Nitrate reduction by mixed iron(II-III)

0 downloads 0 Views 745KB Size Report
May 28, 2014 - et al. 2010). While ferrous hydroxides Fe(OH)2 and siderite. FeCO3 compounds reduced very slowly nitrate (Matocha et al. 2012; Ottley et al.
w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

Available online at www.sciencedirect.com

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

Nitrate reduction by mixed iron(II-III) hydroxycarbonate green rust in the presence of phosphate anions: The key parameters influencing the ammonium selectivity Marjorie Etique a,b, Asfaw Zegeye a,b,1, Brian Gregoire a,b, Cedric Carteret a,b, Christian Ruby a,b,* a Universite de Lorraine, Laboratoire de Chimie Physique et Microbiologie pour l'Environnement, LCPME, UMR 7564, 405 rue de Vandœuvre, F-54600 Villers-les-Nancy, France b CNRS, Laboratoire de Chimie Physique et Microbiologie pour l'Environnement, LCPME, UMR 7564, 405 rue de Vandœuvre, F-54600 Villers-les-Nancy, France

article info

abstract

Article history:

The reduction of nitrate anions by a mixed FeIIeFeIII carbonated green rust (GR) in aqueous

Received 3 February 2014

medium is studied as a function of the initial pH and the initial concentrations of iron,

Received in revised form

phosphate and nitrate. The influence of these parameters on the fraction of nitrate

14 May 2014

removed and the production of ammonium is investigated by the help of statistical

Accepted 18 May 2014

experimental designs. The goal is to determine experimental conditions that maximize the

Available online 28 May 2014

þ fraction of NO 3 removed and concomitantly minimize the production of NH4 . Increasing

the phosphate concentration relatively to the initial FeII concentration inhibits the Keywords:

reduction of nitrate probably due to a surface saturation of the lateral sites of the GR

Denitrification

crystals. The kinetics of the reaction is greatly enhanced by increasing the initial pH at 10.5,

Green rust

however it leads to a global increase of the NHþ 4 production. A partial saturation of the

Iron

surface sites by phosphate leads to a global decrease of selectivity of the reaction towards

Layered double hydroxide

II ammonium. The evolution of the ratio of the NHþ 4 concentration to the Fe concentration

Passive treatment

confirms that the NO 3 species are only partially transformed into ammonium. Interest-

ingly at an initial pH of 7.5, the selectivity of the reaction towards NHþ 4 is often lower than ~30%. The reduction of nitrate by carbonated GR differs from the behavior of other GRs incorporating Cl, F and SO2 4 anions that fully transform nitrate into ammonium. Finally, if GR is intended to be used during a passive water denitrification process, complementary dephosphatation and ammonium treatments should be considered. © 2014 Elsevier Ltd. All rights reserved.

 de Lorraine, Laboratoire de Chimie Physique et Microbiologie pour l'Environnement, LCPME, UMR * Corresponding author. Universite s-Nancy, France. Tel.: þ33 (0)3 83 68 52 20. 7564, 405 rue de Vandœuvre, F-54600 Villers-le E-mail address: [email protected] (C. Ruby). 1  de Lorraine e CNRS, Laboratoire Interdisciplinaire des Environnements Continentaux. LIEC, UMR 7360, 15 Present address: Universite s-Nancy, France. avenue du Charmois, BP 70239, F-54500 Vandœuvre-le http://dx.doi.org/10.1016/j.watres.2014.05.028 0043-1354/© 2014 Elsevier Ltd. All rights reserved.

30

1.

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

Introduction

Nitrate contamination of groundwater is a result of intensive agriculture and industrial waste effluent discharge (Puckett, 1995; Showers et al. 2008). It has specific effects on infants leading methemoglobinemia and nitrate has also suspected carcinogenic properties on human health (Fan and Steinberg, 1996; Hoering and Chapman, 2004). It impacts the environment, especially in smothering the aquatic ecosystem through the growth of algae and their decomposition causing hypoxia and concomitant eutrophication (Robertson and Groffman, 2007). To remove nitrate from water numerous physicochemical and biological processes have been developed. Among these methods, biological denitrification (Kapoor and Viraraghavan, 1997; Mateju et al. 1992; Rivett et al. 2008), catalytic hydrogenation (Constantinou et al. 2007) and electrochemical reduction (Li et al., 2010; Talhi et al. 2011) have proved to selectively reduce nitrate to non-toxic dinitrogen molecules. Biological denitrification is more often performed in water purification plants situated in urban area and needs the control of microorganisms' activity and relatively costly infrastructures. Such treatments are not always suitable in rural area where municipal wastewater collection system and the construction of water-treatment plants could be very costly. Therefore, reed beds filters are sometimes implemented that are relatively efficient for removing organic matter from wastewater. However, nitrate removal by horizontal reed bed filters was found to be poor (O'Luanaigh et al., 2010) and ammonium species are generally partially converted into nitrate species in vertical reed bed filters (Schonerklee et al. 1997). Therefore, complementary redox mediated chemical nitrate reduction treatments using low cost materials should be used. A particular effort was devoted to study the reactivity towards nitrate of metallic iron Fe0 or other FeII containing compounds (Alowitz and Scherer, 2002; Cheng et al. 1997; Choe et al. 2004; Huang et al. 1998; Rakshit et al. 2008; Matocha et al. 2012; Xu et al., 2012; Zhang et al. 2010). While ferrous hydroxides Fe(OH)2 and siderite FeCO3 compounds reduced very slowly nitrate (Matocha et al. 2012; Ottley et al. 1997), significant NO 3 reduction rate was observed either for mixed FeIIeFeIII green rusts (Hansen et al. 2001, 1996; Robertson and Groffman, 2007) or for metallic Fe0 (Alowitz and Scherer, 2002; Cheng et al. 1997; Choe et al. 2004; Huang et al. 1998; Zhang et al. 2010). Green rusts (GR) are Fe(II)eFe(III) double hydroxysalt minerals constituted by positively charged hydroxide layers  2 separated by an interlayer of anions, An (e.g.: CO2 3 , SO4 , Cl ) and water molecules. Their general formulae was established to be {FeII(1x)FeIIIx(OH)2}xþ{(x/n)AnmH2O}x and their redox flexibility enabled them to reduce relevant organic and inorganic pollutants. The mechanism of oxidation of GR and their surface charge properties were the subject of several studies (Guilbaud et al. 2013; Ruby et al. 2010, 2006). GR with high FeII:FeIII ratio (typically FeII:FeIII of ~2:1) incorporating monovalent anions were shown to reduce nitrate more quickly than GR incorporating divalent anions (Hansen et al. 2001). It was even possible to enhance significantly the kinetics of nitrate reduction by adding soluble metallic species such as Cu2þ or Agþ in contact with GR (Choi and Batchelor, 2008). However, despite the fact that GR were able to reduce rapidly soluble

nitrate species, NO 3 species were quasi-exclusively transformed into more toxic ammonium NHþ 4 species (Choi and Batchelor, 2008; Hansen et al. 2001, 1996). This observation is a serious impediment to further application using these compounds for water denitrification processes. For instance, if one considers the water treatment performed in vertical reed beds filters, the oxic conditions lead to the transformation of the  NHþ 4 species into NO3 species that should further be treated and transformed into gaseous N species, preferentially N2 because N2O is a strong greenhouse gas. A tertiary treatment in which the reduction of nitrate anions would be reduced back into NHþ 4 should obviously be avoided. Therefore, the aim of this work is to explore the ability of hydroxycarbonate green rust {GR(CO2 3 )} to reduce nitrate by limiting the formation of ammonium. For this purpose, the systematic influence of 4 parameters e initial nitrate, ferrous and phosphate concentrations and initial pH e on the selectivity of GR(CO2 3 ) to transform nitrate into ammonium are studied through full factorial and Box-Behnken designs (Leardi, 2009; Ferreira et al. 2007). Implications of this study on the potential use of green rust for passive water treatment technology are also discussed.

2.

Materials and methods

II 2.1. Preparation of GR(CO2 3 ) by coprecipitation of Fe III and Fe species

GR(CO2 3 ) was synthesized using a coprecipitation method by adding a mixture of sodium hydroxide and sodium carbonate to a mixture of iron(II) and iron(III) salts (Bocher et al. 2004). In an anoxic chamber (95% N2/5% H2, Coy Laboratory Product Inc., Grass Lake, MI, USA), ferrous sulphate heptahydrate FeSO4$7H2O and ferric sulphate pentahydrate Fe2(SO4)3$5H2O were dissolved within 30 ml of pure water (Purelab Option-Q, Elga LabWater, Antony, France) in a 100 ml flask in the presence of various concentrations of phosphate NaH2PO4 : 0; 0.5; 1 mM. The ferric molar fraction x ¼ [FeIII]0/([FeII]0 þ [FeIII]0) was set at x ¼ 0.33 that corresponds to an FeII : FeIII ratio of 2 : 1. Then 30 ml of an appropriate concentrations of the basic {NaOH, Na2CO3} solution ([OH]0/{[FeII]0 þ [FeIII]0} ¼ 2 and II III II III so[CO2 3 ]0/{[Fe ]0 þ [Fe ]0} ¼ 7/6) was added to the Fe eFe lution. A bluish-green rust precipitate appeared immediately as confirmed by X-ray diffraction (Figure S1-a). At this stage, the pH of the GR1(CO2 3 ) suspension was close to 11 and it was readjusted to fixed values at 7.5, 9, 10.5 by adding a few drop of a concentrated 6 M H2SO4 solution. Once the pH was constant, the flask was sealed with thick butyl rubber stoppers before bubbling argon through the suspension for 30 min to outgas nitrogen. The reaction was started when 20 ml of Ar-bubbled NaNO3 solution was added resulting in initial nitrate concentrations of 0.81, 2.02 and 3.23 mM. Crimp seals were applied and the vials were placed into an incubator for 7 days where the temperature was fixed at 30  C.  2.2. Analyses of FeII ions and soluble NHþ 4 and NO3 species

At regular time intervals, the reaction mixtures were sampled using syringes. Total (dissolved and particulate) Fe(II) and total

31

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

Fe (Fe(II) þ Fe(III)) (extractable with 2 M HCl) were analyzed at 562 nm by the ferrozine method (Viollier et al. 2000) in an anoxic chamber. Ammonium was measured by the colorimetric salicylate/nitroprusside procedure (Baethgen and Alley, 1989) (Protocol S1) after dilution of the sample in pure water and filtering (0.22 mm, Acrodisc Supor®, Gelman, Pall Corporation). Nitrate concentrations were determined by ion chromatography (IC) after dilution of the sample in pure water and filtering. IC analyses were carried out with a Metrohm 882 Compact IC plus instrument equipped with a high-pressure pump, sequential (Metrohm CO2 suppressor MCS) and chemical (Metrohm suppressor MSM II for chemical) suppression modules and a conductivity detector. The separation was performed on a Metrosep ASupp5-250 column packed with polyvinyl alcohol particles functionalized with quaternary ammonium group (5 mm particle diameter) and preceded by a guard column (Metrosep A supp 4/5 guard) and an RP 2Guard column to remove traces of organic compounds. The mobile phase consisted of a mixture of 3.2 mM Na2CO3 (SigmaeAldrich, 99.5%) and 1 mM NaHCO3 (SigmaeAldrich, 99.7%) in pure water. The flow rate was 0.7 mL min1 and the sample loop volume was 20 mL.

2.3.

Design of experiments

Previous experiments performed by using hydroxychloride GR and hydroxysulphate GR have shown that GR was

transformed into magnetite by reducing nitrate into ammonium (Hansen et al. 2001). In a first approach, such a reaction can be hypothesized for GR (CO2 3 ) and leads to the following Equation (1):  2 II III þ þ 4FeII4 FeIII 2 ðOHÞ12 CO3 þ NO3 ¼ 8Fe Fe2 O4 þ NH4 þ 6H þ 4CO3

þ 19H2 O (1) One goal of this study was to reduce the maximal quantity of nitrate; therefore the majority of the reactions studied here (26 runs over a total of 30 runs) was performed with a FeII initial concentration present in GR(CO2 3 ) in excess in comII parison to nitrate concentration, i.e. a ratio [NO 3 ]0:[Fe ]0 lower than 1:16 ¼ 6.25% if one considers reaction (1). The range of nitrate concentration used in this study, i.e. 1 or 0.8 mM  [NO 50 mg L1  [NO 3 ]0  200 mg L 3 ]0  3.2 mM, corresponds to level measured in contaminated groundwater collection wells (Aquilina et al. 2012; Pauwels et al. 2000; Smith et al. 1999). The lower limit of nitrate concentration corresponds also to the upper limit of 50 mg L1 fixed by the European Union (European Commission, 1998). The corresponding range of FeII concentration was therefore fixed between 42 and 167 mM. The upper limit of phosphate concentration (1 mM or 95 mg L1) chosen in this study plays a stabilizing role for GR ðCO2 3 Þ (Bocher et al. 2004). Note that an excess amount of phosphate in groundwater, even at 0.5 mg L1, may cause eutrophication (Farmer, 2004). For

Table 1 e Parameters of the full factorial design and the Box-Behnken experiments. 3 M) [NO 3 ]0 (10

[FeII]0 (103 M)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

0.8 3.2 0.8 3.2 0.8 3.2 0.8 3.23 0.8 3.2 0.8 3.2 0.8 3.2 0.8 3.2 2

42 42 167 167 42 42 167 167 42 42 167 167 42 42 167 167 104

18 19 20 21 22 23 24 25 26 27 28 29 30

2 2 2 2 2 2 2 2 2 2 2 2 2

42 167 42 167 42 167 42 167 104 104 104 104 104

Runs

3 [PO3 M) 4 ]0 (10

pHi

II [NO 3 ]0/[Fe ]0 (%)

24 factorial design 0 7.5 1.9 0 7.5 7.7 0 7.5 0.48 0 7.5 1.9 1.0 7.5 1.9 1.0 7.5 7.7 1.0 7.5 0.48 1.0 7.5 1.9 0 10.5 1.9 0 10.5 7.7 0 10.5 0.48 0 10.5 1.9 1.0 10.5 1.9 1.0 10.5 7.7 1.0 10.5 4.9 1.0 10.5 1.9 0.5 9.0 1.9 3 factors Box-Behnken experiments 0 9.0 4.8 0 9.0 1.2 1.0 9.0 4.8 1.0 9.0 1.2 0.5 7.5 4.8 0.5 7.5 1.2 0.5 10.5 4.8 0.5 10.5 1.2 0 7.5 1.9 1.0 7.5 1.9 0 10.5 1.9 1.0 10.5 1.9 0.5 9.0 1.9

II [PO3 4 ]0/[Fe ]0 (%)

 [PO3 4 ]0/[NO3 ]0 (%)

0 0 0 0 2.4 2.4 0.6 0.6 0 0 0 0 2.4 2.4 0.6 0.6 0.5

0 0 0 0 123 31 123 31 0 0 0 0 123 31 123 31 24.7

0 0 2.4 2.4 1.2 0.3 1.2 0.3 0 1.0 0 1.0 0.5

0 0 49.5 49.5 24.7 24.7 24.7 24.7 0 49.5 0 49.5 24.7

32

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

 t

¼



NO 3

 0

     NO NO 3 t 3 0

(2)

1 ) is the initial nitrate concentration and where [NO 3 ]0 (mole L [NO ] is the nitrate concentration measured after a given time 3 t of reaction t. And Equation (3):

R NHþ 4

 t

¼

      þ NHþ NO 4 t  NH4 0 3 0

(3)

1 where [NHþ ) is the initial ammonium concentra4 ]0 (mole L þ tion and [NH4 ]t is the ammonium concentration measured after a time of reaction t. The ammonium selectivity can easily be calculated from Equations (2) and (3). Indeed, it is represented by the following Equation (4):

S NHþ 4

 t

¼



        þ NHþ NO 4 t  NH4 0 3 0  NO3 t

(4)

where the denominator of Equation (4) represents the quantity of consumed nitrate and the numerator the quantity of produced ammonium. As it will be observed later, some of the reactions did not  lead to measurable nitrate removal {[NO 3 ]t e [NO3 ]0 ~ 0} and concomitantly the eventual ammonium production is not detectable. For this reason, the values of S ðNHþ 4 Þt were not be chosen as a response of the factorial design (undetermined þ values). On the contrary, the values of XðNO 3 Þt and RðNH4 Þt are mathematically well defined for all the set of experiments and therefore were chosen as the two responses of the factorial design. Interestingly, the parameters that lower the  values of RðNHþ 4 Þt and increase the value of XðNO3 Þt are suitable indicators as they will induce the desired decrease of the values of S(NHþ 4 )t. The four independent factors of this study II are the initial nitrate ([NO 3 ]0), ferrous ([Fe ]0) and phosphate 3 ð½PO4 0 Þ concentrations and the initial pH (pHi). In a first step, each factor was studied at two levels e low level and high level e corresponding to run 1 to run 16 (Table 1). Central point (Figure S2) was performed 4 times in order to evaluate the standard deviation of each parameter (run 17). In a second step, a 3 factors Box-Behnken design (Figure S2) was realized at a constant nitrate concentration of 2 mM in order to determine more accurately the influence of 3 of the 4 initial parameters, i.e. [FeII]0, ½PO3 4 0 and pHi (run 18 to run 29). Central point was repeated 3 times to evaluate the standard deviation of each parameter (run 30). All runs were performed in 2 replicates; the total number of runs increased therefore to 60. The results of the factorial design were analyzed and interpreted by Minitab® Statistical Software, version 14 (http:// www.minitab.com/) to estimate which parameter minimized the transformation from nitrate to ammonium with an acceptable reduction of nitrate.

Results

3.1.

Nitrate reduction and ammonium production ratio

Typical time evolutions of the normalized nitrate and  and ammonium concentrations, i.e. ½NO 3 t =½NO3 0 þ þ ½NH4 t =½NH4 0 , are shown (Fig. 1). The full set of experiments is divided into 4 groups as following: - Experiments of group A (Table 2) are characterized by a negligible fraction of nitrate removed and a negligible ammonium production ratio (XðNO 3 Þ7 days ~ 0% and Þ ~ 0%), RðNHþ 7 days 4

(a) 100

Normalized nitrate concentration [NO3 ] / [NO3 ]0 (%)

X NO 3

3.

80

60

40

Run 5 Run 4 Run 24 Run 12

20

0 0

2

4

6

8

6

8

Time (days)

(b) Normalized ammonium concentration + [NH4 ] / [NO3 ]0 (%)

successful system optimization, a 24 full factorial design was carried out (Figure S2). It consists on a simultaneous study of the effects that four parameters, i.e. initial nitrate, ferrous and phosphate concentrations and initial pH, may have on two responses, i.e. the fraction of nitrate removed X(NO 3 )t and the ratio of produced ammonium to the initial nitrate concentration R(NHþ 4 )t. Here, both the fraction of nitrate removed and ammonium production ratio represent important indicators corresponding respectively to the following Equation (2):

Run 5 Run 4 Run 24 Run 12

100

50

0 0

2

4

Time (days)

Fig. 1 e Typical evolutions of the nitrate reduction (a) and the ammonium production (b) rates for different experimental conditions of the initial nitrate concentration, the initial quantity of ferrous iron and phosphate, and the starting pH (Table 1): run 5 (square), run 4 (circle), run 24 (triangle) and run 12 (diamond). Experimental conditions are: [NO 3 ]0 ¼ 0.8 mM, [FeII]0 ¼ 42 mM, [PO3 4 ]0 ¼ 0 mM, pH ¼ 7.5 for run 5; II 3 [NO 3 ]0 ¼ 3.2 mM, [Fe ]0 ¼ 167 mM, [PO4 ]0 ¼ 0 mM, II  pH ¼ 7.5 for run 4; [NO3 ]0 ¼ 2 mM, [Fe ]0 ¼ 42 mM,  [PO3 4 ]0 ¼ 0.5 mM, pH ¼ 10.5 for run 24; [NO3 ]0 ¼ 3.2 mM, II 3 [Fe ]0 ¼ 167 mM, [PO4 ]0 ¼ 0 mM, pH ¼ 10.5 for run 12. Error bars are standard deviations and the full lines correspond to the curve fitting using a pseudo first order law.

33

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

þ þ Table 2 e Fraction of nitrate removed X(NO 3 ), ammonium production ratio R(NH4 ) and selectivity S(NH4 ) divided into 4 groups (A, B, C &D).

Runs

3 [NO M) 3 ]0 (10

[FeII]0 (103 M)

3 [PO3 M) 4 ]0 (10

X(NO 3) % (t ¼ 7 days)

R(NHþ 4) % (t ¼ 7days)

S(NHþ 4) % (t ¼ 7days)

7.5

0

0

u.v.a

Group

1

0.8

42

2 5 6 13 14 20 21 22 29

3.2 0.8 3.2 0.8 3.2 2.0 2.0 2.0 2.0

42 42 42 42 42 42 167 42 104

0 1.0 1.0 1.0 1.0 1.0 1.0 0.5 1.0

7.5 7.5 7.5 10.5 10.5 9 9 7.5 10.5

0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0

u.v. u.v. u.v. u.v. u.v. u.v. u.v. u.v. u.v.

A

4 7 8 17b 18 19 23 26 30c

3.2 0.8 3.2 2.0 2.0 2.0 2.0 2.0 2.0

167 167 167 104 42 167 167 104 104

0 1.0 1.0 0.5 0 0 0.5 0 0.5

7.5 7.5 7.5 9 9 9 7.5 7.5 9

18.5 10.4 9.6 13.9 6.4 7.6 12.6 8.0 6.9

15.6 13.0 7.9 4.9 4.4 6.5 4.8 3.5 1.9

84 ~100 82 35 69 85 38 44 27

B

3 24 27

0.8 2.0 2.0

167 42 104

0 0.5 1.0

7.5 10.5 7.5

40.5 32.6 22.5

8.1 8.5 1.7

20 26 7

C

9 10 11 12 15 16 25 28

0.8 3.2 0.8 3.2 0.8 3.2 2.0 2.0

42 42 167 167 167 167 167 104

0 0 0 0 1.0 1.0 0.5 0

10.5 10.5 10.5 10.5 10.5 10.5 10.5 10.5

72.8 74.6 92.1 97.1 92.1 79.8 91.7 91.2

55.6 62.7 86.8 86.9 76.3 89.6 84.0 50.2

a b c

0

pHi

76 84 94 90 83 ~100 92 55

D

u.v. means undetermined value. Run 17 is the central point (Figure S2) of the full factorial design. The values correspond to the mean value of 4 repetitions. Run 30 is the central point of the Box-Behnken (Figure S2). The values correspond to the mean value of 3 repetitions.

- Experiments of group B (Table 2) are characterized by low fraction of nitrate removed and a low ammonium proþ duction ratio (X(NO 3 )7days < 20% and R(NH4 )7days < 20%), - Experiments of group C (Table 2) are characterized by a moderate fraction of nitrate removed and a very low ammonium production ratio (20% < XðNO 3 Þ7days < 50% and RðNHþ 4 Þ7days < 10%), - Experiments of group D (Table 2) are characterized by a high fraction of nitrate removed and a high ammonium production ratio (XðNO 3 Þ7days > 50% and ) > 50%). R(NHþ 7days 4 þ The whole set of XðNO 3 Þ7days and RðNH4 Þ7days values and the corresponding group are summarized (Table 2). For most of the experiments of group D, the nitrate reduction is almost completed with XðNO 3 Þ values higher than 90%. Therefore, the ammonium concentration measured after 7 days of reaction of the experiments of group D reaches values close to the initial nitrate concentration. As proposed previously for studying the kinetics of ammonium production during the reduction of nitrate by sulphated and chlorinated GR (Hansen, 2001), it is possible to adjust the variation of the

values of [NHþ 4 ]t by a first order rate law with the following Equation (5):     þ (5) NHþ 4 t ¼ NH4 max  f1  expð  kobs  tÞg where [NHþ 4 ]max represents the maximal ammonium concentration obtained when t / þ∞. The corresponding variation of the nitrate concentration is a simple exponential decay corresponding to the following Equation (6):      NO 3 t ¼ NO3 0  expð  kobs  tÞ

(6)

þ Typical adjustments of the ½NO 3 t and ½NH4 t curves with such first order laws are shown (Figure S3). The value of ½NHþ 4 max was not specified as the last data point and it was adjusted as a free parameter. The estimated values of  kobsðNHþ 4 Þ and kobsðNO3 Þ determined by using respectively Equations (5) and (6) are given for all the experiments of group D (Table 3). The values determined by both methods are relatively close for all the runs and average values of ¼ 3 ± 0.6  106 s1 and kobs(NO 3) þ 6 1 kobs(NH4 ) ¼ 2.4 ± 0.8  10 s are measured.

34

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

Table 3 e Pseudo order kinetic rate constants and corresponding standard deviations of experiments of group D. Runs

3 M) [NO 3 ]0 (10

9 0.81 10 3.23 11 0.81 12 3.23 15 0.81 16 3.23 25 2.0 28 2.0 Mean values of the eight runs

3 [PO3 M) 4 ]0 (10

42 42 167 167 167 167 167 104

0 0 0 0 1.0 1.0 0.5 0

1 6 kobs(NO 3 ) (s )  10

3.1 ± 2.4 ± 3.5 ± 3.5 ± 2.4 ± 2.4 ± 3.7 ± 4.3 ± 3±

1 6 kobs(NHþ 4 ) (s )  10

3.7 ± 1.6 3.2 ± 0.05 2.3 ± 0.3 1.7 ± 0.7 ea 1.4 ± 0.5 2.5 ± 1.1 2 ± 1.2 2.4 ± 0.8

0.3 0.1 0.1 0.2 0.3 0.2 0.5 0.3 0.6

Undetermined value due to too high error bars.

3.2. Selectivity of the nitrate reduction towards ammonium The selectivity of the reaction S(NHþ 4 )t can easily be visualised by plotting the values of R(NHþ ) as a function of X(NO t 4 3 )t (Fig. 2, Table S2). Points situated at the bottom-left of the figure correspond to  the coupled values of (R(NHþ 4 )1day, X(NO3 )1day). On the right hand þ side are presented the values R(NH4 )7days and X(NO 3 )7days values obtained after 7 days of reaction time. Intermediate values obtained at 2, 3 and 4 days for different runs are also presented. Points situated on the proximity of the segment [OM] for which þ R(NHþ 4 ) ¼ X(NO3) correspond to a selectivity S(NH4 )t of ~100% and points situated near the segment [ON] correspond to a selectivity S(NHþ 4 )t of ~0%. The segment [OP] for which the selectivity S(NHþ 4 )t ¼ 50% is also presented. Results of the experiments of group A are not represented because the values of S(NHþ 4 )t are not determined due to the quasi-absence of nitrate reduction (X(NO 3 )t ~ 0%). The most interesting experiments are those of group C for which the S(NHþ 4 )t values are relatively low (~5% < S(NHþ 4 )t < ~30%) meaning that a major proportion of NO3 species are not transformed into ammonium for runs 3, 24 and 27. Higher selectivities (~50% < S(NHþ 4 )t < ~100%) are measured for experiments of groups B and D. The S(NHþ 4 )7days values obtained for all the experiments are summarised in Table 2.

3.3.

Evolution of the FeII concentration

The green rust reactivity was also evaluated by measuring the evolution of the total FeII concentration of the suspension. Among the 30 experiments, only 5 runs present a detectable decrease of the FeII concentration in the presence of nitrate (runs 9, 10, 12, 16 and 28). For all the other experiments, the FeII concentration did not vary (see for instance curves of runs 1 and 15 in Fig. 3). Such a behavior can easily be explained by the fact that for these experiments either the fraction of nitrate removed was too low (group A) or the initial FeII concentration was too high in comparison to the initial nitrate concentration in order to detect any variation of the FeII concentration, e.g. runs 12, 13, 16, 17, 19; 21, 23, 25e30 for which II the [NO 3 ]0:[Fe ]0 ratio is lower than 2%.

3.4. Statistical analyses of the full 24 factorial design and Box-Behnken design The design matrix of uncoded values for the factors and the responses in terms of fraction of nitrate removed (X(NO 3 )) and

ammonium production ratio (R(NHþ 4 )) were established on the data of Table 1. A Pareto plot is drawn for each response to visually represent the absolute values of the effects of main parameters and their interactions (Fig. 4). The factors that overpass the indicated reference line are statistically significant with a confidence level of 90%. Here, the reference line stands at 11.28 and the initial pH, the ferrous iron and the phosphate concentrations have a significant impact on the X(NO 3 ) values (Fig. 4a), whereas only pHi and the ferrous concentration affect significantly the R(NHþ 4 ) response (Fig. 4b). The sign of the main effect gives its direction. For the fraction of nitrate removed, the most influential parameters, i.e. [FeII]0 and pHi, influence positively the X(NO 3 ) values, while the phosphate concentration ½PO3 4 0 has a negative impact (Fig. 4a). This means that the fraction of nitrate removed grows when the [FeII]0 and pHi values are high, and ½PO3 4 0 values are low. Nevertheless, the interaction between these three parameters has a positive repercussion on the X(NO 3 ) values. For the ammonium production ratio, the most influential parameters are the [FeII]0 concentration and pHi values that influence positively the R(NHþ 4 ) values (Fig. 4b). Therefore, a high value of [FeII]0 and/or pHi leads to a fast production of ammonium during the reduction of nitrate by GR(CO2 3 ). For all the experiments of the full factorial design,

100

+ Ammonium production ratio R(NH4) %

a

[FeII]0 (103 M)

Group B

80 Group C

60

S(NH4+) = 100 %

Run 4 Run 17 Run 24

7 days

Run 3 Run 12

Group D

S(NH4+) = 50 %

Run 11 Run 28

40 1 day

20

0 0

20

40

60

80

100

Fraction of nitrate removed X(NO3) %

Fig. 2 e The selectivity towards ammonium of the nitrate ¡ reduction is visualized in the R(NHþ 4 ) ¼ f (X(NO3 )) diagram. The two lines corresponding to an ammonium selectivity ¡ þ S(NHþ 4 ) ¼ R(NH4 )/X(NO3 ) of 50% and 100% are showed. Experimental conditions of each run are given in Table S2.

35

180 170 160 150 140 130 120 110

II

[ Fe ] (mM)

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

-

(a) run 15

Pareto plot for X(NO ) 3

D B C B*C*D B*D C*D B*C A*C*D A*B*C*D A*B A A*D A*B*D A*B*C A*C

25 20

run 12

15

+ 11.28

45 40

run 1

35

run 9

10 5 0

30 -5

25 20

run 10 0

1

2

3

4

5

6

7

- 11.28

Reaction time (Days) Fig. 3 e Typical evolution of the concentrations of ferrous iron during the nitrate reduction by the hydroxycarbonated green rust for different run numbers. Error bars are standard deviations. Experimental conditions of each run are computed in Table 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

only one (run 3 of group C in Table 2) leads to an appreciable reduction of nitrate together with a low ammonium production ratio. The Box-Behnken is designed to refine the effect of the main parameters, i.e. [FeII]0, pHi and ½PO3 4 0 . In this approach, the initial amount of nitrate was fixed at 2 mM and only the three parameters ([FeII]0, pHi and ½PO3 4 0 ) were varied (Table 1, runs from 18 to 30). Two supplementary experiments  present R(NHþ 4 ) and X(NO3 ) values that fulfill the criteria of group C (Table 2, runs 24 and 27). Unfortunately, the linear þ regression of both responses X(NO 3 ) and R(NH4 ) is not consistent for the full factorial design (Table S3). Even with the Box-Behnken, a second order polynomial regression is not suitable to model the responses (Table S4).

4.

Discussion

4.1.

Kinetics of the nitrate reduction

4.1.1.

Critical role of the phosphate concentration

The data exhibit clearly the strong influence of the phosphate concentration on the fraction of nitrate removed (Table 2). Significant nitrate decrease, i.e. X(NO 3 ) values higher than ~ 70%, is only measured for experiments where the ratio II ½PO3 4 0 :[Fe ]0 is equal or lower than 0.6% (see runs of group D in Tables 2 and 1). On the contrary, all the experiments where the II ratio ½PO3 4 0 :[Fe ]0 is fixed at 2.4% are characterized by the absence of nitrate reduction (X(NO 3 ) ~ 0% for runs 5, 6, 13, 14, 20 and 21 of group A). In a previous study (Bocher et al. 2004), the relative quantity of phosphate adsorbed onto GR ðCO2 3 Þ crystals that led to a quasi-saturation of the surface sites was

-10 -15 +

(b)

Pareto plot for R(NH ) 4

25

20

15

+ 11.28

10

5

D B B*D C C*D B*C*D B*C A*B*C*D A A*C*D A*D A*B A*B*C A*C A*B*D

0

-5

- 11.28

-10

Fig. 4 e Pareto charts for standardized effects for X(NO 3) II  3 and R(NHþ 4 ). The parameters [NO3 ]0, [Fe ]0, [PO4 ]0 and pHi of the full factorial design was symbolized by A, B, C and D, respectively. Interactions between each independent parameter A, B, C and D were considered up to the fourth degree. The reference line presents a fixed value at ±11.28 for a confidence interval of 90%.

II estimated to be in the range ~1%  ½PO3 4 0 :[Fe ]0  ~2% (see Fig. 5 of (Bocher et al. 2004)). The study showed also that the PO3 4 anions was preferentially adsorbed on the lateral surface sites of the hexagonal GR ðCO2 3 Þ crystals and led to a chemical stabilisation of GR ðCO2 3 Þ by preventing its partial transformation into a {Fe3O4, Fe(OH)2} mixture. According to these previous data, it can be suggested that GR ðCO2 3 Þ crystals chemically stabilised by phosphate anions are also nonreactive towards nitrate due to the barrier effect of P species adsorbed on the lateral faces of crystals. Moreover, a partial saturation of these surface sites by phosphate species induces also a lowering of the nitrate reduction (e.g. one may compare run 15 to run 11 or run 8 to run 3 in Tables 1 and 2), as confirmed

36

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

(Hansen, 2001). Note also that in our experiments the FeII concentration is in excess relative to the nitrate concentration II (½NO 3 0:[Fe ]0 ~ 0.02  0.08) and that a reverse situation is chosen in the experiments of Hansen et al. (2001) ([NO 3 ]0 :[FeII]0 ~ 1.47). Two supplementary experiments performed II 2 at much higher [NO 3 ]0:[Fe ]0 with GRðCO3 Þ lead to a gradual increase of the reaction kinetics (Figure S5). All the values of kobsðNHþ 4 Þ measured in this work are summarized in table 4 as II a function of pHi and the (½NO 3 0:[Fe ]0) ratio.

4.2. Selectivity towards ammonium from the nitrate reduction 4.2.1.

Fig. 5 e Evolution of the nitrate concentration consumption as a function of the iron(II) consumption for time reaction situated between 4 and 7 days. The lines correspond to the reactions described in Table S1.

by the negative repercussion of the ½PO3 4 0 parameter on the X(NO 3 ) values of the statistical analyses (Fig. 4a).

4.1.2. Parameters influencing the values of the kinetic constants 6 1 The mean values kobs(NHþ s measured for 4 ) ¼ 2.4 ± 0.8  10 2 GR ðCO3 Þ of experiments of group D are of the same order of magnitude as the values measured previously for GR ðSO2 4 Þ at 6 1 s (Hansen et al. a pH of ~8, i.e. kobsðNHþ 4 Þ ¼ 9.5 ± 3.8  10 2001). However, the experiments of group D are all performed at a pH of 10.5 and a significant decrease of the reaction kinetics is observed at lower pH in agreement with the positive influence of the pH on the XðNO 3 Þ values (Fig. 4a). Therefore, the question arises whether the nature of the 2 intercalated anions (SO2 4 or CO3 ) or other experimental parameters govern the kinetics of the reaction. In order to answer the above mentioned question, the reduction of ni2 trate by GRðSO2 4 Þ and GRðCO3 Þ is studied in identical experimental condition, i.e. a pH of 7.5, ½NO 3 0 ¼ 3.2 mM, [FeII]GR ¼ 167 mM. (Figure S4). The effect of the intercalated 2 anions, i.e. CO2 3 or SO4 , is relatively limited and the kinetics of nitrate reduction of GR(CO2 3 ) is only slightly lower than the  one observed for GRðSO2 4 Þ. Rate constants kðNO3 Þ of 7 1 7 1 6.1 ± 0.7  10 s and 4.3 ± 0.4  10 s are estimated for 2 GRðSO2 4 Þ and GRðCO3 Þ according to Equation (6). These results show that the most influential parameters identified in this study, in particular the pH, governs the reaction kinetics by another divalent anions has and that substituting CO2 3 only a minor effect. As proposed previously, it is necessary to substitute the divalent anion by a monovalent anion such as Cl to observe a significant increase of the reaction kinetics

Analyses of nitrate and FeII consumptions

Most of the selectivity values SðNHþ 4 Þ are significantly lower than 100% (Table 2, Fig. 3). It reveals that nitrate is not fully transformed into ammonium and that reaction (1) alone is only a part of the occurring processes. Raman spectroscopy demonstrates that the main oxidation product is in fact magnetite as proposed in Equation (1) (Figures S1-b & S6) and in agreement with a previous work (Das and Hendry, 2011). Nevertheless, the eventual formation of small quantity of nanocrystals of ferric oxyhydroxide cannot be completely ruled out and it is difficult to identify such disordered compounds with X-ray diffraction. Therefore, other chemical reactions leading to the formation of a ferric oxyhydroxide of formula FeOOH instead of magnetite and to dinitrogen N2(g), nitrous oxide N2O or nitric oxide NO(g) instead of NHþ 4 are also considered (Table S1). The eventual formation of nitrite ions NO-2 is not taken into account here since these species were not detected with ion chromatogII raphy. The expected ratio ½NO 3 :[Fe ]GR corresponding to the amount of consumed nitrate ions over the amount of consumed FeII species during all the reactions presented in table S1 are 1:8, 1:5, 1:4 and 1:3 when NHþ 4 , N2, N2O and NO are formed, respectively. Note that the predicted II ½NO 3 :[Fe ]GR ratios are independent of the nature of the final oxidation product, i. e. Fe3O4 or FeOOH. The evolution of the nitrate consumption as a function of the iron(II) consumption for several runs is presented in Fig. 5. As expected, experiments that exhibit an ammonium selectivity close to 100% fits well the line of slope 1:8, confirming the occurrence of reaction 1. It suggests also that other species, e.g. N2, N2O or NO, are formed for the experiments having lower ammonium selectivity values (e.g. runs 24 & 28).

4.2.2. Analyses of the experiments of group C (S(NHþ 4 ) < ~30%) The three experiments of group C, i.e. runs 3, 24 and 27, are particularly interesting since they exhibit the lowest SðNHþ 4Þ values. In alkaline conditions (pHi ¼ 10.5), it appears that the presence of an adequate quantity of phosphate, i.e. II ½PO3 4 0 :[Fe ]0 ~ 1%, is necessary to lower the selectivity; indeed II þ  for ½PO3 4 0 :[Fe ]0 ratio lower than 1%, S(NH4 ) values higher ) values of experiments than 70% were obtained (see the S(NHþ 4 II  :[Fe ] ratio of 2.4%, the of group D in Table 2) and for ½PO3 0 4 0 reduction of nitrate was completely inhibited (see runs 13 and 14 in Table 2). Therefore, one may hypothesize that a partial saturation of the lateral surface sites of the GRðCO2 3 Þ crystals by phosphate anions is favourable to the observed lowering of

37

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

Table 4 e Comparison of the ammonium pseudo order kinetic rate constant of the reactivity of hydroxycarbonate and II hydroxysulphate green rusts towards nitrate at different pH and [NO 3 ]0:[Fe ]0 ratios. Type of GR II [NO 3 ]0:[Fe ]0 pHi 6 1 s kobs(NHþ 4 )  10

GR(CO2 3 ) (this work)

GR(CO2 3 ) (this work)

GR(CO2 3 ) (this work)

GR(CO2 3 ) (this work)

GR(SO2 4 ) (this work)

GR(SO2 4 ) (Hansen et al., 2001 )

0.02e0.08 10.5 2.4 ± 0.8

0.29 10.5 4.2 ± 0.7

0.48 10.5 6.6 ± 0.5

0.02 7.5 ~0.40 ± 0.07a

0.02 7.5 ~0.60 ± 0.07a

1.4e 7 6e8 9.5 ± 0.4

 Estimated values by assuming that kobs(NHþ 4 ) ~ kobs(NO3 ). This assumption is in agreement with the fact that the ammonium selectivity S(NHþ ) is higher than 80% for these experiments. 4 a

the selectivity SðNHþ 4 Þ. In neutral condition at a pH of 7.5, relatively low values of SðNHþ 4 Þ were obtained even in the absence of phosphate, e.g. SðNHþ 4 Þ ¼ 26% for run 3. Such behaviour was not observed for green rust incorporating other anions such as F, Cl, and SO2 4 (Hansen et al. 2001).

4.3. Implications of the results for passive water treatment Filtration materials such as sand or pouzzolana coated with green rust could be considered for future applications such as water denitrification. Generally, domestic wastewater contains both phosphate and nitrate anions that are not systematically eliminated before water is released in natural environment. The experiments presented here demonstrate that phosphate anions may strongly lower the fraction of nitrate removed by green rust, i.e. when phosphate concentration is in the range 10e40 mg L1. Therefore, for future filtration applications, removing phosphate in a first reactor before performing the denitrification with green rust in a second separated reactor is recommended. Adsorption of phosphate on low cost material such as hydroxyapatite, steel slag or ferric oxyhydroxides coatings could be used for the first step of this process (Barca et al. 2012; Bellier et al. 2006; Naille et al. 2013). A second difficulty is the presence of remaining ammonium in the outflow of the second reactor, since a selectivity S(NHþ 4 ) of ~0% is never observed in the whole set of the presented experiments. Mixing the filtration materials with ammonium adsorbent such as clays could be a possible way to eliminate residual ammonium (Cho et al. 2010; Li et al., 2010; Zhang et al. 2011). Moreover, the effect of strong complexing agent other than phosphate, e.g. silicates anions or humic substances, should be considered in natural conditions. Finally, the calculated rate constants 7 1 þ s to kðNO 3 Þ and kðNH4 Þ situated in the range of ~5  10 ~3  106 s1 can be used to estimate the order of magnitude of contact time between green rust and nitrate inside the filtration reactor. If the goal is for instance to reduce by a factor 5 the initial nitrate concentration, e.g. to diminish the nitrate concentration from 50 mg L1 to 10 mg L1, a retention time situated between ~6 days and ~37 days would be necessary if Equation (6) is used for the calculation. The retention time values estimated here exist only in extensive wastewater treatment plants such as horizontal bed filters or constructed wetlands where green rust could be incorporated. Such estimation is valid only if an equivalent concentration of total iron is used in the water treatment filter in comparison with the value used in the laboratory, i.e. between 3.5 and 14 g L1.

5.

Conclusion

The key parameters influencing the fraction of nitrate þ removed (NO 3 ) and the ammonium production ratio R(NH4 ) are in order of importance the pH, the initial concentration of ferrous ions in GR ([FeII]0) and the initial phosphate concentration in the aqueous medium ([PO3 4 ]0). More precisely, the reduction of nitrate by GR is completely inhibited for a ratio II [PO3 4 ]0/[Fe ]0 of 2.4%.The two responses are strongly correlated with each other and an optimal set of parameters that concomitantly induce an increase of the X(NO 3 ) values combined to a decrease of the RðNHþ 4 Þ values cannot be easily determined. However, singular experiments lead to a moderate fraction of nitrate removed (20% < XðNO 3 Þ7days < 50%) combined to a low ammonium production ratio (RðNHþ 4 Þ7days < 10%). Therefore, it is important to prioritise the responses in order to adjust the parameters at adequate values. Two water treatment strategies could be proposed: (i) the first one consists of choosing the fraction of nitrate removed X(NO 3 ) as a priority response. This treatment is optimised in alkaline condition, high concentration of FeII and low concentration of phosphate. It would require an additional phosphate adsorption pre-treatment and an ammonium adsorption post-treatment. (ii) the second strategy consists of choosing the ammonium production ratio R(NHþ 4 ) as a priority response. This treatment is optimised in neutral condition, low concentration of FeII and high phosphate concentration (with II [PO3 4 ]0:[Fe ]0 < 2.4%). Depending on the concentration of phosphate present in the water to be treated, the phosphate pre-treatment and the ammonium adsorption post-treatment could be avoided or more limited. Because a moderate lowering of fraction of nitrate removed seems to be necessary to lower the ammonium selectivity, efficient denitrification would requisite a contact time between green rust and nitrate of a few weeks. Modifying the green rust surface properties in order to increase the fraction of nitrate removed and to lower the ammonium production ratio would take the research to a new level allowing potential application of green rust for passive water treatment.

Acknowledgments The Agence Nationale de la Recherche ANR is gratefully acknowledged for financial support (ECOTECH2009 e

38

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

le my for N 0994C0103). The authors acknowledge Dr. K. Barthe the ion chromatography analyses.

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.watres.2014.05.028.

references

Alowitz, M.J., Scherer, M.M., 2002. Kinetics of nitrate, nitrite, and Cr(VI) reduction by iron metal. Environ. Sci. Technol. 36 (3), 299e306. Aquilina, L., Vergnaud-Ayraud, V., Labasque, T., Bour, O., nat, J., Ruiz, L., de Montety, V., De Ridder, J., Roques, C., Mole Longuevergne, L., 2012. Nitrate dynamics in agricultural catchments deduced from groundwater dating and long-term nitrate monitoring in surface e and groundwaters. Sci. Total Environ. 435e436, 167e178. Baethgen, W.E., Alley, M.M., 1989. A manual colorimetric procedure for measuring ammonium nitrogen in soil and plant Kjeldahl digests. Commun. Soil. Sci. Plant Anal. 20 (9e10), 961e969. Barca, C., Gerente, C., Meyer, D., Cliazarenc, F., Andres, Y., 2012. Phosphate removal from synthetic and real wastewater using steel slags produced in Europe. Water Res. 46 (7), 2376e2384. Bellier, N., Chazarenc, F., Comeau, Y., 2006. Phosphorus removal from wastewater by mineral apatite. Water Res. 40 (15), 2965e2971. Bocher, F., Gehin, A., Ruby, C., Ghanbaja, J., Abdelmoula, M., Genin, J.M.R., 2004. Coprecipitation of Fe(II-III) hydroxycarbonate green rust stabilised by phosphate adsorption. Solid State Sci. 6 (1), 117e124. Cheng, I.F., Muftikian, R., Fernando, Q., Korte, N., 1997. Reduction of nitrate to ammonia by zero-valent iron. Chemosphere 35 (11), 2689e2695. Cho, D.W., Chon, C.M., Jeon, B.H., Kim, Y., Khan, M.A., Song, H., 2010. The role of clay minerals in the reduction of nitrate in groundwater by zero-valent iron. Chemosphere 81 (5), 611e616. Choe, S.H., Liljestrand, H.M., Khim, J., 2004. Nitrate reduction by zero-valent iron under different pH regimes. Appl. Geochem. 19 (3), 335e342. Choi, J., Batchelor, B., 2008. Nitrate reduction by fluoride green rust modified with copper. Chemosphere 70 (6), 1108e1116. Constantinou, C.L., Costa, C.N., Efstathiou, A.M., 2007. The remarkable effect of oxygen on the N2 selectivity of water catalytic denitrification by hydrogen. Environ. Sci. Technol. 41 (3), 950e956. Das, S., Hendry, M.J., 2011. Application of Raman spectroscopy to identify iron minerals commonly found in mine wastes. Chem. Geol. 290 (3e4), 101e108. European Commission, 1998. The Implementation of Council Directive 91/676/EEC Concerning the Protection of Waters Against Pollution Caused by Nitrates From Agricultural Sources. Communities, Office for Official Publications of the European Communities. L-2985 Luxembourg. Fan, A.M., Steinberg, V.E., 1996. Health implications of nitrate and nitrite in drinking water: an update on methemoglobinemia occurrence and reproductive and developmental toxicity. Regul. Toxicol. Pharmacol. 23 (1), 35e43. Farmer, A.M., 2004. In: Valsami-Jones, E. (Ed.), Phosphorous in Environmental Technology: Principles and Applications. IWA, pp. 174e191. Ferreira, S.L.C., Bruns, R.E., Ferreira, H.S., Matos, G.D., David, J.M., Brandao, G.C., da Silva, E.G.P., Portugal, L.A., dos Reis, P.S.,

Souza, A.S., dos Santos, W.N.L., 2007. Box-Behnken design: an alternative for the optimization of analytical methods. Anal. Chim. Acta 597 (2), 179e186. Guilbaud, R., White, M.L., Poulton, S.W., 2013. Surface charge and growth of sulphate and carbonate green rust in aqueous media. Geochim. Cosmochim. Acta 108, 141e153. Hansen, H.C.B., 2001. Environmental chemistry of Iron(II)-iron(III) LDHs (green rusts). In: Rives, V. (Ed.), Layered Double Hydroxides: Present and Future. Nova Science Publishers, Huntington, N.Y, pp. 469e493. Hansen, H.C.B., Guldberg, S., Erbs, M., Bender Koch, C., 2001. Kinetics of nitrate reduction by green rustsdeffects of interlayer anion and Fe(II):Fe(III) ratio. Appl. Clay Sci. 18 (1e2), 81e91. Hansen, H.C.B., Koch, C.B., NanckeKrogh, H., Borggaard, O.K., Sorensen, J., 1996. Abiotic nitrate reduction to ammonium: key role of green rust. Environ. Sci. Technol. 30 (6), 2053e2056. Hoering, H., Chapman, D., 2004. Nitrate and Nitrite in Drinking Water. IWA Publishing, London. Huang, C.P., Wang, H.W., Chiu, P.C., 1998. Nitrate reduction by metallic iron. Water Res. 32 (8), 2257e2264. Kapoor, A., Viraraghavan, T., 1997. Nitrate removal from drinking water e Review. J. Environ. Eng. Asce 123 (4), 371e380. Leardi, R., 2009. Experimental design in chemistry: a tutorial. Anal. Chim. Acta 652 (1), 161e172. Li, J.F., Li, Y.M., Meng, Q.L., 2010a. Removal of nitrate by zerovalent iron and pillared bentonite. J. Hazard. Mater. 174 (1e3), 188e193. Li, M., Feng, C., Zhang, Z., Chen, R., Xue, Q., Gao, C., Sugiura, N., 2010b. Optimization of process parameters for electrochemical nitrate removal using Box-Behnken design. Electrochimica Acta 56 (1), 265e270. Mateju, V., Cizinska, S., Krejci, J., Janoch, T., 1992. Biological water denitrification e a review. Enzyme Microb. Technol. 14 (3), 170e183. Matocha, C.J., Dhakal, P., Pyzola, S.M., 2012. In: Sparks, D.L. (Ed.), The role of Abiotic and Coupled Biotic/Abiotic Mineral Controlled Redox Processes in Nitrate Reduction, Advances in Agronomy, vol. 115, pp. 181e214.  le my, K., Mallet, M., Ruby, C., 2013. Proce  de  de Naille, S., Barthe riau de filtration, PCT/EP2013/066651. fabrication d'un mate O’Luanaigh, N.D., Goodhue, R., Gill, L.W., 2010. Nutrient removal from on-site domestic wastewater in horizontal subsurface flow reed beds in Ireland. Ecol. Eng. 36, 1266e1276. Ottley, C.J., Davison, W., Edmunds, W.M., 1997. Chemical catalysis of nitrate reduction by iron(II). Geochim. Cosmochim. Acta 61 (9), 1819e1828. Pauwels, H., Foucher, J.C., Kloppmann, W., 2000. Denitrification and mixing in a schist aquifer: influence on water chemistry and isotopes. Chem. Geol. 168 (3e4), 307e324. Puckett, L.J., 1995. Identifying the major sources of nutrient water-pollution. Environ. Sci. Technol. 29 (9), 408Ae414A. Rakshit, S., Matocha, C.J., Coyne, M.S., 2008. Nitrite reduction by siderite. Soil. Sci. Soc. Am. J. 72, 1070e1077. Rivett, M.O., Buss, S.R., Morgan, P., Smith, J.W.N., Bemment, C.D., 2008. Nitrate attenuation in groundwater: a review of biogeochemical controlling processes. Water Res. 42 (16), 4215e4232. Robertson, G.P., Groffman, P., 2007. Nitrogen transformations. In: Paul, E.A. (Ed.), Soil, Microbiology, Biochemistry, and Ecology. Springer, New York, USA, pp. 341e364. Ruby, C., Abdelmoula, M., Naille, S., Renard, A., Khare, V., OnaNguema, G., Morin, G., Genin, J.M.R., 2010. Oxidation modes and thermodynamics of FeIIeIII oxyhydroxycarbonate green rust: dissolution-precipitation versus in-situ deprotonation. Geochim. Cosmochim. Acta 74 (3), 953e966. Ruby, C., Upadhyay, C., Gehin, A., Ona-Nguema, G., Genin, J.M.R., 2006. In situ redox flexibility of FeIIeIII oxyhydroxycarbonate

w a t e r r e s e a r c h 6 2 ( 2 0 1 4 ) 2 9 e3 9

green rust and fougerite. Environ. Sci. Technol. 40 (15), 4696e4702. Schonerklee, M., Koch, F., Perfler, R., Haberl, R., Laber, J., 1997. Tertiary treatment in a vertical flow reed bed system e a full scale pilot plant for 200e600 P.E. Water Sci. Technol. 35 (5), 223e230. Showers, W.J., Genna, B., McDade, T., Bolich, R., Fountain, J.C., 2008. Nitrate contamination in groundwater on an urbanized dairy farm. Environ. Sci. Technol. 42 (13), 4683e4688. Smith, C.N., Payne, W.R., Pope, J.D., Winkie, J.H., Parrish, R.S., 1999. A field study to compare performance of stainless steel research monitoring wells with existing on-farm drinking water wells in measuring pesticide and nitrate concentrations. Chemosphere 38 (4), 875e889. Talhi, B., Monette, F., Azzouz, A., 2011. Effective and selective nitrate electroreduction into nitrogen through synergistic parameter interactions. Electrochimica Acta 58, 276e284.

39

Viollier, E., Inglett, P.W., Hunter, K., Roychoudhury, A.N., Van Cappellen, P., 2000. The ferrozine method revisited: Fe(II)/ Fe(III) determination in natural waters. Appl. Geochem. 15 (6), 785e790. Xu, J., Hao, Z.W., Xie, C.S., Lv, X.S., Yang, Y.P., Xu, X.H., 2012. Promotion effect of Fe2þ and Fe3O4 on nitrate reduction using zero-valent iron. Desalination 284, 9e13. Zhang, J.H., Hao, Z.W., Zhang, Z., Yang, Y.P., Xu, X.H., 2010. Kinetics of nitrate reductive denitrification by nanoscale zero-valent iron. Process Saf. Environ. Prot. 88 (6), 439e445. Zhang, Y., Li, Y.M., Li, J.F., Hu, L.J., Zheng, X.M., 2011. Enhanced removal of nitrate by a novel composite: nanoscale zero valent iron supported on pillared clay. Chem. Eng. J. 171 (2), 526e531.