Nanofiltration for the Treatment of Coke Plant Ammoniacal Wastewaters

Separation and Puriﬁcation Technology 76 (2011) 303–307

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

Contents lists available at ScienceDirect provided by Elsevier - Publisher Connector

Separation and Puriﬁcation Technology

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

Nanoﬁltration for the treatment of coke plant ammoniacal wastewaters

Christa Korzenowski a,b, Miguel Minhalma a,c, Andréa M. Bernardes b, Jane Zoppas Ferreira b, Maria Norberta de Pinho a,∗

a Instituto Superior Técnico, ICEMS, Universidade Técnica de Lisboa (IST/ICEMS/UTL), Lisbon, Portugal b Programa de Pós-Graduac¸ão em Engenharia de Minas, Metalúrgica e de Materiais (PPGEM), Universidade federal do Rio Grande do Sul (UFRGS), Porto Alegre, RS, Brazil c Instituto Superior de Engenharia de Lisboa (ISEL), Lisbon, Portugal

article info abstract

Article history: This work addresses the treatment by nanoﬁltration (NF) of solutions containing NaCN and NH4Cl at Received 20 July 2010 various pH values. The NF experiments are carried out in a Lab-Unit equipped with NF-270 mem- Received in revised form 25 October 2010 branes for model solutions that are surrogates of industrial ammoniacal wastewaters generated in the Accepted 26 October 2010 coke-making processes. The applied pressure is 30 bar. The main objective is the separation of the

compounds NaCN and NH4Cl and the optimization of this separation as a function of the pH. Mem- Keywords: brane performance is highly dependent on solution composition and characteristics, namely on the pH. Coke plant wastewaters In fact, the rejection coefficients for the binary model solution containing sodium cyanide are always Ammonium Cyanides higher than the rejections coefficients for the ammonium chloride model solution. For ternary solutions Nanofiltration (cyanide/ammonium/water) it was observed that for pH values lower than 9 the rejection coefficients to Fractionation ammonium are well above the ones observed for the cyanides, but for pH values higher than 9.5 there is a drastic decrease in the ammonium rejection coefficients with the increase of the pH. These results take into account the changes that occur in solution, namely, the solute species that are predominant, with the increase of the pH. The fluxes of the model solutions decreased with increased pH. © 2010 Elsevier B.V. Open access under the Elsevier OA license.

1. Introduction eral authors reported difficulties in obtaining complete oxidation of NH3 in this kind of wastewaters. It seemed that nitrifying bacteria In the past years special attention has been given to industrial are inhibited by high cyanide concentrations [4]. The introduction ammoniacal wastewaters generated in the coke-making processes. of such wastewaters to an activated sludge system may result in Wastewaters containing ammonia are generated on coke-making loss of activated sludge viability and may change sludge commu- processes, since ammonium is removed from the exhaust gas by nity structure. Stringent limits for coke oven effluents have been adsorption onto water in order to reduce its concentrations to set in countries where these wastewaters represent a major prob- acceptable levels in the gas outlet of the plant, i.e. to around lem, such as Germany, and the corresponding guidelines include 0.1gm−3. In a typical coke plant, the wastewaters from the gas- the requirements for coke oven effluents before being discharged washing towers and the condensed waters from the coke oven have to a water receiver and prior to mixing with other effluents. Very high ammonia concentration, and besides that they have toxic and low concentrations are required for certain pollutants like cyanides carcinogenic substances, such as phenol, cyanide and thiocyanate, (<0.03 mg L−1) [5]. that are biologically refractory organic compounds and considered In comparison with the large amount of literature regarding priority pollutants [1–3]. The industry uses different methods for conventional treatments for ammoniacal wastewaters contami- ammonium removal and the choice is highly dependent on its con- nated with priority pollutants like cyanides, the literature is very centration and type of contaminants. The more frequently used scarce regarding to non-conventional and more adequate treat- ones are the biological treatments that generally involve nitrifi- ments for ammoniacal wastewaters contaminated with priority cation/denitrification processes. Various bacterial strains are able pollutants [6–10]. to perform cyanide, thiocyanate and NH3 oxidation. However, sev- The pressure-driven membrane separation processes (micro- filtration, ultrafiltration, nanofiltration and reverse osmosis) and namely nanofiltration (NF) have been playing a major role in the ∗ Corresponding author at: Department of Chemical and Biological Engineering, development of advanced wastewater treatments. In fact NF has Instituto Superior Técnico, ICEMS, Universidade Técnica de Lisboa (IST/ICEMS/UTL), the unique feature of selective permeation to target components Av. Rovisco Pais, 1049-001 Lisbon, Portugal. Tel.: +351 218417488; of multi-component solutions, making possible the confinement in fax: +351 218499242. the permeate stream of pollutants like the cyanides and therefore E-mail address: [email protected] (M.N. de Pinho).

Table 1 Membrane characteristics of NF270.

Composition top layer Polyamide MWCO (Da) 270 Roughnessa (A)˚ 28 Contact angle (◦)27 Zeta potential (mV) pH 7 −19 pH 10 −24

a Roughness measured with non-contact mode AFM on a scan area of 1 ␮m × 1 ␮m.

where B is a parameter characteristic of a given membrane/solute system and is defined by B = DAm˚/L. In steady-state the flux of solute A trough the membrane is also Fig. 1. Concentration profiles of solute A in a nanofiltration membrane and in the given by: fluid phases adjacent to the membrane. Where L is the membrane thickness, C , Ab J = J C CAm and CAp are the concentrations of solute A in the feed, at the membrane surface A p Ap (4) (fluid phase side) and in the permeate and CAm and CAp are the concentrations of solute A inside the membrane at the feed and permeate side, respectively. where Jp is the permeate flux. Combining Eqs. (3) and (4) and using the definition of intrinsic decontaminating the ammoniacal feed solutions [11,12]. This fea- rejection, the following equation is obtained: ture arises from the fact that NF is ruled not only by sieving J f = p (5) mechanisms but also by electrostatic interactions. Jp + B The main objective of this study is the investigation of nanofiltration capability of fractionating ammonium/cyanide containing 3. Materials and methods solutions, with respect to the confinement of the ammonium and cyanide in two separate streams that can be object of more specific 3.1. Model solutions treatments. The goal is to investigate the influence of the pH solution on the performance of the nanofiltration of coke plant effluents Binary aqueous model solutions containing CN− (MS1) and + namely the rejection coefficients to ammonium and cyanide ions. NH4 (MS2) in a concentration similar to the ones founded on coke For that, NF permeation experiments are carried out for wastewaters were used. Aqueous ternary solutions of ammonium aqueous binary solutions of ammonium and of cyanide (ammo- and cyanide were also prepared (MS3). Model solution 1 contains −1 −1 nium/water and cyanide/water) and for aqueous ternary solutions 0.18 g L of NaCN, model solution 2 contains 18.8 g L of NH4Cl of ammonium and cyanide (ammonium/cyanide/water). The NF and model solution 3 contains 0.18 g L−1 of NaCN and 18.8 g L−1 of experiments are carried in a pH range from 7 to 11. The comparison NH4Cl. The pH was varied between 8 and 11 and sodium hydroxide between binary and ternary solutions results will show if there is solutions were used for pH adjustment. The reagents used were of a change in the solute/membrane interactions when the different analytical purity and the water was deionized. solutes are mixed together as is the case in the real effluent solutions. The solute/membrane interactions may be described by the 3.2. Membrane solution/diffusion model, where a B parameter is used to quantify the affinity of a given solute to the membrane. The nanofiltration membrane selected was the NF270, supplied by Filmtec Corp., Minneapolis, MN (USA). It is a polyamide thin- 2. Theory film composite membrane and was characterized in terms of the hydraulic permeability, Lp, and in terms of rejection coefficients to The diffusive permeation of a component A through a NF mem- reference solutes – NaCl and Na2SO4. These experiments were car- brane (represented in Fig. 1) is described by Fick equation: ried out at a transmembrane pressure of 30 bar, 25 ◦C, a Feed flow − ∂C rate of 9.2 L min 1. Some of the characteristics of this membrane J =−D A (1) A Am ∂x are summarized in Table 1 [15]. where J is the diffusive flux of solute A, D is the solute diffusion A Am 3.3. Nanofiltration permeation experiments coefficient inside the membrane, CA is the solute concentration and x is the distance inside the membrane. The NF experiments were performed in a nanofiltration plant The relationship between solute A concentrations in the fluid (DSS Lab-Unit M20) with 0.072 m2 of membrane surface area. phase adjacent to the membrane and in the membrane side is given Membrane conditioning was carried out through the circula- by the partition coefficient defined as: − tion of deionized water (conductivity <1 ␮Scm 1) pressurized at C 30 bar for 2 h. This avoids pressure effects on membrane structure ˚ = A (2) CA in subsequent experiments. The nanofiltration experiments were carried out in total recircu- The integration of Eq. (1) along the membrane, considering the lation mode, where the permeate and the concentrate streams were following boundary conditions: recirculated to the feed tank. Permeation experiments of aqueous x = 0 CAm = CAm model solutions of sodium cyanide and ammonium chloride were x = LCAp = CAp carried out in order to study the variation of permeate fluxes and of the solute rejection coefficients with the pH. The operating condi- results in [13,14]: tions were: temperature 25 ◦C, transmembrane pressure of 30 bar −1 D ˚ and feed circulation flow rate of 9.2 L min . The initial volume of J = Am (C − C ) = B (C − C ) (3) A L Am Ap Am Ap feed solution for all experiments was 5 L. C. Korzenowski et al. / Separation and Purification Technology 76 (2011) 303–307 305

Fig. 2. Hydraulic Permeability for NF270 membrane. Transmembrane pressures = 10–30 bar. Feed ﬂow rate = 9.2 L min−1. Membrane surface area: 0.072 m2. Fig. 3. Possible species formed in the model solution 3. Temperature: 25 ◦C.

The results relative to the variation of the permeate fluxes of The rejection coefficients to a particular solute were determined model solutions 1–3 with pH are displayed in Fig. 4. by Eq. (6), where CA feed and CA permeate are the solute concentration According to the results it can be seen that with the increase of in the feed and in the permeate, respectively. the pH of the solution the permeate flux decreases for all model C − C solutions. It is also observed that the permeate fluxes for the solu- A feed A permeate fA = (6) C tions containing ammonium are lower than the one observed for A feed the model solution containing only the cyanides, this is due to the 3.4. Analytical methods higher concentration of ammonium in the model solutions that translates into a higher value of the osmotic pressure, thus leading The solutions were analyzed for different parameters: pH, to lower permeate fluxes. conductivity, cyanide and ammonia concentrations. pH was deter- The results relative to the variation of the rejection coefficients mined by a pH meter (Crison 202), conductivity was determined to ammonium and cyanide ions as a function of pH in the experi- by a conductivimeter (Crison 525). The ammonia concentration ments with model solutions 1–3 are displayed in Fig. 5. was determined by a spectrophotometric method (UV-1700 Phar- The fractionation of cyanides/ammonium solutions is extremely maSpec, Shimadu) [16], and cyanide concentration was determined dependent on solution chemistry, which in turn is a function of the pH. In fact, the binary solutions NF experiments (Fig. 5, left) by titration with AgNO3 [17]. have shown that the rejections coefficients for the model solution 1 (sodium cyanide) are always higher than the rejections coeffi- 4. Results and discussion cients for the model solution 2 (ammonium chloride), and that in both cases there is an increase in the rejection coefficients with the The membrane was characterized in terms of hydraulic perme- increase of the pH, ranging from 40 to 75% for MS1 and 28 to 34% for ability (L ). Pure water permeation flux (PWP) was measured at p MS2. It is also observed that the variation is much more pronounced transmembrane pressures (P) of 10, 15, 20, 25 and 30 bar. The to cyanides than to ammonium. These results are due to the fact membrane hydraulic permeability (L ) was obtained from the slope p that for higher pH values the HCN/CN− equilibrium (see Fig. 3)is of the straight line, PWP versus P. The result for the membrane displaced towards the CN− form, and therefore because the mem- used in this study is 10.251 kg h−1 m−2 bar−1 (Fig. 2). brane is negatively charged there is an increase in cyanide repulsion The rejection coefficients to reference solutes – NaCl and Na SO 2 4 leading to higher rejection coefficients. Regarding the NH +/NH – are presented in Table 2. 4 3 equilibrium, the increase of pH leads to a displacement towards Using the Hydra Medusa program [16] it was possible to see the predominant species in solution as a function of feed solution pH. These species are shown in Fig. 3. The figure shows that it is not expected the formation of complexes in the model solution 3 and only anions, neutral molecule and cations are present. It is also observed that for pH values around 9.0–9.5, there is a change in the − + predominant species in solution for the CN /HCN and NH4 /NH3 + pairs. In fact, for values lower than 9 the HCN and NH4 are the predominant species in solution, while for values higher than 9,5 − there is a shift and the CN and NH3 forms become predominant.

Table 2

Rejection coefﬁcients to reference solutes – NaCl and Na2SO4 for the NF270 membrane. Transmembrane pressure = 30 bar. Feed ﬂow rate = 9.2 L min−1. Membrane surface area: 0.072 m2. Temperature: 25◦C.

Solute Feed concentration (mg/L) f (%)

NaCl 2000 84 Na2SO4 2000 98 Fig. 4. Permeate ﬂux variation as a function of pH for model solutions 1–3. 306 C. Korzenowski et al. / Separation and Puriﬁcation Technology 76 (2011) 303–307

Fig. 5. Rejection coefﬁcients to ammonium and cyanide ions as a function of pH, in the binary (left) and ternary (right) model solutions.

Fig. 6. Variation of B parameter with the pH for NaCN in binary (MS1) and ternary Fig. 7. Variation of B parameter with the pH for NH4Cl in binary (MS2) and ternary (MS3) solutions. (MS3) solutions.

solution, there is a drastic variation of the ammonium B parameter the NH3 form (see Fig. 3), which have a slightly lower affinity to + as a function of the pH and for pH lower than 9 its value is very the membrane than the NH4 form and therefore there is a slight increase in its rejection coefficient. low and then increases strongly for pH values higher than 9.5. This Regarding the ternary model solution (Fig. 5, right) the varia- means that the affinity of NH3 towards the membrane increases tion of the rejection coefficients to cyanide and to ammonium with rapidly with pH. The analysis of these results shows that depend- the pH, present a totally different behavior than the one observed ing on which solute one wants to concentrate in the concentrate for the binary model solutions. In fact, for pH values lower than 9, stream a special attention has to be taken in terms of the solu- the rejection coefficient to ammonium is higher than the one to tion pH. The predominant species in solution depends greatly on cyanide, around 75% and 45%, respectively. For pH values higher the pH value, because it dictates how the solutes interact with the than 9.5, there is a drastic reduction on the rejection coefficient membrane and other solutes leading to higher or lower rejection to ammonium, which drops to nearly zero, while there is only a coefficients, and more precisely on the fact of the pH being higher slight decrease on the cyanide rejection coefficient. This behavior or lower than 9. is correlated to the change in the predominant species present in solution, as depicted in Fig. 3, which leads to a totally different set 5. Conclusions of interactions between the solutes and the membrane. For higher pH values there is an increase in solution of the negatively charged The treatment of coke plant wastewaters presents a major − CN anions and also an increase of the NH3 form, and for this par- problem as there are different prioritary pollutants, such as ticular set of conditions there is a shift of the affinity of NH3 towards ammonium and cyanides, mixed together that render the specific the membrane and this leads to a drastic decrease in its rejec- solute treatments inefficient. The fractionation by nanofiltration of tion coefficient, and therefore leads to cyanides enrichment and cyanides/ammonium solutions is extremely dependent on solution ammonium depletion in the feed stream. In order to quantify this chemistry, which in turn is a function of the pH. In fact, the binary change in terms of solute/membrane affinity, the solution/diffusion solutions NF experiments have shown that the rejections coeffi- model was applied (Eq. (5)) and the B parameter was calculated for cients for the model solution 1 (sodium cyanide) are always higher the cyanide and ammonium for both binary and ternary solutions, than the rejections coefficients for the model solution 2 (ammo- these results are presented in Figs. 6 and 7. Because the operat- nium chloride), and that in both cases there is an increase in the ing conditions were set in order to minimize the concentration rejection coefficients with the increase of the pH, ranging from 40 polarization phenomena the apparent rejection coefficients were to 75% for MS1 and 28 to 34% for MS2. These results would lead considered to be equal to the intrinsic rejection coefficients. to expect that cyanide retention and ammonium depletion in the The results presented in Figs. 6 and 7 show that for the binary concentrate would be feasible, but the ternary solutions NF exper- solutions there is a decrease of the B parameter with the pH iments have shown that the ions rejection behavior may differ in a increase, for both cyanides and ammonium. For the ternary model very drastic way from the results obtained for the binary solutions. C. Korzenowski et al. / Separation and Purification Technology 76 (2011) 303–307 307

In the ternary solutions, for pH values lower than 9 the rejection [5] C.A. Papadimitriou, P. Sâmaras, G.P. Sakellaropoulos, Comparative study of coefficient to ammonium chloride is 75% and to sodium cyanide phenol and cyanide containing wastewater in CSTR and SBR activated sludge reactors, Bioresour. Technol. 100 (1) (2009) 31–37. is 45%. It means that the ammonium chloride is mainly in the [6] R. Marr, M. Koucar, Recovery of ammonia from industrial wastewater, Int. concentrate stream and the sodium cyanide is mainly in the per- Chem. Eng. 33 (3) (1993) 416–425. meate stream. Therefore one can envisages, for example, the use [7] C. Ratanatamskul, C. Chiemchaisri, K. Yamamoto, The use of zeolite-iron column for residual ammonia and phosphorus removal in the effluent from a mem- of a stripping operation to process the concentrate stream and an brane process as an on-site small-scale domestic wastewater treatment, Water advanced oxidation process to destroy the cyanide salt in the per- Sci. Technol. 31 (9) (1995) 145–152. meate stream. At pH higher than 9.5 due to the low rejections to [8] I. Vázquez, J. Rodríguez, E. Maranón,˜ L. Castrillón, Y. Fernández, Simultane- both salts the use of specific treatments is not feasible technically. ous removal of phenol, ammonium and thiocyanate from coke wastewater by aerobic biodegradation, J. Hazard. Mater. B 137 (2006) 1773–1780. [9] B. Beler-Baykal, M. Oldenburg, I. Sekoulov, The use of ion exchange in ammo- Acknowledgements nia removal under constant and variable loads, Environ. Technol. 17 (1996) 717–726. [10] S.H. Lin, Y.L. Yen, Ammonia and nitrate removal from sea water by ozonation, The authors thank CAPES – Brazil and FCT – Portugal for financial Environ. Technol. 18 (1997) 65–73. support. Support from project REEQ/764/EQU/2005 is gratefully [11] M. Minhalma, M.N. de Pinho, Integration of nanofiltration/steam stripping for the treatment of coke plant ammoniacal wastewaters, J. Membr. Sci. 242 (2004) acknowledged. 87–95. [12] M. Minhalma, M.N. de Pinho, Development of nanofiltration/steam stripping for the treatment of coke plant ammoniacal wastewaters, Desalination 149 References (2002) 95–100. [13] H.K. Lonsdale, C.E. Milstead, B.P. Cross, F.M. Graber, J.A. Hunter, W.S. Gilam, H.E. [1] M.K. Ghose, Complete physico-chemical treatment for coke plant effluents, Podall, L.M. Kindley, Study of rejection of various solutes by RO membranes, Water Res. 36 (2002) 1127–1134. US Office of Saline Water Research and Development Progress Report #477, [2] M. Minhalma, Synthesis and optimization of processes for the recovery of 1969. industrial wastewaters with ultrafiltration and nanofiltration. PhD Disserta- [14] R. Rautenbach, R. Albrecht, Membrane Processes, John Wiley & Sons, UK, 1989. tion, IST, Technical University of Lisbon, 2001. [15] K. Boussu, C. Kindts, C. Vandescateele, B. Van der Bruggen, Applicability of [3] A. Rancano,˜ Biological treatment of coke plant wastewaters. PhD Dissertation, nanofiltration in the carwash industry, Sep. Purif. Technol. 54 (2007) 139–146. University of Oviedo, 2000. [16] I. Puigdomenech, Hydra Medusa: Make Equilibrium Diagrams Using Sophisti- [4] M. Lay-Son, C. Drakides, New approach to optimize operational conditions for cated Algorithms, Royal Institute of Technology, Stockholm, Sweden, 2001. the biological treatment of a high-strength thiocyanate and ammonium waste: [17] American Public Health Association, Standard Methods for the Examination of pH as key factor, Water Res. 42 (3) (2008) 774–780. Water and Wastewater, 16 ed., 1995, Washington.