Iron functionalized silica particles as an ingenious sorbent for removal of fluoride from water

The paucity of safe drinking water remains a global concern. Fluoride is a pollutant prevalent in groundwater that has adverse health effects. To resolve this concern, we devised a silica-based defluoridation sorbent from pumice rock obtained from the Paka volcano in Baringo County, Kenya. The alkaline leaching technique was used to extract silica particles from pumice rock, which were subsequently modified with iron to enhance their affinity for fluoride. To assess its efficacy, selected borehole water samples were used. Scanning electron microscopy, X-ray diffraction, Fourier transform infrared and X-ray fluorescence spectroscopy was used to characterize the sorbent. The extracted silica particles were 96.71% pure and amorphous, whereas the iron-functionalized silica particles contained 93.67% SiO2 and 2.93% Fe2O3. The optimal pH, sorbent dose and contact time for defluoridation of a 20 mg/L initial fluoride solution were 6, 1 g and 45 min, respectively. Defluoridation followed pseudo-second-order kinetics and fitted Freundlich's isotherm. Fluoride levels in borehole water decreased dramatically; Intex 4.57–1.13, Kadokoi 2.46–0.54 and Naudo 5.39–1.2 mg/L, indicating that the silica-based sorbent developed from low-cost, abundant and locally available pumice rock is efficient for defluoridation.

www.nature.com/scientificreports/ et al. demonstrated that it is feasible to isolate cost-effective silica particles from pumice volcanic rock using an alkaline extraction protocol at low temperatures. This method yielded 94% pure amorphous silica nanoparticles with a high specific surface area (422 m 2 g −1 ) and a mean pore diameter of 5.5 nm that was used as a support material for catalysis 29 . As previously stated, defluoridation has been accomplished using a variety of techniques and adsorbents. However, based on review of the literature, we are unaware of any reports of silica extracted from pumice rock and then modified with iron for fluoride removal from water. Therefore in this study, silica-based defluoridation sorbent was prepared by isolating silica particles from pumice rock via alkaline leaching then its surface modified with Fe 3+ (hard acid) to increase selectivity towards F − (hard base), and used to evaluate fluoride removal from water. Batch experiments were used to evaluate the kinetics and isotherm of fluoride adsorption, as well as the effects of pH, contact time, dosage and initial fluoride concentration on fluoride removal. The efficacy of the adsorbent was then assessed using borehole water samples.

Materials and methods
Study area and sample collection. With the assistance of a geologist, approximately 5 kg of pumice rock was collected at random in a clean well-label polythene sampling bag from Paka volcano in Baringo County, Kenya (36° 10′ 59″ E and 0° 55′ 14″ N).
Chemicals and standards. The following analytical grade chemicals and reagents were used in this study: HCl, NaOH, H 2 SO 4 , NaF, pH buffers and total ionic strength adjustment buffer (TISAB) bought from Sigma-Aldrich through Kobian Scientific Limited in Kenya and used without further purification. Furthermore, deionized water was used throughout.
Fluoride analysis. Fluoride levels were assessed using an ion-selective electrode (ISE) model (Elit 9801) in accordance with the American Public Health Association's standard protocol 30 .
Pretreatment of pumice rock. Pumice rock samples were thoroughly cleaned with deionized water, dried and crushed. The ground powder was then passed through a 180 μm sieve to obtain uniform particle sizes, which were subsequently activated in a muffle furnace model (STT-1200C-3.5-12) at 500 °C for 3 h.
Extraction of silica particles (SPs) from pumice rock. Silica particles were recovered in triplicate from pumice rock using a low-temperature alkaline leaching protocol described by Mourhly et al. 29 . In brief, 10 g of ground pumice was refluxed with 300 mL of 3 M NaOH at 100 °C for 4 h while stirring at 300 rpm to dissolve the silicate and form a Na 2 SiO 3 solution 31 . To recover Na 2 SiO 3 , the slurry was filtered with ashless filter paper (Whatman No 41). The filtrate was then acidified with drops of 5 M H 2 SO 4 to pH 7 while vigorously stirring to form silica gel 32 . Prior to filtration and thorough washing, the silica gel was aged overnight. The silica gel was then dried overnight at 110 °C before being refluxed with 1 M HCl for 3 h at 100 °C to remove any soluble minerals such as Fe, Al, Ca, and Mg. The suspension was filtered, thoroughly washed, and dried overnight at 110 °C. The final product was activated for 3 h in a muffle furnace at 550 °C to yield very fine white silica particles (SPs) powder.
Silica yield. The amount of silica recovered from from pumice rock was calculated using Eq. (1) 33 .
The average weight of silica in pumice rock is the product of the average weight of pumice rock used in the extraction and the average percent SiO 2 obtained from XRF analysis.

Modification of SPs with iron.
The silica particles were iron-coated according to the methodology established by Ref. 34 . In a 50 mL solution containing 1 g of Fe(NO 3 ) 3 ·9H 2 O, 10 g of silica particles were dissolved. The pH of the solution was adjusted to 7 with 0.5 M NaOH and then stirred at room temperature for 1 h. The mixture was centrifuged, and the resulting particles were thoroughly washed and dried overnight at 105 °C. Finally, the Fe-coated silica particles (FCSPs) were activated in a muffle furnace for 6 h at 500 °C before being stored in a clean plastic container.
Characterization. The bulk chemical composition of pumice rock, silica particles (SPs) and Fe-coated silica particles (FCSPs) were determined using X-ray fluorescence (XRF) spectrophotometer (Rigaku ZSX Primus II). For phase identification, an X-Ray diffractometer (XRD) model (Rigaku MiniFlex II) with copper radiation (CuKα = 1.5418 Å) operating at 15 mA and 30 kV was used to record diffractograms between 2θ of 3° and 50°, with a step size of 0.02 at 2 s per step. The functional groups were identified using a Shimadzu fourier transform infrared (FTIR) spectrophotometer (IRAffinity-1S) in attenuated total reflectance mode, with spectra recorded between 4000 and 400 cm −1 with a resolution of 4 cm −1 . The morphology of the silica particles was examined using a scanning electron microscope (JCM-7000-JEOL).
Adsorption studies. A batch experiment was conducted at room temperature to determine the optimal pH, sorbent dose, contact time and initial fluoride concentration for fluoride removal using FCSPs. Equations (2) (1) SPs yield(%) = Average weight of exctracted SNPs (g) Average weight of silica in pumice rock (g) × 100 www.nature.com/scientificreports/ and (3) were used to calculate the amount of fluoride adsorbed at equilibrium ( q e ) and the percentage of fluoride removed 35 .
where M (g) is the sorbent mass, V (L) is the volume of the solution, q e (mg/g) is the amount of fluoride adsorbed at equilibrium, C o and C e (mg/L) is the initial and equilibrium fluoride concentrations, respectively 36 .
Optimization of pH. The effect of pH on fluoride removal was investigated using 1.5 g of FCSPs and 250 mL of a 20 mg/L fluoride solution. The pH was varied from 2 to 10 using 0.05 M HCl and 0.05 M NaOH. The solutions were stirred at room temperature for 90 min before being filtered with Whatman No. 42 filter paper. The residual fluoride concentration in the filtrate was then determined using an ion-selective electrode (ISE).
Optimization of sorbent dose. The effect of sorbent dose on defluoridation was evaluated by equilibrating various sorbent doses (0.2-2.5 g) with 250 mL of a 20 mg/L fluoride solution at the optimum pH of 6. The solutions were stirred at room temperature for 90 min before being filtered with Whatman No. 42 filter paper. The residual fluoride concentration in the filtrate was then determined using an ISE.
Optimization of contact time. The adsorption capacity of FCSPs as a function of time was studied using 250 mL of a 20 mg/L initial fluoride solution at optimal pH (6) and sorbent dose (1 g) by varying contact time from 5 to 90 min. After stirring the solutions for a predetermined time at room temperature, they were left to settle for 2 min before being filtered with Whatman No. 42 filter paper. The concentration of residual fluoride in the filtrates was then determined using an ISE.
Optimization of initial fluoride concentration. The effect of initial fluoride concentration on defluoridation was investigated using optimal pH (6), dose (1 g) and contact time (45 min), and the initial fluoride concentration was varied from 2 to 60 mg/L. After stirring the solutions for 45 min at room temperature, they were left to settle for 2 min before being filtered with Whatman No. 42 filter paper. The concentration of residual fluoride in the filtrates was then determined using an ISE.
Adsorption isotherms. In this study, the Langmuir and Freundlich models were used to interpret adsorption data 37 . Freundlich model usually describes a heterogeneous system based on assumption that sorption takes place in several sites and as the number of adsorbates increases, the surface binding energy decreases exponentially which implies a multilayer formation. The model is expressed by Eqs. (4) and (5) 38 .
where C e (mg/L) is the concentration of fluoride at equilibrium. q e (mg/g) is the amount of fluoride adsorbed per unit mass of adsorbent. K F (mg/g) is the Freundlich coefficient indicating sorbent sorption capacity. 1/n (unitless) is the constant, signifying surface heterogeneity or adsorption intensity with a value ranging from 0.1 to 1 39 . The Langmuir model essentially describes a monolayer type of adsorption and it is expressed by Eq. (6) 40 .
where q e (mg/g) is the amount of fluoride adsorbed per unit mass of adsorbent. C e (mg/L) is the concentration of fluoride at equilibrium. q max (mg/g) is the maximum monolayer adsorption capacity. K L is the Langmuir constant depicting adsorbent affinity towards the adsorbate. The value of the separation factor (R L ) expressed by Eq. (7) indicates the suitability of the Langmuir model to fit the data: The value of R L indicates whether the isotherm is favourable (0 < R L < 1), unfavourable (R L > 1), linear (R L = 1) or irrevesible (R L = 0).
Kinetics models. Pseudo-first-order and pseudo-second-order kinetics models were used to investigate the rate and mechanism of defluoridation 39 . Pseudo-first-order is ideal for simple sorption processes in which saturation occurs in 20-30 min 41 and it is expressed by Eq. (8) 42,43 .
(4) q e = K F C 1/n e (Non-linear form) where q t and q e are fluoride concentrations (mg/g) at a time (t) and equilibrium, respectively, and K 1 (min −1 ) denotes the rate constant. Plotting log q e − q t versus time yields a straight line and the values for q e and K 1 are determined from the intercept and slope, respectively 43 .

Removal of fluoride from real water samples. Borehole water samples collected from Tiaty in Baringo
County, Kenya, were utilized to evaluate the efficiency of FCSPs in defluoridation. Apart from filtration with Whatman No. 42 filter paper, the samples were used without any other treatments. The initial fluoride levels were determined, then defluoridation was performed using the optimal sorbent dose (1 g) and contact time (45 min). The residual fluoride levels were then determined.
Regeneration studies. A batch desorption experiment was done according to Rafigue and colleagues with slight modification to evaluate the ability of adsorbents to be regenerated and recycled 13 . Five consecutive cycles of adsorption-desorption experiments were done using 0.1 M NaOH as a desorbing agent. The spent sorbent was soaked in NaOH for 2 h, washed with deionized water until the washed water pH was 7 then dried in an oven at 90 °C for 4 h. A fluoride solution of 20 mg/L initial concentration was used with optimum sorbent dose (1 g) and contact time (45 min).

Results and discussion
Silica yield. From an average of 9.978 g of pumice rock used, 5.296 g silica particles (SPs) were recovered.
According to Eq. (1). This implies that silica particle extraction from pumice rock via alkaline leaching is viable. Previous research has revealed a similar outcome 29 .
Characterization. XRF analysis. XRD analysis. An X-ray diffractometer was used to identify the minerals present in pumice rock, SPs, and FCSPs. According to the diffractograms in Fig. 1, pumice rock comprises crystalline phase minerals, primarily anorthoclase, feldspar and quartz 45 .
The extracted silica particles exhibited a single broad peak from 2θ of 15° to 30°, centered at 2θ of 22°, which is a distinctive feature of amorphous silica 46 . The absence of crystalline peaks previously observed in pumice   Fig. 3 demonstrate that the extracted silica particles were spherical and agglomerated together to form clusters. This denotes amorphous silica and is consistent with XRD data (Fig. 1). A similar finding was made when silica particles were extracted from pumice rock 29 .

SEM analysis. The SEM micrographs in
Adsorption studies. Effect of pH. The effect of pH on the removal of fluoride from water by FCSPs was investigated, and the results are shown in Fig. 4.
As illustrated in Fig. 4, fluoride sorption rose from 41.6% at pH 2 to an optimum of 83.4% at pH 6, and then decreased as pH increased further. The pH of the solution is an important parameter in the adsorption process since it regulates the sorbent's surface charge and the degree of ionization of the adsorbate 49 . The reduced sorption www.nature.com/scientificreports/ capacity at low pH could be due to the generation of weakly ionizing hydrofluoric acid, which decreases the availability of free fluoride ions for electrostatic interactions with Fe 3+ on the sorbent surface 7,49 . The declines in sorption capacity from 83.4 to 19.6% with pH rises from 6 to 10 may be attributed to competition for the active site on the adsorbent between OHand Fions due to their similar ionic sizes and charges 24 . Furthermore, the decrease in sorption capacity at alkaline pH can be due to the electrostatic repulsion of fluoride ions with the negatively charged adsorbent surface 9 .
Effect of sorbent dose. The effect of sorbent dose on defluoridation was investigated by varying the sorbent dosage from 0.2 to 2.5 g at the optimal pH of 6. Figure 5 depicts the outcomes.
The results show that increasing the sorbent dose from 0.2 to 1.0 g increases fluoride removal from 56.4 to 85.8%. According to Nehra and co-workers, this is most likely owing to the availability of a greater number of unoccupied active sorption sites and the existence of more surface areas for sorption 50 . However, increasing the sorbent dose from 1.0 to 2.5 g has no discernible effect on sorption capacity, presumably due to sorbent agglomeration or overlap, which reduces the availability of active sorption sites at higher sorbent doses 51 . In earlier studies, most adsorbents showed a similar trend 14,52 .
Effect of contact time. The effect of contact time on fluoride removal was studied by varying contact time from 5 to 90 min using optimum pH (6) and sorbent dose (1 g). Figure 6 depicts the results.
Fluoride sorption increased rapidly in the beginning, from 49.2 to 84.5% at 5 and 45 min (Fig. 6). The presence of a higher number of vacant active sites and a fluoride concentration gradient may be responsible for the initial high fluoride sorption rate 49 . After 45 min, there were negligible changes in fluoride uptake, presumably due to a decrease in the number of active sites and fluoride concentration 14 .   Figure 7 depicts the outcomes. Fluoride absorption is greater when the initial fluoride concentration is lower than when the initial fluoride concentration is higher (Fig. 7). This means that the sorbent's capability diminishes as the initial fluoride concentrations rise. This could be ascribed to sorbent active site saturation as a result of a larger fluoride-tosorbent active site ratio 53 . Previous research has also shown that as the initial fluoride concentration increases, the sorbent's fluoride removal ability diminishes 41,54,55 . Adsorption isotherms. The Freundlich and Langmuir models were used to interpret the data from adsorption experiment. The plots are presented in Figs. 8 and 9, respectively. Table 2 shows that the experimental data fit better to the Freundlich isotherm model (R 2 = 0.989) than the Langmuir isotherm (R 2 = 0.941). The values of 1/n (0.419) between 0.1 and 1.0 and n (2.384) between 1 and 10 confirmed the high bond strength between the adsorbate and adsorbent, as well as the heterogeneous nature of the adsorbent surface. Furthermore, the low value of 1/n indicates the heterogeneity of the adsorbent surface 13 . The small value of the Langmuir constant (K L ), 0.277 L/mg, implies a low heat of adsorption 56 . The R L value of  (Table 2), which is between 0 and 1, indicates favorable experimental conditions for sorption. According to the Langmuir model, q max is 8.913 mg/g ( Table 2).

Kinetics of defluoridation.
The rate as well as mechanism of defluoridation was evaluated using pseudofirst-order and pseudo-second-order kinetics models. The plots are presented in Figs. 10 and 11, respectively. The linear regression plots show that the experimental data fit best to the pseudo-second-order model, which has a higher correlation coefficient of R 2 = 0.992 (Table 3), than the pseudo-first-order model (R 2 = 0.988).
The fit of this data to a pseudo-second-order model shows that adsorption occurs via chemisorption caused by electrostatic attractions or, more likely, ion exchange processes 54,57 . These findings are consistent with the majority of previous studies on fluoride removal using various adsorbents, as shown in Table 4.

Application of FCSPs to real water samples. Water samples collected from Tiaty Baringo County in
Kenya were utilized to examine the efficacy of FCSPs in defluoridation; the findings are displayed in Fig. 12.
The FCSPs adsorb a reasonable amount of fluoride from water, up to the WHO criterion of 1.5 mg/L 61 . However, the percent fluoride removal was lower than what could be obtained using the model solution, which is ascribed to competition for the sorbent active sites with other potential anions commonly found in groundwater such as PO 4 3   www.nature.com/scientificreports/ Regeneration studies. Five adsorption-desorption cycles were performed to assess the adsorbent's ability to be regenerated and reused. The adsorption efficiency decreased with the number of cycles, but not significantly (Fig. 13). This implies that the adsorbent can be recycled several times without losing its efficiency, which is an important factor to consider when choosing an adsorbent.