The performance of calcined serpentine to simultaneously remove fluoride, iron and manganese

To solve the problem of high fluoride, iron and manganese concentrations in groundwater, serpentine (Srp) was modified by metal salt impregnation, acid-base activation and calcination, and the effects of these three modifications on removal performance of Srp were compared. Specifically, the effects of the calcined serpentine (Csrp) dose, reaction time, pH, and temperature on the removal performance of F , Fe2þ and Mn2þ on Csrp were analysed. An isothermal adsorption model and adsorption kinetic equation were established and confirmed through SEM, EDS, XRD and FTIR spectroscopy to analyse the mechanism of removing F , Fe2þ and Mn2þ by Csrp. The results show that when 3 g/L Csrp was used to treat water samples with 5 mg/L F , 20 mg/L Fe2þ, and 5 mg/L Mn2þ (pH of 6, reaction temperature of 35 °C, and time of 150 min), the removal rates of F , Fe2þ, and Mn2þ were 94.3%, 99.0%, 98.9%, respectively. The adsorption of F , Fe2þ and Mn2þ on Csrp follows the quasi-second-order kinetic equation and Langmuir isotherm adsorption model. After 5 cycles of regeneration of Csrp, Csrp can still maintain good properties of fluoride,iron and manganese removal.

Fluoride is widely present in minerals, natural water systems and geological deposits and enters the food chain through human consumption or vegetation (Jin et al. 2016). The excessive intake of fluoride affects human health and causes fluorosis (Cai et al. 2015). Iron and manganese ions exist in nature in the form of dissolved single ions (Fe 2þ , Mn 2þ ) or in the form of undissolved Fe(OH) 3 and MnO 2 (Barloková & Ilavský 2010). Fe 2þ and Mn 2þ form hydroxides after oxidation, which affects the colour, taste and turbidity of water (Homoncik et al. 2010). These insoluble hydroxides can produce toxic derivatives in the body and cause many physical diseases. The World Health Organization stipulates that the maximum limit of fluoride is 1.5 mg/L (Zawar et al. 2020). The EU Directive 2020/2184 sets the following limits for aesthetic reasons: Fe , 0.2 mg/L and Mn , 0.05 mg/L. Therefore, the removal of fluoride, iron, and manganese from groundwater is an urgent need to solve the problem of using groundwater that contains fluoride, iron, and manganese as drinking water.
There are many studies on the separate removal of fluoride, iron and manganese both at home and abroad (Kwakye-Awuah et al. 2019;Yadav et al. 2019;Aziz et al. 2020). However, there are few studies on their simultaneous removal, and thus, it is essential to explore how to simultaneously remove fluoride, iron and manganese in groundwater (Zhang et al. 2009). At present, the methods for treating fluoride-containing groundwater mainly include ion exchange, electrolysis, reverse osmosis, nanofiltration and adsorption (Abri et al. 2019). The main approaches to treat groundwater containing iron and manganese include filtration, contact oxidation, microbial treatment, ion exchange, ultra/microfiltration membranes, and adsorption (Chaturvedi & Dave 2012;Patil et al. 2016). Among the above methods, the adsorption method is the first choice to remediate groundwater pollution due to its low cost, high adsorption capacity, and low energy consumption (Meski et al. 2019). Optimizing adsorbents for the simultaneous removal of fluorine, iron and manganese is the key to research and application. Natural minerals are inexpensive and have good chemical stability; thus, they are often used to adsorb fluoride or heavy metal ions from water. Serpentine, as a magnesium-rich silicate mineral, has abundant reserves in China and is widely distributed throughout areas such as the Liaoning Province and Shanxi Province. Serpentine itself contains a large number of active groups, such as hydroxyl groups and unsaturated Si-O-Si, O-Si-O, magnesium and hydrogen bonds, which can be combined with metal ions to undergo ion exchange and surface coordination reactions (Cattaneo et al. 2003;Shaban et al. 2018). Thus, the adsorption performance of modified serpentine can be significantly improved. Mobarak et al. (2019) used chemically modified serpentine to remove Cr 6þ and F À at rates of 87.31% and 94.72%, respectively. Huang et al. (2017) used mechanically activated serpentine to remove Cu 2þ from water, and this modified serpentine exhibited a significant increase in adsorption capacity compared to natural serpentine.
Based on the above results, this article first studies the modification of Srp and selects the best modification method. The effects of the Csrp dose, reaction time, pH, and temperature on the removal effect of F À , Fe 2þ , Mn 2þ were also explored. An isothermal adsorption model and adsorption kinetic equation were established to investigate the adsorption behaviour of Csrp for F À , Fe 2þ and Mn 2þ . The adsorption performance of Csrp for F À , Fe 2þ and Mn 2þ in water was compared with other adsorbents. The removal mechanism of Csrp for F À , Fe 2þ and Mn 2þ was revealed by SEM, EDS, XRD, and FTIR spectroscopy, which provides a reference for further studies on the simultaneous removal of F À , Fe 2þ and Mn 2þ from groundwater.
after the reaction. Fourier transform infrared (FTIR, AVATAR 330) spectroscopy was used to analyse changes in the functional groups before and after adsorption by the sample.

Preparation of modified serpentine
Srp was modified by the following methods: Impregnation of a metal salt: Two hundred millilitres of aluminium sulfate solution and calcium chloride solution with mass fractions of 1, 2, 3, 4, 5, 6, and 7% were prepared. Then, 10 g of 120 mesh Srp was added to each solution, which was fully stirred for 3 h, aged for 24 h, filtered to remove the Srp, washed with distilled water and dried for later use.
Acid-base activation: Two hundred millilitres of hydrochloric acid solution and sodium hydroxide solution with mass fractions of 1, 2, 3, 4, 5, 6, and 7% were prepared. Then, 10 g of 120 mesh Srp was added to each solution, fully stirred for 3 h, aged for 24 h, filtered to remove the Srp, washed with distilled water and dried for later use.

Preparation of dealkalized Csrp
Ten grams of Csrp was placed in a 500 mL conical flask, and a certain amount of distilled water was added to dealkali the particles. The distilled water in the conical bottle was replaced at regular intervals until the particles no longer release alkalinity. The dealkalized particles were put into the oven to dry until a constant weight was reached.

Test water sample
The groundwater in Fuxin City, Liaoning Province, China contains high concentrations of fluoride (1.5-5.0 mg/L), iron (3.5-19.5 mg/L) and manganese (0.9-4.8 mg/L), and the pH is between 6.0 and 6.8. Due to the low oxygen content in the groundwater, iron and manganese are not easily oxidized and are often dissolved in the groundwater in the forms of Fe 2þ and Mn 2þ . Thus, they were used as the research objects in this work. Considering the instability and complexity of actual groundwater, experimental water samples were configured to simulate the groundwater quality in the Fuxin area. The mass concentrations of F À , Fe 2þ and Mn 2þ in the water samples were 5 mg/L, 20 mg/L and 5 mg/L, respectively, and the pH was adjusted to 6.5. In addition, to prevent Fe 2þ and Mn 2þ from being oxidized during the adsorption process, a nitrogen cylinder was used to keep the shaker box filled with nitrogen. Xie et al. (2015) also used this method when studying the adsorption of Fe 2þ and S 2by fermented rice husks.

Experiment method
Modified Srp to remove fluoride, iron and manganese experiment Briefly, 150 mL of a composite water sample was added to an Erlenmeyer flask. Then, 300 mg of differently modified Srp was added to the flask, shaken, and then allowed to react for 120 min in a shaker at 180 rpm and 25°C. Finally, the mass concentrations of F À , Fe 2þ , and Mn 2þ were measured.

Experiments on the influence of Csrp removal performance
The effects of the dose, reaction time, pH and reaction temperature on the removal of F À , Fe 2þ and Mn 2þ by Csrp were investigated through batch tests. The adsorption capacity and removal rate of F À , Fe 2þ and Mn 2þ by Csrp can be calculated using Equations (1) and (2): where q e is the adsorption capacity at equilibrium, mg/g; c 0 is the initial concentration of solution, mg/L; c e is the concentration of solution at adsorption equilibrium, mg/L; V is the volume of solution, L; and m is the mass of adsorbent, g.

Experiments on the influence of dealkalized Csrp removal performance
A total of 450 mg of dealkalized Csrp was weighed and added into 150 mL of a composite water sample containing F À , Fe 2þ and Mn 2þ . The reaction was carried out in a constant temperature shaker at 35°C, and the samples were sampled and filtered at a predetermined time to determine the concentrations of F À , Fe 2þ and Mn 2þ in the filtrate.

Adsorption kinetics
Briefly, 450 mg of Csrp was weighed and added to 150 mL of a composite water sample containing F À , Fe 2þ and Mn 2þ . The oscillating reaction was carried out under optimal adsorption conditions (pH of 6, temperature of 35°C). Samples were taken and filtered at predetermined times to determine the concentrations of F À , Fe 2þ and Mn 2þ in the filtrate. Quasi-first-order and quasi-second-order rate equations are often used to describe and analyse the kinetic process of solid-liquid adsorption, and the fitted correlation coefficient R 2 is used for fitting evaluation (Ijagbemi et al. 2010). The equations are as follows (Equations (3) and (4)): Quasi-first-order kinetic equation: Quasi-second-order kinetic equation: where q t (mg/g) is the amount of adsorbent at time t (min) and k 1 and k 2 are the adsorption rate constants, min À1 .

Adsorption isotherm
Briefly, 450 mg of Csrp was weighed and added to 150 mL of a composite water sample containing F À , Fe 2þ and Mn 2þ . The sample was allowed to react at 25°C, 30°C, and 35°C for 150 min before being filtered. The concentrations of F À , Fe 2þ and Mn 2þ in the filtrate were measured. Langmuir and Freundlich isotherm models were used to evaluate the maximum adsorption capacity and adsorption difficulty of Csrp for F À , Fe 2þ and Mn 2þ (Matouq et al. 2015). The equations are as follows (Equations (5) and (6)): Langmuir isotherm adsorption equation: Freundlich isotherm adsorption equation: where q m is the adsorption capacity at adsorption saturation, mg/g; K L is the Langmuir constant, L/mg; and K F and n are Freundlich empirical constants.

Desorption and regeneration
To evaluate the availability of Csrp, Csrp with adsorbed ions was regenerated with 0.1 M NaCO 3 and 0.1 M HNO 3 solutions. Before this process, the adsorbed Csrp was pretreated, the unadsorbed ions on the surface of the Csrp were washed off with deionized water, filtered, and the Csrp was dried. The dried Csrp was added to 150 mL of Na 2 CO 3 solution, and the mixture was shaken at 35°C and 180 rpm for 150 min. Then, the Csrp was washed with deionized water to remove Na 2 CO 3 from the surface. Subsequently, 150 mL of nitric acid solution was added to Csrp, and the mixture was shaken at 35°C and 180 rpm for 150 min. Next, Csrp was washed with deionized water to remove nitric acid from the surface. Finally, the Csrp was dried at 110°C for 2 h. After drying, Csrp was regenerated, and the adsorption test was carried out again for completely desorbed Csrp for a total of five cycles.

RESULTS AND DISCUSSION
Optimization the Srp modification methods Natural Srp was modified by metal salt impregnation, acid-base activation and calcination, and the performance of the modified Srp in removing fluoride, iron and manganese was compared. The results are shown in Figure 1. As shown in Figure 1(a), when the reaction time was less than 120 min, the removal rates of F À , Fe 2þ and Mn 2þ by natural Srp all increased with increasing reaction time. As the reaction time continued to increase, the removal rates of Fe 2þ and Mn 2þ remained stable, while the removal rate of F À showed a downward trend. This is because with increasing reaction time, the increase in OH À released by Srp increases the electrostatic repulsion towards F À , which affects the removal of F À . In summary, 120 min was determined to be the best reaction time for natural Srp to adsorb F À , Fe 2þ , and Mn 2þ . At this time, the removal rates of F À , Fe 2þ , and Mn 2þ were 73.4%, 92.1%, and 89.8%, respectively.
As shown in Figure 1(b), the removal rate of F À by metal salt-impregnated Srp increased with increasing metal salt concentration. This is because aluminium sulfate and calcium chloride adhere to the pores and surface of Srp, and when in contact with F À in solution, aluminium fluoride and calcium fluoride are generated and adsorbed on Srp, thus reducing the concentration of F À in solution (Zhai et al. 2010). The removal rates of Fe 2þ and Mn 2þ decreased with increasing metal salt concentration due to the enrichment of a large amount of Al 3þ and Ca 2þ in the modified Srp. These metal ions combine with the active groups in Srp to form complexes, occupying the effective sites for Fe 2þ and Mn 2þ adsorption on Srp. Additionally, due to a large amount of positively charged ions, the electrostatic repulsion towards Fe 2þ and Mn 2þ increases, which Uncorrected Proof affects the removal of Fe 2þ and Mn 2þ . In summary, fluoride, iron and manganese removal effect on the same concentration of metal salt-impregnated Srp were opposite; thus, the simultaneous removal of all three ions could not be achieved. Figure 1(c) shows that the removal rates of F À , Fe 2þ , and Mn 2þ are improved by modifying Srp with a proper amount of acid and base. This is because hydrochloric acid can dissolve impurities in Srp pores, thus improving its adsorption performance. Sodium hydroxide can chemically react with silica inside Srp to dissolve silicon, reduce the zeta potential, and improve the adsorption performance of Srp (Chen et al. 2021). When the concentration of hydrochloric acid exceeded 4%, the removal rates of F À , Fe 2þ , and Mn 2þ began to decrease, which may be due to the excessive acidity destroying the pores of Srp and resulting in reduced adsorption performance (Gong et al. 2020). When the concentration of sodium hydroxide exceeded 2%, the F À removal rate showed a downward trend. This is because with increasing alkalinity, some surface particles are negatively charged and electrostatically repel F À , which reduces the adsorption of F À on Srp. The removal rates of Fe 2þ and Mn 2þ increased with increasing sodium hydroxide concentration because OH À can form Fe(OH) 2 and Mn(OH) 2 precipitates with Fe 2þ and Mn 2þ , thereby increasing the removal of Fe 2þ and Mn 2þ . The acid-base activated Srp could not achieve the simultaneous removal of fluoride, iron and manganese.
As shown in Figure 1(d), when the calcination temperature did not exceed 500°C, the removal rates of F À , Fe 2þ and Mn 2þ gradually increased with increasing calcination temperature. This is because the surface water and interlayer water of Srp will be lost after calcination at a high temperature, which further enhances the hydrophilicity of the silicon oxygen structure in Srp and increases the contact performance between Srp and various ions. When the temperature exceeded 500°C, the removal rates of F À , Fe 2þ and Mn 2þ began to decrease because the high temperature may change the internal skeletal structure of Srp, resulting in changes in its crystal lattice, morphology and phase and a reduction in the removal of F À , Fe 2þ and Mn 2þ . Cao et al. (2017) used thermally activated Srp to adsorb cadmium and reached a similar conclusion. The adsorption capacity of thermally activated Srp was higher than that of natural Srp, and its adsorption capacity was strongly dependent on the activation temperature. After analysis and comparison, Srp calcined at 500°C showed the best simultaneous removal of F À , Fe 2þ and Mn 2þ .
In conclusion, compared with natural Srp, the removals of F À , Fe 2þ and Mn 2þ by Srp modified with metal salt impregnation, acid-base activation and calcination all improved. Overall, calcination at 500°C showed the best effect among the various modifications, with F À , Fe 2þ and Mn 2þ removal rates of 91.8%, 95.7% and 94.7%, respectively.
Analysis of the factors influencing the removal of F À , Fe 2þ and Mn 2þ on Csrp

Influence of the adsorbent dose
The pH was kept at 6.5, the temperature was 25°C, and the reaction time was 120 min. The effects of the Csrp dose on the removal rates of F À , Fe 2þ and Mn 2þ were analysed, and the results are shown in Figure 2(a). When the dose of Csrp did not exceed 450 mg, the removal rates of F À , Fe 2þ , and Mn 2þ increased with an increasing dose of Csrp. This is because with an increasing Csrp dose, the relative specific surface area and adsorption sites of Csrp increase; thus, the removal rates of F À , Fe 2þ and Mn 2þ gradually increase (Xie et al. 2020). When the dose of Csrp exceeded 450 mg, the removal rate of F À began to decrease, and the removal rates of Fe 2þ and Mn 2þ tended to be stable. This was because the OH À released into the water by Csrp increased when the Csrp dose was increased. It can also be seen from the figure that the pH of the solution after the reaction increased as the Csrp dose increased. A large amount of OH À and F À resulted in electrostatic repulsion and a decrease in the F À removal rate. However, since the concentrations of Fe 2þ and Mn 2þ in the solution were constant, the removal rates of Fe 2þ and Mn 2þ remained basically unchanged as the Csrp dose continued to increase. In conclusion, when the Csrp dose was 450 mg (3 g/L), the removal effect of F À , Fe 2þ and Mn 2þ was the best; the removal rates of F À , Fe 2þ and Mn 2þ were 92.3%, 97.2% and 97.5%, respectively.

Effect of the reaction time
The Csrp dose was kept at 450 mg, the pH was 6.5, and the temperature was 25°C while the Csrp reaction time was changed to analyse its influence on the removal rates of F À , Fe 2þ and Mn 2þ . The results are shown in Figure 2(b). Over the period of 30-150 min, with increasing reaction time, the removal rates of F À , Fe 2þ and Mn 2þ increased significantly. This is because during the initial stage of the reaction, the surface of the adsorbent contains a large number of adsorption sites, which fully combine with each ion in the test water sample and result in a large increase in the removal of each ion. When the reaction time exceeded 150 min, the removal rate of F À began to decrease, while the removal rates of Fe 2þ and Mn 2þ tended to be stable. Considering the adsorption effects of F À , Fe 2þ and Mn 2þ , the optimal reaction time was determined to be 150 min, and the removal rates of F À , Fe 2þ and Mn 2þ were 93.1%, 98.1% and 98.5%, respectively.

Influence of the initial pH
The Csrp dose was maintained at 450 mg, the reaction time was 150 min, and the temperature was 25°C while the initial pH of the water sample was changed to analyse its effect on the removal rates of F À , Fe 2þ and Mn 2þ by Csrp, as shown in Figure 2(c). The removal efficiencies of F À , Fe 2þ and Mn 2þ were poor when the pH was less than 4, probably due to the loss of the Mg 2þ active sites of Csrp at lower pH (Sun et al. 2011). When the pH was 6, the F À removal rate reached its maximum because the Csrp surface was protonated and a positive charge was obtained. This increases the electrostatic attraction to F À and reduces the F À concentration. When the pH continued to increase, the F À removal rate began to decline because with increasing alkalinity, OH À in the solution began to compete with F À for adsorption, thereby decreasing the number of available adsorption sites. The removal rates of Fe 2þ and Mn 2þ increased with increasing pH. Due to the increase in pH, the content of OH À in solution increased, which promoted the removal of Fe 2þ and Mn 2þ . After comprehensive consideration, the optimal reaction pH was determined to be 6, and the removal rates of F À , Fe 2þ and Mn 2þ were 93.8%, 98.3% and 99.2%, respectively.

Effect of the reaction temperature
The Csrp dose was kept at 450 mg, the reaction time was 150 min, and the pH was 6 while the reaction temperature was changed to analyse its influence on the removal rates of F À , Fe 2þ and Mn 2þ . The results are shown in Figure 2(d). When Uncorrected Proof the reaction temperature was between 20°C and 35°C, the removal rates of F À , Fe 2þ and Mn 2þ increased with increasing temperature because the increase in temperature increased the activity of the reactants. From 35 to 45°C, the removal rate of F À began to decrease, and the removal rates of Fe 2þ and Mn 2þ increased. This was caused by the increase in reaction temperature promoting the release of OH À from Csrp. When the temperature exceeded 45°C, the removal rates of F À , Fe 2þ and Mn 2þ all began to decrease. This may be due to the weakness of the adsorptive forces between the active site of the adsorbent and the adsorbate species and between the adjacent molecules of the adsorbed phase (Al-Anber & Al-Anber 2008). After comprehensive analysis, the optimal reaction temperature was determined to be 35°C, and the removal rates of F À , Fe 2þ and Mn 2þ were 94.3%, 99.0% and 98.9%, respectively.
Our research group previously studied a new type of adsorbent, Srp/HAP (Li et al. 2021a(Li et al. , 2021b, whose F À , Fe 2þ , and Mn 2þ removal rates reached 98.6%, 99.9%, and 99.8%, respectively. Although the removal effect of Csrp is slightly lower than that of Srp/HAP, the preparation process of Csrp is simple, and the preparation cost is low. The most important thing is that the quality of the effluent treated by Csrp meets the drinking water standards set by the World Health Organization. Therefore, Csrp is also an excellent adsorbent for removing fluoride and iron and manganese.

Analysis of removal of F À , Fe 2þ and Mn 2þ by dealkalized Csrp
It can be seen from Figure 3 that the removal rates of F À , Fe 2þ and Mn 2þ by dealkalized Csrp were lower than those by Csrp. This result may be due to the release of alkalinity by Csrp during the reaction, which caused some Fe 2þ and Mn 2þ to be removed by precipitation. The removal of Fe 2þ and Mn 2þ by dealkalized Csrp was due to adsorption. The removal rates of Fe 2þ and Mn 2þ reached 95.2% and 96.1% after 150 min, and these rates are only 3.8% and 2.8% lower than that of Csrp. It shows that the removal of Fe 2þ and Mn 2þ by Csrp mainly occurs via adsorption, and precipitation is only a small factor. The slight decrease in the F À removal rate may be due to the decrease in the OH À content in the dealkalized Csrp, which reduces the ion exchange between F À and OH À .

Kinetic analysis
The Csrp dose was kept at 450 mg, the pH was 6, and the reaction temperature was 35°C. Samples were taken at pre-set times to determine the concentrations of F À , Fe 2þ and Mn 2þ in the filtrate, and kinetic equation fitting was performed. The fitting curves are shown in Figure 4, and the fitting parameters are shown in Table 1.
It can be seen from the fitting curves in Figure 3 and the fitting results in Table 1 that the adsorption of F À , Fe 2þ and Mn 2þ by Csrp is more in line with the quasi-second-order kinetic equation. The correlation coefficients (R 2 ) of the quasi-secondorder kinetics for all ions are greater than 0.99, and the theoretical equilibrium adsorption capacities are closer to the  Uncorrected Proof experimental results. The quasi-second-order kinetic equation assumes that the adsorption rate is constrained by chemisorption , indicating that the removal of F À , Fe 2þ and Mn 2þ by Csrp is mainly by chemisorption.

Adsorption isotherm
The effects of different initial concentrations of F À , Fe 2þ and Mn 2þ on the adsorption capacity of Csrp are shown in Figure 5 when the dose of Csrp was 450 mg, the pH was 6, and the reaction time was 150 min. Figure 5 shows that the adsorption capacities of F À , Fe 2þ and Mn 2þ on Csrp increased with increasing reaction temperature and initial ion concentrations. With increasing initial ion concentrations, the active adsorption sites on the surface of the adsorbent gradually tend to be saturated, and thus, the adsorption capacity tends to plateau. Table 2 shows the fitting results of the isothermal adsorption model. The correlation coefficient R 2 fitted by the Langmuir model for each ion is greater than 0.98, and the correlation coefficient R 2 fitted by the Freundlich model is greater than 0.96. The Langmuir isotherm model can better simulate the adsorption of F À , Fe 2þ and Mn 2þ by Csrp, and the adsorption process of each ion conforms to monolayer adsorption. At 35°C, the maximum monolayer saturated adsorption capacity q m values of Csrp for F À , Fe 2þ and Mn 2þ were 6.6774 mg/g, 11.0473 mg/g and 9.9315 mg/g, respectively. Moreover, as seen from Table 3, Csrp has a higher adsorption capacity than most other adsorbents reported; thus, Csrp can be considered a potentially highly efficient adsorbent with the ability to synchronously remove F À , Fe 2þ and Mn 2þ from aqueous solutions.

Analysis of adsorption regeneration
After 5 regeneration cycles, the removal rates of F À , Fe 2þ and Mn 2þ by Csrp are shown in Figure 6. As shown in Figure 6, after five cycles of regeneration, the removal rates of F À , Fe 2þ and Mn 2þ by Csrp decreased, which may be related to the incomplete elution of adsorbents. However, the removal rates of F À , Fe 2þ and Mn 2þ remained above 83%, indicating that Csrp still Adsorbents q e of F À (mg/g) q e of Fe 2þ (mg/g) q e of Mn 2þ (mg/g) References

Evaluating the cost of the treatment
Both the effects of the treatment and the cost of the water treatment process should be considered. It can be seen from Table 4 that compared with membrane treatment and other adsorbent, Csrp water treatment has a lower cost, and its regeneration cost is reasonable.
Microscopic characterization and adsorption mechanism analysis of Csrp X-ray diffraction analysis Figure 7 shows the X-ray diffraction patterns of Srp and Csrp before and after adsorption. Figure 7 shows that Srp is mainly composed of Mg 3 Si 2 (OH) 4 O 5 and that there are 3 characteristic peaks at 2θ values of 12.16°, 24.68°, and 35.56°. The diffraction peaks of Csrp were consistent with natural Srp, indicating that the Srp structure was almost unchanged after calcination at 500°C. After adsorption, new characteristic peaks appeared in Csrp, which were analysed by XRD software as Fe 3 Si 2 (OH) 4 O 5 , Mn 3 Si 2 (OH) 4 O 5 and MgSiF 6 ·6H 2 O. It is speculated that F À can exchange ions with OH À and form organosilicon bonds with unsaturated Si-O-Si bonds. Additionally, the ion-exchange reaction of Fe 2þ and Mn 2þ with Mg 2þ occurred to form these three new substances (Li et al. 2003).

SEM and EDS analysis
As shown in Figure 8(a), the surface of Srp presents a porous curled structure, and flaky structures are superimposed on each other on the surface. There are obvious textures and concave-convex structures among the textures. This special structure increases the specific surface area of Srp and gives Srp a certain physical adsorption capacity. As shown in Figure 8(b), found that the structure of Srp did not change significantly at a lower thermal activation temperature (,550°C). The XRD analysis results of Srp before and after calcination were confirmed. As shown in Figure 8(c), after the treatment of the test water sample, the pores on the surface of Csrp became significantly smaller, and the amount of fine flocculent crystals increased. It is speculated that F À , Fe 2þ and Mn 2þ may physically adsorb on the surface of Csrp or that a surface coordination reaction may have taken place. Using an energy spectrometer to further determine the elements in the material, it can be seen from Figure 8(d) and 8(e) that Srp mainly contained O, Mg, and Si. After calcination, the content of O decreased, while the contents of magnesium and silicon increased, indicating that the dehydroxylation process occurred during the thermal activation of Srp. The EDS after adsorption (Figure 8(f)) showed F, Fe, and Mn, while the weight fraction of O decreased from 53.43% before adsorption to 46.70%, and the weight fraction of Mg decreased from 25.52% before adsorption to 17.53%. This further demonstrated that the ion-exchange reactions between F À and OH À and between Fe 2þ /Mn 2þ and Mg 2þ are in good agreement with the XRD analysis results.

FTIR spectroscopy
To understand the changes in the functional groups of Srp and Csrp before and after adsorption, FTIR spectroscopy was performed, and the results are shown in Figure 9. The infrared absorption peak of Csrp was very similar to that of natural Srp, indicating that the structure of serpentine did not change significantly after calcination and further confirming the XRD and SEM analysis results. The absorption peaks at 630.72 cm À1 and 3,676.32 cm À1 belonged to the Mg-OH stretching vibration. The absorption peaks of Csrp after adsorption were significantly weakened at these two points, indicating that the ionexchange reactions between F À and OH À and between Fe 2þ /Mn 2þ and Mg 2þ occurred. In addition, the vibration peaks at 455.20 cm À1 and 985.62 cm À1 were attributed to Si-O bond stretching vibrations. After the reaction, the vibration peak of Csrp Si-O was weakened. This shows that F À , Fe 2þ , and Mn 2þ may undergo surface coordination with the unsaturated Si-O-Si bond. The absorption peak at 1,452.4 cm À1 was CO 3 2À (Selim et al. 2018), which may be because natural Srp absorbs Uncorrected Proof CO 2 in the atmosphere and reacts to form carbonate groups. During the adsorption test, CO 3 2À combined with H þ in the solution to produce CO 2 gas that escaped; thus, the vibration peak of Srp at 1,452.4 cm À1 disappeared after the reaction.

Analysis of the removal mechanism
The removal of fluoride, iron and manganese by Csrp is mainly chemical adsorption, including ion exchange and surface coordination, and a small amount of metal ions are removed by precipitation. The specific analysis is as follows: (1) Ion exchange: At the beginning of the reaction, there was more H þ in the solution, which made the Csrp surface positively charged, and F À could be adsorbed to the Csrp surface by electrostatic attraction. As the reaction went on, the pH of the solution increased, resulting in more OH À in the solution, and Fe 2þ and Mn 2þ also accumulated on the Csrp surface due to electrostatic attraction. The Csrp then released an equivalent ion to exchange with it. The reaction equations are shown in (7) (8) and (9): OH À Mg À OH þ 2F À ! F À Mg À F þ 2OH À 3Fe 2þ þ Mg 3 Si 2 (OH) 4 O 5 ! Fe 3 Si 2 (OH) 4 O 5 þ 3Mg 2þ (8) 3Mn 2þ þ Mg 3 Si 2 (OH) 4 O 5 ! Mn 3 Si 2 (OH) 4 O 5 þ 3Mg 2þ (9) (2) Surface coordination: The unsaturated Si-O-Si bond has high chemical activity, can displace iron and manganese from the solution, and can adsorb fluoride to make it fixed on the surface. We found in FTIR analysis that the absorption peak of Si-O-Si was weakened, which further confirmed this speculation. The specific process is shown in Equations (10)-(12): ; Si À O À Si ; þ NaF !; Si À O À Na þ ; Si À F (10) ; Si À O À Si ; þ Fe À OH !; Si À O À Fe þ ; Si À OH (11) ; Si À O À Si ; þ Mn À OH !; Si À O À Mn þ ; Si À OH (3) Precipitation: Some iron and manganese may have been removed by precipitation because Csrp releases alkalinity during the reaction, resulting in an increase in the pH of the solution system, which in turn induces heavy metal ions to precipitate. Guo & Yuan (2000) also obtained similar conclusions in the serpentine adsorption test for heavy metal ions.
The specific removal process is shown in Figure 10.