Silicic Acid Removal by Metal-Organic Frameworks for Silica-Scale Mitigation in Reverse Osmosis

Reverse osmosis (RO) membranes are susceptible to silica scaling, resulting in irreversible degradation of membrane performance. This work covered the fabrication of MIL-101(Fe) for silicic acid adsorption to alleviate the silica scaling of RO membranes. The effect of pH, mixing time and initial concentration on silicic acid adsorption of MIL-101(Fe) was appraised in detail. The adsorption experiments demonstrated that MIL-101(Fe) possessed an excellent adsorption ability for silicic acid with the maximum adsorption capacity reaching 220.1 mgSiO2·g−1. Data fitting confirmed the pseudo-second-order equation and Freundlich equation were consistent with silicic acid adsorption on MIL-101(Fe). Finally, a simulated anti-scaling experiment was carried out using a feed solution pretreated by MIL-101(Fe) adsorption, and the permeance exhibited a much lower decline after 24 h filtration, confirming that MIL-101(Fe) exhibits an excellent application potential for silica-scale mitigation in RO systems.


Introduction
Water resource shortages have become an international problem worthy of concern in industrial production and daily life [1,2], seriously affecting household living quality and global sustainable development [3,4]. To solve this problem, RO systems have been widely used due to their ease of operation, efficient water treatment and low environmental pollution [5][6][7]. However, inorganic scaling is a commonly encountered bottleneck for high water recovery of RO systems, leading to a reduction in flux and extra cost of energy consumption and operation.
Inorganic pollution generally results from the deposition of salts on the surface or in the pores of membranes when the ion concentration exceeds its solubility product. Salts with low solubility, including calcium carbonate (CaCO 3 ) [8,9], calcium sulfate (CaSO 4 ) [10,11] and sodium metasilica (Na 2 SiO 3 ) [12,13], are the most common inorganic scales on the membrane surface. Among them, silica scale is worthy of more attention because of the high concentration of SiO 2 in water (20-60 mg·L −1 ) [14]. The formation mechanism of silica scale has been researched, and it was confirmed that the interaction between silicic acid and the membrane surface had a decisive effect on the formation rate of scale [15,16]. Kempter et al. surveyed the formation process of silica scales with atomic force microscopy (AFM). After the membrane was immersed in a Na 2 SiO 3 solution (250 mg/L) for 1 h, SiO 2 particulates with a diameter of 20-30 nm began to appear, and the membrane surface was completely covered with SiO 2 after 9 h of immersion [17].
During the actual operation of an RO device, cleaning-in-place is generally applied to remove fouling [18]. Acidic cleaning can dissolve inorganic scales, and alkaline cleaning can remove organic fouling and biofouling [19]. However, silica scale generated by the

MIL-101(Fe) Fabrication and Characterization
MIL-101(Fe) was fabricated through a simple solvothermal method using FeCl 3 ·6H 2 O and H 2 BDC as the monomers [43,46]. In the first step, FeCl 3 ·6H 2 O (4 mmol, 1.08 g) and H 2 BDC (4 mmol, 0.665 g) were added into DMF (30 mL). After stirring at room temperature for 30 min, the mixture was transferred to an autoclave, heated at 110 • C for 20 h and then cooled naturally. Finally, the reddish-brown precipitate was washed using DMF and ethanol three times each and then dried at 80 • C under vacuum overnight.

Evaluation of Adsorption Performance
An adsorption experiment was performed to investigate the silicic acid adsorption ability of MIL-101(Fe). First, 0.5 g·L −1 MIL-101(Fe) was mixed with 120 mg·L −1 silicic acid solution, and the solution was sampled periodically until adsorption equilibrium. Then, the silicic acid concentration could be obtained by querying the standard curve. The adsorption rate (R) was calculated with Equation (1).
where C 0 (mg·L −1 ) and C e (mg·L −1 ) correspond to silicic acid concentrations in the initial and equilibrium solutions, respectively. The instantaneous adsorption capacity (q) was calculated using Equation (2).
where C (mg·L −1 ) is the silicic acid concentration, V (L) is the volume of silicic acid solution and W (g) is the quality of MIL-101(Fe). Silico-molybdic acid spectrophotometry was used to determine the concentration of silicic acid. Specifically, 2 mL of ammonium molybdate solution (0.1 g·mL −1 ) and 1 mL of 18 wt% HCl solution were added to 20 mL of diluted sample. The solution was mixed evenly and then allowed to stand for 5 min. Afterwards, 1.5 mL of oxalic acid solution (1 mol/L) was added and mixed again. One minute later, 2 mL C 10 H 9 NO 4 S was added and set aside for 10 min. Silicic acid was quantified using UV-Vis spectroscopy at 420 nm. The concentration could be obtained by querying the standard curve of silicic acid.

Membrane Silica-Scaling Experiment
The silica-scaling experiment was carried out utilizing a cross-flow filtration RO device. The feed solution for the experiment was prepared by mixing 2000 mg/L NaCl, 770 mg/L CaCl 2 , 500 mg/L MgCl 2 and 340 mg/L Na 2 SiO 3 (the concentration of silica was 167 mg/L) with a pH value of 6.5 ± 0.1. The feed solution with or without the pretreatment of MIL-101(Fe) was used to investigate the permeance decline of the RO membrane during 24 h of continuous filtration. The filtration experiment was conducted at 25 ± 1 • C and the initial flux was adjusted to~62 L·m −2 ·h −1 by regulating the pressure.

Structural Morphology of MIL-101(Fe)
As seen from the SEM image presented in Figure 1a, the synthesized MIL-101(Fe) presented a smooth octahedral structure, which was consistent with the structure reported previously [47,48]. The XRD curve of MIL-101(Fe) shown in Figure 1b exhibited intense peaks at 2θ values of 9.3 • , 12.48 • , 18.76 • and 21.9 • , corresponding to the previous reports [49,50], confirming the well-developed MIL-101(Fe) crystals. The PSD was characterized using the Brunauer-Emmett-Teller (BET) method and analyzed through density functional theory (DFT). It can be seen from Figure 1c that three narrow peaks are observed in the PSD profile (19.1 Å, 20.3 Å and 25.1 Å) [51], indicating the existence of micropores and mesopores in MIL-101(Fe). evenly and then allowed to stand for 5 min. Afterwards, 1.5 mL of oxalic acid solution (1 mol/L) was added and mixed again. One minute later, 2 mL C10H9NO4S was added and set aside for 10 min. Silicic acid was quantified using UV-Vis spectroscopy at 420 nm. The concentration could be obtained by querying the standard curve of silicic acid.

Membrane Silica-Scaling Experiment
The silica-scaling experiment was carried out utilizing a cross-flow filtration RO device. The feed solution for the experiment was prepared by mixing 2000 mg/L NaCl, 770 mg/L CaCl2, 500 mg/L MgCl2 and 340 mg/L Na2SiO3 (the concentration of silica was 167 mg/L) with a pH value of 6.5 ± 0.1. The feed solution with or without the pretreatment of MIL-101(Fe) was used to investigate the permeance decline of the RO membrane during 24 h of continuous filtration. The filtration experiment was conducted at 25 ± 1 °C and the initial flux was adjusted to ~62 L·m −2 ·h −1 by regulating the pressure.

Structural Morphology of MIL-101(Fe)
As seen from the SEM image presented in Figure 1a, the synthesized MIL-101(Fe) presented a smooth octahedral structure, which was consistent with the structure reported previously [47,48]. The XRD curve of MIL-101(Fe) shown in Figure 1b exhibited intense peaks at 2θ values of 9.3°, 12.48°, 18.76° and 21.9°, corresponding to the previous reports [49,50], confirming the well-developed MIL-101(Fe) crystals. The PSD was characterized using the Brunauer-Emmett-Teller (BET) method and analyzed through density functional theory (DFT). It can be seen from Figure 1c

Chemical Composition of MIL-101(Fe)
The chemical compositions of MIL-101(Fe) and sodium metasilica were determined using ATR-FTIR analysis. It can be seen from Figure 2 that the characteristic peaks of MIL-101(Fe) included carboxyl groups (1658 cm −1 , 1593 cm −1 and 1388 cm −1 ), benzene groups (1498 cm −1 ), and chelate bonds formed by Fe 3+ with carboxyl groups (553 cm −1 and 748 cm −1 ) [47,52], proving the successful preparation of MIL-101(Fe). Moreover, the characteristic peaks of sodium metasilica included Si-O-Si stretching vibrations (1022 cm −1 , 707 cm −1 and 587 cm −1 ) [53,54], indicating that silicic acid could be adsorbed on the surface of MIL-101(Fe). XPS spectra were further performed to analyze the surface elements and chemical bonds of MIL-101(Fe). As seen from Figure 3, the C 1s peak could be divided into three peaks at 284.8 eV, 286.5 eV and 288.4 eV, which were attributed to C-C, C-O and C=O bonds, respectively. In addition, the peaks at 530.2 eV, 725.5 eV and 711.7 eV could be assigned to Fe-O bonds, Fe 2p1/2 and Fe 2p3/2, respectively. The Fe 2p1/2 and Fe 2p3/2 peaks with a satellite signal at 717.1 eV indicated the presence of Fe 3+ within MIL-101(Fe) [55]. Peaks at 709.0 eV and 722.6 eV were not observed, suggesting that Fe 2+ was not incorporated within MIL-101(Fe).

Silicic Acid Adsorption of MIL-101(Fe)
3.3.1. Effect of Initial pH on Silicic Acid Adsorption Figure 4a shows that the adsorption ability of MIL-101(Fe) to silicic acid was closely related to the initial pH of the sodium metasilica solution. Sodium metasilica mainly existed in the form of silicic acid molecules when the pH was below 9.83, resulting in a low adsorption capacity. When the pH reached 11.42, silicic acid existed in the form of ions, XPS spectra were further performed to analyze the surface elements and chemical bonds of MIL-101(Fe). As seen from Figure 3, the C 1s peak could be divided into three peaks at 284.8 eV, 286.5 eV and 288.4 eV, which were attributed to C-C, C-O and C=O bonds, respectively. In addition, the peaks at 530.2 eV, 725.5 eV and 711.7 eV could be assigned to Fe-O bonds, Fe 2p 1/2 and Fe 2p 3/2 , respectively. The Fe 2p 1/2 and Fe 2p 3/2 peaks with a satellite signal at 717.1 eV indicated the presence of Fe 3+ within MIL-101(Fe) [55]. Peaks at 709.0 eV and 722.6 eV were not observed, suggesting that Fe 2+ was not incorporated within MIL-101(Fe). XPS spectra were further performed to analyze the surface elements and chemical bonds of MIL-101(Fe). As seen from Figure 3, the C 1s peak could be divided into three peaks at 284.8 eV, 286.5 eV and 288.4 eV, which were attributed to C-C, C-O and C=O bonds, respectively. In addition, the peaks at 530.2 eV, 725.5 eV and 711.7 eV could be assigned to Fe-O bonds, Fe 2p1/2 and Fe 2p3/2, respectively. The Fe 2p1/2 and Fe 2p3/2 peaks with a satellite signal at 717.1 eV indicated the presence of Fe 3+ within MIL-101(Fe) [55]. Peaks at 709.0 eV and 722.6 eV were not observed, suggesting that Fe 2+ was not incorporated within MIL-101(Fe).

Silicic Acid Adsorption of MIL-101(Fe)
3.3.1. Effect of Initial pH on Silicic Acid Adsorption Figure 4a shows that the adsorption ability of MIL-101(Fe) to silicic acid was closely related to the initial pH of the sodium metasilica solution. Sodium metasilica mainly existed in the form of silicic acid molecules when the pH was below 9.83, resulting in a low adsorption capacity. When the pH reached 11.42, silicic acid existed in the form of ions,   Figure 4b shows the changing trend of pH after adding MIL-101(Fe) to silicic acid solution under different pH conditions. Since both silicic acid and MIL-101(Fe) were negatively charged under alkaline conditions, the electrostatic attraction between them cannot explain the adsorption process well. Thus, we held the opinion that the high adsorption capacity of MIL-101(Fe) to silicic acid when the pH reached 9.83 might be due to the formation of Fe-OH groups, and then these groups reacted with -OH groups of silicic acid to form Fe-O-Si bonds. As the adsorption progressed continuously, the OH-in the solution was largely consumed, resulting in a significant decrease in the pH value after the adsorption.

Effect of Mixing Time and Kinetics on Silicic Acid Adsorption
The effect of mixing time shown in Figure 5a exhibited the fast adsorption of silicic acid on MIL-101(Fe). At the beginning of the adsorption process, the adsorption capacity increased rapidly, and then reached an invariable value. In the first 6 min, the adsorption capacity reached 125.9 mgSiO2·g −1 , and then slowed down. The sample reached adsorption saturation at about 2 h, and the maximum adsorption capacity was 220.1 mgSiO2·g −1 .
To quantify the adsorption rate, the pseudo-first-order and pseudo-second-order equations were used to describe the process of adsorption, as show in Figure 5b,c. Model parameters such as the adsorption rate constant and equilibrium capacity could be calculated according to the intercept and slope of the fitted curve, as shown in Table 1. The pseudo-first-order and pseudo-second-order equations are normally expressed as Equations (3) and (4): where qe (mg·g −1 ) and qt (mg·g −1 ) are the mass of silicic acid adsorbed at equilibrium and at time t (min), respectively; k1 and k2 are the adsorption rate constant; and t (min) is the mixing time.  Figure 4b shows the changing trend of pH after adding MIL-101(Fe) to silicic acid solution under different pH conditions. Since both silicic acid and MIL-101(Fe) were negatively charged under alkaline conditions, the electrostatic attraction between them cannot explain the adsorption process well. Thus, we held the opinion that the high adsorption capacity of MIL-101(Fe) to silicic acid when the pH reached 9.83 might be due to the formation of Fe-OH groups, and then these groups reacted with -OH groups of silicic acid to form Fe-O-Si bonds. As the adsorption progressed continuously, the OH-in the solution was largely consumed, resulting in a significant decrease in the pH value after the adsorption.

Effect of Mixing Time and Kinetics on Silicic Acid Adsorption
The effect of mixing time shown in Figure 5a exhibited the fast adsorption of silicic acid on MIL-101(Fe). At the beginning of the adsorption process, the adsorption capacity increased rapidly, and then reached an invariable value. In the first 6 min, the adsorption capacity reached 125.9 mgSiO 2 ·g −1 , and then slowed down. The sample reached adsorption saturation at about 2 h, and the maximum adsorption capacity was 220.1 mgSiO 2 ·g −1 .   Table 1 shows that the coefficient of determination (R 2 ) of both equations was greater than 0.99, indicating a good fit of the data. The adsorption process of MIL-101(Fe) to silicic acid was consistent with the pseudo-second-order kinetic equation due to its higher R 2 , indicating that chemisorption played a leading role during the adsorption process. To quantify the adsorption rate, the pseudo-first-order and pseudo-second-order equations were used to describe the process of adsorption, as show in Figure 5b,c. Model parameters such as the adsorption rate constant and equilibrium capacity could be calculated according to the intercept and slope of the fitted curve, as shown in Table 1. The pseudo-first-order and pseudo-second-order equations are normally expressed as Equations (3) and (4): where q e (mg·g −1 ) and q t (mg·g −1 ) are the mass of silicic acid adsorbed at equilibrium and at time t (min), respectively; k 1 and k 2 are the adsorption rate constant; and t (min) is the mixing time.  Table 1 shows that the coefficient of determination (R 2 ) of both equations was greater than 0.99, indicating a good fit of the data. The adsorption process of MIL-101(Fe) to silicic acid was consistent with the pseudo-second-order kinetic equation due to its higher R 2 , indicating that chemisorption played a leading role during the adsorption process.    Table 1 shows that the coefficient of determination (R 2 ) of both equations was greater than 0.99, indicating a good fit of the data. The adsorption process of MIL-101(Fe) to silicic acid was consistent with the pseudo-second-order kinetic equation due to its higher R 2 , indicating that chemisorption played a leading role during the adsorption process. Figure 6a exhibits the effect of the initial concentration of silicic acid on the adsorption performance. With the initial concentration increasing gradually, MIL-101(Fe) showed a rapid adsorption of silicic acid at the beginning, and then the adsorption rate decreased due to the limited active sites. Langmuir and Freundlich isotherm models [56] are normally expressed as Equations (5) and (6), respectively: Langmuir and Freundlich isotherm models [56] are normally expressed as Equations (5) and (6), respectively:

Effect of Initial Concentration and Adsorption Isotherm Study
where q m (mg·g −1 ) is the maximal adsorption capacity, C e (mg·L −1 ) is the equilibrium concentration and b (L·mg −1 ) is the equilibrium constant in the Langmuir model. K f (mg·g −1 ) and n are the empirical coefficients in the Freundlich model. In general, the Langmuir model assumes that only monomolecular layer adsorption occurs on the solid surface, which means that the adsorbate only interacts with the surface of the adsorbent, whereas the Freundlich model is applicable for multilayer adsorption, in which the adsorbates can interact with each other. It can be seen from Figure 6b,c and Table 2 that the R 2 of both isotherm models are above 0.98, indicating that both monomolecular layer adsorption and multimolecular layer adsorption occurred during the adsorption of silicic acid on the surface of MIL-101(Fe). Thus, the adsorption process of silicic acid on MIL-101(Fe) involves two steps, as can be seen in Figure 7. First, silicic acid would bind directly to the unsaturated ion sites, which was the main mechanism of the rapid adsorption. As the adsorption proceeded, all the ion sites were occupied, resulting in the free silicic acid interacting with the adsorbed silicon to form Si-O-Si bonds. where qm (mg·g −1 ) is the maximal adsorption capacity, Ce (mg·L −1 ) is the equilibrium concentration and b (L·mg −1 ) is the equilibrium constant in the Langmuir model. Kf (mg·g −1 ) and n are the empirical coefficients in the Freundlich model. In general, the Langmuir model assumes that only monomolecular layer adsorption occurs on the solid surface, which means that the adsorbate only interacts with the surface of the adsorbent, whereas the Freundlich model is applicable for multilayer adsorption, in which the adsorbates can interact with each other. It can be seen from Figure 6b,c and Table 2 that the R 2 of both isotherm models are above 0.98, indicating that both monomolecular layer adsorption and multimolecular layer adsorption occurred during the adsorption of silicic acid on the surface of MIL-101(Fe). Thus, the adsorption process of silicic acid on MIL-101(Fe) involves two steps, as can be seen in Figure 7. First, silicic acid would bind directly to the unsaturated ion sites, which was the main mechanism of the rapid adsorption. As the adsorption proceeded, all the ion sites were occupied, resulting in the free silicic acid interacting with the adsorbed silicon to form Si-O-Si bonds.   Figure 8a shows the effect of temperature on the adsorption of silicic acid. Obviously, the adsorption capacity to silicic acid increased with increasing temperature, indicating the benefits of heating on silicic acid adsorption.  Figure 8a shows the effect of temperature on the adsorption of silicic acid. Obviously, the adsorption capacity to silicic acid increased with increasing temperature, indicating the benefits of heating on silicic acid adsorption.

Effect of Temperature and Adsorption Thermodynamics Study
where qm (mg·g −1 ) is the maximal adsorption capacity, Ce (mg·L −1 ) is the equilibrium concentration and b (L·mg −1 ) is the equilibrium constant in the Langmuir model. Kf (mg·g −1 ) and n are the empirical coefficients in the Freundlich model.
In general, the Langmuir model assumes that only monomolecular layer adsorption occurs on the solid surface, which means that the adsorbate only interacts with the surface of the adsorbent, whereas the Freundlich model is applicable for multilayer adsorption, in which the adsorbates can interact with each other. It can be seen from Figure 6b,c and Table 2 that the R 2 of both isotherm models are above 0.98, indicating that both monomolecular layer adsorption and multimolecular layer adsorption occurred during the adsorption of silicic acid on the surface of MIL-101(Fe). Thus, the adsorption process of silicic acid on MIL-101(Fe) involves two steps, as can be seen in Figure 7. First, silicic acid would bind directly to the unsaturated ion sites, which was the main mechanism of the rapid adsorption. As the adsorption proceeded, all the ion sites were occupied, resulting in the free silicic acid interacting with the adsorbed silicon to form Si-O-Si bonds.   Figure 8a shows the effect of temperature on the adsorption of silicic acid. Obviously, the adsorption capacity to silicic acid increased with increasing temperature, indicating the benefits of heating on silicic acid adsorption. The Van't Hoff equation was applied to calculate thermodynamic parameters, as shown in Equations (7)-(9) [57]: where k is the adsorption coeffcient; T (K) is the thermodynamic temperature; ∆H (kJ·mol −1 ) is the adsorption enthalpy; ∆S (J·mol −1 ·K −1 ) is the adsorption entropy; R (J·mol −1 ·K −1 ) is gas constant; and ∆G (kJ·mol −1 ) is the adsorption Gibbs free energy.
The fitting diagram and thermodynamic parameters can be seen in Figure 8b and Table 3. ∆H > 0 demonstrates that the adsorption process is endothermic, which means that the temperature rise is beneficial to accelerate the adsorption. ∆G < 0 indicates that the adsorption of silicic acid on MIL-101(Fe) can occur spontaneously. ∆S > 0 signifies that this adsorption is a process of entropy increase. Table 3. Thermodynamic parameters for silicic acid adsorption on MIL-101(Fe).  Figure 9 shows the SEM images and elemental mapping of MIL-101(Fe) before and after the adsorption of silicic acid. The Si signal on the surface after the adsorption was significantly stronger than that before the adsorption, mainly due to the coverage of ion sites with silicic acid. The elemental composition of MIL-101(Fe) was also investigated using elemental mapping. Specifically, the weight concentration of oxygen on the surface increased from 22.70% to 25.11% while that of silicon increased from 0.62% to 3.34%. It can be proven using elemental composition analysis that silicic acid can not only bind to the unsaturated ion sites, but also interact with the adsorbed silicon, which was important evidence of the adsorption mechanism of silicic acid on MIL-101(Fe).
The fitting diagram and thermodynamic parameters can be seen in Figure 8b and Table 3. ΔH > 0 demonstrates that the adsorption process is endothermic, which means that the temperature rise is beneficial to accelerate the adsorption. ΔG < 0 indicates that the adsorption of silicic acid on MIL-101(Fe) can occur spontaneously. ΔS > 0 signifies that this adsorption is a process of entropy increase. 3.3.5. Adsorption Mechanism Study Figure 9 shows the SEM images and elemental mapping of MIL-101(Fe) before and after the adsorption of silicic acid. The Si signal on the surface after the adsorption was significantly stronger than that before the adsorption, mainly due to the coverage of ion sites with silicic acid. The elemental composition of MIL-101(Fe) was also investigated using elemental mapping. Specifically, the weight concentration of oxygen on the surface increased from 22.70% to 25.11% while that of silicon increased from 0.62% to 3.34%. It can be proven using elemental composition analysis that silicic acid can not only bind to the unsaturated ion sites, but also interact with the adsorbed silicon, which was important evidence of the adsorption mechanism of silicic acid on MIL-101(Fe).
Recycle and regeneration experiments of MIL-101(Fe) were conducted immediately after the 2 h adsorption test. After washing MIL-101(Fe) with 0.05 M NaOH used as a regeneration reagent to release the silicic acid three times, ~75.6% of the adsorbent could be recycled and it showed ~60% recovery of the adsorption capacity.

Silica-Scale Mitigation on RO Membranes
To investigate the potential application of MIL-101(Fe) for silica-scaling mitigation in RO systems, a control experiment was designed. With other conditions being the same, the feed solution was pretreated with MIL-101(Fe) to adsorb silicic acid. As shown in Figure 10a, when the feed solution was not pretreated, the RO membrane showed a flux decline of around 23% after 24 h continuous filtration. Comparatively, when the Recycle and regeneration experiments of MIL-101(Fe) were conducted immediately after the 2 h adsorption test. After washing MIL-101(Fe) with 0.05 M NaOH used as a regeneration reagent to release the silicic acid three times,~75.6% of the adsorbent could be recycled and it showed~60% recovery of the adsorption capacity.

Silica-Scale Mitigation on RO Membranes
To investigate the potential application of MIL-101(Fe) for silica-scaling mitigation in RO systems, a control experiment was designed. With other conditions being the same, the feed solution was pretreated with MIL-101(Fe) to adsorb silicic acid. As shown in Figure 10a, when the feed solution was not pretreated, the RO membrane showed a flux decline of around 23% after 24 h continuous filtration. Comparatively, when the feed solution was pretreated by the adsorption of MIL-101(Fe), the membrane flux only decreased by 13% after 24 h continuous filtration. Therefore, it can be concluded that if the feed solution is pretreated, less silica scale will form during the RO process, mainly because a large amount of silicic acid in the feed solution is adsorbed by MIL-101(Fe). Moreover, the membrane surface after filtration was observed by SEM, as presented in Figure 10b.
In the non-pretreatment group, silica scale was concentrated on the membrane surface, while in the MIL-101(Fe) adsorption pretreatment group, there was a visible peak-valley structure and less scale on the RO membrane surface.
NaOH [18] with a flow rate of 1.2 L/min. After 2 h of cleaning, deionized water was used to rinse the membrane and filtration system. The membrane flux recovery rate was calculated using the initial flux and the flux after cleaning. As shown in Figure 10d, the water flux recovery rates of the membranes reached 96.1% for the pretreatment group and 84.7% for the non-pretreatment group, suggesting that chemical cleaning using an NaOH solution could effectively restore water flux by increasing silica solubility and depolymerizing colloidal silica [23]. In addition, the adsorption pretreatment of the feed solution favored an improvement in the efficiency of chemical cleaning, since the polymerization degree of silicic acid was greatly reduced after the adsorption [29]. Thus, the pretreatment of the feed solution based on the adsorption of MIL-101(Fe) became a more efficient approach for mitigating silica scale on RO membranes.

Conclusions
In this work, the metal-organic framework MIL-101(Fe) was successfully synthesized for the adsorption of silicic acid. Owing to its numerous unsaturated ion sites, MIL-101(Fe) exhibited excellent adsorption ability for silicic acid. The results of the adsorption experiment showed that the maximum adsorption capacity of MIL-101(Fe) for silicic acid could reach 220.1 mgSiO2·g −1 . On account of the positive correlation between OHand adsorbed silicic acid, the adsorption process of silicic acid on MIL-101(Fe) should involve two steps. Firstly, silicic acid binds directly to the unsaturated ion sites. As the adsorption proceeds, all the ion sites are occupied, resulting in the free silicic acid interacting with the adsorbed silicon to form Si-O-Si bonds. When the feed solution of RO systems was pretreated with the adsorption of MIL-101(Fe), the membrane flux only decreased by 13% after 24 h of continuous filtration, indicating that the adsorption treatment with MIL-101(Fe) could effectively be applied for silica-scaling mitigation in RO systems.  Moreover, after 24 h of silica scaling, the membrane was cleaned with 0.1 M NaOH [18] with a flow rate of 1.2 L/min. After 2 h of cleaning, deionized water was used to rinse the membrane and filtration system. The membrane flux recovery rate was calculated using the initial flux and the flux after cleaning. As shown in Figure 10d, the water flux recovery rates of the membranes reached 96.1% for the pretreatment group and 84.7% for the non-pretreatment group, suggesting that chemical cleaning using an NaOH solution could effectively restore water flux by increasing silica solubility and depolymerizing colloidal silica [23]. In addition, the adsorption pretreatment of the feed solution favored an improvement in the efficiency of chemical cleaning, since the polymerization degree of silicic acid was greatly reduced after the adsorption [29]. Thus, the pretreatment of the feed solution based on the adsorption of MIL-101(Fe) became a more efficient approach for mitigating silica scale on RO membranes.

Conclusions
In this work, the metal-organic framework MIL-101(Fe) was successfully synthesized for the adsorption of silicic acid. Owing to its numerous unsaturated ion sites, MIL-101(Fe) exhibited excellent adsorption ability for silicic acid. The results of the adsorption experiment showed that the maximum adsorption capacity of MIL-101(Fe) for silicic acid could reach 220.1 mgSiO 2 ·g −1 . On account of the positive correlation between OH − and adsorbed silicic acid, the adsorption process of silicic acid on MIL-101(Fe) should involve two steps. Firstly, silicic acid binds directly to the unsaturated ion sites. As the adsorption proceeds, all the ion sites are occupied, resulting in the free silicic acid interacting with the adsorbed silicon to form Si-O-Si bonds. When the feed solution of RO systems was pretreated with the adsorption of MIL-101(Fe), the membrane flux only decreased by 13% after 24 h of continuous filtration, indicating that the adsorption treatment with MIL-101(Fe) could effectively be applied for silica-scaling mitigation in RO systems.