sorption of aroma compounds in poly(octylmethylsiloxane) (poms)
TRANSCRIPT
Journal of Membrane Science 254 (2005) 259–265
Sorption of aroma compounds in poly(octylmethylsiloxane) (POMS)
Thomas Schafera, Andreas Heintzb, 1, Joao G. Crespoa, ∗a Department of Chemistry, REQUIMTE/CQFB, Faculdade de Ciˆencias e Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal
b Department of Physical Chemistry, Universit¨at Rostock, 18051 Rostock, Germany
Received 1 June 2004; received in revised form 21 October 2004; accepted 1 December 2004Available online 24 February 2005
Abstract
The equilibrium partitioning of model aroma compounds into a membrane polymer, poly(octylmethylsiloxane) (POMS) is studied. Linearsorption isotherms were determined in the concentration range relevant to subsequent pervaporation studies, both in aqueous solution aswell as in presence of 10 wt.% of ethanol. Multi-component sorption studies confirmed sorption coefficients obtained with binary solutions,indicating that possible solute–solute interactions had no apparent effect on the partitioning of the solutes into the membrane.© 2005 Elsevier B.V. All rights reserved.
K
1
ctctbvt
µ
wltte
sa
j
y anght. Anhy-
ely
ofsucholy-) orenaac-
ymerand. A
hodsand
cro-omation,
0d
eywords:Pervaporation; Sorption; Aroma recovery; Poly(octylmethylsiloxane)
. Introduction
When two phases of different physico-chemical natureontact each other, a mixture between both will occur untilhe thermodynamic equilibrium is achieved. In the case of aross-linked polymer contacting a liquid feed solution, as ishe case in pervaporation, the mixing process will occur solelyy solutes partitioning into polymer matrix, rather than viceersa. Assuming constant pressure, this mixing is describedhermodynamically by
i,f = hi,f + si,f · T = µi,m = hi,m + si,m · T (1)
ith µi,f: chemical potential of solute componenti in theiquid feed solution;µi,m: the corresponding chemical po-ential in the polymer matrix system;hi,f, hi,m, andsi,f, si,m:he corresponding partial molar enthalpies and partial molarntropies, all at temperatureT.
Eq. (1) states that both the affinity of the polymer for theolute and the sorption (mixing) entropy determine whetherny partitioning of the solute into the polymer matrix occurs.
Hence, in absence of any solute–polymer affinity driven bexothermic enthalpy of solution in the polymer, mixing mistill take place if compensated for by a change in entropyexample is the sorption of a polar compound, water, into adrophobic polymer, poly(dimethylsiloxane) (PDMS), soloccurring owing to entropy effects rather than by affinity[1].
Although phenomenologically plausible, modellingsorption can be intricate due to complex phenomena,as, for example, a rearrangement (relaxation) of the pmer chains upon solute uptake (“membrane swelling”formation of solute clusters within the polymer, phenomwhich in themselves are thermodynamically driven. Forcurate mathematical modelling, data on the solute–polinteraction are hence required on the molecular levelcannot be tackled in a straightforward, macroscopic formwide variety of thorough studies have investigated metto model the solute–polymer interactions with reviewscritical evaluations to be found elsewhere[2–5].
The aim of this work, however, was to determine mascopically the sorption coefficients of dilute solutes fran aqueous phase into the polymer used for pervapor
∗ Corresponding author. Tel.: +351 21 294 83 85; fax: +351 21 294 83 85.E-mail addresses:[email protected] (A. Heintz),
[email protected] (J.G. Crespo).1 Fax: +49 381 498 6502.
namely poly(octylmethylsiloxane) (POMS), in particularwith regard to the application of POMS as a selective mem-brane layer during the pervaporation of aroma compoundsfrom an ongoing wine-must fermentation. The emphasis was
d.
376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserveoi:10.1016/j.memsci.2004.12.047260 T. Schafer et al. / Journal of Membrane Science 254 (2005) 259–265
hereby not only on gathering data on the partitioning, butalso on highlighting in how far the partitioning is affected bya changing feed matrix, as it occurs during the fermentationof wine-must.
2. Theory
Following from Eq.(1), and by introducing the definitionof the chemical potential, the sorption of a solute into thepolymer is thermodynamically described by the equality ofthe chemical potentials
µi,m = µi,0 + RT ln(ai,m) = µi,f = µi,0 + RT ln(ai,f )
from which followsai,m
ai,f= 1 ⇔ xi,mγi,m = xi,fγi,f ⇔ xi,m
xi,f= γi,f
γi,m= Si(T )
(2)
with ai,m, xi,m, γ i,m: activity, mole fraction, and activity co-efficient, respectively, of solutei in the membrane surface;ai,f, xi,f, γ i,f: activity, mole fraction, and activity coefficient,respectively, of solutei in the liquid phase adjacent to themembrane surface;Si : sorption coefficient ofi; all precedingparameters at a given temperatureT.
ratioo mera vityc
portp eightf of ar tioncc
S
w -bm re-s -mm low-il er,m nt,m atS
asss ntalc lised,
stemming from the fact that by using mass concentrationsthe activity of the solutes in both the liquid and the mem-brane phase are not accounted for.
For dilute solutions as applied in this study, it can be shownthat the thermodynamic sorption coefficient can be approxi-mated by
Sthi = x∗
i,m
x∗i,f
=n∗
i,m/(n∗i,m + no
polymer)
(n∗i,f/n∗
i,f + nosolvent)
= γ∗i,f
γ∗i,m
≈ n∗i,m
n∗i,f
nosolvent
nopolymer
= Smi
Mpolymer
Msolvent(4)
Hence, even with the accurate molecular weight of the poly-mer unknown but under the assumptions made, thermody-namic sorption coefficients can be inferred from mass sorp-tion coefficients for comparison reasons between differentsolutes, although the absolute values might not be exact.
In the following, sorption coefficients will be calculated inbinary and multi-component solutions as they occur in wine-must fermentation, with special regard to the role of ethanol.
3. Materials and methods
POMS-polymer (GKSS Research Centre, Germany) pre-p tos -f hesev hicha f thes ttomo wast s ac-c lv tions(
an-t de-g nds,s initiala cal-c t ofa ecteda th orw re allo ud-i nta-t . Thea d be-f orp-t assb
rdingt con-c myla also
The sorption coefficient is hence determined by thef the mole fractions of the solute in the membrane polynd the liquid, or of the inverse of the respective actioefficients of the solute in both phases.
In practice, it is more convenient to describe transhenomena in pervaporation based on, for example, w
ractions, in particular, because the molecular weightespective polymer is often not known, and its determinaumbersome[6]. Instead of usingSi from Eq.(2) it is moreonvenient to introduce the mass sorption coefficientSmass
i :
massi = w∗
i,m
w∗i,f
=m∗
i,m/m∗polymer
m∗i,f/m∗
solution
=m∗
i,m/(mopolymer+
∑ni m
∗i,m)
m∗i,f/(mo
solvent+∑n
i m∗i,f )
≈m∗
i,m/mopolymer
m∗i,f/mo
solvent(3)
ith w∗i,m, w∗
i,f : equilibrium weight fractions ofi in the memrane and the liquid, respectively;m∗
i,m, m∗i,f : equilibrium
ass of componenti in the membrane and the liquid,pectively;m∗
polymer, m∗solution: equilibrium mass of the poly
er and the solution, respectively;mopolymer, mo
solvent: initialass of polymer and the solvent, respectively. The fol
ng assumptions were made: with regard tomosolvent a neg-
igible amount of solvent sorbs in the membrane polymosolvent≈ m∗
solvent; the mass of polymer remains constaopolymer ≈ m∗
polymer; the system is highly dilute such thmassi can be approximate by the last term in Eq.(3).
Contrary to thermodynamic sorption coefficients, morption coefficients are valid only under the experimeonditions employed and can therefore not be genera
ared according to[7] in iso-octane and poured gently intatic headspace vials (10 and 20 cm3) sealed with teflonaced septa until yielding the polymer mass desired. Tials were used due to their low cross-sectional area wllowed minimising the headspace. After evaporation oolvent, a smooth layer of POMS was obtained on the bof the vial as a cast thin, homogenous film. The polymer
hen cross-linked at about 363.15 K during several hourording to a procedure described elsewhere[7]. Any residuaolatiles were desorbed during 24 h under vacuum condi10 Pa).
After preliminary experiments, the ratio between the quities of polymer and solution were varied according to theree of partitioning expected for different aroma compouuch that aroma concentration differences between thend the equilibrium value lay within a suitable range forulation and hence a minimum error. A defined amounn aqueous solution containing a defined quantity of selroma compounds was then added to the polymer, wiithout the presence of ethanol. The chemicals used wef analytical grade (Sigma–Aldrich, USA). Sorption st
es were conducted at 291.15 K (the wine-must fermeion temperature during the pervaporation experiments)roma content of the aqueous solution was determine
ore and after equilibration with the polymer and the sion (partitioning) coefficient then determined from a malance according to Eq.(2).
The concentration range studied was chosen accoo solute concentrations as occurring during wine-mustentration[8]. For ethyl hexanoate, hexyl acetate, isoacetate and linalool, sorption studies were conducted
T. Schafer et al. / Journal of Membrane Science 254 (2005) 259–265 261
at concentrations slightly above the maximum concentrationfound in wine-must (about 2 mg kg−1) for analytical reasons.The time given for equilibration was about five days for allsolutes, although it had been found later by time-dependentsorption trials, using ethyl hexanoate as a model solute, thatthe equilibrium partitioning could be completed in less thanan hour.
The solute concentration was determined by staticheadspace as was described previously[8], or byliquid–liquid extraction as follows: to 5 ml of the aqueoussolution 1 ml of chloroform was added in a teflon-sealed vial,subsequently shaken using a vortex-mixer at maximum speedfor 1 min. After phase separation, which occurred within afew minutes, the organic phase was transferred to a 1 ml GC-vial together with few microsieve granules of 5A (Merck,Germany) in order to bind any residual water which couldinterfere with the analysis. The GC-vial was capped witha teflon-seal and the solute concentration determined us-ing a Chrompack 9001 gas chromatograph equipped witha Chrompack 9010 autosampler (both Chrompack-Varian,USA). Sample injection was carried out in the Direct Injec-tion Mode using an FFAP-column of 0.53 mm inner diameter,fit with a non-polar retention gap (both Chrompack-Varian,USA). The sample volume was chosen between 1 and 3�ldepending on the sensitivity required, and GC operating con-d pu-r Pa;i in-u e:4
tiond f thes datao a so-l is a
rather brittle material that can partially disintegrate due toabrasion during handling or agitation of the solution. Thetiny amounts of polymer that might detach during handlingare significant in view of the amount of solute absorbed.Another advantage of casting the polymer film on the bot-tom of the vial was that it could be guaranteed that thepolymer was definitely always completely immersed in theliquid.
For comparison reasons, activity coefficients of the com-pounds studied in this work were determined, in dilute aque-ous solution, using a static headspace technique[9]. The sametechnique was employed for determining the co-solvent ef-fect of ethanol on the activity of the aroma compounds indilute aqueous solution.
4. Results and discussion
Fig. 1A and B depicts the weight fraction of the modelaroma solutes in the membrane as the function of their re-spective equilibrium weight fraction in the contacting liquid.The mass sorption data follow apparently a linear function,as could, in fact, be expected considering the high dilutionof the model solutes as well as the flexibility of the mem-brane polymer. Similar observations were made for the sorp-t S)b dT esh nc outo r andt Theh as ob-s hol( the
F solutio ane andf nalool (© la
itions were as follows: carrier gas: helium, 99.99999%ity (Ar Liquide, Portugal); column head pressure: 100 knitial column temperature: 308.15 K, held for three mtes; temperature rise: 20 K min−1; final column temperatur73.15 K; detector temperature: 533.15 K.
It should be pointed out that the quality of the sorpata obtained in this way (POMS cast on the bottom otatic-headspace vials) exceeded by far the quality ofbserved using pieces of polymer immersed in the arom
ution (data not shown), owing to the fact that POMS
ig. 1. Sorption of model solutes in POMS from binary aqueous feedeed: (A) ethyl hexanoate (�), hexyl acetate (�), isoamyl acetate (�) and lilcohol (�).
ion of aroma compounds in polydimethylsiloxane (PDMy Lamer et al.[10] and more recently by Trifunovic anragardh in POMS[11] for concentrations up to 10 timigher than those depicted.Table 1summarises the sorptiooefficients calculated from the linear regression carriedn the sorption data depicted, including the standard erro
he correlation coefficient, as well as data from literature.ighest standard errors and hence scatter in the data werved for the fusel oil alcohols isobutyl and isoamyl alcoup to 10% of the sorption coefficient), which stems from
ns, given by the respective equilibrium weight fractions in the membrthe); (B) 1-hexanol (�), ethyl acetate (©), isoamyl alcohol (�) and isobuty
262 T. Schafer et al. / Journal of Membrane Science 254 (2005) 259–265
Table 1Sorption coefficients Si for the model aroma compounds, as well as standarderror S.E. and correlation coefficientr2; literature data calculated from[11],and as reported in[10]
Si S.E. r2 Si [11] Si [10]
Ethyl acetate 5.4 0.2 0.973 1.3 4.1Isoamyl acetate 70.0 0.6 0.999 50.0Ethyl hexanoate 241.3 1.9 0.998 264.6Hexyl acetate 227.9 1.9 0.997 188.0Isobutyl alcohol 1.0 0.1 0.933 0.8Isoamyl alcohol 1.8 0.1 0.9541-Hexanol 6.4 0.2 0.986 3.7Linalool 27.5 0.7 0.980
fact that the low partitioning of the alcohols into the mem-brane polymer was more difficult to determine analyticallythan that of stronger sorbing compounds.
In contrast to the model aroma solutes, the sorption ofwater and ethanol in POMS was studied using pure liquids,measuring the solute uptake of the initially dry membranepolymer at equilibrium. Because the partitioning of both so-lutes was expected to be very low, using pure liquids was theonly method to obtain reliable liquid sorption data. Repeti-tive experiments gave on the basis of Eq.(2) for pure wa-ter a sorption coefficient of 5× 10−4 ± 1.3× 10−4, and forethanol 0.017± 0.005.
The sorption data presented inTable 1are of the same orderof magnitude of data reported in literature, being, in fact,in better agreement with data obtained using PDMS ratherthan with data reported on the same polymer as used in thiswork [11]. Similarly, the water uptake measured in this workwhen contacting pure water, 500 mg kg−1, was close to thatreported for PDMS, namely 1100 and 700 mg kg−1 from[10]and[1], respectively, but more than four times lower than thatreported on POMS by[11], which was 2000–3000 mg kg−1.
This deviation seems surprising at first sight becausePOMS differs from PDMS inasmuch as 40 mol.% of the
methyl-groups are substituted by octyl-groups[7], render-ing the polymer more hydrophobic and supposedly result-ing in a diminished water-uptake. However, the variationsfound might be insignificant and rather reflect the difficultyand margin of error with which the sorption of low-affinitysolutes is determined by liquid sorption experiments. Theymay furthermore be explained by different degrees of cross-linking during the polymerisation of the respective polymers.This restricts considerably the usefulness of sorption coef-ficients so obtained for the modelling of fluxes accordingto the solution–diffusion mechanism, as the error associatedtranslates directly into a corresponding error of the partialflux.
Sorption data measured for multi-component solutions(data points inFig. 2A and B) also followed a linear sorptionisotherm and could be fit very well by sorption coefficientsdetermined from binary solutions (lines inFig. 2). This indi-cated that within the solute feed concentration range studied,possible solute–solute interactions were apparently not af-fecting the individual partitioning of the solutes.
4.1. Extrapolation of sorption coefficients
It has been found for dilute aqueous solutions[12] thatthe logarithm of the activity coefficients of a homologouss ctiven pliesa thee sl
wt
F d solu brane at and lini ng bina
ig. 2. Sorption of model solutes in POMS from multi-component feehe feed: (A) ethyl hexanoate (�), hexyl acetate (�), isoamyl acetate (�)sobutyl alcohol (�). Lines indicate sorption data from experiments usi
eries of compounds increases linearly with the respeumber of carbon atoms. Such a linear relationship imccording to the definition of the chemical potential thatxcess chemical potential of solution,�µE
solution, increaseinearly with every CH2-group:
�µEsolution
RT= �(ln γi)
ith R: universal ideal gas constant;T: temperature;γ i : ac-ivity coefficient of solutei.
tions, given by the respective equilibrium weight fractions in the memndalool (©); (B) 1-hexanol (�), ethyl acetate (©), isoamyl alcohol (�) andry aqueous solutions (Fig. 1).
T. Schafer et al. / Journal of Membrane Science 254 (2005) 259–265 263
Fig. 3. Liquid activity coefficientγ i,f (left) and membrane activity coefficientγ i,m (right) of the alcohols (©), esters (�), as well as of water and ethanol, as afunction of the number of carbon atoms.
Applied to activity coefficients in dilute aqueous solutionof the respective aroma compounds studied in this work, asimilar observation is made (Fig. 3A), with only ethanol andlinalool diverting from the straight line depicted. On the ba-sis of Eq.(4), and under the assumptions made, the mem-brane activity coefficients of the respective solutes can becalculated for an arbitrary molecular weight of the poly-mer (here assumed:Mpol = 104 g mol−1) and likewise de-picted as a function of the number of carbon atoms of thesolute (Fig. 3B).
It can be seen that the excess chemical potential of solutionincreases linearly with the number of carbon atoms for boththe alcohols (except ethanol) and the esters at a very simi-lar slope, but at a different order of magnitude. This revealsthe preferential partitioning of esters, rather than alcohols,into POMS. It furthermore confirms observations made pre-viously[10] that the magnitude of solute uptake by the mem-brane cannot solely be attributed to the respective activity ofthe solute in the contacting liquid.
The calculated activities in the membrane for ethanol asthe most polar alcohol and water, the latter being the mostpolar compound studied, are shown for completeness. How-ever, it should be kept in mind that sorption data of thesecomponents were obtained from pure solutions.
Attention is drawn to the fact that the calculated membranea erw e re-a d fort r thes e ac-tt solu-t em-b ouss spons mer
do not change. Otherwise, a non-linear increase with increas-ing carbon-number would have to be expected.
4.2. Effect on solute uptake by the membrane
As a highly polar compound, the sorption of pure water inPOMS may be considered merely driven by entropy, ratherthan by a favourable physico-chemical interaction with themembrane polymer. It may hence be justified to assume thatthe membrane polymer, saturated with water, still remains inan unswollen state.
In contact with pure water, and with a measured waterequilibrium sorption coefficient of 5× 10−4 kg kg−1, POMStheoretically absorbs an amount of water that correspondsapproximately to a liquid volume of 0.5 cm3 kg−1 of poly-mer, with the latter being calculated using the density of wa-ter at 293.15 K (Table 2). The aim was now to determinefor which affinity-solute feed concentrations the membranepolymer might be as little swollen as it is in contact withpure water. For this purpose, the corresponding solute weightfraction in the membrane was calculated using the respec-tive solute liquid densities and the sorption coefficients ob-tained from binary solution measurements (Table 2). Thisprocedure did not imply that the solutes sorbed were as-sumed to exist in a liquid state within the polymer. With thes thev a ro-b asilya
ag-n , fory anep onw gly,t oreti-c ax-
ctivity coefficientsincreasewith increasing carbon numbithin a homologous series. This appears plausible as thrrangement of the polymer chains induced and require
he uptake of the solute will be more pronounced the largeolute molecule, resulting in increasing solute membranivity coefficients with increasing carbon number. InFig. 3B,he linear increase of the excess chemical potential ofion, here represented by the logarithm of the solute mrane activity coefficient, suggests that within a homologeries of components the thermodynamic phenomena reible for the solution of the solute in the membrane poly
-
welling of the polymer assumed to be correlated witholume of solute taken up, the liquid density appearedust parameter for which literature data are reliable and eccessible.
As could already be expected from the orders of mitude of difference between the sorption coefficientsielding the same solute (liquid) volume in the membrolymer (0.5 cm3 kg−1), the equilibrium feed concentrati∗i,f of the different solutes varies considerably. Interestin
he solute feed concentrations so calculated for a theally unswollen membrane coincide rather well with the m
264 T. Schafer et al. / Journal of Membrane Science 254 (2005) 259–265
imum respective solute concentrations in wine-must. Only 1-hexanol, linalool and ethanol differ considerably: 1-hexanoland linalool possess a concentration in wine-must even lowerthan calculated inTable 2, namely about 1 mg kg−1; ethanol,in contrast reaches a feed concentration about four times thatlisted (up to 10 wt.%). With regard to each aroma compoundindividually, the POMS-membrane can still be expected tobe in a practically unswollen state when contacting the wine-must feed solution, or at most in a scarcely swollen state[14],as it takes up as much affinity solute as water. This observa-tion is significant because, for example, as little as 2 mg kg−1
of ethyl hexanoate in aqueous solution result theoreticallyin the same volume of solute being taken up by the mem-brane as 23,256 mg kg−1 of ethanol (or 2.3 wt.%). Althoughseemingly trivial, this underlines how little the character ofmembrane mass transport phenomena may be inferred based
Table 2Liquid densitiesδi [13] for calculating the equilibrium mass fractionw∗
i,m (in
ppm, or mg kg−1) of solutei in the membrane corresponding to a water equi-librium partitioning volume of 0.5 cm3 kg−1 polymer (POMS), given alongwith the sorption coefficientSi for subsequently determining the respectiveequilibrium feed mass fractionw∗
i,f
δia (g cm−3) w∗
i,m
(mg kg−1)
Si w∗i,f (mg kg−1)
Ethyl acetate 0.9003 451 5.4 84IEHII1LEW
verys
Table 3Ratio between the solute activity coefficientγ i,10% in presence 10 wt.% ofethanol and in aqueous solution (at 343.15 K, see text), and sorption coeffi-cientsSi,10%for the model aroma compounds studied in presence of 10 wt.%ethanol, as well as the respective standard error S.E., correlation coefficientr2, and ratio between sorption coefficients in presence of 10 wt.% ethanol,Si,10%, and in aqueous solution,Si (values taken fromTable 1)
γ i,10%/γ i Si Si,10% S.E. r2 Si,10%/Si
Ethyl acetate 0.74 5.4 5.0 0.3 0.875 0.93Isoamyl acetate 0.59 70.0 61.5 1.0 0.994 0.88Ethyl hexanoate 0.54 241.3 198.5 1.5 0.999 0.82Hexyl acetate 0.50 227.9 183.0 1.6 0.997 0.80Isobutyl alcohol 0.77 1.0 0.9 0.1 0.741 0.90Isoamyl alcohol 0.70 1.8 1.7 0.2 0.767 0.941-Hexanol 0.57 6.4 5.5 0.2 0.969 0.86Linalool 0.47 27.5 20.4 0.4 0.986 0.74
on solute feed concentrations. For this purpose, data on theactual solute concentrationin the membrane are indispens-able.
4.3. Ternary water ethanol solutions of aromacompounds
During wine-must fermentation, the ethanol concentrationincreases from 0 to 10 wt.% owing to the yeast metabolism.Because ethanol acts as a co-solvent for all aroma compoundsin aqueous solution, the respective solute activity coefficientsdecrease in the feed solution.Table 3(first column) givesthe ratio between the solute activity coefficients in presenceof 10 wt.% of ethanol and in aqueous solution, as was de-termined by static headspace chromatography. As a conse-quence, and assuming the solute activity coefficient in themembrane being determined by solute–polymer interactionsonly, the respective sorption coefficient would be expected todecrease according to Eq.(4).
Ff(
soamyl acetate 0.8670 434 70.0 6thyl hexanoate 0.8710 436 241.3 2exyl acetate 0.8710b 436 227.9 2
sobutyl alcohol 0.8081 402 1.0 402soamyl alcohol 0.8092 405 1.8 225-Hexanol 0.8136 408 6.4 64inalool 0.8700 436 27.5 16thanol 0.7893 395 0.017 23256ater 0.9982 500 0.0005 1000000a Given for 293.15 K.b Density assumed to be equal to that of ethyl hexanoate owing to aimilar molecular structure.
ig. 4. Sorption of model solutes in POMS from ternary aqueous feed solu ium weighractions in the membrane and the feed: (A) ethyl hexanoate (�), hexyl acetate (� e©), isoamyl alcohol (�) and isobutyl alcohol (�).
tions containing 10 wt.% of ethanol, given by the respective equilibrt), isoamyl acetate (�) and linalool (©); (B) 1-hexanol (�), ethyl acetat
T. Schafer et al. / Journal of Membrane Science 254 (2005) 259–265 265
Fig. 4A and B depict the sorption of the model aromacompounds in the presence of 10 wt.% ethanol in the feedsolution. Linear sorption isotherms were determined and therespective calculated sorption coefficients are listed inTable 3(third column).
In general, a decrease of the sorption coefficients can beobserved illustrated by the ratio between the sorption coef-ficient measured in the presence of 10 wt.% of ethanol andwithout ethanol,Si,10%/Si . It should be noted that this obser-vation is strictly valid only for compounds whose sorptioncoefficients have a low standard error associated, as is thecase for the esters. However, when comparing the decreaseobserved with the one that would be expected according to thesolutes’ reduced activity in the feed solutionγ i,10%/γ i , thenthe decrease in sorption coefficients is found to be much lesspronounced than predicted. From this observation it followsdirectly that in this case also the membrane activity coeffi-cient of the solutes must be lower in the presence of ethanol.
Such a decreased membrane activity could be explainedby solute–solute interactions in the membrane polymer inaddition to existing solute–polymer interactions. Recallingthat the amount of water sorbed in POMS was in the orderof magnitude of the amount of aroma and ethanol sorbed,solute–water–ethanol interactions could be imagined to occuralsowithin the membrane.
5
sucha n bee t ofa ag-n tionr ppar-e enceo f ther ssere ts.
tallyd ora-t w-i ore-o oughs lid-i eri-m orp-t eedc com-p orp-t isl them needt piri-c er-
ating conditions, then the feed solution during sorption stud-ies should reflect as much as possible that of the pervapora-tion experiment. If, however, intrinsic solute–polymer inter-actions are to be investigated, then the sorption experimentsneed, in fact, be conducted in the presence of the target soluteonly.
Acknowledgement
We thank GKSS Research Centre (Geesthacht, Ger-many) for the kind donation of poly(octylmethylsiloxane).T. Schafer would like to acknowledge grant numberBD/9100/96 from Fundac¸ao para a Ciencia e Tecnologia(Portugal), as well as financial support from IBET (Oeiras,Portugal), and Clausmeier (Susel, Germany).
References
[1] J.M. Watson, M.G. Baron, The behaviour of water in poly-(dimethylsiloxane), J. Membr. Sci. 116 (1996) 47.
[2] A. Heintz, H. Funke, R.N. Lichtenthaler, Sorption and diffusion inpervaporation membranes, in: R.Y.M. Huang (Ed.), PervaporationMembrane Separation Processes, Elsevier Science Publishers, Ams-terdam, The Netherlands, 1991.
[3] A. Jonquieres, L. Perrin, S. Arnold, P. Lochon, Comparison of UNI-lar
lingginsembr.
ess,
r &
ofmericBrill
e-enta-99)
ion,tugal,
[ rp-their
[ on216
[ con-ods,
[ 78th
[ resis-embr.
. Conclusions
When contacting dilute aqueous solutions of aromas,s a wine-must fermentation, the membrane polymer caxpected to be practically unswollen, with the amounffinity-solutes taken up being of the same order of mitude as that of water. Within the solute feed concentraange investigated, no solute–solute interactions were ant during multi-component sorption studies. The presf ethanol in the feed solution resulted in a decrease oespective solute sorption coefficients, however, to a lextent than expected from solute feed activity coefficien
It became evident that the usefulness of experimenetermined sorption coefficients for modelling pervap
ion fluxes is limited with regard to low-affinity solutes ong to the considerable experimental error associated. Mver, sorption experiments as presented in this work, althtraightforward in their execution, possess a limited vaty as they allow only macroscopic observations. If exp
ental conditions during both the pervaporation and sion experiments differ considerably with regard to the fomposition, experimental results obtained may not beatible. This applies all the more when determining s
ion coefficients byvapoursorption experiments, as in thatter case no bulk solvent partitions additionally into
embrane polymer. Results from sorption experimentsherefore be handled with care: if the purpose is the emal “modelling” of pervaporation fluxes under certain op
QUAC with related models for modelling vapor sorption in pomaterials, J. Membr. Sci. 150 (1998) 125.
[4] A. Jonquieres, L. Perrin, A. Durand, S. Arnold, P. Lochon, Modelof vapour sorption in polar materials: comparison of Flory–Hugand related models with the ENSIC mechanistic approach, J. MSci. 147 (1998) 59.
[5] P.J. Flory, Principles of Polymer Chemistry, Cornell University PrIthaca, NY, USA, 1953.
[6] F. Rodriguez, Principles of Polymer Systems, fourth ed., TayloFrancis, USA, 1996.
[7] D. Fritsch, G. Bengtson, K.W. Boddeker, Pervaporationaqueous–organic and organic–organic mixtures using elastopolymer membranes, in: J. Kahovec (Ed.), Macromolecules,Academic Publishers, Leiden, The Netherlands, 1992.
[8] T. Schafer, G. Bengtson, H. Pingel, K.W. Boddeker, J.P.S.G. Crspo, Recovery of aroma compounds from a wine-must fermtion by organophilic pervaporation, Biotechnol. Bioeng. 62 (19412.
[9] T. Schafer, Recovery of wine-must aroma by pervaporatPh.D. thesis, Universidade Nova de Lisboa, Caparica, Por2002.
10] T. Lamer, M.S. Rohart, A. Voilley, H. Baussart, Influence of sotion and diffusion of aroma compounds in silicone rubber onextraction by pervaporation, J. Membr. Sci. 90 (1994) 251.
11] O. Trifunovic, G. Tragardh, The influence of permeant propertiesthe sorption step in hydrophobic pervaporation, J. Membr. Sci.(2003) 207.
12] R.G. Buttery, J.L. Bomben, D.G. Guadagni, L.C. Ling, Somesiderations of the volatilities of organic flavor compounds in foJ. Agric. Food Chem. 19 (1971) 1045.
13] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics,ed., CRC Press, Boca Raton, FL, 1997.
14] E. Bode, M. Busse, K. Ruthenberg, Considerations on interfacetances in the process of permeation of dense membranes, J. MSci. 77 (1993) 69.