Defluoridation of groundwater using mechanochemically-activated clay soil

Groundwater is widely preferred for drinking. In some rural communities in South Africa and some Asian countries, the groundwater contains high concentrations of fluoride. Fluoride in drinking water can be beneficial or detrimental to human health, depending on its concentration and the total amount ingested. Its presence is beneficial to the calcification of dental enamel, in particular for children under eight years old. However, excess fluoride intake causes dental and skeletal fluorosis (Sorg 1978). The World Health Organisation (WHO) has established that the maximum acceptable concentration of fluoride in drinking water is 1.5 mg-F/L, beyond which ingestion is harmful for human health (WHO 2011). Moreover, it has been shown that the intelligent quotient (IQ) of children living in fluoride-rich areas is relatively low (Lu et al. 2000).

In South Africa, waters available in the Western Cape, Limpopo, KwaZulu-Natal and North-West provinces are rich in fluoride. Groundwaters in Limpopo, North-West, Northern Cape, Western Cape and KwaZulu-Natal provinces require partial defluoridation so that they are safe to drink, because they contain fluoride at concentrations exceeding 3 mg/L. These provinces also have large populations still living in rural areas, most of whom drink ground- and surface water (Ncube & Schutte 2005).

Several defluoridation techniques have been suggested, and can be classified as adsorption, ion-exchange, membrane processes and precipitation/coagulation (Hassen 2007). Adsorption is the preferred technique for rural communities because it is cost-effective and the materials are readily available. Activated alumina (AA) (Johnston & Heijnen 2001), bone char (Rao 2003), fly ash (Piekos & Paslawska 1999), activated carbon (AC) (Abe et al. 2004), and clay soil (Kamble et al. 2009) have all been used to defluoridate groundwater.

Mechanochemistry is the study of chemical and physicochemical transformations of substances in all states of aggregation induced by mechanical energy (Wieczorek-Ciurowa & Gamrat 2007). Mechanochemical activation is one such method, and involves simple grinding of an inorganic material, leading to physical disintegration with the formation of active surfaces as well as changes in physicochemical behaviour (Meenakshi et al. 2008). During mechanochemical activation of kaolinites using an oscillating disc mill, Meenakshi et al. (2008) reported specific surface enhancement from 15.11 m²/g (raw kaolinite) to 32.43 m²/g. As new active surfaces were formed, fluoride adsorption capacity increased, from 0.096 to 0.106 mg/g.

In this study, the effectiveness of mechanochemically activated aluminosilicate-rich clay for fluoride removal was examined. The defluoridation trials were performed in batch mode under different experimental conditions.

Aluminosilicate-rich clay soil was obtained from Mukondeni Village, Vhembe district, South Africa. All reagents, including Total Ionic Strength Adjustment Buffer III (TISAB III), were obtained from Monitoring and Control Laboratories (Pty) Ltd (VWR International Ltd, UK) and Sigma-Aldrich Co. LLC, South Africa.

One litre of 1,000 mg-F/L stock solution was prepared by dissolving 2.21 g of analytical grade sodium fluoride (NaF) in 1 litre of Milli-Q (Type 1 ultrapure) water in a volumetric flask and adjusting the volume by adding more Milli-Q water. Lower concentration fluoride solutions were prepared from the stock by serial dilution.

Two hundred grams of clay soil were washed with Milli-Q water at a ratio of 1:10 w/v. The mixture was stirred and left to stand for about 12 hours to allow the soil particles to settle. After settling, the supernatant with floating particles was decanted. The procedure was repeated until the clay was free of floating particles. The upper clay layer was scooped into a clean beaker from which the clay paste was centrifuged to remove the supernatant. The residue was dried in petri dishes in an oven at 105°C for 12 hours. Finally, the dried clay was milled to a fine powder, sieved and stored in plastic bags for future use.

A RETSCH R5 200 ball mill (RETSCH, Haan, Germany) was used for mechanical treatment of the clay at the University of Venda, South Africa.

The morphology was determined using a Hitachi X-650 Scanning Electron Micro-analyser (Hitachi, Tokyo, Japan) equipped with CDU lead detector at 25 kV at the University of Cape Town, South Africa.

The specific surface analysis was carried out by the Brunauer–Emmett–Teller (BET) method and the micropore volume was determined using nitrogen adsorption/desorption isotherms collected at liquid nitrogen temperature (77 °K). Both the specific surface and micropore volume analyses were carried out at the Council for Scientific and Industrial Research (CSIR), Pretoria, South Africa.

The functional groups such as Si-O-H and Si-O-Si were analysed at the University of Venda, South Africa using a Bruker Tensor 27 FT-IR and the OPUS Data Collection Program (Bruker, Billerica, MA, USA). Clay sample mineralogy was analysed at the University of Pretoria, South Africa, using a PANalytical X'Pert Pro powder (Malvern Panalytical, Almelo, Netherlands) diffractometer in θ–θ configuration with an X'Celerator detector.

The elemental composition of the mechanochemically-activated clay soil was determined at Stellenbosch University, South Africa, using a PANanalytical Axios instrument. The exchangeable cations were also determined at Stellenbosch University by Inductively Coupled Plasma-Mass Spectrometry (ICP-MS).

The pH at point-of-zero charge was determined using 0.01 and 0.1 M KCl solutions. Sixty millilitres of the respective solutions were measured into plastic containers and the pH adjusted using 0.1 M HCl and 0.1 M NaOH. Aliquots of 50 mL of the solutions with known pH were then measured into clean and dry 100 mL plastic bottles. A mass of 0.6 g of clay was weighed into each bottle, corked and shaken at 200 rpm for 24 hours. The pH of each mixture was measured after the equilibration time. The change in pH was determined and plotted against the initial pH of the KCl solution. Cation Exchange Capacity (CEC) was determined using ammonium acetate buffers at pH 7, and identified by measuring the concentrations of Ca²⁺, K⁺, Mg²⁺ and Na⁺ as exchangeable cations on the surface of aluminosilicate clay soil that had been mechanochemically activated for 30 minutes.

The pH of the mixture was adjusted using 0.1 M HCl or 0.1 NaOH. A known mass of adsorbent was dispersed in 90 mL of 10 mg-F/L fluoride solution in a 250 mL plastic bottle, which was then closed tightly and shaken at 200 rpm for a fixed time. The pH was measured before and after shaking. The reacted suspension was centrifuged for 5 minutes at 40 rpm and the supernatant analysed for residual fluoride, using a four-standard calibrated ORION fluoride ion-selective electrode. (TISAB III was added to both the standards and samples at a ratio of 1:10 v/v, 30 minutes before analysis.) The effects of pH (range 2 to 12), contact time (1 to 90 minutes), adsorbent dosage (0.1 to 1 g), adsorbent concentration (4.5 to 90 mg-F/L), temperature (298, 309 and 318 K), and chemical stability of the clays in adsorption media were all evaluated.

The scanning electron microscope (SEM) images of raw and mechanochemically-activated aluminosilicate clay soil treated for 5, 10, 15, and 30 minutes are presented in Figure 1 and show the honeycombed, spongy, and porous and irregular nature of the surfaces.

The images show the decreasing clay soil particle size with increased milling time. The largest particle sizes were obtained at milling times of 3 and 5 minutes, while the smallest were found after 30 minutes' milling.

The BET results are presented in Table 1. The specific surface and pore volume of the raw clay were relatively low. Those of the activated material increased with corresponding increase in milling time. The pore size of the soil decreased, however, with increasing milling time.

Previous researchers have employed mechanochemical activation to increase the specific surface (Meenakshi et al. 2008), as well as increasing the material's surface energy and surface reactivity, and, consequently, its chemical activity (Thompson et al. 1993), which would be expected to increase the adsorbent's sorption capacity.

FT-IR spectra of the clay soil were prepared to identify possible changes in the functional groups associated with the aluminosilicate structure arising from different activation periods (Figure 2). The stretching vibrations of Si-O-H bands were noted at 998 cm⁻¹ in material activated for 5 minutes and at 910 cm⁻¹ for 10 minute activation, while the same absorbance was noted at 996 cm⁻¹ for clay soils activated for 15 and 30 minutes. Such changes may be related to different activation times due to changes in the release of functional groups necessary for interaction with fluoride. A similar trend was noted for Si-O-Si stretching and bending vibrations. The peaks observed in the range from 1,200 to 400 cm⁻¹ were characteristic of mixed metals. Peak height increased with milling time, so that the highest peak recorded was for material milled for 30 minutes (Figure 2).

Experimental milling times did not exceed 30 minutes. The BET results showed that the specific surface of the adsorbent increased with milling time and that the best noted was achieved with material milled for 30 minutes. The FT-IR images also confirm that a 30 minute milling time achieved the best mechanochemical activation of those studied.

The XRD spectra in Figure 3 show the main mineral components, quartz and albite (a plagioclase feldspar), in the clay soil. Other minerals occurring in minor quantities included vermiculite, talc, muscovite and actinolite.

Quartz is the most abundant mineral phase by weight, at about 31%, followed by plagioclase (29%). Actinolite was present at about 5% by weight.

The XRF major oxide analysis of clay soil milled for 30 minutes – Table 2 – showed that silica (SiO₂) was the commonest component followed by alumina (Al₂O₃) at a ratio of 3.87:1. The minor components included MnO, Cr₂O₃ and P₂O₅.

The trace element analysis results showed that chromium, barium and many other elements were present at trace concentrations. Their analyses improved understanding of the adsorbent elemental composition.

The pH at point-of-zero charge (pHpzc) of an adsorbent is significant in the adsorption process. It represents the pH at which the net surface charge on the adsorbent is zero (Valdivieso et al. 2006). The Mukondeni red clay soil has the high pHpzc characteristic of clay materials dominated by aluminosilicates (Gitari et al. 2015). Figure 4 shows the plot patterns for the adsorbent in KCl solutions of different concentrations and the points where they cross the horizontal axis (representing zero charge). The pHpzc is the abscissa for which ΔpH equals zero (Izuagie et al. 2016). The mean pHpzc was 6.6 ± 0.1.

The trend of exchangeable cation concentrations was Mg²⁺ > Ca²⁺ > K⁺ > Na⁺, at pH 7.

The pH of aqueous solutions affects and controls sorption capacity due to its influence on the sorbent's surface properties and the ionic forms of the pollutants in solution. Figure 5(a) shows the proportional effect of pH on fluoride removal and adsorption capacity. At lower pH, fluoride removal was higher, possibly because increasing the electropositivity of the adsorbent surface enhanced attraction for the negatively charged fluoride ions (Izuagie et al. 2016). Above pH_e 2.41, proportional fluoride removal decreased, perhaps because of the formation of hydroxyl groups, which competed with fluoride ions for uptake by clay soil materials (Izuagie et al. 2016). The optimum pH for fluoride removal from solution was at the lowest pH_e of 2.41.

The study showed that the highest fluoride removal occurred at 1 minute contact time for the different adsorbent doses (Figure 5(b)). The values at other equilibration times, although a little lower than that at 1 minute, did not vary appreciably from one another. Equilibrium is thought to have occurred after about 90 minutes. The high rate of fluoride adsorption in the first minute of equilibration was likely due to the high fluoride gradient at the adsorbent-solution interface at that point.

The effects of adsorbent dose on fluoride removal and adsorption capacity are shown in Figure 5(c), and were significant. Increasing the adsorbent mass at constant fluoride concentration decreased adsorption capacity. Proportional fluoride removal increased with increasing adsorbent dose because more active sites were available.

Figure 5(d) shows that fluoride removal decreased with increasing initial fluoride solution concentration. This was possibly because, as the adsorbent dose was constant, there was a limit to the amount of fluoride that could be adsorbed at any particular concentration. Therefore, the unadsorbed/adsorbed fluoride ratio increased with increasing initial concentration.

The adsorption capacity increased with increasing initial fluoride concentration. The almost linear trend of the plots indicates that the adsorbent surface was far from being saturated with adsorbed fluoride within the limit of fluoride concentrations used.

Adsorption models reflect the relationship between the amount of a solute adsorbed at constant temperature and its concentration in the equilibrium solution (Yousef et al. 2011).

Several isotherm equations are available for analysing experimental adsorption equilibrium data. Two of them, those of Langmuir and Freundlich, were applied to the fluoride adsorption equilibrium data of the clay soil studied.

The experimental data fit best into the Freundlich model (Table 3), because the coefficient of correlation, R², is closer to 1. The Freundlich model is applicable to multi-site adsorption onto rough or smooth surfaces. In fact, as the values of 1/n and K_L are both less than 1, the adsorption fits reasonably well with both the Langmuir and Freundlich models – see Figures 6(a) and 6(b). This indicates that the adsorbent surface is partly homogeneous and partly heterogeneous.

The adsorption kinetics data were analysed using both the pseudo-first-order and pseudo-second-order models.

The pseudo-first-order model was tested by fitting the adsorption data but the plot of against t did not give straight lines. In other words, the model was not applicable to the sorption process as shown in Figure 6(c).

The value was plotted against time t for adsorbent doses of 0.1, 0.3 and 0.4 g, yielding straight lines – Figure 6(d). The data fit well to the pseudo-second-order equation, indicating that fluoride adsorption was by chemisorption.

The calculated and experimental values of q_e are compared in Table 4. The results confirm that the adsorption data fit well on the pseudo-second-order kinetic model.

The pHpzc values at 0.01 and 0.1 M KCl were 6.67 and 6.53 respectively, with a mean of 6.6 ± 0.1. While at pH 6.60 the surface charge is neutral, below 6.6 ± 0.1 it is positive and, above it, negative.

Fluoride removal only occurred at low pH, when the fluoride ions were attracted to the positively charged clay surface – ref Equation (14).

A comparison of the adsorption capacity of the clay soil used in this study with that of other adsorbents in common use is presented in Table 5. It was observed that the clay soil used in this study had lower adsorption capacity than commercially prepared montmorillonite, bentonite, or mechanochemically activated kaolinites. On the other hand, the clay soil studied had a slightly higher adsorption capacity for fluoride than red mud – a solid state industrial by-product of the process of alumina extraction from bauxite (Lv et al. 2013).

The SEM images showed that milling tends to break down elongated, high aspect ratio particles, yielding smaller particle size from longer milling time, the latter promoting fluoride adsorption. The specific surface and pore volume of the mechanochemically activated clay soil increased with increased milling time, while the pore size decreased.

In FT-IR analysis of the clay soil, the stretching vibrations of Si-O-H bands were noticed at absorbance 998 cm⁻¹ for 5 minute activation and 910 cm⁻¹ for 10 minutes, while absorbance at 996 cm⁻¹ arose from activation times of 15 and 30 minutes. These changes in Si-O-H stretching with respect to different activation times may be due to changes in the release of functional groups necessary for fluoride interaction.

The major mineral components of the clay soil are quartz and albite. Other minerals that occur in relatively lesser quantities include vermiculite, talc, muscovite and actinolite. The XRF analysis showed that silica (SiO₂) is present at the highest concentration followed by alumina (Al₂O₃).

The mean pHpzc is 6.6 ± 0.1. Below 6.6 ± 0.1 the surface charge is positive and, above it, negative. The CEC concentration trend was Mg²⁺ > Ca²⁺ > K⁺ > Na⁺.

Mechanochemical activation had significant impact on adsorption efficiency. Fluoride adsorption capacity depends on the contact time, adsorbent dose, fluoride concentration, and pH. Maximum fluoride uptake was achieved with 1 minute of agitation of 0.6 g/100 mL at 200 rpm.

Proportional fluoride removal decreased with increasing initial F⁻ concentration. Maximum fluoride removal (41%) was obtained at the pH_e of 2.41.

The adsorption data fit better into the Freundlich than the Langmuir model. Pseudo-first-order and pseudo-second-order models were tested and the latter was the more appropriate for fluoride sorption.

The fluoride adsorption process mechanism with the clay soil was also evaluated.

The financial support from WRC Project No. K5/2363/3, NRF Project No. CSUR13092849176, Grant No. 90288 and THRIP Project No. TP12082610644 Research & Innovation Directorate, University of Venda and Sasol Inzalo Foundation is acknowledged.