Big Carrousel mechanism of copper removal from ammoniacal wastewater through supported liquid membrane

Separation and Purification Technology 54 (2007) 104–116

Big Carrousel mechanism of copper removal from ammoniacal wastewater through supported liquid membrane N.M. Kocherginsky a,b , Qian Yang a,∗ a

Department of Chemical and Biomolecular Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore b Division of Bioengineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore Received 6 June 2006; received in revised form 16 August 2006; accepted 17 August 2006

Abstract Although “Big Carrousel” mechanism was suggested long time ago [1], there was no experimental work demonstrating its applicability. In this work, copper recovery from industrial ammoniacal wastewater using flat supported liquid membranes (SLM) was chosen as one of the most well known and practically important examples to demonstrate the limitations of “Small Carrousel” mechanism to describe facilitated ion transport through SLM. In this work, LIX54, one of well-established extractants for copper, was used as a carrier in the liquid membrane phase to extract and transfer copper. The dependences of the copper transmembrane flux on carrier concentrations may be explained by the “Small Carrousel” model: reaction of the carrier and copper occurs on the membrane/water interface and LIX54 stays in the membrane, carrying Cu ions in the membrane in one direction and two protons in another direction. Nevertheless, a quantitative explanation of experimental copper transmembrane flux as a function of feed pH and initial copper concentration cannot be based on the same values of physico-chemical parameters determined by simulation. A more advanced model is required. A detailed theoretical model – referred to as “Big Carrousel” – for facilitated transport through flat membranes is developed. Both diffusion of a copper complex with ammonia in an aqueous stagnant layer and fast reactions of the carrier and copper species in the aqueous reaction layer are accounted for in this model. This model, in which the carrier moves slightly out from the organic membrane in the aqueous reaction layer and then transfers from one aqueous phase to another through the membrane before finally moving back, is called “Big Carrousel”. Both reactions of LIX54 in the aqueous phase and the fact that Cu ions form complexes with ammonia have not been considered in previously published papers, describing the transport of simple Cu ions from ammoniacal aqueous solutions through liquid membranes. Mathematical model simulation demonstrated that only “Big Carrousel” model gives satisfactory quantitative description of all experimental results. Finally, high selectivity for Cu over other cations and long-term stability in a hollow fiber supported liquid membrane system for ammoniacal wastewater treatment make the SLM technology promising for practical industrial applications. © 2006 Elsevier B.V. All rights reserved. Keywords: Ammoniacal wastewater; Copper removal; Supported liquid membrane; Small Carrousel mechanism; Big Carrousel mechanism

1. Introduction Compared with conventional separation processes such as solvent extraction, supported liquid membranes (SLM) have significant advantages, including the possibility to conduct both processes of extraction and re-extraction in one technological step, relatively high membrane surface area per unit volume, small volume of organic phase and low organic species (extractant and solvent) usage, etc. Recent studies have demon-

∗

Corresponding author. Tel.: +65 65166433; fax: +65 67791936. E-mail address: [email protected] (Q. Yang).

1383-5866/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.seppur.2006.08.019

strated that SLM may be an effective tool for metal ions reclamation [2,3], pharmaceutical product recovery [4], chiral separation [5], organics removal [6], sea water desalination [7], analytical instrumentation [8,9], etc. Feasibility studies of copper separation from aqueous solutions by SLM have also been conducted worldwide [10–12], which is determined both by simplicity of the process and its practical potential and importance for microelectronic industry, generating a lot of wastewater and spent etching solutions with high copper content. Spent etching solution with a high content of both ammonia and copper is formed after printed circuit board (PCB) etching with ammoniacal etching solutions. The spent solution is then

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

Nomenclature HR CuR2 D J k Kd Kex,f Kex,s KF Kf Ks mCuR2 mHR Sm V

extractant/carrier LIX54 copper–carrier complex diffusion coefficient transmembrane flux overall mass transfer coefficient dissociation constant of NH4 + effective extraction equilibrium constant for the Cu(NH3 )4 2+ /LIX54 system extraction equilibrium constant for the Cu2+ /LIX54 system equilibrium constant for complex formation in the Cu(NH3 )4 2+ /LIX54 system equilibrium constant for complex formation in the Cu2+ /LIX54 system stability constant for copper ammonia complex effective distribution coefficient of copper–carrier complex effective distribution coefficient of the carrier membrane surface area volume

Greek letters δ thickness ε porosity τ tortuosity Subscripts a aqueous phase f feed phase i inter phase m membrane phase r reaction zone s strip phase

treated with an acid (usually HCl) so that the main part of copper precipitates as copper oxide and hydroxide. After this treatment the spent solution still contains some copper. This solution and also solution after washing of PCBs form the industrial ammoniacal wastewater. Typically, the copper concentration in this wastewater is around several hundred ppm, but it must to be reduced to less than 5 ppm in order to be reused or discharged safely. Removal of copper ions through SLM is made possible due to the chemical potential difference between H+ ions in the feed and strip solutions. The initial pH of the feed copper solution is less acidic and its acidity decreases with time due to electroneutral 2H+ /Cu2+ exchange through the liquid membrane. The disadvantage of this process is that copper transfer stops because the chemical potential difference of H+ ions vanishes. In this process the pH value in the feed becomes more acidic and final Cu2+ concentration in this solution is not low enough because at equilibrium chemical potential difference of copper ions is balanced by chemical potential difference of H+ ions.

105

Earlier we have demonstrated that during treatment of copper-containing ammoniacal solutions the pH of the feed – in contrast to usual aqueous solutions with hydrated Cu2+ ions – does not decrease. Instead, it stays approximately constant due to the buffer effect of NH3 /NH4 + [13]. As the result it was not necessary to adjust the feed pH in order to keep the large H+ gradient, and the process could be conducted until a very low copper content in the treated water was reached. Comparison of these two processes is given in Fig. 1. Facilitated ion transport through SLMs is commonly described based on the idea that ions in aqueous solutions are reacting through the interface with the chelating agent (carrier) located in the organic membrane phase. The simple physicochemical mechanism of the reaction between the two species located in different phases and separated by liquid/liquid interface remains behind the scenes. In this paper we reveal that this simple mechanism cannot systematically explain the totality of experimental results. A more detailed and physically reasonable mechanism of copper removal from typical ammoniacal waste aqueous solutions is suggested based on the idea that chelating agent is able to diffuse from the organic into the aqueous phase where ion exchange reactions are taking place. The mechanism by which the carrier moves from one aqueous phase through the membrane into another before returning and completing the cycle is called the “Big Carrousel”. The simpler mechanism by which the carrier stays in the membrane and two ion exchange reactions take place on the water/membrane interfaces corresponds to the term of “Small Carrousel” [1]. In this work, copper recovery from industrial ammoniacal wastewater using flat supported liquid membranes (SLM) was chosen as one of the most well known and practically important examples to demonstrate the applicability of “Big Carrousel” mechanism to describe facilitated ion transport through SLM. Mathematical model simulations demonstrated that only “Big Carrousel” model gives satisfactory quantitative description of all the experimental dependences of the copper transmembrane flux on carrier concentration, initial feed copper concentration and feed pH. Finally, an effective hollow fiber supported liquid membrane system for copper removal from ammoniacal wastewater was investigated. High selectivity for Cu(II) over other cations and its long-term stability in the ammoniacal wastewater treatment process demonstrate that the system has practical potential for industrial applications and can be important for PCB production. 2. Experimental 2.1. Reagents Sulfuric acid, hydrochloric acid and sodium hydroxide (Merck, USA) were all of Reagent grade. LIX54 was supplied by Cognis Corporation, USA and was diluted in kerosene (Aldrich, USA) as a carrier in liquid membrane phase. Unless otherwise specified, the stripping phase used was 2 M sulfuric acid. All chemicals were used as received without any further purification. Light blue ammoniacal wastewater (0.02 M Cu(II), 5.5 mM Zn(II), 0.37 mM Ni(II), 1.8 ␮M Cd(II), total NH3 around 0.4 M

106

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

Fig. 1. Kinetics of pH and copper concentration changes in feed solution. Strip: 2 M H2 SO4 ; organic membrane phase: LIX54 33% (v/v) in kerosene. (A) Copper solution without ammonia; (B) copper-containing ammoniacal solution.

and pH above 7. Cl− concentration can be as high as 0.5 M) and dark blue spent ammoniacal etchant solution (Cu content up to 2.5 M with trace amount of other cations, total NH3 near 11 M and pH around 9.5), both without any precipitates, were kindly supplied by a PCB producing company in Singapore. 2.2. Experimental setup and SLM preparation 2.2.1. Flat sheet supported liquid membrane Effective membrane surface area (Sm ) in the horizontal flat membrane system (Fig. 2) was 10 cm2 . Volume of the strip solution was 20 ml, and that of feed solution was 100 ml. Both aqueous phases were mechanically stirred with magnetic stirrers at 250 rpm. The stirring of both aqueous solutions was aimed to reduce the mass transfer resistances in the aqueous un-stirring layers and also provide homogeneous environments in both feed and strip solutions when sampling. The polymer porous support used for horizontal flat membrane system was Fluoropore (PTFE) membrane filter (Millipore, USA) with an average thickness of 50 ␮m, 70% porosity and average pore size 0.2 ␮m. To prepare supported liquid membrane, the membrane filter was immersed into organic solution (LIX54 in kerosene under predetermined ratio). If the weight gain in membrane after

impregnation was close to what could be expected based on the known porosity and the density of the carrier, the membrane pores were considered to be fully impregnated by the organic carrier. After fully wetted by the organic solution, membrane filter was then taken out and extra oil on the membrane surface was removed by tissue paper. Then the impregnated filter was fixed in the membrane module. The experimentally determined copper transmembrane flux was calculated based on the Cu concentration increase in strip solution versus time: J=

dC V dt Sm

(1)

where dC is the change in Cu(II) concentration in the strip over time interval dt and V is the strip solution volume. 2.2.2. Hollow fiber supported liquid membrane A laboratory scale microporous polypropylene hollow fiber module (Liqui-Cel Extra-Flow 2.5 in. × 8 in. membrane contactor, Hoechst Celanese, USA) was employed to prepare the hollow fiber supported liquid membrane. Detailed specification of the hollow fiber membrane module was listed in Table 1. The liquid membrane phase (33% (v/v) LIX54 in kerosene in this case) was impregnated in fiber wall by pumping organic membrane phase through the contactor using a peristaltic pump

Table 1 Specifications of bench scale hollow fiber membrane module

Fig. 2. Horizontal flat membrane system.

Fiber type

Polypropylene

Number of fibers, N Fiber internal radius, ri (␮m) Fiber outer radius, ro (␮m) Effective module outer diameter, da (cm) Effective module inner diameter, di (cm) Effective pore size, rp (␮m) Porosity, ε (%) Tortuosity, T Effective fiber length, L (cm) Effective surface area, Sm (m2 )

9950 120 150 4.67 2.2 0.03 40 2.5 15 1.4

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

107

(Cole-Parmer, USA) in a recycling mode for about 1 h. The extra oil was washed out with de-ionized water. 2.3. Analytical methods Copper concentration was measured by an inductively coupled plasma spectroscopy (ICP-AES) (ICP Optima 3000, Perkin-Elmer, USA). In some cases copper solutions were diluted so as to coincide with the measuring range. Values of pH were measured with a Cyberscan 200 digital pH meter and a combined pH glass electrode. Total NH3 concentration (free NH3 together with Cu–NH3 complex) in the ammoniacal copper solution was determined by Kjeldahl method [14]. This method involves the separation of ammonia from the sample of interest through distillation. After the distillation procedure, the distilled ammonia would react with the predetermined concentration of sulfuric acid. The experiment was then followed by the titration of the sulfuric acid with standard concentration of sodium hydroxide to determine the amount of sulfuric acid reacted and to calculate the concentration of ammonia. Chloride concentration was measured with Cl− ion selective electrode (Metrohm, Singapore). The interfacial tension between ammoniacal wastewater and organic ´˚ membrane phase (LIX54 in kerosene) was measured by FTA125 (First Ten Angstroms, USA) using the drop shape method. 3. Results and discussion 3.1. The influence of carrier concentration on copper transmembrane flux Fig. 3 shows that the rate of copper transfer through the membrane increases gradually with increase of the carrier concentration, reaching a maximum at concentration above 20% (v/v) in kerosene. Further increase of the flux may be hindered by reduced copper–carrier diffusion coefficientDCuR2 ,m due to an increase of membrane viscosity [10]. In the subsequent experiments, 33% (v/v) LIX54 in kerosene was used as the organic membrane phase. A higher carrier inventory provided is aimed to

Fig. 4. Influence of feed copper concentration on copper flux. Strip: 2 M H2 SO4 ; organic membrane phase: LIX54 in kerosene 33% (v/v). (Simulation based on Eq. (23) and Eq. (34), simulation parameters are given in Table 3.)

improve the stability (lifetime) of liquid membrane which may be slightly dissolved out from the pores of membrane matrix in the process. 3.2. The influence of feed copper concentration on copper transmembrane flux Fig. 4 shows the relationship between copper flux and its concentration (range from 2.3 × 10−4 to 0.4 M) in the ammoniacal feed solution as well as its simulation (see later). The different concentrations of feed copper solutions were prepared by diluting spent ammoniacal etchant solution or ammoniacal wastewater and the pHs were kept at 7.25 ± 0.2. The experimental results (Table 2) show that when copper concentration is lower than 0.1 M, the flux is first order with respect to feed copper concentration and the rate limiting step is determined by the diffusion of copper ions in the stagnant aqueous layers [11,15,16]. When copper concentration is higher than 0.1 M, the flux reaches plateau. The carrier on the feed side of the membrane in this case is saturated with copper [10,15] and transport is limited by the diffusion of copper–carrier complexes through the membrane phase. Jmax,m and Jmax,a are the maximum possible copper fluxes through the membrane and stagnant aqueous layer with given copper concentration in the feed solution, respectively. They are defined as 0

Jmax,m =

DCuR2 ,m [HR]m δm 2

(2)

Table 2 Comparisons on J/Jmax,a and J/Jmax,m with different copper concentrations in the feed solution

Fig. 3. Copper transmembrane flux as a function of carrier concentration. Feed: ammoniacal wastewater with 0.021 M Cu(NH3 )4 2+ at pH ∼7; Strip: 2 M H2 SO4 ; organic membrane phase: LIX54 in kerosene with different volume ratios. (Simulation based on Eq. (23) and Eq. (34), simulation parameters are given in Table 3.)

Feed copper cone (M)

J/Jmax,a

2.30 × 10−4

1.00 1.00 0.99

2.30 × 10−2 5.00 × 10−2

1.00 × 10−1 2.00 × 10−1 4.00 × 10−1

J/Jmax,m

Rate-limiting step Diffusion in the stagnant layer of aqueous phase

0.98 0.99 0.99

Diffusion in membrane phase

108

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

Jmax,a =

DCu,a [Cu(NH3 )2+ 4 ]a δa

(3)

The maximum flux Jmax,m is 4.9 × 10−8 mol/cm2 s (Fig. 4). This value is determined by diffusion through the membrane. The membrane thickness in this case was 50 ␮m and the carrier concentration was 0.55 M as 33% (v/v) LIX54 in kerosene. From Eq. (2), it gives the value of diffusion coefficient for the copper–carrier complex DCuR2 ,m in the membrane phase near 8.9 × 10−7 cm2 /s. The DCuR2 ,m value calculated based on Wilke–Chang equation (Eq. (4)) for monomeric form is slightly higher (2.8 × 10−6 cm2 /s) [17]: 0 DAB =

7.4 × 10−8 × (MB )0.5 T η × VA0.6

(4)

0 is the diffusion coefficient of solute A (CuR ) in solwhere DAB 2 vent B (kerosene), MB the molecular weight of solvent kerosene (∼200 g/mole), T the temperature (299 K), η the viscosity of solution (2.28 cP), and VA (666.6 cm3 /mole) is the molar volume of solute (CuR2 ) at its normal boiling temperature according to LeBas method [18]. Taking into consideration the tortuosity 2.5 and porosity 70% of the membrane filter, the effective diffusion coefficient of the copper–carrier complex in the membrane phase is 1.0 × 10−6 cm2 /s, in consonance with the experimental values. Further we can assume that in stagnant aqueous solution diffusion coefficients of the copper–ammonia complex and the copper–water complex are approximately the same, and equal to 6.5 × 10−6 cm2 /s [6]. Based on the experimental mass transfer coefficient, equal to 8.37 × 10−4 cm/s (slope of the insert in Fig. 4) we can estimate the thickness of this stagnant aqueous layer δa equal to 77 ␮m, which is comparable to 110 ␮m at stirring 400 rpm [6] and 26 ␮m at stirring 1200 rpm [19].

3.3. The influence of pH in feed solution on copper transmembrane flux Equilibrium distribution of copper between aqueous solution with ammonia and organic phase with LIX54 depends on pH in the feed [20–24]. Copper extraction with LIX54 can reach almost 100% at pH near 8 but decreases at pH in acidic media [20,24,25]. Fig. 5 demonstrates qualitatively similar behavior for transmembrane Cu transport from the ammoniacal wastewater with pH adjusted by HCl and ammonia aqueous solution. Below we will develop theoretical models for describing the copper facilitated transport through flat membrane based on the experimental dependences of copper transmembrane flux on carrier concentrations, pH and initial copper concentrations in feed solution. Our experimental values of diffusion coefficient in the membrane and thickness of aqueous unstirred layer are used later in detailed simulation of all results.

Fig. 5. Copper flux as a function of pH in the feed. Feed: ammoniacal etching wastewater with 0.021M Cu(NH3 )4 2+ at different pH. Strip: 2 M H2 SO4 ; Organic membrane phase: LIX54 in kerosene 33% (v/v). (Simulation was based on Eq. (23), (24) and Eq. (34), 35, simulation parameters are given in Table 3.)

3.4. Modeling of mass transfer of copper species through SLM 3.4.1. Description of transmembrane Cu transport based on facilitated “Small Carrousel” mechanism It is usually assumed that the heterogeneous reaction of an ion with a carrier takes place only on/at the membrane surface. The ion-carrier complex moves through the membrane, and then participates in the ion-exchange reaction with H+ on another surface. Finally the protonated carrier comes back. This process of facilitated counter-transport is based on the idea that the carrier and metal ion are in different phases but still they react. This simplified model may be termed a “Small Carrousel”. It is different in comparison to another mechanism by which the carriers exit the membrane, react with ions in an aqueous solution and then transport back into and through the membrane. Same sequence of elementary steps takes place near the opposite interface, and the whole process is called “Big Carrousel” [1]. To model the mass transfer of copper in the “Small Carrousel” process, we as usual assume that the system is in a steady state and it means that: 1. Linear concentration gradients of the copper–ammonia complexes exist through an aqueous stagnant layer and the copper–carrier complexes through the membrane phase (Fig. 6), respectively [11,22,26,27]. 2. Chemical reaction with the carrier at the aqueous/membrane interface is fast in comparison to the diffusion process, so that it is possible to assume the existence of extraction equilibrium at the interface [10]. 3. H+ ions transferred from the strip to the feed solution are bound by ammonia and as the result feed pH stays approximately constant in the process (Fig. 1B). Specifics of the system described in this paper is that copper ions in the feed ammoniacal solutions are in a complex with ammonia molecules, while in the strip phase they are in a complex with water. This asymmetry of the system was not considered in the previous papers on membrane transport, but because of this asymmetry it is possible to conduct the process and decrease Cu content in the feed practically to zero (Fig. 1B).

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

109

2. Copper extraction with LIX54 at the feed/membrane interface: + Cu(NH3 )2+ 4 + 2HR ⇔ CuR2 + 2NH4 + 2NH3 ,

[CuR2 ]fi · [NH3 ]2f · [NH+ 4 ]f

2

Kex,f =

2

[Cu(NH3 )2+ 4 ]fi · [HR]fi

(6)

where Kex,f (M2 ) represents the effective equilibrium constant for the extraction of copper from tetra ammonia complex at the feed/membrane interface; overbar indicates species in the organic phase. Equilibrium between NH3 and NH4 + is described by the following dissociation constant Kd (M) [31]: Fig. 6. Schematic description of copper transport through SLM with “Small Carrousel” model.

To simplify equations we assume that diffusion coefficients of the copper–ammonia complex and the hydrated copper ions are approximately the same, and equal to DCu,a . It is well known that Cu2+ in aqueous solutions with ammonia is forming octahedral complexes from Cu(NH3 )(H2 O)5 2+ to Cu(NH3 )4 (H2 O)2 2+ and the fifth and sixth ammonia do not bind strongly with Cu2+ because of Jahn–Teller effect [28,29]. It was demonstrated that Cu complex with four NH3 is the predominate form in ammoniacal solutions at copper concentration near 0.01 M and ammonia above 0.3 M at pH 8.9 [24]. This complex is characterized by high value of stability constant Ks of copper–tetra ammonia complex (∼1012.46 M−4 ) [30] and copper loading from ammoniacal solutions by the organic phase with LIX54 involves Cu(NH3 )4 2+ . Additional equilibrium processes with the formation of a complex with three ammonia molecules at pH near 7 and also different mixed copper complexes with ammonia, OH− and Cl− are possible at lower pH. It is important that these complexes are much less stable than Cu(NH3 )4 2+ and are at equilibrium with it. If we accept that LIX54 reacts mainly with Cu(NH3 )4 2+ , it would mean that other complexes are transformed in Cu(NH3 )4 2+ and thus do not change the total kinetics. Further for the sake of simplicity and also because the extraction constants of Cu(II) with LIX54 are known only for Cu(NH3 )4 2+ and Cu2+ species, we will present the mechanism based on the existence of Cu(NH3 )4 2+ species in the ammoniacal feed and Cu2+ in the highly acidic strip solution. This simplification is certainly correct at pH above 8, where Cu(NH3 )4 2+ becomes the dominant species. In this case we can suggest the following description: 1. The flux of copper in the stagnant layer of aqueous feed solution is DCu,a 2+ J1 = ([Cu(NH3 )2+ 4 ]f − [Cu(NH3 )4 ]fi ) δa

(5)

where DCu,a is the diffusion coefficient of copper–ammonia complex in the aqueous solution; subscript a indicates species in the aqueous phase; subscripts f and fi indicate species in the bulk feed solution and near the feed/membrane interface; δa is the effective thickness of aqueous stagnant layer.

Kd =

[NH3 ][H+ ] = 5.56 × 10−10 [NH+ ] 4

(7)

The stability constant Ks (M−4 ) of copper–tetra ammonia complex is [30]: Ks =

[Cu(NH3 )4 2+ ] = 10+12.46 [Cu2+ ][NH3 ]4

(8)

The values of stability constants for Cu(II) complexes with one, two and three ammonia are 1.9 × 104 , 7.4 × 107 and 7.41 × 1010 , respectively [30]. Earlier Lazarova has studied the kinetics of copper extraction with LIX54 from aqueous ammonia solutions using a rotating diffusion cell and published the value of the extraction equilibrium constant equal to 8 × 10−10 [22]. Alguacil et al. also studied copper recovery from ammoniacal/ammonium sulphate medium by LIX54, but according to this group extraction equilibrium constant was 7 × 10−7 [20]. In both papers copper extraction from the ammoniacal solution was described based on simple extraction of copper ions:  Kex,f =

[CuR2 ] · [H+ ] [Cu2+ ][HR]

2

(9)

2

Three orders of value difference in these two groups can be explained based on the fact that copper in the ammoniacal solutions reacts with LIX54 in the form of copper–tetra ammonia complex [CuR2 ]fi · [H+ ]f

2

Kex,f =

2

[Cu(NH3 )2+ 4 ]fi · [HR]fi

 = Kex,f ·

[NH3 ]4f Kd2

·

[NH3 ]4f Kd2 (10)

 is only an effective constant and its value Evidently, Kex,f should strongly depend on ammonia content. Recently Ismael et al. reported that the extraction constant for copper extraction from Cu(NH3 )4 2+ in ammoniacal media with LIX54 is 102.89 M2 . The nonideality of the aqueous phase was taken into consideration by applying the Pitzer model and the extended Debye–Huckel equation and the use of the thermodynamic equilibrium constant allowed the satisfactory

110

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

prediction of isotherms for copper extraction from ammoniacal media with LIX54 [24]. This more accurate value of extraction constant for Cu(NH3 )4 2+ /LIX54 system is used further for simulation. 3. The complex LIX54-copper formed at the feed/membrane interface diffuses through the membrane: DCuR2 ,m J2 = ([CuR2 ]fi − [CuR2 ]si ) δm

(11)

where DCuR2 ,m is the diffusion coefficient of copper–carrier complex in the membrane; subscript m indicates species in the membrane phase; δm is the thickness of liquid membrane and subscript si indicates species in the membrane near membrane/strip interface. 4. Copper ions are re-extracted at the interface membrane/acidic strip phase without ammonia: CuR2 + H2 SO4 ⇔ CuSO4 + 2HR, Kex,s =

[CuR2 ]si · [H+ ]s [Cu2+ ]si

2

2 · [HR]si

= Kex,f Ks Kd2

(12)

Substituting Eqs. (17) and (18) into Eq. (11), we finally have 2

4 + ([Cu(NH3 )2+ 4 ]f · [HR]fi /[NH3 ]f [H ]f ) 2

J=

2

−Ks ([Cu2+ ]s [HR]si /[H+ ]s ) 2

(δm /DCuR2 ,m )(1/Kex,f · Kd2 ) + (δa /DCu,a ) 2

2

× (([HR]fi /[NH3 ]4f · [H + ]f ) − (Ks · [HR]si /[H+ ]s )) (19) 2

2

At equilibrium, when [HR]fi = [HR]si and J = 0, Eq. (19) gives  + 2 [Cu(NH3 )2+ [H ]f 1 4 ]f = Ks · · [NH3 ]4f = (20) + 2+ [Cu ]s [H ]s EF where EF indicates enrichment factor. This important relationship can also be derived based on thermodynamic considerations. Total transport process for ammoniacal wastewater treatment can be described as a simple pseudochemical reaction: + + 2+ [Cu(NH3 )2+ 4 ]f + 2[H ]s ⇔ [Cu ]s + 2[NH3 ]f + 2[NH4 ]f

We can consider the next cycle:

where Kex,s is the extraction equilibrium constant for copper ion in Cu2+ /LIX54 system; subscript s indicates species in the strip solution. 5. The carrier at the strip/membrane interface diffuses back to the membrane/feed interface: J3 =

DHR,m ([HR]si − [HR]fi ) δm

(13)

where DHR,m is the diffusion coefficient of carrier in the organic liquid membrane phase. 6. The stripped copper ions permeate through an aqueous stagnant layer and enter the bulk of stripping phase. Diffusion coefficient of copper species in this case is the one for an aqueous complex: J4 =

DCu,a ([Cu2+ ]si − [Cu2+ ]s ) δa

0

[CuR2 ]fi =

[Cu(NH3 )2+ 4 ]f

Jδa − DCu,a



(15)

(16)

2

Kex,f · Kd2 · [HR]fi [NH3 ]4f · [H+ ]f

2

(17)

For the strip side we have slightly simpler expression:  [CuR2 ]si =

[Cu2+ ]s +

Jδa DCu,a



2

Kex,s · [HR]si [H+ ]s

2

+4RT ln[NH3 ]f + μ0H+ + 2RT ln[H+ ]f

And finally for copper–ammonia complex we have  + 2 [Cu(NH3 )2+ [H ]f 4 ]f = K [NH3 ]4 s [H+ ]s [Cu2+ ]s

From Eqs. (5), (6), (15) and (16), we have: 

+2RT ln[H + ]s = μ0Cu2+ + RT ln[Cu2+ ]s + μ0NH3

It means that  + 2 [Cu2+ ]f [H ]f = 2+ [Cu ]s [H+ ]s

Also there is a mass balance in the liquid membrane phase between HR and CuR2 : 2[CuR2 ]fi + [HR]fi = 2[CuR2 ]si + [HR]si = [HR]m

μ0Cu2+ + RT ln[Cu2+ ]f + μ0NH3 + 4RT ln[NH3 ]f + μ0H+

(14)

Based on the steady state assumption we have: J3 J1 = J2 = = J4 = J 2

Equilibrium between the second and the fourth states can be easily described based on chemical potentials:

(18)

Both kinetic and thermodynamic analyses give the same equation for equilibrium. It demonstrates that it is possible to transfer copper from ammoniacal solution against of its concentration gradient and thus to have an active transport. The simplest way to accumulate copper in the strip solution is to keep higher H+ concentration gradient between feed and strip solutions, but an increase of ammonia concentration in the feed makes this process less favorable. As an example for ∼1 M NH3 and pHf = 10 in the feed and pHs = 0 in the strip solutions enrichment factor EF can be higher than 104 . Too high ammonia concentration in the feed solution will decrease enrichment and hinder copper transfer from the feed to the strip side. Earlier Kyuchoukov et al. have also found

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

that ammonia depresses the extraction of copper with LIX54 when a significant excess of ammonia present in the aqueous solutions [23]. If the stripping pH is low enough, effective distribution coefficient of copper between the strip phase and membrane is much lower than that between the feed phase and the membrane [32]. In this case only one-way transport from the feed solution through the membrane phase is important and Eq. (19) can be simplified to Eq. (21): Ja =

[Cu(NH3 )2+ 4 ]f

(21)

(δa /DCu,a ) + (δm /DCuR2 ,m )(1/Kex,f · Kd2 ) 2

× ([NH3 ]4f [H+ ]f /[HR]fi ) 2

Ja is determined by total of two resistances, i.e. that of the aqueous stagnant layer and the membrane phase. The difference of J and Ja describes the role of the process reversibility. Combination of Eqs. (2), (11) and (16) gives:   J 0 (22) [HR]fi = [HR]m 1 − Jmax,m and then Ja =

(δa /DCu,a ) + (δm /DCuR2 ,m ) · ([NH3 ]4f · [H+ ]f /Kex,f · 2

0 2

(23)

It is easy to show that if no ammonia is presented in the feed or pH of the feed solution is relatively low, which corresponds to the transport of simple Cu2+ ions from an acidic phase, monodirectional copper flux is described by slightly simpler Eq. (24): [Cu2+ ]f (δa /DCu,a ) + (δm /DCuR2 ,m ) · ([H+ ]f /Kex,s · 2

(24)

0 2

([HR]m ) (1 − (Ja /Jmax,m ))2 ) According to Eq. (23), when the aqueous stagnant layer resistance is dominant and the membrane role can be neglected, copper flux Ja is first order with respect to the feed concentration of copper–tetra ammonia complex, and is described by Eq. (3): Ja =

DCu,a [Cu(NH3 )2+ 4 ]f δa

In slightly more general case when copper concentration in the feed solution is low and Ja  Jmax,m we have: Ja =

[Cu(NH3 )2+ 4 ]f (δa /DCu,a ) + (δm /DCuR2 ,m ) · ([NH3 ]4f · 0 2

[H+ ]f /Kex,f · Kd2 ([HR]m ) ) 2

Table 3 Parameters necessary for theoretical simulation Parameters

Value

DCu,a DHR,r DCuR2 ,r DCuR2 ,m Kex,s Kex,f Kd Ks δm δa δr log mCuR2 log mHR

6.5 × 10−6

References cm2 /s

[6] Calculated based on Hayduk and Laudie method [18] This work [20] [24] [31] [30] Catalog This work This work [34] [34]

5.35 × 10−6 cm2 /s 3.58 × 10−6 cm2 /s 8.9 × 10−7 cm2 /s 7 × 10−7 102.89 M2 5.56 × l0−10 M 1012.46 M−4 50 ␮m 77 ␮m at 250 rpm 10−6 to 10−5 cm −0.46 pH + 7.95 −0.27 pH + 6.06

In the opposite situation when the feed copper concentration is high, the carrier is fully loaded with copper. In these condi0 0 tions [HR]fi /[HR]m ∼ = 0 and [CuR2 ]fi ∼ = [HR]m /2. Using Eqs. (5) and (6), after transformation of Eq. (23) we can show that the flux is now described by Eq. (2): 0

[Cu(NH3 )2+ 4 ]f Kd2 ([HR]m ) /(1 − (Ja /Jmax,m ))2 )

Ja =

111

= k · [Cu(NH3 )2+ 4 ]a Here k is the overall mass transfer coefficient, which depends on pH and ammonia content in the feed solution as well as the carrier concentration.

Ja = Jmax,m =

DCuR2 ,m [HR]m · δm 2

As it should be expected, the rate of copper transfer reaches maximum and is independent on the feed copper concentration. Fig. 3 shows that the experimental flux dependence on the carrier concentration can be satisfactory simulated by this model based on the set of parameters given in Table 3. Nevertheless one can see that copper flux dependences on feed copper concentration (Fig. 4) and feed pH (Fig. 5) cannot be described by this model: using the parameters listed in Table 3 we have the first term in the denominator of Eq. (23) always much larger than the second term, which cannot explain the flux plateau at high copper concentrations. The “Small Carrousel” model also cannot predict the flux decrease in more acidic feed solutions. To do this we have to account for more complex “Big Carrousel” model presented below. 3.4.2. Description of transmembrane Cu transport based on facilitated “Big Carrousel” mechanism If the carrier is not very hydrophobic, it can leave the interior of the membrane. This implies that both metal ions and chelating agents may be in the same aqueous phase where chemical reactions take place and transport through the membrane proceeds by a mechanism, which may be called “Big Carrousel” [1]. This type of the description is based on a common idea for the reactions of the substances in the same phase and without introduction of the effective parameters like interface rate constants for the processes between substances located in two different liquid phases. Very low concentrations of the carrier and corresponding complex with the metal ion in the aqueous phase mean high transport resistance of this elementary step, playing an important role in the actual mass transfer process. According to usual macro-kinetics ideas the location of the chemical reaction should be in the vicinity of membrane/water

112

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

where Kf is the formation constant of the complex CuR2 in LIX54/Cu2+ system in the area of contact stagnant/reaction layer. The corresponding equations for flux are J1 =

DCu,a 2+ ([Cu(NH3 )2+ 4 ]a − [Cu(NH3 )4 ]r ) δa

DHR,r DHR,r J2 = ([HR]i − [HR]r ) = δr δr



(29)

[HR]i − [HR]r mHR

 (30)

DCuR2 ,r ([CuR2 ]r − [CuR2 ]i ) δr   DCuR2 ,r [CuR2 ]i [CuR2 ]r − = δr mCuR2

J3 = Fig. 7. Schematic description of copper transport through SLM with “Big Carrousel” model.

interface because of high distribution coefficients of organic species, their slow diffusion and fast reactions. The strip pH is low enough in membrane experiments with 2 M H2 SO4 as the strip and the effective distribution coefficient of copper between the membrane phase and the stripping phase is much lower than that between the feed phase and the membrane, thus, the concentration of the copper–carrier complex in the membrane phase at the stripping side may be considered negligible compared to that at the feed side [32]. To simplify the kinetic model, the description below is only for mono-directional transport from the feed solution through the membrane phase (Fig. 7). We will also use classical two-film model [33], assuming that the reaction of LIX54 and copper–ammonia complex is instantaneous. It means when they meet each other in the area of contact stagnant/reaction layer, formation of the complex CuR2 can be described by the equilibrium constant KF : KF =

[NH3

]2

· [CuR2 ]r · [NH4

J4 =

(32)

J1 = 21 J2 = J3 = J4 = J Similar to the “Small Carrousel” kinetic model (Eqs. (22) and (23)), we can show that     0 1 [HR]m 2δr mHR 1− ·J (33) [HR]r = + mHR Jmax,m DHR,r [HR]0 m

and

2 [Cu(NH3 )2+ 4 ]r · [HR]r

[Cu(NH3 )2+ 4 ]a (δa /DCu,a ) + ((δr /DCuR2 ,r ) + (δm /DCuR2 ,m · mCuR2 ))· 2

[NH3 ]4 · [CuR2 ]r · [H+ ]a

2 2 [Cu(NH3 )2+ 4 ]r · [HR]r · Kd

2

0

[NH3 ]4 · [H+ ]a /KF · Kd2 ([HR]m /mHR )

2

=

DCuR2 ,m [CuR2 ]i δm

where subscript r indicates species at the aqueous/reaction layer interface; δr is the thickness of reaction layer. At the steady state:

Ja =

+ ]2

(31)

0

2

× (1 − (1/Jmax,m + 2δr · mHR /[HR]m · DHR,r ) · J) (34)

(25)

Based on Eq. (6) the relationship between Kex,f and KF can be described as mCuR2 · KF (26) Kex,f = m2HR

Further, similar to Eq. (24) for transport from acidic copper solution without ammonia, copper flux is

where mHR and mCuR2 are effective distribution coefficients of the carrier and copper–carrier complex, respectively. It was demonstrated that both LIX54 and its copper complex are partially soluble in aqueous solutions [22,34]. In diluted acidic copper solutions without ammonia:

([H+ ]a /Kf · ([HR]m /mHR ) (1 − (1/Jmax,m )

Kf =

2 [CuR2 ]r [H+ ]a [Cu2+ ]a [HR]2r

(27)

mCuR2 Kf m2HR

(28)

and Kex,s =

Ja =

[Cu2+ ]a (δa /DCu,a ) + ((δr /DCuR2 ,r ) + (δm /DCuR2 ,m · mCuR2 ))· 2

0

0

2

2

+2δr · mHR /[HR]m · DHR,r ) · J)

(35)

One can see that in the case of “Big Carrousel” model the denominator has two additional terms, related to the reaction layer thickness δr . At low copper concentrations, the second term in the denominator of Eq. (34) is much less than the first term and the calculated flux is first order with copper concentration. The two additional terms make the “Big Carrousel” model able to predict flux plateau at high copper concentrations (Fig. 4). In these conditions when J = Jmax,m , the existence of δr makes the

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

value of second term in the denominator comparable to the first one, which results in flux saturation at high copper concentrations. In comparison, in the “Small Carrousel” model though it is possible to get simulated flux Ja very close to the experimental Jmax,m at high copper concentrations, the second term of the denominator of Eq. (23) is still negligible compared to the first one, which leads to the proportionality of the flux to copper concentration. In addition, Eq. (34) has three additional parameters in the denominator, δr , mHR and mCuR2 , which make the model much more sensitive to pH changes in the feed (Fig. 5) than it was in the case of “Small carousel”. It is well known that the distribution of species in aqueous and/or organic phases is varied with pH changes. According to [34], the correlation between feed aqueous pH and distribution coefficients is log mCuR2 = −0.46 pH + 7.95 and log mHR = −0.27 pH + 6.06. Experimental value of the flux Jexp and published correlations for distribution coefficients of LIX54 and its complex with Cu as a function of pH were used as starting values in right hand side of Eq. (34) for simulation. Optimization of parameters δr was conducted to fit the simulated flux to experimental one using Microsoft Excel builtin “solver” function. After optimization, the range of reaction layer thickness is between 10−6 and 10−5 cm at the experimental feed pHs from 4 to 9. Using these optimized values at different values of feed pH with “Big Carrousel” model, we are able to give satisfactory quantitative description of pH dependence presented in Fig. 5. According to the developed model when the reaction is much faster compared to the diffusion, the resistance of the reaction layer could be negligible and the “Big Carrousel” model (Eq. (34)) is reduced to the “Small Carrousel” (Eq. (23)). In this case δcr DCuR2 ,r



δm DCuR2 ,m · mCuR2

(36)

and 2δcr · mHR 0 [HR]m

· DHR,r



1 Jmax,m

(37)

where δcr is the critical thickness of reaction layer. Substituting kinetic parameters from Table 3 into expressions (36) and (37), we have for example that at pH around 7 and if also δcr  3.7 nm the “Big Carrousel” can be reduced to the “Small Carrousel” model. In addition δcr is inversely proportional to mHR and mCuR2 , both of which increase at lower pH, resulting in even smaller δcr value at more acidic feed solutions. On the contrary, when the reaction layer thickness is much larger than the critical value, the chemical reaction developed in the chemical reaction layer cannot be neglected. Therefore, the “Big Carrousel” model in this case is superior to describe the actual mass transfer process. For example, as the result of parameters optimization to fit the experimental data, calculated reaction layer thickness was much higher than the corresponding critical thickness: It was practically constant and equal to ∼5 × 10−6 cm at pH from 8.6 to 5 and only then it decreased to ∼1 × 10−6 cm at pH 4, which are much lager than the critical thickness from ∼10−7 to ∼10−8 cm

113

at corresponding feed pH from 8.6 to 4. That is why only “Big Carrousel” model is able to describe the experimental decrease of the flux in the more acidic feed solutions, whereas the “Small Carrousel” model gives over-estimated results and it is more significant at lower feed pH. Earlier it was demonstrated that the thickness of reaction layer for simple reversible ion exchange reactions should be determined by the square root of the ratio D/(k+ c+ + k− c− ), where subscripts + and − mean the forward and reversed reaction, respectively and k is the rate constant [1,35]. When concentration of H+ is too small the thickness of reaction layer should be constant, and then it should start decreasing at lower pH, which agrees well with our simulation results. The reaction layer plays an important role in the actual mass transfer and was discussed in a recent review regarding interfacial aspects of metal extraction in liquid–liquid system [36]. In some cases to describe kinetics of metal extraction facilitated by ligands it is necessary to assume diffusion of the extractant into the aqueous phase where it reacts with the metal ion and then the complex transfers back into the organic phase. Evidently this mechanism is similar to the “Big Carrousel” model developed for the membrane carrier in this paper. 3.5. Ammoniacal wastewater treatment through a hollow fiber supported liquid membrane system Although it is useful for the fundamental kinetics and mechanism studies, the flat sheet SLM system is not suitable for practical application due to its small membrane surface area. In contrary, hollow fiber SLM represents a very attractive solution to the need of applying SLM with very high throughputs. After liquid membrane preparation, the ammoniacal wastewater was pumped through the lumen side of hollow fiber membrane contactor once-through at various flow rates from 55 to 390 ml/min, whereas the strip solution co-currently was fed in the shell side in a recirculation mode at 70 ml/min. Both inlet and outlet concentrations of the targeted metal species (Cu(II), Zn(II), Ni(II), Cd(II)) in the tube side of membrane contactor were measured to get corresponding overall

Fig. 8. Exploration of experimental overall mass transfer coefficients over extended period of time for ammoniacal wastewater treatment in a hollow fiber SLM.

114

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

Fig. 9. Schematic description of long term stability of SLM.

mass transfer coefficient P according to P=

Q Cin ln Sm Cout

(38)

where Cin and Cout are the inlet and outlet concentrations of metal ions in the lumen side, respectively; Q is the volumetric flow rate in the tube side. The separation factor of copper over other cations is defined as the ratio of their overall mass transfer coefficients. It was found that the light blue ammoniacal wastewater became a colorless one with copper concentration less than 5 ppm when it was pumped through the lumen side of hollow fiber membrane contactor once through at a volumetric flow rate lower than or equal to 70 ml/min. This means the ammoniacal copper solutions after single pass through treatment can be reused as the rinse water or be discharged safely. The experimental results also show a highly selective separation of Cu(II) over other cations can be achieved at feed flow rate of 70 ml/min: the separation factor of Cu(II) over Cd(II) is 103 and the value is 17 for Cu(II) over Zn(II). However, the selectivity of Cu(II) over Ni(II) is not very high with a separation factor less than 4. This agrees with the fact that LIX54 can extract both metal ions well in ammoniacal solution [37]. In addition, the selective separation of copper with LIX54 will be somehow decreased with increasing feed flow rate. This may be attributed to the decreasing residence time of cations contacted with this highly selective extractant for copper over other cations. It was demonstrated in our previous work that in experiments with flat SLM after one month of exploitation the transmembrane flux was approximately the same. The electrical resistance and capacitance of the impregnated membrane filter were kept satisfactorily same over one month [13]. The stability of hollow fiber SLM system for ammoniacal wastewater treatment over an extended period of time was also investigated in this work: the wastewater was pumped once-through the lumen side, whereas the strip solution co-currently was recirculated in the shell side at the same volumetric flow rate as feed. The co-current flow mode and the same volumetric flow rates were adopted in order to provide a stable SLM system and to avoid the instability caused by the hydrodynamic velocity difference between tube and shell sides of the membrane contactor. Volumetric flow rates lower than 50 ml/min were not used because of too small

wastewater volume purified in the unit of time. On the other hand with too high flow rates the supported liquid membrane could be unstable. Therefore, the same volumetric flow rates of 70 ml/min in both feed and strip streams were employed in the membrane stability study. The membrane stability for the treatment of ammoniacal wastewater was evaluated by monitoring mass transfer coefficients over extended period of time. The experiment was conducted with the same membrane module almost for one month. The mass transfer coefficient decreases by half mainly in the first 1 to 2 weeks but then it stays practically constant (Fig. 8). The membrane could be washed, dried, reimpregnated and reused after that. It was experimentally determined that an interfacial tension between the ammoniacal wastewater and organic membrane (33% (v/v) LIX54 in kerosene) was around 26.5 mN/m. According to Laplace–Young equation the corresponding breakthrough pressure, defined as the minimum transmembrane pressure required to displace the impregnating phase out of the largest pores, was as high as 2.3 atm. Therefore the hydrodynamic pressure caused by pumping both aqueous solutions could not push the organic membrane easily out of the pores. The high breakthrough pressure makes them relatively stable in the microporous support. With time due to shear force caused by continuous pumping of the aqueous solutions in both tube and shell side, part of impregnated liquid membrane may be removed from the pores and the corresponding empty space is occupied by the surrounding aqueous solutions. Further loss of the impregnated liquid is stopped when stagnant aqueous layers are formed inside the micropores of the polymeric support (Fig. 9). In this case though the carrier moves out into the very thin reactive layer, it does not have time to move further through the whole stagnant layer of water in the pore and returned back due to “Big Carrousel” mechanism. All these factors make the system working, which is not observed in many other liquid membrane based systems. 4. Conclusions Facilitated active copper transport through the flat supported liquid membrane (SLM) from aqueous ammoniacal solution in exchange to two H+ ions has been investigated. To describe kinetics of the process a new “Big Carrousel” model was

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

developed and compared with usually assumed mechanism of facilitated transport. Mathematical model simulation demonstrated that only “Big Carrousel” model, based on the ability of the carrier to leave the membrane and to react with copper ammonia complexes in aqueous solutions, gives satisfactory quantitative description of all experimental results, including the dependence of copper transmembrane flux on carrier concentration, flux plateau at high feed copper concentrations and the decrease of copper flux at lower feed pH. Copper forms complexes with ammonia and if the ammonia concentration in the feed solution is too high, it hinders copper transfer from feed to strip side. Simple thermodynamic description of enrichment and role of ammonia is presented. Hollow fiber SLM system has been demonstrated as an effective way to remove copper from ammoniacal wastewater. Copper in the wastewater can be reduced to less than 5 ppm by pumping it once-through the lumen side of a Liqui-Cel Extra-flow 2.5 in. × 8 in. membrane contactor at low flow rate. The treated aqueous solution can be reused as the rinse water for PCB production or discharged directly in compliance with the government environmental regulations. The high selectivity of copper over other metal cations and long-term stability of the hollow fiber SLM system are promising for its practical industrial application for ammoniacal wastewater treatment. Acknowledgements Financial support from Agency for Science, Technology and Research, Singapore is gratefully acknowledged (Grant No. R279-000-164-305). Thanks are also given to Dr. Grischenko, A.B. and Mr. Zhang Yankun for their valuable work on earlier stages of this project. Cognis Corporation, USA kindly supplying of LIX54 is also highly appreciated. References [1] A.V. Mogutov, N.M. Kocherginsky, Macrokinetics of facilitated transport across liquid membranes. 1. Big Carousel, J. Membr. Sci. 79 (2/3) (1993) 273–283. [2] Z.F. Yang, A.K. Guha, K.K. Sirkar, Novel membrane-based synergistic metal extraction and recovery processes, Ind. Eng. Chem. Res. 35 (4) (1996) 1383–1394. [3] N.M. Kocherginsky, Q. Yang, S. Lalitha, Recent advances in supported liquid membrane technology. Presented at Recent Advances in Separation & Purification Techniques for Biological & Pharmaceutical Products Development, Nanyang Technological University, Singapore, 21–23 February, 2005. [4] M. Grote, B. Haciosmanoglu, M. Bataineh, J. Nolte, Separation of drug traces from water with particular membrane systems, J. Environ. Sci. Health Part a: Toxic/Hazard. Sub. Environ. Eng. 39 (4) (2004) 1039–1053. [5] J.D. Clark, B.B. Han, A.S. Bhown, S.R. Wickramasinghe, Amino acid resolution using supported liquid membranes, Sep. Purif. Technol. 42 (3) (2005) 201–211. [6] R.S. Juang, J.D. Chen, H.C. Huan, Dispersion-free membrane extraction: case studies of metal ion and organic acid extraction, J. Membr. Sci. 165 (1) (2000) 59–73. [7] M.M. Naim, A.A. Monir, Desalination using supported liquid membranes, Desalination 153 (1/3) (2003) 361–369. [8] J.A. Jonsson, L. Mathiasson, Membrane extraction in analytical chemistry, J. Sep. Sci. 24 (7) (2001) 495–507.

115

[9] J.J. Ubl, J.Y. Chatton, S.H. Chen, J.W. Stucki, A critical evaluation of in situ measurement of mitochondrial electrical potentials in single hepatocytes, Biochim. Biophys. Acta-Bioenergetics 1276 (2) (1996) 124– 132. [10] P.A. Ohara, M.P. Bohrer, Supported liquid membranes for copper transport, J. Membr. Sci. 44 (2/3) (1989) 273–287. [11] G.R.M. Breembroek, A. van Straalen, G.J. Witkamp, G.M. van Rosmalen, Extraction of cadmium and copper using hollow fiber supported liquid membranes, J. Membr. Sci. 146 (2) (1998) 185–195. [12] F. Valenzuela, C. Basualto, C. Tapia, J. Sapag, Application of hollow-fiber supported liquid membranes technique to the selective recovery of a low content of copper from a Chilean mine water, J. Membr. Sci. 155 (1) (1999) 163–168. [13] N.M. Kocherginsky, A. Grishchenko, Method for metal recovery from aqueous solution. US Patent 6,521,117 B2 (2003). [14] T.H. Irvine, Chemical Analysis of Electroplating Solutions: A Theoretical Approach, Chemical Publishing Co, New York, 1970. [15] A.B. Dehaan, P.V. Bartels, J. Degraauw, Extraction of metal-ions from waste-water—modeling of the mass-transfer in a supported-liquidmembrane process, J. Membr. Sci. 45 (3) (1989) 281–297. [16] S.Y.B. Hu, J.M. Wiencek, Emulsion-liquid-membrane extraction of copper using a hollow-fiber contactor, AIChE J. 44 (3) (1998) 570– 581. [17] C.R. Wilke, C. Pin, Correlation of diffusion coefficients in dilute solutions, AIChE J. 1 (2) (1955) 264–270. [18] W.J. Lyman, W.F. Reehl, D.H. Rosenblatt, Handbook of Chemical Property Estimation Methods: Environmental Behavior of Organic Compounds, McGraw-Hill, New York, 1982. [19] F.J. Alguacil, M. Alonso, A.M. Sastre, Modelling of mass transfer in facilitated supported liquid membrane transport of copper(II) using MOC-55 TD in Iberfluid, J. Membr. Sci. 184 (1) (2001) 117–122. [20] F.J. Alguacil, M. Alonso, Recovery of copper from ammoniacal/ammonium sulfate medium by LIX 54, J. Chem. Technol. Biotechnol. 74 (12) (1999) 1171–1175. [21] W. Mickler, E. Uhlemann, Liquid-liquid-extraction of copper from ammoniacal solution with Beta-diketones, Sep. Sci. Technol. 27 (12) (1992) 1669–1674. [22] Z. Lazarova, Study on the kinetics of copper LIX54 system using a rotating diffusion cell, Solv. Extraction Ion Exchange 13 (3) (1995) 525–540. [23] G. Kyuchoukov, M.B. Bogacki, J. Szymanowski, Copper extraction from ammoniacal solutions with LIX 84 and LIX 54, Ind. Eng. Chem. Res. 37 (10) (1998) 4084–4089. [24] M.R.C. Ismael, M.L.F. Gameiro, J.M.R. Carvalho, Extraction equilibrium of copper from ammoniacal media with LIX 54, Sep. Sci. Technol. 39 (16) (2004) 3859–3877. [25] M.J. Zapatero, M.P. Elizalde, J.M. Castresana, Acid–base equilibria and metal extraction properties of 1-phenyl-1,3-decanedione, Anal. Sci. 7 (6) (1991) 935–940. [26] P.R. Danesi, E.P. Horwitz, G.F. Vandegerift, Mass transfer rate through liquid membranes: interfacial chemical reactions and diffusion as simultaneous permeability controlling factors, Sep. Sci. Technol. 16 (2) (1981) 201–211. [27] P.R. Danesi, G.F. Vandegerift, E.P. Horwitz, Simulation of interfacial two-step consecutive reactions by diffusion in the mass-transfer kinetics of liquid-liquid extraction of metal cations, J. Phys. Chem. 84 (1980) 3582–3587. [28] F.A. Cotton, G. Wilkinson, Advanced Inorganic Chemistry, Wiley, New York, 1988, p. 769. [29] N.S. Ahmetov. General and Inorganic Chemistry, Moscow, 1988. [30] J. Bjerrum, G. Schwarzenbach, L.G. Sillen, in: L.G. Sillen, A.E. Martell (Eds.), Stability Constants of Metal–Ion Complexes, International Union of Pure and Applied Chemistry, London, 1964. [31] D.F. Shriver, P.W. Atkins, Inorganic Chemistry, Oxford University Press, New York, 1999. [32] P.R. Danesi, Separation of metal species by supported liquid membranes, Sep. Sci. Technol. 19 (11/12) (1984) 857–894. [33] C.J. Geankoplis, Mass Transport Phenomena, Holt, Rinehart and Winston, Inc., New York, 1972.

116

N.M. Kocherginsky, Q. Yang / Separation and Purification Technology 54 (2007) 104–116

[34] L. Boyadzhiev, S. Alexandrova, Recovery of copper from ammoniacal solutions by rotating film pertraction, Hydrometallurgy 35 (1) (1994) 109–121. [35] A.V. Mogutov, N.M. Kocherginsky, Macrokinetics of facilitated transport through liquid membranes. 2. Stirring, J. Membr. Sci. 86 (1/2) (1994) 127–135.

[36] G.W. Stevens, J.M. Perera, F. Grieser, Interfacial aspects of metal ion extraction in liquid–liquid systems, Rev. Chem. Eng. 17 (2) (2001) 87– 110. [37] LIX 54- A new reagent for metal extraction from ammoniacal solutions, Henkel Corporation, Tuczon, Arizona, 1975.