Removal of fluoride from aqueous solution by porous Vaterite calcium carbonate nanoparticles

This study is based on a simple, low-cost and a novel approach towards the removal of excess fluoride ions from aqueous solution by absorbing fluoride on porous vaterite calcium carbonate nanoparticles (PVCCNPs) synthesised using ethylene glycol-water soft template method. SEM images clearly show the porous nature of aggregated nanoparticles present in the dry powder. Physicochemical properties of synthesised PVCCNP and fluoride on PVCCNP was characterised further by FTIR, XRD, XRF, EDX, and TGA-DTG. Fluoride removal by PVCCNPs from 100.00 ml of 10.0 mg l−1 NaF solution with 0.500 g of PVCCNPs, determined using a fluoride ion-selective electrode, indicates that around 90% removal is achieved within 1 h thus reducing the level to desired 1 ppm. The pseudo-second order kinetic model has a better fit to describe the adsorption of fluoride on PVCCNP than pseudo-first order model. The Langmuir isotherm model is more appropriate to describe the equilibrium behaviour of the adsorption process, than the Freundlich model. Given that the value of n (Freundlich constant) is greater than 1 (3.07) and RL value is in the range of 0 < RL <1 (0.014–0.024) implies that the adsorption process is spontaneous and fluoride ions are favourably adsorbed on PVCCNPs. Langmuir model shows that the maximum adsorption capacity of fluoride is 1.956 mg g−1. Excess fluoride in drinking waters causes several severe ill-health effects and filter media based on these nanoparticles can be used to remove fluoride down to safe and required levels to tackle these health problems. As such, PVCCNPs-based filter can be designed to remove fluoride in drinking waters. This may be a way for controlling fluorosis and many other diseases associated with excess fluoride present in drinking waters.


Introduction
Fluorine is one of the most reactive, lightest and the most electronegative halogen on the Earth. Fluorine occurs in nature as fluoride (F − ) ions and is found in air (as gases or particulates), in water (present in the form of hydrated F − ions), in soils and in living organisms [1]. Fluorine compounds are used in the metallurgical industry, mainly in the production of aluminium and manufacturing of glass [2], ceramics [3] and phosphate fertilizers [4]. Fluorine is present in 32 minerals as fluorides with or without carbonate, silicate, and phosphate and hydroxyl anions of alkali, alkaline earth, aluminium, bismuth, manganese or rare-earth cations [5]. Exchange of OH − for F − and vice versa in minerals is quite possible due to their similarities in size and charge and hence weathering of fluoride containing minerals add fluoride ions to water resources particularly under alkaline conditions [5].
Main sources of fluoride are drinking waters and a relatively small amount of fluoride comes from food items such as sea fish and tea [6]. Once absorbed, fluoride is quickly distributed throughout the body via the blood circulation [6]. In general, the level of fluoride content in the plasma is affected by the rates of bone accretion and dissolution, and by the renal clearance rate of fluoride ions. Renal excretion is the major pathway of fluoride removal from the human body. There is no tubular secretion of fluoride. Generally fluoride ions are filtered from the plasma by the glomerulus and then partially reabsorbed. There are several factors which can influence urinary fluoride excretion including urinary pH, urinary flow, and glomerular filtration rate [7]. Fluoride ions are not accumulated in soft tissues [8,9] generally except when there is a higher concentration because of partial re-absorption. Fluoride has both beneficial and detrimental effects on human health, which depending upon the concentration of fluoride present in drinking waters [10]. According to the WHO standards, the recommended desirable fluoride limit in drinking water is 1.0 ppm and maximum permissible limit is 1.5 ppm, though this value may vary slightly depending upon temperature and climate [11]. Recently, the recommended level of fluoride present in drinking water has been brought down to 0.7 ppm [12][13][14].
As such, removal of fluoride to the desired levels from drinking waters is an important precaution to be taken to prevent many diseases related to excess fluoride. There are several methods used for the removal of fluoride from contaminated waters, which include coagulation-precipitation, adsorption [15,16], ion-exchange [17], electrocoagulation, membrane filtration, membrane separation [18], electrolysis, and so on. Among these methods, adsorption is the cheapest and easiest way to remove fluoride. Some of the adsorbents used to remove fluoride in water include activated alumina [19], activated carbon [20], activated alumina coated silica gel [21], calcite [22], activated saw dust [23], activated coconut shell powder [24], coffee husk [25], rice husk [26], tricalcium phosphate [27], bone charcoal [28], activated soil sorbent [29], defluoron and so on. So far, up to 90% of fluoride removal has been achieved from the adsorption technique, and the adsorbent can be regenerated by flushing with a hydroxide solution. When the nanoparticles are used the sludge generated is minimal since nanoparticles are expected to have very large specific adsorption capacity. Calcium ions have good affinity to fluoride ions and some calcium-based nanoparticles have been investigated for the removal fluoride ions from water [30].
As such, we report, in this paper, the synthesis of highly porous nanoparticles of calcium carbonate predominantly in its spherical vaterite form with some calcite nanoparticles [31] and their application in defluridation of drinking water.

Materials and methods
All the chemicals were purchased from Sigma Aldrich and they were of analytical grade and used without further purification.

Synthesis of porous vaterite calcium carbonate nanoparticles (PVCCNPs)
A solution of Ca(CH 3 COO) 2 (25.00 ml, 0.50 M), H 2 O (10.00 ml) and ethylene glycol (EG) (25.00 ml) labelled as solution A and a solution of NaHCO 3 (25.00 ml, 0.50 M), H 2 O (10.00 ml) and EG (25.00 ml) labelled as solution B were prepared. The solution A was added to solution B in drop-wise manner using a dropping funnel and mixed slowly under constant stirring. After 24 h, the resultant precipitate was filtered and washed first with ethanol and then with distilled water. Product obtained was dried, at 100°C, for 4 h, in a vacuum oven. Nanoparticles and other end products obtained were characterized by Laser Light Scattering based Particle Size Analysis (CILAS Particle Size Analyser NANO DS), XRD (Siemens D5000 x-ray powder diffractometer), XRF, FT-IR (Shimadzu IR-Prestige 21 Instrument with the KBr pellet method, SEM (Environmental SEM with EXAS Facilities) studies, TGA analysis(TA Instruments SDTQ600 Thermo-Gravimetric Analyser) The surface area of the PVCCNPs was determined by the novel BET instrument constructed in our research group.

Preparation of total ionic strength adjustment buffer II (TISAB II)
500.00 ml of distilled water, 57.00 ml of glacial acetic acid and 58.000 g of sodium chloride were mixed in a beaker, at 28°C. The mixture was stirred to dissolve and allowed to cool down to room temperature. The pH of the solution was adjusted between 5.0 and 5.5 with 5 N sodium hydroxide solution. The pH-adjusted solution was transferred to a 1000.0 ml volumetric flask and diluted to the mark with distilled water.
2.3. Preparation of fluoride standard solutions 2.210 g of NaF solid was dissolved in 1.00 l deionized water in a volumetric flask to prepare a solution of 1000ppm stock fluoride solution. This stock solution was diluted ten-fold to prepare 100.0 ppm F -(aq) bulk solution. This bulk solution was diluted in respective factors to prepare 10.00, 12.00, 14.00, 15.00, 16.00, 18.00 and 20.00 ppm F − (aq) standard solutions.4.00 ml of TISAB II prepared was added to 4.00 ml standard sample each and mixed well. 5.00 ml of each of these samples was pipetted out and filtered. Then 4.00 ml of filtrate of each sample was pipetted out, separately, and 4.00 ml of TISAB II was added to each of these samples. The Fluoride concentrations of these samples were measured using Ion Selective Electrode which was calibrated using 1.00, 2.00 and 10.00 ppm F − (aq) standard samples.

Fluoride removal by PVCCNP
1000.00 ml of 10.00 ppm fluoride solution was prepared by appropriately diluting 1000 ppm solution. Then 0.500 g of PVCCNPs was added to 100.00 ml of the above solution while stirring and stirring was continued for several hours. At each 1 h intervals, 5.00 ml of the sample was withdrawn and its fluoride concentration was determined by the calibrated fluoride ion-selective electrode. The removal process of Fluoride by PVCCNPs depends on several experimental parameters. These parameters are the initial pH of the aqueous Fluoride solution, PVCCNP dosage, shaking time, settling time and temperature. The optimum values determined separately by changing one parameter and others are kept constant. The all experiments are carried out at normal room temperature (298 K).

Studies on the effect of settling time
In this study, 10.00 ppm Fluoride standard solution was used. As above procedure 0.050 g of PVCCNPs was added to 20.00 ml Fluoride sample and shaken for 1 h. After 30 min settling time, 5.00 ml pipetted out for Fluoride analysis. The procedure was repeated for 60, 90, 120, 150, 180, 240 and 300 min time intervals by keeping other factors constant.

Studies on the effect of PVCCNPs dosage
In this study 10.00 ppm Fluoride standard solution was used. 9 samples of 20.00 ml were separated from this standard solution and 0.050, 0.060, 0.070, 0.080, 0.090, 0.100, 0.125, 0.150 and 0.200 g of PVCCNPs were added to above samples respectively. After 1 h shaking and 2.5 h settling time 5.00 ml from each sample were pipetted out for Fluoride analysis.

Isotherm and kinetics studies
A concentration series of Fluoride from 10.00 ppm to 20.00 ppm was prepared using 100.00 ppm standard Fluoride solution. 0.110 g of PVCCNPs was added in to 20.00 ml to each sample and shake for 1 h. After 2.5 h settling time 5.00 ml from each sample were pipetted out for Fluoride analysis. In this study the Fluoride concentrations of samples were measured by 1:1 dilution with distilled water and resulted measurement was multiplied by 2. The Langmuir and Freundlich, isotherm equations were used to model the equilibrium behaviour of the Fluoride adsorption process. The pseudo-first order kinetic model and pseudo-second order kinetic model were used to analyse the kinetics of the absorption. Figure 1(a) shows a photograph of the prepared. PVCCNPs after vacuum drying. As shown, the sample consists of a fine powder which was dispersed in water using poly(ethylene glycol) (PEG) as the stabilizer. Laser light scattering based particle size analysis reveals that these particles are in a fairly narrow size range in the nanoscale between 5 nm and 50 nm with an average particle size of 11.5 nm. (figure 1(b)).

Characterization of synthesized PVCCNPs
XRF analysis detects only Ca in this sample (figure 2). Since the XRF measures elements with atomic number >13, it is incapable of detecting C and O. The fact that it does not detect any other element it clearly shows that the sample is free of impurity atoms with at least atomic number >13. Figure 3 shows EDAX mapping and elemental analysis of PVCCNPs that were used to remove fluoride from water. As seen in these figures, the sample consists of Ca, C, O and F whose elemental mapping is shown in this order in figure 3 (top). Shown in bottom of the figure are the EDAX spectrum and the part of the SEM image used to get these data. It is interesting to note that Ca, C, O, F atomic percentages are 38.6, 22.6, 28.3 and 10.5, respectively, suggesting considerable conversion of CaCO 3 to CaF 2 possibly on the surface of particles.
FT-IR Spectrum, given in figure 4(a), clearly shows major bands centred at 746, 874, 1088 and 1413 cm −1 wavenumbers. The first three IR-bands correspond to those of vaterite polymorph of calcium carbonate since standard IR bands of vaterite should appear at 745.8, 877 and 1084 cm −1 [32]. As can be seen from XRD in and (224) crystallographic planes, respectively, can be clearly observed. Except for some other peaks such as 29.47°, 47.92°corresponding to calcite polymorph can also be seen due to vaterite to calcite conversion [33]. Application of the Debye-Scherrer equation (equation (1)) gives the average crystallite size to be 23.0 nm which is in the range of that obtained from solution phase particle size analysis.
Where, t is the average crystallite size. K is a dimensionless shape factor, with a value close to unity. The shape factor has a typical value of about 0.9, but varies with the actual shape of the crystallite. l is the X-Ray wavelength, b is the line broadening at half the maximum intensity (FWHM), after subtracting the instrumental line broadening, in radians. q is the Bragg angle.
Morphological studies of the PVCCNPs were done by SEM analysis and an image is shown in figure 5(a) which shows that large number of vacuum dried individual PVCCNPs are aggregated to highly porous spherical large micrometer-sized particles. The fact that particles in suspension contain discrete particles in the size range between 5-50 nm with 11.5 nm average particle size, it is possible that these aggregated particles can be broken down to essentially discrete single particles and they can be stabilized using EG. The SEM image shown in   TGA (shown in figure 6) was carried out on fluoride incorporated PVCCNP samples from room temperature to 1000°C. TGA thermogram shows that a mass loss of ∼15% was observed in the temperature range from 50°C to 150°C, which is due to the removal of moisture physically absorbed in the sample. A massloss of around 20% in the temperature range 200°C-300°C can be attributed to release of structurally bound  water. Combustion of any ethylene glycol molecules trapped within the nanoparticles or adsorbed onto nanoparticles is observed in the temperature range 300°C to 600°C [34]. Weight loss beyond 840°C is due to the decomposition of CaCO 3 releasing CO 2 gas [35].

Use of PVCCNPs for fluoride removal from water
In this study, known amount of PVCCNPs was shaken in a fixed volume of fluoride ion solution of known concentration at a constant temperature (usually 25°C). There are several factors that could affect the fluoride removal in this process. These include the shaking time, time allowed to settle the particles after shaking for a fixed period of time (settling time), pH value of the solution, and initial fluoride concentration or the amount of PVCCNPs used for a fixed volume of fluoride solution with exactly the same initial concentration. Figure 7 shows the percentage removal F − from 20.00 ml of 10.03 ppm concentrated fluoride ion solution by 0.050 g of PVCCNPs at 25°C for shaking times as 10, 20, 30, 40, 50, 60, 70, 80 and 90 min each flowed by 1 h settling time.
Percentage of fluoride removal increases rapidly during first 40 min which reaches a maximum around 35% after about 60 min and levels off at this equilibrium value. Since in this system, the equilibrium has been reached after 60 min of shaking time, the effect of settling time on fluoride removal was studied by keeping shaking time at 1 h, and the results are depicted in figure 8. It shows that the settling time has no drastic effect on fluoride removal. These results indicate that more PVCCNPs are required for further removal of fluoride under these conditions. Shown in figure 9 is the percentage removal of fluoride from 20.00 ml of 9.92 ppm fluoride ion solutions as a function of amount of PVCCNPs. As shown, fluoride removal increase as the amount of PVCCNP is increased at the fixed initial amount of fluoride reaching equilibrium at 95% when 0.100 g or more of  PVCCNPs is used. The maximum fluoride removal capacity can then be calculated to be around 1.96 mg fluoride ions per 1 g of PVCCNPs.

Adsorption Isotherm studies
In this study, all parameters were kept constant by considering results described above. Shaking and settling time were alarmed to 1 h and 2.5 h, respectively, and 0.110 g PVCCNPs were treated with each sample containing 12.20, 14.10, 15.00, 16.10, 18.14 and 20.02 ppm of F − (aq). These concentrations were measured by 1:1 dilution with distilled water to be the linear dynamic calibrated range of the ISE. Data pertinent to initial F − concentration, C 0 in ppm, equilibrium F − concentration, C e in ppm, and specific adsorption by PVCCNPs, q e in mg/g, and the calculated data required to fit to Langmuir and Freundlich Adsorption Isotherm.
Langmuir adsorption isotherm plot was constructed by plotting C e /q e in g/L versus C e in mg l −1 since the linearized form of the Langmuir Adsorption Isotherm is as shown in equation (2) where K and q m are Langmuir constant in L mg −1 and maximum saturated adsorption capacity in mg g −1 .   (table 1). Langmuir isotherm model can be used to predict whether the adsorption is favourable or not [36]. The separation factor R L which is defined by equation (3) is a key factor determining the favourability of the adsorption. If R L >1, it is an unfavorable adsorption, when 0<R L <1 adsorption is favourable, if R L <0 adsorption is irreversible and when R L =1 it is then called linear adsorption.
Within the C 0 range investigated, R L values lie in the range from 0.014 to 0.024 indicating that fluoride ions are favourably adsorbed on PVCCNPs. Linearized form of the Freundlich adsorption isotherm takes the form shown in equation (4) where K F is the Freundlich constant and n is a constant which explains the relationship between adsorption capacity and adsorption intensity, respectively  The value of n signifies the interactions between the adsorbent (PVCCNP) and the ion (F − ), Interactions between the adsorbent and pollutant is high if n value is high. If n=1, it indicates the linear adsorption for all active sites of the adsorbent [37]. The 'n' for the fluoride adsorption to PVCCNP is 3.07 confirming the adsorption process desirable and physical process. Figure 10(b) shows the Freundlich adsorption isotherm plot and table 1 presents the both Langmuir and Freundlich isotherm results. The plot is not very linear (R 2 =0.9211) as that of Langmuir adsorption isotherm plot (R 2 =0.9988) though there is some linearity can be seen. Thus Langmuir model has a better ability to describe the adsorbent isotherm behaviour than Freundlich isotherm. However it indicates the possibility for physisorption on the chemisorbed monolayer of fluoride ions that are already on the active sites of the PVCCNP surfaces. It may also be due to some precipitated CaF 2 on PVCCNPs [38].

Kinetic studies
The kinetic models can be used to determine the mechanisms of adsorption processes. In this study, pseudo-first and pseudo-second order models were used to study the mechanism of fluoride on PVCCNP [39].
Basic equation for Pseudo-first order adsorption: The integrated pseudo-first-order rate equation is written as Basic equation for Pseudo-second order adsorption: By integrating the equation at boundary conditions (t=0 to t=t and q t =0 to q t =q t ) gives Where q e and q t are the amount of adsorbed fluoride at equilibrium and any time t (mg fluoride/g solid material), k 1 (min −1 ) and k 2 (gmg −1 min −1 ) are the equilibrium rate constant of firstand second-order sorption respectively, and t is the shaken time (min). k 1 can be calculated plotting the linear plot of log (q e -q t ) versus t (equation (6)) ( figure 11(a)) A larger adsorption rate constant k 1 usually denotes a quicker adsorption rate. k 2 can be determined by plotting t/q t against t based on (equation 10) ( figure 11(b)). The larger the k 2 value, the slower the adsorption rate [39]. The correlation coefficient (R 2 ) of the pseudo-first order kinetic model for adsorption of fluoride with initial concentrations of 10 ppm is 0.2488 and that for pseudo-second order kinetic model ts 0.9882. The closeness of the R 2 values to 1, indicates that the pseudo-second order kinetic model is more suitable than the pseudo-first order model for describing the data. Considering the fact that the reaction is following pseudo-second order  [38]. Important point to realize here is that by taking appropriate amount of PVCCNPs (0.01 g) the fluoride concentration can be reduced to 1 ppm, to the required fluoride levels in drinking waters, from 20.0 ml of 10 ppm solution. As such, the mass of nanoparticles required to remove fluoride in 10 l water sample that is commonly used per day in household requirements is only 10 g even if this water is contaminated with 10 ppm of fluoride ions. In environmental samples, the contamination does not usually exceed 2-3 ppm and occasionally~7 ppm has been detected in just one tube well waters. Fluoride concentrations in drinking waters of selected 30 dug and tube wells of Moneragala District of Sri Lanka are shown in figure 12(a). Analysis of show that only 10% of these samples have less than 0.5 ppm of F − which means 90% have over 0.5 ppm, 73% have over 1 ppm, 37% have over 2 ppm, 20% have over 3 ppm, 6.7% have over 4 ppm and 3.3% have over 5 ppm of F − . Notably, people drinking water from these wells over a long period have caught the chronic kidney disease on unknown aetiology (CKDU). This shows that there is some correlation between fluoride in drinking waters and the occurrence of CKDU. However, there are no evidence to confirm that fluoride alone is causing CKDU except that two reports showing excess fluoride and calcium together damaging kidneys of laboratory mice [40].
To get an idea about the possible effect of excess fluoride in drinking water on the prevalence of CKDU, we have selected 30 reference drinking water resources. This was done by carefully selecting wells by taking into  account of the fact that there were no recorded CKDU patients who drink water from these water sources. Data are shown in figure 12 (b). Interestingly, statistical analysis indicates that 13%>0.5 ppm, 87%<0.5 ppm, 10% between 0.5 ppm and 1.0 ppm, and just one reference point had 2.86 ppm which means almost all have F − concentration below recommended maximum concentration of 1.5 ppm except just one [41,42].
Therefore, on average even if 5 ppm is the level of contamination, mass of PVCCNPs required to reduce that to 1 ppm level would be only 5 g. Calcium ions required for the synthesis of PVCCNPs can be obtained from commonly available, cheap and mundane minerals such as calcite and dolomite. We have already developed methods to separate calcium and magnesium components and silica impurities from local dolomite and developed methods to prepare various polymorphs of calcium carbonate nanoparticles and their polymer nanocomposites [33]. As such, simple and low-cost filters can be designed and developed to remove excess fluoride from drinking waters in household scale to prevent the occurrence of various chronic diseases due to exposure to excess fluoride. This is perhaps the way forward for preventing dental and skeletal fluorosis and even effects of fluoride to other organs such as heart and kidneys.

Conclusions
This paper describes a soft-templated approach to the synthesis of spherical nanoparticles of vaterite, their characterization and application in the removal of excess fluoride in contaminated waters. Even if the contamination level is 5 ppm the mass of vaterite form of the precipitated calcium carbonate required to reduce level down to 1 ppm is only 5 g. Physicochemical properties of synthesised PVCCNP and fluoride on PVCCNP was determined using FTIR, SEM, XRD, XRF, EDX, and TGA-DTG characterization techniques. The pseudosecond order kinetic model was better able to describe the adsorption of fluoride on PVCCNP than pseudo-first order model. The Langmuir isotherm model, in contrast to the Freundlich model, is more able to describe the equilibrium behaviour of the adsorption process. Given that the value of n (Freundlich constant) obtained is greater than 1 (3.07) and R L value is in the range 0<R L <1 (0.014-0.024) imply that the adsorption process is spontaneous and fluoride ions are favourably adsorbed on PVCCNPs. Langmuir model shows that the maximum adsorption capacity of fluoride is 1.956 mg g −1 . Excess fluoride in drinking waters causes several severe ill-health effects and filter media based on these nanoparticles can be used to remove fluoride down to safe and required levels to tackle these health problems.