- Email: [email protected]
Pervaporation of water—ethanol through ion exchange membranes
Journal of Membrane Science, 22 (1985) 333-344 Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands
PERVAPORATION MEMBRANES*
A. WENZLAFF, GKSS Research
OF WATER-ETHANOL
K.W BGDDEKER Center,
2054
THROUGH
333
ION EXCHANGE
and K. HATTENBACH
Geesthacht
(W. Germany}
(Received April 13, 1984, accepted m revised form September 18, 1984)
Summary The performance of ion exchange membranes in sorption and pervaporation of waterethanol has been studied as function of membrane type (ion loading) and feed composition, and has been compared with the performance of cellulose trlacetate. A relatively uniform pattern of preferential water sorption for all membranes studied, reasonably excepting the H’ form and CTA, contrasts sharply with distinct differences in selectivity and flux between cation exchange and anion exchange membranes. The key to an understanding of the mteractions involved appears to be the activity of the sorbed ethanol, which is increased in case of cation exchange membranes, and lowered in case of anion exchange membranes. Increased ethanol activity is traced to a salting-out effect ultimately leading to phase separation within the membrane fluid, as evidenced by a disproportionately high ethanol flux and the adverse effect of selective diffusion reducing overall selectivity. Lowering the ethanol activity enhances selectivity, thus favoring amon exchange membranes for the separation under consideration. When projected to comparable thickness, anion exchange membranes are superior to cellulose triacetate, the OH- form being the most promising of the membranes studied.
Introduction
Pervaporation may become the last of the long-recognized membrane separation effects to reach practical application, happily reconciling thermal and membrane separation processing. Transport phenomena in pervaporation are more complex than in other membrane processes because of the multiple interactions between feed constituents and membrane polymer manifested by swelling. “Selective sorption” to produce swelling constitutes one of two mechanisms of separation in pervaporation, the second mechanism being “selective diffusion” of the sorbed species across the membrane. Desorptlon from the downstream side into the vapor phase, being essentially complete, is generally considered to
*Paper presented in part at the 4th Symposium on Synthetic Membranes m Science and Industry, Tubingen, W. Germany, September 6-9, 1983.
0376-‘7388/85/$03
30
o 1985 Elsevier Science Publishers B.V.
334
be non-selective. Swelling and diffusive mobility are interrelated, and both are affected by the feed composition in contact with the membrane. Observing that water is a prefered solvent for electrolytes whereas ethanol is not, the selectivity of ion exchange membranes (considered as stationary electrolytes) towards sorption and pervaporation of water-ethanol mixtures was studied. Comparison of the compositions of sorbed membrane fluid and permeate (pervaporate) at corresponding immersion and feed compositions, respectively, is expected to yield information on the prevailing mechanism of separation and transport. Membranes The ion exchange membranes used are identified as CMV (cation exchange) and AMV (anion exchange) by Asahi Glass Co., Japan; they were selected because of their temperature resistance and high ion exchange capacity [ 11. Both membrane types consist of a crosslinked styrene/butadiene base, mechanically stabilized by a poly(viny1 chloride) backing. The fixed ion exchange functions are sulfonic acid (cation exchange) and quarternary ammonium moieties (anion exchange), respectively. Pertinent membrane data are given in Table 1. The following counter ion loadings were used in the present study: CMV:
H’, Li’, Na+, K’, NH,‘, Mg’+, A13’
AMV:
OH-, Cl-, SCN-, SO:-
There is very little variation of wet thickness with the nature of the counter ion. For comparison with previous work [2,3], cellulose triacetate (CTA) membranes were used, as suggested by the high rate of water vapor transmission of this polymer [4]. Homogeneous CTA membranes of 40 pm dry thickness were cast from dichloromethane solution. Membrane thickness directly affects pervaporation flux, but has only a minor effect on selectivity. Among cellulose acetates, the efficiency of water-ethanol separation improves with increasing degree of acetylation and with increasing chain length (solution viscosity). TABLE
1
Properties of commercial Membrane
CMV AMV
ion exchange membranes [l]
Ion exchange capacity (edkg)
Water content (wt. %)
Thickness incl. backing (pm)
2.4 1.9
25 19
150 140
335
Experimental Sorption (imbibition) and pervaporation experiments were carried out on mixtures of 10 to 97 wt.% ethanol in 10% increments under no-recovery conditions. With two membranes (AMV--SO:- and CTA), pervaporation was performed also at 99 wt.% ethanol to establish the selectivity trend. Temperatures were 20°C in sorption, and 60°C in pervaporation, the effect of temperature on pervaporation selectivity being insignificant m the range of 20 to 60°C. Membrane conditioning (swelling) on gradually increasing the ethanol fraction of the feed is complete within minutes, whereas conditioning in the reverse direction takes hours. Conditioning times were verified by observing the time dependence of pervaporation performance. The effect is attributed to the preferentially absorbed water producing a swelling pressure which is readily released as water is replaced by ethanol. Immersion samples of typically 100 cm2 membrane area were blotted dry, weighed, and quickly transferred into a micro-distillation unit which had been purged with dry air prior to receiving the sample. The fluid released from the membrane was collected in a micro cold trap at liquid mtrogen temperature while maintaining the sample at 60°C for 2 hours. The sorbate collected was weighed and analyzed by gas chromatography. Sorption and desorption of water-ethanol by ion exchange membranes was found to be completely reversible through repeated cycles. For the pervaporation experiments, reverse osmosis test cells, modified for downstream low pressure operation, were used (Fig. 1). The heat capacity
Fig. 1. Pervaporation test cell, membrane area 20 cm’.
336
of the massive steel cells stabilizes the process temperature, which was maintained by thermostating the feed reservoir. Experimental conditions were: Membrane area per cell, 20 cm’; 4 cells in series; feed flow rate, 50 l/h; feed temperature, 6O”C, with control runs at 20 and 40°C; downstream pressure, 20 mbar; running-in period, 1 hr; permeate collection in U-traps in liquid nitrogen; typical collection time, 1 hr; analysis of the weighed permeate samples by gas chromatography or Karl Fischer titration. Selectivity The equilibrium absorption at 20°C of water and ethanol by various ion exchange membranes and cellulose triacetate (CTA) as function of immersion (feed) composition is illustrated in Fig. 2. Presentation is in terms of weight of sorbate per 100 gram of dry membrane, backing included (PHR = per 100 CMV-H+
Et I
a 2
x
AMV-OH-
LO 30 20 10
b
u-l
20
LO
60
80
I
I
I
I
20
LO
60
80
CMV-Na+ I
AMV- CII
LO 30 20 10
AMV-
CMV- Mg 2+
20
CMV-AI
I
LO
80
I
I
60
80
CTA
3+
I
20
60
LO
SO,‘-
Feed
I
20
LO
60
80
Concentratron(wt%H20)
Fig. 2. Sorption of water and ethanol by ion exchange membranes and cellulose triacetate as function of immersion composition (PHR = per 100 parts resin)
337
parts resin). As expected, preferential sorption of water is observed throughout the concentration range, with only minor differences between ion loadings of either membrane type being apparent. The exceptionally high absorption of water and ethanol by the strongly acidic H’ cation exchange membrane is attributed to a dominating effect of repulsive forces between the fixed charges of the ion exchange matrix [5]. Moreover, absorption of ethanol is enhanced by acidification. The resulting separation effect in terms of concentration of absorbed water vs. concentration of water in the immersion (feed) liquid is shown in Figs. 3 and 4 (dashed curves denote sorbate). On the presumption that equilibrium absorption resembles the entrance condition in actual pervaporation, these curves represent the “selective sorption”‘contribution to the separation effect.
Feed
20
40
60
80
20
LO
60
80
Concentrahon Iwt “/oHz01
Fig. 3. Separation of water-ethanol by selective sorption (- - -) (-) through cation exchange membranes.
and by pervaporation
The separation of water-ethanol by pervaporation through the membranes under study is represented in Figs. 3 and 4 alongside the sorption curves (solid curves denote permeate), the concentrations now referring to permeate composition vs. feed composition, water being the preferentially pervaporating species. In the range studied, 20 to 6O”C!,temperature has little effect on the selectivity of the membranes. Interpretation of the curves in Figs. 3 and 4 is as follows. If sorbate and permeate curves coincide, “selective sorption” is the only mechanism of
338
separation operative. If, on the other hand, the two curves differ, there is an additional contribution of “selective diffusion” to the overall separation effect. Significant differences between sorption and pervaporation selectivity are observed with cation exchange membranes, excepting the H’ form; the pervaporation selectivity is lower than the selectivity due to sorption (Fig. 3). Selective diffusion, rather than improving the overall separation effect, thus counteracts selective sorption, as has been observed by others [6]. The enhanced mobility of the absorbed ethanol in case of cation exchange membranes is reflected by the flux behavior discussed below. With anion exchange membranes a minor shift in the opposite direction is noticed, selective diffusion slightly improving the overall separation effect (Fig. 4). Neither effect is apparent with the H’ cation exchange membrane or with cellulose triacetate.
80 60 LO 20
20
LO
60
80
20
Feed Concentrahon
LO
60
80
(wt %H20)
Fig. 4. Separation of water-ethanol by selective sorption (- --) (-) through anion exchange membranes and cellulose triacetate.
and by pervaporation
The separation factors, defined as ratio of water-ethanol fractions of permeate and feed, are shown as function of feed composition in Fig. 5, for cation exchange membranes, and Fig. 6, for anion exchange membranes, respectively. By tendency, cation exchange membranes in water-ethanol pervaporation exhibit close-lying separation factors of the order of 1 to 10, with optimum separation performance at an intermediate range of feed compositions (Fig. 5). The H’ cat,ion exchange membrane again behaves exceptionally. The separation factors derived for anion exchange membranes are consistently higher, and reach optimum values at low water (high ethanol)
339
concentrations of the feed (Fig. 6). Highest water-ethanol separation is observed with the SO:- anion exchange membrane, with a separation factor of 70 at a narrow range of feed concentrations around 97%. At still higher ethanol feed concentration, separation factors decline sharply.
10
20
30
LO
50
60
70
80
90
C
wt % Hz0 Fig. 5. Separation factors in water-ethanol membranes.
(A-B)
pervaporation
through cation exchange
n
sok2-
10
20
30
LO
50 C wt % Hz0
Fig. 6. Separation factors in waterethanol membranes and cellulose triacetate.
60
(A-B)
70
80
90
pervaporation
through anion exchange
340
Flux The total pervaporation flux of water-ethanol through ion exchange membranes at 60°C as function of feed composition is presented in Fig. 7, for cation exchange membranes, and Fig. 8, for anion exchange membranes, respectively. A reciprocal relationship between flux and selectivity is anticipated, and is generally observed, although there are particularities in detail.
80 60
10
20
30
LO
50
60
70
80
90
C wt % Hz0 Fig. 7. Gross pervaporation flux of water-ethanol as function of feed composition.
through
cation
exchange
membranes
5C Lc -0
*%
30 y” 20 10
10
20
30
LO
P
50
60
70
80
90
L
wt %HzO Fig. 8. Gross pervaporation flux of water-ethanol through cellulose triacetate as function of feed composition.
anion exchange
membranes
and
341
Gross fluxes with cation exchange membranes show a considerable spread; the sudden permeability increase at feed concentrations between 15 and 20% water (80 to 85% ethanol), observed in case of the Na+, K’ and NH: forms, is reversible (Fig. 7). Flux values through anion exchange membranes are altogether lower and more concise, and decrease regularly as the water content of the feed is lowered (Fig. 8). The gross pervaporation fluxes provide little insight into the transport mechanism; however, they may be rationalized by considering the individual contributions. This is illustrated in Figs. 9 and 10, in which the total flux, J, is divided into the fluxes of water, JHzO, and ethanol, JEtOH, for typical membranes under study. For an analysis of the flux curves, it is recalled that, excepting the H+ form and CTA, sorption of water-ethanol by ion exchange membranes follows a relatively uniform pattern of preferential water sorption (Fig. 2) not suggestive of the striking difference in flux behavior between cation exchange and anion exchange membranes. This difference is apparent in the exceptionally steep increase of ethanol permeation through cation exchange membranes (Fig. 9) as compared to the almost unaffected ethanol flux through anion exchange membranes (Fig. 10). Water flux, by comparison,
60
CMV f MQ” 60
Feed
20 50 60 Cmcentratlon
Fig. 9. Pervaporation flux of water-ethanol down into partial fluxes.
80 (wt %H
through cation exchange membranes; break-
342 AMV-OH-
“In-m-r-7
20
40
60
80
100
20
AWV-SO:-
LO
60
80
100
20 40 60 Concentmtlon
80 (wt%
100 Hz01
CTA
60
20
LO
60
80
100
Feed
Fig. 10 Pervaporation flux of water-ethanol through anion exchange membranes and cellulose triacetate; breakdown into partial fluxes.
correlates to a first approximation with the water content of the feed for all membranes under consideration (albeit not for cellulose triacetate). The disproportionate increase of ethanol permeability reflects an increase in ethanol activity as would be the result of salting-out of the nonelectrolyte out of aqueous solution upon addition of salt [7]. It is therefore suggested that the enhanced permeability of ethanol in cation exchange membranes is the result of a salting-out effect within the membrane fluid. Phase separation, which is the ultimate salting-out effect, would produce two completely uncoupled sorbed phases, the relative diffusivities of which determine the process, Since, on account of lower ionic interactions, the ethanol phase is higher in diffusivity than the aqueous phase, the salted-out ethanol is expected to dominate flux and selectivity of pervaporation through cation exchange membranes. This is corroborated by the extreme ethanol permeability observed in case of the monovalent cation exchange membranes (CMV-Na’, K+, NH:) at low water content of the feed (Fig. 7). Enhanced mobility of the sorbed ethanol fraction also reduces the overall separation effect through selective diffusion, as is revealed by comparing the selectivities of sorption and pervaporation (Fig. 3). Salting-out leading to decoupling of fluxes is
343
thus seen to be an adverse effect in the pervaporation of water-ethanol through ion exchange meinbranes. Membrane evaluation Considering the effect of ion loading on the observed performance, it must be concluded that there are different interactions fortuitously producing a fairly similar pattern of selective sorption for all ion exchange membranes studied, with the reasonable exception of the H+ form. The basic interactions may be summarized as follows. With cation exchange membranes, interaction of water with dissociated cations 1s strongest, with consequent increase in activity of the ethanol; with anion exchange membranes, ionic hydration is weaker; however, there are significant interactions of water with ethanol and with the membrane polymer. In line with the condition that lowering the activity of the sorbed ethanol improves selectivity, anion exchange membranes are found to show considerably higher separation effects than cation exchange membranes. Optimum separation factors as function of feed composition, along with the corresponding flux values at 6O”C, are given in Table 2. The higher the selectivity, the narrower the range of feed compositions to which it applies. TABLE 2 Feed composition, separation factor and flux at point of highest selectivity in wate? ethanol pervaporation through anion exchange membranes and cellulose triacetate at 60°C Membrane
H, O/EtOH (wt.%)
Separation factor, (Y
Flux (kg/m*-d)
AMV-so:-
3197 7193 15185 20/80 13187
70 53 38 27 14
2.0 9.2 11.2 20.0 22.0
AMV-OHAMV-SCNAMV-clCTA
The high permeation rate through cellulose triacetate in spite of its relatively low water sorption is due to the lower thickness of this membrane [2], 40 pm for CTA vs. 140 pm for the AMV types. When projecting the performance of the anion exchange membranes to an equal footing of 40 pm, assuming an inverse proportionality between flux and thickness, the OHform clearly emerges as the most promising of the membranes studied. It is concluded that low-thickness anion exchange membranes, the OH- form in particular, exhibit inherently favorable properties for the dehydration of ethanol by pervaporation.
344
References 1 K. Kneifel and K. Hattenbsch, Properties and long-term behavior of ion exchange membranes, Desalination, 34 (1980) 77-95. 2 K.W. Boddeker, A. Wenzlaff and D. Cavigelli, Membranverfahren in der Bioalkoholgewinnung, Report GKSS 81/E/62,1982. 3 M.H.V. Mulder, J. Oude Hendrikman, H Hegeman and C.A. Smolders, Ethanol-Water separation by pervaporation, J. Membrane Sci., 16 (1983) 269-284. 4 R.E. Kesting, Synthetic Polymeric Membranes, McGraw-Hill, New York, 1971, p. 279. 5 F. Helfferich, Ionenaustauscher, Verlag Chemie, Weinheun, 1959, pp. 127, 439. 6 S.-T. Hwang and K. Kammermeyer, Membranes in Separations, Wiley, New York, 1975, p. 107. 7 G.N. Lewis, M. RandaIl, K S. Pitzer and L. Brewer, Thermodynamics, 2nd edn., McGraw Hill, New York, 1961, p. 584.
PERVAPORATION MEMBRANES*
A. WENZLAFF, GKSS Research
OF WATER-ETHANOL
K.W BGDDEKER Center,
2054
THROUGH
333
ION EXCHANGE
and K. HATTENBACH
Geesthacht
(W. Germany}
(Received April 13, 1984, accepted m revised form September 18, 1984)
Summary The performance of ion exchange membranes in sorption and pervaporation of waterethanol has been studied as function of membrane type (ion loading) and feed composition, and has been compared with the performance of cellulose trlacetate. A relatively uniform pattern of preferential water sorption for all membranes studied, reasonably excepting the H’ form and CTA, contrasts sharply with distinct differences in selectivity and flux between cation exchange and anion exchange membranes. The key to an understanding of the mteractions involved appears to be the activity of the sorbed ethanol, which is increased in case of cation exchange membranes, and lowered in case of anion exchange membranes. Increased ethanol activity is traced to a salting-out effect ultimately leading to phase separation within the membrane fluid, as evidenced by a disproportionately high ethanol flux and the adverse effect of selective diffusion reducing overall selectivity. Lowering the ethanol activity enhances selectivity, thus favoring amon exchange membranes for the separation under consideration. When projected to comparable thickness, anion exchange membranes are superior to cellulose triacetate, the OH- form being the most promising of the membranes studied.
Introduction
Pervaporation may become the last of the long-recognized membrane separation effects to reach practical application, happily reconciling thermal and membrane separation processing. Transport phenomena in pervaporation are more complex than in other membrane processes because of the multiple interactions between feed constituents and membrane polymer manifested by swelling. “Selective sorption” to produce swelling constitutes one of two mechanisms of separation in pervaporation, the second mechanism being “selective diffusion” of the sorbed species across the membrane. Desorptlon from the downstream side into the vapor phase, being essentially complete, is generally considered to
*Paper presented in part at the 4th Symposium on Synthetic Membranes m Science and Industry, Tubingen, W. Germany, September 6-9, 1983.
0376-‘7388/85/$03
30
o 1985 Elsevier Science Publishers B.V.
334
be non-selective. Swelling and diffusive mobility are interrelated, and both are affected by the feed composition in contact with the membrane. Observing that water is a prefered solvent for electrolytes whereas ethanol is not, the selectivity of ion exchange membranes (considered as stationary electrolytes) towards sorption and pervaporation of water-ethanol mixtures was studied. Comparison of the compositions of sorbed membrane fluid and permeate (pervaporate) at corresponding immersion and feed compositions, respectively, is expected to yield information on the prevailing mechanism of separation and transport. Membranes The ion exchange membranes used are identified as CMV (cation exchange) and AMV (anion exchange) by Asahi Glass Co., Japan; they were selected because of their temperature resistance and high ion exchange capacity [ 11. Both membrane types consist of a crosslinked styrene/butadiene base, mechanically stabilized by a poly(viny1 chloride) backing. The fixed ion exchange functions are sulfonic acid (cation exchange) and quarternary ammonium moieties (anion exchange), respectively. Pertinent membrane data are given in Table 1. The following counter ion loadings were used in the present study: CMV:
H’, Li’, Na+, K’, NH,‘, Mg’+, A13’
AMV:
OH-, Cl-, SCN-, SO:-
There is very little variation of wet thickness with the nature of the counter ion. For comparison with previous work [2,3], cellulose triacetate (CTA) membranes were used, as suggested by the high rate of water vapor transmission of this polymer [4]. Homogeneous CTA membranes of 40 pm dry thickness were cast from dichloromethane solution. Membrane thickness directly affects pervaporation flux, but has only a minor effect on selectivity. Among cellulose acetates, the efficiency of water-ethanol separation improves with increasing degree of acetylation and with increasing chain length (solution viscosity). TABLE
1
Properties of commercial Membrane
CMV AMV
ion exchange membranes [l]
Ion exchange capacity (edkg)
Water content (wt. %)
Thickness incl. backing (pm)
2.4 1.9
25 19
150 140
335
Experimental Sorption (imbibition) and pervaporation experiments were carried out on mixtures of 10 to 97 wt.% ethanol in 10% increments under no-recovery conditions. With two membranes (AMV--SO:- and CTA), pervaporation was performed also at 99 wt.% ethanol to establish the selectivity trend. Temperatures were 20°C in sorption, and 60°C in pervaporation, the effect of temperature on pervaporation selectivity being insignificant m the range of 20 to 60°C. Membrane conditioning (swelling) on gradually increasing the ethanol fraction of the feed is complete within minutes, whereas conditioning in the reverse direction takes hours. Conditioning times were verified by observing the time dependence of pervaporation performance. The effect is attributed to the preferentially absorbed water producing a swelling pressure which is readily released as water is replaced by ethanol. Immersion samples of typically 100 cm2 membrane area were blotted dry, weighed, and quickly transferred into a micro-distillation unit which had been purged with dry air prior to receiving the sample. The fluid released from the membrane was collected in a micro cold trap at liquid mtrogen temperature while maintaining the sample at 60°C for 2 hours. The sorbate collected was weighed and analyzed by gas chromatography. Sorption and desorption of water-ethanol by ion exchange membranes was found to be completely reversible through repeated cycles. For the pervaporation experiments, reverse osmosis test cells, modified for downstream low pressure operation, were used (Fig. 1). The heat capacity
Fig. 1. Pervaporation test cell, membrane area 20 cm’.
336
of the massive steel cells stabilizes the process temperature, which was maintained by thermostating the feed reservoir. Experimental conditions were: Membrane area per cell, 20 cm’; 4 cells in series; feed flow rate, 50 l/h; feed temperature, 6O”C, with control runs at 20 and 40°C; downstream pressure, 20 mbar; running-in period, 1 hr; permeate collection in U-traps in liquid nitrogen; typical collection time, 1 hr; analysis of the weighed permeate samples by gas chromatography or Karl Fischer titration. Selectivity The equilibrium absorption at 20°C of water and ethanol by various ion exchange membranes and cellulose triacetate (CTA) as function of immersion (feed) composition is illustrated in Fig. 2. Presentation is in terms of weight of sorbate per 100 gram of dry membrane, backing included (PHR = per 100 CMV-H+
Et I
a 2
x
AMV-OH-
LO 30 20 10
b
u-l
20
LO
60
80
I
I
I
I
20
LO
60
80
CMV-Na+ I
AMV- CII
LO 30 20 10
AMV-
CMV- Mg 2+
20
CMV-AI
I
LO
80
I
I
60
80
CTA
3+
I
20
60
LO
SO,‘-
Feed
I
20
LO
60
80
Concentratron(wt%H20)
Fig. 2. Sorption of water and ethanol by ion exchange membranes and cellulose triacetate as function of immersion composition (PHR = per 100 parts resin)
337
parts resin). As expected, preferential sorption of water is observed throughout the concentration range, with only minor differences between ion loadings of either membrane type being apparent. The exceptionally high absorption of water and ethanol by the strongly acidic H’ cation exchange membrane is attributed to a dominating effect of repulsive forces between the fixed charges of the ion exchange matrix [5]. Moreover, absorption of ethanol is enhanced by acidification. The resulting separation effect in terms of concentration of absorbed water vs. concentration of water in the immersion (feed) liquid is shown in Figs. 3 and 4 (dashed curves denote sorbate). On the presumption that equilibrium absorption resembles the entrance condition in actual pervaporation, these curves represent the “selective sorption”‘contribution to the separation effect.
Feed
20
40
60
80
20
LO
60
80
Concentrahon Iwt “/oHz01
Fig. 3. Separation of water-ethanol by selective sorption (- - -) (-) through cation exchange membranes.
and by pervaporation
The separation of water-ethanol by pervaporation through the membranes under study is represented in Figs. 3 and 4 alongside the sorption curves (solid curves denote permeate), the concentrations now referring to permeate composition vs. feed composition, water being the preferentially pervaporating species. In the range studied, 20 to 6O”C!,temperature has little effect on the selectivity of the membranes. Interpretation of the curves in Figs. 3 and 4 is as follows. If sorbate and permeate curves coincide, “selective sorption” is the only mechanism of
338
separation operative. If, on the other hand, the two curves differ, there is an additional contribution of “selective diffusion” to the overall separation effect. Significant differences between sorption and pervaporation selectivity are observed with cation exchange membranes, excepting the H’ form; the pervaporation selectivity is lower than the selectivity due to sorption (Fig. 3). Selective diffusion, rather than improving the overall separation effect, thus counteracts selective sorption, as has been observed by others [6]. The enhanced mobility of the absorbed ethanol in case of cation exchange membranes is reflected by the flux behavior discussed below. With anion exchange membranes a minor shift in the opposite direction is noticed, selective diffusion slightly improving the overall separation effect (Fig. 4). Neither effect is apparent with the H’ cation exchange membrane or with cellulose triacetate.
80 60 LO 20
20
LO
60
80
20
Feed Concentrahon
LO
60
80
(wt %H20)
Fig. 4. Separation of water-ethanol by selective sorption (- --) (-) through anion exchange membranes and cellulose triacetate.
and by pervaporation
The separation factors, defined as ratio of water-ethanol fractions of permeate and feed, are shown as function of feed composition in Fig. 5, for cation exchange membranes, and Fig. 6, for anion exchange membranes, respectively. By tendency, cation exchange membranes in water-ethanol pervaporation exhibit close-lying separation factors of the order of 1 to 10, with optimum separation performance at an intermediate range of feed compositions (Fig. 5). The H’ cat,ion exchange membrane again behaves exceptionally. The separation factors derived for anion exchange membranes are consistently higher, and reach optimum values at low water (high ethanol)
339
concentrations of the feed (Fig. 6). Highest water-ethanol separation is observed with the SO:- anion exchange membrane, with a separation factor of 70 at a narrow range of feed concentrations around 97%. At still higher ethanol feed concentration, separation factors decline sharply.
10
20
30
LO
50
60
70
80
90
C
wt % Hz0 Fig. 5. Separation factors in water-ethanol membranes.
(A-B)
pervaporation
through cation exchange
n
sok2-
10
20
30
LO
50 C wt % Hz0
Fig. 6. Separation factors in waterethanol membranes and cellulose triacetate.
60
(A-B)
70
80
90
pervaporation
through anion exchange
340
Flux The total pervaporation flux of water-ethanol through ion exchange membranes at 60°C as function of feed composition is presented in Fig. 7, for cation exchange membranes, and Fig. 8, for anion exchange membranes, respectively. A reciprocal relationship between flux and selectivity is anticipated, and is generally observed, although there are particularities in detail.
80 60
10
20
30
LO
50
60
70
80
90
C wt % Hz0 Fig. 7. Gross pervaporation flux of water-ethanol as function of feed composition.
through
cation
exchange
membranes
5C Lc -0
*%
30 y” 20 10
10
20
30
LO
P
50
60
70
80
90
L
wt %HzO Fig. 8. Gross pervaporation flux of water-ethanol through cellulose triacetate as function of feed composition.
anion exchange
membranes
and
341
Gross fluxes with cation exchange membranes show a considerable spread; the sudden permeability increase at feed concentrations between 15 and 20% water (80 to 85% ethanol), observed in case of the Na+, K’ and NH: forms, is reversible (Fig. 7). Flux values through anion exchange membranes are altogether lower and more concise, and decrease regularly as the water content of the feed is lowered (Fig. 8). The gross pervaporation fluxes provide little insight into the transport mechanism; however, they may be rationalized by considering the individual contributions. This is illustrated in Figs. 9 and 10, in which the total flux, J, is divided into the fluxes of water, JHzO, and ethanol, JEtOH, for typical membranes under study. For an analysis of the flux curves, it is recalled that, excepting the H+ form and CTA, sorption of water-ethanol by ion exchange membranes follows a relatively uniform pattern of preferential water sorption (Fig. 2) not suggestive of the striking difference in flux behavior between cation exchange and anion exchange membranes. This difference is apparent in the exceptionally steep increase of ethanol permeation through cation exchange membranes (Fig. 9) as compared to the almost unaffected ethanol flux through anion exchange membranes (Fig. 10). Water flux, by comparison,
60
CMV f MQ” 60
Feed
20 50 60 Cmcentratlon
Fig. 9. Pervaporation flux of water-ethanol down into partial fluxes.
80 (wt %H
through cation exchange membranes; break-
342 AMV-OH-
“In-m-r-7
20
40
60
80
100
20
AWV-SO:-
LO
60
80
100
20 40 60 Concentmtlon
80 (wt%
100 Hz01
CTA
60
20
LO
60
80
100
Feed
Fig. 10 Pervaporation flux of water-ethanol through anion exchange membranes and cellulose triacetate; breakdown into partial fluxes.
correlates to a first approximation with the water content of the feed for all membranes under consideration (albeit not for cellulose triacetate). The disproportionate increase of ethanol permeability reflects an increase in ethanol activity as would be the result of salting-out of the nonelectrolyte out of aqueous solution upon addition of salt [7]. It is therefore suggested that the enhanced permeability of ethanol in cation exchange membranes is the result of a salting-out effect within the membrane fluid. Phase separation, which is the ultimate salting-out effect, would produce two completely uncoupled sorbed phases, the relative diffusivities of which determine the process, Since, on account of lower ionic interactions, the ethanol phase is higher in diffusivity than the aqueous phase, the salted-out ethanol is expected to dominate flux and selectivity of pervaporation through cation exchange membranes. This is corroborated by the extreme ethanol permeability observed in case of the monovalent cation exchange membranes (CMV-Na’, K+, NH:) at low water content of the feed (Fig. 7). Enhanced mobility of the sorbed ethanol fraction also reduces the overall separation effect through selective diffusion, as is revealed by comparing the selectivities of sorption and pervaporation (Fig. 3). Salting-out leading to decoupling of fluxes is
343
thus seen to be an adverse effect in the pervaporation of water-ethanol through ion exchange meinbranes. Membrane evaluation Considering the effect of ion loading on the observed performance, it must be concluded that there are different interactions fortuitously producing a fairly similar pattern of selective sorption for all ion exchange membranes studied, with the reasonable exception of the H+ form. The basic interactions may be summarized as follows. With cation exchange membranes, interaction of water with dissociated cations 1s strongest, with consequent increase in activity of the ethanol; with anion exchange membranes, ionic hydration is weaker; however, there are significant interactions of water with ethanol and with the membrane polymer. In line with the condition that lowering the activity of the sorbed ethanol improves selectivity, anion exchange membranes are found to show considerably higher separation effects than cation exchange membranes. Optimum separation factors as function of feed composition, along with the corresponding flux values at 6O”C, are given in Table 2. The higher the selectivity, the narrower the range of feed compositions to which it applies. TABLE 2 Feed composition, separation factor and flux at point of highest selectivity in wate? ethanol pervaporation through anion exchange membranes and cellulose triacetate at 60°C Membrane
H, O/EtOH (wt.%)
Separation factor, (Y
Flux (kg/m*-d)
AMV-so:-
3197 7193 15185 20/80 13187
70 53 38 27 14
2.0 9.2 11.2 20.0 22.0
AMV-OHAMV-SCNAMV-clCTA
The high permeation rate through cellulose triacetate in spite of its relatively low water sorption is due to the lower thickness of this membrane [2], 40 pm for CTA vs. 140 pm for the AMV types. When projecting the performance of the anion exchange membranes to an equal footing of 40 pm, assuming an inverse proportionality between flux and thickness, the OHform clearly emerges as the most promising of the membranes studied. It is concluded that low-thickness anion exchange membranes, the OH- form in particular, exhibit inherently favorable properties for the dehydration of ethanol by pervaporation.
344
References 1 K. Kneifel and K. Hattenbsch, Properties and long-term behavior of ion exchange membranes, Desalination, 34 (1980) 77-95. 2 K.W. Boddeker, A. Wenzlaff and D. Cavigelli, Membranverfahren in der Bioalkoholgewinnung, Report GKSS 81/E/62,1982. 3 M.H.V. Mulder, J. Oude Hendrikman, H Hegeman and C.A. Smolders, Ethanol-Water separation by pervaporation, J. Membrane Sci., 16 (1983) 269-284. 4 R.E. Kesting, Synthetic Polymeric Membranes, McGraw-Hill, New York, 1971, p. 279. 5 F. Helfferich, Ionenaustauscher, Verlag Chemie, Weinheun, 1959, pp. 127, 439. 6 S.-T. Hwang and K. Kammermeyer, Membranes in Separations, Wiley, New York, 1975, p. 107. 7 G.N. Lewis, M. RandaIl, K S. Pitzer and L. Brewer, Thermodynamics, 2nd edn., McGraw Hill, New York, 1961, p. 584.