Improved Recovery of an Ester Flavor Compound by Pervaporation

Food Microbiology and Engineering Laboratory (INRA), Institut National ..... Figure 6 Pervaporation with flash condensation: water fluxes collected in...
1 downloads 0 Views 145KB Size
4458

Ind. Eng. Chem. Res. 1999, 38, 4458-4469

Improved Recovery of an Ester Flavor Compound by Pervaporation Coupled with a Flash Condensation Arnaud Baudot* and Miche` le Marin† Food Microbiology and Engineering Laboratory (INRA), Institut National Agronomique Paris-Grignon, F-78850 Thiverval-Grignon, France

The recovery of an aroma compound, ethyl acetate, highly diluted in model aqueous feeds was carried out with a pervaporation stage coupled to a flash condensation. The permeate, at the vapor state, was selectively fractionated between two condensers in series with decreasing temperatures. A global modeling of the coupling system was proposed, taking into account the extraction step (with transmembrane mass transfer equations) as well as the recovery step (with flash vapor-liquid equilibrium curves). It was shown that the inert gases flowing in the permeate circuit (which were due to unavoidable air leakage) were inducing a dilution of the condensable permeate vapors, leading to a significant lowering of the dew point of the permeate. If compared to pervaporation with conventional total permeate condensation, the coupling of pervaporation with a flash condensation led to a significant enhancement of the selectivity for the recovery of ethyl acetate. Introduction Pervaporation is a separation process based on a selective transport through a dense membrane associated with an evaporation of the permeate on the downstream side of the membrane. This phase change is obtained by lowering the partial pressure of the permeate, either by gas sweeping or, more generally, with a vacuum pump. The permeate, at the vapor state, is then recovered by condensation. Pervaporation with organophilic membranes is a promising alternative process to distillation or solvent extraction for the separation and concentration of diluted organic compounds extracted from aqueous feeds. This process has shown many advantages regarding the field of the bio-industries: it is suited for the extraction of temperature-sensitive organic volatile compounds such as aroma compounds and is especially well-fitted for the coupling with a fermenter. Indeed, this extraction technique displays many advantages for a continuous extraction of valuable volatile organic compounds from a bioreactor: there is no membrane fouling due to microorganisms (because of the nonporous nature of the membrane) and gentle pressure and temperature are applied in the feed liquid. Moreover, an improvement of the bioreactor productivity is generally observed due to the continuous removal of inhibiting volatile organic compounds.1 Recently, pervaporation has been applied to the continuous extraction of aroma compounds produced by biotechnology, which are highly valuable because of their natural-grade type.2-4 Apart from the numerous pervaporation studies that have been carried out on model solutions, several authors have applied pervaporation for the recovery of aroma compounds from real industrial liquid media. For instance, poly(dimethylsiloxane) (PDMS)-based mem* To whom correspondence should be addressed. Present address: Unilever Research Vlaardingen, Olivier van Noortlaan 120, 3133 AT Vlaardingen, The Netherlands. E-mail: [email protected]. † E-mail: [email protected].

branes were used for the extraction of valuable fragrances out of liquid waste effluents issued from the processing of perfume essential oils.5 Pervaporation was also applied for the recovery of a flavor-enriched extract from apple essence.6 Other applications of pervaporation concern the extraction of volatile organic compounds from liquid wastes issued from the food industry. Souchon et al.7 showed that thick PDMS membranes were particularly selective for 3-octanone, a spicy flavor characteristic of meat cooking vapor condensates. Commercial pervaporation membranes were also tested for the recovery of aldehydes, higher alcohols, and esters from apple juice evaporator condensates.8 As a last example, the pervaporation technique was applied to the deodorization of cauliflower blanching wastewaters.9 The polyether-block-amide dense membranes supplied by GKSS (Geesthacht, Germany) proved to be particularly selective toward the sulfur-containing organic volatile fraction that was causing the olfactive disturbance generated by these effluents. The compounds that were extracted preferentially were mainly S-methyl thioesters (ranging from S-methyl thioacetate to Smethyl thiobutanoate) and methyl sulfides (dimethyl monosulfide, dimethyl disulfide, dimethyl trisulfide, and dimethyl tetrasulfide). Surprisingly, the pervaporate was displaying a flavor profile characteristic of certain surface-ripened French cheeses and it was foreseen to valorize the sulfur organic extract as a flavoring cocktail for industrial sauces or food dressings. In a recent review on the recovery of aroma compounds by pervaporation,10 it has been shown that the extraction through organophilic pervaporation membranes is particularly selective for hydrophobic highboiling aroma compounds, such as lactones or aromatic (i.e., benzene derivatives) flavor compounds. However, as a general rule, the pervaporation process is hardly more selective than one stage of distillation with diluted aroma compounds displaying a moderate molecular weight (MW < 100 g‚mol-1). While benefiting from the technical advantages of the extraction by pervaporation (particularly fitted for the extraction of traces of temperature-sensitive organic compounds), it is possible to

10.1021/ie990095h CCC: $18.00 © 1999 American Chemical Society Published on Web 10/07/1999

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999 4459

enhance the selectivity of the recovery of the desired compounds by coupling the pervaporation with another operation unit. Because the permeate on the downstream side of the pervaporation membrane is at the vapor state, connecting the downstream side compartment of the pervaporation module with a flash-condensation system constitutes one of the best options. With this system, the permeate vapors coming out of the membrane module are selectively fractionated between two condensers in series with decreasing temperatures. The less volatile fraction (as a general rule water) is mainly recovered in the first (warmer) condenser while the most volatile fraction (i.e., the aroma compounds) mainly remains at the vapor state when in contact with the first condenser and is then recovered within the second (colder) condenser. The first experiments with a “pervaporation/two-stage condensation” coupling system11 were reported in 1989. The pervaporation technique was used in order to lower the alcohol content of wines. A two-stage condensation system was used on the downstream side of the membrane module in order to recover preferentially the most volatile flavors (esters and higher alcohols) that were also permeating through the membrane. This fraction was meant to be reincorporated in the dealcoholized wine in order to compensate the flavor loss during the pervaporation operation. A double condensation system has also been used in order to fractionate solvents that were extracted through a pervaporation membrane from water wastes produced by the polymer industry.12 The mechanisms that rule the fractionation of the permeate in a two-stage condensation system at the downstream of a pervaporation membrane were first studied with model aqueous feed solutions containing 10% ethanol.13 By this way, it was intended to simulate the extraction of bioethanol from a fermentation broth. A “flash” fractionation mechanism was occurring if the temperature of the first condenser was comprised between the dew point and the bubble point characteristic of the gaseous permeate mixture flowing through this condenser. This means that the permeate was fractionated, on one hand, in a boiling liquid collected in the first condenser and, on the other hand, in a vapor phase that was recovered in the colder second condenser. The composition of the two fractions can be described thanks to the calculation of the isobaric vapor-liquid equilibria curves. The composition of both permeate fractions can be obtained from the intersection of the isotherm line corresponding to the temperature of the first condenser with the dew curve (composition in condenser 1) and with the bubble curve (composition in condenser 2) characteristic of the studied gaseous mixture. As an illustration, some results concerning the extraction of ethanol with a pervaporation-flash condensation system14 are displayed in Figure 1. The composition of the permeate was around 40% w/w ethanol, meaning that the mass-related enrichment factor β of the pervaporation membrane was equal to 4. With a (condensable) permeate vapor pressure equal to 2000 Pa, flash condensation of the permeate could be achieved if the temperature in the first condenser was between 7 and 14 °C. As an example, if the temperature in the first condenser was 10 °C, the fraction collected in the first condenser was depleted in ethanol (around 18% w/w) whereas the fraction in the second condenser was significantly enriched in ethanol (65%). In this case, the coupling system was more than

Figure 1. Pervaporation with flash condensation: permeate distribution as a function of the isobaric vapor-liquid equilibrium curve (GFT PDMS 1060 membrane; feed composition, water + 10% w/w ethanol; Tfeed ) 30 °C; pressure of the condensable vapors, 2000 Pa; temperature of the first condenser, 10 °C; after Beaumelle14).

50% more selective for ethanol than the pervaporation membrane alone. In this paper, we present results concerning the performances of a “pervaporation-two-stage condensation” coupling system applied to the recovery of a model aroma compound, ethyl acetate, highly diluted in water (in order to simulate a model food or biological feed). Whereas the previous published works concerning “pervaporation-two-stage condensation” coupling systems11-13 were mainly descriptive, a quantitative modeling of this technique is being proposed for the first time. It combines the modeling of the extraction step (transmembrane mass transfer) with the recovery step (ruled by the vapor-liquid equilibrium in the first condenser). The role of the inert gases, only previously studied in the total condensation configuration15 (i.e., when the temperature in the first condenser was lower than the bubble point), has been established when carrying out a flash condensation under vacuum. Experimental Section Pilot Plant. The experimental study was carried out on a pilot plant which is presented in Figure 2. The membrane effective area was 0.1 m2. The plate and frame module, supplied by Le Carbone GFT, was modified in order to neglect the boundary layer effect in the liquid feed.15 As a high flow rate was used in the liquid feed, the variation of the feed temperature was also neglected. At the downstream face of the membrane, the vacuum was maintained with a Alcatel 2021A vacuum pump placed after the condensers. The value of the total downstream pressure was measured by a pressure gauge (MKS 220 CA, range 0-10 000 Pa) located in the bell jar containing the membrane module. The total downstream pressure was regulated with a dry-air inlet which was located between the condensers and the vacuum pump. It should be pointed out that the air injected at this point was entirely stripped away by the vacuum pump and did not flow backward to the permeate circuit. Thus the dry air injected at the upstream of the vacuum pump did not flow into the permeate stream and did not participate to the dilution

4460

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999

Figure 2. Schematic representation of the pervaporation pilot plant.

of the permeate vapors. The permeate circuit was designed in the laboratory and consisted of two parallel condensation systems: the first one with a liquid nitrogen total condenser and the other with two variable-temperature stages of condensation in series. The vapors were circulated from the membrane module to the condensers through stainless steel pipe of an internal diameter of 4.5 cm. In the liquid nitrogen condensation circuit, the condenser was located 1.3 m far from the module. In the two-stage condensation circuit, the distance between the first condenser and the module was 1.5 m and the distance between the first and the second condenser was 0.9 m. When the twostage condensation circuit was used, the first condenser was operated between ambient temperature and -25 °C while the second one was operated at a lower temperature (generally ca. -70 °C) to avoid permeate loss through the vacuum pump. The heat transfer in the condensers was considered as nonlimiting due to sufficient exchange area. Membranes. PDMS 1070 composite membranes, supplied by Le Carbone-GFT, were used. They are composed of a 30 µm thick silicalite-grafted poly(dimethylsiloxane) film supported by a polyacrylonitrile layer coated on a polyester nonwoven. The same set of membranes was used during all the experiments. It is well-known that the conditioning of the pervaporation membrane has a strong influence on the membrane performances.16 That is why the experimental fluxes were obtained after 2 or 3 days of pervaporation in order to stabilize the sorption of the membrane in contact with the feed mixture. The chemical stability of the membranes was checked between each experiment, measuring pure water flux at reference operating conditions. Model Solutions. Ethyl acetate is widely encountered in liquid food or biological media, such as fruit

juices6,8,17 or alcoholic beverages.18,19 This molecule was chosen as a model aroma compound primarily because its thermodynamic properties were well-known.20 Ethyl acetate (Prolabo, 99.5% purity) was dissolved in doubleosmosed water to prepare the model solutions. The feed tank was filled with ca. 10 L of mixture. The retentate was recirculated to the feed tank during the experiments, and the time for each run was small enough to keep the concentration in the feed constant and equal to 500 ppm (w/w). Analytical Methods. Quantitative analysis of the concentration in ethyl acetate in the aqueous samples was obtained by direct injection on a 3-m-long Chromosorb 101 packed column fitted on a Erba Science Mega HRGC 5300 chromatograph. Before injection, the samples were diluted at 50% v/v with an internal standard aqueous solution containing 1000 ppm w/w butanol. The injector temperature was set at 175 °C, the oven temperature was constant at 175 °C, and the FID detector temperature was fixed at 180 °C. Air was flowing through the column at 200 mL‚min-1. Thermodynamic Properties. The thermodynamic properties of the studied compounds were needed for the following reasons: to appreciate the driving force that was inducing the transport of the permeants through the membrane; to calculate the mass-related enrichment factor βVLE(T) of the vapor-liquid equilibrium characteristic of the feed. βVLE(Tfeed) ) MEAxEA(feed)γEA(feed)pEA°(Tfeed)

/

MEAxEA(feed)γEA(feed)pEA°(Tfeed) + Mwxw(feed)γw(feed)pw°(Tfeed) wEA(feed) (1)

x, γ, and M correspond respectively to the molar fraction

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999 4461 Table 1. Water-Ethyl Acetate Binary Mixture: Infinite Dilution Activity Coefficient and Antoine Law Constants of Each Component component

M (g‚mol-1)

γ∞

A

B

C

T range (°C)

ref

water ethyl acetate

18 70

8.65 70

10.1962 9.67438

1730.63 1473.995

233.46 238.629

0 to 100 -43 to 77

20 20

Figure 3. Pervaporation with total condensation: fluxes of water (A) and ethyl acetate (B) as a function of the total permeate pressure and the feed temperature (GFT PDMS 1070 membrane; feed composition, water + 500 ppm w/w ethyl acetate).

(mol‚mol-1), the activity coefficient (dimensionless), and the molar mass (g‚mol-1) of each compound in the feed. The subscripts “w” and “EA” refer to water and ethyl acetate. wEA(feed) is the mass fraction of ethyl acetate in the feed (g‚g-1). Tfeed is the temperature in the feed. By comparing the enrichment factors of the vapor-liquid equilibrium of the feed (βVLE) and the pervaporation operation (βPV ) wEA(permeate)/wEA(feed)), it was possible to appreciate if the presence of a pervaporation membrane at the interface between the liquid feed and the vaporous permeate was effectively leading to a separation more selective than a stage of distillation. The thermodynamic properties of the studied compounds were also needed to calculate the dew and bubble curves characteristic of the VLE in the first condenser. Those data were needed to predict the fractionation of the permeate in the two-stage condensation system. The thermodynamic constants of the compounds in the feed and the permeate are listed in Table 1. The saturated vapor pressures of the pure compounds p°(T) (Pa) were calculated with the Antoine equation, which constants A, B, and C are given in Table 1 (T being expressed in °C):

p°(T) ) 10A-[B/(C+T)]

(2)

It should be pointed out that the values of the Antoine constants for water are normally valid only for temperatures higher than 0 °C. However, we made several calculations with temperatures slightly lower than this limit. Nevertheless, for temperatures between 0 and -10 °C, the relative difference between the values of water-saturated vapor pressures extrapolated from the Antoine law and the measured values21 was always lower than 12%. In this study, we also assumed that the values of the infinite dilution activity coefficients of each compound were independent of the temperature. We considered that the temperature dependence of the vapor-liquid equilibria was mainly related to the variation of the

saturated vapor pressures rather than to the variation of the activity coefficients. We based this assumption on the results obtained by Carelli et al.,22 who showed that a temperature variation between 25 and 50 °C had only a moderate effect on the value of the infinite dilution activity coefficient of ethyl acetate in model aqueous solutions. Moreover, the value of the infinite dilution activity coefficient of ethyl acetate in water estimated with the UNIFAC method23-25 proved to be very slightly dependent on the temperature (7% variation in the -10/15 °C range). Results and Discussion Modeling of the Transmembrane Mass Transport. These experiments were carried out at two feed temperatures (30 and 50 °C), and the permeate pressures were varied between 250 and 2500 Pa. The permeate was recovered by total condensation with liquid nitrogen. Each experiment was doubled. The standard errors on the water fluxes and the ester fluxes are respectively 3% and 5%. In Figure 3, it can be seen that, whatever the feed temperature, an increase of the permeate total pressure led to a decrease of the fluxes. This was due to the increase of the partial pressures of the permeants at the downstream side of the membrane and thus, to a decrease of the driving force (eqs 3 and 4). Indeed, as the mass transport was isothermal (as a phase change occurred) and induced by a chemical potential gradient, it was possible to express the transmembrane species fluxes as a function of the partial pressure difference of each permeant between both sides of the membrane. In this case, the expression of the driving force inducing the mass transport through the membrane is continuous whatever the considered phase (diluted in the feed, dissolved in the membrane polymer, or in the vapor state in the permeate). It was shown on the present installation that the sum of the partial pressures of the permeants at the downstream side of the membrane was always equal to the total permeate pressure measured in the bell jar containing the mem-

4462

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999

Table 2. Pervaporation of Water + 500 ppm w/w Ethyl Acetate: Transmembrane Mass Transfer Coefficient of Each Permeant and Comparison between the Enrichment Factor of the Pervaporation Membrane (βPV) and the Enrichment Factor of the Vapor-Liquid Equilibrium (βVLE) Characteristic of the Feed (GFT PDMS 1070 Membrane) Tfeed (°C)

kw × 107 (mol‚m-2‚Pa-1‚s-1)

kEA × 107 (mol‚m-2‚Pa-1‚s-1)

βPV

βVLE

30 50

1.8 ((0.3) 1.3 ((0.05)

3 ((0.5) 1.8 ((0.05)

230 280

180 225

brane module.26 According to Raoult’s law, the mass transport can thus be represented as the following:

Jw ) kw(pw(feed) - ywpT) ) kw(xw(feed)γw(feed)pw°(Tfeed) - ywpT) (3) JEA ) kEA(pEA(feed) - yEApT) ) kEA(xEA(feed)γEA(feed)pEA°(Tfeed) - yEApT) (4) with

yEA ) 1 - yw )

JEA JEA + Jw

(5)

J, k, and y are respectively the transmembrane flux (mol‚m-2‚s-1), the transmembrane mass transfer coefficient (mol‚m-2‚Pa-1‚s-1) and the molar fraction (mol‚mol-1) at the downstream side of the membrane of each permeant. pT is the total permeate pressure (Pa) measured in the bell jar containing the membrane module. pw(feed) and pEA(feed) are the hypothetical partial pressures of water and ethyl acetate in equilibrium with the respective species in the feed. For both studied feed temperatures, the transmembrane transfer coefficients were calculated from eqs 3-5. These coefficients were independent of the total permeate in the studied range of permeate pressures. Their mean values are displayed in Table 2. To provide an homogeneous representation of the transport properties of the membrane materials toward each permeant, the transmembrane coefficients have been expressed in mol‚m-2‚Pa-1‚s-1. From Table 2, it becomes obvious that the feed temperature had a significant influence on the intrinsic transport properties of the membrane. Indeed, whereas the feed temperature increase led to a raising of the transmembrane partial fluxes, it surprisingly induced a lowering of the transmembrane transfer coefficients. This could only be attributed to a lowering of the permeant sorption inside the membrane at higher temperature. Thus, the doubling of the fluxes of water and ester, when increasing the feed temperature from 30 to 50 °C, was mainly attributed to the increase of the partial pressure of the considered permeants at the upstream side of the membrane. The modeling of the species fluxes was satisfactory, whatever the operating conditions in the feed or in the permeate (Figure 3). The selectivity of the pervaporation membrane, expressed in terms of mass-related enrichment factor, βPV, was constant all over the permeate pressure range.

βPV )

JEAMEA/(JEAMEA + JwMw) wEA(feed)

(6)

It appeared that the pervaporation membrane PDMS 1070, provided by GFT, was only slightly more selective

than the vapor-liquid equilibrium characteristic of the feed mixtures at 30 and 50 °C (Table 2). The use of a two-stage condensing device was thus justified in order to improve significantly the global selectivity of the process. Further comments and more in-depth studies about the mass transport properties or the selectivity of silicalite-filled pervaporation membranes in contact with volatile organic compounds (which were not the primary goal of the present study) can be found in other previous works.27-30 Characterization of the Inert Gas Flow in the Vacuum Circuit. Due to the dismountable nature of the installation on which the experiments were realized, air leakages into the permeate circuit could hardly be avoided. It was established that leakages were taking place mainly at the level of the bell jar containing the membrane module. As will be shown later in this paper, the resultant inert gases (oppositely to the condensable permeate vapors) had a great influence on the recovery of the permeate by condensation. This very important technical issue has almost never been discussed in the literature. Another example of the hindering of the permeate recovery in pervaporation induced by the presence of inert gases in the permeate circuit has been reported very recently by Scha¨fer et al.31 Indeed it has been established by these authors that, when coupling a pervaporation stage with a fermenter, a significant part of the permeate could be stripped out of the condenser by the carbon dioxide flowing in the permeate circuit (which was initially diluted in the fermentation must and which permeated through the membrane). Consequently, we decided to characterize carefully the quantity of inert gases that were flowing from the bell jar through the permeate circuit to the vacuum pump. To do so, during each of the previously described “pervaporation with total condensation” experiments, the total molar flow n˘ T (mol‚s-1) entering the bell jar was calculated. The value of n˘ T was obtained thanks to the total pressure increase (∆pT) in the bell jar after isolating it from the rest of the permeate circuit during a short time (∆t) (generally 20 s in order to avoid an important total permeate pressure build-up in the bell jar, so that the transmembrane permeant flux was constant). Indeed, the ideal gas law led to the following equation:

n˘ T )

∆n ∆pT V ) ∆t ∆t RT

(7)

with V being the volume of the bell jar containing the membrane module (m3), T, the temperature in the permeate side (K) and ∆n/∆t, the molar flow accumulating inside the module (mol‚s-1). The inert gas flow, characteristic of the equipment design, was obtained by deducing the condensable flow ∑in˘ i (obtained from the experiments described in the previous section) from the total molar flow n˘ T calculated with eq 7:

n˘ inert ) n˘ T -

∑i n˘ i ) n˘ T - A∑i Ji

(8)

The inert gas flow n˘ inertwas found to be independent of the feed temperature (30 or 50 °C) or the total permeate pressure pT (in the 250-2500 Pa range). The value of n˘ inert was 3 × 10-5 mol‚s-1 on the studied configuration.

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999 4463

Figure 4. Pervaporation with flash condensation: composition of the fraction collected in the second condenser as a function of the temperature in the first condenser (GFT PDMS 1070 membrane; feed composition, water + 500 ppm w/w ethyl acetate; pT ) 2000 Pa; (A) Tfeed ) 30 °C/(B) Tfeed ) 50 °C).

Two-Stage Condensation: Distribution of the Permeate. The coupling experiments were carried out at a constant total pressure pT equal to 2000 Pa. The feed temperatures were equal to either 30 or 50 °C, and the first condenser temperature was varied over an appropriate range of temperature (namely between -10 and 10 °C). The isobaric dew and bubble curves of a water-ethyl acetate at 2000 Pa were obtained from a program designed in the laboratory. This software permits the estimation of isothermal or isobaric vaporliquid, as well as liquid-liquid, equilibria data of more than 80 binary water/aroma compound solutions. The model used for the description of the evolution of the activity coefficient of both components as a function of the composition of the medium was the three-suffix Margules equation.32 The variation of the composition of the fraction collected in the second condenser as a function of the operating conditions is represented in Figure 4. Condensation in the first condenser (and consequently permeate fractionation) occurred for first temperatures lower than -2 °C (Tfeed ) 30 °C) or 6 °C (Tfeed ) 50 °C). Above these limit temperatures, the permeate was completely recovered in the second condenser which, in this case, acted as a single total condensation step. At first glance, it was obvious that the distribution of the permeate did not fit with the isobaric vapor-liquid equilibrium curves corresponding to a total condensable vapor pressure of 2000 Pa (dashed dew curves in Figure 4). Indeed, the condensation in the first condenser was achieved at a temperature by far lower than what could be expected with the dew curve corresponding to a pressure of 2000 Pa. This discrepancy could be explained by the dilution of the condensable vapors by the inert gases flowing in the permeate circuit. Thus, the sum of the partial pressures of the condensable vapors circulating between the membrane module and the first condenser was not equal to the total permeate pressure measured into the bell jar but, following Raoult’s law, was limited to

pw + pEA )

n˘ w + n˘ EA p n˘ w + n˘ EA + n˘ inert T

(9)

Taking only into account the estimated condensable vapor partial pressures (eq 9), it appeared that the

Figure 5. Schematic flow sheet of pervaporation-flash condensation coupling system (see eqs 10a-10e′ for details).

composition of the fraction recovered in the second condenser was fitting well with the dew point curve characteristic of the operating conditions in the first condenser (plain dew curves in Figure 4). The fractionation of the permeate between the condensers was thus effectively ruled by a flash vapor-liquid equilibrium. For similar operating conditions in the permeate circuit, the dew point in the first condenser was always higher with a 50 °C feed temperature than at 30 °C, because the flow in condensable vapors was higher at 50 °C than at 30 °C. As the flow of inert gases was constant and independent of the feed temperature, the proportion of condensable vapors flowing from the output of the membrane module to the first condenser was thus higher at 50 °C (eq 9). These results clearly demonstrated the dilution role played by the inert gases flowing in the permeate circuit. This phenomenon induced a significant lowering of the dew point of the permeate vapors in the first condenser. Global Modeling of the “Pervaporation/Flash Condensation” Coupling System. The combination of the equations describing the mass transport through the membrane and the permeate fractionation mecha-

4464

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999

nisms led to the overall modeling of the performances of the “pervaporation/two-stage-condensation” process (eqs 10a-10e′). The meaning of the abbreviations used

xw(1)γw(1)pw°(T1) + xEA(1)γEA(1)pEA°(T1) ) pT - pinert ) n˘ w(1) + n˘ w(2) + n˘ EA(1) + n˘ EA(2) p (10a) n˘ w(1) + n˘ w(2) + n˘ EA(1) + n˘ EA(2) + n˘ inert T n˘ w(2) n˘ w(2) + n˘ EA(2)

) xw(1)γw(1)p°w(T1)

(xw(1)γw(1)p°w(T1) + xEA(1)γEA(1)p°EA(T1))

) xw(2) (10b)

n˘ EA(2)

) n˘ w(2) + n˘ EA(2) xEA(1)γEA(1)p°EA(T1)

) xEA(2) ) (xw(1)γw(1)p°w(T1) + xEA(1)γEA(1)p°EA(T1)) 1 - xw(2) (10b′)

n˘ w(1) + n˘ w(2) ) JwS )

(

kwS pw(feed) -

n˘ w(1) + n˘ w(2) n˘ w(1) + n˘ w(2) + n˘ EA(1) + n˘ EA(2)

n˘ EA(1) + n˘ EA(2) ) JEAS )

(

kEAS pEA(feed) -

n˘ EA(1) + n˘ EA(2)

)

p n˘ w(1) + n˘ w(2) + n˘ EA(1) + n˘ EA(2) T (10c′)

xw(1) ) xEA(1) )

)

pT (10c)

n˘ w(1) n˘ w(1) + n˘ EA(1)

n˘ EA(1) n˘ w(1) + n˘ EA(1)

) 1 - xw(1)

(10d)

(10d′)

γw(1) ) exp((2 ln γ∞w - ln γ∞EA)(1 - xEA(1))2 + 2 ln γ∞EA - ln γ∞w (1 - xEA(1))3) (10e) γEA(1) ) exp((2 ln γ∞EA - ln γ∞w)(xEA(1))2 + 2 ln γ∞w - ln γ∞EA (xEA(1))3) (10e′) in eqs 10a-10e′ is provided in the schematic flow sheet of the pervaporation-flash condensation coupling system, displayed in Figure 5. The proposed model was based on two major assumptions: 1. The condensation in the first condenser was ruled by a “flash” vapor-liquid equilibrium. This mechanism is described by eq 10a (see Table 1 and eq 2 for the numerical expression of the saturated vapor pressure of each permeant). With the proposed series of experiments, the experimental operating conditions were chosen so that a flash condensation of the permeate vapors (diluted by the inert gases) occurred. It should nevertheless be pointed out that this is not the only possible solution to obtain a fractionation of the permeate between two condensers in series. Indeed, Marin et al.13 showed that even if the temperature in the condenser was lower than the bubble point of the permeate, a fractionation of the permeate could still occur. This

was due to a stripping by the inert gases of the vapor fraction in equilibrium with the liquid permeate collected in the first condenser toward the second condenser.15 2. The transport of the vapors between the two condensers was only convective. In this case, the molar fraction of each component at the vapor state in the first condenser was equal to the molar fraction of each component collected in the second condenser (eqs 10b and 10b′). This supposed that there was no diffusive resistance toward mass transport between the two condensers. We based this assumption, on one hand, on the fact that the vapors were flowing under vacuum (and thus the diffusive resistance in the vapors is sharply decreased if compared to ambient pressure) and, on the other hand, on the fact that the inert gases were promoting convection in the permeate circuit. Equations 10c and 10c′ are the simple mass transport equations (eqs 3 and 4) that were established previously for each permeant. The values of the transmembrane transfer coefficients kw and kEA used in eqs 10c and 10c′ were obtained from the first series of experiments (pervaporation with total condensation; cf. Table 2). It is necessary to be precise here that, if the dilution of the permeate vapors has been observed in the permeate circuit, this dilution phenomenon did not occur inside the membrane module. Indeed, as previously mentioned, it was experimentally shown on the present installation that the partial pressure of the permeants at the downstream side of the membrane was always equal to the total permeate pressure, over a very broad range of operating conditions.26 To do so, pure water pervaporation experiments were carried out and the water partial pressure inside the permeate compartment of the membrane module was measured. It appeared that the water partial pressure at the downstream side of the membrane was always equal to the total permeate pressure. It can be easily assumed that, whereas the inert gases and the condensable vapors were flowing in the same direction through the permeate circuit (from the outlet of the membrane module to the vacuum pump), the inert gases could hardly diffuse against the condensable vapors flowing from the downstream side of the membrane to the outlet of the module. As a consequence, the flow of inert gases entering the vacuum bell jar n˘ inert was not incorporated in the denominator used to calculate the permeate side mole fraction in eqs 10c and 10c′. The molar fractions of water and ethyl acetate in the first condenser are described in eqs 10d and 10d′. Equations 10e and 10e′ describe the evolution of the activity coefficient of each compound as a function of the composition in the first condenser (three-suffix Margules equation). Because of the very low temperature in the second condenser (-70 °C), no significant mass loss of permeate to the vacuum pump was observed. For each series of experiments, the system of equations was solved with the “optimization” toolbox of Matlab 4.2 (The Mathworks Inc., Natick, MA), using the Gauss-Newton method. The results of the calculation are compared to the experimental results in Figures 6 and 7. In these figures, the partial flows that were recovered in each condenser have been divided by the membrane area and expressed in terms of fluxes, in order to keep those results homogeneous with the total condensation ex-

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999 4465

Figure 6. Pervaporation with flash condensation: water fluxes collected in each condenser as a function of the temperature in the first condenser (GFT PDMS 1070 membrane; feed composition, water + 500 ppm w/w ethyl acetate; pT ) 2000 Pa; (A) Tfeed ) 30 °C/(B) Tfeed ) 50 °C).

Figure 7. Pervaporation with flash condensation: ethyl acetate fluxes collected in each condenser as a function of the temperature in the first condenser (GFT PDMS 1070 membrane; feed composition, water + 500 ppm w/w ethyl acetate; pT ) 2000 Pa; (A) Tfeed ) 30 °C/(B) Tfeed ) 50 °C).

periments displayed previously. Whatever the operating conditions, the model fitting with the experimental results was satisfactory. The differences in fluxes observed inside each series of figures was, of course, due to the difference in feed temperature. Thanks to the very favorable thermodynamic properties of the water-ethyl acetate mixture, a very efficient fractionation of the two compounds between the two condensers was observed. Indeed, the majority of the water was recovered in the first condenser whatever the temperature of the first condensation step (Figure 6) while ethyl acetate was mainly recovered in the second condenser (Figure 7). Comparison of Overall Performances: Calculation of Enrichment Factors and Rony’s Extents of Separation. To describe the overall performances of the coupling system, the apparent enrichment factor in the second condenser β(2) was related to the concentration at the input of the process, i.e., the feed concentration:

β(2) )

n˘ EA(2)MEA/(n˘ EA(2)MEA + n˘ w(2)Mw) wEA(feed)

(11)

When the temperature in the first condenser was above the value of the dew point of the condensable vapors, the permeate was wholly recovered in the second

condenser. In this case, the apparent enrichment factor in the second condenser was then equal to what was obtained by conventional total condensation. However, below the dew point, a decrease of the first condenser temperature induced an important lowering of the flow of water recovered in the second condenser (Figure 6). Consequently, the apparent enrichment factor β(2) in the second condenser, where ethyl acetate was the major fraction, sharply increased (Figure 8) when the temperature in the first condenser was decreased. At lower temperatures, the apparent enrichment factor in the second condenser could reach values 4 times higher than the enrichment factor of the pervaporation membrane with a total condensation step. Moreover, in Figure 8, the increase of the feed temperature from 30 to 50 °C induced a shifting of the apparent enrichment factor curve to the right side (which was synonymous with a more effective fractionation of the permeate with similar operating conditions in the permeate circuit). This phenomenon was of course due to the increase of the dew point of the permeate (Figure 4). Attention should nevertheless be paid to the fact that if a lower first condenser temperature induced an increase of the apparent selectivity in the second

4466

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999

condenser, it also induced a lowering of the yield of the recovery of the ester in the second condenser. Indeed, it was noticeable that the flux of ethyl acetate recovered in the second condenser diminished with the temperature in the first condenser (Figure 7). In other words, while both benefitted to the global efficiency of the process, the selectivity and the yield of recovery of the desired compound in the second condenser had an antagonistic effect. Whereas the enrichment factor allows only a qualitative characterization of the efficiency of a separation process, the extent of separation ξ, introduced by Rony,33 provides a unifying separation index that combines both the qualitative and quantitative performances of separation techniques. In membrane technology, this index has been largely used for the characterization of gas separation stages.34-37 From the definition given by Rony,33 the extent of separation of the pervaporation step alone (with a total recovery of the permeate), if considered as a continuous extraction step, is equal to

ξPV )

[

] |

|

yw A(JEA + Jw) yEA abs ) Q˙ feed xEA(feed) xw(feed) yw yEA θPV abs (12) xEA(feed) xw(feed)

|

|

where Q˙ feed is the overall feed flow entering the membrane module (mol‚s-1) and θPV the stage cut of the pervaporation stage (dimensionless), namely equal to the ratio between the total mass flow that crosses the membrane barrier and the total mass flow entering the membrane module. The extent of separation is in fact a combination of two terms. The first one is the stage cut of the pervaporation stage, while the second term gives an indication on the actual separation performances of the pervaporation step. The expression of the extent of separation can be further refined, after the expression proposed by Sirkar38 for single entry barrier separation processes. Indeed, it comes easily from eq 12:

ξPV )

[

A(JEA + Jw)yw

Q˙ feed(1 - xEA(feed))

]

abs

|

yEA(1 - xEA(feed))

xEA(feed)(1 - yEA)

|

-1

(13)

and thus, as xEA(feed) , 1

ξPV )

[

AJw

Q˙ feed(1 - xEA(feed))

]

|

abs

yEA(1 - xEA(feed))

|

-1 = xEA(feed)(1 - yEA) AJw abs|RPV - 1| (14) Q˙ feed

[ ]

where RPV is the separation factor of the pervaporation stage alone (RPV ) yEA(1 - xEA(feed))/(xEA(feed)(1 - yEA)). The composite nature of the extent of separation becomes even clearer in eq 14. The first term is the fraction of water originally present in the feed and transferred through the membrane. The second part of the equation is directly linked to the separation factor RPV of the pervaporation stage. The first term is directly proportional to the membrane area and gives an indication of the stage capacity of the pervaporation operation, whereas the second term only depends on the intrinsic

Figure 8. Pervaporation with flash condensation: apparent enrichment factor in ethyl acetate in the second condenser as a function of the temperature in the first condenser (GFT PDMS 1070 membrane; feed composition, water + 500 ppm w/w ethyl acetate; pT ) 2000 Pa).

separation ability of the pervaporation membrane. At 2000 Pa, the extent of separation of the pervaporation operation was equal to 0.05 at 30 °C feed temperature and equal to 0.1 at 50 °C feed temperature. The low values of the extent of separation of the pervaporation step were mainly due to the unfavorable stage cut of the pervaporation operation with the chosen operating conditions. It should be repeated here that this was achieved on purpose, as it was intended to maintain the feed concentration constant during each experiment. By analogy with eq 14, the performances of the “pervaporation-flash condensation” system could be evaluated thanks to the calculation of the extent of separation in the second condenser, ξ(2):

ξ(2) )

[

n˘ w(2)

Q˙ feed(1 - xEA(feed))

]

abs

|

xEA(2)(1 - xEA(feed))

xEA(feed)(1 - xEA(2))

|

-1

(15)

To compare both techniques, the ratio between the respective extents of separation of the “pervaporationflash condensation” coupling system and the pervaporation step alone are featured in Figure 9. When the temperature in the first condenser is lower than the dew point of the permeate vapors (i.e., respectively -2 and 6 °C at 30 and 50 °C feed temperatures), the fractionation of the permeate between the two condensers led to a lowering of the extent separation of the extraction operation. This means that, whereas the coupling was more selective than the pervaporation stage alone, the decrease in yield of ester recovery overcame the overall increase in selectivity. This trend becomes obvious when simplifying eq 15. As xEA(feed) , 1, eq 15 gives

ξ(2) =

[ ] n˘ w(2)

Q˙ feed

abs

|

xEA(2)

xEA(feed)(1 - xEA(2))

|

-1

(16)

By replacing the molar fraction of ester and water in the second condenser by the species flows ratios, it comes from eq 16

ξ(2) =

[ ] |

n˘ EA(2) 1 abs - n˘ w(2) Q˙ feed xEA(feed)

|

(17)

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999 4467

Figure 9. Pervaporation with flash condensation: comparison of Rony’s extent of separation of the pervaporation stage alone (ξPV) and the “pervaporation-flash condensation” coupling system (ξ(2)) as a function of the temperature in the first condenser (GFT PDMS 1070 membrane; feed composition, water + 500 ppm w/w ethyl acetate; pT ) 2000 Pa).

As the molar fraction of the ester in the feed is extremely low, the term n˘ EA(2)/xEA(feed) is always between 800 and 1200 times higher than n˘ w(2), whatever the operating conditions. Consequently, the flow of water recovered in the second condenser can be neglected in eq 17, leading to

ξ(2) =

[ ] |

n˘ EA(2) 1 abs Q˙ feed xEA(feed)

|

(18)

The extent of separation in the second condenser ξ(2) is thus mainly depending on the flow of ester recovered in the second condenser. From Figure 7, it can be observed that the lower the temperature in the first condenser the lower the flow of ester recovered in the second condenser (Figure 7), and consequently the lower the extent of separation of the “pervaporation-flash condensation” coupling system (Figure 9). Nevertheless, the decrease in extent of separation remained moderate as long as the temperature in the first condenser remained close to the dew point. Indeed, for both feed temperatures, if the temperature in the first condenser was 5 °C lower than the dew point of the permeate vapors, the extent of separation of the “pervaporation-flash condensation” system was barely 20% lower than the pervaporation stage alone, whereas the resulting enrichment factor was 4 times higher (Figure 8). In the authors opinion, such operating conditions correspond to an acceptable compromise between selectivity and yield of recovery. However, with lower temperatures in the first condenser, the recovery yield decreases dramatically leading to significantly lower values of extent of separation (Figure 9), despite an apparent increase of the coupling system selectivity (Figure 8). Concluding Remarks The aim of this study was the improvement of the recovery by pervaporation of a model aroma compound, ethyl acetate, highly diluted in aqueous feeds. This was achieved by fractionating the permeate with a flash condensation between two condensers in series. Due to the very favorable thermodynamic properties of the

water-ethyl acetate mixture, the enrichment factor observed in the second condenser (where the main fraction of ethyl acetate was recovered) was up to 4 times higher than the pervaporation operation with a conventional single total condensation step. Apart from the proven selectivity enhancement, the two-stage condensation system offers two other advantages. First, because the temperature in the first condenser is by far higher than the second condenser (which normally acts as a total condensation device), the addition of the flash condensation step should not induce a major excess energy consumption in comparison with a conventionally operated pervaporation system. Second, the increase of the selectivity can be achieved without hindering dramatically the yield of recovery of the ester in the second condenser, if the fractionation is monitored optimally. This supposes of course that the vapor-liquid equilibrium properties of the permeants are well-known, as well as the ratio of inert gases flowing in the permeate circuit. It was been shown that the choice of the optimal operating conditions should be based on the consideration of both the enrichment factor and Rony’s extent of separation. A possible technical improvement of the two-stage condensation system could be the continuous recirculation of the fraction recovered in the first condenser to the pervaporation feed. This would lead to an optimal configuration of the two-stage condensation system, as the only output of the process system would come from the second condenser, in which the selectivity is maximal. This modification is of course realistic only with larger scale installations. From a practical point of view, the flash condensation device could also be applied to the separation of the organic part of the permeate in several fractions with various volatilities. By this way, pervaporation with flash condensation could prove to be more energy-saving and less damaging for the fractionation of flavor essences or essential oils than successive distillation steps. As a last remark, because of its high selectivity, the proposed coupling system could also prove to be useful for environmental deeds such as the extraction and recycling of organic pollutants highly diluted in aqueous industrial wastes. Acknowledgment The authors thank Electricite´ de France (EDF-DER, Centre des Renardie`res, Route de Sens, BP 1, F-77250 Moret-sur-Loing, France) for its financial support and GFT (GFT GmbH, Gewerbegebiet Heinitz, D-6680 Neunkirchen Heinitz, Germany) for providing the membranes. Nomenclature A ) first constant of the Antoine law (dimensionless) B ) second constant of the Antoine law (°C) C ) third constant of the Antoine law (°C) J ) species transmembrane flux (mol‚m-2‚s-1) k ) transmembrane mass transfer coefficient (mol‚Pa-1‚ m-2‚s-1) M ) molar mass (g‚mol-1) n˘ ) total flow circulating in the permeate circuit (mol‚s-1) n˘ (1) ) partial flow recovered in the first condenser (mol‚s-1) n˘ (2) ) partial flow recovered in the second condenser (mol‚s-1) n˘ inert ) partial flow of inert gases flowing in the permeate circuit (mol‚s-1)

4468

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999

p ) partial pressure (Pa) p° ) saturated vapor pressure (Pa) pinert ) inert gas partial pressure in the permeate circuit (Pa) ppm ) part per million (10-6) pT ) total permeate pressure (Pa) Q˙ feed ) overall feed flow entering the membrane module (mol‚s-1) S ) membrane area (m2) t ) time (s) T ) temperature (°C) T1 ) temperature in the first condenser (°C) x ) mole fraction in the liquid phase (mol‚mol-1) y ) mole fraction in the permeate at the downstream side of the membrane (mol‚mol-1) v/v ) volume fraction w ) mass fraction (g‚g-1) w/w ) weight fractions Greek Symbols R ) separation factor (dimensionless) β ) enrichment factor (dimensionless) γ ) activity coefficient (dimensionless) γ∞ ) activity coefficient at infinite dilution (dimensionless) θPV ) stage cut of the pervaporation stage (dimensionless) ξ ) Rony’s extent of separation (dimensionless) Subscripts w ) refers to water EA ) refers to ethyl acetate feed ) in the feed perm ) in the permeate at the downstream side of the membrane (1) ) in the first condenser (2) ) in the second condenser Superscripts PV ) refers to the pervaporation step alone (with total condensation of the permeate) VLE ) refers to the vapor-liquid equilibrium in the feed

Literature Cited (1) Strathmann, H.; Gudernatsch, W. Pervaporation in biotechnology. In Pervaporation membrane separation processes; Huang, R. Y. M., Ed.; Elsevier: Amsterdam, The Netherlands, 1991. (2) Ishii, N.; Matsumura, M.; Kataoka, H.; Tanaka, H.; Araki, K. Diacetyl fermentation coupled with pervaporation using oleyl alcohol supported liquid membrane. Bioprocess Eng. 1995, 13, 119. (3) Lamer, T.; Spinnler, H. E.; Souchon, I.; Voilley, A. Pervaporation: an efficient process for benzaldehyde recovery from fermentation broth. Proc. Biochem. 1996, 31, 533. (4) Souchon, I. Continuous extraction of lactones produced by microbiological means. Ph.D. Dissertation, University of Burgundy, Dijon, France, 1994. (5) Hue, X.; Charbit, G.; Charbit, F. La pervaporation applique´e au traitement des rejets des distilleries d’huiles essentielles. In Re´ cents progre` s en Ge´ nie des Proce´ de´ s; Jalut, C., Ed; Tech & Doc Lavoisier: Paris, France; 1993; Vol. 30, p 309. (6) Zhang, S. Q.; Matsuura, T. Recovery and concentration of flavour compounds in apple essence by pervaporation. J. Food Process Eng. 1991, 14, 291. (7) Souchon, I.; Godiard, P.; Voilley, A. Membrane processes for the treatment of food industry waste: recovery of volatile organic compounds by pervaporation. Odours VOC’s J. 1995, 1, 124. (8) Bengtsson, E.; Tra¨gårdh, G.; Hallstro¨m, B. Concentration of apple juice from evaporator condensate using pervaporation. Lebensm. Wiss. Technol. 1992, 25, 29. (9) Baudot, A.; Souchon I.; Martin, N.; Marin, M. Application de la pervaporation au traitement d’effluents des industries alimentaires: extraction de compose´s d’aroˆme des eaux de blanchiment de chou-fleur. Ind. Alim. Agric. 1998, 115, 17.

(10) Baudot, A.; Marin, M. Pervaporation of aroma compounds: comparison of membrane performances with vapourliquid equilibria and engineering aspects of process improvements. Trans. Inst. Chem. Eng., Part C (Food Bioprod. Process.) 1997, 75, 117. (11) Escudier, J. L.; Le Bouar, M.; Moutounet, M.; Jouret, C.; Barille`re, J. M. Application and evaluation of pervaporation for the production of low alcohol wines. In Proc. 3rd Int. Conf. Pervap. Process. Chem. Ind.; Bakish, R., Ed; Bakish Materials Corp.: Englewood, NJ, 1988; p 387. (12) Bo¨ddeker, K. W.; Bengtson, G.; Pingel, H.; Siegert, H.; Sufkte, T. Treating synthetic polymer dispersion by organophilic pervaporation. In Proc. 7th Int. Conf. Pervap. Process. Chem. Ind.; Bakish, R., Ed; Bakish Materials Corp.: Englewood, NJ, 1995; p 333. (13) Marin, M.; Hammami, C.; Beaumelle, D. Separation of volatile organic compounds from aqueous mixtures by pervaporation with multi-stage condensation. J. Food Eng. 1996, 28, 225. (14) Beaumelle, D. Pervaporation process applied to the separation of organic compounds diluted in aqueous solutionssAnalysis of mass transfer. Ph.D. Dissertation, National Agronomic Institute of Paris-Grignon, Paris, France, 1994. (15) Baudot, A.; Marin, M. Dairy aroma compounds recovery by pervaporation, J. Membr. Sci. 1996, 120, 207. (16) Marin, M.; Kalantzi, K.; Gibert, H. Pervaporation process: membrane conditioning and experimental mass transfer analysis. J. Membr. Sci. 1992, 74, 105. (17) Bo¨rjesson, J.; Karlsson, H. O. E.; Tra¨gårdh, G. Pervaporation of a model apple juice aroma solution: comparison of membrane performance. J. Membr. Sci. 1996, 119, 229. (18) Cabranes, C.; Moreno, J.; Mangas, J. J. Cider production with immobilized Leuconostoc oenos. J. I. Brewing 1998, 104, 127. (19) Ciani, M.; Maccarelli, F. Oenological properties of nonSaccharomyces yeasts associated with wine-making. World J. Microbiol. Biotechnol. 1998, 14, 199. (20) Gmehling, J.; Onken, U.; Artl, W.; Grenzheuser, P.; Weidlich, U.; Kolbe, B.; Rarey, J. Vapour-liquid equilibrium data collection: DECHEMA: Dortmund, Germany, 1991; Part 1. (21) Perry, R. H.; Chilton, C. H. Chemical engineer’s handbook; McGraw-Hill: Kagokusha, Tokyo, Japan, 1984. (22) Carelli, A. A.; Crapiste, G. H.; Lozano, J. E. Activity coefficients of aroma compounds in model solutions simulating apple juice. J. Agric. Food Chem. 1991, 39, 1636. (23) Fredenslund, A.; Jones, R. L.; Prausnitz, J. M.; Groupcontribution estimation of activity coefficients in non ideal liquid mixtures. AIChE J. 1975, 21, 1086. (24) Gmehling, J.; Li, J.; Schiller, M. A modified UNIFAC (Dortmund) Model. 2. Present parameter matrix and results for different thermodynamic properties. Ind. Eng. Chem. Res. 1993, 32, 178. (25) Gmehling, J.; Lohmann, J.; Jakob, A.; Joh, R. A modified UNIFAC (Dortmund) Model. 3. Revision and extension. Ind. Eng. Chem. Res. 1998, 37, 4876. (26) Baudot, A. Pervaporation and two-stage condensation: application to the extraction of aroma compounds. Ph.D. Dissertation, National Agronomic Institute of Paris-Grignon, Paris, France, 1997. (27) Baudot, A.; Souchon, I.; Marin. M. Total permeate pressure influence on the selectivity of the pervaporation of aroma compounds. J. Membr. Sci. 1999, 158, 167. (28) Dotremont, C.; Goethaert, S.; Vandecasteele, C. Pervaporation behaviour of chlorinated hydrocarbons through organophilic membranes. Desalination 1993, 95, 177. (29) te Hennepe, H. J. C.; Smolders, C. A.; Bargeman, D.; Mulder, M. H. V. Exclusion and tortuosity effect on alcohol/water separation by zeolite-filled PDMS membranes. Sep. Sci. Technol. 1991, 103, 211. (30) Vankelekom, I. F. J.; De Beukelaer, S.; Uytterhoeven, J. B. Sorption and pervaporation of aroma compounds using zeolitefilled PDMS membranes. J. Phys. Chem. B 1997, 101, 5186. (31) Scha¨fer, T.; Rodrigues, C. M.; Crespo, J. P. S. G. Recovery of natural aroma compounds by pervaporation during a winefermentation. Presented at ICOM ’99, Toronto, June 1999; oral presentation (Food & Beverage Processing Session). (32) Prausnitz, J. M.; Lichtenthaler, R. N.; de Azevedo, E. G. Thermodynamics of Fluid-Phase Equilibria, 2nd ed.; Prentice Hall Inc.: Englewoods Cliffs, NJ, 1986. (33) Rony, P. R. The extent of separation: on the unification of the field of chemical separations. In AIChE Symposium Series;

Ind. Eng. Chem. Res., Vol. 38, No. 11, 1999 4469 Li, N. N., Malulevicius, E. S., Fels, M., Eds; American Institute of Chemical Engineers: New York, 1972; Vol. 120, p 89. (34) McCandless, F. P. Stage extent of separation in ideal countercurrent recycle membrane cascades. J. Membr. Sci. 1999, 154, 15. (35) McCandless, F. P. Separation factors in permeation stages. J. Membr. Sci. 1984, 19, 101. (36) McCandless, F. P. The extent of separation in single permeation stages. J. Membr. Sci. 1984, 17, 323. (37) Kao, Y. K.; Qiu, M. M.; Hwang, S. T. Critical evaluation

of two membrane gas permeator designs: continuous membrane column and two strippers in series. Ind. Eng. Chem. Res. 1989, 28, 1514. (38) Sirkar, K. K. On the composite nature of the extent of separation. Sep. Sci. 1977, 12, 211.

Received for review February 8, 1999 Revised manuscript received August 12, 1999 Accepted August 19, 1999 IE990095H