Sie sind auf Seite 1von 9

ARTICLE IN PRESS

WAT E R R E S E A R C H

41 (2007) 458 466

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Phosphate and potassium recovery from source separated urine through struvite precipitation
J.A. Wilsenach, C.A.H. Schuurbiers, M.C.M. van Loosdrecht
Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628BC, Delft, The Netherlands

ar t ic l e i n f o
Article history: Received 10 July 2006 Received in revised form 9 October 2006 Accepted 11 October 2006 Available online 28 November 2006 Keywords: Crystallisation Precipitation Phosphate Potassium Recovery Struvite Urine

A B S T R A C T

Phosphate can be recovered as struvite or apatite in uidised bed reactors. Urine has a much higher phosphate concentration than sludge reject water, allowing simpler (and less expensive) process for precipitation of phosphates. A stirred tank reactor with a special compartment for liquid solid separation was used to precipitate struvite from urine. Magnesium ammonium phosphate as well as potassium magnesium phosphate are two forms of struvite that were successfully precipitated. Liquid/solid separation was very effective, but the compaction of struvite was rather poor in the case of potassium struvite. Crystals did not form clusters and maintained the typical orthorhombic structure. Ammonium struvite had slightly lower efuent phosphate concentrations, but an average of 95% of inuent phosphate was removed regardless of ammonium or potassium struvite precipitation. Fluid mechanics is believed to be important and should inform further work. & 2006 Elsevier Ltd. All rights reserved.

1.

Introduction

Phosphate in wastewater has to be removed to prevent eutrophication of surface waters. Chemical phosphate removal is expensive and the produced inorganic solids complicate the sludge treatment, which further adds to costs (Paul et al., 2001). Metal phosphates, such as FePO4 , are generally considered to be unavailable as plant fertilizer. For other reasons too (e.g. heavy metal content) most wastewater sludge containing phosphate eventually ends in landlls. Since phosphate rock is a nite resource, phosphate should be recovered from liquid wastes in more sustainable systems. Nowadays, biological excess phosphate removal is well understood and has been implemented in advanced biological nutrient removal (BNR) plants in many countries. Organisms that can take up phosphate in excess of their nutrient requirement, release this excess phosphate in anaerobic
Corresponding author. Fax: +27 21 888 2682.

conditions, such as sludge digesters or the anaerobic compartments of advanced BNR plants. The phosphate concentration in sludge reject water can be quite high, i.e. 8595 g P=m3 (e.g. von Munch and Barr, 2001; Battistoni et al., 2002; Jaffer et al., 2002). This may lead to precipitation and scaling in pipes that increase operating and maintenance costs (Neethling and Benisch, 2004). Controlled precipitation, however, allows phosphate recovery. Phosphate can also be recovered directly via a phosphate pump from the anaerobic compartment of advanced BNR pants, although at lower concentrations than sludge reject water (Brandse et al., 2001). Fluidised bed reactors have been developed to crystallise Ca3(PO4)2, which could be a secondary ore in industrial phosphorus production (Eggers et al., 1995; Giesen, 1999). Boundary conditions include degassing, additional seeding material and ow control, which all add to the complexity and costs. Struvite, MgNH4 PO4 6H2 O, on the other hand, can

E-mail address: JWilsenach@csir.co.za (J.A. Wilsenach). 0043-1354/$ - see front matter & 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2006.10.014

ARTICLE IN PRESS
WAT E R R E S E A R C H

41 (20 07) 45 8 466

459

be used directly as a slow-release fertilizer and has a potentially higher market value (von Munch and Barr, 2001; Ueno and Fujii, 2001). Urine contains around 80% of the total nitrogen (N), 70% of the potassium (K) and up to 50% of the total phosphate (P) loads in municipal wastewater, but adds less than 1% of the volume (Larsen and Gujer, 1996). Modern no-mix toilets and waterless urinals have been developed to collect urine separately and largely undiluted (Larsen et al., 2001). The immediate benets of urine separation would be an increased capacity and better efuent quality of existing treatment plants, with a lower overall resource consumption (Wilsenach and van Loosdrecht, 2003). Source separated urine could be treated at a central plant, but this raises the problem of transport. One solution could be de-central treatment (e.g. in ofce blocks or hospitals), after which the treated liquid would be discharged via existing sewers. Fluidised bed reactors are relatively complicated processes even at centralised treatment plants (e.g. Abe, 1995; Ueno and Fujii, 2001; von Munch and Barr, 2001; Adnan et al., 2004) and would hardly be viable on a smaller scale. The economies of scale dictate that de-central processes should therefore be simpler in general and relatively maintenance free. The aim of this study was to design and investigate a low-tech system for struvite recovery from source separated urine. The issue of de-central or central treatment is left open for discussion. Simple and efcient low-tech processes are obviously benecial to all. The hydrolysis of urea releases ammonium and bicarbonate in urine, which increases the pH and determines the concentration of ions involved in equilibria of chemical speciation. The phosphate concentration PO3 therefore 4 increases in urealysed urine at the same total inorganic phosphate concentration of around 800 g P=m3 (Ciba Geigy, 1977). This leads to supersaturation of struvite, which has been found to precipitate naturally in urine collection systems with all the available Mg2 (Udert et al., 2003c, b). Biological nitrication of urine removes all the available bicarbonate (Udert et al., 2003a), but since the molar ratio of NH :K:P:Mg in urine is roughly 260:13:6:1 (e.g. Ciba Geigy, 4 1977; Grifth et al., 1976), ammonium in nitried urine would still be sufcient for struvite precipitation with alkalinity dosing. In the case of complete ammonium oxidation, or nitrogen removal, potassium struvite KMgPO4 6H2 O could be precipitated instead. A continuous stirred tank reactor has been recovering potassium struvite for more than ve years from animal manure in the Netherlands (Schuiling and Anrade, 1999). This reactor is however not optimised for good settling characteristics. The outow of struvite particles from such reactors to liquid/solids separation devices could lead to scaling in downstream conduits. We therefore designed and tested a lab-scale precipitator that incorporated a special compartment for liquid/solid separation. The effects of operating parameters (e.g. hydraulic retention time (HRT), mixing intensity and pH) on the performance of the precipitator are discussed. Differences between ammonium struvite and potassium struvite precipitation and recovery were also examined.

2.
2.1.

Materials and method


Synthetic urine mixture

Thermodynamics and kinetics of precipitation in synthetic urine do not differ from real urine (Ronteltap et al., 2003). Synthetic urine according to Grifth et al. (1976) was therefore used in all experiments. The phosphate concentration was increased by 50% above this recipe in order to be more representative of most recent studies. The composition of the urine mixture is shown in Table 1. Small amounts of urease were added to hydrolyse urea, which increased the pH to around 9.4. Some natural precipitation occurred in the inuent (with Mg and Ca in urine) leaving a P concentration of around 750 mg P/l.

2.2.

Batch tests

The use of different magnesium additives to recover MgNH4 PO4 6H2 O (MAP) as well as KMgPO4 6H2 O (KMP) were investigated in 250 ml stirred and unstirred asks, adding MgO and MgCl2 at different Mg:P ratios. KMP precipitation was investigated with a different synthetic urine solution, containing no urea and only a small amount of NH4 Cl, which is closer to the chemical composition after nitricationdenitrication of urine. The initial P concentration was around 460 mg P/l and NH N was around 40 mg N/l. The lower P 4 concentration simulates the conditions in biological treatment of urine, in which some dilution prevents microbial inhibition. In KMP batch experiments with MgCl2 , the pH was initially increased from 7.4 to 9.4 through NaOH addition. No extra base was added for experiments with MgO addition. Batches were run for 48 h.

2.3. Continuous stirred tank reactor (CSTR) for struvite precipitation and settling
Fig. 1 illustrates two alternatives for liquid/solid separation in the experimental CSTR. The initial set-up was used in experiments for continuous MAP precipitation, in which the treated liquid ows inwards after liquid/solid separation (Fig. 1a). A second set-up, believed to be an improvement, was

Table 1 Composition of synthetic urine Salt


CaCl2 2H2 O MgCl2 6H2 O NaCl Na2 SO4 Na3 citrate 2H2 O Na2 (COO)2 KH2 PO4 KCl NH4 Cl NH2 CONH2 (urea) C4 H7 N3 O (creatinine)

g/l
0.65 0.65 4.60 2.30 0.65 0.02 4.2 1.60 1.00 25.0 1.10

mM
4.4 3.2 78.7 16.2 2.6 0.15 30.9 21.5 18.7 417 9.7

ARTICLE IN PRESS
460
WA T E R R E S E A R C H

41 (2007) 458 466

a
Influent Mg2+ Effluent

b
Influent and Mg2+ Acid/Base Effluent

Struvite

Struvite

Fig. 1 Schematic drawing of struvite precipitator with cross sections through alternative liquid/solid separation devices. (a) Inward ow device, with efuent owing from reactor wall upwards in single efuent line. Used for MAP precipitation with inuent pH49 and Mg2 dosed in separate line. (b) Outward ow device, with efuent owing upwards towards the reactor wall and out through three efuent lines (only two shown). Used for KMP precipitation, with inuent pHo7 and Mg2 mixed with inuent.

used (Fig. 1b), in which liquid ows outwards after liquid/solid separation. The reactor was made of acrylic glass and had a height of 300 mm and a diameter of 100 mm. The volumes of the precipitation and settling sections were both 0.77 l. An axial ow impeller (propeller-type) was submerged halfway into the precipitation zone and connected to a variable speed motor. In the case of MAP precipitation, the inuent had a pH of 9.4 and MgCl2 (1 M) was added continuously with a separate dose pump. Inuent for KMP precipitation had a pH of around 6 and the required MgCl2 was mixed directly into the inuent. A pH probe was placed directly into the reactor just below the liquid surface. For experiments with KMP, the pH was monitored on-line and controlled by addition of 1.0 M NaOH or 0.1 M HCl solutions. Set points for minimum and maximum pH values dened a narrow band of 0.2 pH units. Samples were taken directly from the precipitation zone. All experiments were done at room temperature, i.e. 2324 C.

3.
3.1.

Results
Batch tests

The most important results from batch tests are shown in Fig. 2. In the case of MAP precipitation, an excess ammonium concentration and bicarbonate buffer ensured a near constant pH of 9.4. With addition of MgCl2 , the phosphate removal is linear with an increasing Mg:P ratio, and 99% removal was achieved with MgCl2 : P 1. Practically, the same result was obtained with MgO addition (not shown in Fig. 2). In the case of phosphate removal through KMP, results are somewhat different regarding the pH. With MgCl2 as magnesium source, the pH decreased with increasing phosphate removal, presumably according to K Mg2 HPO2 ! KMgPO4 H . 4 (1)

2.4.

X-ray diffraction (XRD) and analysis

Samples of both MAP and KMP precipitants were mixed with ethanol and ground in a mortar. A small amount was added to a substrate consisting of a silicon single crystal wafer. After evaporation of the ethanol, a thin layer of sample is obtained with a weight of about 10 mg. The XRD measurements were performed on a Bruker-AXS D5005 diffractometer equipped with an incident beam CuKa1 monochromator and a Braun position sensitive detector (PSD). The 2y range was 570 with a stepsize of 0:04 and a counting time per step of 1 s. All ammonium and phosphate concentrations were measured using standard commercial Dr. Lange spectrophotometer equipment.

The decrease in the pH to 8.2 resulted in a relatively low phosphate removal, even with an overdose of magnesium (P removal was only 75% with Mg:P 2). However, with further base addition to pH 9.1, the P removal efciency increased to 95% with Mg:P 1. With MgO as magnesium source, no additional base was required. Although the reaction is presumably the same as Eq. (1), the oxide neutralises acid and results in a slight net increase in pH. At Mg:P 1, the pH exceeded 9, which was sufcient for complete phosphate removal, similar to MAP precipitation. With Mg:P 1, the ammonium concentration decreased from 41 mgNH N/l 4 initially to 22 mg NH N/l at equilibrium. This indicates that 4 less than 1% of the phosphate was removed as MAP, the greatest percentage presumably being KMP. With further MgO additions (Mg:P 1.5 and 2) the pH increased further, but without further effect on the P removal efciency. Although Miles and Ellis (2001) reported lower phosphate removal

ARTICLE IN PRESS
WAT E R R E S E A R C H

41 (20 07) 45 8 466

461

efciencies (through struvite) at pH 10, due to the NH NH3 4 equilibrium moving towards NH3 , this is not important in untreated urine with excess ammonium concentration, and irrelevant regarding potassium struvite. Summing up, almost all phosphate could be removed at Mg:P 1, regardless of the magnesium source or whether ammonium or potassium struvite is precipitated. Conrming previous research, pH 9 or higher is crucial for complete struvite precipitation from urine (e.g. Ronteltap et al., 2003).

3.2. Precipitation efciency in continuous operation: MAP and KMP


In the baseline experiment, the struvite precipitator was lled with untreated urine and the inuent pumps for synthetic urine and dissolved MgCl2 were started. The resulting HRT was 2 h, based on the volume of the precipitation chamber. The Mg:P ratio was 4.2. The soluble P efuent concentration

KMP (MgCl2) MAP (MgCl2) pH (MgCl2) KMP (MgO) pH (MgO) P - removal efficiency (%) 100 75 50 25 0 0 0.5 1.0 Mg2+:Ptot
Fig. 2 Phosphate removal efciencies from batch experiments for potassium struvite precipitation (KMP) with different Mg2 additives and ammonium struvite (MAP) with constant pH 9:4 and MgCl2 .

10 9 8 7 pH

1.5

2.0

dropped to 18 mg P/l at steady state (precipitation efciency of 97.5%), which was reached within 3 h of continuous operation. This is in line with ndings from Ronteltap et al. (2003) who determined that struvite is more soluble in urine than in other liquids and that efuent concentrations of around 16 g P=m3 could be expected with an initial Mg:P ratio of 1. The XRD analysis showed that precipitant was predominantly struvite, with some chloride salts. Table 2 shows the most important results of all further experiments, which were all started with the efuent from previous experiments in order to reach steady state quicker. With a Mg:P ratio of 2.6, the efuent P concentration was quite low, i.e. 10 mg P/l. However, the decrease in removal efciency with a Mg:P ratio of 1.1 was almost insignicant compared to the inuent concentration. Whether the precipitator was stirred (mixed) or not, had virtually no effect on the soluble P efuent concentration (experiments 1 and 2). The precipitator was operated at different HRTs, i.e. 0.5, 0.9 and 2.0 h, for continuous periods of up to one day. No difference could be found in precipitation efciency for different HRTs. Experiment 5 was in fact a series of seven experiments where the effects of different mixing speedsfrom 50 to 600 rpmwere investigated. Mixing speed had practically no effect on the precipitation efciency. In experiments 810, small amounts of ammonium were added to the inuent, representative of conditions after nitricationdenitrication of urine. Remaining ammonium was partly removed with precipitation and nal efuent concentrations with an average 18 mg NH N were measured 4 (inuent ammonium was 52 mg NH N). This means that 4 with efuent phosphate concentrations around 38 mg P/l (inuent phosphate was 320 mg P/l for experiments 810), only 17% of phosphate was precipitated as MAP, with the rest presumably KMP. The XRD analysis conrmed that the precipitant was predominantly struvite. The MAP samples from the experiments 17 all had the same line pattern, which correlated perfectly with struvite. KMP samples all gave the same line pattern, which was slightly different from MAP samples, but still had a very strong correlation with struvite. The XRD analysis of KMP samples revealed no other

Table 2 Summary of operational parameters and results of struvite precipitation Exp. number Struvite type pH Mg:P (feed) HRT (h) Mixing speed (rpm)
100 0 100 100 50600 100 100 100 300 100

Efuent conc. (mg P/l)


9 11 25 24 33 2:0 26 25 35 35 62

Volume index (ml/g P)


44 5 500 310 370

Scaling on wall (% of inuent P)


(Fig. 3) 1.2 5.2

1 2 3 4 5 6 7a 8a 9a 10a
a

MAP MAP MAP MAP MAP MAP MAP KMP KMP KMP

9.4 9.4 9.4 9.5 9.4 9.4 9.4 9.0 9.0 8.7

2.6 2.6 1.6 1.1 1.1 1.1 1.3 1.3 1.3 1.3

2.0 2.0 0.5 2.0 0.57 0.92 1.55 1.63 1.63 1.56

Outward ow device (Fig. 1b).

ARTICLE IN PRESS
462
WA T E R R E S E A R C H

41 (2007) 458 466

crystal formations. Precipitant scraped off the impeller blades was also predominantly struvite.

3.3. 3.3.1.

Liquid/solid separation in continuous operation Scaling of struvite

Excessive scaling occurred in all MAP experiments. At least 50% of the total precipitant (by volume) had to be removed mechanically from the impeller blades, reactor wall and the outside of the internal efuent pipe. Fig. 3 shows the effect of mixing speed on the collection of MAP precipitant in experiment 5, with total duration of 135 min (i.e. 4 times the HRT). Higher mixing speeds in general led to a higher percentage of scaling on the reactor wall, etc. At best, 50% of precipitant collected directly in the settler (50 rpm) while at worst only around 30% collected directly in the settler (400, 500 and 600 rpm). By contrast, very little scaling occurred during experiments with KMP precipitation. Still, a thin layer of precipitant was observed on the reactor wall. After identical KMP experiments, with equal inuent ows and equal total running times (8 h), but different mixing speeds of 100 and 300 rpm (i.e. experiments 8 and 9, respectively), the reactor was ushed twice with tap water. The reactor was lled with distilled water and the pH was reduced to 5.8. The reactor was mixed for 24 h and then drained into a clean vessel where the phosphate concentration was measured. Based on observation of the acrylic glass reactor walls, lowering the pH was effective to completely dissolve this layer of precipitant from the reactor wall. The P concentration after lowering the pH in experiment 8 (100 rpm) was 16 mg P/l, while after experiment 9 (300 rpm) the P concentration was 74 mg P/l. This correlates to, respectively, 1.2% and 5.2% of the inuent load that precipitated on the reactor wall, indicating the effect of higher mixing speed on wall scaling.

present in the efuent. The total phosphate efuent concentration was around 51 mg P/l at 50300 rpm, and 78 mg P/l at 600 rpm. When these concentrations are compared to that in Table 2 (experiment 5), it seems that around 6% of inuent phosphate was present in particulate form in the efuent. At higher mixing speeds (500 and 600 rpm), transfer of mixing power from the precipitation chamber to the settler was evident and resulted in bigger losses of nes in the efuent. This was improved with the outward ow device (Fig. 4a). Precipitant can be seen in the precipitation compartment with a hazy liquid. The impeller and connecting shaft are barely visible through the liquid/solid mixture. A white powdery substance (struvite) can be seen in the compaction compartment, with the settling zone almost completely transparent. Settling particles, leaving the precipitation compartment, can be seen in the detail photograph (Fig. 4b). Practically, no precipitant was lost and no blockages occurred during any of the experiments with the outward ow device (experiments 710).

3.3.3.

Struvite volume index of MAP and KMP precipitant

3.3.2. Differences between precipitation and removal efciency


Although good precipitation efciencies were observed for all experiments involving MAP precipitation, nes were always

Volume precipitant settler/ total precipitant volume

0.6

0.4

0.2

The struvite volume index of the combined MAP precipitant (collected in settler plus the scaling removed) was determined for the different mixing speeds. This was determined by calculating the volume occupied by the total precipitant and then ltering and weighing the mass of precipitant. We assumed all precipitant to be struvite, based on the XRDanalysis. The average volume occupied by MAP after a load of 2.5 g P was 103 12 ml. The volume index for various mixing speeds was 44 ml/g P on average for the seven runs in experiment 5 (50, 100, 200, etc. up to 600 rpm), with small differences between runs. These are still only indicative gures for comparative purposes. The volume index of KMP precipitant was determined based on the volume occupied by the precipitant, and the load of P removed. The volume occupied by KMP, after a load of 1.4 g P was already around 450 ml. The volume index for KMP was between 310 and 500 ml/g P, for different operating parameters shown in Table 2, i.e. roughly ten times higher than the MAP volume index. In experiment 10, the pH was lowered to 8.7. It was believed that this would decrease the supersaturation to favour secondary crystal growth instead of primary nucleation, thereby increasing the crystal sizes and improving settling. The volume index was however not much improved, compared to that of MAP precipitant from experiment 5. Although practically no precipitant was lost during experiments 710, removal of precipitant from the compaction zone was problematic. When precipitant was released and ow became turbulent, the precipitant was re-suspended and mixed uniformly through the settling zone.

200 400 600 Mixing speed (rpm)

4.
4.1.

Discussion
Precipitation efciency

Fig. 3 Ratio of MAP precipitant volume in the settler to the total MAP precipitant volume after mechanical removal from impeller, internal efuent pipe and reactor walls, for different mixing speeds (refer Fig. 1a for liquid/solid separation device).

Struvite precipitation is in general a very efcient way of phosphate removal from urine. Even when stoichiometric and kinetic parameters were critical at the same time (i.e. Mg:P 1.1 and HRT 0.6 h), the efuent P concentration

ARTICLE IN PRESS
WAT E R R E S E A R C H

41 (20 07) 45 8 466

463

Fig. 4 Struvite precipitator: (a) general view of experimental set-up during precipitation, settling and compaction, with outward ow device for liquid/solid separation; (b) detail showing a plume of KMP crystals being discharged from the precipitation chamber into the settling zone; (c) detail showing the inward ow device for liquid/solid separation, with scaling on reactor wall and axial ow impeller clearly visible (MAP).

was still low, and only slightly higher than with oversupply of magnesium, or with longer contact time (compare experiment 5 to experiments 3 and 4). However, this is of academic interest only, because the precipitation efciency was still 96%. Potassium struvite precipitation KMgPO4 6H2 O was shown to be almost as efcient in phosphate removal as the more familiar MgNH4 PO4 6H2 O. Although the efuent P concentration was a little higher for KMP (38 vs. 25 mg P/l), this is still not signicant compared to the inuent concentration. Batch tests have also shown that complete KMP removal (99%) is possible. Although not investigated in this study, precipitation kinetics could play a more important role with potassium struvite. The practical insignicance of HRT on precipitation efciency, found in this study, is similar to ndings of others. Battistoni et al. (2002) used an empirical double saturation model to show that beyond a certain minimum contact time (around 30 min) only pH plays a role in nucleation efciency. In untreated and hydrolysed urine, however, the pH is already between 9.4 and 9.5. This means that magnesium addition would always lead to struvite precipitation in untreated urine, where the excess ammonium concentration apparently drives struvite precipitation to ensure high removal efciencies (Stratful et al., 2001). In untreated urine, the NH :P ratio is 43, while the K:P ratio is 4 only 2. The work by Ronteltap et al. (2002) suggests that the thermodynamics and kinetics of potassium struvite precipitation in urine would be different, not only from potassium struvite precipitation in animal manure, but also from ammonium struvite in urine. The different ion activity products in different liquids would then explain the slightly

better precipitation efciency of MAP relative to KMP. A more fundamental and quantitative understanding of potassium struvite precipitation in urine is still lacking.

4.2.

Crystallisation and recovery efciency

Crystallisation efciency is just as important as good precipitation efciency. Crystal retention time is therefore an important parameter. This was pointed out by Battistoni et al. (2002), who also clearly differentiated between precipitation and nucleation, i.e. secondary crystal growth sufciently large to prevent outow of nes (precipitant) from a uidised bed reactor. In this study, increasing the mixing speed to maintain more solids in suspension for a larger effective precipitation area only resulted in transfer of mixing power to the settling zone of the inward ow (Fig. 1a). Since the opening between the two compartments was at the perimeter, the liquid shear stress would have been highest there, i.e. at the outside of the vortex. This problem was eliminated by the outward ow device with the opening at the centre (Fig. 1b). Although the transfer of mixing power to the sedimentation zone was eliminated, an increase in mixing speed still only led to more precipitation on walls. Struvite has a specic density of about 1.7 and would therefore be transported outwards under inuence of centrifugal forces. Fig. 5a shows the typical needle-like orthorhombic structure of struvite from a MAP sample produced in experiment 7 (outward ow device). Increasing the crystal size through lower supersaturation (i.e. lower pH) was only partly successful. Fig. 5c shows a rare example of a crystal formed at pH 8.7,

ARTICLE IN PRESS
464
WA T E R R E S E A R C H

41 (2007) 458 466

Fig. 5 Microscope images of struvite crystals from precipitator with outward ow device: (a) experiment 7; typical needlelike MAP crystals; (b) experiment 9; typical KMP crystals, pH 9.0; (c) experiment 10; example of one big KMP crystal, pH 8.7; (d) experiment 8; general view of KMP crystals, pH 9.0 and (e) experiment 10; general view of KMP crystals, pH 8.7. Images (a)(c) are 400 times magnied. Images (d) and (e) are 50 times magnied. Samples were dried and salt can be seen in all images.

which is substantially bigger than the crystals shown in Fig. 5b (formed at pH 9.0, experiment 9). The comparison of the average thickness of crystals in Fig. 5d (KMP, pH 9.0, experiment 8) and Fig. 5e (KMP, pH 8.7) is more representative and shows that limiting the supersaturation could play some role in struvite crystal growth. Crystals formed at pH 8.7 are not much longer but noticeably thicker. No obvious differences were observed between the sizes and shapes of crystals produced in experiment 9, at 300 rpm, and experiment 10. Better mixing and the limiting of supersaturation might therefore be equally important in order to increase crystal size. Even though the crystals settled well (refer Fig. 4a and b) the compaction of KMP crystals, which still had a very high struvite volume index, could not be improved in these experiments. These gures are believed to be representative of the experiments, but they still give mostly qualitative information. The crystal in Fig. 5c has dimensions very similar to that of von Munch and Barr (2001). The shorter and more bulky crystals would lead to better compaction compared to the longer and thinner crystals. Crystals produced by von Munch and Barr (2001) were all still individual struvite crystals (i.e. no clusters like Ueno and Fujii, 2001; Adnan et al., 2004), with a median size of 110 mm and length-to-thickness ratios of 34. The images in Fig. 5 show crystals with length-to-thickness ratios of 10 or more. The large difference between the volume index of experiment 5 (44 ml/g P) and that of experiments 810 (around 400 ml/g P) is also illustrated by comparison of Fig. 4a with Fig. 4c. Whereas the impeller and shaft are barely visible in Fig. 4a due to crystals in suspension, almost no crystals in suspension can be seen in Fig. 4c. Since no ow bafes were installed in the precipitation compartment, the vertical ow component of mixing is believed to have been small in comparison to the rotational movement of the liquid, which approached

that of solid body rotation. Although the crystals were kept in suspension, there could have been little movement of particles relative to each other. The effectiveness of mixing could therefore have been poor and an increase of mixing speed could have had less effect to improve the mixing. The main difference between the precipitation chambers in the two alternative congurations is the presence of the efuent pipe in the inward ow device. We speculate that better mixing, caused by eddies around the efuent pipe, would increase the number of particle collisions. Scaling around the efuent pipe can also be seen in Fig. 4c. A further difference in physical operation between the inward and outward ow devices was the dosing of magnesium, as illustrated in Fig. 1. In the outward ow device, magnesium was already mixed with the inuent, which had a relatively low pH. This situation would evidently have resulted in localised areas with a high degree of struvite supersaturation in the precipitation chamber, especially with poor mixing, which could have favoured primary nucleation. The struvite volume index was not measured in experiment 7 (MAP in outward ow device). Fig. 5a and b give no evidence to suggest that these MAP crystals have different particle shapes and sizes than KMP. This would suggest that differences in the struvite volume index is rather due to mechanical differences discussed above, than differences between MAP and KMP.

4.3.

Future application and further research

The potassium struvite produced at the calf manure treatment plant in Putten, The Netherlands is still separated as slurry, rather than solids (Verhoek, 2005). Struvite crystals were also shown to cluster together, forming composite crystals with typical diameters of 2025 mm (Schuiling and Anrade, 1999). These crystals are produced from two CSTRs in series without any bafes and have a typical volume index of

ARTICLE IN PRESS
WAT E R R E S E A R C H

41 (20 07) 45 8 466

465

6070 ml/g P (based on KMP) after thickening in a storage tank (Verhoek, 2005). One obvious difference between the potassium struvite from calf manure and urine (this study) is the presence of many nes in the calf manure inuent, such as pieces of animal hair and organic matter, which could possibly improve crystal growth. Virtually all precipitation takes place in the rst of this series of reactors, but the downstream reactor and storage tank could improve the coagulation or clustering of precipitant. Crystals could also grow further through ageing. Recycling of precipitated crystals to the precipitation zone was not investigated in this study. This would possibly improve secondary crystal growth. Mechanical abrasion would eventually break down long thin crystals, promoting growth of thicker crystals. Addition of a MgO:P ratio of around 0.30.4 to biologically treated urine, would not only leave more than 50% of phosphate in treated urine in solution, but would also increase the pH only to around 8.5. Although this was not investigated, it could prevent supersaturation, which would lead to better secondary growth. A 50:50 mixture of MgO and MgCl2 could be the ideal magnesium additive for recovering phosphate for treated urine. This would have to be investigated in more detail. This study indicates that downstream dewatering is still required after simple struvite precipitation. Good mixing seems to be crucial to ensure secondary crystal growth instead of primary nucleation. Further research should focus on quantifying the effects of different mixing scenarios on secondary crystal growth. This should include the issues of bafe size and distribution, impeller type, reactor dimensions and the possibility of partial crystal recirculation. The operational problem of scaling could be turned into an advantage, if for instance a plastic lm could be devised to cover bafes, which could then easily be removed. If say 90% of the ammonium in urine is removed (or converted to nitrite/nitrate) in a biological step, the remaining ammonium could be removed as struvite. Even if too little ammonium remains for complete phosphate removal through MAP precipitation, potassium is sufcient for complete phosphate recovery through KMP precipitation. At this stage, implementation of struvite recovery on a de-central scale seems to be inopportune. In a transitional period, the mixing of some urine with supernatant would be the most logical step in improving phosphate recovery.

of mixing and suspension of particles should be increased, while the mixing speed should be decreased to limit centrifugal force and wall scaling. Localised supersaturation could be prevented by diffuse dosing of magnesium, e.g. in a fractal manifold, rather than point dosing. An approach using computational uid dynamics would benet further studies.

Acknowledgements
This study was nancially supported by the STOWA (Dutch Foundation for Applied Water Research). The XRD analysis was done by Niek van der Pers, Material Sciences (TU Delft).
R E F E R E N C E S

5.

Conclusions

Struvite precipitation from urine (or urine mixed with sludge reject water) should be further developed as a process downstream of biological N-removal. This allows recovery of some potassium and at the same time act as a polishing process for improved ammonium removal. Addition of MgO provides sufcient alkalinity for struvite precipitation in nitried urine. The precipitation of struvite is less of a problem than engineering the fate of the precipitant itself. The particle size of precipitant appears to be increased by the separate and combined effects of limiting supersaturation and good mixing. These should be examined further. The effectiveness

Abe, S., 1995. Phosphate removal from dewatering ltrate by MAP process at Seibu treatment plant in Fukuoka City. Sewage Works in Japan, pp. 5964. Adnan, A., Dastur, M., Mavinic, D.S., Koch, F.A., 2004. Preliminary investigation into factors affecting controlled struvite crystallization at the bench scale. J. Environ. Eng. Sci. 3, 195202. Battistoni, P., de Angelis, A., Prisciandaro, M., Boccadoro, R., Bolzonella, D., 2002. P removal from anaerobic supernatants by struvite crystallization: long term validation and process modelling. Water Res. 36 (8), 19271938. Brandse, F., van Loosdrecht, M.C.M., 2001. Sewage treatment and methods for phosphate recovery in BCFS process. In: Proceedings from Second International Conference on Phosphate Recovery for Recycling from Sewage and Animal Wastes. Noordwijkerhout, The Netherlands. Ciba Geigy, J., 1977. Wissenschaftliche Tabellen Geigy, eighth ed. Korperussigkeiten, Basel. Eggers, E., Dirkzwager, A.H., van der Honing, H., 1995. Full-scale experiences with phosphate crystallisation in a CRYSTALACTOR. Water Sci. Technol. 23 (46), 819824. Giesen, A., 1999. Crystallisation process enables environmental friendly phosphate removal at low costs. Environ. Technol. 20 (7), 769775. Grifth, D.P., Musher, D.M., Itin, C., 1976. The primary cause of infection induced urinary stones. Invest. Urol. 13 (5), 347350. Jaffer, Y., Clark, T.A., Pearce, P., Parsons, S.A., 2002. Potential phosphorus recovery by struvite formation. Water Res. 36, 18341842. Larsen, T.A., Gujer, W., 1996. Separate management of anthropogenic nutrient solutions (human urine). Water Sci. Technol. 34 (34), 8794. Larsen, T.A., Peters, I., Alder, A., Eggen, R., Maurer, M., Muncke, J., 2001. Re-engineering the toilet for sustainable wastewater management. Environ. Sci. Technol. 35 (9), 192A197A. Miles, A., Ellis, T.G., 2001. Struvite precipitation potential for nutrient recovery from anaerobically teated wastes. Water Sci. Technol. 43 (11), 259266. Neethling, J.B., Benisch, M., 2004. Struvite control through process and facility design as well as operation strategy. Water Sci. Technol. 49 (2), 191199. Paul, E., Laval, M.L., Sperandio, M., 2001. Excess sludge production and costs due to phosphorus removal. Environ. Technol. 22 (11), 13631371. Ronteltap, M., Biebow, M., Maurer, M., Gujer, W., 2003. Thermodynamics of struvite precipitation in source separated urine. In: Proceedings from the Second International Symposium on Ecological Sanitation, Lubeck, Germany. Schuiling, R.D., Anrade, A., 1999. Recovery of struvite from calf manure. Environ. Technol. 20 (7), 765768.

ARTICLE IN PRESS
466
WA T E R R E S E A R C H

41 (2007) 458 466

Stratful, I., Scrimshaw, M.D., Lester, J.N., 2001. Conditions inuencing the precipitation of magnesium ammonium phosphate. Water Res. 35 (17), 41914199. Udert, K.M., Fux, C., Munster, M., Larsen, T.A., Siegrist, H., Gujer, W., 2003a. Nitrication and autotrophic denitrication of source separated urine. Water Sci. Technol. 48 (1), 119130. Udert, K.M., Larsen, T.A., Biebow, M., Gujer, W., 2003b. Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Res. 37 (11), 25712582. Udert, K.M., Larsen, T.A., Gujer, W., 2003c. Estimating the precipitation potential in urine-collecting systems. Water Res. 37, 26672766.

Ueno, Y., Fujii, M., 2001. Three years experience of operating and selling recovered struvite from full-scale plant. Environ. Technol. 22 (11), 13731381. Verhoek, J.A., 2005. Personal communication with technologist. Mestverwerking Gelderland, Jansbuitensingel 15, Arnhem. von Munch, E., Barr, K., 2001. Controlled struvite crystallisation for removing phosphorus from anaerobic digester sidestreams. Water Res. 35 (1), 151159. Wilsenach, J., van Loosdrecht, M., 2003. Impact of separate urine collection on wastewater treatment systems. Water Sci. Technol. 48 (1), 103110.

Das könnte Ihnen auch gefallen