Liquid Separation by Membrane Pervaporation: A Review - Industrial

1048

Ind. Eng. Chem. Res. 1997, 36, 1048-1066

Liquid Separation by Membrane Pervaporation: A Review Xianshe Feng† and Robert Y. M. Huang* Department of Chemical Engineering, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1

Pervaporation is one of the most active areas in membrane research, and the pervaporation process has been shown to be an indispensable component for chemical separations. In this paper, the recent development in pervaporation membranes and pervaporation processes is reviewed, and some outstanding questions involved in membrane pervaporation are discussed with emphasis on the following issues: mass transport in the membrane, membrane material selection, concentration polarization in the boundary layer, pressure buildup in hollow fiber membranes, asymmetric and composite membranes, and the activation energy for permeation. We attempt to provide insight into this dynamic field and to highlight some of the outstanding problems yet to be solved or clarified. Background During the past decade, industrial membranes have established themselves as indispensable components of chemical processing industries. Membrane-based technology is currently regarded as a new frontier of chemical engineering and has been widely used for the purification, concentration, and fractionation of fluid mixtures. Pervaporation is a relatively new membrane separation process that has elements in common with reverse osmosis and membrane gas separation. In pervaporation, the liquid mixture to be separated (feed) is placed in contact with one side of a membrane and the permeated product (permeate) is removed as a lowpressure vapor from the other side (Figure 1). The permeate vapor can be condensed and collected or released as desired. The chemical potential gradient across the membrane is the driving force for the mass transport. The driving force can be created by applying either a vacuum pump or an inert purge (normally air or steam) on the permeate side to maintain the permeate vapor pressure lower than the partial pressure of the feed liquid. Vacuum pervaporation, which is customarily referred to as the standard pervaporation, is the most widely utilized mode of operation, while inert purge pervaporation is normally of interest if the permeate can be discharged without condensation. Besides these two modes of operation, there are several other process variants, including thermal pervaporation, perstraction or osmotic distillation, saturated vapor permeation, and pressure-driven pervaporation (Franken et al., 1990; Neel, 1991; Goncalves et al., 1990). Some of them are really process hybrids rather than process variants. Recently, electrically induced pervaporation has also been attempted by Timashev et al. (1994). Though pervaporation is one of the most popular areas of current membrane research, the concept of pervaporation separation is not new. The phenomenon of pervaporation was first observed by Kober (1917), who originated the term in a publication reporting selective permeation of water from aqueous solutions of albumin and toluene through collodion (cellulose * Author to whom correspondence may be addressed. Phone: (519)885-1211. Fax: (519)746-4979. E-mail: [email protected]. † Current address: Alberta Research Council, P.O. Box 8330, Edmonton, Alberta, Canada T6H 5X2. S0888-5885(96)00189-3 CCC: $14.00

Figure 1. Schematic diagram of the pervaporation process. (a) Vacuum pervaporation, (b) purge gas pervaporation.

nitrate) films. The usefulness of pervaporation for separation and concentration was recognized in 1935 by Farber. However, the first known quantitative work on pervaporation was published by Heisler et al. (1956) for the separation of water/ethanol mixtures using a cellulose membrane. It was the work of Binning and co-workers (1958, 1961, 1962) that established the principles and highlighted the potential of pervaporation technology. Although the research work was continued for several years and many patents were obtained, the permeation flux was too low to be economically useful. A breakthrough was achieved in the early 1980s when Gesellschaft fu¨r Trenntechnik (GFT) Co. developed a composite membrane comprised of a thin layer of crosslinked poly(vinyl alcohol) supported on a porous poly(acrylonitrile) substrate. A pervaporation process for dehydrating ethanol was then commercialized. In the following years, substantial work was done that widened the research scope to many liquid mixtures and a variety of membranes. Membranes made of both synthetic polymers and derivatives of natural polymers have been tested for the separation of various liquid mixtures including, for example, alcohols-water, acetone-water, methanol/methyl tert-butyl ether, methanol/ pentane, toluene-heptane, and isomeric xylenes. The applications of pervaporation can be classified into three categories: (i) dehydration of organic solvents, (ii) removal of organic compounds from aqueous solutions, and (iii) separation of anhydrous organic mixtures. Currently, pervaporation has been commercialized for two applications: one is the dehydration of © 1997 American Chemical Society

Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997 1049

alcohols and other solvents, and the other is the removal of small amounts of organic compounds from contaminated waters. In the latter application, pollution control and solvent recovery are effected simultaneously. There are also some other promising applications such as aroma recovery and beer dealcoholization in the food industry (Lee and Kalyani, 1991; Karlsson and Tragardh, 1993, 1994; Lamer et al., 1994) and product recovery from fermentation broths for enhanced bioconversions (Friedl et al., 1991; Groot et al., 1991a,b; Vrana et al., 1993; Dettwiler et al., 1993; Geng and Park, 1994). The separation of organic-organic mixtures is presently the least developed application of pervaporation because of the problems normally associated with membrane stability under relatively harsh conditions, but it represents the largest opportunity for energy and cost savings. The separation of methanol/methyl tert-butyl ether/C4’s azeotropes is now being actively investigated due to commercial interest of producing octane enhancers for gasolines (Chen et al., 1989; Farnand and Noh, 1989; Shah et al., 1989; Doghieri et al., 1994; Nakagawa and Matsuo, 1994; Park et al., 1995; Chen and Martin, 1995). Once a suitable membrane is available, pervaporation can also find a niche in some reversible reactions such as esterification and condensation reactions to remove one or more product species selectively, thereby shifting the equilibrium toward the product side (David et al., 1991a,b, 1992; Okamoto et al., 1991, 1993; Bagnell et al., 1994). The application of pervaporation has now been extended to include dehydration of aqueous electrolyte solutions (Schaetzel et al., 1993). Interestingly, membrane pervaporation is also being investigated as a chemical sensor in instrumental analysis (Mattos and de Castro, 1994; Papaefstathiou et al., 1995; Papaefstathiou and de Castro, 1995). Compared to other membrane separation processes, pervaporation is in a far less advanced state. So far, research has been mainly concentrated on dense flat membrane sheets. Most of them are found to be selective for water permeation, and only a few are selective to the permeation of organic compounds. The permeation flux in pervaporation is generally low. This problem can, in principle, be solved or compensated by reducing the effective membrane thickness and/or increasing the membrane area packing density. From an application point of view, asymmetric membranes and hollow fiber configuration are of great interest. Unfortunately, little effort has been made to prepare defectfree integral asymmetric membranes, and few reports have been found in the literature on the systematic study of hollow fiber membrane configuration. The understanding of pervaporation separation is incomplete. It is thought at present by some researchers that the preferential sorption of a component in the membrane is the prerequisite to the preferential permeation of the component. Based on this idea, several approaches for selection of membrane materials have been proposed. On the other hand, an ideal sorption of liquid in polymer is generally assumed to describe the mass transfer using the commonly accepted solutiondiffusion model. This controversy affects proper understanding of the pervaporation mechanism and appropriate selection of membrane materials. For a membrane to be efficient for a specific separation, it is always desirable to have a membrane with good permeability and selectivity. However, the hydrodynamic conditions of the flow for feed and, sometimes, for permeate cannot be overlooked. Concentration

polarization is inherent to all membrane processes, and the boundary layer effect is expected to be more significant for highly permselective membranes. Although there are already numerous studies showing the significance of concentration polarization in pervaporation, some fundamental questions are not yet well addressed. The literature on pervaporation is extensive. Many reports and articles provide a substantial amount of information on the development of pervaporation technology (Neel, 1991; Fleming and Slater, 1992; Zhang and Drioli, 1995). Among others, the book edited by Huang (1991) provides an extensive treatment of theory and practice of the process. In this paper, we attempt to give a review of the current status of pervaporation membranes and pervaporation separation processes, with emphasis on highlighting some outstanding issues. Characteristics of Pervaporation Pervaporation separation is governed by the chemical nature of the macromolecules that comprise the membrane, the physical structure of the membrane, the physicochemical properties of the mixtures to be separated, and the permeant-permeant and permeantmembrane interactions. Pervaporation transport is usually described to be a three-step process: solution-diffusion-evaporation. The separation is based on the selective solution and diffusion, i.e., the physical-chemical interactions between the membrane material and the permeating molecules, not the relative volatility as in distillation. Therefore, pervaporation is commonly considered to be a profitable complement to distillation for the separation of azeotropic and close-boiling mixtures, which requires at present the use of energy-intensive processes. Pervaporation can be operated at low feed pressures and at ambient temperature or even below this, and no additional chemicals are needed for separation. Therefore, pervaporation can be applied in biotechnology to the concentration to heat-, stress-, and/or chemicalsensitive biochemicals. The early study of Farber (1935) demonstrated the effectiveness of pervaporation for concentrating dilute protein and enzyme solutions. Unlike reverse osmosis, pervaporation transport is not limited by osmotic pressure because the driving force for mass transfer through the membrane is provided by lowering the chemical potential of the permeate stream on the downstream side, and consequently, the feed pressure is not critical. For example, pervaporation can be used for concentrating ethanol in an aqueous solution from 85 to more than 99 wt %, while an extremely high operating pressure is needed to overcome the osmotic pressure if reverse osmosis is used. For a given membrane and a given liquid mixture, both the permeation flux and separation factor in pervaporation are higher than in reverse osmosis (Tanimura et al., 1990). As in reverse osmosis, the liquid in contact with the membrane tends to dissolve into it and cause membrane swelling. Swelling tends to alter the membrane properties and generally leads to higher permeability and lower selectivity. On the permeate side of the membrane, the partial vapor pressure of a component affects its permeation rate significantly. Hence, the downstream vapor pressure must be maintained as low as economically feasible to maximize the driving force for the permeation. This issue is particularly important in designing hollow fiber membrane units because the buildup in permeate vapor pressure could be devastating to the separation performance.

1050 Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997

Unlike reverse osmosis and membrane gas separation, pervaporation involves a phase change of permeating species from the liquid to the vapor state. Consequently, energy is needed for the vaporization of the permeate. Thus, from an energy consumption point of view, pervaporation is especially promising when the concentration of the preferentially permeating species in the feed is low. In principle, the heat of vaporization required for permeation can be supplied either in the feed liquid or by a sweeping fluid on the permeate side or directly to the membrane. Direct heating of the membrane has been tested recently by Wnuk and Chmiel (1992) and Boddeker et al. (1993). It has not yet been determined which way is more efficient. Pervaporation membrane devices are modular in construction. There is no significant economy of scale, so they can be used in either large or small processing capacity. Moreover, it is easy to integrate pervaporation units with other suitable techniques so that the hybrid processes would be more effective than a scheme where the full separation is effected by either technique alone. Compactness, flexibility, simplicity, and versatility are some other strong points of the pervaporation process. In a comprehensive assessment of fluid separation techniques, pervaporation is ranked the third highest among the 31 techniques under evaluation (Bravo et al., 1986). Requirements for Membranes Pervaporation is a rate-controlled separation process. In developing pervaporation membranes, three issues must be addressed: (i) membrane productivity, (ii) membrane selectivity, and (iii) membrane stability. Membrane productivity is a measure of the quantity of a component that permeates through a specific area of membrane surface in a given unit of time. Membrane productivity is frequently characterized by permeation flux, J, which relates the product rate to the membrane area required to achieve the separation. Note that permeation flux depends on both the intrinsic permeability and the effective thickness of a membrane. The commercialization of the pervaporation technique is, to a large extent, attributed to the engineering approach of making thin membranes in asymmetric and composite forms. When describing the selectivity of a membrane for the separation of a mixture composed of components A and B, the separation factor is defined as

R)

(1 -Y Y) (1 -X X)

(1)

where X and Y are the molar fractions of the more permeable component A in the feed and permeate, respectively. Note that the numerical value of R is independent of the concentration units used, as being the ratio of ratios. When the separation factor is unity, no separation occurs; when it approaches infinity, the membrane becomes perfectly “semipermeable”. It is the membrane selectivity that forms the basis for separating a mixture. It should be pointed out that only when the concentration polarization is negligible will the selectivity expressed by eq 1 be an intrinsic property of the membrane. Otherwise, the feed concentration on the membrane surface has to be substituted for X; concentration polarization will be discussed later. Generally, membrane permeability and selectivity have to be determined experimentally.

Occasionally, membrane selectivity is expressed in terms of the enrichment factor, β, which is simply defined as the ratio of concentrations of the preferentially permeating species in the permeate and in the feed. Unlike R, the numerical value of β depends on the concentration units used. R is a more significant term than β from the physicochemical point of view, although the β term is sometimes more convenient to use especially when dealing with very dilute feed solutions. Since there is usually a trade-off between membrane permeability and selectivity, Huang and Yeom (1990) introduced a composite parameter to evaluate the overall performance of a membrane, namely, the pervaporation separation index (PSI), which was expressed as the product of permeation flux and separation factor. Consider that when R ) 1, no separation occurs, but the corresponding PSI may still be very large, depending on the flux, as is the case of a highly porous membrane. Therefore, PSI is redefined as J multiplied by R - 1; a PSI of zero means either zero flux or zero separation (Huang and Feng, 1993a). Membrane stability is the ability of a membrane to maintain both the permeability and selectivity under specific system conditions for an extended period of time. Membrane stability is affected by the chemical, mechanical, and thermal properties of the membrane. When considering polymeric membranes for the separation of anhydrous organic mixtures, the membrane stability is of prime importance. Mass Transport in Membranes A proper understanding of the membrane separation mechanism may provide direct information on the research and development of an appropriate membrane. Because of the complicated penetrants-membrane interactions, it is difficult to formulate a single explanation for the complex transport process. There are principally two approaches to describing mass transport in pervaporation: (i) the solution-diffusion model and (ii) the pore flow model. The solution-diffusion model is accepted by the majority of membrane researchers (Kataoka et al., 1991a,b; Wijmans and Baker, 1995). According to this mechanism, pervaporation consists of three consecutive steps: (i) sorption of the permeant from the feed liquid to the membrane, (ii) diffusion of the permeant in the membrane, and (iii) desorption of the permeant to the vapor phase on the downstream side of the membrane (see Figure 2). In general, solubility and diffusivity are concentration dependent. A number of mathematical equations for mass transport have been formulated on the basis of Fick’s diffusion equation using different empirical expressions of concentration dependence of solubility and/or diffusivity. However, these equations cannot be taken for granted unless they are used within the experimentally established range for which the relationships expressed for diffusion and thermodynamic equilibria are applicable. The transport of a single component through a nonporous homogeneous membrane has been relatively well described. The concentration dependence of diffusivity is often expressed by exponential or linear forms (Fels and Huang, 1971; Greenlaw et al., 1977; Brun et al., 1985a,b). Assuming thermodynamic equilibria exist at both membrane interfaces, the steady-state flux equation can be readily derived on the basis of Fick’s


Figure 2. Schematic representation of the pervaporation transport mechanism. (a) Solution-diffusion model, (b) pore flow model.

equation for one-dimensional diffusion normal to the membrane surface. For binary mixtures, the mass transport is complicated by the permeant-permeant and permeantmembrane interactions, and no overall explaining theory exists. Lee (1975) used concentration-independent diffusion coefficients and constant solubilities for both permeating species. As a result, it is derived that the permeation flux is a linear function of permeate pressure and the separation factor is essentially equal to the ratio of pure-component permeabilities. This treatment is exactly the same as that used in the permeation of noble gases through rubbery polymer membranes. Being the oversimplified version of the solution-diffusion model, it can hardly be applied to the pervaporation of liquid mixtures. Using the same approach as in single-component pervaporation, the solution-diffusion model has been modified by introducing different empirical parameters, most of which arise from the concentration dependence of diffusivities. Assuming that the diffusivities of individual permeants are proportional to the total concentration of permeants in the membrane, Greenlaw et al. (1977) presented a simple model for the pervaporation of hexane/heptane mixtures that behave almost ideally. However, this model does not apply to nonideal cases such as the pervaporation of alcohol/water mixtures. The concentration dependence of diffusivity is due at least in part to the plasticizing action of the permeants on the polymer, while different components may have different plasticizing effects. Hence, it is generally not appropriate to assume the contribution of permeants to their diffusivities to be linearly additive.

Further, as commonly observed, diffusivities are very sensitive to permeant concentration, especially when the membrane has a strong affinity to the permeating species. A simple linear relationship is often inadequate to describe the concentration dependence of diffusivity. On the same basis as the model developed by Greenlaw et al. (1977) but with the hypothesis that diffusivities are exponentially dependent on permeant concentration, Brun et al. (1985b) proposed a “six-coefficient exponential model”. The permeation flux is expressed by two complicated differential equations. All the model parameters have to be determined by fitting the flux equations to experimental data. Because of the large number of empirical parameters involved in the transport equations, many different sets of parameters can be fit to a given set of pervaporation data (Sferrazza et al., 1988). Consequently, the significance of a particular parameter becomes suspect, and the model is of little predictive or interpretive value. The uncertainty in the determination of the model parameters would be reduced by conducting sorption experiments in addition to pervaporation experiments so that the parameters related to membrane interface equilibrium could be determined independently. But this is only true for the case where liquid sorption in the membrane obeys Henry’s law because a constant solubility was assumed in the model derivation. Therefore, the use of this model is restricted to such pervaporation systems where ideal sorption is expected but variation of diffusivity with concentration may nevertheless occur. Mulder and Smolders (1984, 1985a,b, 1986) presented a more complex model which took into account coupling effects on both sorption and diffusion aspects. The permeation flux is described by two coupled nonlinear differential equations which include permeant-permeant and permeant-polymer interaction parameters. The permeant-permeant interaction parameter is a function of the permeant concentration and can be calculated from the excess free energy of mixing data. However, the parameter so obtained may be totally different, and sometimes, as the authors’ data showed, even an opposite trend in the concentration dependence can result, depending on the equations used in the calculation. Thus, the significance of the interaction parameter is suspect. Further, both permeant-permeant and permeant-polymer interaction parameters are evaluated from the liquid free energy of mixing data and the swelling data of the individual liquids in the membrane, respectively. The interaction between the permeating components in the presence of membrane and the interaction between a permeant and the membrane in the presence of the other permeant are not fully addressed. Moreover, this model supposes the knowledge of diffusivity as a function of permeant concentration, which is in fact one of the most difficult problems yet to solve in this mechanistic approach. Consequently, the functional relationships between operational variables and the membrane performance are difficult to find. Therefore, the practical use of the model is limited. Fels and Huang (1971) pursued an alternative approach. It is based on an extension of the free-volume theory for diffusion of organic substances in polymers and takes into account the effect of one component of a liquid mixture on the diffusion of the other component. Discrepancy between theory and experiment still exists primarily due to the inaccuracies of free-volume parameters and the assumption of ideal sorption as well as

1052 Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997 Table 1. Comparison of the Pervaporation and Evapomeation Performance for the Separation of Water from Isopropyl Alcohol Using a Chitosan Membrane at 30 °C isopropyl alcohol, wt % in feed 10 30 50 70 90 95 a

pervaporation flux, kg/(m2‚h) separation factor 3.62 2.25 1.33 0.75 0.20 0.12

1.4 7.0 23.4 87.4 491 1096

evapomeation flux, kg/(m2‚h) separation factor 0.053 0.048 0.043 0.037 0.019 0.014

11.6 27.8 88.3 236 980 1816

separation factor of liq evaporationa 0.06 0.12 0.26 0.51 1.09 1.30

Calculated from vapor-liquid equilibrium data.

the neglecting of permeant-permeant interactions. Recently, the Fels and Huang model was modified by introducing the interaction parameters on the basis of the Flory-Huggins thermodynamics (Huang and Rhim, 1992; Rhim and Huang, 1989, 1992). A further improvement was made by Yeom and Huang (1992) to account for the effect of flux coupling. Using the experimentally obtainable parameters, the modified model enables prediction and interpretation of pervaporation performance for a given separation system. However, the same concerns about the interaction parameters as discussed for Mulder and Smolders’ model are also pertinent here. Moreover, in this model, the concentrations of permeating species at the permeate side of the membrane were assumed to be zero. Consequently, this model can hardly be used to quantify the effect of permeate pressure on permeation rate. Nevertheless, the free-volume approach has shown to be very useful for practical purposes. Taking into account free volume and other factors, Doong et al. (1995) proposed a comprehensive model for multicomponent permeation. With further improvement, the free-volume approach would get a better insight into the transport in pervaporation processes. Blume et al. (1990) presented a different approach. To facilitate mathematical treatment, they considered pervaporation as a combination of a liquid evaporation step and a vapor permeation step. The normal solution-diffusion model for gas permeation was simply applied to describe vapor permeation. It follows that the overall separation factor is given as the product of the separation factor of liquid evaporation and the separation factor of membrane vapor permeation. If this is true, then (i) the separation favor for evapomeation (a term which refers to the permeation of vapor evaporated from feed liquid; the membrane is not in direct contact with the feed solution) should be the same as for pervaporation and (ii) a membrane should exhibit the same selectivity to vapor permeation and liquid pervaporation for a mixture at azeotropic composition. From the experimental data of Uragami et al. (1988), Uragami and Takigawa (1990), Uragami and Saito (1989), Uragami and Morikawa (1992), Suematsu et al. (1989), and Kataoka et al. (1991a,b) as well as ours (shown in Table 1), it can be seen that this is not the case no matter whether the membranes under study are water selective or organic selective. Pervaporation is not a simple combination of liquid evaporation step and a vapor permeation step. Membrane swelling by a liquid is more significant than by a vapor of the same species, and thus, it is expected that the membrane will exhibit different diffusivity to a penetrant in the liquid and vapor states. Besides the difference in vapor and liquid solubilities, an opposite sorption selectivity has been observed. A sorption study for the system isopropyl alcohol/water/cellulose acetate showed that isopropyl alcohol is preferentially sorbed from the liquid phase,

while water is preferentially sorbed from the vapor phase (Deng et al., 1990). Although Blume et al.’s approach allows an intuitive illustration on how the separation factor changes due to the membrane relative to simple evaporation, it is difficult to make a reliable prediction of the pervaporation separation behavior. All the models mentioned above are based on the solution-diffusion mechanism. In essence, pervaporation is treated as a diffusion-controlled process, and the fluids on both sides of the membrane are considered to be in equilibrium with their respective membrane interfaces. It is not explained where the phase change occurs. Recently, Matsuura and co-workers have proposed a transport model applicable to pervaporation on the basis of the pore flow mechanism (Okada and Matsuura, 1991, 1992; Okada et al., 1991). It is assumed that there are a bundle of straight cylindrical pores on the membrane surface. The mass transport by the pore flow mechanism also consists of three steps: (i) liquid transport from the pore inlet to a liquid-vapor phase boundary, (ii) evaporation at the phase boundary, and (iii) vapor transport from the boundary to the pore outlet (see Figure 2). The distinguishing feature of the pore flow model is that it assumes a liquid-vapor phase boundary inside the membrane, and pervaporation is considered to be a combination of liquid transport and vapor transport in series. There are different views regarding the existence of pores in pervaporation membranes. The solutiondiffusion model considers the pores as passageways allowing communication between the upstream and the downstream membrane face by Knudsen flow or viscous flow mechanism. An effective solution-diffusion membrane has no pores but relies on the thermally agitated motion of chain segments comprising the polymer matrix to generate penetrant-scale transient gaps in the matrix, thereby allowing diffusion from the upstream to the downstream side of the membrane. The permeation is the net result of the random jumps of the penetrant in the membrane. On the other hand, the pores in the pore flow model are defined as the space between nonbound material entities in the polymer matrix through which mass transfer takes place. The equivalent size of such a pore is expressed by any distance (however small) greater than zero (Sourirajan and Matsuura, 1985). The pores in the membranes effective for pervaporation and gas separation are angstrom sized, and these pores are thus not directly observable, making it difficult to physically see whether pervaporation membranes have pores or not. However, it is clear that the pore concept in the pore flow model is not exactly the same as usually perceived from the standpoint of the solution-diffusion model. At present, it would be fair to recognize that the two models represent two different approaches to the description of pervaporation transport. Both models predict cor-


rectly that membranes with pores large enough for Knudsen or viscous flow to occur have no or little selectivity. The membrane pores in the pore flow model and the aforementioned transient gaps in the solutiondiffusion model, both on the molecular scale, describe essentially the same thing, that is, the two-dimensional space for the movement of penetrant molecules. In consideration that (i) the physical structure of the membrane is accounted for explicitly in the pore flow model, (ii) the position of the phase change of permeant in the membrane is clearly addressed in the pore flow model, (iii) several main features of pervaporation observed experimentally can be explained semiquantitatively by the pore flow model, and (iv) good models that enable description and prediction of pervaporation transport are still lacking, the pore flow model should be appreciated, although the present quantitative expression of the pore flow model is at a preliminary stage because some macroscopic concepts such as viscosity and friction constant are used in the derivation of the model equations (Okada and Matsuura, 1991), while fluid continuity does not necessarily hold when the pores are very small. It is important to investigate pervaporation from a study of diffusivities and sorption relations in order to elucidate the transport mechanism. However, the present knowledge on diffusivities and sorption properties is very limited, both experimentally and theoretically. Although numerous models have been proposed, many of them are solely based on qualitative observations, and vigorous verification by experimental data is lacking. Each model works for some systems, but none is general enough to make good descriptions and predictions for most systems. It is important to note that ideal sorption has frequently been assumed in formulating the transport equations and the nonideal behavior in pervaporation has been attributed to the diffusivity aspect. It is thus irrelevant to incorporate a lot of parameters in an attempt to characterize nonideal diffusion without considering how nonideal sorption (e.g., preferential sorption) influences selective permeation. On the other hand, some researchers believe that preferential sorption is the prerequisite to preferential permeation, and the effect of diffusion aspect is overlooked. Comparison of preferential sorption with permeation results indicates that, while preferential sorption is favorable to preferential permeation, the separation can occur either through selective sorption or through diffusivity difference (Huang and Feng, 1992; Okuno et al., 1992, 1993; Zhang and Drioli, 1995). Boundary Layer Effect: Concentration Polarization Concentration polarization is a common phenomenon in membrane processes since membrane separation is based on the difference in the permeation rates of different permeating components. Due to retention of the slow permeating component on the membrane surface, the concentration of the fast permeating component on the membrane surface is lower than that in the bulk phase, while the opposite is true for the slow component. This is shown schematically in Figure 3. Concentration polarization generally leads to a lower productivity and a lesser extent of separation. However, the boundary layer effect is often assumed to be insignificant for most of the current pervaporation membranes because the permeation fluxes are usually low, and little attention has been paid to the significance

Figure 3. Schematic diagram of the boundary layer effect: concentration polarization.

of concentration polarization until recently (Psaume et al., 1988; Gref et al., 1992; Nijhuis et al., 1991; Raghunath and Hwang, 1992a,b). Currently, the resistance-in-series model is mainly used to describe concentration polarization (Nijhuis et al., 1991; Raghunath and Hwang, 1992a,b; Gref et al., 1992). According to this model, the overall masstransfer resistance (1/kt) is the sum of the membrane resistance (1/km) and the liquid boundary layer resistance (1/kl):

1 1 1 ) + kt kl km

(2)

where kt, km, and kl are the overall mass-transfer coefficient and the mass-transfer coefficients in the membrane and the boundary layer, respectively. The nature of the driving force for mass transport in both the boundary layer and the membrane has been assumed to be identical. By assuming constant diffusion and sorption coefficients in the membrane, the boundary layer effect can be investigated experimentally in two different ways (Raghunath and Hwang, 1992a,b): one way is to study the variation of flux with membrane thickness at constant hydrodynamic conditions, and the other is to analyze the influence of hydrodynamic conditions on the flux for a given membrane. In the latter case, if the Reynolds number influences flux significantly, then the boundary layer mass-transfer resistance is important. The changing membrane thickness method allows determination of the boundary layer resistance and the intrinsic membrane permeability, but strictly speaking, it is restricted to homogeneous membranes since the membrane resistance had been assumed to be proportional to the membrane thickness. However, this approach is not without reservation. Only the molecular diffusion in the boundary layer is considered, and the convective transport is overlooked. The resistance-in-series model applies only to the more permeable component whose concentration gradient in the boundary layer is in the same direction as mass transport. Because of the concentration polarization, the concentration of the less permeable component on the membrane surface is higher than in the bulk feed, which means the boundary layer has a “negative” resistance to the mass transfer of this component. Further, only when the liquid sorption in the membrane


Figure 4. Plot showing the concentration polarization index Xs/ Xf as a function of v/k and β (Feng and Huang, 1994).

obeys Henry’s law does the linear additive relationship of the mass-transfer resistances hold. As mentioned earlier, such an ideal case of sorption is rarely encountered in pervaporation systems. The resistance-inseries model approach emphasizes the relative significance of the mass-transfer resistance of the membrane and the boundary layer to the fast permeating component, and the effect of membrane selectivity is not fully taken into account. It is obvious that concentration polarization results from the selective permeation of different components through a membrane. For a nonselective membrane, regardless of the system hydrodynamics, there will be no polarization in concentration. The current version of the resistance-in-series model itself cannot explain well the origin of concentration polarization. Nevertheless, the resistance-in-series model has found applicability in pervaporation separation of dilute organic compounds in aqueous solutions where the concentration polarization has little effect on the permeation rate of water, the slow permeating component, since water is present in abundance and its concentration change due to boundary layer effect is negligible. An alternative approach was pursued by Feng and Huang (1994) on the basis of the classical “film theory”. It is shown that the significance of concentration polarization is determined not only by the membrane permeability and hydrodynamic conditions but also by the membrane selectivity, as expressed by the following equation:

Xs/Xf ) 1/[β - (β - 1) exp(-v/k)]

(3)

where Xs/Xf is the ratio of the mole fraction of the fast permeating species on the membrane surface and in the bulk feed. β is the “intrinsic” enrichment factor of the membrane based on the concentration that the membrane “sees”; β is related to the observed “apparent” enrichment factor βa by βa ) β(Xs/Xf). v and k are the overall molar average velocity of mass transport and the mass-transfer coefficient in the boundary layer, respectively. The ratio v/k is a measure of mass-transfer resistance of the boundary layer relative that of the membrane. Obviously, for finite values of k, Xs/Xf < 1; when k is infinity or when β is unit, Xs/Xf ) 1. The ratio Xs/Xf serves as an index to measure the severity of concentration polarization. The usefulness of eq 3 can be illustrated using the calculation results of concentration polarization index as a function of β and v/k, as shown in Figure 4. It shows that concentration polarization is intensified with an increase in membrane permeability and selectivity and a decrease in the mass-transfer coefficient in the

boundary layer. When v/k is sufficiently small, Xs/Xf approaches unity, an indication of negligible concentration polarization. This is in agreement with the conclusion of the resistance-in-series model. The magnitude of the mass-transfer coefficient in the boundary layer (k) is mainly determined by the chemical nature of the permeating species and the hydrodynamic conditions. Although the mass-transport velocity (v) is not very high for most current membranes, the membrane selectivity (β) can be extremely large, especially for the removal of trace compounds from bulk solutions. Therefore, the concentration polarization in pervaporation should not always be overlooked even for membranes with moderate permeability. In practice, the significance of concentration polarization can be diagnosed using eq 3. Firstly, assume a numerical value of the concentration polarization index. The quantities v and β are obtainable from pervaporation experimental data. k can be estimated from appropriate correlations for the masstransfer coefficient. Then calculate the concentration polarization index Xs/Xf from eq 3. Compare the calculated with the assumed value, and repeat the above procedure until an agreement is reached. Based on the Xs/Xf value so obtained, one can judge the severity of the boundary layer effect. Because little control over the diffusion coefficient in the boundary layer can be achieved, fluid management techniques aimed at promoting mixing of the feed solution near the membrane surface so as to reduce the boundary layer thickness is the practical approach of increasing k in order to minimize the concentration polarization. There has been a notion that concentration polarization is more significant when the target component that preferentially permeates through a membrane is present in the feed at lower concentrations (Huang, 1991). According to eq 3, the concentration polarization index is dependent only on the parameters β and v/k. The observed effect of the feed concentration on concentration polarization is due to the concentration dependence of membrane permselectivity and the boundary layer mass-transfer coefficient. Concentration polarization is not necessarily intensified when the concentration decreases. However, because the relationship between the maximum possible enrichment factor β and the concentration is that of an inverse proportionality, the boundary layer effect is not likely to pose a severe problem for high feed concentrations. Mass transfer is accompanied with heat transfer in pervaporation. The heat needed for a phase change of permeant from the liquid to the vapor phase normally comes from the feed solution. This leads to a temperature gradient in the direction of permeate flow. The heat transfer could be described, in analogy to mass transfer, by a resistance-in-series model. Gooding (1986) showed that the heat-transfer resistance is mainly dominated by the boundary layer. The temperatur drop occurs primarily between the bulk feed and the feed-membrane interface, not across the membrane. The magnitude of the temperature drop depends naturally on the permeation flux, the latent heat of the permeate, and the heat-transfer coefficient in the boundary layer. Karlsson and Tragardh (1995) estimated that for water permeation through a silicone membrane, the difference between the temperature at the feedmembrane interface and that of the bulk feed is less than 3 °C when the Reynolds number that characterizes the hydrodynamic condition of feed flow is above 100. Pervaporation performance is influenced by tempera-


ture, and thus, promoting feed mixing near the membrane surface seems to be an efficient way of reducing the temperature drop. Concentration polarization and the temperature drop are due to the boundary layer effect; both aspects should be taken into consideration in the design of pervaporators. Polymers for Pervaporation Membranes Polymeric materials are overwhelmingly utilized for pervaporation membranes. Three types of polymer membranes can be classified: glassy polymer membranes, rubbery polymer or elastomeric membranes, and ionic polymer membranes. By tendency, glassy polymers are suitable for making water-selective membranes used for solvent dehydration, and rubbery polymer membranes are favorable to the selective removal of organic compounds from water. However, it is interesting to note the exception that some polyacetylene derivatives, which are glassy polymers, are preferentially permeable to organic compounds over water (Masuda et al., 1990). Poly(1(trimethylsilyl)-1-propyne), or PTMSP, has been shown to be more alcohol selective than poly(dimethylsiloxane) (PDMS), a representative of organic-selective membranes. For the separation of organic mixtures, it is not yet very clear whether glassy or rubbery polymers are more appropriate, and both types of polymers show some pervaporation selectivity. Ionic polymers contain ionic groups that are neutralized by counterions. They may be viewed as crosslinked polyelectrolytes. Ionic membranes, which can be subdivided into cationic and anionic, are normally water selective due to their affinity to water. Ionic polysaccharides have been shown to be potential materials for making dehydration membranes. A membrane with low hydrophilicity generally exhibits a low water flux in dehydration, but some membranes made of polymers with very high hydrophilicity such as poly(vinyl alcohol) (PVA) and poly(acrylic acid) (PAA) need cross-linking for improved stability and selectivity. The polymer materials for dehydration membranes should maintain a proper balance of hydrophilicity and hydrophobicity. The techniques of controlling the hydrophilicity-hydrophobicity balance have been discussed (Huang, 1991). Improvement in membrane performance can often be achieved by polymer modification, for which several techniques have been developed, including cross-linking, grafting, blending, copolymerization, and incorporation of adsorbent materials. Interpenetrating polymer networks (IPN’s) are a unique type of polymer blend. IPN membranes of hydrophilic/hydrophobic, cationic/anionic, and glassy/ rubbery constituents offer another way of improving membrane performance (Lee et al., 1991d). Such membranes can be prepared either by simultaneous polymerization of both networks or by sequential polymerization in which a polymer is swollen in a monomer followed by in situ polymerization of the monomer. Many polymers that can be formed into membranes have been investigated in terms of pervaporation properties, and the list of polymer membrane materials is virtually endless insofar as possible chemical varieties are concerned. Feng (1995) conducted an extensive survey of the pervaporation membranes and separation systems studied in the journal literature during the past half decade; the research work reported in the pre-1990 literature has been documented by Neel (1991). It is shown that currently silicone rubber based polymers

(primarily poly(dimethylsiloxane)) are mainly used for the selective permeation of organic compounds from aqueous solutions, and PTMSP and other siliconecontaining polyacetylene derivatives are under development as potential membrane materials (Nagase et al., 1990, 1991a,b; Kang et al., 1994). Generally, silicone rubber membranes exhibit limited selectivity for some mixtures such as lower alcohols-water and acetic acidwater (Netke et al., 1995). To improve permselectivity, it has been attempted to fill the membrane with organophilic adsorbent (Bartels-Caspers et al., 1992; Vankelcom et al., 1995). As such, the sorption capacity and/or sorption selectivity will be enhanced due to the adsorbent fillers. However, strong adsorption will cause immobilization of the permeating species, leading to a reduction in permeation flux. Thus, a suitable adsorbent filler should have proper organophilicity, hydrophobicity, and pore size characteristics. Poly(1-(trimethylsilyl)-1-propyne) is an extraordinary glassy polymer. It has ∼25% voids which may be linked through chainto-chain gaps at least ∼3 Å wide (Srinivasan et al., 1994). Gas permeabilities through the polymer are orders of magnitude larger than in other glassy polymers due to its loose microstructure and high mobility of the pendant groups (Odani and Uyeda, 1991). However, during pervaporation operations, both permeability and selectivity decline with operation time (Hino et al., 1991; Camera-Roda et al., 1991, 1992). The problem associated with membrane stability sustained applicability of this unique material for producing industrial pervaporation membranes. For making dehydration membranes, PVA- and PAAbased polymers are the most widely used materials, while chitosan and aromatic polyimide materials are attracting great interest. PVA is a 1,3-diglycol polymer whose hydroxyl groups have strong interactions with water through hydrogen bonding. It is one of the very few high-molecular-weight water-soluble resins and can easily be cross-linked either chemically or thermally. Of all the membranes for pervaporation separation of aqueous organic mixtures, PVA-based membranes have been studied most intensively. Most of the research is centered on modification of PVA for improved permselectivity and stability and performance testing for various perspective applications (Huang, 1991). The commercial membrane of GFT Co. for solvent dehydration is made from chemically cross-linked PVA. PAA is another polymer suitable for preparing waterselective membranes. It has a high charge density based on the carboxyl groups, which are readily available for cross-linking and salt formation with alkaline metals. Multivalent cations such as Al3+, Cr3+, Ca2+, and Mg2+ can be used to induce ionic cross-linking for improved resistance of PAA to dissociation in aqueous solutions (Zhao and Huang, 1990). Alternatively, PAA can undergo ionization. It has been observed that by converting the acidic form of PAA to an alkali-metal salt form, both permeability and selectivity can be enhanced (Maeda et al., 1991a; Karakane et al., 1991). However, the membrane suffers from the problem of long-term stability due to elution of the alkaline-metal ions out of the membrane, which leads to regeneration of polyacrylate to its acid form. Maintaining an appropriate pH of feed solution would immobilize the alkaline-metal ions in the membrane, but this method has limited applicability from an application point of view. Another way to stabilize the polyacrylate membrane is to use certain polycations, instead of alkali metals, to form a


stable polyion complex (Karakane et al., 1991; Tsuyumoto et al., 1991). This method has been shown to be quite effective. BP International of the U.K. and Daicel Chemical of Japan are developing PAA-based membranes for solvent dehydration. Chitosan is the N-deacetylated product of chitin, the second most abundant natural polymer next to cellulose. It is a linear polymer comprised primarily of glucosamine. The reactive amino groups and the primary and secondary hydroxyl groups can be used for chemical modifications. Chitosan is considered as a versatile material for various applications partly due to the large varieity of useful forms that either are commercially available or can be readily obtained in the laboratory. Chitosan membranes have high water permselectivity and solvent stability. Chitosan in its free amine form is insoluble in water at neutral pH’s. However, in some acidic solutions (e.g., acetic acid), the free amino groups (NH2) become protonated to form water-soluble chitosan-acid salts. Treatment of the latter with an alkaline solution converts the cationic amine groups (NH3+) into the NH2 form, thereby accomplishing regeneration of chitosan to the free amine form (Feng and Huang, 1996). Therefore, chitosan membranes can be prepared by casting an aqueous solution of an appropriate chitosanacid salt, followed by neutralization with an alkaline solution (Uragami and Shinomiya, 1992; Goto et al., 1994; Feng and Huang, 1996). Chemically modified chitosan membranes have been studied extensively by Lee and co-workers for separation of ethanol-water and acetic acid-water mixtures (Lee, 1993; Lee and Shin, 1991; Lee et al., 1991a-c, 1992). Aromatic polyimides are well-known for their excellent thermal stability, mechanical strength, and chemical resistance. They are conventionally made by polycondensation of aromatic dianhydrides and diamines to form a soluble poly(amic acid) which is then converted to the polyimides by intramolecular condensation. Many imide polymers are extremely resistant to solvent dissolution, and hence, these polymers are of particular interest for preparing membranes used for organicorganic separations where membrane stability is often a primary consideration due to the relatively harsh conditions. However, the strong chemical resistance sometimes imposes restrictions to selection of suitable solvents in membrane preparation by the conventional solution casting method. In this case, the membrane can be prepared by casting the poly(amic acid) solution before imidization. Recently, a chemical vapor deposition and polymerization (CVDP) technique was developed for making polyimide membranes (Yanagishita et al., 1993a,b). The aromatic dianhydrides and diamines are effused as molecular beams and deposited simultaneously on the surface of a membrane substrate, and the deposited layers of the monomers are then converted to polyimide by heating. Since the membrane is formed by in situ polymerization in a “dry” environment, the CVDP technique is especially useful for preparing composite polyimide membranes that are solventresistant. Ube Industries of Japan is developing various polyimide membranes for dehydration of organic solvents and separation of organic mixtures. Selection of Polymer Membrane Materials According to the solution-diffusion model, membrane permeability is determined by diffusivity and solubility, and thus, membrane selectivity is determined by sorption selectivity and diffusion selectivity. While smaller

permeating molecules normally exhibit larger diffusivity, the solubility is often influenced by the chemical affinity of the permeating species to the membrane material. Consider the separation of aqueous-organic mixtures where the water molecules are smaller than organic compounds. A hydrophilic membrane favors both solubility and diffusivity for selective permeation of water, while an organophilic membrane must have a large solubility to the organic compound in order to permeate the organic compound preferentially because of the unfavorable diffusion selectivity. This may be the reason that many polymers are selective to water permeation, while only a few are selective to the permeation of organic compounds. Currently, there is no well-established criterion for the selection of membrane materials, and the materials for pervaporation membranes are normally selected empirically. Polymers with high selectivity are often preferred for further study because the disadvantage associated with low permeability can be partly compensated by introducing asymmetricity to the membrane structure, thereby reducing the effective thickness of the membrane. Several possible approaches for membrane material selection have been attempted, as discussed below. Solubility Parameter Approach. Based on the perception that preferential sorption is the prerequisite to the preferential permeation (Mulder et al., 1985; Wenzlaff et al., 1985; Mulder and Smolders, 1985a, 1986, 1991), the solubility aspect of pervaporation has been emphasized in the selection of membrane materials. The solubility of a component in the membrane is determined primarily by the chemical nature of the membrane material and the permeating molecules and can be described qualitatively using the solubility parameter. Considering a permeant A and a membrane M, the greater their mutual solubility, the smaller the difference in their solubility parameters, ∆AM. Lloyd and Meluch (1985) used the ratio ∆AM/∆BM as a measure of preferential sorption for the permeants A and B in the membrane M and as an index for membrane material selection. It is suggested that if it is desired to permeate component B and reject component A, the membrane material should be selected to maximize the ratio ∆AM/∆BM. However, there are some restrictions in using the solubility parameter approach based on the following considerations: (i) There is no guarantee at all that preferential sorption leads to preferential permeation, and the difference in diffusivity also affects (and in some cases determines) the permeation selectivity. According to the solution-diffusion model, the permeation selectivity (R) is given as the product of the sorption selectivity (RS) and diffusion selectivity (RD). All three situations have been observed experimentally: (i) RS > 1 and RD > 1, (ii) RS > 1 and RD < 1, and (iii) RS < 1 and RD > 1 (Masuda et al., 1990; Okuno et al., 1993, 1995). Obviously, in the latter two cases where the membrane exhibits opposite selectivity to sorption and diffusion, the preferential permeation is determined by which one of the sorption and diffusion selectivities is dominating. Hence, the solubility parameter approach may be misleading for the situations where selective diffusion dominates the separation. (ii) Even in some cases where preferential sorption determines the separation, the solubility parameter approach sometimes does not work at all, as reported by Lee et al. (1989), who studied the separation of ethanol and chloroform from aqueous


solutions using different membranes. (iii) There are some restrictions in using the solubility parameter. The solubility parameter predicts the mixing of solvent and polymer from the properties of pure substances, which means the interactions involved in a ternary system, namely, membrane M-permeant A-permeant B, are not fully taken into account. (iv) The numerical value of the solubility parameter of a polymer is usually obtained by applying group contributions using the numerical values assigned to the various structural groups in the polymer repeating unit. The numerical value of the solubility parameter for a given polymer is subject to some indefinite variations depending on the source of the data involved. In sorption studies of ethanol-water in some copolymer membranes, the predicted preferential sorption on the basis of the solubility parameter approach is shown to be the opposite of the experimental results (Wei, 1993). In spite of these limitations, the solubility parameter approach is convenient to use as a first estimate in the selection of polymer membrane materials. Surface Thermodynamics Approach. Surface thermodynamics, which was originally proposed by van Oss et al. (1983), was also tried to be used in membrane material selection by Lee et al. (1989). This approach requires the use of two interfacial free-energy parameters, ∆F123 and ∆F132, which represent the surface free energy between water and membrane in the presence of organic compound and that between water and organic compound in the presence of the membrane, respectively. These parameters are calculated using the interfacial tension values that are available in the literature. According to the surface thermodynamics, a negative value of ∆F123 implies preferential sorption of the organic component into the polymer; a large value of ∆F132 suggests good separation of the organic component from water in the polymer. However, a critical examination of the data of Lee et al. (1989) reveals that this is not always the case. For example, all the ∆F123 values are negative for the ethanol-water-polymer systems where some of the polymers are selective to ethanol sorption while others are selective to water sorption; the ∆F123 values are positive for the chloroform-water-polymer systems where the polymers exhibit preferential sorption of chloroform. On the other hand, the systems having large ∆F132 values do not necessarily exhibit large separation. Since the surface thermodynamics approach emphasizes interfacial properties and ignores the kinetic factor (diffusion), it is unable to predict which component permeates through the polymer preferentially. Moreover, this approach suffers from similar limitations as the solubility parameter approach does. Liquid Chromatography Approach. The forces that lead to the physicochemical interactions involved in pervaporation systems are present in other situations where the permeating components and the membrane material are in contact with one another, for example, liquid chromatography (LC). The LC column can be packed with the powder of the polymer membrane material. Using solvent A as the carrier, the retention time (or volume) for solvent B in the column can be obtained by injecting a small amount of solvent B into the carrier stream. The retention data are a measure of the interaction between solvent B and the membrane material in the presence of solvent A on a relative scale. Similarly, the retention data for solvent A in the LC column can be determined by using solvent B as the

carrier. For a given separation of A from B, the membrane material should be selected so that the polymer exhibits large differences in the retention data of the two components. Sourirajan and Matsuura (1985) have used this approach to characterize polymer materials for reverse osmosis and ultrafiltration membranes, but little work has been conducted so far with respect to pervaporation membranes (Balint et al., 1993). Contact Angle Approach. Farnand and Noh (1989) used the contact angle method in a preliminary screening of membrane materials for the separation of methanol from C4 hydrocarbons. It was intended to measure the contact angle of methanol with the membrane surface in order to give an approximate determination if the membrane could be used to reject or attract methanol. This approach is similar to the surface thermodynamics approach in nature. However, it seems that the experimental results do not show an obvious relation between the contact angle and the pervaporation performance. Polarity Parameter Approach. On the basis of the comparison of the membrane polarity in terms of Dimroth’s solvent polarity parameter value (Et at 25 °C), Shimidzu and Yoshikawa (1991) observed that if the membrane polarity is close to the water polarity (63.1 kcal/mol), the membrane tends to be water selective and the separation factor tends to decrease as the polarity parameter of the membrane deviates from that of water. As no knowledge of precise polymer structure is required, the Et value is one of the promising parameters for use in the development of membrane materials, in particular for newly synthesized membranes. This approach has been applicable to the so-called “fixed carrier membranes” used for water-ethanol separations. However, its validity has yet to be tested for other membrane systems. Asymmetric and Composite Membranes In the development of high-flux membranes, much effort has been made to introduce asymmetricity into the membrane structure in order to reduce the effective thickness of the membrane. These membranes generally have a thin dense skin layer supported on a microporous substrate, and thus, the permeation flux is substantially enhanced. All industrially important membranes are structurally asymmetric. An asymmetric membrane can be formed in an integral form or in a composite form. The difference between the two types of asymmetric membranes mainly lies in whether the skin and the substrate are made from the same polymer material. A composite membrane consists of a skin layer and a substrate that are made separately and from two different materials. In practice, integrally skinned asymmetric membranes are often simply called asymmetric membranes. Asymmetric Membranes. Asymmetric membranes are typically prepared from a single polymer solution via the phase inversion process. In this process, a homogeneous polymer solution is transformed into a two-phase system in which a polymer-rich solid phase forms the rigid membrane structure, while a polymerpoor liquid phase forms the voids. To prepare this kind of membrane, the polymer should be dissolved in a suitable solvent. The homogeneous solution is formed into a film in such a way that the desired shape (e.g., flat, tubular, or hollow fiber) is obtained. The polymer solution is treated in a specific way to precipitate the polymer, followed by a suitable drying process.


In the precipitation step, several techniques have been developed to achieve phase inversion (Wijmans and Smolders, 1986). They include (i) precipitation from the vapor phase, (ii) precipitation by controlled evaporation, (iii) precipitation by immersion in a nonsolvent, (iv) thermal precipitation, and (v) a combination thereof. The immersion precipitation technique, first developed in 1960 by Loeb and Sourirajan for the preparation of reverse osmosis membranes, is the most common technique to prepare asymmetric membranes for various applications. The variables involved in the procedure for membrane preparation and their effects on the structure of the resulting membranes are discussed by Kesting (1985). The feature of the technique is that the cast polymer film is immersed into a nonsolvent bath. Then the polymer precipitates as a result of solvent exchange with the nonsolvent. A dense skin forms on the top surface of the cast film due to rapid polymer precipitation. The skin layer slows down the entry of nonsolvent into the underlying polymer solution, and consequently, the polymer precipitates much more slowly, forming a porous substructure. This process, involving solvent-nonsolvent exchange, is also termed a wet-phase inversion process. It is generally considered that the separation properties of an asymmetric membrane are primarily determined by the top skin layer. Therefore, most developments in asymmetric membranes are focused on the minimization of the skin layer thickness. However, the reduction in the skin layer thickness is accompanied by an increase in the possibility of skin layer defects (large pores) that lead to a decrease in selectivity. Many asymmetric membranes made by the traditional phase inversion process are shown to be not very selective for pervaporation. It is frequently assumed that the porous substrate acts only as a mechanical support, and there is a notion that such asymmetric membranes are unsuitable for pervaporation (Neel, 1991; Heinzelmann, 1991). However, it should be noted that in pervaporation, the permeate side is maintained at vacuum and the flow of permeate vapor through the pores of the substrate is likely to the follow the Knudsen flow mechanism. The resistance of the substrate cannot always be neglected. Experimental work shows that through proper control and adjustment of the membrane preparation conditions, integrally skinned asymmetric membranes with high flux and satisfactory selectivity for pervaporation separation can be produced by the phase inversion technique from various polymers, including poly(parabanic acid) (Maeda et al., 1991b), nylon 4 (Lai et al., 1994), polyimide (Tanihara et al., 1992; Yanagishita et al., 1994, 1995), poly(ether imide) (Huang and Feng, 1993a,b), polysulfone (Li et al., 1994; Koops et al., 1994), poly(sulfone amide) (Tsujii et al., 1992), polycarbonate (Lee et al., 1994), and poly(vinylidene fluoride) (Jian and Pintauro, 1993). The selectivity achievable in an asymmetric membrane can be analyzed using the resistance model (Huang and Feng, 1993b). The mass flow can be considered analogous to an electric flow in a circuit consisting of three resistance components in a seriesparallel configuration (Figure 5). A permeating species must first pass through the skin layer and then through the pores and the polymer matrix in the substrate. The overall permeation resistance Rt is given by

Rt ) R1 + R2R3/(R2 + R3)

(4)

where R1 is the resistance of the skin layer and R2 and

Figure 5. Resistance configuration of mass transport through an asymmetric membrane.

Figure 6. Calculated ratio of R/Rint as a function of relative resistances, showing the selectivity achievable in an asymmetric membrane (Huang and Feng, 1993b).

R3 are the resistances of the pores and the polymer matrix, respectively, in the porous substrate. Using the relative resistance of the skin layer and the substrate, Huang and Feng (1993b) demonstrated that the selectivity achievable in an asymmetric pervaporation membranes is determined not only by the resistance of the skin layer and substrate but also by the relative resistance of the polymer matrix and the pores in the substrate. This is illustrated in Figure 6, where the ratio of the separation factor R in an asymmetric membrane to the intrinsic separation factor Rint is plotted as a function of relative resistances. Clearly, if the resistances offered by the polymer matrix and the pores in the substrate are not well balanced, even a membrane with a defect-free skin layer may still exhibit low selectivity. Thus, the low selectivity observed for some asymmetric membrane is not always due to a defective skin layer, and the notion that phase inversion asymmetric membranes are unsuitable for pervaporation is not a general statement. The resistance model approach, though not a mechanismic approach to membrane separation, has practical utility to membrane development because it correlates membrane selectivity and the resistance components lying in the path of mass transport through the membrane and offers a means of determining these resistances qualitatively, at least in relative terms. Composite Membranes. Two major steps are involved in the preparation of composite membranes: casting of the microporous support first, followed by deposition of the selective dense layer (barrier) on the surface of the microporous support. One of the advantages of using the composite approach is that different polymers may be used as the barrier layer and the


porous support, which allows a combination of properties that may not be available in a single material. Several methods have been developed to prepare composite membranes (Heinzelmann, 1991), including (i) casting of the barrier layer and membrane support separately followed by lamination, (ii) direct coating of a polymer solution onto a support film followed by an appropriate posttreatment, and (iii) in situ formation of the barrier layer on a microporous support film. Currently, direct coating of a polymer solution onto a microporous support is widely used for producing composite membranes. Additional variations of this method include the use of reactive monomers in place of the polymer and incorporation of cross-linking agents or additives in the polymer solution. It is believed that the commercial pervaporation membranes used by GFT (Deutsche Carbone) and Membrane Technology & Research (MTR) Co. are manufactured by this technique. The porous membrane support is usually prepared using the aforementioned phase inversion technique. It has to be emphasized that the support layer of the composite membrane should be highly porous so that its resistance to mass transport is small. However, the pores must be sufficiently small in order to prevent the coating solution from filling into the pores because the pore sealing will decrease the permeation flux of the resulting membrane. Several parameters affect the coating technique for the preparation of composite membranes. Such parameters include (i) the composition, viscosity, and surface tension of the coating solution, (ii) the pretreatment of the support, (iii) the choice of appropriate coating methods (e.g., dip coating, spray coating, spin coating, etc.), and (iv) the drying conditions. As an alternative to direct solution coating, the lamination approach is useful especially when direct coating is inappropriate due to problems such as insufficient wetting of the support film by the coating solution or incompatibility of the support with the solvents. Further, the lamination approach is effective in transport studies since the thickness of the laminating layer is much easier to determine than that of a coating layer. However, the preparation of membrane laminates is a delicate operation because the two layers should be closely contacted, and any space between the two layers may deteriorate the membrane performance. In situ formation of the ultrathin top layer on the surface of a highly porous support by means of interfacial polymerization (Parthasarathy et al., 1994) or plasma polymerization (Hirotsu and Arita, 1991; Masuoka et al., 1992a,b; Matsuyama et al., 1994a,b) is in perspective. For this type of composite membranes, the top layer, which can be formed from a wide range of monomers, is primarily responsible for the separation. Thus, the permeability and selectivity may be solely determined by the dense layer. Moreover, plasma polymerization can be used to polymerize not only the compounds having double bonds but also many other organic compounds. The structure of the plasma polymers depends strongly on the plasma parameters such as the plasma-forming gases (including both monomers and inert gas) and the discharge conditions. The properties of the plasma polymers are quite different from those of polymers by free-radical polymerization. This opens up new possibilities to prepare and modify membranes that surpass the properties of the membranes synthesized by conventional polymerization. Naturally, a successful plasma polymer composite mem-

brane will also rely on an appropriate porous support. Yamaguchi et al. (1991, 1992, 1993, 1994) developed a plasma graft filling polymerization technique to prepare membranes where the micropores of a porous substrate film are filled with a plasma-grafted polymer. In the plasma graft filling polymerized membranes, the porous substrate is inert to feed solution and the grafted polymer is responsible for the permselectivity. Consequently, membrane swelling is restrained by the substrate matrix. These membranes are reported to have high permeability and selectivity for the separation of organic-organic mixtures and removal of dissolved organics from water. Although the plasma chemistry of polymers has been studied extensively during the past 3 decades, only recently has the fabrication of plasma polymer composite membranes become available on a pilot-plant scale due to technical difficulties and reproductivity problems. The resistance model approach has also been applied to analyzing pervaporation performance of composite membranes (Gudernatsch et al., 1991). Since the skin layer and the substrate are formed separately, the substrate can be characterized individually. The substrate is typically prepared from engineering plastics which are normally water selective. It is natural to anticipate that when the skin layer is selective to organic compound permeation, the overall performance of the composite membrane may be either water selective or organic selective, depending on whether the skin or the substrate dominates the separation. This is the case for acetic acid/water separation by PDMS-coated asymmetric poly(ether imide) membranes (Bai et al., 1993). Interestingly, Deng et al. (1994) reported that when a water-selective asymmetric polyamide membrane is laminated with PDMS, the membrane becomes more selective to water permeation. Their explanation to this phenomenon is that placing a PDMS membrane on top of the polyamide membrane prevents swelling of the polyamide membrane by the feed liquid, making the latter membrane contribute more effectively to selectivity. Heintz and Stephan (1994a,b) proposed a generalized solution-diffusion model for pervaporation through composite membranes by taking into account concentration polarization, coupled diffusion through the dense active layer, and the influence of the porous support layer. Hollow Fiber Membrane Configuration Flat membranes housed in plate-and-frame modules and spiral-wound modules are extensively used for pervaporation. Hollow fiber membranes, an alternative to flat membranes, are not well developed despite their wide acceptance in other membrane processes such as reverse osmosis and gas separations. Hollow fiber membrane permeators are constructed similar to the shell-and-tube heat exchanger. The feed solution can be introduced to either the shell side or the bore side, depending on the system requirements. Both modes of operation have been used. Hollow fiber membranes have the following advantageous characteristics: (i) the membrane packing density (i.e., membrane area per unit module volume) can be much higher than that achievable with flat membranes; (ii) in contrast to flat membranes which need mechanical support, hollow fiber membranes are self-supporting; and (iii) the hollow fibers themselves form the vacuum vessel if the shellfed mode of operation is used. However, when the


permeate is withdrawn from the fiber bores, the permeate pressure buildup inside the hollow fiber is normally more significant than is the case with flat membranes because of the small size of the channel for permeate flow. The pressure buildup reduces the driving force for pervaporation. The effects of permeate pressure buildup can be reduced by using larger fiber diameters, but the high membrane area packing density will be sacrificed. Thus, optimization of the fiber dimensions (length and diameter) is important to module performance. The importance of permeate pressure buildup in shellside-fed hollow fiber pervaporation membranes is well recognized, but research work on quantifying the significance of the pressure buildup is scanty. In most experimental investigations on pervaporation with hollow fiber membranes reported to date, short fiber lengths (generally 104 (Colman and Naylor, 1991). In the shell-fed mode of operation, the turbulence in the feed flow can be more easily achieved by using the well-established techniques in tube-and-shell heat exchangers to manipulate the hydrodynamic conditions. Obviously, of the two modes of operation with shell feed and tube feed, one is not necessarily superior to the other. Various engineering aspects have to be considered simultaneously to optimize the pervaporation system. Activation Energy for Pervaporation Transport The phase change of the permeating species is one of the most distinguishing features of pervaporation. However, except for the obvious fact that sufficient energy needs to be supplied to prevent a temperature drop during pervaporation, the effect of enthalpy change accompanied by the phase change on the permeation has not been elaborated. The experimental data of the temperature dependence of the permeation flux generally exhibits an Arrhenius-type relation

J ) J0 exp(-EJ/RT)

(10)

where EJ has been considered to be the activation energy for permeation in the literature. Strictly speaking, this is not correct, as will be discussed below. Based on the solution-diffusion model, the flux equation can be written phenomenologically, in analogy to gas permeation, as (Wijmans and Baker, 1995)


Ji )

Pi (p - pil) l i0

(11)

where p0 and pl are the partial vapor pressure of the permeant in liquid feed and vapor permeate, respectively, and l is the membrane thickness. P is the permeability coefficient of the membrane with respect to the driving force expressed in terms of partial vapor pressure and is related to the solubility coefficient (S) and diffusivity coefficient (D):

P ) DS

(12)

where D and S are normally dependent on temperature and the temperature dependence can be expressed as

D ) D0 exp(-ED/RT)

(13)

S ) S0 exp(-∆HS/RT)

(14)

Thus, the following relation results

P ) P0 exp(-EP/RT)

(15)

where EP ()ED + ∆HS) is the activation energy of permeation, which is a combination of the activation energy of diffusion (ED) and the enthalpy of dissolution (∆HS) of the permeant in the membrane; P0 is a preexponential factor equal to D0 multiplied by S0. In practice, membrane permeance (P/l) is more useful, especially for asymmetric and composite membranes. Rearranging eqs 11 and 15 yields

P/l ) J/∆p ) (P0/l) exp(-Ep/RT)

(16)

where ∆p is the transmembrane partial pressure difference. Since temperature influences both the membrane permeability and the driving force for mass transport, the activation energy (EP) that characterizes the temperature effect on the permeability coefficient of a membrane should be evaluated from the slope of the ln(J/∆p) vs 1/T plot instead of ln J vs 1/T; EJ obtained from the ln J vs 1/T plot is a compounded parameter characterizing the overall temperature dependence of permeation flux. This is a simple and almost obvious observation, but it seems to have escaped the attention of researchers dealing with this phenomenon. Since the permeate pressure in pervaporation is generally low, ∆p is largely determined by the saturated vapor pressure (ps). Based on the classical ClausiusClapeyron equation, the temperature dependence of ps can also be approximated by an exponential equation

ps ) A exp(-∆HV/RT)

(17)

where ∆HV is the molar heat of vaporization. Consequently, a simple yet useful rule of thumb for estimating EP is to subtract the heat of vaporization (∆HV) from the EJ value, namely,

EP ) EJ - ∆HV

(18)

because evaluating EJ from the ln J vs 1/T data is much simpler than evaluating EP from the ln(J/∆p) vs 1/T data. This explicitly shows how the enthalpy change due to the phase change in pervaporation influences the permeation behavior. Note that eq 18 applies when the permeate pressure is sufficiently low compared to the vapor pressure over the feed liquid. Otherwise, eq 16 has to be used to evaluate EP.

It needs to be mentioned that the numerical values of EJ have shown to be in the range of 4-92 kJ/mol (Huang, 1991), which overlaps with the ∆HV range for water and many organic substances of interest for pervaporation separation (Reid and Sherwood, 1966). It is thus suggested that the negative value of EP may result. This has been found to be true for ethanol/water permeation through polyion and chitosan membranes (Karakane et al., 1991; Zhang and Drioli, 1995). While ED is generally positive, ∆HS is usually negative for the exothermic sorption process. When the negative ∆HS dominates over the positive ED, a negative value of EP occurs, indicating that the membrane permeability coefficient decreases with increasing temperature despite the fact that the permeation flux normally increases with an increase in temperature due to the contribution of increased driving force. The transmembrane driving force is set by the feed liquid mole fraction, temperature, and permeate vacuum. Besides the partial pressure difference, the masstransport driving force can be expressed in other equivalent terms (Wijmans and Baker, 1995). In such cases, the EP value could be equally well obtained as long as the flux is normalized for the driving force in plotting the Arrhenius relation. Summary Pervaporation is an emerging membrane separation process. Recent developments in pervaporation membranes and pervaporation processes are reviewed in this paper, and some important questions involved in membrane pervaporation are discussed. In particular, the following issues are emphasized: mass transport in the membrane, membrane material selection, concentration polarization in the boundary layer, pressure buildup in hollow fiber membranes, asymmetric and composite membranes, and the activation energy for permeation. Applications of pervaporation in industrial separations are briefly addressed. It is difficult, if not impossible, to give a complete in-depth review of all aspects of pervaporation, whose technology is still developing very rapidly. This paper attempted to provide an insight into this dynamic field and to highlight some of the key problems yet to be solved or clarified. Literature Cited Bagnell, L.; Cavell, K.; Hodges, A. M.; Mau, A. W. H.; Seen, A. J. Use of catalytically active pervaporation membranes in esterification reactions to simultaneously increase product yield, membrane permselectivity and flux. J. Membrane Sci. 1994, 85, 291. Bai, J.; Fouda, A. E.; Matsuura, T.; Hazlett, J. D. A study on the preparation and performance of polydimethylsiloxane-coated polyetherimide membranes in pervaporation. J. Appl. Polym. Sci. 1993, 48, 999. Balint, T.; Nagy, E.; Kraxner, M. Study of interaction between butyl alcohols and cellulose-acetate polymers with reverse osmosis, high-pressure liquid-chromatography and pervaporation methods. J. Membrane Sci. 1993, 78, 101. Bartels-Caspers, C.; Tusel-Langer, E.; Lichtenthaler, R. N. Sorption isotherms of alcohols in zeolite-filled silicone rubber and in PVA-composite membranes. J. Membrane Sci. 1992, 70, 75. Binning, R. C.; James, F. E. Permeation. A new commercial separation tool. Pet. Eng. 1958, 30, C14. Binning, R. C.; Lee, R. J.; Jennings, J. F.; Martin, E. C. Separation of liquid mixtures by permeation. Ind. Eng. Chem. 1961, 53, 45. Binning, R. C.; Jennings, J. F.; Martin, E. C. Removal of water from organic chemicals. U.S. Patent 3,035,060, 1962.

Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997 1063 Blume, I.; Wijmans, J. G.; Baker, R. W. The separation of dissolved organics from water by pervaporation. J. Membrane Sci. 1990, 49, 253. Boddeker, K. W.; Bengtson, G.; Pingel, H.; Dozel, S. Pervaporation of high boilers using heated membranes. Desalination 1993, 90, 249. Bravo, J. L.; Fair, J. R.; Humphrey, J. L.; Martin, C. L.; Seibert, A. F.; S. Fluid Mixture Separation Technologies for Cost Reduction and Process Improvement; Noyes Data: Park Ridge, NJ, 1986. Brun, J. P.; Larchet, C.; Bulvestre, G.; Auclari, B. Sorption and pervaporation of dilute aqueous solutions of organic compounds through polymer membranes. J. Membrane Sci. 1985a, 25, 55. Brun, J. P.; Larchet, C.; Melet, R.; Bulvestre, G. Modeling of the pervaporation of binary mixtures through moderately swelling, non-reacting membranes. J. Membrane Sci. 1985b, 23, 257. Camera-Roda, G.; Bottino, A.; Capannelli, G.; Costa, G.; Starti, G. C. Pervaporation of organic compounds through poly[1(trimethylsilyl)-1-propyne]. In Proc. 5th Int. Conf. Pervaporation Process in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991; pp 495-507. Camera-Roda, G.; Bottino, A.; Capannelli, G.; Costa, G.; Starti, G. C. The effect of a filler on poly[1-(trimethylsilyl)-1-propyne] membrane performances. In Proc. 6th Int. Conf. Pervaporation Process in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1992; pp 451-467. Chen, W. J.; Martin, C. R. Highly methanol selective membranes for pervaporation separation of methyl t-butyl ether/methanol mixtures. J. Membrane Sci. 1995, 104, 101. Chen, M. S. K.; Markiewicz, G. S.; Venugopal, K. G. Development of membrane pervaporation TRIM process for methanol recovery from CH3OH/MTBE/C4 mixtures. AIChE Symp. Ser. 1989, 85 (No. 272), 82. Colman, D. A.; Naylor, T. The influence of operating variables on flux and module design in a high performance pervaporation system. In Proc. 5th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991; pp 143-161. David, D. O.; Gref, R.; Nguyen, Q. T.; Neel, J. Pervaporationesterification coupling. I. Basic kinetic model. Chem. Eng. Res. Des. 1991a, 69, 341. David, D. O.; Gref, R.; Nguyen, Q. T.; Neel, J. Pervaporationesterification coupling. II. Modeling of the influence of the operating parameters. Chem. Eng. Res. Des. 1991b, 69, 355. David, D. O.; Nguyen, Q. T.; Neel, J. Pervaporation membranes endowed with catalytic properties, based on polymer blends. J. Membrane Sci. 1992, 73, 129. Deng, S.; Shiyao, B.; Sourirajan, S.; Matsuura, T. A study of the pervaporation of isopropyl alcohol/water mixtures by cellulose acetate membranes. J. Colloid Interface Sci. 1990, 136, 283. Deng, S.; Sourirajan, S.; Matsuura, T. A study of polydimethylsiloxane/aromatic polyamide laminated membranes for separation of acetic acid/water mixtures by pervaporation process. Sep. Sci. Technol. 1994, 29, 1209. Dettwiler, B.; Dunn, I. J.; Heinzle, E.; Prenosil, J. E. A simulation model for the continuous production of acetoin and butanediol using bacillus-subtilis with integrated pervaporation separation. Biotechnol. Bioeng. 1993, 41, 791. Doghieri, F.; Nardella, A.; Sarti, G. C.; Valentini, C. Pervaporation of methanol-MTBE mixture through modified poly(phenylene oxide) membranes. J. Membrane Sci. 1994, 91, 283. Doong, S. J.; Ho, W. S.; Mastondrea, R. P. Prediction of flux and selectivity in pervaporation through a membrane. J. Membrane Sci. 1995, 107, 129. Farber, L. Applications of pervaporation. Science 1935, 82, 158. Farnand, B. A.; Noh, S. H. Pervaporation as an attractive process for the separation of methanol from C4 hydrocarbons in the production of MTBE and TAME. AIChE Symp. Ser. 1989, 85 (No. 272), 89. Fels, M.; Huang, R. Y. M. Theoretical interpretation of the effect of mixture composition on separation of liquids in polymers. J. Macromol. Sci., Phys. 1971, B5, 89. Feng, X. Studies on Pervaporation Membranes and Pervaporation Processes. Ph.D. Thesis, University of Waterloo, Waterloo, Ontario, Canada, 1995. Feng, X.; Huang, R. Y. M. Concentration polarization in pervaporation separation processes. J. Membrane Sci. 1994, 92, 201. Feng, X.; Huang, R. Y. M. Permeate pressure build-up in shellsidefed hollow fiber pervaporation membranes. Can. J. Chem. Eng. 1995, 73, 833.

Feng, X.; Huang, R. Y. M. Separation of water from ethylene glycol by a chitosan/polysulfone composite membrane. J. Membrane Sci. 1996, 116, 67. Fleming, H. L.; Slater, C. S. Pervaporation. In Membrane Handbook; Ho, W. S. W., Sirkar, K. K., Eds.; Van Nostrand Reinhold: New York, 1992; pp 103-159. Franken, A. C. M.; Mulder, M. H. V.; Smolders, C. A. Pervaporation process using a thermal gradient as the driving force. J. Membrane Sci. 1990, 53, 127. Friedl, A.; Qureshi, N.; Maddox, I. S. Continuous acetonebutanol-ethanol (ABE) fermentation using immobilized cells of clostridium-acetobutylicum in a packed-bed reactor and integration with product removal by pervaporation. Biotechnol. Bioeng. 1991, 38, 518. Geng, Q.; Park, C. H. Pervaporative butanol fermentation by Clostridium acetobutylicum B18. Biotechnol. Bioeng. 1994, 43, 978. Goncalves, M. D. C.; Windmoller, D.; Erismann, N. D. M.; Galembeck, F. Pressure-driven pervaporation. Sep. Sci. Technol. 1990, 25, 1079. Gooding, C. H. Modeling of some relatively simple concepts on the heat transfer aspects of pervaporation. In Proc. 1st Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1986; pp 121-132. Gooding, C. H.; Teague, K. G. Permeate pressure drop characteristics of hollow fibre and spiral wound pervaporation modules. In Proc. 4th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1989; pp 262-273. Gooding, C. H.; Hickey, P. J.; Crowder, M. L. Mass transfer characteristics of a new pervaporation module for water purification. In Proc. 5th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991; pp 237-249. Goto, M.; Shiosaki, A.; Hirose, T. Separation of water/ethanol vapor mixtures through chitosan and crosslinked chitosan membranes. Sep. Sci. Technol. 1994, 29, 1915. Greenlaw, F. W.; Shelden, R. A.; Thompson, E. V. Dependence of diffusive permeation rates on upstream and downstream pressures. II. Two component permeant. J. Membrane Sci. 1977, 2, 333. Gref, R.; Nguyen, Q. T.; Neel, J. Influence of membrane properties on system performance in pervaporation under concentration polarization regime. Sep. Sci. Technol. 1992, 27, 467. Groot, W. J.; den Reyer, M. C. H.; dela Faille, T. B.; van der Lans, R. G. J. M.; Luyben, K. C. A. M. Integration of pervaporation and continuous butanol fermentation with immobilized cells. 1. Experimental results. Chem. Eng. J. 1991a, 46, B1. Groot, W. J.; den Reyer, M. C. H.; dela Faille, T. B.; van der Lans, R. G. J. M.; Luyben, K. C. A. M. Integration of pervaporation and continuous butanol fermentation with immobilized cells. 2. Mathematical modeling and simulations. Chem. Eng. J. 1991b, 46, B11. Gudernatsch, W.; Menzel, Th.; Strathmann, H. Influence of composite membrane structure on pervaporation. J. Membrane Sci. 1991, 61, 19. Heintz, A.; Stephan, W. A generalized solution diffusion model of the pervaporation process through composite membranes. 1. Prediction of mixture solubilities in the dense active layer using the Uniquac model. J. Membrane Sci. 1994a, 89, 143. Heintz, A.; Stephan, W. A generalized solution diffusion model of the pervaporation process through composite membranes. 2. Concentration polarization, coupled diffusion and the influence of the porous support layer. J. Membrane Sci. 1994b, 89, 153. Heinzelmann, W. Fabrication methods for pervaporation membranes. In Proc. 5th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991; pp 22-30. Heisler, E. G.; Hunter, A. S.; Siciliano, J.; Treadway, R. H. Solute and temperature effects in the pervaporation of aqueous alcoholic solutions. Science 1956, 124, 77. Hickey, P. J.; Gooding, C. H. Modeling spiral wound pervaporation modules for pervaporation removal of volatile organic compounds from water. J. Membrane Sci. 1994, 88, 47. Hines, A. L.; Maddox, R. N. Mass Transfer Fundamentals and Applications; Prentice Hall: Englewood Cliffs, NJ, 1985. Hino, T.; Ohya, H.; Hara, T. Removal of halogenated organics from their aqueous solutions by pervaporation. In Proc. 5th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991; pp 423-436.

1064 Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997 Hirotsu, T.; Arita, A. Plasma graft polymerization of N,N-dimethylaminoethyl methacrylate and water-ethanol separation by pervaporation through the grafted membranes. J. Appl. Polym. Sci. 1991, 42, 3255. Huang, R. Y. M., Ed. Pervaporation Membrane Separation Processes; Elsevier: Amsterdam, 1991. Huang, R. Y. M.; Yeom, C. K. Pervaporation separation of aqueous mixtures using crosslinked poly(vinyl alcohol). 2. Permeation of ethanol-water mixtures. J. Membrane Sci. 1990, 51, 273. Huang, R. Y. M.; Feng, X. Pervaporation of water/ethanol mixtures by aromatic polyetherimide membrane. Sep. Sci. Technol. 1992, 27, 1583. Huang, R. Y. M.; Rhim, J. W. Prediction of pervaporation separation characteristics for methanol-pentane-nylon-6-poly(acrylic acid) blended membrane system. J. Membrane Sci. 1992, 71, 211. Huang, R. Y. M.; Feng, X. Dehydration of isopropanol by pervaporation using aromatic polyetherimide membranes. Sep. Sci. Technol. 1993a, 28, 2035. Huang, R. Y. M.; Feng, X. Resistance model approach to asymmetric polyetherimide membranes for pervaporation of isopropanol/water mixtures. J. Membrane Sci. 1993b, 84, 15. Jian, K.; Pintauro, P. N. Integral asymmetric poly(vinylidene fluoride) (PVDF) pervaporation membranes. J. Membrane Sci. 1993, 85, 301. Kang, Y. S.; Shin, E. M.; Jung, B.; Kim, J-J. Composite membranes of poly(1-trimethylsilyl-1-propyne) and poly(dimethyl siloxane) and their pervaporation properties for ethanol-water mixture. J. Appl. Polym. Sci. 1994, 53, 317. Karakane, H.; Tsuyumoto, M.; Maeda, Y.; Honda, Z. Separation of water-ethanol by pervaporation through polyion complex composite membrane. J. Appl. Polym. Sci. 1991, 42, 3229. Karlsson, H. O. E.; Tragardh, G. Aroma compound recovery with pervaporation. Feed flow effects. J. Membrane Sci. 1993, 81, 163. Karlsson, H. O. E.; Tragardh, G. Aroma compound recovery with pervaporation. The effect of high ethanol concentrations. J. Membrane Sci. 1994, 91, 189. Karlsson, H. O. E.; Tragardh, G. Heat transfer and temperature polarization in pervaporation. In Proc. 7th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1995; pp 171-181. Kataoka, T.; Tsuru, T.; Nakao, S.; Kimura, S. Membrane transport properties of pervaporation and vapor permeation in ethanolwater system using polyacrylonitrile and cellulose acetate membranes. J. Chem. Eng. Jpn. 1991a, 24, 334. Kataoka, T.; Tsuru, T.; Nakao, S.; Kimura, S. Permeation equations developed for prediction of membrane performance in pervaporation, vapor permeation and reverse osmosis based on the solution-diffusion model. J. Chem. Eng. Jpn. 1991b, 24, 326. Kesting, R. E. Synthetic Membranes: A Structural Perspective, 2nd ed.; Wiley: New York, 1985. Kober, P. A.; Pervaporation, perstillation and percrystallization. J. Am. Chem. Soc. 1917, 39, 944. Koops, G. H.; Nolten, J. A. M.; Mulder, M. H. V.; Smolders, C. A. Integrally skinned polysulfone hollow fiber membranes for pervaporation. J. Appl. Polym. Sci. 1994, 54, 385. Lai, J.-Y.; Chu, Y.-H.; Huang, S.-L.; Yin, Y.-L. Separation of wateralcohol mixtures by pervaporation through asymmetric nylon 4 membrane. J. Appl. Polym. Sci. 1994, 53, 999. Lamer, T.; Rohart, M. S.; Voilley, A.; Baussart, H. Influence of sorption and diffusion of aroma compounds in silicone rubber on their extraction by pervaporation. J. Membrane Sci. 1994, 90, 251. Lee, C. H. Theory of reverse osmosis and some other membrane permeation operations. J. Appl. Polym. Sci. 1975, 19, 83. Lee, Y. M. Modified chitosan membranes for pervaporation. Desalination 1993, 90, 277. Lee, E. K.; Kalyani, V. J. Process of treating alcohol beverage by vapor-arbitrated pervaporation. U.S. Patent 5,013,447, May 1991. Lee, Y. M.; Shin, E. M. Pervaporation separation of water-ethanol through modified chitosan membranes. 4. Phosphorylated chitosan membranes. J. Membrane Sci. 1991, 64, 145. Lee, Y. M.; Bourgeois, D.; Belfort, G. Sorption, diffusion and pervaporation of organics in polymer membranes. J. Membrane Sci. 1989, 44, 161.

Lee, Y. M.; Shin, E. M.; Chung, C. N. Pervaporation separation of water-ethanol through modified chitosan membranes. 3. Sulfonated chitosan membranes. Polym. (Korea) 1991a, 15, 497. Lee, Y. M.; Shin, E. M.; Noh, S. T. Pervaporation separation of water-ethanol through modified chitosan membranes. 2. Carboxymethyl, carboxyethyl, cyanoethyl, and amidoxime chitosan membranes. Angew. Makromol. Chem. 1991b, 192, 169. Lee, Y. M.; Shin, E. M.; Yang, K. S. Pervaporation separation of water-ethanol through modified chitosan membranes. 1. Chitosan-acetic acid and -metal ion complex membranes. Polym. (Korea) 1991c, 15, 182. Lee, Y. K.; Sohn, I. S.; Jeon, E. J.; Kim, S. C. IPN membranes for the pervaporation of ethanol-water mixture. Polym. J. 1991d, 23, 427. Lee, Y. M.; Nam, S. Y.; Kim, J. H. Pervaporation of water-ethanol through poly(vinyl alcohol)/chitosan blend membrane. Polym. Bull. 1992, 29, 423. Lee, K. R.; Liu, M. J.; Lai, J. Y. Pervaporation separation of aqueous alcohol solution through asymmemtric polycarbonate membrane. Sep. Sci. Technol. 1994, 29, 119. Li, S. G.; Koops, G. H.; Mulder, M. H. V.; van den Boomgaard, T.; Smolders, C. A. Wet spinning of integrally skinned hollow fiber membranes by a modified dual-bath coagulation method using a triple orifice spinneret. J. Membrane Sci. 1994, 94, 329. Lipski, C.; Cote, P.; Fleming, H. Transverse feed flow for hollow fibers significantly improves mass transfer at low energy consumption. In Proc. 5th Int. Conf. Pervaporation Proc. in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991; pp 134-142. Lloyd, D. R.; Meluch, T. B. Selection and evaluation of Membrane materials for liquid separations. In Materials Science of Synthetic Membranes; Lloyd, D. R., Ed.; American Chemical Society: Washington, DC, 1985; pp 47-79. Loeb, S.; Sourirajan, S. Sea water determination by means of a semipermeable membrane, University of California, Los Angeles, Report No. 60-60, 1960. Maeda, Y.; Tsuyumoto, M.; Karakane, H.; Tsugaya, H. Separation of water/ethanol mixtures by pervaporation through hydrolyzed polyacrylonitrile hollow fiber membranes. Polym. J. 1991a, 23, 501. Maeda, Y.; Tsuyumoto, M.; Karakane, H.; Tsugaya, H. Separation of water-acetic acid mixtures by pervaporation through aromatic polymer membranes. In Proc. 5th Int. Conf. Pervaporation Processes in Chem. Ind.; Bakish, R., Ed.; Bakish Materials Corp.: Englewood, NJ, 1991b; pp 31-44. Masawaki, T.; Ohno, T.; Taya, M.; Tone, S. Separation of butanol from butanol-oleyl alcohol mixture by pervaporation with poly(dimethylsiloxane) hollow-fiber membrane. J. Chem. Eng. Jpn. 1992, 25, 257. Masuda, T.; Takatsuka, M.; Tang, B. Z.; Higashimura, T. Pervaporation of organic liquid-water mixtures through substituted polyacrylene membranes. J. Membrane Sci. 1990, 49, 69. Masuoka, T.; Iwasubo, T.; Mizoguchi, K. Pervaporation membranes for ethanol-water mixture prepared by plasma polymerization of fluorocarbons. 2. Perfluoropropane membranes. J. Membrane Sci. 1992a, 69, 109. Masuoka, T.; Iwasubo, T.; Mizoguchi, K. Pervaporation membranes for ethanol-water mixture prepared by plasma polymerization of fluorocarbons. 3. Perfluoropropane membranes. J. Appl. Polym. Sci. 1992b, 46, 311. Matsuyama, H.; Kariya, A.; Teramoto, M. Characteristics of plasma polymerized membrane from octamethyltrisiloxane and its application to the pervaporation of ethanol-water mixture. J. Membrane Sci. 1994a, 88, 85. Matsuyama, H.; Kariya, A.; Teramoto, M. Effect of siloxane chain lengths of monomers on characteristics of pervaporation membranes prepared by plasma polymerization. J. Appl. Polym. Sci. 1994b, 51, 689. Mattos, I. L.; de Castro, M. D. L. Study of mass transfer efficiency in pervaporation process. Anal. Chim. Acta 1994, 298, 159. Mulder, M. H. V.; Smolders, C. A. On the mechanism of separation of ethanol/water mixtures by pervaporation. I. Calculations of concentration profiles. J. Membrane Sci. 1984, 17, 289. Mulder, M. H. V.; Smolders, C. A. Preferential sorption versus preferential permeability in pervaporation. J. Membrane Sci. 1985a, 22, 155. Mulder, M. H. V.; Smolders, C. A. On the mechanism of separation of ethanol/water mixtures by pervaporation. II. Experimental concentration profiles. J. Membrane Sci. 1985b, 23, 41.

Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997 1065 Mulder, M. H. V.; Smolders, C. A. Pervaporation, solubility aspects of the solution-diffusion model. Sep. Puri. Method. 1986, 15, 1. Mulder, M. H. V.; Smolders, C. A. Mass transport phenomena in pervaporation processes. Sep. Sci. Technol. 1991, 26, 85. Mulder, M. H. V.; Franken, T.; Smolders, C. A. Separation of isomeric xylenes by pervaporation through cellulose ester membrane. J. Membrane Sci. 1985, 11, 349. Nagase, Y.; Ishihara, K.; Matsui, K. Chemical modification of poly(substituted acetylene). II. Pervaporation of ethanol/water mixture through poly(1-trimethylsilyl-1-propyne)/poly(dimethylsiloxane) graft copolymer membrane. J. Polym. Sci., Polym. Phys. Ed. 1990, 28, 377. Nagase, Y.; Sugimoto, K.; Takamura, Y.; Matsui, K. Chemical modification of poly(substituted acetylene). VI. Introduction of fluoroalkyl group into poly(1-trimethylsilyl-1-propyne) and improved ethanol permselectivity at pervaporation. J. Appl. Polym. Sci. 1991a, 43, 1227. Nagase, Y.; Takamura, Y.; Matsui, K. Chemical modification of poly(substituted acetylene). V. Alkylsilylation of poly(1-trimethylsilyl-1-propyne) and improved liquid separating property at pervaporation. J. Appl. Polym. Sci. 1991b, 42, 185. Nakagawa, K.; Matsuo, M. Process for producing ether compounds. U.S. Patent 5,292,963, 1994. Neel, J. Introduction to pervaporation. In Pervaporation Membrane Separation Processes; Huang, R. Y. M., Ed.; Elsevier: Amsterdam, 1991; pp 1-109. Nijhuis, H. H.; Mulder, M. H. V.; Smolders, C. A. Removal of trace organics from aqueous solutions. Effect of membrane thickness. J. Membrane Sci. 1991, 61, 99. Odani, H.; Uyeda, T. Theories of sorption and transport in polymer membrane. Polym. J. 1991, 23, 467. Okada, T.; Matsuura, T. A new transport model for pervaporation. J. Membrane Sci. 1991, 59, 133. Okada, T.; Matsuura, T. Predictability of transport-equations for pervaporation on the basis of pore-flow mechanism. J. Membrane Sci. 1992, 70, 163. Okada, T.; Yoshikawa, M.; Matsuura, T. A study on the pervaporation of ethanol-water mixtures on the basis of pore flow model. J. Membrane Sci. 1991, 59, 151. Okamoto, K.; Semoto, T.; Tanaka, K.; Kita, H. Application of pervaporation to phenol-acetone condensation reaction. Chem. Lett. 1991, 1, 167. Okamoto, K.; Yamamoto, M.; Otoshi, Y.; Semoto, T.; Yano, M.; Tanaka, K.; Kita, H. Pervaporation-aided esterification of oleicacid. J. Chem. Eng. Jpn. 1993, 26, 475. Okuno, H.; Morimoto, K.; Uragami, T. Permselectivities of poly(vinyl chloride) membrane for binary alcohol mixtures in pervaporation. Polym. Bull. 1992, 28, 683. Okuno, H.; Nishimoto, H.; Miyata, T.; Uragami, T. Behavior of permeation and separation for aqueous organic acid solutions through poly(vinyl chloride) and poly(vinyl chloride)-co-(vinyl acetate) membranes. Makromol. Chem. 1993, 194, 927. Okuno, H.; Nishida, T.; Uragami, T. Preferential sorption and permeation of binary liquid mixtures in poly(vinyl chloride) membrane by pervaporation. J. Polym. Sci., Polym. Phys. 1995, 33, 299. Papaefstathiou, I.; de Castro, M. D. L. Approaches for improving the precision and sensitivity of continuous pervaporation processes. Anal. Lett. 1995, 28, 2063. Papaefstathiou, I.; Tena, M. T.; de Castro, M. D. L. On-line pervaporation separation process for potentiometric determination of fluoride in dirty samples. Anal. Chim. Acta 1995, 308, 246. Park, H. C.; Ramaker, N. E.; Mulder, M. H. V.; Smolders, C. A. Separation of MTBE-methanol mixtures by pervaporation. Sep. Sci. Technol. 1995, 30, 419. Parthasarathy, A.; Brumlik, C. J.; Martin, C. R.; Collins, G. E. Interfacial polymerization of thin polymer films onto the surface of a microporous hollow-fiber membrane. J. Membrane Sci. 1994, 94, 249. Psaume, R.; Aptel, P.; Aurelle, Y.; Mora, J. C.; Bersillon, J. L. Pervaporation: importance of concentration polarization in the extraction of trace organics from water. J. Membrane Sci. 1988, 36, 373. Raghunath, B.; Hwang, S.-T. Effect of boundary layer mass transfer resistance in the pervaporation of dilute organics. J. Membrane Sci. 1992a, 65, 147.

Raghuanth, B.; Hwang, S.-T. General treatment of liquid boundary layer resistance in the pervaporation of dilute aqueous organics through tubular membranes. J. Membrane Sci. 1992b, 75, 29. Rautenbach, R.; Albrecht, R. The separation potential of pervaporation. 2. Process design and economics. J. Membrane Sci. 1985, 25, 25. Reid, R. C.; Sherwood, T. K. Properties of Gases and Liquids; McGraw-Hill: New York, 1966. Rhim, J. W.; Huang, R. Y. M. On the prediction of separation factor and permeability in the separation of binary mixtures by pervaporation. J. Membrane Sci. 1989, 46, 335. Rhim, J. W.; Huang, R. Y. M. Prediction of pervaporation separation characteristics for the ethanol-water-nylon-4-membrane system. J. Membrane Sci. 1992, 70, 105. Schaetzel, P.; Favre, E.; Nguyen, Q. T.; Neel, J. Mass transfer analysis of pervaporation through an ion exchange membrane. Desalination 1993, 90, 259. Sferrazza, R. A.; Escobosa, R.; Gooding, C. H. Estimation of parameters in a sorption-diffusion model of pervaporation. J. Membrane Sci. 1988, 35, 125. Shah, V. M.; Bartels, C. R.; Pasternak, M.; Reale, J. Opportunities for membranes in the production of octane enhancers. AIChE Symp. Ser. 1989, 85 (No. 272), 93. Shieh, J. J.; Huang, R. Y. M. A modified solution diffusion model for pervaporation. Presented at the 7th Annual Meeting of North American Membrane Society, Portland, OR, May 2024, 1995. Shimidzu, T.; Yoshikawa, M. Synthesis of novel copolymer membranes for pervaporation. In Pervaporation Membrane Separation Processes; Huang, R. Y. M., Ed.; Elsevier: Amsterdam, 1991; pp 321-361. Sourirajan, S.; Matsuura, T. Reverse Osmosis/Ultrafiltration Process Principles; National Research Council Canada: Ottawa, 1985. Srinivasa, R.; Auvil, S. R.; Burban, P. M. Eludicating the mechansim(s) of gas transport in poly[1-(trimethylsilyl)-1-propyne) (PTMSP) membranes. J. Membrane Sci. 1994, 86, 67. Suematsu, H.; Harada, K.; Takoka, T. Separation of ethanolwater mixtures by pervaporation through cellophane membrane. Membrane (Maku) 1989, 14, 337. Tanihara, N.; Tanaka, K.; Kita, H.; Nakamura, A.; Kusuki, Y.; Nakagawa, K. Vapor permeation separation of water-ethanol mixtures by asymmetric polyimide hollow fiber membrane modules. J. Chem. Eng. Jpn. 1992, 25, 388. Tanimura, S.; Nakao, S.; Kimura, S. Ethanol-selective membrane for reverse osmosis of ethanol/water mixture. AIChE J. 1990, 36, 1118. Timashev, S. F.; Valuev, V. V.; Salem, R. R.; Strugatshaga, A. G. Pervaporation induced by electric current. J. Membrane Sci. 1994, 91, 249. Tsujii, S.; Yamada, H.; Takegami, H. Separation of water-ethanol mixtures through poly(sulfone amide) membranes. J. Polym. Sci. Technol. Jpn. 1992, 49, 13. Tsuyumoto, M.; Karakane, H.; Maeda, Y.; Tsugaya, H. Development of polyion complex hollow fiber membrane for separation of water-ethanol mixtures. Desalination 1991, 80, 139. Uragami, T.; Saito, M. Studies on syntheses and permeabilities of special polymer membranes. 68. Analysis of pervaporation and separation characteristics and new technique for separation of aqueous alcoholic solutions through alginic acid membranes. Sep. Sci. Technol. 1989, 24, 541. Uragami, T.; Takigawa, K. Permeation and separation characteristics of ethanol-water mixtures through chitosan derivative membranes membrane by pervaporation and evapomeation. Polymer 1990, 31, 668. Uragami, T.; Morikawa, T. Permeation and separation characteristics of alcohol-water mixtures through poly(dimethyl siloxane) membranes by pervaporation and evapomeation. J. Appl. Polym. Sci. 1992, 44, 2009. Uragami, T.; Shinomiya, H. Concentration of aqueous dimethyl sufoxide solutions through a chitosan membrane by pervaporation with a temperature difference. J. Membrane Sci. 1992, 74, 183. Uragami, T.; Saito, M.; Takigawa, K. Studies on synthesis and permeabilities of special polymer membranes. 69. Comparison of permeation and separation characteristics for aqueous alcoholic solutions by pervaporation and new evapomeation methods through chitosan membranes. Makromol. Chem. Rapid Commun. 1988, 9, 361.

1066 Ind. Eng. Chem. Res., Vol. 36, No. 4, 1997 van Oss, C. J.; Visser, J.; Absolom, D. R.; Omenyi, S. N.; Neumann, A. W. The concept of negative Hamaker coefficients. II. Thermodynamics, experimental evidence and applications. Adv. Colloid Interface Sci. 1983, 18, 133. Vankelecom, I. F. J.; Depre, D.; De Beukelaer, S.; Uytterhoeven, J. B. Influence of zeolites in PDMS membranes. Pervaporation of water/alcohol mixtures. J. Phys. Chem. 1995, 99, 13193. Vrana, D. L.; Meagher, M. M.; Hutkinsand, R. W.; Duffield, B. Pervaporation of model acetone-butanol-ethanol fermentation product solutions using polytetrafluoroethylene membranes. Sep. Sci. Technol. 1993, 28, 2167. Wei, Y. Pervaporation with Latex Membranes. Ph.D. Thesis, University of Waterloo, Waterloo, Ontario, Canada, 1993. Wenzlaff, A.; Boddeker, K. W.; Hattenbach, K. Pervaporation of water-ethanol through ion exchange membranes. J. Membrane Sci. 1985, 22, 333. Wijmans, J. G.; Smolders, C. A. Preparation of asymmetric membranes by the phase inversion process. In Synthetic Membranes: Science, Engineering and Applications; Bungay, P. M., Lonsdale, H. K., de Pinho, M. N., Eds.; D. Reidal: Amsterdam, 1986; pp 39-56. Wijmans, J. G.; Baker, R. W. The solution-diffusion model: a review. J. Membrane Sci. 1995, 107, 1. Wnuk, R.; Chmiel, H. Direct heating of composite membranes in pervaporation and gas separation processes. J. Membrane Sci. 1992, 68, 293. Yamaguchi, T.; Nakao, S.; Kimura, S. Plamsa-graft filling polymerizationsPreparation of a new type of pervaporation membrane for organic liquid-mixtures. Macromolecules 1991, 24, 5522. Yamaguchi, T.; Nakao, S.; Kimura, S. Solubility and pervaporation properties of the filling-polymerized membrane prepared by plasma-graft Polymerization for pervaporation of organic-liquid mixtures. Ind. Eng. Chem. Res. 1992, 31, 1914. Yamaguchi, T.; Nakao, S.; Kimura, S. Design of pervaporation membrane for organic-liquid separation based on solubility control by plasma-graft filling polymerization technique. Ind. Eng. Chem. Res. 1993, 32, 848. Yamaguchi, T.; Yamahara, S.; Nakao, S.; Kimura, S. Preparation of pervaporation membranes for removal of dissolved organics

from water by plasma-graft filling polymerization. J. Membrane Sci. 1994, 95, 39. Yanagishita, H.; Nakane, T.; Nozoye, H.; Yoshitome, H. Preparation of polyimide composite membrane by chemical vapor deposition and polymerization technique (VCDP). J. Appl. Polym. Sci. 1993a, 49, 565. Yanagishita, H.; Nozoye, H.; Nakane, T. Preparation of 6FDAbased polyimide composite membrane by CVDP process. Desalination 1993b, 90, 55. Yanagishita, H.; Maejima, C.; Kitamoto, D.; Nakane, T. Preparation of asymmetric polyimide membrane for water/ethanol separation in pervaporation by the phase inversion process. J. Membrane Sci. 1994, 86, 231. Yanagishita, H.; Kitamoto, D.; Nakane, T. Separation of alcohol aqueous solution by preparation using asymmetric polyimide membrane. High Perform. Polym. 1995, 7, 275. Yeom, C. K.; Huang, R. Y. M. Modeling of pervaporation separation of ethanol-water mixtures through crosslinked poly(vinyl alcohol) membrane. J. Membrane Sci. 1992, 67, 39. Zhang, S.; Drioli, E. Pervaporation membranes. Sep. Sci. Technol. 1995, 30, 1. Zhang, S. Q.; Fouda, A. E.; Matsuura, T. A study on pervaporation of benzyl alcohol solution by polydimethylsiloxane membrane. J. Membrane Sci. 1992, 70, 193. Zhao, X. P.; Huang, R. Y. M. Pervaporation separation of ethanolwater mixtures using crosslinked blended poly(acrylic acid)nylon 66 membranes. J. Appl. Polym. Sci. 1990, 41, 2133.

Received for review April 1, 1996 Revised manuscript received June 12, 1996 Accepted June 12, 1996X IE960189G

X Abstract published in Advance ACS Abstracts, February 15, 1997.

Liquid Separation by Membrane Pervaporation: A Review - Industrial

Recommend Documents