Preliminarily comparative performance of removing bisphenol-S by ferrate oxidation and ozonation

Bisphenol-S (BS) has recently raised public concerns for its adverse effect on the health safety and ecological security. BS concentrations were detected in many water resources, ranging from 10 ng L−1 to 300 μg L−1, so that various purification techniques have been sought to remove BS. This study investigated the performance of ozonation and ferrate oxidation in the degradation of BS since they are both promising oxidants with high redox potential among water treatment chemicals. It was observed that both ozone and ferrate can achieve over 99% of BS concentration reduction and up to 22.5% of DOC reduction for dosing 0.036 mM of either ferrate or ozone. The vibrio fisheri toxicity exhibited a decline in the treated samples after ozonation or ferrate oxidation. According to the mass spectra analyzed, the degradation pathways were proposed and oxidation products (OPs) were identified. BS degradation by ozonation and ferrate oxidation followed a similar route and four common OPs (OP-249; OP-497-a; OP-497-b, and OP-201) were detected. While ferrate treatment produced one more intermediate (OP-217), ozonation did not, which is attributed to the intensified decomposition of BS by ozonation. The major impact of this study is that ferrate treatment is comparable to the ozonation in removing BS, and further research continuing from this study is necessary to explore the BS removal in various waters with more complex matrixes (e.g., high natural organic matter contents), to investigate BS degradation mechanisms in depth, and to conduct pilot-scale and full-scale trials to establish operational database in running ferrate oxidation and/or ozonation for the treatment of BS in practical world.


INTRODUCTION
Endocrine disrupting chemicals (EDCs) have received major public concerns in terms of their adverse effects on the water quality and public health. For human population exposed to EDCs, many endocrine-related diseases and disorders have been reported, for example, the incidence of genital malformations has increased over time at an unexpectedly high rate 1,2 . Bisphenol-A was once world widely used in industry but after its reorganization as one of EDCs, it has been phased out in the commercial use. Bisphenol-S (BS, or 4,4-sulfonyldiphenol) is then regarded as a safe replacement to bisphenol-A due to its higher thermal stability and proposed high biodegradability [3][4][5] , and has been widely used in the production of epoxy resins and also used as a chemical additive to produce colourfast agents, leather tanning agent, dye dispersants and fiber improvers 6 . However, based on fatal mouse tests, recent researches showed that BS exhibits high endocrine negative effect on human and aquatic species and the studies suggested that BS possesses higher toxicity than bisphenol-A and other bisphenol analogs 3,4,7,8 . In addition, BS concentrations were detected in a range from 10 ng L −1 to 300 μg L −1 in many water resources including raw water and waste water effluent [9][10][11] , which has caused concerns on the BS adverse effects on the aquatic environment and human health.
Several cleaning techniques have been tested for the removal of BS. For example, the activated sludge reactor and constructed wetland have shown high efficiency of removing BS but their performance was uncontrollable due to varying metabolisms when the ambient conditions changed 12,13 . The adsorptive technique can be considered as one of options, with 50-80% removal achievable but the optimized performance was achieved under relatively high restricted conditions 14 . Besides, many homogeneous and heterogeneous chemical oxidation techniques have been tested including the activated persulfate with CuCo 2 SO 4 and permanganate coupled with iodide, which all show the acceptable performance in the removal of BS but with limitations [15][16][17] . And then, the alternative treatment technologies are still sought to tackle the BS pollution issues.
Ferrate (Fe(VI)) is a promising cleaning chemical for water and waste water treatment, which predominantly attributed to its high oxidation capacity and coagulation function after its rapid decomposition and dissociation in water [18][19][20] . The ferrate treatment is highly recommended due to its less adverse effect on human and ecosystems 18,21 , and has been studied to clean the H 2 S odor, remove turbidity, color and heavy metals and to inactivate lethal bacteria 22,23 . Currently, the water treatment by ferrate is receiving more attention at the prospect of removing residuals of pharmaceuticals, EDCs and personal care products in water and waste water effluents 18,19,24 .
On the other hand, ozone is acknowledged as a strong reagent applied in water treatment 25 , which has greater oxidation capacity than permanganate, chlorine dioxide, chlorine gas, oxygen and hypochlorite 26,27 . As an oxidant in water treatment, ozone not only plays main role in disinfection 27 , but can react with unsaturated bonds and aromatic and amino groups and then degrade various organics 27,28 . The mechanism of ozonation is the formation of various reactive oxygen species (ROS) and free radicals when ozone is decomposed, which hold fairly high redox potentials, especially the hydroxyl radical can exert maximum redox potential of 2800 mV 27 . Table 1 shows the basic information of ferrate and ozone.
Due to the problematic issues caused by BS, this study aimed to conduct the comparative study of BS removal with ozonation and ferrate oxidation. Specifically, we investigated at bench scale, (1) the reduction of BS concentrations, (2) mineralization efficiency of BS, (3) resulting toxicity after the treatment, (4) the potential oxidation products (OPs) formation, and (5) the degradation pathways of BS, by ozonation and ferrate oxidation. Table 2 shows that the ozonation and ferrate oxidation both achieved the high BS reduction, together with removal efficiency of other EDCs by ferrate oxidation and ozonation. In this study, for its original concentrations of 10 µg L −1 and 100 µg L −1 , respectively, BS was reduced >99% by either ozonation or ferrate treatment for a chemical dose of 0.036 mM at pH 7. The dose of 0.036 mM of ferrate at pH 7 was chosen for the comparison with the performance of ozonation since doses of ferrate greater than 0.036 mM would not increase the BS reduction further (Supplementary Fig. 1) and neutral pH conditions produced similar BS reduction to that at the acidic condition but high reduction than that at the alkali condition ( Supplementary Fig. 2).

Reduction of BS concentration
Mineralization of BS (DOC removal) DOC removal indicates the extent of mineralization of BS in water samples. Figure 1 shows that comparing to the concentration reduction of BS, the DOC removal of BS was relatively lower; 17.2 ± 1.91% and 22.5 ± 1.93%, respectively, by ozonation and ferrate treatment at a dose of 0.036 mM.
Toxicity assessment Figure 2 shows the comparative toxicity in untreated BS samples and in the BS samples after treatment by ozonation and ferrate. The resulting toxicity of BS solution decreased after the treatment. It can be observed that the toxicity declined more with ferrate treatment compared with that by ozonation, the average resulting toxicity was reduced by 23% after ozonation and by 69% after ferrate treatment.
The toxicity results in this study were consistent to a reported study 29 , where the resulting toxicity was significantly reduced after BS was treated by the peroxymonosulfate with the addition of Fe-SBA-15 under UV photolytic conditions. However, in another study 16 , the toxicity increased when decreasing in BS concentration after BS was treated by peroxydisulfate with sonolysis, suggesting the more toxic oxidation products were generated.  Degradation pathways Figure 3 shows the detected OPs and the suggested degradation pathways of ozonation and ferrate treatment. The initial OP-1 (m/z: 249), OP-2a (m/z: 497) and OP-2b (m/z: 497) were detected in the treated samples, which were also found in the previous studies 15,17,30 . In degradation process, the hydroxylate group of BS was attacked by either ferrate or ozone, and two bisphenol-S groups were combined to form OPs-2a, 2b, which is one of common stages during the degradation, and OPs-2a, 2b are consistent with that produced from other oxidation processes 16,30 . OP-4 (m/z: 201) was detected in the treated samples by both ferrate and ozone. It can be the partial cleavage of one aromatic ring but the second ring and sulphonyl group in BS remained. Figure 3 also shows that OP-4 could be derived directly from OP-1 (m/z: 249) and OP-2 (m/z: 497) in the processes of ozonation and ferrate oxidation, and indirectly from intermediate OP-3 (m/z: 217) in the ferrate oxidation. Due to the relatively limited references linking to OP (4) compared to other OPs, the specific molecular structure and functional groups in OP (4)  ) have high oxidation numbers which act as electron-acceptors too, then ferrate anions can captures the chemical bonding electrons of BS.   Table 2, as well as BS, other selected endocrine disrupting chemicals (EDCs) with ERMs groups were reduced their concentrations greatly (at least >95%) by the reaction with ferrate and ozone, suggesting that the proposed electron-capturing by either ozone molecules/free radicals or ferrate species plays predominate role in the treatment of ERMs' abundant EDCs.

As shown in
However, the mineralization of BS or DOC reduction was lower than the reduction of BS concentration, suggesting the formation of substantial oxidation products (OPs) in the processed samples. The slightly higher BS mineralization (DOC reduction) by ferrate compared with that by ozonation (Fig. 1) could be attributed to the additional ferric coagulation, which was simultaneously occurred when ferrate was decomposed to water. The declined toxicity after treatment (Fig. 2) suggests that the BS degradation derived OPs had lower toxicity comparing to the BS itself. Lower toxicity after ferrate treatment was achieved in comparison with that after ozonation and this is most likely due to the formation of bromated-halogenated derivatives produced in ozonation when bromide was presented in water 32 (Table S2). Instead, ferrate barely reacts with bromide ions thus averting the critical formation of potential carcinogenic bromate and other halogenated by-products such as trihalomethanes (THM) 33 . The formation of OPs in both ferrate treatment and ozonation was detected in this study (Fig. 3) and the BS degradation pathways were proposed as shown in above. Interestingly, OP-217 was not presented in the ozonated samples (Fig. 3) and this can be attributed to the higher re-dox potential of ozone, which resulted in the intensified cleavage rate of heterocyclic group in the stated intermediate OP-217. For supporting this explanation, though, further studies are required. When the toxicity and adverse environmental effect of BS have been recognized and the potential treatment techniques are sought to tackle the BS pollutions, the major impact of this study is that ferrate is comparable to the ozonation in the treatment of BS presenting in water, and the performance of ferrate was even better than ozonation, such as low residual toxicity in the treated samples. Further research continuing from this study is necessary to explore BS removal in various waters with more complex matrixes (e.g., high natural organic matter contents), to investigate BS degradation mechanisms in depth, and to conduct pilot-scale and fullscale trials to establish operational database in running ferrate for the treatment of BS in practical world.

Preparation of ozonation reactor and ferrate treatment and procedures
Mobile ozone generator with 3 micro cells ELEK 4.4 and DC power supply device were purchased from Innovate Geraetetechnik GmbH Rheinbach, Germany. The calibration of ozone concentration was conducted based on the Supplementary Method 1. Peristaltic pump (Heidoph Pump drive 5001) was applied for loading lake water into the circulation system ( Fig. 4 and Supplementary Table 1), the quality of lake water can be viewed in Supplementary Table 2. IKA IKAMAG combing RCT heated magnetic stirrer was used for mixing the water at the required rate. When the ozone concentration reached to 0.03 mM (2 mg L −1 ) (tested by DPD colorimetric method in Supplementary Method 2), ozone generation was stopped and the given volume of BS stock solution (Table 3) was spiked into the ozone saturated lake water to start the reaction. After 15 min ozonation, the reaction was quenched by either adding 0.1 mL of 0.1 M thiosulphate or the injection of nitrogen gas for 4 min. Each test was triplicated. The nitrogen-gas quenched samples were collected for LC-MS analysis, and the thiosulphate quenched samples were collected for DOC analysis.
The strength of the purchased ferrate (K 2 FeO 4 ) was determined using the UV-vis spectrometric method at 505 nm and calculated by Eq. (1), and it was 23%. The ferrate working solutions were freshly prepared before the experiment; 324.64 mg K 2 FeO 4 was mixed with 0.005 M NaOH solution to make a total of 5 mL of ferrate working solution at concentration of 4.0 g Fe L −1 . To achieve the required ferrate doses, 0.4 mL of the stated ferrate working solution was dosed into 800 mL of the test sample to achieve Fe (VI) dose of 0.036 mM.
A 10 µg L −1 and 100 µg L −1 of BS test solutions were prepared by pipetting 1 mL and 10 mL of 20 mg L −1 of each stock solution respectively, into 2 L volumetric flasks and added up to 2 L with lake water; three magnetic stirrers were used for the jar tests. The stated volume of ferrate working solution was dosed into the pre-prepared 10 µg L −1 and 100 µg L −1 test solutions. The mixing programming was set up as shown in Table 4.   34 . All untreated and treated samples were prepared with 20% NaCl to make up to 2% salinity. Vibrio fischeri bacteria were rehydrated and evenly shaken with reactivation reagent, the reactivated bacteria solution were stored in the refrigerator for 30 mins under temperature between 2°C and 8°C for the purpose of stabilization. The experiment was conducted following the protocols of ISO 11348-3:2007 (E). All test samples were prepared three times, and the toxicity test of each sample was triplicated. The results were calculated by adapting the protocol of ISO 11348-3 as shown in Eq. (2) inh% ¼ IC30ÀI30 Where the inh% in Eq. 2 denotes the inhibition of luminescence. f k30 is the correction factor controlling for the natural attenuation of luminescence, which is based on the negative control (2% NaCl solution). I 30 is the actual luminescence intensity after adding the test sample for 30 min, I0 is the luminescence intensity before adding the test sample, I 0 is the initial luminescence reading of the negative control without adding test sample, and I k30 is the luminescence reading of the negative control after adding test sample for 30 min. I c30 is the initial luminescence reading corrected with consideration of effect by time. The tests were validated based on the ISO 11348-3:2007 (E); the correction factors f k30 in this study distribute among range between 0.6 and 1.8, and the inhibition (ihn%) of positive control (5.29 g L −1 K 2 Cr 2 O 7 in 2% NaCl solution) was in the range from 20 to 80%.
Analytical technology DOC in samples was measured by Elementar Vario TOC analyzer. The measurement was conducted when the temperature of the combustion tube reached 850°C, the loading gas flow was 200 mL min −1 . The liquid chromatographic mass spectro-meter (LC-MS) employed was Thermo Scientific Q-Exactive Orbital-trap mass spectrometer with Accela 1200 LC and Acquity HSS T3 column (2.1 × 100 mm, 1.8 μm). The mobile phase was a gradient mixture of solvent A and solvent B: solvent A was made of 0.05% formic acid in MilliQ water, and solvent B was made of 0.05% formic acids in HPLC-grade methanol. For the analysis of BS concentration, the equilibration started with elution of 95% solvent A and 5% solvent B at flow rate of 0.3 mL min −1 for 0.25 min. Then, elution of solvent A started decrease till reached to 20% with solvent B increased to 80% at 6 min, and this elution remained constant shortly till 6.95 min. Finally, the solvent A rapidly increased to 95% with solvent B dropped to 5% till whole elution completed at 10 min. The injection volume of sample was 100 μL, and the overall running time was 10 min. The detected retention time of bisphenol-S was approximately 6.43 min. The monoisotopic mass of the protonated molecular ion of BS was 251.0372 and one identifying fragment of BS in mass spectra was 156.9953 (C 6 H 5 O 3 S). The BS analytical method was validated; the standard recoveries were 85-125% which meets the acceptable range 35 , and the spiking standard concentrations used to establish recovery percentages were 5, 15, 30, 50, and 300 μg L −1 , respectively. In addition, the detection limit for BS was set up as 25 ng L −1 , the limit of quantification was 25 ng L −1 , and the total instrumental run-time of analysis per one batch of samples was 6 h.

DATA AVAILABILITY
All data generated in the study are presented in this communication and/or the Supplementary Information. Additional data related to this paper may be requested from the authors.