Defluoridation of calcium-rich groundwater using iron oxide nanoparticles

People drink groundwater containing substantial levels of fluoride (Chandrajith et al. 2012; Weragoda & Kawakami 2017), exceeding that recommended by the World Health Organisation (1.5 mg/L) (WHO 2017). Fluoride and its interactions with some other ions have been implicated as trigger factors of the ‘chronic kidney disease of unknown etiology’ (CKDu) prevalent in Sri Lanka (Wasana et al. 2016; Dissanayake & Chandrajith 2017; Dharma-wardana 2018). Over-extraction of groundwater, particularly during dry spells in tropical climates is highly likely to result in elevated fluoride and calcium levels. High calcium concentrations instigate a profound synergistic effect with fluoride on CKDu (Wasana et al. 2016). Fluoride exerts direct and calcium dose-dependent cytotoxic effects on human renal proximal tubular epithelial cells, and extracellular calcium chelation markedly attenuates fluoride-induced cell–death (Zager & Iwata 1997). The oxidative stress instigated by fluoride on human kidney cells also induces apoptosis and promotes the production of reactive oxygen species (anion superoxide, hydrogen peroxide, peroxynitrite, hydroxyl radicals), resulting in decreased cellular antioxidant defence mechanisms against oxidative kidney cell damage (Barbier et al. 2010).

Since about 2010, many researchers have studied fluoride removal using adsorption (Bhatnagar et al. 2011; Jayarathna et al. 2015), precipitation (Jadhav et al. 2015), ion exchange (Robshaw et al. 2019), electrodialysis (Belkada et al. 2018), coagulation and microfiltration (Da Conceição et al. 2015), and membrane filtration (Camacho et al. 2013). Among these, defluoridation using nanomaterial-based adsorption has gained considerable attention because of its efficiency. In particular, iron oxide nanoparticles (FeONs) have been widely studied (Raul et al. 2012). FeONs often favour fluoride and calcium adsorption, but the effect of calcium on defluoridation using FeONs does not appear to have been reported in the literature. In this study, the defluoridation potential of FeONs in both calcium-free and -rich groundwater was investigated. The study mainly focused on the different removal mechanisms of fluoride and calcium, adsorption isotherms, and kinetic studies of fluoride and calcium, using single- and multi- solute batch experiments.

Nano zero-valent iron was synthesised following the modified method of Petala et al. (2013), using 0.18 M FeCl₃ and 0.94 M NaBH₄. FeONs were synthesised by purging compressed air (3.5 L/min) through the solution containing nano zero-valent iron particles. Finally, they were extracted by centrifuging, rinsing three times with ethanol and oven drying at 80 °C (Gui et al. 2012).

The morphology and elemental composition of the FeONs were analysed, before and after the adsorption experiments, using Environmental Scanning Electron Microscopy (ESEM, Carl Zeiss, EVO 18, Secondary Electron Microscope, Germany) coupled with Energy-Dispersive X-ray Spectroscopy (EDAX, Element EDS system, AMETEK Materials Analysis Division, USA). FeON phase identification was performed by X-ray Powder Diffraction (XRD-D8, ECO, Advance Bruker Diffractometer with filtered Cu Kα radiation, Germany). Fourier transform-infrared spectroscopy (FT-IR, ALPHA Bruker, Germany) was performed in adsorption mode at ambient temperature in the spectral range 500 to 4,000 cm⁻¹ to identify the FeONs' functional groups.

Single-solute batch experiments were carried out to determine the FeON dosage at which maximum fluoride and calcium removal occurred (optimum FeON dosage). Dosage was varied from 0.04 to 7.20 g/L on the basis of preliminary studies. The operating parameters were – sample volume, 100 mL; contact time, 1 hour; initial fluoride and calcium concentrations, 5 and 300 mg/L, respectively; initial pH, 6.5; temperature, 28 °C; and, stirring speed, 150 rpm. The FeON dosage for maximum removal of fluoride and calcium was determined as 1.25 g/L. Subsequently, a series of batch experiments was conducted using the optimum FeON dosage, to determine the fluoride and calcium adsorption isotherm and kinetic behaviour. In these studies, the initial concentrations of fluoride and calcium were varied from 1 to 20 mg/L, and 100 to 500 mg/L, respectively. All other experimental conditions were similar to those used when determining the optimum FeON dosage. The theoretical amounts of fluoride and calcium adsorbed at equilibrium, (Q_e)_ad and the actual amounts removed, (Q_e)_exp, were calculated to determine the extent of chemical precipitation and co-precipitation.

Multi-solute batch experiments were carried out in the presence of both fluoride and calcium to determine the optimum FeON dosage for maximum removal of fluoride and calcium. The operating parameters used were similar to those used in the single-solute batch studies and the optimum FeON dosage was determined as 1.25 g/L.

The differing amounts of fluoride and calcium removed by co-precipitation, chemical precipitation and adsorption were estimated by varying their respective concentration combinations. In CKDu prevalent areas in Sri Lanka, higher concentrations of fluoride and calcium have been reported during the dry season (June to August) than in the wet season. The combinations of fluoride and calcium concentrations considered for the multi-solute studies were based on seasonal variations typically encountered in the dry season. Twenty groundwater samples were collected during the dry season and the fluoride (1.1 to 9.0 mg/L) and calcium (140.0 to 900.0 mg/L) concentrations determined. Based on prevailing conditions in the CKDu affected areas, the experimental fluoride and calcium concentrations were set in the ranges 0.9 to 9.2 and 136.0 to 980.0 mg/L, respectively, and 10 fluoride/calcium concentration combinations were prepared accordingly. The concentration ranges were determined to exceed the natural equivalents slightly.

The first fluoride/calcium combination (0.9 and 136 mg/L, respectively) was prepared without FeONs and kept under quiescent conditions for 24 hours at ambient temperature to allow CaF₂ precipitation. After 24 hours, the mixture was filtered (Whatman^TM 1003-110 Grade 3, 6 μm) and the residual fluoride concentration in the filtrate determined. FeONs (1.25 g) were subsequently added to 100 mL of the filtrate, and the mixture stirred continuously (150 rpm) at ambient temperature. The solution's residual fluoride and calcium concentrations were determined after one-hour contact time. The mixture's pH was monitored continuously for 72 hours. It always exceeded 5, confirming that no CaF₂ dissolution occurred. The same procedure was followed for all other fluoride/calcium combination solutions.

For different fluoride/calcium combinations without FeON, fluoride supersaturation levels were estimated and compared with the respective residual fluoride concentrations, to determine whether or not chemical precipitation had occurred, and estimate the quantities of CaF₂ precipitated. The solubility product (K_sp) of CaF₂ [K_sp = (Ca²⁺) (F⁻)²], which is 3.5 × 10⁻¹¹ (mol/L)² at 25 °C (Nasr et al. 2014), was used to determine fluoride supersaturation levels for different calcium concentrations. XRD and ESEM-EDX analyses were performed to confirm CaF₂ precipitation, and the quantities used to estimate the corresponding amounts of fluoride and calcium precipitated. The amounts of fluoride and calcium adsorbed onto FeONs were estimated assuming monolayer adsorption phenomena. As the amounts of fluoride and calcium both adsorbed and precipitated were known, the amounts co-precipitated were estimated from the total amounts of these species removed from solution.

Residual fluoride concentrations were analysed by ion chromatography (Metrohm, 930 Compact IC Flex, Switzerland) with a mobile phase of 3.2 mmol/L Na₂CO₃ + 1 mmol/L NaHCO₃, and a flow rate of 0.7 m/L/min. Residual calcium concentrations were determined using the EDTA titrimetric method based on USEPA Method No. 130.2 (USEPA 1983). Deionised water (18.2 Ω/cm) from a deioniser (Smart Plus-N, Heal Force, China) was used in all experiments.

Fluoride and calcium supersaturation are a prerequisite for CaF₂ precipitation (Nath & Dutta 2010). Thus, fluoride removal depends on both the calcium concentration and the initial fluoride concentration, implying lower likelihood of precipitation for low initial fluoride concentrations.

Figure 1(a) shows the XRD spectrum of synthesised FeONs, exhibiting peaks at 30.4°, 35.8°, 43.7°, 57.6°, and 63.1°, which correspond to diffraction of the (220), (311), (400), (511), and (440) planes, respectively, of magnetite nanoparticles. Others have made similar observations for XRD characterisation (Lin et al. 2005; Xu et al. 2007; Wu et al. 2011; Raul et al. 2012; Mohseni-Bandpi et al. 2015; Shukla et al. 2015). The average crystallite size obtained for this study was 46.3 nm, confirming that a significant fraction of FeONs was within the nano-scale range. XRD analysis [Figure 1(b)] of CaF₂ shows peaks at 28.2° (111), 46.9° (220), 55.7° (311), 68.6° (400) and 75.8° (331), confirming the formation of CaF₂ precipitate (Pandurangappa et al. 2010; Tahvildari et al. 2012) in the binary mix, with the average particle size 96.3 nm.

The ESEM image of synthesised FeONs (Figure 2(a)) shows the presence of interlinked flake-like aggregates, indicating agglomeration, possibly because of magnetic interaction between particles. The EDX analysis (Figure 2(b)) shows that the surface was composed of Fe and O with no impurities, and the narrow diffraction peaks confirm the crystalline structure. Similar observations on the agglomeration and crystalline structure formation of FeONs are reported by Lin et al. (2005) and Azzam et al. (2016). ESEM analysis of FeONs after fluoride adsorption in single-solute batch-studies shows spherical agglomerated particles with sharp edges (Figure 3(a)) and EDX analysis confirms the presence of fluoride adsorbed on FeON surfaces (Figure 3(b)). Figure 4(a) depicts the morphology of the FeON surface after multi-solute batch-studies, where both fluoride and calcium adsorbed and co-precipitated, which are visible as fluffy and irregular shapes. A magnified ESEM image (nm-scale) shows a nodular-like pattern on the FeON surface, which is ascribed to calcium co-precipitation with FeONs (Figure 4(b)). EDX analysis confirmed the presence of both fluoride and calcium on the FeON surface (Figure 4(c)).

ESEM-EDX analysis of CaF₂ (Figure 5(a) and 5(b)) shows that the CaF₂ formed was agglomerated and porous, with polycrystalline nanoparticles. The larger particles exhibited different, spherical, surface perturbations (Pandurangappa et al. 2010).

The FT-IR spectrum of FeONs shows both the specific peaks reflecting the surface functional groups and the surface's complex nature. Figure 6 shows the FT-IR spectrum of FeONs (a) before adsorption, (b) after fluoride adsorption (single-solute studies), and (c) after fluoride and calcium adsorption (multi-solute studies). Before adsorption, there were broad bands around 1,033, 1,632, and 3,417 cm⁻¹, attributed to N–H and O–H bending, and O–H stretching vibrations, respectively (Sharma & Jeevanandam 2013). N–H functional groups are indicated because of the use of NH₄OH solution to keep the pH between 10 and 11 during FeON synthesis. The band at 1,434 cm⁻¹, assigned to the symmetric and asymmetric bending of the C–H bond, may be ascribed to the ethanol added for FeON synthesis and storage (Qiu et al. 2011; Ayob & Abdullah 2012). The broad peak at 928 cm⁻¹ arises from the bending vibrations of O–Fe–O groups (Jayarathna et al. 2015), and the narrow band around 575 cm⁻¹ represents the Fe–O vibrations of magnetite nanoparticles (Xu et al. 2018).

After fluoride adsorption, the functional groups on the FeONs' surface had changed, and the broad peak around 3,000 to 4,000 cm⁻¹ had weakened and shifted. The hydrogen-bonding strength was also diminished, and hydroxyl groups became much freer from their initial hydrogen-bonding character (Jayarathna et al. 2015), so that the new, isolated, surface –OH peak appeared around 3,824 cm⁻¹. The band at 1,632 cm⁻¹ was shifted to 1,631 cm⁻¹, which was attributed to the presence of –OH groups even after fluoride adsorption onto FeONs (Jayarathna et al. 2015; Mohseni–Bandpi et al. 2015). The intensity of the band at 1,383 cm⁻¹ was increased sharply, perhaps representing the characteristic peak of fluoride adsorption (Patnaik et al. 2016; Liu et al. 2018). The results generally indicate that the surface hydroxyl groups play a significant role in fluoride adsorption.

After fluoride and calcium adsorption, the broad peak at 3,417 cm⁻¹ was decreased and shifted, with new peaks appearing at 3,406 and 3,205 cm⁻¹, attributed to the symmetric and asymmetric vibration of Fe–F and Fe–O–Ca–F bonds, respectively. A new peak appeared at 3,804 cm⁻¹, and the peak at 1,632 cm⁻¹ was shifted to 1,630 cm⁻¹, indicating the presence of –OH groups after fluoride and calcium adsorption. In multi-solute adsorption, the sharp peak at 1,383 cm⁻¹, attributed to Fe–F bonding, was decreased and shifted, and a broad peak appeared assigned to the asymmetric F–Ca–O–Fe bridge. This might arise because calcium occupies the active sites available for fluoride, showing a new peak around 500 cm⁻¹, ascribed to Ca–O and Ca = O bonds (Galván-Ruiz et al. 2009) formed due to calcium adsorption and co-precipitation. The precipitation of CaF₂ in the multi-solute batch experiments could not be verified with FT-IR, as CaF₂'s Ca–F stretching vibration occurs at 443 cm⁻¹ (Tahvildari et al. 2012) – i.e., beyond the analytical range in this study.

The possible mechanism of fluoride and calcium removal in single-solute batch-studies was identified as adsorption by FeONs. Therefore, attributes affecting fluoride and calcium adsorption by FeONs were studied.

FeON dosage was varied from 0.04 to 7.20 g/L (Figure 7) to study its effect on fluoride and calcium adsorption. Fluoride adsorption showed little variation at FeON dosages exceeding 1.25 g/L, when it was 74%. Maximum calcium removal (39%) was observed at 1.25 g-FeONs/L with minor variation beyond that. For every FeON dosage, the total dissolved iron (Fe) concentration remaining after one-hour contact time with fluoride or calcium was measured by ion chromatography-mass spectrometry (ICP-MS) (Model: Agilent 7900 ICP-MS system, Japan, USA). The data showed that dissolution of Fe from the FeONs was less than 0.001 mg/L when dosage was increased, and that minute amounts of Fe dissolution did not affect FeON adsorption capacity. Increasing FeON dosage did, however, increase particle density per unit volume, increasing interaction between FeON particles and leading to agglomeration. This decreased the FeONs' adsorption capacity by decreasing the number of active sorption sites. Similar results have been reported for defluoridation with different adsorbent dosages of magnetite-chitosan composite (Mohseni-Bandpi et al. 2015).

The effect of initial concentration on fluoride adsorption was studied for the range 1 to 20 mg-F/L, over which fluoride adsorption increased from 0.04 to 5.3 mg/g. Calcium adsorption increased from 0.04 to 3.34 mg/g, at constant FeON dosage, for the initial calcium concentration range 100 to 500 mg-Ca/L.

The equilibrium isotherm and kinetic model parameters for fluoride and calcium adsorption by FeONs are presented in Table 1. The R² values for the Langmuir isotherm adsorption model for both fluoride and calcium are 0.99, the high values indicating monolayer adsorption onto a surface with a finite number of active adsorption sites. The maximum monolayer adsorption capacities (Q_m) of FeONs based on the Langmuir model were 28.98 mg-F/g and 1.13 mg-Ca/g. The equilibrium parameter (R_L) was between 0.37 and 0.48 for fluoride, and 0.52 and 0.75 for calcium, indicating that the monolayer adsorption mechanism was favoured. The Freundlich constants (n) for fluoride and calcium were 1.35 and 1.79, respectively, indicating that F and Ca monolayer adsorption were moderately uncertain and weak, respectively. In single-solute batch experiments, monolayer adsorption was, therefore, preferred for both species. The R² values for the BET adsorption isotherm for fluoride and calcium were 0.85 and 0.91, respectively, implying that the experimental data do not follow the BET isotherm assumption of multilayer formation during adsorption by FeONs. The equilibrium binding constant and adsorption energy variation for the Temkin model are given in Table 1. The Temkin model's R² values for F and Ca – 0.87 and 0.91, respectively – show that the model does not fit the experimental data well. The Dubinin-Radushkevich model's R² values for F and Ca were 0.93 and 0.75, respectively, showing that this model also failed to fit the experimental data well.

The adsorption kinetics of fluoride onto FeONs are described well by a pseudo-second-order model (R² = 0.99 to 1; Table 2). Hence, the rate-limiting factor of fluoride and calcium adsorption by FeONs is chemisorption, confirming covalent bonding by sharing or exchange of electrons between adsorbate and adsorbent (Arshadi et al. 2014). The results show that the experimental Q_e values (Q_e(Exp)) derived from the mass balance equation (Equation (1)) are approximately equal to the Q_e values calculated (Q_e(Cal)) with the pseudo-second-order equation (Equation (12)). As the experimental and calculated Q_e values are approximately equal, in single-solute batch-studies for both fluoride and calcium, adsorption was pronounced and removal due to co-precipitation could not be observed.

The possible fluoride removal mechanisms in multi-solute batch-studies were identified as chemical precipitation as CaF₂ and adsorption by FeONs. For calcium removal they were identified as co-precipitation with oxide and hydroxide functional groups of FeONs, chemical precipitation as CaF₂ and adsorption by FeONs. Different physicochemical conditions in the mixture governed these mechanisms, so the potential for adsorption, co-precipitation and chemical precipitation was determined separately for different fluoride and calcium combinations.

The monolayer adsorption capacity of FeONs for fluoride and calcium removal was determined using constant 1.25 g-FeON/L dosage with different concentration combinations of fluoride and calcium (Figure 8). The fluoride and calcium concentrations were varied between 1.12 and 19.3 mg-F/L, and 136 and 980 mg-Ca/L, respectively. The monolayer adsorption capacity of fluoride for the corresponding combinations of fluoride and calcium, was between 0.07 and 1.63 mg for a 100 mL solution. For calcium it was between 0.29 and 0.49 mg (Figure 9). The results showed that for the combination containing 1.12 mg-F/L and 136 mg-Ca/L, total fluoride removal by FeONs occurred only through monolayer adsorption, with limited CaF₂ precipitation due to the low fluoride concentration. For other fluoride and calcium combinations, total fluoride removal was attributed to the formation of CaF₂ through chemical precipitation apart from some monolayer adsorption through chemisorption.

Calcium and fluoride tend to form insoluble CaF₂ (K_sp = 3.45 × 10⁻¹¹) via a stoichiometric reaction in the pH range 4.17 to 10 (Nasr et al. 2014). The solubility of CaF₂ is independent of pH above 4.17 but increases significantly at lower pH levels.

Measurement of residual fluoride concentration temporal variations showed that much of the CaF₂ formed within an hour of mixing provided that the residual fluoride and calcium concentrations exceeded supersaturation levels. Similar observations were made by Nasr et al. (2014). Fluoride removal by chemical precipitation was in the range 0.04 to 0.13 mg for 100 mL of solution with increasing F and Ca levels (Figure 8). Therefore, total fluoride removal due to adsorption and chemical precipitation was determined as between 0.11 and 1.76 mg from corresponding initial fluoride concentrations of between 0.11 and 1.93 mg. The calcium content removed by chemical precipitation and co-precipitation was between 0.14 and 0.05 mg, and 2.76 and 55.46 mg, respectively, when F and Ca concentrations increased (Figure 9). Total calcium removal due to adsorption, chemical precipitation and co-precipitation was, therefore, in the range 3.20 to 56.00 mg, with initial calcium amounts between 13.60 and 98.00 mg. The difference between adsorption and co-precipitation depends mainly on the geometry of the adsorbate surface (Corey 1981), where co-precipitation is considered a 3-dimensional process leading to higher removal capacity than adsorption alone. The formation of mixed metal hydroxides, which is leading to enhanced metal ion removal at high orders of magnitude has been attributed to co-precipitation over adsorption (Crawford et al. 1993).

Fluoride and calcium removal by FeONs in single-solute batch-studies was ascribed only to adsorption, and calcium co-precipitation was not observed. Adsorption studies suggested that both fluoride and calcium follow the Langmuir isotherm model and pseudo-second-order kinetics, indicating monolayer adsorption and chemisorption, respectively. The maximum monolayer adsorption capacities of FeONs were 28.98 and 1.13 mg/g for fluoride and calcium, respectively. In multi-solute batch-studies, fluoride removal was attributed to both adsorption and chemical precipitation as CaF₂ in almost similar magnitudes. For calcium removal in multi-solute batch-studies, co-precipitation, precipitation, and adsorption were the removal mechanisms, with co-precipitation the predominant contributor.

The study confirmed that FeONs are effective in removing fluoride by adsorption, in the absence or presence of calcium. Irrespective of the calcium concentration in raw water, the use of FeONs is worthwhile to lessen the combined effect of fluoride and calcium on CKDu.

No potential conflicts of interest were reported by the authors.

The Senate Research Committee (SRC) Grant of the University of Moratuwa (Grant No. SRC/ND/15/01) and the National Research Council (NRC), Sri Lanka (Grant No. 15-056) supported this research study.