development and characterization of uf membranes

7/27/2019 Development and Characterization of Uf Membranes

1/169

DEVELOPMENT AND CHARACTERIZATION OF UF MEMBRANESAND

RELATION OF MEMBRANE PROPERTIES TO ABSORPTIVEFOULING

Prepared by :Mark M. Clark,

Corine Combe, Elisabeth Molis, YonghunLee,Kerry Howe, Kwang-Soo Kim, Manish Kum ar, Pascale Lucas, and Y ingge

The University of IllinoisAt Urbana-Champaign

205 N. Mathews AvenueUrbana, IL 61801

Sponsored by:National W ater Research Institute

and Lyonnaise de Eaux


2/169

CONTENTS

Part iDiffusion an d Partitioning of Humic Acid in a Porous Ultrafiltration Membrane,

published in Journal of Membrane Science, 1998and

The Effect of CA Membrane Properties on Adsorptive Fouling of Humic Acid,published in Journal of Membrane Science, 1999

Part IIResults of SPEES/PES-PS Series Membranes

Part IIIResults of PVP-PS Series Membranes

Part IVThe Effect of pH and Ionic Strength on the Diffusion Coefficient of Humic Acid,

submitted to Environmental Science and Technology, 1999


3/169

Part I

Diffusion and Part i t ioning of Hum ic Acid in a Porous (Jltrafiltration Membrane,published in Journal of M e mbr a ne Science, 1998

andThe Effect of CA Membrane Properties on Adsorptlve Fouling of Humic Acid,

published in Journa l of Me mb rane Sc ience , 1999


4/169

journalofMEMBRANESCIENCEELSEVIER J o u r n a l of Membrane Science 154 (1999} 73-87

The effect of CA membrane properties onadsorptive fouling by humic acidC. Combe3, E. Molisa, P. Lucasa, R. Rileyb, M.M. Clarka>*

"Department of Civil am! Environmental E ngineering, University of Illinois, 205 North Matltews Avenue, Urbana, IL 6ISOI, USA^Separation Systems Technology, 490! Moreno Blvd. San Diego, CA 92! 17, USAReceived 9 April 1998; received in revised form 28 May 1998; accepted 31 Augusl 1998

AbstractCellulose acetate membranes with varied charge, hydrophobicity, porosity, an d pore size have been developed by annealing,

hydrolysis and oxidation of a basic cellulose acetate membrane. The effects of these modifications were characterized bypoly(ethylene glycol) retention, contact angle and streaming potential measurements, and atomic force microscopy. Thebehavior of th e membranes dur ing humic acid adsorption exper iments ha s also been studied. The results show that humic acidadsorption occurs both inside th e pore and on the membrane surface. Experiments at different pH show th e importance ofsolution properties on humic acid adsorption through modification of membrane and humic acid charge. Although manymembrane characteris tics ar e modified by hydrolysis an d oxidation, nei ther t reatment prevented humic acid adsorption on theCA membranes . The most effective surface t rea tment was with an an ion ic polymer , whi ch significantly reduced adsorption ofhumic acid. 1999 Elsevier Science B.V. All rights reserved.Keywords: Humic acid; Membrane preparation an d slruclure; Water treatment; Surface characterization

1. IntroductionHumic substances are abundant in natural waters.They are the result of chemical an d biological degra-dation of plant and animal residues and the synthesisactivities of microorganisms [1]. Humic and fulvjc

acids represent the major fraction of dissolved naturalorganic matter in aquatic environments. They areresponsible fo r natural water color and for in i t ia t ingphotochemical transformations of both organic com-pounds an d trace met als [2], These substances are alsoimportant constituents of the organic colloidal phase

"Corresponding author. Tel: +1-217-333-3629; fax: + 1-217-333-6968; e-mail: [email protected]

and are one of the major fouling agents during filtra-tion of surface waters in reverse osmosis [3], nanoftl-tration [4], ultrafiltration [5,6], and microfillration [7],Although membrane fouling by relatively well char-acterized tnacromolecules such as proteins has beenextensively studied [8,9], mechanisms of fouling byhumic substances are not yet well understood becauseof the heterogeneous nature of these macromolecules.According to previous studies, membrane fouling byhumic substances is influenced by the characteristicsof the humic substances an d membrane, (h e hydro-dynamic conditions, and the chemical composition ofthe feed water. Understanding of these factors isessential for better control of membrane fouling byhum ic acid and other types of natural organic matter.

037G-7388/99/$ - see f r on t mailer (r) 1999 Elsevier Science B.V. Al l rights reserved.P I I ; 8 0 3 7 6 - 7 3 8 8 ( 9 8 ) 0 0 2 6 8 - 3


5/169

74 C. Combe et al./Joiintcit of Membrane Science 154 (1999) 73-87

Several studies have focussed on humic acid proper-ties. Cornel et ai . [ 1 J found that pH an d ionic strengthaffect th e molecular size dis tr ibut ion of humic acid.Because of the presence of numerous carboxylic an dphenolic -OH functi onal groups, humic substancesar e usually negatively charged in aqueous solutions atneutral to high pH. This charge also varies withphysico-chemical properties of the solution such asionic strength and humic concentrat ion. The theoryproposed by Ghosh and Schnitzer [10] suggests tw oconfigurations fo r h u m i c acid macromolecules: aflexible linear macromolecule at very lo w ionicstrength, high pH, and low solution conce ntration,and a rigid compact spherocolloidal macromoleculeat high ionic strength, low pH, and high solutionconcentration. Membrane fouling by humic sub-stances may be explained in part by the effect ofsolution chemistry on molecular configuration.Membrane pore size, porosity, charge, and hydro-phobic/hydrophi l ic character wil l affect foul ing byorganic matter. For example, adsorption of hum ic acidseems to be enhanced on hydrophobic membranes[11]. Hong an d Elimelech f 12] remark that foul ing onnanotiltration membranes cannot be explained by thesame mechanisms as those used in the case of foulingof ultrafiltration or rnicrofillration membranes. Forcharged molecules transport phenomena like back-diffusion or charge repulsion ar e enhanced in nanofil-tration compared to ultrafiltration an d microfiltration.

The objective of this work is to investigate th e effectof membrane properties on h u m i c acid adsorption.Several variations of the basic cellulose acetate mem-brane were developed by surface modif icat ion of themembrane. Th e effects of modifications of porehydrophobicity, charge, pore size, and porosity onh u m i c acid adsorption were studied. Electrostatican d hydrophobic interact ions between humic acidan d the membrane surface were taken into accountto explain adsorption results.

2. Experimental2.7. Membrane preparation an d structure

Cellulose acetate is the result of esleriiication ofcellulose by acetic acid or anhydr ide. The most com-mon cellulose acetale is cellulose triacetate. This ester

Table 1Polymer mixture composit ion for the membrane syn ihesis

15%A snlkisCA 39.8-3 (77.446%)CTA Eastman 2314-44% Acelyl (22.54%)Average aeelyl content (40%)DioxaneAcetoneMclhanolMaleic acid

results from th e replacement of al l or port ion of thehydroxyl units on the cellulose chain with acetylgroups. The acetyl content affects the hydrophobici tyof the membrane.

The mem branes used in th is research were manu-factured by Separation Systems Technology, SanDiego, CA. They are made from cellulose acetateus ing th e phase inve rsion process, in w hich a p o l y m eris dissol ved in an appropria te solven t and cast as a 0.1-1 mm thick film. The polymer mixture composit ion ispresented in Table 1. A non-solvent is then contactedwith this l iquid film, causing phase separation andprecipitation. Jn our case the film is immersed into acold water gelation bath . The m em bran es are cast onthe surface of a non-woven polyester support, whic h iscommonly used in the industry. The membranes werefabricated on both a laboratory cast ing m ach i n e and acont inuous commercial - type cas t ing mach ine .

2.2. Membrane surface modificationIn order to study th e influence of membrane surface

properties on fouling, several series of modified CAmembranes have been developed.

2.2.7. Annealing: modification of porosity, andpore size

Cellulose acetate has a negative coefficient ofexpansion, and it is possible to increase th e densi tyof the skin layer by ann eali ng the membrane atdifferent temperatures, In our work CA membraneshave been annealed at temperatures varying between40C an d 80C by immersion in a water bath. Theeffect of annealing temperature is then measured interms of porosity and pore size modification.


6/169

C. Combe el al./Journal of Membrane Science 154 (1999) 73-87 75

2.2.2. Hydrolysis and oxidation: modification ofhydrophilic/hydrophobic nature.

Increased hydrophobicity is general ly correlatedwith increased fouling by htunic acid. Hydro phobici tyis directly related to the acelyl content of the mem-brane and can be reduced by breaking the ester l inkbetween cellulose and acetyl groups. Treatment withcaustic effectively reduces acetyl content of the mem-brane below 39.8%, a level that cannot normally beachieved in th e phase inversion process because of theinsolubil i ty of cellulose acetate in common organicsolvents at these acetyl levels. In our research, hydro-lysis was carried out by treating the CA membranewith a solution of 0.1 M NaOH with 1% M eO Hfor varying t imes of 1, 5, and l O m i n . Membranehydrophobicity is estimated by contact angle measure-ments .Surface oxidation can also be used to decrease CAmembrane hydrophobic i ty by i nc reas ing ca rboxyicontent. Oxidation has been carried out with a strongoxidizing agent, 35 % H2O2. Oxida t ion takes place atthe anhydroglucose linkages, an d should affect hydro-phobicity an d pore/surface charge by replacing hydro-xyl groups by acid groups. In our study the oxidat iontimes were 30, 60, and 120 min.

After hydrolys i s an d oxidat ion t reatments , th eCA membranes were annealed by immers ion in awater bath at a fixed temperature between 40C an d80C.2.2.3. Polymer treatment: surface chargemodification

After annea l ing , some of the membranes werecontacted with a dilu te solution of a vinyl acetate/acrylic acid copolymer , called ''Colloid 189". AtpH=4, this anionic polymer adsorbs onto the surfaceof cellulose acetate, bu t can dcsorb above pH=7. Thistreatme nt will increase the hydrophi l ic character of themembrane surface and increase the negative charge.This treatment is called sizing and is known in theindus try to increase the rejection of reverse osmo sismembranes. Sizing is investigated here to see if itaffects adsorptive fouling of CA membranes by h u m i cacid.

After membrane cas t ing an d treatm ent, water per-meation rate was measured al different pressures; thenthe CA membranes were characterized with the dif-ferent methods presented below.

Table 2Solute radius calculated by empir ica l Lcnlscli's equation [ I3|MW (Da) 301)rs (nm) 0.5

2.3. PEG retentionIn order to characterize the membrane pore size,

molecular weight cut-off (MWCO) was determinedusing model solutes. Synthet ic polymers , such aspolyethylene glycol (PEG), dexlran, polyvinylpyrro-lidone are usually chosen because they do not interacts trongly with th e membrane material . In this study w eused a group of four PEG, with molecula r we ight s of300, 600, 3350 an d 10000 Da purchased from Sigma(S t Louis, MO). Lentsch [ 13 J used an empir ical equa-tion to calculate the Stokes radius of PEG. Experi-menta l PEG diffusion coefficient measuremen t s werecombined with the Stokes-Einstein law to give th efol lowing relation:r, = 0.045 M ( (I)where rK is the PEG Stokes radius (nm); M is th emolecular weight (Da). The Stokes radi i calculated forthe PHG used in this s tudy ar e presented in Table 2.

Fil tration tests were performed using a 28.7 cm2diameter UF cell with magne t ic s t i r rcr ( A m i c o n ,Minne tonka , MI). The pressure was appl ied via acompressed nitrogen tank, and was varied betw een103.4 and 344.7 kPa (15 and 50 psi). The concentra-tion of PEG was 10 mg/1 and th e pH=6.5. Membranewater permeabi l i ty was measured before an d aftereach fi l trat ion exper iment in order to check fo r mem-brane fouling by PEG. For each applie d pressure , PEGconcent ra t ions in the retcntate and permeate weremeasured by a TOC analyzer (Phoenix 8000, Dohr-m a n n ) and PEG retentions were calculated using thefo l lowing equa t ion:

CpCo" (2)

Because of concentrat ion polar izat ion occur r ing inth e relenlatc side, an in t r ins ic retent ion mus t bedefined by

(3)


7/169

76 C. Combe et al./Journal of Membrane Science 154 (1999) 73-87where Cm repre sent s the concentration at the mem -brane surface [14J. The film model is used to relate Rand R*:In 1 -RR In

1 -R '~R r (4)

R* is obtained by the intercept of the slope ofln(( 1 -/?)//?) versus flux J. The plot of intrinsic reten-tion versus flux is used to determine th e membraneM WC O .

In order to get the average corresponding poreradius, we used a model of stcric exclusion as pre-sented by Ferry [15"|. When size exclusion is the onlyselectivity phenomena the intrins ic retenti on extra-polated to no flux is equal to

with A = (i' s/rp) where rs is the solute radius an d rp isth e pore radius.The first approximation is to consider that th emembrane pores are all the same size. In this casean average pore radius can be calculated from thevalues of 7?* determined by the retention experimentsan d PEG radius. This method a llows the calculation ofth e pore radius of the homogeneous membraneequivalent to the membrane tested. This is a goodmethod in our case where we want to compare dif-ferent membranes.2.4. AF M characterization

Atomic force microscopy (AFM) is used for directanalysis of membrane surface properties, such as poresize distrib utio n, pore shape, and surface roughness .This techn ique does not require any special prepara -tion of the sample, and the sample can be imaged ineither air or liquid.The CA membranes were imaged under water sinceth e membranes should be studied in their workingenvironment in order to gel the best resolution [ 16, 17 1.Th e wet membranes are studied in contact mode. Inthis case th e probe ti p scans across the sample surfacean d is in direct physic al contact with the sample. It canthen respond to very short-range repuls ive interactio nswith th e sample.The AFM observations were performed on a Topo-metrix TMX 200 1 inst rument . Canti levers, made fromsilicon nitride, are 200 i.im long an d have a t r iangular

shape. The spr ing constant of the narrow canti leve r is/:==0.032 N/m. The tip is also made of silicon nitridean d has a pyramidal shape. Square pieces (5 mmx5 mm) of the membrane were cut and attached to themagnet ic holder using doubled-sided tape. The wetmembrane surface was studied by contacting the sur-face with a drop of ul t rapure water after samplemount ing . Fo r studies in l iquids , a liquid tube scanneris used. The l iquid tube scanner allows simple andr ou t ine i m ag i n g in l iquids, without th e need fo r ancil-lary equipment or a closed chamber [18].

Pore sizes ar e estimated from AFM images an dcalculated by AFM surface analysis software (SPMv.3.06). Line profiles are selected to traverse the AFMimage and pass through the pores. To determine poresize, th e operator must choose points on each heightprofile where the pore is considered to start and end.The pore radius is then th e horizontal distancebetween these points f 18]. Th e porosity of the mem-brane was estimated us ing th e measured average porediameter and the "lakes analysis" included with thesoftware.2.5. Contact angle measurements

The basis contact angle measurement is related toth e three-phase equi l ib r ium that occurs at the contactpoint of the solid/ l iquid/vapor or so l id / l iqu id / l iqu idinterface [ J 9 J . In our case, where th e phases ar emembrane/water/air , th e contact angle is a measureof wetlability of the membrane, i.e., th e capacity ofwater to be adsorbed on the membrane |20|. This ca nbe interpreted as the hydrophobici ty/hydrophi l ic i ty ofthe membrane. A contact angle of 0 corresponds to anideal hydrophi l ic surface.

Th e sessile drop method was used in our research[211. This method, also called (h e water dropletmethod, is based on m easur in g the contact anglebetween a water droplet and the membrane surfaceusing a goniometer. Prior to measurement, mem braneswere stored in petri dishes at 4C. They were soaked inmilli-Q water for 3 h, with three water changes, an dthen soaked in 5% NaCl solutio n for 1 h. All mem-branes were then flushed in the stirred cell in order toremove an y trace organic compounds used by themanufacturer to prevent drying and preserve th emembrane. The membranes were cu t into 20 m m x1.5 mm strips an d mounted in a holder that was placed


8/169

C. Combe ct at/Journal of Membrane Science 154 (1999) 73-87 77

in a glass chamber. A water droplet ( 1 . 8 u J ) wasdeposited on the membrane wi th a microliter syr inge.A light source was placed behind the sample and thel ight was focussed thro ugh a slit. The contact ang lewas measured with the goniometer. Reported contactangles are the average of 7-10 measurements . Theerror on the measurement is equal to 15%.2.6, Streaming potential measurements

In th e case of charged membranes , the charges ar edistributed on the mem brane surface and on the poresurface. When a charged solute passes through amembrane, th e charges located on the pore surfaceca n be important . It is possible to have some informa-t ion about the net charge of the pore surface, andtherefore about th e charge dis tr ibut ion inside th eelectrochemical double layer, by measuring the zctapotentia l . The zeta potential represents th e potentiallocated at the shearing plane between th e compactlayer attached to the pore wall and the mobile diffuselayer. It can be calculated from streaming potentialmeasurements with the Helmholtz-Smoluchowskiequation [22]:

T/K&.E (6)with is the zeta potential, 77 the viscosity, K theconductivity, e0 th e conductivity, e0 th e permit t ivi ty ofvacuum, e r th e relat ive dielectr ic constant of thesolution, /\E the electrical field potential and AP isth e differential pressure.

This equation shows a l inear relation between th edifferential pressure AP applied across the membranean d the electrical field A measured because ofcharged species accu mula tion. This equa tion impliesthere is no overlap of double layers inside the pore,that is the ratio between the pore rad ius and Debyelength is assumed to be very large. The streamingpotential is measured at constant ionic strength,7=0.01 in our experiments. Considering the mem-brane pore radius and the solution velocity throug hth e pores, the pore flow is laminar [11]. It is imp ortan tto recall here that this equ atio n only applies to an idealcase.

Nystrom et al. [23] have described the principle ofs t reaming potential measurement and its application.The instrument constructed for this research consists

of two half -cyl indr ical cells of 42 cm" , connected to aCapsuhel ic differential pressure gage (Dwyer Instru-ment , Michigan City, IND). The accuracy of the gageis 0.34 kPa (0.05 psi). Silve r chlo ride electrodes,placed in each half-cell, measure th e potential dropacross th e membrane. They are made of copper mesh(0.01 m m width opening) cut into disks (17 and 29 mmdiameter) soldered to a copper wire, silver plated andchloridized following th e method of Brown [24J.

The electrical field A is measured with a Fluke 45Dua l Multimeter (Fluke an d Phil l ips , Everet t , W A )characterized by an impute impedance of 1 MSi. Fo reach measurement, the membrane is cut into a circlean d placed on the larger lower electrode that acts as asupport . Trapped ai r bubbles are removed from th ecell. The solution (7=0.01, var ious pH) is constantlyflowing through the cell at about 350 ml /min . Thetemperature is maintained constant at 240.5Cbetween the different exper iments . For each dif feren-tial pressure AP, the conductivi ty K an d th e poten t ia ldrop AE are measured, Streaming potent ia l dataobtained by this method ar e function of physico-chemistry (ionic strength and pH).2.7. Adsorption experiments2.7.L Preparation of the natural organic matter

The organic matter used for this study was a soilhumic acid supplied by Aldrich (Mi lwaukee , WI). Thecleaning procedure was described by Hering andMorel [25] with the addition of a dialysis step, asexpla ined by Hong and Elimelech [12]. The hum ic sa ltwas dissolved in deionized water an d precipitated byth e addition of concentrated HC1. The suspension wascentrifuged, an d fol lowing centrifugation, t h e super-natant was discarded and the precipitate resuspendcdin 3 M HC1. The supernatant gives an inten se color onaddition of potassium thiocyanate solution (KSCN),indica t ing th e presence of Fe(lII) contaminat ion. Theprocedure was repeated 10 t imes with acid an d twicewith a water rinse. This cleaning procedure wasfollowed by dia lysis against water to further purifyth e precipitate. The solid was placed in a regeneratedcellulose dialysis bag hav ing a MWCO of 50000 Da(Fisher Scientif ic) and was dia lyzed against deionizedwater until the conductivity of the solution was lessthan 10 j-iS/cm (more than 7 days). The dialysis solu-t ion w as replaced every day.


9/169

78 C. Combe ct all'Journal of Membrane Science 154 (J999) 73-87

2.7.2. Adsorption experimentsAn 8 mg/1 solution of liuniic acid, prepared in

10~3 M calcium chloride was used in the experiments .The pH was adjusted with 0.1 M HC1. The solutionwas placed in a glass ja r with th e membrane on the top,th e skin side facing th e interior of the jar. A Teflonsheet covered by five sheets of aluminum foil was thenplaced on the membrane. The jar was closed andturned upside down, in order to contact the solutionan d membrane. The a luminum sheets are required toavoid an y leakage from th e jar. The jar is than placedon a gyrator shaker in the dark at 222C and it isshaken during 3 days.

After th e adsorption period, the humic acid con-centration of the supernatant was measured by U Vabsorption (254 nm). However, th e very small differ-ence between the concentration of the solution beforeand after the adsorption experiments could not beresolved by Ihis technique. Hence, th e membranewas rinsed and placed in a 2 ( ) m l of 0.1 M NaOHduring 5 h in order to recover all the adsorbed h u m i cacid by desorption [26]. The concentrat ion of thissolution was then measured by UV absorption ana ly-sis.

The fractional uptake of humic acid in the mem-brane at e qui l ibri um was calculated byfractional uptake = -UV0 - UVf (7)where UV0 is the absorbancc before adsorptionexperiment and UV f is the absorbance after adsorptionexperiment .

Clark and Lucas [18] proposed a model based ondiffusion and adsorption of hum ic acid materials froma mixed solution into a one-dime nsional porous med-ium. They int roduced a interaction parameter, Km,given by

hKm ~ -,a (8)where K is the partition coefficient, /; is the solutiondepth (in the jar), in is the membrane thickness, andwhere a is calculated fromU V 0 -UV,, 1 (9)For a given membrane, Km characterizes th e strengthof interaction between th e humic acid an d th e mem-

brane. The main benefit of the parameter Km overs imply present ing the fractional u p t a k e is that Kmtakes into account the so lut ion volume.

Kinetic experiments were performed with a solutionof 8 mg/1 humic acid prepared in 10 M ca lc iumchloride, pH=5.5, 6.5, 7.5, on the fol lowing mem-branes: Membranes annealed at 40C, 60"C and 80C. Membranes hydrolyzcd during 5 min and then

annealed at 40C, 60C an d 80C. Membranes oxidized during 1 h and then annealed

at 40C, 60C and 80"C. Annealed membranes sized with Colloid 189.

3. Results

3.1. AnnealingTh e water f luxes are presented in Fig. I for the

different anneal ing temperatures . Th e decrease ofmembran e permeabi li ty with increased anneal in gtemperature was expected because of the negativecoefficient of expansion of cellulose acetate. In orderto find if this decrease is due to pore radius and/orporosity modif icat ion, the retention at A/5344.7 kPa(50 psi) for four different PEG fractions was carriedout ; this is presented in Fig. 2. The increase of reten-tion wi th increas ing anne al ing temperature is relatedto a decrease of pore radius . According to theseresults, th e molecular weight cut-off (MWCO) ofmembranes annealed at 40C, 60C an d 80C ar e520, 410 and 300 Da, respectively. Bouchard et al.[4 ] found a similar MWCO around 400 Da for CAnanofi l t ra t ion membranes . Th e pore radii calculatedfrom PE G retent ion and PEG Stokes-Einslein radiiare presented in Table 3. It is also possible to calculateporosity by assuming that the fluid follows Poiseuille'slaw ins ide th e pores. The porosity values are calcu-lated according to the equat ion:

(10)

with 77 is the water viscosity [0.89.10 3 kg/ms(7' 20C)], / is the assumed membrane thickness(0.25 (am), Lp is the water permeability (slope ofmembrane flux versus A/3), an d rp is the pore radiusdetermined by PEG retention exper iments .


10/169

C. Combe el al./Journal, of Membrane Science 154 (1999) 73-87

J (Um 2 h)79

* 4 0 C 60CA 8 0 C

Hydrolysis10min, T=40Cx Oxidation 2h,T=40C

20 30AP (psi) 50

Fig. 1. Water flux for different membranes. Effect of annealing temperature and treatment (hydrolysis and oxidation).

100 1000PEG MW (Da)

10000

Fig. 2. PEG retention curves fo r different membranes. Effect ofanneal ing temperature and treatment (hyd ro lys i s an d oxidat ion) .PEG concentration: 10 mg/1.

Th e value chosen to approximate the membranethickness comes from the literature, where thicknessvalues between 0.05 and 0.3 u,m have been d eterminedfo r ultrafiltration and nanofiltration membranes

[27,28]. Using this method, Table 3 indicates a slightdecrease of porosity with increased annealing tem-perature.

AFM images of the 40C, 60C, and 80C annealedmembranes are presented in Fig. 3(a)-(c). We do notsee a great difference between the membranes. Fromthe AFM images, pore radii can he calculated by theTopometrix software (Table 3) . They follow theexpected order, i.e., a decrease of pore radius withincreasing annealing temperature. However, pore radiicalculated from the AFM images are very large com-pared to those calculated from PEG retention curves. Itseems probable that the AFM software overestimatespore radius because o f the difficulty in determining theedge of the pore on the image. Dietz et al. [29] studi ed(he inf luence of tip shape on AFM images of mem-branes. They proposed a model taking into accountpore and lip shape and showed that AFM measure-ments can overestimate pore diameter by 13-80% inthe case of pores with rounded edges. The tip tends tomeasure (.he pore diameter at the pore entrance, whichleads to over-estimation of pore diameter. The tip size

Table 3M WCO an d pore radius calculated from PE G retent ion, an d porosily, calculated by Poiscuillc's law and AFM softwareMembrane MWCO

(g/mol)Pore ratlins fromrejection ( n m )

Porosi ly fromPoiscuille law (%)

Pore radius fromAFM (nm) Porosity fromAF M (% )

Annealed 40C 520 ! .62Annealed 60QC 41 0 1.56Annealed 8(TC 300 1.57Hydrolyzed 10 min, annealed 40"C 930 1.7Oxidized 2 h, annealed 40'C 830 1.69

0.70.40.21.40.5

13.29.13

11.515

6.23.146.59


11/169

so C. Combe et al./Journal of Membrane Science 154 (1999) 73-87

Wctogre*

44133

221.67

Onni 7,44 nmOw n 22l,67nm 4*3.:nm

(b)

600

Ow n 250 mn 500 nm

498.8

249.4 o.

Qttm249.4 nm 498.8 nm

Fig. 3. AFM images of membranes annealed at: (a) 40C; (b)60C; (c) 80C.

an d shape also affect the measurement of membranesurface roughness. The same observation has beenobtained with track-etched polycarbonate m e m b r a n e s[3() | . In th is case th e pores arc recti l ineal- but w i t h aslight incl ine compared to the actual membrane sur-face. They show that resolut ion in depth depends onlip thickness an d shape, wi th an increase in resolut ionof depth with reduced thickness . The pore radii deter-mined by the AFM software ar e often larger than thosedetermined by cal ibrated molecule retent ion [311.These resul ts show th e c omple x i ty of us ing AFM lodetermine membrane pore radius.

Contact angle data, presented in Table 4, show lhatth e membrane surface is not very hydr ophi l i c . Purecellulose acetate is usually considered a hydroph i l i cmater ia l . However, as explained in Section 2.1 , cel-lulose acetate is commonly cellulose triacetate wherethe hydroxyl units have been replaced by acetylgroups . Compared to pure cel lulose acetate, this mod-i f ication decreases th e free OH-groups, and hence th ehydi 'ophi l ic i ty . Comparable values have been mea-sured by Jucke r and Clar k |32]. Z h a n g et al. [33] alsoshow the di f f i cu l ty in interpret ing contact angle databecause of changes in membrane surface energies andbecause of the variation of contact angle with mem-brane roughness . Annealing has a negligible effect oncontac t angle and hydrophobicity, with only a sl ightincrease in contact angle wi t h decreasing pore size.

The curves for zeta potential versus pH show thatzeta potent ia l does not vary with annea l ing t empera -ture (Fig. 4). Membrane pores ar c negatively chargedin (h e range of pH5.5-8.5, decreasing wi th pH.Therefore, we conclude that anne al ing has no effecton the surface chemistry of the pores.

The results of adsorption experiments show that theamount of h u m i c acid adsorbed is lower when anneal-in g temperature is increased (Table 5). Therefore,adsorption could be related to pore radius, since th eresults presented above show tha t pore ra d ius is theonly paramete r affected by anneal ing [32]. This cou ldmean that adsorption occurs inside the pores. With the40"C and 60 annealed membranes the adsorpt iondecreases wi th increas ing pH. Zcta p otent ial m easure-ments showed an increase of the negative charge ofmembrane pores with increas ing pH. H umic acid isnegatively charged at high pH because of ionizedCOOH groups . The electrostatic repulsion betweenthe negative pore charge and the negative macro-


12/169

C. Combe a al./Journal of Membrane Science 154 (1999) 73-87 81Table 4Effect of anneal ing temperature on average contact an g l e dataA nnea l ing temperature (C)Contact angle ()

4054.18.2

50568.4

6054.7 8.2

7056.38.4

-15 -

Contact angle

5 5.5 6 6.5 7 7.5 8 8.5 9PH

Fig. 4. Variatio n of zela potential with pH for membranes annealedat different temperatures.

molecule leads to a decrease in adsorption of humicacid on the membrane surface [32]. No effect of pH onadsorption is visible with the 80C annealed mem-brane. In this case, th e pore size has a more impor tantinf luence than pore charge.

Contact angles have been measured on fouledmembrane (F ig . 5). The values are close to thosedetermined on clean membranes. The hydrophobic/hydrophi l ic character of the membrane is not reallyaffected by fo ulin g. It is interesting to compare theseresults w i t h those obtained by Jucker and Clark [32]on two hydrophobic ultra filtration membranes. Theyreported a significant decrease of contact angle with

50 -40302010 H

0 -0 0.5 1 1.5 2 2.5 3 3.5 4Adsorbed mass(%)

Fig, 5. Contact angle on memb ran es fouled by h u m i c acid. Effectof annealing temperature and comparison with c l ean memb ran es .

increasing adsorbed mass. In that study, m e m b r a n e swere contacted with Suwannee Rive r hu mi c and fulvicacid. In our experiments , th e hum ic ac id ( A ldr ich) isextracted from the soil. The average mo lecular weigh tof this humic acid as reported by different authors isusually larger than 50000 Da |12]. Typically thehydrophobic i ty of humic substances increases withincreasing molecular weight and decreasing acidity.Hong and Elimelcch [12] found a higher ca rboxyl i cacidity for Suwannee River humic acid than forAldrich hum ic acid. These results suggest that Aldr ich

Table 5Effect of annea l ing temper ature and hydrolysis on adsorption results (K m and % of adsorbed mass)Membrane

Annealed 40r'Annealed 60 Annealed 80"Hydrolyzcd 5Hydrolyzed 5Hydrolyzed 5

CCCm in , annealed 40 l"Cm in , annealed 60Cmin, annealed 80C

pH=5.5Km0.1550.1930.0660.1330.2050.107

% Adsorbed mass2.743.391.192.363.591.91

(2.78,(3.33,(1 .16,(2.41,(3.52,(1-87,

2.68)3.46)1 . 2 1 )2.31)3.66)1.95)

pH=6.5Km0.1450.1850.0760.1070.1370.131

% Adsorbed mass2.57 {2.52,3.25 (3.18,1.36 (1.33,1.91 (1.87,2.43 (2.38,2.33 (2.28,

2.62)3.31)1.39)1 .95)2.48)2.38)

pH=7.5Km0.1060.1330.0910.1230.13!0.143

% Adsorbed mass1 .89(1 .85 ,2.36 (2 .31,1.63 (1 .6 , 12.19 (2 .15 ,2.33 (2.28,2.53 (2.48,

1 .93)2.41).66)2.23)2. 38)2.5S)

H um ic acid concentration^ mg/1, pH5.5, 6.5, an d 7.5, hydrolys is timc=5 m i n .


13/169

82 C. Combe el at. / Journal of Membrane Science 154 (1999) 73-87J (l/

35

3025 -20 -15 -10 -50

0

* Hydrolysis 1 min* Hydrolysis 5 min* Hydrolysis 10 min non hydrolyzed

Fig. 6. Water flux fo r hydrolyzed membranes, Effect of hydrolysistime and annealing temperature

humic acid may be more hydrophobic than SuwanneeRiver humic acid. Therefore, the lack of decline incontact angle observed on fouled membranes may bedue to the greater hydrophobicity of the soil humicacid. These results underline th e difference in beha-vior of these tw o humic acids.3.2. Hydrolysis

Fig. 6 shows tha t hyd rolys is can have a great effectof permeability on the more porous membranes.Exper iments fo r different times of hydrolysis showan increase of water flow of as much as 100% for

membranes annealed at 40C an d 60"'C. This effect isnot visible for the membrane a nnea led at 80C. Th eMWCO of th e hydrolyzed 40C annealed m e m br a nedetermined from PE G retention is estimated lo be930 Da, which corresponds to a pore radius of 0.94 nm(Fig. 2). P orosi ty calc ulated by Poiseuille's la wincreases with hydrolys is , from 3.8% to 4.7%(Table 3). The comparison between pore radius cal-culated from PE G retention and AFM measurementsshows again a higher value for AFM (Table 3).

According to contact angles measu rem ents hydro-lysis has a dramatic effect on hydrophobicity(Table 6). The longer the treatment the smaller thecontact ang le. Treatment by a caustic agent to reduceacetyl content is used in the industry to produce morehydrophilic regenerated cellulose acetate membranes.We have found very low contact angles for a com-mercial regenerated cellulose acetate membrane [32].We do not see a great change of contact angle withannealing temperature, which supports our previousobservations that annealing does not affect pore sur-face chemistry.For 40C an d 60C annealing temperatures, zelapotential of the hydrolyzed membranes decreases inabsolute value an d becomes positive after a certaintime of hydrolysis (>10 min). The data presented inFig. 7 are for an annealing temperature of 60C.Therefore hydrolysis reduces th e pore charge, an dth e mem brane becomes more neutral . Since negat ivemembrane charge is often considered to reduceorganic fouling, hydrolysis may increase adsorptivefouling.

Table 6Contact angle data for membranes after hydrolys is or oxidat ion and ann ealin g at di ffe rent temperatureAnneal ing t emperatureTO40

60

80

Type of t reatment

NoneHydrolysis 1 minHydrolysis 5 minHydrolysis 1 0 minNoneHydrolysis 1 minHydrolys is 5 minHydrolysis 1(1 minNoneHydrolysis 1 minHydro lysis 5 minHydrolysis 10 min

Contact Anneal ingangle () temperature ("Q54.1 4028.223.320.854.7 6030282658.2 8028.318.615.7

Type of t r ea tm en t

N oneOxidat ion 30 minOxidation 1 hOxidat ion 2 hNoneOxidation 30 minOxidat ion I hOxidation 2 hN oneOxidation 30 minOxidation 1 hOxidation 2 h

Contaetangle ( )54.130.530.33054.731.431.731.758.2313131.4


14/169

Zeta potential (mV )C Combe et a!./Journal of Membrane Science 154 (1999) 73-87

J {l/m 2h)

-Hydrolysis 10min- Hydrolysis 5min-Hydrolysis 1min- non hydrolyzerj

7PH

7.5 8.5

Fig. 7. Variat ion of zcta potential with pH Tor hydrolyzei!membranes . Effect of hydrolysis l ime for membranes annealed at60CC.

* Oxidation 30 min* Oxidation 60 min'Oxidation 120 min* nonoxidized

10 20 30 40AP (psi)

50

Fig. 9. Water f l u x for oxidi/.ed membranes. Effect of oxidat ionl im e on membrane annealed at 4() 'JC.

Contact angle90807060504030,2 0 ;

10 -

0

/*

non hydrolized 80Cnon hydrolized 60C

non hydrolized 40C

- Hydrolized 5min, T=40C- Hydrolized 5min, T=60C- Hydrolized Smin, T=BOC

1.5 2 2.5Adsorbed mass (%)

3.5

Fig. 8. Effect of hydrolysis on contact angle. Evolution withannealing temperature an d adsorbed mass.

However, no significant change in humic acidadsorption wa s observed with hydrolys is (Table 5) .Although hydroph obi city decreases with hydrolysis(according to contact angle measurements), we do notobserve any decrease in adsorption. This result isdiscussed in Section 4. Fig. 8 shows that the contactangle increases with the mass of humic acid adsorbed,reaching an asymptotic value similar to the values onthe annealed (unhydrolyzed) membranes.

3.3. OxidationA slight decline of water permeability is observed

after oxidation of the membrane (Fig. 9). Unreportcddata for membranes annealed at 60C an d 80"C alsoshow that oxidation time ha s little effect on membranepermeability: the main effect of treatment by H2O2 isalready visib le after 30 min. Although a decrease ofwater perm eabil ity is observed by com pari ng the 40"Cannealed membranes before and after oxidation, PEGretention experiments show an increase in pore radius(Table 3). A lower porosity value is calculated for theoxidized membrane compared to the membraneannealed at the same temperature without treatment.According to these resulls oxidation leads to some-what ambiguous increase in pore size and decrease inmembrane porosity. In this case also, AFM overesti-mates pore radius.

The con tact angle mea surem ents presented inTable 6 show a significant decrease of contact anglevalues after oxidation. Oxidation times greater than30 min have no effect on contact angle, and thereforeno effect on surface hydrophilicily. By compar ingresults obtained on hydrolyzed and oxidized mem-branes th e ul t imate hydrophi l ic i ty is greater afterhydrolysis than oxidation (Table 6) .Although time of oxidation does no t seem to havean y affect on water permeabil i ty an d membranehydrophobiclty, it s effect on pore charge is visible(Fig. 10). Streaming potential measurements show a


15/169

C. Combe el til./Journal of Membrane Science 154 (1999) 73-87zeta potential (mV)

-10

-15

" Oxidalion SOminOxidation 60 min

"Oxidation 120 minnon oxidized

7PH

7.5 8.5

Fig. 10. Variation of zcta poten t ia l wi th pH for oxidizedmembranes. Effect of oxidation t ime fo r membranes annealed at60C.

more negative pore charge when oxidation timeincreases. This effect is jus t the opposite to whatwas observed fo r hydrolysis. We might, therefore,anticipate that oxidation of cellulose acetate willdecrease fouling by humic acid.However, adsorption experiments results show anincrease of NOM adsorption on the oxidized mem-branes (Table 7). Despite the decrease in membranehydrophobici ty and the more negative pore charge,adsorption of NOM increases on oxidized CA mem-brane surfaces.3.4. Treatment with anwnic polymer

Some of the previously studied membranes weretreated by a polymer, in order to see if there would be

an effect on adsorption of humic acid. Adsorptionexperiments carried out on these membran es show agreat decrease of hum ic acid adsorption (Table 8). Thedata are obtained at pH4 for the treated mem branean d al pH=5.5 fo r untreated membranes. In near ly allcases adsorption on the treated membranes is lowerthan adsorption on untreated membranes. At pH=4th e polymer adsorbs on the membr ane surface, andapparent ly prevents or greatly decreases the ability ofhumic acid to interact with the membrane surface.

4. Discussion

In th e membrane literature, it is comm on to associ-ate increased negative charge and increased hydro-phil ic i ty with decreased adsorptive fouling by na tur a lorganic matter . Yet in this study of adsorption ofh u m i c acid on modified CA membranes, this viewwas not at all clearly supported. For the hydrolyzcdmembranes, the apparent hydrophilicity was increasedaccording to sessile drop contact angle measurements,but the negative charge as measured willi streamingpotential was reduced. There was little change inadsorbabili ty of hum ic acid compared to the unbydro-lyzed mem branes. One could postulate a canc ellatio nof beneficial effects here, i.e., th e posit ive effect ofincreased hydro philic ity was offset by the negat iveeffect of decreased negative charge. On (he other hand,oxidat ion of the membrane surface increased theapparent hydrophilicity and increased th e negativecharge, tw o effects tha t would norm ally be consideredbeneficial; yet, adsorption of humic acid wasincreased.

Table 7Effect of an n ea l i n g temperature and oxidation on adsorption results (Km and % of adsorbed mass)Membrane

AnnealedAnnealedAnnealed

OxidizedOxidi/.edOxidized

40"6080

l h ,I h ,l h ,

CCCannealed 40Cannealed 60DCannealed 80rC

pH-5.5Km0.1550.1930.066

0.2770.3510.339

% Adsorbed mass2.743.391 . 1 9

4.796.005.81

(2.68,(3.32,(1 .17,

(0.47,(5.88,(5.69,

2.79)3.46)1.21)

0.49)6.12)5.93)

pH-6.5Km0.1450.1850.076

0.2290.3630.314

% Adsorbed mass2.57 (2.52,3.25 (3.18,1.36(1 .33 ,

4.00 (3.92,6.19 (6.07,5.40 (5.29,

2.62)3.31)1.39)

4.08)6.31)5.51)

pH=7.5Km0.1060.1330.0910.2470.2170.295

% Adsorbed mass1.89 (1.85, 1.93)2.36 (2.31. 2 .41)1.63 (1.6, 1.66)4.30(4.21,4.4)3.80 (3.72. 3.88)5.09 (4.99. 5.19)

Hu m ic acid concentration8 mg/1, pH=5.5, 6.5, an d 7.5, oxidation t i m c = l h. The values in parenthesis represent the max im um and m i n i m u mvalues of adsorbed mass.


16/169

C. Combe c! at. /Journal of Membrane Science 154 (1999) 73-87 85Table 8Effect of treatment by Colloid 189 on adsorption (Km and % of adsorbed mass)Non-treated with Colloid 189membranesAnnealed 40CHydrolyzed 5 min, annealed 40COxidized ! h, annealed 40"CAnnealed 60CAnnealed 80"CHydrolyzed 5 min, annealed 8(TCOxidized 1 h, annealed SO'^C

Adsorbed mass(% )2.6 (2.55, 2.65)23 (2.25, 2.35)4.8 (4.7, 4.9)3.6 (3.53, 3.67)1.3 (1.27, 1.33)1.8 (1.76, 1.84}5.8 (5.68, 5.92)

Km

0.1550.1330.2770.1930.0660.1070.339

Treated with Colloid189 membranesAnnealed 40"CHydrolyzed 10 min. annealed 40^0Oxidized 4 h, annealed 4()"CAnnealed 60"CAnnealed 80"CHydrolyzed 10 min , annealed 80"COxidized 4 h, annealed 80C

Adsorbed mass(% )2.9 (2.84, 2.96)1-52(1.49, 1.55)3.56 (3.49, 3.63}2.7 (2.64, 2.75)1.92(1 .88 , 1.96)3.49 (3.42, 3.56)3 . 99( 3 . 91 , 4 . 07)

Km

0.1050.0540.1290.0970.0680.1260.145

Humic acid conccntra l ion-8 mg/1, pH4. Comparison with untreated membranes . Tin; values in parenthesis represent th e m a x i m u m an dm in im u m values of adsorbed mass.

To explain these results, we consider more carefu llythe role of calcium in adsorption of humic acid, andthe measurement and interpretation of zeta potential.First, several investigators [12,18,32] have shown thatcalcium has an important role in increasing adsorptionof humic acid on membranes. In the soil scienceliterature, calcium is known to complex with h u m i cmaterials. Hence, calcium is often considered to act asa bridge between th e negatively charged membranesurface and negatively charged carboxyl groups on thehumic acid. Althou gh calcium concentration was notvaried in this study of CA membranes, calcium waspresent in the background electrolyte during adsorp-tion experiments. Reg arding streaming potential mea-surements, it should be recalled that there has not beena clear explanation of the negative charge of manypresumably uncharged m embra ne surfaces like cellu-lose acetate (recall Fig. 4). One of the most likelyexplanations is that the negative charge is due to I headsorption of hydroxyl ions or chloride ions from thestrong electrolyte (0.01 M KC1) used in the streamingpotential measurement [32,34]. Probstein explainsthat the negative charge of hydrocarbons is causedby th e preferential adsorption of anions such as Cl~,over simple cations. This adsorption is enhanced bythe hydrophobic nature of the interface with th e water[35]. A study o f the structu re of adsorbed water incellulose acetate membranes shows the imporlance ofH-bonds between the membrane and the water orhydroxyl groups; th e lower th e acetyl content (com-parable to the more hydrolyzed membranes in ourstudy), the greater the sorption of water and the morehydrophilic the membrane [36]. Our results show thatthe zeta potential is iess negative after hydrolysis, that

is when the membrane hydrophobici ty is decreased.According to Probstein this can be explained by thedecrease in adsorption of chloride ions, due to theincreased hydrophilicity. Therefore, if it is postulatedthat adsorption of humic acid is dominated by chargeinteractions, then as observed, adsorption of humicacid (in the absence of the strong KC1 electrolyte) isno t affected by hydrolysis, since there is really nochange in charged surface groups during hydrolysis -only an apparent change du e to interactions of themembrane surface with chloride or hydroxyl ions. Onth e other hand, for the oxidized membrane s, oxid atio nis a clear mechanism fo r increasing (h e concentrat ionof charged surface carboxyl groups; this leads to thesignificantly increased negative charge measured afteroxidation. If as has been suggested, calcium forms abridge between negative surface groups on the mem-brane an d negative groups on the humic acid, thenoxidation increases the number of adsorption sites,which leads to increased adso rpti on (whi ch is indeedobserved).

5. Conclusion

Several vari ati on s of a basic cellulose acetate mem-brane have been developed by surface modiiication ofIhe membrane . Th e results of this study show thattechniques t ike AF M , streaming potential, contactangle measurements, and PEG retention chartedchanges in membrane surface properties. Annea l inghas a great effect on membrane p ermeabili ty, withoutchanging pore charge or hydrophobicity. Hydrolysisgreatly increases membrane porosity, decreases th e


17/169

86 C. Combe el di./Journal of Membrane Science 154 (1999) 73-87charge of the membranes, and increases hydrophi l i -city, but has little effect on adsorption of humic acid.Oxidation had a small effect on porosity, but signifi-cant ly increased hydrophilicity and the negative mem-brane charge; adsorption of hum ic acid was increasedon the oxidized membranes. Only by pre-adsorplion ofan anionic polymer on the membrane surface isadsorption of humic acid on the membrane surfacedecreased.

Finally, th e results of this work have challengedsome of the conventional wisdom on foul ing of mem-branes by natural organic matter. Normally, decreasedmembrane hydrophobici ty an d increased (negalivc)charge are assumed to reduce fouling by negativelycharged organic foulants. W e found exactly thesedesirable characteristics for the oxidized membranesstudied here, nevertheless, adsorplive fouling byhumic acid was actually increased on (he oxidizedmembranes. We have attempted to explain this beha-vior by a careful consideration of charge-relatedadsorption and the role of calcium in adsorption ofhumic acid.

6. NomenclatureCm solute concentration at the surface of th e

membrane (mol/l or mol/m)Cp concentration in the permeate (mol/l)C0 solute concentration in the feed (mol/l)AE electric field potential (V)J flow density, called the total volume flow

rate ("im/s)k mass transfer coefficient (m/s)Km adsorption parameter (m)/ membrane thickness (m )Lp membrane hydra ulic permeability (m s/Kg)M molecular weight (Da)A.P pressure gradient (Pa)rp pore radius (m )rK Stokes solute radius (m)R* intrinsic retention coefficientR observed retention coefficientGreek symbolse membrane porosity: empty volume/total

volume

e


18/169

C Combe el at./Journal of Membrane Science 154 (1999) 73-87 87[12] S. Hong, M. Elimclecli, Chemical and physical aspects of

NO M fouling on nano f i l t r a l i o n membranes, .1. Mcmbr. Sci.132(1997) 159.

[13] S. Lenlsch, P. Aimar, J.L. Orozco, Separation albnmin-PEG:transmission of PEG through u l t r a f i l t r a t i o n membranes ,Biotechnol. Bioeng. 41 (1993) 1039.[14] W.M. Deen, Hindered transport of large molecules in liquid-f i l l ed pores, AIChE J. 33 (1987) 1409.

[15] J.D. Ferry, Ul l r a f i l t e r membranes an d ultra filtration, Chem.Rev. 18 (1936)373 .

[16] P. Dielz, P. Hansma, K. Herrmann, O. Inacker, H. Lehmann,Atomic force microscopy in air and under water, Ultramicro-scopy 35 (1991) 155.[17] A. Bessiercs, Elude des proprieles fonetionncllcs el slructur-ales des membranes synlhctiques pa r retention de moleculescalibrees et microscopies a champ proche, These dc doctoral,Unlversite Paul Sabatier, Toulouse, France, 1994,f!8] M.M. Clark, P. Lucas, Diffusion an d parti t ioning of humicacid in a porous ultnifillralion membrane, J. Mcmbr. Sci. 143(1998) 13.|!9] J.D. Andrade, L.M. Smith, D.E. Grcgonis, Th e contact angleand interface energetics, in: J.D. Andrade (Ed.). Surfaces andlnterfacial Aspects of Biomcdical Polymers, Plenum Press,New York, 1985, p. 249.

[20] P.C. Hiemenz, in : JJ . Lagowski (Ed.), Principles of Colloidand Surface Chemistry, 2nd cd., Dekkcr, New York, 1986,p. 287.

[21] J.D. Chen, N. Wada, Edge profiles and dynamic contactangles of a spreading drop, J. Colloid Interface Sci. 148(1992) 207.

[22] RJ. Hunter, in: R.H. Ollcrwiil, R.L. Rowcll (Eds.), ZetaPotential in Colloid Science, Principle and Applicat ions ,Academic Press, London. 198i.

123] M. Nystrom, M. Lindstrom, E. Mathiasson, Streamingpotential as a tool in th e characterization of ultrafiltrationmembranes, Colloids Surf. 36 (1989) 297.

[24] A.S. Brown, A type of silver chloride electrode sui tablefor use in dilute solutio ns, J. Am. Chem. Soc. 56 (1934)646.

[25] J.G. Hering, F.M.M. Morel, Humic acid complexation ofcalcium an d copper. Environ. Sci. Techno!. 22(10) (1988)1234.

|26] K.L. Jones, PhD thesis, the John Hopk ins Univ ersity ,Baltimore, MD , 1996.[27] R. Rilcy, .I.O. Gardner, U. Merten. Cellulose acetate

membranes : electron microscopy of structure, Science143(3608) (1964 ) SOI.[28] A. Braghetla, F.A. DiGiano, W.P. Ball, Nanoli l t rat ion of

natural organic matter: pH and ionic strength effects, J.Environ. Eng. (1997) 628.

129] P. Dietz. P.K. Hansm a, 0. Inack er, H.-D. L ehmann . K.-H.Herrmann, Surface pore structures of micro- an d ultrafiltra-tion membranes imaged with [he atomic force microscope,J. Mcm br. Sci. 65 (1992) 101.

[30J A. Chahboun, R. Coralger. F. Ajustron, J. Beauvillain, P.Aimar. V. Sanchez, Co mpara tive s tudy of micro- andul t raf i l [ ra t ion membranes using STM, AF M and SEMtechniques, Ultramicroscopy 41 (1992) 235.

|31] A. Bessicres. M. Meireles, R. Coratger, J. Beauvillain, V.Sanchez, Investigations of surface properties of polymericmembranes by near field microscopy, J. Membr. Sci. 109(1996)271.[32] C. Juckcr , M.M. Clark, Adsorpt ion of i u j u a t i c lu i in icsubslances on hydrophobic ultrailItralion membranes, J .Membr. Sci. 97(1994) 37.

[33] W. Zhang , M. Wahlgrcen, B. Sivik, Membra ne characteri za-tion by the contact angle technique. II . Characterization ofUF-mcmbranes an d comparison between the captive ai rbubble an d sessile drop as methods to obtain water contactangles, Desalination 72 (1989) 263.[34] M. Nystrom, P. Jarvinen, Modification of polysuj foneultrafillration membranes with UV irradiation an d hydro-philicily increasing agents, J. Membr. Sci. 60 (1991) 275.

[35J R.F. Probslein, Physicochcmical Hydrodynamics : An Intro-duction, 2nd ed., Wiley, New York, 1994, p. 212.

[36] W.A.P. Luck, K. Rangsr iwatananon, Th e s t ructure ofadsorbed water in cellulose acetate membranes. ColloidI 'olym. Sci. 275 (1997) 1018.


19/169

journal ofMEMBRANESCIENCEELSEVIER Journal of Membrane Science 143 (1998) 13-25

Diffusion and partitioning of humic acid in aporous ultrafiltration membrane

Mark M. Clark*, Pascale LucasDepartment of Civil and Environmental Engineering, University of Illinois, 205 N. Malhews Avenue, Urbuna, IL 61801, USA

Received 23 June 1997; received in revised form 8 December 1997; accepted 9 December 1997

AbstractSurface imag in g of wet polysu l fonc (PM30) ultrafiltration membranes with atomic force microscopy indicates a relatively

smooth but porous surface in the aqueous environment. The average pore radius was 9.70.8 nm , and the average porositywas 10%. Therefore, for the case of adsorption of h u mic acid on the membrane (J. Membr. Sci. 97 (1994) 37-52), wehypothesize here that the transport of humic materials in th e membrane can be modeled as one-dimensional d i f f u s i o n andlinear partitioning in a porous medium. An interaction parameter was developed using the model to characterize the strengthof in teract ion between the h u mic acid and (h e membrane as a function of pH and calc ium concentration. The mass of h u micacid adsorbed and the calculated interaction parameter were found to increase w i t h decreasing pH and increasing calciumconcentration; hence, the s trength of interaction between h u mic acid and membrane is now simply parameterized. The kineticsof adsorption of humic acid was also modeled, and fitted diffusion time scales showed that the h u mic acid diffusion coeff ic ientincreased wi th decreasing pH and increasing calcium concentra t ion, which is consis tent w i t h a compaction of th e h u m i cmolecu le . These effects were also consistent with a previous study of h u mic acid diffusion coefficients in activated carbon(J. Colloid Interface Sci. 110 (1) (1986) 149-164). 1998 Elsevier Science B.V.Keywords: Diffusion; Fouling; Ultrafiltration; Atomic force microscopy; Adsorption

1. IntroductionClark and Jucker [1J and Jucker and Clark [2]studied the adsorption of aquatic humic substanceson PM30 (polysulfone, 30,000 Da MWCO) andXM50 (poly (aery lonitrile-co-vinyl chloride),50,000 Da MWCO) ultrafiltration membranes in ashaken cell without permeation. It was found thatthe degree of adsorption of humic and fulvic acid*Corrcsponding author. Fax: 217 333-9464; e-mail:

[email protected]

depended on several physlcochemical factors. First,adsorbed mass wa s considerably higher on the moreporous of the two membranes (PM30). Measurementsof apparent pore zeta potential for both membranesshowed that th e initially negative pore zeta potentialincreased (became less negative) rather quickly withamou nt of mass adsorbed, but then plateaued as humicmaterials continued to adsorb. This was interpreted tomean that pore adsorption sites (or other roughnessfeatures) adsorbed humic materials quickly; a weakerand longer-term adsorption continued either on lowerenergy sites or as a second layer deposited on the first.

0376-7388/98/S19.00 1998 Elsevier Science B.V. Al l rights reserved.PlI 3 0 3 7 6 - 7 3 8 8 ( 9 7 ) 0 0 3 3 2 - 3


20/169

14 MM . Clark, P. Lucas/Journal of Membrane Science 143 (1998) 13-25

Th e chemistry of th e system also strongly affectedadsorption of humic materials. As pH was loweredfrom 6 to 4.8, mass adsorbed increased significantly.Jucker and Clark [2] showed usi ng octanol/waterpartition coefficients that the solubil i ty and hydrophi-licity of hum ic substances decreases from pH 6 to 4,which is due likely to the protonation of phenolic an dcarboxylic hydroxyl groups on the humic materials.St reaming potential measurement showed that thepore zeta potential of virgin membranes also increasessignificantly (becomes less negative) fo r that samepH change. So th e following picture emerges: Thehumic m aterials have a net negative charge at neut ralpH; as the pH is lowered, the pore zela potentialincreases (becomes less negative), (he negative chargeof the humic materials is neutralized, and the electro-static barrier to adsorption of h u m i c materials isdecreased. Jones an d O'Melia [3] found a parallelincrease in adsorption of Suwanncc humic acid on aregenerated cellulose acetate UF m embran e as pH waslowered.

Calcium is an important environmental com ponent ,and its complexation by hum ic materials has been longrecognized [4]. Calcium is also implicated in foulingof hydrophobic membranes [2,5,6J. In Jucker andClark's [2] work, calcium was also found to signifi-cantly enhance adsorption of hum ic mater ials . X-rayPhotoelectron Spectroscopy (XPS) measurements onth e PM30 membrane surface after adsorption sug-gested th at the calcium was complexed w ith the h u m i cacid. The role of calcium was interpreted as either abridge between th e negative surface an d hum ic mate-rials, or as a means of altering the ionic strength,hence, compacting or coiling the humic polym er chain[7]. A smaller polymer conformation could alsopermit a more dense adsorbed layer [8].

Jucker an d Clark [2] did not attempt to model th ekinetics of humic acid adsorption on the membranes,nor did they parameter ize th e strength of the humic-membrane interaction. In this follow-up to their work,we modeJ adsorption (emetics assuming a linear iso-therm, and consider that adsorption proceeds in amanner s imilar to one-dimensional diffusion an d par-t i t ioning in a porous medium [9J. We focus in thispaper on the adsorption of hum ic acid on the PM30membrane, a system for which excellent kinetic dataare available [10]. Comparison of humic acid mole-cular sizes with pore size statistics gathered with

atomic force m icroscopy im agin g of th e PM30 mem-brane support th e modeling of humic acid transport asdiffusion an d par t i t ioning in a porous medium. Aparameter related to the classical parti t ion coeff ic ientis developed to character ize th e strength of the h u m i cac id-membrane interaction, whi le th e model permi t sus to interpret the effects of solution chem istry on th ehumic acid diffusion coefficient.

2. ExperimentalThe polysulfone PM30 membrane had a nomina l

molecular-weight cutoff of 30 kDa. The PM30 mem-branes were cleaned in a multistep process with Mill i -Q water an d 5% NaCl [2]. The cleaned membraneswere stored skin side down in Mil l i - Q water in a dark6C chamber, and membranes were never allowed todry prior to any analyses.

A Topometrix, TMX 2000 Explorer, atomic forcemicroscope was used to acquire images of the PM30membrane surface in the contac t mode. Square pieces(5 mm x 5 mm) of the membrane were cut andattached to the magnetic holder using double-sidedtape. The wet membrane surface was studied bycontac t ing th e surface with a drop of ul l rapure waterafter sample mount ing . For the studies in l iquids , aliquid tube scanner was used. The l iquid tube scannerallows simple imaging in l iquids, w ithout anc i l l a r yequipment or closed chambers .The canti levers were 200 um long and of t r iangu larshape; the spring con stant of the cant ileve r used herewas &0.032 N/m. The lips were made from siliconnitr ide, and had a pyramidal shape. Scans were madeon more or less rand om ly selected me mb rane areas of1000 x 1000, 500 x 500, and 200 x 200 nm .

The manu facturer 's software al lowed computat ionof various statistics related to the surface roughness,inc lud ing th e average height , ar i thmetic mean de via-tion from th e average height, the RMS roughnessheight, the bearing ratio, and the ratio of surface areato projected area. The roughness statistics reported inthe next section were based on two scanned areas of500 x 500 nm. Pore size was also determined. In th isanalysis, line profiles arc selected to traverse the AFMimage an d pass through th e pores. To determine poresize, the operator mu st choose the poin t on each h e i g h tprofile where the pore is considered to start and end. In


21/169

MA/. Clark, P. Lucas /Journal of Membrane Science 143 (1998) 13-25 15the analysis reported here, the pore radius is thensimply th e distance between tw o cursors, one at thebottom of the pore profile, and the other at the verticallocation corresponding to the local 50% bearing ratio.The statistics on pore radi us reported in the nextsection were based on measurement of more than74 pores on 5 images of 500 x 500 nm and200 x 200 nm membrane areas. The number of poresand porosity of the membrane were estimated usingthe "lakes" analysis included with the software. Inthis analysis, tw o 500 x 500 nm and one200 x 200 nm areas were analyzed using a bearingratio of 75%.Th e reader is referred to the previous paper [2] for athorough discussion of the adsorption experiments.Solutions of either 8 or 25 mg/1 humic aeid solutionswere made up using reference standards from theInternational Humic Substances Society (USA) dis-solved in distilled-deionized water. Solutions werebuffered with 10 ~3 M sodium phosphate dibasic,and pH was adjusted with IO"1 M HC1. Calciumconcentration was varied by adding the appro priateamount of calcium chloride. A 250 ml solution ofhumic acid was poured into a 473 ml (1 6 oz ) clearmedium round glass jar, and a cleaned membrane wasplaced skin side down on the top of the glass bottle,which nearly perfectly supported the membrane alongits outside edge. A Teflon sheet was then placed overthe top of the jar, followed by several sheets ofaluminum foil to ensure a good final seal betweenthe glass threads and the plastic cap. After the plasticcap was screwed on, the jar w as inverted an d placed ona gyratory shaker used extensively in carbon adsorp-tion studies in our laboratories. Therefore, in ouradsorption experiments, humic substances ar e con-tacted with th e skin side of the membrane under non-permeation condit ions, where the solution over lyin gthe membrane is effectively mixed over relativelyshort time scales. Blank solutions (everything thesame except no membrane), were used to correctfor an y humic acid adsorbed to the glass (negligible).Every day, the shaker was stopped, and 5 ml of solu-tion wa s withdrawn for ultraviolet (UV) absorptionanalysis (254 nm) and computation of depletion ofhumic acid from the solution. Kinetic experimentswere performed with new membranes after eachchange in chemical conditions. The temperature foral l adsorption experiments was 222C, and all long-

term adsorption and kinetics experiments were con-ducted in a dark chamber.

3. Model developmentBornzin and Miller [1IJ and Jones and O'Mclia [3]

modeled adsorption of organic compoun ds on vario uspolymer an d membrane surfaces as diffusion from anunmixed l iquid phase to a H at and non-porous surface,so-called "static adsorption". In their systems, th econcentration gradient at the membrane surfacedeclines over time as more organic material diffusesto the membrane surface. However, as microscopicanalyses show (Section 4), the surface of PM30 UFmembranes studied here is quite porous. In addition,like most adsorption experiments, the adsorptionexperiments of Jucker an d Clark [ 2 j were performedwith a mixed liquid phase. Therefore, in this work, wemodel adsorption of humic materials from a mixedsolution into a one-dimensional porous medium,where the concentration boundary condition at themembrane surface is determined by the overall deple-tion of organics from solution. Diffusion in a porousmedium can be described with th e one-dimensionaldiffusion equation,

(1)where D e(T is interpreted as an effective diffusioncoefficient an d x is perpendicu lar to the m embran esurface. The boundary conditions for diffusion from awell-mixed reservoir are (1) C=Q throughout themembrane at time=0, and (2) the flux of materialinto the membrane results in a uniform decrease inconcentration of humic materials in the well-mixedreservoir of depth // ,

Oc _ dcat oxCrank [12] provides an analytical solution tothis problem in terms of the mass adsorbed at some

t ime t, M,, divided by the mass adsorbed at equ ili-brium, M c:M '- ex p (3)Here a is related to the fractional uptake of adsorbate


22/169

16 M.M, Chirk, P. IMCCIS/Journal of Membrane Science 143 (1998) 13-25

in the membrane at equilibrium,fractional uptake of

humic materials by membrane (4 )at equil ibriu m

The qn are the roots of the generating funct ion,tan 9,, = aqn (5)an d r is the diffusion lime scale,

(6)D,effwhere m is the effective membrane thickness fo radsorption. Th e model described by Crank [12Jassumes a linear isotherm, which was generallyobserved in the experiments of Juckcr and Clark[2]. Th e linear partition coefficient K is related toa, the solution depth h, and the mem brane thickness asK = h (7)

Since m is difficult to characterize in adsorptionexperime nts, we will later use a parameter whi ch is theproduct of K and m,Km - *a (8)

Th e product of AT and m is subsequently referred toas the "interaction parameter". The equil ibr iu m frac-tional uptake in Eq. (4 ) was determined from th e UVmeasurement as ,

fractional uptake ofi .. i r. (, UV n-U V fhumic materials by membrane U V | 'at equilibrium (9)

where UVy is the initial UV absorbance of the humicacid solution, and UV/- is the final UV measurement atthe end of the kinetics experiments. For a value of adetermined from Eq. (9) and the experiments, Eq. (5)was plotted and the local roots were determined usingthe secant method (Mathcad, Mathsoft, Cambridge,MA). For non-zero values of M,/MC, the series inEq. (3) converged quickly . The best fit of Eq. (3) tothe experimental data was found by adjusting theparameter r to minimize the variance between mea-sured an d modeled Mt/Me.

The summary, ou r model assumes that th e solutionabove th e membrane is well mixed over lime scales

which are short relative to diffusion and adsorption inthe membrane itself . I t is reasonable to expect lhatt ime scales fo r m i x i n g in the overlying fluid arc of theorder of seconds to minutes, whereas it is observed inthe following section that time scales for uptake ofhumic acid by the membranes are of the order of days.Hence in our modeling, mass transfer limitation isassumed to occur only within th e membrane itself(i.e., no l iquid-phase mass transfer l imitation).

4. Results

4.1. AFM images an d surface statistics fo r PM30membrane

Fig. 1 shows a scan of a 500 nm square section ofthe surface of a new, wet, PM30 membrane. Theaverage of the roughness heights is Zavc=(). 12 nm,with a m axim um roughness height of Zmax=0.23 nm .The arithmetic mean of the deviation in height fromthe average height is 7?,,0.04 nm, while the RMSheight is Zrnis:=0.05 nm. The ratio of surface area toprojected area is A/ AP1.004. Comparison of theseroughness statistics with other studies is complicatedby the fact that different AFM manufacturers do notalways calculate equivalent roughness statistics, andbecause to our knowledge, there are no other AFMstudies of the PM30 membrane. Bowen et al. [13]studied a 25,000 MWCO polyethersul fonc membranewith non-contact AFM. Although they did not reportquantitative roughness statistics, th e apparent Zmaxvalues on a 100 x 100 nm scan were very sim ilar tothose reported in this work. OurZave , Zmax , 7?a an d Znnsvalues are approximately an order of magnitude lessthan comparable statistics from tapping mode AFManalysis of dry PPO gas separation membranes castfrom chloroform and TCE [14]. Non-contact AFMscans of comparable areas of dry Nucleopore poly-carbonate (N0015) UF m emb ranes also indicatedroughness statistics about an order of m ag n i t u d egreater than the PM30 results presented here [15].There is evidence which suggests that roughnessstatistics ar e influenced by the area of the scannedsample. On scans of larger 1 p.m square areas of th ePM30 membrane, some roughness statistics werefound to increase (Z,lve=0.20 nm, Zmoix=0.37 nm ,tf a=0.06 nm). Pradanos et al, [151 found roughly an


23/169

MM. Clark, P. Lucas/Journal of Membrane Science 143 (1998) 13-25

500 nm

17

0.23 nm

0 nm250 nm

0 nmT0 nm 250 nm 500 nmFig. 1. Top view of 500 x 500 nm section of wet PM30 membrane using atomic force microscopy.

order of magnitude larger R^ values on 1 nm . squareareas than on 200 nm square areas of the Nueleoporemembrane. On AFM scans of 5 u,m square pieces ofpolypiperazineamide fully aromatic polyam ide nano-and brackish water filtration membranes, Safarik [16]found a very grainy surface with much greater rough-ness statistics than those for the PM30 UF membranestudied here. The variation in roughness statistics withscanned area is probably caused by two factors. First,statistically speaking, a larger sampling area is morel ikely to sample larger roughness elements than asmal ler area. Second, we have found that there aredifferent scales of roughness on a given membranesurface, an d scans of progressively larger areas lend toencounter ever larger scales of roughn ess. Finally, theissue of the effect of wet and dry imaging environ-ments on roughness values is beginning to receiveattention in the literature; one study of a 40 kDaMWCO sulfonated polysulfone membrane indica tessignificantly smaller roughness elements in theaqueous environment than in the dry environment,

an effect which may be related to polymer dehydra t ion[17].

Results for pore size analysis for the PM30 mem-brane ar e shown in Fig. 2. The average pore radius ofthe PM30 membrane was found to be 9.690.83 nm(95% confidence interval). The porosity of the PM30membrane was found to be 10%, with 344 pores periim2. Fane et al. [18] assumed pore diameters for thePM30 membrane of 3.5-4 nm in the i r hydr au l ic esti-mation of PM30 porosity. Using TEM , Fane and Fell[19] fou nd pore radii of 6 nm and porosities of 2% forth e PM30 membrane, whi le using FESEM, Kim et al.[20J found average pore radii of 2 nm and averageporosity of 5.9% fo r (h e same membrane. All techni-ques used to measure pore size and porosity sufferfrom certain limitations an d artifacts. For example ,both electron an d atomic force microscopic techni-ques can not discriminate between true pores an ddead-end pores or other pore-like features. The micro-scopic techniques are also not opt imized to determineeffective pore diameters whi ch mig ht resu lt from


24/169


1 3

if

5

n

7

??

Is

i:

9

If

11 13

H1 11 , ii , i ,15 17 19 21 23 25

pore radius (nm)Fig. 2. Pore size dis t r ibut ion of PM30 membrane.

necking down o f a pore at lower levels in the relativelydense skin layer. H ydraulic technique s for estima tingporosity must first determine (or assume) the capil larylength, diameter, an d tortuosity f21]. TEM andFESEM are done on dry membranes, and pores onth e dehydrated membrane may be smaller than on ahydrated membrane.

The measurements of pore size reported here areconsistent with other recent AFM measurements . Forexample, Dietz et al . [22] found 350 pores per u,m2 ona 10,000 MWCO polysulfone membrane. Bowen etal.'s [13] study of dry 25,000 MWCO polyethersul-fone membranes indicated average pore diameters of5.1. nm . Fritzsche et al. [23] studied 30,000 MWCOpolyethersulfone membranes and found pore dia-meters in the range of 15-25nm. Wetted 30,000M W C O polyacrylonitrile membranes were studiedby Fritzsche et al. [24] who found pore diametersin the 12-20 nm range.Based on small-angle X-ray scattering at high ionicstrength (/~10~'), th e radius of gyration of theSuwannee humic acid molecules is given as1.11 nm [25], with a molecular weight range of 5-l O k D a [26]. Using membrane separation techniques(Amicon YM series UF membranes) at lower ionicstrengths (/~5xlO~4), Jucker and Clark [2] foundsomewhat larger apparent molecular weight for the

Suwannee hum ic acid of 0.5-30 kDa. (Approxim ately96% of the Suwa nnee humic acid passed throug h th e30 kDa M WC O YM30 membrane.) The differences inapparent molecular weights at such widely differentionic strength is expected, since hu mi c acid confor-mation is known to depend on variables l ike pH an dionic strength |7|. Considering th e above information ,it is clear that th e humic acid can pass into th e PM30pores. This provides jus t i f icat ion for our model ing ofhumic acid adsorption as diffusion an d par t i t ion ing ina porous medium.4.2. Interaction between humic acid and PM30membranes

A summary of the appropr iate exper im ental condi-t ions an d measured parameters ar e shown in Table 1.Note from th e table th e general increase in f ract ionaluptake of h u m i c acid wi th decreasing pH; as discussedin the Section 1, these results are consistent with adecreasing electrostatic barr ier between humic acidan d membrane with lowered pH . Hum ic acid contain sapproximately 4. 9 ineq/g of carbo xylic and 2.9 meq/gof phenolic hydroxyl groups [27]. The carboxylicgroups are deprotonated at neutral pH. Hence, asthe pH is lowered, these groups become prolonaledand the overall negative charge of the molecu les is


25/169

MM. Clark, P. Lucas/Journal of Membrane Science 143 (1998) 13-25Table 1Experimental conditions, and measured and filled model para-meters

C0

88882525

PH6.04.86.04.86.04.8

[Ca]

11113434

180ISO

Fractionaluptake0.120.290.320.390.490.54

a

7.02.52.11.61.00.85

r(d)2.989.48.17.67.8

Km(cm)0.972.93.34.56.88.3

decreased. This decreasing solubility of the humicacid with decreasing pH was verified through mea-surement of octanol-water partition coefficients forthe humic acid. As the pH was lowered from 6 to 4, thepartition coefficient increased significantly, indicatingthat the hum ic materials become less water so luble atthe lower pH [1 ]. Jucker and Clark [2J also found anincreased (less negative) pore zeta po ten tial as the pHwas lowered over the same pH range. Hence the neteffect of the pH decrease is a decreased electrostaticbarrier to adsorption of humic acid. A similar inter-pretation of th e effect of pH was employed by Mal-thiasson [28] for protein (BSA) adsorption onpolysulfone membrane, and by Braghetta [29] forSuwannee River humic material adsorption on a sul-fonaled polysulfone nanofiltralion membrane.

Th e increase in adsorption of humic acid withincreasing ca lc ium concentration is consistent withseveral possible mechanisms. First, it is well knownfrom soil science that the humic materials complex orchelate metal ions such as calci um [4]. Metal ions arethought to attach to negatively charged groups likedeprotonated carboxyl groups. This can result inexpulsion of hydration water, coiling of the molecule,and a reduction in charge. The decreased charge of thehumic acid may allow for increased adsorptionthrough a net decrease in the electrostatic barrierbetween the humic acid and the membrane. In addi-tion, if the molecular conformation is indeed smaller,this would allow for a more dense packing of humicacid during physisorption [8j. The other model for theeffect of calcium is simply an ionic strength effect; theincrease in ionic strength with added calcium chloridedecreases th e shielding between negatively chargedgroups on the humic material. This allows a smaller

molecular conformation and a more dense adsorbedlayer.The main parameter characterizing the strength of

the humic acid-membrane interaction in our model isthe partition coefficient, K. To calculate this param-eter, both th e membrane thickness and the solutiondepth are required. Although the depth of solutionabove th e membrane was known (h=l.\, th ethickness of the membrane m - essentially the thick-ness of the diffusion domain - is a more difficultparameter to specify because of the asymmetry of themembrane. There are various estimates and measure-ments of the thickness of asymmet ric membran es andskin layers, but even with accurate estimates of suchthicknesses, th e effective diffusion thickness wouldstill be unknown in our experiments; therefore, wehave preferred the more accessible calculation of theproduct of the membrane thickness and the partitioncoefficient, Km . For lack of a better name, we call itth e "interaction parameter". F or experiments wit h thesame membrane (i.e., constant in) it should be linearlyrelated to the partition coefficient. Hence, the inter-action parameter offers a direct comparison of thestrength of interaction between the humic acid andPM30 membrane under different chemical conditions.As seen in Fig. 3, the interaction parameter increases

ECDcn(Xc

1 - rt-0 -

50 100 150 200

Calcium Concentration, mg/LTig. 3. Effect of pH and calcium concentration on interactionbetween l iumic acid an d PM30 membrane.


26/169


significantly as the pH decreases from 6.0 to 4.8 and asth e calcium concentration increases from 1 1 to180mg/l. Hence the interaction parameter parallelsan d parameterizes the strength of the humic acid-membrane interaction. Since it is based on an acceptedadsorption model fo r diffusion an d parti t ioning in aporous medium, it offers a new and logical parameterwith which to characterize membrane-organic inter-actions.4.3. Adsorption kinetics

Figs. 4-6 are plots of the normalized adsorbed mass(Mi/Mc) vs. time for six different condition s of pH andcalcium concentration. For all experiments but on e

PM30, CO - 8, pH - 6.0, Ca - 111.0 -

PM30, CO - 8, pH - 6.0, Ca - 341.0

0.0 11-

0 1 2 3 4 5 B(a) Time, days

PM30, CO - 8, pH - 4.8, Ca - 111.0 -r

Jatanod

0.0 -n0 2 4 6 a 10 12 14 16

(b) Time, daysFig. 4. Kin e t ic s o f h u m ic acid adsorplion on PM30 membrane; (a )pH 6.0, and (b) pH 4.8. Cn-8 mg/1 and [Ca]=l I mg/l .

0 2 4 6 B 10 12 14 16(a) Time, days

PM30, CO - 8, pH - 4.8, Ca - 341.0

datnmocal fi .

(b ) 0 2 4 6 B 10 1? 14 16Time, daysFig. 5. Kinetics of h u m i c acid adsorption on PM 3 0 membrane: (a )pH 6.0, an d (b) pH 4.8. C0=8 mg/l an d [Cal^34 m g/ l .

which was terminated prematurely - experiment 1 inTable 1, and Fig. 4(a) - the time scale of the adsorp-tion process seems to be roughly of the same order ofmagnitude (Table 1). The model approximations ofth e kinetic data for experiments 1-6 are also shown inFigs. 4-6. In most cases, the fit of the model to thekinetic data is quite reasonable. Th e model fit t ingprocedure yields apparent diffusion l ime constants ,T, and these are indicated in Table 1. Except for theprematurely termin ated first experiment, the adsorp-tion time scales are fairly lightly clustered, with anaverage value of 8.2 d, and standard deviation of 0.7 d;therefore, the adsorption time constants do not seem tobe clearly correlated to chemical conditions. Further,the reasonable model f i t s suggest that the adsorption ofhumic acid from th e mixed solution is indeed diffu-sion-limited in the membrane, where the intensity of


27/169

MM. Clark, R Lucas/Journal of Membrane Science 143 (1998) 13-25 21PM30, CO - 25, pH - 6.0, Ca - 18

PM30, CO = 25, pH - 4,8, Ca - 18'1.0 -

D /

7

2

1

'"TDD

1

datamode

1

fit

'0 2 4 S 8 10 12

(b) Time, daysFig. 6. Kinetics of humic acid adsorption on PM30 membrane: (a)pH 6.0, and (b) pH 4.8. C0=25 mg/1 and [CaJ=180 mg/1.

th e interaction between humic acid an d membranesurface is captured by the interaction parameter devel-oped here.

5. Discussion

Studies of the fouling of membranes by naturallyoccurring organic comp ounds have been hampe red bylack of a quantitative understanding of the roles ofconvection, diffusion, and adsorption phenomena[30]. The experiments an d model presented here focuson the diffusion an d adsorption aspects of the mem-brane fouling problem. The effective membrane thick-ness for diffusion was not known in the present study,and it will continue to be a difficult parameter to

characterize for asymmetric membranes . However,we can at least say for the same membrane that th einteraction parameter developed here should be sen-sitive to variations in hum ic mate r ia l -membrane inter-actions caused by var iations in system c hemistry. Thisstudy showed marked increases in th e in teract ionparameter with increasing calcium concentrationan d decreasing pH.

One can get an idea about th e magnitude of thepartit ion coefficient by subst i tu t ing in to Eq. (7) anest imated effective membrane th ickne ss in. Fo r exam-ple, we might consider that th e m i n i m u m effect ivemem brane thickness could correspond to th e verticaldistance between cursors used in the pore size analysisof Section 4.1. However, since th e AFM tip can on lychart th e depth of the pore profile to a l imi ted extent,the actual pore depths ca n no t at present be determinedwith AFM. Another (probably large) estimate for thedepth of the diffusion domain would be the mem brane"skin thickness" . Hanemaaijeretal . [31] indicate sk inthicknesses fo r polysulfone UP membranes on theorder of 100 urn from high resolution SEM. Thendiv id ing th e Km estimates in Table 1 by ; > i = I OO nm ,we find partition coefficients ranging from a minimumvalue of A"=9.7xl04 (less favorable adsorption con-di t ions - Experiment 1) to a m a x i m u m v a l u e of/=8.3xl03 (more favorable adsorption condit ions- Experiment 6). So assuming that the effectivemembrane thickness in of the PM30 membrane is aconstant regardless of the chemical condit ions studiedhere, th e resul ts in Table 1 suggest that variations inchemical conditions studied here can lead to changesin th e partition coefficient of nearly on e order ofmagnitude.

Because of a lack on confidence in the effect ivemembrane thickness, we have elected to present "dif-fusion time scales", r (Table 1) , rather than attempt tocompute an effective diffusion coefficient in Eq. (6).Considering th e values of r calculated above, an dassuming a constant m, Eq. (6) yields an est imateof the rat io of (he effective diffusion coefficients.However, since the effective diffusion coefficient ina porous medium is a complicated funct ion of theporous medium tortu osity and the retardation of thediffusing species, it is useful to develop th e modelfurther to see if any clear physicochemical parameterscan be estimated (viz., th e molecular dif fusion coeffi-cient of the humic acid).


28/169


In porous medium, Dc{f is often presented as [32](10)R

where D is the molecular diffusion coefficient of thehumic material in water, and T' is the tortuosity of thediffusion path. R in Eq. (10) is the retardation coeffi-cient,

( I Dwhere e is the membrane porosity. For large retarda-tion factors (#>1) typical of strongly adsorbingsolutes, and for constant values of m and T',Eqs. (6), (10) and (11) can be manipulated to yieldfor the ratio of molecular diffusion coefficients at twodifferent pH values,D(pH2)

(12)eff

-(pHl)m(pH2)e(pHl)(PH2 )(pH2) k

If we further assume that the membrane porosity isno t affected by small variations of physicochemicalconditions, i.e., e(p H 1 ) = e ( l ) H 2 ), then Eq. (12) reducesto

(13) > ( p H 2 )D(pHl) =

Eq. (!3) could also be derived fo r changes in cal-cium concentration or ionic strength at constant pH(just substitute ionic strength for pH in above for-mula). With Eq . (13), we can now calculate ratios ofapparent molecular diffusivities of the humic acid atth e differ ent physicochemical conditions in the abovetests. As suggested above, we can look at diffusivitiesat different pH values (for constant ionic strength), orat differe nt ionic strength values (for constant p H). Forexample, using Eq. (13), we can make the followingcomparisons of the effect of a pH excursion from 6.0to 4.8 at constant ionic strength:7=1.5 X 1 0 ~ 2 M ([Ca] - 180mg/l) :

pH6 -> 4.8 yields Dpll4*/D6- = 1 . 1 8/ = 3.97 x 10~3 M ([Ca] = 33 mg/1) :

PH6 -> 4.8 yields DpR4-s/D6-Q = 1.56

Fig. 7. Effect of pH and ionic strength on diffusion coefficient oflarge molecular weight fraction of Aldrich humic acid. Reproducedfrom Cornel, Summers, an d Roberts, J. Colloid Interface Sci. 11 0(1 ) (1986) 160. Reprinted with permission granted by AcademicPress.

These calculations suggest a number of things.First, regardless of the ionic strength, the diffusivityof humic acid is increased for a decrease in pHfrom 6.0 to 4.8. This is consistent with a com-paction of the humic molecules due to neutra l i -zation of carboxylic groups [7]. Second, th ecalculation suggests that th e effect of a pHdecrease from 6.0 to 4.8 is greater at lower ionicstrength. This is also reasonable if it is hypothesizedthat ionic strength also causes compaction of thehumi c molecules, perhaps through shi eld ing ofcharged groups on the molecule [4]. Data on the effectof pH and ionic strength on the diffusion coefficien t ofhumic acid are limited. However, Cornel et al . [33]found pH and ionic strength effects very sim ilar to th echanges described above in a study of the adsorptionof humic acids by activated carbon. Fig. 7 is repro-duced from their study. Unfortunately, the humic acidstudied by Cornel et al. (Aldrich Humic A cid, Ald r ichChemical , Milwaukee, Wl) was extracted from adifferent source (ban th e Suwanncc humic acids usedby Jucker and Clark [2J; hence, it is not useful to domore than simp ly note the consistenc y between thepresent predictions and the measurements of Cornelet al. [33].With the data from Table 1, we can also calculatethe effect of an ionic strength change at different pHvalues. For example:


29/169

MM. Clark, P. Lucas/Journal of Membrane Science 143 (1998) 13-25

= 4.8: / = 3.97x l(T 3M 1.5 x I (T 2 M ,yields D aoi5M//) ao0397M= 1.92

3H - 6 . 0 : 7 = 3.97x l ( T Myields Do.oi5M/Z)o.oo397M

1.5 x 1(T2M,2_56

These calculations suggest as before, an

development and characterization of uf membranes

Documents