Performance and mass transfer of aqueous fluoride removal by a magnetic alumina aerogel

Abstract

Magnetic alumina aerogel (MAA) was successfully synthesized and evaluated for F⁻ removal from water. The adsorbent was characterized by field emission-scanning electron microscopy (FE-SEM), energy dispersive X-ray spectroscopy (EDX), X-ray diffraction (XRD), vibration sample magnetometry (VSM), X-ray photoelectron spectroscopy (XPS), FTIR, and Raman spectroscopy. The magnetic alumina aerogel sample has a BET surface area of 215.1 m² g⁻¹ and meso- to macro-sized pores ranging from 50 nm to 200 nm. A saturation magnetization of 19.8 emu g⁻¹ was observed, which makes the separation of the adsorbent realizable after batch adsorption. Boehmite and magnetite phases were identified in the adsorbent. A core–shell structure is favored and additional studies are suggested. Aggregated aerogel nanoparticles composed of boehmite are the inner portion. Magnetite nanoparticles discretely and partly cover the alumina by leaving portions of alumina's surface exposed to the ambient environment. This adsorbent has a moderately high adsorption capacity of 32.1 mg F⁻ per g adsorbent at pH 5.0. Batch studies revealed that fluoride adsorption followed the pseudo-second-order kinetics model. The intraparticle mass diffusion was fitted well using a homogeneous surface diffusion model (HSDM), and the intraparticle surface diffusion coefficients (D_s) were numerically determined. The evolution of dimensionless radial F⁻ concentration profiles inside particles was also simulated. General co-existing anions did not inhibit F⁻ uptake by MAA except for a slight inhibition by HCO₃⁻ and PO₄³⁻. Spiking experiments demonstrated that MAA effectively removed F⁻ when treating simulated well water contained 1.83 mg L⁻¹ F⁻. Al 2p and Fe 2p XPS spectra of MAA before and after F⁻ removal demonstrated that both the alumina and iron oxide phases contributed to F⁻ surface adsorption. The adsorption performance and easy separation of MAA show this adsorbent is a promising candidate for fluoride removal from water.

---

1. Introduction

Drinking water is often the main source of fluoride intake by humans in contaminated areas where high fluoride groundwater or surface water concentrations occur.¹ More than 200 million people worldwide rely on drinking water with fluoride concentrations exceeding the 1.5 mg L⁻¹ World Health Organization (WHO) guideline.² There are 200 million people in 27 nations across the globe facing issues of excess fluoride in drinking water.³ China and India are the most affected countries with nearly 35 and 26 million people are at fluoride risk, respectively.⁴

Effective mitigation methods for removing fluoride from drinking water include precipitation/coagulation,⁵ adsorption,^6,7 ion exchange,⁸ reverse osmosis,⁹ and electrodialysis.^4,10,11 Fluoride adsorption is attractive because of its effectiveness, simplicity, flexibility and moderate cost.¹² It is suitable for polishing the fluoride concentration in effluents to meet the regulation criteria and for suppling fluoride-free water to small communities and scattered rural households.^11,13 Activated alumina is the most widely commercialized and used adsorbent for fluoride removal over the past two decades. It has been intensively tested in both lab and field applications. Granular activated alumina fixed beds represent the optional available scheme for aqueous fluoride mitigation. Nevertheless it has several disadvantages including low adsorption capacity, poor mechanical strength and mass loss from fixed beds over long operation periods.¹⁴ Therefore, frequent bed regeneration or replacement is required.

Alumina's intrinsic ability to bind F⁻ has spurred considerable past and present work to develop novel alumina-based materials with good performances. Crystalline and amorphous mesoporous alumina,^15–17 alumina nanotubes,¹⁸ alumina nanofibers,¹⁹ alumina foams,²⁰ alumina embedded composites (carbon nanotube, graphite, polymer, etc.)²¹ and alumina aerogels²² etc. were recently reported for hazardous ions including F⁻ removal. Among these, alumina aerogels have attracted more recent attention based on its high F⁻ binding affinity, unique aerogel properties, and acceptable cost.

Barriers still exist for field applications of novel alumina aerogel adsorbents. These adsorbents are usually prepared as powders, fine particles, fresh precipitates, flocs or even gels.²³ Traditional filtration and sedimentation separation methods are unreliable or cost ineffective for these adsorbents. Turbidity and released metals must be eliminated from water after their use. Solutions for these issues include (1) coating, loading, impregnation and entrapment of the active adsorbents in or on carriers, (2) directly granulating the adsorbents with binders, or (3) using a magnetic separation during operation.²⁴ Magnetic separation allows short adsorbent contact time in stirred slurries and high separation efficiency.

In this study, alumina aerogel was amended with magnetite to become a magnetic hybrid adsorbent. It then served as a novel adsorbent for aqueous fluoride removal. The fluoride adsorption potential of this magnetic aerogel adsorbent was systematically evaluated under various initial F⁻ concentrations, solution pHs, reaction times and co-existing potential interfering adsorbates. In addition, defluoridation performance was evaluated using simulated F⁻-containing groundwater. Finally, the F⁻ binding mechanism to the adsorbent was explored using XPS, FTIR and Raman spectroscopic measurements.

2. Materials and methods

2.1. Materials

All chemicals were analytical grade reagents. The F⁻ stock solution was prepared with deionized water using NaF. F⁻-bearing solutions were freshly prepared by diluting a F⁻ stock solution with distilled water.

2.2. Adsorbent preparation

Alumina aerogel was synthesized via a modified procedure.²⁵ AlCl₃·6H₂O (4.35 g) was dissolved in a mixture of 6.21 mL absolute ethanol and 4.00 mL deionized water until it completely dissolved. Then 7 mL of propylene oxide was quickly added into the solution under continuous stirring for 1 min. The transparent solution turned to gel in approximately 20 s after adding propylene oxide. The resultant homogeneous solution was sealed and maintained for 5 min at 60 °C, which led to further gelation. After gelling, the wet gel was soaked into a five-fold volume of isopropanol and aged for 48 h at 60 °C. The gel was subsequently dried by evaporation at 60 °C for 48 h under ambient pressure. After drying, this product was ground into powder for further use. The specific surface area of alumina aerogel is measured to be 320.3 m² g⁻¹.

The magnetic alumina aerogel was prepared following a modified procedure by Kassaee et al.²⁶ (NH₄)₂Fe(SO₄)₂·6H₂O (4.33 mmol) and NH₄Fe(SO₄)₂·12H₂O (8.66 mmol) were dissolved in 200 mL deionized water under vigorous stirring. Then alumina aerogel (1 g) was added into the stirred solution until it was highly dispersed. At room temperature, aqueous ammonia (25%) was added into the dispersion quickly until its pH reached 10 under mechanical agitation. A black precipitate was formed. Then the temperature was raised to 50 °C and stirring was continued for another 30 min. The precipitate was separated and washed with deionized water for 2 days, and then washed ten times with ethanol. After that, the precipitate was dried at 100 °C, and then ground for further characterization and F⁻ adsorption evaluation. The final yield of the adsorbent was ca. 6.0 g.

2.3. Adsorbent characterization

The specific surface area, pore volume and pore size distribution of the pristine alumina aerogel and magnetic alumina aerogel were determined by the N₂ adsorption desorption isotherms collected on a Tristar II 3020 surface area analyzer (Micromeritics, USA). The pore size distributions of the samples were derived from the adsorption branches of the isotherms based on the Barrett–Joyner–Halenda (BJH) model.^27–29 XRD patterns of the samples were characterized on an XRD-7000 (Shimadzu, Japan) using Cu K_α radiation source. Samples were scanned at a speed of 2° min⁻¹ from 10° to 80° operating at 40 kV and 40 mA. The bulk density was calculated by weighing samples with a measured volume. Magnetic properties of the magnetic alumina aerogel were measured using a VSM 7307 vibrating sample magnetometer (LakeShore, USA). Surface morphologies were observed using a JEOL-2011 high resolution transmission electron microscope (JEOL, Japan). Thermogravimetry (TG) and differential scanning calorimetry (DSC) analyses were performed on a STA 449 F1 simultaneous thermal analyzer (Netzsch, Germany). XPS spectra were measured on an Axis Ultra photoelectron spectrometer (Kratos, UK). An Al-Kα anode radiation source was used for excitation. The XPS results were calibrated by using a C 1s energy of 284.6 eV. XPS data processing and peak fitting were performed using a nonlinear least-squares fitting program (XPSPeak software 4.1, R. W. M. Kwok). Fourier transform infrared (FTIR) spectra of magnetic alumina aerogel, before and after adsorption, were recorded at room temperature on a Spectrum GX spectrophotometer (Perkin Elmer, USA) at a resolution of 2 cm⁻¹.

2.4. Batch adsorption experiments

The adsorption isotherms were carried out by varying the initial concentrations (10–110 mg L⁻¹) of fluoride with a fixed adsorbent dose (0.3 g L⁻¹) in 250 mL polyethylene bottles. The total volumes were 100 mL. The suspensions were shaken at 25 ± 1 °C and a 130 rpm shaking speed. Sample pH was maintained at 5.0 ± 0.2 by manual adjustment with 0.01 mol L⁻¹ HCl and NaOH every 2 h. The effect of pH on fluoride removal was studied using 250 mL polyethylene bottles with a total volume of 100 mL. The initial fluoride concentration was 42.2 mg L⁻¹ and the adsorbent dose was 0.3 g L⁻¹ at 25 ± 1 °C. The solution pH was manually adjusted at 2 h intervals and maintained in the range 3 to 10. After shaking for 24 h, the residual F⁻ concentration was determined with a fluoride-selective electrode connected to an ion meter (Metrohm 809, Swiss).

The fluoride stock solution and deionized water were added into two 2000 mL high density polyethylene (HDPE) bottles, to generate initial F⁻ concentrations of 32.4 mg L⁻¹ and 55.7 mg L⁻¹, respectively, with total volumes of 1000 mL. Then, a 0.3 g L⁻¹ does of adsorbent was added. The suspension's pH was manually adjusted and maintained at 5.0 ± 0.2 throughout the experiment at 2 h intervals. Mixtures were swirled at 130 rpm and maintained at 25 ± 1 °C. Approximately 4 mL aliquots were taken from the suspension at predetermined intervals. The samples were immediately filtered through a 0.45 μm membrane and analyzed.

The effects of coexisting anions (Cl⁻, SO₄²⁻, HCO₃⁻, NO₃⁻, PO₄³⁻, SiO₄⁴⁻, NO₂⁻) on fluoride adsorption were also investigated. The initial F⁻ concentration was fixed at 5 mg L⁻¹ with a total volume of 100 mL, and the adsorbent dosage was 0.3 g L⁻¹. The concentrations of spiked anions are provided in Table S1.† The pH of the mixtures was manually adjusted and maintained at 5.0 ± 0.2 during 24 h of shaking.

3. Results and discussion

3.1. Characterization of the aerogel adsorbents

The magnetic alumina aerogel adsorbent (MAA) presents puffy cotton-like shapes in its SEM image (Fig. 1(a)). MAA's TEM image (Fig. 1(b)) also exhibits randomly interconnected networks made up of nanometer-sized fibrous flakes. The former are similar to the previously reported shapes of alumina aerogel plus clustered nanoparticles.³⁰ Irregular sphere-shaped, aggregated ∼10 nm diameter magnetite nanoparticles cover the fibrous flaked alumina aerogel.

Fig. 1 (a) SEM and (b) TEM images of the magnetic alumina aerogel samples.

The XRD pattern of MAA (Fig. 2) exhibits two sets of characteristic peaks. Peaks at 28.18° and 49.16° are due to boehmite diffractions (PDF no. 83-2384). Peaks at 18.26°, 30.27°, 35.53°, 43.38°, 57.47° and 63.05° are diffractions of magnetite (PDF no. 02-1035). The magnetic hysteresis loop of MAA (Fig. 3) displays a maximum saturated magnetization of 19.9 emu g⁻¹.

Fig. 2 XRD pattern of the magnetic alumina aerogel sample.

Fig. 3 Magnetization curve of the magnetic alumina aerogel sample and a demonstration of magnetic separation.

The N₂ adsorption–desorption isotherm is presented in Fig. 4. The BET surface areas, pore properties and elemental atomic percentages for pristine magnetite, aerogel alumina and magnetic aerogel alumina samples and F⁻-adsorbed magnetic aerogel alumina samples before and after 3 h-grinding are provided in Tables S2† and 3. This MAA adsorbent has a high BET surface area (215.1 m² g⁻¹) and meso-sized pores ranging from 2 nm to 20 nm. The surface area and pores result from the combined boehmite aerogel (320.3 m² g⁻¹) and magnetite (89.9 m² g⁻¹) deposited onto it. This combination inevitably decreases the BET surface area of the non-magnetic alumina aerogel from 320.3 m² g⁻¹ to 215.1 m² g⁻¹. The pore volume also dropped from 1.0 mL g⁻¹ to 0.4 mL g⁻¹, which means that magnetite nanoparticles might deposit on the surface of the alumina, or a portion formed within the alumina pores. The change is a trade-off between losing partial surface and pore properties and getting magnetic separation ability.

Fig. 4 N₂ adsorption–desorption isotherms and pore size distributions (inset plots) for magnetic alumina aerogel.

The results in Tables S2 and S3† showed that (1) enlarged surface areas for all samples after grinding, (2) the existence of both Al and Fe elements on the surface, and (3) F in MAA slightly decreased after grinding. That means both alumina aerogel and magnetite phases are present at the surface. A core–shell structure is favored for the MMA adsorbent as illustrated in Fig. S1.† MAA was prepared through two steps including aerogel alumina synthesis and magnetic coating by magnetite. The pore average diameters of the alumina aerogel and magnetite are too close to each other to show any difference (Table S2†). However, the pore average diameter of magnetic alumina aerogel drops a lot in comparison to alumina aerogel, which is consistent with the formation of an iron oxide coating or core/shell system. The surface area of the magnetic alumina aerogel only goes up modestly after grinding, which is consistent with a core/shell where the shell is incomplete. Overall, these results seem to be consistent with, but not definitive for the core/shell structure. Additional studies like depth-profiling measurements using Ar⁺ sputtering XPS or electron energy loss spectroscopy (EELS) would be useful to confirm this supposition. Further, less fluoride was observed after grinding because the majority of the bound fluorine was originally adsorbed near the surface for both phases and especially so for the magnetite.

The TG and DSC curves of MAA (Fig. 5) were divided into three main regions (1) 0–200 °C; (2) 200–450 °C and (3) 450–1200 °C. The weak endothermic peak around 200 °C in the first region is attributed to the evolution of physically adsorbed water accompanied by a weight loss of 3.40%. The second region includes the thermal conversion of boehmite (γ-AlO(OH)) to corundum (Al₂O₃) occurring with a large endotherm which peaks at a wide range of 200–450 °C. This is accompanied by a much larger 11.7% calculated weight loss. The phase transition can be described by the chemical eqn (1) as follows:

2AlO(OH) → γ-Al₂O₃ + H₂O (1)

Fig. 5 Thermogravimetric and differential scanning calorimetry (TG-DSC) curves of the magnetic alumina aerogel sample under flowing argon.

The theoretical weight loss is 15%. This is higher than the observed value (11.7%), which indicates an incomplete conversion. The third thermal decomposition region appears in the vicinity from 450 °C to 1200 °C accompanied by a small weight loss (3.39%). It accounts for the pyrolysis and final decomposition of the residual organics in the adsorbent. These organics were introduced during synthesis.

3.2. Batch adsorption experiments

3.2.1. Adsorption isotherms. The effect of adsorbent dose on F⁻ adsorption (Fig. S2†) was evaluated first. The results definitively showed that the removal performance increased with increasing adsorbent doses, since more active sites were available. The adsorption capacity dropped a lot when the dosage was less than 0.3 g L⁻¹, but not much above this dosage. So a dosage of 0.3 g L⁻¹ was selected as the optimum parameter to minimize the effect of adsorbent dose on F⁻ adsorption. This was used in all further experiments.

The F⁻ adsorption isotherm on MAA (Fig. 6) was fitted using the Langmuir, Freundlich and Sip isotherm models. These results are shown in Table 1. The Freundlich model fits the experimental data reasonably well, yielding a correlation coefficient (R²) above 0.994. This Freundlich model fit is presented in Fig. 6. The good fit suggests that F⁻ uptake by MAA may not be restricted to monolayer adsorption. Second-layer or multilayer adsorption may have occurred. This adsorption occurred on heterogeneous and “amorphous” surfaces of both boehmite and magnetite, which have different adsorption energies. Heterogeneous and “amorphous” surfaces have more available surface sites than homogenous and “crystalline” surfaces at high equilibrium concentrations. Thus, MAA is a relatively good adsorbent with potential for aqueous F⁻ mitigation when compared with other previously reported alumina or magnetic adsorbents (Table 2).

Fig. 6 The F⁻ adsorption isotherm onto magnetic alumina aerogel. Initial F⁻ concentrations, 10–110 mg L⁻¹; adsorbent dose, 0.3 g L⁻¹; total solution volumes, 100 mL; pH, 5.0 ± 0.1; temperature, 25 ± 1 °C, shaking time, 24 h.

Table 1 Adsorption isotherm parameters for F⁻ adsorption on magnetic alumina aerogel (MAA)

Equations	Parameter
Langmuir	qmax (mg g^−1)	30.25
	b (L mg^−1)	0.106
	RL^2	0.968
Sip isotherm model	a (L g^−1)	8.824
	b	0.279
	c (L mg^−1)	0.014
	Rs^2	0.993
Freundlich qe = kCe^1/n	k ((mg L^−1) (L mg^−1)^1/n)	8.69
	n	3.704
	RF^2	0.994

---

Table 2 Comparison of the magnetic alumina aerogel with reported alumina adsorbents in removal of F⁻ from water

Adsorbent	Adsorbent state and size	Concentration range (mg L^−1)	Dose g L^−1	Temperature (°C)	pH	Eq. time (h)	Capacity (mg g^−1)	Reference
Magnetic alumina aerogel	Powder	10–110	0.3	25	5.0	24	32.1	Present study
Hydrous ferric oxide doped calcium alginate beads	Powder	5–35	0.5	42	7.0	4	8.9	24
MgO-coated magnetite	Powder	5.6–25	2.5	25	6.0	2	11.0	25
Fe–Ti oxide nanoadsorbent	Powder	50	0.5	25	6.9	12	47.0	26
Fe–Al–Ce nanoadsorbent	Powder	42	0.5	25	7.0	36	2.8	27
Mesoporous alumina with improved properties	Powder	5–55	4.0	25	7.0	5	8.3	28
Fe(III) carboxylated chitosan	Fine powder	3–35	4.0	25	6.0	6	4.2	29
Cellulose coated hydroxyapatite	Powder	1–10	2.5	25	6.5	6	3.4	30
Aluminum-supported carbon nanotubes	Fine powder	0–50	2.0	25	7.0	—	27.0	31
Magnetite lanthanum oxide	Powder	1–30	2.0	25	7.0	24	0.5	32

---

3.2.2. Kinetics of F⁻ uptake by MAA. Fig. 7(a) shows the time dependence of fluoride uptake by MAA at two initial F⁻ concentrations (32.4 and 55.7 mg L⁻¹). Fluoride kinetic adsorption data onto MAA was fitted using pseudo-first-order³¹ and pseudo-second-order rate models.³² These two models are expressed in eqn (2) and (3), respectively.where q_e and q_t are the amounts of fluoride adsorbed at equilibrium and at any time t (mg g⁻¹ solid material), k₁ (min⁻¹) and k₂ (g mg⁻¹ min⁻¹) are the rate constants for pseudo first- and second-order sorption respectively, and t is the shaking time (min).

Fig. 7 Effect of contact time on F⁻ adsorption on and within the magnetic alumina aerogel adsorbent (MAA). Initial F⁻ concentrations, 32.4 mg L⁻¹ and 55.7 mg L⁻¹; adsorbent dose, 0.3 g L⁻¹; total solution volumes, 1 L; pH 5.0 ± 1; temperature, 25 ± 1 °C.

The linear and nonlinear pseudo-first-order plots were presented in Fig. S3† and 7(a), respectively. The calculated rate constants and related parameters are listed in Tables S4 and 3. The large value of the correlation coefficient (R² > 0.985) of the pseudo-second-order fit favors this kinetics model. However, this equation is only a generalized model describing the decline profile for residual F⁻ concentration in solution during adsorption. Further details about adsorbate mass transfer inside pores and on the particles' inner-surfaces are not revealed. However, it is important to understand pollutant uptake behavior into porous aerogel-based adsorbents.

Table 3 Adsorption kinetic parameters for aqueous F⁻ adsorption on the magnetic alumina aerogel (MAA)

Concentration of F^− (mg L^−1)	Pseudo-first order			Pseudo-second order
	k1 (min^−1)	qe (mg g^−1)	R1^2	k2 (g (mg^−1 min^−1))	qe (mg g^−1)	R2^2
32.4	0.490	12.6	0.898	0.045	14.4	0.949
55.7	1.570	33.9	0.932	0.060	37.1	0.985

---

F⁻ mass transfer inside porous materials was therefore described using the homogeneous surface diffusion model (HSDM).^33,34 This model uses three simultaneous steps including: (i) external (or film) mass transfer of solute molecules from the bulk solution to the adsorbent particle surface; (ii) surface diffusion along the particle inner surface; (iii) sorbate adsorption on the internal surface sites. In this study, external film barrier to mass transfer is negligible in comparison to surface diffusion resistance in a well-stirred or shaken batch reactor.³⁵ Thus, the model simplified to a Fick's second law-controlled diffusion model. In a 1-D spherical coordinate system, the homogeneous surface diffusion model is described by eqn (4).^36,37

The initial and boundary conditions are given as,

q = 0, t = 0 (6)

q = q_e, r = R_p, t = infinity (7)

The average adsorbate concentration inside the adsorbent is,

The partial differential eqn (4) was solved numerically using the FlexPDE 6.37 student version software.³⁸

The HSDM fittings are shown in Fig. 7(a)–(c) and Table 4. The best fit curves are obtained by minimizing an objective function, the smaller root mean-square error (RMSE) value between the experimental data and model estimations. This equation is expressed as,

where C_A,exp and C_A,model are the experimental and model predictions of F⁻ concentrations, respectively, obtained in solution.

Table 4 The surface diffusion parameter (D_s) in the HSDM model for aqueous F⁻ adsorption on the magnetic alumina aerogel (MAA)

C0 (mg L^−1)	m (g)	Ce (mg L^−1)	qe (mg g^−1)	Ds (cm^2 s^−1)	RMSE
32.4	0.3	26.8	14.47	0.39 × 10^−11	1.411
55.7	0.3	43.4	37.11	0.1 × 10^−10	2.159

---

The low RMSE values indicated that the HSDM model fit to the kinetic datasets reasonably well (Table 4). The mass transfer parameter in the HSDM model (D_s) increased from 0.39 × 10⁻¹¹ cm² s⁻¹ to 0.10 × 10⁻¹⁰ cm² s⁻¹ as the initial F⁻ concentration rose from 32.4 mg L⁻¹ to 55.7 mg L⁻¹. This D_s enlargement at higher F⁻ concentrations was due to faster surface diffusion driven by the larger concentration gradient from the outer-surface to the spherical core.

Fig. 7(b) and (c) present the evolution of F⁻ concentration front profiles inside adsorbent particles versus time. The F⁻ diffusion dynamics can be compared at different times and different radius positions inside the particle for each initial F⁻ concentration from the outer-surface to the spherical center along the normal direction. At the higher initial F⁻ concentration (55.7 mg L⁻¹), the diffusion of F⁻ within particles was significantly faster than at the lower initial F⁻ concentration (32.4 mg L⁻¹). This was caused by the concentration-induced linear driving forces (LDF) mentioned above.

3.2.3. Effect of pH on fluoride removal. Fig. S4† shows the influence of pH on fluoride removal by MAA. The adsorbent's capacity decreased with increasing pH, especially when the solution pH increased above 4.0. This pH dependent removal efficiency presented a typical anion adsorption “edge” curve.⁶ It showed a pattern similar to the trends of the zeta potentials. The pH_pzc of this adsorbent is ca. 7.5. The pH_pzc of magnetite and boehmite were reported to be 6.3–8.5 and 8.5–8.7,^39–42 respectively. Each component contributes to the pH_pzc value of this composite adsorbent. Hence, its surface will be positively charged when the solution pH is below pH = 7.5 and vice versa. HF has a pK_a value of 3.14 and F⁻ is dominant when solution pH is above 4.5 (Fig. S5†). The reduction of removal efficiency at alkaline pH values (>7.5) probably is due to the increased electrostatic repulsion forces between the negatively charged surface and F⁻ anion. When solution pH is below 7.5 and decreases further, the adsorbent surface will become increasingly positive, F⁻ gradually turns to HF (aq), and HF (aq) becomes dominant. Thus, the electrostatic repulsion forces will decline from both of these changes, which in turn enhances the adsorption of F⁻ at low pH values.

The zeta potentials of MAA in the absence and presence of 25 mg L⁻¹ and 50 mg L⁻¹ F⁻ as a function of pH were shown in Fig. 8. The pH_pzc curve with higher initial F⁻ concentration shifted to lower pH values. Shifts in the pH_pzc and reversal trends of the zeta potentials with increasing ion concentrations are considered an evidence of strong specific ion adsorption and inner-sphere surface complex formation.⁴³ Therefore, it was assumed that inner-sphere surface complexes formed on MAA after F⁻ adsorption. In the context of this paper, an inner-sphere complex would mean the formation of Al–F and/or Fe–F bonds at adsorption sites.

Fig. 8 Zeta potentials of the magnetic alumina aerogel (MAA) adsorbent in different pH solutions with an adsorbent dosage at initial F⁻ concentrations of 25.0 mg L⁻¹ and 50.0 mg L⁻¹, adsorbent dose, 0.03 g L⁻¹, temperature, 25 ± 1 °C, and shaking time, 24 h.

3.2.4. Effects of coexisting substances. Natural groundwater always contains numerous aqueous compositions, some of which can compete for sorption sites and decrease the removal efficiency of an adsorbent. F⁻ adsorption on MAA in the presence of common co-existing ions was investigated at pH 5 (Fig. 9). Anions including SO₄²⁻, SiO₄⁴⁻, NO₃⁻, NO₂⁻ and Cl⁻ did not cause inhibition of F⁻ adsorption by MAA while HCO₃⁻ and PO₄³⁻ caused a very small inhibition. These two anions have consequential binding affinities to Al–OH and Fe–OH surface sites. Anions like SO₄²⁻, SiO₄²⁻ and NO₃⁻ at 200 mg L⁻¹ even caused small enhancements rather than inhibition of fluoride adsorption. This was promoted by the increased ionic strength of the medium.

Fig. 9 Effect of co-existing anions on F⁻ adsorption on the magnetic alumina aerogel (MAA) adsorbent. Initial F⁻ concentrations, 5.0 mg L⁻¹; adsorbent dose, 0.3 g L⁻¹; total solution volumes, 100 mL; pH 5.0 ± 0.1; temperature, 25 ± 1 °C, and shaking time, 24 h (the dashed lines correspond to the removal efficiencies of the adsorbent without the co-existing anions present).

3.3. Treatment of simulated well water

F⁻-contaminated well water usually contains multiple anion and cation components, which make F⁻ removal more complicated in field applications than in lab testing using deionized water. F⁻ mitigation from simulated well water using MAA was evaluated and the results are presented in Fig. 10. The properties of the simulated well water are summarized in Table S5.† This water contains 1.83 mg L⁻¹ of F⁻ and co-existing ions like AsO₄³⁻, SO₄²⁻, Fe³⁺, Al³⁺, Na⁺ and K⁺ (Table S5†). Within 5 min, the residual F⁻ in solution was lowered to well below the WHO guideline of 1.5 mg L⁻¹, with an MMA does of 1 g L⁻¹. These results clearly demonstrate that this MAA adsorbent is effective for F⁻ removal when treating this simulated well water, with promise for F⁻ mitigation in real applications.

Fig. 10 Removal of F⁻ from spiked well water by the magnetic alumina aerogel (MAA) adsorbent. Initial F⁻ concentrations, 1.83 mg L⁻¹; adsorbent dose, 1 g L⁻¹; total solution volume, 1000 mL; pH, 7.57; temperature, 25 ± 1 °C, and shaking time, 0–30 h (see Table S2† for As(V), Cl⁻, Fe³⁺, SO₄²⁻, Na⁺, K⁺, and Al³⁺ concentrations).

The magnetic adsorbent can be operated in a magnetic separation-enhanced sequencing batch (MSES) mode including four steps. These include (1) influent fill (Fig. S6(a)†), (2) adsorption (Fig. S6(b)†), (3) magnetic separation (Fig. S6(c)†) and (4) treated water withdrawl (Fig. S6(d)†). This mode fully utilizes and benefits from the magnetic separation property and the fast kinetics of fine MAA powders.

3.4. Characterization the adsorbent before and after F⁻ adsorption

3.4.1. XPS characterization. The full scan, F 1s, Al 2p and Fe 2p XPS spectra of MAA, both before and after 40 mg L⁻¹ of F⁻ adsorption, are shown in Fig. 11(a)–(d), respectively. The appearance of F 1s peaks after F⁻ adsorption confirmed that F⁻ was bound to the adsorbent (Fig. 11(a) and (b)). Due to F⁻ adsorption, the Al 2p and Fe 2p peaks were shifted to the higher binding energies of 74.2 eV and 711.2 eV from their original binding energy positions of 74.4 eV and 711.5 eV, respectively. This indicated that F⁻ reacted with both Al- and Fe-hydroxyl groups on the surface, and that both boehmite and magnetite phases contributed to F⁻ removal by formation of Al–F and Fe–F bonds.

Fig. 11 (a) Full scan, (b–d) high resolution scans (b) Al 2p, (c) F 1s and (d) Fe 2p XPS spectra of the magnetic alumina aerogel before and after 40 mg L⁻¹ of F⁻ adsorption at pH 5.0 ± 0.1.

3.4.2. FTIR and Raman spectra. FTIR spectra of MAA before and after F⁻ adsorption are provided in Fig. 12(a). Bands at 1631, 1071, 881, 735, 633 and 486 cm⁻¹ were observed. Similar band patterns were observed before and after adsorption except that all band intensities slightly decreased after F⁻ adsorption. Broad bands around 3400 cm⁻¹ and 1631 cm⁻¹ were assigned to the stretching modes of OH bands related to free water (surface adsorbed water) and the bending mode of H–O–H bound, respectively.⁴⁴ Bands at 1071 cm⁻¹ were assigned to the stretching bands of C–O–H groups from residual organics in MAA.⁴⁵ Bands at 735 cm⁻¹ and 486 cm⁻¹ resulted from the Al–OH torsional mode and from the Al–O stretching mode in AlO₆ octahedral sites of boehmite, respectively.⁴⁶ Bands at 633 cm⁻¹ were corresponding to the Fe–O stretching vibration in the crystalline lattice of Fe₃O₄.⁴⁷ After F⁻ adsorption, the observed tiny changes in the IR spectra were consistent with, but were not convincing, that F⁻ containing bonds had formed. Fe–F stretching vibration and Al–F stretching vibrations were reported to appear at 490 cm⁻¹ and above 500 cm⁻¹,^48,49 respectively. They were not observed for the samples after F⁻ was adsorbed. The densities of these bonds formed at the surface are low. Therefore, the mole fractions of Al–F or Fe–F bonds are too low to be observed since IR only samples surface and some small depth below the surface. Fig. 13 presents the Raman spectra of MAA before and after F⁻ adsorption at two initial concentrations (40 and 80 mg L⁻¹). The Raman shift at 307 cm⁻¹ was attributed to the Al–O stretching vibration in AlO(OH). The Raman shift at 939 cm⁻¹ was assigned to the C–C stretch mode of the organics remaining in MAA. These were introduced during preparation since several organic starting materials were used.

Fig. 12 FTIR spectra of magnetic alumina aerogel before and after 40 mg L⁻¹ of F⁻ adsorption at pH 5.0 ± 0.1.

Fig. 13 Raman spectra of magnetic alumina aerogel before and after 40 and 80 mg L⁻¹ of F⁻ adsorption at pH 5.0 ± 0.1.

3.4.3. Suggested mechanisms for fluoride adsorption by MAA. Mechanisms of F⁻ uptake by MAA are illustrated in Fig. 14(a) and (b). In acid media, the surface of MAA can be protonated and positively charged when the solution pH is below the pH_pzc value (7.5). This is illustrated in Fig. 14(a) for the alumina surface. After protonation, H₂O will be released to solution (Path A). F⁻ is attracted to the positive metal site and bonds to Al (Path A1). Alternatively, HF coordinates to the metal center followed by proton loss to water (Path A2). Either path ends up with F⁻ replacing the hydroxyl on a surface aluminum atom. Alternatively, Path B illustrates an associative mechanistic route. Here, F⁻ coordinates to the Al site with its attached hydroxyl group protonated. This tetracoordinated intermediate then loses water. Fig. 14(b) illustrates the situation at higher (more basic) pH values. Hydroxyl ions deprotonate surface aluminum hydroxyl group, which generates a negatively charged surface repelling F⁻. However, F⁻ can associatively coordinate with neutral surface aluminum sites to form tetracoordinated intermediates. These can lose hydroxide to generate Al–F sites where F⁻ is permanently adsorbed.

Fig. 14 Illustration of the F⁻ binding on magnetic alumina aerogel's surface. Similar mechanism route can exist for surface iron hydroxide sites on the magnetite surfaces.

4. Conclusions

Magnetic alumina hydrogel (MAA) adsorbent was fabricated by amending alumina aerogel with magnetite. This adsorbent has both good F⁻ removal performance and is easily separated after adsorption. MAA exhibited a moderately high, 32.1 mg g⁻¹, adsorption capacity at pH 5.0. This is higher than many reported alumina adsorbents. Boehmite and magnetite phases were identified in the adsorbent. The uptake of F⁻ by MAA was well described using pseudo-second-order kinetics, and the intraparticle mass transfer could be fitted using the HSDM mode. The intraparticle diffusion parameter, D_s increased when initial F⁻ concentrations were raised, which is attributed to the concentration-induced linear driving force within the particles. Decreasing solution pH enhanced F⁻ removal by MAA, by increasing electrostatic attraction between the surface and F⁻ anion and decreasing repulsions that increase at higher pH. This is consistent with the zeta potential testing results. General co-existing anions did not inhibit F⁻ removal except for HCO₃⁻ and PO₄³⁻ which gave only small adverse effects. In addition, MAA effectively mediated F⁻ removal from simulated well water containing 1.83 mg L⁻¹ of F⁻. XPS measurements confirmed that F⁻ reacted with both Al- and Fe-hydroxyl groups on the surface. No obvious FTIR band changes were observed due to the small fraction of fluoride in the volume of the adsorbent sampled by IR. In all, MAA is an effective and easily prepared adsorbent with moderate cost suitable for fluoride remediation from water.

Nomenclature

C	Concentration of F^− at t time (mg L^−1)
C0	Initial concentration of F^− in aqueous solution (mg L^−1)
Ce	Concentration of F^− at equilibrium (mg L^−1)
Cexp	Concentration of F^− recorded in the experiments (mg L^−1)
Cpred	Concentration of F^− predicted with the simulation (mg L^−1)
Ds	Intraparticle surface diffusion coefficient (cm^2 s^−1)
q̄	Average mass of F^− adsorbed per unit mass of adsorbent (mg g^−1)
q	Mass of F^− adsorbed per unit mass of adsorbent (mg g^−1)
qe	Mass of F^− adsorbed per unit mass of adsorbent at equilibrium (mg g^−1)
m	Mass of adsorbent (g)
V	Volume of the F^− solution (L)
Rp	Radius of MAA particle (cm)
r	Distance in radial direction of MA particle (cm)
t	Time (min)

Greek letters

ρp	Apparent particle density (g mL^−1)
εp	Void fraction of mesoporous alumina

Acknowledgements

This work was supported by the Fundamental Research Funds for the Central Universities (2015ZCQ-HJ-02) and the Beijing Municipal Science and Technology Project (Z151100001415008).

References

1. L. Craig, A. Lutz, K. A. Berry and W. Yang, Sci. Total Environ., 2015, 532, 127–137.
2. M. Amini, K. Mueller, K. C. Abbaspour, T. Rosenberg, M. Afyuni, K. N. Møller, M. Sarr and C. A. Johnson, Environ. Sci. Technol., 2008, 42, 3662–3668.
3. N. Mumtaz, G. Pandey and P. K. Labhasetwar, Crit. Rev. Environ. Sci. Technol., 2015, 45, 2357–2389.
4. S. V. Jadhav, E. Bringas, G. D. Yadav, V. K. Rathod, I. Ortiz and K. V. Marathe, J. Environ. Manage., 2015, 162, 306–325.
5. H. Wang, R. Li, C. Fan, J. Feng, S. Jiang and Z. Han, J. Fluorine Chem., 2015, 180, 122–129.
6. Z. Jin, Y. Jia, K.-S. Zhang, L.-T. Kong, B. Sun, W. Shen, F.-L. Meng and J.-H. Liu, J. Alloys Compd., 2016, 675, 292–300.
7. S. Wen, Y. Wang and S. Dong, RSC Adv., 2015, 5, 89594–89602.
8. B.-S. Zhu, Y. Jia, Z. Jin, B. Sun, T. Luo, X.-Y. Yu, L.-T. Kong, X.-J. Huang and J.-H. Liu, Chem. Eng. J., 2015, 271, 240–251.
9. J. Shen and A. I. Schäfer, Sci. Total Environ., 2015, 527–528, 520–529.
10. L. J. Banasiak and A. I. Schäfer, J. Membr. Sci., 2009, 334, 101–109.
11. A. Bhatnagar, E. Kumar and M. Sillanpää, Chem. Eng. J., 2011, 171, 811–840.
12. S. Jagtap, M. K. Yenkie, N. Labhsetwar and S. Rayalu, Chem. Rev., 2012, 112, 2454–2466.
13. X. Dou, Y. Zhang, H. Wang, T. Wang and Y. Wang, Water Res., 2011, 45, 3571–3578.
14. J. Hao, M.-J. Han, C. Wang and X. Meng, Microporous Mesoporous Mater., 2009, 124, 1–7.
15. X. Dou, D. Mohan, X. Zhao and C. U. Pittman Jr, Chem. Eng. J., 2015, 264, 617–624.
16. W. Li, D. Chen, F. Xia, J. Z. Y. Tan, P.-P. Huang, W.-G. Song, N. M. Nursam and R. A. Caruso, Environ. Sci.: Nano, 2016, 3, 94–106.
17. W. Li, C.-Y. Cao, L.-Y. Wu, M.-F. Ge and W.-G. Song, J. Hazard. Mater., 2011, 198, 143–150.
18. H. Wei, S. Deng, Q. Huang, Y. Nie, B. Wang, J. Huang and G. Yu, Water Res., 2013, 47, 4139–4147.
19. A. Mahapatra, B. G. Mishra and G. Hota, Ind. Eng. Chem. Res., 2013, 52, 1554–1561.
20. X. Wang, J. Li, Y. He, T. Qian and Q. Shen, Colloids Surf., A, 2014, 461, 178–183.
21. W. Yang, Q. Tang, J. Wei, Y. Ran, L. Chai and H. Wang, Appl. Surf. Sci., 2016, 369, 215–223.
22. Y. Zhong, Y. Kong, X. Shen, S. Cui, X. Yi and J. Zhang, Microporous Mesoporous Mater., 2013, 172, 182–189.
23. A. Yuchi and K. Matsuo, J. Chromatogr. A, 2005, 1082, 208–213.
24. T. Kaneyoshi, J. Magn. Magn. Mater., 2011, 323, 2483–2486.
25. L.-a. Wu, X. Qiao, S. Cui, Z. Hong and X. Fan, Microporous Mesoporous Mater., 2015, 202, 234–240.
26. M. Z. Kassaee, E. Motamedi and M. Majdi, Chem. Eng. J., 2011, 172, 540–549.
27. L. Cao, D. Chen, W. Li and R. A. Caruso, ACS Appl. Mater. Interfaces, 2014, 6, 13129–13137.
28. J. C. Groen, L. A. A. Peffer and J. Pérez-Ramírez, Microporous Mesoporous Mater., 2003, 60, 1–17.
29. H. Li, W. Li, Y. Zhang, T. Wang, B. Wang, W. Xu, L. Jiang, W. Song, C. Shu and C. Wang, J. Mater. Chem., 2011, 21, 7878–7881.
30. G. Zu, J. Shen, X. Wei, X. Ni, Z. Zhang, J. Wang and G. Liu, J. Non-Cryst. Solids, 2011, 357, 2903–2906.
31. M. S. Chiou and H. Y. Li, Chemosphere, 2003, 50, 1095–1105.
32. Y. S. Ho and G. McKay, Process Biochem., 1999, 34, 451–465.
33. S. Baup, C. Jaffre, D. Wolbert and A. Laplanche, Adsorption, 2000, 6, 219–228.
34. R. Leyva-Ramos, J. Rivera-Utrilla, N. A. Medellin-Castillo and M. Sanchez-Polo, Chem. Eng. J., 2010, 158, 458–467.
35. G. A. Heeter and A. I. Liapis, J. Chromatogr. A, 1997, 760, 55–69.
36. Q. Zhang, J. Crittenden, K. Hristovski, D. Hand and P. Westerhoff, Water Res., 2009, 43, 1859–1866.
37. Y. L. Yeh, C. M. Liao, J. S. Chen and J. W. Chen, Appl. Math. Model., 2001, 25, 593–611.
38. A. Schnepf, T. Schrefl and W. W. Wenzel, J. Plant Nutr. Soil Sci., 2002, 165, 713–718.
39. M. Kosmulski, J. Colloid Interface Sci., 2011, 353, 1–15.
40. M. Kosmulski, J. Colloid Interface Sci., 2009, 337, 439–448.
41. A. V. Klebanov, N. F. Bogdanova, L. E. Ermakova and M. P. Sidorova, Colloid J., 2001, 63, 568–572.
42. E. Tombácz, Á. Dobos, M. Szekeres, D. H. Narres, E. Klumpp and I. Dékány, Colloid Polym. Sci., 2000, 278, 337–345.
43. H. Cai, L. Xu, G. Chen, C. Peng, F. Ke, Z. Liu, D. Li, Z. Zhang and X. Wan, Appl. Surf. Sci., 2016, 375, 74–84.
44. K. Nakamoto, Infrared and Raman spectra of inorganic and coordination compounds, part a, theory and applications in inorganic chemistry, 6th edn, 2009.
45. S. Dong, X. Dou, D. Mohan, C. U. Pittman Jr and J. Luo, Chem. Eng. J., 2015, 270, 205–214.
46. A. Boumaza, L. Favaro, J. Lédion, G. Sattonnay, J. B. Brubach, P. Berthet, A. M. Huntz, P. Roy and R. Tétot, J. Solid State Chem., 2009, 182, 1171–1176.
47. J. H. Kwon, L. D. Wilson and R. Sammynaiken, Synth. Met., 2014, 197, 8–17.
48. Y. Shen, X. Wang, H. Hu, M. Jiang, Y. Bai, X. Yang and H. Shu, RSC Adv., 2015, 5, 38277–38282.
49. R. König, G. Scholz, K. Scheurell, D. Heidemann, I. Buchem, W. E. S. Unger and E. Kemnitz, J. Fluorine Chem., 2010, 131, 91–97.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra23532h

---