Journal of Membrane Science, 76 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA (1993) 61-71 61 zyxwvuts Elsevier Science Publishers B.V., Amsterdam Porosity and protein adsorption of four polymeric microfiltration membranes Kenneth M. Perssona, Gustav0 Capannellib, Aldo Bottinob and Gun Trligardh” “Department of Food Engineering, University of Lund, P.O. Box 124, S-221 00 Lund (Sweden) bInstitute of Industrial Chemistry, University of Genoa, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED COFSO Europa 30, I-16132 Genoa (Italy) (Received M arch 4,1992; accepted in revised form September 9,1992) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQP Abstract Four different microfiltration (M F) membranes were analysed with respect to porosity and protein adsorption. Liquid displacement porosity (LDP) and adsorption of “C-labelled /3-lactoglobulin were used as characterization methods. LDP showed a clear pore size distribution for all membranes analysed, with maximally 10% of the pores having at least 90% of the permeability. The protein adsorption came to equilibrium within 30 minutes, giving decreased permeabilities for all membranes. The permeability loss could not be modelled as a simple membrane pore restriction, but the protein load corresponded to a monolayer of protein adsorbed on and in the membranes. Keywords: porous and microporous membranes; porosimetry; liquid displacement porosity; protein adsorption - Introduction Microfiltration is an important concentration/separation unit operation with applications in for example protein purification, cell harvesting, cold sterilization and water purification. The crossflow mode should in theory diminish the membrane surface clogging tendency, which would mean better performance during microfiltration, but in practice some problems occur, such as concentration-polarization and inevitable adsorption. An examination of the physical properties of the memCorrespondence to: K.M . Persson, Department of Food Engineering, University of Lund, P.O. Box 124, S-221 00 Lund (Sweden). 0376-7388/93/$06.00 brane might increase the knowledge about these problems and could help improve crossflow microfiltration performance. The porosity of the membranes and the protein adsorption can be used for characterizing the different membrane materials, although results from dead-end conditions not automatically can be translated to crossflow. Theory Liquid displucementporosity If a membrane (or any porous material) is wetted with a liquid, and then another liquid, which is immiscible with the first one, is forced by pressure through the membrane, the result- 0 1993 Elsevier Science Publishers B.V. All rights reserved. 62 zyxwvut K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 ing flux is due to transport only through those pores that are larger than the pore sizes according to Cantor’s equation: pr=2 6cosql (1) Because the liquid wets the membrane, the contact angle, Q, can be assumed to be zero, and Cantor’s equation can be simplified into the following: pr=26 (la) At the pressure pi = 2 6/ri, the permeable pores have a radius r > ri, and the flow is given by Poiseuille’s law: k=l (2) It should be remembered, that this law holds for laminar flow and cylindrical pores. Most symmetric MF membranes exhibit a tortuous pore structure (see Figs. 1 and 2 ) . Although the cross-sectional radius varies along the way through the membrane, there is always some minimal radius which limits the transport (Fig. 3). This limiting pore size is measured through liquid displacement porosity. The total area of the permeable pores with r>ri iS Ai= i rnkrg (3) k=l The permeability is Ji/pi. From eqns. (la) and ( 2 ) , the unit area and pore number distribution for the membrane can be obtained [ 1,2]. When comparing three different pore size determining methods for polysulfone MF membranes (Poiseuille-Knudsen, mercury intrusion and liquid displacement porosity), Bottino et al. found only the liquid displacement porosity applicable for proper characterization [ 3 1. Protein fouling studies Originally highly porous microfiltration mem- Fig. 1. A scanning electron micrograph of the surface of a PVDF hydrophilic membrane. Magnification 10.000 X . Photo: Helgot Olsson. K.M. Persson et al/J. Membrane Sci. 76 (1993) 61-71 63 Fig. 2. A scanning electron micrograph of the surface of Nylon-66 membrane. Magnification 10,000 X . Photo: Helgot Olsson. Fig. 3. The tortuous paths through a microfiltration membrane. PVDF-hydrophilic, 500 and 2500 times magnification. Photo: Helgot Olsson. branes were used in the application of sterile filtration. The membranes were intended for simple uses only. Total sterility of the products and low cost membranes were the two most important requirements. When the crossflow mode of operation seemed to work in ultrafiltration applications, it was natural to test it for MF. But it did not function very well, since the membranes became fouled almost instantly, and the high water permeability could not be recovered, even after cleaning. An investigation of the flat sheet MF-membranes, their protein adsorption isotherms, and their poros- ities could provide some useful information about this problem. Protein adsorption isotherms for some ultrafiltration (UF ) membranes [ 41, and some MF membranes [ 5,6] have been reported. For UFmembranes it was shown that at protein concentrations about 5%, the protein interacted with the membrane surface and with adsorbed protein layers, possibly giving multilayer effects. Matthiasson [4] described it as a surface attachment process. Bowen and Gan [ 51 studied BSA adsorption on polyvinylidene fluoride zy 64 K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 zyxwvutsr membranes, and found that the adsorption was Liquid displacement porosity rapid initially, reaching equilibrium after 30 min to 3 hr. The adsorption isotherms indiA computer driven LD-porosimeter made at cated two different adsorption sites, one of high the Institute of Industrial Chemistry in Genoa affinity and the other of low affinity. At pH 7, (Capannelli et al. [ 1 zyxwvutsrqponmlkjihgfedcbaZYXW ] was used for the porosity they did not find any adsorption at all, which study. The membranes were soaked for 16 hr in partly may have been due to their calculations. the water phase obtained from demixing at They based their protein load on a membrane 20 oC of a water/n-pentanol mixture prior to be surface area measured by multipoint nitrogen placed into the porosimetric cell. The wetting BET adsorption. The surface area obtained by liquid was displaced from the membrane pores the BET-method was larger than the available with the alcohol phase, by a HPLC-pump. The adsorption area for proteins, since protein molresults were analysed according to the theory ecules were by several orders of magnitudes discussed above. Some membranes were pholarger than the nitrogen, and could not diffuse tographed with a scanning electron microscope. as easily. Persson and Nilsson [6] measured the Protein adsorption isotherms amount of whey protein deposited indirectly, as deposit resistance after static protein exposure. They found that the deposit resistance j&lactoglobulin, crystallized and lyophilized depended on the membrane pore size: the larger once, obtained from Sigma Chemicals, was lathe nominal pore size, the less the deposit rebelled with 14C-formaldehyde through reducsistance. Comparing Nylon-66 membranes with tive alkylation (Dottavio-Martin and Ravel cellulose acetate under equal static conditions [5] ). 14C-formaldehyde, 25 ~1 of 400 PCi, was indicated that the former had approximately a added to 3 ml 50%-protein solution in 0.04 M 10 times larger resistance than the latter. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHG potassium phosphate buffer pH 7.0. The mixture was incubated at 25 ’ C for 4 hr and at 4” C for 12 hr. During the incubation, the mixture Materials and methods zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA was stirred. After the incubation, the solution was dialysed for 48 hr, in order to remove all M embranes free formaldehyde. The labelled protein was mixed with unlabelled protein in the ratio 1: 19 in a 0.005 M Four different MF membranes in flat sheet phosphate buffer solution at pH 7. The protein configurations were studied: solution was placed in cylindrical tubes, which - Pall Ultipore Nylon-66 NR, 0.2 pm. were directly positioned over the tested mem- Millipore GSVP cellulose acetate (CA), 0.22 branes in a multi-batch module (described by pm. Nilsson [ 81). The membranes were exposed to - Millipore GVWP (poly (vinylidene fluoride ) the protein mixture for 15 and 30 min and 1 and (PVDF), hydrophilic, 0.22 ,um. 2 hr, respectively, after which the membranes - Millipore GVHP poly (vinylidene fluoride ) were rinsed twice, first in buffer solution, then (PVDF), hydrophobic, 0.22 ,um. in double distilled water. The rinsing lasted 10 All pore sizes were nominal. The PVDF memmin, to ensure that no protein solution rebranes were made of the same material, but the mained inside the membranes. The amount of surface of the hydrophilic one was, according protein absorbed was measured as disintegrato the manufacturer, modified. K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED 65 tions per minute (dpm) from the labelled proteins directly on the membranes in a LKBTechnology 1214 Rackbeta scintillation detector. An approximate level of error of about 5% less dpm with protein adsorbed onto membranes, compared to protein in free solution was reported by Nilsson [ 81. A check-up of a dilution series with different proportions of labelled/non-labelled protein displayed a linear relation between the amount of labelled protein in solution and the amount of deposited protein on a Nylon membrane under constant total protein concentration, on the basis of which it was concluded that no specificity in the interactions had to be considered. Also, a control with j$lactoglobulin adsorption on a well characterized chromium surface, (kindly supplied by Thomas Arnebrant and Marie Wahlgren, Dept. of Food Technology, Lund, Sweden) was made, giving an adsorption of 1.2 mg/m’ chromium surface after 2 hr exposure to 1% protein solution, to be compared with the maximump-lactoglobulin adsorption of 1.3 mg/ m2 measured with ellipsometry [ 91. mean pure water permeability for each membrane. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ Results and discussion The porosities of the membranes expressed as cumulative relative pore numbers, relative pore areas and relative permeabilities are shown in Figs. 4-7. From these results it is apparent that the CA membrane has the sharpest pore distribution. Comparing the porosity results shown in Figs. 5-7, one can see that the Nylon66 has a number of pores of smaller size. The contribution of these pores to the membrane permeability is negligible, although they are relevant when pore area is considered. The zyxwvutsr Permeabilities The permeabilities of the fouled membranes were measured as pure dead-end water flux at 20°C (293 K) and 95 kPa transmembrane pressure before and immediately after 15 min, 30 min, 1 hr and 2 hr exposure to fi-lactoglobulin in the same type of module as described above for adsorption measurements, but with the modification that the permeate could be drained off through small channels under the membranes. In this case, the protein was not labelled. The water permeabilities were measured as pure water flux at 20°C after 30 min at three different pressures (20 kPa, 37 kPa and 95 kPa) , and were calculated as the mean of the values for the three cases. The relative permeabilities were simply calculated as the mean fouled membrane permeability divided by the RADI U S r. nnl Ii Fig. 4. Cumulative pore number ( * ), pore area (0 ) and permeability ( + ) for a cellulose acetate membrane. I lo2 10 RADI U S 3 c 10 nm 4 3 Fig. 5. Cumulative pore number ( * ), pore area (0 ) and permeability ( -I- ) for a PVDF-hydrophilic membrane. K.M . Persson et al.jJ. M embrane Sci. 76 (1993) 61- 71 TABLE 1 Maximum and minimum pore radii according to liquid displacement porosity and maximum pore radii according to bubble point measurements Membrane 10 RADIUS 3 c 10 nm 4 3 Fig. 6. Cumulative pore number ( * ), pore area ( 0 ) and permeability ( + ) for a PVDF hydrophobic membrane. CA PVDF phi1 PVDF phob Nylon-66 Liquid displacement porosity Bubble point max. radius (pm) min. radius (pm) max. radius’ (pm) 0.49 1.10 1.26 1.26 0.09 0.15 0.19 0.10 no value 0.72 0.80 0.52 “Values according to Gekas and Hallstrijm, 1990 [lo]. TABLE 2 Membrane permeabilities at 20°C RADIUS c nm Membrane Permeability (m/Pa-secX 10’) CA PVDF phi1 PVDF phob Nylon-66 29.1 f3.1 17.5 I!I1.4 21.5k2.5 34.0 + 3.9 zyxwvutsrqponmlkjihgfedcba 3 Fig. 7. Cumulative pore number ( * ), pore area ( 0 ) and permeability ( + ) for a Nylon-66 membrane. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 12001 ” permeability of each of the membranes, and especially of the Nylon-66 membrane seems mainly to be controlled by the pores of larger size. The porosities and permeabilities correspond acceptably to the maximum bubble point pore [lo], reported in the last column of Table 1. Pure water flux is highest for the Nylon-66 membrane, which also has the largest permeability according to our measurements, see Table 2. Hydrophilic PVDF has the lowest pure water flux, lower than hydrophobic PVDF; this membrane has probably smaller pores due to the surface modification. Protein adsorption isotherms are shown in Figs. 8-11. In all cases, the margin of error is estimated at 5 20%, with 5 percent-units of the II 0 20 40 60 80 100 120 Fig. 8. The protein adsorption isotherms for the cellulose acetatemembrane case. Proteinconc.: (m) O.l%, (+ ) l%, (x) 10%. error due to variations in the membrane [ 81, 5% due to the method and 10% as a result of experimental variations. The large amounts of material adsorbed at the different protein concentrations investigated could be attributed to K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIH 67 600, according to its agent in Sweden. The amounts given in Figs. 8-11 are expressed as material per unit membrane area. A closely packed layer of fi-lactoglobulin corresponds to 2.7 mg/m2 assuming a monomer molecular weight of 18,000 with a diameter of 3.6 nm and I hexagonally packed [91. This would yield a toI tal membrane surface for the investigated ma0 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 100 120 20 40 60 60 terials of at least 30-400 m2 per m2 of memFig. 9. The protein adsorption isotherms for the PVDF hybrane, assuming close-packed monolayers, see drophilic membrane. Protein cont.: (m) O.l% , (+ ) l% , Table 3. (x) 10% . The adsorption isotherms appear to be stable after a 30 min exposure to protein solutions. 1200X -X j&Lactoglobulin displayed the highest adsorpPVDF hydrophobic tion on the cellulose acetate and the hydrophobic PVDF membrane. The lowest protein adsorption was detected for the hydrophilic PVDF membrane. Since cellulose acetate is generally regarded as the most hydrophilic material of the four mentioned above, the high value was prob0 20 40 60 60 120 ably due to its large internal area. The protein thelmin load results can be interpreted as indicating that Fig. 10. The protein adsorption isotherms for the PVDF for cellulose acetate, hydrophobic PVDF and hydrophobicmembrane.Proteinconc.: (m) O.l% , (+) l% , Nylon-66 membranes, a monolayer is built up (x) 10% . on the membrane surface, but that for hydrophilic PVDF much less than a monolayer is formed. The lower adsorption for protein on hydrophilic PVDF may partly be caused by the low porosity (small inner surface) of this type 6 of membrane. In contrast, the hydrophobic PVDF membrane, which is basically the same membrane as the hydrophilic one, but unmoPVDF hydrophilic 0 20 40 60 80 100 120 TABLE 3 Fig. 11. The protein adsorption isotherms for the Nylon66membrane.Proteinconc.: (m) O.l% , (+) l% , (x) 10% . zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGF Estimated minimum total internal membrane surface, based on an assumption of a closed-packed monolayer the fact that proteins not only adsorbed onto the surfaces but also diffused into the porous structure. Bowen and Gan [5] reported a specific surface of 387 (in m2 total internal surface/m2 planar surface) for Millipore 0.22 pm hydrophylic PVDF membrane. The Pall 0.2 pm Nylon-66 membrane has a specific surface of M embrane CA PVDF phi1 PVDF phob Nylon-66 Total adsorption (mg/m”) M inimum internal surface (m2/m2) 1149 a4 1157 743 426 31 429 275 K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 zyxwvutsr 68 dified, displays a high protein adsorption, probably due to a pronounced hydrophobic interaction between protein and the hydrophobic surface. The difference in membrane hydrophility between the two PVDF membrane types was described with contact angles by Gekas and Hallstrom [lo]. Obviously enough, hydrophilization of the PVDF membrane has genuinely changed its surface properties. The importance of the membrane’s inner surface can be seen in terms of the adsorption results. The SEM micrograph of the hydrophilic PVDF membrane (Fig. 1) displays a surface of relatively low porosity compared with the Nylon-66 membrane (Fig. 2); a small net area for deposition, as in the case of the PVDF, means a lower degree of net protein deposition. The adsorption isotherms clearly display some similarities for the different membrane materials (Fig. 12 ). With exposure at low concentration, only modest adsorption of protein occurs, while at 10% protein, higher levels are reached. Permeabilities after protein exposure are shown in Figs. 13 and 14, expressed as relative permeability, the quotient of pure water permeability after exposure and before. Maximum and minimum values vary within 10% of al g 0.4 . . . . ..___________....................................................................................... f OJ 0 a 40 60 time 80 Fig. 13. The relative water permeability position for the cellulose acetate and hydrophilic cont.: ) l% , (A ) after protein membrane PVDF membrane (m) O.l% , (+ 100 (dashed lines). Protein I ‘$ . . . . . . . . . . . . . . . ..___.................................................................................... o.4 - t “I 0 2-O 40 60 so lb0 li0 time Fig. 14. The relative water permeability and Nylon-66 membrane zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA O.l% , (+) 3 de- (solid lines) 10% . after protein position for the hydrophobic PVDF membrane “.L I 120 l% , (A ) de- (solid lines) (dashed lines). Protein cont.: (n ) 10% . at 1 2.0 the mean permeabilities reported in the figures. A monolayer of /3-lactoglobulin would rez 3 2.4duce the permeability by 3-4%, assuming an E m 2.2even distribution in all pores and using a sim_o zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA ple pore restriction model for the permeability reduction. The much higher reductions that were found could either be explained by nonvalidity of the pore restriction model, or by postulating a small number of larger aggregates c o nc e ntra tio n being present in the protein solutions. If these Fig. 12. The logarithmic relationship between amount of aggregates clog some of the larger pores, the toadsorbed protein and concentration for the four membrane tal permeability reduces drastically. This is exmaterials after 2 hr of exposure. ( n ) CA, ( + ) PVDF Phil, emplified in Fig. 15, in which the theoretical ( x ) PVDF phob, ( 0 ) Nylon-66. 1 2.6 zyxwvutsrq K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 69 zyxwvutsr permeability for a hydrophobic PVDF memreported by Persson and Nilsson [ 61, is probbrane is shown. The even distribution of one ably due to their choice of fouling mixture. They protein bilayer or of two (hypothetical) bilayused a whey protein concentrate, in which large ers does not reduce permeability significantly, aggregates were likely to be present. Here, the whereas a general reduction in the number of blocking of the large pores of the Nylon-66 pores by 20% drastically reduces permeability. membranes drastically reduced flux. The celSince the protein solution was not recircululose acetate membrane seems less sensitive lated, the time during which the protein was to pore blocking. under the influence of high shear was short, Generally no extra adsorption effect due to some parts of a second. It is well known, neverdenaturation could be distinguished from adtheless, that protein under high shear undersorption. “Soft” proteins can adsorb more easgoes conformational changes. Bowen and Gan ily due to an extra driving force emanating from [ 111 studied three different MF-membranes, the conformational changes the protein underwhich all displayed greater permeability loss at goes in leaving the solvated stage and adsorbhigh pumping velocities, than at low; they exing onto a surface [12,13]. However, j?-lactoplained the difference in terms of differences in globulin is not generally regarded as a soft applied shear. protein. The high permeability of Nylon-66 The higher amount of protein deposited on membranes fits in very well with its high poNylon-66 and cellulose acetate membranes as rosity and the considerable size of its largest compared with the hydrophilic PVDF mempores. The consequently high initial flux in Clbrane, is compensated by the higher total tering, however, also means the accumulation permeability of the first two. Thus, the memof fouling materials at the surface, and consebrane material is not of any considerable imquently rapid concentration polarization and portance for the separation performance as a fouling. Membranes of comparatively low powhole. The low flux for Nylon-66 membranes rosity do not show this dramatic sequence as compared with cellulose acetate membranes, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFE clearly. It might be valuable to compare MF with UF. The latter membrane type is mostly asymmetr1 ical, has lower porosity and subsequently lower 0.9 water permeability. Nilsson [ 81 found a 50-75% i 0.8 3 permeability loss for polysulfone membranes t with a nominal molecular weight cut-off of 20,000 after exposure to j?-lactoglobulin of various concentrations in the dead-end mode. A typical UF membrane has a pore size distribuI:.numbsrdstI tion from 0.5 to 10 nm. The internal membrane area is only slightly larger than the planar area, meaning that the protein adsorption mainly is at the feed-membrane interface. The interacpore sizehm zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA tion between protein-monomer and UF memFig. 15. The nominal permeability, permeability after the brane is controlling the pore clogging, while a adsorption of one bilayer in the pores, after the adsorption MF membrane pore scarcely is blocked by one of two layers in the pores, after a loss of 20% of the pores monomer. and after the loss of 40% of the pores, and the nominal pore If the product cost is high, then it may be imnumber distribution for hydrophilic PVDF membrane. T 70 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 portant to minimize protein adsorption, i.e. loss of protein when engineering membrane processes. Since the reduction in water permeability is unaffected and all membranes studied showed approximately the same relative permeability reduction, membrane costs seem to be the most relevant process parameter in general. One should also bear in mind the importance of choosing membranes with good chemical stability, for regular and effective cleaning. In the case of tangential microfiltration, the membranes should have a narrow pore size distribution in order that the whole membrane can be used for the separation purposes, and a hydrophilic surface for a minimum of protein adsorption and deposition. It is neither an advantage nor a disadvantage to choose a membrane with a very high pure water permeability, but the hydrodynamics must be optimized in order to avoid high foulant concentrations at the membrane. Conclusions Liquid displacement porosity is an easy and comprehensive porosimetric method. The high pure water permeabilities for microfiltration membranes reported in the literature are to a large extent caused by a small number of large pores. When these pores are plugged, or the surface above them is covered with a fouling layer, the permeabilities, and the filtration capacity, fall rapidly. Different materials have different adsorption behaviour: hydrophilic PVDF has the lowest protein adsorption of the four materials studied. The permeability loss after protein adsorption seems to originate from the influence of a small number of larger protein aggregates that clog some of the large pores important for the flow of liquid. Acknowledgement Marie Wahlgren at the Dept. of Food Technology, and Prof. Petr Dejmek at the Dept. of Food Engineering, University of Lund, are acknowledged for help and valuable critisism when preparing the paper. The Ernhold LundStrom foundation and the National Swedish Board for Technical Development are acknowledged for financial support. List of symbols A pore area (m”) flux (m/set) J zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLK L equivalent pore length (m) n number of pores P L v, rl pressure (Pa) pore radius (m) interfacial tension (N/m) contact angle viscosity of the testing liquid (Pa-set ) References G. Capannelli, I. Becchi, A. Bottino, P. M oretti and S. M unari, Computer driven porosimeter for ultrafiltration membranes, in: K.K. Unger et al. (Eds.), Characterization of Porous Solids, Elsevier Science Publishers B.V., Amsterdam, The Netherlands, 1988, p. 283. S. M unari, A. Bottino, G. Capannelli and P. M oretti, M embrane morphology and transport properties, Desalination, 53 (1985) 11. A. Bottino, G. Capannelli, P. Petit-Bon, N. Cao, M . Pegoraro and G. Zoia, Pore size and pore-size distribution in microfiltration membranes, Sep. Sci. Technol., 26 (1991) 1315. E. M atthiasson, M acromolecular adsorption and fouling in ultrafiltration and their relationships to concentration polarization, Thesis, Lund, Sweden, 1984. W .R. Bowen and Q. Gan, Properties of microfiltration membranes: Adsorption of bovine serum albumin at polyvinylidene fluoride membranes, J. Colloid Interface Sci., 144 (1991) 254. K.M . Persson and J.L. Nilsson, Fouling resistance models in M F and UF, Desalination, 80 (1991) 123. K.M. Persson et al./J. Membrane Sci. 76 (1993) 61-71 71 7 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA D. Dottavio-Martin and J.M. Ravel, Radiolabeling of 11 W.R. Bowen and Q. Gan, Properties of microfiltration proteins by reductive alkylation with (Y) formalmembranes: Flux loss during constant pressure perdehyde and sodium cyanoborohydride, Anal. Biochem., meation of bovine serum albumine, Biotechnol. Bioeng., 38 (1991) 688. 87 (1978) 562. 12 M. Wahlgren and T. Arnebrant, Protein adsorption 8 J.L. Nilsson, A study of ultrafiltration membrane to solid surfaces, TIBTECH, 9 (1991) 201. fouling, Thesis, Lund, Sweden, 1989. 13 W. Norde and J. Lyklema, Protein adsorption and 9 T. Nylander, Proteins at the metal/water interface. bacterial adhesion to solid surfaces: A colloid-chemiAdsorption and solution behaviour, Thesis, Lund, cal approach, Colloids and Surfaces, 38 (1989) 1. Sweden, 1987. IO V. Gekas and B. Hallstrom, Microfiltration membranes, Crossflow transport mechanisms and fouling studies, Desalination, 77 (1990) 195.