Preparation of mesoporous alumina electro-generated by electrocoagulation in NaCl electrolyte and application in fluoride removal with consistent regenerations

2.1 Chemicals

All the chemicals used in both the electrochemical and the adsorption processes were of analytical grade and purchased from Sigma-Aldrich (France and India). All the solutions were prepared using double distilled water. Aluminum used for electrodes was 1050 Å of the registered international designation purchased from Ridings and Dafrodis (Bouc-Bel-Air, France).

2.2 Electro-synthesis of EGA_NaCl

Electrolyses were carried out in a reactor of 1.5 L of capacity with magnetic stirring. The electrolyte was 1 L of an aqueous solution of 0.1 M NaCl. Electrodes were aluminum plates of 11.5 cm × 4.7 cm × 0.4 cm dimensions. The gap between the electrodes was of 1 cm. Electrodes were connected to a digital DC supply (2303 GPS-type) with voltage and current range of 0–32 V and 0–4 Å, respectively. A digital ammeter and voltmeter were used to measure the current or the voltage. Before electrolysis, the aluminum electrodes were immersed in 0.1 M NaOH for 10 min. Then these electrodes were rinsed with distilled water and dried at 105 °C for 10 min. A typical electrolysis is as follows. Freshly prepared electrodes were weighed, then immersed in the electrolyte and finally connected to the DC supply. The electrolyses were performed with a constant current. The solution was magnetically stirred. At regular intervals, samples of solution were taken to carry out measurements of pH, turbidity, and conductivity with an electronic pH-meter (Metrohm), a turbidimeter (Hach) and a conductivity meter (Radiometer Meterlab CDM 210), respectively. The electrolysis course was investigated with a current of 0.3 or 0.7 Å. The preparation of EGA_NaCl for the fluoride removal studies was performed with a current of 0.7 Å for 2 h. At the end of the electrolyses, the electrogenerated solid was collected by filtration and then rinsed three times with distilled water under agitation and finally dried in an oven at 105 °C for 24 h. About 1.9 g of dried material was obtained. The used electrodes were rinsed with distilled water and then dried at 105 °C for one hour and finally weighed.

2.3 Instrumental characterization

The samples were analyzed by X-ray powder diffraction (XRD) using a PANalytical Empyrean diffractometer (θ–θ Bragg–Brentano geometry) working with the Cu Kα radiation (λ Kα₁ = 1.5406 Å, λ Kα₂ = 1.5444 Å). Data were collected with a PIXcel1D detector, in the angular range 5–80° (2θ). Match! software (Cristal Impact, Germany) was used to display XRD diagrams and to identify phases by comparison with reference database. The infrared spectra were recorded using a Thermo Scientific Nicolet iS5 FT-IR spectrometer with iD7 ATR accessory. A small amount of the ground sample was laid down on the diamond crystal. Infrared spectra were recorded from 4000 to 400 cm⁻¹. Omnic software was used to display spectra. Observations with a Scanning Electron Microscope (SEM) and element analysis by Energy Dispersive Spectroscopy (EDS) were performed using Jeol JSM 6400 equipment operating at an accelerating voltage of 20 kV.

The physical and structural characteristics were determined by nitrogen adsorption/desorption at 77.5 K using Micromeritics ASAP 2020 surface area and porosity size analyzer.

2.4 Analytical techniques

Fluoride removal experiments were performed with dilution of a homemade stock solution of 100 mg L⁻¹. The defluoridation experiments were carried out by the batch method at temperatures of 300 K, 313 K and 323 K, various pH values, EGA_NaCl doses and fluoride concentrations of 2.78, 4.41 and 5.31 mg L⁻¹. The residual fluoride concentrations were determined with a fluoride ion selective electrode (Orion, number 9609, USA) connected to an ion meter (Orion, model 290A, USA). The fluoride samples were diluted to 1:1 with a commercial buffer (TISAB II, Orion). The specific amount of fluoride adsorbed was calculated from Eq. (1) where Q_e is the adsorption capacity at the equilibrium (mg g⁻¹), C₀ and C_e are respectively the initial and the equilibrium fluoride concentration (mg L⁻¹), V is the volume of the aqueous solution (L) and W is the mass (g) of EGA_NaCl.

The pH of point of zero charge (pH_PZC) was determined by the following method. 100 mg of EGA_NaCl was added to 50 mL of 0.05 M NaNO₃ deoxygenated aqueous solutions with a pH range of 1–12. Blanks with no adsorbent were also run in the same initial pH values. After magnetically shaking for 24 h at room temperature, the final pH was measured. The pH_PZC was determined as the pH of the NaNO₃ solution that remained unchanged after contact with EGA_NaCl. The isoelectric point (IEP) for EGA_NaCl was determined by potentiometric titration with the support of electrophoretic technique using Zetasizer Nano with 633 nm He-Ne laser equipped with an MPT-2 autotitrator (Malvern, U.K). Series of 0.01 M and 0.001 M NaCl solutions (0.1 g per 50 mL) from initial pH 2 – 12 were mechanically agitated for 24 h at room temperature. Later, the above solutions were calculated for ζ potentials using Smoluchowski equation.

The effect of co-existing anions was investigated during defluoridation experiments which were conducted for fluoride concentration of 2.78 mg L⁻¹ for an equilibrium time of 35 min at 300 K. The various co-ion solutions of 50, 100 and 200 mg L⁻¹ were prepared with sodium salts of chloride, nitrate, hydrogen carbonate, sulfate and phosphate. Regeneration experiments were conducted using 0.05 M NaOH solution with the spent EGA_NaCl for the time of 35 min. The sorption of fluoride using the initial fluoride concentration of 2.78 mg L⁻¹ at an optimized pH was initially carried out for the optimized equilibrium time and followed by the regeneration study. The adsorption and regeneration cycles were attempted up to six operations. Validation of kinetic and isotherm models was made using the experimental data. The determination of thermodynamic parameters such as ΔG^o, ΔH^o and ΔS^o was also calculated.

2.5 JMP statistical software

The statistical method called Response Surface Methodology (RSM) uses the quantitative results from appropriate experiments for determining regression model equations and suitable operating conditions (Alam et al., 2007). RSM is the incorporation of mathematical and statistical techniques for modeling and analysis of problems where the response of interest is influenced by several variables (Montgomery, 2001).

A standard RSM design called full factorial design (FFD) was applied in this work to study the effect of variables on the adsorption of fluoride from aqueous solution using EGA_NaCl in a batch mode experiments. The FFD was widely used for fitting a linear and quadratic model. By using this method, modeling is possible with the requirement of only a minimum number of experiments. It is not necessary in the modeling procedure to know the detailed reaction mechanism since the mathematical model is empirical. Generally, the number of factors under study allowed the choosing of FFD rather than a fractional factorial design that further reduces the number of experiments (Montgomery, 2001), which may decrease the reliability of the model. The fluoride sorption efficiency onto EGA_NaCl made the subject of the study, and was chosen as the model response (Y, mg g⁻¹). The main factors affecting this sorption efficiency are the temperature, EGA_NaCl dose and the contact time. For this modeling, all experiments are provided in order to achieve the number mⁿ FFD model. Each variable is investigated at three levels. Meanwhile, as the number of factors (n) increases, there would be a rapid increase in the number of runs for a complete replicate of the design. In this case, main effects and interactions may be estimated by full factorial designs running by all the experiments suggested by models. The individual second-order effects can be estimated separately by the FFD and only the third-order effect is supposed to be insignificant. The resulting responses and the corresponding parameters are modeled and optimized using ANOVA. Basically this optimization process involves three major steps such as, performing the statistically designed experiments, estimating the coefficients in a mathematical model and, predicting the response and thereby checking the adequacy of the model.

In the above equations, Y is the predicted response, b_o the constant coefficient, bi the linear coefficients, bij the interaction coefficients, bii the quadratic coefficients and xi, xj are the coded values of the adsorption of fluoride onto EGA_NaCl variables. The central points are chosen such that they allow readability, which ensures that the variance of the model prediction is constant at all points equidistant from the design center (Box and Hunter, 1975). The number of experiments (N) is defined by the choice of the number of factors and their levels. In the case of a three – level study with three factors, the number of experiments is equal to N (equal to 3³). Thus the number of experience allowed to study all the probabilities and possible configurations between the three factors and three levels. Hence, the total number of tests (N) required for the three independent variables (m) at the three levels (n) is 27 (N = 3³).

3.1 Preparation of EGA_NaCl

The electrolyses were carried out at room temperature in 0.1 M NaCl solution. The electrolysis current was controlled at 0.7 A. After 120 min of electrolysis the mass loss of the anode was 0.565 g, it was larger than the mass loss calculated at 0.469 g by using the Faraday law. The cathode showed also a mass loss of 0.182 g. These mass losses for both electrodes are usual observations in electrolysis with aluminum electrodes. The larger mass losses are due to concomitant corrosion which contributes to the formation of 1.936 g of the electrolysis products.

Owing to the formation of EGA_NaCl, the turbidity of the solution nearly increased linearly as shown in Fig. 1A. At the end of electrolysis, the turbidities are proportional to the current. This observation and the linear increase may be explained by the formation of homogeneous particles.

Close

Fig. 1 Influence of current on turbidity (A), pH (B) and conductivity (C) during electro-generation of EGA_NaCl. Conditions: Volume of NaCl – 1 L; Concentration of NaCl – 0.1 M; Temperature – 300 K; Time – 120 min; Anode and Cathode – Aluminium.

Export to PPT

At the beginning of electrolysis, the pH increased and reached an approximate constant value as shown in Fig. 1B. The pH increase was larger for larger current. This increase in pH was explained by the involvement of chloride anion in competition with hydroxide anion leading to incorporation of chloride into the aluminum hydroxides (Eqs. (5) and (6)) while the released hydroxide anion increases the pH (Tchomgui-Kamga et al., 2013).

In agreement with the pH increase, concomitant increase in the conductivity was observed as shown in Fig. 1C. These conductivity increases can be attributed to the electrical mobility of hydroxide anion which is larger than the one of chloride anion.

3.2 Characterization of EGA_NaCl

The SEM images of EGA_NaCl are presented in Fig. 2(A)–(C). Despite the mixture of crystallized aluminum compounds in EGA_NaCl, SEM images show almost homogenous particles which are planar particles of about 0.1–1 μm. This structure offers large surface area for the access of fluoride anions. The EDS results showed that Al and O were the main elements detected. Besides these elements, Cl was also detected at about 2 atomic%. The absence of Na, or its very low concentration, is in agreement with the occurrence of the exchange reactions of Eqs. (5) and (6) which explain the increase in pH during the electrolysis (Tchomgui-Kamga et al., 2013).

Close

Fig. 2 Scanning electron micrographs with EDAX for EGA_NaCl at different magnifications (A, B and C).

Export to PPT

The XRD patterns of EGA_NaCl powder are presented in Fig. 3A and B. There are a lot of peaks in which it is possible to identify the presence of aluminum oxide hydroxide or boehmite AlOOH, characterized by small broad peaks at 2θ = 14.5° and 28.3° for the most noticeable ones. The other peaks correspond to three aluminum hydroxides Al(OH)₃. The enlargement of the angular scale of the diffractogram (Fig. 3B) shows the presence of gibbsite, bayerite and nordstrandite. This composition of EGA_NaCl is confirmed by the IR spectrum (Fig. 4) which exhibits distinct bands in the 3000–3700 cm⁻¹ range, assigned to the stretching vibration of O—H bonds of the different polymorphs of aluminum hydroxides and also of the boehmite (Elderfield and Hem, 1973; Music et al., 1998; Ogorodova et al., 2012). It is remarkable that four crystallized aluminum compounds are obtained in the same electrolysis. Their formation cannot be explained by three origins of aluminum: the corrosion of the anode and the cathode and the anodic dissolution of the anode.

Close

Fig. 3 X-ray diffraction pattern of EGA_NaCl (A) Expanded pattern of 2θ between 10 and 30 (B).

Export to PPT

Close

Fig. 4 FTIR spectrum of EGA_NaCl.

Export to PPT

The nitrogen adsorption-desorption analysis for EGA_NaCl and the textural properties based data are shown in Table 3. According to IUPAC classification (Betiha et al., 2011), the EGA_NaCl material shows Type IV isotherms characteristic of H4 type (Fig. 5A) with the desorption branch leading to the lower closure point at P/P_o is equal to 0.46. In general, this relative pressure is inferred to be almost independent of the nature of the porous EGA_NaCl but depends mainly on the adsorptive (Betiha et al., 2011). The Type H4 loop is not supposed to exhibit any limiting adsorption at high P/P_o with aggregates organized as an assemblage of particles which tend to be loosely coherent. These EGA_NaCl aggregates are recognized to resemble plate – like particles giving rise to slit – like pores (Sing et al., 1985). The isotherm hysteresis demonstrates that a pronounced condensation step in the relative pressure range (P/P_o) 0.1–0.2 suggests the narrow – sized mesoporous materials (Benítez-Guerrero et al., 2014). The following second condensation of nitrogen associating the nitrogen uptake at the relative pressure (P/P_o) greater than 0.8 indicates the possible isothermal sorption due to large voids between EGA_NaCl particles. The distribution of pore size of EGA_NaCl is illustrated in Fig. 5B. It shows a narrow curve in the range of ca. 1.6–2.4 nm with the maximum pore volume of 0.015 cm³ g⁻¹. This is followed by another gradation between 2.5 nm and 7.0 nm with pore volume of 0.0075 cm³ g⁻¹. Then the gradation in pore volume reaches 0.0125 cm³ g⁻¹ and thereafter a declination was observed up to 40 nm. The average pore diameter was observed to be 6.64 nm. The recorded BET surface area of 114.31 m² g⁻¹ for the crystalline phases contained EGA_NaCl is close to that of the amorphous in – situ Al₂O₃·xH₂O (118.24 m² g⁻¹) with low crystalline nature (Liu et al., 2011). Fang et al., 2016 reported the surface area (111.65 m² g⁻¹) for the calcined Al(OH)₃ of amorphous type was increased about 100 times to that of raw Al(OH)₃. The enrichment of surface area was achievable when the process was maintained at 500 °C for 240 h. But 50% of the time was required for the preparation of EGA_NaCl in different crystalline phases by this electro-synthesis. The textural properties of EGA_NaCl are represented in Table 3.

Close

Fig. 5 Nitrogen adsorption – desorption isotherm of EGA_NaCl (A) Pore size distribution of EGA_NaCl (B).

Export to PPT

3.3 Fluoride adsorption studies

3.3.1

3.3.1 Influence of pH

The parameter which influences the sorption of fluoride at the water – sorbent interfaces is significantly pH and is associated with pH_zpc (Jin et al., 2015). Fluoride sorption onto alumina is strictly governed by pH and particularly at weakly acidic pH levels (Ku and Chiou, 2002). Either low or high pH conditions are favorable for the dissolution of Al species. Based on theories, solutions with pH < 5.5, Al³⁺ ions appear whereas pH > 8 aluminates are formed.

Solution pH could affect the surface properties of EGA_NaCl and hence the protonation – deprotonation degree of oxygen – containing functional groups directly (Zou et al., 2016).

The trend of pH versus adsorbed fluoride (Fig. 6A) begins from pH 2 with the low fluoride uptake amount of 8.95 mg g⁻¹ (24.1% of fluoride removal). Later, the plot gradually elevates and reaches a maximum fluoride uptake of 16.33 mg g⁻¹ at pH 6.55. Then the curve continued with a descending trend toward the low fluoride uptake amount of 5.20 mg g⁻¹ at pH 11.5. The pH value of 6.55 less than pH_zpc of 6.7 (Fig. 7A) favors a positively charged nature due to excess H⁺ on the EGA_NaCl surface which facilitates the coulombic attractive force with the fluoride ions (Eqs. (7) and (8))

(7)

$$\text{EGA}\equiv \text{OH}+{\text{H}}^{+}\rightleftarrows \text{EGA}\equiv {\text{OH}}_2^{+}$$

(8)

$$\text{EGA}\equiv {\text{OH}}_2^{+}+{\text{H}}^{+}\leftrightarrow \text{EGA}\equiv {\text{OH}}_2^{+}\dots {\text{F}}^{-}$$ where ≡OH₂⁺ is the positive hydroxo group on the surface of EGA_NaCl.

Close

Fig. 6 pH influence (A) dose influence (B) temperature [F]_o influence in mg L⁻¹ (C) for the fluoride sorption onto EGA_NaCl Conditions: (A) EGA dose – 0.1 g; Time – 35 min; [F]₀ – 2.78 mg L⁻¹; T – 300 K. (B) pH – 6.55; Time – 35 min; [F]₀ – 2.78 mg L⁻¹; T – 300 K. (C) pH – 6.55; Time – 35 min; EGA dose – 0.1 g.

Export to PPT

Close

Fig. 7 pH_zpc (A), plot of pH versus Zeta potential (B). Conditions: (A) EGA dose – 0.1 g; Volume of NaNO₃ – 0.05 L; Concentration of NaNO₃ – 0.05; Time – 24 h (B) EGA dose – 0.1 g; Volume of NaCl – 0.05 L.

Export to PPT

It was reported (Hao and Huang, 1986) that, at pH greater than 7, the sorption of fluoride onto EGA_NaCl takes place through a ligand exchange mechanism as Eq. (9),

(9)

$$\text{EGA}\equiv \text{OH}+{\text{F}}^{-}\rightleftharpoons \text{EGA}\equiv \text{F}+{\text{OH}}^{-}$$

Fig. 7B shows the plot of pH versus zeta potential (mV) for EGA_NaCl which determines the Isoelectric point (IEP) for EGA_NaCl. It is a point at which the electrophoretic mobility is zero or a reverse sign representation of zeta potential. Assumption of aged alumina surface coated by amorphous Al(OH)₃ can be substantiated from zeta potential measurements and solid – solution equilibria. From the measurements, the values of IEP were determined to be 7.4 and 7.6 respectively for 0.01 M and 0.001 M NaCl solutions respectively. The zeta potential values decreased with the increasing pH values. The electrophoresis measurements showed the IEP shift toward lower pH due to the significant adsorption of background NaCl electrolyte. Consequently, the positively charged EGA_NaCl surface gets neutralized with the anions of the electrolyte and causes the decrease in the IEP. The development of significant positive charges at pH 3.0–6.5 indicates the high sorption efficiency of EGA_NaCl toward negatively charged fluoride ions and further suggested that the governing mechanism may be electrostatic in nature.

In the case of EGA_NaCl, pH greater than 6.55 followed a gradual declination up to pH 10 with 10.08 mg g⁻¹ (27.2% of fluoride removal) of fluoride uptake and then a sudden fall at pH 11.5 with the removal of 14.3%. At high pH values, exceeding the pH 11.5, the gradual distortion of hydroxo groups leads to form a surface with increasing negative charge. In addition, the OH⁻ ions compete against fluoride for the active sites and the electrostatic repulsion caused between negatively charged fluoride and the EGA_NaCl surface. Due to the above facts, fluoride removal at high pH was studied with appreciable reduction. The lower fluoride removal capacity at acidic pH levels may be attributed to the HF ion formation (Raichur and Basu, 2001) or complexation of fluoride ions with dissolved aluminum. The decreasing sorption of fluoride below pH 5 can also be associated with alumina dissolution promoted by fluoride (Nordin et al., 1999) and made clear by Reyes Bahena et al. through the solubility – pH diagram of α-Al₂O₃. In addition, the fluoride enhanced solubility of aluminum hydro(oxides) is possible at low pH conditions due to the formation of predominant aluminum – fluoro complexes (AlF₂⁺, AlF²⁺, AlF₃(aq), and AlF₄⁻) in solution. In support of the present study, several researchers explored the optimal pH level between pH 6 and pH 7 and obtained the maximum defluoridation efficiency (Gong et al., 2012; Kumar et al., 2011; Yang et al., 2014; Zhang and Jia, 2016; Tripathy et al., 2006; Du et al., 2014; Wang et al., 2009a,b).

3.3.4

3.3.4 Inhibition by interfering ions

The purpose of this study is to explore the inhibition to fluoride sorption onto EGA_NaCl by other interfering ions such as chloride, nitrate, phosphate, sulfate and hydrogen carbonate as these major anions are present in groundwater. Among the interfering anions, hydrogen carbonate was the most inhibiting and chloride was the least inhibiting anions. The inhibition percentage increased with the increase in the concentration of anions (50–200 mg L⁻¹) as shown in Fig. 8A. The maximum negative interference by chloride, nitrate, sulfate, phosphate and hydrogen carbonate was recorded as 15.6%, 17.9%, 38.2%, 43.0% and 77.0% respectively for the anionic concentration of 200 mg L⁻¹. Among the interfering ions, the highest inhibition due to hydrogen carbonate was associated with the raise in the pH level (8.5–11.0) toward basic condition and the competency for the same sorption site as well. The maximum inhibition due to hydrogen carbonate for fluoride sorption using aluminum adsorbents has already been reported by early researchers (Patankar et al., 2013; Nigussie et al., 2007; Kamble et al., 2010). The interference of phosphate to fluoride sorption may be associated with the enabled formation of aluminum phosphate under alkaline conditions as phosphates have strong affinity over aluminum (III) species (Ayoob and Gupta, 2009). The inhibitory action of sulfate may be envisaged due to the formation of partial outer and inner sphere complex formation and its divalent nature repulsion against fluoride (Sujana et al., 1998). The interference patterns of chloride and nitrate (low affinity ligands) indicated the formation of weaker bonds with the active sites on the EGA_NaCl’s surface through outer sphere mechanism. Hence lower inhibitory influence on fluoride sorption was observed due to chloride and nitrate ions.

Close

Fig. 8 Influence of co – ions during sorption of fluoride onto EGA_NaCl (A) and Regeneration of EGA_NaCl with first fluoride removal as reference (B). Conditions: (A) EGA dose – 0.1 g; Time – 35 min; T – 300 K (B) EGA dose – 0.1 g; Time – 35 min; T – 300 K; Regenerating solution – 0.05 M NaOH; V_NaOH – 0.05 L.

Export to PPT

3.4 Kinetic and isotherm models

Table 1 shows the kinetic and isotherm data for the fluoride sorption onto EGA _NaCl controlled at different temperatures (300 K, 313 K and 323 K) and initial fluoride concentrations (2.78 mg L⁻¹, 3.41 mg L⁻¹ and 5.31 mg L⁻¹).

The fluoride kinetics based on pseudo – order – model is validated using the linearized – integral form (Lagergren, 1898) to describe adsorption in solid – liquid systems, where the adsorption depends on the solid capacity (Ho, 2004). The pseudo – first – order rate constant (k₁, min⁻¹) and q_e (mg g⁻¹) are calculated from the slope and the intercept of Eq. (11) respectively.

The pseudo – second – order model (Ho, 2006) can be expressed by the following Eq. (12) and is based on the assumption that the rate-limiting step may be chemical sorption or chemisorption involving valence forces through sharing or exchanging electrons between the sorbent and sorbate.

Fig. 9(A)–(C) shows the plot of t/q_t versus t and from which the pseudo – second – order rate constant (k₂, g mg⁻¹ min⁻¹), initial sorption rate (h, mg g⁻¹ min⁻¹) and equilibrium sorption capacity (q_e, mg g⁻¹) can be determined. It is apparent that the second – order equilibrium sorption capacity (q_e) increased with respect to initial fluoride concentration and temperature. The kinetic data of pseudo – second – order model revealed that the influence of initial fluoride concentration doubled the q_e values (2.2 times at 300 K and 2.1 times at 313 K) from 3.41 mg L⁻¹ to 5.31 mg L⁻¹. At 300 K and 313 K, the influence between the first two concentrations (2.78 mg L⁻¹ and 3.41 mg L⁻¹) was lower than between the second and third fluoride concentrations (3.41 mg L⁻¹ and 5.31 mg L⁻¹). The influence of fluoride concentration seems to be reversed at 323 K, where about 1.95 fold (for the first two fluoride concentrations) was recorded against 1.27 fold between 3.41 mg L⁻¹ and 5.31 mg L⁻¹. The direct proportionality of both temperature and fluoride concentration in driving the initial sorption rate (h) could be well established from the calculated data. An appreciable compliance of pseudo – second – order model than the pseudo – first – order model could be corroborated from the regression coefficient values (Table 1).

Close

Fig. 9 Plots of kinetic models for different fluoride concentrations: Pseudo – second – order (A, B, C) Intra – particle – diffusion (D, E, F) Elovich (G, H, I); [F]₀ – Initial fluoride concentration.

Export to PPT

The following Eq. (13) for intra – particle diffusion determines the rate parameter for fluoride sorption onto EGA_NaCl (Mi et al., 2012).

Elovich model (Aharoni and Tompkins, 1970) is one among the kinetic models to describe chemisorption and Eqs. (14) and (15) are as follows:

The simplified form developed by Chien and Clayton (1980) is,

The adsorption isotherms express the specific relation between the concentration of adsorbate and its degree of accumulation onto adsorbent surface at a constant temperature. Based on the adsorption equilibrium data, three adsorption isotherm equations viz., Langmuir isotherm, Freundlich isotherm and DKR were examined for understanding the thermodynamics of fluoride sorption onto EGA_NaCl. Based on the Langmuir isotherm (Langmuir, 1916), an adsorbent is assumed to have the monolayer adsorption on uniform homogeneous surface having identical sites without lateral interactions (Marani et al., 2016).

Langmuir Eqs. (16) and (17) are as follows:

The sorption capacity (Q°) is the amount of adsorbate at complete monolayer coverage (mg g⁻¹), C_e is the residual fluoride concentration at equilibrium (mg L⁻¹) and b (L mg⁻¹) is the Langmuir isotherm constant that relates the affinity of binding sites and the energy of adsorption. The respective values of Q° and b were determined from the slope and the intercept of the straight line plot of C_e/q_e versus C_e (Table 1). The sorption capacity (Q°) increased from 4.683 mg g⁻¹ to 16.267 mg g⁻¹ on increasing the temperature from 300 K to 323 K (Fig. 10A). For the increase of 1.78 times in ‘b’, the sorption capacity was 3.34 times at 323 K. The influence of temperature on the sorption capacity confirmed the endothermic nature of sorption of fluoride onto EGA_NaCl. The regression values greater than 0.995 approved the excellent validation of the Langmuir model by suggesting the fluoride sorption on a homogeneous surface and preferably a monolayer. This observation is in agreement with the fluoride sorption onto mesoporous alumina, studied by Xu et al. (2016). The equilibrium saturation point characterized by the plateau of Langmuir model assumes the monolayer sorption on the sites without further sorption. In addition, Langmuir theory made an inverse relation between the distance and intermolecular attraction among molecules. Thus, another parameter R_L, a dimensionless constant separation factor (Weber and Chakravorti, 1974) can be calculated with the following Eq. (18),

Close

Fig. 10 Langmuir model (A) Plot of ln K_d versus 1/T (B).

Export to PPT

Based on the Freundlich model (Freundlich, 1906) dealing with the multilayer sorption on a heterogeneous surface, Eq. (19) where, the Freundlich constants K_F (intercept) and n (as 1/n from the slope) are measures of adsorption capacity (mg g⁻¹) and adsorption intensity (or surface heterogeneity) respectively. The magnitude of 1/n is referred as a joint measure of both the relative magnitude and the diversities of energies associated with a particular sorption process (Karickhoff, 1981; Weber et al., 1992).

Evidently, from the decreased values of 1/n from 0.247 to 0.137, it is inferred that the surface heterogeneity is directly proportional to the temperature. In addition, the values of 1/n between 0 and 1 confirm the favorable conditions for the sorption of fluoride onto EGA.

On the other hand, the value of n > 1 at all the temperatures indicates the strong affinity between EGA and fluoride. The obtained linearity for the plot ln q_e versus ln C_e based on the regression values implies that the compliance of Freundlich model (R² ≥ 0.86) is less than the Langmuir model and is represented in Table 1.

In order to comprehend the nature of adsorption, equilibrium data were tested with Dubinin–Radushkevich (DR) isotherm. The linearized DR equation is represented by Eq. (20) (Dubinin, 1960)

3.5 Evaluation of thermodynamic parameters

The significance of thermodynamic analysis such as ΔH, ΔS and ΔG lies in its illustration of sorption mechanism and to study the nature of the process whether spontaneous or non-spontaneous and endothermic or exothermic. This parametric analysis is based on the calculation of temperature – dependent adsorption isotherms. The calculation of standard free energy change is possible from Eqs. (22) and (23),

The thermodynamic parameters are shown in Table 2. The negative ΔG values tend to increase with the increase in temperature and substantiate the feasibility of fluoride sorption onto EGA_NaCl. The positive values of ΔH indicated the endothermic nature of fluoride sorption. The tendency of endothermic nature was increased from 2.78 mg L⁻¹ to 3.41 mg L⁻¹ but decreased thereafter. It corroborates the decreasing tendency of endothermic nature while increasing the concentration from 3.41 mg L⁻¹ to 5.41 mg L⁻¹. The values of ΔH^o greater than 25 kJ mol⁻¹ were inferred to be governed by chemical forces during the fluoride sorption (Gopal and Elango, 2007).

Likewise, the increased ΔS values indicated the increased randomness (at the solid/solution interface) of about 3.12 times from 2.78 mg L⁻¹ to 3.41 mg L⁻¹. This may be due to the displacement of water molecules by fluoride species which leads to gain more translational entropy so as to allow the prevalence of randomness in the system (Doula et al., 2000). Conversely, the translational entropy loss at higher concentration of 5.41 mg L⁻¹ may be the pertinent reason for the decrease in ΔS value.

The Arrhenius Eq. (24) is helpful to understand the nature of sorption based on physical forces with E_a value no more than 4.2 kJ mol⁻¹ or chemical forces with E_a values between 8.4 kJ mol⁻¹ and 83.7 kJ mol⁻¹ (Aksu, 2002).

3.6 JMP statistical analysis

The present FFD system on the sorption of fluoride onto EGA_NaCl is applied to explore the effect of variables in the sorption process. FFD was applied to develop correlation between the variables operating aqueous fluoride solution and the EGA_NaCl. The details about the range and levels of independent variables are given in the inserted table of Fig. 11. Two models were selected and applied using JMP statistical software version 10.0.0 (SAS Institute Inc., Cary, NC, USA).

Close

Fig. 11 The combined effect of temperature and fluoride concentration (A) time and fluoride concentration (B) on the adsorption of fluoride at pH 6.55 and EGA dose of 0.075 g.

Export to PPT

The performance of regression analysis in fitting the response function of fluoride sorption (mg g⁻¹) was done. FED corroborates the influence of each factor on response and the interdependency among factors. The significant factors which affect the sorption of fluoride were determined by the performance of ANOVA. The variables denoted with coded values represent the sorption of fluoride as a function of time (X₁), temperature (X₂) and initial fluoride concentration (X₃). The final empirical models such as linear and quadratic models (as expressed by Eqs. (3) and (4)) in terms of coded factors for fluoride sorption are shown by Eqs. (25) and (26) respectively.

In Table 4 is represented the design of experiment with the observed and predicted removal of fluoride for quadratic and linear models. The observed values are the measured response data for a particular run whereas the predicted are based on model evaluation with the use of approximating functions. The variance of original observations is constant for all values of the response as inferred through the random scatter plot between Studentized residuals and predicted fluoride sorption.

The data analysis of variance (ANOVA) for the factorial design is summarized in Table 5. The values obtained as “Prob > F” less than 0.05 indicate that the terms are termed statistically significant. The factors such as time (X₁), temperature (X₂), fluoride concentration (X₃) and the squared interaction of fluoride concentration (X₃²) are statistically significant at the 95% confidence level in the case of quadratic model. The coefficient of determination (R²) of linear and quadratic model was 0.929 and 0.984 respectively and indicated an appreciable fit between predicted values and experimental data points. It can be corroborated that 92.9% (linear model) and 98.4% (quadratic model) of variation in the fluoride sorption are explained by the independent variables. The difference of R² and R²_adj within ∼0.2 is considered with reasonable agreement. In view of our case, the differences in R² and R²_adj values obtained for linear and quadratic models are 0.023 and 0.009 respectively. However, these values are higher for quadratic model (0.984 & 0.975) than the linear model (0.929 & 0.906). Hence the validation of quadratic model with the fluoride sorption onto EGA_NaCl fits well to that of the linear model. It was justified that the insignificant terms interfere in the fair correlation coefficients and are likely the variable range with limited number of experiments and the nonlinear influence of parameters on the process response (Cronje et al., 2011; Sahu et al., 2009).

The investigation on the influencing three factors on fluoride sorption is depicted using the response surface methodology 3D plots (Fig. 11A and B). The combined effect of fluoride concentration and time at an optimized dose of 0.075 g and pH of 6.55 illustrates an initial raise of fluoride sorption in a gradual manner which later appeared to have a steep raise. The increase in fluoride sorption reaches the maximum value of greater than 70 mg g⁻¹. Similarly, the effect of fluoride concentration and temperature at an optimized pH and EGA_NaCl dose demonstrated the possible increase in fluoride sorption both with respect to initial concentration and temperature and hence the plots appear almost identical. But the steep raise in fluoride sorption for time – fluoride interaction is greater than that for concentration – temperature interaction. Hence it may be inferred that the influence of fluoride concentration is more predominate than the temperature. The plot for the combined effect of time and temperature for the fluoride concentration of 2.78 mg L⁻¹ at an optimized pH and dose shows the gradual raise of fluoride sorption with the increase in temperature. However, the pattern shows the raise of fluoride sorption at lower temperature is gradual than the one at higher temperature appearing to have a steep pattern. Based on the mathematical equations developed for linear and quadratic models, the optimized fluoride sorption (at pH 6.55 and EGA_NaCl dose of 0.075 g) conditions according to linear model are 313 K and 5.31 mg L⁻¹ (initial fluoride concentration), and according to quadratic model are 303 K and 5.31 mg L⁻¹ of initial fluoride concentration. Nevertheless, the above optimal conditions for the fluoride sorption of 53.07 mg g⁻¹ (predicted value is 53.11 mg g⁻¹) for the quadratic model and 45.25 mg g⁻¹ (predicted value is 45.13 mg g⁻¹) for the linear model were observed. The lower temperature and higher fluoride sorption (at the same optimized dose and pH) seems suitable as per the conformity of quadratic model.

3.7 Regeneration of spent EGA_NaCl

An efficient regeneration process includes the factors such as, a remarkable removal, avoidance of excess regenerating solution and appreciable number of cycle operations. In our study, after optimizing the required sodium hydroxide of 0.05 M, the cycle of operation was admirable up to 5 times. The recorded removal efficiency was 99.3%, 97.9%, 97.1%, 96.1% and 95.5% for the first, second, third, fourth and fifth regenerations respectively (Fig. 8B). The difference in the removal percentage between first and fifth regeneration is about 3.8%. But on proceeding to sixth regeneration, the removal efficiency was dropped to 82.4% with a difference of 13.1%. Even though the following regenerations stand interrogative for further substantive efficiency, the retention more than 80% of removal after sixth regeneration is quite appealing. Although the regeneration was consistent in the beginning up to five cycles, the continuous interaction under basic condition during the sixth regeneration process could be the reason for initiating the dissolution of EGA_NaCl. Liu et al. (2011) studied the soluble Al species as Al(OH)₄^– ions under basic conditions at pH > 9 using in – situ Al₂O₃·xH₂O. The regeneration study attempted by Du and coworkers (Du et al., 2016) using 1 × 10⁻⁴ M NaOH was justifiable up to 2 cycles due to the Al loss from aluminum hydr(oxide) amended molecular sieves. Similar observation was reported using ordered mesoporous alumina (OMA 850) which could sustain the removal efficiency of about 80% after three cycles of operations (Ayoob and Gupta, 2009).

3.8 Comparison with other aluminum adsorbents

Since EGA_NaCl is obtained from electrolysis with aluminum electrodes and its defluoridation efficiency can be compared to other defluoridation adsorbents containing aluminum. It can also be compared with electrocoagulation using aluminum electrodes.

Concerning the comparison, many papers give a table with adsorption capacities in the range 0.85–16.2 mg g⁻¹ (Jin et al., 2015) or 0.37–8.40 mg g⁻¹ (Goswami and Purkait, 2012). Some aluminum compounds show a greater adsorption capacity like 110 mg g⁻¹ for freshly prepared aluminum hydroxides (Liu et al., 2011), or 62.5 mg g⁻¹ for nanoscale aluminum hydroxide (Adeno et al., 2014) and 56.8 mg g⁻¹ for bayerite/boehmite nanocomposites (Jia et al., 2015). A good comparison of defluoridation (Table 6) adsorbents needs more consideration than the adsorption capacity. Important results such as the equilibrium time and the efficiency of the regeneration of spent adsorbents must be taken into account. From this point of view, EGA_NaCl is an interesting and efficient adsorbent.

The comparison of a defluoridation adsorbent with an electrocoagulation is rarely discussed in the papers. With the results of the pioneering work of Ming et al. (1987), and assuming the formation of Al(OH)₃, it is possible to calculate a fluoride removal capacity of 30.46 mg g⁻¹. After five regeneration experiments (Fig. 8B) EGA_NaCl showed a fluoride removal capacity of 95.68 mg g⁻¹ which correspond to a ratio Al/F of 2.8 instead of the ratio of 8 given by Ming et al. (1987). The chemical regeneration of EGA_NaCl which is an electro-generated compound allows a more efficient defluoridation treatment than electrocoagulation.