the ZnFe2O4 migrated from ZnFe2O4 to g-C3N4. This is due to the lower valence band potential and more negative conduction band potential of g-C3N4. The photo-induced charge pairs effectively migrate from the opposite direction, promoting the separation efficiency and enhancing the photocatalytic activity. The responsibilities of the photogenerated electrons (\({e}^{ - }\)) mainly have the following aspects: (1) \({e}^{ - }\) was favored for the generation of active species (\(\cdot {\text{OH}}\) and \(\cdot {\text{O}}_{{2}}^{ - }\)) and H2O2 (Eqs. (5–7)); (2) The self-redox reaction process of Fe ions was realized by using the reducibility of \({e}^{ - }\), accelerating the regeneration from Fe3+ to Fe2+. On one hand, some transferred electrons in the CB combine with adsorbed O2, yielding \(\cdot {\text{O}}_{{2}}^{ - }\) radicals for photocatalytic reaction. On the other hand, some electrons react with O2 and resolved \({\text{H}}^{ + }\) to produce H2O2, which will be further reduced into \(\cdot {\text{OH}}\) radicals. The iron-based nanomaterials show photo-Fenton reaction, due to that the ferric ions in the crystal lattice can absorb the chemical oxidants (H2O2) and create highly reactive \(\cdot {\text{OH}}\) radicals for photocatalytic reaction. However, in photo-self-Fenton system, the degradation efficiency is usually higher than the Fenton reaction and photo-Fenton reaction because of the fast conversion of Fe3+ into Fe2+ ions, high utilization and activation efficiency of self-production H2O2 [30, 42]. A self-redox reaction process of the Fe ions induces the H2O2 to give \(\cdot {\text{OH}}\) radicals. Fe3+ can not only participate in the photo-self-Fenton reaction, but also promote the separation of photogenerated electron–hole pairs via gains electrons. Finally, the pollutants are gradually decomposed by \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\) into degraded products.

Fig. 8
figure 8

Band structure and photocatalytic mechanism of the ZFO/CN heterostructures

Full size image

4 Conclusions

In summary, magnetic ZnFe2O4/g-C3N4 heterojunctions were constructed as visible-light-driven photocatalyst. In the heterojunctions, tiny ZnFe2O4 nanoparticles were anchored on the g-C3N4 nanosheets, which can sufficiently contact with each other, inducing a higher specific surface area and plentiful active sites for photocatalysis. The specific structure and synergistic effect make ZnFe2O4/g-C3N4 heterojunctions behave improved photocatalytic properties in comparison with the onefold g-C3N4 nanosheets or ZnFe2O4 nanoparticles, which can efficiently photodegradate three different pigments of methylene blue, methyl orange and rhodamine B. The loading tiny ZnFe2O4 nanoparticles can greatly enhance the visible-light absorption capacity and accelerate the photo-induced electron–hole separation of g-C3N4 nanosheets, leading to boosted photocatalytic performance. Photocatalytic mechanism studies show that photogenerated electrons play a crucial role in the photocatalytic system. Besides, the ZnFe2O4/g-C3N4 heterojunctions are a photo-self-Fenton system, which can spontaneously produce H2O2 under illumination, forming active species of \(\cdot {\text{O}}_{{2}}^{ - }\) and \(\cdot {\text{OH}}\) radicals. The synergistic effect of \(\cdot {\text{OH}}\) and \(\cdot {\text{O}}_{{2}}^{ - }\) radicals is responsible for degradation organic pollutants. In addition, the ZnFe2O4/g-C3N4 heterojunction photocatalysts show excellent stability after cycling, and can be easily separated by an applied magnet, which is important for practical applications.