Desalination 344 (2014) 137–143 Contents lists available at ScienceDirect Desalination journal homepage: www.elsevier.com/locate/desal Silica removal to prevent silica scaling in reverse osmosis membranes S. Salvador Cob a,b,c,⁎, B. Hofs a, C. Maffezzoni a, J. Adamus b, W.G. Siegers a, E.R. Cornelissen a, F.E. Genceli Güner d, G.J. Witkamp a,b,c a KWR Watercycle Research Institute, P.O. Box 1072, 3430 BB Nieuwegein, The Netherlands Biotechnology Dept., Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands Wetsus, Centre for Sustainable Water Technology, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands d Process & Energy Dept., Delft University of Technology, Leeghwaterstraat 39, 2628 CB Delft, The Netherlands b c H I G H L I G H T S • • • • Different methods to remove silica from solution were investigated. Al3 + was the most efficient precipitant for silica, removing up to 99% of silica. A strongly basic anion exchange resin removed silica up to 94%. Monitoring residual silica and Al3 + is crucial to prevent scaling in membranes. a r t i c l e i n f o Article history: Received 3 October 2013 Received in revised form 13 March 2014 Accepted 17 March 2014 Available online xxxx Keywords: Scale control High recovery Precipitation Aluminum ions Ion exchange a b s t r a c t Reverse osmosis membranes are increasingly used in drinking water treatment. However, the production of a concentrate stream is the main disadvantage of its application. Increasing the recovery of the membranes in order to have the smallest amount of concentrate possible is an attractive approach. In the absence of bivalent cations in the feed water, silica and silica-derived precipitants are limiting factors in high-recovery reverse osmosis operations. The removal of silica in a separate pretreatment process might be the solution. Several methods were tested to remove silica. Precipitation of silica with Fe(OH)3, Al(OH)3 and silica gel was investigated, and also the removal of silica using a strongly basic anion (SBA) exchange resin. Al(OH)3 was the most effective precipitant for silica, removing nearly all of the molecularly dissolved silica. However, a residual amount of aluminum remained in solution, and aluminosilicate colloids were not removed. The use of the SBA exchange resin also showed a good performance, removing up to 94% of the silica. However, further investigations, such as checking whether the residual small amounts of silica and aluminum can still cause scaling in the membrane, need to be conducted. © 2014 Elsevier B.V. All rights reserved. 1. Introduction Nowadays membrane technology is progressively used to produce drinking water [1]. Nanofiltration (NF) and reverse osmosis (RO) produce high quality water by removing pathogens, organic micropollutants, colloids, natural organic matter and salts. However, NF and RO have several drawbacks, like membrane fouling and the production of concentrate. One of the major foulants of RO membranes is silica. Its ubiquitous presence in natural waters can complicate desalination processes because of its high scaling potential [2–4], and the complex chemistry of silica adds to the difficulty of this problem. Silica solubility is low, about 120 mg/L in water at pH 7 and at 25 °C. It is usually assumed that as long as the ⁎ Corresponding author at: KWR Watercycle Research Institute, Groningenhaven 7, P.O. Box 1072, 3430 BB Nieuwegein, The Netherlands. Tel.: +31 306069581. E-mail address: sara.salvador.cob@kwrwater.nl (S. Salvador Cob). http://dx.doi.org/10.1016/j.desal.2014.03.020 0011-9164/© 2014 Elsevier B.V. All rights reserved. concentration of Si(OH)4 is below 120 mg/L there is no polymerization, but it could be that solutions with lower concentrations might nucleate less soluble polymeric species of lower solubility [5]. Furthermore, the presence of iron and aluminum decreases the solubility of silica [5]. Once silica polymerizes, deposition on the membranes is likely and it is difficult to remove it by cleaning. Therefore, preventing the occurrence of silica scaling is preferred. In our previous research we studied the feasibility of achieving a high RO recovery in a system composed of a cation exchange (CIEX) resin, followed by treatment with NF and RO, which treated the NF concentrate [6]. We found that at N94% total system water recovery, silica scaling became the limiting factor for the investigated water type. The silica concentrations in the bulk solution of the RO feed varied between 80 and 200 mg/L SiO2. One of the options to prevent silica scaling consists of removing silica in a separate pretreatment process. Several investigations have been 138 S. Salvador Cob et al. / Desalination 344 (2014) 137–143 Fig. 1. Simplified scheme of the CIEX–NF–RO system. conducted in this field. Silica can be removed from water by precipitation with multivalent metal hydroxides, such as Fe(OH)3, Al(OH)3 and Mg(OH)2. This treatment removes both soluble and colloidal silica [3,7,8]. Other methods of silica removal, such as electrocoagulation and in-line coagulation/ultrafiltration have been also investigated and were able to remove up to 65 and 80% of the silica, respectively [9,10]. Silica can also be removed with anion exchange [11] or by chemical (lime or caustic soda) softening at pH N 10 [12,13]. The removal of silica by chemical softening is in co-precipitation with Mg(OH)2. Thus, to remove silica with chemical softening, it should be applied before CIEX — as CIEX removes nearly all Mg2+(aq). The aim of this study is to find the most suitable method and conditions to remove silica in order to be able to reach very high recoveries with RO without silica scaling. In the present paper we report different experiments to remove silica with Fe(OH)3, Al(OH)3, silica gel and a strongly basic anion (SBA) exchange resin, performed with synthetic water, tap water, and water extracted from a pilot plant [6], paying special attention to the amounts of residual silica and residual precipitant after the treatment. With the results obtained we can design a pretreatment step to remove silica, avoiding silica scaling in the RO membrane. 2. Materials and methods Two different kinds of experiments were performed in an attempt to remove silica in the most efficient way. The first method was the removal of silica by precipitation and the second one by means of a strongly basic anion (SBA) exchange resin. These analyses were done immediately after sampling, following the silicomolybdate Hach method 8185. Total silica (reactive and non-reactive) was measured by means of Ion Couple Plasma Atomic Emission Spectrometry (ICP-AES), using an ICP analyzer (Spectro Arcos). In the experiments using AlCl3·6H2O, the dissolved silica was too low to be determined spectrophotometrically, thus, the remaining silica was total silica measured only with ICP-AES. The concentrations of aluminum remaining in the treated water were determined after filtration with a 0.22 μm filter, by ICP-AES. 2.1.2. Water type The experiments were performed using different types of water: deionized water and water extracted from a pilot plant described elsewhere [6]. The pilot achieved high recoveries in the treatment of tap water by subsequent application of CIEX, NF and RO (on the NF concentrate). In the pilot the water was recirculated over both the NF and RO in local loops, in order to get sufficient cross-flow. Three water types were extracted from the pilot, to investigate the efficiency and impact of precipitation using Al3+ for the removal of silica at three different possible stages in the pilot installation. The extracted water types were: the tap water feeding the pilot (‘tap water’), water from the NF recirculation loop (‘NF recirculation’) and water from the RO recirculation loop (‘RO recirculation’). A simplified scheme of the system is shown in Fig. 1. The average composition of these streams, including the concentrations of silica, is given in Table 1. By performing precipitation experiments in three different water types, the most optimal location for silica precipitation in a CIEX–NF–RO system can be assessed. 2.1. Silica removal by precipitation 2.1.1. Experimental procedure Silica removal experiments were performed using 250 mL plastic covered beakers. The solutions were placed in the beakers and the required reagent (see below) was added, if necessary. The pH was checked with a pH meter (Radiometer Copenhagen PHM95) and adjusted to the desired value by addition of (1 M) HCl or (1 M) NaOH (Sigma Aldrich). Supersaturated silica solutions (200 mg/L) were prepared by dissolving the corresponding amount of Na2SiO3·5H2O (Sigma Aldrich) in deionized water. The FeCl3·6H2O, AlCl3·6H2O and silica gel used to precipitate the silica were from JT Baker, Merck and Sigma Aldrich, respectively. The silica gel had a particle size ranging from 9.5 to 11 μm and its pore size ranged from 50 to 76 Å. The solutions were stirred continuously at 600 rpm. The samples for analysis were collected every 10 or 30 min, depending on the experiment, with a plastic syringe and filtrated with a 0.22 μm Millipore filter. All the experiments were done in duplicate. In all the experiments using deionized water, the pH was adjusted to 8.5 ± 0.3. Dissolved silica, also known as reactive silica, was analyzed with a spectrophotometer at 452 nm wavelength (Hitachi U-2900). 2.1.3. Precipitate characterization and particle size distribution For the design of a separate process for removing silica it is important to characterize the precipitate formed [14]. Information about particle size and shape is needed to design the proper filtration unit to follow the precipitator. Ultrafiltration, for instance, might be required to filter residual colloids after precipitation. Table 1 Average composition in the different water streams tested (all in mg/L except pH). Ca2+ Na+ Mg2+ K+ Al3+ Fe3+ Cl− DOC HCO− 3 SiO2 pH Tap water NF recirculation RO recirculation 70 15 5.98 1.23 0.02 0.00 10 1.90 271 18 8.1 0.40 677 0.03 0.57 0.12 0.03 47.5 16 1540 20 8.3 2.31 4300 1.21 3.37 0.50 0.10 285 101 10,300 140 8.8 S. Salvador Cob et al. / Desalination 344 (2014) 137–143 139 Table 2 Tap water average quality (all concentrations as mg/L). mg/L Ca2+ Na+ Mg2+ Al3+ K+ Fe3+ Ba2+ Cr2+ Cu2+ Sr2+ B3+ Cl− HCO− 3 SO2− 4 SiO2 DOC 70 14 5.86 0.02 1.25 0.01 0.02 0.01 0.05 0.23 0.01 9.19 277 b2 20 1.95 Fig. 3. Soluble silica concentration over time at different silica gel concentrations (pH 8.5). When SI N 0 the compound is supersaturated and there is risk of scaling [15]. Two techniques were used to characterize qualitatively the precipitate formed in the experiments using the RO concentrate, powder X-ray diffraction (XRD) and scanning electron microscope (SEM). The precipitate was collected on a 0.22 μm filter. XRD patterns were recorded in a Bragg–Brentano geometry in a Bruker D5005 diffractometer equipped with a Huber incident-beam monochromator and a Braun PSD detector. Data collection was carried out at room temperature using monochromatic Cu radiation (Kα1 λ = 0.154056 nm) in the 2θ region between 10° and 90°, step size 0.038° 2θ. Data evaluation was done with the Bruker program EVA. For the SEM analysis a JEOL-6480LV (JEOL Company) equipped with a Noran system SIX X-ray microanalysis (EDX) system (Thermo Electron Corporation) was used to determine the structure and the composition. The samples were coated with a thin (10 nm) Au layer. An accelerating voltage of 6 kV was used for SEM observation and 10 kV for the EDX analysis. The particle size of the precipitate was monitored using a Microtrac S3500 Particle Size Analyzer (Anaspec Solutions). Liquid samples were extracted from the beaker and poured into methanol. Ultrasound was applied to avoid agglomeration of the particles. 2.1.4. Phreeqc calculations Phreeqc-2 software was used to calculate the saturation index (SI) of the sparingly soluble inorganic salts in water using the database Wateq4f. The SI is defined as   SI ¼ log IAP=Ksp ð1Þ where, IAP is the ion activity product and Ksp is the thermodynamic solubility product. Fig. 2. Soluble silica concentration over time after addition of different concentrations of iron (pH 8.5). 2.2. Silica removal using a strong basic anion exchange resin Silica can also be removed with a SBA exchange resin. To test this method, we performed a bench scale experiment. The resin used for this test (Lewatit MonoPlus M500) was in the OH− form. 2.2.1. Experimental procedure A plastic tank was filled with 20 L of locally available tap water with a known silica concentration. Then, 100 mL of resin was poured into a small column 18 cm high and 5 cm in diameter. A peristaltic pump (Masterflex L/S) was connected between the tank and the top of the column using a silicone tube. The flow was set at 55 mL/min and it was regularly checked with a stopwatch and a graduated cylindrical flask. The silica concentration in the discharge of the column was monitored with a silica analyzer (Alert Colorimeter ADI 2004 from Applikon) in order to determine the breakthrough of silica. The pH was also measured using a Meterlab Standard pH meter PHM210. 2.2.2. Water type The water for this experiment was locally available drinking water. This water was produced from groundwater at plant Tull en 't Waal (Water Supply Company Vitens) by aeration and rapid sand filtration with addition of polyaluminum chloride and without post-chlorination. The quality of the water is given in Table 2 (published before in [6]). The water had an average pH of 8.1. Fig. 4. Soluble silica concentration over time combining Fe3+ and silica gel (pH 8.5). 140 S. Salvador Cob et al. / Desalination 344 (2014) 137–143 3.1.2. Precipitation with aluminum in deionized water Different amounts of AlCl3·6H2O (amount of Al3 + given) were added to 200 mg/L silica solutions. The changes in the dissolved silica concentration in time are shown in Fig. 5. Silica removal at pH 8.5 by addition of Al3+ was very fast, compared to removal by addition of Fe3+ and/or silica gel. After 10 min, most of the silica was already removed for all the different aluminum doses. The maximum removal of silica, 99.9%, was achieved after 1 h, by addition of 400 mg/L of Al3+, with a residual silica concentration 0.29 mg/L and residual aluminum of 1.27 mg/L. Increasing the amount of Al3 + from 125 to 400 mg/L significantly increased the performance, from 77.3 to 99.9% removal of silica. Fig. 5. Soluble silica concentration over time at different Al3+ concentrations (pH 8.5). 3. Results and discussion 3.1. Silica removal by precipitation 3.1.1. Precipitation with iron and silica gel in deionized water Different amounts of Fe3+ were added to 200 mg/L silica solutions at a pH of 8.5 and the subsequent change in the concentration of dissolved silica was monitored, as is shown in Fig. 2. In the control experiment, silica concentration dropped to 180 mg/L after 4 h and remained constant. This is known as the pseudoequilibrium concentration of silica [16]. For both Fe3+ concentrations, the soluble silica concentration in the solution after 6 h was about 130 mg/L so about 35% of silica precipitated with iron. Increasing the iron concentration did not increase the silica removal. The same procedure was repeated, with different amounts of silica gel added (Fig. 3). The soluble silica concentration in the solution after 6 h was about 140 mg/L for both silica gel concentrations, thus about 30% of silica was deposited in the silica gel seeds. A third set of experiments was conducted where both Fe3+ and silica gel were added (Fig. 4). 145 mg/L of soluble silica remained in the solution 6 h after adding 100 mg/L of Fe3 + and 2000 mg/L of silica gel; therefore, 28% of the initial silica was removed. The removal efficiency for silica gel and Fe3+ we found is lower than that found by Bremere et al. [8], who found a removal of about 50% in 6 h. This is probably due to the different pH, as we conducted our experiments at pH 8.5, whereas Bremere et al. worked at pH 7. 3.1.3. Precipitation with Al3+ using water from a pilot plant After observing that Al3+ was far more efficient than the other precipitants in reducing the silica content, more experiments were performed with Al3+ and different water types produced by a pilot installation, described in detail elsewhere [6]. Three water types were extracted from the pilot plant: the tap water feeding the pilot, and the NF and RO recirculation streams. Therefore, the optimal location for silica removal could be studied. For the tap water and NF recirculation the pHs were adjusted to 8, 8.5 and 9, which is the optimal range for the removal of silica with Al3+ [17]. For the RO recirculation the pH was constant at 8.8, due to the high amount of bicarbonate (~10,000 mg/L) and carbonate (~300 mg/L) present in the water buffering the solution. Table 3 gives an overview of all the conducted experiments, including the projected concentrations of SiO2 and Al3+ in the RO feed of the pilot plant after silica removal. The silica concentration in the tap water was 18 mg/L. The pH went down to 5.7 upon addition of 30 mg/L of aluminum (as Al3+ by adding AlCl3·6H2O) and to 4.8 upon addition of 60 mg/L of aluminum. The pH was subsequently adjusted to the values given in Table 3. The highest amount of silica removal, 98% after 1 h, was reached at pH 9 with the addition of 60 mg/L of aluminum. Silica removal increased with increasing Al3+ concentration from 30 to 60 mg/L and with increasing pH from 8 to 9. However, the residual dissolved Al3+ concentration in the treated water also increased with increasing pH and initial Al3+ concentration (at the same pH). The increased residual Al3 + with increasing pH is probably due to the higher proportion and solubility of Al(OH)− 4 at higher pH. The same procedure as for the tap water was followed with the NF recirculation water. The silica concentration in the NF recirculation was 20 mg/L and the pH went down to 7.1 and 6.7 for 30 and 60 mg/L of added Al3+, respectively. In the NF recirculation the silica removal by addition of Al3 + showed the same trends as observed in the Table 3 Summary of the experiments performed. Al3+ added (mg/L) Tap water NF recirculation RO recirculation 30 30 30 60 60 60 30 30 30 60 60 60 240 300 400 pH 8 8.5 9 8 8.5 9 8 8.5 9 8 8.5 9 8.8 8.8 8.8 Silica removal (%) 90 93 95 95 97 98 82 86 86 90 94 95 79 79 80 Residual SiO2 (mg/L) 1.77 1.16 0.81 0.77 0.46 0.37 3.39 2.61 2.61 1.91 1.21 0.88 29.5 29.5 27.9 Residual Al3+ (mg/L) 0.64 2.20 5.61 0.74 2.49 7.15 0.69 1.82 1.56 0.84 1.65 2.38 1.78 1.34 0.76 Projected concentrations in RO feed SiO2 (mg/L) Al3+ (mg/L) 17.9 11.7 8.2 7.8 4.7 3.7 40.3 31 31 22.7 14.4 10.5 29.5 29.5 27.9 17.3 60 152 20 67 193 3.10 8.2 7 3.80 7.4 10.7 1.78 1.34 0.76 141 S. Salvador Cob et al. / Desalination 344 (2014) 137–143 Table 4 Residual concentration of HCO− 3 in the RO concentrate after 1 h as a function of Al3+ added at t = 0. Al3+ added (mg/L) HCO− 3 (mg/L) 240 300 400 8000 7500 6500 treatment of the tap water. The highest silica removal, 95%, was reached at pH 9 with the addition of 60 mg/L of Al3 +. The residual dissolved Al3+ concentration was lower in treated NF recirculation (2.38 mg/L) than in treated tap water (7.15 mg/L). For the RO recirculation, with an initial silica of 140 mg/L, the silica removal was similar with varying amounts of added Al3 + from 240 to 400 mg/L. However, the residual dissolved Al3+ concentration decreased from 1.78 to 0.76 mg/L with an increasing amount of added Al3+. The HCO− 3 concentration in the RO recirculation also decreased with an increasing amount of added Al3+ (Table 4). The projected concentrations in the RO feed (Table 3) are calculated based on known rejections for Al3+ and silica for the NF and RO membranes in the pilot installation. If silica is removed in the tap water feeding the pilot, the stream will be further concentrated in the NF and RO steps, therefore the silica and, specially, Al3 + concentrations are expected to increase considerably. The projected amount of Al3+ in this case (17–193 mg/L) is too high to consider treatment of the tap water with Al3+ as a feasible option. The saturation indices of various aluminosilicates would probably be high (see Section 3.1.3.1), and aluminosilicates would cause scaling on the RO membrane. After treatment of the NF recirculation by addition of Al3+, the treated NF recirculation water would be further concentrated in the RO step. The projected concentrations of SiO2 and Al3 + in the RO feed are the same as the concentrations after treatment, as the treated water would be directly fed to the RO. The coprecipitation of silica with aluminum depends on several factors, of which the pH and the state of silica (colloidal or molecularly dissolved) are probably the most important. Okamoto et al. [17] showed that 45–100 mg/L of colloidal silica can be precipitated by minor amounts of added aluminum sulfate at pH 4 to about pH 5. The amount of added aluminum sulfate is critical with respect to precipitation: if the ratio silica to aluminum exceeds a certain value, colloidal silica cannot be precipitated. This is probably because positively charged aluminum ions cause charge compensation and bridging (mostly due to AlOH2+ which has a maximum contribution to the aluminum species at a pH of about 4.6 [18]) by adsorption onto the negatively charged silica colloids (at pH N 4). At higher pH, this mechanism is no longer possible as the aluminum ion species become neutral, or negatively charged. Okamoto et al. [17] also showed that the precipitation of dissolved silica on the other hand, requires the addition of large amounts of aluminum sulfate and has an optimum at pHs 8–9. This is in apparent contradiction with the results obtained by Gallup [19], who showed that 1000 mg/L of dissolved silica can be precipitated by minor amounts of aluminum ions in the pH range of 6–9. However, the relatively high concentration of silica that Gallup used will inevitably lead to polymerization of dissolved silica to colloidal silica in most of the used pH range and T. Thus, most likely aluminosilicate colloids were precipitated by the aluminum ions, which explains the different behavior with respect to the pH range of precipitation and the required dose of aluminum observed in the two papers. This is probably because silica and aluminosilicate colloids have very different surface properties with respect to the isoelectric point [20]. In our experiments, the silica concentration was relatively low for the deionized water, tap water and NF recirculation. Thus, removal of molecularly dissolved silica by the addition of Al3+ was possible down to low residual concentrations of silica. The fact that some aluminosilicates were already supersaturated in the RO recirculation, even before treatment (before addition of Al3 +, see Section 3.1.3.1), probably led to the formation of aluminosilicate colloids. The removal of molecularly dissolved silica and the silica and aluminosilicate colloids by addition of Al3 + proceeds via different mechanisms, at different pH values. Therefore, the removal of total silica by addition of Al3 + from the RO recirculation at pH 8.8 probably succeeded only in removing the molecularly dissolved silica, and did not remove the aluminosilicate colloids. Due to the presence of about 100 mg/L dissolved organic matter (DOC) in the RO recirculation it was not possible to use the colorimetric method to determine the concentration of monomeric silica. The concentration of silica remaining in solution after filtration was determined with ICP-AES. Thus, the total silicon concentration was determined in this case (from which the total silica concentration was calculated), and not just the molecularly dissolved silica concentration. For the RO recirculation the amount of silica removed was constant with increasing the amount of Al3+ added. The residual Al3+ decreased slightly with increasing the amount of Al3+ added. Probably, the presence of bicarbonate plays an important role in capturing the excess aluminum and in determining the residual aluminum concentration. The higher the dose of Al3 +, the higher the removal of HCO− 3 . The higher the bicarbonate concentration was (tap water = 270 mg/L; NF concentrate = 1500 mg/L), the lower the residual aluminum concentration was. In the RO recirculation, specifically, there was a high concentration − of HCO− 3 , around 10,000 mg/L. Possible removal mechanism for HCO3 are the formation of an unknown solid phase composed of at least aluminum, carbonate and other ions, or absorption onto formed alumi3+ nosilicate colloids [21]. As much more HCO− 3 was removed than Al was added, the formation of a solid phase seems more likely. This also explains why the HCO− 3 concentration decreased with increasing amounts of aluminum added. 3.1.3.1. Phreeqc calculations. The projected concentrations in the RO feed for the experiments with the NF recirculation adding 60 mg/L of Al3+ and the RO recirculation (Table 3) were put into Phreeqc to calculate the saturation indices of various minerals. The saturation indices in the RO recirculation before any treatment have been also added. The results of the calculations are given in Table 5. From the saturation indices' values we can see that there is high supersaturation of inorganic compounds (minerals) containing Al, Si and O in all cases. These indices are considerably higher before addition of Al3+. The lowest supersaturation values were found after removal of silica in the NF recirculation at pH 9 and in the RO recirculation after addition of 400 mg/L Al3+. These results showed that the supersaturation of these minerals at the inlet of the RO membrane would possibly lead Table 5 Saturation indices in the RO feed before and after removing silica in the NF and RO recirculation (albite: NaAlSi3O8; kaolinite: Al2Si2O5(OH)4; pyrophyllite: Al2Si4O10(OH)2). Saturation indices Name Albite Kaolinite Pyrophyllite Before addition of Al3+ 4.58 6.51 10.62 NF recirculation (60 mg/L Al3+) RO recirculation pH 8 pH 8.5 pH 9 240 mg/L Al3+ 300 mg/L Al3+ 400 mg/L Al3+ 2.44 7.35 9.46 2.26 5.2 8.2 1.69 5.46 6.79 2.26 5.2 7.42 2.22 5 7.28 1.97 4.51 6.79 142 S. Salvador Cob et al. / Desalination 344 (2014) 137–143 to aluminosilicate scaling. Furthermore, this emphasizes the importance of minimizing not only the SiO2, but also the residual dissolved Al3+ concentration. 3.1.3.2. Precipitate characterization and particle size distribution. The precipitate formed 1 h after adding 400 mg/L of aluminum to the RO concentrate was characterized with XRD and SEM–EDX. The XRD analysis indicated that the precipitate was amorphous, and no crystalline form was identified. The SEM image (Fig. 6) showed a homogenous precipitate with many very small particles (≪1 μm) with some larger particles (of about 0.5 μm). The picture looked similar to previous SEM images of an aluminum-silicate scaled RO membrane [6]. The EDX results (Table 6) showed that the precipitate contained C, O, Al, Na and Si. The ratio Al–Si found was 8.6, which was in agreement, approximately, with the ratio of the amount of aluminum added and the amount of silica precipitated (7.9). The carbon percentage did not match with the precipitated carbon present in HCO− 3 (it should be ~ 26%). The reason might be that the HCO− 3 precipitate was partially smaller than the filter size (0.22 μm). The particle size distribution of the precipitate formed in solution 1 h after addition of 400 mg/L of aluminum to the RO recirculation was measured with a Microtrac S3500 particle size analyzer (Fig. 7). The precipitate apparently had a trimodal distribution, with very small particles of around 0.16 μm, medium sized particles of about 0.52 μm and a very small amount of relatively large particles, which was in agreement with the SEM analysis. The smallest particles could already be seen easily in the SEM picture and consist of about Table 6 EDX analysis of the Si–Al precipitate. Element Atom% C O Al Na Si Cl Fe 7.0 71.9 14.6 4.2 1.7 0.4 0.0 55.1 vol.% of the precipitate volume. The medium sized particles, around 0.52 μm, take up 41.4 vol.% and the largest particles, of about around 3.45 μm, take up only 3.5 vol.%. The measured particle size distribution is split up into two main populations, which probably correspond to two different types of colloids. The smallest particles are close in size to the pore size of the filter (0.22 μm), and these are probably the solid phase composed of aluminum and bicarbonate, that would explain why there is carbon missing in the EDX analysis. The larger particles at around 0.5 μm are probably composed of the silica that was dissolved in the RO recirculation and the added aluminum. The aluminosilicate colloids that were already present in the RO recirculation were probably smaller than the filter size. To completely remove silica from the RO recirculation thus would require application of both Al3+ (to remove molecularly dissolved silica) and ultrafiltration (to remove colloids). In conclusion, the precipitation of silica with Al3+ seems to be an efficient way of removing molecularly dissolved silica from solution. However, the residual concentration of Al3+ should be as low as possible as aluminosilicates have about ten times lower solubility than silica [22]. Thus, removal of silica by Al3+ may prevent silica scaling, but may cause aluminosilicate scaling. From the Phreeqc calculations one can see that after removing silica from the solution there is still supersaturation of aluminosilicates. Therefore, further research should focus on investigating the possibility of scaling in the RO membrane after silica removal. 3.2. Silica removal using a strong basic anion exchange resin Another option to remove silica is with an SBA exchange resin. The main advantage of removing silica by anion exchange is the absence of a need to introduce multivalent cations, like Fe3 + and Al3 +. The anion exchange resin was placed in the small column and the tap water (see Table 2) was pumped through the column in fixed bed mode with a linear velocity of 3.2 m/s. We monitored the silica concentration before and after the resin and the pH after the resin. The change in the pH and silica concentration in the effluent of the anion exchange column is shown in Fig. 8, as a function of the treated bed volumes. The silica concentration in the feed to the column was constant (18 mg/L) and the pH was 8.1. Removal of silica was, on average, 94% before breakthrough, which was at about 50 bed volumes. The pH of the treated water after the removal of silica was 11.8, and after the Fig. 6. SEM image of the Si–Al precipitate (left) and blow-up of the SEM image (right). Fig. 7. Particle size distribution of the Si–Al precipitate. S. Salvador Cob et al. / Desalination 344 (2014) 137–143 143 (www.wetsus.nl). Wetsus is funded by the Dutch Ministry of Economic Affairs, the European Union Regional Development Fund, the Province of Fryslân, the City of Leeuwarden and the EZ/Kompas program of the “Samenwerkingsverband Noord-Nederland” and by the Joint Research Programme of the Dutch Water Companies. The authors would like to thank the participants of the research theme “Clean Water Technology” for the discussions and their financial support. We would like to thank Michel van den Brink for the ICP analysis, Arie Zwijnenburg for the SEM–EDX analysis and Ruud Hendrikx for the XRD analysis. Finally, we also would like to thank the reviewers for their comments, which helped to improve the quality of this paper. References Fig. 8. Silica concentration and pH in the effluent of the anion exchange column as a function of the treated bed volumes. breakthrough it reached pH 8. The overshoot in silica concentration at 70–100 bed volumes occurs because the tap water contains many anions whose affinity for the anion exchange resin is higher than that of silica. If the water contained only silica the breakthrough would be delayed by about 50 bed volumes. The low selectivity of SBA exchange resins for silica (typically b1) [23] makes this option less attractive for large scale operations, due to the great volumes of resin needed. Furthermore, the pH during the operation reached a value of 12, so after the resin a pH correction would be needed. A smarter option is probably to only treat a part of the stream with the SBA exchange resin. The removed silica, and the increased pH lead to a higher solubility of silica. This in turn means a higher recovery can be reached by treatment with RO before silica scaling becomes limiting. This solution can easily be applied in practice, but would require fairly large amounts of NaOH for the regeneration of the SBA exchange resin. 4. Conclusions If bivalent cations have been removed from water, silica remains as a recovery limiting scalant in RO membranes. Silica must be removed to allow high recoveries (depending on the feed water quality, N94%) in NF–RO systems. In this work different methods have been tested: precipitation with Fe3 +, Al3 + and silica gel, and the use of a strong basic anion exchange resin. Precipitation with Al3 + seems to be the most efficient method to remove molecularly dissolved silica. Unfortunately, colloidal aluminosilicates are not removed from brines by addition of Al3+, at the tested pH of 8.8. Silica removal is thus possible by either SBA exchange resin, or by a combination of ultrafiltration and Al3+ dosing. However, for the latter case, further research should be conducted in order to find out if the small amounts of remaining silica and aluminum in solution are sufficient to cause scaling in RO membranes. Acknowledgments This work was performed in the TTIW-cooperation framework of Wetsus, Centre of Excellence for Sustainable Water Technology [1] L.F. Greenlee, et al., Reverse osmosis desalination: water sources, technology, and today's challenges, Water Res. 43 (9) (2009) 2317–2348. [2] R. Sheikholeslami, S. Tan, Effects of water quality on silica fouling of desalination plants, Desalination 126 (1–3) (1999) 267–280. [3] R. Semiat, I. Sutzkover, D. Hasson, Scaling of RO membranes from silica supersaturated solutions, Desalination 157 (1–3) (2003) 169–191. [4] M. Badruzzaman, et al., Impacts of silica on the sustainable productivity of reverse osmosis membranes treating low-salinity brackish groundwater, Desalination 279 (1–3) (2011) 210–218. [5] R.K. Iler, The Chemistry of Silica; Solubility, Polymerization, Colloid and Surface Properties, and Biochemistry, 1979. [6] S. Salvador Cob, et al., Silica and silicate precipitation as limiting factors in highrecovery reverse osmosis operations, J. Membr. Sci. 423–424 (15) (2012) 1–10. [7] R. Sheikholeslami, et al., Pretreatment and the effect of cations and anions on prevention of silica fouling, Desalination 139 (1–3) (2001) 83–95. [8] I. Bremere, et al., Prevention of silica scale in membrane systems: removal of monomer and polymer silica, Desalination 132 (1–3) (2000) 89–100. [9] H.-H. Cheng, S.-S. Chen, S.-R. Yang, In-line coagulation/ultrafiltration for silica removal from brackish water as RO membrane pretreatment, Sep. Purif. Technol. 70 (1) (2009) 112–117. [10] W. Den, C.-J. Wang, Removal of silica from brackish water by electrocoagulation pretreatment to prevent fouling of reverse osmosis membranes, Sep. Purif. Technol. 59 (3) (2008) 318–325. [11] M.B. Sik Ali, et al., Silica removal using ion-exchange resins, Desalination 167 (2004) 273–279. [12] C.J. Gabelich, et al., High-recovery reverse osmosis desalination using intermediate chemical demineralization, J. Membr. Sci. 301 (1–2) (2007) 131–141. [13] A. Rahardianto, et al., High recovery membrane desalting of low-salinity brackish water: Integration of accelerated precipitation softening with membrane RO, J. Membr. Sci. 289 (1–2) (2007) 123–137. [14] O. Söhnel, J. Garside, Precipitation Basic Principles and Industrial Applications, 1992. [15] C.A.J. Appelo, D.P., Geochemistry, Groundwater and Pollution, Second ed. CRC Press. Taylor & Francis Group, 2005. [16] B.A. Fleming, Kinetics of reaction between silicic acid and amorphous silica surfaces in NaCl solutions, J. Colloid Interface Sci. 110 (1) (1986) 40–64. [17] G. Okamoto, T. Okura, K. Goto, Properties of silica in water, Geochim. Cosmochim. Acta 12 (1–2) (1957) 123–132. [18] D. Panias, P. Asimidis, I. Paspaliaris, Solubility of boehmite in concentrated sodium hydroxide solutions: model development and assessment, Hydrometallurgy 59 (1) (2001) 15–29. [19] D.L. Gallup, Aluminum silicate scale formation and inhibition: scale characterization and laboratory experiments, Geothermics 26 (4) (1997) 483–499. [20] M. Kosmulski, pH-dependent surface charging and points of zero charge. IV. Update and new approach, J. Colloid Interface Sci. 337 (2) (2009) 439–448. [21] C. Su, D.L.S., In situ infrared speciation of adsorbed carbonate on aluminum and iron oxides, Clay Clay Miner. 45 (6) (1997) 814–825. [22] D.L. Gallup, Aluminum silicate scale formation and inhibitor (2): scale solubilities and laboratory and field inhibition tests, Geothermics 27 (4) (1998) 485–501. [23] The Dow Chemical Company, DOWEX ion exchange resins — Technical InformationForm No. 177-01755-0207.