Removal of Contaminants of Emerging Concern from Wastewater Using an Integrated Column System Containing Zero Valent Iron Nanoparticles

: Non-steroidal anti-inﬂammatory drugs (NSAIDs) and endocrine disruptors (EDCs) are among the most important categories of contaminants of emerging concern (CECs), and many advanced technologies have been developed for their elimination from water and wastewater, including nano-zero valent iron (nZVI). This study investigates the performance of nZVI synthesized from green tea extracts and incorporated into a cationic resin (R-nFe) in the removal of four selected NSAIDs, namely ibuprofen (IBU), naproxen (NPX), diclofenac (DCF), and ketoprofen (KFN), and an EDC, namely bisphenol A (BPA). Column experiments were conducted to evaluate the effect of various operating parameters, including initial CECs concentration, contact time, pH, addition, and dose of sodium persulfate (PS). To the authors’ knowledge, this is the ﬁrst time that environmentally friendly produced nZVI has been combined with PS in column experiments for the removal of CECs from wastewater. With a contact time of 2.2 min, PS = 1 mM, and inﬂuent pH = 3.5, 27–72% of IBU, 70–99% of NPX, 70–95% of DCF, 28–50% of KFN, and 61–91% of BPA were removed during a 12-day operation of the system, while the initial concentration of each substance was 5 µ g/L. Therefore, it is anticipated that the proposed system could be a promising post-treatment technology for the removal of CECs from wastewater.


Introduction
In recent decades the occurrence of micropollutants in the surface and groundwater has become an issue of great concern worldwide. Micropollutants or emerging pollutants (EPs), also known as emerging contaminants or contaminants of emerging concern (CECs), are synthetic or natural compounds that enter the environment and can have adverse effects on humans and ecosystems [1], also posing a threat to the sustainability of freshwater resources. They include a wide range of substances, such as pharmaceuticals and personal care products, agrochemicals, such as pesticides and herbicides, and other industrial chemicals, such as plasticizers, flame retardants, food additives, etc. [2]. Although they are found in trace concentrations (ng/L to µg/L) [3,4] in water bodies and wastewater, their continuous discharge and stability lead to bio-accumulation in ecosystems and living organisms and biomagnification in some species through the trophic chain [5], increasing awareness of their potential toxicity. Furthermore, synthetic chemicals can be transformed into intermediate products with different physicochemical properties and ecotoxicological profiles, depending on the environment in which they occur (e.g., groundwater, surface water, sediments, wastewater treatment plants, drinking water facilities) [6]. Endocrine disrupting chemicals or endocrine disruptors (EDCs) and non-steroidal anti-inflammatory drugs (NSAIDs) are two main subgroups of CECs, while according to researchers [7,8] experiments and pilot scale applications for the in situ remediation of TCE [60], heavy metals [61][62][63][64][65][66], PCE [67], PAHs [68], explosive materials [69], and dyes [70,71], most of the experiments on the treatment of CECs are related to batch testing, and research is very sparse [72,73], with continuous flow experiments simulating real conditions. This study aims to fill this gap by using nZVI in column systems to evaluate the removal efficiency of selected NSAIDs and EDCs from wastewater, with and without the addition of an oxidative reagent that involves AOPs. To achieve this goal, nano zero valent iron was incorporated into a cationic resin through an environmentally friendly synthesis method where the adsorbed trivalent iron is reduced to zero valent though the application of polyphenols produced by green tea extract. The resin not only serves as a carrier material for the nZVI to ensure that the nanoparticles do not escape together with the treated medium, but also as a suitable sorbing medium for column systems. Other researchers have used biopolymers, such as alginate beads, to cage NPs [74]. In addition, the resin does not allow the agglomeration of the nanoparticles, and the polyphenols act both as a reducing and capping agent, prolonging the reactivity of the nZVI. The synthesized nanocomposite was named R-nFe. To the authors' best knowledge, this is the first study where green teaproduced nZVI is used in combination with an oxidative reagent in column experiments to treat CECs from wastewater. The target compounds of the present study were ibuprofen (IBU), naproxen (NPX), diclofenac (DCF), and ketoprofen (KFN) for the NSAIDs category, and bisphenol A (BPA) for the EDCs group. Here, BPA is a synthetic estrogen used to harden polycarbonate plastics and epoxy resins [75], whereas IBU, NPX, DCF, and KTP are pharmaceutical compounds commonly used to relieve pain and inflammations. The DCF and IBU compounds are the most frequently detected NSAIDs globally, with NPX and KFN following [76]. The main objectives of the present study were the investigation of the effects of different parameters, such as contact time, initial concentration of target compounds, influent pH, synergy and dose of an oxidative reagent, and the development of an integrated treatment system. During the experimental period, two (2) potential oxidative reagents were tested in batch experiments as potential candidates to be combined with the nanocomposite material, namely sodium persulfate (PS) and hydrogen peroxide (H 2 O 2 ).

R-nFe Synthesis
The R-nFe composite material was synthesized following the method proposed by Toli et al. [52], and a pretreatment step using 1 M NaCl solution was also performed, since it has a positive effect on the removal efficiency of the pollutants, as recently proposed by Panagou et al. [77]. The synthesis procedure consisted of the following steps: (1) the cationic resin Amberlyst 15 H + in its wet form was agitated at 200 rpm for 2 h with NaCl solution (1 M) in order to reduce the acidity. The resin that was produced mainly carries on Na + ions and is called R-Na. (2) R-Na was agitated with FeCl 3 6H 2 O solution 0.05 M at 200 rpm for 4 h. Here, Fe 3+ replaced Na + ions due to the higher electrostatic affinity from the sulfonic groups of the resin. The resin with the adsorbed Fe 3+ species is called R-Fe. (3) The R-Fe was agitated with green tea extract for 20 h with the aim of reducing the Fe 3+ species to Fe 0 by polyphenols, generating the final composite material that is called R-nFe. Finally, (4) the R-nFe was treated with 1 M NaCl solution with a mixing ratio of 100 g R-nFe/1000 mL NaCl solution as a pre-treatment step that was performed 1 day before the use of the R-nFe in each experiment. The same nanocomposite material has been used in a previous study, and information about the morphology of the resin beads and elemental analysis of the observed surfaces is provided there in detail [77]. The SEM analysis indicated that iron content was uniformly distributed, while the TEM analysis showed that iron nanoparticles had a spherical shape and size in the order of 20-40 nm. An EDS analysis was also performed in the same study.

Wastewater Source and Characteristics
All column experiments were performed using tertiary treated wastewater from the main wastewater treatment plant of Athens (Psyttaleia). The main quality parameters of the wastewater are summarized in Table 1.

Experimental Set-Up
Within the context of this experimental work, two types of experiments were conducted-batch tests and continuous flow experiments (column experiments)-in order to study the effect of different parameters on the removal of selected CECs with a combination of nanotechnology (nZVI) and advanced oxidation processes (AOPs). The target compounds of the present study were ibuprofen (IBU), naproxen (NPX), diclofenac (DCF), ketoprofen (KFN), and bisphenol A (BPA).

Batch Experiments
The goal of the batch experiments was the selection of the most suitable oxidative reagent for combination with R-nFe material. The synergistic effect of two oxidative reagents, sodium persulfate (PS) and hydrogen peroxide (H 2 O 2 ), with the R-nFe material was investigated. The matrix used was distilled water. The initial concentration of the compounds was 10 µg/L, while the R-nFe concentration was 15 g/L. The pH was adjusted to an acidic value of 3-3.5. Schott glass bottles of 1 L each were used. Three different concentrations (1, 2, and 3 mM) were tested at two different contact times (30 and 60 min) at a stirring speed of 200 rpm. In addition, the combination of 1 mM PS with 1 mM H 2 O 2 , in the presence of R-nFe, was performed to investigate whether the cheaper H 2 O 2 can partially replace PS. Control experiments were also performed using only PS or only H 2 O 2 at three different concentrations (1, 2, and 3 mM) and only R-nFe at 15 g/L. In Table 2, a detailed description of the experimental conditions in every batch experiment is provided.

Column Experiments
In the continuous flow study, experiments were carried out using four columns made of cell cast acrylic material (plexiglass), with a height of 1.40 m and a circular cross-section with an inner diameter of 0.05 m, in different configurations. Each column was filled with R-nFe and, to prevent material loss, a 1 cm layer of glass wool was placed at the bottom. The plastic tank feeding the system contained tertiary effluent, and target pollutants were added at the desirable concentration. Continuous downward flow was applied using a peristaltic pump, at a constant flowrate of approximately 26 L/day. Each column was equipped with a sampling port located at the bottom of each column.
The column experiments were conducted in two phases. In the first phase, the experiments were conducted with four columns connected in series in order to investigate the effect of contact time and initial concentration of the target compounds. The contact times tested were 2.2, 4.4, 6.6, and 8.8 min, while the initial concentrations of CECs were 1 and 5 µg/L. These values were chosen according to the range of the concentration of emerging contaminants in the treated wastewater that mostly varies from a few ng to a few µg per liter [5,19,78,79]. The general configuration is shown in Figure 1a. Practically, this set-up enabled the simulation of four different bed heights, with the bed height of the last column being four times taller (40 cm) than the first column's bed height (10 cm). In the second phase, the column experiments were continued with the two columns of the original system working in parallel, while the initial concentration of the compounds was 5 µg/L. In this phase, the effects of pH adjustment on the influent matrix and contact time at a controlled pH were investigated by performing an experiment with pH adjustment to acidic values for two alternative contact times (2.2 and 4.4 min). Furthermore, column experiments were performed at acidic pH with PS addition which, according to the batch experiments, was the optimum oxidative reagent. The PS was supplied to the columns with a peristaltic pump (Shenchen LabM6, Baoding Shenchen Precision Pump Co., Ltd., Baoding, China). The flow rate of the PS solution was 1.2 L/d. The general configuration is shown in Figure 1b. The effect of PS dose was investigated by performing two experiments at a contact time (CT) of 2.2 min with concentrations of 1 and 5 mM, chosen in accordance with the concentration range of PS used in the batch experiments and in other studies [29,75,[80][81][82][83].
All experiments (column and batch) took place under aerobic conditions (DO = 6-8 mg/L), and the temperature ranged from 17-26 • C. The main parameters of each experimental run of the column system are summarized in Table 3.

Elaboration of Data
Contact time was calculated using the following Equation (1): while θ (the external porosity) was given by the following Equation (2): where ρ bulk is the bulk density, which is equal to the following: where A is the surface area of the cross-section equal to 20 cm 2 , H is the height of the filling material, M is the mass of the R-nFe material, ρ particle is the particle density equal to 1278 kg/m 3 , and Q is the flowrate equal to 26 L/d.

Sampling Procedure
Sampling was performed at predetermined time intervals. In the column experiments, liquid samples were collected from the sampling port at the bottom of each column, while in the batch tests, samples were withdrawn from the Schott bottles. Collected samples were then filtered through 0.45 µm membranes, acidified with 2 N HCl to a pH of 2.5, and stored at 4 • C for a maximum of 48 h before analysis.

Analytical Methods
The qualification and quantification of the target compounds (IBU, NPX, DCF, KFN, and BPA) was carried out according to the analytical method developed by Samaras et al., 2011 [84]. The main steps of this method are as follows: (i) filtration of the samples, (ii) acidification (pH = 2.5), (iii) addition of the surrogates, (iv) solid phase extraction where elutions of 6 mL of ETH are produced, (v) evaporation of the elusions to dryness under nitrogen purge (N 2 ), (vi) derivatization with 10 µL pyridine and 50 µL BSTFA + 1% TMCS at a bathing device of 70 • C for 20 min, and, finally, (vii) GC-MS analysis. Here, BPA-d16 and MCF are used as surrogates for the EDC and NSAIDs, respectively. The gas chromatograph (GC) was a 7890A, and the mass selective detector (MSD) was a 5975C from Agilent Technologies. The software used was Agilent ChemStation.

Effect of Target Compounds Initial Concentration
The effect of the target compounds' initial concentration on CECs removal was investigated by conducting experiments at initial concentrations of 1 and 5 µg/L. Figure 2 shows the evolution of DCF and BPA during the operating time of the column system for the contact time of 8.8 min. The figures for the evolution of IBU, NPX, and KFN are provided in the Supplementary Materials ( Figure S1). It is observed that with the increase in the initial concentration from 1 to 5 µg/L, the performance of the R-nFe column for the removal of IBU, NPX, DCF, and KFN declined. The ratio of the mass of the pollutant removed per mass of the nanocomposite material (R-nFe) at the end of each experiment (Table 4) increases for all the target compounds at the higher concentration (C 0 = 5 µg/L), especially in the case of BPA. The main possible removal mechanisms involved are adsorption into the surface of (hydro)oxides of nZVI or into the resin, oxidation by radicals through a Fenton-like mechanism with a small contribution depending also on the availability of H + ions and reduction for the case of DCF. The ability of the R-nFe to remove the target NSAIDs is reduced faster in the case of the 5 µg/L initial concentration, indicating an exhaustion of the active sites for adsorption and other mechanisms (oxidation, reduction) which occur to a lower extent. However, BPA molecules continued to be removed at the highest initial concentration, indicating no saturation until 192 h of operation, showing an enhanced process longevity and removal ability. Regarding other studies in the literature, Sulaiman and Al-Jabari [73] noticed a slight reduction in the removal of DCF when the initial concentration increased in their column tests using sand and nZVI supported on bentonite. It is emphasized that this is the second time that nZVI has been evaluated in column systems for the removal of CECs. Panagou et al. [77] observed that with the increase in NSAID from 1 to 10 µg/L, R-nFe removal efficiency is significantly reduced at a dose of 15 g/L during batch experiments. Consistent with this finding, Ali et al. [85] observed during batch tests that the removal efficiency of amoxicillin by activated carbon from pomegranate peel coated with nZVI decreased with the increase in the concentration of amoxicillin due to the reduction in and saturation of active sites. Dehghani et al. [86] also found during batch tests that with the increase in the initial BPA concentration from 2 to 6 mg/L, removal efficiency of BPA using nZVI-chitosan slightly increased from 84% to 90%. They concluded that removal efficiency improved because more BPA molecules could be removed by the available surface area of nZVI-chitosan.
It seems that increasing the initial concentration of contaminants plays an important role in the longevity of the process. The performance of the system in terms of removal efficiency towards operating time declined for the IBU, NPX, DCF, and KFN when 5 µg/L was tested instead of 1 µg/L, while this was not the case for BPA. The initial concentration of 5 µg/L was adopted for the following experiments.

Effect of Contact Time
In this section, the effect of contact time is discussed, and the results for DCF and BPA are illustrated in Figure 3, while the graphs for the other compounds are shown in the Supplementary Materials ( Figure S2). These experiments were conducted with no pH control. In order to investigate whether the effect of contact time was crucial for the performance of the column and whether lower contact times may facilitate similar removals, improving cost efficiency and treatment time, four different contact times were investigated with values between 2.2-8.8 min, at an initial CEC concentration of 5 µg/L. The decrease in contact time from 8.8 min to 2.2 min indicated a slight decrease in the removal efficiencies of NSAIDs achieved with R-nFe material; however, the ratio of pollutant mass removed per mass of nanocomposite material (R-nFe) did not significantly increase with the increase in contact time (Table 5). In general, all NSAID removal efficiencies were low for all CTs studied. More specifically, the concentration of IBU at the outlet of the R-nFe column with CT 2.2 min was approximately equal to that of the influent tank after 6 h of continuous operation, whereas for NPX, DCF, and KFN, this was observed at 24 h. Even with the longest contact time (CT = 8.8 min), the R-nFe column seemed insufficient to remove the selected NSAIDs after a period of 3 days of continuous operation. On the other hand, the effect of CT was more profound in the case of BPA. The performance of the R-nFe column was significantly better for BPA removal, with a removal efficiency of 90% after 2 h of continuous operation and approximately 10% after 72 h with contact time of 2.2 min. As shown in Table 5, the mass ratio of BPA removed per mass of R-nFe material remained almost constant for all CT studied. This enhanced performance was also evident in the case of the highest contact time studied (CT = 8.8 min), as the R-nFe material could no longer remove BPA after 192 h of continuous operation. Table 5. The ratio of mass of pollutant removed per mass of the nanocomposite material (R-nFe) at the end of each experiment, conducted at four different contact times (initial concentration of each pollutant 5 µg/L, without pH adjustment).

Mass of Pollutant Removed per Mass of R-nFe (µg/g) at the End of Each
Experiment (C 0 = 5 µg/L) As mentioned above, these experiments were conducted with no pH control. A progressive increase in the pH value was observed in all columns, starting from acidic values (approximately 2.1-4.9) until it progressively reached the neutral pH of the inlet tank. Interestingly, the removal efficiencies in the case of the target compounds belonging to NSAIDs decreased gradually and concurrently with the increase in the pH value to alkaline levels. Therefore, it is anticipated that the role of pH may be more profound than the role of contact time for NSAID removal for the range of CTs studied. On the other hand, in the case of BPA, the removal efficiency remained high, even when the effluent pH was increased at values similar to the ones of the influent pH. This result was consistent for all the different contact times investigated. Thus, in the case of BPA, the effect of contact time is considered to be more significant than the effect of pH, showing that adsorption plays an important role in BPA removal. This could be attributed to the fact that pKa values of all selected NSAIDs vary from 4.14-4.59, whereas the pKa value of BPA is significantly higher (9.6). Thus, while the molecules of NSAIDs are changing their form from neutral to anionic [87], the form of BPA remained neutral in the pH range from 2.1-7.4 [88]. Furthermore, the nZVI surface charge is considered to be neutralized at pH values around 7, while for lower pH values, the nZVI surface charge is positively charged [73,89]. In addition to pH, which as shown affects the extend of electrostatic interactions and adsorption of the compounds to the R-nFe material, dissolved oxygen may also exert an effect on CEC removal. It should be underlined that all columns were operated under aerobic conditions and, thus, dissolved oxygen may have triggered iron corrosion, resulting in higher concentrations of Fe +2 and Fe +3 ions in the liquid, thus, promoting the occurrence of Fenton reactions, as reported by other researchers [89,90]. In order to further study the effect of pH on CEC removal by the R-nFe columns, an additional experiment was conducted at controlled pHs, as described in the next section.

Effect of pH Adjustment and Contact Time
The effect of pH on CEC removal was further investigated by conducting column experiments at lower CT values (2.2 min and 4.4 min). The acidic influent pH favored the removal efficiency of the target compounds, as illustrated in Figure 4 for DCF and BPA, and in the Supplementary Materials ( Figure S3) for the other target compounds. The removal efficiencies for controlled and uncontrolled pH are displayed for the two alternative contact times. One possible mechanism attributed to this pH effect could be the activation of the Fenton mechanism. At acidic pH, hydroxyl radicals can be formed according to Equations (4) and (5), where hydrogen peroxide can be generated from nZVI reacting with oxygen. Indeed, Fe 2+ can be formed by the oxidation of nZVI through Equation (4), but also through Equations (6)- (8). In addition, acidic pH favors nZVI corrosion and dissolution of the passive film that gradually forms around the nanoparticles of zero valent iron, leading to the formation of more Fe 2+ ions, which are used via the Fenton mechanism to create HO radicals [29,82,91].
In addition, the formation of iron oxides and hydroxides is promoted under neutral and alkaline conditions, leading to the passivation of nZVI and, thus, preventing iron corrosion and contact with pollutants [92]. Despite the positive effect of pH adjustment to obtain acidic values, the removal efficiency remains low to moderate after 2 days of operation, suggesting that an oxidative reagent may be required. The positive effect of pH adjustment to obtain acidic values, especially at pH 3, for the removal of NSAIDs and EDCs by nZVI, was also observed in our previous work in batch experiments [77]. Furthermore it is consistent with results of batch experiments reported by other researchers in the literature [29,77,86,[93][94][95][96]. For example, Bao et al. [93] found that the degradation rate of BPA by nZVI supported on biochar was reduced (from 99.5% to 52.2%) when the initial pH of the solution was increased from 4 to 11, with a pH of 3 chosen as the optimal value. They also observed a similar performance for the BPA degradation in bare nZVI. Wei et al. [95] used E. faecalis in order to bio-recover Pd, which was then treated with Fe 3 O 4 and nZVI to produce bio-Pd/Fe 0 attached to Fe 3 O 4 nanoparticles for DCF degradation. According to their experiments conducted at a pH range of 3-9, they observed that acidic conditions favor DCF degradation. High degradation efficiencies of 94.69% and 99.65% in 20 min and 40 min, respectively, were obtained at pH 3, while they were very low at pH 7 and pH 9.
The pH adjustment to acidic values (3.5) enhanced the performance of the system, as evidenced by the higher ratio of mass of each compound removed per mass of R-nFe (Table 6). However, an increase in contact time from 2.2 to 4.4 min could not significantly enhance the removal ability and the process longevity of the system. As previously mentioned, acidic pH can trigger Fenton-like reactions, causing oxidation of the compounds. However, without the addition of an oxidative reagent the duration of this phenomenon is limited. For BPA, higher removal efficiencies were recorded on the first day of operation for both CT = 2.2 min (50%) and CT = 4.4 min (63%). On subsequent days, removal efficiencies decreased to 5% and 31% for CTs equal to 2.2 min and 4.4 min, respectively. For NSAIDs, the highest IBU and DCF removal efficiencies were obtained on the first day of operation. The IBU removal was 74% and 79% for CTs equal to 2.2 min and 4.4 min, respectively. The DCF removal was 81% and 86% for the two CT values investigated. Removal efficiencies followed the same trend for all the CECs studied and gradually decreased to values between 0% and 45% by the end of the experiments (Figure 4). Table 6. The ratio of mass of pollutant removed per mass of the nanocomposite material (R-nFe) at the end of each experiment, with and without pH adjustment to 3.5 for the contact times of 2.2 min and 4.4 min (initial concentration of each pollutant, 5 µg/L).

Mass of Pollutant Removed per Mass of R-nFe (µg/g) at the End of
Each Experiment (C 0 = 5 µg/L)

Oxidative Reagent Selection
Setting the inlet pH to acidic values (around 3.5) ameliorated the removal efficiency for most of the tested compounds, but for BPA and IBU it remained quite low. Within the context of this study, batch tests were carried out to evaluate the effect of H 2 O 2 and PS sole addition on CECs removal, or combined with nZVI particles, as an activation agent.
The synergistic effect of R-nFe with the oxidative reagents appears to be a promising technology for the removal of CECs from water and wastewater. Many researchers have investigated the synergy of nZVI with PS and H 2 O 2 in batch tests performed either exclusively with each reagent [29,75,82,83,93,97,98] or with their combination [81,99]. Most publications reported that optimum results were obtained at an acidic pH of 3.
As shown in Figure 5, the control experiments showed that the two oxidative reagents alone could not effectively remove the target compounds, even at the highest concentration tested (3 mM) and the longest contact time (60 min). The effectiveness of the synergy between R-nFe material and oxidative reagent is quite high, especially for PS. The superiority of PS over H 2 O 2 was recorded for DCF, BPA, and NPX. The removal efficiency of DCF and BPA was below 60% for R-nFe/H 2 O 2 at all concentrations tested (1, 2, and 3 mM) for contact times of 30 and 60 min, respectively. On the other hand, even at the lowest dose of 1 mM, the addition of PS (R-nFe/PS system) achieved the removal of more than 80% of DCF at 30 min contact time and more than 70% of BPA after 60 min contact time. As can be seen in Figure 6, the removal of NPX by R-nFe/PS time was more than 80% after a 30 min contact time, while R-nFe/H 2 O 2 contributed to the removal of almost 60-70% with no significant differences between the different doses tested. The simultaneous combination of H 2 O 2 and PS with R-nFe showed a similar performance to the R-nFe/PS at 2 mM and achieved a high removal efficiency (more than 80%) for DCF, BPA, and NPX for both contact times tested (30 and 60 min). Furthermore, IBU and KFN recorded a moderate removal (about 50%) by R-nFe/PS at 1 mM for both of the tested contact times.
With the addition of H 2 O 2 , hydroxyl radicals (Equation (5)) could be formed by the divalent iron produced by nZVI oxidation (Equations (4), (6)-(8)) in a Fenton-like process, as described in previous section. Furthermore, in the presence of protons (H + ) and dissolved oxygen, H 2 O 2 could also be generated as depicted in Equation (4).
The synergistic action of R-nFe with PS leads to the formation of sulfate radicals and possibly, to a lesser extent, hydroxyl radicals. Sulfate radicals can be formed directly (Equation (9)) or indirectly (Equations (10) and (11)) from nZVI if persulfate ions are present. Hydroxyl radicals can be formed by the reaction of sulfate radicals with hydroxyl groups (Equation (12)) or water (Equation (13)) [81] or by the direct reaction of nZVI with persulfate ions and water (Equation (14)) [29]. In addition, a heterogeneous Fenton reaction process is possible (Equations (4) and (5)). Both types of radicals (SO 4 − and HO) are assumed to be present during the treatment process involving the addition of PS, with sulfate radicals predominating [29]. The degradation of pollutants by SO 4 − is mainly by single-electron transfer, addition-elimination, and H-atom abstraction. The addition of HO to C=C leads to H-atom abstraction from C-H, N-H, or O-H [81]. The SO 4 − radicals tend to react with the aromatic ring whereas HO reacts with both the aromatic ring and the aliphatic chain [100]. When combining H 2 O 2 with PS, both radical species (HO and SO 4 − ) are expected to play an important role.
Since PS addition seemed to be more efficient than H 2 O 2 addition for the removal of most of the target compounds, it was decided to complement the nZVI system with the addition of a sodium persulfate solution.

Effect of PS Dose at Controlled pH
Taking into account the findings of the experiments presented in Sections 3.3 and 3.4, i.e., that the removal efficiency of the selected compounds is enhanced by acidic pH, while CT does not significantly affect removal and high removal rates could only be obtained by the synergistic effect of R-nFe with PS, it was decided to perform additional column experiments with a CT of 2.2 min, an acidic pH of 3.5, and PS addition. Two different PS concentrations (1 mM and 5 mM or 238 mg/L and 1191 mg/L) were investigated in the column set-up. Figure 7 illustrates the effect of the PS concentration on the removal efficiency of DCF and BPA, while the graphs for the other compounds can be found in the Supplementary Materials ( Figure S4). For IBU, the addition of 1 mM of PS resulted in steady, approximately 60% removal even after 250 h of operation. At the high PS dose (5 mM), the removal efficiency remained around 80% throughout the experiment. In the case of NPX, BPA, and DCF, the system showed high removal efficiency, which was 80-90% throughout the experiment for both PS concentrations without significant differences, which can be attributed to the highly reactive SO 4 − free radicals. In contrast, KTP was not as readily degraded in wastewater, with average removal efficiencies of about 35% and 65% for PS 1 mM and PS 5 mM, respectively. With respect to the ratio of mass of the pollutant removed per mass of R-nFe, it is obvious that the addition of PS enhanced the performance of the system for all the studied substances, as illustrated in Table 7, indicating the importance of oxidation by radicals. For instance, the ratio for DCF compound removal increased to 6.70 µg DCF/g R-nFe with the addition of 1 mM PS, whereas the obtained removal without PS addition at the same contact time was equal to 0.18 and 3.38 µg DCF/g R-nFe at pH = 7 and 3.5, respectively. The improvement of CEC degradation achieved by coupling nZVI with PS has been studied in batch tests, as mentioned in Section 3.4, but this is the first time that the synergistic action of nZVI with PS is being evaluated in column tests for these recalcitrant organics. The results indicate that in the studied range of PS, the addition of a higher dose of PS (five times higher) did not significantly improve the removal efficiency or the longevity of the process for NPX, DCF, and BPA. As described in the previous section in detail, the main mechanisms involved in the removal of the target compounds are oxidation by sulfate radicals and possibly hydroxyl radicals, and adsorption on the surface of the (hydro)oxides of nZVI or in the resin pores.

Conclusions
The present work focused on proposing an integrated system for the removal of selected NSAIDs and an EDC in continuous flow column experiments. Based on the results, nZVI appears to be a promising technology for the removal of CECs from tertiary wastewater and could be used as a post-treatment method in WWTPs. This is the first time that nZVI has been studied and optimized for the removal of CECs in column tests assessing the longevity of the system's performance. It should be underlined that advanced treatment technologies are of increasing importance globally but also specifically in Europe, in view of the anticipated requirement for quaternary wastewater treatment in order to eliminate micro-pollutants that is included in the European Commission proposal (https://environment.ec.europa.eu/publications/proposal-revised-urbanwastewater-treatment-directive_en, accessed on 11 January 2023) for the revision of the Urban Wastewater Treatment Directive (UWWTD) 91/271/EEC.
According to the results, the optimum system's configuration consists of the nZVI column with pH adjustment of the influent at values around 3-3.5 and the addition of persulfate at a concentration of 1 mM. This configuration appeared to be very promising as it was able to achieve constantly high removals for almost all target compounds with values ranging between 70-99% for NPX, 70-95% for DCF, and 61-91% for BPA throughout the experiments, while IBU and KFN exhibited lower average removals of 58% and 35%, respectively. Future research should concentrate on the evaluation of the sustainability of the proposed system, mostly focusing on the regeneration of the resin. Furthermore, the pH control requirement of influent and effluent wastewater hinders the application and the cost efficiency of the process. Therefore, more studies are needed in order to evaluate the possibility of operating efficiently at higher pH values.