Separation of acetone–butanol–ethanol (ABE) from dilute aqueous solutions by pervaporation

Separation and Purification Technology 42 (2005) 273–282

Separation of acetone–butanol–ethanol (ABE) from dilute aqueoussolutions by pervaporation

Fangfang Liu1, Li Liu, Xianshe Feng∗

Department of Chemical Engineering, University of Waterloo, Waterloo, Ont., Canada N2L 3G1

Received 1 June 2004; received in revised form 26 August 2004; accepted 30 August 2004

Abstract

In acetone–butanol–ethanol (ABE) fermentation, which is a potential process of producing feed stock chemicals and liquid fuels fromrenewable biomass, product inhibition is a severe problem affecting the bioconversion. This study is concerned with the separation ofacetone–butanol–ethanol (ABE) from dilute aqueous solutions by pervaporation. Poly(ether block amide) (PEBA 2533) membranes were used.The separation of binary acetone–water, n-butanol–water and ethanol–water mixtures by the membranes was initially carried out using a rela-tively thick (100 �m) membrane to evaluate the membrane permselecivity, which was found to be in the order of n-butanol > acetone > ethanol.Ttsfb©

K

1

soulctfsrth

H

1d

he effects of feed composition, operating temperature and membrane thickness on the membrane performance were studied. It was shown thathe boundary layer effect became significant when a thinner membrane (30 �m) was used, especially for n-butanol separation. In addition, theeparation of quaternary acetone–n-butanol–ethanol–water mixtures was also studied, and the potential of the membrane for ABE extractionrom dilute aqueous solutions was demonstrated. From an application point of view, the recovery of butanol from the fermentation broth coulde a niche application for the membrane.

2004 Elsevier B.V. All rights reserved.

eywords:Pervaporation; Membrane; Poly(ether block amide); Acetone–butanol–ethanol

. Introduction

Fermentation is an attractive process for producing feedtock chemicals from renewable biomass. The productionf butanol by acetone–butanol–ethanol (ABE) fermentationsed to be one of the largest bioprocesses until the 1950s, butater it was replaced by the less expensive petroleum-basedhemical synthesis. In recent years, interest in bio-based bu-anol has been revived primarily due to concerns with fossiluel depletion, and microbial production of butanol is con-idered to be a potential source of liquid fuels. There is aelatively wide range of substrates suitable for ABE fermen-ation [1], but the process suffers from severe product in-ibition, which is one of the primary reasons that the tradi-

∗ Corresponding author. Tel.: +1 519 888 4567; fax: +1 519 746 4979.E-mail address:[email protected] (X. Feng).

1 Present address: College of Chemical and Pharmaceutical Engineering,ebei University of Science and Technology, China.

tional batch process of ABE fermentation is not economicallyviable. The low concentration of the fermentative products(<5 wt%, depending on the fermentation process) means notonly a cost intensive product separation but also a large vol-ume for downstream processing and waste water treatment[2]. Since butanol is less volatile than water, the separationof butanol from dilute aqueous solutions by distillation is un-favorable; it is estimated that at a butanol concentration of<5%, the energy consumption required for butanol purifica-tion will exceed the energy content of the butanol recovered[1].

As a result, in order to make the fermentation processeconomically attractive, more efficient butanol recovery pro-cesses are needed. Several methods (including gas stripping,adsorption, extraction, membrane distillation, perstraction,and pervaporation) (see, for example [3]) have been inves-tigated during last decade in order to improve the recoveryof ABE from the fermentation broth. Among these methods,pervaporation appears to be particularly promising. It is based

383-5866/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
oi:10.1016/j.seppur.2004.08.005

274 F. Liu et al. / Separation and Purification Technology 42 (2005) 273–282

Table 1Membranes used for butanol separation pertinent to ABE fermentation

Membranes Operation mode References

PDMS Vacuum pervaporation [5,6,13,15,20]Sweep gas pervaporation [10,11,17,19]

PDMS filledwithsilicalite

Vacuum pervaporation [5,6,9,12–14]

EPDR Vacuum pervaporation [5]SBR Vacuum pervaporation [5]PMS Vacuum pervaporation [20]PTMSP Vacuum pervaporation [7,16,18,20]Porous PP

(pore size0.2 �m)

Sweep gas pervaporation [1,3]

Porous PTFE(pore size0.1–0.45 �m)

Sweep gas pervaporation [8]

PDMS: poly(dimethyl siloxane); SBR: styrene butadiene rubber; EPDR:ethylene propylene diene rubber; PTMSP: poly[-1-(trimethylsilyl)-1-propyne]; PP: propylene; PTFE: polytetrafluoroethylene and PMS:poly(methoxy siloxane).

on the selective permeation of the ABE components througha membrane in preference to water. Pervaporation can be cou-pled with fermentation so that the inhibitory products fromthe fermentation broth can be removed continuously as soonas they are formed, thereby enhancing the process produc-tivity. Only the membrane permeated components undergoliquid–vapor phase change during pervaporation, and froman energy consumption point of view, butanol recovery bypervaporation is more economical than distillation. Unlikeperstraction, a membrane process that requires an additionalseparation step for product recovery from the extractants, per-vaporation does not involve external mass separating agent,and thus there is no harmful effect on the microorganismsin the fermentation broth. In addition, pervaporation mem-branes are generally non-porous; in the case of asymmetriccomposite membranes where a dense skin layer is supportedby a microporous substrate, it is the non-porous skin thatis in contact with the feed solution. Consequently, the fer-mentation medium can be retained by the membrane withoutclogging the pores of asymmetric membranes.

Pervaporation separation of ABE components from afermentation broth derives from selective permeation ofABE through the membrane in preference to water, andorganophilic membranes are generally required [4]. How-ever, very few polymer membranes are available for this ap-pwicoitbsb

lated to ABE separation. Further, membranes prepared frompoly[-1-(trimethylsilyl)-1-propyne], which is a glassy poly-mer with a large free volume, were also found to be selectiveto organic compound permeation [7,16,18,20]. Although mi-croporous membranes made from hydrophobic polypropy-lene [1,3] and polytetrafluoroethylene [8] have been used,they generally do not exhibit a high selectivity as the sep-aration is based on the flow of the ABE and water vaporsthrough the pores of the membrane. It is essentially a mem-brane striping process where the hydrophobic membrane pri-marily functions as a barrier that prevents the aqueous so-lution from entering the membrane pores [21]. In contrastto the nonporous membranes in pervaporation, the porousmembranes used in the stripping process rarely govern theseparation.

Poly(ether block amide) (PEBA) is a group of copoly-mers comprising of flexible polyether segments and rigidpolyamide segments. Depending on the nature and the rel-ative content of the two segments, certain PEBA polymershave attracted great interest as a promising membrane mate-rial. PEBA membranes have shown excellent selectivity forthe extraction of aroma compounds from water by pervapora-tion, especially for the enrichment of esters from dilute aque-ous solutions [22,23]. Boddeker et al. [24], who carried out ancomparative study on pervaporation of four isomeric butanolstflpmo

tplaPwstsosPspalPmbmtasro

lication. Polydimethylsiloxane (PDMS) is so far the mostidely used organophilic membrane material, as illustrated

n Table 1, which summarizes the membranes used in re-ent studies on ABE removal from model butanol solutionsr fermentation broths. Silicalite has been used as a fillern the PDMS membranes to improve the membrane selec-ivity. In addition to the PDMS-based membranes, other rub-ery polymer membranes such as poly(methyhoxy siloxane),tyrene butadiene rubber, and ethylene propylene diene rub-er [5,20] have also been investigated for pervaporation re-

hrough a PEBA 40 membrane, showed that the permeationux of the PEBA membrane was higher than the PDMS andolyether-based polyurethane membranes. Moreover, PEBAembranes were also reported to be effective for the removal

f phenol from phenolic resin wastewater streams [25,26].In this study, PEBA 2533 was chosen as the membrane ma-

erial for ABE separation. It contains 80 wt% organophilicoly(tetramethylene glycol) soft segments and 20 wt% ny-on 12 hard segments. This material has a considerably highffinity to butanol, and as a mater of fact, butanol can dissolveEBA 2533 at elevated temperatures [27]. Consequently, itas expected that the membranes would yield a good flux and

electivity in consideration of its favorable solubility proper-ies to butanol, the component to be separated from aqueousolutions. In spite of some research on PEBA membranes forrganic separation from water by pervaporation, a literatureearch showed that there has not been any study on usingEBA membranes for ABE extraction from dilute aqueousolutions. The objective of this study is to explore the ap-licability of utilizing PEBA 2533 membranes for the sep-ration of acetone–butanol–ethanol from dilute aqueous so-ution, pertinent to ABE removal from fermentation broths.ervaporation is a rate controlled process, and for a givenembrane material, a high permeation flux can be achieved

y using thin membranes. We have recently developed a newethod of preparing ultrathin PEBA 2533 membranes (as

hin as 0.3 �m) [27]. However, considering that the bound-ry layer effect (i.e. concentration polarization) will be moreignificant with thinner membranes, it was decided to useelatively thick membranes in the present study for the sakef evaluating the permselectivity of the PEBA 2533 mem-

F. Liu et al. / Separation and Purification Technology 42 (2005) 273–282 275

branes. The boundary layer effect was addressed by evalu-ating the separation performance using membranes with dif-ferent thicknesses. The effects of operating conditions on themembrane performance were also investigated.

2. Experimental

2.1. Materials

Poly(ether block amide) (PEBAX® 2533) was kindly pro-vided by Antofina Canada. Acetone, n-butanol and ethanolwere supplied by Fisher Scientific. All solvents were ofreagent grade and used directly without further purification.De-ionized water was used in preparing the aqueous feedsolutions for the pervaporation experiments.

2.2. Membrane preparation

PEBA 2533 was dissolved in n-butanol at 80 ◦C undervigorous stirring to form a homogenous polymer solutioncontaining 5 wt% of the polymer. After the polymer solutionwas kept at room temperature without disturbance for oneday to remove the air bubbles entrapped, it was cast onto aglass plate using a casting knife to form a film of the polymerstefotM

2

pdochpttswsmocTteTmc

spheric pressure, whereas the permeate pressure was main-tained below 5 mmHg using a vacuum pump. The permeationrate was determined gravimetrically by weighing the perme-ate sample collected over a given period of time. Both thefeed and permeate compositions were analyzed by a Hewlett-Packard gas chromatograph (HP5890 series II) equipped witha thermal conductivity detector. In many cases involving thepermeation of butanol, the content of butanol in the perme-ate exceeded its solubility limit at room temperature, andthe permeate sample was phase separated. Under these cir-cumstances, the permeate sample was diluted with deionizedwater when determining the overall permeate composition.

During a pervaporation run, the quantity of permeate re-moved by the membrane was kept below 0.5% of the initialfeed load so as to maintain an essentially constant feed com-position. A steady state of pervaporation was considered tohave been reached when the permeation rate and permeateconcentration became constant. All the experimental data re-ported were obtained at steady state of pervaporation. Thepermeate concentration, permeation flux and selectivity wereused to characterize the membrane performance. The mem-brane selectivity can be characterized by the separation fac-tor, which is similar to that used in distillation, defined asαAB = (yA/yB)/(xA/xB), where x and y are the mass fractionsof the permeant in the feed and permeate, respectively, andt[

3

tatmaw

3

tsgwotr

mtatin

olution with a uniform thickness. The film, together withhe glass plate, was placed in an oven at 70 ◦C for 24 h tovaporate the solvent. Then the dry membrane was peeled offrom the glass plate, followed by further drying in a vacuumven at 50 ◦C for 2 days to remove any residue solvent. Thehickness of the resulting dry membrane was measured by a

itutoyo digital micrometer.

.3. Pervaporation

The pervaporation experiments were carried out in a staticervaporation cell, and the experimental procedure has beenescribed elsewhere [28]. The permeation cell was comprisedf two detachable stainless steel parts. The upper part of theell, a cylindrical chamber of 3.3 cm in diameter and 31 cmigh, acted as the feed compartment. The membrane was sup-orted by a porous sintered stainless steel plate embedded inhe lower part of the cell. The lower and the upper parts ofhe permeation cell were set in proper alignment, and a pres-ure tight seal between the membrane and the permeation cellas formed using two rubber O-rings. The feed solution was

tirred by a magnetic stirrer located about 3 mm above theembrane surface. Vacuum was applied to the permeate side

f the membrane, and the permeate vapor was condensed andollected in a Pyrex cold trap immersed in liquid nitrogen.he permeate was sampled periodically (every 2.5 h) to de-

ermine the permeation rate and permeate composition. Theffective area of the membrane for permeation was 13.9 cm2.he operating temperature was controlled by circulating ther-ostatted water through a water jacket surrounding the feed

ompartment. In all experiments, the feed was kept at atmo-

he subscripts A and B represent the two permeating species29].

. Results and discussion

As mentioned before, permeation flux and separation fac-or are two important parameters that can be used to char-cterize the separation performance. In order to evaluatehe effects of operating conditions on the separation perfor-

ance, the membrane was tested for the separation of binarycetone–water, butanol–water and ethanol–water mixtures asell as quaternary acetone–butanol–ethanol–water mixtures.

.1. Effect of feed concentration

Pervaporative enrichments of the ABE solvents (i.e. ace-one, n-butanol and ethanol) from their respective aqueousolutions through the PEBA 2533 membrane were investi-ated first. The thickness of the membrane used in the testsas 100 �m unless specified otherwise. The concentrationsf the organic components in the feed solutions were main-ained at no more than 5 wt%, a concentration range that iselevant to typical ABE fermentation processes.

The total permeation flux and the partial organic per-eation fluxes as a function of feed concentration for

he pervaporation of binary ABE–water mixtures (i.e.cetone–water, n-butanol–water, and ethanol–water mix-ures) are shown in Fig. 1. At a given organic concentrationn the feed, the membrane permeability follows the order of-butanol > acetone > ethanol. This is in agreement with the


Fig. 1. Effect of feed concentration on water and organic fluxes forthe pervaporation of binary acetone–water (�), n-butanol–water (�) andethanol–water (©) mixtures. Operating temperature 23 ◦C.

speculation that the selective permeation is primarily due tothe preferential sorption. At an elevated temperature, PEBA2533 can be dissolved in n-butanol, but not in acetone orethanol, indicating the strong affinity between n-butanol andthe polymer. This is further supported by the results of sorp-tion experiments carried at 23 ◦C, which showed that the sol-vent uptake in the polymer (in g solvent/g polymer) is 6.83,0.71 and 0.56 for butanol, acetone and ethanol, respectively.It is believed that the relatively high permeability ofn-butanolderives primarily from its high solubility in the membrane.In addition, unlike ethanol and acetone which are completelywater miscible, butanol is only partially miscible with wa-ter. This means that when an aqueous butanol solution is incontact with the membrane, the forces that retain butanolmolecules in water are relatively weak. Therefore, for thepervaporation separation of the binary organic–water mix-tures, compared to the acetone–water and ethanol–water sys-tems, the interactions between the permeating species for n-butanol–water permeation favors the permeation of butanol,in spite of the relatively large size of the butanol molecules.

The data in Fig. 1 show that increasing the content of or-ganic compound in the feed will increase the permeation fluxof the organic compound; the increase in acetone and ethanolfluxes is almost linear, whereas the feed concentration affectsn-butanol flux more significantly. Similar trends have beenrtms

will enhance the permeation. In the concentration range stud-ied, the water flux was found to be relatively constant becausethe concentration of water, which is the major component, inthe feed did not change substantially.

The above observations have been reported for other per-vaporation systems. In a study on pervaporation of aqueousbutanol solutions through a GFT PDMS membrane, Jon-quieres and Fane [13] noticed that an increase in the feedbutanol concentration will increase the butanol flux morethan proportionally when the feed concentration was above2.5 wt%, while the water flux was almost independent ofthe feed concentration in a concentration range of 0–5 wt%at 40 ◦C. Favre et al. [6] studied the pervaporation of agroup of alcohols (including butanol, 2-methyl-2-propanol,and ethanol) through a PDMS membrane at 40 ◦C, and theyfound that the water flux was essentially independent of feedcomposition over the feed concentration range (0–10 wt%)studied, and the alcohol flux increased linearly with the feedalcohol concentration. The fluxes of alcohols having lim-ited miscibilities with water (e.g. butanol) were higher thanthe flux of an alcohol that is completely water-miscible (i.e.ethanol). Feng and Huang [28] investigated pervaporationseparation of low concentration isopropanol from water bya PDMS membrane and also found that the isopropanol fluxwas a linear function of feed concentration, while the waterfl

somtttob

Ffe

eported by Boddeker et al. [24], who studied the pervapora-ion of four butanol isomers through a homogenous PEBA 40

embrane at 50 ◦C. This is understandable because the higholubility tends to plasticize or swell the membrane, which

ux was relatively constant.The permeate concentration of the ABE components ver-

us the feed concentrations for the pervaporation of binaryrganic–water mixtures is shown in Fig. 2. Clearly, the per-eate concentration of ABE increases with an increase in

he concentrations of the organic compounds in the feed, andhe relative magnitude of organic enrichment conforms tohe order of their permeation fluxes. At a feed concentrationf 4 wt% organic compound, the overall concentration of n-utanol in the permeate is 31 wt%, as compared to a perme-

ig. 2. Effect of feed concentration on the permeate concentration of ABEor the pervaporation of binary acetone–water (�), n-butanol–water (�) andthanol–water (©) mixtures. Operating temperature 23 ◦C.


ate concentration of 15 wt% ethanol and 9 wt% acetone forethanol–water and acetone–water separation, respectively. Itshould be pointed out that the binary n-butanol–water sys-tem exhibits partial miscibility. At 25 ◦C, the solubility ofn-butanol in water is 7.7 wt%, while the solubility of water inn-butanol is 20.1 wt% [30]. Apparently, for the pervaporationseparation of n-butanol from water, if the butanol-enrichedpermeate stream falls within the immiscibility gap, it will,upon condensation, separate into two phases: the organicphase is substantially enriched in n-butanol (about 80 wt%),and the aqueous phase contains about 92 wt% water. Thephase separation will obviously further contribute to the over-all separation. After the phase separation, the aqueous phasecould be recycled to the feed stream to increase the butanolrecovery.

Fig. 3 shows the selectivity of the membrane for the per-vaporation of binary ABE–water mixtures at various feedconcentrations. The membrane selectivity tends to be higherat lower feed organic concentrations because of the weakerswelling effect. In the concentration range of 0.5–5 wt% or-ganic compounds in the feed, the membrane exhibited a per-meation selectivity of 9–13, 4–6 and 2–3 for separation ofn-butanol–water, acetone–water and ethanol–water, respec-tively. Note that the selectivity presented in Fig. 3 is basedon the overall permeate concentration. For n-butanol–watersraacns

npp

Fpe

Fig. 4. A comparison of permeate concentration with vapor phase concen-tration that would be obtained at vapor–liquid equilibrium at 25 ◦C.

equilibrium, the PEBA membrane used does not have su-perior selectivity to distillation. However, in pervaporation,only the permeate, which is a fraction of the feed, undergoesa liquid–vapor phase change, while the whole feed streamneeds to be evaporated in distillation. In addition, pervapora-tion can be operated in a single-pass mode, whereas distilla-tion involves multiple phase changes under reflux. Therefore,from an energy consumption point of view, pervaporation isstill potentially advantageous to distillation, especially whenthe light components to be distilled off are the minor compo-nents in the feed, which is the case for ABE recovery fromdilute aqueous solutions.

3.2. Effect of temperature

The effects of operating temperature on the overall andpartial permeation fluxes for the pervaporation of binaryorganic–water mixtures through the membrane are illustratedin Fig. 5, which shows the temperature dependence of the per-

eparation, the phase separation of the permeate stream caneadily be used to further augment the overall separation, ands a result the overall selectivity will be much higher. For ex-mple, at 1 wt% n-butanol in the feed, the overall selectivityan be calculated to be 405 based on the concentration of the-butanol enriched phase in a decanter that induces the phaseeparation of the permeate stream.

It is well known that distillation is a conventional tech-ique for liquid separation. As shown in Fig. 4, which com-ares the permeate concentration with the vapor phase com-osition that would be obtained on the basis of vapor–liquid

ig. 3. Effect of feed concentration on ABE–water selectivity for theervaporation of binary acetone–water (�), n-butanol–water (�) andthanol–water (©) mixtures. Operating temperature 23◦C.


Fig. 5. Effect of temperature on permeation flux for the pervaporation ofbinary acetone–water (�), n-butanol–water (�) and ethanol–water (©) mix-tures. Feed organic concentration 5 wt%.

meation follows the typical Arrhenius type of relation. Thepartial flux of water increases with an increase in the tempera-ture for pervaporation of all the three binary mixtures. In addi-tion, the magnitude of temperature dependence of water flux,which is affected by the organic compound present in the feed,is found to be in the order of n-butanol > ethanol > acetone.The organic compounds present in the feed will cause mem-brane swelling, which tends to make water permeation easier.For the pervaporation of the three binary mixtures considered,the difference in the water permeation rate tends to be largerat higher temperatures. These results are an indication thatthe membrane performance is influenced by the interactionsbetween permeating molecules.

The temperature dependence of partial fluxes of the or-ganic compounds follows a different trend. When the temper-ature increases, the permeation flux of n-butanol decreases,whereas the fluxes of acetone and ethanol increase, and therate of change in ethanol flux is more significant than that ofacetone. The apparent activation energy for the permeation ofthe three organic compounds in the binary mixture pervapo-ration can be calculated to be 8.33, −7.33, and 23.7 kJ/molfor acetone, n-butanol, and ethanol, respectively. Note thatthe activation energy values determined from the Arrheniusplot of flux versus reciprocal temperature has accounted forthe effect of temperature on the driving force for the perme-a

be regarded as a quantity characterizing the overall effect oftemperature on the permeation flux.

Based on the solution–diffusion model, the sorption anddiffusion are the two major steps in pervaporation transportthat control the permeation. In cases where the diffusion isthe rate-controlling step, an increase in temperature will in-crease the permeation flux. However, a positive temperaturedependence of permeation flux is not necessarily an indica-tion of a diffusion dominating process because the drivingforce for pervaporation is also increased due to the increasedvapor pressures. This appears to be the case for the pervapo-ration of binary ethanol–water and acetone–water mixtures.Based on the analysis of Feng and Huang [29], one will findthe activation energy of permeation (excluding the effect oftemperature on the driving force) are negative for the perme-ation of both ethanol and acetone, which implies the effect ofdiffusion on the permeation is not more significant than thatof preferential sorption. This is understandable because themolecular sizes of ethanol and acetone are relatively small,and their diffusion through the membrane is relatively easy.Therefore, the observed temperature dependence of perme-ation fluxes of ethanol and acetone is the overall effects oftemperature on their diffusivities as well as the vapor pres-sures.

To understand what renders the permeation behavioroapUbsdcdfrAwamfnmfitoaotpshw

ft
tion [29]. Thus, the activation energy for permeation can
f n-butanol different from ethanol and acetone perme-tion, one may look into the system differences in terms ofermeant–permeant and permeant–membrane interactions.nlike water-miscible ethanol, which has strong hydrogenonding force, the hydroxyl groups represent a relativelymall portion in n-butanol molecules. Even though the hy-roxyl portion forms hydrogen bonds to water, these cannotompensate for the significant disruption of the strong hy-rogen bonds among the water molecules required in orderor the butanol molecules to move between water molecules,esulting in a poor miscibility between butanol and water.s the temperature increases, the solubility of n-butanol inater increases. This means that at a given concentration

n increase in temperature will help retain more n-butanololecules in water, resulting in a decrease in the repulsive

orce between n-butanol and water molecules that facilitates-butanol molecules entering the membrane. Moreover, per-eant sorption in a polymer is generally exothermic. As arst approximation, the activation energy can be considered

o be the sum of the heat of sorption and the activation energyf diffusion. If the heat of sorption is so dominant over thectivation energy of diffusion that it prevails against the effectf temperature on the vapor pressure, then the overall activa-ion energy for permeation will be negative and a decrease inermeation rate with an increase in temperature will occur. Aimilar trend has been also observed in the pervaporation de-ydration of dimethyl carbonate, which is partially miscibleith water, using chitosan membranes [31].The permeate concentration as a function of temperature

or the pervaporation of the three binary mixtures is illus-rated in Fig. 6. In consideration of the different permeation


Fig. 6. Effect of temperature on ABE concentration in the permeate forthe pervaporation of binary acetone–water, (�) n-butanol–water (�) andethanol–water (©) mixtures. Feed organic concentration 5 wt%.

behavior of the ABE compounds, one expects that the tem-perature will have different effects on the permeate concen-tration, which is determined by the relative permeation rate ofwater and the organic compound. It is shown that as the tem-perature increases, the concentration of ethanol in the perme-ate increases, while the permeate acetone concentration de-creases slightly. Interestingly, for butanol–water separation,an increase in temperature from 23 to 70 ◦C will decrease thepermeate concentration of n-butanol from 40.3 to 11.6 wt%because of the increase in water flux and the decrease in n-butanol flux. The same trend was observed for the selectivityof the membrane for ABE–water pervaporation, as shown inFig. 7. The membrane selectivity for n-butanol–water sepa-ration is more sensitive to the temperature than do the sep-arations of acetone–water and ethanol–water. Note that theselectivity for n-butanol–water permeation is based only onthe overall composition of the permeate and the overall pro-cess selectivity will be much higher if one takes into accountthe phase separation that contributes to additional separa-tion substantially. In the typical temperature range of ABEfermentation (30–35 ◦C), the membrane exhibited a selec-tivity of approximately 10, 4, and 2.5 for the separation of

Fte

Table 2Pervaporation performance of the membranes with different thicknessesfor separation of binary mixtures (feed concentration ∼5 wt%, temperature23 ◦C)

Binary mixture Thickness(�m)

Organic flux(g/m2 h)

Water flux(g/m2 h)

Selectivity

Acetone–water 30 21.4 118.7 3.3100 4.7 22.7 4.2

n-Butanol–water 30 42.2 136.8 5.9100 19.1 46.2 8.2

Ethanol–water 30 13.1 104.4 2.5100 4.4 32.8 2.4

n-butanol–water, acetone–water and ethanol–water mixtures,respectively.

3.3. Effect of membrane thickness

As mentioned earlier, pervaporation is a rate-controlledprocess. In order to obtain a higher permeation flux, thinnermembranes are desired. However, the concentration polar-ization in the boundary layer tends to become severe as thepermeation flux increases. Due to retention of the less perme-able species on the membrane surface, a concentration gradi-ent is built up in the unstirred layer adjacent to the membranesurface. The concentration polarization generally leads to adecreased permeation flux and a lesser extent of separation.

In the above studies on the permselectivity of the mem-brane, a substantially thick membrane (100 �m) was usedto render the boundary layer effect insignificant. In order toevaluate the effect of membrane thickness on the separationperformance, a thinner membrane (30 �m) was also testedfor the separation of binary ABE–water mixtures, and theexperimental results are shown in Table 2; for comparisonpurposes the separation data of the 100 �m thick membraneare also presented. Throughout the experiments, the hydrody-namic conditions of the feed remained the same. The thinnermembrane exhibited a higher permeation flux and a lowersemat

aifltfctlpffd

ig. 7. Effect of operating temperature on ABE–water selectivity forhe pervaporation of binary acetone–water, (�) n-butanol–water (�) andthanol–water (©) mixtures. Feed organic concentration 5 wt%.

electivity than the thicker membrane. If the boundary layerffect were negligible and the membrane behaved ideally, theembrane selectivity would remain the same and the perme-

tion flux would be inversely proportional to the membranehickness. This is obviously not the case.

To better illustrate the boundary layer effect, the perme-tion flux as a function of membrane thickness is plottedn Fig. 8, where the solid lines represent the permeationux that would be obtained if it is inversely proportional

o the membrane thickness, assuming the boundary layer ef-ect to be negligible with the thick (100 �m) membrane. Itan be seen that the permeation fluxes of ethanol and acetonehrough the 30 �m thick membrane are close to the calcu-ated results, but the flux of n-butanol is much lower thanredicted. These results indicate that the boundary layer ef-ect is more significant for n-butanol–water separation thanor ethanol–water and acetone–water separations. This is un-erstandable because the significance of the concentration


polarization is influenced by both the permeation flux andthe membrane selectivity [32]. At a given hydrodynamicconditions of the feed on the membrane surface, the perme-ation flux and membrane selectivity for ethanol–water andacetone–water separation are lower than for n-butanol–waterseparation. An increase in permeation flux and/or selectivitywill intensify the effect of concentration polarization. In ad-dition, at a given temperature, the diffusivity of n-butanol inwater is lower than the diffusivity of acetone and ethanol. Thiswill also contribute to the more severe concentration polar-ization in the boundary layer for n-butanol–water separation.Therefore, in the development of thin membranes for im-proved permeation productivity, one should look into strate-gies of minimizing the concentration polarization as well,especially for membranes with very high intrinsic permsele-civities.

It should be mentioned that in the concentration bound-ary layer, the concentration of the more permeable specieson the membrane surface is lower than that in the bulk feed,while the opposite holds for the less permeable species. Thus,the boundary layer effect will cause the partial flux of theslow permeating species to increase, unless the slow perme-ating species is the major component in the mixture such thatthe concentration polarization does not result in a significantchange in its concentration on the membrane surface. In thelt

Fblodt

Fig. 9. A comparison of experimental data (�) of permeation flux withthat would be obtained without concentration polarization (solid line) forn-butanol–water permeation through the 30 �m thick membrane at varioustemperatures.

separation, the water flux is slightly lower than what wouldbe expected without considering the concentration polariza-tion. This is primarily due to the fact that because of theboundary layer effect, the concentration of n-butanol on themembrane surface is decreased, which leads to a lesser de-gree of membrane swelling by n-butanol. As a result, waterpermeation is restrained. The coupling effect in pervapora-tion due to permeant–permeant interactions has been wellrecognized [4].

While it is difficult to measure the mass transfer coefficientin the boundary layer directly, the liquid phase mass transfercoefficient can be estimated to be normally on the order of10−5 m/s [32]. The significance of the boundary layer effectcan be diagnosed by estimating the concentration polariza-tion index from the pervaporation data [32]; if the bound-ary layer effect is negligible the index will be approximatelyequal to 1. It can be calculated that for the pervaporationof n-butanol–water through the 30 �m thick membrane, theconcentration polarization index is around 0.83, indicatingthe concentration polarization is indeed not negligible. Con-sidering that only one data point was available in Fig. 8 for n-butanol–water pervaporation through the thin membrane, inorder to verify experimentally that the above observed bound-ary effect is not due to experimental error, the permeationfluxes at different temperatures have been determined, andtieot

3 s

atter case, the flux of the slow permeating species will remainhe same. It can be seen from Fig. 8 that for n-butanol–water

ig. 8. Permeation flux of acetone, n-butanol, ethanol and water vs. mem-rane thickness for the pervaporation of binary ABE–water mixtures. Solid
ines represent the permeation flux that would be observed in the absencef concentration polarization, and the symbols represent the experimentalata: (�) ethanol, (×) acetone, and (�) n-butanol. Feed organic concentra-ion 5 wt%.
bq

he results are shown in Fig. 9. Clearly the permeation fluxs lower than that would be obtained if the boundary layerffect is insignificant, and this further confirms the presencef concentration polarization for butanol–water permeationhrough the thin (30 �m) membrane.

.4. Pervaporation of quaternary aqueous ABE mixture

The separation performance of the PEBA 2533 mem-rane (thickness 100 �m) for the separation of ABE fromuaternary ABE–water mixtures at 23 ◦C was investigated,


Table 3Membrane performance for ABE separation from quaternary ABE–watermixtures

1 2 3 4

Feed concentrationa (wt%)Ethanol 1.57 4.99 0.64 10.2Acetone 0.90 1.07 0.62 1.07n-Butanol 1.11 1.04 1.91 1.03

Permeation flux (g/m2 h)Ethanol 1.37 3.70 0.73 12.2Acetone 1.23 1.42 1.05 2.24n-Butanol 3.83 3.04 6.60 5.26Water 34.6 26.8 25.4 38.9

Permeate concentrationa (wt%)Ethanol 3.34 1.06 2.18 20.9Acetone 2.99 4.06 3.11 3.82n-Butanol 9.33 8.69 19.5 8.97

SelectivityEthanol–water 2.4 2.6 4.4 2.7Acetone–water 3.8 4.6 6.5 4.7n-Butanol–water 9.6 10.1 13.2 11.6

a Balance water.

and the experimental results are shown in Table 3. In gen-eral, the pervaporation data for quaternary mixtures are con-sistent with those obtained for binary mixtures separation,and the membrane selectivity (with respect to water perme-ation) still follows the order of n-butanol > acetone > ethanol.Again, the n-butanol–water selectivity is calculated fromthe overall concentration of butanol in the permeate, andwhen the phase separation occurs, the separation factorfor the actual overall separation process would be muchhigher.

Comparing the experimental data for Runs 2 and 4, onecan see that an increase in ethanol concentration in the feedincreased the partial fluxes of both n-butanol and acetone.The partial flux of water also increased in spite of the de-crease in water content in the feed, and there was no signif-icant change in the membrane selectivity. However, a com-parison of the data for Runs 1 and 2 shows that depend-ing on the specific composition of the feed, the partial fluxof water may decrease as the feed ethanol concentration in-creases. Further, the data for Run 3 show that the membraneselectivity tends to be higher when the concentrations ofethanol and acetone in the feed are relatively low even atrelatively high concentrations of n-butanol, which can swellthe membrane more significantly than ethanol and acetone.All these results indicate that there exist coupling effectsapNstpasc

and acetone, from the ABE fermentation broth by pervapo-ration.

4. Conclusions

The separation of acetone–butanol–ethanol (ABE) fromdilute aqueous solutions by pervaporation using PEBAmembranes was investigated. It is relevant to product ex-traction from ABE fermentation, where the product com-ponents are inhibitory to the bioconversion. The mem-branes were shown to be preferentially permeable to theorganic compounds over water. The separation of binaryacetone–water, n-butanol–water and ethanol–water by themembranes was carried out initially to evaluate the mem-brane permselecivity, which was found to follow the order ofn-butanol > acetone > ethanol. The effects of feed composi-tion, operating temperature and membrane thickness on themembrane performance were studied. An attempt was madeto increase the permeation flux by reducing the membranethickness, and it was observed that the improvement in theseparation performance was compromised by the boundarylayer effect, which became more significant as the perme-ation flux was increased. This was especially the case in n-butanol–water separation for which the membrane showed ahettfslmtb

A

naHs

R
mong the permeating species in the system because of theermeant–permeant and permeant–membrane interactions.evertheless, it is shown that the membrane had a fairly high
electivity for n-butanol separation even when the concen-ration of ethanol is as high as 10 wt%. Butanol is the mainroduct of acetone–butanol–ethanol fermentation, and it islso the primary inhibitory product affecting the bioconver-ion. From an application point of view, the PEBA membranean be used to extract butanol, and to a lesser extent ethanol

igh permselecivity. In addition, the separation of quaternarythanol–acetone–n-butanol–water mixtures was studied, andhe results were generally consistent with those obtained withhe binary mixture separation. The potential of the membraneor ABE extraction from dilute aqueous solutions was demon-trated. It was shown that the membrane was particularly se-ective to butanol, which is the primary product and also the

ost inhibitory one in ABE fermentation. From an applica-ion point of view, the butanol recovery from the fermentationroth could be a niche application for the membrane.

cknowledgement

Research support from the Natural Sciences and Engi-eering Research Council (NSERC) of Canada is gratefullycknowledged. One of the authors (F.L.) also wishes to thankebei University of Science and Technology for a scholar-

hip.

eferences

[1] A. Friedl, N. Qureshi, I.S. Maddox, Continuousacetone–butanol–ethanol (ABE) fermentation using immobi-lized cells of Clostridium acetobutylicumin a packed bed reactorand integration with product removal by pervaporation, Biotechnol.Bioeng. 38 (1991) 518–527.

[2] S.A. Leeper, Membrane separation in the production of alcohol fuelsby fermentation, in: W.C. McGregor (Ed.), Membrane Separationsin Biotechnology, Marcel Dekker, 1986, pp. 161–200.


[3] N. QureshiI., I.S. Maddox, A. Friedl, Application of continuous sub-strate feeding to the ABE fermentation. Relief of product inhibitionusing extraction, perstraction, stripping, and pervaporation, Biotech-nol. Prog. 8 (1992) 382–390.

[4] K.W. Boddeker, G. Bengtson, Selective pervaporation of organicsfrom water, in: R.Y.M. Huang (Ed.), Pervaporation Membrane Sep-aration Processes, Elsevier, 1991, pp. 437–460.

[5] J.D. Kawedia, V.G. Pangarkar, K. Niranjan, Pervaporative strip-ping of acetone, butanol and ethanol to improve ABE fermentation,Bioseparation 9 (2000) 145–154.

[6] E. Faver, T.Q. Nguyen, S. Bruneau, Extraction of 1-butanol fromaqueous solutions by pervaporation, J. Chem. Technol. Biotechnol.65 (1996) 221–228.

[7] A.G. Fadeev, M.M. Meagher, S.S. Kelley, V.V. Volkov, Fouling ofpoly[-1-(trimethylsilyl)-1-propyne] membranes in pervaporative re-covery of butanol from aqueous solutions and ABE fermentationbroth, J. Membr. Sci. 173 (2000) 133–144.

[8] D.L. Vrana, M.M. Meagher, R.W. Hutkins, B. Duffield, Pervapo-ration of model acetone–butanol–ethanol fermentation product solu-tions using polytetrafluoroethylene membranes, Sep. Sci. Technol.28 (1993) 2167–2178.

[9] N. Qureshi, M.M. Meagher, J. Huang, R.W. Hutkins, Acetone bu-tanol ethanol (ABE) recovery by pervaporation using silicalite-silicone composite membrane from fed-batch reactor of Clostridiumacetobutylicum, J. Membr. Sci. 187 (2001) 93–102.

[10] N. Quershi, H.P. Blaschek, Production of acetone butanol ethanol(ABE) by a hyper-producing mutant strain of Clostridium beijerinckiiBA101 and recovery by pervaporation, Biotechnol. Prog. 15 (1999)594–602.

[11] Q. Geng, C.-H. Park, Pervaporative butanol fermentation by Clostrid-

[

[

[

[

[

aqueous solutions by pervaporation through poly(1-trimethylsilyl-1-propyne), J. Membr. Sci. 186 (2001) 205–217.

[17] M.A. Larrayoz, L. Puigjaner, Study of butanol etraction throughpervaporation in acetobutylic fermentation, Biotechnol. Bioeng. 30(1987) 692–696.

[18] A.G. Fadeev, S.S. Kelley, J.D. McMillan, Y.A. Selinskaya, V.S.Khotimsky, V.V. Volkov, Effect of yeast fermentation by-productson poly[1-(trimethylsilyl)-1-propyne] pervaporative performance, J.Membr. Sci. 214 (2003) 229–238.

[19] N. Qureshi, H.P. Blaschek, Fouling studies of a pervaporation mem-brane with commercial fermentation media and fermentation broth ofhyper-butanol-producing Clostridium beijerinckiiBA101, Sep. Sci.Technol. 34 (1999) 2803–2815.

[20] P.J. Hickey, F.P. Juricic, S. Slater, Effect of process parameters onthe pervaporation of alcohols through organophilic membranes, Sep.Sci. Technol. 27 (1992) 843–861.

[21] W.S.W. Ho, K.K. Sirkar (Eds.), Membrane Handbook, Van NostrandReinhold, 1992.

[22] P. Sampranpiboon, R. Jiraratananon, D. Uttupap, X. Feng, R.Y.M.Huang, Pervaporation separation of ethyl butyrate and isopropanolwith poly(ether block amide) (PEBA) membranes, J. Membr. Sci.173 (2000) 53–59.

[23] M.K. Djebbar, Q.T. Nguyen, R. Clement, Y. Germain, Pervaporationof aqueous ester solutions through hydrophobic poly(ether-block-amide) copolymer membranes, J. Membr. Sci. 146 (1998) 125–133.

[24] K.W. Boddeker, G. Bengtson, H. Pingel, Pervaporation of isomericbutanols, J. Membr. Sci. 54 (1990) 1–12.

[25] M. Kondo, H. Sato, Treatment of wastewater from phenolic resinprocess by pervaporation, Desalination 98 (1994) 147–154.

[26] K.W. Boddeker, H. Pingel, K. Dede, Continuous pervaporation of

[

[

[

[

[

[

ium acetobutylicumB18, Biotechnol. Bioeng. 43 (1994) 978–986.

12] N. Quershi, M.M. Meagher, R.W. Hutkins, Recovery of butanol frommodel solutions and fermentation broth using a silicalite/siliconemembrane, J. Membr. Sci. 158 (1999) 115–125.

13] A. Jonquieres, A.G. Fane, Filled and unfilled composite GFT PDMSmembranes for the recovery of butanols from dilute aqueous so-lutions: influence of alcohol polarity, J. Membr. Sci. 125 (1997)245–255.

14] J. Huang, M.M. Meagher, Pervaporative recovery of n-butanol fromaqueous solutions and ABE fermentation broth using thin-filmsilicalite-filled silicone composite membranes, J. Membr. Sci. 192(2001) 231–242.

15] W.J. Groot, M.C.H. den Reyer, T.B. de la Faille, R.G. van der Lans,K.C. Luyben, Integration of pervaporation and continuous butanolfermentation with immobilized cells. I. Experimental results, Chem.Eng. J. Biochem. Eng. J. 46 (1991) B1–B10.

16] A.G. Fadeev, Y.A. Selinskaya, S.S. Kelley, M.M. Meagher, E.G.Litvinova, V.S. Khotimsky, V.V. Volkov, Extraction of butanol from

aqueous phenol on a pilot plant scale, in: R. Bakish (Ed.), Proceed-ings of the 6th International Conference Pervaporation Processesin the Chemical Industry, Bakish Materials Corp., Englewood, NJ,1992, pp. 514–519.

27] L. Liu, A. Chakma, X. Feng, A novel method of preparing ultra-thin poly(ether block amide) membranes, J. Membr. Sci. 235 (2004)43–52.

28] X. Feng, R.Y.M. Huang, Separation of isopropanol from water bypervaporation using silicone-based membranes, J. Membr. Sci. 74(1992) 171–182.

29] X. Feng, R.Y.M. Huang, Liquid separation by membrane pervapo-ration: a review, Ind. Eng. Chem. Res. 36 (1997) 1048–1066.

30] J. Kroschwitz, M. Howe-Grant (Eds.), Encyclopedia of ChemicalTechnology, Wiley, 1998.

31] W. Won, X. Feng, D. Lawless, Pervaporation with chitosan mem-branes: separation of dimethyl carbonate/methanol/water mixtures, J.Membr. Sci. 209 (2002) 493–508.

32] X. Feng, R.Y.M. Huang, Concentration polarization in pervaporationseparation process, J. Membr. Sci. 92 (1994) 201–208.

Documents

Separation of acetone–butanol–ethanol (ABE) from dilute aqueous solutions by pervaporation