Calcium hydroxide recycled from waste eggshell resources for the effective recovery of fluoride from wastewater

In the hunt of waste recovery pathways, eggshells emerged as a potential adsorbent for fluoride because they contain plenty of calcium. However, as the main component, calcite has weak interaction with fluoride. In this study, calcium hydroxide was derived from waste eggshells successfully by an aging treatment with moisture for fluoride recovery from water. The X-ray diffraction (XRD) and infrared spectroscopy (FT-IR) analyses indicate that CaO in calcined egg shells (AEG900) is completely converted to calcium hydroxide. The adsorption experiments showed that the adsorption capacity of AEG900 for fluoride was improved by nearly 29.21% compared with the calcined eggshells without the aging treatment. In the batch experiment, the temperature effect is the most significant for the adsorption process, and nearly a half increment of removal rate is achieved by increasing the temperature by 30 °C. Further research revealed that the adsorption process fitted well with the pseudo-second order model and the Langmuir–Freundlich isotherm model, with a maximum adsorption capacity of 370.15 mg g−1. Moreover, precipitation was regarded as the main step for fluoride removal mechanism based on the calculated results of the surface complexation model. X-ray photoelectron spectroscopy (XPS) results showed that the stable fluorite formed in situ of AEG900 avoids calcium loss in water. Finally, AEG900 was applied in fluoride removal with real-life groundwater and industrial wastewater, and the results showed that the final fluoride concentration could meet the WHO requirement and industrial wastewater discharge standard.


Introduction
Fluoride generated from natural, anthropogenic, and industrial activities is widespread in different water sources. Especially, ouride will lead to different health effects for humans depending on the concentration. As an essential trace element for bone mineralization and enamel formation, uoride is benecial for strengthening the bone and teeth of young children at the level of 0.5 to 1.0 mg L −1 in drinking water. 1 However, the excessive ingestion of uoride will cause dental uorosis, skeletal uorosis, and other diseases; hence, the World Health Organization (WHO) suggests uoride concentration in drinking water below 1.5 mg L −1 . 2,3 Recently, most community water systems in the US have made more serious requirements for uoride concentration, which has been changed from 0.7 $ 1.2 mg L −1 to 0.7 mg L −1 . However, with increasing industrial activities, particularly in the semiconductor manufacturing and smelting industry, thousands of mg L −1 of uoride is discharged into natural water. 4,5 More than 30 mg L −1 in groundwater is detected, which is hazardous for water environment safety. Moreover, according to the statistics from Jadhav et al., China and India are most threatened by uoride, which nearly occupy 48.30% and 36.62% population of the selected countries. 6 Aer realizing the reason for uorosis, tremendous research studies have been carried out to keep uoride in water at an acceptable level. 7,8 The common treatment methods for uoride removal include chemical precipitation/coagulation, membrane separation, ion exchange, and adsorption. Precipitation/coagulation can be traced back to around 90 years ago. The famous Nalgonda process was developed to remove uoride in drinking water, which included two steps, the formation of insoluble uoride with lime and the coagulation treatment with alum. This process is of low cost and has high removal effectiveness, in which the initial concentration of uoride with 109 mg L −1 in groundwater can be reduced to 8 mg L −1 aer 35 min. 9 However, in over 10 years of practical applications, some serious drawbacks were found in the Nalgonda process, especially the negative effect of alum and excessive sludge. 10 Membrane technology is considered a reliable method for uoride removal in recent years, of which reverse osmosis has become popular to supply safe drinking water and over 95% of uoride can be removed by membrane separation, and the effluent concentration can be under 0.3 mg L −1 . 11 As a relatively new technique, membrane separation involves high costs for skillful operators, lm materials, and maintenance, which presents challenges for small-scale water treatment systems. Based on the US Environmental Protection Agency (EPA) report, the removal rate of uoride can be up to 100% by the adsorption process, which is recognized as one of the best available technologies for uoride control. 12 As the EPA recommended adsorbent, activated alumina is widely used to remove uoride and has been used in large-scale systems. 13,14 Although this process owns high selectivity and low cost, to achieve a high adsorption performance, a narrow pH range of 5.0-6.0 is required in this process because activated alumina will dissolve at pH < 5, and excessive aluminum will remain in the effluent causing secondary pollution. 15 At pH > 7, hydroxide and other anions will be more easily adsorbed than uoride. 16 On the other hand, while many synthesized adsorbents (mesoporous or modied alumina) display a high removal rate of uoride, their preparation methods are relatively complex and expensive. 17 Moreover, few commercial alumina samples can reduce uoride concentration under the safety limit for drinking water. For example, as Dayananda reported, the commercial alumina only could remove 39% uoride in water with the initial uoride concentration of 5 mg L −1 aer 10 h. 18 Hence, numerous adsorbents have been developed in recent years to achieve a high-efficient, low-cost, and stable adsorption process for uoride removal, in which bio-adsorbent is a potential candidate.
Bio-adsorbents are derived from industrial and agricultural wastes, which is a mode of resource utilization for wastes. Eggshells are abundant food waste from daily life, the food industry, and poultry farm. Apart from their use as fertilizers and feed additives, there are still a large number of eggshells without appropriate disposal. 19 Currently, EPA has ranked eggshells at 15 th as one of the environmental problem makers, and more than 100 000 dollars has been paid for the landll disposal of eggshells in the USA every year. 20 Due to being rich in calcium carbonate and a natural porous structure, eggshells have a potential application to adsorb heavy metals, 21 organic dyes, 22 and anions. 23 As a calcium-based material, eggshells own high uoride affinity and biocompatibility. In 2012, Bhaumik et al. rstly tried to use eggshells to adsorb uoride with 5 to 20 mg L −1 in water, and the maximum uoride adsorption capacity was 1.09 mg g −1 . 24 They found that the chemical adsorption acted as the main step, and a lower pH would release more Ca 2+ from calcite on the eggshell surface to benet uoride adsorption. While eggshells display a signicant advantage for uoride removal over other bio-adsorbents, there are fewer studies on this system. In 2021, Lee et al. restarted this topic, and they found that the CaCO 3 in natural eggshells changed to CaO by thermal treatment, 70% uoride could be removed within 15 min, and the maximum adsorption capacity of thermally treated eggshells could reach 258.28 mg g −1 . 25 Aer reviewing the relevant studies about the waste eggshell recycling as bioadsorbent, most studies were focused on discussing CaO as the main active component to capture anions in water. 26 In fact, CaO as a solid component in the calcined eggshells is unstable and hygroscopic, leading to an exothermic reaction with water and causing calcium ion loss in water.
Herein, we attempted to recycle Ca(OH) 2 from calcined eggshells and explore their application as a bio-adsorbent for uoride removal in water. To our best knowledge, this is the rst study reporting Ca(OH) 2 as the main active component, which is derived from waste eggshells and their application in uoride removal. The prepared eggshells with different treatment temperatures and aging treatment with moisture were characterized by different physicochemical methods to reveal the effect of composition and structural changes. Then, the batch experiments were conducted to discuss the factors inuencing the uoride removal by eggshells with Ca(OH) 2 as the main component and found that temperature signicantly affected the removal rate rather than pH, as reported in the literature. Moreover, using the equilibrium model calculation and XPS analysis, the uoride removal mechanism and the contribution of adsorption and precipitation by Ca(OH) 2 based eggshells were proposed. Finally, the adsorption tests were conducted with real-life industrial wastewater and groundwater to prove their potential for different application backgrounds for uoride removal.

Materials
Details of the information for materials are described in ESI. †

Preparation of the adsorbent materials
The collected eggshells were washed to remove grease and other impurities, and the eggshell membrane was peeled and dried at 100 C for 24 h. Then, eggshells were ground into small pieces and sieved by a 200-mesh screen to obtain an eggshell powder with a particle size of around 70 mm, as shown in Fig. 1(a). The sieved eggshells were used in the adsorption experiment as the contrast sample named EG. Then, the eggshell powder was heated to 500, 600, 700, 800, and 900 C in the muffle furnace under air atmosphere for 4 h. As shown in Fig. 1(b)-(f), the obtained samples were denoted as ES500, ES600, ES700, ES800, and ES900, respectively. All calcined samples were stored in a dryer before use in the adsorption experiments, and ES800 and ES900 were selected for further treatment. The aging treatment was conducted in moisture, in which ES800 and ES900 were stored in a box with a relative humidity of 70% for one week. It is worth noting that before the sample characterization and adsorption experiments, the aged samples were dried at 80 C for 24 h to remove free water.

Adsorption performance
2.3.1 Batch experiment. The adsorption experiments were conducted with uoride solution in batches to determine the effects of different parameters. A typical batch experiment was carried out as follows. A certain concentration of uoride solution and adsorbents was added to a 50 mL of polypropylene centrifugal tube. Aer adjusting the pH, the samples were placed in a shaking incubator operating at 200 rpm and a set temperature for 12 h. Aer completed adsorption, the samples were centrifuged, and the supernatants were ltered with a 0.45 mm nylon microporous membrane. To measure the residual concentration of uoride, the ltrate was diluted 50 times to meet the detection limit of ion chromatography (761, Metrohm, Switzerland). The effect of co-existing ions was investigated by adding 10 mM NaHCO 3 , NaSO 4 , K 2 HPO 4 , and NaCl to the initial uoride solution, respectively. The nal concentration of uoride (C e (mg L −1 )) was recorded, and the removal rate of uoride and the adsorption capacity of adsorbents were calculated as follows: where C 0 (mg L −1 ) is the initial concentration of uoride, m (g) is the mass of adsorbents, and V (L) is the solution volume. All experiments were repeated thrice with the error range within 3%, and the average value was discussed.

Adsorption isotherms.
To describe the adsorption behavior of adsorbents, a series of adsorption experiments were conducted. Briey, 1 g L −1 adsorbent was added into the uoride solution with the concentration of 100 to 1000 mg L −1 with pH at 7, and then placed in a shaking incubator operating at 200 rpm and 20 C for 12 h, ensuring to reach the adsorption equilibrium. The residual concentration of uoride was measured and tted by the Langmuir (eqn (3)), Freundlich (eqn (4)), and Langmuir-Freundlich (eqn (5)) isotherm models. 27 Langmuir model: Freundlich model: Langmuir-Freundlich model: where q m (mg g −1 ) and q m,LF (mg g −1 ) are the maximum adsorption capacity of the Langmuir model and Langmuir-Freundlich model, respectively, n is the Freundlich constant, and n LF is the index of heterogeneity varying from 0 and 1. K L , K F , and K FL are the binding constants for the Langmuir model and Freundlich model, respectively.

Adsorption kinetics.
Kinetics experiments were carried out for 1 g L −1 adsorbent and 600 mg L −1 uoride solution at pH 7 and 20 C, and the adsorption time was set from 10 min to 12 h. The residual concentrations of uoride at different times were measured and tted by the pseudo-rst order (eqn (6)) and pseudo-second order (eqn (7)) equations, respectively.
Pseudo-rst-order model: Pseudo-second-order model: where q t (mg g −1 ) is the adsorption amount at different times (t), k 1 (h −1 ) is the pseudo-rst-order constant, and k 2 (g (mg −1 h −1 )) is the pseudo-second-order constant.

Characterization method
Scanning electron microscopy (SEM, Phenom™ ProX Desktop SEM) was used to observe the surface morphology of eggshells aer thermal treatment. The specic surface area and pore size of eggshells were measured by the Accelerated Surface Area and Porosimetry System (SSA-4200, Beijing Builder, China) and calculated based on the Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) methods. The Fourier transform infrared spectroscopy (FT-IR, Nicolet is 50, USA) was employed to measure the functional groups on eggshells, and the tablet method with KBr was adopted with the measuring wavelength from 400 to 4000 cm −1 . The X-ray diffraction (XRD) analysis was conducted to determine the crystal structure and components of the sample by a multi-function diffractometer (DX-2700BH, Haoyuan, China), equipped with a copper (Cu Ka) radiation (l ¼ 1.541Å) scanning from 10 to 80 . The synchronous thermal analysis (TGA/DSC1, Mettler Toledo, Switzerland) was used to show the mass change of eggshells aer thermal treatment at different temperatures. The measurement temperature was from 30 to 1000 C with an increased rate of 20 C min −1 under air atmosphere. Moreover, X-ray photoelectron spectroscopy (XPS, Thermo Scientic K-Alpha, UK) of eggshells before and aer uoride adsorption was measured.

Screening of adsorbent
The effects of calcination temperature and humidity on the preparation of adsorbents are discussed in this section. As shown in Fig. 1, the adsorbents were different visually with different treatment temperatures. With the increase in calcination temperature, the color of eggshells becomes darker and darker until it is black below 700 C. However, below 800 C, the color of the powder will turn white, and all of them become white below 900 C. Aging treatment also can affect the color of the powder, and the white color disappears in AES800. Those color changes reect the change of components in eggshells aer the treatment. Fig. 2(a) shows the TGA curve of eggshells, and a slight mass variation happens before 600 C with the weight loss of 3.88%, which indicates organic matter in eggshells has become charred, and some of them escaped as CO 2 . Then, a big drop of mass loss is found at about 42.55% until 800 C, which means the thermal decomposition of CaCO 3 occurs according to eqn (8). Aer 800 C, the eggshell mass remains stable, meaning the decomposition of CaCO 3 has been completed. Moreover, a distinct inverted peak occurs between 700 to 850 C in the DSC curve (Le axis in Fig. 2(a)), which is ascribed to the endothermic decomposition of CaCO 3 . It is worth pointing out that the total weight loss of eggshells was 46.43% in this study, which is higher than in a reported work with 40%. 23 Different calcination environments could cause this difference. In this study, air was chosen because of easier operation, and organic matter would become CO 2 instead of char.
CaCO 3 (s) + heat / CaO(s) + CO 2 (g) (8) Fig. 2(b) shows the uoride adsorption on eggshells with different treatment processes. The equilibrium concentration of uoride decreases from 590.11 to 317.33 mg L −1 by increasing the calcination temperature to 900 C. Obviously, natural eggshells and calcined eggshells obtained below 600 C are not feasible for uoride removal, and only 3.68% uoride is removed for ES600. However, when the calcination temperature increases from 700 to 900 C, the adsorption capacity of eggshells escalates from 89.07 mg g −1 (ES700) to 282.67 mg g −1 (ES900), which implies that the component change of eggshells may play the main role for the improvement of uoride adsorption. A similar tendency is also found in Lee's work. 25 In their study, the adsorption capacity of ES800 (18.54 mg g −1 ) was similar to ES900 (18.56 mg g −1 ); however, in our study, the adsorption capacity of ES900 is obviously larger than that of ES800 (230.42 mg g −1 ). Apart from the effect of different calcination conditions, BET results could explain this discrepancy to a certain extent. Indeed, the specic surface area (SSA) of ES900 (2.785 m 2 g −1 ) was more than twice that of ES800 (1.08 m 2 g −1 ), which could provide more active sites for uoride. Moreover, aged eggshells display higher adsorption capacity for uoride, which increases to 333.70 mg g −1 for AES800 and 365.24 mg g −1 for AES900, respectively. This enhancement is considerable, in which the equilibrium concentration drops to 234.76 mg L −1 for AES900, meaning 61% uoride could be removed even with a high initial concentration. Eqn (9) is the main reaction that happens in the aging stage in eggshells, where CaO adsorbing moisture from air would convert to hydrated lime, which is more favorable for uoride removal.
In order to identify the components formed in the eggshells at different calcination temperatures, XRD patterns of each adsorbent were obtained, as shown in Fig. 3(a). The results conrm that when the calcination temperature is below 700 C, the major component in the samples is calcite (JCPDS No. 05-0586). When eggshells are heated under 800 C, lime is formed in ES800 (JCPDS No. 37-1497). By comparing the XRD patterns of ES800 and ES900, it can be deduced that almost all CaCO 3 converts to CaO at 900 C because only the pattern of lime can be found in the ES900 XRD pattern, and the same result was reported in Risso's study, who attempted to obtain CaO catalysts from eggshells at 900 C. 28 Unlike that described in Lee's study, Ca(OH) 2 is not present in fresh ES800 and ES900 due to strict dry storage conditions. 25 Conversely, Ca(OH) 2 (JCPDS No. 04-0733) is formed aer storing in humid air due to eqn (9), and CaO almost disappears in aged samples, as shown in Fig. 3(b). Moreover, the proportion of Ca(OH) 2 in AES900 is obviously higher than that in AES800, which could be the reason for a higher adsorption capacity for uoride for AES900, as mentioned in Fig. 2(b). Panagiotou et al. suggested that Ca(OH) 2 was more soluble than CaO or CaCO 3 , so soluble Ca in water would be more reactive with the target ions, and the reason for the removal of ions was precipitation rather than adsorption. 29 To prove that the adsorption happened on eggshells, the XRD result of AES900 aer uoride adsorption to equilibrium is shown in Fig. 3(b), and CaF 2 (JCPDS No. 35-0816) can be found from the main diffraction patterns, which indicates that the ion exchange happens and generates uorite on the surface of AES900 based on eqn (10). 30 Moreover, the peak of Ca(OH) 2 almost disappears, but the peak of CaCO 3 can still be found in the XRD spectrum of AES900 aer adsorption. Hence, Ca(OH) 2 plays the main role in uoride removal, and aging treatment with humid air is very important to improve the adsorption capacity of calcined eggshells.
To obtain more information about the structure of adsorbents, the FT-IR spectra of adsorbents thermally treated above 800 C were measured, as shown in Fig. 4(a). Before adsorption, the exhibited adsorption peaks include 3633, 3440, 1411, 1050, and 871 cm −1 for all samples. Among them, the sharp peak at 3633 cm −1 and the broad peak at 3440 cm −1 are attributed to the bending vibrations of -OH from Ca(OH) 2 and crystalline hydrate, respectively. 31,32 A higher relative intensity peak at 3633 cm −1 can be found in the aged samples but not in calcined samples at 800 or 900 C, which is in agreement with the XRD result that more Ca(OH) 2 is generated by aging treatment. The peaks at 1411 and 1050 cm −1 and the peaks at 709 and 2513 cm −1 in ES800 and AES800 belong to the vibrations of C-O bonds. 33 Moreover, the peaks at 2976, 2859, and 1791 cm −1 are attributed to the C]O bonds from CO 3 2− , and the abovementioned bands indicate the presence of CaCO 3 in ES800 and AES800. 31 However, the peaks belonging to CaCO 3 are not found in the sample calcined 900 C, which could be explained by a lower relative peak intensity. Moreover, the sharp bands at 871 cm −1 correspond to the Ca-O bond, and the broad peaks at 498 cm −1 are from the Ca]O bonds. 34 Aer adsorption, the peak of -OH bond at 3633 cm −1 disappears in AES900, which indicates that the active site of Ca-OH on the surface reacted with F − . The peak of Ca-O bond at 871 cm −1 becomes weak and the peak of Ca]O bonds disappears and these indicate the reaction happened between Ca and F. The change of AES900 before and aer adsorption conrms the key role of Ca(OH) 2 for uoride adsorption again. 35 Furthermore, the SEM image of  AES900 was taken, as shown in Fig. 4(b). Aer calcination, the pore size of the eggshell powder became smaller (<5 mm), which is also reected in the BET data (SSA increased from 0.515 to 2.785 m 2 g −1 ). Moreover, the common pores from natural eggshells could not be found in the SEM image because the average pore size was 12.8 nm for AES900. 36 Moreover, the nal pH and stability of adsorbents also were discussed, and the results are listed in Table S1. † During the adsorption experiment, the initial pH was around 6.66, and the nal pH was around 10.51 for natural eggshells and eggshells with calcination temperature below 700 C. There is no obvious difference in the nal pH for eggshells with or without aging treatment, which are all around 12.01. The nal pH increase is a common phenomenon in Ca-based materials for uoride removal and also happens in the case of sepiolite. 37 The stability of adsorbents was evaluated by measuring the weight loss of adsorbents during the adsorption experiment and calculated as TDS (%) shown in Table S1. † TDS displayed the proportion of weight loss of adsorbents to raw adsorbents aer adsorption of uoride. From the data in Table S1, † it is clear that the stability of natural eggshells and calcined eggshells obtained below 700 C are very weak, and more than 50% weight loss happened in the uoride adsorption process. Combined with pH change, it could be deduced that CaCO 3 as the main component in these materials would dissolve in water with a weak acidic condition (pH ¼ 6.66) causing mass loss, and the increase in the nal pH reects the consumption of H + during the dissolution of calcite. 38,39 Moreover, eggshells obtained by thermal treatment above 800 C show higher stability. The values of TDS for EG800 and EG900 are 14.33% and 10%, respectively. As the main reason, CaO in those materials reacts with water to form Ca(OH) 2 , and meanwhile, Ca dissolves in water, causing mass loss, and Santos et al. gave a similar explanation for the weight loss of EG800. 23 In contrast, the stability of aged eggshells is the best, and the values of TDS are 4.67% and 3.00% for AEG800 and AEG900, respectively. In this system, the reaction between uoride and Ca(OH) 2 might have a major role, and forming CaF 2 covered on the surface of eggshells would prevent Ca loss.
Hence, above all, AES900 displays an excellent adsorption capacity and better stability for the removal of high concentrations of uoride in water, and the adsorption performance of AEG900 with different effects will be discussed in the next section comprehensively.

Batch adsorption experiments
3.2.1 Effects of pH, temperature, other ions on adsorption. The adsorbent dosage, pH, temperature, and other ions as the main operating parameters were discussed, and the results are shown in Fig. 5. Fig. 5(a) shows that with an increase of AES900 dosage from 0.33 to 3 g L −1 , the removal rate of uoride increases from 25.92% to 99.86% for 600 mg L −1 uoride solution. It is worth noting that the increment rate gradually decreases, especially between the dosage of 2.33 (97.93%) and 3 g L −1 (99.86%), in which the removal rate of uoride only increases by 1.93%. It can be explained by the weak driving force for adsorption in a low concentration of uoride (the nal concentration is 10.55 and 0.83 mg L −1 , respectively). 40 On the other hand, the adsorption capacity decreases with increasing dosage of AES900, which means the increase in unsaturated active sites, and excess active sites will not improve uoride adsorption signicantly. 41 Moreover, compared with other studies, the adsorption capacity is 199.72 mg g −1 for 3 g L −1 of AES900, which is higher than other adsorption systems for uoride removal. For example, the adsorption capacity of CaO 20 @Al 2 O 3 with 3 g L −1 for 600 mg L −1 uoride was about 120 mg g −1 . 42 Fig. 5(b) shows the effect of temperature on uoride adsorption. With increasing temperature from 10 to 40 C, the removal rate of uoride increases from 35.0 to 87.44%. This increment is signicant (nearly 52.44%) within a 30 C increase compared with other studies. 43,44 This is probably because the increasing temperature could increase the diffusion of uoride to the adsorbent and enhance the reaction between Ca(OH) 2 and F − . Fig. 5(c) shows the effect of initial pH for uoride adsorption and nal pH in solution. With an increase in pH, the adsorption capacity increases from 348.89 mg g −1 at pH 3, peaking at 364.77 mg g −1 at pH 5 and then decreases to 333.86 mg g −1 at pH 11. This tendency is similar to Saini's work and their plots of the speciation curve of uoride in solution at different pH could be helpful in understanding this phenomenon. 43 Firstly, at pH below 3, more than half of uoride will present as HF, which will further form a stable association to reduce the concentration of free uoride ions. 45 When pH increases to 5, most of the uoride will present as free uoride ions, which is favorable for reaction with the adsorbent. On the other hand, in weak acidic conditions, more active sites will explore and carry a positive charge, which will capture uoride ions easily by electrostatic attraction. 39 At a basic pH, while all the uorides are present as free ions, the adsorption capacity still decreases, which is mainly due to the physical changes on the surface of AES900. In alkaline conditions, the surface of the adsorbent will carry a negative charge, which will repel the uoride ions. 39 Moreover, OH − can act as a competitor for uoride adsorption, and more OH − will occupy the active sites, leading to a decrease in adsorption capacity for uoride. 46 Apart from uoride ions, other ions also exist in the natural water and also can adsorb on eggshells acting as a competitor in the uoride adsorption process. Herein, chloride, bicarbonate, sulfate, and hydrogen phosphate as typical competitive ions were discussed. Fig. 5(d) shows the effects of other ions on the adsorption of uoride. It can be found that co-existing ions have negative effects on uoride adsorption and the adsorption capacity of uoride decreases in the order of HPO 4 2− > HCO 3 − > SO 4 2− > Cl − , which is similar to other studies. 25,47 Among them, the negative effects of chloride and sulfate are not signicant, and the adsorption capacities of uoride only decrease by 2.42% and 5.77%, respectively. Bicarbonate and hydrogen phosphate have remarkable inhibitive effects on uoride adsorption, and the adsorption capacities of uoride decrease by about 36.32% and 89.06%, respectively. The main reason for the competitive adsorption of bicarbonate is the increase in pH, which is adjustable before the adsorption process. However, the negative effect of hydrogen phosphate is inescapable, and to ensure the feasibility of uoride removal, dephosphorization treatment for the inow is necessary.

Adsorption kinetics.
In order to understand the rate controlling step of mass transfer and the effect of contact time for uoride adsorption on AEG900, the pseudo-rst and pseudo-second order models were applied to determine the adsorption rate of uoride on AEG900. Fig. 6 depicts the measured adsorption capacity as a function of contact time and their tting curves with two kinetic models. The result shows that the adsorption capacity increases sharply within the rst 100 min, and active sites are fast occupied by uoride. Then, with the decrease of active sites, the adsorption capacity increases slowly near the adsorption equilibrium until 720 min. Based on the R 2 listed in Table 1, pseudo-second order model is better for describing the adsorption process of uoride on AES900 (R 2 > 0.989), and the calculated q e is similar to the experimentally measured value with a relative error of 3%. The pseudo-second order model assumes that chemisorption is the controlling step, which describes the sharing or exchanging charge between adsorbent and adsorbate to generate new compounds. Hence, precipitation, diffusion, and surface adsorption could happen in the adsorption of uoride onto AEG900. Moreover, the experimental data also is tted in the intra-particle diffusion model (eqn (S1) †), and the tting result is shown in Fig. S1. † It can be found that the adsorption process can be divided into three continuous steps involving surface diffusion, intra-particle diffusion, and the equilibrium stage of adsorption and desorption. None of these three plots passes through the original point (C 1 s 0), meaning the internal diffusion is not a single step controlling the whole adsorption process. It also proves that the adsorption of uoride onto AES900 involves more than one mechanism. 46 3.2.3 Adsorption isotherm. The effect of initial concentration for uoride adsorption on AEG900 was investigated, as shown in Fig. 7(a). The adsorption capacity increases rstly from 87.57 to 349.56 mg g −1 as the initial concentration increases from 100 to 500 mg L −1 , and then with further increasing initial concentration, the adsorption capacity remains constant. Inversely, the removal rate decreases from 88.28% to 38.71%, with the initial concentration increasing from 300 to 900 mg L −1 . A slight increase in removal rate occurs within the initial concentration between 100 to 300 mg L −1 , which means a low initial concentration may affect the uoride adsorption due to the lack of driving force, so for the removal of uoride with a low initial concentration in water, the feasibility of AEG900 should be discussed further.
Moreover, the experimental data is tted in isotherm models, including Langmuir, Freundlich, and Langmuir-Freundlich models, as shown in Fig. 7(b), and the obtained parameters are listed in Table 2. Clearly, the Freundlich model is not suitable for describing the uoride adsorption onto AES900 with R 2 > 0.87. While the R 2 of tting the Langmuir curve is higher than 0.94, the simulated q m is 380.93 mg g −1 with a 4.11% relative error for the experimental value. The tting result by the Langmuir-Freundlich model has the highest correlation (R 2 > 0.96), and the calculated q m,LF only has a 1.37% relative error for experimental value. Langmuir-Freundlich model is favorable for heterogeneous systems with the variable density function, which is determined by n LF . 48 If n LF equals 1, the adsorbent material is homogeneous, and the Langmuir-Freundlich model is reduced to the Langmuir model. In this study, n LF is 0.34, indicating AEG900 is a heterogeneous material to adsorb uoride.

Removal mechanism analysis
3.3.1 Surface complexation modeling. According to the discussion result of batch adsorption experiments, the kinetic analysis points that the adsorption process of uoride on AEG900 tted well with the pseudo-second order models, which means adsorption and co-precipitation both participate in the removal of uoride. Moreover, the adsorption isotherm curve ts well with the Langmuir-Freundlich model, indicating the adsorption process is monolayer adsorption, and the adsorbent is a heterogeneous material. Therefore, the contribution degree of adsorption and precipitation should be discussed rst to reveal the whole removal mechanism of uoride by AEG900. Herein, the Visual MINTEQ 3.1 soware was employed to calculate the chemical equilibrium to predict the coprecipitation behavior between uoride and calcium ions. 46 Firstly, the set concentration of uoride ions and the experimentally obtained total dissolved concentration of calcium ions were used as inputs for an aqueous system. The total dissolved concentration of calcium ions was measured as 266 mg L −1 for 1 g L −1 AEG900. Then, a solid phase in this aqueous system by the three planes model (Calcite-CDM) to form a complexation surface was added. The input parameters for the complexation surface are listed in Table S2, † including the SSA, dosage, and equilibrium pH from the experimental data. As one of the output results, saturation indices (SI) for minerals are obtained to predict the formation potential of possible components in this system, of which the positive value means oversaturation, zero means apparent equilibrium, and the negative value means undersaturation. When the initial uoride concentration is 600 mg L −1 , the SI of uorite, lime, and portlandite is 0, -16.19, and −6.00, respectively, which indicates that only  uorite is a possible co-precipitated product. Moreover, the equilibrium mass distribution of uoride and calcium ions between dissolved and precipitated phases predicted by Visual MINTEQ 3.1 are summarized in Table S3, † of which the equilibrium mass distribution of uoride is plotted in Fig. 8. As shown in Fig. 8, when the initial uoride concentration is between 50-200 mg L −1 , more than 95.67% uoride will be precipitated with calcium to form uorite. As uoride concentration increases, the proportion of dissolved uoride will increase from 1.83% to 72.0% due to the excess uoride ions, and the proportion of precipitated uoride will decrease correspondingly. By combining the predicted and experimental data, the amount of uoride at different phases is summarized in Table 3. Obviously, the dominant mechanism of uoride removal for AEG900 is co-precipitation when the initial uoride concentration is below 200 mg L −1 . Fluoride ions will form CaF 2 preferentially, with sufficient calcium ions dissolving from AEG900. Moreover, the predicted precipitation value is higher than the experimental value because the adsorption equilibrium is hard to reach at a low initial concentration within 12 h. When the initial uoride concentration increases over 200 mg L −1 , the co-precipitation and adsorption will both act as the removal mechanism for uoride, in which the co-precipitation is predominant. We also measured the calcium concentration in solution for the initial uoride concentration at 600 mg L −1 and found that the residual calcium concentration was less than 1.33 mg L −1 . Hence, it proves again that most of the calcium ions will react with uoride to form co-precipitated products.

XPS analysis.
Based on the analysis result of surface complexation modeling, precipitation plays a predominant role for uoride removal. As previously reported, while quick lime could remove uoride efficiently under suitable alkaline conditions, the separation and disposal of produced sludge will cause some extra load for the treatment system. 49 However, based on the XRD results shown in Fig. 2(a), uoridate precipitation could happen on the eggshell surface. To further reveal the removal mechanism of uoride by eggshells, XPS analysis was carried out for AEG900 before and aer the adsorption of uoride. As shown in Fig. 9(a), besides Ca 2p, O 1s and C 1s signals can be found in the low energy regions of the full XPS spectra in the two samples, and F 1s signal can be clearly found in AEG900 aer uoride removal, which conrms the migration of uoride from water to AEG900. The intensity of the F 1s signal also is considerable compared with that of other elements, and by a rough calculation, the proportion of uoride atoms is about 35.36% in AEG900. Moreover, the F 1s spectra were tted and the binding energy of the F 1s signal is 685.32 eV, as shown in Fig. 9(e), while the binding energy of NaF (684.50 eV) is lower, which implies the reaction between calcium and uoride ions to form CaF 2 rather than the crystallization of NaF onto AEG900. 32 The Ca 2p spectra also was tted and decomposed into different species as shown in Fig. 9(c). It is clearly seen that the peaks of Ca 2p 1/2 and Ca 2p 3/2 shi before and aer the uoride removal due to the higher electronegativity of uoride. For raw AEG900, the binding energies of Ca 2p in 347.32 and 350.62 eV belongs to Ca(OH) 2 and those of 346.58 and 349.88 eV belong to CaCO 3 . 50 Based on the tted result, the contribution of Ca(OH) 2 is 43.07% and that of CaCO 3 is 56.93%, which is in accordance with the XRD result as shown in Fig. 2(b). Aer uoride removal, the Ca 2p signal can be divided into two species, including CaCO 3 with the binding energies of 346.85 and 350.15 eV and CaF 2 with the binding energies of 348.09 and 351.63 eV. 25 It can be inferred   Moreover, as shown in Fig. 9(b), the change of C 1s spectra for AEG900 before and aer adsorption are slight, and the two peaks belong to CaCO 3 and C-C at 289.26 and 284.98 eV, respectively. For the O 1s spectra in Fig. 9(d), the peaks of metal carbonate and metal hydroxide can be decomposed with the binding energies at 531.34 and 532.88 eV, respectively, for raw AEG900. However, aer the removal of uoride, the metal hydroxide peak disappeared in the O 1s spectra, implying the dissociation of hydroxide accompanied by calcium dissolution from AEG900 in water. This conclusion is consistent with the FT-IR result of AEG900, as shown in Fig. 3, in which a sharp peak of -OH groups at 3633 cm −1 disappeared aer the removal of uoride. In conclusion, XPS analysis indicates that a strong interaction occurs between calcium and uoride to form a stable component on the adsorbent, in which Ca(OH) 2 is consumed preferentially compared with CaCO 3 .
3.3.3 Mechanism of uoride removal by AEG900. On the basis of material characterization, the batch adsorption experiment, and model analysis, the removal mechanism of uoride by AEG900 is shown in Scheme 1. By thermal treatment, part of CaCO 3 in eggshells will convert to CaO, which could improve the dissolution of Ca ions to capture uoride ions in water. However, compared with Ca(OH) 2 , the ability to release Ca ions for CaO is weaker, so by the aging treatment under a humid environment, CaO can convert to Ca(OH) 2 with higher dissolution capacity. Hence, unlike the common adsorbent recovery from eggshells, Ca(OH) 2 becomes the core component in AEG900. Combined with the results in the single-factor experiment, this adsorption process ts with the pseudo-second kinetic and Langmuir-Freundlich models well, indicating the presence of chemisorption and that the adsorbent material is heterogeneous. By the mass equilibrium analysis, dissolution, co-precipitation, and adsorption are involved in the removal process of uoride by AEG900. Based on the surface complexation modeling predicted, precipitation is the dominant mechanism for uoride removal. On the other hand, while the difference in SSA for eggshells treated under different conditions is slight, the removal efficiency of uoride is enhanced by the aging treatment, indicating that adsorption is not a key point for uoride removal. So, based on model prediction, adsorption only works under a high uoride concentration. Moreover, the in situ precipitation of uorite is found based on the characterization results of XPS and XRD. Correspondingly, the mass loss of AEG900 before and aer use is negligible. Hence, the removal mechanism of uoride by AEG900 is precipitation and adsorption, in which in situ precipitation preferentially occurs on the surface of AEG900.  a Calculated from initial uoride concentration minus nal uoride concentration and the calculated adsorbed uoride concentration. b Calculated from predicted total dissolved uoride minus nal uoride concentration in the experiment. If the result is negative, the value of adsorbed uoride is regarded as 0.

Evaluation of uoride removal in real-life water
To show the feasibility of AEG900 in different application environments, the uoride removal experiments were conducted with different concentration levels of uoride in deionized water, real groundwater, and industrial wastewater, and the results are shown in Fig. 10. According to the experimental results, we can nd that when the dosage of AEG900 is 3 g L −1 , the nal concentration of uoride can reach 0.83 mg L −1 , meeting the guidelines of WHO (1.5 mg L −1 ). To investigate its feasibility for low uoride concentration removal, the same dosage of AEG900 was used to treat uoride with an initial concentration of 10 mg L −1 , and the nal uoride is 1.43 mg L −1 . A satisfying result can be obtained, while the low initial concentration will limit the driving force for the interaction between uoride ions and adsorbents. Additionally, AEG900 was applied to remove uoride in different real water samples. The groundwater was collected in Chengdu city, China, and the industrial water was collected from the photovoltaic industry, and the detailed characterizations are listed in Table S4. † Based on the results of pre- Fig. 9 XPS curves of AEG900 before and after fluoride adsorption (a), C 1s spectrum (b), Ca 2p spectrum (c), O 1s spectrum (d) for AEG900 before and after fluoride adsorption and F 1s spectrum (e) for AEG900 after fluoride adsorption.

Scheme 1
The removal mechanism of fluoride by AEG900. experiments, the optimum conditions were chosen for different water samples, and the removal rates of uoride were 99.77% for industrial wastewater and 75.54% for groundwater. It is clear that AEG900 is efficient in treating uoride in groundwater to meet the WHO standards. For photovoltaic wastewater, lime precipitation is the most mature method for uoride removal, but the nal uoride concentration is commonly in the range of 20 to 100 mg L −1 , which is higher than the emission standard (15 mg L −1 ). 51 However, in this study, by in situ precipitation on AEG900, the lower nal uoride concentration can be achieved without a turbidity increase in wastewater.
To sum up, AEG900 is feasible for uoride removal within a wide range of uoride concentrations and can reduce uoride to the permissible concentration suggested by WHO. It also can be used in different application scenarios, and by adjusting the adsorption condition, it is feasible for industrial wastewater or groundwater treatment.

Comparison of other adsorbents reported in the literature
To conrm the superiority of AEG900, the uoride adsorption experiment was conducted under the same condition as Lee's study. 25 The adsorption capacity is 56.84 mg g −1 for AEG900 (F − ] 0 ¼ 200 mg L −1 and [adsorbent] ¼ 3.33 g L −1 ), which is nearly 3 times that of eggshells treated under 800 C for 4 h as Lee et al. reported (18.54 mg g −1 ). Moreover, the equilibrium adsorption capacities of uoride by different adsorbents in the literature are plotted in Fig. 11. To make the comparison meaningful, the adsorption capacities were compared with different dosages of adsorbents and the initial concentration of uoride higher than 100 mg L −1 was discussed in these studies. Obviously, AEG900 owns outstanding adsorption capacity for uoride removal, especially compared with other absorbents recovered from waste. For example, under similar conditions, the adsorption capacity of AEG900 (370.15 mg g −1 ) was nearly three times that of nepheline prepared from kaolinite (125.0 mg g −1 ). 44 Moreover, compared with other calcium-based materials, AEG900 also shows higher adsorption performance. For example, the adsorption capacity of CaSO 4 $2H 2 O was 128.08 mg g −1 with 1.5 g L −1 dosage, which is even less than the adsorption capacity of AEG900 for uoride with 3 g L −1 dosage. 32 Besides, some synthetic adsorbents displayed higher adsorption performance than AEG900, such as calcium hydroxide nanorods (450 mg g −1 ). Compared with AEG900 recovered from eggshells, nano-size calcium hydroxide with higher SSA and the content of Ca(OH) 2 is favorable for the adsorption and precipitation of uoride in water, but meanwhile, the higher preparation cost and the risk of nanoparticle leaching remains. 46 Hence, AEG900, with its low-cost, easy preparation method, and high uoride removal efficiency, displays application potential for uoride removal.

Conclusion
In this study, a modied preparation method was proposed to improve uoride removal in an aqueous solution by Ca(OH) 2 derived from calcined eggshells. By an aging treatment in a humid environment, CaO as the main active component in calcined eggshells could convert to Ca(OH) 2 to realize a great degree of improvement for uoride removal, wherein the uoride removal rate of EG900 increased nearly by 29.21%. The single-factor experiments demonstrate that AEG900 owns the characteristic of calcium-based adsorbents for uoride removal, and the adsorption equilibrium process was well tted with the Langmuir-Freundlich isotherm models with the maximum adsorption capacity of 370.15 mg g −1 . Based on the adsorbent material characterization, surface complexation model analysis,  and adsorption performance results, the adsorption and precipitation both contribute to the uoride removal mechanism, in which the contribution of precipitation is bigger. Fluorite as a product is formed in situ, which avoids increased turbidity in the lime precipitation method. Moreover, satisfactory results of uoride removal by AEG900 in different initial uoride concentrations, real-life groundwater, and photovoltaic industrial wastewater were achieved, and the nal uoride could reach the requirement of WHO and the discharge standard of industrial wastewater. Overall, by aging treatment, calcined eggshells were endowed with outstanding deuorination capacity, which could be applied as a potential adsorbent for uoride removal.

Conflicts of interest
There are no conicts to declare.