dynamically formed poly (vinyl alcohol) ultrafiltration membranes with good anti-fouling...

Journal of Membrane Science 169 (2000) 17–28

Dynamically formed poly (vinyl alcohol) ultrafiltrationmembranes with good anti-fouling characteristics

Li Na ∗, Liu Zhongzhou, Xu ShuguangPolymer Membrane Division, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences,

P.O. Box 2871, Beijing, 100085, PR China

Received 9 July 1999; received in revised form 21 September 1999; accepted 8 October 1999

Abstract

The anti-fouling poly (vinyl alcohol) TFC membranes were dynamically prepared when an aqueous solution containingpoly (vinyl alcohol), cross-linking agents and additives passed through porous substrate membranes under definite pressurefollowed by heat treatment resulting in cross-linking reaction and drying. By means of dead-end filtration the poly (vinylalcohol) solution deposited on the surface and entered the pores of porous substrate and thus an even poly (vinyl alcohol)gel layer was formed on both the external and internal surface. The effects of poly (vinyl alcohol) concentration, dynamiccoating time, additive concentration, cross-linking agent concentration, curing time and support membrane on pure water fluxand protein rejection of the resultant membranes were studied. Membrane morphology was characterized by SEM studies.The modified membranes were investigated in ultrafiltration experiments with pepsin to assess their resistance to fouling.The experimental results prove that modified membranes with an even poly (vinyl alcohol) hydrogel layer show dramaticallyhigh anti-fouling characteristics compared to inadequately modified and unmodified membranes. Moreover, it is possiblethat a series of dynamic membranes within broader molecular weight cut-off can be readily obtained by suitable control andcombination of various preparative conditions. ©2000 Elsevier Science B.V. All rights reserved.

Keywords:Ultrafiltration; Anti-fouling; Dynamic membrane; Poly (vinyl alcohol); Hydrogel layer

1. Introduction

Today, more and more attention in membrane sepa-ration technology has been focused upon solutions todecrease and eliminate irreversible membrane foulingbecause of its serious hindrance to the developmentand application of membrane science and technology.Much important research work is on the technique ofsurface modification which is based on the principlethat increasing hydrophilicity of membrane surface

∗ Corresponding author. Fax:+86-0106-2923441.E-mail address:[email protected] (L. Na).

can generally improve the ability to resist fouling byinhibiting non-specific binding between the membranesurface and retained molecules, particularly proteins.

Poly (vinyl alcohol) (PVA) polymer, with highlyhydrophilic character, good film-forming propertiesand outstanding physical and chemical stability, is akind of excellent membrane material for preparationof a hydrophilic membrane. In fact, considerable workhas been carried out in the area of PVA–RO compos-ite membrane, where PVA or PVA co-polymer wasprepared as a selective skin layer of composite ROmembrane. The PVA skin layer provides the resultingmembranes with high water permeation rate, good

0376-7388/00/$ – see front matter ©2000 Elsevier Science B.V. All rights reserved.PII: S0376-7388(99)00327-0

18 L. Na et al. / Journal of Membrane Science 169 (2000) 17–28

anti-fouling nature, excellent integrity in acidic andalkaline environments and remarkable resistance toabrasion.

For example, Petersen [1] reviewed the publishedactivity over the period 1961–1993 on RO membranesmade with PVA. Immelman et al. [2] prepared bothflat-sheet and tubular composite RO membranes bydepositing aqueous solutions of PVA and potassiumperoxydisulphate on asymmetric poly (arylether sul-phone) [PES] substrate membranes. Lang et al. [3,4]prepared PVA composite membranes for RO with var-ious cross-linking agents including aldehydes, dialde-hydes and malic acid. The anti-fouling low-pressureRO membranes-LF10 series commercialized by NittoDenko Corporation in 1997 indicated the directionin the development of the RO membrane. Theirmembranes were formed by coating a traditionallymade polyamide support membrane with a PVA layerthat not only eliminates the negative charge of thesupport membrane but also provides hydrophilicityand chlorine resistance, thereby largely improvingthe anti-fouling characteristics. Besides this, PVA isalso widely used as the material of selective skin ofcomposite pervaporation membranes.

The studies on preparation of PVA membranesapplied to UF are relatively rare compared to that ofPVA membranes for reverse osmosis or pervaporation.Since the UF process is usually used in separationand concentration of foods, pharmaceuticals and bio-logical substances, especially for the concentration ofdilute protein solutions and separation of proteins withvarious molecular weights, it requires more exploita-tion of the use of PVA in UF membranes to reduceadsorption of proteins or other macromolecules con-cerned on the membrane surface. Few reported workshave thus far been carried out on PVA-UF membranes.Li and Barbari [5] inferred that the use of cross-linkedhydrogel for the modification of UF membranes was alarge unexplored area. Hydrogels provide smooth, hy-drophilic surfaces with minimal protein-binding andcan be manipulated easily in terms of thickness andthe degree of cross-linking to fit desired UF needs.Li and Barbari prepared composite UF membranesby spin-coating cross-linked PVA hydrogel onto re-generated cellulose UF membranes that providedhigh anti-fouling to bovine serum albumin (BSA) asindicated by nearly 100% pure water flux recovery.They considered that since the PVA hydrogel surface

exhibited minimal protein adsorption, any irreversiblefouling that manifested itself in less than 100% purewater recovery was associated with the surface ofregenerated cellulose support. This was consistentwith the study of Fane et al. [6] who inferred thatalthough the adsorption was largely on the top sur-face and the amount of protein adsorbed within thepores could be a minor fraction of the total quan-tity adsorbed, it could have a profound effect onmembrane performance. Nabeet et al. [7] proved bytheir experiments that aggregation in the pores oron the pore surface could be an important mech-anism of flux decline and that the chemistry ofthe substrate membrane surface directly affectedthe flux decline during the ultrafiltration of proteinsolution.

So, if the hydrophilicity of the internal surface ofmembrane is increased as well as that of the exter-nal surface through suitable surface treatment, it isexpected that the resultant elimination of macroso-lute adsorption within the pores and of blockage ofthe passage should lead to markedly improved resis-tance to fouling and increase in separation efficiencyfor macromolecular mixtures and membrane perme-ation stability. However, most composite membraneshave hydrophobic support and much attention has beenfocused on the hydrophilicity and formation of theultra-thin skin layer of the composite membranes butnot on the hydrophilicity and internal pore fouling ofthe support membrane.

The present study focuses on the development ofgood anti-fouling PVA-based TFC membrane for UF.The PVA composite membranes were formed whenan aqueous solution containing PVA, cross-linkingagents and additives passed through porous sub-strate membranes under pressure. By means of thisdynamic coating method the PVA casting solutiondeposited on the surface and entered the pores of theporous substrate by pressure and deposited on theinternal surface of the support. Then the deposit wassubmitted to heat treatment leading to cross-linkingreactions and drying. Thus an even cross-linked PVAultra-thin layer formed on the external and internalsurface. Both UF and MF membranes have been usedas porous substrates. In order to obtain satisfactoryflux, macroporous substrates with high pure waterflux and high porosity are preferable. By ultrafil-tration experiments with protein feed solution, the

L. Na et al. / Journal of Membrane Science 169 (2000) 17–28 19

Table 1The support membranes characteristics

Membrane designation Material Total thicknessa (mm) Nominal MWCOb (Dalton)/pore diameter (mm) Water fluxc (l/m2 h)

PA-1 PAN 190 60,000 460±20PV-1 PVDF 165 150,000 800±20PV-2 PVDF 150 500,000 1800±50NL-1 Nylon 155 1mm 2200±100d

a Including the non-woven support of thickness approx. 112mm.b Based on above 90% rejections of proteins.c The test conditions are 0.1 MPa, room temperature.d The test conditions are 0.05 MPa, room temperature.

membranes were characterized with respect to theirresistance to fouling and their ‘cleanability’.

2. Experimental

2.1. Materials

Fully hydrolyzed PVA powder having a degreeof polymerization of 1799 was obtained from Bei-jing Hongxing Chemical Industry Factory. Fullyhydrolyzed PVA powder having average molecularweight of 11,000–31,000 was obtained from Aldrich.Glutaraldehyde (GA), the cross-linking agent, wasobtained as a 50 wt% GA aqueous solution. The glu-taraldehyde concentration in this paper all refers tothe concentration of 50 wt% GA. Methanol, aceticacid and sulfuric acid were of reagent grade andwere used without further purification. The water wasde-ionized.

The flat-sheet microfiltration membranes and ultra-filtration membranes used as the support membraneswere made by traditional phase-inversion methodon polyester non-woven fabric in our laboratory and

Fig. 1. Flow sheet of the experimental apparatus.

workshop. In principle, almost all porous polymericmembranes had high water flux and permitted perme-ation of PVA. The support membranes used in thisstudy are listed in Table 1. Prior to use, the supportmembranes were soaked in distilled water to removethe glycerol-wetting agent and the support membranewas tested to determine the pure water flux and thenstrictly selected to ensure that its water flux waswithin the given range for its type as indicated inTable 1. A discussion of casting conditions of sup-port membranes falls beyond the scope of this paper.The support membranes were pre-treated by beingimmersed in a 60 vol% aqueous solution of alcoholfor 24 h to increase its hydrophilicity and then coatedaccording to the following procedures.

2.2. Apparatus and modification procedures

The dynamic coating of support membranes andprotein fouling experiments were carried out by usinga 50 ml stirred dead-ended filtration cell with 10 cmmembrane area. A flow sheet of the experimental ap-paratus is shown in Fig. 1.


Aqueous PVA solution prepared by dissolving thepolymer in distilled water completely and then be-ing filtered was added with additive, glutaraldehydeas cross-linking agent, methanol as quencher, aceticacid as buffer, sulfuric acid as catalyst. The mixturewas stirred until homogeneous and heated to 45◦C.Then the 50m1 mixture was quickly poured into thestirred cell in which a support membrane had beenloaded, and was pressured through the membrane byusing a nitrogen cylinder. During a pre-determined pe-riod of filtration, the dynamic membrane was formed.The membrane was taken out and cross-linked in anoven at 50◦C for a pre-determined period. Finally, themembrane was swollen in distilled water and storedfor performance experiments. The particular prepara-tion and testing conditions are indicated in the figuresand tables in Section 3.

2.3. Membrane performance test

2.3.1. Permeability and retention studiesIn order to study the effects of various prepara-

tive conditions on the performance of the resultantcomposite membranes, three primary parameters weredetermined: pure water flux (l/m2 h), protein aqueoussolution flux (l/m2 h) and the protein rejection (Re)which is defined asRe =1− Cp/Cb, whereC is theconcentration of protein in the permeation andCbis the concentration of protein in the bulk solution.The protein rejection and molecular weight cut-off(MWCO), defined as above 90% rejection of proteins,were measured using two kinds of proteins: pepsin andbovine serum albumin (BSA). The molecular weightand isoelectric points (PI) of these proteins are listedin Table 2. Flux data were measured at room temper-ature under pre-determined operating pressure usingthe apparatus of Fig. 1. Protein rejection values wereobtained using UV at 280 nm (Model UV-120-02,Shimadzu).

Table 2Molecular weights and isoelectric points of the proteins used

Protein Molecular weight Isoelectric point

Pepsin 35,000 Below 1.0BSA 65,000 4.8

2.3.2. Fouling study methodsThe fouling experiments were consecutively carried

out with the apparatus of Fig. 1:1. The pure water fluxes were measured volumetri-

cally until the fluxes remained constant for threesuccessive readings (over 5 min).

2. The UF experiment with an aqueous solutionof protein was performed until the fluxes re-mained constant for three successive readings(over 5 min). The nominal molecular weight ofthe chosen protein is smaller than the MWCOof the tested composite membranes so as to in-dicate the internal fouling situation of the testedmembranes. Pepsin aqueous solution was usedto assess the resistance of membranes of 60,000(MWCO) to fouling.

3. The membrane was taken out and briefly rinsedwith de-ionized water for approximately 5 min.and the pure water fluxes were measured againuntil the flux remained constant for three succes-sive readings (over 5 min).

4. The same way as Step (2).5. The same way as Step (3).

2.3.3. SEM studiesA number of scanning electron micrographs was

recorded to compare the morphology of the modifiedmembranes with that of the unmodified membranes.All samples were soaked in 50% glycerol aqueous so-lution for at least 10 h, dried in vacuum, fractured inliquid nitrogen and gold-coated. Fig. 2 (a, b) showsthe top surface and cross-section of NL-1 supportmembrane magnified×2000, respectively. Fig. 3(a, b)shows the top surface and cross-section of a modifiedmembrane magnified×2000, respectively. Fig. 4(a,b) shows the top surface of another modified mem-brane magnified×10000 and cross-section magnified×2000.

3. Results and discussions

3.1. Effects of various preparative conditions onmembrane performance

The permeability and structures of PVA–TFC mem-brane were much different from that of the substrate


Fig. 2. (a) Scanning electron micrograph of the top surface ofNL-l support membrane magnified×2000. (b). Scanning electronmicrograph of the cross-section of NL-1 support membrane mag-nified ×2000.

membrane since a PVA cross-linked layer on both theexternal and the internal surface covered by the sub-strate membrane. Various preparative conditions usedin the process can cause more or less effects on theperformance of the resultant membrane and should bestudied in detail.

Fig. 3. (a) Scanning electron micrograph of the top surface ofthe PVA composite membrane magnified×2000. (b) Scanningelectron micrograph of the cross section of the PVA compositemembrane magnified×2000.

3.1.1. Effect of PVA concentrationThe effect of PVA concentration in the casting so-

lution on membrane performance was studied usingPA-2 membrane as substrate, which was dynamicallycoated with coating solutions containing from 0.3 to1 wt% PVA whereas other preparative conditions were


Fig. 4. (a) Scanning electron micrograph of the top surface ofthe PVA composite membrane magnified×10,000. (b) Scanningelectron micrograph of the cross-section of the PVA compositemembrane magnified×2000.

fixed. The UF experimental data illustrated in Fig. 5show that an increase in PVA concentration resulted ina decrease in pure water flux and an increase in BSAretention. The increase of PVA concentration meansthat more PVA penetrates into and through the mem-brane during formation procedures. Under the pressure

Fig. 5. Effect of PVA concentration and dynamic coating timeon the performance of PVA-1799 coated PV-2 membrane. Condi-tions of membrane preparation: concentration of glutaraldehyde,0.3(vol%); volume concentration of methanol, acetic acid and sul-furic acid, 0.2, 0.3, 0.1%; the additive concentration, 12 vol%;operating pressure, 0.2 MPa; curing time: 7 h. Conditions of UFtesting: 0.1 MPa, room temperature.

one part of PVA passes through the pores and leavesthe membrane while other part of PVA stays on both,the external and internal surface of the membrane asa boundary layer. An increase in concentration pro-vides more PVA into the boundary layer and becomescross-linked. This leads to the result of smaller porediameters, which is responsible for the reduction inwater flux and the increase in protein retention.

3.1.2. Effect of dynamic coating timeIn the preparation process the dynamic coating time

of PVA aqueous solution through support membranecan have an evident effect on membrane performance.To study the effect, the coating time was varied from1 to 13 min under pressure of 0.2 MPa. The resultsin Fig. 5 show that BSA retention increased whereaspure water flux decreased with increasing coatingtime. During the formation procedure one part of PVApasses through the substrate membrane and leaves themembrane under pressure and other part of PVA stayson both the external and internal surface and forms adynamic layer on both the external and internal sur-face. It is expected that when filtration time increases,


Fig. 6. Effect of glutaraldehyde concentration in PVA castingsolution on the performance of PVA-1799 coated PV-2 membrane.Conditions of membrane preparation: PVA-1799 concentration,0.3 wt%; the additive concentration, 12 vol%; volume concentrationof methanol, acetic acid and sulfuric acid, 0.02, 0.03, 0.0 1%;operating pressure, 0.2 MPa; curing time, 5 h. Conditions of UFtesting: 0.1 MPa, room temperature.

more PVA deposits on the coating layer. The smallerpores on the substrate membrane would be affectedinitially and be filled quickly and the dynamic layeron the top surface of the substrate membrane willeven cover the smaller pores with increasing the filtra-tion time. In the meanwhile, the bigger pores are notcovered completely with the dynamic layer, but theiractual pore diameters decrease because the coatinglayer formed on the wall of the pores is thicker withthe increasing of filtration time. All of these causea decrease in the water flux and a lower molecularweight cut-off (MWCO).

3.1.3. Effect of cross-linking agent concentrationFig. 6 illustrates the effect of glutaraldehyde con-

centration in PVA casting solution on the performanceof the composite membrane. The glutaraldehyde con-centrations were varied from 0.1 to 1.2 vol% and theother parameters were fixed. Increasing glutaralde-hyde concentration led to increasing BSA retentionand decreasing pure water flux of PVA compositemembrane. It is unclear whether cross-linking results

Fig. 7. Effect of additive concentration on the performance ofPVA-1799 coated PV-2 membrane. Conditions of membrane prepa-ration: PVA-1799 concentration in casting solution, 0.3 wt%; glu-taraldehyde concentration, 0.3 vol%; the volume concentration ofmethanol, acetic acid, sulfuric acid, 0.02, 0.03, 0.0 1%; operat-ing pressure, 0.2 MPa; curing time, 7 h. Conditions of UF testing:0.1 MPa, room temperature.

in more coating or a denser coating. There are somereports on the structure of PVA cross-link includingthe effect of cross-linking agent on the structure ofPVA hydrogel, such as the study of Burczak et al.[8]. They illustrated that the cross-linked PVA hydro-gel consists of networks of macromolecular chains(macromolecular mesh). Their experimental resultsdemonstrated that the increase in the concentrationof the added cross-linking agent formed the densercross-linked PVA hydrogel network. According totheir study, it is expected that the addition of GA re-sults in denser coating at the same PVA concentrationand the density will increase with the increase of GAconcentration, which results in increase in BSA anddecrease in pure water flux.

3.1.4. Effect of additive concentrationThe effect of additive concentration on membrane

performance is shown in Fig. 7. The additive used wasPEG-400. An increase in additive concentration appre-ciably increased pure water flux of composite mem-brane, but slightly decreased BSA retention. PEG-400


Table 3Effect of curing time on the performance of composite membranea

Curing time 1 hour 4 hours 7 hours

PWP (l/m2 h) 330.6 310.2 290.4Re (BSA) (%) 90.7 92.3 94.1

a Conditions of membrane preparation: PVA concentration,0.6%; glutaraldehyde concentration, 0.3% vol.; the volume con-centration of methanol, acetic acid, sulfuric acid, 0.02, 0.03, 0.01%; dynamic coating time, 5 min; operation pressure, 0.2 MPa.Conditions of UF testing: 0.1 MPa, room temperature.

is a pore-forming agent here. When PVA compositemembrane was soaked in distilled water during the for-mation procedures, PEG was separated out from thePVA membrane and dissolved into the distilled water.So increasing PEG concentration results in formingmore pores and this may give a chance to form largepores, thereby leading to an increase in pure water fluxand the slight decrease in protein retention.

3.1.5. Effect of curing timeThe effect of curing time on the performance of the

membrane is shown in Table 3. The longer curing timeresulted in higher BSA retention, but lower flux. Thismay be because curing induced water removal fromthe gel structure and is responsible for a closer positionof PVA chains. However, the change is not very sig-nificant compared to those caused by the effects of theprocedures shown in Sections 3.1.1–3.1.4. This maybe related with the curing temperature, 45◦C, whichwas not very high.

3.1.6. Effect of support membraneSince the pore wall of substrate membrane was

covered with a cross-linked PVA layer and the resul-tant pore diameters were changed markedly, it is nat-ural that the water flux of PVA–TFC membranes wasmuch lower than those of their substrate membranes,and strongly affected by the performance of sub-strate membranes. The experimental results shown inTable 4 illustrated that the water flux of the compositemembrane increased, whereas the protein rejectiondecreased with an increase in pore sizes of substratemembranes when all remaining variables were fixedat pre-determined values. It is quite evident that theeffect of substrate on the flux and retention of thecomposite membrane was considerably large among

Table 4Effect of support membrane on the performance of compositemembranea

Support membrane PA-1 PV-1 PV-2 NL-1

PWP (l/m2 h) <10 100 190 1460b

Re (BSA) (%) – 99.9 97.2 –

a Conditions of membrane preparation: PVA-1799 concentra-tion, 1 wt%; glutaraldehyde concentration, 0.7 vol%; the volumeconcentration of methanol, acetic acid, sulfuric acid, 0.04, 0.06,0.02%; dynamic coating time, 3 min; operation pressure, 0.2 MPa.Conditions of UF testing: 0.1 MPa, room temperature.

b The test conditions are 0.05 MPa, room temperature.

the preparative conditions. Obviously the substrateswith high water flux are preferable and it is possi-ble to produce a PVA–TFC membrane with broaderMWCO range by manipulating and controlling vari-ous preparative conditions when the substrate mem-branes with large pore diameter and high porosity areapplied. Just considering the effect of substrate mem-brane, glutaraldehyde was chosen as a cross-linkingagent because PVA cross-linking by glutaraldehydecan take place at room temperature and heat treatmentat high temperature can be avoided which can causethe most significant flux loss as proved by Lang etal. [4]. Since the performance of some membranes isaffected less by heat treatment at high temperature,such membranes can be chosen as substrates whenthe cross-linking agents must react with PVA at hightemperature. So the process can use other kinds ofthe cross-linking agents without much loss of waterflux of PVA composite membrane caused by heattreatment at high temperature.

3.1.7. Effect of post-treatmentDuring the preparation of PVA composite mem-

branes, after the cross-linking reaction of PVA fora pre-determined period, the membrane was oftensoaked in distilled water to make the additive sep-arated out and the PVA layer reach swelling equi-librium. Here, in order to investigate the stability ofthe PVA composite membrane in hot water, after themembrane was soaked in 30◦C distilled water for2 days, it was then soaked in 80◦C distilled waterfor 24 h post-treatment. Table 5 shows the prepara-tive conditions of two PVA composite membranes,1′ and 2′. Test the pure water fluxes and protein re-tention of the two membranes with post-treatment


Table 5Conditions of membrane preparation to study the effect of post-treatment

Membrane designation 1′ 2′

Substrates PV-1 PV-2PVA-1799 concentration (wt%) 1 1Glutaraldehyde concentration (vol%) 0.6 0.6Concentration of methanol, acetic acid, sulfuric acid (vol%) 0.04, 0.06, 0.02 0.04, 0.06, 0.02Additive (PEG-400) Concentration (vol%) 12 12Dynamic coating time (min) 5 5Operation pressure (MPa) 0.2 0.2Curing time (h) 7 7

and the two membranes without post-treatment. Theexperimental results shown in Fig. 8 illustrate thatthe fluxes of PVA membranes with post-treatmentare rather higher than that of the membranes with-out post-treatment while the difference of proteinretention between them is not large. Protein retention

Fig. 8. Effect of post-treatment on the performance ofPVA-1799-coated PV-1 and PV-2 membranes.

of the membranes after post-treatment still remainsrather high. This result may be related to the factthat the additive, PEG, is separated out more com-pletely and some short-chain PVA molecules notcross-linked during cross-linking reaction, may dis-solve into hot water during soaking. On the otherhand, that the protein retention of the membranes isstable after treatment in 80◦C hot water indicates thestability of the PVA composite membrane. The aboveresults suggest that this post-treatment can improvethe general performance of the PVA composite mem-brane. PVA is a kind of water-soluble polymer andthe membrane prepared with PVA may dissolve inwater during application processes if PVA does notproceed with the cross-linking reaction. The chemicalcross-linking of PVA in an aqueous solution leadsto the formation of an insoluble polymeric network.So it is expected that the cross-linked layer lastslonger than an uncross-linked layer. In the followingwork we will further study the stability of the PVAcomposite membrane in other media under certainconditions.

3.2. Preparation of PVA composite membrane withdifferent MWCO

Figs. 2–4 show the differences of morphologyamong support membrane and its modified compositemembranes. The support membrane was NL-l mem-brane made in our lab. The preparatory conditions ofmodified membranes are shown in Table 6. It can bedistinguished that the average pore size of modifiedmembranes shown in Fig. 3(a) was not much smallerthan that of support membrane shown in Fig. 2(a)and the pore size of modified membrane shown in


Table 6Preparative conditions of membranes of Figs. 3 and 4

Membrane Membrane of Fig. 3 Membrane of Fig. 4

Type Modified membrane Modified membranePVA-1799 concentration, (wt%) 0.7 3PVA (M.W. 11,000–31,000) concentration (wt%) 0.3 2Glutaraldehyde concentration, (vol%) 0.5 1Concentration of methanol, acetic acid, sulfuric acid (vol%) 0.04, 0.06, 0.02 0.1, 0.15, 0.05Additive (PEG-6000) concentration (wt%) 3 3Dynamic coating time ( min) 2 10Operation pressure (MPa) 0.05 0.1Curing time (h) 7 7

Fig. 4(a) was so small that it could not be observedwith SEM magnified×10000. This resulted from thedifference in preparative conditions between the twomodified membranes. So it is evident that PVA com-posite membranes with different diameters or MWCOcan be prepared by suitable control of various prepar-ative conditions and it is possible to dynamicallymake a series of PVA composite membranes withinbroad molecular weight cut-off.

Fig. 4 clearly shows a thick and smooth dynamiclayer on the top surface while the coating thicknessof the membrane in Fig. 3 is relatively small. Fromthis it can be suggested that with the increase of fil-tration time or/and PVA concentration, the dynamiclayer on the top surface increases quickly and createsa denser layer, and plays a much more important rolein the performance of the PVA composite membrane.It can be expected that the combination of a very largepore size, low concentration and short filtration timeproduce the thinner PVA coating layer on the mem-brane surface, as shown in Fig. 3. The top dynamiclayer is not a continuous layer so the pores can remainlarge.

The authors of this paper think that this dynamicformation method should not be limited to applica-tion just for preparing flat-sheet membranes for UF.It can be extended to the application for preparingcomposite membranes in other configurations in situ,such as tubular composite membranes and hollow fibercomposite membranes. In fact, the authors have pre-pared hollow fiber PVA composite membranes withMWCO 35,000 in situ by this method and have provedthat the composite membranes have good anti-foulingcharacteristics.

3.3. Fouling studies

Short-time fouling tests for the modified and un-modified membranes have been performed by themethods described in Section 2.3.2. The experimentalresults are shown in Figs. 9 and 10.

Fig. 9 shows typical flux versus time profiles formodified thin-gel composite membrane 3′ (MWCO,60,000) and unmodified polysulfone membrane(MWCO, 60,000) that exhibit approximately the same

Fig. 9. Flux vs. time of modified membrane 3′ and unmodifiedpolysulfone membrane that exhibit approximately same rejectionof 0.1% pepsin aqueous solution.

Fig. 10. Flux vs. time of two different modified membranes 4′ and5′ that were dynamically coated for 3 and 1 min, respectively.


Table 7Conditions of membrane preparation to study the anti-fouling characteristics

Membrane designation 3′ 4′ 5′

Substrates PV-2 PV-2 PV-2PVA-1799 concentration (wt%) 0.5 0.8 0.8Glutaraldehyde concentration (vol%) 0.2 0.3 0.3Concentration of methanol, acetic acid, sulfuric acid (vol%) 0.02, 0.03, 0.01 0.02, 0.03, 0.01 0.02, 0.03, 0.01Additive (PEG-400) Concentration (vol%) 12 12 12Dynamic coating time (min) 3 3 1Operation pressure (Mpa) 0.2 0.2 0.2Curing time (h) 6 6 6

rejection of 0.1% pepsin solution, 71.7 and 70.7% re-spectively. The preparative conditions of the modifiedmembrane 3′ are listed in Table 7. Fig. 9 shows thatmodified membrane 3′ exhibits stable pure water fluxand protein solution flux during filtration and as goodas 100% water flux recovery while the polysulfonemembrane exhibits relatively rapid flux decline andlow flux recovery. It shows that the modified mem-brane has excellent ‘cleanability’ and anti-foulingcharacteristics over a few cycles. Flux decline withtime of suitably modified membranes appears to becontrolled by the reversible formation of a proteinlayer adjacent to the hydrogel and compaction ofmembranes under pressure during filtration.

Fig. 10 shows typical flux versus time profiles fortwo different modified thin-gel composite membraneswhich exhibit 90.7 and 83.8% rejections of 0.1% BSA,respectively. The preparative conditions of the twomembranes are listed in Table 7. The only differencebetween the preparation processes of the two mem-branes consists in dynamic coating time: 3 and 1 min,respectively. From the above content, we can knowthat when other preparation conditions are fixed, in-creased coating time leads to the decrease of the flux.So the flux of Membrane 5′ should be higher than thatof Membrane 4′. Fig. 10 shows that the flux of Mem-brane 5′ is higher than that of Membrane 4′ for thefirst 60 min, but after 60 min the flux of Membrane 5′is rather similar to that of the Membrane 4′. This indi-cates greater flux decline of Membrane 5′. As the dif-ference of the coating times of the two membranes, 3and 1 min. respectively, is not very large, the differenceof the fluxes between the two membranes appears to benot very large too, including at 150 min stage, accord-ing to Fig. 10. However, the degree of pure water fluxrecovery of Membrane 5′ is rather lower compared tothat of Membrane 4′. This proves that Membrane 4′

is better than Membrane 5′ in anti-fouling character-istics. In addition, the membranes have been removedfrom service at different times because it took differ-ent time for the two membranes to reach stable fluxand it appears to be easier for Membrane 4′ to reachstable flux than the Membrane 5′. From Fig. 10 wecan distinguish the general decline situation of the twomembranes. It can be proved that the more dynamiccoating time, the more the stability of the flux and thegreater degree of water flux recovery. The degree ofpure water flux recovery of Membrane 4′ can reach asmuch as 100% according to experimental data whilethat of Membrane 5′ is low, which may result fromthe too short dynamic coating time. It is expected thatwhen the dynamic coating time is too short, the PVAlayer on the surface of substrate membrane is not uni-form and some part of substrate membrane surface isexposed to feed solution which results in fouling andless ‘cleanability’.

4. Conclusion

1. The anti-fouling PVA–TFC membranes can bedynamically prepared when an aqueous solutioncontaining PVA, cross-linking agents and addi-tives pass through porous substrate membranesunder definite pressure followed by cross-linkingreaction and drying.

2. The effects of preparative conditions on resultingcomposite membranes were investigated. Increas-ing PVA concentration of the casting solution, dy-namic coating time, glutaraldehyde concentrationand curing time, as well as decreasing additiveconcentration, all result in a decrease in flux andan increase in protein rejection.


3. Due to the flexibility of the manipulation in theprocess, PVA composite membranes with differ-ent molecular weight cut-off can be readily ob-tained by suitable control of various preparativeconditions and it is possible to dynamically makea series of PVA composite membranes with broadmolecular weight cut-off.

4. On the basis of flux stability and pure waterflux recovery tested with pure water and proteinaqueous solution, modified membranes coatedwith an even PVA hydrogel layer show dramat-ically high anti-fouling characteristics and good‘cleanability’ compared to inadequately modifiedmembranes and unmodified membranes. The ad-equately modified membrane shows recoverablepure water flux and protein solution flux thatmakes them suitable for repeat usage applicationsespecially for the concentration or separation ofprotein solutions.

Acknowledgements

This work was supported by the Chinese Academyof Sciences, Project KZ951-A1-201-02, KZ95T-05.

References

[1] R.J. Petersen, Composite RO and NF membrane, J. Membr.Sci. 83 (1993) 81–150.

[2] E. Immelman, D. Bezuidenhout, R.D. Sanderson, Poly (vinylalcohol) gel sub-layers for reverse osmosis membranes.III. Insolubilization by cross-linking with potassiumperoxydisulphate, Desalination 94 (1993) 115–132.

[3] K. Lang, T. Matsuura, G. Chowdhury, Preparation andtesting of polyvinyl alcohol composite membranes for reverseosmosis, Can. J. Chem. Eng. 73 (1995) 686–692.

[4] K. Lang, S. Sourirajan, T. Matsura, G. Chowdhury, A studyon the preparation of polyvinyl alcohol thin-film compositemembranes and reverse osmosis testing, Desalination 104(1996) 185–196.

[5] R.H. Li, T.A. Barbari, Performance of poly (vinyl alcohol)thin-gel composite ultrafiltration membranes, J. Membr. Sci.105 (1995) 71–78.

[6] A.G. Fane, C.J.D. Fell, A.G. Waters, Ultrafiltration of proteinsolutions through permeable membranes – The effect ofadsorption and solution environment, J. Membr. Sci. 16 (1983)211–224.

[7] A. Nabe, E. Staude, G. Belfort, Surface modification ofpolysulfone ultrafiltration membranes and fouling by BSAsolutions, J. Membr. Sci. 133 (1997) 57–72.

[8] K. Burczak, T. Fujisato, M. Hatada, Y. Ikada, Proteinpermeation through poly (vinyl alcohol) hydrogel membranes,Biomaterials 15 (1994) 231–238.

dynamically formed poly (vinyl alcohol) ultrafiltration membranes with good anti-fouling...

Documents