Performance and properties of coking nanofiltration concentrate treatment and membrane fouling mitigation by an Fe(ii)/persulfate-coagulation-ultrafiltration process

Coking nanofiltration (NF) concentrates, as typical wastewater with high salinity and refractory organics, have become one of the greatest challenges for “near-zero emission” processes. In our study, an advanced oxidation technology based on ferrous iron/persulfate (Fe(ii)/PS) and polyferric sulfate (PFS) coagulation coupled with ultrafiltration (UF) was used to treat NF concentrates and mitigate membrane fouling. Based on batch experiments, the optimal parameters of Fe(ii)/PS were obtained, during which we discovered that the slow reaction stage of total organic carbon (TOC) removal followed first-order degradation kinetics. Under the optimal reaction conditions, Fe(ii)/PS could efficiently mineralize 69% of organics in coking NF concentrates. In order to eliminate the iron floc generated in the Fe(ii)/PS step, a small amount of PFS (0.05 mM) was added to coagulate the iron floc, which could further improve the effluent quality so that the turbidity, iron content and TOC were significantly reduced by 79.18%, 98% and 21.79% respectively. Gas chromatography coupled with time-of-flight mass spectrometry (GC × GC-TOFMS) and fluorescence excitation-emission matrix spectrometry (EEM) were performed to characterize the removal of phenols, PAHs, quinolines and humic acids in NF concentrates which were responsible for UF membrane fouling. Moreover, scanning electronic microscopy (SEM) and atomic force microscopy (AFM) were conducted to study the surface of the UF membrane after treatment of NF concentrates. The result exhibited that the organic pollutants deposited on the UF membrane surface were reduced by Fe(ii)/PS-PFS pretreatment, and UF membrane flux was thus enhanced. Our results show the potential of the approach of applying Fe(ii)/PS-PFS-UF in NF concentrate treatment.


Introduction
The coal chemical industry is characterized by high water consumption and high organic loads. 1 Large quantities of coking wastewater with a high content of recalcitrant compounds are produced every year. 2 To date, "near-zero emission" processes have been developed to reduce wastewater emissions and control the use of fresh water based on the characteristics of the intercepting screening effect and chargeability of NF. 3 However, the secondarily generated NF concentrates with high salinity and resistant pollutants still need further handling, 4,5 and the problem of membrane fouling also needs to be addressed. 6 The control of membrane fouling basically includes three approaches, i.e. modication of the membrane, 7,8 pretreatment of feed water and cleaning of the fouled membrane. 9 Among these methods, pretreatment of the feed water is the most direct method for industrial use.
At present, coagulation, 10 adsorption, 11 and pre-oxidation 12,13 are commonly applied to NF concentrate pretreatment. However, due to the complex composition of NF concentrates, the treatment efficiency was always quite low by coagulation and adsorption. 14,15 Besides, these treatment methods would mean extra energy consumption, waste of resources and secondary pollution in the long term. 16 As for the pre-oxidation approaches, the treatment period was quite long 12 by the photo-Fenton method. The ozone oxidation method could not be applied to large-scale practice as its strong oxidation corrodes the membrane module, 13 and might even form byproducts. Instead, there are other pretreatment options such as integrating pretreatment processes to enhance membrane performance and reduce fouling. Integrated pretreatment like ultraltration (UF) followed by nanoltration (NF) comprehensively takes the advantages of each treatment and avoids respective defects 17,18 which could effectively remove some bacteria, colloids, macromolecular organics, 19,20 reduce NF membrane fouling and enhance the permeate ux of the system. In addition, the compact process, high degree of automation and stable water quality are regarded by the water treatment industry as the most promising treatment processes. 21,22 However, the UF approach still faces the same problems that the complex NF concentrates are likely to cause rapid pore blockage. Therefore, nding an efficient, inexpensive, energy-saving resource pretreatment method to further mitigate membrane fouling and prolong membrane life will provide more possibilities for the "near-zero emission" process.
Recently, methods for removing refractory organic pollutants by sulfate radical (SO 4 À c)-based advanced oxidation processes (SR-AOPs) have been widely studied. 23,24 SR-AOPs have been increasingly applied to degrade organic pollutants such as atrazine, 25 (3)), a nontoxic AOP system with a low cost, 30 has the potential to be employed into NF concentrates pretreatment. While benecial, suspended Fe(III) oc will not be completely precipitated, resulting in adsorption and pore blocking of membrane. 31,32 Besides, as the membrane pore size is narrowed, various pollutants are deposited on the surface, and thus form a dense fouling layer on the membrane. 33 Under this condition, coagulation is considered as an indispensable step to be adopted prior to UF to effectively remove the suspended Fe(III) oc. 34 In this study, with the effect of low-dose coagulants, the turbidity of effluent, electrical conductivity and the concentration of iron content reduced by 79.18%, 8.83% and more than 98%, respectively. The TOC was further reduced. The results indicated that the preliminary combination of Fe(II)/PS and PFS was more effective than a single treatment due to the removal of suspended Fe(III) oc and could control irreversible fouling of the membrane. 35 Since most of the previous studies used synthetic wastewater in batch experiments, 36,37 more detailed research is needed to investigate the feasibility of the Fe(II)/PS system using authentic wastewater. Besides, it is of great importance to explore the optimal conditions for realizing efficient removal of refractory organics from complex real wastewater. The objectives of this study were: (1) to elucidate the efficiency of Fe(II)/PS-PFS reducing TOC in coking NF concentrates; and (2) to identify the mechanism of organics removal in coking NF concentrates and UF membrane fouling mitigation. Our results are expected to provide a potential pretreatment technology for coking NF concentrates treatment.

Materials
NF concentrates, which were produced in a double-membrane treatment process (the detailed process is shown in Fig. S1 †), were collected from Baogang coking plant, Shanghai, China.
The main water quality (raw water) is shown in Table 1.
Ferrous sulfate was employed as the activator to generate persulfate radicals. PFS was used as a coagulant. The reagents in this experiment were analyzed pure and purchased from the Sinopharm Chemical Reagent Co., Ltd. The remaining water was prepared from an ultrapure water system (Nanopure Millipore System, D11911, Thermo Scientic). A polyethersulfone (PES) membrane (Microdyn-Nadir, Germany) with pore sizes of 0.05 mm was employed in the ltration experiment. The characteristics of the employed membranes are shown in Table S1. †

Experimental setup
Fe(II)/PS and PFS coagulation was employed for feed water pretreatment, and retreated water samples were subsequently used for UF membrane ltration.
The oxidation experiments were carried out batch wise in 250 mL glass bottles using 100 mL coking NF concentrates. During this step, the pH of the solution was adjusted to 3.0. 38,39 The samples were shaken at 200 rpm. PS was added from 0 mM to 4 mM of the stoichiometric dose, 40 Fe(II) was added from 0 mM to 6 mM of the stoichiometric dose, and the reaction was maintained for 2 h. Different reaction times were maintained to examine the effect of the initial PS dose and the initial Fe(II) dose on the Fe(II)/PS process. PFS of 0.5 mM was added into 100 mL of the oxidizing effluent to start the coagulation reaction. The pretreatments were performed with rapid mixing for 1 min at 200 rpm, followed by slow mixing for 20 min at 50 rpm. Then, the samples were taken, ltered and immediately analyzed. All of the experiments were conducted in triplicate. The results shown are the averages of the duplicates. The ltration experiment was performed in ltration cells in the dead-end mode (details are shown in Fig. S2 †). The PES membrane ltration was conducted in a ltration cell (MSC300, Mosu Science Equipment, Shanghai) with an effective volume of 300 mL. A nitrogen gas bottle connected to the ltration cell was used to maintain a constant transmembrane pressure of 100 kPa. UF membranes were placed at the bottom of the ltration cell during the experiment, and each ltration test was run for three cycles.

Analytical method
pH was measured by a pH meter. A Hach 2100N turbidimeter (Loveland, CO) was used for the turbidity analyses. Electrical conductivity was obtained by a conductivity meter. A Hach DR-2800 spectrophotometer was used for the colorimetric analyses. The analysis was carried out using either inductive coupled plasma emission spectrometer (ICP). Chloride concentrations was determined using potentiometric titration methods. The hydrophilicity/hydrophobicity of the membranes was indicated by measuring their water contact angles using a water contact angle analyzer (Dataphysics OCA 15EC, DE). The membrane porosity 3 (%) was calculated by a gravimetric method 41 and mean pore size r m (nm) was determined using the Guerout-Elford-Ferry equation 41,42 based on the data of porosity and pure water ux. Zeta potential was measured by a Zeta Potential Analyzer (Zetasizer Nano ZS 90, UK). The dissolved organic carbon (DOC) concentration (aer 0.45 mm membrane ltration) of the water samples was determined using a TOC analyzer (Multi N/C 2100, JENA). The TOC degradation efficiency was calculated as follows: In eqn (1), C 0 ¼ TOC concentration before oxidation, mg L; C ¼ TOC concentration aer oxidation, mg L À1 . EEM was used to characterize the uorescent components in the water. The EEM spectra were generated using a uorescence spectrophotometer (F4600, Hitachi, Japan) with excitation (Ex) wavelengths of 200-550 nm and emission (Em) wavelengths of 300-650 nm. The EEM spectrum of ultrapure water was subtracted from each EEM sample to remove most of the Raman scatter peaks. 43 Molecular distribution was determined by Gel Permeation Chromatography (GPC), Agilent Technologies, USA (TSK gel: G3000PWXL; column no. S0127; temperature: 40 C; detector: UV254).
Refractory organic pollutants were analyzed using a Pegasus 4D GC Â GC-TOFMS instrument (LECO Corp., St. Joseph, MI, USA) to detect compounds in complex samples. This system utilizes a xed quad-jet modulator consisting of two liquid nitrogen jets and two pulsed hot-air jets to trap and refocus compounds eluted from the rst dimension (1D) column. 44 The modulation period was 2.5 s. The hot pulse duration was 0.60 s.
Helium was used as the carrier gas at a constant ow rate of 1 mL min À1 . The mass spectrometry (MS) transfer line temperature was 330 C. Ionization was performed with an electron ionization (EI) energy of 70 eV and an ion source temperature of 250 C. The collected mass range was 50-550 amu with an acquisition rate of 200 scans per s aer a solvent delay of 450 s. 45 Attenuated total reectance-Fourier transform infrared spectroscopy (ATR-FTIR) (Nicolet 8700, Thermo Electron Corporation, USA) with a resolution of 4 cm À1 in the range of 400-4000 cm À1 was used to obtain information about the functional groups of the membrane surfaces. Membrane samples were held between the diamond plate and the pressure column with the separation layer facing the beam.
The surface and cross-section of the membranes were visualized by SEM imaging. Cross-sectional samples were prepared by mechanically fracturing the membrane in liquid nitrogen. These samples were then dried and sputter coated with a 5 nm thick Au/Pd layer under an argon atmosphere to obtain the necessary conductivity. Microscopic analyses were performed at 12 kV using a Phenom proX SEM-EDS (the USA).
The surface roughness of the membranes was investigated by atomic force microscopy (AFM, CSPM5500) with a noncontact mode. Roughness parameters such as root-meansquare roughness (R q ), mean roughness (R a ) and maximum roughness (R z ) were quantied from the topography images of the 10 mm Â 10 mm area.

Filtration performances
Water ux was carried out at 25 C in a ltration cell (MSC300, Mosu Science Equipment, Shanghai), with an effective test area of 36 cm 2 . An electronic balance is linked to a computer to automatically log weight data every 5 s. The permeation ux (J) was calculated using the following equation: 46 where V is the permeate volume (L), A is the effective membrane area (m 2 ), and Dt is the permeation time (h).  (5) and (6)) which were unfavorable for coagulation. 47 Therefore, excessive amounts of PS could not improve the mineralization of TOC, which was attributed to the fact that SO 4 À c might react with excess PS to produce a persulfate radical

Results and discussion
with an oxidizing power lower than SO 4 À (eqn (9)). 48 This is consistent with the observations made by other researchers. 49 To better evaluate the process parameters for pretreatment performance, TOC-based kinetic analysis has been conducted to understand the mineralization behavior of organic pollutants. 54,55 The following rate equation has been applied to describe the kinetics of TOC mineralization in the process.
In eqn (10), C t ¼ the concentration of TOC at the oxidation time, mg L; K ¼ reaction rate constant; T ¼ reaction time.
As shown in Fig. 1b and d, the ln(C 0 /C)-t plot showed the TOC degradation at different PS concentrations and Fe(II) concentrations, respectively. And the results in Fig. 1a indicated that TOC removal could be divided into two stages: the rapid reaction stage and the slow reaction stage. In the rst 5 min, the main reaction happening in the solution was eqn (3), and the rapidly generated SO 4 À c was used for the mineralization of TOC.
Aer 5 min of reaction, Fe(II) mainly changed to Fe(III) by eqn (4). Under this condition, TOC removal in the solution was relatively stable, and the plot of (ln(C 0 /C)) versus time (inset of Fig. 1b) showed a linear relationship, indicating that the TOC degradation followed the rst-order kinetic model. 3.1.2 The effect of PES for coagulation. The range of important parameters for water quality monitoring is summarized in Table 1. Although the TOC of the oxidizing effluent of Fe(II)/PS was signicantly reduced compared with NF concentrates, high turbidity, electrical conductivity and iron content still existed in the solution. Thus, the coagulation process was conducted. As a result, the turbidity and electrical conductivity of the coagulation effluent substantially reduced by 79.18% and 8.83%, respectively, in comparison to the oxidizing effluent of Fe(II)/PS. The TOC in the effluent further reduced by 21.79%, and the nal iron content was 0.24 AE 0.02 mg L À1 , being reduced by more than 98%. The combination of Fe(II)/PS and PFS was more effective than a single treatment and could compensate for the single Fe(II)/PS process by effective removal of suspended Fe(III) oc with the addition of a low-dose coagulant, which can further improve the quality of product water, mitigate pore blocking and form a cake layer. Among the selected coagulants, 34 Wu et al. 56 concluded that PFS was the most effective in controlling membrane fouling. With the nal Fe(II) concentration reducing to 0.3 mg L À1 , no membrane fouling occurred. Moreover, the formation of ocs with good permeability on the membrane surface was benecial to mitigating the fouling during the coagulation-ultraltration process. 47 So coagulation is considered a destabilization process of colloidal particles. 57 Fig. 2a, the MW distribution of the NF concentrates is shown, which implied that biopolymers (MW > 10 kDa), humic substances and other small MW organic pollutants are presented in raw water. As expected, extensive organic pollutants over a wide MW range have been removed in comparison to the effluent aer coagulation. The results indicated that the Fe(II)/PS-PFS coupling process achieved substantial removal of organic pollutants in the MW range between 1 kDa and 10 kDa. The uorescence EEM spectra of organics in different systems are presented in Fig. 2b. The peaks A (Ex/Em 330/400 nm), B (Ex/Em 280/390 nm) and C (Ex/Em 250/ 410 nm) exhibited high intensities in the EEM spectra which were associated with humic-like substances 59 and soluble microbial byproduct-like substances (SMP-like). A SMP-like substance is dened as a soluble extracellular polymeric substance (EPS), mainly containing small carbonaceous compounds derived from the original substrate 60 and proteinlike substances. 59 The intensity of the uorescence peaks, especially peak A followed the intensive sequence of the raw water > the effluent aer Fe(II)/PS-PFS pretreatment, indicating that the coupling process could remove macromolecular organics such as humic-like, SMP-like and protein-like substances in coking NF concentrates. The result was similar to the previous chapter (Section 3.1).

3.2.2
The removal performance of refractory organics. GC Â GC-TOFMS was applied to detect the removal of refractory organic pollutants under optimal conditions (pH ¼ 3.0, [PS] ¼ 4 mM and [Fe(II)] ¼ 2 mM, PFS ¼ 0.5 mM). The result is shown in Fig. 3. Refractory organic pollutants of 22 and 243 were measured, accounting for 80.69% and 14.96% of the total area, respectively. 44,61 Aer Fe(II)/PS-PFS pretreatment, the aggregate area reduced by 76.27%, and the TOC removal efficiency was the same as that with the Fe(II)/PS-PFS coupling process. At the same time, the peak area of seven principal organic pollutants (shown in Table S2 †) were listed, which revealed several refractory organics, such as phenol, p-nitrophenol, indole, methylquinoline, polycyclic aromatic hydrocarbons (PAHs) and a small amount of butanones, butenals and phthalates. Among them, phenol, PAHs and quinolines were the main constituent substances of TOC, which had a high concentration in the coking wastewater. 62 Fig. 3b showed that the intensities of refractory organics were much weaker in the effluent, indicating that the organic pollutants with complex molecular structures experienced open-loop and chain-breaking oxidation by Fe(II)/ PS-PFS. Thus, we could draw the conclusion that refractory organics such as phenols, PAHs, quinolines and humic acids could be effectively degraded under optimal conditions by Fe(II)/ PS-PFS coupling technology. Table 1 summarizes the relevant parameters of the water quality. Although the TOC of the permeate of raw water was slightly reduced compared with NF concentrates, there was also high TOC, turbidity, electrical and conductivity. Apparently, compared with the oxidizing effluent of Fe(II)/PS and PFS coagulation, the turbidity of the permeate of UF substantially reduced by 86.27%, the electrical conductivity substantially reduced by 8.83% with TOC reducing by 72.55%, and iron content was 0.19 AE 0.02 mg L À1 , reduced by more than 98%. As mentioned previously, the preliminary combination of Fe(II)/PS and PFS was shown to be effective for the UF process. To comprehensively assess the performance of the Fe(II)/PS-PFS-UF system, essential characterizations are necessary to elucidate the changes of the UF membrane in the reaction process.

Fe(II)/PS-PFS-UF process
3.3.1 Investigation of the chemical composition on the membrane surface. FTIR analysis was applied to analyse the nature of the organics on membranes. 63 The spectra of the virgin membrane and two kinds of fouled UF membranes are presented in Fig. 4. The spectrum of the virgin membrane was similar to the specic ATR-FTIR absorbance peaks typical for PES membranes. 64 However, the decrease in the absorbance intensity of typical peaks, and the appearance of additional peaks both suggested that the membranes were severely fouled. Based on Fig. 3, the spectra of the two kinds of fouled UF membranes had similar peaks. Generally, there were four distinct absorption peaks in the picture. The main absorption bands were in the range of 3440-3245 cm À1 (-OH stretching), which indicated that both raw water and effluent contained -OH (alcohol or phenol, most likely phenolic hydroxyl groups). The difference was that the raw water has a weak absorption peak, which was the unsaturated double bond (C]C) and the triple bond (C^C) stretching vibration in the range of 2500-1900 cm À1 . The results indicated that there were double bonds, aromatic and other compounds which contained methyl groups and alkyl groups in the raw water and the effluent from the coking wastewater. This was consistent with the analysis results of phenols, quinolines and other organic compounds detected by GC Â GC. Moreover, compared with the membranes of NF concentrates, the peaks of the membranes aer Fe(II)/PS-PFS pretreatment decreased in both the absorbance intensity and peak quantity, which suggested that the membrane fouling could be addressed.
3.3.2 Investigation of the membrane morphology and structure. In order to investigate the effect of Fe(II)/PS-PFS pretreatment on the microstructure of the membranes, SEM micrographs and AFM analysis of PES membranes with different feed water have been obtained. The SEM images are displayed in Fig. 5. When pollutants were ltrated through the UF membrane aer Fe(II)/PS-PFS pretreatment, there was a cake layer on the membrane surface. Some pores remained on the membrane surface due to the smaller molecular weights of pollutants to the pores diameter of UF membrane, such as dissolved organic and inorganic ions 65 (Fig. 5a). Meanwhile, Fig. 5b exhibits the cross-section aer ltration, the thickness of which was approximately 57.1 mm. Additionally, when pollutants were ltrated through the UF membrane without pretreatment, the membranes were irregular and rough, and cake layers were deposited on the surface of the membranes, including large particles and some foulants 66 (Fig. 5c). Besides, Fig. 5d showed the cross-section aer ltration, the thickness of which was approximately 17.2 mm. However, compared with the membrane of the Fe(II)/PS-PFS-UF process, the parallel UF membrane surface seemed to be denser and more compact, and more organics were deposited, indicating the addition of PFS can reduce the deposition of foulants on the membrane surface. Fig. 6 shows the three-dimensional surface AFM images of the UF membrane surfaces. The roughness parameters of the surfaces of the UF membranes are given in Table 2. Roughness parameters could be obtained with the AFM analysis soware. There exists mean roughness (R a ), root mean square of Z data (R q ) and maximum roughness (R z ). It was observed that the roughness parameters of the NF concentrates membranes were   Table 2). The R a value decreased from 85.7 nm (NF concentrates) to 36.4 nm (effluent of Fe(II)/PS-PFS), which was possibly due to the large particles and more organic foulants from NF concentrates within the membrane surface. Xu et al. 67 reported that a membrane with smoother surface possesses  greater antifouling capability. 68,69 Therefore, the membranes with Fe(II)/PS-PFS pretreatment turned out to have a potential antifouling tendency, which was consistent with the ux recovery results of the membranes depicted in the later part. This may be caused by the degradation and transformation of the macro-molecular and refractory organics aer Fe(II)/PS-PFS pretreatment. In addition, pollutants and ferric colloids aer advanced oxidation in the solution aggregated into large particles and precipitated by occulation of PFS (Table 1). Coagulants were considered to change the particle-size distribution of organic matters in the feed towards larger fractions, with a notable reduction in colloidal matter. 70 Therefore, Fe(II)/ PS-PFS coupling technology can be used to improve membrane fouling.
3.3.3 Investigation of the membrane performances. The ux curve proles during the ltration of coking NF concentrates are shown in Fig. 7. As for the NF concentrates, the ux substantially decreased in the initial ltration phase and subsequently reached a plateau, and the specic ux was nally reduced to 0.185. When Fe(II)/PS-PFS was added to the feed water, the ux decline during the ltration was slightly alleviated, and a specic ux of 0.438 was obtained at the end of ltration. The reasons why Fe(II)/PS-PFS techniques could effectively maintain the ux are as follows. First, this may be caused by the inuence of organics in the NF concentrates. 71 The degradation and transformation of the refractory organics aer Fe(II)/PS-PFS pretreatment can mitigate the ux decline ( Fig. S5 † mechanism of Fe(II)/PS-PFS). Second, this may be due to the reduction of macromolecule organics such as humic-like, SMP-like and protein-like substances by Fe(II)/PS-PFS pretreatment. Yuan 72 studied the pollution of humic acid to the hydrophilic ultraltration membrane. It was found that the adsorption and deposition of humic acid on the surface of the membrane could cause serious membrane fouling. Third, Fe(III) was generated during Fe(II)/PS oxidation, and the coagulation effect of Fe(III) and PFS could be utilized for membrane fouling control and to improve the cake layer structure. 73 Yu et al. 74 demonstrated that Fe(III) caused natural organic matter to aggregate and form large ocs, lowering the thickness and density of the cake layer during ltration.
In order to further explain the treatment process, the forming fouling mechanism of the membrane is shown in Fig. 8. The shape and structure of the obtained surface greatly depended on organic pollutants such as phenol, quinoline and humic acid from NF concentrates deposited on the membrane surface. With the high concentration of organic pollutants in the NF concentrates, substantial organics were attached to the membrane surface aer ltration. These organics tended to   accumulate at the membrane pores and block the pores. Then, a new layer was gradually generated with an increasing amount of organics deposited on the membrane surface, creating a strong resistance to water permeation. Therefore, organics were probably the reason for the antifouling property of the Fe(II)/PS-PFS pretreatment. As an effective pretreatment method, Fe(II)/PS-PFS can be effective in improving the permeability for the mitigation of membrane fouling.

Conclusions
This study proposed a Fe(II)/PS-PFS coupling technology to achieve both the efficient removal of organics in raw water and the mitigation of membrane fouling. The effect of PS concentration and Fe(II) concentration for TOC degradation and its dynamics were analyzed. The optimal experimental parameters of Fe(II)/PS pretreatment technology are conrmed ([PS] ¼ 4 mM and [Fe(II)] ¼ 2 mM, and pH ¼ 3.0). In addition, the quality of the treated water and membrane fouling can be efficiently improved by added 0.5 mM PFS. Organic pollutants deposited on the membrane surface and plugged in the membrane pores were obviously reduced with the transformation of refractory organics and macromolecule organics. It is worth mentioning that membrane ux was also signicantly improved, further conrming that the ux declining could be effectively mitigated by Fe(II)/PS-PFS coupling technology. As an effective pretreatment method, Fe(II)/PS-PFS exhibits good performance in the mitigation of membrane fouling, which may have great potential in NF concentrate treatment.

Conflicts of interest
There are no conicts to declare.