Adsorptive fouling of modified and unmodified commercial polymeric ultrafiltration membranes

Adsorptive fouling of modified and unmodified commercial polymeric ultrafiltration membranes

Journal of Membrane Science 160 (1999) 65±76 Adsorptive fouling of modi®ed and unmodi®ed commercial polymeric ultra®ltration membranes J. Lindau, A.-...

273KB Sizes 0 Downloads 29 Views

Journal of Membrane Science 160 (1999) 65±76

Adsorptive fouling of modi®ed and unmodi®ed commercial polymeric ultra®ltration membranes J. Lindau, A.-S. JoÈnsson* Department of Chemical Engineering 1, Lund University, PO Box 124, SE-221 00 Lund, Sweden Received 11 September 1998; received in revised form 6 January 1999; accepted 8 February 1999

Abstract The fouling tendency, due to adsorption on the pore walls, of two pairs of modi®ed and unmodi®ed ultra®ltration membranes, with similar observed retentions determined by dextran and gel permeation chromatography, was studied. The membranes investigated were made of modi®ed and unmodi®ed polyaramide (PA) and modi®ed and unmodi®ed polyvinylidene ¯uoride (PVDF). The PVDF membrane was surface-modi®ed and the PA membrane was made from a modi®ed polymer solution. Membrane modi®cation was used to reduce fouling by adsorption. Octanoic acid was used as the fouling substance, representing a large number of small, hydrophobic compounds. It is demonstrated in this investigation that membrane modi®cation is not always successful. It was determined that at lower concentrations of octanoic acid, the modi®ed PA membrane exhibits a smaller fouling tendency than the unmodi®ed PA membrane, while the result is reversed for concentrations above 60% of the saturation concentration. The fouling tendency of the unmodi®ed PVDF membrane is much lower than that of the modi®ed PVDF membrane at all concentrations. The cross-sections of the membranes were visually examined with scanning electron microscopy, but no difference could be observed between the modi®ed and unmodi®ed membranes. The membranes were also examined with Fourier transform infrared spectroscopy. The spectra of the two PA membranes were different, while no difference was observed for the unmodi®ed and surface-modi®ed PVDF membranes. Remains of octanoic acid were found in the membranes, although they had been thoroughly rinsed with deionized water and the initial pure water ¯ux was recovered. # 1999 Elsevier Science B.V. All rights reserved. Keywords: Fouling; Polyaramide; Polyvinylidene ¯uoride; Modi®ed membranes

1. Introduction A major limiting factor in pressure-driven membrane processes is the ¯ux decline due to fouling. The causes of fouling are very disparate, and vary depending on the nature of the membrane and solute used, as well as on the operating parameters, but can generally be divided into the build-up of a cake layer on the *Corresponding author. Fax: +46-46-222-4526.

membrane surface, blocking of the pores, and adsorption on the membrane surface or the pore walls. Much effort is being devoted in an effort to understand the mechanisms of fouling and to modify membranes so as to be less susceptible to fouling. However, there is still a considerable lack of knowledge concerning the fouling tendencies of modi®ed membranes. Protein fouling is a major problem in, for example, the food and biotechnology industries and has therefore been studied extensively. Several investigations concerning the fouling tendency of modi®ed and

0376-7388/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 9 9 ) 0 0 0 7 6 - 9

66

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

unmodi®ed membranes using proteins as fouling substance have been published [1±10]. For proteins, it is often considered that there are two main mechanisms of irreversible fouling: pore narrowing, as a result of protein adsorption and pore plugging. Brink et al. [6], among others, have studied the anti-fouling action of pre-adsorbed polymers on ultra®ltration (UF) membranes of different cut-off and on micro®ltration (MF) membranes. They concluded that the reduced protein fouling of the UF membrane was a result of the fact that the pre-adsorbed polymer sterically prevented the proteins from entering the pores and thereby adsorbing on the pore walls. This effect could not be seen in the MF membranes, as their pores were too large to be blocked by the polymer. Kim and Fane [7] investigated surface hydrophilized UF membranes using aqueous proteins. They found that, in general, surface modi®cation enhanced the ¯ux when the modi®ed membranes were compared with their untreated counterparts. They also found that hydrophilic membranes were not necessarily easier to clean. The modi®ed membranes did, however, exhibit a slower ¯ux decline than the unmodi®ed ones. Some investigations have also been performed on the fouling tendency of modi®ed and unmodi®ed membranes using substances other than proteins, for example by Vigo et al. [11], who used exhausted cutting oil emulsions to test a modi®ed polyvinylidene ¯uoride (PVDF) membrane. Modi®ed and unmodi®ed UF and nano®ltration membranes have also been studied in the removal of undesirable substances from the clear ®ltrate in a paper machine water circulation system [12]. The membranes were modi®ed with long-chained, negatively charged polyelectrolytes. Modi®cation with dextran sulphate (DEXSU) gave greater reductions of lignin residuals and chemical oxygen demand (COD) and also increased the ¯ux, because of the favourable electrostatic repulsive conditions achieved. Dal-Cin et al. [13,14] tested the performance of a number of different membrane materials in treating a pulp mill ef¯uent containing 432 mg/l of fatty acid and resins. Among the membrane materials tested were modi®ed and unmodi®ed PVDF from different manufacturers. Adsorption tests with no applied pressure were performed and the water ¯uxes before and after adsorption were compared. It appeared that the modi®ed membrane was slightly less susceptible to fouling by adsorption, but there was a

considerable variation in the results from the different membrane samples. It is well known and accepted that hydrophilic membranes have a smaller fouling tendency than hydrophobic ones [15±21]. Changing of the hydrophobicity of the membrane by including a hydrophilic copolymer could thus make the membrane less susceptible to fouling. However, when Hamza et al. [22] modi®ed polyethersulphone membranes by using macromolecules as additives, the membrane surface became more hydrophobic, not more hydrophilic. They found that the modi®ed membranes had higher ¯uxes when treating oil/water emulsions. Their results indicate that formation of the gel layer was reduced by the hydrophobic modi®cation of the membrane. In this investigation, the fouling tendency of two pairs of modi®ed and unmodi®ed UF membranes was studied, during treatment of a solution containing a small hydrophobic solute. The membranes investigated were made of PVDF and polyaramide (PA) and modi®ed in two different ways. The PVDF membrane was modi®ed by chemical binding (grafting) of a cellulosic polymer onto the surface of a pre-formed support membrane of PVDF [23], while the PA membrane was modi®ed to become more hydrophilic by blending the polymer solution with a hydrophilic copolymer before the membrane was cast. The PVDF membrane was thus surface-modi®ed whereas the internal membrane matrix of the PA membrane was altered. Most previous fouling studies on modi®ed membranes have been concerned with large molecules such as proteins, oil emulsions or very complex liquids like pulp mill ef¯uents. In this investigation, a fatty acid, octanoic acid, was used as a model substance, representing a large number of small, hydrophobic compounds which can be found in, for example, fermentation broth, oily waste water, pulp and paper mill ef¯uents and liquid food. 2. Equipment and methods 2.1. Membranes and equipment The experiments were performed in a cross-¯ow module equipped with a circular, ¯at membrane with an area of 0.00196 m2. The permeate ¯ow was continuously measured with a PhaseSep ¯ow meter. All the

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76 Table 1 Membranes used in the investigation (NMWCO is that given by the membrane manufacturers) Membrane

Material

Manufacturer

NMWCO (Da)

PA20 PA20H FS81 Etna20A

PA Modified PA PVDF Modified PVDF

Hoechst Hoechst Dow Denmark Dow Denmark

20 000 20 000 6 000 20 000

instruments, ¯ow meter, pressure meters and thermocouple, were connected to a computer which allowed continuous registration of ¯ux, pressure and temperature. The membranes investigated are listed in Table 1. The reason for choosing the FS81 membrane which has a lower nominal molecular weight cut-off (NMWCO) than the other membranes was that the observed retentions of dextran by the FS81 and Etna20A membranes were about the same. Since the ¯ux reduction due to adsorption is affected by the pore size [24±26], it is important to compare membranes with similar retentions. The cross-sections of the membranes were visually examined with a scanning electron microscope, Jeol JSM-840A. 2.2. Chemicals and analyses Octanoic acid of pro analysi grade, supplied by Merck, was used as the model substance. The molecular mass of octanoic acid (caprylic acid, C8H16O2) is 144 Da. Its solubility in water at 258C is 4.7 mM [27] or 0.7 g/l [28]. The pH of a 4.7 mM octanoic acid solution is about 3.6. The amount of octanoic acid in the feed solution during the ¯ux reduction experiments was determined as the COD, using Dr. Lange LCK 214 cuvettes and a pocket ®lter photometer. The relation between the octanoic acid concentration, C (g/l) and the COD (mg/l), is Cˆ4.110ÿ4 COD. The concentration of octanoic acid is represented in the ®gures as the relative concentration, C/Csat, where Csat is the saturation concentration of octanoic acid at 258C, i.e. 0.7 g/l. The dextran fractions used for retention measurements were T10, T40 and T70 obtained from Pharmacia Biotechnology, Sweden. Permeate and retentate samples were analysed by gel permeation chromatography using the gel Sepharose CL-6B, obtained from

67

Pharmacia LKB Biotechnology, Sweden. The column used was made of glass and was jacketed to maintain a constant temperature. HPLC equipment from Shimadzu (LC-6A, SIL-6A and SCL-6A) was used to pump the eluent at a ¯ow rate of 0.5 ml/min and to inject the sample. A refractometer from Erma (FRC7510) was used for detection. The analyses were carried out at 188C using deionized water as eluent containing 1 mg/l NaN3, to prevent bacterial growth. 2.3. Performance of experiments 2.3.1. Flux reduction The in¯uence of the feed concentration of the octanoic acid on the ¯ux was investigated. All the experiments were performed at 258C, 100 kPa and a retentate recirculation rate of 95 l/h, corresponding to an average velocity across the membrane of 1.5 m/s. Both the retentate and permeate were recirculated to the feed tank. Before the in¯uence of octanoic acid on the ¯ux was studied, the membranes were conditioned for a total period of four days, at the same temperature and pressure as those used during the experiment. This was done for all four membranes investigated to rinse out the preservatives and to obtain a stable pure water ¯ux (PWF). After three days of conditioning, the observed dextran retention was determined. After the retention determination, the dextran was rinsed out with 54 l of lukewarm, deionized water and fresh water was circulated through the module overnight. The PWF was taken as the ¯ux the following morning. The octanoic acid was added in small doses every 30 min and the ¯ux continuously registered. The amount of acid added each time was dependent on the ¯ux and was adjusted to obtain as smooth a ¯ux reduction as possible. Before each new addition, a sample of the feed solution was withdrawn for analysis. When the ¯ux was zero, close to zero or the solubility limit of octanoic acid was reached, the octanoic acid was rinsed out with 54 l of lukewarm water and fresh water was circulated through the system throughout the night. 2.3.2. Dextran retention The retention was determined using the equipment described above. The test solution was prepared by dissolving 1 g/l of a mixture of 40 wt% T10, 35 wt%

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

68

T40 and 25 wt% T70 in deionized water. Sodium azide, NaN3, was added to prevent bacterial growth (1 mg/l). It is important to minimize the concentration polarization when studying the retention of, for example, dextran [29,30]; therefore, a dilute solution and as low a ¯ux as possible, with maintained circulation velocity, were used. The test conditions were 258C and a pressure corresponding to a ¯ux of 40 l/m2 h. The cross-¯ow velocity over the membrane was the same as in the octanoic acid experiments. Both the retentate and the permeate were recirculated to the feed tank. The dextran solution was ultra®ltered for 2 h before the samples were collected. A standard curve was obtained using pure dextran fractions of T10, T40 and T70. The molecular masses of the dextran molecules were found to be proportional to their residence time inside the column, which is in agreement with the results obtained by Andersson et al. [31].

The used samples had been subjected to octanoic acid and then rinsed with deionized water, according to the procedure described above. Membranes to be analysed were completely dried at room temperature in a desiccator and examined directly in the spectrometer. The FTIR spectra were recorded on a Bruker IFS 66 spectrometer. The FTIR measurements were performed in two ways. With the thin, transparent FS81, PA20 and PA20H membranes, the top layer of the membranes was peeled off and the transmission of these samples was measured. The opaque Etna20A membrane is too thick for transmission measurements and instead overhead attenuated total re¯ection (ATR) was used for the analysis of this membrane. The difference between these two types of measurements is that during transmission measurements the whole sample is analysed, whereas when using ATR±FTIR only the surface of the sample is analysed.

2.4. FTIR analysis

3.1. Modified and unmodified PVDF membranes

Fourier transform infrared spectroscopy (FTIR) analysis was carried out on new and used membrane samples. The new membrane samples were rinsed with deionized water for 24 h before the analysis.

3.1.1. Influence of the octanoic acid concentration on the flux The in¯uence of the octanoic acid concentration on the ¯ux of the PVDF membranes is shown in Fig. 1 as

3. Results

Fig. 1. Relative flux vs. concentration of octanoic acid of the FS81 and Etna20A membranes.

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

relative ¯ux vs. relative concentration of octanoic acid in the feed solution. Upon increasing the octanoic acid concentration, the ¯ux of the unmodi®ed PVDF membrane, FS81, initially decreased but then started to increase to reach a maximum ¯ux plateau at C/Csat extending from 0.55 to 0.8. When the concentration was further increased, the ¯ux again decreased but the relative ¯ux was still 0.8 at C/Csatˆ0.95, as can be seen in Fig. 1. The relative ¯ux of the modi®ed PVDF membrane, Etna20A, decreased slowly until C/Csatˆ0.6, and then started to decrease more rapidly. The ¯ux never

69

became zero, even though an emulsion was formed at concentrations above the saturation concentration. After rinsing with deionized water, the PWF of the Etna20A membrane was slightly higher than the PWF before exposure to octanoic acid, as shown in Fig. 2(a). The PWF of the FS81 membrane after rinsing was signi®cantly higher than the initial PWF, Fig. 2(b). It can be seen in Fig. 2 that the ¯ux sometimes increased after a sudden drop, but this cannot be seen in Fig. 1. The reason for this is that the ¯ux sometimes dropped rapidly directly after the addition of octanoic acid, and then stabilized. Since

Fig. 2. Flux of the (a) Etna20A (b) FS81 membrane during stepwise addition of octanoic acid and ensuing rinsing with water.

70

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

Fig. 3. Observed retention of the Etna20A and FS81 membranes.

the samples were withdrawn 30 min after the addition of octanoic acid, the ¯uxes were stable and this phenomenon is therefore not observed in Fig. 1. 3.1.2. Observed dextran retention The observed retentions of a 1000 ppm mixture of 40 wt% T10, 35 wt% T40 and 25 wt% T70 were determined for the membranes investigated. The observed retentions of the Etna20A and the FS81 membranes are quite similar, although their NMWCOs are different, Fig. 3.

3.2. Modified and unmodified PA membranes 3.2.1. Influence of the octanoic acid concentration on the flux For the PA membranes it was found that at lower concentrations of octanoic acid, below C/Csat0.6, the reduction in ¯ux of the modi®ed membrane, PA20H, was less than for the unmodi®ed, PA20, membrane. However, at concentrations above C/Csat0.6, the PA20H membrane rapidly lost its ¯ux while the ¯ux decline for the PA20 membrane was slower, as shown

Fig. 4. Relative flux vs. concentration of octanoic acid of the PA20 and PA20H membranes.

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

in Fig. 4. The ¯ux of the PA20 membrane did not approach zero until it was in the vicinity of the saturation concentration, whereas the corresponding low ¯ux of the PA20H membrane was reached at C/ Csat0.75. After rinsing the membranes with deionized water, the PWF of the PA20 membrane was almost totally regained, as can be seen in Fig. 5(a). The PWF of the PA20H membrane could not be regained by simply rinsing with water, Fig. 5(b).

71

3.2.2. Observed dextran retention A slight difference was found in the observed retention of the two PA membranes with the same NMWCO, see Fig. 6. 3.3. SEM images The cross-sections of the four membranes investigated were examined with scanning electron microscopy (SEM). No visual difference could be seen

Fig. 5. Flux of the (a) PA20 (b) PA20H membrane during stepwise addition of octanoic acid and ensuing rinsing with water.

72

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

Fig. 6. The observed retention of a mixture of T10, T40 and T70 for the unmodified PA20 membrane and the modified PA20H membrane.

between the modi®ed and unmodi®ed membranes in either pair. 3.4. FTIR There was a signi®cant difference between the spectra of the unmodi®ed PA20 and modi®ed

PA20H membranes, see Fig. 7. For the modi®ed PA membrane, PA20H, an absorbance of O±H at 3446 cmÿ1 was observed. Both membranes show N±H stretching modes at 3300 cmÿ1 and C±H stretching vibrations at 3025 cmÿ1. Both membranes exhibited absorbance at 2925 and 2950 cmÿ1, although the absorbance was more pronounced for the PA20H

Fig. 7. FTIR spectra of new PA20 and PA20H membranes.

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

73

Fig. 8. FTIR spectra of a used and a new PA20H membrane.

membrane. The PA20 membrane spectrum has a small peak at 2860 cmÿ1 while the PA20H membrane spectrum has one absorbance peak at 1420 cmÿ1 and another at 1290 cmÿ1. This difference in spectra is in accordance with the fact that the PA20H membrane is made from a different polymer blend than the PA20 membrane. The only signi®cant difference between new and used PA20H membranes is the C=O stretching band at 1720 cmÿ1, originating from remains of octanoic acid in the used membrane, see Fig. 8. No signi®cant difference could be observed between the spectra of the FS81 and Etna20A membranes (Fig. 9) which means that the modifying layer on the Etna20A membrane is too thin to be detected. This was also found to be the case by Stengaard [23]. As for the PA20H membrane, the only signi®cant difference between new and used FS81 membranes was the C=O stretching band at 1720 cmÿ1, see Fig. 10. 4. Discussion In this investigation, it was found that the ¯ux reduction of the modi®ed PVDF membrane was more pronounced than for the unmodi®ed membrane. The surface modi®cation of the PVDF membrane thus had no positive effect on the fouling behaviour when

treating small hydrophobic solutes. The reason for this is that the octanoic acid molecule is not retained by the membrane. Instead, it passes into the membrane pores where it can be adsorbed on the pore walls. In a case such as this, a modi®ed layer on the surface of the membrane does not have the effect of steric hindrance suggested by Brinck et al. [6]. Regarding the PA membranes, it was found that the modi®ed membrane had a lower fouling tendency at lower octanoic acid concentrations than the unmodi®ed membrane, but the ¯ux deteriorated very rapidly at concentrations above 60% of the saturation concentration of the acid. This ¯ux reduction behaviour indicates that the pore size distribution of the modi®ed PA membrane is narrower than that of the unmodi®ed PA membrane. The result of the PA membranes underlines the importance of testing the fouling tendency of a membrane at different concentrations of the fouling substance. A membrane with only a small ¯ux reduction at low concentration can exhibit a severe loss of ¯ux at higher concentrations due to, for example, a narrower pore size distribution. 5. Conclusions The number of commercially available modi®ed membranes is steadily increasing. The aim of mem-

74

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

Fig. 9. FTIR spectrum of (a) an FS81 membrane; (b) an Etna20A membrane.

brane modi®cation is, in most cases, to obtain a lowprotein-binding membrane by rendering the membrane more hydrophilic. However, modi®cation of a UF membrane does not necessarily lead to a better fouling behaviour when treating a complex solution, containing not only protein but also fatty acid, for example. In this investigation, the fouling of unmodi®ed and modi®ed membranes when treating a fatty acid solution was compared. The membranes were modi®ed by

including a hydrophilic copolymer and by surface modi®cation. A signi®cant difference was found between the FTIR spectra from an unmodi®ed PA membrane and a PA membrane that had been modi®ed by the addition of a copolymer to the blend, whereas no signi®cant difference was observed between the spectra from an unmodi®ed PVDF membrane and a surface-modi®ed PVDF membrane. The fouling of the modi®ed PVDF membrane was more pronounced than the fouling of the unmodi®ed

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76

75

Fig. 10. FTIR spectra of new and used FS81 membranes.

membrane. The ¯ux reduction of the modi®ed PA membrane was lower than that of the unmodi®ed membrane at concentrations below 60% of the saturation concentration, but more pronounced at higher concentrations. It is thus important to use different concentrations of the fouling substance when investigating the fouling tendency of a membrane to avoid erroneous conclusions. The pure water ¯ux of all membranes, except the modi®ed PA membrane, was recovered by simply rinsing the membrane with deionized water. Although the pure water ¯ux was recovered, remains of fatty acid were detected in the rinsed membranes on using FTIR analysis. Acknowledgements The authors wish to thank Helen DeÂrand at the Department of Chemical Engineering II, for her help with the FTIR analyses. References [1] K.J. Kim, A.G. Fane, C.J.D. Fell, The performance of ultrafiltration membranes pretreated by polymers, Desalination 70 (1988) 229±249.

[2] M. NystroÈm, Fouling of unmodified and modified polysulfone ultrafiltration membranes by ovalbumin, J. Membr. Sci. 44 (1989) 183±196. [3] K.J. Kim, A.G. Fane, C.J.D. Fell, The effect of Langmuirblodgett layer pretreatment on the performance of ultrafiltration membranes, J. Membr. Sci. 43 (1989) 187±204. [4] M. NystroÈm, M. Laatikainen, K. Turku, P. JaÈrvinen, Resistance to fouling accomplished by modification of ultrafiltration membranes, Progr. Colloid Polym. Sci. 82 (1990) 321±329. [5] M. NystroÈm, P. JaÈrvinen, Modification of polysulfone ultrafiltration membranes with UV irradiation and hydrophilicity increasing agents, J. Membr. Sci. 60 (1991) 275± 296. [6] L.E.S. Brink, S.J.G. Elbers, T. Robbertsen, P. Both, The antifouling action of polymers preadsorbed on ultrafiltration and microfiltration membranes, J. Membr. Sci. 76 (1993) 281±291. [7] K.J. Kim, A.G. Fane, Performance evaluation of surface hydrophilized novel ultrafiltration membranes using aqueous proteins, J. Membr. Sci. 99 (1995) 149±162. [8] M. Ulbricht, H. Matuschewski, A. Oechel, H.G. Hicke, Photo-induced graft polymerization surface modifications for the preparation of hydrophilic and low-protein-adsorbing ultrafiltration membranes, J. Membr. Sci. 115 (1996) 31±47. [9] A. Nabe, E. Staude, G. Belfort, Surface modification of polysulfone ultrafiltration membranes and fouling by BSA solutions, J. Membr. Sci. 133 (1997) 57±72. [10] N. Ehsani, S. Parkkinen, M. NystroÈm, Fractionation of natural and model egg-white protein solutions with modified and unmodified polysulfone membranes, J. Membr. Sci. 123 (1997) 105±119. [11] F. Vigo, G. Capannelli, C. Uliana, S. Munari, Modified polyvinylidene fluoride membranes suitable for ultrafiltration

76

[12]

[13]

[14]

[15] [16] [17] [18] [19] [20] [21]

J. Lindau, A.-S. JoÈnsson / Journal of Membrane Science 160 (1999) 65±76 purpose applications, La Chimica e L'industria 64(2) (1982) 74±79. J. Nuortila-Jokinen, S. Luque, L. Kaipia, M. NystroÈm, The effect of ultra- and nanofiltration on the removal of disturbing substances in the paper machine water circulation system, BHR Group Conf. Ser. Publ. 3 (1993) 203±213. M.M. Dal-Cin, F. McLellan, C.N. Striez, C.M. Tam, T.A. Tweddle, A. Kumar, Membrane performance with a pulp mill effluent: Relative contributions of fouling mechanisms, J. Membr. Sci. 120 (1996) 273±285. M.M. Dal-Cin, C.N. Striez, T.A. Tweddle, C.E. Capes, F. McLellan, H. Buisson, Effect of adsorptive fouling on membrane performance: Case study with a pulp mill effluent, Desalination 101 (1995) 155±167. A.-S. JoÈnsson, B. JoÈnsson, The influence of nonionic and ionic surfactants on hydrophobic and hydrophilic ultrafiltration membranes, J. Membr. Sci. 56 (1991) 49±76. A.-S. JoÈnsson, Concentration polarization and fouling during ultrafiltration of colloidal suspensions and hydrophobic solutes, Sep. Sci. Technol. 30(2) (1995) 301±312. C. JoÈnsson, A.-S. JoÈnsson, Influence of the membrane material on the adsorptive fouling of ultrafiltration membranes, J. Membr. Sci. 108 (1995) 79±87. A.-S. JoÈnsson, Fouling during ultrafiltration of a low molecular weight hydrophobic solute, Sep. Sci. Technol. 33(4) (1998) 503±516. P. Mourot, M. Oliver, Comparative evaluation of ultrafiltration membranes for purification of synthetic peptides, Sep. Sci. Technol. 24 (5±6) (1989) 353±367. M.K. Ko, J.J. Pellegrino, Determination of osmotic pressure and fouling resistances and their effects on performance of ultrafiltration membranes, J. Membr. Sci. 74 (1992) 141±147. D. Doulia, V. Gekas, G. TraÈgaÊrdh, Interaction behaviour in ultrafiltration of nonionic surfactants. Part I. Flux behaviour, J. Membr. Sci. 69 (1992) 251±258.

[22] A. Hamza, V.A. Pham, T. Matsuura, J.P. Santerre, Development of membranes with low surface energy to reduce the fouling in ultrafiltration applications, J. Membr. Sci. 131 (1997) 217±227. [23] F.F. Stengaard, Characteristics and performance of new types of ultrafiltration membranes with chemically modified surfaces, Desalination 70 (1988) 207±224. [24] J. Lindau, A.-S. JoÈnsson, R. Wimmerstedt, The influence of a low-molecular hydrophobic solute on the flux of polysulphone ultrafiltration membranes with different cut-off, J. Membr. Sci. 106 (1995) 9±16. [25] A.-S. JoÈnsson, J. Lindau, R. Wimmerstedt, J. Brinck, B. JoÈnsson, Influence of the concentration of a low-molecular organic solute on the flux reduction of a polyethersulphone ultrafiltration membrane, J. Membr. Sci. 135 (1997) 117±128. [26] J. Lindau, A.-S. JoÈnsson, A. Bottino, Flux reduction of ultrafiltration membranes due to adsorption of a lowmolecular-weight hydrophobic solute: correlation between flux behaviour and pore size, J. Membr. Sci. 149 (1998) 11± 20. [27] G. Solomons, Fundamentals of Organic Chemistry, 2nd ed., Wiley, Singapore, 1986, p. 666. [28] A.W. Ralston, C.W. Hoerr, The solubilities of the normal saturated fatty acids, J. Org. Chem. 7 (1942) 546±555. [29] R. Nobrega, H. de Balmann, P. Aimar, V. Sanchez, Transfer of dextran through ultrafiltration membranes: a study of rejection data analysed by gel permeation chromatography, J. Membr. Sci. 45 (1989) 17±36. [30] H. de Balmann, R. Nobrega, The deformation of dextran molecules. Causes and consequences in ultrafiltration, J. Membr. Sci. 40 (1989) 311±327. [31] M. Andersson, A. Axelsson, G. Zacchi, Determination of the pore-size distribution in gels, Bioseparation 5 (1995) 65±72.