Highly efficient removal of Sb(V) from water by franklinite-containing nano-FeZn composites

The existence of toxic and carcinogenic pentavalent antimony in water is a great safety problem. In order to remove antimony(V) from water, the purpose of this study was to prepare a novel graphene nano iron zinc (rGO/NZV-FeZn) photocatalyst via hydrothermal method followed by ultrasonication. Herein, weakly magnetic nano-Fe–Zn materials (NZV-FeZn, GACSP/NZV-FeZn, and rGO/NZV-FeZn) capable of rapid and efficient Sb(V) adsorption from water were prepared and characterised. In particular, rGO/NZV-FeZn was shown to comprise franklinite, Fe0, and graphite. Adsorption data were fitted by a quasi-second-order kinetic equation and Langmuir model, revealing that among these materials, NZV-FeZn exhibited the best Sb removal performance (543.9 mgSb gNZV-FeZn−1, R2 = 0.951). In a practical decontamination test, Sb removal efficiency of 99.38% was obtained for a reaction column filled with 3.5 g of rGO/NZV-FeZn. Column regenerability was tested at an initial concentration of 0.8111 mgSb L−1, and the treated water obtained after five consecutive runs complied with the GB5749-2006 requirement for Sb. rGO/NZV-FeZn was suggested to remove Sb(V) through adsorption-photocatalytic reduction and flocculation sedimentation mechanisms and, in view of its high cost performance, stability, and upscalable synthesis, was concluded to hold great promise for source water and wastewater treatment.

www.nature.com/scientificreports/ raw materials (except for NaBH 4 and graphene) are not expensive. Still, graphene can be prepared by the relatively inexpensive method of cooling-coupled ball milling, while NaBH 4 can be replaced by other reducing agents, and activated carbon powder with a cheaper special formula can replace the synthesised materials on a large application scale. Therefore, the developed method allows excellent Sb sorbents to be prepared at a relatively low cost. Figure 1b presents the effects of five auxiliaries, including HCOONa, white polymeric aluminium chloride (PAC-02), and anionic polyacrylamide (PAM), on the efficiency of Sb(V) removal by GAC SP /NZV-FeZn, revealing that this efficiency decreased in the following order: V > VI > II > IV > III > I. However, considering that raw water may contain other ions and organic substances/pigments, PAC-02 and PAM additives were concluded to be well suited for water treatment.

Batch processing experiments. Effects of auxiliaries on Sb(V) removal efficiency.
Effect of pH on Sb(V) removal efficiency. pH strongly affected the efficiency of Sb(V) removal by rGO/NZV-FeZn (Fig. 1c), which was ascribed to the decrease in Sb(V) solubility under alkaline conditions; the optimal pH was determined as 6.45.
Effects of illumination and oxygen on Sb(V) removal efficiency. Figure 1d presents the effects of illumination and oxygen on the efficiency of Sb(V) removal, revealing that for short treatment times, the effects of anaerobic + dark conditions were small and similar to those of anaerobic + light conditions. However, for a longer treatment time of two days, the best and second best effects were observed for anaerobic + dark and oxygen + light conditions, respectively. At this point, one should mention that the conditions were not strictly anaerobic or dark, as small amounts of oxygen and light were still present. For example, under dark conditions, the sample was covered by a box to block sunlight, i.e. it was still exposed to a certain amount of scattered light. Moreover, exposure to light occurred during sampling.
Adsorbent regenerability testing. The batch adsorption process was repeated six times at an adsorbent loading of 10 g L −1 and initial Sb concentration of 9.173 mg L −1 , with each set of experiments carried out in triplicate. The evaluated adsorbents performed well during the first five runs (Fig. 2).
Sorption isotherms and kinetics. Figure 3a,b present Sb(V) sorption kinetics and isotherms, revealing that Sb(V) removal was best described by a quasi-second-order kinetic model. In comparison, the Langmuir model was a more suitable thermodynamic model, as the reliability of the thermodynamic equation is not as good as that of the dynamic equation (evaluated in terms of R 2 ). The maximum Sb(V) adsorption capacities of NZV-FeZn, GAC SP /NZV-FeZn, and rGO/NZV-FeZn were determined as 543.9, 341.1, and 391.1 mg g −1 (R 2 = 0.951, 0.969, and 0.965), respectively. Although the adsorption capacity of GAC SP /NZV-Fe-Zn was lower than that of rGO/NZV-FeZn, the former adsorbent was better suited for industrial-scale production, as GAC SP is significantly cheaper than rGO.
Continuous processing experiments. The results of column regenerability evaluation are presented in Fig. 4 and Table 1. Specifically, Fig. 4 presents the performances of columns with (a) 0.3 g of rGO/NZV-FeZn at an initial Sb concentration of 0.9324 mg L −1 and (b) 3.5 g of rGO/NZV-FeZn at an initial Sb concentration of 0.8111 mg L −1 . Table 1 presents the contents of Fe, Zn, Al, and Sb in the effluents collected during six-fold treatment for columns with an adsorbent loading of 3.5 g.
For the column in Fig. 4a, the effect of the first five treatments was better than that of the sixth treatment. In this case, the lower adsorbent loading and higher initial concentration of Sb resulted in a relatively low volume of effluent satisfying the condition of [Sb] = 0.005 mg L −1 (i.e. C e /C 0 = 0.00536). However, if C e /C 0 is set to 0.1055 (i.e. [Sb] = 0.09837 mg L −1 (Actual measured value) and Sb removal efficiency = 83.4%), the column performance  www.nature.com/scientificreports/ decreases only slightly during the first five runs, with the respective effluent volumes satisfying this condition determined as 1800, 1200, 800, 600, and 500 mL, respectively. For the column in Fig. 4b, good performance was observed for all six runs. As the inflection point was set to 0.005 mg Sb L −1 , the effluents collected during the first five runs met the requirements of GB5749-2006. The corresponding Sb removal efficiency equalled 99.38%, and the volumes of the first five effluents satisfying this condition equalled 3725, 3725, 3725, 3300, and 1600 mL, respectively. However, after five-fold regeneration, the requirement of 0.005 mg L −1 could not be met. This result is consistent with that obtained for the column in Fig. 4a under batch conditions (Fig. 2). As the maximal allowable Sb level in drinking water is very low (0.005 mg L −1 according to GB5749-2006), the adsorbent amount should exceed 0.3 g to comply with this guideline.
Fe, Zn, Al, and other ions in the effluent of the column with an adsorbent loading of 3.5 g were monitored by ICP-MS to determine the suitability of the chosen material for drinking water purification. Table 1 shows that   Given that Fe, Zn, and Al are washed off the regeneration column during operation, the adopted method is suitable for drinking water treatment, which is in line with the use of powdered materials to remove pollutants from tap water. The unused recycled materials can be utilised for wastewater treatment to increase the material utilisation efficiency and decrease treatment costs.
Effects of temperature and sunlight on adsorbent stability. The effects of storage time, temperature, and sunlight on the Sb(V) removal performance of the selected adsorbents are presented in Fig. 5. The maximum treatment capacities of freshly synthesised NZV-FeZn (4#) and rGO/NZV-FeZn (5#) were approximately 380 and 270 mg g −1 , respectively, and were both observed at an initial Sb concentration of 131.18 mg L −1 . For sealed and dried NZV-FeZn stored for 7 months (2#) and rGO/NZV-FeZn stored for 7 months (3#), the maximal adsorption capacities equalled 258.2 and 245.8 mg g −1 , respectively, and were observed at initial Sb concentrations of 95.56 and 68.18 mg L −1 , respectively. The above values were lower (by 32.05 and 8.96%, respectively) than those of freshly synthesised materials (4# and 5#). The maximum adsorption capacity of rGO/NZV-FeZn stored for 19 months (1#) equalled 124.4 mg g −1 at an initial Sb concentration of 50.36 mg L −1 , i.e. was lower than that of freshly synthesised rGO/NZV-FeZn (5#) by 67.26%. At the time of certification, the shelf life of nano-sized Fe and Zn and the materials loaded on graphene is more than 7 months, whereas the shelf life of nano-sized Fe and Zn loaded on graphene is longer.
Thus, if NZV-FeZn and rGO/NZV-FeZn are stored in sealed containers in a dry nitrogen atmosphere and tested in the absence of sunlight at 20 °C, the storage period is at least 7 months. The adsorption capacity of rGO/ NZV-FeZn stored for 19 months was only ~ 33% of that of the freshly prepared material. When NZV-FeZn and rGO/NZV-FeZn were tested under sunlight at 20-26 °C, their Sb adsorption capacities decreased by < 5% after 7-month storage. Notably, these materials could be preserved for 19 months under the above optimal conditions (dry oxygen-free atmosphere, sunlight) and retailed 66% of their capacity at low initial loadings.   XRD. The crystalline properties and composites of the three samples including the new material (1#) and stored material for one year (2#) of rGO/NZV-FeZn and the used material (3#) of rGO/NZV-FeZn(Sb) (which was treated by photocatalysis and air drying at low temperature ) were studied using XRD method. Figure 8 shows XRD patterns for rGO/ NZV-FeZn (new), rGO/NZV-FeZn (after one year) and the used material of rGO/ NZV-FeZn(Sb), and the numbers in brackets on the curve represent the crystal surface number (hkl) of XRD characterization material. The primary crystallite size of NZVs-FeZn microspheres were determined from the strongest XRD peak by Scherrer's formula 35 : Based on comparisons to a database of XRD samples, the results suggest that the crystalline material in the XRD characterisation results show that sample of 1# contains graphite-2H(11, C, 30 ± 3A), franklinite, syn (12,(Zn0.803  . By comparing the newly synthesized materials with the materials stored at room temperature for one year, it is found that the composition of the material has a little change. The zero valent iron in the old material (2#) is missing from the XRD characterization, but addition of ferrous sulfite. In addition, compared with untreated materials(2#), XRD result shows that composition of the used material (3#) of rGO/NZV-FeZn(Sb) had some changes. In which, zero valent iron basically disappears, part of divalent iron becomes trivalent iron, and the molecular formula of nano iron and zinc also had some changes, but the structure of nano iron and zinc has not been changed, the graphite crystal in the treatment process has also become smaller, and the reducibility is reduced in the used material of rGO/NZV-FeZn(Sb).
Electrochemical characterisation of rGO, NZV-FeZn, and rGO/NZV-FeZn. Further insights into the photocatalytic properties of rGO, NZV-FeZn, and rGO/NZV-FeZn were provided by electrochemical characterisation. Specifically, transient photocurrent response and electrochemical impedance spectroscopy (EIS) photocurrent measurements were performed to probe the efficiency of electron-hole separation. Figure 10a shows the transient photocurrent responses of rGO, NZV-FeZn, and rGO/NZV-FeZn in several light cycles under full-wavelength artificial light irradiation. Under the same conditions, the photocurrent response intensities of bare rGO and NZV-FeZn were significantly lower than that of rGO/NZV-FeZn, which indicated that the heterojunction constructed with rGO/NZV-FeZn can promote the separation of photogenerated electrons and holes. Figure 10b presents the Nyquist plots (real impedance (Z) vs. imaginary impedance (Z″')) of rGO, NZV-FeZn, and rGO/NZV-FeZn obtained under full-wavelength artificial light irradiation. Generally, charge transfer efficiency is negatively correlated with arc radius. This radius was smaller for rGO/NZV-FeZn than that for NZV-FeZn, which indicated that charge transfer was faster in the former case. Thus, the heterojunction formed between rGO and NZV-FeZn can significantly promote the separation and transfer of photogenerated charges, thereby extending their lifetime. Overall, all three tested materials exhibited good electrochemical properties 40 .
UV-vis diffuse reflectance spectroscopy and PL spectra of the three samples. Figure 11a presents the UV-vis diffuse reflectance spectra of rGO, NZV-FeZn, and rGO/NZV-FeZn. Since graphene, as well as NZV-FeZn and rGO/NZV-FeZn, were conductive, their spectral data were processed according to conductor type. Pure NZV-FeZn had an obvious visible light absorption edge at ~ 630 nm. Interestingly, enhanced absorption was www.nature.com/scientificreports/  www.nature.com/scientificreports/ observed within the range of 800-1200 nm, and rGO exhibited a broad absorption peak (200-1200 nm) due to light scattering by black graphite 41 . rGO/NZV-FeZn showed a relatively smooth and broad absorption peak at 400-1200 nm. Compared with that of NZV-FeZn, the spectra of rGO and rGO/NZV-FeZn became smoother at absorption edges, which was attributed to light scattering by rGO. The absorption edge of rGO/NZV-FeZn was located at 848 nm (Fig. 11a), which showed that the loading of rGO on NZV-FeZn resulted in absorption edge broadening and enhanced the absorption capacity and utilisation efficiency of visible light. The band gaps (E g ) of rGO, NZV-FeZn, and rGO\NZV-FeZn were determined as 0, 2.68, and 2.2 eV, respectively from (ahν) 2 versus hν plots (Fig. 11b). This indicates that hybridisation with rGO changed the energy-band structure of NZV-FeZn, i.e. reduced the band gap. Fig.13c presented the PL emission spectra of rGO, NZV-FeZn, and rGO/NZV-FeZn. The PL spectrum could reflect the recombination rate of photogenerated electron and hole in photocatalyst. The lower PL intensity indicated that the separation effect of photogenerated electron hole pair is better. The major emission wavelength of samples was located at 657-658 nm when excited at 370 nm. Apparently, the photoluminescence spectra of rGO and NZV-FeZn displayed strong intensity, demonstrating low separation efficiency of photoelectron and hole pairs. More importantly, compared with rGO and NZV-FeZn, rGO/ NZV-FeZn showed a significant decrease of photoluminescence intensity, which proved that construction of composite materials promoted the separation of the photogenerated charges, and inhibited the electron and hole recombination, and finally improved the photocatalytic efficiency.
Energy-band properties. The redox capabilities of the photogenerated electrons and holes arriving at the semiconductor surface depend on the valence band (VB) and conduction band (CB) potentials, respectively. Therefore, Mott-Schottky plots are suitable for the analysis of flat-band potentials and were used to characterise the energy-band properties of the synthesised adsorbents (Fig. 12). rGO, NZV-FeZn, and rGO/NZV-FeZn displayed distinct n-and p-type semiconductor characteristics 42

Elucidation of the Sb(V) removal mechanism.
Reductive flocculation hypothesis. The efficiency of Sb(V) removal by a 3.5-g rGO/NZV-FeZn column significantly decreased in the presence of 10 mg L −1 KMnO 4 (Fig. 13), as in this case, Mn was removed in preference to Sb. The concentration of Mn in the effluent was below the detection limit (1 ppb). This result, together with the fact that flocculants were present in Sb-contaminated  Photocatalytic hypothesis. Figure 14 shows the Sb removal performances of freshly prepared rGO/NZV-FeZn and NZV-FeZn at different temperatures and variable illumination conditions. At an adsorbent loading of 0.1 g L −1 , the Sb removal performance of NZV-FeZn (a1) and rGO/NZV-FeZn (b1) could be enhanced using high initial Sb concentrations, temperature of 20 °C, and sunlight (the measured temperature equalled 26 °C (a3 and b3)). A dynamic balance was reached at 20 °C in the absence of sunlight at a loading of 0.1 g L −1 (a2 and b2) or at 26 °C in the presence of sunlight (30 °C) at a loading of 0.05 g L −1 (a1 and b1). However, at an ambient temperature of 26 °C in the presence of sunlight (30 °C), the average temperature was calculated as 28 °C, and the maximum adsorption capacities of both materials (a1, b1) were much higher than those (a4, b4) obtained at 28 °C in the absence of sunlight. Therefore, sunlight was concluded to activate the removal of Sb by both adsorbents, i.e., this removal was photocatalytic in nature.
Photocatalysis allows solar energy to be directly converted into chemical energy and is therefore viewed as a green and sustainable solution to energy production and environmental problems. Temperature is usually a vital factor controlling the kinetics and thermodynamics of a given reaction. Menga et al. (2018) showed that the photocatalytic activity of three catalysts increased with increasing temperature 45 , in line with our results.  www.nature.com/scientificreports/ Physical adsorption. Subsequently, we probed the textural properties of adsorbents and correlated them with Sb removal efficiencies ( Table 3). All three adsorbents had high zeta potentials and specific surface areas, and Sb adsorption thereon was speculated to involve physisorption. Notably, hybridisation with rGO increased the specific surface area and the absolute zeta potential of Fe-Zn nanoparticles, i.e., provided more active sites and a higher reduction potential. Figure 15 presents the adsorption-photocatalytic reduction and flocculation sedimentation mechanisms of Sb(V) removal by rGO/NZV-FeZn. The enhanced photocatalytic activity of this adsorbent was mainly attributed to its broad photoabsorption and the effective separation of photogenerated electron-hole pairs. The central catalytic component (NZV-FeZn) is excited under full-wavelength artificial light irradiation because of its narrow band gap (2.68 eV). The zero-valent Fe in the above composite is a strong reductant. Additionally, rGO can act as a reductant and provides a narrower band gap (2.2 eV). The removal of Sb(V) was assumed to primarily involve chemical reduction, physisorption, flocculation, and photocatalytic action, which was demonstrated experimentally. Therefore, the following mechanism was proposed. Irradiation of the semiconducting zinc ferrite ((Zn 0.803 Fe 0.197 )(Fe 1.824 Zn 0.166 )O 4 ) affords electrons that reduce Sb(V) to Sb(III), with additional electrons provided by the conversion of Fe 0 to Fe(II) and Fe(III). Sb(III) and Fe(III) ions react with water to afford the corresponding hydroxides, which precipitate on the adsorbent under the flocculating action of polyaluminum chloride. Finally, rGO is believed to accelerate the transfer of photoelectrons to Sb(V) because of its large specific surface area and good electrical conductivity.

Discussion
Weakly magnetic NZV-FeZn, GAC SP /NZV-FeZn, and rGO/NZV-FeZn effectively and rapidly removed Sb(V) from water, and rGO/NZV-FeZn contained franklinite, zero-valent Fe, and graphite. The adsorption of Sb(V) was described using quasi-second-order and Langmuir models. Among the tested materials, the highest adsorption capacity of 543.9 mg Sb g −1 was observed for NZV-FeZn (R 2 = 0.951). However, rGO/NZV-FeZn was selected for continuous Sb(V) removal because of its higher yield and efficiency (see Supporting Information). At an initial Sb(V) concentration of 0.8111 mg L −1 , the reaction column constructed with 3.5 g of rGO/NZV-FeZn exhibited good Sb removal efficiency (99.38%) and regenerability during the first five operations before the inflection point met the GB5749-2006. As the spent catalyst was not dispersed in the treated water, it was concluded to be safe, and further characterisation provided important insights into the mechanism of Sb(V) removal. Thus, rGO/NZV-FeZn was concluded to be a cheap, stable, and efficient material for Sb removal from source water or waste water on both small and large scales. XRD results showed that rGO/NZV-FeZn contains not only spinel ((Zn 0.803 Fe 0.197 )(Fe 1.824 Zn 0.166 ) O 4 ) but also a small amount of zero-valent Fe and graphene. The experimental Table 3. Selected textural properties of adsorbents.

Methods
Materials. All reagents were purchased from commercial sources and used without further purification.  44 . NZV-FeZn-based adsorbents were prepared by liquid-phase reduction as described elsewhere 44 . For comparison, unsupported pristine NZV-FeZn was synthesised without the addition of PPG or GAC SP according to the following equation: The synthetic protocol was optimised as follows. A suspension of 100-mesh rGO or GAC SP (0 or 10 g) in a degassed aqueous solution of 1.1 M FeSO 4 , 0.5 M ZnSO 4 , and 1% PEG-6000 (100 mL) was ultrasonicated at 80 kHz for 120 min. Subsequently, a freshly prepared solution of NaBH 4 in degassed water/absolute ethanol (30:70, v/v; NaBH 4 :(Zn 2+ + Fe 2+ ) = 2:1, mol/mol) was added to the suspension at 50-60 drops min −1 , and the mixture was allowed to react for 2 h. The solid was separated via centrifugation at 10000 rpm for 15 min, dispersed in 200-250 ml deionised water, and sedimented by centrifugation at 10000 rpm for 15 min. The rinsingsedimentation procedure was performed one more time, and the solid was vacuum-dried in an oven and stored in sealed vials under argon for subsequent experiments. The following batch numbers were used: NZV-FeZn 2018091803) was used as a Sb removal agent at a loading of 1 g L −1 . System I contained 2 mM Sb with no auxiliary; system II contained 2 mM Sb, 2 mM HCOONa, 10 mg L −1 PAC-02, and 0.5 mg L −1 PAM; system III contained 2 mM Sb and 2 mM HCOONa; system IV contained 2 mM Sb, 10 mg L −1 PAC-02, and 0.5 mg L −1 PAM; system V contained 2 mM Sb and 10 mg L −1 PAC-02; and system VI contained 2 mM Sb and 0.5 mg L −1 PAM.
Effect of pH on Sb(V) removal. The pH of original solution (2 mM Sb without auxiliaries) was adjusted to 4.51, 5.45, 6.49, 7.49, or 8.49 with 0.01 M HCl, 0.05 M NaOH, and ultrapure water, and was checked using a pH meter. The sample volume was adjusted to 20 mL, and the specimens were denoted as I-V. Each specimen (5 mL) was treated with 5.0 mg of rGO-NZV-FeZn (No. 2018091803) and left to stand for 22 h, centrifuged for 2 min at 12,000 rpm, diluted 40,000-fold, and probed by ICP-MS. All the experiments were performed in triplicate.
The effects on Sb(V) removal of sunlight and oxygen were probed using a 25-mL transparent and capped scaled centrifugal tube containing 10 mL of contaminated water with an initial Sb concentration of ~ 1 mM and an adsorbent loading of ~ 1 g L −1 . First, 138.94 mg of NZV-FeZn (No. 201901801) was added upon mixing to 134 mL of the Sb solution with an initial Sb concentration of ~ 1 mM, and the suspension was divided into 10-mL portions that were placed into 25-mL centrifugal tubes for testing the effects of oxygen and sunlight. Each experiment was performed in triplicate. When the adsorbent was mixed with Sb-containing water, the vial was opened and immediately capped again. Then, the specimens were exposed to light or placed in the dark for 4 or 24 h, sampled, centrifuged, and probed by ICP-MS as described above.  Batch processing was performed as follows. Sb-containing water (9.173 mg Sb L −1 , 25 mL) and the adsorbent of choice (250 mg) were mixed for 15 min at room temperature, and the suspensions were centrifuged at 10,000 rpm for 15 min. The supernatant was collected and analysed by ICP-MS. During regeneration, the precipitate obtained after centrifugation was treated with dilute hydrochloric acid (5 mL, 0.1 MHCL) for 30 min. Then, the mixture www.nature.com/scientificreports/ was supplemented with water (35 mL) and centrifuged for 10 min at 10,000 rpm for three times. The supernatant was collected and analysed by ICP-MS. The process was repeated six times.
Sorption isotherms and kinetics. Sb(V) removal was performed at adsorbent loadings of 0.05-1.0 g L −1 using NZV-FeZn (1#, 2019042801), GAC SP /NZV-FeZn (2#, 2019042802), and rGO/NZV-FeZn (3#, 2019042803). Sb(V)-polluted water with initial Sb concentrations of 0-420 mg L −1 was prepared from pure water and Sb-K tartrate. Typically, 50 mL of the Sb(V) solution and 5.0-50 mg of the adsorbent were mixed and reacted for 24 h at room temperature. Samples taken at certain time intervals were centrifuged at 12,600 rpm for 2 min and appropriately diluted for analysis by ICP-MS (Agilent 1260-7700e; Agilent Technologies Co. Ltd., USA). The adsorption capacities (or handling capacities; q e ) and the total Sb removal efficiencies were calculated as follows: where X is the sorbent loading (g L −1 ), and C 0 and C e (mg L −1 ) are the initial and equilibrium concentrations of Sb(V), respectively. Pseudo-first-order and pseudo-second-order models were used to characterise the kinetics of Sb(V) adsorption 46 . with the integrated forms of these models given as follows: where q t is the Sb(V) adsorption capacity at time t (mg g −1 ), and k 1 (s −1 ) and k 2 (g mg −1 s −1 ) are the pseudo-firstorder and pseudo-second-order rate constants, respectively.
where q m is the maximum adsorption capacity (mg g −1 ), K L is the Langmuir adsorption equilibrium constant (L mg −1 ), K F and n are Freundlich constants representing the adsorption capacity (L −1/n mg −(1−n) g −1 ) and the adsorption intensity varying with the degree of heterogeneity, respectively, and K L ' and n' are the Langmuir adsorption equilibrium constant (L mg −1 ) and the adsorption intensity, respectively 50 .

Continuous processing experiments.
To meet the requirement of GB 5749-2006 ([Sb] < 0.005 mg L −1 ), we probed the effects of reducing the initial concentration and increasing the adsorbent loading, revealing that the latter strategy is more effective than the former. Specifically, a glass column (3.5 cm in diameter and 60 cm in height) was packed with 0.3 or 3.5 g of rGO/NZV-FeZn (2019042803) and 35 g of K-04 (granular activated carbon for pure water sterilisation) to a fixed bed depth of ~ 15.0 cm using ~ 1.5 g (3 × 0.5 g) of cotton wool and 10.0 g of macroporous resin soaked in ethanol. Aqueous Sb(V) solutions were prepared with initial concentrations of 0.8111-0.9324 mg L −1 and contained PAC-02 (10 mg L −1 ) and PAM (0.5 mg L −1 ) as auxiliaries. During operation, ~ 5000 mL of Sb-contaminated water was passed through the column at a flow velocity of 9-10.0 mL min −1 . The initial 35 mL of the effluent was discarded, and samples were subsequently collected every 200 mL (± 15 mL) and analysed by ICP-MS to determine the concentrations of Fe, Zn, Al, and Sb. During regeneration, the column was washed with 200 mL of pure water after each operation and then soaked in 50-100 mL of 0.1 M HCl for > 20 h. Column regenerability was evaluated in terms of the ratio of Sb (Ce/C0 < or = 0.2) level in the effluent water of different volumes and initial water at different moving times.
Effects of temperature and sunlight on adsorbent stability. Samples  (3) q e = (C 0 − C e )/X, www.nature.com/scientificreports/ The control materials (NZV-FeZn and rGO/NZV-FeZn) were synthesised from the same raw materials at different times by the same method (except for rGO loading) and were placed in a compact bag that was evacuated in a desiccator and then filled with nitrogen. Sb(V) solutions containing the desiccant were prepared with pure water and Sb-K tartrate at initial concentrations of 0-140 mg L −1 . The Sb(V) solution (50 or 100 mL) was treated with NZV-FeZn or rGO/NZV-FeZn (5.0 mg), and the mixture was allowed to react for 48 h at 20 °C in the presence or absence of light. Subsequently, the samples were centrifuged at 12,000 rpm for 2 min, and the concentrations of Sb(V) in the supernatants were determined by ICP-MS. The adsorption (or handling) capacity and the total Sb(V) removal efficiency were calculated as described above.
Characterisation. The surface morphologies of NZV-FeZn, GAC SP /NZV-Fe-Zn, and rGO/NZV-FeZn were examined using SEM (JEOL Ltd., Japan), while particle sizes were determined using TEM (JEOL Ltd., Japan). TEM and HR-TEM images were acquired on a JEOL JEM-2100F instrument. XRD patterns were recorded on an Ultima IV diffractometer at a scan rate of 2° min −1 within a 2θ range of 10-90° using Cu K-beta radiation and an accelerating voltage of 40 kV. Brunauer-Emmett-Teller (BET) surface areas, zeta potentials, and total Fe contents were determined according to GB/T 19587-2017 (determination of specific surface area of solid substances by gas adsorption BET method), GB/T 32666-2016 (general rules for gel particle zeta electroanalysis by electrophoresis), and JT/T 015-1996 general principles of inductively coupled plasma atomic emission spectrometry), respectively. All electrochemical measurements were performed on a CHI 660E electrochemical workstation (Shanghai Chenhua Instrument Co. Ltd., China) using a three-electrode system at room temperature. The working electrode was prepared on fluoride-tin oxide (FTO) conductor glass that was cleaned by sequential sonication in chloroform, acetone, and ethanol (30 min for each solvent). Adsorbent powder (5.0 mg) was mixed with pure water (0.5 mL) upon 30-min sonication, and 0.1 mL of the resulting slurry was diluted to 2 mg mL −1 with pure water (0.3 mL) and 0.5 wt% Nafion (0.1 mL, CAS:66796-30-3,alfa). The diluted slurry was dropped onto FTO glass and dried at 60 °C to afford a working electrode with an exposed area of 1 cm 2 . A Pt plate and a Ag/ AgCl (3 M KCl) electrode were used as counter and reference electrodes, respectively. Aqueous Na 2 SO 4 (0.2 M) was used as the electrolyte. Mott-Schottky plots were constructed for the abovementioned three-electrode system using the impedance-potential technique. Element contents and valence states were investigated using XPS (ESCALAB 250xi). Diffuse reflectance spectra were recorded on a UV-vis-NIR scanning spectrophotometer (UV3600, Shimadzu) using an integrating sphere.
Elucidation and verification of the Sb(V) removal mechanism. Based on the fast and efficient removal of Sb(V) by the tested adsorbents, we suggested that this removal occurs via reductive flocculation, photocatalysis, and physisorption, with further details provided below.
Reductive flocculation-based removal. Two Sb removal columns were loaded with 3.5 g of rGO/NZV-FeZn (2019042803) and 35 g of K-04 as described above. Sb-contaminated water (1.2-1.4 mg Sb L −1 ) containing PAC-02 (10 mg L −1 ) and PAM (0.5 mg L −1 ) as auxiliaries was divided into two parts. One part was used directly, while the other was supplemented with KMnO 4 (10 mg L −1 ). The flow rate was controlled at 9-10 mL min −1 , and a comparative test was carried out. Sampling of 15 mL per tube at 250-mL intervals, 35 ml waste liquid was discharged before sampling. In addition to 100 times dilution of the original solution, other effluent samples were detected using ICP-MS with the original solution to test the Sb content.
Photocatalytic removal. Freshly synthesised NZV-FeZn (d, 2019111401) and rGO/NZV-FeZn (e, 2019111402) were used at loadings of 0.05-0.1 g L −1 . Sb(V) solutions had initial concentrations of 0-140 mg L −1 . The Sb(V) solution (50 or 100 mL) was mixed with the adsorbent (5.0 mg) and reacted for 48 h under different conditions. The same volume of Sb solution with a different initial concentration was placed in a clean glass bottle (150 mL) as a blank control. After 48 h, the samples were centrifuged at 12,000 rpm for 2 min, and the concentrations of Sb(V) in the supernatant were analysed by ICP-MS. The adsorption (or handling) capacity and the total Sb(V) removal efficiency were calculated as described above. Each experiment was performed in triplicate. The reaction conditions were as follows. A: Reaction temperature = 20 °C, no additional sunlight. B: Temperature in reaction chamber = 20 and 26 °C in the absence and presence of additional sunlight, respectively; in the absence of sunlight, the sample was returned to the chamber at 20 °C. C: Reaction temperature = 28 °C, no sunlight. D: Temperature in reaction chamber = 26 and 30 °C in the absence and presence of additional sunlight, respectively; when there was no sunlight (6:15 PM to 8:15 AM), the sample was returned to the chamber at 26 °C. Statistical analysis. Each set of batch sorption experiments was conducted in triplicate. One-way ANOVA with Dunnett's post hoc test and Tukey's multiple comparison test was conducted using GraphPad Prism 5.0 software to determine the statistical significance of the results.