Fluoride removal by Ca-Al-CO3 layered double hydroxides at environmentally-relevant concentrations

(cid:1) Ca e Al e CO 3 layered double hydrox- ides showed af ﬁ nity for F (cid:3) at 2 e 12 mg/L. (cid:1) A higher F (cid:3) removal capacity at lower pH and lower temperature was observed. (cid:1) Only marginal de ﬂ uorination improvements by calcination/micro- wave treatment. (cid:1) Fast F (cid:3) uptake during the initial 20min and little anions competition was observed. (cid:1) F (cid:3) removal capacity is not necessarily re ﬂ ected in speci ﬁ c surface area. In this study, F (cid:3) removal by Ca e Al e CO 3 layered double hydroxides (LDHs) was investigated at environmentally-relevant concentration ranges (2 e 12 mg/L) to below the WHO guideline, with an emphasis on the effect of LDHs ’ modi ﬁ cation, as well as the effects of initial F (cid:3) concentration, adsorbent dose, pH, temperature and co-existing ions. Ca e Al e CO 3 LDHs, either untreated, calcined or microwave treated, showed af ﬁ nity for the removal of F (cid:3) from synthetic groundwater with capacities of 6.7 e 8.4 mg F (cid:3) /g LDHs at groundwater-relevant pH, with a higher F (cid:3) removal capacity at lower pH ( < 8) and lower temperature (12 (cid:4) C, as compared to 25 (cid:4) C & 35 (cid:4) C). Since calcination and microwave treatment resulted in only marginal de ﬂ uorination improvements, using untreated LDHs appears the practically most feasible option. For the untreated LDHs, competition with Cl (cid:3) and NO 3 (cid:3) was not observed, whereas at higher HCO 3 (cid:3) and SO 42 (cid:3) concentrations ( > 250 mg/L) a slight reduction in F (cid:3) removal was observed. This study indicates the potential of Ca e Al e CO 3 LDHs as a cost-effective F (cid:3) removal technology, particularly when locally sourced and in combination with low-cost pH correction.


r a c t
CaeAleCO 3 layered double hydroxides showed affinity for F À at 2 e12 mg/L. A higher F À removal capacity at lower pH and lower temperature was observed.
Only marginal defluorination improvements by calcination/microwave treatment. Fast F À uptake during the initial 20 min and little anions competition was observed. F À removal capacity is not necessarily reflected in specific surface area.

Introduction
Fluoride (F À ) is essential for the growth of teeth and bones, but excessive F À intake causes dental and skeletal fluorosis (Edmunds and Smedley, 2013). Fluorosis caused by F À contamination of groundwater has been reported in 35 nations over the world from Africa, South Asia, the Middle East, North, Central and South America, and Europe (Ayoob et al., 2008). Although a multitude of technologies such as coagulation, adsorption, ion exchange, electrochemical, and membrane processes have been investigated for F À removal and some of them showed promising F À removal capacities, many of them indulge in high installation and maintenance costs, complex treatment processes, and by-product pollutants (Ayoob et al., 2008;Osterwalder et al., 2014;Bhatnagar et al., 2011). Adsorption is a preferable F À removal method because of the ease of operation, cost-effectiveness, and potential effective F À removal capacity, especially for developing countries (Mohapatra et al., 2009;Kanno et al., 2014). Layered double hydroxides (LDHs) are a class of anionic clays that are being applied in various fields such as flame-retardant fillers, catalysts, drug carriers and adsorbents (Li and Duan, 2006). LDHs have attracted considerable attention because of their potential high defluoridation capacity and cost-effectiveness (Sun et al., 2017). The general formula of LDHs is [M 2þ 1-x M 3þ x (OH) 2 ] xþ [A mÀ x/m •nH 2 O], where M 2þ corresponds to a divalent metal, such as Mg 2þ , Zn 2þ or Ni 2þ , M 3þ corresponds to a trivalent metal, such as Al 3þ , Fe 3þ or Cr 3þ , x is the ratio of M 3þ /(M 2þ þM 3þ ), and A mÀ corresponds to an exchangeable anion such as CO 3 2À , Cl À , NO 3 À (Cavani et al., 1991;Ingram and Taylor, 1967). The presence of a large number of exchangeable anions and sizable interlayer spaces makes LDHs attractive for the removal of dissolved contaminants from water (Mandal and Mayadevi, 2008a), including F À , Cl À , NO 3 À , BrO 3 À , PO 4 3À , As(III), Pb(II), Cr(VI), Cu(II) and Cd(II) (Batistella et al., 2011;Chitrakar et al., 2011;Halajnia et al., 2012;Das et al., 2006;Lazaridis et al., 2004;P erez et al., 2006;Caporale et al., 2013;Lv et al., 2009). In order to enhance the removal capacity of LDHs, several methods have been used to modify LDHs, such as metal oxide amendment, acid treatment and thermal activation (Batistella et al., 2011;Lv, 2007;Zhang et al., 2013). Thermal treatment using a muffle furnace (Lv, 2007), so-called calcination, causes the destruction of the layered structure of LDHs, however, this structure is afterwards recovered upon hydration . Microwave radiation heating is an alternative method that heats the material from inside out (Bhatnagar et al., 2013), and has been applied for activated carbons (Yuen and Hameed, 2009). Microwave radiation would reduce the treatment time compared to calcination, consequently resulting in reduced energy consumption and CO 2 emission (Huang et al., 2011). Microwave treatment has been applied for synthesis and construction research (Benito et al., 2009), but not for LDHs' modification.
Different types of LDHs have been studied for the removal of F À from aqueous solutions, e.g., MgeAl, NieAl, ZneAl, LieAl, MgeCr, MgeFe, FeeMgeAl LDHs (Batistella et al., 2011;Wang et al., 2007;Kameda et al., 2015;Chang et al., 2011;Lv et al., 2007;Du et al., 2014;Mandal and Mayadevi, 2008b). However, most of the previous studies were carried out at high initial F À concentration (>50 mg/L), whereas F À concentrations in actual groundwater ranges from 0.1 to 22 mg/L, and most frequently below 10 mg/L (Wen et al., 2013). In addition, F À concentrations need to be reduced to <1.5 mg/L to comply with the World Health Organisation (WHO) guideline for drinking water (World Health Organization, 2004). Among the various types of LDHs, MgeAl LDHs and their calcined products are the most studied. However, F À removal by CaeAl LDHs and their modified products is rarely reported, while CaeAl LDHs is expected to have a better defluorination efficiency than MgeAl LDHs considering the affinity of calcium and aluminum towards F À (Ghosal and Gupta, 2015).
It was the aim of this study to investigate whether CaeAleCO 3 LDH is a suitable alternative for F À removal from groundwater, specifically in the low, environmentally-relevant initial F À concentration range (2e12 mg/L), to concentrations below the WHO guideline. For this reason, the performance of CaeAleCO 3 LDH as F À adsorbent was investigated under different environmentallyrelevant water compositions (including, initial F À concentration, pH, temperature and co-existing ions). In addition, calcination and microwave treatment were investigated as CaeAleCO 3 LDH modification methods with respect to their effect on LDH surface properties (including, XRD, FTIR, BET and pH zpc ).

Chemicals and materials
All reagents including HCl, NaOH, NaF, NaCl, NaNO 3 , NaHCO 3 and Na 2 SO 4 were of analytical grade (Sigma-Aldrich). Deionized water was used throughout the experiments and treatment processes. Ca 4 Al 2 (OH) 12 CO 3 $nH 2 O (n ¼ 4e5) LDH (ACTILOX®CAHC) was obtained from Nabaltec (Germany) with a Ca/Al ratio of 1.86. Before use, CaeAleCO 3 LDHs was dried in an oven at 105 C for 12 h.

Calcination and microwave treatment of LDHs
The calcined CaeAleCO 3 LDHs were obtained by calcining CaeAleCO 3 LDHs (5 g) in a muffle furnace at 500 C for 2 h and cooling in a desiccator. Microwave radiation heating was carried out in a commercial microwave oven with 1000 W output at 2450 MHz (Samsung MS28J5215AB) with suitable adjustment (Supplementary Information (SI), Fig. S1). The microwave treatment was carried out in a quartz bowl fixed in the chamber of the microwave oven. The temperature of the LDHs after microwave irradiation was measured immediately after treatment, using an infrared radiation thermometer (TROTEC BP21). 5 g of CaeAleCO 3 LDHs sample was treated by microwave irradiation for 15 min and then cooled in a desiccator to room temperature for further use.

Characterization methods
The crystalline structure of CaeAleCO 3 LDH and its modified products before and after the experiments was characterized using a Philips PW 1830 powder X-ray diffractometer with a 2q range of 5e70 . Fourier transform infrared spectroscopy (FTIR) spectra of CaeAleCO 3 LDH and its modified products were recorded by a fourier transform infrared spectrometer (Spectrum TM 100 Optical ATR-FTIR), following pelletization of the materials. The textural properties of CaeAleCO 3 LDH and its modified products were characterized using a surface area analyzer (Micrometrics Gemini VII 2390 V1.03). Brunauer-Emmett-Teller (BET) surface areas of CaeAleCO 3 LDH and its modified products were determined by N 2 adsorption-desorption method. The pore-size distribution was determined by the Barret-Joyner-Halender (BJH) method.
The pH at the point of zero charge (pH zpc ) of CaeAleCO 3 LDH and its modified products was determined by the pH drift method (Müller et al., 1985). The initial pH of NaCl (0.01 mol/L) was adjusted from 3.0 to 13.0 by addition of 0.1 mol/L HCl or NaOH, followed by LDHs sample (0.1 g) addition to the solution. After stirring at room temperature for 24 h, the final pH (pH final ) was measured. The pH zpc of the CaeAleCO 3 LDH and its modified products was determined from the plots pH final versus pH initial . The point at which pH final equal pH initial was taken as the pH pzc .

Batch adsorption experiments
A stock solution of F À (NaF as 1000 mg F À /L) was diluted to different working solutions. For adsorption isotherm experiments, CaeAleCO 3 LDH and its modified products (0.1 g) were dispersed in 100 mL of F À solutions (2e12 mg F À /L), with the pH being adjusted to 8 at the experiment start using HCl (1 M). After stirring (Magnetic stirrer at 100 rpm) for 24 h, the solutions were filtered by a microfiltration membrane (0.45 mm). The concentrations of F À were analyzed by Metrohm 881 ion chromatography (IC) with a Metrosep A Supp 5 column (eluent: 3.2 mM Na 2 CO 3 and 1 mM NaHCO 3 ; flow rate: 0.7 mL/min). The effects of batch adsorption parameters including pH, adsorbent dose, co-existing anions, were investigated at an initial F À concentration of 10 mg/L. The effect of initial pH was investigated at 6, 7, 8, 9, and 10 (±0.1). The effect of adsorbent dose was studied at pH 8 with adsorbent doses of 0.1, 0.2, 0.5, 1, 1.5 and 2 g/L. The batch experiments of the effect of temperature were carried out at different temperatures (12 ± 1, 25 ± 1 and 35 ± 1 C). The effect of co-existing anions (Cl À , NO 3 À , HCO 3 À , SO 4 2À ) on F À adsorption was studied by varying their concentrations (0, 50, 100, 250, 500 mg/L) with an initial F À concentration of 10 mg/L at pH of 8, by dosing NaCl, NaNO 3 , NaHCO 3 and Na 2 SO 4 . For kinetic studies, CaeAleCO 3 LDHs samples (1 g) were dispersed in 1 L F À solutions (10 mg/L) and stirred for 5 h under the initial pH of 8. Samples (3 mL) were collected at selected time intervals (0,1,2,3,4,5,10,20,30,40,50,60,120,180,240 and 300 min), then filtered and analyzed as mentioned before.
The F À uptake (q: mg/g) at equilibrium and non-equilibrium contact times was calculated using the following equation: where C 0 and C are initial and F À concentrations after the experiment (mg/L), respectively; V (L) is the solution volume; and m (g) is the adsorbent mass. An overview of experimental conditions is given in Table 1. All the batch experiments and kinetic studies were carried out in duplicates. For the results of isotherm, the averaged data were reported.

Adsorption isotherm and kinetic models
Assuming strictly adsorptive removal processes, the equilibrium data were fitted by the Langmuir and Freundlich isotherm models. The kinetic models can be used to investigate the possible mechanism of adsorption and potential rate controlling steps (Gupta and Bhattacharyya, 2011). Four most commonly used models, namely the pseudo-first-order, pseudo-second-order, the Elovich equation, and the intraparticle diffusion models were fitted to the kinetic experimental data. Details on the equations and plotting can be found in the SI.
The fitting of experimental data to adsorption isotherm and kinetic models was accomplished by using OriginPro 9.0. The calculation of saturation-index of calcium fluoride (CaF 2 ) was carried out on PHREEQC (Dhiman and Keshari, 2006;Parkhurst and Appelo, 2013). Detailed descriptions of simulation can be found in the SI.

Characterization of CaeAleCO 3 LDHs
The differences in chemical and physical properties of untreated, calcined and microwave treated CaeAleCO 3 LDHs were compared by XRD, FTIR, and N 2 gas adsorption to study the effects of modification. Table 2 shows the BET surface area, pore volume, pore diameter and pH pzc of CaeAleCO 3 LDH and its modified products. The surface area was less than 10 m 2 /g, which is lower compared to other inorganic sorbents, such as activated alumina (250 m 2 /g) (Ghorai and Pant, 2004). After calcination, the surface area decreased slightly while the average pore size increased. Both calcination and microwave treatment increased the pore volume of LDHs.
The XRD patterns of CaeAleCO 3 LDH and its modified products are presented in Fig. 1 (A). The diffraction pattern of CaeAleCO 3 LDH ( Fig. 1 A (a)) shows sharp and symmetric peaks at lower 2q values (003 and 006), which are characteristic of LDH's crystalline structure (Ramírez-Llamas et al., 2015). These peaks disappeared after calcination ( Fig. 1 A (b)) and microwave treatment ( Fig. 1 A  (c)), indicating the layered structure was destroyed, while an increase of the Al(OH) 3 and CaCO 3 peaks was observed (Sun et al., 2017). Upon re-hydration, the peaks of 003 and 006 are usually re-occurring (Ramírez-Llamas et al., 2015), underlining the recoverability of temperature-treated LDH materials. After aqueous F À uptake, however, LDHs crystalline structure characteristic peaks (003 and 006) were not observed while characteristic peaks of CaCO 3 became stronger, indicating the LDHs crystalline structure was not rebuilt.
The FTIR spectra of CaeAleCO 3 LDH and its modified products are shown in Fig. 1 (B). The bands between 3600 and 3300 cm À1 are due to the vibrations of OH groups in the adjacent layers, the interlayer and/or adsorbed water (Noorjahan et al., 2015). The peaks at 1416 and 1362 cm À1 are due to the vibrations of sorbed CO 2 and interlamellar CO 3 2À , respectively (Lv et al., 2006;Cai et al., 2012). The bands between 1000 and 600 cm À1 can be attributed to characteristic vibrations of calcium and aluminum oxides (Lv et al., 2006;Das et al., 2003) (Fig. 1 B(a)). After calcination and microwave treatment, the vibrations of OH groups (3600-3300 cm À1 ) mostly disappeared indicating the loss of water due to thermal treatment  Fig. 1 B(b)), which is in line with the XRD results. After calcination, the band at 1362 cm À1 disappeared and the band at 1416 cm À1 shifted to a lower frequency of 1411 cm À1 with a higher intensity indicating the loss of CO 3 2À species and the transformation of CO 3 2À to CO 2 , which indicates that the layered structure of LDHs was destroyed by calcination, confirming the XRD findings.

F À adsorption equilibrium
The adsorption isotherm results, given in Fig. 2, indicate that CaeAleCO 3 LDH and its modified products show affinity for F À within the low initial F À concentration range of 2e12 mg/L, up to F À loadings of~1% LDH weight. Although the differences are small, both modified LDHs exhibit a higher adsorption capacity than the raw LDH, especially at initial F À concentrations >10 mg/L. However, it should be noted that both thermal treatments involve the loss of internal surface-bound water and calcination involves the loss of CO 3 2À , which do not occur when drying (105 C) of the untreated LDH (cf. 3.1). Thus, the total mass of CaeAleCO 3 LDH added to the batches is somewhat higher for the calcined/microwave-treated LDHs, and consequently, the LDH-mass-related performance differences are smaller than those shown in Fig. 2. The isotherms are likely to be L-type curves without strict plateau (Limousin et al., 2007), however, the relatively low C e range of the experiment restricts the curves to a mostly linear range.
The fitting results indicate that Freundlich model can well fit the experimental data (R 2 > 0.98) (Table S1). At the C e of 1.5 mg/L, the F À adsorption capacities of calcined, microwave treated, and untreated CaeAleCO 3 LDHs are 8.4, 7.4, and 6.7 mg/g, respectively. For comparison, previous research has shown a F À removal capacity of 1.84 mg/g for LDHs type MgeAleCO 3 , at 5 mg/L (Wang et al., 2007). For MgeAleCO 3 LDH also higher capacities have been reported (319.8 mg/g), but these experiments were conducted at higher initial F À concentrations (up to 2500 mg/L) .
The equilibrium isotherms of F À uptake on CaeAleCO 3 LDH and its modified products at different temperatures (12 ± 1, 25 ± 1 and 35 ± 1 C) are shown in Fig. 3. The F À adsorption capacity of CaeAleCO 3 LDH and its modified products decreased with the increase in temperature, which indicates that F À uptake on CaeAleCO 3 LDH and its modified products is an exothermic process (Errais et al., 2012). Fig. 4 shows the effect of contact time on F À uptake by CaeAleCO 3 LDH and its modified products. For all CaeAleCO 3 LDH and its modified products, the F À uptake increased rapidly during the initial 20 min, with a slower rate thereafter. Microwave treated and calcined CaeAleCO 3 LDHs present a higher F À removal capacity than untreated CaeAleCO 3 LDHs. The experimental data fits to kinetic models are presented in the SI (pseudo first-order, pseudo-second-order, Elovich equation, and intraparticle diffusion, Fig. S2, Table S2), of which the pseudo second-order model resulted in the best data representation (R 2 > 0.992, Fig. S2b). The calculated adsorption capacity (q e ,cal) of CaeAleCO 3 LDHs, calcined CaeAleCO 3 LDHs and microwave treated CaeAleCO 3 LDHs was 6.38 mg/g, 7.18 mg/g and 7.22 mg/g, respectively, which are close to the experimental values (6.31, 7.09 and 7.16 mg/g, respectively). These values are lower than those obtained from Fig. 2 (~8 mg/g). A possible reason could be the acidification order, which was LDHs before acidification for the isotherm studies, while the opposite for the kinetic studies (same amount of acid dosed in both experimental series). As previously discussed, the thermal treatments might have resulted in an overestimation of the respective performances as compared to the raw LDH (cf. 3.2).

Effect of co-existing anions
Groundwater also contains common anions such as Cl À , NO 3 À , HCO 3 À , SO 4 2À , potentially competing with F À in adsorption (Kang et al., 2018). Especially the introduction of potentially competing anions by acidification (e.g. Cl À by adding HCl) has been ignored in  most previous studies. The impact of different anions on the defluorination performance of the unmodified LDH at variable doses is shown in Fig. 5. The data show that Cl À and NO 3 À hardly affected the defluorination performance of CaeAleCO 3 LDHs, especially when their concentrations were lower than 100 mg/L. This observation is consistent with other findings for F À removal by metal composite (Kang et al., 2018;Xiang et al., 2014;Wang et al., 2017;Tang and Zhang, 2016). One possible reason could be that chemical reactions (e.g. the formation of AlF x complexes, precipitation of CaF 2 ) are involved in the removal process (Kang et al., 2018) and F À has a stronger affinity for LDH adsorbents than Cl À and NO 3 À (Tang and Zhang, 2016;Loganathan et al., 2013). At higher concentrations of 250 and 500 mg/L, HCO 3 À and SO 4 2À had a slight effect on the removal process. When the concentrations of HCO 3 À and SO 4 2À increased from 0 to 500 mg/L, the F À removal percentage decreased from 83% to 71% and 77%, respectively. The weak effect of HCO 3 À and SO 4 2À at low concentrations (<100 mg/L) could be due to the large number of available adsorption sites. The stronger effect of HCO 3 À might be due to CO 3 2À in the interlayer of the LDH firstly being converted into HCO 3 À , then exchanged with F À , but the presence of a high concentration of HCO 3 À potentially prevents the conversion process.

pH effect and buffering by LDHs
CaeAleCO 3 LDHs showed a noticeable effect on the pH of the solution. With the dose of 0.1e2 g/L, the unmodified CaeAleCO 3 LDH, and the calcined and microwave treated products increased the pH of the solution (from approximately 6.5) to 10.8e11.5, 10.1e11.6 and 10.8e11.7, respectively, indicating the release of hydroxyl ions from LDHs. In order to investigate the effect of adsorbent dose (and the corresponding pH effect) on F À removal, the concentration of CaeAleCO 3 LDHs was varied from 0.1 g/L to 2 g/L with the initial F À concentration of 10 mg/L, as presented in Fig. S3.
The results clearly indicate that for all LDHs, with the increase in adsorbent dose, the adsorption capacity gradually decreased from around 30 mg/g to 5 mg/g (also corresponding to the data in Fig. 2). However, the F À removal percentage considerably increased from 20% to 90% with the increase of adsorbent dose from 0.1 g/L to 2 g/L. Fig. 6 shows the result of F À uptake by CaeAleCO 3 LDH and its modified products over an initial pH range of 6e10. The CaeAleCO 3 LDH and its modified products presented a higher defluoridation capacity at a lower pH. A similar phenomenon was observed in other adsorbents, such as alumina (Viswanathan and Meenakshi, 2010), hydroxyapatite (Jim enez-Reyes and Solache-Ríos, 2010) and KMnO 4 -modified activated carbon (Daifullah et al., 2007). One explanation is that the surface of adsorbents was positively charged with the decrease of pH due to their high pH pzc (Table 2, Fig. S4) (Das et al., 2003). The pH pzc of untreated, calcined and microwave treated CaeAleCO 3 LDHs were 12.5, 12.3 and 12.7, respectively. At pH below pH pzc , the surface of the LDHs has a net positive charge which is favorable for F À adsorption (Wu et al., 2015). This is due to the fact that hydroxyl groups on the surface (≡M) of LDHs were protonated at low pH and the sorption was dominated by the electrostatic interaction (Wu et al., 2017). The process could be expressed as follows (Kang et al., 2013): In addition, at higher pH, more hydroxyl groups on the surface of LDHs could compete with F À due to their similar ion radius (Feng et al., 2004).
The sharp increase in adsorption capacity when lowering the pH from 9 to 8 can potentially be explained by the HCO 3 À d CO 3 2À equilibrium. At pH 9, there is mainly CO 3 2À which is not beneficial for F À removal because LDHs have a stronger affinity for divalent ions (e.g. CO 3

2À
) than monovalent ions (e.g. F À ) . At pH 8, however, CO 3 2À converted to HCO 3 À which is beneficial for F À removal. Although both HCO 3 À and F À are monovalent, HCO 3 À only shows a slight reduction in F À removal at high concentration (>250 mg/L) (Fig. 5). This hypothesis is supported by the finding that the pH effect is less prominent for the calcined CaeAleCO 3 LDHs (cf. Fig. 6), because its interlayered CO 3 2À was decomposed  into CO 2 during the calcination (Ramírez-Llamas et al., 2015).

Discussion
CaeAleCO 3 LDHs are capable of removing F À at environmentally-relevant concentrations with practically feasible LDH dosages. F À removal capacities for untreated, calcined and microwave treated CaeAleCO 3 LDHs were similar, although XRD and FTIR spectra indicated (minor) changes to the LDH surface. The BET specific surface area of CaeAleCO 3 LDH and its modified products (<10 m 2 /g) was lower than the surface areas obtained for MgeAleCO 3 LDHs (196.8 m 2 /g) (Batistella et al., 2011), ZneAleCl LDHs (92.6 m 2 /g) (Mandal and Mayadevi, 2009), ZneCreNO 3 LDHs (12e26 m 2 /g) (Koilraj and Kannan, 2013) and LieAl LDHs (37.24e51.27 m 2 /g) (Tao et al., 2012). Nevertheless, the removal capacity was higher for CaeAleCO 3 LDHs compared to most other LDHs, i.e., 5 mg/g (initial F À ¼ 6 mg/L, pH ¼ 8) for CaeAleCO 3 LDHs versus 1.84 mg/g (initial F À ¼ 5 mg/L, pH ¼ 6) for MgeAleCO 3 LDHs (Wang et al., 2007) at similarly environmentally relevant F À concentrations. This illustrates that F À adsorption capacity of LDHs is not necessarily reflected in BET measurements with N 2 , which may be explained by the fact the diameter of F À (1.33 Å) (Shannon, 1976) is less than half the diameter of N 2 (3.64 Å) (Kentish et al., 2008). Also, although the removal of F À by CaeAleCO 3 LDHs shown here appears promising, the exact pathway for F À removal remains partly undisclosed. Both adsorption and anion exchange on the LDH surface might occur, but under specific conditions, also (surface) precipitation of CaCO 3 and/or Al(OH) 3 could occur (see XRD in Fig. 1A). The results of the PHREEQC calculations show that the precipitation of CaF 2 can occur at higher Ca 2þ (>10 mg/L) and F À (>4 mg/L) concentrations (Fig. S5). As such, CaCO 3 /Al(OH) 3 precipitation processes on the surface of LDH cannot be excluded and could hypothetically contribute to F À removal through enhanced adsorption. The pH dependency of F À uptake by CaeAleCO 3 LDHs indicates that the CO 3 2--HCO 3 balance might be of importance to the F À removal pathway and requires further investigation. For example, CO 3 2À in the interlayer of LDHs might be converted to HCO 3 À , making exchanges with F À more favorable. While further studies on the exact removal pathway of F À on CaeAleCO 3 LDHs are needed, this study has presented clear evidence that effective F À removal is feasible in the lower concentration ranges as well as at practically applicable LDH dosages, and potential competition by other typical anions appears to be advantageously low. For application it is, however, crucial to develop methods for granulation (e.g., by binding with sodium alginate, vinyl alcohol or a clay binder such as kaolin), as well as for pH control to achieve optimal performance. The strong buffering capacity of CaeAleCO 3 LDHs boosts the operating pH upwards (>pH 11), which is undesirable from a removal capacity perspective. Achieving lower operating pH (<pH 8) should be further examined, e.g. by preacidification of groundwater prior to LDH treatment. From an economic standpoint, at an F À removal capacity of 6.7 mg/g, a kilogram of CaeAleCO 3 LDHs can treat about 788 L of F À contaminated water (initial F À ¼ 10 mg/L; pH ¼ 8; LDHs costs ¼ V 3.8e5.2  Microwave treated Ca-Al-CO 3 LDHs Fig. 6. Effect of initial pH on the F À removal by CaeAleCO 3 LDH and its modified products. (Initial F-¼ 10 mg/L; pH ¼ 6e10; adsorbent dose ¼ 1 g/L; T ¼ 25 ± 1 C). per m 3 water; acidification costs ¼ V 200.7 per m 3 water; for calculations refer to the SI), which is comparable to the widely applied active alumina (AA) whose limited defluorination capacity should be noted (normally <2 mg/g at neutral pH with initial F À of 10 mg/ L) (Mondal and George, 2015). In addition, the costs could be lower by using local raw materials (e.g. CaCO 3 , Ca(OH) 2 , Al(OH) 3 ) to synthesize LDHs. In combination with an environmentally friendly and cost-effective acidification method (e.g. aeration), CaeAleCO 3 LDH appears an attractive alternative for F À removal.

Conclusions
CaeAleCO 3 LDHs, either untreated, calcined or microwave treated, showed affinity for the removal of F À from synthetic groundwater in the environmentally-relevant lower concentration ranges (2e12 mg/L) to below the WHO guideline. F À removal capacities at near-neutral pH were 6.7e8.4 mg F À /g LDHs, with a higher capacity at lower pH (<8) and lower temperature (12 ± 1 C). Defluorination capacity of CaeAleCO 3 LDHs is higher than that of MgeAl LDHs (q ¼ 1.84 mg/g, initial F À ¼ 5 mg/L, pH ¼ 6) and active alumina (<2 mg/g at neutral pH with initial F À of 10 mg/L) at similarly environmentally relevant F À concentrations and the capacity is not necessarily reflected in BET measurements with N 2 . Given only marginal defluorination improvements by calcination/ microwave treatment, applying untreated LDH appears to be the practically most feasible option. Competition with Cl À and NO 3 À was not observed, whereas at higher HCO 3 À and SO 4 2À concentrations (>250 mg/L), a slight reduction in F À removal was observed. It is recommended to further investigate CaeAleCO 3 LDHs as a costeffective F À adsorbent with local raw materials, particularly in combination with pH correction to enhance its performance.