Struvite precipitation from urine with electrochemical magnesium dosage

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

Available online at www.sciencedirect.com

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

Struvite precipitation from urine with electrochemical magnesium dosage Alexandra Hug, Kai M. Udert* Eawag, Swiss Federal Institute of Aquatic Science and Technology, U¨berlandstrasse 133, 8600 Du¨bendorf, Switzerland

article info

abstract

Article history:

When magnesium is added to source-separated urine, struvite (MgNH4PO4$6H2O) precipi-

Received 11 June 2012

tates and phosphorus can be recovered. Up to now, magnesium salts have been used as the

Received in revised form

main source of magnesium. Struvite precipitation with these salts works well but is

17 September 2012

challenging in decentralized reactors, where high automation of the dosage and small

Accepted 18 September 2012

reactor sizes are required. In this study, we investigated a novel approach for magnesium

Available online 27 September 2012

dosage: magnesium was electrochemically dissolved from a sacrificial magnesium electrode. We demonstrated that this process is technically simple and economically feasible

Keywords:

and thus interesting for decentralized reactors. Linear voltammetry and batch experiments

NoMix technology

at different anode potentials revealed that the anode potential must be higher than
0.9 V

Magnesium ammonium phosphate

vs. NHE (normal hydrogen electrode) to overcome the strong passivation of the anode. An

(MAP)

anode potential of
0.6 V vs. NHE seemed to be suitable for active magnesium dissolution.

Electrochemistry

For 13 subsequent cycles at this potential, we achieved an average phosphate removal rate

Magnesium electrode

of 3.7 mg P cm
2 h
1, a current density of 5.5 mA cm
2 and a current efficiency of 118%.

Decentralized treatment

Some magnesium carbonate (nesquehonite) accumulated on the anode surface; as

Nesquehonite

a consequence, the current density decreased slightly, but the current efficiency was not affected. The energy consumption for these experiments was 1.7 W h g P
1. A cost comparison showed that sacrificial magnesium electrodes are competitive with easily soluble magnesium salts such as MgCl2 and MgSO4, but are more expensive than dosing with MgO. Energy costs for the electrochemical process were insignificant. Dosing magnesium electrochemically could thus be a worthwhile alternative to dosing magnesium salts. Due to the simple reactor and handling of magnesium, this may well be a particularly interesting approach for decentralized urine treatment. ª 2012 Elsevier Ltd. All rights reserved.

1.

Introduction

The separate collection and treatment of urine e also called NoMix technology e is a novel approach designed to improve the efficiency of nutrient removal (Wilsenach and van Loosdrecht, 2004) or its recovery (Maurer et al., 2003) in urban wastewater management. Previous studies have shown that this technology is a good alternative to conventional wastewater treatment, but the transport of the separated

urine can be highly demanding of resources. Decentralized or small on-site reactors for urine treatment would consequently be more suitable for implementing urine separation than large centralized treatment plants (Larsen et al., 2009). Phosphate recovery is a preeminent goal of urine treatment, because the known sources of mineral phosphorus are being depleted at a rapid rate (Cordell et al., 2009). Phosphate removal from urine is a technically simple process. During urine storage, urea is microbially degraded resulting in

* Corresponding author. Tel.: þ41 58 765 5360; fax: þ41 58 765 5389. E-mail addresses: [email protected] (A. Hug), [email protected] (K.M. Udert). 0043-1354/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2012.09.036

290

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

a strong pH increase to values of around 9. At elevated pH values, struvite (MgNH4PO4$6H2O) and calcium phosphate (mainly octacalcium phosphate, Ca8H2(PO4)6$5H2O) precipitate spontaneously (Udert et al., 2003a). During storage, at least 28% of the phosphate precipitates until all magnesium and calcium in the urine is consumed (Udert et al., 2003b). By adding additional magnesium ions, for instance as MgCl2, MgO, MgSO4 or bittern, the remaining phosphate precipitates as struvite (Etter et al., 2011). Phosphate can also be removed from urine by adding iron or aluminium ions, which are the usual precipitants for chemical phosphate removal in conventional wastewater treatment. Instead of adding salts, iron or aluminium can also be dosed electrochemically by dissolving a sacrificial electrode (Chen, 2004). This approach is especially interesting for small decentralized reactors, because the dosage can be easily controlled via the current, and the volume of the precipitant is minimal. Electrochemical dissolution of iron or aluminium has mainly been used to supply cations for the coagulation of particulate substances, for example in drinking water treatment (Holt et al., 2005). Some researchers have recently tested its application for removing phosphate from urine. Ikematsu et al. (2006) described the electrochemical precipitation of iron phosphate from synthetic fresh urine in a reactor, which first oxidized urea and organic substances at a dimensionally stable anode (DSA) and, after changing the polarization, precipitated phosphate by dissolving the iron electrode. The authors worked with urine samples at different dilutions. In experiments with fresh and stored urine, Zheng et al. (2009a) achieved nearly 100% phosphate removal by electrochemically dissolving iron at a ratio of 1.7e2.5 g Fe g P
1 (which corresponds to a molar ratio of 0.94e1.4 mol Fe mol P
1). Phosphate removal was slower in stored urine than fresh urine, which can be explained by slower precipitation kinetics at higher pH values. In experiments with synthetic stored urine at different pH values, Zheng et al. (2009b, 2010) achieved the highest phosphate removal rates at the lower end of the pH range (pH 5 and 6), irrespective of whether iron or aluminium was used as the precipitant. To our knowledge, the use of magnesium electrodes for phosphate precipitation has not yet been described in the scientific literature, although a patent exists for electrochemical magnesium dissolution for struvite production from manure (Egner and Brynioc, 2007). However, struvite precipitation can have some important advantages over iron phosphate or aluminium phosphate precipitation: First, the conditions in stored urine are ideal for fast and complete precipitation of phosphate, second, agricultural studies have shown that struvite has a significantly higher phosphate availability than iron or aluminium phosphates (Ro¨mer, 2006). Struvite is therefore a favorable product for recycling phosphate directly to agriculture. Magnesium electrodes may not have been used for phosphate precipitation so far, but they are widely used for cathodic protection to prevent corrosion of more noble metals or alloys such as steel (Andrei et al., 2003). Magnesium has very low resistance to corrosion or oxidation. Its use as a sacrificial anode is therefore limited to applications in environments with high resistivity such as pipelines in soils,

fresh-water storage tanks (including boilers) or ships in fresh water (Gurrappa, 2005). Besides very fast anodic dissolution, non-galvanic corrosion is another phenomenon that limits the applicability of magnesium as a sacrificial anode. However, both effects increase the efficiency of magnesium for electrochemical phosphate removal: fast dissolution increases the rate of precipitation and non-galvanic corrosion reduces the energy demand for magnesium release. The hypothesis of this study was that electrochemical magnesium dissolution is a technically feasible and economically viable process for phosphate removal from stored urine. To test this hypothesis, we conducted wellcontrolled electrochemical batch experiments at constant anode potentials and compared the costs of electrochemical magnesium dissolution with those for magnesium salts.

2.

Materials and methods

2.1.

Urine composition

For all experiments, we used real source-separated urine from the men’s collection tank of Eawag’s main building Forum Chriesbach (Table 1). In this solution, urea was completely degraded and the magnesium and calcium concentrations were very low, because struvite and calcium phosphate had already precipitated spontaneously in the urine collecting pipes and the storage tank (Udert et al., 2003a). The low ammonia concentration was due to volatilization (Goosse et al., 2009).

2.2.

Reactor setup

All experiments were conducted in a single-compartment reactor with a maximum liquid volume of 1 L. A schematic illustration of the experimental system is given in Fig. 1. The anode was a magnesium plate (type MgAl3Zn1m%; EN_1754:1997, 1997) and the cathode a steel plate (X5CrNi18-10; EN_10088-1, 1995). Both electrodes were flat and their rear sides were covered. The size of the exposed electrode surface

Table 1 e Average composition of the stored urine used in the experiments. The average and standard deviations were calculated from measurements of nine different samples. The urine was taken from the men’s collection tank in Eawag’s main building, Forum Chriesbach. Stored urine pH Conductivity COD TIC Mg PO4-P NH4-N Cl SO4 Na K Ca

[e] [mS cm
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1] [mg L
1]

8.9 25.0 4500 1250 1.6 197 2540 3060 721 1990 1980 16.5

           

0.1 1.2 910 120 0.5 16 330 140 25 230 450 3.6

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

Potentiostat pH Conductivity Temperature

Ag/AgCl

Mg

H2

Steel

e-

e-

291

batch experiment. In a sequencing batch experiment, the anode potential was set at
0.6 V vs. NHE and the same electrode was used for 13 consecutive cycles. Each cycle lasted for 2 h. We chose the same experimental duration for each cycle (2 h) to be able to compare the performances directly. At the beginning of each cycle, the reactor was filled with new urine solution. The cycles were run during the day with no break, but in the evening the reactor was emptied and the electrodes were dried in the ambient air during the night. This operation was chosen because it represents a hypothetically practical application.

H+ Mg

2.5.

2+

Mg 2+ NH 4

+

PO 43 -

Struvite

Fig. 1 e Schematic illustration of the experimental setup. Magnesium was electrochemically dissolved and precipitated as struvite, thus removing nearly all the phosphate and a small part of the ammonium in urine. Hydrogen is the likely product at the steel cathode.

depended on the type of experiment. The distance between the electrodes was 5.5 cm. In all experiments, the anode potential was controlled with a potentiostat (PGU 10V-1A-IMP-S, Ingenieurbu¨ro Peter Schrems, Mu¨nster, Germany). The anode potential was measured against an Ag/AgCl electrode (product number 6.0724.140, Metrohm, Zofingen, Switzerland). The reference electrode was set close to the anode surface by using a Luggin-Haber capillary. The reactor liquid could be stirred magnetically (RCT basic, Ika Labortechnik, Staufen, Germany). The temperature was not controlled in any experiment and had an average value and standard deviation of 22.7  1.3  C.

2.3.

Voltammetry

Linear sweep voltammetric measurements were conducted on the magnesium electrode over a range between
1.8 V and þ1 V vs. NHE. The scan rate was chosen as 0.5 mV s
1. The electrode had a free surface area of 1.25 cm2 (0.5 cm  2.5 cm). To ensure identical starting conditions, we used a new magnesium plate in each experiment. The urine volume in the reactor was 0.3 L and the solution was not stirred.

2.4.

Electrolysis experiments

Phosphate removal rates at different anode potentials were determined with electrolysis experiments. Each electrode had a free surface area of 19.4 cm2 (2.2 cm  8.8 cm). The reactor was filled with 1 L of urine and stirred at approximately 275 rpm. Single batch experiments were conducted with fixed anode potentials (
1.2 V,
0.8 V,
0.6 V and
0.2 V vs. NHE) as well as with a single magnesium electrode without any electrical connection. A new magnesium electrode was used for every

Sampling

We took samples for a complete urine analysis according to Table 1 before and after each experiment. During the experiments, single samples were taken at constant time intervals to analyze the phosphate and magnesium. The samples were filtered (0.45 mm, MachereyeNagel, Du¨ren, Germany), diluted at least five times and immediately acidified with nitric acid to prevent further precipitation.

2.6.

Analysis of precipitates

After each batch experiment, some precipitated solids were recovered by filtration (0.45 mm, MachereyeNagel, Du¨ren, Germany) and dissolved in hydrochloric acid to determine their composition by wet chemical analysis. The samples were analyzed for ammonium, magnesium, potassium, sodium, calcium and phosphate. The precipitate of one experiment and the anode surface films of two experiments were analyzed by X-ray diffraction (XRD). This analysis was conducted at the Swiss Federal Laboratories for Material Science and Technology (EMPA) in Du¨bendorf, Switzerland, with an X’Pert PRO diffractometer (PANalytical B.V., Almelo, Netherlands) using Cu Ka radiation (45 kV and 40 mA). Data evaluation was performed with HighScorePlus 2.2 provided by PANalytical. The electrode surfaces were also investigated with a scanning electron microscope (SEM) (NovaNanoSEM 230, FEI, Hillsboro, USA) at EMPA.

2.7.

Measurements

Chloride and sulfate were analyzed by ion chromatography (IC, Metrohm, Herisau, Switzerland), while the cations (magnesium, potassium, calcium and sodium) were determined by inductively coupled plasma optical emission spectrometry (ICP OES, Ciros, Spectro Analytical Instruments, Kleve, Germany). Ammonium and phosphate were measured photometrically with a flow injection analyzer (FOSS, Hillerød, Denmark). Additionally, we measured the ammonium, phosphate and chemical oxygen demand (COD) with cuvette tests (LCK 303, LCK 350, LCK 614, Hach-Lange, Berlin, Germany). The standard deviation for all measurements was below 3%. The pH, temperature and electrical conductivity were measured continuously with electrodes (LF 96, pH 196, WTW, Weilheim, Germany) and were recorded at 5-s intervals on a data logger (Memograph M, RSG40, Endress þ Hauser, Reinach, Switzerland). The cell voltage was also recorded on the data logger, while the current was measured and recorded with the

292

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

potentiostat. In the experiment without a counter electrode, the anode potential was measured directly on the data logger by using a high impedance converter.

2.8.

Calculations

2.8.1.

Electrode potentials

All electrode potentials were measured against an Ag/AgCl reference electrode and were later converted to potentials against normal hydrogen electrode (NHE). For the conversion, 0.21 V was added to the measured potentials vs. Ag/AgCl.

2.8.2.

mMg;dissolved $100% mMg;current

(1)

CE current efficiency [%]; mMg, dissolved mass of magnesium dissolved [g]; mMg, current magnesium dissolved according to the current [g]. The dissolved mass of magnesium (mMg, dissolved) was calculated as the sum of the removed phosphate (assuming that 1 mol of magnesium precipitates with 1 mol of phosphate) and the accumulated dissolved magnesium MMg þ mMg ðtÞ
mMg ðt0 Þ mMg;dissolved ðtÞ ¼ ðmP ðt0 Þ
mP ðtÞÞ $ MP

(2)

mP(t) measured mass of phosphate at time t in the reactor [g]; mMg(t) measured mass of magnesium at time t in the reactor [g]; MMg molar mass of magnesium 24.3 [g mol
1]; MP molar mass of phosphorus 31.0 [g mol
1]. The expected magnesium dissolution based on the measured current (mMg, current) was calculated on the basis of Faraday’s law of electrolysis assuming that two electrons are transferred for the dissolution of one magnesium ion mMg;current ðtÞ ¼

MMg Q $ z F

(3)

z number of electrons transferred in the reaction, z ¼ 2; F Faraday constant 96,485 [C mol
1]; Q electric charge [C], Z t calculated as Q ¼ I ds; I electric current [A] 0

2.8.3.

Mg:P dosage ratio

The molar Mg:P dosage ratio was calculated as the amount of magnesium dissolved during the experiment divided by the initial mass of phosphate in urine. Dosage ratio ¼

2.8.4.

mMg;dissolved ðtÞ MP $ mP ðt0 Þ MMg

SI ¼

log10 ðIAPÞ   log10 Ksp

(4)

Computer simulation of mineral saturation

We used the PHREEQC Interactive software (Version 2.15.0; Parkhurst and Appelo, 1999) to calculate the saturation of struvite and nesquehonite. The basic database wateq4f.dat was extended with terms for acetate (data taken from the

(5)

IAP ion activity product; Ksp solubility product of the respective mineral

2.8.5.

Current efficiency

The current efficiency (CE) was calculated by dividing the effectively dissolved magnesium (mMg, dissolved) by the value expected from the measured current (mMg, current). An efficiency of 100% indicates that all the current is used to dissolve the magnesium. CE ¼

minteq.dat database) and struvite (solubility constant pK ¼ 13.15; Taylor et al., 1963). Acetate was used as a model substance for chemical oxygen demand (COD). We used the concentrations given in Table 1 as the initial conditions and fitted the ion balance by adjusting the acetate concentration. The saturation index SI is given by

Cost estimate

We compared the cost of a magnesium anode with that of magnesium oxide (MgO), magnesium chloride (MgCl2 anhydrous) and magnesium sulfate (MgSO4 anhydrous). For all magnesium salts, we used prices for technical grade chemicals. For the magnesium salts, we only considered the material costs, while for the magnesium electrode we also included the costs for the electric power. Energy costs for other reactor parts such as dosage pumps and stirrers, or installation and maintenance costs, were not considered.

3.

Results

3.1.

Voltammetry

We used voltammetric measurements to identify suitable magnesium potentials for dissolving magnesium in sourceseparated urine. From the starting potential (
1.8 V vs. NHE) up to a potential of around
1.6 V vs. NHE (Fig. 2A), the current density was negative, which means that the magnesium electrode acted as a cathode and received electrons from oxidation processes at the steel electrode. Hydrogen was the most probable product at the magnesium electrode (Song, 2005). At a potential of
1.6 V vs. NHE, the current changed direction: the magnesium electrode became the anode and the steel electrode the cathode. The potential with zero electron transfer is commonly called the corrosion potential (Ecorr) (Song et al., 1997a). Between
1.5 and
0.9 V vs. NHE the current hardly increased at all, a phenomenon generally attributed to the formation of a passive hydroxide film (Makar and Kruger, 1990). This passivation range ended at a potential of
0.9 V vs. NHE, when the current increased strongly. As the activation of the magnesium electrode is accompanied by the initiation of irregular corrosion pits, this potential is commonly referred to as the pitting potential (Ept) (Song et al., 1997a). At a scan rate of 0.5 mV s
1 (Fig. 2A), the current reached a peak soon after passing Ept. Later, the current dropped slightly and approached a constant value. We refer to this potential region as the secondary passivation (Gulbrandsen, 1992).

3.2.

Batch experiments

On the basis of the voltammetric measurements (Fig. 2A), we chose five anode potentials to investigate struvite precipitation in different regions of magnesium electrode dissolution:

293

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

Fig. 2 e (A) Linear voltammetry in stored urine (upward scan, scan rate: 0.5 mV sL1) and the anode potentials in the batch experiments, and (B) phosphate concentrations during the batch tests at different anode potentials. The phosphate removal increased with higher anode potentials.

first, at the corrosion potential (Ecorr), second, in the first passivation region between Ecorr and Ept, third, at the potential with maximum current density, and fourth and fifth, at two potentials in the secondary passivation region. For the experiment at Ecorr, the magnesium electrode was not connected to the steel electrode. Under these conditions, the magnesium corrosion was already high: 75% of the phosphate was removed within 8 h (Fig. 2B). The magnesium potential increased from a starting value of below
1.5 V to
1.4 V vs. NHE at the end of the experiment. In the experiments with applied anode potentials (
1.2 V,
0.8 V,
0.6 V and
0.2 V vs. NHE), the current density and the phosphate removal rate increased with the anode potential (Table 2). For the anode potentials
0.8 V,
0.6 V and
0.2 V vs. NHE, the magnesium dissolution rate was nearly constant for the duration of the experiment, whereas the dissolution rates for the lowest potentials, Ecorr and
1.2 V vs. NHE, decreased with time. Since the application of higher anode potentials requires higher cell voltages, the energy needed to remove a certain amount of phosphorus also increased with the anode potential. In all experiments with applied anode potentials, the current efficiency (calculated on the basis of the release of Mg2þ) was above 100%. During the batch experiments, a film built up on the anodes. The film was thin in all experiments, but its color depended on the anode potential: it was white for the experiments with Ecorr and
1.2 V vs. NHE, but dark grey for higher anode potentials. In Fig. 3, the phosphate removal (as a percentage of the initial phosphate) is shown as a function of the Mg:P dosage

(data in Table S1, supplementary information). The phosphate removal increased linearly with the Mg:P ratio up to a ratio of 1.1 mol mol
1 but did not depend on the applied anode potential. The observed phosphate removal was lower than the simulated values, assuming thermodynamic equilibrium with struvite. However, in a sample which was removed from the reactor and left to react for three days, the final phosphate removal was close to the simulated value (Fig. 3). In all experiments, the pH increase was lower than 0.05 units until the minimum phosphate concentration was reached. Stronger pH increases were only recorded later in the experiment (Figure S1, supplementary information). The average molar ratios in the precipitates collected at the end of each batch experiment agreed well with the composi2þ 2þ and PO3
tion of struvite: the molar ratios of NHþ 4 :Mg 4 :Mg were 0.96  0.04 mol mol
1 and 1.02  0.04 mol mol
1 respectively. The content of sodium, potassium and calcium was low: 0.07  0.01 mol mol
1, 0.04  0.01 mol mol
1 and 0.04  0.01 mol mol
1 respectively. We analyzed the precipitate from the experiment with an applied anode potential of
0.6 V vs. NHE with XRD (Figure S2, supplementary information). The only mineral phase detected was struvite: no other magnesium mineral, such as magnesium oxide, magnesium hydroxide or magnesium carbonate, was observed.

3.3.

Sequencing batch experiment

A sequencing batch experiment was conducted to determine whether a similarly high performance could be achieved

Table 2 e Performance data of the experiments at various magnesium potentials. The current density and cell voltage were measured at a residual dissolved phosphate concentration of 50 mg P LL1. The current efficiency and energy demand were calculated for the time required to remove 75% of the initial phosphate. Anode potential [V] vs. NHE Ecorr
1.5 to
1.4
1.2
0.8
0.6
0.2

Phosphate removal rate [mg P cm
2 h
1]

Steady state current density [mA cm
2]

Current efficiency [%]

Cell voltage [V]

Energy demand [W h g P
1]

0.9 1.7 3.2 4.2 5.8

0 0.8 4.8 6.6 7.5

n.a. 222 119 122 136

0
0.16 0.62 1.11 1.50

0 0.1 1.0 1.7 2.2

294

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

was 118  10%. The storage and drying of the electrodes overnight had no obvious effect on phosphate removal, current density or current efficiency.

3.4.

Fig. 3 e Phosphate removal as a function of the Mg:P dosage ratio. The phosphate removal in the batch experiments was lower than the maximum removal simulated with PHREEQC. By allowing a longer contact time, the simulated values were almost reached (star). In this case, the additional contact time was three days.

when the same magnesium electrode was used several times in a row. The experiment consisted of 13 cycles, each of them lasting 2 h. The achieved phosphate removal varied between 59 and 84%, which was mainly due to the variation of the initial phosphate content (between 180 and 226 mg L
1). A higher phosphate removal would have been achieved by allowing longer reaction times. In every cycle, the current reached a constant level after a short adjustment phase (steady state current, Fig. 4). The current densities were highest in the first three experiments (6.5  0.1 mA cm
2), and then dropped to a lower level of 5.2  0.5 mA cm
2, possibly because precipitates accumulated at the electrode. The average current density for all experiments was 5.5  0.7 mA cm
2. We achieved an average phosphate removal rate of 3.7  0.5 mg P cm
2 h
1. The observed phosphate removal rate was faster than expected based on the measured current densities, resulting in a current efficiency above 100% for all experiments: the average current efficiency

The magnesium and steel electrodes used in the sequencing batch experiment were examined with SEM (Fig. 5). Before the start of the experiment, the magnesium surface showed straight multidirectional scratches. The magnesium also contained aluminium (white spots, Fig. 5). Examination after the experiment revealed that the steel electrode was covered by struvite particles of various sizes. The magnesium surface was also covered with minerals: These were mostly magnesium carbonate interspersed with some characteristic struvite crystals. The minerals on the magnesium electrodes were scratched off and analyzed with XRD (Figure S3, supplementary information). This analysis confirmed the SEM findings and the magnesium carbonate was identified as nesquehonite (MgCO3$3H2O). To determine the composition of the passivation layer built during the first passivation, we also examined the mineral deposition on the magnesium electrode from the batch experiment at
1.2 V vs. NHE. The identified minerals were mainly nesquehonite, elemental aluminium and some struvite (Figure S4, supplementary information).

3.5.

Simulation of nesquehonite and struvite saturation

To better understand the formation of nesquehonite and struvite, we simulated the saturation of nesquehonite and struvite at increasing magnesium concentrations (Fig. 6). The background concentration was chosen according to Table 1 with acetate as the COD source. In the simulation, the charge of the added Mg2þ was balanced with OH
, which caused a pH increase. At the initial pH value (8.9), nesquehonite was undersaturated, while struvite was slightly oversaturated. Above a pH value of 9, however, nesquehonite also became oversaturated and the saturation index remained above 0 up to the maximum simulated pH value. The saturation of struvite decreased when a pH value of 9.4 was exceeded. At pH values above 12.8, struvite was undersaturated, while nesquehonite was still oversaturated. The decrease of the struvite saturation is due to a decrease of the ammonium (NHþ 4 ) concentration.

3.6.

Fig. 4 e Steady state current density (A) and current efficiency (B) during the sequencing batch experiments (four days). The steady state current density was 5.5 ± 0.7 mA cmL2. The current efficiency was 118 ± 10%.

Mineral deposition on electrodes

Cost comparison of magnesium dosage

On the basis of our experimental results and literature data, we carried out a preliminary cost comparison of struvite production with electrochemical magnesium dosage and direct dosage of MgO, MgCl2 and MgSO4 (Table 3). The magnesium costs comprise the raw material costs but are also affected by the magnesium content and the dosage ratio needed for a certain degree of phosphate removal. Etter et al. (2011) achieved 90% phosphate removal with a dosage ratio of Mg:PO4 of 1.1 mol Mg mol P
1 for MgSO4 and bittern. This 90% removal includes the phosphate precipitation as well as the filtration efficiency. We used the same efficiency for MgCl2 and the magnesium anode calculations. MgO, as a salt

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

295

Fig. 5 e SEM pictures of the electrode surfaces used in the sequencing batch experiment: a bare magnesium surface before the experiments (left), the magnesium electrode after 13 cycles (middle) and the steel electrode after 13 cycles (right). The magnesium electrode contained aluminium (white spots) before the experiment; after the experiment, it was covered with magnesium carbonate (MgCO3$3H2O, nesquehonite) and some interspersed struvite particles. The steel electrode was covered with struvite.

of rather low solubility, is usually dosed at a higher ratio (Abegglen, 2008; Antonini et al., 2009). We consequently increased the dosage ratio to 1.5 mol Mg mol P
1. The costs of the electrochemical magnesium dosage were mainly governed by the material costs; the additional energy costs for the electrochemical process were negligible. In our comparison, MgO turned out to be by far the cheapest option ð0:85 V kg
1 struvite Þ, while MgSO4 and the magnesium anode cost about the same ð3:45 and 4:50 V kg
1 struvite Þ, and MgCl2 was the most expensive magnesium source ð6:80 V kg
1 struvite Þ.

4.

Discussion

4.1. Mechanism of electrochemical magnesium dissolution We observed two different kinds of galvanic corrosion: macro and microgalvanic corrosion (Song and Atrens, 2003).

Fig. 6 e PHREEQC simulation of the saturation index of nesquehonite (MgCO3$3H2O) and struvite (MgNH4PO4$6H2O), and the magnesium concentration as a function of the pH value. Background concentrations according to Table 1.

Macrogalvanic corrosion occurred when the magnesium electrode was connected to the steel electrode: magnesium dissolved and released Mg2þ ions to the solution, while hydrogen was produced at the steel electrode. The macrogalvanic corrosion increased when the anode potential was raised by applying a voltage. Microgalvanic corrosion, also known as local cell action (Andrei et al., 2003), was the cause of magnesium dissolution at Ecorr (i.e. when no electrons were withdrawn from the magnesium), but it was also observed at higher potentials. Microgalvanic corrosion proceeds on a microscale and the anode and cathode are located on the magnesium electrode itself. So this process occurs without any electrical connection to another electrode. Impurities or second phases (e.g. Mg17Al12) are cathodic to the magnesium matrix and cause the dissolution of pure magnesium (Song and Atrens, 2003). Microgalvanic corrosion is one explanation for the observed current efficiencies above 100%. Another explanation for the high current efficiencies comes from Song and Atrens (2003). These authors proposed that magnesium is either oxidized directly to Mg2þ or to Mgþ (Eq. (6)), which is subsequently oxidized in the bulk solution to Mg2þ (Eq. (7)) Mg/Mgþ þ e


(6)

Mgþ þ H2 O/Mg2þ þ OH
þ 0:5H2

(7)

According to these equations, the number of effectively transferred electrons range between 1 and 2 per magnesium ion (Hoey and Cohen, 1958; Makar and Kruger, 1990; Song, 2005), while a two-electron transfer is generally assumed for calculating the current efficiency (Eq. (3)). Another process that contributes to the release of magnesium to the solution is known as the chunk effect (Andrei et al., 2003; Marsh and Schaschl, 1960; Straumanis and Bhatia, 1963). It describes the mechanical loss of metal pieces from the electrode surface. The loss of chunks has to be expected after the magnesium electrode has been strongly corroded (Song and Atrens, 2003). However, the chunks may not only be pure magnesium particles, but also can be impurities or second phases. The dimensions of the chunks are usually very small

296

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

Table 3 e Cost comparison for a sacrificial magnesium anode and different magnesium salts. MgO is the cheapest option, followed by MgSO4, the magnesium anode and MgCl2. Magnesium compound Raw magnesium costs Magnesium contentc Magnesium required Magnesium costs Additional energy requirement Energy costsf Total costs a b c d e f

Mg anode ½V kg
1 source  ½kgMg kg
1 source  ½kgMg kg
1 struvite  ½V kg
1 struvite  ½kWh kg
1 struvite  ½V kg
1 struvite  ½V kg
1 struvite 

MgO

a

b

MgCl2 anhydrous b

MgSO4 anhydrous

32.50 0.90 0.12d 4.35 0.4

3.00 0.54 0.17e 0.90 e

14.10 0.25 0.12d 6.80 e

5.50b 0.12 0.12d 3.45 e

0.10 4.45

e 0.90

e 6.80

e 3.45

Magnesium sacrificial anode (1.8 kg) as quoted by Segelladen.de, Germany (August 2011). Produced by BDH PROLABO, price quoted by VWR International, Germany, for a 25 kg bag of technical grade material (August 2011). According to the certificate of specification. Assuming 90% phosphate precipitation at a dosage ratio of 1.1 mol Mg mol P
1. Assuming 90% phosphate precipitation at a dosage ratio of 1.5 mol Mg mol P
1. Assuming an electricity price of 0.3 V kWh
1.

(6  10
5 mm diameter) (Straumanis and Bhatia, 1963). The chunk effect was not quantified in this study. Microgalvanic corrosion and the release of monovalent ions increase the current efficiency and hence the struvite precipitation performance, whereas the loss of chunks has the opposite effect if they do not dissolve.

4.2.

Passivation

The first passivation between
1.5 and
0.9 V vs. NHE is usually explained by the formation of a magnesium hydroxide film (Eq. (8)) from dissolved magnesium ions (Makar and Kruger, 1990; Song and Atrens, 2003). Mg2þ þ 2OH
/MgðOHÞ2

(8)

This film is formed at low potentials and reduces the magnesium dissolution (Song and Atrens, 2003). The surface film on magnesium is not as stable as surface films on other metals, such as iron (Makar and Kruger, 1990). Anions, especially chloride, reduce the stability of the passivation layer (Song et al., 1997b). Corrosion starts at flaws in the surface film. When the anode potential exceeds the pitting potential, the areas without surface film grow, thus accelerating the magnesium dissolution. It is likely that the high chloride content of urine prevented the formation of a strong passivation layer. At a potential of
1.2 V vs. NHE, which was substantially lower than Ept (
0.9 V vs. NHE), the current density was considerable (0.8 mA cm
2, Table 2). The deposit layer at
1.2 V vs. NHE mainly contained nesquehonite, struvite and aluminium, but no Mg(OH)2 was detected: its fraction might simply have been too low for detection. The detection of aluminium is probably a sampling artifact: together with the passivation layer, some material from the electrode itself might have been scratched off during sampling. The magnesium electrode was also passivated at higher anode potentials (secondary passivation). The formation of the second layer explains the peak current observed in the voltammetric experiments shortly after Ept had been passed (Fig. 2A). After passing the peak, the current stabilized at slightly lower values, because diffusion through the porous

passivation layer limited magnesium dissolution. According to a semi-quantitative analysis of the X-ray diffraction pattern the secondary passivation layer contained more than 90% nesquehonite and less than 10% struvite (Figure S3, supplementary information). Gulbrandsen (1992) and Jo¨nsson et al. (2007) also observed the formation of magnesium carbonates, including nesquehonite, during magnesium corrosion in carbonate solutions and humid air respectively. In the experiments conducted by Gulbrandsen (1992), nesquehonite whiskers of 2e3 mm length formed a porous layer, which reduced the current by a factor of ten. In our experiments, the nesquehonite whiskers were much smaller (0.1 mm, Fig. 5), which probably explains why the currents did not decrease significantly in most cases. A substantial current decrease was only observed in the experiments performed within the first passivation. While the passivation layers consisted mainly of nesquehonite, the particles in the bulk solution consisted almost exclusively of struvite. The occurrence of nesquehonite at the electrode can be explained by higher pH values at the electrode surface caused by the production of hydroxyl ions (Eq. (7)). Our equilibrium simulations with PHREEQC (Fig. 6) support this hypothesis: nesquehonite was undersaturated at the pH values and magnesium concentrations in the bulk solution (pH: 8.9, maximum measured magnesium concentration: 150 mg L
1). However, a small increase above a pH of 9.0 and magnesium concentrations of 240 mg L
1 would have been sufficient to cause nesquehonite formation.

4.3.

Performance at different anode potentials

The batch experiments basically showed two different regions of magnesium dissolution, namely within the first and secondary passivations. This contrasts with the voltammetric experiments, where we observed a potential range with maximum currents. However, these peak currents were probably only attained because the secondary passivation layer was not yet fully formed, and diffusion-limiting conditions were not yet reached. Dissolution of magnesium within the first passivation should be prevented in practical applications, as the current

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

densities were low and even decreased with time. To avoid the first passivation, the anode potential must be above Ept. In the case of our experiments that would be above
0.9 V vs. NHE. However, Ept depends strongly on the electrolyte solution and the magnesium alloy (Gulbrandsen, 1992; Song et al., 1997a). Ept therefore needs to be determined by linear voltammetry, for instance, when different solutions or magnesium alloys are used. The secondary passivation layer built up above Ept, but it was porous and did not impede magnesium dissolution to the same extent as the primary passivation layer. Indeed, the current densities stabilized at high values (Fig. 4A, Table 2). The current density is an important factor in designing a reactor for electrochemical dissolution, because higher current densities lead to smaller reactor units (Chen, 2004). However, at least three other factors are crucial for optimal reactor performance: (1) stable long-term performance (resilience), (2) high current efficiency, and (3) low energy consumption. To ensure stable operation and minimize maintenance, Chen (2004) recommended current densities of 2e2.5 mA cm
2 for aluminium and iron electrodes. In our experiments with magnesium electrodes, such current densities were achieved at anode potentials slightly above Ept. At
0.8 V vs. NHE, the current density was already 4.7 mA cm
2 at an applied cell voltage of 1.62 V. Instead of setting a constant anode potential, which requires the use of a reference electrode, magnesium can also be dissolved by applying a constant current density. The current efficiency was above 100% in all experiments due to microgalvanic corrosion, the release of monovalent ions and the chunk effect (Song and Atrens, 2003). At potentials higher than Ept, there was no obvious trend in the current efficiencies either at different anode potentials (Table 2) or over different cycles (Fig. 4B). The energy consumption increased with higher anode potentials due to the higher voltages and currents. High currents and voltages should therefore be prevented, which is best done by keeping the anode potential not too far above Ept. The energy consumption did not exceed 2.2 W h g P
1 in our study, which is low compared to Zheng et al. (2009a), who reported energy consumptions above 8 W h g P
1 for iron electrodes at their recommended current densities of 40 mA cm
2 and a gap size of 5 mm between the electrodes. The same authors also found that the gap size had a crucial influence on energy consumption, which tripled when the gap was increased from 0.5 cm to 4 cm. A gap smaller than 5.5 cm between the magnesium and steel electrodes would therefore reduce the energy consumption. However, the gap should be wide enough to ensure sufficient mixing. The sequencing batch experiment over 13 cycles revealed that continuous magnesium dissolution is possible at a potential as low as
0.6 V vs. NHE. After a decrease of current density in the first three cycles, we achieved an average current density of 5.2  0.5 mA cm
2 and a phosphate removal rate of 3.7  0.5 mg P cm
2 h
1 for the subsequent cycles. The passivation with nesquehonite might be the reason for the slight decrease of the current density, but it did not lead to a breakdown of the process. The lower current density also reduced the phosphate removal rate. The current efficiency, however, was not affected by the layer formation. Nevertheless, we cannot exclude that the accumulation of nesquehonite will result in a considerable decrease of the

297

phosphate removal rate and the current efficiency during long-term operation. Hence, further research should focus on how magnesium carbonate depositions can be minimized or removed regularly. In conclusion, anode potentials just above Ept seem to be optimal, because considerable current densities are already achieved, the current efficiency is high and the energy consumption low. For practical applications, where only the current can be controlled, the optimal anode potential should be determined in preliminary experiments. For our experimental setup, a current density of around 5 mA cm
2 is optimal.

4.4.

Cost comparison

Electrochemical magnesium dosage from a sacrificial magnesium electrode turned out to be more expensive than the dosage of MgO, but is in the same range as MgCl2 and MgSO4. This result shows that electrochemical dosage of magnesium is competitive with the dosage of most magnesium salts, although many aspects, such as installation and maintenance costs, were neglected in this cost comparison. However, it is likely that the magnesium source will be the major cost of struvite production over the long-term, as was shown in a cost estimation for a field reactor in Nepal (Etter et al., 2011). As the costs depend strongly on the magnesium price, cost savings can be achieved by minimizing magnesium losses. This requires that (1) struvite is the main precipitation product, (2) struvite precipitation in the reactor is almost complete, (3) struvite is efficiently separated from urine, and (4) overdosage of magnesium is prevented. Struvite was the main precipitation product, even though some magnesium precipitated with carbonate to nesquehonite on the electrode surface. However, magnesium carbonate was only detected in the layer attached to the anode and not in the bulk liquid, which means that only a small fraction of the dissolved magnesium precipitated as magnesium carbonate. In our experiments, struvite precipitation was incomplete (Fig. 3). However, it almost reached its maximum thermodynamic value when the solution was left to react for a few additional hours. An additional tank for precipitation and sedimentation of struvite can consequently increase the overall struvite recovery. Alternatively, magnesium could be dosed in the influent pipe to a reactor. The turbulence in the pipe would help to minimize the passivation of the electrode (Chen, 2004) and the accumulation of magnesium carbonate. Continuous dosage of magnesium during electrochemical dissolution probably results in higher overall struvite recovery than a one-time salt dosage. Continuous dosage means low supersaturation, supporting the growth of large crystals which can be easily recovered by filtration or sedimentation (Ronteltap et al., 2010). Emptying the reactor is the easiest way to stop magnesium dissolution after sufficient magnesium has been dosed. As long as the metallic magnesium is in contact with the urine, electrochemical magnesium dissolution will continue due to electron transfer to the cathode or to local cell action. Our sequencing batch experiment has shown that intermittent air-drying of the electrode did not impede the magnesium dissolution in the next batch.

298

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

Although electrochemical magnesium dosage is more expensive than dosing magnesium oxide, the handling is simpler. Winker et al. (2011) reported that magnesium oxide had to be packed in small bags of polyvinyl alcohol to prevent the incrustation of the powder in the screw feeder. Electrochemical magnesium dosage involves direct dissolution of the magnesium in the solution and requires no mechanical feed mechanism. Such a system is particularly interesting for decentralized reactors, because the reactor setup promises to be simple and scalable in size.

5.

Conclusions

 Electrochemical magnesium dosage is a feasible process for struvite precipitation if the electrode potential is kept above Ept (
0.9 V vs. NHE in our study). An anode potential slightly above Ept is optimal because the energy consumption increases with the anode potential. Linear voltammetry can be used for the fast and reliable detection of Ept.  Electrochemical struvite precipitation has a high current efficiency (above 100%) and low energy consumption (1.7 W h g P
1 at a potential of
0.6 V vs. NHE).  Sequencing batch experiments revealed that electrochemical magnesium dissolution works for multiple treatment cycles. Although magnesium carbonate accumulated at the anode, the current density decreased only slightly. Further experiments are needed to determine whether mineral deposition on the electrodes could be problematic over extended operation times.  With respect to material costs, dissolution of a sacrificial magnesium electrode can compete with the dosage of easily soluble magnesium salts (MgCl2 and MgSO4) but is more expensive than MgO dosage.  Overall struvite recovery can be completed if the solution is stored for several hours before filtration. The solution should not be stored in the dosage chamber, because the magnesium metal will continue to dissolve as long as it is in contact with the electrolyte.  Thanks to the simple reactor setup and the easy handling of the magnesium source, electrochemical struvite precipitation is a promising technology for small, decentralized reactors.

Acknowledgments We thank Alain Bourgeois and Michel Walker for their preliminary studies, which were extremely helpful for the planning and realization of this work. Dr. Thomas Suter (EMPA) supported our study with laboratory materials and valuable advice. We also would like to thank Dr. Ralf Ka¨gi and Vanessa Sternitzke for the REM and XRD analysis and Claudia Ba¨nninger and Karin Rottermann for the chemical analyses.

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.watres.2012.09.036.

references

Abegglen, C., 2008. Membrane bioreactor technology for decentralized wastewater treatment and reuse. PhD thesis. Swiss Federal Institute of Technology Zurich, Switzerland. Andrei, M., di Gabriele, F., Bonora, P.L., Scantlebury, D., 2003. Corrosion behaviour of magnesium sacrificial anodes in tap water. Materials and Corrosion e Werkstoffe und Korrosion 54 (1), 5e11. Antonini, S., Paris, S., Clemens, J., 2009. Nitrogen and phosphorus recovery from human urine. In: Arnold, U., Gresens, F. (Eds.), Closing Nutrient Cycles in Decentralised Water Treatment Systems in the Mekong Delta. University of Bonn, Germany. Chen, G.H., 2004. Electrochemical technologies in wastewater treatment. Separation and Purification Technology 38 (1), 11e41. Cordell, D., Drangert, J.O., White, S., 2009. The story of phosphorus: global food security and food for thought. Global Environmental Change: Human and Policy Dimensions 19 (2), 292e305. Egner, S., Brynioc, D., 2007. Reactor for the production of magnesium ammonium phosphate, and method for producing magnesium ammonium phosphate from liquid manure or ammonium-containing exhaust gases. Patent No. WO 2007/009749 A1. EN_1754:1997, 1997. Magnesium and Magnesium Alloys e Magnesium and Magnesium Alloy Anodes, Ingots and Castings. European Committee for Standardization, Brussels. EN_10088-1, 1995. Stainless Steels. Part 1: List of Stainless Steels. European Committee for Standardization, Brussels. Etter, B., Tilley, E., Khadka, R., Udert, K.M., 2011. Low-cost struvite production using source-separated urine in Nepal. Water Research 45 (2), 852e862. Goosse, P., Steiner, M., Udert, K.M., Neuenschwander, W., 2009. NoMix-Anlage. Erste Monitoringergebnisse im Forum Chriesbach ((NoMix Toilet System. First monitoring results in Forum Chriesbach)) (in German), English summary. Gas Wasser Abwasser, 7/2009, pp. 567e574. 2
Gulbrandsen, E., 1992. Anodic behaviour of Mg in HCO
3 /CO3 buffer solutions. Quasi-steady measurements. Electrochimica Acta 37 (8), 1403e1412. Gurrappa, I., 2005. Cathodic protection of cooling water systems and selection of appropriate materials. Journal of Materials Processing Technology 166 (2), 256e267. Hoey, G.R., Cohen, M., 1958. Corrosion of anodically and cathodically polarized magnesium in aqueous media. Journal of the Electrochemical Society 105 (5), 245e250. Holt, P.K., Barton, G.W., Mitchell, C.A., 2005. The future for electrocoagulation as a localised water treatment technology. Chemosphere 59 (3), 355e367. Ikematsu, M., Kaneda, K., Iseki, M., Matsuura, H., Yasuda, M., 2006. Electrolytic treatment of human urine to remove nitrogen and phosphorus. Chemistry Letters 35 (6), 576e577. Jo¨nsson, M., Persson, D., Thierry, D., 2007. Corrosion product formation during NaCl induced atmospheric corrosion of magnesium alloy AZ91D. Corrosion Science 49 (3), 1540e1558. Larsen, T.A., Alder, A.C., Eggen, R.I.L., Maurer, M., Lienert, J., 2009. Source separation: will we see a paradigm shift in wastewater handling? Environmental Science & Technology 43 (16), 6121e6125. Makar, G.L., Kruger, J., 1990. Corrosion studies of rapidly solidified magnesium alloys. Journal of the Electrochemical Society 137 (2), 414e421. Marsh, G.A., Schaschl, E., 1960. The Difference effect and the chunk effect. Journal of the Electrochemical Society 107 (12), 960e965.

w a t e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 8 9 e2 9 9

Maurer, M., Schwegler, P., Larsen, T.A., 2003. Nutrients in urine: energetic aspects of removal and recovery. Water Science and Technology 48 (1), 37e46. Parkhurst, D.L., Appelo, C.A.J., 1999. User’s Guide to PHREEQC (Version 2) e A Computer Program for Speciation, Batchreaction, One-dimensional Transport, and Inverse Geochemical Calculations. Water-Resources Investigations Report 99-4259. US Geological Survey, Denver CO, USA. Ro¨mer, W., 2006. Plant availability of P from recycling products and phosphate fertilizers in a growth-chamber trial with rye seedlings. Journal of Plant Nutrition and Soil Science 169 (6), 826e832. Ronteltap, M., Maurer, M., Hausherr, R., Gujer, W., 2010. Struvite precipitation from urine e influencing factors on particle size. Water Research 44 (6), 2038e2046. Song, G., 2005. Recent progress in corrosion and protection of magnesium alloys. Advanced Engineering Materials 7 (7), 563e586. Song, G., Atrens, A., 2003. Understanding magnesium corrosionda framework for improved alloy performance. Advanced Engineering Materials 5 (12), 837e858. Song, G., Atrens, A., St John, D., Wu, X., Nairn, J., 1997a. The anodic dissolution of magnesium in chloride and sulphate solutions. Corrosion Science 39 (10e11), 1981e2004. Song, G., Atrens, A., Stjohn, D., Nairn, J., Li, Y., 1997b. The electrochemical corrosion of pure magnesium in 1 N NaCl. Corrosion Science 39 (5), 855e875. Straumanis, M.E., Bhatia, B.K., 1963. Disintegration of magnesium while dissolving anodically in neutral and acidic solutions. Journal of the Electrochemical Society 110 (5), 357e360. Taylor, A.W., Gurney, E.L., Frazier, A.W., 1963. Solubility products of magnesium ammonium and magnesium potassium

299

phosphates. Transactions of the Faraday Society 59 (487), 1580e1584. Udert, K.M., Larsen, T.A., Biebow, M., Gujer, W., 2003a. Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Research 37 (11), 2571e2582. Udert, K.M., Larsen, T.A., Gujer, W., 2003b. Estimating the precipitation potential in urine-collecting systems. Water Research 37 (11), 2667e2677. Wilsenach, J.A., van Loosdrecht, M.C., 2004. Effects of separate urine collection on advanced nutrient removal processes. Environmental Science & Technology 38 (4), 1208e1215. Winker, M., Paris, S., Heynemann, J., Montag, D., 2011. Erste Ergebnisse der Implementierung der Urin-, Braun und Grauwasserbehandlung im Eschborner GIZ Hauptgeba¨ude ((First implementation results of the urine, brown- and greywater treatment in the GIZ headquarters in Eschborn)) (in German). Leibzig, Deutschland. Zheng, X.-y., Kong, H.-N., Wu, D.-y., Wang, C., Li, Y., Ye, H.-r., 2009a. Phosphate removal from source separated urine by electrocoagulation using iron plate electrodes. Water Science and Technology 60 (11), 2929e2938. Zheng, X.-Y., Yan, L., Ye, H.-R., Li, J., Zhu, S.-Y., Chen, H.-H., Wang, C., Kong, H.-N., 2009b. Effects of initial pH and current density on the phosphate removal from source separated urine by electrocoagulation using iron plate electrodes. In: 3rd International Conference on Bioinformatics and Biomedical Engineering, iCBBE, Beijing. Zheng, X.Y., Shen, Y.H., Ye, H.R., Yan, L., Zhang, Y.J., Wang, C., Kong, H.N., 2010. Phosphate removal from source-separated urine by electrocoagulation using aluminium plate electrodes. Fresenius Environmental Bulletin 19 (4A), 635e640.