Bioresource Technology 100 (2009) 74–78Contents lists available at ScienceDirect Bioresource Technology journal homepage: www.elsevier.com/locate/biortech Effect of mixing on spontaneous struvite precipitation from semiconductor wastewater Daekeun Kim, Jinhyeong Kim, Hong-Duck Ryu, Sang-Ill Lee * Department of Environmental Engineering, Chungbuk National University, Cheongju, Chungbuk 361-763, Republic of Korea a r t i c l e i n f o a b s t r a c t The objective of this study was to investigate on the effect of mixing intensity (G) and mixing duration (td) on struvite precipitation in the chemical mechanical polishing (CMP) wastewater generated from the semiconductor manufacturing process. Batch-scale experiments revealed that struvite crystallization was affected by both G and td. The mixing effect was to enhance the mass transfer of solute to the crystals in the process, resulting in the improvement of struvite crystallization and growth. By forming struvite, removal efficiencies of N and P increased logarithmic with the multiple values of G and td, i.e., Gtd. Insufficient mixing energy with the Gtd value less than 105 caused an increase in the formation potential of unexpected precipitate unlike to pure struvite, causing a decrease in removal efficiencies of N and P in the process. At the Gtd value over 106, struvite precipitation was not restricted by fluoride, of which high level inherently contained in the CMP wastewater. The study results can be taken into consideration in the design and operation of the struvite precipitation process for both nutrient (N and P) removal and recovery. Ó 2008 Elsevier Ltd. All rights reserved. Article history: Received 27 December 2007 Received in revised form 13 May 2008 Accepted 17 May 2008 Available online 2 July 2008 Keywords: Struvite Mixing Nitrogen Phosphorus Fluoride 1. Introduction The semiconductor manufacturing process generates a large quantity of hazardous wastes, which include organic solvents, acids, bases, salts, heavy metals, fine suspended oxide particles, and other organic compounds (Huang and Liu, 1999; Lai and Lin, 2003). The management of these wastes has become an important issue in the industry. Especially, since the high levels of ammonia, phosphate, and fluoride are presented in the wastewater generated from the process of chemical mechanical polishing (CMP), the treatment of the CMP wastewater is still a pending question in the industry. It is believed that the biological application is inadequate in the CMP wastewater because it has various toxic substances which inhibit active nitrification by nitrifying bacteria (Lin and Kiang, 2003). As an alternative, the precipitation of nitrogen and phosphorus in a struvite crystal would be attractive in treating high level of ammonia and phosphate in the stream (Kabdasli et al., 2000; Kim et al., 2007; Nelson et al., 2003). Struvite (Magnesium ammonium phosphate, MgNH4PO4 Á 6H2O) crystallized as a white orthorhombic crystalline structure, which is composed of three components, i.e., magnesium, ammonium, and phosphate, in equal molar concentrations (Lee et al., 2003). The factors affecting the struvite crystallization reaction include pH, temperature, supersat* Corresponding author. Tel.: +82 043 261 2469; fax: +82 043 272 2469. E-mail address:
[email protected] (S.-I. Lee). 0960-8524/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.biortech.2008.05.024 uration of solutes (NHþ ; Mg2+; PO3À ), the presence of crystalliza4 4 tion nucleus, the type of magnesium source, and impurity ions (Bouropoulos and Koutsoukos, 2000; Kim et al., 2007). In terms of nutrient (i.e., nitrogen and phosphorus) recovery, struvite gained from the stream can be taken for its agricultural reuse. In our previous study (Ryu et al., 2008), the field-scale application of struvite precipitation has been conducted in the CMP wastewater treatment. The system performance was evaluated with respect to the operational parameters, i.e., pH, solute concentration, and impurity ions (other ions except for NHþ , Mg2+, and 4 PO3À ). It is worthwhile to note from our unpublished study that 4 the CMP wastewater showed possible toxicity because organic solvents are unavoidably contained in the wastewater through the polishing process, while any toxicity was not detected in the struvite samples obtained through the precipitation process. It was also observed from our previous study (Ryu et al., 2008) that the reaction of struvite formation apparently varied at the different mixing strength applied in the system (Details have been not reported). It is general in crystallization process that crystal formation is primarily made by nucleation from crystal embryos and nucleus growth in sequence (Doyle and Parsons, 2002; Jones, 2002). The generation of a nucleus depends on the intensity of initial mixing, and its growth relies on mixing duration (Jones, 2002). Ohlinger et al. (1999) observed in their study of the struvite precipitation process that the mixing intensity had a significant effect on struvite crystal nuclei development. Wang et al. (2006) demonstrated that both sturivite crystal size and settling were D. Kim et al. / Bioresource Technology 100 (2009) 74–78 75 significantly affected by mixing strength. However, details about the effect of mixing on struvite precipitation are still limited in our best knowledge. Further investigation should be made to provide information to improve the practical application of struvite precipitation in the CMP wastewater. It is also worthwhile to note that since the CMP wastewater contains a high level of fluoride, the effect of fluoride on the struvite formation should be considered in the process. For this purpose, batch-scale experiments were designed to identify the mixing effect on struvite precipitation in the CMP wastewater. The goal was to investigate (1) the effect of mixing intensity (G) and duration (td) on struvite precipitation with respect to the removals of ammonia and phosphate; (2) the effect of initial concentration of nitrogen in the sample as a function of the multiple value of G and td; (3) mass balance analysis and (4) fluoride removal as a function of the multiple value of G and td. 2. Methods 2.1. Characteristic of CMP wastewater The CMP wastewater tested was obtained from the semiconductor manufacturing plant in Cheongju, Korea. It had pH of 2.9, NH4–N concentration of 100 mg/L (7.1 mM), and PO4–P concentration of 286 mg/L (9.2 mM) on the average. High concentration of fluoride was found about 172 mg/L (9.1 mM) on the average. The characteristics of the sample wastewater were summarized in Table 1. 2.2. Materials For the struvite reaction, magnesium chloride (MgCl2 Á 6H2O) solution was used as an alternate source of magnesium ions. In all experiments, magnesium was fed to reach at the molar ratio of NH4:Mg of 1:1.2 because struvite precipitation occurs when the combined concentrations of the three components, i.e., NH4, Mg, and PO4, exceed the struvite solubility limit (Ohlinger et al., 1998). 2.5 N NaOH was used to adjust pH to 9 in the solutions. In the previous study (Ryu et al., 2008), the optimum pH for struvite formation was found to range from 9 to 10. All chemicals used were analytical grade. 2.3. Experimental setup The standard jar tester with six paddles was used in all experiments. Jars were made of acrylic plastic with dimensions of 11.5  11.5  21 cm and held 2.0 L of liquid. The paddles at the end of each stirrer shaft, which were made of stainless steel, had a diameter of 7.6 cm and a height of 2.5 cm. The stirring device included a tachometer and a controller so as to adjust the revolution rates ranging from 0 to 360 rev/min. All experiments were carried out at ambient laboratory temperature. In all experiments, struvite formation proceeded by simultaneous addition of excess magnesium sources and the sample solution, and the buffering reagent was finally fed. This experimental procedure provided the optimum condition to reach the thermodyTable 1 The characteristics of the wastewater samples used in this study Cations Na NHþ –N 4 K+ 2+ Mg Ca2+ pH a + namic equilibrium of struvite precipitation (Kim et al., 2007). Mixing intensity (G) values tested in this study varied from 49 sÀ1 to 940 sÀ1. The applied mixing duration (td) ranged from 2 min to 90 min. The applied G value was calculated based on the interaction between the revolutions of paddle and the velocity gradient based on the suggestion by Cornwell and Bishop (1983), in which study the velocity gradient had the linear correlation to the impeller speed. 2.4. Analytical method The concentration of ammonia nitrogen was measured by using a UV-Spectrophotometer (HACH DR-4000, Loveland, CO, USA). Orthophosphate level was determined by the ascorbic acid method. The analyses of ions including fluoride were done by using an ion chromatograph (DX-80, Dionex, USA). For the separation of anion, an IonPac AG4A-SC (50 mm  4 mm) guard column and an IonPac AG4A-SC (250 mm  4 mm) analytical column were used. The eluent was 1.8 mM Na2CO3/1.7 mM NaHCO3 at a flow rate of 2 mL/ min. For cationic separation, an IonPac CG412 (50 mm  4 mm) guard column and an IonPac CS12 (250 mm  4 mm) analytical column were used. 20 mM Methane sulfonic acid was used as the eluent at a flow rate of 1 mL/min. The precipitate crystal structure was determined with an X-ray diffraction (XRD, Model DMS2000 system, SCINTAG) and a scanning electron microscopy (SEM, Model S-2500 C, Hitachi) equipped with an energy-dispersive X-ray analyzer (EDX, MS2, Sigma). All soluble parameters were determined in samples filtered through 0.45 lm membrane filter. 3. Results and discussion 3.1. Effect of mixing intensity (G) and mixing duration (td) Fig. 1 describes the effect of values of G and td on the removal of nitrogen and phosphorus in the process of struvite precipitation. Overall removal performances of nitrogen and phosphorus were observed to be similar. At low values of G and td, removal efficiencies of nitrogen and phosphorus were unexpectedly low. In contrast, as those values were increased, removal efficiencies of nitrogen and phosphorus correspondingly increased. Interestingly, the removal curves presented in Fig. 1 well corresponded to the contour line of the multiple values of G and td (i.e., Gtd), indicating that high value of Gtd was unavoidable to achieve sufficient removal performance of nitrogen and phosphorus from the sample. However in Fig. 1, a little different observation was inquiringly found at the 70% PO4–P removal line as compared with that for NH4–N. It should be noted that phosphorus concentration was much higher than that of nitrogen in the sample; struvite crystallizes in equal molar concentrations of NH4, PO4, Mg; and mixing energy affects struvite crystallization. These facts might have remarkable relations with the removal of nitrogen and phosphorus through struvite formation. Detailed discussion is made in the following sections. In Fig. 2, removal efficiencies of nitrogen and phosphorus were plotted as a function of the Gtd value. The data were obtained from the sample with initial concentrations of 94 mg NH4–N/L (6.7 mM NH4–N) and 293 mg PO4–P/L (9.5 Mm PO4–P). It is apparent in Fig. 2 that both removal patterns of nitrogen and phosphorus were a logarithmic increase with the Gtd value. Nitrogen removal efficiencies increased to the 80% level, but remained at this level while the G  td value was at least 106. For phosphorus, the removal efficiency increased to the 70% level while the G  td value reached at 106. Overall, high mixing energy was effective in the removal of nitrogen and phosphorus by forming struvite. Wang et al. (2006) demonstrated that relatively high mixing energy could shorten the time needed for the development of struvite crystal nuclei. It Concentration range (mg/L) 6.8–19.5 74.9–179 2.0–10.1 1.8–2.5 5.0–12.2 2.9a Anions F ClÀ NOÀ 3 PO3À –P 4 2À SO4 À Concentration range (mg/L) 129.3–199.72 6.9–11.8 8.3–11.7 209.1–360.4 0–12.7 Dimensionless. 76 D. Kim et al. / Bioresource Technology 100 (2009) 74–78 Fig. 1. Effect of mean velocity gradient (G) and mixing duration (td) on NH4–N removal (a) and PO4–P removal (b) in the process of struvite precipitation: the dotted line represents the contour line for the multiple value of G and td value (Gtd). Initial concentrations in the sample were 94 mg/L (6.7 mM) and 299 mg/L (9.7 mM) for nitrogen and phosphorus, respectively. Fig. 2. NH4–N and PO4–P removal efficiencies as a function of the multiple values of G and td. Initial concentrations in the sample were 94 mg/L (6.7 mM) and 293 mg/L (9.5 mM) for nitrogen and phosphorus, respectively. R2 values of both regression curves were over 0.99. tial concentrations of nitrogen, removal performance seems to have a similar pattern. For each case, nitrogen removal efficiency logarithmically increased with the Gtd value. The maximum removal efficiencies were 98%, 88% and 86% for initial nitrogen concentrations of 179 mg/L (12.8 mM), 110 mg/L (7.9 mM), and 94 mg/L (6.7 mM), respectively. It was also shown in Fig. 3 that the removal efficiency was observed to be relatively high when much higher initial concentration of nitrogen was contained in the sample. Further discussion is made that even at a low initial concentration of nitrogen in the sample, the mixing energy enhanced the nitrogen removal. It is worthwhile to note that the conflict intensity among the solute particles might become stronger as the solute concentration increased. Hence at a low level of the Gtd value applied, the conflict intensity among the solute particles can be a driving force to generate the crystal nucleus. In other words, the particles in solution can accelerate the speed of crystal nuclei generation while the solute concentration increased. Bouropoulos and Koutsoukos (2000) observed that the induction time to reach the crystallization had an inverse relation with the solution saturation, demonstrating that spontaneous nucleation is possible at supersaturation level. was also interestingly observed from Fig. 2 that below 2  105 of Gtd, phosphorus removal efficiencies were much better than those of nitrogen. In contrast, this performance was in reverse at the Gtd value above 2  105. This observation might be caused by the mixing effect. Further discussed is made in Section 3.3 of mass balance analysis. Microscopic examination of the struvite crystals by using SEM illustrates that the morphology of the crystals precipitated was different depending on the Gtd value applied, while at the same value of Gtd it was alike regardless of the matrix of G and td (Detailed data were not shown). The observation revealed that the morphology of the struvite crystal can be affected by the Gtd applied, but it would be apparently stable regardless of each value of mixing intensity (G) and duration (td) only if the multiple values of G and td are the same. 3.2. Effect of initial concentration of nitrogen in sample It should be noted that the sample targeted had a wide range of nitrogen concentration. Fig. 3 illustrates that at three different iniFig. 3. Effect of initial concentration of nitrogen in the sample on nitrogen removal. R2 values of regression curves were over 0.98. D. Kim et al. / Bioresource Technology 100 (2009) 74–78 77 Momberg and Oellermann (1992) found that at a low ammonia nitrogen concentration, the growth of struvite precipitate was restricted due to the limitation of the transport of ammonium ions to the crystal surface. In terms of mass transfer coefficient as correlations of dimensionless numbers, Sherwood number is roughly proportional to mixing intensity (G) based on the observation by Garcia-Ochoa and Gomez (1998). This demonstrates that by accelerating the mixing in the process, the potential of struvite precipitation can be enlarged due to an enhancement of mass transport of ammonium ions in the system. 3.3. Mass balance analysis In this analysis, it was assumed that chemical precipitation is the only removal mechanisms in the present process, and the ideal molar ratio of Mg:P:N is to be 1:1:1 (Lee et al., 2003) only if the precipitate after the reaction is pure struvite. Hence the comparison of molar mass of three components removed from the sample might give details about the resulting precipitate. The sample, in which pure struvite was precipitated, had to provide a molar mass of three components to be 1:1:1. Fig. 4a shows the comparison of molar mass of three components removed from the sample at different Gtd values. The observed molar ratios of three components were different, depending on the applied Gtd value. While the applied Gtd values were less than 105, molar mass of magnesium was greater than or similar to those of phosphorus and ammonium. In contrast, while the applied Gtd values were over 105, it was smaller than others. Fig. 4b indicates the corresponding removal efficiency of nitrogen in the sample. Nitrogen removal efficiency was relatively high while the applied Gtd values ranged over 105. The precipitate obtained at Gtd of about 2  105 was characterized by using an X-ray diffractrogram, revealing that the peaks at this case exhibited several peaks indicative of the presence of struvite. However, the precipitates at other range of Gtd were unlike to that of struvite. Inconsistent mass balance presented in Fig. 4 might be resulted by unexpected precipitates. Since the excess mass of external magnesium sources was fed in the system in order to exceed the sturvite solubility limit, magnesium ions can predominately react with OHÀ ions in the solutions readily to produce amorphous products such as Mg(OH)2. The generation of this product correspondingly attributed to relatively high molar ratio of magnesium to ammonium removed from the solution. Further discussion is made that in the condition of excess concentration of magnesium and phosphate, magnesium preferentially react with phosphate to produce different magnesium phosphates such as hannayite (Mg3(NH4)2H4(PO4)4 Á 8H2O), shertelite (Mg(NH4)2H2(PO4)2 Á 4H2O), bobbierite (Mg3(PO4)2 Á 8H2O), and newberyite (MgHPO4 Á 3H2O). Seckler et al. (1996) discussed in their study of the process of calcium phosphate precipitation that the possibility of magnesium phosphate formation can increase up to 45% only if there is no calcium ion. If calcium ion exists, calcium ion might preferentially react with phosphate to produce hydroxyapatite, dicalcium phosphate, and octacalcium phosphate. In the present study, calcium concentration ranged from 5 mg/L to 12.2 mg/L in the CMP wastewater as shown in Table 1. Golubev et al. (2001) demonstrated that the unseeded precipitation kinetics of magnesium phosphate might not only depend on the operational parameters such as the pH, phosphate, and magnesium concentration in the system, but is also affected by the stirring condition including induction period and mixing intensity. Therefore, these unexpected cases were mainly due to insufficient mixing energy applied in the process. This also explains the observation in Fig. 2 that nitrogen removal efficiencies overwhelmed those of phosphorus over 2  105. Since nitrogen concentration was lower than that of phosphorus in this study, nitrogen concentration would be a limiting factor in struvite precipitation with respect to struvite solubility product. This is worthwhile only if sufficient mixing energy was provided in the system. In other cases, pure struvite precipitation may be limited because of poor crystallization and formation of amorphous precipitate or different magnesium phosphates, which was mainly caused by insufficient mixing energy applied. 3.4. Fluoride removal Since the CMP wastewater had a high level of fluoride, it is important to investigate the effect of fluoride on the struvite formation. Fig. 5 shows the removal efficiencies of fluoride in the sample by struvite precipitation at the applied different mixing energy, which was represented by the Gtd value. Fluoride removal performance was different to those of nitrogen and phosphorus observed in Fig. 2. At the Gtd value of 105, fluoride removal was observed to be about 20%. However, it decreased as an increase in the Gtd value, and finally reached at the 1% level while applying over 106 of the Gtd value. Based on the regression analysis as shown in Fig. 5, the relation between the fluoride removal and the Gtd value applied can be explained by as following: log [percent removal of fluoride] = À0.61 log [Gtd] + 4.2, r2 = 0.84. Fig. 4. Mass balance analysis of three components (N, P, and Mg) in the precipitation process: (a) Molar ratio of three components in precipitate; (b) NH4–N removal as a function of molar ratio of PO4–P removed to NH4–N removed. 78 D. Kim et al. / Bioresource Technology 100 (2009) 74–78 effect in the struvite precipitation process. In the process, the removal patterns of nitrogen and phosphorus was a logarithmic increase with the Gtd value. The mixing effect was to enhance the mass transfer of ammonium ions in the solution, which induced the potential of struvite crystallization and growth to increase. The resulting removal of nitrogen and phosphorus was effectively enhanced in this process. At the Gtd value over 105, the precipitates were close to pure struvite. Insufficient mixing energy provided the formation of the unexpected precipitate unlike to struvite. High level of fluoride in the CMP wastewater was not an operational problem in the application of the struvite precipitation process in case high mixing energy was applied. Acknowledgements This research was supported both by the Brain Korea (BK) 21 Project which is administrated by the Korean Ministry of Education and the Eco-Technopia 21 Project which is administrated by the Ministry of Environment, Republic of Korea. References At a relatively low value of Gtd, fluoride would be involved in the struvite crystallization reaction. Fluoride can be considered as one of impurities in this reaction. It has been demonstrated by other studies (Kabdasli et al., 2006; Le Corre et al., 2005) that some foreign ions can be easily adsorbed on the surface of crystals, and hence the crystal growth is likely to be retarded. It has been reported by Ryu et al. (2008) that at high concentrations of fluoride, fluoride inhibited struvite crystallization because fluoride may react with magnesium, forming magnesium fluoride (MgF2). However, while the process was made at a relatively high Gtd value, fluoride would not work for an impurity in struvite crystallization because sufficient mixing energy amplifies the potential of struvite crystallization as well as its growth. In terms of solubility product, struvite (Ksp = 10À13.12 at 25 °C) tended to form easily as compared with magnesium fluoride (Ksp = 10À8.1 at 25 °C) because solutions were supersaturated with three compounds, i.e., Mg2+, NH4+, and PO3À . This can explain why the fluoride removal trends was oppo4 site of nitrogen and phosphorus removal. The other explanation can be made by analyzing the precipitate crystallized. It is also important to note that fluoride was not detected in the precipitates by using EDX. This demonstrates that the only removal mechanism of fluoride in the system would be the weak adsorption on the surface of the precipitate, hence the potential of the adsorption of fluoride on the precipitate became weak as an increase in the mixing energy, probably due to an upward tendency of its desorption. 4. Conclusions This study elucidated the effects of the mixing energy, i.e., mixing intensity (G) and duration (td) in the struvite precipitation process in order to enhance the removal of N and P from the CMP wastewater. Both G and td were the physical variables in the process. The obtained results provide information to improve the practical application of struvite precipitation in the industry. Considering the agricultural use of struvite, the present study gave feasible information of the mixing effect on struvite precipitation in order to obtain pure struvite from the stream. The detailed findings had been drawn as following The Gtd value, the multiplying value of mixing intensity (G) and mixing duration (td), was a useful parameter to explain the mixing Bouropoulos, N.C., Koutsoukos, P.G., 2000. Spontaneous precipitation of struvite from aqueous solutions. Journal of Crystal Growth 213, 381–388. Cornwell, D.A., Bishop, M.M., 1983. Determining velocity gradients in laboratory and field scale systems. Journal of the American Water Works Association 75, 470–475. Doyle, J.D., Parsons, S.A., 2002. Struvite formation, control and recovery. Water Research 36, 3925–3940. Garcia-Ochoa, F., Gomez, E., 1998. Mass transfer coefficient in stirred tank reactors for xanthan gum solutions. Biochemical Engineering Journal 1, 1–10. Golubev, S.V., Pokrovsky, O.S., Savenko, V.S., 2001. Homogeneous precipitation of magnesium phosphates from seawater solutions. Journal of Crystal Growth 223, 550–556. Huang, C.J., Liu, J.C., 1999. Precipitate flotation of fluoride-containing wastewater from a semiconductor manufacturer. Water Research 33, 3403–3412. Jones, A.G., 2002. Crystallization Process System, first ed. Butterworth-Heinemann, Oxford. Kabdasli, I., Parsons, S.A., Tunay, O., 2006. Effect of major ions on induction time of struvite precipitation. Croatica Chemica ACTA 79, 243–251. Kabdasli, T.O., Ozturk, I., Yilmaz, S., Arikan, O., 2000. Ammonia removal from young landfill leachate by magnesium ammonium phosphate precipitation and air stripping. Water Science & Technology 41, 237–240. Kim, D., Ryu, H.-D., Kim, M.-S., Kim, J., Lee, S.-I., 2007. Enhancing struvite precipitation potential for ammonia nitrogen removal in municipal landfill leachate. Journal of Hazardous Materials 146, 81–85. Lai, C.L., Lin, S.H., 2003. Electrocoagulation of chemical mechanical polishing (CMP) wastewater from semiconductor fabrication. Chemical Engineering Journal 95, 205–211. Le Corre, K.S., Valsami-Jones, E., Hobbs, P., Parsons, S.A., 2005. Impact of calcium on struvite crystal size, shape and purity. Journal of Crystal Growth 283, 514–522. Lee, S.I., Weon, S.Y., Lee, C.W., Koopman, B., 2003. Removal of nitrogen and phosphate from wastewater by addition of bittern. Chemosphere 51, 265–271. Lin, S.H., Kiang, C.D., 2003. Combined physical, chemical and biological treatments of wastewater containing organics from a semiconductor plant. Journal of Hazardous Materials 97, 159–171. Momberg, G.A., Oellermann, R.A., 1992. The removal of phosphate by hydroxyapatite and struvite crystallization in South Africa. Water Science & Technology 26, 987–996. Nelson, N.O., Mikkelsen, R.L., Hesterberg, D.L., 2003. Struvite precipitation in anaerobic swine lagoon liquid: effect of pH and Mg:P ratio and determination of rate constant. Bioresource Technology 89, 229–236. Ohlinger, K.N., Young, T.M., Schroeder, E.D., 1998. Predicting struvite formation in digestion. Water Research 32, 3607–3614. Ohlinger, K.N., Young, T.M., Schroeder, E.D., 1999. Kinetics effects on preferential struvite accumulation in wastewater. Journal of Environmental EngineeringASCE 125, 730–737. Ryu, H.-D., Kim, D., Lee, S.-I., 2008. Application of struvite precipitation in treating ammonium nitrogen from semiconductor wastewater. Journal of Hazardous Materials 156, 163–169. Seckler, M.M., Bruinsma, O.S.L., Van Rosmalen, G.M., 1996. Calcium phosphate precipitation in a fluidized bed in relation to process conditions: a black box approach. Water Research 30, 1677–1685. Wang, J., Burken, J.G., Zhang, X.Q., 2006. Effect of seeding materials and mixing strength on struvite precipitation. Water Environment Research 78, 125–132. Fig. 5. Fluoride removal in the process of struvite precipitation: Initial concentrations in the sample were 131 mg/L (6.9 mM), 78 mg/L (5.6 mM), and 191 mg/L (6.2 mM) for fluoride, nitrogen, and phosphorus, respectively.