80 (1991) 71-84 Elsevier Science Publishers B.V., Amsterdam
Pervaporation and Dehydration of WaterEthanol Mixtures by Means of Cuprophan Hollow Fiber Membranes JIN SHENG* and J.C. MORA** Syrinx Research Institute, Level 61, MLC Centre, Sydney 2&W (Austmlia). Tel: 61-071-527644; Fax: 61-071-524402
(Received December 8, 1990)
Cuprophan hollow fiber membranes were used in the pervaporation of water-ethanol mixtures. The experimental work was aimed at studying the possibility of breaking the azeotrope and dehydrating the water-ethanol mixtures. The experimental results obtained under various operating conditions are presented and discussed. The transfer model based on the studies of experimental results is also presented in this paper. Key words: Pervaporation,
hollow fiber membrane, ultrafiltration
is an efficient
for the fractionation
mixtures. As a membrane separation method with no phase change, it consumes less energy and has very few undesirable side effects on the product quality [1,2]. Research work on pervaporation is now widely carried out both in laboratories and the chemical industry and has resulted in considerable advances over recent years. Industrial scale pervaporators are now readily available commercially [3-51.
*To whom correspondence
should be addressed.
**Professor of National Polytechnic Institute, Toulouse (France). Current mailing address: Asian Institute of Technology, GPO Box 2754, Bangkok 10501 (Thailand).
0 1991 Elsevier Science Publishers B.V.
Pervaporation is different from other membrane separation processes due to the change in physical state of the penetrate (pervaporate) from liquid to vapor in the downstream interface of the membrane. The accessory equipment needed for the process includes a vacuum system or an inert gas stream to ensure efficient mass transfer and pervaporate removal. The solution-diffusion model is generally used to interpret the mass transfer mechanism in pervaporation. The selectivity of the membrane is inversely proportional to the pervaporate flux. It is usually the more important parameter to take into consideration rather than flux density because it is generally hard to obtain both a high pervaporate flux and a high selectivity at the same time. Due to this fact, a large membrane area is needed for a pervaporator if production demand is high. In recent years research work has concentrated on the development of suitable membrane materials for separating the water-ethanol mixtures. A number of polymers have been studied. These include cellulose acetate and its derivates [6,7], poly(N-vinylpyrrolidone) , polyacrylamide , ionexchange membrane (Nafion” [lo] and polyvinyl alcohol [ll]. Another area of interest has been membrane and module configuration. The hollow fiber membrane generally seems better than flat membrane sheets because from these it is easy to manufacture modules with a high membrane surface/module volume ratio. The problem of non-ideal flow of liquid mixtures in the membrane module can also be reduced, compared with that in the flat and spiral wound membrane modules. These effects result in the feasibility of fractionation of the water-ethanol mixtures using hollow fiber membranes, and it is for this purpose that the experiments reported in the present investigation were carried out. The mass transport phenomenon was studied, and from this a transfer model was established to predict production data.
The hollow fiber membrane used in these experiments was composed of cuprammonium regenerated cellulose (Cuprophan). The outside diameter of the fiber was 200 pm and with a membrane area of a single cartridge (Fig. 1) of 0.9 m2. The membrane modules used contained two or eight single cartridges connected in series, defined herein as Module-Al and Module-A2, respectively.
Feed (Liquid) Fig. 1. Single Cuprophan hollow fiber membrane cartridge.
Reservoij of ~atða~ol
brvaporate Rotameter apparatus:
Fig. 2. Pervaporation installation.
A schematic diagram of the pervaporation installation is shown in Fig. 2. The membrane module was placed in a temperature controlled chamber in which the temperature was maintained identical with that of the water-ethanol mixture in the inlet position of the membrane module. Each was carried out for 1 h. The samples of pervaporate, feed and retentate were analysed
with a refractometer. The pervaporate was condensed and recovered using a cold water-glycerol mixture (- 12”C). The flow rate of the liquid mixtures stream (Q) varied from 5-20 l/h. The pervaporate
flux J (mole/m2s) is calculated using Eqn. 1:
J = m/St
where m, S and t are the amount of recovered pervaporate (mole), membrane surface area in the module (m2) and experimental time (s), respectively. The separation factor (Yis defined as follows : 0
Y/WY) x /( 1 -x)
where y and x are the molar fractions of water in the pervaporate liquid mixture, respectively.
(2) and in the
RESULTS AND DISCUSSION
Experimental results The experimental results for pervaporate flux and the separation factor are presented in Figs. 3-6. It can be seen that the pervaporate flux J is a concentration dependent function. It increases as the water composition x increases. A higher flux is obtained when x is greater than 0.5. The variation of the separation factor cxin relation to the water composition x is inversely proportional to the pervaporate flux. The membrane favours water penetration with lower water compositionsx. Higher values of the separation factor ar (higher than 6) were obtained only when x is lower than 0.1. It was found that both the pervaporate flux J and the separation factor a! change significantly as the water composition x approaches 0.2. This composition is close to but differs from the azeotropic point (x = 0.1057). This phenomenon may be caused by special interactions between the membrane being used because similar phenomena have been reported previously by other researchers working with different membranes [6,7,10, 12-141.
15.0 Experimental Results : 0 w. (Q - 5 l/h! 0 c( (Q -20 l/h) Te-35’C S-1.8 a 2 Module-AI
Experimental Results : rti(Q51/h) l @ (Q -20 I/h) Te-30°C S-1.8 m2 Module-A1
& 8 3
1 Q 6.0
Calculated Results 1’
Calculated Results 1’ /’
Te = 30 “C
b. Te = 3J%
1.0 *II 0 2
Fig. 3. Effect of feed composition on separation factor (Module-Al).
. Module-AZ Te-30°C S-7.2 m 2 Experimental Results m o( IQ - 5 I/h) 2.5. g! (Q -12 l/h) l &IQ -20 I/h)
. Module-A2 Te-35’C S-7.2 m 2 3.0-A Experimental Results 0 Cf(Q - 5 I/h) A d (Q -12 l/h) 0 M (Q -20 I/h)
C+culated Results /’
Te = 30°C
Fig. 4. Effect of feed composition on separation factor (Module&U).
Te = 35’C
0.8 1.0 *H20
Experimental Results : 0 J(Q- 51/h) 0 J (Q -20 I/h) Te-35% S-1.b m *
Experimental Results : m J (Qb- 5 I/h) c, F. 25.0 l J (Qb-20 I/h) Te-30*c S-Lam2 5,” Module-AI f ri
Calculated Results 0.4
b. Te = 35’C
a. Te = 30°C
1.0 XII 0 2
Fig. 5. Effect of feed composition on penetrant flux (ModuleAl). 7.5 k
S-7.2 m 2
),j ,*: c c
5.0 ,s: 4.5 u
3.0 2.5 2.0 l
J (Q -20 I/h)
Te-35’C S-7.2 m ’
d 6.0 3 5.5
g,g $ &
i lated Results
_J 1 0
,o 0 A
Experimental Results 0 J (Q - 5 I/h) A J (Q -I2 I/h) 0 J (Q -20 I/h)
1.0 0.0 0.2 0.4 0.6 0.8 1.0 XHO a. Te = 3o”c 2 b. Ts - 35*C Fig. 6. Effect of feed composition on penetrant flux (ModuleA2). 0.2
It can be seen that the membrane favours the penetration of ethanol rather than water when the water composition x is higher than 0.2. This composition may be considered as the “transition point” for the pervaporation of water-ethanol mixtures. It is seen from Figs. 3-6 that both the pervaporate flux and separation factor obtained in Module-Al (S = 1.8 m2) are higher than those obtained in ModuleLA2 (S = 7.2 m2) under similar experimental conditions. The effect of temperature on mass transfer is also shown in Figs. 3-6. The relationship between the pervaporate flux and the temperature of liquid mixture can be interpreted by Arrhenius’ law. It may be seen that the separation factor does not change significantly when the temperature of the liquid mixture stream is varied. The heat transport phenomenon in this case is important because large decreases in temperature ( > 7” C) were observed in the experiment, not only in the liquid mixture stream but also between the liquid mixture stream and the pervaporate vapor. The phenomenon has been studied in detail .
Transfer model It was interesting to study the variation in pervaporate flux [J(x)] and separation factor [cz(x)]due to water composition x. It may be seen in Figs. 3 and 4 that the variation of separation factor Q in relation to the variation of water composition x can be interpreted by the following model shown in polynomial form: a(x)
where C,, C2 and c3 are characteristic parameters of the model. The discrete experimental results are analysed by the least quadratic method to obtain parameters C,, C2 and C,. The parameters obtained are presented in Table I.
TABLE I Characteristic parameters of model (r(x)
Membrane surface area in a module (m2)
Range of flow rate
- 0.550 - 0.495
It is observed in Figs. 5 and 6 that the variation of pervaporate flux in relation to water composition x can be also interpreted by a model shown in polynomial form, which is similar to Eqn. 3. However, the influence of variations in temperature of the liquid mixture stream Te (“C) and flow rate Q(l/h) on the pervaporate flux should also be taken into consideration. This is realized by replacing parameter C3 in Eqn. 3 with flow rate Q and temperature Te dependent function y(x,Te,Q) and Arrhenius form function E(Te) shown as follows: J(x) =
c;(x +Cz’)”+ CTY (x,RQ) We)
or J(X) =
c; (x + c;)2 + C; v(x,%Q)
R(273.15 + Te)
where C;, Ci and C; are characteristic parameters of the model.!(x); Ep and R are the apparent activation energy (kJ/mol) and universal gas constant (kJ/mol “K), respectively. The values are obtained from the numerical treatment of experimental results. They are presented in Table II. TABLE II Characteristic parameters of model J(x) Membrane surface area in a module
Range of flow rate
0.292 - 0.068
1.589 lo* 4.630 lo5
The function incorporates the effects of temperature and the flow rate of the liquid mixture streams on mass transfer. It contains a temperature dependent function f(x,Te) and a flow rate dependent function g(x) shown in Eqn. 6: Y(4 %Q) = F(x,WS-)
where f(x,Te), g(x) are given in the following form: f(x, Te) = [Te /30]ctx+c’
G(x) = [Q/5] C~+c~
with C4, C5, C6 and C7 characteristic parameters. In the case where Module-A2 is used, the function g(x) has a minor effect on mass transfer as compared with function f(x,Te), and consequently it is neglected in the model J(x). Due to this, Eqn. 8 may be simplified into the following form:
,t(x, re, Q)
The values of parameters (74, C5, C6 and C7 are obtained from the numerical treatment of experimental results through simulation studies. The results are presented in Table III. It can be observed in Figs. 3-6 that there is good agreement between the results calculated through models J(x) and or(x) and those obtained experimentally.
The experimental results show that the mass transfer mechanism has both pervaporation and ultrafdtration components. This is expected because Cuprophan is not a truly dense membrane, unlike the ones which are commonly used in the pervaporation process. However, it is important to know the effects of the two phenomena on the mass transfer mechanism. Furthermore, it needs to be determined whether these two mechanisms take place at the same time and whether one of them can play a predominant part in the mass transfer mechanism under the particular conditions being used. The experimental results show that, at low water compositionx, penetrant fluxes are low but the separation factors are high. The penetrant evaporates instantaneously and is removed from the membrane surface. This phenomenon indicates that the pervaporation effect in" this case plays the major role in the mass transfer mechanism. Due to the hydrophilic nature of Cuprophan, a high water composition (x > 0.3) results in a more swollen membrane than when low water compositions are used. This effect relaxes the membrane structure and makes the penetration of liquid much easier. As a result, this effect causes an increase
of penetrant flux and decrease in separation factor. This phenomenon becomes more noticeable as the water composition x increases. As the penetrant flux increases, part of the penetrant may not be evaporated and removed instantaneously from the membrane surface. Consequently, a thin liquid film may form on the membrane surface in the downstream side interface. This effect makes the mass transfer mechanism transform from a liquid-vapor one into a liquid-liquid one. This is the ultrafiltration effect. Evidently, in this case the ultrafiltration effect plays the main part in the mass transfer mechanism, while pervaporation plays only a minor role. The ultrafiltration effect becomes more important when water compositionx increases. It enhances the penetrant flux and decreases the separation factor because a reduced pressure cannot be seen directly by the membrane surface. However, this effect is not an important consideration because there is no economical advantage in carrying out fractionation of water-ethanol mixtures by membrane processes at such concentrations. There is a transition composition (x -0.2) between these two composition fields where both pervaporation and ultrafiltration effects play their parts in mass transfer. This phenomenon is shown in Fig. 7. The experimental results show that both the pervaporate flux and separation factor have large rates of change in this field. 15.0 ‘d 2$
s 3 (d
Module-Al Te=30CC S-1.8 m 2 Experimental Results : 8 (Q - 5 l/h) l (Q -20 I/h) A (Q -12 l/h) Brvaporation
Fig. 7. Effect of feed compo&tion on the mass transfer mechanism.
These effects provide evidence that the fractionation of water-ethanol mixtures starting at low water composition (~~0.2) by means of hollow fiber membranes is feasible. It was found that the pervaporate fluxes increase as the flow rate of the liquid mixtures increases. In other words the pervaporate flux depends on the Reynolds number of the fluid flow. This phenomenon is shown in Figs. 5 and 6, respectively. It is probable that this is due to the characteristics of the membrane. The variation of Reynolds number which corresponds to the different flow rates of water-ethanol liquid mixtures is given in Fig. 8; it takes into account the physical properties of the water-ethanol mixtures vs. water composition x.
2 Fig. 8. Variation of the Reynolds numbers vs. feed composition.
The experimental results indicate that the separation factor CYdoes not change significantly when the flow rate of liquid mixture changes. This suggests that the interaction between the membrane and water, due to the hydrophilic nature of the membrane, has a greater effect on the mass transfer than the physical structure of the membrane, e.g., the pore size and tortuosity of the membrane. Results also indicate that the pervaporate flux and separation factor obtained in the larger surface area membrane module (S = 7.2 m2) are lower than those obtained in the smaller surface membrane module (S = 1.8 m2). This may be caused by the heat transport phenomenon and nonideal flow of the liquid mixtures at the upstream interface of the membrarie.
It is commonly considered that the heat of vaporization for pervaporate is supplied by the liquid mixture. This causes a temperature drop in the liquid mixture. To interpret this phenomenon, it can be assumed that the membrane surface in each module is divided in small sequential zones. Each zone produces a local pervaporate flux. Evidently the temperature of liquid mixture as well as local pervaporate flux contacting the membrane in these zones decreases longitudinally according to Arrhenius’ law [ 161. Due to the fact that the penetrant flux presented in Figs. 5 an 6 is the mean value of all local fluxes produced in the module, the penetrant flux obtained in the larger surface area module is consequently lower than that obtained from the smaller area. On the other hand, it can also be considered that the pervaporate recovered in Module-A2 may be produced at a mean temperature Te” which is between temperatures Te and Ts. This assumption may be helpful in explaining the different between the penetrant fluxes obtained in ModuleAl and Module-A2 shown in Figs. 5 and 6. In addition, the effect may indicate that a very large membrane surface area may not be suitable for a pervaporator unless there are enough intermediate heating cells in the upstream side of each membrane cell. The effect on the mass transfer mechanism due to the non-ideal flow of liquid mixture has been studied previously . This effect may be very important for a large size flat configuration membrane module. Contrarily, this problem is less significant and is minimized in modules using hollow fiber membranes. CONCLUSIONS The
results from the present investigation show that hollow fiber membranes can be used in the fractionation and dehydration of water-ethanol mixtures by means of the pervaporation process. Transfer models are proposed on the basis of the studies on experimental results. They incorporate a number of variables such as water composition, temperature and the flow rate of liquid mixtures. They can be used as a simulation basis for optimizing the design of a pervaporator. The results estimated using the mathematical models are in good agreement with the ones obtained experimentally. The experimental results indicate that both heat transport and non-ideal flow phenomena of the liquid mixture in the membrane module have significant effects on the mass transfer mechanism. ACKNOWLEDGEMENT The experimental work was carried out in the laboratory of E.N.S.I.G.C., National Polytechnic Institute, Chemin de la Loge, 3 1078 Toulouse (France).
1 P. Aptel, N. Challard and J. Neel, Application of the pewaporation process to separate azeotropic mixtures, J. Membrane Sci., 1(1976) 271. 2 I. FreMesson, G. Tragardh and B. Hahn-Hagerdal, Pervaporation and ethanol upgrading-a literature review, Chem. Eng. Commun., 45 (1986) 277. 3 A.H. Ballweg, H.EA. Bruchke, N.H. Schneider and GA. Tusel, Pervaporation membrane-an economical method to replace conventional dehydration and rectitlcation column in ethanol distilleries. Paper presented at the 5th Inter. Fuel Tech. Symp., May, 1982, Auckland, New Zealand. 4 U. Sander and P.-B. Soukup, Practical experienceswith pervaporation systems for liquid and vapor separation. Paper presented at the 3rd Inter. Conf. on Pewaporation Proc. in the Chem. Ind., Sept. 1988, Nancy, France. 5 A. Asada, Dehydration of organic solvents: some actual results of pewaporation plants in Japan. Paper presented at the 3rd Inter. Conf. on Pervaporation Proc. in the Chem. Ind., Sept. 1988, Nancy, France. 6 E. Nagy, 0. Borlai and J. Stehnaszek, Pewaporation of alcohol-water mixtures on cellulose hydrate membrane, J. Membrane Sci., 16 (1983) 79. 7 M.H.V. Mulder, J.O. Hendrikman, H. Hegeman and CA. Smolders, Ethanol-water separation by pervaporation, J. Membrane Sci., 16 (1983) 269. 8 P. aptel, J. Cuny, J. Jozefnowitz, G. Morel and J. Neel, Liquid transport through membrane prepared by grafti of polar monomers onto polytetrafluroethylene fii. I. Some fractions of liquid mixtures by pervaporation, J. Appl. Poly. Sci., 16 (1972) 1061. 9 R. Fries and J. Neel, J. Chim. Phys., 494 (l%S). 10 J. Neel, W. Kujawski, Q.T. Ngyuen and Z.H. Ping, Mechanism of pervaporation: selectivity of ion-exchange membrane for the separation of water-ethanol mixtures. Paper presented at the 3rd Inter. Conf. on Pervaporation Proc. in the Chem. Ind., Sept. 1988, Nancy, France. 11 N.R. Jarvis and R.Y.M. Huang, Separation of liquid mixtures by using polymer membrane, II. Permeation of aqueous alcohol solution through cellophane and poly(viny1 alcohol), J. Appl. Poly. Sci., 14 (1970) 2341. 12 K.W. Boddeker and A. Wenzlaff, Pewaporation with ion-exchange membranes. Paper presented at the 1st Inter. Conf. on Pervaporation Proc. in the Chem. Ind., Feb. 1986, Englewood, New Jersey, USA. 13 J. Neel, Dehydration des melange eau-ethanol par pervaporation. Paper presented at the annual meeting of “Membrane Club-EDG,” Jan. 1987, Paris, France. 14 T. Uragami, Permeation and separation characteristics of aqueous alcohol solution through hydrolic and hydrophobic membranes by pervaporation and evaporation. Paper presented at the 3rd Inter. Conf. on Pervaporation Proc. in the Chem. Ind., Sept. 1988, Nancy, France. 15 J. Sheng and J. Mora, Study the heat transport phenomenon in pervaporation by means of highly compact hollow fiber bundle. Paper presented at the 4th Inter. Conf. on Pewaporation Proc. in the Chem. Ind., Dec. 1989, Ft. Lauderdale, Florida, USA. 16 J. Sheng, R.S. Bes and J.C. Mora, Separation of water-ethanol systems by pervaporation: the two dimensional transport phenomenon by means of highly compact hollow fiber bundle. Paper presented at the 2nd annual national meeting of the North American Mem. Sot., June 1988, Syracuse, NY, USA. 17 B. Wandelt, Etude du fonctionnement d’un module de pervaporation op6rant en continu pour deshydrater un m6lange hydro-organique, Intemat. Research Report, L.C.P.M.E.N.S.I.C., Aprill988, Nancy, France.