Journal of Membrane Science, 5 (1979) 235-251 0 Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands
CHARACTERIZATION PART I. WATER
and JEAN NEEL
Ecole Nationale Supgrieure des Industries Chimiques, Lorraine, 1, rue Grandville, 54000 Nancy (France) (Received
Znstitut National Polytechnique
13, 1978; accepted in revised form April 13, 1979)
Summary The hydraulic permeabilities of several commercial ultrafiltration membranes with respect to water and organic solvents have been measured in the absence of solutes. Darcy’s law was used to describe the convective transport of water and organic solvents through the ultrafilters. The permeability coefficient defined by this law could be used to characterize the membrane behaviour towards one class of solvent. Comparative studies of organic-solvent permeabilities for two important commercial membranes, Rhone-Poulenc IRIS 3038 and 3042, are reported. The irreversible structure modification exhibited by the IRIS 3038, after annealing in dioxane, hydrocarbons, or mixtures of polar solvents, is discussed. The solvating ability of these solvents allowed an increase in the mobility of the polymer chains and a rearrangement to a more stable configuration occurred. On the contrary, the IRIS 3042 was initially in a stable state and annealing in organic solvents had no effect on its water permeability.
The ability of a semi-permeable membrane to filter out soluble macromolecules from solution has been known for more than a century. With the advent of new efficient synthetic membranes, ultrafiltration is now beginning to be adopted as a larger scale unit process [l]. The availability of organic-solvent resistant membranes increases the interest in applying ultrafiltration to non-aqueous media. As an example, the regeneration of waste lubricating oils by ultrafiltration has been studied on a pilot-plant scale
It must be recognized that most of the significant papers in this field deal with aspects of the application of the process. Further effort is now necessary to ensure the full development of this separation technique: the need for more data to aid designers and manufacturers of ultrafiltration equipment is now *Present address: Laboratoire de Chimie-Physique et Electrochimie, 118, route de Narbonne, 31400 Toulouse, Prance.
Universite Paul Sabatier,
evident . Membrane performance, as commonly quoted by most manufacturers, is derived from the pure water flux and the rejection coefficient obtained after tests with a few proteins. With such data it is not possible to predict the ability of a membrane to filter out other solutes from aqueous or non-aqueous solutions. Because the primary characteristic of a membrane is its pore size distribution -which defines selectivity and solvent flow - significant efforts have recently been made to determine the membrane structure of hemofilters . The aim of our present work in ultrafiltration is to determine how the nature of the solvent and solute affect the performance of a membrane. In this first part, we report the effect of the nature of the solvent on volume flow in the absence of any solute. For two membranes, Rhbne-Poulenc IRIS 3042 and 3038, the modification of permeability after annealing in various solvents is discussed. Experimental A. Membranes The membranes tested were as follows: RhGne-Poulenc IRIS 3022, 3042, 3038 and 3069, Millipore PTGC, Amicon UM 10, and D.D.S. GR 6. Table 1 provides a list of their characteristics, as given by the manufacturers. The membranes IRIS 3042 and 3038 (Fig. l), which were extensively evaluated in this study, were supplied by the Soci&Q Rhbne-Poulenc. They are cast from a solution of the two polyelectrolytes in N,N-dimethylformamide. After quenching in water they are annealed in hot water .
CH / 3 -kX,-CH-),-(-CH,-C-j,
-(U-i,-?H-$,-6CH2-CH-),,, LH> :O,-
Fig. 1. Chemical composition of the RhSnePoulenc IRIS 3038 and 3042 membranes. AN 69: copolymer of acrylonitrile and sodium methallyl sulphonate. ANP: copolymer of acrylonitrile and 2-methyl-Svinylpyridine quaternized with methyl sulphate.
symmetric asymmetric asymmetric asymmetric asymmetric composite asymmetric
IRIS IRIS IRIS IRIS UM 10 PTGC GR6
10 000 20 000
10 000 9 600
40 20 20 20
1000 7 000 16 000 4 000
anionic polyelactrolyte polyelectrolyte complex polyelectrolyte complex anionic polyelectrolyte
000 000 000 000
Mol. wtb cut off
‘Water fluxes were brought back to the values under 2 bars, bMolecular weight cut off levels are relative to proteins in aqueous solution. ‘DDS = De Danske Sukkerfabrikker.
3069 3042 3038 3022
Characteristics of tested commercial ultrafilters as given by manufacturers
B. Sample preparations 1. Water permeability measurements
The membranes were stored either with glycol as a wetting agent, or in a solution of 1% formalin to prevent microbial growth, or both. Samples were initially washed with freshly distilled water for 30 min. Residual matter was removed by flushing freshly distilled water through the membranes after they had been installed in the cell. 2. Permeability measurements with water-miscible organic solvents After determination of the water permeability, the sample was taken out of the cell, blotted with filter paper, and directly immersed in the organic liquid under study. To ensure the complete elimination of imbibed water, the membranes were soaked in four lots of solvent, the solvent being changed every 15 minutes. After washing, the sample was tested. Since the conditioning time can influence the mass-transfer properties, other samples were stored in the solvent, between tests, for different lengths of time (from one hour to one month). 3. Permeability measurements with water-immiscible organic solvents After testing with water, the sample was first conditioned in an intermediate solvent miscible both with water and with the solvent under study. Ethanol was used in the case of benzene and toluene, acetone in the case of heptane, decane, and chloroform. After four washings in the intermediate solvent, the membrane was immersed in the water-immiscible liquid and the conditioning procedure was repeated, 4. Successive treatments in organic solvents and water
Some tests were carried out to follow the variation of the permeability when two solvents were used alternately. The membrane was conditioned as described above before each experiment. C. Experimental
apparatus and test procedure
A batch stirred cell (Amicon 401 S) with turbulent flow at the membranesolution interface was used in all the measurements. The feed volume was 300 ml, the effective area was 39.2 cm*, and the rotational speed was 300 rpm. The cell was provided with a heating jacket in which water was circulated at a temperature controlled (? O.l"C) by means of a sensitive thermostat. The flow of nitrogen used to pressurize the feed was governed by a precise valve. A mercury manometer accurate to within it: 1 mmHg, was located near both the valve and the cell. The ultrafiltration equipment is schematically shown in Fig. 2. To test a film sample after conditioning, the cell was filled with the liquid under study and the temperature was allowed to reach the desired value. The pressure was then applied. The times required to collect 25 ml or 50 ml of
Fig. ‘2.Ultrafiltration equipment: (1) applied nitrogen pressure; (2) differential mercury manometer; (3) ultrafiltration cell; (4) heating jacket; (5) magnetic stirring table; (6) vent valve; (7) thermowell. ultrafiltrate sample were measured. The volume flux was considered as stabilized when the difference between three successive measured times was less than 5%.
D. Chemicals All the organic solvents were of analytical grade. Prior to use they were filtered with 0.15 pm filter (Sartorius SM 11608). Dioxane was carefully dried over sodium after distillation. The viscosities of pure solvents were taken from the Handbook of Chemistry and Physics (The Chemical Rubber Company, Cleveland, Ohio, 1970, 50th edn.) and those of mixtures were measured with an Ubbeholde viscometer.
Results and discussion A. Ultrafiltration of pure water The ultrafiltration rates of water were measured at different temperatures, hts of hydraulic fhx Jv against the viscosity 77are shown in Fig. 3. The data illustrate an observation made previously by Pace et al.  with Abcor HFA 306 membranes and by Grieves et al.  with Amicon UM membranes: the Water flux is inversely proportional to the viscosity. Figure 4 shows the transport measurements for the same membranes when the pressure is increased at constant temperature. The well known linear relation between Jv and P is observed with all membranes tested. These results indicate that the convective transport of water through an ultrafilter can be described by Darcy’s law as used for Porous media: Jv = B,AP/qe
5c- ~______-_---------~=L!& 1Pz -____ --------
111 1.2 1 / viscosity iJ ( cp-1)
Fig. 3. Water flux versus temperature and viscosity for some ultrafiltration membranes. (e) IRIS 3038 (AP = 1 atm); (0) IRIS 3042 (AP = 2 atm); (0) UM 10 (AP = 2 atm); (A) IRIS 3069 (AP = 2 atm).
Fig. 4. Water flux versus pressure at 20°C for some ultrafiltration membranes: (0) IRIS 3042; (A) UM 10; (x ) IRIS 3069.
(e) IRIS 3038;
where J, is the volume flux density (ml/cm’-s), AP is the hydraulic pressure difference (atm), e is the thickness of the porous medium (cm), 77is the viscosity of the fluid (cP) and B, is the permeability coefficient of the porous material. In general, the membranes are not homogeneous and to take into account the asymmetry of the media, it is more consistent to define a permeability coefficient as:
This intrinsic permeability
coefficient for a membrane
can be expressed
the relation: L;I, = Jv g/AP It kust be noted that Li q can be easily related conductivity coefficient &, by the relation:
(3) to the widely used hydraulic (4)
L, = L”p,q 177
Li?, 1 is the volume flux density of pure water with a viscosity of 1 CP across on; square cm of membrane when the applied pressure difference is 1 ah. %l is expressed in cm3-cP/cm’ s-atm. The values of LAa for the tested commercial membranes are given in Table 2. Knowing L p,t7, the flux of water can be calculated at any temperature and pressure by equation (3a):
Jv = L;,,, AP/v
The values of J, in liter/m’-day-atm are also shown in Table 2. JV values for PTGC and GR 6 membranes are in fair agreement with values given by the manufacturers and reproducible water fluxes were difficult to obtain; variation of +30% was observed by changing the sample. It must be noted that the reported J, values were measured with a fresh membrane sample just after washing it with newly distilled water. TABLE
end measured water fluxes in comparison with the water fluxes given
by the manufacturers
for the tested membranes Membrane
(cm -cP/cm*-aatm) Flux of water at 20% and 1 atm (I/day-m’-
UM 10 3.0
atm) (this work) Flux of water at 20°C and 1 atm (l/day-m*-
atm) (from Table 1)
B. Ultrafiltration of organic solvents The hydraulic permeabilities of the membranes were measured with several organic liquids. The selection of the solvents tested was based on the need to cover a wide range of potential applications. Water-miscible solvents (alcohols, acetone and dioxane) and water-immiscible solvents (hydrocarbons, chloroform) were selected. The results will be expressed in terms of a dimensionless parameter CYcalled the “permeability ratio”:
organic solvent hydraulic
of the membrane:
(6) Thus, the CYratio is a measure of the effect of the organic solvent on the structure of the membrane. For example, a.> 1 will mean that the liquid “sees” a looser membrane than water does and 01= 1 (i.e. L”p rl = L”p,i, ) will show that the permeability is not modified when the membrane is used successively in water, then in the solvent.
1. Ultrafiltration of water-miscible solvents: alcohols and acetone The ultrafiltration rates were first measured as function of viscosity and pressure. Typical results are represented in Figs. 5 and 6. The linearity of these plots shows that Darcy’s law is valid for the system membrane/alcohol or acetone. The effect of the length of conditioning time was then checked. No variation of permeability was detected for a period of 30 days. Values for the permeability ratio (Yare given in Table 3. The IRIS 3069, 3042 and 3038 membranes keep the same permeability after changing from water to alcohols or acetone. However, the other membranes are more permeable in these solvents than in water. The increase in permeability could be the result of a better solvation of the polymeric chains, which would induce an expansion of the membrane network.
20°C:(0) IRIS 3038; (0) IRIS 3042.
Fig. 6. Ethanol flux versus temperature and viscosity: IRIS 3042 (AP = 0.4 atm).
(0) IRIS 3038
0.2 atm); (0)
Permeability ratios of alcohols and acetone for the tested membranes Membrane IRIS 3022
2. Ultrafiltration of water-immiscible solvents: aromatic and aliphatic hydrocarbons, chloroform Ultrafiltration data for water-immiscible solvents were generally recorded using IRIS 3042 and 3038 membranes because of the similarity of their transport properties in water and in the intermediary solvents (ethanol or acetone) chosen for the conditioning operations. Figure 7 shows the variation of the permeability ratio (Ywith the conditioning time. For the two membranes a strong decrease is observed during the first days. But, except for decane, constant permeabilities are obtained after the initial decreases. The same constant rate was reached more quickly when the conditioning operations were performed at higher temperatures (45°C). This observation shows that the solvent exchange in the membrane between acetone or ethanol and the water-immiscible solvent is governed by an activation energy. Conflicting results were obtained in the measurement of the initial ultrafiltration rates. Such discrepancies may be attributed to variations in the conditioning procedure (poor temperature or time controls). No systematic
III ” time
s! 21 r = .8 9
IRIS 3038 21d
1 10 I
/‘, time (day)
Fig. 7. Change in permeability ratio with conditioning time in several organic solvents for IRIS 3042 and 3038 membranes; temperature: 20°C; pressure = 0.4 atm. (A) n-decane; (0) n-heptane; (0) benzene; (X ) ethanol and methanol; (0) chloroform; d = days.
studies were undertaken during the present work because, after equilibrium, the observed permeabilities were the same within the range of experimental errors ( f 5%). The permeability ratios of benzene, toluene, heptane, decane and chloroform are listed in Table 4. Through the IRIS 3042 membrane, hydrocarbon permeabilities are higher than water permeability (CY> 1). The opposite occurs for the IRIS 3038. These variations reflect important structural changes, but it is surprising to find opposite trends for the two membranes, since they are prepared by association of the same anionic and cationic copolymers (Fig. 1).
245 TABLE 4 Permeability ratios of some water-immiscible and IRIS 3038
organic solvents for the membranes IRIS 3042
Viscosity at 20°C (cP)
01for the membrane IRIS 3042
01for the membrane IRIS 3038
The only difference between the two ultrafilters is the relative concentration in the two polyelectrolytes: the IRIS 3038 has equal numbers of negative and positive charges, while the IRIS 3042 has an excess in anionic groups. A test run series was carried out using the two membranes successively in hydrocarbon and in water. The evolution of the permeability ratio is shown in Fig. 8. While the IRIS 3042 tends to recover its initial water permeability after IRIS
Fig. 8. Change in permeability ratio (Yafter successive treatment in water (W) and hydrocarbon (S) for IRIS 3042 and 3038 membranes; temperature: 20°C; pressure = 0.4 atm. Each value of o was calculated by reference to the initial _I& value.
each treatment in toluene or hexane, the water and toluene permeability of the IRIS 3038 membrane exhibits a strong tendency to decrease when the number of tests increases. 3. Ultrafiltration of water-ethanol and methanol--acetone mixtures The permeability ratios of the IRIS 3042 were measured with waterethanol (Table 6) and methanol-acetone (Table 6) mixtures. The results show that the porous structure of this membrane is very stable in the studied binary mixtures. On the other hand, the permeability of the IRIS 3038 decreases sharply as a function of the volume of permeate passed. Typical results are shown in Fig. 9. This change is irreversible since after transferring the samples back to water, the permeability ratio remained at the same low value. 4. Ultrafiltration of 1,4-dioxane In spite of the total solubility of dioxane in water, the dioxane permeability of the IRIS membranes was very different from those found for alcohols or acetone. Figure 10 shows the effect of the conditioning time on permeability ratios for the IRIS 3042 and 3038. It is seen that the permeability decreases to reach a low limit value. This effect may be attributed to the high hygroscopy of the aprotic liquid: dioxane is able slowly to remove the hydration water from the membrane. As a consequence, an important structural change occurs. Other tests were carried out with addition of water to the dioxane. The membranes were initially equilibrated in dry dioxane; increasing amounts of water were then added to the dioxane feed. For each composition, the ultrafiltration rate was measured after the conditioning operation (one hour in the binary mixture). A stable value of the flux was immediately obtained. TABLE
and (Yfor some commercial membranes
1.0 1.0 1.25 1.0 1.4 1.4 1.4 1.4 0.9 0.2
Value of a for: methanol ethanol n-butanol acetone n-heptane n-decane benzene toluene chloroform dioxane
1.0 1.0 1.0 0.7 0.8 0.9 0.9 0.45 0.05
Permeability ratios (CY)of water-ethanol and methanol-acetone IRIS 3042. ~~ _Mixture composition: Valve of (Y Ethanol or acetone (vol. %) Water-ethanol mixtures 0
for the membrane
1.00 0.93 1.06 0.97 0.99
20 40 80 100
1.02 0.98 1.01 1.02
Fig. 9. Change in permeability ratio with permeate volume passed for two organic solvent mixtures with IRIS 3038 membrane; temperature: 20°C; pressure = 0.4 atm. (0) methanoiacetone (methanol, 80 vol.%); (A) ethanol-water (ethanol, 80 vol.%); (0) water.
Graphs of the permeability ratios as functions of the volume fraction of water in the dioxane solution are shown in Fig. 11. Only 4% water is needed to recover the initial permeability of the IRIS 3042. But the IRIS 3038 seems to have been irreversibly modified in dry dioxane since the Q:limit value does not exceed 0.5. The results illustrate the dependence of the permeability of the membranes
248 .5 -
I 10 time
Fig. 10. Change in permeability ratio with conditioning time in dry dioxane for IRIS 3042 and IRIS 3038 membranes; temperature: 20°C; pressure = 0.4 atm. (0) IRIS 3038; (0) IRIS 3042.
Fig. 11. Change in permeability ratio with addition of water in dioxane for IRIS 3042 and IRIS 3038 membranes initially conditioned in dry dioxane; temperature = 20°C; pressure = 0.4 atm. (0) IRIS 3038; (0) IRIS 3042.
on the hydration state of the polymeric chains. Halary et al.  have shown that during the annealing in hot water of cellulose diacetate reverse-osmosis membranes, water-polymer bonds were broken to form polymer-polymer interactions. As a result, the water content decreases and the membrane shrinks to give a less porous structure. It may be assumed that during flushing of the IRIS membranes with dry dioxane, the affinity between water and dioxane molecules is strong enough to remove the solvation water so that a modification of the chain configuration
occurs. With the presence of small quantities of water in the solvent, the polar sites of the membranes will be rehydrated and the polymer chains tend to return to their initial configuration (IRIS 3042) or to adopt a new arrangement more stable than the initial configuration (IRIS 3038).
5. Some comments about the stability of the IRIS 3038 and 3042 membranes in organic solvents The preceding results show that great differences in structure and stability can be induced simply by a change in relative concentration of the two polyelectrolytes associated to form the membrane. When cast from a mixture with the same number of anionic and cationic charges (IRIS 3038), the membrane was more permeable but less stable than the membrane obtained from a solution where the anionic polyelectrolyte was in excess (IRIS 3042). The effect of the annealing of the IRIS 3038 in hydrocarbons, dioxane or polar solvent mixtures was always an irreversible decrease in permeability. On the other hand, the same treatment of the IRIS 3042 only induced a reversible modification. To clarify the reasons for such a large difference, a study of the factors governing the formation of the membranes would be necessary. However the observations reported in this paper can supply a first insight into the observed phenomena. In the casting solution, the polymer chains are more solvated by the N,Ndimethylformamide than are the ionic sites. Thus the counter-ions interact strongly with the ionic charges of the polyelectrolytes. During the quenching in water, the solvation of the polymer chains decreases while the greater solvation of the counter-ions increases their mobility and allows an enhancement of the interactions between the charged groups of the chains. The counterions are rapidly released  and precipitation occurs. When the quantities of ionic groups of opposite charges are equivalent (IRIS 3038), the rate of precipitation will be high. This will lead to a porous structure since the random distribution of the chains in their extended conformation within the solution will be partially retained. Not all the fixed cationic groups are neutralized by the fixed anionic groups. The remaining free charged sites will be well solvated by water. During the annealing of the membrane in a liquid which has enough solvation power to increase the chain mobility to a sufficient extent, new interactions between the free charged sites will occur leading to a less hydrated, less porous but more stable structure. On the other hand, a deficiency in one of the two polyelectrolytes (IRIS 3042) will decrease the number of potential interactions between the charged groups and thus the precipitation rate will be lowered. The time prior to complete coagulation will be sufficient to allow a rearrangement of the chains to a stable thermodynamic state in which all the fixed anionic charges are neutralized by the cationic charges. So a further annealing will not induce any structural change. This analysis is in close agreement with the experimental results: (i) The water permeability coefficient of the IRIS 3038 was initially twice the water permeability coefficient of the IRIS 3042 (Table 2).
(ii) While the st ructure of the IRIS 3042 was not modified after treatment in dioxane or methanol/acetone mixtures, the water permeability of the IRIS 3038 decreased (Figs. 9 and 10) to reach the value observed for the other membrane (0 = 0.5). Qualitatively, the same trends were observed after treatment in ethanol-water mixtures (Fig. 9) and after successive treatments in toluene and water (Fig. 8). Conclusions Measurements of the water permeability of several commercial ultrafiltration membranes as a function of pressure and temperature have shown that the convective flux of water was proportional to the pressure and inversely proportional to the viscosity. Thus, the permeability coefficient of Darcy’s law:
Jv= L”p,q AP/V is an intrinsic characteristic of a given membrane. Furthermore, Darcy’s law can be also used to describe the permeability to organic solvents (permeability coefficient Lf&). For several binary membranesolvent systems, the two permeability coefficients Lbl, and Lg,v have the same value. For other systems, the value of the permeabihty ratio (Y: must be determined. Thus a membrane may be characterized by L”p,rl and the values of the coefficient 01(Table 5). Examination of the properties of two membranes (Rhsne-Poulenc IRIS 3038 and 3042) leads us to conclude that the mechanism of the formation of a membrane must be carefully studied in order to clarify the reasons for the structural changes which can occur in organic solvents. Acknowledgements Thanks are due to Delegation G&&ale h la Recherche Scientifique et Technique for supporting the investigation and to SociCte Rhbne-Poulenc providing ultrafiltration membranes. The authors wish to thank Mr. X. Marze for helpful discussions.
References P.A. Bailey, Ultrafiltration - The current state of the art, Filtr. Sep., 14 (1977) 213. D. Defives, R. Avrillon, C. Miniscloux, R. Roullet and X. Marie, Regeneration des huiles lubrifiantes usagges par ultrafiltration, Inf. Chim., 175 (1978) 127. J. Murkes, Semi-permeable membranes. A need for performance standards, Chem. Proc. Rep., (Jan) (1974) 67. E. Klein, F.F. Holland, A. Donnaud, A. Leboeuf and K. Eberle, Diffusive and hydraulic permeabilities of commercially available cellulosic hemodialysis films and hollow fibers, J. Membrane Sci., 2 (1977) 349.
251 5 X. Marze, S.U.C.R.P., FrenchPatent, 71 24235 (1973). 6 G.W. Pace, M.J. Schorin, M.C. Archer and D.J. Goldstein, The effect of temperature on the flux from a stirred ultrafiltration cell, Sep. Sci., 11 (1976) 65. 7 R.B. Grieves, D. Bhattacharaya, W.G. Schomp and J.L. Bewley, Membrane ultrafiltration of a non ionic surfactant, AIChEJ., 19 (1973) 766. 8 J.L. Halary, C. Noel and L. Monnerie, Processus physicochimiques intervenant dans la preparation des membranes de diacetate de cellulose. Recherche de critere pour l’optimisation des performances, Desalination, 13 (1973) 251. 9 A.S. Michaels, Polyelectroiyte complexes, Ind. Eng. Chem., 57 (1965) 32.