Comparison of pressure driven transport of ethanol/n-hexane mixtures through dense and microporous membranes

Chemical Engineering Science 64 (2009) 3914 -- 3927

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.e lsev ier .com/ locate /ces

Comparisonofpressuredriven transport of ethanol/n-hexanemixtures throughdenseandmicroporousmembranes

Siavash Darvishmanesh∗, Jan Degrève, Bart Van der BruggenDepartment of Chemical Engineering, Laboratory for Applied Physical Chemistry and Environmental Technology, Katholieke Universiteit Leuven, W. de Croylaan 46, B-3001 Heverlee,Belgium

A R T I C L E I N F O A B S T R A C T

Article history:Received 30 March 2009Received in revised form 18 May 2009Accepted 20 May 2009Available online 29 May 2009

Keywords:NanofiltrationSolvent resistant polymeric membraneDiffusionConvectionSolubility parameterViscositySurface tensionDielectric constant

The solvent flux was measured in binary mixtures of ethanol and n-hexane for nine solvent-stablepolymeric membranes in range of reverse osmosis (RO) to ultrafiltration (UF) (GE AK Osmonics, Dow102326, GE DK Osmonics, MPF-34, STARMEM�122, STARMEM�240, NF30, NTR7450, NF-PES-010). GC-analyses of feed and permeate samples showed a separation factor close to 1, which indicates the solventtransport occurs by convection or by coupled diffusion through the membranes. The effect of viscosity,surface tension, di-electric constant and solubility parameter of solvent on permeation rate was studiedfor four categories of membranes, i.e. RO membranes, dense nanofiltration (NF) membranes, semi-porousNF membranes and micro-porous NF membranes. While viscosity seems to be a main transport param-eter (similar composition of feed and permeate), higher fluxes of ethanol compared to n-hexane (withlower viscosity) confirmed that the transport may occur through coupled diffusion. The influences of thesolvent–membrane interaction parameters such as surface tension, polarity and solubility parameters ofsolvent and membranes for dense membranes were investigated. The effect of solvent membrane inter-action by means of solubility parameters was more pronounced compared to surface tension since therespective surface tensions of solvents are close to one another (�ethanol =21.9, �n-hexane =17.9). Partial per-meabilities were studied as well to evaluate the influence of each component on permeation of the other.Unexpected results were observed for MPF-34, NF30 and NTR7450. Further investigation confirmed thattheir polymeric structure changed in contact with the solvents. Hydrophobic STARMEM� membranes,which are expected to have a higher permeability for apolar solvents showed higher fluxes for ethanolcompared to n-hexane. The similar solubility parameter of these membranes and ethanol may increasethe permeation rate of ethanol molecules through membranes. For porous membranes viscosity was rec-ognized as the key transport parameter, while affinity between membrane and solvent has a lower effect.

© 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Nanofiltration (NF) in non-aqueous media is a recent membraneprocess, which can replace traditional separation systems in thechemical, pharmaceutical and biotechnology industry where sol-vents are used in production. In comparison with NF in aqueousmedia, performance of solvent resistant nanofiltration (SRNF) is lessexamined. Lack of solvent-resistant materials was for a long time amajor problem. However, by introducing solvent stable material inrecent years (Schmidt et al., 1999), a number of researchers havecarried out studies on the fundamentals of transport phenomenaand industrial applications of SRNF systems (Van der Bruggen et al.,2004, 2006; Li et al., 2008).

∗ Corresponding author. Tel.: +32 16322349; fax: +32 16 322991.E-mail address: [email protected] (S. Darvishmanesh).

0009-2509/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.doi:10.1016/j.ces.2009.05.032

For the development of new applications it is necessary to findout the parameters determining the process performance, and tothoroughly understand the transport mechanism of solutes andsolvents through porous and dense SRNF membranes. This shouldeventually lead to a transport model to predict fluxes and rejections,and can be used for a physico-chemical interpretation of SRNF mem-branes (Katleen Boussu et al., 2007). Attempts to understand trans-port mechanisms resulted into different transport models; however,there is no agreement on which model has to be used for SRNFmembranes. It is still not clear whether transport (both solvent andsolute) occurs by viscous flow or diffusion. Robinson et al. (2004) as-sumed viscous transport to relate the permeability of various alkanesthrough a dense polydimethylsiloxane (PDMS). Yang et al. (2001)showed that viscous flow could not describe their measurements.Han et al. (2003) used the MPF50 membrane in a solvent membraneextractor and suggested that solution–diffusion type models may bemore appropriate than pore-flow models for describing transport of

http://www.sciencedirect.com/science/journal/ces

http://www.elsevier.com/locate/ces

mailto:[email protected]

S. Darvishmanesh et al. / Chemical Engineering Science 64 (2009) 3914 -- 3927 3915

Fig. 1. The influence of swelling on a dense (a) and a porous (b) membrane.

solvents through SRNF membranes. Machado et al. (1999) measuredthe effect of temperature on the acetone flux through MPF-50 andMPF-60 membranes. It was seen that temperature had a consider-able effect on solvent flux. It was discussed that a rise in tempera-ture increases permeation flux through either a reduction in solventviscosity or an increase in solvent diffusion coefficient or by an in-crease in polymer chain mobility. Later, they successfully correlatedthe temperature effect on solvent flux through an Arrhenius plot(Machado et al., 2000).

As mentioned above, several groups have studied the perfor-mance of SRNF membranes for non-aqueous systems. Swelling of thepolymeric membrane is often reported. As result of swelling, moredifficulties to understand the transport mechanism arise comparedto aqueous systems. Robinson et al. (2004) examined fluxes of or-ganic solvents through a polydimethylsiloxane composite nanofil-tration membrane for a selection of n-alkanes, i-alkanes and cycliccompounds. They showed that swelling is a good indicator of per-meation. Due to swelling, so-called channels are formed and the sol-vent flux increases. The swelling effect was described in terms ofmembrane polymeric chain reorganization by Van der Bruggen et al.(2002). The effect of solvent on membranes was assessed by mea-suring the pure water flux and the maltose rejection for hydrophilicand hydrophobic membranes before and after exposure in differentclasses of solvents. Their results showed that the membrane per-formance shifts towards lower rejections for the hydrophilic mem-branes and towards higher pure water fluxes for the hydrophobicmembrane. The results were explained by a reorganization of themembrane structure due to the clustering of hydrophobic and hy-drophilic zones within the active layer of the membranes. Scanningelectron microscopy images also confirmed the reorganization ofpolymeric network. The influence of swelling also was argued byEbert (2005). For dense membranes polymeric chains move furtherapart during swelling, thus increasing the free volume; the mem-brane becomes more open. This would result in lower rejections. Onthe opposite side, when a porous membrane swells, the pores be-come narrower. The membrane becomes `less open', which resultsin higher rejections. This principle is illustrated in Fig. 1.

In pressure driven membrane processes NF is related to bothreverse osmosis (RO) and ultrafiltration (UF) membranes. Duo tothis spectrum, either a loose or dense structure of NF-membranescan be expected (Fig. 2). Starting at the UF-end of the spectrum,solvent transport occurs through the pores while solute separationdepends on sieving phenomena. Flux and rejection are empirically

linked with feed viscosity, membrane pore size, etc. (Bowen andJulian, 2005). On the other hand, many of the tighter NF membranes,and definitely RO-membranes, are considered to have a dense top-layer, where transport happens in free volume elements betweenthe polymer chains. In this kind of NF membranes the transportand separation mechanism in the dense top layer is different fromthe porous support layer and usually described by the simultaneoussolution and diffusion phenomena (Bhanushali and Bhattacharyya,2003; Bhanushali et al., 2001; Dijkstra et al., 2006). Differences ineither solubility or diffusivity give preferential permeation. Solu-bility depends primarily on differences in the chemical nature ofthe permeating species whereas diffusivity is determined largelyby the size and shape of these molecules and the degree of thediffusing species aggregate within the polymer (Robert and Huang,1968). Robinson et al. (2004) also found that the chemical naturehad a significant effect by observing the relative permeability ofhydrocarbon pairs of similar size and shape. They found that thepermeability of n-hexene (an equivalent paraffin) was about threetimes that of cyclohexane (cyclic paraffin) despite having the samenumber of carbon atoms. Molecular shape dominates in chemi-cally similar components but size and shape has little influence onpermeability when differences in chemical or solubility character-istics are very large. Binning (1961) found that when there wereconsiderable differences in molecular size, shape and chemical na-ture (e.g., benzene and methanol), solubility was the main factordetermining membrane selectivity. Shape and size effects pre-dominate for chemically similar molecules (alcohols, . . . ). However,molecules with large differences in chemical nature are affectedmore by parameters such as solubility than shape and size. Mate-rials with similar values of � are likely to be miscible. As ethanoland n-hexane are solvents with different chemical nature, theirpermeation strongly depends on solubility. Robinson et al. (2004)observed no change in the composition of the binary mixtures ofnon-polar solvent (n-heptane/xylene and n-hexane/cyclohexane)for a laboratory made PDMS membrane before and after filtration.Similar results were reported by Geens et al., (2005a) for binarymixtures of polar solvent (water and alcohol). The following twogeneral trends were observed through literature study: (1) in bi-nary permeation of two species of a homologous series, the lowermolecular weight species permeates preferentially (Machado etal., 1999, 2000; Geens et al., 2005a) and (2) shape and size effectspredominate for chemically similar molecules. However, moleculeswith large differences in chemical nature are affected more by

3916 S. Darvishmanesh et al. / Chemical Engineering Science 64 (2009) 3914 -- 3927

Fig. 2. Schematic representation of a nanofiltration membrane.

solvent–membrane interaction parameters such as solubility,than shape and size (Robinson et al., 2004; Machado et al., 2000;Bhanushali and Bhattacharyya, 2003; Bhanushali et al., 2001,2002;Tarleton et al., 2006a,b; Reddy et al., 1996). The solubility of a sol-vent in polymeric membrane can be expressed by the followingequation (Danner and High, 1993):

�12 = �s + Vm

RT(�1 − �2)

2 (1)

The entropic contribution, (�s), is the inverse of the coordinationnumber for the lattice structure, which is found to be between 0.3and 0.4 (Danner and High, 1993). The �'s are the Hildebrand solubil-ity parameters and these can be calculated from group contributionmethods for the solvent (1) and the polymer (2). Similar solubilityparameters indicate a good compatibility of the polymer and the sol-vent (Barton, 1991). For a particular temperature, the first term ofthe equation is constant. Therefore, the solubility would be expectedto increase with the decrease in the difference (�1 − �2)

2. Bartonreported the solubility parameter values for wide range of solventsand materials (Barton, 1990). Thus solvent–membrane interactionswere found to be highly complicated, which compromise the devel-opment of a predictive model for SRNF. As pioneers in the study ofsolvent transport through SRNF-membranes, Machado et al. (2000)introduced the resistance-in-series model, based on pore flowmodelapproach. Solvent molecular volume, viscosity and surface tensiondifference between solvent and membrane surface were indicatedas key parameters. Their model did not include the swelling effect.Later Bhanushali et al. (2001) developed a solution–diffusion basedmodel for hydrophobic membranes. A solvent-sorption parameterwith solvent molecular volume, viscosity and membrane surface en-ergy together were the flux determining factors. Geens et al. (2006)introduced a new model by adopting the Bhanushali model. In thismodel the solvent-sorption parameter was replaced by the differ-ence in surface tension between the solvent and the membrane.Dijkstra et al. (2006) used the solution diffusion with imperfectionsmodel and the Maxwell–Stefan transport equation to study the per-meation of pure solvents from different chemical families such asalcohols, ketones, alkanes and also binary solvent of hydrocarbonsthrough laboratory made PDMS membranes. These models, whichconsist of both diffusive and viscous flow types, acceptably fitted ontheir results. Unfortunately, these kinds of models are not applica-ble on commercially available membrane due to lack of informationabout the membranes material and structure.

From the literature, it is evident that different interpretations ontransport mechanism of organic solvent can be made, due to thewide diversity of membranes that can be applied (ranging from ROto UF). Therefore, in this study, fluxes of binary mixtures (ethanoland n-hexane) using commercial NF-membranes in the range of UF

to RO are studied. Transport phenomena can be investigated fromfiltration experiment with binary mixtures of ethanol and n-hexanesince they have completely different properties (polarity, viscosity,surface tension, solubility parameter, . . . ). Specific goals for this re-search are: (1) to obtain a physico-chemical interpretation of trans-port mechanism in SRNF; (2) to identify the solvent and membraneparameters influencing transport in SRNF and (3) to understand therole of pore size (membrane type).

2. Materials and methods

2.1. Solvents

Two organic solvents, ethanol and n-hexane, were selected forthis study. Although these solvents are different in most physicalproperties, they are completely miscible. The solvents were of ana-lytical grade, and were obtained from Across Organic (Across, Bel-gium). Table 1 lists the most important physical properties of thesesolvents.

2.2. Membranes

Commercial membranes from different manufacturers were se-lected. Table 1 presents the relevant membrane properties. Themembranes have been categorized base on their molecular weightcut-off (MWCO). MWCO is a parameter often used as an indicationfor the rejections which is the molecular weight of a reference com-ponent that is rejected for 90%. MWCO depends on the characteriza-tion technique (Mulder, 2003). However, in non-aqueous nanofiltra-tion a larger number of membrane–solvent–solute interactions areobserved. It was reported that solute rejections are mostly lower inorganic solvents than in aqueous solution. The MWCO, characterizedin aqueous solution, is therefore not of use in other solvents. Thiswas confirmed by experimental data on solute rejection in organicsolvents (Yang et al., 2001; Whu et al., 2000). However, MWCO stillcould be a good indicator of the membrane porosity state.

GE AK OSMONICS and GE DK OSMONICS are made of polyamide,STARMEM� series membranes are polyimide, NF30 and NF-PES-010are polyethersulfone and NTR7450 is sulfonated polyethersulfone.MPF-34 is a proprietary product. All membranes were supplied in adry form except MPF-34.

Most of the membranes are specified to be resistant in ethanoland n-hexane. To verify this, the stability of the all membranes wastested by immersing the membranes sheets in glassy Petri dishes,filled by ethanol or n-hexane. No visual damage was observed forthese membranes after one week of exposure.


Table 1Physico-chemical properties of the pure solvent used.

Solvent Molecularweight (g/mol)

Density (g/ml) Viscosity(mPa s)

Surface tension(mN/m)

Solubilityparameters(MPa1/2)

Dielectric con-stant

n-hexane 86.20 0.65 0.33 17.98 26.6 1.90Ethanol 46.10 0.78 1.08 21.99 14.9 24.90

ster

litec

h 74

50

ster

litec

h 74

50

N2

Fig. 3. Schematic diagram of dead-end stirred cells.

Table 2Membranes used in this study and membrane characteristic.

Membrane Manufacturer Pore size Membrane class Nature Applied pressure (bar)

GE AK OSMONICS GE Osmonics ∗ RO Hydrophilic 30Dow 102326 FILMTEC� ∗ RO Hydrophilic 30GE DK OSMONICS GE Osmonics 150–300 MWCOa NF (dense) Hydrophilic 20MPF-34 Koch membrane 200 MWCOb NF (dense) Hydrophilic 20STARMEM�122 (MET) 220 MWCOc NF (dense) Hydrophobic 20STARMEM�240 (MET) 400 MWCOc NF (semi-porous) Hydrophobic 20NF30 Nadir 400 MWCOd NF (semi-porous) Hydrophilic 20NTR7450 FILMTEC� 700–800 MWCO NF (porous) Hydrophilic 20NF-PES-010 Nadir 1000 MWCOd NF (porous) Hydrophilic 10

∗ denotes 99% rejection of monovalent ion salts.aBased on rejection of magnesium sulfate in water.bBased on rejection of sodium chloride water.cBased on rejection of normal alkanes dissolved in toluene.dBased on rejection of lactose in water.

2.3. Membrane cell and permeation experiments

Fig. 3 shows a scheme of the experimental set up. Solvent fluxeswere measured at room temperature in dead-end mode with a Ster-litech HP4750 Stirred Cell. The active membrane area of each cellwas 14.6 cm2 and the capacity of the vessel up to 300mL. The nitro-gen cylinder coupled with the pressure regulator was connected tothe top of each vessel to pressurize the cells. An applied transmem-brane pressure was varied from 10 to 30bar depending on mem-brane type which is indicated in Table 2. Concentration polarizationat the membrane surface is minimized by driven a Teflon-coatedmagnetic stir bar on top of the membrane. Membranes permeabil-

ity was a function of pressure and determined by measuring thepermeate volume over a constant time period. Binary mixtures ofethanol/n-hexane were used (0, 25, 50, 75 and 100wt%) as solvent.Prior to the experiments the membranes were pretreated in ethanolfor at least 24h. Then the residual solvents frommanufacturing wereflushed out with pure ethanol as well. Membrane fluxes through onemembrane sample were measured every 10min for 1h. The similarmembrane was used for a complete series of feed composition andeach series is repeated four times with different samples of the samemembrane sheet. Therefore, the obtained permeability is the aver-age of 20 experimental values. The maximum experimental errorswere up to 5% in all cases.


Fig. 4. (a) Viscosity, (b) surface tension and (c) dielectric constant of binary mixture of ethanol/n-hexane.


2.4. Solvent and membrane analysis

The viscosity, surface tension and dielectric constant of solventmixture was determined by Aspen properties� modeling software(Aspen Technology Inc., 2002a). UNIFAC-LLE model was chosen asequation of state to estimate mixture properties (Carlson, 1996). Tocalculate the dielectric constant, a FORTRAN subroutine was writ-ten based on Bruggeman's formula (View and Govinda, 1988) andsuccessfully applied into the Aspen properties� (Aspen TechnologyInc., 2002b). Fig. 4 shows the estimated viscosity, dielectric constantand surface tension of the ethanol/n-hexane binary mixtures at 25 ◦Cas a function of the composition.

2.5. Sample composition analysis

The composition of different mixtures of ethanol and n-hexane were determined using a Perkin–Elmer Autosystem XL(Perkin–Elmer Analytical Instruments, Shelton, USA) gas chro-matograph with a WCOT FUSED SILICA 50M column with CPSIL 8CB coating and flame ionization detector. The externalstandard method was applied for the analysis of the samplescompositions.

Fig. 5. (a) Normalized permeability of RO membranes and (b) n-hexane separation factor.

Ethanol and n-hexane have different diffusion velocity througha polymeric material used to manufacturing above membranes. Asignificant contribution of diffusional transport would cause separa-tion between the two solvents. Consequently, the feed and permeatecomposition would be different.

For a binary mixture consisting of components A and B, the sep-aration factor is given by Eq. (2), where yA and yB are the permeatecompositions and xA ad xB are the feed or retentate compositions.These compositions can be described by means of mole fractions,mass fractions or volume fractions.

�A/B = yA/yBxA/xB

(2)

The separation factor is chosen in such a way that its value islarger than one and so that component A permeates preferentially.If � = 1, no separation can be achieved.

3. Results and discussion

3.1. RO membranes

GE AK OSMONICS and Dow 102326 are considered to be com-pletely non-porous membrane as their MWCO is below measur-able standards. Fig. 5 shows the permeability of ethanol/n-hexane


Fig. 6. Partial permeability of RO membranes: (a) GE AK Osmonics and (b) DOW 102326.

Fig. 7. Physical interpretation of coupled diffusion transport for binary mixture of ethanol and n-hexane.

mixtures and the separation factor of the membranes as a functionof feed composition.

The largest fluxes were observed for ethanol, which has a higherviscosity than n-hexane. Although the higher permeation rate ofethanol can be explained by the effect of solvent surface tension,

its effect is not enough to account for the low permeation of n-hexane, with a very low viscosity. The permeation of solvent throughmembranes can be affected by membrane–solvent solubility, andchanges in diffusion rate due to membrane swelling. The first twosteps of the permeation process may involve dissolution of solvent


Fig. 8. (a) Normalized permeability of dense NF membranes and (b) separation factor of ethanol and hexane.

molecules into the membrane top layer and then diffusion and/orconvection of thesemolecules through themembrane polymeric ma-trix (Mulder, 2003). Differences in solubility or diffusivity (and/orconvectivity) give preferential permeation. As mentioned before,solubility depends primarily on differences in the chemical natureof the permeating solvent and polymeric structure of membrane;whereas diffusivity and convectivity are determined largely by thesize, shape, viscosity of solvent molecules and the degree the speciesaggregate within the polymer (Bhanushali et al., 2001; Robert andHuang, 1968; Reddy et al., 1996). Smaller solubility parameter dif-ferences between solvent and membrane indicate that a strongersolvent–polymer interaction takes place (Burke, 1984). The solubil-ity parameter difference between ethanol (26.6MPa1/2) and Desal-AK (polyamide = 25MPa1/2) is much smaller than (14.9MPa1/2) forn-hexane. So stronger ethanol–membrane interaction takes place,causing a higher flux of ethanol. Fig. 6 shows the partial permeabil-ity of ethanol and n-hexane for both membranes. In both cases thereis sharp decrease of the amount of ethanol permeation by additionof n-hexane to the feed. It also shows that addition of n-hexane hasa stronger effect compared to addition of ethanol on n-hexane. Asethanol has more affinity for polyamide membranes, the additionof n-hexane has a stronger effect on the ethanol permeability. N-hexane has a lower affinity for this membranes and its permeabilityis therefore less influenced by the addition of ethanol.

The separation factors close to one indicate that transport occurscompletely by convection or by coupled diffusion. On the one hand,the separation is not good enough to conclude that transport is basedon diffusion. On the other hand, viscous transport also can be ques-tioned, when permeability of pure ethanol is much higher than forpure n-hexane. For binary mixtures of solvents, the transport behav-ior is generally complex (Geens et al., 2005a). The complexity mayarise from strong interactions between the solvents molecules, in-teraction of the solvent with different segments of the membranestructure, and plasticization effects of the solvents molecules on thepolymeric matrix. In this case, coupled diffusion generally occurs,which implies that that diffusivity of one solvent is influenced bythe presence and movement of the other solvents (Meares, 1979).Uchytil et al. (1996) showed coupled diffusion of acetic acid andwater in polyvinyl alcohol membranes by using a differential per-meation method. Ni et al. (2001) observed the coupling effects inthe diffusion of water and ethanol through a polyamide membrane.These effects were explained by comparing the calculated coupledflux with non-coupled flux, which was calculated by integration ofFick's first law, using independent diffusion coefficients. They usedthe Maxwell–Stefan mass transfer equation for modeling diffusionthrough the membrane.

Here, higher permeation of ethanol is observed with no signifi-cant separation, whichmay indicate that transport occurs by coupled


Fig. 9. Partial permeability of dense NF membranes: (a) GE DK Osmonics, (b) MPF-34 and (c) STARMEM�122.

diffusion. Fig. 7 illustrates the physical interpretation of coupleddiffusion transport for binary mixture of ethanol and n-hexane.The ethanol molecules create conducive condition for transport ofn-hexane through the membrane. In absence of ethanol, the perme-ation flux drops dramatically to zero (Fig. 6).

3.2. Dense NF membranes

Membranes with MWCO from 100 to 300Da can be consideredas dense NF membranes. This group of membrane consists of Desal-DK, MPF-34 and STARMEM�122. Fig. 8 gives the permeability and


Fig. 10. (a) Normalized permeability of semi-porous NF membranes and (b) separation factor of ethanol and hexane.

separation properties of these membranes as a function of feed com-position.

The normalized permeability decreased gradually for GE DKOsmonics and STARMEM�122. This was not expected sinceSTARMEM�122 is a hydrophobic membrane, which is supposedto have affinity toward apolar solvents. The STARMEM�122 is apolyimide membrane, which is inherently semi-polar with bothhydrophilic (four oxygen and two nitrogen atoms) and hydrophobicsites (benzene ring) in its chemical structure. The solubility param-eter value of polyimide (23.2–26.8MPa1/2) is in close proximity toethanol (Silva et al., 2005). Therefore, a higher flux of ethanol canbe observed.

MPF-34 shows a different behavior. There is a sharp increase innormalized permeability for MPF-34. The pure n-hexane permeabil-ity is four times larger than pure ethanol. Swelling of the polymericmatrix may increase the permeation rate dramatically. MPF-34 pos-sibly lost its structure in contact with the solvent and the micro-valleys appear through the dense matrix. Consequently, the averagepore size increases. Van der Bruggen et al. (2002), demonstrated thisassumption by SEM images for MPF-44 and MPF-50 membranes. Inthe case of a highly swollen membrane, the key transport param-eter for porous structure shifts from solubility and surface tensionto viscosity. In the case of GE DK OSMONICS and STARMEM�122where swelling dose not change the transport nature; solvent mix-

ture transport takes place by coupled diffusion since no separationoccurs (Fig. 8b).

Fig. 9 shows the partial permeabilities of these three mem-branes. For the GE DK OSMONICS and STARMEM�122 mem-branes, similar curves were obtained. This was unexpected, sinceboth GE DK OSMONICS and MPF-34 are hydrophilic membranes,whereas STARMEM�122 is hydrophobic. Similar conclusions couldbe drawn here. Polyamide (GE DK OSMONICS) and polyimide(STARMEM�122) membranes have a similar solubility parametervalue (26.2MPa1/2 and 23.2–26.8MPa1/2, respectively). Therefore,both membranes show a similar affinity toward ethanol and n-hexane. It can be seen that the influence of n-hexane on partial per-meability of ethanol is large for these membranes, whereas oppositeresults were obtained for MPF-34. The large pores appearing afterswelling of this membrane increased the permeability of n-hexane.The ethanol permeability was not influenced by n-hexane addition.At low fractions of n-hexane, where there is a little swelling, thefeed remains polar and attraction between the membrane surfaceand the polar solvent is more important than viscosity. Therefore,the ethanol permeation remains in its plateau where the massfraction of n-hexane in feed is below 0.5.

The permeation experiments show that for membranes witha dense structure, surface tension and solvent–membrane affinityfactors (polarity and solubility) are the key transport parameters,


Fig. 11. Partial permeability of semi-porous NF membranes: (a) STARMEM�240 and (b) NF30.

instead of the solvent viscosity, which has a minor effect. It wasobserved that for RO, dense NF membranes (except MPF-34), thenormalized permeability decreases by decreasing the dielectricconstant and surface tension of solvent. However, the viscosity ofthe solvent mixture has a minor effect on the permeation rate. Incase of MPF-34, viscosity is the governing transport parameter. Thepossible reason for this occurrence is that, MPF-34 lost its structurein contact with ethanol and n-hexane. To confirm this hypothesis,pure ethanol filtration experiments have been done after the lastmeasurements, to verify whether the ethanol permeation rate issimilar before and after contact the membrane by n-hexane. Thepermeability of ethanol was observed to be nearly two times largerthan before exposure to n-hexane.

3.3. Semi-porous NF membranes

Experimental permeabilities and separation factors result forSTARMEM�240 and NF30 are presented in Fig. 10.

STARMEM�240 permeability was much the same asSTARMEM�122 (both are polyimide membrane). It may be possiblethat the viscosity was not the dominating transport parameters forSTARMEM� membranes, since the flux of ethanol is higher than ofn-hexane. In these membranes, the affinity between n-hexane and

the membrane is lower than for ethanol, indicating that the perme-ation of ethanol is larger than of n-hexane. Separation factors forSTARMEM�240 are same as for STARMEM�122 and a similar con-clusion could be drawn (transport by coupled diffusion). The othermembrane, NF30, which is known as a semi-hydrophilic membrane(Geens et al., 2005b) has a comparable MWCO than STARMEM�240.The pure n-hexane permeation is higher than ethanol, which con-firms the importance of viscosity. The hydrophilicity of membranesurface reduced the effect of solvent viscosity. However, the effectis not enough for a membrane with this MWCO to overcome theviscosity effect.

The partial permeability of these membranes is presented inFig. 11. For STARMEM�240 a similar curve than for STARMEM�122was obtained. Ethanol and n-hexane influence their mutual perme-ability in the sameway as NF30. As it was shown in Fig. 11b the effectof ethanol on n-hexane permeability is more pronounced, which con-firms the importance of viscosity in transport through NF30 mem-branes.

The post experiment by ethanol was carried out for thesemembranes as well. The ethanol permeability was identical forSTARMEM�240 after experimental procedure, whereas it was 1.2times lower for NF30. The same result was also reported by Geenset al. (2004) for pure water filtration after pretreatment in ethanol


Fig. 12. (a) Normalized permeability of porous NF membranes and (b) separation factor of n-hexane.

and n-hexane. The solubility parameters of n-hexane and NF30(polyethersulfone = 12.9MPa1/2) Vidotti et al. (2007) are close toone another. Interactions between n-hexane and the polyethersul-fone polymeric matrix of the NF30 membrane change its structure.

3.4. (Micro) porous NF membranes

NTR7450 and NF-PES-010 are considered (micro) porous NFmembranes with MWCO near UF membranes. Solvent transportthrough micro porous media such as (micro) porous NF and UFmembranes can be characterized by the pore flow model (Lenckiand Williams, 1995), where viscosity is the key transport parameter.Fig. 12 presents experimental permeability and separation factor forthese membranes.

Although viscosity is supposed to be a dominant factor for thetwo membranes, only NF-PES-010 follows the concept. The puren-hexane permeability for NF-PES-O10 is almost three times morethan pure ethanol permeability, while the n-hexane viscosity isthree times lower than for ethanol. Furthermore, no separation wasachieved for this membrane.

For NTR7450 the result is unexpected. The overall permeationamount is low and is even close to zero for n-hexane (lowest viscos-ity). The top layer of this membrane after exposure to ethanol wasvisually observed to shrink dramatically. This corresponds to the as-sumption that the pore size decreases. Hydrophilic membranes withsmaller pores have higher resistance against the semi polar (ethanol)and non-polar (n-hexane) solvent (Geens et al., 2005a). Therefore,the permeation graph decreases gradually with increasing n-hexanecomposition in the feed. The repulsion effect becomes more sensi-ble for n-hexane compare to ethanol, therefore n-hexane moleculesrejected as illustrated in Fig. 12b.

Partial permeabilities of these membranes are presented inFig. 13. The slope of the partial ethanol permeability curve declinedsharply for NTR7450. It appears that the addition of n-hexanestrongly affects the partial ethanol permeation, when ethanol ad-dition has not any effect on n-hexane curves. Back to above point,the hydrophilic nature of the membrane strongly resists againstnon-polar solvents. As a result, the ethanol partial permeabilitydecreases for this membrane. On the other hand, n-hexane hasno effect on ethanol permeability curve. However, the addition of


Fig. 13. Partial permeability of porous NF membranes: (a) NTR7450 and (b) NF-PES-010.

ethanol decreases the n-hexane permeability due to the increaseof the mixture viscosity. It is expected that for porous membraneswith no visual shrinking, viscosity determines the permeation rate.

Permeation experiments through NF-PES-010 membrane withhigh MWCO confirmed the rule of viscosity for transport throughporous membrane. The polarity effect is less pronounced here com-pared to viscosity. The NTR7450 membrane shows a poor stabilityin contact with organic solvents, therefore it could not be includedhere.

4. Conclusion

It could be concluded that the following parameters control per-meation rate through membranes: solvent solubility parameter, di-electric constant (polarity), surface tension and solvent viscosity assolvent property; solubility parameter, surface tension, polymericmaterial and definitely resistance of the membrane structure in con-tact with solvent media.

It was also concluded that the proportion of each effect on thesolvent transport is related to the pore size of membrane (generallyexpresses as MWCO, although this is not accurate description) aswell as stability of membrane in solvent media.

Separation factors of ethanol and n-hexane for all membrane areclose to one. This implies that transport takes place practically en-

tirely by convection or coupled diffusion. The permeation of binarymixture of ethanol/n-hexane showed that transport through densemembranes (solvent stable) occurs by couple diffusion, while forporous membranes transport has a convective nature. The study ofthe permeation showed the most important parameters that con-tributed in transport of solvent through the membranes. The effectof these parameters is strongly related to the membrane structure.Here theMWCOof themembrane could be an indicator (although notaccurate) for the porosity of the membranes. Swelling of some mem-branes causes a unexpected permeation result. Visual deformationof the membrane's top layer also could be considered as instabilityof the polymeric structure, however, it could be confirmed throughpermeation experiments. However, it was shown that permeationthrough dense membranes is more affected by mutual affinities ofmembrane and solvent, whereas viscosity is the major transport pa-rameter for porous membranes.

Acknowledgments

The Research Council of the K.U. Leuven is gratefully acknowl-edged for their financial support to this work (OT/2006/37). SiavashDarvishmanesh also wishes to acknowledge the assistance of Chris-tine Wouters for gas chromatography analysis.


References

Aspen Technology Inc., 2002a. Introduction to Aspen Properties� . Training Manual,AspenTech.

Aspen Technology Inc., 2002b. Aspen Properties� . System management andcustomization guide, AspenTech.

Barton, A.F.M., 1990. CRC Handbook of Polymer–Liquid Interaction Parameters andSolubility Parameters. CRC Press, Boca Raton, FL. ISBN 0-8493-3544-2, hardcover, CRC handbook, 771.

Barton, A.F.M., 1991. CRC Handbook of Solubility Parameters and Other CohesionParameters.

Bhanushali, D., et al., 2001. Performance of solvent-resistant membranes for non-aqueous systems: solvent permeation results and modeling. Journal of MembraneScience 189, 1–21.

Bhanushali, D., Kloos, S., Bhattacharyya, D., 2002. Solute transport in solvent-resistantnanofiltration membranes for non-aqueous systems: experimental results andthe role of solute–solvent coupling. Journal of Membrane Science 208, 343–359.

Bhanushali, D., Bhattacharyya, D., 2003. Advances in solvent-resistant nanofiltrationmembranes. Advanced Membrane Technology 984, 159–177.

Binning, R., 1961. Separation of liquid mixtures by permeation. Industrial &Engineering Chemistry 53, 45.

Bowen, W.R., Julian, W.S., 2005. Modelling the performance of nanofiltrationmembranes. In: Schafer, A.I., Fane, A.G., Waite, T.D. (Eds.), Nanofiltration:Principles and Applications. Elsevier, Oxford, UK.

Burke, J., 1984. Solubility parameters: theory and application. Available from:〈http://cool-palimpsest.stanford.edu/byauth/burke/solpar/〉.

Carlson, E.C., 1996. Don't gamble with physical properties for simulations. ChemicalEngineering Progress 92, 35–46.

Danner, R.P., High, M.S., 1993. Handbook of Polymer Solution Thermodynamics.Design Institute for Physical Property Data, American Institute of ChemicalEngineers. ii, New York. p. 184.

Dijkstra, M.F.J., Bach, S., Ebert, K., 2006. A transport model for organophilicnanofiltration. Journal of Membrane Science 286, 60–68.

Ebert, K., 2005. Solvent stable membranes developed at GKSS. In: XXV SummerSchool of the European Membrane Society.

Geens, J., Van der Bruggen, B., Vandecasteele, C., 2004. Characterisation of the solventstability of polymeric nanofiltration membranes by measurement of contactangles and swelling. Chemical Engineering Science 59, 1161–1164.

Geens, J., et al., 2005a. Polymeric nanofiltration of binary water–alcohol mixtures:influence of feed composition and membrane properties on permeability andrejection. Journal of Membrane Science 5, 255–264.

Geens, J., et al., 2005b. Solute transport in nonaqueous nanofiltration: effectof membrane material. Journal of Chemical Technology & Biotechnology 80,1371–1377.

Geens, J., Van der Bruggen, B., Vandecasteele, C., 2006. Transport model for solventpermeation through nanofiltration membranes. Separation and PurificationTechnology 48, 255–263.

Han, S.J., et al., 2003. Insights into the transport of toluene and phenol throughorganic solvent nanofiltration membranes. Separation Science and Technology38, 1899–1923.

Katleen Boussu, J.D.B., Charles, D., Marc, W., Kelvin, G.L., Diederik, D., Steliana,A., Ivo, F.J., Vankelecom, C.V., Van der Bruggen, B., 2007. Physico-chemicalcharacterization of nanofiltration membranes. Chem. Phys. Chem. 8, 370–379.

Lencki, R.W., Williams, S., 1995. Effect of nonaqueous solvents on the flux behaviorof ultrafiltration membranes. Journal of Membrane Science 101, 43–51.

Li, X., et al., 2008. Multifunctional membranes for solvent resistant nanofiltration andpervaporation applications based on segmented polymer networks. The Journalof Physical Chemistry B 112, 16539–16545.

Machado, D.R., Hasson, D., Semiat, R., 1999. Effect of solvent properties on permeateflow through nanofiltration membranes. Part I: investigation of parametersaffecting solvent flux. Journal of Membrane Science 163, 93–102.

Machado, D.R., Hasson, D., Semiat, R., 2000. Effect of solvent properties on permeateflow through nanofiltration membranes: Part II. Transport model. Journal ofMembrane Science 166, 63–69.

Meares, P., 1979. Transport through polymer membranes from the liquid-phase.Berichte Der Bunsen-Gesellschaft-Physical Chemistry Chemical Physics 83,342–351.

Mulder, M., 2003. Basic principles of membrane technology.Ni, X.S., Dong Ceng, F.H., 2001. Coupled diffusion of water and ethanol in a polyimide

membrane. Polymer Engineering & Science 41, 1440–1447.Reddy, K.K., et al., 1996. Membrane concentration and separation of L-aspartic acid

and L-phenylalanine derivatives in organic solvents. Separation Science andTechnology 31, 1161–1178.

Robert, Y.M., Huang, V.J.C.L., 1968. Separation of liquid mixtures by usingpolymer membranes. I. Permeation of binary organic liquid mixtures throughpolyethylene. Journal of Applied Polymer Science 12, 2615–2631.

Robinson, J.P., et al., 2004. Solvent flux through dense polymeric nanofiltrationmembranes. Journal of Membrane Science 230, 29–37.

Schmidt, M., et al., 1999. Nanofiltrationsmembranen fur Trennprobleme inorganischen Losungen. Chemie Ingenieur Technik-CIT 71, 199–206.

Silva, P., Han, S., Livingston, A.G., 2005. Solvent transport in organic solventnanofiltration membranes. Journal of Membrane Science 262, 49–59.

Tarleton, E.S., et al., 2006a. The influence of polarity on flux and rejectionbehaviour in solvent resistant nanofiltration—experimental observations. Journalof Membrane Science 278, 318–327.

Tarleton, E.S., Robinson, J.P., Salman, M., 2006b. Solvent-induced swellingof membranes—measurements and influence in nanofiltration. Journal ofMembrane Science 280, 442–451.

Uchytil, P., et al., 1996. Diffusion of acetic acid and water through poly(vinylalcohol)membranes. Coupling effects. Polymer 37, 93–100.

Van der Bruggen, B., Geens, J., Vandecasteele, C., 2002. Influence of organic solventson the performance of polymeric nanofiltration membranes. Separation Scienceand Technology.

Van der Bruggen, B., et al., 2004. Determination of parameters affecting transportin polymeric membranes: parallels between pervaporation and nanofiltration.The Journal of Physical Chemistry B 108, 13273–13279.

Van der Bruggen, B., et al., 2006. Characteristics and performance of a “Universal”membrane suitable for gas separation, pervaporation, and nanofiltrationapplications. The Journal of Physical Chemistry B 110, 13799–13803.

Vidotti, S.E., Chinellato, A.C., Pessan, L.A., 2007. Effects of antiplasticization on thethermal, volumetric, and transport properties of polyethersulfone. Journal ofApplied Polymer Science 103, 2627–2633.

View, T.O.C., Govinda, R., 1988. Dielectric constant of binary mixtures of liquids.In: 1988 Annual Report Conference on Electrical Insulation and DielectricPhenomena, p. 357.

Whu, J.A., Baltzis, B.C., Sirkar, K.K., 2000. Nanofiltration studies of larger organicmicrosolutes in methanol solutions. Journal of Membrane Science 170, 159–172.

Yang, X.J., Livingston, A.G., Freitas dos Santos, L., 2001. Experimental observations ofnanofiltration with organic solvents. Journal of Membrane Science 190, 45–55.

http://cool-palimpsest.stanford.edu/byauth/burke/solpar/

Comparison of pressure driven transport of ethanol/n-hexane mixtures through dense and microporous membranes

Documents