Salting-out effect on facilitated transport membranes for CO₂ separation: From fluoride salt to polyoxometalates

Abstract

Salting-out effect on facilitated transport membranes for CO₂ separation has been revealed. The salts, especially fluoride salt and polyoxometalates, contained in the membrane may decrease the permeance of CO₂ to some extent but remarkably increase the selectivity of CO₂/N₂, as a result of synergistic interactions between facilitated transport and salting-out.

---

In 1967, Ward and Robb firstly introduced a facilitated transport membrane for CO₂ separation by immobilizing an aqueous bicarbonate–carbonate solution onto a porous support to achieve a separation factor of 1500 for a CO₂/O₂ mixture.¹ Because of its potential to overcome the trade-off between the permeability and selectivity, the facilitated transport membrane has attracted considerable attention worldwide for CO₂ separation.² However, these supported liquid membranes (SLMs) suffer from solution loss from the porous support, making it impossible for large-scale industrial applications.³ Since then, great efforts have been made to develop facilitated transport membranes with fixed-site-carriers (FSCs) which exhibit much higher stability because of the chemical bonding to the polymer backbones.^4–12 It was also found that the FSC polymeric membranes showed higher permselectivity in a water-swollen condition than in a dry condition,^5–8 as CO₂ molecules in combination with moisture produce HCO₃⁻, which is then quickly and selectively transported through the membrane.^13–15 The membrane is hence capable of combining the advantages of both supported liquid membrane and solid polymeric membrane.¹⁵

On the other hand, the addition of salts into a polymeric membrane has proved practical to improve its permselectivity for gas separation. For example, the addition of NH₄SCN to a crosslinked poly(vinyl alcohol) membrane modified by methacrylic acid resulted in the NH₃/N₂ selectivity over 1000 as compared to 3 for the membrane without salts.¹⁶ Similarly, the selective permeation of 1-butene over n-butane was demonstrated for a crosslinked poly(vinyl alcohol) membrane containing AgNO₃.¹⁷ In a patent published in 2008, Hägg et al. firstly added NH₄F to the FSC facilitated transport membrane based on polyvinylamine for CO₂ separation, resulting in the CO₂/CH₄ selectivity as high as 1143. It was speculated that the water molecule became more basic than pure bulk water when it was hydrogen-bonded to a fluoride ion, which had an increased affinity for CO₂ that led to an increased concentration of HCO₃⁻ in the membrane and a consecutively increased transport of CO₂.¹⁸ It should be noticed that, as a matter of fact, the hydrolysis of NH₄F will make its aqueous solution mildly acidic, which is contrary to that of NaF (mildly alkaline). Even if for the latter, the alkalescence caused by the hydrolysis of fluoride ions should be negligible in the alkali polyvinylamine matrix. Since that, it is interesting to clarify the real role of fluoride ions and the mechanism that contributes to such a high permselectivity.

Following the above question, the facilitated transport effect of fluoride ions needs to be validated in the current study first. We designed the experiment by immersing a thermally crosslinked water-swollen poly(vinyl alcohol) membrane that possesses P_CO2 = 0.79 GPU and α_CO2/N2 = 79 originally, into a 0.5 M NaF shoultion, which led to P_CO2 = 0.53 GPU and α_CO2/N2 = 500.¹⁹ The effect of NaF on the membrane was to reduce the CO₂ permeance by a factor of 1.5 and to reduce the N₂ permeance by a factor of 9.4. The decreased permeance of CO₂ indicates that the effect of facilitated transport by hydrolysis of fluoride ions could be negligible even if the membrane was based on such a non-alkali poly(vinyl alcohol) matrix. The decreased permeance of N₂ by a factor of 9.4 suggests that the solubility of N₂ in the membrane with NaF must be 9.4 times less than that in the original membrane without NaF, as the transport of N₂ through membranes was only by Fickian diffusion, indicating a salting-out effect. On the basis of the hydration theory of the salt effect in aqueous solutions,²⁰ this salting-out effect in the water-swollen poly(vinyl alcohol) membrane could be approximately attributed to the preferential attraction between the ions (Na⁺ and F⁻) and the water-swollen poly(vinyl alcohol) which should decrease the numbers of “free” water molecules and poly(vinyl alcohol) macromolecules available to dissolve the gas (N₂). In contrast with N₂, the decreased permeance of CO₂ just by a factor of 1.5 profited from the favorable water–CO₂ interactions in water-swollen membranes.^5–8 With the addition of NaF, water molecules adsorbing in the inter-chain spaces of poly(vinyl alcohol) would be further stabilized through hydration in the form of such as [Na(H₂O)_n]⁺, HOH⋯F⁻, etc., which may decrease the number of active water molecules available to interact with CO₂ and thus decrease the CO₂ permeance, in a sense indicating a salting-out effect similar to that of N₂. Consequently, the high selectivity of 500 of the membrane could be considered as the result of the salting-out effect which would decrease the solubility of CO₂ to a certain extent but decrease the solubility of N₂ remarkably. In other words, this effect would decrease the permeance of CO₂ to some extent but increase the selectivity remarkably.

In order to further explore this phenomenon, we performed a series of experiments by adding NaF, NaCl and NaBr respectively to the facilitated transport glutaraldehyde-crosslinked polyallylamine membrane with original P_CO2 = 8.4 GPU and α_CO2/N2 = 50, leading to the results shown in Table 1. After the addition of NaX (X = F⁻, Cl⁻, or Br⁻), both CO₂ and N₂ permeances decreased as expected. The corresponding P_CO2 values of NaF, NaCl and NaBr were similar, indicating that the hydrated cations ([Na(H₂O)_n]⁺) on one side of hydration could exert a dominant influence on the decrease of CO₂ permeance though the hydrated anions ([X(H₂O)_n]⁻), as the other side of hydration are quite different, e.g., in hydrogen bonding strength (F⁻ ≫ Cl⁻ > Br⁻). The distinctively increased selectivity for NaF indicates that the hydrogen bonding (Y–H⋯X (Y = N, or O)) could play a dominant role in the decrease of N₂ permeance, and thus for NaCl and NaBr the indistinctively increased selectivity could be ascribed to the weaker hydrogen bonding interaction because of their smaller electronegativity and larger atom radius. By that means, in order to get the best permselectivity, the anion of a salt with the largest electronegativity and the smallest radius would be preferred; F⁻ is just the best choice.

Table 1 Comparison of NaX (X = F⁻, Cl⁻, or Br⁻) effect on permselectivity for the glutaraldehyde-crosslinked polyallylamine (M_w = 15 000) membranes cast on polyacrylonitrile supports (MWCO = 20 000)^ab

	Aqueous solution	NaF (0.5 M)	NaCl (0.5 M)	NaBr (0.5 M)
a Membrane preparation: cast a polyallylamine solution on supports, dried at room temperature, crosslinked with glutaraldehyde, and dipped in the respective solution of NaX for 24 h. b All the data measured at room temperature and 1 bar. c Permeance, in units of GPU, which is 10^−6 cm^3 (STP)/(cm^2 s cmHg).
P CO2 (GPU)^c	8.4	2.8	3.1	3.1
α CO2/N2	50	1400	62	51

---

Further investigation was performed using oxysalts. Experiments were carried out by the addition of CH₃COONa, Na₂SO₄ and Na₃[PW₁₂O₄₀] to the glutaraldehyde-crosslinked polyallylamine membranes with original P_CO2 = 8.4 GPU and α_CO2/N2 = 50, leading to the results shown in Table 2. With the increase of oxygen atoms in the structural formula of these oxysalts, both CO₂ and N₂ permeances decreased while the selectivities increased. For CH₃COONa, the P_CO2 value was close to that of NaX in view of the same concentration of hydration cations ([Na(H₂O)_n]⁺), further indicating that the hydrated cations could have the dominant influence on the decrease of CO₂ permeance. For Na₂SO₄, both CO₂ and N₂ permeances were lower than that of CH₃COONa due to the increased hydrated cations ([Na(H₂O)_n]⁺) and the increased hydrogen bonds (Y–H⋯O (Y = N, or O)), respectively; the selectivity of 199, which was obviously higher than CH₃COONa as well as NaCl and NaBr, benefited from the increased number and strength of hydrogen bonds, further indicating the dominant role of hydrogen-bonding in the selectivity. For Na₃[PW₁₂O₄₀], due to the restriction of its maximum solubility, the concentration of 0.02 M was chosen to carry out the experiment. Interestingly, although its hydrated cations were far less than that of 0.5 M NaX, CH₃COONa and Na₂SO₄, the value of CO₂ permeance for Na₃[PW₁₂O₄₀] was the lowest among these salts; meanwhile, the number and the strength of hydrogen bonds were likewise far less than that of the other salts, while Na₃[PW₁₂O₄₀] exhibited the remarkable selectivity similar to that of NaF. The result seemed difficult to explain by using the hydration theory which has well explained the results for NaX, CH₃COONa and Na₂SO₄. It should be pointed out that the explanation based on the hydration theory for the salting-out effect is simplified because it ignores the significant effect the ions could play on the structures of the surrounding water molecules or polymeric macromolecules, considering that an electrolyte with some unique structure could organize the structure of matrix to decrease the gas solubility (salting-out) in terms of an entropy effect.^20,21 As is known, polyoxometalates (POMs) are nano-sized metal–oxygen cluster species with a diverse compositional range and an enormous structural variety;^22–25 one of the important properties of POMs is their capability to accept and stabilize various and large numbers of guest molecules in their crystal lattice.²⁴ Consequently, the result for Na₃[PW₁₂O₄₀] should be ascribed to the unique structure of [PW₁₂O₄₀]³⁻ (Fig. 1),²⁶ which could organize the structure of the water-swollen membrane to play a significant role in the decrease of both CO₂ and N₂ permeances. Compared with F⁻, importantly, POMs with an enormous structural variety²⁵ could give a wider platform for the salting-out effect to improve the permselectivity of membranes. Furthermore, the higher permselectivity of POMs at the lower concentration than that of fluoride salt indicates the stronger interaction between polyoxometalates and membranes, and thus the better stability in membranes. However, fluoride salt also has a good long-term performance in membranes, which has been demonstrated by Hägg et al..¹⁸

Fig. 1 Ball and stick representation of [PW₁₂O₄₀]³⁻ with a typical Keggin structure.²⁶ The color code is as follows: phosphorus (purple), tungsten (black) and oxygen (red).

Table 2 Comparison of oxysalt effect on permselectivity for the glutaraldehyde-crosslinked polyallylamine (M_w = 15 000) membranes cast on polyacrylonitrile supports (MWCO = 20 000)^ab

	Aqueous solution	CH3COONa (0.5M)	Na2SO4 (0.5 M)	Na3[PW12O40] (0.02 M)^d
a Membrane preparation: cast a polyallylamine solution on the supports, dried at room temperature, crosslinked with glutaraldehyde, and dipped in the respective solution for 24 h. b All the data measured at room temperature and 1 bar. c Permeance, in units of GPU, which is 10^−6 cm^3 (STP)/(cm^2 s cmHg). d Restricted from maximum solubility (<0.1 M at room temperature), 0.5 M is not applicable to Na3[PW12O40].
P CO2 (GPU)^c	8.4	3.4	2.0	1.7
α CO2/N2	50	83	199	1700

---

From the above, the salting-out effect on FSC facilitated transport membranes for CO₂ separation has been revealed by a series of experiments from fluoride ions (NaF) to polyoxometalates (Na₃[PW₁₂O₄₀]). It is an effect which may decrease the permeance of CO₂ to some extent but remarkably increase the selectivity, especially for fluoride and polyoxometalates (fluoride > chloride > bromide; polyoxometalates > sulfate > acetate), as a result of synergistic interaction between facilitated transport and the salting-out. As is known, solute diffusion across the facilitated transport membrane can take place by two mechanisms: (I) diffusion of the uncomplexed species, or (II) diffusion of the carrier–solute complex.²⁷ From this point of view, the experiment results in the current study indicate that the salting-out effect will have a dominant influence on the mechanism (I). Consequently, the N₂ permeance will decrease remarkably because the transport of N₂ across the membrane is by mechanism (I). Meanwhile, the CO₂ permeance will just decrease to some extent because the CO₂ transport is principally attributed to the mechanism (II). In addition, the salt concentration is certainly critical for the salting-out effect. Taking the salting-out effect of Na₃[PW₁₂O₄₀] on the facilitated transport glutaraldehyde-crosslinked polyallylamine membrane for example, the concentration of 0.02 M was chosen to carry out the experiment, resulting in P_CO2 = 1.7 GPU and α_CO2/N2 = 1700; when the concentration of 0.05 M was chosen, it resulted in P_CO2 = 1.4 GPU and α_CO2/N2 = 3200.

To go a step further, we discuss the salting-out effect on the water-swelling behavior of these FSC polymeric membranes. The FSC polymeric membranes showed higher permselectivity in a water-swollen condition than in a dry condition.^5–8 Indeed the swelling helps for CO₂ facilitated transport, considering the favorable water–CO₂ interactions which brings about the increased permeability along with an increase in membrane swelling as well as the increased water content in the membrane,¹⁵ but at the same time the swelling leads to a larger inter-chain spacing in the polymer, and thus a drastic increase in gas diffusion, which goes together with a reduced performance of the membranes in terms of selectivity. For example, we performed the experiment on a fully swollen uncrosslinked polyallylamine membrane by dipping in water for 3 h, resulting in P_CO2 = 195 GPU and α_CO2/N2 = 2.²⁸ In an attempt to overcome the instability of such a membrane, a proper chemical-crosslinking is an effective way to restrict the swelling. However, another issue emerged as the crosslinking through chemical valence bonding is often accompanied with remarkably decreased permeance of CO₂ because of the loss of amino groups and the decreased water content. This is documented from the data in Table 1 or 2, P_CO2 = 8.4 GPU and α_CO2/N2 = 50 after glutaraldehyde-crosslinking. From another point of view, if the adsorbed water molecules could be further availably stabilized in the swollen membrane by some means, the disadvantage of swelling can be overcome while its advantage can be maintained. Salts can be just the good candidate to stabilize the water molecules through hydration. We performed the experiment on the fully swollen uncrosslinked polyallylamine membrane by dipping in a 0.5 M NaF solution for 3 h, leading to P_CO2 = 47 GPU and α_CO2/N2 = 218.²⁹ The similar experiment was carried out by dipping the membrane into a Na₃[PW₁₂O₄₀] solution (0.02 M), leading to P_CO2 = 68 GPU and α_CO2/N2 = 160.³⁰ The results indicated that salts could offset the over-swelling disadvantage and impart the membrane with both high permeance and selectivity.

Finally, coming back to the beginning, in 1967 Ward and Robb firstly introduced the facilitated transport membrane for CO₂ separation by immobilizing an aqueous bicarbonate–carbonate solution into a porous support to get a CO₂/O₂ separation factor of 1500.¹ This supported liquid membrane (SLM) can be considered as the first example of the salting-out effect for CO₂ separation. The effect of bicarbonate–carbonate mixture in the membrane lowered the O₂ permeance by a factor of 40 but raised the CO₂ permeance by a factor of 2. This result seems inconsistent with our discussed salting-out effect which would decrease not only the N₂ permeance but also the CO₂ permeance according to the hydration theory. It should be noticed that in such a liquid membrane the hydrolysis of the bicarbonate–carbonate mixture, which is different from that of those weakly hydrolytic salts such as NaF and CH₃COONa reported in this work, could play a dominant role in the facilitated transport of CO₂ and should never be neglected. However, the salting-out effect still holds true anyway, considering that if the salting-out effect of the bicarbonate–carbonate mixture was taken out, the CO₂ permeance would have to be raised by a factor of more than 2. It can be visualized that if a highly hydrolytic salt such as Na₂CO₃ was added into the FSC polyallylamine membrane, the macroscopical decrease or increase of CO₂ permeance would lie on alkali competition between polyallylamine matrix and the highly hydrolytic salt.

In summary, we have for the first time revealed the salting-out effect on facilitated transport membranes for CO₂ separation. This effect may open a simple but practical way to balance the conflict between the permeance and selectivity in pursuit of high-permselective CO₂ separation membranes. Furthermore, to the best of our knowledge, polyoxometalates (POMs) as a large class of compounds were firstly introduced for CO₂ separation in this article. Because of their unique properties including molecular composition, shape, size, charge density and redox potentials, polyoxometalates can be attractive additives of membranes for CO₂ separation.

This research grant was supported by the Singapore National Research Foundation under its Environmental & Water Technologies Strategic Research Programme and administered by the Environment & Water Industry Programme Office (EWI) of the PUB (EWI RFP 0901-IRIS-04-06). We are also grateful to the Singapore Economic Development Board for funding the Singapore Membrane Technology Centre.

References

1. W. J. Ward and W. L. Robb, Science, 1967, 156, 1481.
2. J. H. Meldon, P. Stroeve and C. Gregoire, Chem. Eng. Commun., 1982, 16, 263.
3. F. F. Krull, C. Fritzmann and T. Melin, J. Membr. Sci., 2008, 325, 509.
4. H. Matsuyama, M. Teramoto and K. Iwau, J. Membr. Sci., 1994, 93, 237.
5. H. Matsuyama, K. Hirai and M. Teramoto, J. Membr. Sci., 1994, 92, 257.
6. M. Yoshikawa, K. Fujimoto, H. Kinugawa, T. Kitao and N. Ogata, Chem. Lett., 1994, 243.
7. T. Yamaguchi, L. M. Boetje, C. A. Koval, R. D. Noble and C. N. Bowman, Ind. Eng. Chem. Res., 1995, 34, 4071.
8. H. Matsuyama, M. Teramoto and H. Sakakura, J. Membr. Sci., 1996, 114, 193.
9. H. Matsuyama, M. Teramoto, H. Sakakura and K. Iwai, J. Membr. Sci., 1996, 117, 251.
10. T. Yamaguchi, C. A. Koval, R. D. Noble and C. N. Bowman, Chem. Eng. Sci., 1996, 51, 4781.
11. H. Matsuyama, K. Masui, Y. Kitamura, T. Maki and M. Teramoto, Sep. Purif. Tech., 1999, 17, 235.
12. H. Matsuyama, A. Terada, T. Nakagawara, Y. Kitamura and M. Teramoto, J. Membr. Sci., 1999, 163, 221.
13. A. Brunetti, F. Scura, G. Barbieri and E. Drioli, J. Membr. Sci., 2010, 359, 115.
14. Z. Wang, M. Li, Y. Cai, J. Wang and S. Wang, J. Membr. Sci., 2007, 290, 250.
15. L. Deng and M. -B. Hägg, J. Membr. Sci., 2010, 363, 295.
16. G. P. Pez, D. V. Laciak, US pat., 4 762 535, 1988.
17. W. S. W. Ho, US pat., 5 062 866, 1991.
18. M. B. Hägg, T. J. Kim, B. Li, US pat., 2008/0156188 A1, 2008.
19. The membrane was prepared by casting an aqueous poly(vinyl alcohol) (M_w = 89,000–124,000) solution onto polyacrylonitrile support (MWCO = 20,000), thermally crosslinking at 90 °C for 1 h, and dipping to water for 24 h. The membrane with NaF was prepared by dipping to a 0.5 M NaF solution for 24 h.
20. F. A. Long and W. F. Mcdevit, Chem. Rev., 1952, 51, 119.
21. J. M. Prausnitz, R. N. Lichtenthaler and E. Gomes de Azevedo, Molecular Thermodynamics of Fluid-Phase Equilibria (3rd ed.), Prentice Hall PTR, Upper Saddle River, N.J., 1999.
22. M. T. Pope and A. Müller, Angew. Chem., Int. Ed. Engl., 1991, 30, 34.
23. Special issue on polyoxometalates, ed. C. Hill, Chem. Rev., 1998, 98.
24. S. Uchida and N. Mizuno, Coord. Chem. Rev., 2007, 251, 2037.
25. D. -L. Long, E. Burkholder and L. Cronin, Chem. Soc. Rev., 2007, 36, 105.
26. J. F. Keggin, Proc. R. Soc. Lond. A, 1934, 144, 75.
27. R. D. Noble and C. A. Koval, ‘Review of Facilitated Transport Membranes’, inMaterials Science of Membranes for Gas and Vapor Separation, ed. Y. Yampolskii, I. Pinnau and B. D. Freeman, John Wiley & Sons, Ltd., Hoboken, N.J., 2006.
28. The membrane was prepared by casting an aqueous polyallylamine (M_w = 15,000) solution onto polyacrylonitrile support (MWCO = 20,000), drying at room temperature, and dipping to water for 3 h.
29. The membrane was prepared by casting an aqueous polyallylamine (M_w = 15,000) solution onto polyacrylonitrile support (MWCO = 20,000), drying at room temperature, and dipping to a NaF solution (0.5 M) for 3 h.
30. The membrane was prepared by casting an aqueous polyallylamine (M_w = 15,000) solution onto polyacrylonitrile support (MWCO = 20,000), drying at room temperature, and dipping to a Na₃[PW₁₂O₄₀] solution (0.02 M) for 3 h.

---

Footnotes

† Electronic Supplementary Information (ESI) available: Experimental set-up for gas permeation test. See DOI: 10.1039/c2ra20882b

‡ The experimental set-up for gas permeation test is shown in the Supporting Information. Modules were tested only in pure gas system. All the data were collected at room temperature with feed pressure 1 bar. Since all the membranes in this study were water-swollen in work condition and thus difficult to determine the exact selective layer thickness, the membrane permeation characteristics were evaluated by measuring the total CO₂ permeance of the composite membrane. Permeance of gas i is defined as the pressure-normalized flux of the gas through a membrane in units of GPU (10⁻⁶ cm³ (STP)/(cm² s cmHg)) given the symbol P_i, as expressed in eqn (1). The ideal selectivity (α) is defined as the ratio of permeance of pure CO₂ over pure N₂, as expressed in eqn (2). (1)
(2)
where J_i is the flux of component i in units of cm³ (STP)/(cm² s), Δp_i is the trans-membrane partial pressure difference in units of cmHg.

---