freeman surface modification of desalination membranes 2011

8/10/2019 Freeman Surface Modification of Desalination Membranes 2011

1/15

Journal of Membrane Science 367 (2011) 273287

Contents lists available atScienceDirect

Journal of Membrane Science

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / m e m s c i

Surface modification of commercial polyamide desalination membranes usingpoly(ethylene glycol) diglycidyl ether to enhance membrane fouling resistance

Elizabeth M. Van Wagner a, Alyson C. Sagle a, Mukul M. Sharma b, Young-Hye La c, Benny D. Freeman a,

a Department of Chemical Engineering, Center for Energy and Environmental Resources, The University of Texas at Austin, Austin, TX 78758, United Statesb Department of Petroleum & Geosystems Engineering, The University of Texas at Austin, Austin, TX 78712, United Statesc IBM Almaden Research Center, 650 Harry Road, San Jose, CA 95120, United States

a r t i c l e i n f o

Article history:Received 16 June 2010

Received in revised form 17 October 2010

Accepted 2 November 2010

Available online 9 November 2010

Keywords:

Fouling resistance

Reverse osmosis

Nanofiltration

Surface modification

Poly(ethylene glycol)

a b s t r a c t

To improve fouling resistance, polyamide reverse osmosis (XLE) and nanofiltration (NF90) membraneswere modified by grafting poly(ethylene glycol) (PEG) diglycidyl ether (PEGDE) to their topsurfacesfrom

aqueous solution. Theeffect of PEGmolecularweight (200vs. 1000)and treatmentsolutionconcentration(1%(w/w) vs.15%(w/w))on waterflux andNaCl rejectionwasmeasured.PEGDEgraftingdensityas well as

surface properties of modified and unmodified membranes, including charge, hydrophilicity and rough-ness,were measuredand compared.The fouling resistanceof modified membranesto charged surfactants

(i.e., sodium dodecyl sulfate (SDS) and dodecyltrimethylammonium bromide (DTAB)) and emulsions ofn-decane and these charged surfactants was compared to that of unmodified membranes. In general,

modified membranes exhibited improved fouling resistance and an improved ability to be cleaned afterfouling compared to unmodified membranes. Fouling resistance increased with increasing PEG molec-

ular weight, but showed little dependence on treatment solution concentration, suggesting that furtherimprovements in membranefoulingresistance might be obtained by usinglower concentrationsof highermolecular weight PEG for surface modification.

2010 Elsevier B.V. All rights reserved.

1. Introduction

The use of polyamide reverse osmosis membranes for desalina-

tion of brackish water and seawater has become common [14].However, fouling remains a serious problem that prevents evenmore widespread use of membranes in water purification appli-cations [57]. Fouling refers to the buildup of contaminants on

a membrane surface, which leads to an increase in mass transferresistance and a decrease in water flux [6].For example, bacterialgrowth on membrane surfaces, termed biofouling, is an obstaclefaced in all desalination plants, and consideration of alternative

water sources presents even more difficult fouling issues [6,7].

In this regard, produced water, the largest waste product associ-ated with oil and gas production, is a vast potential alternativewater source that is often unfit for beneficial use [8]. Contami-

nants in produced water include emulsified oil, surfactants, salt,metals, and treatment chemicals including coagulants and corro-

Corresponding author at: Department of Chemical Engineering, Center for

Energy and Environmental Resources, The University of Texasat Austin, 10100 Bur-

net Road, Building 133, Austin, TX 78758, United States. Tel.: +1 512 232 2803;fax: +1 512 232 2807.

E-mail address:[email protected](B.D. Freeman).

sion inhibitors [8]. Due to fouling concerns, extensive pretreatmentsteps are employed to remove all contaminants except salt beforesending the treated feed water to reverse osmosis modules for

desalination [9]. Thus, there is significant interest in enhancingmembrane fouling resistance to reduce the burden of feed pre-treatment when using membranesfor purification of water sourcescontaining organic as well as inorganic contaminants.

Membrane surface properties including hydrophilicity, rough-ness and charge are reported to influence fouling (i.e., increasinghydrophilicity and decreasing surface roughness and charge arereported to reduce fouling) [10], so surface modification is a

potential means of improving fouling resistance. Poly(ethylene

glycol) (PEG) has proven effective for preventing protein and bac-terial adhesion to surfaces in biomedical applications, due to thehydrophilicity, neutral charge, and steric hindrance imparted by

grafted PEG molecules[1115].The success of PEG in biomedicalfouling applications has led to its use in membrane surface modifi-cation employing both surface coating [10,16] and grafting [1723].For example, Louie et al. coated polyamide reverse osmosis mem-

branes with a block copolymer of Nylon-6 and PEG(PEBAX 1657),which slowed the rate of flux decline in an oil/surfactant/wateremulsion [10], and Sagle et al. reported improved fouling resis-tance for polyamide reverse osmosis membranes coated with

PEG-based hydrogels [16]. Belfer et al. studied redox-initiated rad-

0376-7388/$ see front matter 2010 Elsevier B.V. All rights reserved.

doi:10.1016/j.memsci.2010.11.001
http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.memsci.2010.11.001http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.memsci.2010.11.001http://www.sciencedirect.com/science/journal/03767388http://www.elsevier.com/locate/memscimailto:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.memsci.2010.11.001http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.memsci.2010.11.001mailto:[email protected]://www.elsevier.com/locate/memscihttp://www.sciencedirect.com/science/journal/03767388http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.memsci.2010.11.001


2/15

274 E.M. Van Wagner et al. / Journal of Membrane Science367 (2011) 273287

Table 1

Performance specifications for Dow Water & Process Solutions polyamide reverse osmosis (LE and XLE) and nanofiltration (NF90) membranes.

LE XLE NF90

Water permeance (A) (L/(m2 h bar)) 5.9 (5.07.4) 7.7 (6.59.6) 12.3 (10.515.4)

Salt permeance (B) (L/(m2 h)) 0.51 (0.430.63) 1.3 (1.11.7) 18.9 (16.023.6)

True NaCl rejection (%)a 99.099.3 98.099.0 8595

Water fluxb (L/(m2 h)) 50 (4262) 65 (5581) 107 (91134)

Ranges in parentheses account for acceptable range of water flux around target value (+25/15%)[2426].a NaCl rejection has been corrected for the effects of concentration polarization. NaCl rejection values are the minimum and stabilized rejections expected for each

membrane.b Test conditions:p = 10.3bar; 2000mg/L NaCl feed; 25 C; pH 8.

ical grafting of vinyl monomers, including poly(ethylene glycol)methacrylate (PEGMA), to commercial polyamide reverse osmo-sis membrane surfaces[1721]. Some modified membranes had

lower contact angles and lower average surface roughness thanunmodified membranes, suggesting these modified membranescould demonstrate improved fouling resistance [20,21]. Mickolstreated commercial polyamide reverse osmosis membranes by

submersion in aqueous solutions of poly(ethylene glycol) digly-cidyl ether (PEGDE), giving modified membranes with improvedresistance to fouling by charged surfactants [22].Kang et al. mod-ified the surface of a polyamide reverse osmosis membrane using

aminopoly(ethylene glycol) monomethylether (MPEGNH2), andthe modified membranes had improved resistance to fouling by acharged surfactant[23].

In this study, a commercially available bifunctional PEG,

poly(ethylene glycol) diglycidyl ether (PEGDE), was grafted to thetop surface of polyamide reverse osmosis and nanofiltration mem-branes using an aqueous treatment method. The effect of PEGmolecular weight and concentration on membrane performance

(i.e., water flux and NaCl rejection) is presented, and the surfaceproperties of modified and unmodified membranes are compared.The fouling resistance of modified membranes to charged sur-factants and emulsions of n-decane and charged surfactants is

compared to that of unmodified membranes.

2. Experimental

2.1. Membranes

Three types of polyamide thinfilm composite membranes man-

ufacturedby DowWater& Process Solutions(Edina, MN)were usedin this work. The LE, or low energy, and XLE, or extra low energy,membranes are brackish water reverse osmosis (RO) membranes,while the NF90 is a nanofiltration, or loose RO, membrane. The

water permeance (A), salt permeance (B), NaCl rejection and targetwater flux at common conditions (p =10.3bar, 2000mg/L NaClfeed, 25 C, pH 8) are presented for all three membranes in Table 1[2426]. These membranes are based on interfacial polymerization

ofm-phenylene diamine and trimesoyl chloride, with proprietary

additives and slight changes in polymerization conditions givingrise to differences in water and salt permeance. Commercial mem-branes of different water permeance values were chosen so that

modified and unmodified membranes having similar water fluxcould be compared. Modification decreases water permeance, somembranes with higherwaterpermeance (i.e.,XLE andNF90) wereused for modification and their performance was compared to that

of unmodified membranes having a lower water permeance (i.e.,LE). Additionally, the LE membranes were tested at lower trans-membrane pressure difference (i.e.,p < 10.3 bar), to more closelymatch the initial water flux of the modified XLE and NF90 mem-

branes. The motivation for this approach is based on previousobservations that fouling rate increases with increasing water fluxfor reverse osmosis membranes [27]. Therefore, attempting to com-

pare modified and unmodified membranes of the same (or similar)

flux is useful to minimize the effect of initial water flux on foulingrate.

Membranes were supplied as rolls of glycerin-dried flat sheets

and were stored vertically in a cool, dark place. Before membranesamples were cutfor each test, several (45) rotationsof membranewere unrolled anddiscarded.To removeglycerin, membranes weresoaked in 25% (v) aqueous isopropanol (SigmaAldrich, St. Louis,

MO) solutions for 20 min, then placed in pure water. The soak-ing water was changed three times, and membranes were left tosoak overnight (1624 h, covered to prevent exposure to light)before modification and/or testing. Ultrapure water from a Milli-

pore MilliQ system (18.2 M cm, 1.2 ppb TOC) (Billerica, MA) wasused in all experiments (i.e., soaking, modification and testing).

2.2. Surface modification

Polyamide membranes have free carboxylic acid and pri-mary amine groups (free carboxylic acids on chain ends and onuncrosslinked trimesoyl chloride molecules that hydrolyze follow-

ing polymerization, and free primary amines on chain ends [5])capable of reaction with the epoxy endgroups of poly(ethyleneglycol) diglycidyl ether (PEGDE). The chemical structure of PEGDEis illustrated in Fig. 1. Two different PEG molecular weights

were chosen for study: MW 200 and MW 1000, correspondingto 45 and 2223 repeat ethylene oxide units, respectively. The

molecular weights of the complete molecules, including the PEGand diglycidyl ether groups, were 330 and 1130 g/mol, respec-

tively. XLE membranes were treated with both molecular weightPEGDEs, while NF90membranes were only treated with the highermolecular weight PEGDE. Two different aqueous PEGDE solutionconcentrations were explored in this study (1% (w/w) and 15%

(w/w)), with a total of 10 g of solution used for treatment. WhileMW 200 PEGDE is a liquid, MW 1000 PEGDE is a waxy solid thathad to be melted prior to mixing with water. PEGDE was purchasedfrom Polysciences, Inc. (Warrington, PA) and used as received.

Membranes were modified by a top surface treatment method,which wasdesigned to isolate theactivepolyamidesurface forcon-tact and reaction with the PEGDE solution and prevent adsorptionof PEGDE to the reinforcing fabric backing or the porous support

upon which the polyamide layer is formed. A cartoon illustratingthe apparatusused forthe topsurface treatment method is given inFig.2. An11cm11cm membrane was placedin a petri dishwithaglass casting ring (10cm diameter) on thepolyamide surface, creat-

ing a well for the aqueous PEGDE solution. The appropriate amountofultrapure water washeated to 40 C, immediatelycombined withthe corresponding amount of PEGDE and shaken until the PEGDE

OCH2CH2 OCH2 CH CH2

O

HCH2C H2C

On

Fig. 1. Chemical structure of poly(ethylene glycol) diglycidyl ether (PEGDE). Two

different PEG molecular weights (MW 200 and MW 1000, corresponding ton45

and 2223, respectively) were chosen for study.


3/15

E.M. Van Wagner et al. / Journal of Membrane Science 367 (2011) 273287 275

Fig. 2. Illustration of apparatus used for top surface treatment method.

was completely dissolved. Pressure was applied to the casting ringas the heated aqueous PEGDE solution was poured onto the mem-brane surface, creating a seal preventing the solution from leakingaround the edges of the membrane. After 10 min, the membrane

wasremoved,rinsed several times with pure water,soaked in a 25%(v) aqueous isopropanol solution for10 min to remove anyresidualunreacted PEGDE, rinsed again several times with pure water and,finally, storedin pure wateruntiluse. Unmodifiedmembranes were

treated in the same manner, using 10 g of ultrapure water ratherthan an aqueous solution containing PEGDE.

2.3. ATR-FTIR and XPS

Attenuated total reflectance Fourier transform infrared spec-troscopy (ATR-FTIR) was used to characterize the structure of

modified and control membranes. These experiments were per-formed using a Thermo Nicolet Nexus 470 FTIR with an AvatarSmart MIRacle ATR accessory and a ZnSe crystal (Madison,WI). Spectra were collected in air, in the mid-infrared region

(4004000 cm1), using 128 scans at resolution 4 (1.928cm1

spacing).The background spectrum wasobtained beforeeach mea-surement, and it was subtracted from the membrane spectrum toremove any absorbance peaks due to air. Data analysis was per-

formed using the Omnic software provided with the instrument.

X-ray photoelectron spectroscopy (XPS) was used to charac-terize the surface elemental content of modified and unmodifiedmembranes.Surface scans were performed using an AXIS Ultra DLD

XPS (Kratos Analytical Company, Chestnut Ridge, NY) equippedwith a monochromatic Al K1,2X-ray source (210

9 Torr cham-ber pressure, 15 kV, 150 W). Carbon (1s), nitrogen (1s) and oxygen(1s) were detected using a 45 takeoff angle. Sulfur content was

negligible, indicating that the photoelectron escape depth was lessthan the thickness of the interfacially polymerized polyamide layer(0.1m)[5].The polyamide layer is formed on a porous polysul-fone substrate, so if the photoelectron escape depth is larger than

the thickness of the polyamide layer, one would expect to see thesulfur peak from the underlying polysulfone. A 300m700marea was analyzed, and a charge neutralizer was used to minimizesample charging.

2.4. Graft density

The grafting density of PEGDE on the membrane surface(g/cm2) was measured using a Rubotherm Magnetic Suspen-sion Balance with 1g resolution (Rubotherm GmbH, Bochum,

Germany). A diagram of the apparatus appears inFig. 3.An elec-tronic control unit allows the sample pail to be raised to themeasuring point and lowered to the zero point positions. Mem-brane samples were placed in the sample pail while in the zero

point position and then raised to the measuring point position. Theaverage difference between the measuring point and zero pointbalance readings (five measurements) was taken to be the mass of

the pail and samples.

Balance

Electromagnet

Permanent

magnet

Measuring load

decoupling

Sensor core

and coil

Sample

pail

Fig. 3. Schematic of the magnetic suspension balance used to characterize PEGDE

grafting density.

Twelve 1 in.2 pieces of membrane were used for grafting den-sity experiments, in order to fit the samples into the sample pail.Glycerin was removed from the membranes as described above,and then the membranes were dried under vacuum for 1h and

finallyallowed to equilibrate in airfor 90 min beforemeasurement.Following the mass measurement, membranes were treated withPEGDE,dried under vacuumfor 1 h and allowed to equilibrate in airfor 90 min before measuring the mass of the modified membranes

and sample pail. Subtraction of the mass measured before mod-ification from the mass measured after PEGDE modification wastaken to be the mass of PEGDE on the modified membrane surface.

2.5. Surface charge

Zeta potential was measured using an Anton Paar SurPASS

Electrokinetic Analyzer and associated software (Anton Paar USA,Ashland, VA). Two membrane samples separated by a spacer wereloaded into the clamping cell, creating a channel for electrolyteflow. A 10 mM NaCl solution was used as the background elec-

trolyte. Streaming potential was measured as a function of feed pH,and the FairbrotherMastin approximation was used in the calcu-lationof zetapotential from streaming potential.Before the start ofeach experiment,the feed pH was manually adjusted to pH 9 using

0.1 M NaOH. A 0.1 M HCl solution was then added automaticallyduring each run, to decrease pH from 9 to 3. Four measurements,two ineachflow direction (left toright and right toleft), weremadeat each pH, and the average and standard deviation of those four

values is reported.


4/15


Water

n-decanedroplet

Syringe

Sampleholder

Membrane

Environmentalchamber

Fig. 4. Apparatus used for captive bubble measurements of oil-in-water contact

angles.

2.6. Hydrophilicity

Contact angle analysis was performed using a Ram-Hart

Model 500 Advanced Goniometer/Tensiometer with DROPimageAdvanced software version 2.4 (Ram-Hart Instrument Co., Net-cong, NJ). Fig. 4 illustrates the experimental apparatus used formeasuring oil-in-water contact angles. A strip of membrane was

mounted in a sample holder with the interfacially polymer-ized polyamide layer facing down. A Gilmont Instruments 0.2 mlmicrometer syringe (Cole-Parmer, Vernon Hills, IL) with a hooked

needle wasused to dispense n-decane droplets onto the polyamidesurface of the membrane. Ultrapure water was used as the sur-rounding fluid. The reported contact angle is the average value ofthe left and right side contact angles (measured through the waterphase) obtained for at least three oil droplets placed at different

positions along the length of the membrane. A smaller angle indi-cates a more hydrophilic surface.

2.7. Surface roughness

Surface roughness was analyzed by atomic force microscopy

(AFM) using a Digital Instruments Dimension 3100 atomic forcemicroscope with a Nanoscope IV controller (Woodbury, NY). Sam-

ples were dried under vacuum prior to analysis. AFM imageswere acquired under ambient conditions in intermittent contact

mode at a 1 Hz scan rate and 256256 pixel resolution with sili-con cantilevers (spring constant50 N/m). The tapping mode (i.e.,intermittent contact mode) was used to prevent damage to themembrane surface. Two different positions were analyzed for each

membrane, each over a 5m5m area. Surface roughness wascalculated using the data analysis software provided by the manu-facturer.

2.8. Fouling characterization

All fouling experiments were conducted using a crossflow fil-

tration system. The details of the apparatus and the procedure

followed to clean the system before each experiment have beendescribed previously [28]. Following fouling experiments, an addi-tional cleaning step is necessary to remove organic components

(i.e., surfactant and/or oil) from the system. After circulation of the200 mg/L bleach solution to disinfect the system, a 0.25% (v) aque-ous solution of Nalgene L900 liquid detergent (Fisher Scientific,Pittsburgh, PA) was circulated for 1 h to remove residual organics.

The system was then rinsed several times with pure water, simi-lar to the procedure described previously [28]. In each experiment,three different membranes were tested simultaneously, typicallyone unmodified membrane and two modified membranes. The

permeate stream from each membrane was collected in a beakeron a balance connected to a LabVIEW data acquisition program(National Instruments, Austin, TX) for continuous monitoring of

water flux.

N

CH3

CH3

CH3C12H25

Br

SC12H25O O

O

O

Na

(a) (b)

Fig. 5. Chemical structures of (a) cationic dodecyltrimethylammonium bromide

(DTAB) and (b) anionic sodium dodecyl sulfate (SDS), the surfactants chosen for

use in fouling studies.

Fouling experiments were conducted using either a chargedsurfactant or an oil-in-water emulsion containing a 9:1 ratioof n-decane and a charged surfactant. n-Decane was chosen, as

opposed to a crude oil, to allow controlled measurement of foul-ing with a well-defined and easily reproducible fouling stream. Atotal concentration of 150 mg/L was used regardless of whetherthe foulant was a surfactant alone or an oil-in-water emulsion.

Dodecyltrimethylammonium bromide (DTAB) and sodium dode-cyl sulfate (SDS), cationic and anionic surfactants, respectively,were chosen to investigate the effect of electrostatic interactionson membrane fouling behavior. The chemical structures of the two

surfactants are given inFig. 5.For surfactant tests, the surfactantwas dissolved in 1 L of feed water before being added to the sys-tem. For emulsion tests, the oil and surfactant were blended with1 L of feed water in a Waringthreespeedcommercial blender(War-

ing Products, Torrington, CT) on its highest speed setting for 3minbefore being added to the system.

The procedure followed for all fouling tests is outlined inFig. 6(a), while an illustration of the typical trend in water flux

with time observed duringa fouling experimentis given in Fig.6(b).The applied transmembrane pressure difference (p) was 10.3barin all experiments, unless stated otherwise, and the feed flowrate

Purewaterfeed(pH 8.1- 8.3)

Add 2000ppm NaCl

(pH 8.0- 8.2)

Add foulant(pH 7.9- 8.3)

Start 30 min 1 hr24 hrs

Pure water

(pH 11.5 -

11.9) Decrease pH(2.5 - 2.8)

25.5 hrs

27 hrs

Fresh 2000ppm NaCl

feed (pH7.4 - 7.8)

30 hrs

End

FoulingCleaning

Fluxrecovery

302520151050

Waterflux(L/(m

2h

))

Permeation time (hrs)

Jw(pw)

Jw(NaCl)i

Jw(foul)

Jw(NaCl)f

(b)

(a)

Fig. 6. (a) Timeline and (b) typical observed water flux during a crossflow fouling

experiment.


5/15


in all experiments was 4 L/min, which corresponds to a Reynoldsnumber of 2200 in our crossflow cells. First, water flux was mon-

itored for 30 min in a pure water feed (Jw(pw)) (pH was adjustedto 8 with sodium bicarbonate, NaHCO3). Next, 2000 mg/L NaClwas added to the feed, and water flux (Jw(NaCl)i) and NaCl rejection(RNaCl,i) were measured.At 1 h, the foulant (i.e.,dissolved surfactant

or oil-in-water emulsion) wasintroduced to the system.Water fluxwas monitored for the next 23 h (the final water flux at the end ofthis time isJ

w(foul)), and NaCl rejection (R

foulant) and organic rejec-

tion weremeasured at the end ofthe fouling portion ofthe test.The

fouling feed was then drained from the feed tank and pure waterwas used to flush any residual fouling feed from the system beforerefilling the feed tank with pure water. Next, the membranes weresubjected to a representative, 2-h industrial cleaning protocol to

remove foulants from the membrane surface [20,29]. Foulants thatcan be removed by cleaning contribute to reversible fouling, whileirreversible fouling causes permanent flux loss (i.e., fouling cannotbe reversed by cleaning)[30,31].During the first hour of cleaning,

the feed pH was increased to approximately 12 using NaOH (toremove adsorbed organic and biofoulants), then the feed pH waslowered to below 3 using HCl. After the second hour of cleaning,the feed was again drained from the feed tank and replaced with

an aqueous solution containing 2000mg/L NaCl; using thissolution

as thefeed, theflux recovery (Jw(NaCl)f) and NaCl rejection (Rfoulant,f)after cleaning were determined.

The feed was circulated through the system with no feed fil-

tration at all times during these experiments (continuous feedfiltration removes foulants from the feed water [28]), with theexception of the 2-h cleaning portion, when the feed was con-tinuously filtered through a KX CTO/2 carbon block carbon/5m

particle filter (Big Brand Water Filter, Chatsworth, CA). A systemof valves allowed the feed to be sent either directly from the feedtank to the pump or,duringthe cleaning portion of the experiment,through the filter on each pass from the feed tank to the pump.

The temperature in thefeed tank was maintained at 2425 C atalltimes, except during the first hour of the cleaning procedure (highpH), when the refrigerated bath was turned off.

Sodium chloride rejection was determined by measuring ionconcentration using a conductivity meter (Oakton CON 110Advanced Meter, Fisher Scientific, Pittsburgh, PA). A calibrationcurve allowed determination of NaCl concentration from measuredsolution conductivity. To estimate the true NaCl rejection of the

membrane, the apparent NaCl rejection was corrected for concen-tration polarization using the method described previously, whichuses the pure water flux and the water flux and permeate salt con-centration in a feed solution containing 2000 mg/L NaCl to estimate

the polarization modulus [28,32]. Organic carbon concentrationwas determined using a Shimadzu 5050A Total Organic Carbon(TOC) Analyzer (Columbia, MD). Samples were analyzed immedi-ately after collection, and organic rejection was calculated using

the measured feed and permeate concentrations (mg/L). Feed pH

was monitored using a pH meter (Accumet Research AR25 DualChannel pH/Ion Meter, Fisher Scientific, Pittsburgh, PA).

3. Results and discussion

3.1. Characterization of PEGDE on the modified membrane

surfaces via FTIR and XPS

ATR-FTIR and XPSwere used to qualitatively verifythe presenceof PEGDE on the membrane surface following the surface grafting

protocol. Fig. 7(a) and (b)presents ATR-FTIR spectra in the range of10001200cm1 for the modified and unmodified XLE and NF90membranes, respectively. The polyamide membrane has several

peaks in this region[18],including those at 1080, 1105, 1150 and

(a)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

10001050110011501200

Absorbance

Wavenumber (cm-1)

1

2

3

(b)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

10001050110011501200

Absorbance

Wavenumber (cm-1)

12

3

4

5

Fig. 7. ATR-FTIR spectra in the range 10001200cm1 for (a) the unmodified XLE

membrane (1) and XLE membranes modified with 1% (w/w) MW 200 PEGDE (2),1% (w/w) MW 1000 PEGDE (3), 15% (w/w) MW 200 PEGDE (4), and 15% (w/w) MW

1000 PEGDE (5), and (b) the unmodified NF90 membrane (1) and NF90 membranes

modified with 1% (w/w) MW 1000 PEGDE (2) and 15% (w/w) MW 1000 PEGDE (3).

1170cm1 attributed to the CC stretch of the aromatic rings[33].

The largest peak for PEG, attributed to the ether CO stretch as

well as the CC stretch, occurs around 1100 cm1

[34], and maybe expected to cause a noticeable change in absorbance for thePEGDE-modified membranes, as was observed by Kang et al. for

membranes modified with MPEGNH2 [23]. Fig.7(a) and(b) revealsan increase in absorbance around 1100 cm1 (i.e., between 1080and 1130cm1) with increasing PEG molecular weight and con-centration of the aqueous PEGDE treatment solution. As expected,

peaks at 1150 and 1170 cm1 do not exhibit a systematic increasein absorbance upon PEGDE modification, since PEGDE has mini-mal absorbance at these wavenumbers. The penetration depth ofATR-FTIR is approximately 1m[21], so the fact that the peaks

of the polyamide membrane are visible after PEGDE modificationindicates that the PEGDE graft layer is thinner than 1m.

Table 2presents atomic concentration data from XPS analysis

of an unmodified XLE membrane and XLE membranes modified


6/15


Table 2

XPS measurementsof surface atomic concentration(%) of oxygen, nitrogenand car-

bonmeasured foran unmodifiedXLE membraneand XLEmembranesmodifiedwith

MW 1000 PEGDE.

Membrane O (%) N (%) C (%) O/N

XLE 13.1 11.2 75.7 1.17

1% (w/w) MW 1000 XLE 13.7 10.7 75.6 1.28

15% (w/w) MW 1000 XLE 15.6 10.3 74.1 1.51

100% (w/w) MW 1000 XLE 17.6 10.4 72.0 1.69

with 1% (w/w), 15% (w/w), and 100% (w/w) MW 1000 PEGDE.Since PEGDE contains oxygen but not nitrogen (excluding hydro-gen, PEGDE contains 67% carbon and 33% oxygen), the observedincreasesin oxygencontent anddecreases innitrogen content upon

PEGDE modification are expected, and the ratio of oxygen to nitro-gen on the membrane surface increases. Additionally, the carboncontent of an XLE membrane, 75.7%, is higher than that in PEGDE(67%), so the observed decrease in carbon content with increas-

ing concentration of PEGDE in the surface treatment solution (and,presumably, increasing amounts of PEGDE on the membrane sur-face), is consistent with the oxygen and nitrogen results. Thesetrends (i.e., decreasing nitrogen and carbon content and increas-

ing oxygen content and oxygen to nitrogen ratio) have also been

observed by Kang et al. [23] and Sagle et al. [16] for polyamidemembranes modified with PEG-based materials, so the changes incomposition observed here arereasonable withinthe context of the

previous work done in this area. Furthermore,these XPS results arequalitatively consistent with the ATR-FTIR results.

Interestingly, even an XLE membrane treated with 100% (w/w)MW 1000 PEGDE (i.e., no water used in the treatment solution) still

has a sizeable nitrogen content according to XPS. This result indi-cates that the PEGDE layer is thinner than the photoelectron escapedepth. In general, the photoelectron escapedepth is approximately10nm(at0 takeoff angle) [16], andthe 45 takeoff angle used here

is expected to probe an even thinner surface layer. However, Sagleet al. reported that a 2m thick layer of crosslinked poly(ethyleneglycol) diacrylate on a polyamide membrane still resulted in signif-

icant nitrogen content as measured by XPS [16], soXPS may not bea good indicator of PEGlayerthickness, or the graftedlayermay notbe uniform over the region of membrane probed by XPS. However,XPS does provide qualitative evidence of the presence of PEGDE onthe membrane surface.

3.2. Graft density of PEGDE on the modified membrane surface

Fig. 8 presents the dependence of graft density on the con-

centration of PEGDE in the aqueous treatment solution. Graftdensity increased with increasing PEGDE concentration, and, atleast for membranes modified with 15% (w/w) PEGDE, graft den-sity was larger for membranes modified with higher molecular

weight PEGDE. The graft density data are also qualitatively con-

sistent with the ATR-FTIR and XPS results presented above (i.e.,higher graft density corresponds to higher absorbance in the rangeof 10801130 cm1 and higher oxygen to nitrogen ratio).

The apparent effective PEGDE thickness on the membrane sur-face may be estimated from the measured graft densities (g/cm2)and the density of PEGDE (the density of MW 1000 PEGDE isnot reported by the supplier, so the density of MW 200 PEGDE,

1.15g/cm3 is used as an estimate). The range of graft densitiesobserved (339g/cm2) corresponds to effective thicknesses of26339nm if the thickness of the PEGDE were uniform on thenominal surface area of the membrane. Previous work by Belfer

et al. reported a grafting layer thickness of 20nm for a semi-aromatic piperazine-based polyamide nanofiltration membranegrafted with polyacrylic acid [21], while Louie et al. reported a

coating thickness of 0.3m (300 nm) for a commercial polyamide

0

10

20

30

40

1614121086420

Graftdens

ity(g/cm

2)

PEGDE concentration in water (wt%)

MW 200 XLE

MW 1000 NF90

MW 1000 XLE

Fig.8. Graft densityfor XLEmembranesmodifiedwith1% (w/w) and15% (w/w) MW

200 and MW 1000 PEGDE, and NF90 membranes modified with 1% (w/w) and 15%

(w/w) MW 1000 PEGDE. The unmodified XLE and NF90 membranes (i.e., 0% (w/w)PEGDE concentration in water) are shown in this figure at zero graft density. Error

bars associated with graft density indicate the standard deviation of the results for

at least three membrane samples.

reverse osmosis membrane coated with PEBAX 1657[10],so thelayer thicknesses due to PEGDE grafting appear to be similar to

those reported earlier. Using the measured graft densities and themolecular weightof PEGDE, and approximating the membrane sur-face asa flatsheet, thenumber ofPEGDEmoleculesper unit area canalso be estimated. The measured graft densities suggest the pres-

ence of 15 to over 500 PEGDE molecules per nm2 of membrane,depending on the molecular weight and concentration of PEGDE.Based on estimation of surface charge density from zeta potential

analysis, the total number of reactive amine and carboxylic acidgroupspresent on thesame area of polyamidemembrane surface islikelyto betwo tothree orders ofmagnitude smaller thanthe num-ber of PEGDE molecules present[35].Thus, the amount of PEGDEattached to a membrane surface is apparently notstrictlylimited by

thenumber ofsurfaceamineand carboxylic acid groups, whichsug-gests that additional PEGDE may also adsorb onto the membranesurface, perhaps fillingsurface defects, in addition to grafting to thereactive groups on the membrane surface.

The membrane modification procedure was altered in the graftdensity experiments to accommodate the sample size required forusein themagneticsuspensionbalance.Smallsamples (1 in.2)wererequired to fit in the sample pail, making isolation of the top sur-

face of each of the membranes impractical,so the membranes were

treated by submersion, using as little PEGDE solution as possible tocover all twelve samples. The only difference relative to the topsurface treatment method described in Section 2.2 was the contact

between the PEGDE solution and the reinforcing fabric backing ofthe membrane. Water flux is lower for membranes modified bysubmersion versus top surface isolation, but the post-modificationisopropanol soak and water rinses are expected to remove residual

unreacted PEGDE from the membrane surface and reinforcing fab-ric backing. However, it is possible that not all unreacted PEGDEwas removed (e.g., PEGDE could be adsorbed to the reinforcingfabric backing). The magnetic suspension balance would detect

any residual PEGDE, resulting in apparent grafting density val-ues somewhat higher than those obtained in samples that wereonly modified by exposure of the top surface of the membrane

to PEGDE. Thus, the reported graft densities are upper bounds of


7/15


Table 3

Surface properties (charge, hydrophilicity and roughness) of unmodified and PEGDE-modified membranes.

Membrane Zeta potential at pH 8 (mV) Oil-in-water contact angle () RMS surface roughness (nm)

LE 32.2 0.6 33 2 55 3

XLE 31.2 1.3 35 2 58 5

1% (w/w) MW 200 XLE 30.8 1.0

15% (w/w) MW 200 XLE 31.3 0.4 37 2 70 4

1% (w/w) MW 1000 XLE 29.8 1.4

15% (w/w) MW 1000 XLE 29.8 0.3 32 2 77 3

NF90 32.4 1.1 30 3 47 5

1% (w/w) MW 1000 NF90 26.4 0.3

15% (w/w) MW 1000 NF90 26.3 0.5 32 6 56 1

the graft density for samples used for fouling characterization (i.e.,membranes treated by top surface isolation). Additionally, 70 g ofsolution were required to completely cover thesurface of alltwelve1in.2 membrane samples (0.9 g of solution/cm2 membrane), versus

the 10g used to treat the membranes in all other experiments(where a single continuous area of membrane was modified ina 10 cm diameter casting ring; 0.1 g of solution/cm2 membrane).However, decreasing the amount of treatment solution as much

as possible (to 0.6 g of solution/cm2 membrane) revealed no mea-surable difference in graft density, indicating this difference intreatment procedure had negligible effect on graft density. Finally,mass measurements were done in air, so measured graft densi-

ties may include not only grafted PEGDE, but also water absorbedby the highly hydrophilic PEGDE. Performing the measurementsunder vacuum did not result in lower graft densities, so it appearsthat the amount of any absorbed water is negligible compared to

the amount of PEGDE grafted to the membrane.

3.3. Surface charge, hydrophilicity, and roughness

The measured surface properties of unmodified and modifiedXLE and NF90 membranes, as well as those of the unmodified LEmembrane, are presented inTable 3.The LE membrane is included

because it has lower water flux than theXLE and NF90 membranes,and its fouling behavior will be compared to that of the modifiedXLE and NF90 membranes. This strategy permits a comparison offouling properties in modified and unmodified membranes havingapproximately the same initial flux. The unmodified LE, XLE and

NF90 membranes all have essentially the same zeta potential at pH8 (32mV).Atthe samepH (pH 8)andusing thesame backgroundelectrolyte (i.e., 10 mM NaCl), zeta potential values between10and 40 mV have previously been reported for polyamide desali-

nation membranes[16,36,37].The zeta potential of modified XLEmembranes is essentially equal to that of an unmodified XLE mem-brane. In contrast, the modified NF90membranes have slightly lessnegative zetapotential than theunmodified NF90membrane. Since

PEGDE is a neutral molecule, it is expected to lower the magnitude

of the membranes surface charge, but the charge of the under-lying polyamide is dominant for the modified membranes. Sagleet al. reported that polyamide membranes coated with 2m thick

PEG-based hydrogel layers experienced a 25% decrease in surfacecharge[16],so it is reasonable that the PEG layer thicknesses esti-mated here (26339 nm) would have a minimal effect on surfacecharge, consistent with the results reported inTable 3.

The contact angles for unmodified and highly modified (i.e.,15% (w/w) PEGDE treatment solution concentration) membranesshowed no measurable difference, and the low contact angle val-ues (approximately 30) indicate that all the membranes were

hydrophilic. Louie et al. reported that coating a commercialpolyamide reverse osmosis membrane with PEBAX 1657 hada negligible effect on contact angle, although the actual contact

angle values reported forthe modified andunmodified membranes

(60) were higher, possibly due to differences in measurementconditions (i.e., measurement of air-in-water contact angle versusthe oil-in-water contact angle measured here)[10].

The surface roughness values in Table 3 indicate PEGDE modifi-

cation may cause a slight increase in roughness, perhaps indicatingconformation of PEGDE to the polyamide ridge-and-valley surfacestructure during grafting. Polyamide membrane surface roughnessis typically reported to be approximately 50100 nm [10,21,23,38],

in agreement with thevalues reported in Table 3 forthe unmodifiedLE, XLE and NF90 polyamide membranes. Reports of the effect ofsurface modification on surface roughness vary. While Louie et al.andBelfer et al.reporteddecreases in surface roughnessupon mod-

ification with PEBAX 1657 and PEGMA, respectively, of up to 65%[10,21],Kang et al. observed a 60% increase in surface roughnessupon modification with MPEGNH2[23]. Thus, the increase in sur-face roughness observed here upon PEGDEmodification (1933%)

is reasonable considering the range of results reported previously.Overall, PEGDE modification appears to cause relatively slightchanges in surface charge, hydrophilicity and roughness.

3.4. Water flux and NaCl rejection of modified membranes

The water flux and true NaCl rejection in 2000 mg/L NaCl feed

of the PEGDE-modified and unmodified XLE and NF90 membranesare presented inFig. 9(a) and (b), respectively. The data were col-lected from the initial NaCl feed portion of the fouling experiments(cf.Fig. 6(a), t= 30 min). Modification of the XLE membrane with1% (w/w) MW 200 PEGDE caused a 15% decrease in water flux

and no change in NaCl rejection, while modification with a 15%(w/w) MW 200 PEGDE solution produced a modified XLE mem-brane with 31% lower water flux and 0.2% higher NaCl rejectionthan the unmodified XLEmembrane.Modification of theXLE mem-

brane with MW 1000 PEGDE resulted in a larger decrease in waterflux and larger increase in NaCl rejection than that obtained withMW 200 PEGDE. However, there was minimal change in perfor-mance due to modification with a 15% (w/w) MW 1000 PEGDE

solution relative to modification with a 1% (w/w) MW 1000 PEGDE

solution (44% vs. 37% decrease in water flux and 0.4% vs. 0.3%increase in NaCl rejection compared to an unmodified XLE mem-brane, respectively). Thus, it appears that, at least for the higher

molecularweight PEGDE,the amountof PEGDEgrafted tothe mem-brane surface maybe approaching an upper limit,sincemembranestreated with 1% (w/w) and 15% (w/w) PEGDE solutions have simi-lar performance. Surface modifications causing decreases in water

flux similar to those observed for PEGDE modification have beenreported by Sagle et al. (3540% decrease in water flux observedfor PEG hydrogel-coated membranes)[16]and Belfer et al. (up to30% decrease in water flux for membranes grafted with PEGMA)

[20],although much larger reductions in water flux due to surfacemodification have also been reported (e.g., Louie et al. reportedup to an 80% decrease in water flux for membranes coated with

PEBAX

1657[10]and Mickols observed a 75% decrease in water


8/15


30

40

50

60

70

80

90

100

151050

Jw(NaCl)i(L/(m

2h

))


MW 200 XLE

MW 1000 NF90

MW 1000 XLE

(a)

97

97.5

98

98.5

99

99.5

100

151050

TrueNaClrejection(RNaCl,i)(%)


MW 200 XLE

MW 1000 NF90

MW 1000 XLE

(b)

Fig. 9. (a)Waterflux and (b)true NaClrejectionin 2000mg/LNaCl feed,for unmod-

ified XLE and NF90 membranes (i.e., 0% (w/w) PEGDE concentration in water), XLE

membranes modified with 1%(w/w) and 15% (w/w)MW 200 and MW1000PEGDE,

and NF90 membranes modified with 1% (w/w) and 15% (w/w) MW 1000 PEGDE.

Error bars in the figures indicate the standard deviation of the performance of at

least six membrane samples.

flux for a reverse osmosis membrane treatedby submersion in a 1%(w/w) MW 200 PEGDE solution[22]).Additionally, surface modi-fications that reduce flux often increase rejection[10,16,20,22],so

the increase in rejection accompanying surface modification withPEGDE is reasonable. Increasing rejection could be explained as theresult of plugging surface defects in the polyamide membrane byPEGDE, or possibly as the result of PEGDE interacting with and

influencing the transport properties of the polyamide membrane[16,39].

Plugging of minute surface defects in the unmodified polyamidemembrane by PEGDE could also help explain the minimal differ-

ence in performance for membranes modified with low (1%, w/w)and high (15%, w/w) concentrations of MW 1000 PEGDE. AlthoughPEG is a much more permeable material than the polyamide layer

in desalination membranes, it is still far less permeable than a

defect, so plugging defects with even small amounts of PEGDEcould result in substantial decreases in water flux (e.g., the 37%

flux decline observed for XLE membranes treated with 1% (w/w)MW 1000 PEGDE) and contribute to an increase in rejection. How-ever, PEGDE grafted onto the dense, defect-free polyamide surface,which presumably constitutes the vast majority of the membrane

surface, would have a much smaller effect on water flux, sincethe permeability of the polyamide is believed to be much lowerthan that of PEGDE (e.g., the 7% further flux decline caused bygrafting 15% (w/w) MW 1000 PEGDE onto the XLE membrane

surface).The increase in NaCl rejection observed for XLE membranes

modified with MW 1000 PEGDE inspired the use of NF90 mem-branes for modification. PEGDE modification always decreases

water flux, but using the higher flux NF90 membranes and leverag-ing the NaCl rejection enhancement observed upon modificationcould lead to higher flux modified membranes with NaCl rejec-tion values similar to those of the XLE membranes. Although

modification led to larger relative declines in water flux for theNF90 membranes than the XLE membranes (e.g. 45% vs. 37%decrease in water flux for membranes modified with 1% (w/w)MW 1000 PEGDE, which may indicate a larger number of defects

in the less-selective NF90 membranes), NF90 membranes modi-

fied with MW 1000 PEGDE did have higher water flux than thecomparable modified XLE membranes (cf.Fig. 9(a), 51L/(m2 h) vs.45 L/(m2 h) and 48L/(m2 h) vs. 40 L/(m2 h) for membranes modi-

fied with 1% (w/w) and 15% (w/w) PEGDE solutions, respectively).However, the modified NF90 membranes did not quite reach thesame level of NaCl rejection as the modified XLE membranes (cf.Fig. 9(b)), although their NaCl rejection increased enough (from

97.8% to 98.8%) to fall within the manufacturers target range forthe XLE membranes (98.099.0%, cf. Table 1). The water flux ofthe unmodified NF90 membrane (93 L/(m2 h)) was near the lowerend of the manufacturers target range (91134 L/(m2 h)), while

its NaCl rejection (97.8%) exceeded the expected range (8595%)[26].Lower water flux is typically accompanied by increased NaClrejection due to the inverse relationship between permeability and

selectivity [28,40,41], so the unmodified NF90 performance fallsin line with this tradeoff, suggesting that this particular roll ofNF90 membrane was simply more selective than an average NF90membrane.

Fig. 10presents the dependence of water flux in a 2000 mg/L

NaCl feed on PEGDE grafting density measured using the mag-netic suspension balance. Thebehavior forall modified membranes(MW 200 and MW 1000 XLE and MW 1000 NF90) followed similartrends to those observed in Fig. 9(a), where water flux was pre-

sented as a function of PEGDE treatment solution concentration.For XLE and NF90 membranes modified with MW 1000 PEGDE, alarge decrease in water flux was observed at low grafting densities(i.e.,membranes treated with 1% (w/w) MW 1000PEGDE solution),

and a subsequent large increase in grafting density resulted in lit-

tle additional decrease in water flux (i.e., membranes treated with15% (w/w) MW 1000 PEGDE solution). XLE membranes modifiedwith MW 200 PEGDE, on the other hand, exhibited a more grad-

ualdecline in water flux with increasing grafting density, similar tothe observed dependence of water flux on PEGDE treatment solu-tion concentration for these membranes. The difference in behaviorbetween membranes treated with MW 200 and MW 1000 PEGDE

may be due to the difference in PEG chain length, and could indi-cate that longer chain PEGs fill surface defects more readily thanshorter chain PEGs. The shorter PEG chains of MW 200 PEGDEmay need to be present in higher concentration (i.e., 15% (w/w)

vs. 1% (w/w)) in order to completely fill surface defects, whilethe longer PEG chains of MW 1000 PEGDE may be able to com-pletelyfill surface defectsusing only a 1% (w/w) treatment solution

concentration.


9/15


30

40

50

60

70

80

90

100

403020100

Jw(NaCl)i(L/(m2h

))

Graft density ( g/cm2)

MW 200 XLE

MW 1000 XLE

MW 1000 NF90

Fig. 10. Water flux in a 2000 mg/L NaCl feed versus graft density measured using

the magnetic suspension balance, for XLE membranes modified with 1% (w/w) and15% (w/w) MW 200 and MW 1000 PEGDE, and NF90 membranes modified with 1%(w/w) and 15% (w/w) MW 1000 PEGDE. The unmodified XLE and NF90 membranes

(i.e., 0% (w/w) PEGDE concentration in water) are shown in this figure at zero graft

density.Error barsassociatedwithJw(NaCl)iare definedin Fig. 9; errorbars associated

with graft density are defined inFig. 8.

3.5. Fouling resistance and flux recovery in surfactant fouling

tests

Fig. 11 illustrates representative results from fouling exper-iments. In this example, the fouling of unmodified LE and XLE

membranes as well as an XLE membrane modified with 1% (w/w)MW1000 PEGDE by SDS is presented. Fig. 11(a) presents water fluxas a function of time, where the first point for each membrane cor-

responds toJw(NaCl)i in Fig. 6(b). The water flux data up tot= 2 4 hrepresent the fouling portion of the test (the water flux att= 2 4 hcorresponds toJw(foul)inFig. 6(b)), and the time period t=2730hmonitors the flux recovery aftercleaningthe membranes (thewaterflux at t= 30h is Jw(NaCl)f in Fig. 6(b)). The modified XLE mem-

brane had lower water flux than the unmodified XLE membranethroughout the entire experiment, but it is difficult to comparethe performance of these two membranes directly because mod-ification decreased the initial water flux (Jw(NaCl)i) by 34%. Initial

water flux may impact the extent of fouling observed, since ahigher flux membrane will process (i.e., permeate) a larger vol-ume of feed water (and therefore, be exposed to a larger amountof the foulant) during a given time period than a lower flux mem-

brane. An increase in fouling rate with increasing water flux has

been observed previously for reverse osmosis membranes [27].Although the unmodified XLE membrane had higher water fluxthan the modified XLE membrane (e.g., 35 L/(m2 h) vs. 32 L/(m2 h)

at the end of the 24 h fouling test, cf.Fig. 11(a)), the modified XLEmembrane had higher initial NaCl rejection than the unmodifiedXLE membrane (99.4% vs. 99.1%, respectively, cf. Fig. 9(b)). NaClrejection increased to 99.9% for all membranes during the fouling

test (rejection during fouling tests will be discussed later). Usinglinear interpolation between the initial and final rejection valuesto calculate the amount of NaCl permeated reveals the unmodifiedXLEmembrane permeatedtwice as much salt as an XLEmembrane

modified with 1% (w/w) MW 1000 PEGDE during the 24 h foulingtest, so although the unmodified XLE membrane had higher waterflux, the XLE membrane treated with 1% (w/w) MW 1000 PEGDE

produced higher purity water.

10

20

30

40

50

60

70

80

302520151050

Waterflux(L/(m

2h

))


1% (w/w) MW 1000 XLE

XLE

LE ( p = 9.0 bar)

(a)

(b)

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

302520151050


XLE

1% (w/w) MW 1000 XLE

LE (p = 9.0 bar)

Normalizedwaterflux(Jw(t)/Jw(NaCl)i)

Fig. 11. Typical fouling behavior observed in all fouling tests, using SDS fouling

resultsobtainedfor unmodifiedLE andXLE membranes andan XLEmembrane mod-

ified with 1% (w/w) MW 1000 PEGDE. (a) Water flux and (b) normalized water flux,

as a function of time.

One means of minimizing the effect of water flux on foulingrate is to compare the performance of modified membranes with

that of unmodified membranes of similar initial water flux. Since

all three membranes chosen for study (i.e., LE, XLEand NF90)shareapproximately the same polyamide chemistry, they may exhibitsimilar fouling behavior.Fig. 11(a) also includes the performance

of an unmodified LE membrane, operated at a lower transmem-brane pressure difference (i.e., p = 9.0 bar vs.p =10.3bar usedfor the other membranes) to closely match the initial water fluxof themodified XLEmembrane (4849 L/(m2 h)). During the foul-

ing experiment, the flux decline observed for the unmodified LEmembrane was larger than that of the modified XLE membrane(i.e., water fluxes at the end of the 24 h fouling test were 32 and17 L/(m2 h) for the modified XLE and unmodified LE membranes,

respectively, cf.Fig. 11(a)). The modified XLE membrane also hadhigher water flux after cleaning than the unmodified LE mem-brane (i.e., 43 L/(m2 h) vs. 35L/(m2 h), respectively, cf. Fig. 11(a)).

That is, the modified XLE membrane demonstrated better fouling


10/15


resistance, and flux recovery after cleaning, than an unmodified LEmembrane with nearly identical initial water flux. In addition to

its higher water flux during the fouling test and after cleaning, themodified XLE membrane also had higher initial NaCl rejection thanthe unmodified LE membrane (99.4% vs. 99.1%).

InFig. 11(b), the data fromFig. 11(a) are presented in terms of

normalized water flux, determined by dividing the water flux attime t (i.e.,Jw(t)) byJw(NaCl)i. Flux normalization allows comparisonof membranes withdissimilar initial water fluxes (e.g., the unmod-ified XLE membrane and XLE membrane modified with 1% (w/w)

MW 1000 PEGDE). Although it is difficult to determine which ofthese two membranes has better fouling resistance in Fig. 11(a),Fig. 11(b) demonstrates that the modified XLE membrane retainedmore of its initial water flux than the unmodified XLE membrane

(70% vs. 50%, respectively), and the modified XLE membrane alsoregained more of its lost flux after cleaning (90% vs. 66%, respec-tively). During the early stages of fouling, the flux of the modifiedXLE membrane decreases at a slower rate than that of either the

unmodified LE or XLE membranes, but after about 4h of foul-ing, the fouling rates of all three membranes become very similar.Thus, it is the initial differences in fouling rate between the threemembranes that give rise to the overall order of fouling resistance

observed, with the modified XLE membrane retaining the highest

percentage of itsinitial flux. Onepossibleexplanationis that duringthe early stages of fouling, foulantsurface interactions dominate(and could be very different for different membranes), but as foul-

ing progresses and the membrane surface becomes less accessible,foulantfoulant interactions become more important, and theseinteractions may be expected to be somewhat independent of themembrane.

In the following figures, fouling results will be presented interms of flux retained after fouling (Jw(foul)/Jw(NaCl)i) and regainedafter cleaning (Jw(NaCl)f/Jw(NaCl)i) as a function of initial water flux(Jw(NaCl)i), to account, at least approximately, for the effect of initial

water flux on fouling behavior.The SDS fouling data for all modified XLE and NF90 membranes

as well as unmodified LE, XLE and NF90 membranes are summa-

rized inFig. 12(a) and (b), showing the fraction of flux retainedafter fouling andthe fraction of fluxregained after cleaning,respec-tively. In these and subsequent figures presenting summaries offouling data, filled symbols represent results from modified mem-branes, and unfilled symbols represent results from unmodified

membranes. LE membranes were tested at three different trans-membrane pressures to allow comparison of modified XLE andNF90 membrane behavior to the behavior of unmodified mem-branes with similar initial water fluxes. Comparison of modified

and unmodified membrane performance at the same initial waterflux demonstrates that several modified membranes had betterfouling resistance in SDS tests, based upon fraction of flux retainedafter fouling and flux recovery after cleaning (i.e., fraction of flux

regained), than the unmodified membranes. For example, the XLE

membranes modified with 1% (w/w) MW 1000PEGDEretained andregained approximately 2025% more of their initial flux than didthe LE membranes of similar initial water flux (45 L/(m2 h)). Sim-

ilar improvements in SDS fouling resistance (i.e., flux retention)were observed bySagleet al.forpolyamide membranescoated withPEG-based hydrogels[16].

In general, membranes modified with MW 1000PEGDE demon-

strated better fouling resistance and flux recovery in SDS teststhan membranes modified with MW 200 PEGDE, which could beexplained as the result of steric hindrance (i.e., the longer PEGchains of MW 1000 PEGDE may provide more of a steric barrier

to foulants than the shorter PEG chains of MW 200 PEGDE). Thedifference in molecular weight between MW 200 and MW 1000PEGDE (i.e., 330 g/mol vs. 1130 g/mol) also means that the molar

concentration of MW 200 PEGDE is 3.4 times larger than that of

Fractionoffluxretained(Jw(foul)

/Jw(NaCl)i)

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

30 40 50 60

Jw(NaCl) i

(L/(m2h))

Unmodified

membranes

70

Modified

membranes

80 90 100

(a)

Fracti

onoffluxregained(Jw(NaCl)f/

Jw(NaCl)i)

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

30 40 50 60

Jw(NaCl)i

(L/(m2h))

Unmodified

membranes

70

membranes

80

Modified

10090

(b)

Fig.12. (a)Fraction offlux retained afterfoulingwith150mg/LSDS and (b)fraction

of flux regained after fouling and cleaning, as a function of initial water flux. Defini-

tionof symbols:, LE; , LE (p =9.0bar); , L E (p =8.6bar), , XLE;, NF90;,

1% (w/w) MW 200 XLE; , 15% (w/w) MW 200 XLE; , 1% (w/w) MW 1000 XLE; ,

15% (w/w)MW 1000XLE; , 1%(w/w) MW1000NF90; and , 15% (w/w)MW 1000

NF90. Lineshavebeen drawn throughthe modified andunmodified membranedata

to guide the eye.

MW 1000 PEGDE at the same mass concentration (e.g., 1% (w/w)),so MW200 PEGDEhas 3.4timesas many epoxide endgroupsas MW1000 PEGDE. Since the measured PEGDE graft densities exceeded

the number of surface amine and carboxylic acid groups on themembrane (perhaps indicating PEGDE adsorption as well as graft-ing), most of the epoxide endgroups will be left unreacted. Theepoxide endgroups are not expected to improve membrane foul-

ing resistance, so the higher concentration of endgroups in MW200PEGDE could play a role in the observed dependence of foulingbehavior on PEGDE molecular weight. However, fouling resistancewas less dependent on PEGDE treatment solution concentration,

since membranes modified with 1% (w/w) and 15% (w/w) PEGDEtreatment solutions had similar fouling resistance.

Within the scatter of the data points, the fraction of flux

retained and regained in the SDS fouling tests was essentially


11/15


Afoulant/

Amembrane

0

0.5

1

1.5

2

2.5

3

30 40

Unmodified

50 60

Jw(NaCl)i

(L/(m2h))

membranes

70

Modified

membranes

80 10090

Fig. 13. Ratio of foulant layer permeance to membrane permeance as a function of

initial waterflux duringfouling with 150mg/L SDS. Definitionof symbols:, LE;, LE (p =9.0bar); , LE (p =8.6bar); , XLE;, NF90; , 1% (w/w) MW 200 XLE;, 15%(w/w)MW 200 XLE;, 1%(w/w) MW1000XLE;, 15% (w/w)MW 1000XLE;

, 1% (w/w) MW 1000 NF90; and , 15% (w/w) MW 1000 NF90. Lines have been

drawn through the modified and unmodified membrane data to guide the eye.

independent of initial water flux for the unmodified membranesand decreased with increasing initial water flux for the modi-fied membranes. Using a series-resistance approach [39,42], thepressure-normalized waterflux (i.e.,the permeance)during fouling

is given by:

A =

1

Am+

1

Af

1(1)

where A m is the membrane permeance during fouling (assumedto be equivalent to the membrane permeance in the absence of

fouling) andAfis the permeance of the deposited foulant layer. Thefraction of flux retained during fouling (i.e., Jw(foul)/Jw(NaCl)i) maybe expressed as the ratio of the permeance during fouling to thepermeance in the absence of fouling (i.e., Am):

Jw(foul)Jw(NaCl)i

=1

Am((1/Am)+ (1/Af))=

1

1+ (Am/Af) (2)

Fig. 13presents the ratio Af/Am as a function of initial waterflux, Jw(NaCl)i. For the modified membranes, this ratio decreaseswith increasing initial water flux, possibly indicating the growth

of a thicker or less permeable fouling layer (i.e., decreasingAf) with

increasing water flux, which is reasonable since fouling extent isexpected to increase with increasing initial water flux[27].How-ever,for theunmodifiedmembranes,the ratioAf/Am remains nearly

constant with increasing initial water flux (and, therefore, increas-ingAm), which, ifAmis not influenced by fouling, would indicate anincrease in Af, which is not expected. One possible explanation isthat the permeance of the unmodified membranes (Am) is, in fact,

affected by fouling, i.e., the foulant may interact with the mem-brane, decreasing the membrane permeance during fouling. Thelarger extent of fouling observed for the unmodified membranesas compared to the modified membranes would support the pos-

sibility of interactions between the unmodified membranes andthe foulant which are absent, or at least weaker, in the modifiedmembranes, whose PEG chains may prevent foulants from closely

approaching their surfaces.

Fractionoffluxretained(Jw(foul)

/Jw(NaCl)i)

0

0.2

0.4

0.6

0.8

1

30 40 50

membranes

Unmodified

membranes

Jw(NaCl) i

(L/(m2h))

60

Modified

70 80 90

(a)

Fractionoffluxregained(J

w(NaCl)f/

Jw(NaCl)i)

0

0.2

0.4

0.6

0.8

1

30 40 50

Jw(NaCl)i

(L/(m2 h))

60

Modified

membranes

70 80

Unmodified

membranes

90

(b)

Fig. 14. (a) Fraction of flux retained after fouling with 150 mg/L DTAB and (b) frac-

tion of flux regained after fouling and cleaning, as a function of initial water flux.

Definition of symbols: , LE; , LE (p =9.0bar); , LE (p =8.6bar); , XLE;

, NF90; , 1% (w/w) MW 200 XLE; , 15% (w/w) MW 200 XLE; , 1% (w/w) MW

1000 XLE; , 15% (w/w) MW 1000 XLE; , 1% (w/w) MW 1000 NF90; and , 15%

(w/w) MW1000NF90.Lines havebeendrawnthroughthemodifiedand unmodified

membrane data to guide the eye.

The DTAB fouling results presented in Fig. 14(a) and (b) are sim-ilar to those inFig. 12(a) and (b) for SDS fouling, in that fouling

resistance and flux recovery were better for membranes modifiedwith higher molecular weight PEGDE, and fouling resistance andflux recovery were better in many of the modified membranesthan in unmodified membranes of similar initial water flux. As

in the case of SDS fouling, the unmodified membranes foulingresistance was relatively independent of initial water flux, sug-gesting possible interactions between DTAB and the unmodifiedmembrane surfaces. The improvement in DTAB fouling resistance

demonstrated by PEGDE-modified membranes was similar to thatobserved by Sagle et al.for polyamidemembranes coatedwith PEG-based hydrogels (1015%)[16].Mickols[22]and Kang et al.[23]

also reported similar increases in DTAB fouling resistance during


12/15


short, 23 h fouling tests, for reverse osmosis membranes modi-fiedby submersion in a 0.3% (w/w) MW 200 PEGDE solution and by

treatment witha 5% (w/w) MPEGNH2solution, respectively. Com-parison of the SDS and DTAB fouling behavior of PEGDE-modifiedmembranes to surfactant fouling reports in the literature indicatesthat PEGDE modification is at least as effective as other modifi-

cations. Additionally, the strategy employed here, to compare theperformance of modified and unmodified membranes with simi-lar initial water flux, may provide a clearer picture of the effect ofsurface modification on fouling behavior.

Fouling was more extensive with DTAB than with SDS forall membranes (cf. Figs. 14(a) and 12(a)), which may resultfrom electrostatic interactions. The membrane surfaces are neg-atively charged at the pH of operation (pH 8, cf. Table 3), so

cationic DTAB is attracted to the oppositely charged membranesurface. Anionic SDS has unfavorable electrostatic interactionswith the membrane surface, yet still fouled the membrane, indi-cating the attractive hydrodynamic force (i.e., the water flux

through the membrane that brings foulants in close contact withthe membrane surface) was stronger than the repulsive electro-static force. However, although DTAB caused extensive fouling,much of this fouling was reversible, similar to the case of SDS

(cf.Figs. 14(b) and 12(b)).

3.6. Fouling resistance and flux recovery in emulsion fouling tests

Fig. 15(a) and (b) presents fouling results for membranes testedin an emulsion containingn-decane and SDS. While several mod-ified membranes retained similar fractions of their initial flux inthe emulsion and surfactant-only tests, there was more sample-

to-sample variability in n-decane:SDS tests than in SDS tests, andthe unmodified membranes fouling resistance showed a depen-dence on initial water flux in the emulsion tests that was absent inthe tests using only the surfactant (cf.Figs. 15(a) and 12(a)),which

may indicate that SDS is less likely to interact with the unmodi-fied membrane surfaces when it is emulsified with n-decane. Thecleaning procedure was also much less effective in the emulsion

fouling tests than in the surfactant-only tests. For example, XLEmembranes modified with 15% (w/w) MW 1000 PEGDE recovered90%of their initial flux inthe SDS test(cf. Fig. 12(b)),whilethey onlyrecovered 60% of their initial flux after fouling and cleaning in the

n-decane:SDS emulsion test. Considering that these membranes

retained approximately 60% of their initial flux after fouling, thecleaning procedure had a negligible effect on flux recovery. Thus,emulsion fouling was essentially entirely irreversible while surfac-tant fouling was at least somewhat reversible. In general, higher

molecular weight PEG (i.e., MW 1000 vs. MW 200) gave modifiedmembranes with better resistance to fouling by emulsions ofn-decane and SDS, similar to the behavior observed in the surfactantfouling tests.

Thefoulingresultsforemulsionsof n-decane andDTABare given

inFig. 16(a) and (b). The fouling resistance behavior observed forthe unmodified membranes was similar to that observed in theSDS and DTAB surfactant tests (i.e., fouling resistance was fairly

independent of initial water flux), suggesting the possibility ofinteractionsbetweenthe n-decane:DTAB emulsion and the unmod-ified membrane surfaces. Unlike the SDS emulsion tests, the extentof fouling in the DTAB emulsion tests waslarger than that observed

in the surfactant-only tests (cf. Figs. 16(a) and 14(a)).Fouling wasalso much more severe in DTAB emulsion tests than in SDS emul-sion tests (e.g.,XLE membranesmodified with 15% (w/w) MW 1000PEGDE retained 30% and 60% of their initial fluxes, respectively,

cf. Fig. 16(a) and 15(a)). Thus, DTAB is a more severe foulant forthese membranes than SDS, and the addition ofn-decane appearsto exacerbatefouling.These fouling results are different from those

observed by Sagle et al. for PEG hydrogel-coated polyamide mem-

Fractionoffluxretained(Jw(foul)/Jw(NaCl)i)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

30 40

Unmodified

membranes

50 60

Modified

membranes

Jw(NaCl) i

(L/(m2h))

70 80 10090

(a)

(b)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

10090807060504030

Modified

membranes

Unmodified

membranes

Fractionoffluxregained(J

w(NaCl)f/

Jw(NaCl)i)

Jw(NaCl) i

(L/(m2h))

Fig.15. (a)Fraction of fluxretained afterfouling with a 150mg/L 9:1 n-decane:SDS

emulsionand (b) fractionof flux regained afterfoulingandcleaning, asa function of

initial water flux.Definition of symbols:,LE;, L E (p =8.6bar); ,XLE;, NF90;

, 1%(w/w) MW200XLE;, 15% (w/w)MW 200 XLE;, 1%(w/w) MW1000XLE;,

15% (w/w)MW 1000XLE; , 1%(w/w) MW1000NF90; and , 15% (w/w)MW 1000

NF90. Lineshavebeen drawn throughthe modified andunmodified membranedata

to guide the eye.

branes, where n-decane:SDS fouling was much more extensivethan SDS fouling (i.e., 75% flux decline in emulsion tests vs. less

than 25% flux decline in surfactant tests), and also slightly moreextensive than n-decane:DTAB fouling (i.e., 75% flux decline in

n-decane:SDS vs. 60% flux decline in n-decane:DTAB)[16].How-ever, reports of this type are limited, and further investigation

is required to fully understand the expected fouling behavior ofsurface-modified polyamide membranes in emulsions ofn-decaneand charged surfactants.

Therewas very littledifferencein performance betweenPEGDE-

modified and unmodified membranes in the DTAB emulsion tests,with the exception of the XLE membrane modified with 15%(w/w) MW 1000 PEGDE, which did retain more of its initial flux

after fouling than the unmodified LE membrane operated at sim-


13/15


Fractionoffluxretained

(Jw(foul)/Jw(NaCl)i)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 20 40

Modified

membranes

Jw(NaCl) i

(L/(m2h))

60 80

Unmodified

membranes

100

(a)

(b)

Fraction

offluxregained(Jw(NaCl)f/Jw(NaCl)i)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 20 40

Modified

membranes

Jw(NaCl) i

(L/(m2h))

60 80

Unmodified

membranes

100

Fig.16. (a)Fraction offlux retainedafterfoulingwitha 150mg/L9:1 n-decane:DTAB

emulsion and (b) fraction of flux regained after fouling and cleaning, as a func-

tion of initial water flux. Definition of symbols: , LE; , LE (p =8.6bar), , LE

(p =5.5bar); , XLE;, NF90; , 1% (w/w) MW 200 XLE; , 15% (w/w) MW

200 XLE; , 1% (w/w) MW 1000 XLE; , 15% (w/w) MW 1000 XLE; , 1% (w/w)

MW 1000 NF90, , 15% (w/w) MW 1000 NF90; and, 15% (w/w) MW 1000 XLE

(p = 5.5 bar). Lines have been drawn through the modified and unmodified mem-

brane data (excluding data obtained for p = 5.5 bar) to guide the eye.

ilar initial water flux (30% flux retention versus less than 10%

flux retention), although the cleaning procedure did not result insharp gains in flux, consistent with results from the SDS emulsiontests. Testing this modified membrane at a lower transmembranepressure difference (p =5.5bar, giving an initial water flux of

15L/(m2 h),ratherthantheinitialwaterfluxof40L/(m 2 h) obtainedatp = 10.3 bar) revealed significantly higher flux retention thanthat of an unmodified LE membrane operated at the same trans-membrane pressure difference (60% flux retention versus 32%

flux retention) (cf. Fig. 16(a)). Additionally, at lower transmem-brane pressure difference, the cleaning procedure had a positiveeffect on flux recovery (e.g., an increase of 13%, from 60% of

flux retained after fouling to 73% of flux regained after cleaning

for the 15% (w/w) MW 1000 PEGDE-modified XLE membrane)(cf.Fig. 16(b)).

Fouling rate hasbeen reported to increase with increasing waterflux[27]. The fouling test performed here at significantly lowertransmembrane pressure difference and thus, much lower initialwater flux, indicates that fouling rate also decreases with decreas-

ing water flux, as expected. For highly fouling feeds such as theemulsion ofn-decane and DTAB, fouling may occur so quickly that,beyond the first several minutes of fouling, the observed foul-ing behavior is indicative of transport through a foulant layer on

the membrane surface that acts as the limiting resistance to masstransfer. In this case, the transport properties of the membrane,whether unmodified or modified, have little effect on the observedfouling behavior. Operation at lower transmembrane pressure dif-

ference, and thus lower water flux(decreasing therate at which thefoulant is brought to the membrane surface) could slow the rate offouling enough to allow the fouling resistance of the membrane,not a foulant layer, to determine the observed fouling behavior.

The preliminary experiment performed here confirms the feasi-bility of such an approach to limit hydrodynamic fouling effectsand focus on the effects of surface modification in fouling tests.Other approaches, such as increasing the crossflow shear rate (i.e.,

increasing the feed flowrate to sweep foulants away from the

membrane surface more effectively), have also been suggestedto decrease the fouling rate [43,44]. Although these approacheswere initially proposed as a means of reducing commercial (i.e.,

unmodified) membrane fouling, they could also allow comparisonof modified and unmodified membrane fouling behavior at condi-tions that emphasize the membrane properties, instead of those ofa foulant layer.

3.7. NaCl and organic rejection in fouling tests

Table 4presents the true NaCl rejection for each of the mem-

branesat theendof thefoulingportion ofthe test(i.e.,at t=24h)forthe four fouling conditions studied (i.e., SDS, DTAB, n-decane:SDS,and n-decane:DTAB), andit compares these valuesto theinitial true

NaCl rejection of each membrane (i.e.,measured in 2000mg/L NaClfeed). NaCl rejection increased for all membranes when they werefouled with SDS or DTAB. Following n-decane:SDS fouling, mostmembranes experienced an increase in NaCl rejection, with theexception of the unmodified NF90 membrane, whoseNaCl rejection

decreased. For the case of the n-decane:DTAB emulsion, however,all membranes experienced a decrease in NaCl rejection upon foul-ing, although the NF90 membranes (unmodified and modified)were affected more than the XLE (unmodified and modified) and

LE (unmodified) membranes (NaCl rejection of NF90 membranesdecreased by several percent versus 1% or less for XLEand LE mem-branes). These results suggest in feeds containing an emulsion,modified XLE membranes have similar or better fouling resistance

compared to modified NF90 membranes, with less negative impact

on NaCl rejection.During fouling, deposition of a foulant layer on the mem-

brane surface may further increase the surface NaCl concentration

(i.e., beyond the increase resulting from concentration polariza-tion [6,32,45]), due to hindered back diffusion of NaCl throughthe foulant layer to the bulk feed solution[44].This phenomenon,referred to as cake-enhanced osmotic pressure, leads to higher

NaCl concentration at the membrane surface, higher salt fluxthrough the membrane, and, therefore, lower NaCl rejection[44].The effect of cake-enhanced osmotic pressure increases withincreasing foulant layer thickness (i.e., increasing extent of foul-

ing)[44].Therefore, since all the membranes were severely fouledby the n-decane:DTAB emulsion (flux retention was 30% or lessfor all membranes) and the unmodified NF90 membrane was also

severely fouled by the n-decane:SDS emulsion (16% flux retention),


14/15


Table 4

Initial NaCl rejection and NaCl rejection in the presence of each foulant (measured att=24h)a .

Membrane RNaCl,i (%) RSDS(%) RDTAB (%) Rn-decane:SDS(%) Rn-decane:DTAB(%)

LE 99.1 0.2 99.9 99.70.2 99.80.1 98.90.2

XLE 99.1 0.1 99.9 99.8 99.4 98.20.4

1% (w/w) MW 200 XLE 99.1 0.2 99.9 99.60.2 99.7 98.10.2

15% (w/w) MW 200 XLE 99.3 0.1 99.9 99.70.1 99.7 98.4

1% (w/w) MW 1000 XLE 99.4 0.2 99.9 99.70.1 99.8 98.4

15% (w/w) MW 1000 XLE 99.5 0.3 99.90.1 99.60.3 99.8 98.80.4

NF90 97.8 0.3 99.8 99.8 96.3 91.70.81% (w/w) MW 1000 NF90 98.8 0.2 99.8 99.40.1 99.2 94.8

15% (w/w) MW 1000 NF90 98.8 0.4 99.90.1 99.40.2 99.00.1 92.53.7

a NaCl rejection has been corrected for the effects of concentration polarization, and is, therefore, the true NaCl rejection.

Table 5

Initial NaCl rejection and final NaCl rejection (i.e., measured after fouling and cleaning, att=30h)a .

Membrane RNaCl,i (%) RSDS,f(%) RDTAB,f(%) Rn-decane:SDS,f(%) Rn-decane:DTAB,f(%)

LE 99.1 0.2 99.70.1 99.70.1 99.60.1 99.20.3

XLE 99.1 0.1 99.7 99.7 99.2 98.20.1

1% (w/w) MW 200 XLE 99.1 0.2 99.70.1 99.60.1 99.6 98.20.7

15% (w/w) MW 200 XLE 99.3 0.1 99.70.1 99.70.1 99.6 97.6

1% (w/w) MW 1000 XLE 99.4 0.2 99.70.1 99.70.1 99.7 98.915% (w/w) MW 1000 XLE 99.5 0.3 99.7 99.70.1 99.60.1 98.90.1

NF90 97.8 0.3 99.2 99.3 95.5 94.60.5

1% (w/w) MW 1000 NF90 98.8 0.2 99.30.1 99.20.3 98.7 95.515% (w/w) MW 1000 NF90 98.8 0.4 99.30.1 99.10.3 98.60.1 95.50.1

a NaCl rejection has been corrected for the effects of concentration polarization, and is, therefore, the true NaCl rejection.

it is reasonable that decreasedNaCl rejectionwas observed in these

cases.NaCl rejectionincreased after fouling in cases where fouling was

less severe (i.e., in SDS and DTAB tests for all membranes and in

n-decane:SDS tests for all membranes except the unmodified NF90

membrane),which could indicate that thefouling layer thicknesseswere not large enough for the cake-enhanced osmotic pressureeffect, which tends to decrease rejection, to overcome the effectof surface defect plugging, which tends to increase rejection. Sur-

face defects remaining after PEGDE treatment may be plugged by

foulants adsorbing to the membrane surface, causing the observedincrease in NaCl rejection. NF90 membranes, both unmodified andPEGDE-modified, had lowerNaCl rejections than the corresponding

XLE membranes and experienced larger increases in NaCl rejectionafter fouling with DTAB and SDS, which is reasonable since NF90membranes are expected to have a larger concentration of surfacedefects than XLE membranes.

The final true NaCl rejection values (i.e., after fouling and clean-ing) of the membranes are compared to their initial true NaClrejection values inTable 5. While the final NaCl rejection valueswere generallylower than the corresponding values measured dur-

ing fouling (for the tests with SDS, DTAB and n-decane:SDS (cf.Tables 4 and 5)), they were still generally higher than the ini-tial NaCl rejections, suggesting that while some of the adsorbed

foulant (which potentially plugged surface defects, causing anincrease in NaCl rejection during fouling) was removed by clean-ing, there was also irreversible adsorption of foulant. Observed fluxrecoveries of less than 100% after fouling and cleaning also corrob-orate the hypothesis of some level of irreversible fouling. The final

NaCl rejections measured in the n-decane:DTAB tests (and in the

n-decane:SDS test for the unmodified NF90 membrane) were gen-erally lower than the initial rejection values, which suggests thatthe foulant layer was not removed by the cleaning procedure (i.e.,

the cake-enhanced osmotic pressure effect was significant evenafter cleaning). Negligible flux recovery in these tests supports thehypothesis that a relatively thick foulant layer was still present onthe membrane surface after cleaning.

Within the detection limits of the TOC analyzer, organic

rejection was essentially 100% in all fouling tests. Organic

rejection greater than 95% has been documented previously

for polyamide membranes (i.e., unmodified and coated withPEG-based hydrogels) tested with the same organic foulants[16].

4. Conclusions

An aqueous top surface treatment method was employed tograft poly(ethylene glycol) diglycidyl ether (PEGDE) to the surfaces

of commercial reverse osmosis (XLE) and nanofiltration (NF90)membranes. Water flux decreased due to this surface modifica-tion, as expected, and NaCl rejection increased. However, in manycases, an initial large decrease in water flux was observed forlow grafting density (i.e., for membranes treated with 1% (w/w)

MW 1000 PEGDE), while a much smaller subsequent decrease inwater flux was observed for high grafting density (i.e., for mem-branes treated with 15% (w/w) MW 1000 PEGDE), suggestingthat after a small amount of PEGDE is grafted to the membrane

surface, additional PEGDE has much less of an impact on mem-brane performance. Membranes modified with PEGDE generallydemonstrated improved fouling resistance to charged surfactantsand emulsions containing n-decane and a charged surfactant,

but experienced minimal changes in surface properties (e.g., sur-

face charge, hydrophilicity and roughness). The observed effectscould be related, in part, to steric hindrance imparted by the PEGchains, preventing foulants from closely approaching the mem-

brane surface. Comparison of the fouling behavior of modifiedmembranes to that of unmodified membranes of similar initialwater flux indicated the observed improvements in fouling resis-tance were not due to the decrease in initial water flux caused

by PEGDE modification. Constant flux operation (by adjustmentof transmembrane pressure difference to achieve and maintaina desired water flux during the fouling test), versus the con-stant pressure method used in this work, is commonly used in

field operations and could therefore provide further insight intopredicted membrane fouling resistance. PEGDE molecular weighthad a stronger influence on fouling resistance than did PEGDE

treatment concentration, suggesting modification with lower con-


15/15


centrations (i.e., less than 1% (w/w)) of higher molecular weight(i.e., greater than 1000) PEGDE may be a means of optimizing the

balance between water flux and fouling resistance for modifiedmembranes.

Acknowledgements

This work was prepared with the support of the U.S. Depart-ment of Energy, under Award No. DE-FC26-04NT15547. However,any opinions,findings, conclusions, or recommendations expressedherein are those of the authors and do not necessarily reflect the

views of the DOE. We also gratefully acknowledge partial supportof this research by the National Science Foundation (Grants DMR0423914, IIP-0917971 and CBET-0932781/0931761). The authorsthank Dr. Bill Mickols of Dow Water& Process Solutions for helpful

discussions, Dr. Hugo Celio of the Center for Nano and MolecularScience & Technology at the University of Texas at Austin for per-forming the XPS analysis, and the National Science Foundation forfunding the X-ray photoelectron spectrometer used in this work

(Grant No. 0618242). Graduate student support through a NationalScience Foundation Graduate Research Fellowship and a SandiaNational Laboratories/University of Texas at Austin Excellence inEngineering Fellowship is gratefully acknowledged.

References

[1] G.M. Geise, H.-S. Lee, D.J. Miller, B.D. Freeman, J.E. McGrath, D.R. Paul, Waterpurification by membranes: the role of polymer science, Journal of PolymerScience: Part B. Polymer Physics 48 (2010) 16851718.

[2] L.F. Greenlee, D.F. Lawler, B.D. Freeman, B. Marrot, P. Moulin, Reverse osmo-sis desalination: water sources, technology, and todays challenges, WaterResearch 43 (2009) 23172348.

[3] R.F. Service, Desalination freshens up, Science 313 (2006) 10881090.[4] Review of the Desalination and Water Purification Technology Roadmap, The

National Academies Press, Washington, DC, 2005.[5] R.J. Petersen, Composite reverse osmosis and nanofiltration membranes, Jour-

nal of Membrane Science 83 (1993) 81150.[6] R.W. Baker, Membrane Technology and Applications, 2nd ed., John Wiley &

Sons Ltd., West Sussex, 2004.[7] C. Fritzmann, J. Lwenberg, T. Wintgens, T. Melin, State-of-the-art of reverse

osmosis desalination, Desalination 216 (2007) 176.[8] J.A. Veil, M.G. Puder, D. Elcock, R.J. Redweik, A White Paper Describing Pro-ducedWater fromProduction of Crude Oil,Natural Gas,and CoalBed Methane.Contract No. W-31-109-Eng-38, U.S. Department of Energy National EnergyTechnology Laboratory, Argonne National Laboratory, 2004.

[9] M. C akmakci, N. Kayaalp, I. Koyuncu, Desalination of produced water from oilproduction fields by membrane processes, Desalination 222 (2008) 176186.

[10] J.S. Louie, I. Pinnau, I. Ciobanu, K.P. Ishida, A. Ng, M. Reinhard, Effects ofpolyetherpolyamide block copolymer coating on performance and foulingof reverse osmosis membranes, Journal of Membrane Science 280 (2006)762770.

[11] D.K.Han,K.D.Park, K.-D. Ahn,S.Y.Jeong,Y.H.Kim, Preparationand surfacechar-acterization of PEO-grafted and heparin-immobilized polyeurethanes, Journalof Biomedical Materials Research: Applied Biomaterials 23 (1989) 87104.

[12] J.H. Lee, H.B. Lee, J.D. Andrade, Blood compatibility of polyethylene oxide sur-faces, Progress in Polymer Science 20 (1995) 10431079.

[13] E. Ostuni, R.G. Chapman, R.E. Holmlin, S. Takayama, G.M. Whitesides, A surveyof stru

freeman surface modification of desalination membranes 2011

Documents