Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/seppur
Abstract
The transport number of counterions in ion-exchange membranes (Nafion, Neosepta) has been predicted using the
homogeneous and capillary models of transport. In the case of capillary model, two distributions of ions — described
by the traditional and modified by Gur Boltzmann equations — have been applied. It has been found that the
homogeneous model based on the experimental sorption and the capillary model, with the traditional Boltzmann
equation, predict the transport number relatively well. The only exception is the anion-exchange membrane Neosepta
AM1 in HCl solutions, because of facilitated transport of protons through that membrane. It has been proved that
the sorption given by the Boltzmann equation is always higher than that given by the modified Boltzmann equation
and than the Donnan sorption. The modified Boltzmann equation, successful in predicting partition coefficient in
Nafion 117 membrane equilibrated with H2SO4 solutions and salt mixture solutions, as reported by other authors,
failed here. © 2001 Elsevier Science B.V. All rights reserved.
Nomenclature
1383-5866/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 1 3 8 3 - 5 8 6 6 ( 0 0 ) 0 0 1 6 0 - X
644 S. Koter / Separation/Purification Technology 22-23 (2001) 643–654
Eq. (7) yields: equations of irreversible thermodynamics:
(1+ s̄)ū 01 s̄ ū 0 Ji = Li1(− 9ṽ1)+ Li2(− 9ṽ2)+ Li0(− 9v0)
t( 1 = 1+ 1 + 20 t( 0
N( m
0 0
(1+ s̄)ū 1 + s̄ū 2 ū 1
i= 0, 1, 2 (14)
(1+ s̄)
Eq. 8 s̄t(
1+ 0 (9) In Eq. (14), Ji is the flux of species i; 9ṽi is the
(1+s̄/t1) N( mt1
gradient of electrochemical potential of ion i; and
where 9v0 is the gradient of chemical potential of sol-
s̄ =c̄2/c̄m N( m =c̄0/c̄m (10) vent. The Lik coefficients fulfil the Onsager re-
ciprocity Lik = Lki. Using the definitions Eq. (1)
To estimate t( 1 from Eq. (9), the solvent content and Eq. (3), t( 1 and t( 0 can be expressed as:
and the sorption of electrolyte have to be mea-
sured. The sorption can also be calculated from L11 − L12
t( 1 = , (15)
the Donnan equation: L11 − 2L12 + L22
646 S. Koter / Separation/Purification Technology 22-23 (2001) 643–654
L01 −L02
t( 0 = (16) In Eq. (22) c̄i is the mean concentration of ions i
L11 − 2L12 +L22
in the pore solution, calculated from:
The Lik coefficients, averaged over the cross-sec-
tion of a capillary, are the sum of two compo- c̄i = 2
& 1
c̄i,rrdr i= 1, 2 (23)
nents [17]: 0
Bik = 2R 2
& & &
1
c̄i,rr
1
1 r
c̄k,rrdr 3 i,k= 0, 1, 2
electroosmotic transport number of water calcu-
lated on its basis (Eqs. (16) –(18)), t( 0(R), is equal
0 r pr 0 to the experimental one, t( 0(experimental) [20]:
(18)
t( 0(R)= t( 0(experimental) (24)
It depends on the capillary radius, R, the radial
distribution of species, c̄i,r and on the viscosity of
pore solution, p. The second term in Eq. (17), l( ik, 3. Membranes
describes the transport relative to the center of
mass of pore solution. All the coefficients are The investigated membranes are the cation-ex-
symmetrical, i.e. l( ik =l( ki, Bik =Bki and Lik =Lki. change membranes Nafion 120 (Du Pont de
To calculate Lik, two equations describing the Nemours) and Neosepta CM2 (Tokuyama Corp.),
distribution of ions are chosen. and the anion-exchange membrane, Neosepta
(1) The Boltzmann equation:
AM1. Their chemical composition and the basic
F properties (volume fraction of pores, Vp, concen-
c̄i,r = ci exp −zi (19) tration of fixed charges in the pore solution, c̄m,
RT
water content, N( m) are shown in Table 1.
(2) The Boltzmann equation modified by Gur et The dependence of the experimentally deter-
al. [18]:
mined concentration of sorbed electrolyte in the
pore solution c̄2, the real transport number of
F A 1 1
c̄i,r = ci exp −zi − i − (20) counterions t( 1 and the electroosmotic transport
RT RT mr,r mr,0
number of water t( 0 on the molality of external
In Eq. (20), Ai denotes hydration constant (ANa+ electrolyte are shown in Figs. 1–3. The method of
= 431, ACl− = 296 kJ/mol [18], AH + =555 kJ/ calculating c̄2 is described in Appendix A. t( 1 was
mol [19]), mr,0 is the dielectric constant at r= 0, calculated from t( 1,app and t( 0 using Eq. (2). The
where the gradient of electric potential is zero. experimental data needed to calculate c̄2, t( 1 and t( 0
The dependence of dielectric constant on the elec- were taken from: Nafion 120 membrane [21,22],
tric field in the charged capillary is described by AM1 [23] and CM2 [24].
the Booth equation [3,4]. The concentration profi- Compared to Nafion 120, the Neosepta mem-
les, needed for the calculation of Bik and c̄i, are branes are highly charged — their concentration
found by solving the Poisson equation: of fixed charges exceeds approximately two times
that of Nafion (Table 1). Therefore, the transport
F number of counterions in those membranes is
9(mr 9)= − %z c̄ (21)
m i i i,r much higher and closer to one (Fig. 2). The
with the boundary and electroneutrality exception is the anion-exchange membrane AM1
conditions: in HCl solutions, which rapidly looses its selectiv-
ity when the concentration of HCl increases. Two
((/(r)r = 0 =0 z1c̄1 +z2c̄2 +zmc̄m =0 (22) factors responsible for that effect are: the high
S. Koter / Separation/Purification Technology 22-23 (2001) 643–654 647
Table 1
Characteristics of ion-exchange membranes
Nafion 120 Saponified copolymer of sulphonyl fluoride Na+ 0.39 3.1 17.1
vinyl ether and tetrafluoroethylene
Neosepta CM2 Strongly acidic cation-exchange membrane Na+ 0.33 5.4 10.2
prepared by paste method [29]
H+ 0.27 7.3 7.5
Neosepta AM1 Strongly basic anion-exchange membrane Cl− 0.27 6.4 7.8
prepared by paste method [29]
4.2. Sorption
Table 2
The l (0)
ik coefficients, used to calculate l( ik (see Appendix A), and the relative viscosity of aqueous electrolyte solutions; cs =3 M,
T = 298 K
Electrolyte l (0)
11 (10
−10
mol2/J ms) l (0)
12 (10
−10
mol2/J ms) l (0)
22 (10
−10
mol2/J ms) p/p0 (10−10 mol2/J ms) [30]
a (0)
l ik were estimated from t1 and \ only; l (0)
12 has been assumed to be zero.
tions. With the increase of external concentration modification. This statement results from the
it increases and can exceed the experimental sorp- comparison of the product c̄1,rc̄2,r (z1= z2= 1):
tion. The modified Boltzmann equation gives the
worst estimation of the concentration of sorbed
Boltzmann
c̄1,rc̄2,r =
c 2s \ c 2s exp −
(A1 + A2) 1
−
1
RT mr mr,0
electrolyte (Fig. 6c). This is in contradiction with mod.Boltzmann
the results reported in literature [19]. For the = c̄1,rc̄2,r (25)
membrane Nafion 117 in the wide concentration
remembering that A1 + A2 \ 0 and mr,0 \ mr. Also
range of H2SO4 (0.1 – 3 M), Verbrugge and Hill
the Donnan equation (constant electric potential)
[19] found a good agreement between the experi-
yields lower sorption than the Boltzmann equa-
mental and calculated partition coefficient based tion (variable electric potential), as it is seen in
on the modified Boltzmann equation, whereas Fig. 6a,b. The proof of inequality:
that obtained using the Boltzmann equation was Donnan
about two times higher. The discrepancy between c̄1c̄2(Boltzmann)\ c 2s = c̄1c̄2 (26)
their results and the results presented here is
is given in Appendix C.
probably caused by the much higher diameter of
pores assumed by those authors, equal to the
4.3. Transport number of counterions
Bragg spacing 5.5 nm, as reported by Gierke [27].
One can conclude that the sorption is governed by
The concentration dependence of the ratio of
clusters, whereas the transport by channels con- calculated real transport number of counterions
necting clusters. Using the modified Boltzmann to the experimental one, t( 1/t( 1,exp, is shown in Fig.
equation Bontha and Pintauro [28] also calculated 7 (homogeneous model) and Fig. 8 (capillary
correctly the concentration of counterions in model).
Nafion 117 equilibrated with a mixture of alkali In the case of Nafion 120 membrane, both
metal chlorides (total external concentration 0.1 models predict t( 1 reasonably well. The exception
M). They used the diameter of pores the same as is the modified Boltzmann equation, which signifi-
in [19], whereas the hydration constant for Na+ cantly overestimates t( 1 because of too low sorp-
ions, estimated from their model for ion-solvent tion predicted by that distribution (Fig. 6c). A
energies in a membrane pore, was two times lower good agreement of the simple homogeneous
than that used in this paper (ANa+ =209 com- model with the experiment suggests that more
pared to 431 kJ/mol [18]). However, they did not elaborated theories of transport through Nafion
present the concentrations of sorbed electrolytes, membranes are not necessary for estimating t( 1.
much smaller in magnitude, whose deviation from Using the Donnan equation Gierke and Hsu [11]
the experimental values would be much higher calculated ca. 20% too low CE of Nafion mem-
than that of counterions. branes in the chlor-alkali electrolysis, whereas in
It should be noted that the concentration of this work the deviation of t( 1 is still within 5% for
sorbed electrolyte given by the Boltzmann equa- 4 m NaOH solution (Fig. 7b). The higher devia-
tion is always higher than that calculated by its tion observed in [11] probably results from the
650 S. Koter / Separation/Purification Technology 22-23 (2001) 643–654
applied higher concentration of NaOH. Unfortu- the homogeneous model underestimates t( 1, whereas
nately, the authors did not state for which concen- the capillary model gives satisfactory results — the
tration of NaOH CE was determined. deviation of t( 1 does not exceed 2%.
Regarding the cation-exchange membrane CM2, In the case of the anion-exchange membrane
the transport number of H+ counterions estimated AM1, equilibrated with NaCl solutions, both mod-
by both models agrees very well with the experi- els only slightly overestimate t( 1 (B 3%), whereas
mental one, which is close to 1 in the investigated when the membrane is in contact with HCl solu-
concentration range. This agreement results from tions, the calculated transport number of Cl−
the high mobility of H+, which is the dominating counterions is much too high. Even t( 1 based on the
factor influencing t( 1. In the case of NaCl solutions, experimental sorption, which is larger than in other
systems, exceeds : 30% the experimental value of
t( 1 in 1-m solution of HCl. Evidently, the mobility
of H+ is much higher than that of Cl− in compari-
son to their mobilities in a free solution. Certainly,
the mobility of Cl− in that membrane can also be
to some extent reduced, because also for NaCl
solutions the calculated t( 1 is slightly too high.
However, this effect should be rather small. Taking
into account the Grotthus mechanism of proton
migration [25], one can suppose that the electroos-
motic movement of water forced by Cl− counteri-
ons has a negligible influence on the mobility of
protons. Thus, assuming for H+ coions: ū2 : ū (0)2 ,
Eq. (9) would be changed to:
(1+s̄)ū 01 s̄ ū 02
t( 1 = 1+ t( (27)
(1+s̄)ū 01 + s̄ū 02 N( m ū 01 0
However, the transport numbers of Cl−, calculated
according to Eq. (27). is only a few percent (B
2.5%) smaller than that given by Eq. (9). Thus, it
should be concluded that the migration of H+ in
AM1 is facilitated.
5. Conclusions
2
Fig. 6. c̄2/c̄2,exp versus molality of external solution; (a) the Donnan equation Eq. (11), f 9 /(f( 1f( 2) =1; (b) CD, the Boltzmann equation
Eq. (19); (c) BDH, the modified Boltzmann equation Eq. (20).
Neither homogeneous nor capillary model water, ranging from 0.8 nm (CM2) to 1.3 nm
can predict the transport number of Cl− (Nafion 120), is too small to predict correctly the
counterions in the anion-exchange membrane sorption at low concentrations of external elec-
Neosepta AM1 in HCl solutions because of the trolyte (B0.5 M).
facilitated migration of H+ coions in that mem- It has been proved that the sorption given by
brane. the Boltzmann equation is always higher than
The equivalent pore radius of membrane pores that given by the modified Boltzmann equation
estimated from the electroosmotic transport of and the Donnan equation.
652 S. Koter / Separation/Purification Technology 22-23 (2001) 643–654
n 4 j,k = 1 c̄2, j
M2 z1 2
+ l (0) 1 1
M1 z 22
= c 2s n(2n − 1)(2n + 1)
n4 3
z 20 + 2z1z2 + z0(z1 +z2) (0) n
l12 = l 12 c̄2,k c̄2, j
z2 %
+ (2j −1)(2k − 1) +
j = 1,k = j + 1 c̄2, j c̄2,k
M 1 z2 z M z z (C4)
− 1− 1 l (0) − 2 1 1 − 2 l (0)
M2 z z 11 M1 z z 22
l22 = 1−
z2 2
l (0)
22 − 2
M 1 z2
1−
z2 (0)
l
In Eq. (C4) the last term fulfils the following
relation:
z
M1 z2 2
M2 z z 12
c̄2,k c̄2, j
+
! = 2 for c̄2, j = c̄2,k
(C5)
+ l (0) (B2) c̄2, j c̄2,k \ 2 for c̄2, j " c̄2,k
M2 z
11
In Eq. (B2), zi is the density of ion i, z is the In the case of nonuniform distribution of ions
density of solution. c̄2, j " c̄2,k (the Boltzmann equation) the following
The l( 0k coefficients were calculated from: inequality is fulfilled:
n
2 c̄2,k c̄2, j
% Mil( ik = 0 k = 0,1, 2 (B3) % (2j −1)(2k − 1) +
i=0 j,k = j + 1 c̄2, j c̄2,k
Eq. (B3) results from the definition of flux relative 1
\ (n− 1)n(3n 2 − n−1) (C6)
to the center of mass of solution. 3
654 S. Koter / Separation/Purification Technology 22-23 (2001) 643–654
From Eq. (C6) and Eq. (C4) it results that c̄1c̄2 \ [12] H. Reiss, I.C. Bassignana, J. Membr. Sci. 11 (1982) 219.
c 2s . The above proof is valid for any distribution [13] C. Selvey, H. Reiss, J. Membr. Sci. 23 (1985) 11.
[14] S. Koter, J. Membr. Sci. 108 (1995) 177.
function which locally fulfils the Donnan Eq.
[15] W.H. Koh, H.P. Silverman, J. Membr. Sci. 13 (1983) 279.
(C3). [16] A. Despic, G.J. Hills, Discuss. Faraday Soc. 21 (1956)
121.
[17] S. Koter, Pol. J. Chem. 71 (1997) 1305.
[18] Y. Gur, I. Ravina, A.J. Babchin, J. Coll. Interf. Sci. 64
References
(1978) 333.
[19] M.W. Verbrugge, R.F. Hill, J. Phys. Chem. 92 (1988)
[1] R. Schlögl, Stofftransport durch Membranen, Steinkopff, 6778.
Darmstadt, 1964.
[20] S. Koter, J. Membr. Sci. 166 (2000) 127.
[2] J.C. Fair, J.F. Osterle, J. Chem. Phys. 54 (1971) 3307.
[21] A. Nare˛bska, S. Koter and W. Kujawski, J. Membr. Sci.
[3] A.G. Guzmán-Garcia, P.N. Pintauro, M.W. Verbrugge,
25 (1985) 153; A. Nare˛bska, W. Kujawski, S. Koter, J.
R.F. Hill, AIChE J. 36 (1990) 1061.
Membr. Sci. 30 (1987) 125.
[4] E.H. Cwirko, R.G. Carbonell, J. Membr. Sci. 67 (1992)
[22] A. Nare˛bska, S. Koter, A. Warszawski, T.T. Le, J.
227.
[5] J.L. Anderson, W.H. Koh, J. Coll. Interface Sci. 59 Membr. Sci. 106 (1995) 25.
(1977) 149. [23] A. Nare˛bska, S. Koter, A. Warszawski, T.T. Le, unpub-
[6] Y. Oda, T. Yawataya, Bull. Chem. Soc. Jap. 29 (1956) lished results.
673. [24] N. Agmon, Chem. Phys. Lett. 244 (1995) 456.
[7] R. Arnold, D.A. Swift, Aust. J. Chem. 20 (1967) 2575. [25] D.G. Miller, J. Phys. Chem. 70 (1966) 2639.
[8] N. Lakshminarayanaiah, J. Phys. Chem. 73 (1969) 97. [26] T.D. Gierke, J. Electrochem. Soc. 124 (1977) 319C.
[9] R. Wódzki, A. Nare˛bska, J. Ceynowa, Angew Makromol. [27] J.R. Bontha, P.N. Pintauro, Chem. Eng. Sci. 23 (1994)
Chem. 106 (1982) 23. 3835.
[10] J. Ostrowska, A. Nare˛bska, Coll. Polym. Sci. 261 (1983) [28] S. Koter, unpublished results.
93. [29] Y. Mizutani, J. Membr. Sci. 49 (1990) 21.
[11] W.Y. Hsu, T.D. Gierke, J. Membr. Sci. 13 (1983) 307. [30] B.R. Breslau, I.F. Miller, J. Phys. Chem. 74 (1970) 1056.