Analytica Chimica Acta 387 (1999) 77±84
Selective separation of lanthanides by supported liquid membranes containing Cyanex 925 as a carrier R. Garcia-Valls, M. MunÄoz*, M. Valiente1 Departament de QuõÂmica, QuõÂmica AnalõÂtica, Universitat AutoÁnoma de Barcelona, 08193 Bellaterra, Barcelona, Spain Received 12 May 1998; received in revised form 25 October 1998; accepted 18 November 1998
Abstract The work describes in detail the transport of lanthanides between chloride solutions by a facilitated transport through a supported liquid membrane (SLM) containing Cyanex 925 (a mixture of alkylphosphine oxides, Cyanamid). The chemical nature of this carrier requires a strong difference of chloride content between the two sides of the membrane which leads to a strong rise in the osmotic pressure. The in¯uence of physical and chemical parameters affecting the transport of neodymium was determined including stirring rate and the chemical composition of the feed and stripping solutions. The highest permeability coef®cients (cm minÿ1) were obtained under the following conditions: 1.0 M Cyanex 925 in kerosene (membrane), stirring rate 1200 rpm, feed solution of 2.5 M NaCl at pH 4.0 and a stripping solution of 0.8 M HCl. Under these conditions, the SLM is shown to be effective for the transport and separation of neodymium. Separation of lanthanide mixtures is possible. The selectivity of the system is such that Yb is the most favourably transported and La the least favoured, following the correlation Yb>Tb>NdLa. # 1999 Elsevier Science B.V. All rights reserved. Keywords: Supported liquid membrane; Cyanex 925; Lanthanide
1. Introduction Transport of lanthanides through liquid membranes [1] have been studied using amines [2±4] or acidic reagents [5,6] as carriers. Neutral carriers have scarcely been employed for their transport. TBP has been used at 30% (v/v) in kerosene and with a high concentration of nitrate (salting agent) in the feed solution to enhance the transport [7]. *Corresponding author. Fax: +34-3-581-1985; e-mail:
[email protected] 1 Corresponding author.
This type of reagent has been employed to transport lanthanides on the basis of the following reaction [8]: pLn3 3pClÿ qRorg @
LnCl3 p Rq org
(1)
where Ln3 corresponds to the lanthanide ion, R is the carrier and subindex org denotes those species in the organic phase. The complex is formed in the organic phase between the lanthanide ion and the neutral carrier favoured by the presence of chloride ions. In the stripping phase, the metal ion is recovered by the dissociation of the complex that had been formed in the feed solution/membrane interphase. The complex, when reaching the stripping solution/membrane inter-
0003-2670/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved. PII: S0003-2670(98)00842-3
78
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
phase, is dissociated due to the low chloride concentration in the stripping phase, and the equilibrium shown in Eq. (1) is shifted to the dissociation of the complex. For a system following Eq. (1), the driving force of the related Ln3 transport will be the difference in chloride concentration on each side of the membrane. This will develop an osmotic pressure [9,10]. High values of this pressure have been reported to cause instability of the supported liquid membranes [11]. The reagent Cyanex 925 (Cyanamid) employed in this work is a mixture of alkyl phosphine oxides and behaves as a neutral carrier whose steric properties enhances its selectivity against other reagents of similar characteristics, e.g. trioctylphosphine oxide (TOPO). The use of Cyanex 925 either as extractant for rare earths or as a carrier in a supported liquid membrane (SLM) has not been reported in the literature. This work discusses the results of a systematic study of various chemical and physical parameters affecting the transport, i.e. stirring rate and the chemical composition of feed and stripping solutions. Selectivity in the separation of mixtures of lanthanides has been determined. 2. Experimental 2.1. Reagents and solutions Solutions of 210ÿ3 M Arsenazo III (Fluka) in 1.0± 2.5 M NaCl and 0.2 M sodium formate (Fluka) were used for the determination of neodymium. The determination of mixtures of lanthanides were performed by liquid chromatography (LC) with detection based on Arsenazo III [12]. Neodymium nitrate hexahydrate (Fluka) was taken to prepare stock 0.25 mM metal solution. Appropriate dilutions of this solution in the presence of NaCl were used as feed solutions. Cyanex 925 (15% trioctylphosphine oxide, 85% di(2-(methyl4-dimethyl-pentyl)octylphosphine oxide)), kindly supplied by Cyanamid Canada, was used to prepare organic solutions in kerosene (Fluka) within the range 0.171±1.710 M Cyanex 925. Polymeric microporous membranes of PVDF (Millipore, GVHP 04700), having 0.2 mm pore size and 75% porosity, were used as solid supports for the liquid membrane.
2.2. Apparatus A visible-region spectrophotometer Novospec II LKB (Pharmacia), a peristaltic pump M312 (Gilson), and an injection valve Type 50 (Rheodyne) were used in a ¯ow injection (FI) manifold for the continuous determination of neodymium. A Shimadzu liquid chromatography (LC) apparatus was used for the determination of mixtures of lanthanides. The equipment consisted of two pumps (model LC-9A), a UV± visible spectrophotometer as detector (model SPD6AV) and a recorder (Chromatopac model C-R6A). Appropriate columns of silica base were used for chromatographic determination. A vapour pressure osmometer was used for osmotic pressure measurements (Wescor 5500). An inductively coupled plasma (ICP) atomic emission spectrometer model 3410 (minitorch, ARL) was employed for eventual neodymium determination. 2.3. Procedure The SLM experiments were carried out in a two compartment membrane cell described elsewhere [13]. The liquid membrane was prepared by impregnation of organic solution in the microporous polymer using ultrasonic agitation (1 min). Afterwards, the laminar supported liquid membrane was placed in the round window of the cell, between feed and stripping solutions. Time zero of the experiments is taken as the time when the stirring motors are switched on. The metal concentration in the feed solutions was monitored by using a FIA manifold (see Fig. 1) based on the reaction between Arsenazo III and Nd(III) [12]. The initial conditions for the experiments were: feed solution: 5 mm mlÿ1 Nd(III), 0.2 M NaCl, pH 3.5; stripping solution: 1.0 M HCl; liquid membrane: 0.68 M Cyanex 925 in kerosene; polymeric support: Millipore GVHP04700; stirring speed: 1000 rpm. These conditions were varied systematically in different sets of experiments to determine the in¯uence of the parameters (i.e. stirring rate, composition of the aqueous solutions, carrier concentration in the liquid membrane and osmotic pressure). The selectivity of the system was determined by processing mixtures of La, Nd, Tb and Yb under the optimized conditions: 2.5 M NaCl in the feed solu-
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
79
The permeability coef®cient P is de®ned by P
Fig. 1. Flow injection system employed for the determination of Nd(III). Detector: UV±visible spectrophotometer at 660 nm. Reagent: Arsenazo III. Reaction coils of 1 m length.
tions containing 5 mg ml of each of the four lanthanides; the pH of the feed solution was 4.0. The stripping solution was 1.0 M HCl. The transport of each individual lanthanide, later present in the mixture, was ®rst studied under the optimized conditions. For the mixtures, the transport data for each metal ion were obtained by the determination of the concentration of each lanthanide using the LC set-up. The mobile phase conditions in the chromatographic system were similar to those employed by Cassidy [14]. In this method, the mobile phase is a solution of ahydroxyisobutyric acid as eluent and the detection is done by post-column derivatization using a mixture of Arsenazo III and urea. 3. Results and discussion The transport may take place following the reaction shown below that takes place in two phases: (2)
This reaction competes with HCl extraction by Cyanex 925. Such behaviour makes it necessary to consider a pH gradient as a driving force for the related liquid membrane system. The results of lanthanide transport are reported in the form of permeability coef®cient (P, cm minÿ1) as explained below. Metal ¯ow through the membrane is expressed in terms of Nd(III) ¯ux, J, de®ned by Jÿ
Vf dCf ; Q dt
(4)
Taking into account Eqs. (3) and (4) one can obtain Eq. (5): Pÿ
Vf dCf : Cf Q dt
(5)
By integration of Eq. (5), we obtain
ÿ1
Nd3 3Clÿ Rorg @NdCl3 Rorg
J : Cf
(3)
where Cf is the Nd(III) concentration in the feed solution at time t (min), Vf the feed volume (ml), in our case 200 ml, and Q is the effective membrane area (cm2), in our case 8.51 cm2 (corrected for porosity).
ÿln
Cf QP t; C0 Vf
(6)
where C0 is the initial concentration in the feed solution. From the experimental data the relationship ÿln(Cf/ C0) vs. time is shown to be linear. The corresponding slope determines the value of the permeability by applying Eq. (6). Separate experiments in the absence of carrier have demonstrated that Cyanex 925 is a facilitating carrier as no transport of Nd(III) was observed in these experiments. A systematic study of the liquid membrane system was carried out by varying the corresponding parameter in a certain range while the others were kept constant. Thus, the following parameters have been studied in this work: separation system characteristics: stirring rate, feed conditions, stripping conditions and membrane composition. The selectivity of the system was also determined in separate experiments.
4. Separation system characteristics 4.1. Stirring rate The stirring rate was varied in the range 200± 1400 rpm to determine the in¯uence of hydrodynamic conditions on metal transport. Equal stirring rates were applied to the feed and stripping solutions in each experiment and the resulting permeability coef®cients were determined (see Fig. 2). From these results it is noticed that transport does not show a dependence on the stirring rate over 1200 rpm. Below this stirring rate, the transport is assumed to depend
80
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
Fig. 2. Influence of the stirring rate on the transport of Nd(III) through the SLM using 0.68 M Cyanex 925 in kerosene as carrier. Feed solution was 2.0 M NaCl at pH 4.0 containing 4 mg mlÿ1. Stripping solution was 1.0 M HCl.
Fig. 3. Influence of the acidity of aqueous feed solution, containing 4 mg mlÿ1 of Nd(III) on the transport of Nd(III) in 2.0 M NaCl. Stripping solution was 1.0 M HCl. Membrane was 0.68 M Cyanex 925 in kerosene.
only on chemical conditions and also that diffusion processes have a constant contribution.
loading with the possibility of increasing its permeability over a wider range of pH. Salting agent. The effect of this parameter has been determined in the range 2.0±3.5 M NaCl. The results are given in Fig. 5. It is observed that above 2.5 M the transport of Nd(III) is not affected significantly. These results can also be explained by considering reaction (2) as fundamental for the facilitated transport. An initial increase in the concentration of the salting agent displaces the extraction reaction towards the formation of the transported complex. For chloride concentrations higher than 2.5 M the extraction reaction has reached the maximum yield and the uptake of Nd(III) is not affected by the related excess of chloride, consequently the transport rate remains practically constant above this concentration of Clÿ (the slight decrease shown in Fig. 5 is in the region of 5% which corresponds to the experimental uncertainty determined for the permeability coefficient).
4.2. Feed conditions Two different parameters were varied in the feed solutions, pH and salting agent (NaCl) concentration. pH influence was controlled within the range 2.6± 4.0, where an increase in permeability was observed by increasing the pH (see Fig. 3). This behaviour follows the exchange of H by Nd(III) observed in the liquid±liquid experiments (see Eq. (2)).The observed increase with pH of both extraction and transport rate of Nd(III) (see Figs. 4 and 3, respectively) accounts for the validation of the proposed Eq. (2) because both processes have important differences. Thus, while the solvent extraction is an accumulative process, in the SLM the extraction and stripping of Nd(III) are taking place at the same time. This will lead the membrane to a lower metal
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
Fig. 4. The influence of feed pH on the solvent extraction of Nd(III) by Cyanex 925 (0.68 M) in kerosene. Feed solution was 2.0 M NaCl and 4 mg mlÿ1 Nd(III).
4.3. Stripping conditions The stripping solutions consisted of HCl/NaCl mixtures. This was taken as the best composition after testing other mixtures such as H2SO4/Na2SO4 and HNO3/NaNO3. The HCl concentration was varied from 0.25 to 1.0 M maintaining the total chloride concentration constant at 1.0 M by the appropriate addition of NaCl. Results of these experiments are given in Fig. 6. As observed, a decrease of permeability is obtained with a decrease of HCl concentration. Reaction (2) is also in agreement with these results which can be interpreted from the stripping side as the hydrochloric acid is back-transported by Cyanex 925 [15]. This process helps to release lanthanides into the stripping phase and leaves the carrier molecules ready to continue with their transport cycle. 4.4. Carrier concentration The effect of carrier concentration of Cyanex 925 has been investigated by varying its concentration
81
Fig. 5. Effect of NaCl concentration in feed on the transport of Nd(III) at pH 4.0. The difference between the salt concentrations at both sides of the membrane was 1.5 M. Initial concentration of Nd(III) was 4 mg mlÿ1 and the membrane was a 0.68 M solution of Cyanex 925 in kerosene. The initial stripping solution was 1.0 M HCl and NaCl was added to maintain the salt gradient.
in the liquid membrane from 0.3 to 1.8 M. Results expressed in terms of permeability coef®cient are collected in Fig. 7. As observed, the permeability coef®cient has a maximum at 1.0 M Cyanex 925. This behaviour may be due to the increase of the viscosity of the organic solution at higher carrier concentrations that leads to an increase of the liquid membrane resistance to the diffusion of metal±carrier species [15]. In this case, the increase of the carrier concentration at this point does not compensate for the increase in the solution viscosity. A separate experiment was carried out by using the metal complex formed between Cyanex 925 and Nd(III) already loaded in the organic phase as a liquid membrane. No signi®cant variation of the permeability coef®cient (from 0.0117 to 0.0123 cm minÿ1) was observed. This indicates very rapid kinetics to establish a steady state transport process.
82
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
Fig. 6. Effect of the acidity of the stripping solution (HCl) on the permeability coefficient of Nd(III). The membrane was 0.68 M Cyanex 925 in kerosene, and the feed solution was 2.0±2.75 M NaCl at pH 4.0 containing 4 mg mlÿ1 Nd(III).
Fig. 7. Effect of the concentration of Cyanex 925 in the liquid membrane on the transport of Nd(III). Feed solution was 2.0 M NaCl at pH 4.0 containing 4 mg mlÿ1 Nd(III). The stripping solution was 1.0 M HCl.
4.5. Osmotic pressure
gradient between the two aqueous solutions. Thus, the theoretical osmotic pressure range was set from 24 to 42 atm. The results of these experiments are represented in Fig. 8. It is observed that 30 atm is the maximum value of osmotic pressure under which the SLM does not lose transport properties. This value may be the result of two opposite effects. The increase of salting agent in feed causes a positive effect on the transport following Eq. (1) until a limit value is reached. After a critical value, the osmotic pressure induces the loss of organic phase from the pores of the membrane and it produces a decrease of the active interfacial area [9,10]. Above 30 atm osmotic pressure, the loss of liquid membrane is remarkable and the transport velocity decreases. The transport of water from the stripping solution to the feed due to the gradient of ionic strength may be the cause of breakdown of the Nd(III) transport. However, the driving force requirements for this system produce a minimum of 24 atm of theoretical osmotic pressure to achieve Nd(III) transport. Thus, in this system,
The in¯uence of this parameter was studied as follows. The theoretical osmotic pressure was determined by an indirect technique. Thus, using a vapour pressure osmometer, the osmolality of the solutions was determined. Then, the theoretical osmotic pressure was calculated by applying the general Van't Hoof equation for diluted solutions: cRT; (7) where c stands for the osmolality, the total salt concentration (mol kgÿ1); R is the universal constant for gases (atm Kÿ1 molÿ1), and T is the absolute temperature (K). The theoretical osmotic pressure on the membrane was assumed to be the difference between the feed and stripping osmotic pressures: (8) f ÿ s
cf ÿ cs RT: The in¯uence of the osmotic pressure on Nd(III) transport was studied by varying the salt concentration
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
83
Fig. 9. Transport of lanthanides from a mixture containing 5 mg mlÿ1 of each of the lanthanides La, Nd, Tb and Yb. System conditions were: 1.0 M Cyanex 925 in kerosene, feed solution 2.5 M NaCl at pH 4.0 and stripping solution 1.0 M HCl. The concentration vs. time of the elements in the feed solution was determined by LC.
Fig. 8. Influence of the theoretical osmotic pressure on the transport of Nd(III). Initial conditions were: 0.68 M Cyanex 925 in kerosene as carrier, feed solution was 2.0 M NaCl at pH 4.0 containing 4 mg mlÿ1 Nd(III) and the stripping solution was 1.0 M HCl.
transport of Nd(III) is achieved with minimum loss of the organic phase under a theoretical osmotic pressure ranging from 24 to 30 atm. 5. Selectivity of the system The selectivity for the transport was studied in two different sets of experiments: 1. transport of individual lanthanides; 2. separation of lanthanide mixtures. Both sets of experiments were performed under the best conditions determined in the systematic study. 1. The permeabilities obtained with the individual lanthanides were: Nd (0.0440.001 cm minÿ1), T b ( 0 . 0 6 6 0 . 0 0 1 c m m i n ÿ1 ) a n d Y b (0.2440.001 cm minÿ1); no transport was observed for La. Therefore, the separation factors, calculated from these values as the ratio between
their respective permeabilities, were: Tb/Nd 1.5, Yb/Nd 5.6 and Yb/Tb 3.7. These factors allow one to predict a possible separation of mixtures of these ions. 2. The separation of lanthanides from mixtures of ytterbium, terbium, neodymium and lanthanum, containing 5 mg mlÿ1 of each element was studied under the same conditions as in the individual element experiments. The analyses of the samples were carried out by LC. The results in Fig. 9 show that the velocities of transport obtained are different for each cation. The permeabilities in this case are: Yb (0.250 cm minÿ1), Tb (0.112 cm minÿ1), Nd (0.078 cm minÿ1). No transport was detected for La. The separation factors are Tb/Nd 1.4, Yb/ Nd 3.2 and Yb/Tb 2.2. These factors are slightly different because the permeabilities for Tb and Nd are higher than that in the individual cases. Such small differences may be due to the different chemical conditions in the mixtures, here the total concentration of lanthanides is four times higher. In any case, the order of magnitude is the same and separation factors are still quite high. The analysis of the stripping solutions showed the presence of Yb in the first two hours and no significant presence of the other elements. Therefore, the purification of Yb from the mixtures can easily be achieved.
84
R. Garcia-Valls et al. / Analytica Chimica Acta 387 (1999) 77±84
6. Conclusions
References
The reagent Cyanex 925 has been found to be a successful carrier for the facilitated transport of lanthanides by supported liquid membranes. The conditions for the optimum transport have been determined as: 1.0 M Cyanex 925 in kerosene as the liquid membrane, 1200 rpm of stirring rate, pH4.0 of feed solutions, 0.8 M HCl for the stripping solution and 2.5 M NaCl in feed solutions. The stability of the membrane under these conditions seems to be optimum as it is shown in the experiments related to the osmotic pressure effect. Under these conditions the membrane system has shown a good selectivity in the separation of lanthanides mixtures, the rate of transport being in the order Yb>Tb>NdLa. Puri®cation of individual lanthanides from the studied mixtures can be achieved.
[1] R.D. Nobel, J.D. Way, Liquid Membranes, Theory and Applications, ACS Symposium Series, vol. 347. [2] Z.R. Hua, X. Lei, J. Membr. Sci. 51 (1990) 249. [3] M. Teramoto, T. Sakuramoto, T. Koyama, H. Matsuyama, Y. Miyake, Sep. Sci. Technol. 21(3) (1986) 229. [4] Z. Reihua, W. Dexian, Water Treatment 4 (1989) 165. [5] M. Moreno, A. HrdlicÏka, M. Valiente, J. Membr. Sci. 81 (1993) 121. [6] S. Nakamura, S. Ohashi, K. Akiba, Transport of lanthanoids through supported liquid membrane with Versatic 10 as a mobile carrier, Proceedings of the Symposium on Solvent Extraction, 1989, Tokyo, pp. 101±106. [7] R. Kopunec, T.N. Mahnhand, J. Radioanal. Nucl. Chem. 183 (1994) 181. [8] T.N. Manh, R. Kopunec, J. Radioanal. Nucl. Chem. 195(2) (1992) 219. [9] H.-J. Bart, C. Ramaseder, T. HaselgruÈler, R. Marr, Hydrometallurgy 28 (1992) 253. [10] C. Ramaseder, H.-J. Bart, R. Marr, Sep. Sci. Technol. 28(1)(2)(3) (1993) 929. [11] R. Chiarizia, E.P. Horwitz, K.M. Hodgson, Removal of inorganic contaminants from groundwater, Environmental remediation, ACS Symp. Ser. 509 (1992) 22. [12] M. Rohwer, N. Collier, E. Hosten, Anal. Chim. Acta 314 (1995) 219. [13] C. Palet, M. MunÄoz, S. Daunert, L.G. Bachas, M. Valiente, Anal. Chem. 65 (1993) 1533. [14] R.M. Cassidy, Chem. Geol. 67 (1988) 185. [15] Cyanex 925, Solvent Extraction Reagent, American Cyanamid, Wayne, NJ 07470.201-831-2000.
Acknowledgements This study has been carried out thanks to the ®nancial support of CICYT (Spanish Commission for Research and Development), Projects: MAT-970720-C03 and QUI-96-1025-C03-01. Dr. Ricard Garcia-Valls acknowledges the DireccioÂn General de InvestigacioÂn Cientõ®ca y TeÂcnica for providing a scholarship to carry out his Ph.D. studies (Ref.: PN90 33898421).