Streaming Potential Measurements As a Characterization Method for Nanoﬁltration Membranes

Home , Streaming current, Surface charge

Colloids and Surfaces A: Physicochemical and Engineering Aspects 150 (1999) 247–259

Streaming potential measurements as a characterization method for nanoﬁltration membranes

J.M.M. Peeters, M.H.V. Mulder *, H. Strathmann

Uni6ersity of Twente, Faculty of Chemical Technology, P.O. Box 217, 7500 AE, Enschede, The Netherlands

Received 17 June 1998; accepted 19 October 1998

Abstract

The streaming potentials of two different nanoﬁltration membranes were studied with several electrolyte solutions to investigate the inﬂuence of salt type and concentration on the zeta potential and kinetic surface charge density of the membranes. The zeta potentials decreased with increasing salt concentration, whereas the kinetic surface charge densities increased. The kinetic surface charge densities could be described by Freundlich isotherms, except in one case, indicating that the membranes had a negligible surface charge. The kinetic surface charge density observed was caused by adsorbed anions. Salt retention measurements showed different mechanisms for salt separation for the two investigated membranes. One membrane showed a salt retention that could be explained by a Donnan exclusion type of separation mechanism, whereas for the other membrane the salt rejection seemed to be a combination of size and Donnan excluion. Comparing the results obtained by the streaming potential measurements with those of the retention measurements, it could be concluded that the membrane with the highest kinetic surface charge density showed the Donnan exclusion type of separation, whereas the membrane with the lower surface charge density showed a separation mechanism that was not totally determined by Donnan exclusion, size effects seemed to play a role as well. © 1999 Elsevier Science B.V. All rights reserved.

Keywords: Nanoﬁltration; Salt retention; Streaming potential

1. Introduction responsible for the separation, whereas concerning the separation of charged species, e.g. ions, The separation mechanism of nanoﬁltration both size and charge effects may play a role [3–5]. membranes is assumed to be based on a combina- The charge effects that are partly causing the tion of several processes, like size exclusion, rejection of ions are resulting from interactions charge exclusion and, as sometimes referred to as, between the charged particles and the charged dielectric exclusion [1,2]. In the case of uncharged membrane. These electrochemical interactions be- molecules, mainly size exclusion is thought to be tween membrane and ions can be investigated in different ways. Separation experiments have been * Corresponding author. Tel.: +31-53-4892950; fax: +31- performed, varying parameters like salt concen- 53-4894611; e-mail: [email protected]. tration and membrane ﬂux, resulting in a value

0927-7757/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S0927-7757(98)00828-0 248 J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259

Fig. 1. Schematic drawing of potential decrease as a function of the distance from the surface in an electrolyte solution. for the membrane charge [6–8]. Membrane poten- ions near the membrane will be higher than that tial measurements have been carried out as well to in the bulk of the solution. The charges at and determine the membrane charge density [6,9,10]. adjacent to the surface will cause a potential Furthermore, electrical impedance measurements difference between the region near the surface and have been carried out giving an indication for the the bulk of the solution. The potential decreases resistance of electrolyte transport through the within the solution as a function of the distance membrane [11]. from the charged surface, as shown in Fig. 1. In this paper we will use electrokinetic measure- Three different potentials are shown in this figure, ments to investigate the membrane charge of two the surface potential, 0, the potential at the nanofiltration membranes. Streaming potential Stern plane, d, and the electrokinetic or zeta measurements have been carried out to determine potential, n. All these potentials are defined with the zeta potential and the kinetic surface charge respect to the potential at infinite distance from density of the membranes. To investigate the infl- the surface. Although the surface potential is an uence of the type of the electrolyte and of its important parameter, the potential at the Stern concentration, different types of salts at different plane is practically of more importance. This concentrations were used for the streaming poten- Stern plane is the interface between the fixed part, tial measurements. Finally, the results of the i.e. a layer of immobile ions near the charged streaming potential measurements will be com- surface, and the diffusive (mobile) part of the pared to the results of retention measurements electrical double layer. The potential at the Stern carried out with the same electrolyte solutions as plane is the actual potential influencing the behav- used in the retention measurements. ior of the charged species. However, as this potential cannot be measured directly, the electrokinetic potential is often considered as an adequate sub- 2. Theory stitute. The plane at which the zeta potential is located should be outside the Stern layer, and 2.1. Streaming potential represents the potential at the surface of shear between surface and solution where there is rela- Ions in an electrolyte solution that is brought in tive motion between them. The position of the contact with a charged surface will not be dis- surface of shear is close to, and from practical tributed randomly. The concentration of counter- reasons assumed to be identical to, the Stern J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259 249 mnDP R surface [12]. The zeta potential is shown in Fig. 1 DE = exp (2) str pu R as well. The zeta potential can be determined by 0 th electrokinetic measurements, like streaming poten- In the latter case the streaming potential is not tial and electro-osmosis measurements. dependent on the geometry of the capillary. A streaming potential is the potential difference In the case of a low potential, the relation at zero current caused by the convective flow of between the surface charge density at the hydrody- charge due to a pressure gradient through a namic shear plane and the zeta potential is given charged capillary, plug, diaphragm or membrane. in the following equation: A streaming potential is generated by exerting a mn force on the double layer that has been built up in | d = (3) the solution near the charged surface. Since an s−1 excess of counter charges is present, movement of where k-1 is the Debije length. The Debije length those counter charges causes a current. This cur- can be calculated by Eq. (4): rent which is streaming through the double layer is called the streaming current. The accumulation of ' mkT s−1 = 2 (4) counter charges downstream generates a streaming 4e NAI potential across the capillary which on its turn 2 o causes a conduction current through the capillary with I=0.5 z i ci: where is the permittivity of in the reverse direction. In steady state, the stream- the medium, k the Boltzmann constant, T the ing current equals the conduction current. temperature, e the elementary charge, NA the Measurement of a streaming potential can Avogadro number, I the ionic strength, zi the provide both the zeta potential, n, and the net valency and ci the concentration of species i. charge density, | d, at the hydrodynamic shear Although Eq. (4) was derived for symmetrical plane. The streaming potential is defined to be electrolytes, like 1-1 and 2-2 electrolytes, we will positive if the higher potential is at the high use it for 1-2 and 2-1 electrolytes as well. pressure side. Concerning the membrane charge density, a The relation between the zeta potential, n, the distinction can be made between the actual charge D of the membrane, which originates from mem- streaming potential, Estr, and the pressure difference applied, DP, is given by [12,13]: brane fixed charges, like carboxylic or sulphonic groups, and adsorbed charges. mnDP DE (1) Benavente et al. [14] introduced a method to str = u 2 s discriminate between these two types of charges. p u + 0 r Therefore, it was assumed that three components where m is the permittivity of the medium, p the added to the total charge of the system. First, the S0 viscosity, u the bulk conductivity, u the surface fixed charges at the membrane surface, , second, 0 s Ss conductivity, and r the radius of the pore or the charges of the Stern-layer, , and finally, the capillary or half the width of a slit. charges within the diffusive part of the electric Sd It is difficult to measure directly the surface double layer, . The charge of the Stern-layer is conductivity and therefore, at low electrolyte con- caused by the adsorption of ions. As electroneu- centrations, where the surface conductivity is not trality is assumed, the total charge equals zero. be negligible, Eq. (1) cannot easily be applied. This problem can be circumvented by measuring the %0 +%s +%d =0 (5) actual resistance of the electrolyte solution across the slit or pore, Rexp, and comparing this value with Assuming that the diameter of the pore or the the resistance that can be calculated from experi- width of the slit through which the solution is ments at high concentrations, where the surface streamed is much larger than the Stern-layer thick- s conductivity can be neglected, Rth [13]. Then, Eq. ness, the relation between the charge densities, , (1) can be written as: of each of the layer can be written as: 250 J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259

| 0 +| s +| d =0 (6) and because all charges are spread over an equal area. RT DG =− (12) Both co- and counter-ions can adsorb at ads b charged membrane surfaces due to electrostatic and non-electrostatic interactions. As anions in 2.2. Donnan exclusion the vicinity of non-polar surfaces are less hydrated than cations, they can adsorb more closely to the Electrostatic interaction between ions and surface, resulting often in an excess of negative charges at the membrane surface is one of the charges in the layer nearest to the surface and, mechanisms of nanofiltration membranes retain- therefore, in an apparent negative surface charge ing ions. The repulsion of the ions from the [15,16]. surface can be described by Donnan exclusion. The total molar Gibbs free energy of adsorp- If a charged membrane is put in contact with tion is the sum of an electrostatic and a chemical an ionic solution, ions with the same charge sign D (specific) part, Gc: as that of the membrane (the co-ions) are ex- DG z n DG cluded and cannot pass the membrane, whereas ads = − + c (7) RT z RT the ions with the opposite charge sign as that of + the membrane (counter-ions) are able to pass the The charge of the Stern-layer is caused by the membrane in principle [17]. adsorbed ions. One of the adsorption types by Because of the charge of the membrane, a which the total amount of charge is related to the concentration difference of the ions between the concentration of ions is the Freundlich adsorption solution and the membrane is built up, leading to isotherm, describing adsorption on non-uniform an osmotic pressure difference between the mem- sites. This empirical isotherm describes the rela- brane and the solution. As thermodynamic equi- tion between between the charge density and the librium is assumed, an additional potential across fraction of anions by a power law [14]: the membrane, the Donnan potential, EDon, will | s b compensate this osmotic pressure difference. For (x−)=ax− (8) ideal solutions, the Donnan potential is given by: with a and b being constants, and RT c cw M E = ln i,m (13) x = − H2O Don z F c − z−c[w M +w M −(w −w )M ] i i + + − − + − H2O (9) where R, T, z and F having their usual meaning and with c and c the concentration of species i the fraction of anions in solution. In this equation i i,m in the bulk and the membrane, respec- c is the electrolyte concentration, w the stoichio- tively.Combining the requirements of identical metric number, M the molecular weight, and z chemical potentials in membrane and bulk and of the density, while the subscripts, −, +, and H O 2 electroneutrality, for a 1-1 electrolyte the relation refer to the anion, the cation and water, between concentration of the co-ion in the bulk, respectively. its concentration in the membrane and the fixed The total amount of charge within the mem- charges at the membrane surface, c , can be brane, adsorbed layer and diffusive layer is then: x,m written as [15]: | d | 0 b = +ax− (10) 2 2 cco−ion,m ×cx,m +(cco−ion,m) =(cco−ion) (14) From this equation the number of adsorption sites, N, and the free Gibbs energy of adsorption The Donnan potential is dependent on several factors, e.g.: Gads can be determined according to [14]: * salt concentration; y a sin b * fixed charge concentration in the membrane; N= y (11) z−e b * valence of the co-ion; J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259 251

Fig. 2. Schematic illustration of the streaming potential system.

* valence of the counter-ion. sity of Wageningen (the Netherlands). The mem- With increasing salt concentration (and increas- brane samples are glued upon glass plates ing ionic strength) and decreasing ﬁxed membrane (microscope slides). The dimensions of the mem- charge, the concentration of the co-ion in the brane samples are 76×26 mm. membrane increases, as shown in Eq. (13). This The cell consists of two plexiglass parts contain- decrease of co-ion exclusion from the membrane ing sample channels that hold the glass slides with leads to a lower rejection of the salt, because the the membranes glued on top. Clamps hold both rejection of the co-ion determines the rejection of cell parts together. Under the glass slides silicon the salt. rubber sheets can be put to prevent leakage at the A higher valence of the co-ion and a lower interface between cell and glass slide. The cell valence of the counter-ion will cause an increase parts are separated by a 200 mm thick Teﬂon sheet of the Donnan potential and, therefore, an in- which serves two functions: spacing between sam- crease in retention. ple plates and leak prevention. Platina black elec- trodes are inserted into the chambers at both ends of the cell. 3. Experimental The driving pressure can be applied in either direction and consequently the electrolyte solution 3.1. Streaming potential measurements can pass through the channel. The pressure was varied in the range of 0 to 0.25 bar and was The set-up to determine streaming potentials monitored with an accuracy of 0.01 bar. Stream- was almost identical to that described by Van ing potential measurements were repeated at least Wagenen and Andrade [18]. With this kind of six times by measuring at decreasing pressure. set-up, streaming potentials can be measured The streaming potential was measured using a along a surface, rather than through a surface. In digital multimeter (Simpson, Model 464, Simpson literature streaming potential measurements Electric, Elgin), which had an internal impedance through membranes or other porous species are of 10 GV. described as well [19–23]. In those papers, mainly The electrolyte solutions used were in the range the features of the pores are then determined. of 0.01–10−5 M for the three salts used: NaCl,

Fig. 2 shows a schematic drawing of the cell CaCl2,and Na2SO4 (Merck, PA).The streaming which is used to support the membrane samples potentials were measured from low to high con- for the streaming potential measurements. The centration. For every salt, fresh membranes were cell was manufactured by the Agricultural Univer- used. 252 J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259

Fig. 3. Streaming potential as a function of the applied pressure for different concentrations of NaCl. Membrane: NF45.

Electrolyte conductance was measured by a from pure water to a salt solution. After a stabi- conductivity meter (Microprocessor Conductivity lization time of 2 h the concentration of the feed Meter LF 537, WTW). The pH of the electrolyte was measured and subsequently permeation sam- bulk solutions was measured by a pH meter (691 ples were taken. Salt retention measurements were pH Meter, MetrOhm). Demineralized water started with the lowest concentration. The se-

filtered by a Milli-Q-Plus unit (conductivity 60 nS quence of the salts used was CaCl2, NaCl, and −1 cm ) was used as solvent. The temperature of Na2SO4, respectively. In between retention mea- the system was 20–23°C. The membranes used surements with different salts, the membrane was were two commercially available nanofiltration flushed for at least 2 h with pure water. membranes: ASP35 (Advanced Membrane Tech- nology) and NF45 (Dow-FilmTec). 4. Results and discussion 3.2. Retention measurements 4.1. Streaming potential measurements Retention measurements were carried out with

CaCl2, NaCl, and Na2SO4 (Merck, PA), respec- The streaming potentials of both membranes tively, at three different initial concentrations, decreased while increasing the concentration of 0.001, 0.005, and 0.01 M. The solvent was dem- the electrolyte concentration for all types of elec- ineralized water filtered by a Milli-Q-Plus unit. trolytes. In Fig. 3 this is illustrated using NaCl as The retention measurements were carried out in a electrolyte solution for the NF45 membrane. stirred dead-end filtration set-up at a pressure of 5 The use of different electrolytes of identical bar. The flux through the membrane was auto- ionic strength resulted in different streaming po- matically recorded. tentials and, therefore, in different zeta-potentials. First, all membranes were permeated with wa- The Figs. 4 and 5 show the zeta potentials for ter for 4 h. Then, the feed solution was changed both membranes, as calculated from the stream- J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259 253

Fig. 4. Zeta-potentials for a ASP35 membrane measured with different electrolyte solutions at various concentrations.

| d b ing potentials according to Eq. (2). For the =ax− (15) ASP35 membrane, the zeta potentials determined for the sulphate solution were lower than where the actual charge density of the mem- those for the sodium and calcium chloride solu- brane is neglected and the charge density at the tions. The values for both chloride solutions are Stern-layer equals the charge density at the comparable. For the NF45 membrane, the shear plane. Using a log–log plot with the | d highest zeta-potentials were determined with the parameters and x- at the axes, a straight line Na2SO4 solutions. The zeta-potentials deter- should result with a slope equal to the parame- mined with NaCl were lower, whereas those for ter b, which is indeed the case. As can be seen

CaCl2 were the lowest. Especially in the case of from both figures the differences between the the lower electrolyte concentrations, the ASP35 two lines according to Eqs. (9) and (14) are very membrane showed much higher zeta potentials small, indicating that the actual charge densities for the chloride solutions compared to the NF45 of the two membranes were small as well. The membrane. charge densities at the membrane surface calcu- The kinetic surface charge densities calculated lated by Eq. (9) were in the range of −4× from the streaming potential measurements by 10−5 –2×10−4 Cm−2, which are indeed small Eq. (3) are shown in Figs. 6 and 7. In these compared to the values of the calculated total figures the kinetic surface charge densities of charge density as can be seen from the Figs. 6 NF45 and ASP35 membranes are shown as and 7. functions of the anion fractions in the NaCl and Comparing the ASP35 and the NF45 mem- the CaCl2 solutions to discriminate between the brane, the kinetic surface charge densities of the actual charge of the membrane and the ad- ASP35 membrane calculated from the streaming sorbed charges. The dashed lines within these potential measurements with the chloride solu- figures represent Eq. (9), assuming adsorption tions are higher than those of the NF45 mem- according to the Freundlich isotherm. The solid brane. From Table 1 it can be seen that lines within the figures represent the equation: especially the amounts of adsorption sites, calcu- 254 J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259

Fig. 5. Zeta-potentials for a NF45 membrane measured with different electrolyte solutions at various concentrations.

lated by Eq. (10), differed for both membranes, The results of the data ﬁtting using the Fre- while the adsorption free energies were all in the undlich isotherm are shown in Table 1. same order of magnitude. In the case of the NF45 membrane, the ﬁtting

In the case of the ASP35 membrane the of the Na2SO4 streaming potential data with the ﬁtting of the kinetic surface charge density ver- Freundlich isotherm gave no reliable results. sus the anion fraction resulted in comparable values for the amounts of adsorption sites and 4.2. Retention measurements only slight differences between the free energies of adsorption for both chloride solutions. The Retention measurements carried out with the calculated parameters N and Gads obtained for NF45 and ASP35 membranes showed a different the NaCl and CaCl2 solution in the case of the behaviour for both. For both membranes the NF45 membrane, were comparable as well, al- retention for Na2SO4 is the highest compared to though in this case the free energies of adsorp- that of the other salts (see Figs. 8 and 9), but in tion were identical, whereas the calculated the case of the ASP35 membrane the retention amounts of adsorption sites showed small differ- of NaCl is higher than that of CaCl2, whereas ences. For the former membrane, the calculated for the NF45 membrane it is inversed. amount of adsorption sites was different for the Most probably, the differences in rejection sulphate ions compared to that of the chloride characteristics between the two membranes are ions. For both membranes it seems to be rea- caused by the differences in the determining sep- sonable to assume that the main source for aration mechanisms. In the case of the ASP35 charge at the membrane surface was due to an- membrane the retention results can be inter- ion adsorption, whereas cation adsorption was preted by the Donnan exclusion mechanism, negligible. The same kind of results was ob- based on interactions between the charges of the served for a polysulfone membrane by Be- membrane and those of the ions. The retention navente et al. [14] and for polyamide experiments for the NF45 membrane cannot be membranes by Hernandez-Gimenez et al. [24]. explained exclusively by Donnan exclusion, and J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259 255

Fig. 6. Charge density of the ASP35 membrane as a function of the anion fraction with NaCl and CaCl2 solutions.

most probably size effects play a role for salt tion based on size exclusion only, calcium would separation with this membrane as well. have shown the highest retention. Possibly, the The ASP 35 membrane shows a retention se- retention behavior of the NF45 membrane is de- \ \ quence of R (Na2SO4) R (NaCl) R (CaCl2), termined by a combination of both charge and so the retention for the bivalent anion is the size exclusion. largest, whereas that of the bivalent cation is the smallest. The retention of the 1-1 salt (NaCl) is 4.3. Comparison between streaming potential and in between the other two. If it is assumed now, retention measurements that the ASP35 membrane is negatively charged, the high retention for the sodium sulphate (biva- Comparing the results obtained from the lent co-ion, monovalent counter-ion) and the streaming potential measurements with sodium low(er) retention for calcium chloride (bivalent and calcium chloride for the NF45 and the counter-ion, monovalent co-ion) ﬁt well with the ASP35 membrane, it can be concluded that the Donnan exclusion model. The decrease of the actual surface charge for both membranes is retention with increasing concentration can also only small. However, the adsorbed charge dif- be explained by assuming a decreasing inﬂuence fers largely when comparing the two membranes of Donnan exclusion. and is much higher in the case of the ASP35 The results obtained for the NF45 membrane membrane. The difference in adsorbed charge may not be explained by Donnan exclusion, be- may be the reason for the discrepancy between cause both the retention for the bivalent cation the ion separation characteristics of both mem- and that of the bivalent anion are high. Neither branes. The adsorbed charge at the ASP35 can the retention sequence exclusively be ex- membrane is so large that the membrane be- plained by differences in the size of the different haves in the retention measurements like a nega- ions. Although the sizes of a hydrated sulphate tively charged membrane. As the charge and calcium ion are larger than those of sodium densities for both the sodium and calcium chlo- and chloride, the calcium ion is in turn larger ride solutions are comparable, the difference in than the sulphate [25]. So, in the case of separa- retention seems to be caused by the difference in 256 J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259

Fig. 7. Charge density of the NF45 membrane as a function of the anion fraction determined with NaCl and CaCl2 solutions. valency of the counter-ions for these two salts. 5. Conclusions Because the Donnan-exclusion becomes smaller if the valency of the counter-ion becomes Both streaming and zeta potentials decreased higher, this can explain the higher retention for with increasing salt concentration. The zeta po- the sodium chloride solution. tentials and the calculated charge densities are With a sodium sulphate solution, the charge higher for the ASP35 membrane than for the density of the ASP35 membrane is somewhat NF45 membrane. The actual charge density of smaller than with sodium and calcium chloride both membranes is small and especially at solutions, although the retention of the sodium higher electrolyte concentrations, it is negligible sulphate solution is higher than that of both compared to the adsorbed charges. Anion adsorption seems to be the main source for the chloride solutions. Most probably, this higher apparent membrane charge, whereas cation ad- retention is caused by an increase in repulsion sorption is small. In most cases anion adsorp- due to the valency of the co-ion changing from tion could be described with a Freundlich mono- to bi-valent. adsorption isotherm, only sodium sulphate ad- In the case of the NF45 membrane, the adsorption on the NF45 membrane could not be sorbed charge is not large enough to determine described by this type of isotherm. the salt separation completely. The retention for Retention measurements showed a different calcium chloride is higher than that of sodium salt separation behaviour for the ASP35 and the chloride, which is most probably caused by a NF45 membrane. For the ASP35 membrane the difference in size between the calcium and differences in retention between sodium sul- sodium ion. In the case of a sodium sulphate phate, sodium chloride and calcium chloride solution the charge density is highest, resulting seemed to be caused by Donnan exclusion, in the highest retention for this salt as well. This whereas for the NF45 size effects seems to play may be caused by a combination of larger elec- a role as well, especially when the retentions for trostatic repulsion in the case of the sulphate calcium and sodium chloride are compared. ion and by size exclusion. Most probably, the ASP35 membrane shows J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259 257

Table 1 Parameters a and b, according to Eq. (15), the number of adsorption sites, N, and the free Gibbs energy of adsorption, according to Eqs. (11) and (12), respectively, for the two nanoﬁltration membranes

−2 D −1 Membrane: ASP35 a (C m ) b (−) N (−) Gads (kJ mol )

NaCl 0.36 0.41 1.7×1018 −6.1 18 CaCl2 0.32 0.36 1.6×10 −6.8 17 Na2SO4 0.27 0.41 6.3×10 −6.0 Membrane: NF45 NaCl 0.07 60.37 3.8×1017 −6.7 17 CaCl2 0.08 80.37 4.4×10 −6.7

this Donnan type of retention behavior because Acknowledgements of its high kinetic surface charge. The adsorbed charge at the NF45 membrane is not high The authors would like to thank the Eu- enough to discriminate on the basis of charge ropean Union (Science Programme) for the between sodium and calcium chloride. There- ﬁnancial support for this project. Furthermore, fore, size exclusion seems to determine the the authors would like to thank Maikel van difference in separation between these two Bree, Jeroen van Beckum, and Edwin van der salts. Steeg for their experimental work.

Fig. 8. Retention vs. concentration of various salts for a ASP35 membrane. Pressure: 5 bar. 258 J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259

Fig. 9. Retention vs. concentration of various salts for a NF45 membrane. Pressure: 5 bar.

References of charged cellulosic membranes, J. Membr. Sci. 18 (1984) 285. [1] A. Yaroshchuk, E. Staude, Charged membranes for low [10] R. Takagi, M. Nakagaki, Membrane charge of microp- pressure reverse osmosis properties and applications, orous glass membrane determined by the membrane po- Deslination 79 (1992) 1–15. tential method and its pore size dependency, J. Membr. [2] A.G. Fane, A.R. Awang, M. Bolko, et al., Metal recov- Sci. 111 (1996) 19. ery from waste water using membranes, Water. Sci. [11] J. Benavente, G. Jonsson, Transport of Na2SO4 and Technol. 25 (1992) 5. MgSO4 solutions through a composite membrane, J. [3] S. Alami-Younssi, A. Larbot, M. Persin, J. Sarrazin, L. Membr. Sci. 80 (1993) 275. Cot, Rejection of mineral salts on a gamma alumina [12] H. Lyklema, Fundamentals of Interface and Colloid nanofiltration membrane: application to environmental Science: Part II Solid-Fluid Interfaces, Academic Press, process, J. Membr. Sci. 102 (1995) 123. London, 1996. [4] T. Tsuru, X.L. Wang, S.I. Nakao, S. Kimura, Trans- [13] R.J. Hunter, Zeta Potential in Colloid Science, Aca- port of neutral and charged solutes through nanofiltra- demic Press, London, 1981. tion membranes, in: The International Symposium on [14] J. Benavente, A. Hernańdez, G. Jonsson, Proper and Fiber Science and Technology, Yokohama, Japan, 1994. adsorbed charges on the surfaces of the polysulfonic [5] X.L. Wang, T. Tsuru, M. Togoh, S.I. Nakao, S. support of a composite membrane from electrokinetic Kimura, Transport of organic electrolytes with electro- phenomena, J. Membr. Sci. 38 (1993) 211. static and steric-hindrance effects through nanofiltration [15] C. Werner, H.J. Jacobasch, G. Reichelt, Surface membranes, J. Chem. Eng. Jpn. 28 (1995) 372. charachterization of hemodialysis membranes based on [6] T. Tsuru, S.I. Nakao, S. Kimura, Effective charge den- streaming potential measurements, J. Biomater. Sci. sity and pore structure of charged ultrafiltration mem- Polym. Ed. 7 (1995) 61. branes, J. Chem. Eng. Jpn. 23 (1990) 604. [16] M. Elimelech, W.H. Chen, J.J. Waypa, Measuring the [7] I. Jitsuhara, S. Kimura, Rejection of inorganic salts by zeta (electrokinetic) potential of reverse osmosis mem- charged ultrafiltration membranes made of sulfonated branes by a streaming potential analyzer, Desalination polysulfone, J. Chem. Eng. Jpn. 16 (1983) 394. 95 (1994) 269. [8] X.L. Wang, T. Tsuru, M. Togoh, S.I. Nakao, S. [17] M. Mulder, Basic Principles of Membrane Technology, Kimura, Evaluation of pore structure and electrical Kluwer Academic, Dordrecht, 1996. properties of nanofiltration membranes, J. Chem. Eng. [18] R.A. Van Wagenen, J.D. Andrade, Flat plate streaming Jpn. 28 (1995) 186. potential investigations: hydrodynamics and electroki- [9] Y. Kimura, H.J. Lim, T. Iijima, Membrane potentials netic theory, J. Colloid Interface Sci. 76 (1980) 305. J.M.M. Peeters et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 150 (1999) 247–259 259

[19] M. Nystro¨m, M. Lindstro¨m, E. Matthiasson, Streaming ing potentials of clean and fouled ultrafiltration mem- potential as a tool in the characterization of ultrafiltra- branes, J. Membr. Sci. 88 (1994) 211. tion membranes, Colloids Surf. 36 (1989) 297. [23] N. Kaneko, Y. Yamamoto, Membrane potential in re- [20] M. Nystro¨m, A. Pihlajama¨ki, N. Ehsani, Characteriza- verse osmosis, J. Chem. Eng. Jpn. 9 (1976) 158. tion of ultrafiltration membranes by simultaneous [24] A. Hernańdez-Gimenez, L. Martıńez, M.V. Gomez, streaming potential measurements and flux measure- Concentration dependence of the adsorbed charges and ments, J. Membr. Sci. 87 (1994) 245. the zeta potentials on the walls of neutral and charged [21] A. Hernańdez, F. Martıńez, A. Martin, P. Pra´danos, microporous membranes, J. Colloid Interface Sci. 158 Porous structure and surface charge density on the (1993) 429. walls of microporous alumina membranes, J. Colloid [25] A.L. Horva´th, Handbook of Aqueous Electrolyte Solu- Interface Sci. 173 (1995) 284. tions, Ellis Horwood, Chichester, 1985. [22] C. Causserand, M. Nysto¨m, P. Aimar, Study of stream-