Al₂O₃–Fe₃O₄–expanded graphite nano-sandwich structure for fluoride removal from aqueous solution

Abstract

In this study, a novel Al₂O₃–Fe₃O₄–expanded graphite nano-sandwich adsorbent was prepared to remove fluoride from aqueous solutions. Field emission scanning electron microscope examination was carried out to characterize its microstructure. Fe₃O₄ cubes, with the size of 60–80 nm, and pea-shaped Al₂O₃ particles, with the length of 20–50 nm, were observed on the surface of the expanded graphite. Fourier transform infrared spectra results indicated that new functional groups had occurred on the sandwich surface following the adsorption process. Identified X-ray powder diffraction curves determined that Al₂O₃ was amorphous, while Fe₃O₄ and expanded graphite were crystalline. Vibrating sample magnetometer experiments revealed that the saturation magnetization of the nano-sandwich was 3.7 emu g⁻¹, which was strong enough to separate the sorbents from the solution. The adsorption kinetics followed the pseudo-second-order rate equation with intra-particle diffusion as the rate determining step. The adsorption pattern of the nano-sandwich followed the Langmuir isotherm better than the Freundlich isotherm and the adsorption capacity increased by rising temperature. Under optimized conditions, fluoride removal efficiency reached 96.8%. After two cycles of regeneration with 0.1 mol L⁻¹ NaOH, fluoride removal efficiency could still reach 91.4% and the residual fluoride concentration was lower than 1.5 mg L⁻¹. The sorbents demonstrated great potential in fluoride removal, in terms of higher adsorption efficiency, wider pH adsorption range, good regeneration as well as convenient solid–liquid separation characteristics.

---

Introduction

Fluoride and nitrate contamination in drinking water sources has been a major problem in many countries. With a growing concern about pollution issues in drinking water, fluorides pollution is reprehensive and thus needs to be addressed.^1,2 Trace fluoride is essential and beneficial for an organism, but excessive fluoride intake could cause irreversible or incurable dental and skeletal fluorosis.^3,4 Excessive amounts of fluoride in groundwater have been found in many regions, especially in Asian countries, such as China, India, Pakistan, and Thailand.⁵ It was reported that the concentration of fluoride in these regions was ranging between 1.0 mg L⁻¹ and 35.0 mg L⁻¹,⁶ with most reports being around 5 mg L⁻¹.⁷ The cut off value of fluoride concentration in drinking water is 1.5 mg L⁻¹, according to the standards of the World Health Organization.⁸ Therefore, it is urgent to find effective ways to overcome this issue.

Several methods have been applied to remove excessive fluorides from aqueous solution, with the most important being membrane filtration,⁹ electrocoagulation–flotation,¹⁰ adsorption,¹¹ precipitation,¹² ion-exchange¹³ and the electro-dialysis¹⁴ process. Most of them have drawbacks, such as long reaction time, low fluoride removal efficiency, secondary pollution, and poor recyclability. Therefore, the adsorption route is considered to be the most efficient method for fluoride removal in drinking water.

Aluminium oxides and hydroxides/oxy-hydroxides are abundant as minerals and have been widely used to remove fluorides. Fluoride absorption from aluminum oxides is based in the underlying chemistry process, forming of aluminum–fluoro complexes (monodentate, bidentate complexes or inner- and outer-sphere complexes).^15–19 Among them, activated alumina is an effective aluminium compound that has been categorized by US EPA as the Best Available Technology (BAT) for the removal of fluoride²⁰ due to its high surface area,²¹ amphoteric character, high selectivity towards fluoride, low cost, high abundance and ease of usage. However, the fluoride sorption capacity of many Al-based adsorbents is inefficient, similar to the activated alumina (0.61 mg g⁻¹),²² MCAA (0.15 mg g⁻¹),²³ manganese-oxide-coated alumina (0.43 mg g⁻¹),²⁴ or alumina coated pumice (0.64 mg g⁻¹).²⁵ Moreover, their applications are limited due to the shortcomings of being effective only in extreme pH conditions,²⁶ or the complicated procedures involved in the separation from treated water.^27,28 The objectives of current references hopefully solve some of the typical problems associated with regular defluoridation techniques, as-discussed above.

Among the wide-range of support materials available, graphene, carbon nanotubes, expanded graphite (EG) and its composites^29,30 of the carbon family have been demonstrated as some of the most effective and reliable due to their porous structure, high surface area and presence of surface functional groups. Graphene oxide has been used as the support materials in a supported catalyst for oxidation of organic pollutants.³¹ Carbon nanotube has been used in an electrochemical filter for treatment of aqueous antibiotics tetracycline.³² EG exhibits great promise for potential applications in adsorption for different substances such as methylene blue,³³ oil,³⁴ Sn(II)³⁵ and Cr(VI).³⁶ By combining the advantages of both carbon materials with high specific surface and adsorbent with high affinity to fluoride, mesoporous Al₂O₃/EG composites were prepared to enhanced adsorption of fluoride.³⁷ Hydrous cerium oxide–graphene composite was synthesised for rapid adsorption removal of arsenate,³⁸ which addressed the key limitations of slow adsorption kinetics for arsenic removal. The advantages of the upper integration is the combination of the benefits of both materials. On the contrary, there were no properties of Al₂O₃ and EG which can be used for separation of adsorbent and aqueous solution.

The magnetism separation technology has been extensively used for the removal of heavy metal ions, organic dyes and herbicides from aqueous media.³⁹ And the exceptional magnetic properties of Fe₃O₄ led to its wide use as a chemical compound in separation technology. Therefore, the quick separation of solid/liquid phases can be performed with an external magnetic field and the time required for separation can be dramatically reduced.^40,41

In this study, Al₂O₃–Fe₃O₄–EG (the compound of Al₂O₃, Fe₃O₄ and expanded graphite) nano-sandwich⁴² adsorbent was successfully designed by the co-precipitation method and was employed to remove fluoride from solutions. The nano-sandwich and absorption mechanism were examined via X-ray Diffraction (XRD), Field Emission Scanning Electron Microscopy (FE-SEM), Fourier Transform Infrared spectroscopy (FT-IR) and a Vibration Sample Magnetometer (VSM). The fluoride adsorption experiments of Al₂O₃–Fe₃O₄–EG were comprehensively investigated with different conditions such as pH, adsorbent dose, equilibration time, temperature, initial fluoride concentration, and co-existing anions. This study demonstrates that Al₂O₃–Fe₃O₄–EG can be an effective adsorbent for fluoride removal from aqueous solutions, operating at high separation rates.

Experimental

Materials and reagents

Graphite flakes Pectin (90–99.9%) was purchased from Yichang Xincheng graphite Co., Ltd. (Hubei, China). Sodium fluoride, sodium chloride, potassium permanganate (KMnO₄), ferrous sulfate heptahydrate, ferric chloride hexahydrate, aluminium nitrate, ammonium hydroxide, sodium hydroxide, glacial acetic acid, hydrochloric acid, sulfuric acid, phosphoric acid and nitric acid were of analytical grade and supplied by Sinopharm Chemical Reagent Co., Ltd. All chemicals were used as received without further purification.

Nano-sandwich adsorbents preparation

The nano-sandwich was prepared in three sequential steps:

(1) The preparation of base layer (EG): 6 g graphite flakes were treated with 36 mL vitriol as intercalation compound, and 1.2 g KMnO₄ as oxidant, at 40 °C for 90 minutes. After that, the mixture was rinsed in order to be neutralized for the preparation of the graphite intercalation compounds.⁴³ Then, it was soaked in 36 vol% acetic acid for 2 hours to form expandable graphite. Following rinsing and drying, expandable graphite was expanded in a muffle furnace at 900 °C for 30 s.

(2) The formation of magnetic separation layer (Fe₃O₄) by means of the co-precipitation process: 0.2 mol L⁻¹ Fe²⁺ and Fe³⁺ (molar ratio of 1 : 1.4) along with 0.5 g EG were mixed and dissolved in 85 mL distilled water. Then, at 60 °C, 35 mL dilute NH₄OH solution was added in forms of droplets into the solution to maintain pH = 10. Following 30 min, the products was rinsed with deionized water and dried at 60 °C in an oven.

(3) The formation of the adsorption layer (Al₂O₃): 0.2 g magnetic EG were added into 25 mL of 0.25 mol L⁻¹ Al(NO₃)₃ solution and was mixed by vigorous stirring for 2 hours. Then, the mixture was dried at 100 °C for 7 hours and calcined at 450 °C for 2 hours.

Characterization

XRD (D8-FOCUS) determined the crystalline phases of the samples employing a Cu-Kα radiation (λ = 0.15406 nm). FE-SEM (JSM-5610LV) was used to analyse the morphology of the samples. The FT-IR spectra were recorded in KBr pellets with a Bruker Vertex 70 instrument, in a range between 400 and 4000 cm⁻¹. The magnetic properties were analysed using a vibrating sample magnetometer (VSM, Lake Shore 7410) with an applied magnetic field from −20 kOe to 20 kOe.

Adsorption and regeneration experiments

A stock F⁻ solution (1000 mg L⁻¹) was prepared by dissolving 0.221 g anhydrous sodium fluoride in 100 mL distilled water at room temperature. Experimental solutions were prepared by appropriate dilutions of the above solution. The experimental procedure is illustrated in Fig. 1. Nano-sandwich and fluoride solution with fixed pH values were mixed in polypropylene bottles (Fig. 1a). The contents were constantly shaken during the procedure (Fig. 1b) in a water bath shaker for a period of time and then the sorbents were separated (Fig. 1c). An ion-selective electrode was used to determine the final fluoride concentration. The amount of fluoride adsorbed on the adsorbent at equilibrium state was calculated according to the following equation:where, q_t (mg g⁻¹) is the adsorbed amount at contact time t (min); C₀ and C_t (mg L⁻¹) are the initial and equilibrium concentrations of fluoride ion in the solution, respectively; V (L) is the volume of the solution; and m (g) is the amount of adsorbent added to the solution. For regeneration studies (Fig. 1d), the fluoride-saturated adsorbents were mixed with 0.1 mol L⁻¹ NaOH solution and then were shaken at 30 °C for 6 hours. Following separation, the regenerated nano-sandwich sorbents were dried at 80 °C for 3 hours and then reapplied for fluoride adsorption (Fig. 1a). After that, the whole adsorption-regeneration experiments were performed one more time, as described in Fig. 1.

Fig. 1 Schematic illustration of adsorption and regeneration experimental procedure.

Results and discussion

Characterizations of the composites

In Fig. 2a, there were no obvious diffraction peaks, implying that amorphous Al₂O₃ was present. Fig. 2b shows typical Fe₃O₄ diffraction peaks. The diffraction peaks were located at 26.81° and 54.71°, respectively, and were identified as the EG phase, as shown in Fig. 2c. All characteristic peaks of Fe₃O₄ and EG were identified and no other new peaks appeared in Fig. 2d, which means that Fe₃O₄ and EG layers exist without new materials generating. It could be further expressed that Al₂O₃ in sandwich structure demonstrates an amorphous character from Fig. 2d, which could possess enhanced accessibility to fluoride.²

Fig. 2 Identified XRD patterns of the samples: (a) Al₂O₃, (b) Fe₃O₄, (c) EG, (d) Al₂O₃–Fe₃O₄–EG.

To investigate the adsorption mechanism, FT-IR spectra of nano-sandwich adsorbents before and after adsorption were recorded and presented in Fig. 3. The broad bands detected at 3477 cm⁻¹ and 1641 cm⁻¹ in each sample represent the stretching vibrations of –OH (ref. 44) functional group of H₂O and C=O group in carboxylates, respectively. The peaks at 573, 825, 1743 and 1097 cm⁻¹ were attributed to stretching vibrations of Fe–O, Al–O, C=C and C–O groups, respectively.²⁹ Comparing Fig. 3a with b, the lack of peak at 1743 cm⁻¹ means that Fe₃O₄ was grafted on the surface of EG through the C=C groups. As observed from Fig. 3b and c, the Al₂O₃–Fe₃O₄–EG adsorption band at approximately 1097 cm⁻¹ disappeared, which suggested Al₂O₃ was grafted on the surface of Fe₃O₄–EG via the C–O groups. Fig. 3d shows a Al–F stretching vibration peak at 663 cm⁻¹, which supported the theory of Al³⁺ involvement in the adsorption of fluoride by forming a new coordination compound with F⁻.

Fig. 3 FTIR spectra of the samples: (a) EG, (b) Fe₃O₄–EG, (c) Al₂O₃–Fe₃O₄–EG before adsorption, (d) Al₂O₃–Fe₃O₄–EG after adsorption.

Fig. 4a displays the micro-topography of the nano-sandwich adsorbent, which is presented as a multiple flakes structure that contains EG, Fe₃O₄ and Al₂O₃. In another visual angle, in Fig. 4b, Fe₃O₄ and Al₂O₃ particles either gathered or scattered on EG surface layer. The EDS spectra (Fig. S1†) revealed that the elements on the surface of nano-sandwich were Fe, Al and O, which illustrates that Fe₃O₄ and Al₂O₃ layers were coated on the EG surface. The Fe₃O₄ particles had a size of 60–80 nm and a shape of irregular cubes (Fig. 4c). The nano-sized amorphous Al₂O₃ particles grafted on the EG were 20–50 nm (Fig. 4d).

Fig. 4 FSEM images of, (a) nano-sandwich structure of Al₂O₃–Fe₃O₄–EG, (b) vertical angle view of Al₂O₃–Fe₃O₄–EG, (c) Fe₃O₄–EG and (d) Al₂O₃–EG.

The magnetic test results of Al₂O₃–Fe₃O₄–EG are presented in Fig. 5. The maximum magnetic strength for Al₂O₃–Fe₃O₄–EG is an important characteristic of magnetic separation. According to the VSM magnetization curves, the highest magnetization saturation of 3.7 emu g⁻¹ was obtained and Al₂O₃–Fe₃O₄–EG synthesized here possesses superparamagnetic behavior with strong magnetic power. As indicated in the bottom right of Fig. 5, the separation of adsorbent and solution was extremely clear. This suggests that the magnetic strength of the adsorbent is sufficient to separate the Al₂O₃–Fe₃O₄–EG from the solution without the need for filtration or centrifugation. In addition, the highest magnetization saturation of adsorbent could be adjusted through changing the concentration of Fe²⁺ and Fe³⁺ during the synthetic process.

Fig. 5 VSM magnetization curve of Al₂O₃–Fe₃O₄–EG and the magnetic separation results.

Effect of pH on fluoride removal

The fluoride removal efficiency was slightly decreased by the rise of pH from 2 to 7, and reached 96.8% at pH = 4 (T = 30 °C, t = 120 min, m = 0.2 g and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution) as demonstrated in Fig. 6. The combination of electrostatic adsorption and surface complexion could explain the adsorption mechanism. The pH_pzc of the nano-sandwich was 5.2 and it was significantly important for the electrostatic adsorption. When in acidic medium (above the Point of Zero Charge (PZC)), pH < pH_pzc (5.2), the values of zeta potential were positive. Consequently, the surface of adsorbents was positively charged. The attraction between positively charged surface and negatively charged fluoride ions was attributed to fluoride removal.⁴⁵ However, in alkaline solution, pH > pH_pzc, negatively charged sorbents tend to repel the F⁻ via coulombic repulsion. Furthermore, coordination of Al³⁺ with either F⁻ or OH⁻ when the zeta potentials increased. At this situation, F⁻ and OH⁻ formed a surface competition of Al³⁺. According to the zeta potential measurements and previous studies,^46,47 the following reactions are suggested for pH lower than pH_zpc:

≡MOH_(s) + H_(aq)⁺ = ≡MOH_2(s)⁺ (2)

≡MOH_2(s)⁺ + F_(aq)⁻ = ≡MF_(s) + H₂O (3)

Fig. 6 The effect of pH on fluoride removal efficiency and zeta potential of Al₂O₃–Fe₃O₄–EG at T = 30 °C, t = 120 min, m = 0.2 g and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution.

At pH < 5.2, the hydroxyl groups of the solid adsorbent were protonated, while the electrostatic interaction between fluoride and protonated hydroxyl groups prove the fluoride sorption.

At pH > 5.2, fluoride ions were possibly adsorbed by the following ion-exchange mechanism. The predominant mechanism for F⁻ adsorption might be of the ligand exchange type interactions between the F⁻ and hydroxyl groups.

≡MOH_(s) + F_(aq)⁻ = ≡MF_(s) + OH⁻ (4)

Effect of dosage on fluoride removal

Fig. 7 illustrates the variations of the removal efficiency (%, left axis) and adsorption capacity (mg g⁻¹, right axis) at different adsorbent doses (0.1–0.6 g), at pH = 4, in 30 °C, contact time of 120 min, and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution. As can be observed from Fig. 7, the removal efficiency was significantly improved by an increase in the adsorbent dose, from approximately 91.7%, at the adsorbent dose of 0.1 g, to approximately 97.5%, at a dose of 0.6 g. In contrast. The adsorption capacity had a dramatic decrease by increasing the adsorbent doses. The adsorption capacity was decreased from 2.19 mg g⁻¹ to approximately 0.3 mg g⁻¹, when the adsorbent dose increased from 0.1 g to 0.6 g. This can be explained by the fact that increased adsorbent dose increases the accessibility of active sites on the pores of the Al₂O₃–Fe₃O₄–EG to the fluoride ions, which leads to an enhanced removal efficiency. However, when the vast majority of fluoride ions in the solution are completely adsorbed, the extra adsorbent had no effect on fluoride ions but decreased the adsorption capacity. From all the above, 0.2 g was selected as the ideal dose, where the removal efficiency was 97.5% and the adsorption capacity was approximately 1.2 mg g⁻¹.

Fig. 7 The effect of adsorbent dose on the fluoride removal efficiency and the fluoride adsorption capacity of Al₂O₃–Fe₃O₄–EG at pH = 4, T = 30 °C, t = 120 min and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution.

Effect of time and temperature on fluoride removal

Fig. S2† depicts the effect of contact time on fluoride removal efficiency of Al₂O₃–Fe₃O₄–EG, at pH = 4, at a dose of 0.2 g, in 30 °C and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution. The curve shows that the fluoride removal when time is extended. The reaction occurred acutely after approximately 50 min in the early stage and adsorption achieved dynamic balance within 120 min. The fast adsorption reaction rate of the early stage was caused by the smaller F⁻ concentration on the surface of the adsorbent, which has a large concentration gradient compared to that of the solution. The adsorption rate decreased after the concentration gradient was reduced by the complexation or ion exchange between inner surface and F⁻. Fig. S3† displays the effect of temperature on fluoride removal efficiency of Al₂O₃–Fe₃O₄–EG, at pH = 4, contact time of 120 min, a dose of 0.2 g and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution. The curves indicated that the temperature has a significant influence on the fluoride removal of Al₂O₃–Fe₃O₄–EG, at the lower temperature, where the fluoride removal increases from 82.25% to 99.6% with the temperature increasing from 10 °C to 60 °C. This is attributed to the activity of F⁻ in the solution and the effective ion in the adsorbent, both enhanced by the increasing temperature. Then the improved ion diffusion velocity led to the enhanced ion exchange rate between the F⁻ and the surface of the adsorbent. From all the above, the processing of F⁻ adsorption was endothermic and higher temperature was beneficial to the F⁻ removal. However, taking the actual operation into consideration, T = 30 °C was selected as the ideal temperature where the F⁻ removal efficiency was as high as 96.8%.

Effect of initial fluoride concentration on fluoride removal

Fig. 8 illustrates the effect of initial concentration on fluoride removal efficiency and fluoride adsorption capacity of Al₂O₃–Fe₃O₄–EG, at pH = 4, contact time of 120 min, a dose of 0.2 g, and in 30 °C. As observed in Fig. 8, the fluoride removal efficiency of Al₂O₃–Fe₃O₄–EG decreased with the initial concentration of F⁻. In contrast, the fluoride adsorption capacity increased and achieved a maximum value of 3.35 mg g⁻¹, when the initial concentration of F⁻ was 20 mg L⁻¹. This is due to the fact that when the initial concentration of F⁻ was low and the dose of Al₂O₃–Fe₃O₄–EG was invariable, the active sites on the surface of adsorbent were more than the F⁻. Then the increase of F⁻ concentration improved fluoride adsorption capacity and as a result, full advantage of the extra active sites was implemented. Since that point, fluoride adsorption capacity remained constant, although the initial concentration of F⁻ was continuously increased, because the active sites had already been used up.

Fig. 8 The effect of initial fluoride concentration on fluoride removal efficiency and fluoride adsorption capacity of Al₂O₃–Fe₃O₄–EG at pH = 4, T = 30 °C, t = 120 min and m = 0.2 g.

Effect of co-existing anions on fluoride removal

In natural environment, fluoride often co-exists with other anions, which may compete for adsorption sites and decrease the removal efficiency of the adsorbent. The effect of other anion concentrations on fluoride removal efficiency of Al₂O₃–Fe₃O₄–EG is presented in Fig. 9, at pH = 4, contact time of 120 min, a dose of 0.2 g, and in 30 °C. The effects of different coexisting ions on the fluoride removal efficiency of Al₂O₃–Fe₃O₄–EG had a large difference at various concentrations.^45,48 Cl⁻ and NO₃⁻ hardly influenced the removal of F⁻, while the other three acid radicals (i.e. PO₄³⁻, CO₃²⁻ and SO₄²⁻) had different inhibition effect on F⁻ removal of Al₂O₃–Fe₃O₄–EG at different concentrations. Comparatively, the influence of PO₄³⁻ was the most significant: when the concentration of PO₄³⁻ increased from 20 mg L⁻¹ to 200 mg L⁻¹, the F⁻ removal efficiency decreased from 85.1% to 42.1%. Considering the typical natural concentration range of phosphate in ground water (0–5 mg L⁻¹),⁴⁹ the effect to F⁻ adsorption was limited and the interference from phosphate was not as strong as shown in this study. As the CO₃²⁻ concentration in the solution increased from 20 mg L⁻¹ to 200 mg L⁻¹, the F⁻ removal efficiency decreased from 85% to 50%, vid reduced by 35%. The inhibition effect of SO₄²⁻ on fluoride adsorption was minor, vid, the F⁻ removal efficiency decreased by 15% as the concentration increased from 20 mg L⁻¹ to 200 mg L⁻¹. The main reasons that different co-existing ions had different effect on F⁻ removal are the following: (1) hydrolysis of these three weak acid radicals changed the pH of the solution. The higher the concentration, the stronger the radical hydrolyse and the higher the pH, the greater the fluoride removal efficiency. (2) These three weak acid ions occupied the active sites of the adsorbent surface instead of F⁻. Competing with the F⁻ inhibited the binding between F⁻ and adsorbent. (3) The co-existing ions effect on F⁻ removal efficiency related to the number of their negative charges. The higher amount of negative charges led to increased attraction between the ions and the adsorbent.

Fig. 9 The effect of co-existing anions concentrations on fluoride removal efficiency of Al₂O₃–Fe₃O₄–EG.

Kinetic analysis

Fig. 10a demonstrates the effect of time on the adsorption of fluoride at various temperatures. The fluoride adsorption on the nano-sandwich exhibits an initial rapid increase within the first 30 min and then gradually reaches equilibrium in 2 hours, with a removal efficiency of 96.8%. The high fluoride removal efficiency could be attributed to the high surface area of EG layer and the high fluoride adsorption capacity of Al₂O₃ functional layer.

Fig. 10 (a) Effect of time on the adsorption of fluoride at 30 (), 40 () and 50 °C () (pH = 4, T = 30 °C, m = 0.2 g and an initial fluoride concentration of 5 mg L⁻¹ in 50 mL solution); (b) pseudo-second order kinetic plot for fluoride removal at 30 (), 40 () and 50 °C ().

Two kinetic models, namely pseudo-first-order and pseudo-second-order, were used to investigate the kinetics of fluoride adsorption on the nano-sandwich:

Pseudo-first-order model:

ln(q_e − q_t) = ln q_e − k₁t (5)

Pseudo-second-order model:

where q_t (mg g⁻¹) is the amount of adsorbed fluoride at various time t (min), and q_e (mg g⁻¹) is the amount at equilibrium time. The k₁ (mg g⁻¹ min⁻¹) and k₂ (mg g⁻¹ min⁻¹) are the first-order and second-order rate constants for adsorption, respectively.

The plot of t/q_t versus t at a fixed fluoride concentration, adsorbent dose, pH and temperature resulted in a straight line (Fig. 10b), which certified the applicability of the pseudo-second order model. In Table 1, the kinetic data had a better fit with data of the pseudo-second order model than that of the pseudo-first order model (Fig. S4†), which is explained by the fact that the values of correlation coefficient (R²) for the pseudo-second-order sorption model were higher than those obtained from the pseudo-first-order kinetics. Additionally, experimental values of q_e were in good agreement with the calculated ones by the pseudo-second order model. Thus, intra-particle diffusion contributes to the rate-determining step.⁴⁵

Table 1 Values of the pseudo-first order and pseudo-second order rate constants for fluoride adsorption on nano-sandwich at various temperatures

T/K	qe (exp, mg g^−1)	Pseudo-first-order rate constants			Pseudo-secound-order rate constants
		R^2	k1 (mg g^−1 min^−1)	qe (cal, mg g^−1)	R^2	k2 (mg g^−1 min^−1)	qe (cal, mg g^−1)
303	1.21	0.99	0.025	0.15	0.99	0.37	1.24
313	1.22	0.99	0.026	0.13	0.99	0.43	1.25
323	1.24	0.82	0.051	0.09	0.99	0.81	1.26

---

Adsorption isotherm

The experimental results of the equilibrium isotherm of fluoride adsorption fitted well with the Langmuir and Freundlich isotherm models. These models are represented as follows:where q_e is the equilibrium adsorption capacity (mg g⁻¹); C_e is the equilibrium fluoride concentration in solution (mg L⁻¹); q_m is the maximum fluoride adsorption capacity (mg g⁻¹); K_L is the Langmuir adsorption constant (L mg⁻¹). K_F (mg^1−(1/n) L^1/n g⁻¹) and 1/n are the Freundlich constants, related to the minimum adsorption capacity and adsorption intensity.

The results of fitting Freundlich and Langmuir equations to isotherm curves (Fig. S5†) are presented in Table 2. The estimated value of q_m for fluoride adsorption increased with the increase of temperature, which could be interpreted that the fluoride adsorption on nano-sandwich was an endothermic process.⁵⁰ The n value lies in the range of 1–10 confirming the favourable conditions for fluoride adsorption.⁵¹ By evaluating the correlation coefficients R², the fluoride sorption was well explained by the Langmuir isotherm model. It was implied that all adsorption sites were equally available and with a monolayer surface coverage without interaction between adsorbed species.

Table 2 Values of Langmuir and Freundlich constants for fluoride adsorption on nano-sandwich at different temperatures

T/K	Langmuir			Freundlich
	qm	b	R^2	KF	n	R^2
303	3.38	1.85	0.98	1.83	3.27	0.86
313	3.81	3.72	0.99	2.41	3.37	0.94
323	5.11	4.87	0.99	3.96	2.52	0.95

---

In order to estimate the efficiency of the as-synthesized adsorbent for the removal of fluoride, a comparison was performed between the Al₂O₃–Fe₃O₄–EG adsorbent and other previously reported adsorbents.^{22–25,48,52,53} An analytical comparison (Table 3) presents that when the equilibrium fluoride concentration is 1 mg L⁻¹, the adsorption capacities on the novel nano-sandwich are higher than those of the other aluminum composite sorbents. In addition, the sorption equilibrium could quickly reach over a relatively wide pH range of 2.0–10.0. Thus, it is clear that the nano-sandwich can be considered a promising adsorbent for the removal of fluoride from aqueous solutions.

Table 3 Experimental conditions and adsorption capacities of different adsorbents for the removal of fluoride

Adsorbent	Adsorbent dose (g L^−1)	Solution pH	Equilibrium time (h)	Equilibrium concentration (mg L^−1)	Sorption capacity (mg g^−1)	Ref.
Polyaniline/alumina	0.5	7	0.5	1	1.37	52
Activated aluminum	4	6.5–7	3	1	O.61	22
Manganese dioxide-coated activated alumina (MCAA)	8	7	3	1	0.15	23
Activated aluminum	5	7 ± 0.2	10	1	0.43	24
Aluminum oxide coated pumice	10	6–9	1	1	0.64	25
Chitosan based mesoporous Ti–Al binary metal oxide	4	3–9	24	1	0.22	53
Aluminum and iron oxides dispersed ceramic	20	6	48	1	0.43	48
Al2O3–Fe3O4–EG	2	2–10	2	1	2.19	This study

---

Stability and recoverability

In view of applications, stability and regeneration of the adsorbent is desirable in order to make it cost effective. The fluoride removal efficiency of the nano-sandwich sorbents after two cycles of successive adsorption–regeneration is demonstrated in Fig. 11. The adsorption efficiency decreases slightly from 96.8% to 91.4%, at an initial fluoride concentration of 5 mg L⁻¹, as some Al₂O₃, the active ingredients for fluoride removal, might be dissolved in NaOH solution. Good regeneration ability of this sorbent was the key to its efficiency and cost effectiveness. Fig. 12 indicates the nano-sandwich structure, consisting of three layers. The EG base layer possesses stable physical and chemical properties and delivers large surfaces for adsorption. The middle magnetic Fe₃O₄ layer renders magnetic separation possible. With a saturation magnetization of 3.7 emu g⁻¹ (Fig. 12b), the sorbents could be conveniently separated from the solution (Fig. 12a). The top Al₂O₃ layer was primarily responsible for fluoride adsorption. Due to the fact that Al³⁺ was easily combined with F⁻, the nano-sandwich adsorbent could effectively remove the fluoride in aqueous solution. Stable properties, high fluoride removal efficiency, and good magnetic separation ability makes Al₂O₃–Fe₃O₄–EG nano-sandwich an ideal sorbent for fluoride removal.

Fig. 11 Fluoride removal efficiency of the regenerated nano-sandwich.

Fig. 12 Schematic illustration of the structure of nano-sandwich and the reacting mechanism.

Conclusions

In this study, a promising nano-sandwich structure adsorbent was successfully prepared. It contained three layers of a base layer of EG, a separation layer of Fe₃O₄, and an adsorption functional layer of Al₂O₃. With the parameters of the adsorption experiment (C₀ = 5 mg L⁻¹, pH = 4.0, T = 30 °C, t = 120 min and m = 0.2 g), the fluoride removal efficiency peaked at 96.8%. The equilibrium time data follows the pseudo-second order model, indicating intra-particle diffusion as the rate determining step for fluoride adsorption. The sorbents, after two cycles of regeneration, revealed good fluoride adsorption efficiency of 91.4%. The prepared nano-sandwich showed great potential in water defluoridation due to its stable properties, excellent fluoride adsorption capacity, efficient magnetic separation characteristics and good recyclability.

Acknowledgements

The work was carried out under National Key Research and Development Program of China (No. 2016YFA0201001), National Natural Science Foundation of China (No. 11627801), Natural Science Foundation of Hubei Province (No. 2013CFB412), the financial of China (No. NSFC51102218), the grant of the Opening Project of Zhejiang Research Center of Non-metallic Mineral Engineering Technology (Zhejiang Institute of Geology and Mineral Resource) (ZD2015k02), CUG201510491027 and CUG201610491211.

Notes and references

1. S. Suriyaraj and R. Selvakumar, RSC Adv., 2016, 6, 10565–10583.
2. A. Bansiwal, D. Thakre, N. Labhshetwar, S. Meshram and S. Rayalu, Colloids Surf., B, 2009, 74, 216–224.
3. S. Mandal and S. Mayadevi, Appl. Clay Sci., 2008, 40, 54–62.
4. Q. Guo and E. J. Reardon, Appl. Clay Sci., 2012, 56, 7–15.
5. E. J. Reardon and Y. X. Wang, Environ. Sci. Technol., 2000, 34, 3247–3253.
6. Meenakshi and R. C. Maheshwari, J. Hazard. Mater., 2006, 137, 456–463.
7. A. K. Yadav, C. P. Kaushik, A. K. Haritash, A. Kansal and N. Rani, J. Hazard. Mater., 2006, 128, 289–293.
8. C. Yang, L. Gao, Y. Wang, X. Tian and S. Komarneni, Microporous Mesoporous Mater., 2014, 197, 156–163.
9. P. I. Ndiayea, P. Moulin, L. Dominguez, J. C. Millet and F. Charbit, Desalination, 2005, 173, 25–32.
10. C. Y. Hu, S. L. Lo, W. H. Kuan and Y. D. Lee, Water Res., 2005, 39, 895–901.
11. Y. H. Li, S. G. Wang, X. F. Zhang, J. Q. Wei, C. L. Xu, Z. K. Luan and D. H. Wu, Mater. Res. Bull., 2003, 38, 469–476.
12. N. C. Lu and J. C. Liu, Sep. Purif. Technol., 2010, 74, 329–335.
13. G. Singh, B. Kumar, P. K. Sen and J. Majumdar, Water Environ. Res., 1999, 71, 36–42.
14. S. Lahnid, M. Tahaikt, K. Elaroui, I. Idrissi, M. Hafsi, I. Laaziz, Z. Amor, F. Tiyal and A. Elmidaoui, Desalination, 2008, 230, 213–219.
15. H. Wijnja and C. P. Schulthess, Spectrochim. Acta, Part A, 1999, 55, 861–872.
16. W. Li, R. Harrington, Y. Z. Tang, J. D. Kubicki, M. Aryanpour, R. J. Reeder, J. B. Parise and B. L. Phillips, Environ. Sci. Technol., 2011, 45, 9687–9692.
17. E. Kumar, A. Bhatnagar, U. Kumar and M. Sillanpaa, J. Hazard. Mater., 2011, 186, 1042–1049.
18. D. Peak, J. Colloid Interface Sci., 2006, 303, 337–345.
19. E. J. Elzinga, Y. Z. Tang, J. McDonald, S. De Sisto and R. J. Reeder, J. Colloid Interface Sci., 2009, 340, 153–159.
20. E. Kumar, A. Bhatnagar, W. Hogland, M. Marques and M. Sillanpaa, Chem. Eng. J., 2014, 241, 443–456.
21. B. Kasprzyk-Hordern, Adv. Colloid Interface Sci., 2004, 110, 19–48.
22. S. M. Maliyekkal, S. Shukla, L. Philip and I. M. Nambi, Chem. Eng. J., 2008, 140, 183–192.
23. S. S. Tripathy and A. M. Raichur, J. Hazard. Mater., 2008, 153, 1043–1051.
24. S. M. Maliyekkal, A. K. Sharma and L. Philip, Water Res., 2006, 40, 3497–3506.
25. A. Salifu, B. Petrusevski, K. Ghebremichael, L. Modestus, R. Buamah, C. Aubry and G. Amy, Chem. Eng. J., 2013, 228, 63–74.
26. M. S. Onyango, Y. Kojima, O. Aoyi, E. C. Bernardo and H. Matsuda, J. Colloid Interface Sci., 2004, 279, 341–350.
27. T. Sen, A. Sebastianelli and I. J. Bruce, J. Am. Chem. Soc., 2006, 128, 7130–7131.
28. T. Poursaberi, M. Hassanisadi, K. Torkestani and M. Zare, Chem. Eng. J., 2012, 189, 117–125.
29. H. Jin, J. Yuan, H. Hao, Z. Ji, M. Liu and S. Hou, Mater. Lett., 2013, 110, 69–72.
30. Y. W. Zhu, S. Murali, W. W. Cai, X. S. Li, J. W. Suk, J. R. Potts and R. S. Ruoff, Adv. Mater., 2010, 22, 5226.
31. Y. B. Liu, L. Yu, C. N. Ong and J. P. Xie, Nano Res., 2016, 9, 1983–1993.
32. Y. B. Liu, H. Liu, Z. Zhou, T. R. Wang, C. N. Ong and C. D. Vecitis, Environ. Sci. Technol., 2015, 49, 7974–7980.
33. M. F. Zhao and P. Liu, Desalination, 2009, 249, 331–336.
34. T. Yao, Y. Zhang, Y. Xiao, P. Zhao, L. Guo, H. Yang and F. Li, J. Mol. Liq., 2016, 218, 611–614.
35. M. Yang, Y. H. Zhao, X. Z. Sun, X. T. Shao and D. X. Li, Desalin. Water Treat., 2014, 52, 283–292.
36. R. Nekram and J. Urmila, Res. J. Chem. Environ., 2013, 17, 43–47.
37. H. Y. Jin, Z. J. Ji, J. Yuan, J. Li, M. Liu, C. H. Xu, J. Dong, P. Hou and S. Hou, J. Alloys Compd., 2015, 620, 361–367.
38. L. Yu, Y. Ma, C. N. Ong, J. P. Xie and Y. B. Liu, RSC Adv., 2015, 5, 64983–64990.
39. F. Riahi, M. Bagherzadeh and Z. Hadizadeh, RSC Adv., 2015, 5, 72058–72068.
40. L. Liu, Z. Cui, Q. Ma, W. Cui and X. Zhang, RSC Adv., 2016, 6, 10783–10791.
41. A. Mohseni-Bandpi, B. Kakavandi, R. R. Kalantary, A. Azari and A. Keramati, RSC Adv., 2015, 5, 73279–73289.
42. Z. Xie, Z. He, X. Feng, W. Xu, X. Cui, J. Zhang, C. Yan, M. A. Carreon, Z. Liu and Y. Wang, ACS Appl. Mater. Interfaces, 2016, 8, 10324–10333.
43. W. Zheng and S. C. Wong, Compos. Sci. Technol., 2003, 63, 225–235.
44. P. G. Wu, J. H. Zhu and Z. H. Xu, Adv. Funct. Mater., 2004, 14, 345–351.
45. M. G. Sujana and S. Anand, Appl. Surf. Sci., 2010, 256, 6956–6962.
46. A. Raichur and M. J. Basu, Sep. Purif. Technol., 2001, 24, 121–127.
47. X.-H. Wang, R.-H. Song, H.-C. Yang, Y.-J. Shi, G.-B. Dang, S. Yang, Y. Zhao, X.-F. Sun and S.-G. Wang, Bioresour. Technol., 2013, 127, 106–111.
48. N. Chen, Z. Zhang, C. Feng, D. Zhu, Y. Yang and N. Sugiura, J. Hazard. Mater., 2011, 186, 863–868.
49. L. Y. Chai, Y. Y. Wang, N. Zhao, W. C. Yang and X. Y. You, Water Res., 2013, 47, 4040–4049.
50. R. Chen, Z. Zhang, C. Feng, Z. Lei, Y. Li, M. Li, K. Shimizu and N. Sugiura, Appl. Surf. Sci., 2010, 256, 2961–2967.
51. N. Viswanathan and S. Meenakshi, J. Colloid Interface Sci., 2008, 322, 375–383.
52. M. Karthikeyan, K. S. Kumar and K. Elango, J. Fluorine Chem., 2009, 130, 894–901.
53. D. Thakre, S. Jagtap, N. Sakhare, N. Labhsetwar, S. Meshram and S. Rayalu, Chem. Eng. J., 2010, 158, 315–324.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra19390k

---