All chemicals used were of analytical grade. Ni(NO₃)₂·6H₂O, Fe(NO₃)₃·9H₂O, Zn(NO₃)₂·6H₂O, NaOH, H₂SO₄ (95-98%), HCl (36-38%), C₂H₆O₂ (ethylene glycol), and C₂H₅OH (ethanol) were purchased from Beijing Chemical Works, China. Cu(NO₃)₂·3H₂O and Na₂S₂O₈ were obtained from Tianjin Fuchen Chemical Reagents Factory, China. Co(NO₃)₂·6H₂O and diclofenac sodium were purchased from Shanghai Macklin Biochemical Co. Ltd., China. Mn(NO₃)₂ (50% solution), C₆H₈O₇·H₂O (citric acid), C₂H₃NaO₂·3H₂O (NaAc), and Na₂SO₄ were obtained from Sinopharm Chemical Reagent Co. Ltd., China. AO7 was purchased from Tianjin Guangfu Fine Chemical Research Institute, China.

A series of spinel ferrite MFe₂O₄ (M = Co, Cu, Mn, and Zn) were prepared by sol–gel method (Li et al., 2017) to investigate the effect of M-site metal on PS activation. Then, the three most common methods, sol–gel, solvothermal (Ueda Yamaguchi et al., 2016), and coprecipitation (Jaafarzadeh et al., 2017), were used to study the effect of synthetic method on the properties of ferrite obtained for PS activation, with CuFe₂O₄ as a representative. CuFe₂O₄ synthesized by sol–gel, solvothermal, and coprecipitation methods were denoted as CuFe₂O₄-SG, CuFe₂O₄-ST, and CuFe₂O₄-CP, respectively. The detailed synthetic procedures were described in the Supplementary Information.

X-ray powder diffraction (XRD) of ferrite was carried out on a Rigaku D/max-rc diffractometer using Cu K_α radiation. The morphology of ferrite was observed on a Hitachi S 4700 scanning electron microscope (SEM). N₂ adsorption–desorption analysis was performed on a QuadraSorb Station 4 instrument. X-ray photoelectron spectra (XPS) were measured on a Thermo Fisher Scientific EscaLab 250Xi system with a monochromatic Al K_α source.

The typical experimental steps were as follows: known amounts of ferrite and PS solution were added simultaneously into a 20 mg/L organics solution (AO7 or diclofenac) under magnetic stirring. At known intervals, 3 ml solution was taken using a syringe and filtered through a 0.22 μm filter head. The concentration of the organics in the filtrate was analyzed to evaluate the efficacy of ferrite. The used ferrite was collected using a magnet, washed several times with ethanol and deionized water, and dried for the next run to investigate its reusability. The experiments were done in triplicate.

The concentration of AO7 was determined by TU-1900 UV-visible spectrophotometer (Beijing Persee General Instrument Co., Ltd.) at a maximum absorbance wavelength of 484 nm. The concentration of diclofenac was determined by Ultimate 3000 high performance liquid chromatography (Thermo Fisher Scientific Inc.). The UV detection wavelength was 275 nm, and the mobile phase consisted of acetonitrile and 0.2% acetic acid solution at a volume ratio of 7:3.

The redox property of ferrite was evaluated by oxygen temperature-programmed oxidation (O₂-TPO) and cyclic voltammetry (CV). O₂-TPO was performed from 200 to 500°C at a rate of 10°C/min on a ChemBET Pulsar TPR/TPD instrument (Quantachrome Instruments Inc.) equipped with a thermal conductivity detector (TCD) to measure the change of gas composition. A 5% O₂/He (vol.) gas mixture with a flow rate of 100 ml/min was used in the analysis. CV was conducted on a CHI 760E electrochemical workstation (Shanghai Chenhua instrument Co. Ltd.) with a foamed nickel working electrode, a platinum sheet counter electrode, and a saturated Ag/AgCl reference electrode. Before use, the foamed nickel electrode was dipped in the suspension of ferrite for 10 min to load the catalyst and then air-dried. A mixture of 0.1 mol/L Na₂SO₄ and 0.4 mmol/L PS was used as the electrolyte.

The surface hydroxyl group of ferrite was quantified by the saturated deprotonation method (Ren et al., 2015). In this method, 20 ml of a 0.05 mol/L NaOH solution dispersed with 0.12 g of ferrite was shaken for more than 4 h at 25°C. After separating the solid by filtration, the solution was titrated with a diluted HCl solution.

To investigate the effect of M-site metal on the catalytic performance of spinel ferrite, a series of MFe₂O₄ (M = Co, Cu, Mn, and Zn) were synthesized by sol–gel method. XRD patterns presented in Supplementary Figure 1 confirmed the successful synthesis of the ferrite samples.

Figure 1 clearly shows that the removal efficiencies of organics in different MFe₂O₄ + PS systems were different, but all were higher than that of PS oxidation and MFe₂O₄ adsorption (Supplementary Figure 2). For both AO7 and diclofenac removal (Figures 1A,C), the catalytic performance of ferrite ranked as follows: CuFe₂O₄ > CoFe₂O₄ > MnFe₂O₄ > ZnFe₂O₄. The organics removal processes involved two stages: rapid adsorption-dominated stage (0-1 min) and catalytic degradation stage (> 1 min), which could be fitted by pseudo-first order reaction. The degradation rate constant (Figures 1B,D) was also significantly affected by M-site metal, and the trend was basically consistent with the aforementioned catalytic performance order of MFe₂O₄. In detail, CuFe₂O₄ presented the best and fastest catalytic performance in organics removal. Almost 87.6% AO7 was removed in PS solution coupled with CuFe₂O₄. In comparison, Yue et al. (2016) found that only 53.5% AO7 removal was obtained in Fe₃O₄ + PS system. Moreover, the diclofenac degradation rate constant of CuFe₂O₄ + PS system was about 3.5 times of that of thermally activated PS system at 60°C (Chen et al., 2016). The high removal of AO7 and diclofenac (which have different molecular structures: AO7 is an azo dye and diclofenac is a secondary aromatic amine drug) indicated that CuFe₂O₄ + PS could effectively remove multiple organic pollutants.

These results demonstrated that M-site metal indeed affected the catalytic performance of ferrite. Among the ferrites, CuFe₂O₄ was found to be the best activator of PS for organics removal.

According to above degradation experiments, it has been identified that ferrites with different M-site metals exhibit different catalytic performances. As is known, for PS activation by transition metal, the basic mechanism is chemical reduction of PS through electron transfer (Wacławek et al., 2017). Thus, the reducibility of catalyst is probably the vital factor affecting the effectiveness of PS activation system (Wang and Wang, 2018). Therefore, O₂-TPO and CV were carried out to investigate the redox properties of ferrites with different M-site metals.

Figure 2 shows the O₂-TPO profiles of various MFe₂O₄. The temperature of O₂ consumption surge is an important parameter for evaluating the ease of oxidation-state change of M-site metal ion. As shown, the four ferrites exhibited distinct peaks with temperature increasing from 200 to 500°C, implying the oxidation reaction occurrence of Mⁿ⁺ to M⁽ⁿ⁺¹⁾⁺. The peak temperatures of CuFe₂O₄, CoFe₂O₄, MnFe₂O₄, and ZnFe₂O₄ gradually increased, at 330.8, 353.8, 384.6, and 401.9°C, respectively, which were similar to some other studies. For example, Wang et al. (2011) found that the initial oxidation of CoFe₂O₄ occurred at 350°C resulting from the oxidation of Co. Cihlar et al. (2017) reported that Mn²⁺ in binary oxide was mostly oxidized in the 350–500°C region. The lower oxidation temperature of MFe₂O₄, that was, the easier transition of the oxidation state of M-site metal ion accounted for its better performance in the activation of PS (Su et al., 2017).

To further reveal the redox properties of ferrites, CV curves of different MFe₂O₄ on electrodes were recorded (Figure 3). Except for the couple of redox peaks, the curves of all samples were identical in shape to the control curve, indicating that there was no interference of impossible peak in the solution. CuFe₂O₄ electrode exhibited a well-defined oxidation peak at 0.268 V, which was attributed to the Cu(II)/Cu(III) redox cycle. Likewise, the peaks at 0.350 and 0.401 V were assigned to the oxidation of Co(II) and Mn(II), respectively. ZnFe₂O₄ electrode gave an indistinct oxidation peak at 0.410 V. The lower potential of oxidation peak meant that it was easier for the catalyst to donate electrons, which was favorable for PS activation (Duan et al., 2018). Thus, the reducibility of the ferrites could be ranked as CuFe₂O₄ > CoFe₂O₄ > MnFe₂O₄ > ZnFe₂O₄. Moreover, there was no certain rule about the reduction peak positions of MFe₂O₄ electrodes. However, interestingly, for these ferrite electrodes, the trend of potential separation of redox peaks (ΔE_p) was similar to the trend of oxidation peak potential. CuFe₂O₄ electrode displayed the lowest ΔE_p of 0.179 V. Meanwhile, the ΔE_p values of CoFe₂O₄ and MnFe₂O₄ electrodes were higher at 0.269 and 0.276 V, respectively. In the case of ZnFe₂O₄, it might need much more negative potential than that in the CV curve to make it accept electrons to generate reduction peak. The lower ΔE_p of an electrode indicated the stronger reversibility and more electron transfer of M²⁺/M³⁺ redox reaction (Bard et al., 1980), which were good for the catalytic reaction.

Therefore, from the above results, it can be concluded that M-site metal would affect the catalytic performance of MFe₂O₄ by affecting the reducibility, reversibility of M²⁺/M³⁺ redox couple, and electron transfer on the catalyst surface.

The surface oxygen functional groups of a catalyst might participate in the activation of PS, and thereby affect the catalytic performance of the catalyst (Xiao et al., 2018). Therefore, ferrites with different M-site metals were characterized by FTIR spectra (Figure 4). The broad bands at about 3,416 cm⁻¹ of all samples indicated the obvious presence of hydroxyl group (-OH) (Zhang et al., 2019). The peak at 1,625 cm⁻¹ was due to the deformation vibration of water molecules in the interlayer (Parvas et al., 2014). The two adsorption bands at 1,560 and 1,410 cm⁻¹ (COO- stretching) implied the residual of some citrates in the pores of ferrite. The peak at around 570 cm⁻¹ was associated with the metal–oxygen bond (Zhao et al., 2018).

Among the observed functional groups, -OH should be of particular concern. It was reported that phenol removal was related to the surface hydroxyl concentration of TiO₂ during catalytic ozonation (Song et al., 2010). Hydroxyl group on CuFe₂O₄ surface was found to be critical for radical generation in PMS activation (Guan et al., 2013). Therefore, surface -OH quantities of ferrites with different M-site metals were measured. As shown in Table 1, different M-site metals led to different -OH quantities. Unexpectedly, ferrite with good catalytic performance (e.g., CuFe₂O₄) did not have many surface -OH. This phenomenon was different from the finding of Ren et al. (2015) that MFe₂O₄ containing more surface -OH showed better catalytic performance for PMS. Ren et al. proposed that surface -OH was the main binding site for PMS (surface -OH of ferrite formed hydrogen bond with side -O-OH of PMS), and then PMS accepted electron from metal ion and its O-OH bond was broken to generate SO4•-. The results of this study showed that surface -OH was not crucial for the catalytic performance of MFe₂O₄ in PS system. The reason might be that PS activation involved a different process; SO4•- in MFe₂O₄ + PS system was mainly generated from the fission of middle O-O bond of PS (Wang and Wang, 2018).

According to the above results, CuFe₂O₄ was selected as the representative ferrite to further explore the effect of synthetic method on the physicochemical property and catalytic performance of ferrite for PS activation.

Figure 9 displays AO7 and diclofenac removal in different CuFe₂O₄ + PS systems. As seen, the effectiveness and reusability of the catalyst were significantly influenced by synthetic method.

Figure 9A shows that CuFe₂O₄-SG had the best PS catalytic performance. AO7 removal efficiency in CuFe₂O₄-SG + PS, CuFe₂O₄-ST + PS, and CuFe₂O₄-CP + PS systems were 96.8, 90.1, and 45.3%, respectively. In contrast, only 46.5, 53.4, and 36.3% of AO7 were adsorbed by the corresponding catalyst alone. In the early stage of reaction, the relatively large external surface area of the catalyst enabled CuFe₂O₄-ST to activate PS to remove more AO7. However, as the reaction proceeded, PS and AO7 gradually diffused into the pores of CuFe₂O₄-SG and reacted on the abundant active sites, ultimately leading to a better removal of AO7. Meanwhile, the corrosion and dissolution of the metallic Cu impurity observed on CuFe₂O₄-ST reduced the content of active metal and led to a gradual loss of activation ability of CuFe₂O₄-ST (Li et al., 2019). As for CuFe₂O₄-CP, its adsorption and catalytic capacity were greatly hindered owing to the serious sintering. Moreover, the reusability of CuFe₂O₄ synthesized by different methods was also found to be different (Figure 9B). After three runs, the decrease of AO7 removal in CuFe₂O₄-SG + PS system (19.9%) was lower than that in CuFe₂O₄-ST + PS system (22.0%). AO7 removal efficiency in CuFe₂O₄-SG + PS, CuFe₂O₄-ST + PS and CuFe₂O₄-CP + PS systems became 76.9, 68.1, and 35.8%, respectively. The lower decreasing trend of activity and higher pollutant removal efficiency after repeated uses suggested that CuFe₂O₄-SG had a good reusability. The removal of diclofenac in different CuFe₂O₄ + PS systems (Figures 9C,D) was similar to that of AO7. Once again, CuFe₂O₄-SG showed the best performance, followed by CuFe₂O₄-ST and then CuFe₂O₄-CP.

These results demonstrated that synthetic method would influence the catalytic performance of ferrite from morphological structure, surface area, and element chemical state. Sol–gel method was the ideal one to synthesize ferrite applicable in activated PS process.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.