Simultaneous removal of ﬂ uoride, manganese and iron by manganese oxide supported activated alumina: characterization and optimization via response surface methodology

Fluoride, iron and manganese simultaneous exceedance of standard can be observed in groundwater in northeastern China. This work aims to apply a highly ef ﬁ cient method combining adsorption and oxidation for the synchronous removal of the inorganic ions. An innovative adsorbent (manganese-supported activated alumina) was synthesized by the impregnation method and showed a signi ﬁ cant adsorption capacity better than that of fresh activated alumina. The characterization (scanning electron microscope; Brunauer, Emmett and Teller; X-ray diffraction and Fourier transform infrared spectroscopy) results veri ﬁ ed the successful introduction of MnOOH and MnO 2 , and the improvement of surface microstructure enhanced the removal ability. The effect of single factors, such as pH value, reaction time or dosage on the removal performance has been veri ﬁ ed. The maximum removal ef ﬁ ciencies of ﬂ uoride, iron and manganese were optimized via Response surface methodology considering the independent factors in the range of MO@AA dosage (5 – 9 g/L), pH (4 – 6) and contact time (4 – 12 h). Noted that compared with control, MO@AA exhibited 59.4% of improved ﬂ uoride performance. At pH of 5.79, contacting time of 12 h and 8.21 g/L of MO@AA, ﬂ uoride, iron and manganese removal were found to be 91, 100 and 23%, respectively. Herein, MO@AA was distinguished as good applicability for the treatment of ﬂ uoride-, iron- and manganese-containing groundwater. This of a promising adsorbent by impregnation methods. Characterizations of the adsorbent and the effect of various parameters like adsorption time, adsorbent dosage and initial pH were conducted using manganese oxide supported activated alumina (MO@AA). In addition, Response surface methodology and Design-Expert software were used to determine optimal operating parameters. the data obtained. Then, the mechanism of ﬂ uoride, manganese and iron removal can be discussed.

to the limited set of experimental conditions provided by the software. Meanwhile, the optimal process parameters and operating conditions can be found (Zhang et al. 2020;Zhao et al. 2020). Response surface methodology, which has been confirmed with high accuracy, is suitable for multi-factor experiments.  This paper is mainly describes synthesis of a promising adsorbent by impregnation methods. Characterizations of the adsorbent and the effect of various parameters like adsorption time, adsorbent dosage and initial pH were conducted using manganese oxide supported activated alumina (MO@AA). In addition, Response surface methodology and Design-Expert software were used to determine optimal operating parameters.

Preparation of the adsorbent
As shown in Figure 1, the MO@AA were prepared by the impregnation method (Liping 2011). Briefly, 30 g of the particles were impregnated in 150 mL of MnSO 4 solution (0.05 mol/L) with continuous stirring and heating for 6 h at 115°C (magnetic stirrer), then 2.76 mL H 2 O 2 with a concentration of 30% were immersed. At this point, the particles turned light yellow. Subsequently, 45 mL 25% NH 3 ·H 2 O were poured into the mixture. The brown precipitates, generated on the surface of the samples in this step, were confirmed as MnOOH and MnO 2 . Next, the mixture was washed thoroughly in a constant temperature oscillation incubator with hot DI water (30-40°C) for 40 min at the speed of 150 r/min. After several rounds of flushing, the adsorbent was dried for 4 h at 100°C and cooled to room temperature for further use.

Characterizations of the adsorbent
The surface microstructure of MO@AA was explored by scanning electron microscope (SEM) (S-4800, Hitachi, Japan). When the accelerating voltage was 5 kV and the temperature was constant, the photomicrographs were recorded. The measurement of particle surface areas and pore sizes were conducted by N 2 air-suction desorption on an automatic static physical adsorption instrument (Autosorb-IQ2-MP, Quantachrome, America). The total specific surface was calculated based on the multipoint Brunauer, Emmett and Teller (BET) equation (P/P 0 ¼ 0.005-0.3). The total pore volume was measured at P/P 0 ¼ 0.99. X-ray diffraction (XRD) was measured using an X'Pert Pro instrument (Spectris, Holland). Measurement conditions were as follows: tube voltage 40 kV, tube current 40 mA, Cu K-alpha radiation source, γ ¼ 0.15406 nm, scanning range 5-90°, and scanning at the speed of 5°/min. Lastly, transmission spectra were analyzed using Fourier transform infrared spectroscopy (FTIR) (NICOLET iS50, Thermo Nicolet Corporation, America) for infrared absorption spectra. The infrared range was set to 4,000-400 cm À1 , and each scanned 32 times.

Single-factor study
The prepared MO@AA were used for the removal of fluoride, manganese and iron from the aqueous solution. All trials were conducted under the conditions of room temperature (25 + 0.1°C) and acid condition (hydrochloric acid was used to adjust the pH) by batch scale. Here, 100 mL of fluoride, iron and manganese solution with the initial concentrations of 0.26 mmol/ L, 0.04 mmol/L, and 0.02 mmol/L were mixed with a certain mass of adsorbents. Then, the mixture was shaken in a constant temperature oscillation chamber for a predetermined period of reaction time at the speed of 120 r/min. The residual concentrations of fluoride, manganese, and iron were detected and accessed using fluorometric spectrophotometry and spectrophotometer, respectively. The limits of fluoride, manganese, and iron concentrations from the Standard Test Method for Drinking Water (GB/T 57750-2006), were 0.05 mmol/L, 0.01 mmol/L and 0.002 mmol/L.
In the previous study, pH value, adsorbent dosage, contacting time were basic factors affecting the effect of removal efficiency. Then, based on the range of the single-factor test, Response surface methodology was conducted to obtain the corresponding response values. The concentrations of contaminants were detected at certain time intervals, and the adsorption quantity (q t mg/g) was investigated using the following equation: where C t (mg/L), w (g) and V (L) represent the concentration of fluoride ions at time t (min), the weight of the adsorbent and the volume of the solution, respectively. The removal efficiency of contaminants (%) was calculated using the following mathematical expression: where C o (mg/L) and C t (mg/L) were taken as the initial and final solution concentration.

Box-Behnken experimental design
Box-Behnken experiments were carried out to investigate the interaction of three independent process variables (pH, adsorbent dosage, contact time) and optimize the maximum percent removal efficiency. The scheme, a total of 17 runs, consisted of three levels (low, medium and high). Independent factors, including pH, contact time, adsorbent dosage, were written as A, B, and C. Response value (removal rate of fluoride, iron and manganese) were denoted as Y 1 , Y 2 and Y 3 , respectively. Analysis of variance was generated to manifest the influence of individual linear, quadratic and interaction terms. The applicability of the model was checked using the coefficient of determination (R 2 ) and coefficient of variation (C.V. %). The commonly used second-order polynomial equation can be expressed as: where Y is the response value (%), b i is the regression coefficient, and 1 is the error of the model. The main effects and interactions between factors were determined. Through the Response surface methodology model, the parameters (coefficients of correlation, P-value, F-value, residual analysis and predicted values) were validated by the data obtained. Then, the mechanism of fluoride, manganese and iron removal can be discussed.

. SEM analysis
As depicted in Figure 2, the surface microstructures of fresh alumina and MO@AA were observed by SEM. It manifested that the surface appearance changed obviously in the modification process. As shown in Figure 2(a), the pore structure of AA was inconspicuous with only a few pores and obvious bulks were heaped up on the surface. Whereas, Figure 2(b) shows that after impregnation of AA, certain pores became uniformed and extensive impurities removed, which improved the adsorption performance. In addition, the surface of modified AA presented a convex spinous structure. These alterations that occurred may be possibly for the sake of the introduction of functional groups into the surface of AA by manganese oxides, which destroyed the crystal structure to manifest higher energy of adsorption. Thus, it can be inferred that manganese oxides were successfully loaded on to AA.

BET analysis
To examine the pore properties of AA before and after modification, BET analysis was carried out. As displayed in Table 3, the specific surface increased 24% compared with unloaded alumina beads, which is expected to improve adsorption ability. However, total pore volume and average pore diameter were slightly reduced, attributed to a small amount of manganese oxides entering the pore channel. With respect to the adsorption-desorption isotherm plot shown in Figure 3(a) and 3(b), its nature was similar to type II isotherms with a typical H2 hysteresis loop (Thommes 2016). It was indicated that the process was unrestricted monolayer-multilayer and the MO@AA were mesoporous. Additionally, the N 2 adsorption capacity increased with the build-up of relative pressure, revealing that the pore structure was not damaged during the modification process. The curves of MO@AA did not coincide but formed the larger hysteresis loops, which exhibited better mesoporous properties, being consistent with SEM analysis. The hysteresis loop at high pressure resulted from the occurrence of condensation and evaporation at different relative pressures. The above results indicated that loading with manganese oxides had little effect on the pore structure of the samples.

XRD analysis
XRD measurements were conducted to investigate the main constituent elements and chemicals of AA and modified AA, respectively. XRD analysis of AA before and after treatment is exhibited in Figure 4. It can be concluded that the modification process had little effect on the crystal structure of the samples, indicating that the basic framework of AA showed no obvious change. Taking (Yang et al. 2021). This revealed that the major oxide of fresh AA was Al 2 O 3 and manganese oxide was successfully loaded onto the surface of the modified samples after impregnation.

FTIR analysis
It was well known that the chemical groups involved in the adsorption process were directly affected the performance of fluoride removal. Hence, to verify the functional groups, the FTIR spectroscopy was performed. Figure 5 shows the FTIR spectra of fresh alumina and MO@AA. The peaks located at around 3,455 cm À1 , 1,617 cm À1 , 1,384 cm À1 and 586 cm À1 were observed, respectively. The adsorption band in between 3,455 cm À1 and 3,626 cm À1 was due to -OH stretching vibrations and the peak became sharp. The shift was possibly caused by the increase in hydroxyl groups after manganese oxides loaded. When the concentration of hydroxyl group increased, the association effect would be enhanced, and the stretching vibration peak would become sharp. The intensified water molecule bending vibration at 1,617 cm À1 indicated that the adsorption process might be hydrogen bonding with the hydroxyl group. As can be seen from the images no other chemical groups were formed, which was speculated to be electrostatic adsorption. After adsorption, the sharpness of peaks of the hydroxyl group implied that the ion exchange reaction had occurred and the hydroxyl group had been replaced. The bands at 1,384 cm À1 were CO 3 2À symmetric stretching vibration. This phenomenon can be explained in that AA hydrophilicity inevitably absorbed H 2 O and CO 2 in the air. The peak at 592 cm À1 was attributed to Al-O bond vibration in unmodified AA, while higher bands at 586 cm À1 were due to the combination of Al-O and Mn-O bonds. In general, the peak position of activated alumina before and after modification had little variation, and the peak shape was basically the same, which demonstrated that the basic skeleton of AA had no obvious change in the process of modification. In the process of adsorption, not much had changed for the functional groups, but these obtained an increased -OH stretching vibration peak, so the content of hydroxyl groups and the number of active sites increased.

Effect of contact time
The contacting time is related to the degree of reaction. The longer the reaction time, the better the effect of pollutant removal. As shown in Figure 6, the contact time exhibited a major influence on the pollutant removal. It expressed a similar increasing tendency in the removal of fluoride and iron with a further increment of time. The adsorption capacity of fluoride calculated by Equation (1) was dramatically accelerated in the preliminary stage (0-4 h), which was mainly due to external diffusion. Subsequently, from 4 to 12 h, the adsorption growth rate slowed down. This phenomenon can be explained in that MO@AA was gradually covered by fluoride ions, especially on active sites and pores. Then, in the final stage (12-24 h), the amount of fluoride adsorbed tended to be stable. That is to say, the adsorption came to an equilibrium state, the maximum  achieved fluoride removal efficiency was 98% (Equation (2)). In the process of reducing the concentration of iron and manganese ions, the adsorption and contact oxidation processes worked together. The manganese oxide accumulated on the surface of MO@AA, and the contact oxidation capacity was enhanced. After 24 h of reaction, the maximum iron and manganese removal rates of the effluent water reached 83 and 12% (Equation (2)), respectively. When set as a long period of contact time in acid solution (pH 4), leaching of manganese was observed. This can be explained by the reaction of excess hydrogen ions with MnOOH (Bochatay & Persson 2000):

Effect of adsorbent dosage
With the increase in adsorbent dosage, the active sites and the amount of manganese oxide provided by MO@AA was reinforced, and the fluoride and iron removal rate gradually improved, as observed from Figure 7. This may be due to the enhanced attraction between MO@AA and the contaminants. The fluoride, iron and manganese rates of removal increased 50, 63 and 9% (Equation (2)) in the dosage range selected (1-11 g/L), respectively. It was observed that this further increased the amount of dosage, but removal efficient did not markedly change. This may be due to saturation on the surface of MO@AA.

Effect of pH value
In a strong acid medium (pH , 2.00), HF generation affects the removal of fluoride ions. Here, the initial pH was set from 3.00 to 10.00 (Chen et al. 2021). Apparently, as demonstrated in Figure 8, pH had a significant monitoring force driving the removal process. The increase in pH led to an increase in the removal of fluoride, which could be attributed to the fact that positively charged MO@AA in acidic conditions was combined with negatively charged fluoride ions. When pH increased, the existence of excess OH À ions may compete with negatively charged fluoride ions that weakened the interaction of electrostatic forces (Roy et al. 2018). In general, the removal capacity for fluoride was high at low pH, whereas high pH favors the oxidation of iron and manganese ions. Therefore, pH 4-6 was chosen as the optimum range for pH value for the simultaneous removal of fluoride, manganese and iron.

Analysis of response surface methodology and the model fitting
The ANOVA results for fluoride, iron and manganese are shown in Tables S1-S4 in Supplementary Materials. The values for F and P implied the significance of the fitted equations and the extent to which the original hypothesis is not rejected. The smaller the P-value and the larger the F-value, the more significant is the effect of this item on the response value (Thommes 2016). By analyzing the results obtained, all three Response surface methodology models showed good predictability.
In the regression equation for fluoride (Table S1), the F-value was 9.59, the P-value was 0.0035 , 0.005. Values of P less than 0.005 indicated model terms were significant. In this case, A, B, C, AB, and B 2 were significant factors. Adeq precision measures the signal-to-noise ratio. A value greater than 4 is desirable. Thus, 11.357 indicates an adequate signal. This model can be used to navigate the design space. The values of the correlation coefficient (R 2 ) and adjusted R 2 were 0.9250 and 0.8296, respectively. A high R 2 value (more than 0.8) is expected. There was only 2% of the total variation that was not explained by the model. All demonstrated the good fitness for the model. Results of ANOVA for iron are shown in Table S2. The model F-value of 71.53 demonstrated that the model was accurate. There was only a 0.01% chance that a Model F value this large could occur due to noise. P-values less than 0.05 indicate model terms were significant. In this case, B, C, BC, A 2 and B 2 are significant factors. The sequence of the independent factors was B 2 . C . A 2 . BC . AB . A . AC . C 2 . The R 2 of 0.8680 (.0.8) was in reasonable agreement with the adjusted R 2 of 0.9754 (.0.8). Furthermore, the Adeq precision in this model was 26.409 . 4. It showed that this model fits well with the actual situation, has good stability, high test reliability, and accuracy.
The ANOVA result of manganese model is given in Table S3. The high F-value (347.5) implied the reliability of the model. The existed chance of P , 0.0001 indicated that the model was highly significant. In this manner, A, B, C, AB, A 2 , B 2, and C 2 were significant variables. The high Adeq precision value of 35.942, high R 2 of 0.9924 and adjusted R 2 of 0.9826 were found. It reflected that only 1% of the total variations cannot be explained. Here, this model can be used.
The correlation between actual and predicted values of removal efficiency is displayed in Figure 9(a)-9(c). The points distribution was in the vicinity of a straight line, suggesting that the developed model was adequate in predicting the response variables for the experiment, the reliability of the data was proven. The simulation results showed that the model could be evaluated with a 95% confidence level. To further investigate the effect of pH, and the time of reaction, a multivariate coupling experiment was carried out. The 3D images of Response surface methodology plot shown in Figure 10(a)-10(c) verified the connection of the two parameters on the simultaneous removals of fluoride, iron and manganese. With the increase in pH from 4 to 6, fluoride removal capacity slowed down and the removal efficiency of manganese and iron was enhanced. While, with increase in contacting time, removal rate correspondingly improved. In addition, the slope of the plots was large, and the large plot indicated a huge impact of independent factors on the response values. Figure 11(a)-11(c) shows 3D surface of the impact of pH and dosage on response value. When the pH value rose continually, it an increase was observed in the removal of fluoride along with a decrease in iron and manganese removal efficiency. As the time of reaction was extended, the solid-liquid two-phase system gradually reached an equilibrium, leading to an increase in the removal effect. Maximum efficiency of 94, 100 and 24% was observed in this plot, for 8 g/L of adsorbent.

Effect of variation in contact time and dosage
In Figure 12 Water Science & Technology Vol 84 No 12,3811 (with a higher slope) verified a greater coefficient impact than adsorbent dosage on response value. These results and observations were also confirmed by the ANOVA results presented in Tables S1-S3.

Optimization and validation of the model
To meet the drinking water standards, an optimal condition can be selected based on the Response surface methodology model. In this work, the raw fluoride, iron and manganese ion mass concentrations of 0.26 mmol/L, 0.04 mmol/L, and 0.02 mmol/L were mixed with MO@AA at an initial pH of 5.79, the reaction time of 12 h, adsorbent dose of 8.21 g/L (Table 6). Optimum process condition was 0.955. Under the optimized conditions, the maximum removal rate was achieved. The measured value had an error of only 1, 0 and 1%, with the simulation result. Consistent with the above analysis, the Response surface methodology model has good accuracy and desirability.

Mechanism of fluoride, manganese and iron removal
Compared with the reported adsorbents in Table 2 which showed good performance in the removal of fluoride, iron and manganese, respectively, MO@AA can take effect in the simultaneous removal of these three inorganic solutes. In addition, MO@AA worked effectively over a wider range of pH (3-9). The comparison of fresh alumina and MO@AA is demonstrated in Figure 13. It manifested that the adsorption capacity of fluoride rose sharply on MO@AA as compared to the control. The possible reasons analyzed on the basis of characterization (SEM and BET) were given as follows: the number of active sites enhanced, the specific surface increased, convex spinous structure presented and the pore structure improved. The mechanism of adsorption on MO@AA was based on the experimental results and the analysis of XRD and FTIR. It can be deduced that the adsorption process may be the reason for electrostatic attraction and ion exchange. For one thing, at acidic conditions, the surface hydroxyl group was protonated and led to the surface of MO@AA being electropositive (Kumari et al. 2020). Thus, the negative fluoride ions charged primely onto the positively surface by the electrostatic force as follows: In addition, hydroxyl released when in acidic medium, resulting in the increase in pH. Furthermore, MO@AA absorbed water molecules led to surface hydroxylation (Teng et al. 2009). Active sides enhanced in this process and MnOOH was reduced to Mn(OH) 2 . Simultaneously, fluoride ions exchanged with hydroxyl groups due to the similar hydrated ionic radius. The chemical reaction equations could be expressed as (Yang 2013): R-(Al 2 O 3 ) n Á Mn(OH) 2 þ 2F À ¼ R-(Al 2 O 3 ) n MnF 2 þ 2OH À (10)  The removal of ferrous ions may be due to the oxidation of MnOOH by dissolved oxygen in the aquatic environment, which generated a high-valent manganese compound. Subsequently, MnO·Mn 2 O 7 was further oxidized with divalent iron in solution to form Fe(OH) 3 (Weili 2020 For manganese removal, two steps were involved: (i) the dissolved Mn 2þ adsorbed by MnO 2 ; (ii) the adsorbed Mn 2þ oxidized to high valence manganese compounds. At the beginning of the reaction, influenced by low pH, MO@AA showed poor performance in oxidation. A small amount of Mn 2þ was adsorbed on the surface of the particles (Equation (15)). As the hydroxyl group undergoes ion exchange with the fluoride ion, the hydroxide is released into the aqueous solution, while the high pH facilitates the oxidation of the manganese ion. As the contact time increases, its contact oxidation capacity becomes stronger and stronger, and then contact oxidation for manganese removal is the main focus (Equation (16)-(18)). The equations are shown below (Liping 2011):

CONCLUSIONS
In this work, a novel adsorbent was successfully fabricated on containments reduction. The percent fluoride removal varied from 31% to 91% after the impregnation method. Results obtained from SEM proved better adsorption performance with the convex spinous structure and more active sides on the surface of MO@AA. XRD also manifested the introduction of MnOOH and MnO 2 after impregnation. The specific surface increased 24% compared with unloaded alumina beads on the basis of BET, with respect to the adsorption-desorption isotherm, it verified that the process was multilayered and the modified AA was mesoporous. The FTIR spectroscopy analysis implied that the ion exchange reaction had occurred. By modelling, the effects of the three parameters including pH, contacting time and the amount of dosage were evaluated. Under optimized conditions, maximum removal efficiency reached 90.59%, 100% and 23.46%. Furthermore, it can be deduced that the mechanism of adsorption consisted of electrostatic attraction and ion exchange. The oxidation process played a major role in treating iron and manganese-containing simulated water. Thus, MO@AA showed the great potential of fluoride, iron and manganese removal and was a promising adsorbent for groundwater treatment. In the follow-up trials, MO@AA would be applied in real groundwater to remove the excess fluoride, iron and manganese ions. More cost-effective methods can be explored in the next phase.