A

ftmwcmoesi©

K

1

bflearpic(etrb

0d

Available online at www.sciencedirect.com

Journal of Membrane Science 305 (2007) 176–184

Impact of chemical cleaning on properties and functioning ofpolyethersulfone membranes

Elizabeth Arkhangelsky, Denis Kuzmenko, Vitaly Gitis ∗Unit of Environmental Engineering, Ben-Gurion University of the Negev, POB 653, Beer-Sheva 84105, Israel

Received 1 February 2007; received in revised form 1 August 2007; accepted 3 August 2007Available online 10 August 2007

bstract

In water treatment processes, membrane cleaning is of paramount economic and scientific importance. Traditionally, ultrafiltration membranesouled with organic matter are cleaned with hypochlorite, an operation that usually results in complete restoration of the initial flux. Althoughhe cleaned membranes are deemed suitable for continued operation, membrane integrity may have been damaged (manifested as holes in the

embrane skin) by the cleaning process. To assess the degree of chemical resistance of polyethersulfone (PES)-based membranes, the membranesere fouled with bovine serum albumin and cleaned with various concentrations of sodium hypochlorite. The properties of pristine, fouled, and

leaned membranes were compared by spectrochemical analysis, mechanical testing, microscopy imaging, and streaming potential and permeabilityeasurements. Our results suggest that cleaning does have a major impact on the performance and properties of PES membranes. Effective removal

f the foulant from the membrane surface resulted in more severe fouling and increased electronegativity of the cleaned membrane. The increased

lectronegativity was related to chain scission of the polymer, leading to the formation of phenyl sulfonate. The deterioration in the mechanicaltrength of hypochlorite-treated membranes indicated a loss of membrane integrity. The transmembrane streaming potential technique providednformation on membrane pore size, cleaning efficiency, and membrane integrity.

2007 Elsevier B.V. All rights reserved.

leach

hmrNWmcogh

dt

eywords: Ultrafiltration; Cleaning agents; Fouling; Charge characterization; B

. Introduction

Chemical cleaning of ultrafiltration (UF) polymer mem-ranes is a widely accepted procedure aimed at minimizingux loss due to irreversible fouling. Although there are five cat-gories of cleaning chemicals (caustic, oxidants/disinfectants,cids, chelating agents, and surfactants [1]), hypochloriteemains a popular choice because of its availability, reasonablerice, and capacity to prevent biofouling via efficient clean-ng and biosanitation. A new class of UF membranes made ofellulose triacetate (CA), polysulfone (PSf), polyethersulfonePES), polyacrylonitrile (PAN), or polyvinyldifluoride (PVDF)xhibit enhanced chemical resistance and, as such, they are

reated with hypochlorite at water treatment facilities as a part ofoutine clean-in-place operations. Although the cleaned mem-ranes are considered suitable for continued use, they may

∗ Corresponding author. Tel.: +972 8 6479031; fax: +972 8 6472983.E-mail address: gitis@bgu.ac.il (V. Gitis).

iXtamtt

376-7388/$ – see front matter © 2007 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2007.08.007

solution; Permeability

ave suffered damage, which is manifested as holes in theembrane skin [2]. Previous reports on the effect of hypochlo-

ite treatment on UF membranes indicated a flux increase inaOCl-treated membranes [3,4]. This increase was explained byolff and Zydney [5] in terms of a direct relationship betweenembrane pore size and bleach treatment duration. Reported

hanges in streaming potential [6] and increased hydrophilicityf hypochlorite-cleaned membranes were considered advanta-eous, provided that the UF membranes do not react with theypochlorite blends [3].

A variety of novel instrumental tools have recently been intro-uced to the field of UF membrane surface characterizationo complement conventional flux, contact angle, and stream-ng potential measurements. The new techniques, which include-ray photoelectron spectroscopy (XPS), attenuated total reflec-

ion Fourier transform infrared (ATR-FTIR) spectroscopy, and

tomic force microscopy (AFM), facilitate the investigation ofembrane morphology at length scales spanning a few nanome-

ers to hundreds of micrometers [7]. XPS is a surface-sensitiveechnique that measures the elemental composition (except H)

Memb

atdfAnmdii

rbddchMthlcacttihQfltetfbraami

p1faordccmcwocA0

biUatn

pPgsptait

2

2

(G1aPn(tcah

np–witpf(i(filtration of 50 mL of the feed suspension through the membrane.

Cleaning experiments were performed by soaking pieces ofthe fouled PES membranes in deionized water (MilliQ purity)containing free chlorine, 150 mg/L in Petri dishes with closed

E. Arkhangelsky et al. / Journal of

nd provides chemical binding information for the top 1–5 nm ofhe surface region. ATR-FTIR spectroscopy penetrates to greaterepths (from <200 nm to >1 �m) depending on, among otheractors, the incident wave number and the incident angle [8].lthough less quantitative than XPS, ATR-FTIR provides sig-ificant qualitative detail about the types of functional groups,aking it a highly complementary method to XPS. AFM pro-

uces micrograph images of the membrane surface together withts corresponding physical properties, such as topography andnteratomic forces.

Most membrane characterization efforts have focused oneverse osmosis membranes, and, as a result, very little work haseen done on UF membranes. The scientific literature, however,oes contain some preliminary reports regarding the possibleeformation of UF membranes after hypochlorite treatment:hain breaking in PSf molecules after extended exposure toypochlorite was observed by Rouaix et al. [9] and Gaudichet-aurin and Thominette [10], and the oxidation of CA, leading

o the formation of various end groups, such as carboxyl, alde-yde, and ketone, was reported by Gitis et al. [11]. There is veryittle information in the literature on the effect of hypochloriteleaning on PES membranes, mainly because PES membranesre considered highly tolerant to oxidants (>250,000 mg h/L forhlorine) and to a wide pH range between 2 and 12. Amonghe few reports that have appeared, Wienk et al. [3] describedhe gradual dislodging of polyvinyl pyrrolidone (PVP), usedn the initial casting of PES membranes as a pore former andydrophilizer, but such an effect was viewed as an advantage byin et al. [12], who reported a five-fold increase in membraneux and a narrowed pore size distribution after the hypochlorite

reatment. Thominette et al. [13] reported a decrease in elasticlongation of PES membranes exposed to hypochlorite solu-ion. Kuzmenko et al. [14] reported that effective oxidation withree chlorine resulted in complete restoration of the initial fluxut caused more severe fouling after each cleaning. Althougheports on PES membranes portray the hypochlorite treatments a safe and recommended procedure needed to increase fluxnd minimize fouling, the contradictory reports on PSf and CAembranes imply that the integrity of PES membranes will be

mpaired, despite their stability in hypochlorite solutions.At water treatment facilities the addition of free chlorine is

erformed to either prevent the biofouling by back flushing ofmin with 2–8 mg/L of NaOCl, or to clean the membrane sur-

ace from foulants by soaking in 20–400 mg/L free chlorine forpprox. 1 h. Although at treatment plant the membranes are mostf the time in contact with water and just periodically with chlo-ine, the combined effect of oxidation can be expressed as totalose (concentration × contact time) of hypochlorite. The appli-ability of C·t concept to polymer oxidation was shown [2] onhlorine cleaning of CA membranes. In the current study the PESembranes were dipped into solutions containing 150 mg/L free

hlorine in continuous manner and left to soak. The experimentsere performed at constant temperature and pH to pertain the

xidizing potential. The chosen pH of 7.2 is that at which theoncentrations of hypochlorous and hypochlorite ions are equal.s oxidizing potentials of the two are different (1.49 V versus.94 V, respectively) the more significant degradation in mem-

rane Science 305 (2007) 176–184 177

rane properties is expected with HOCl. Recent reports [9,15]ndicate that changes in pH alter the mechanical strength ofF membranes when lower mechanical strength was observed

t lower pH. It was decided to perform the study in condi-ions where presence of both forms of free chlorine cannot beeglected.

The current study thus aims to systematically compare thehysicochemical properties of pristine and hypochlorite-treatedES membranes. The main question addresses whether theradual alteration of microscopic properties of the membraneurface, such as hydrophilicity and surface charge, affects theerformance of PES membranes. The current study showedhat hypochlorite treatment does, in fact, affect the performancend surface properties of PES membranes. These findings havemportant implications for the design of chemical cleaning pro-ocols for PES membranes used for surface water treatment.

. Experimental

.1. Membrane degradation experiments

PES membranes manufactured by Sterlitech CorporationKent, WA, USA) and Microdyn-Nadir GmbH (Wiesbaden,ermany) with molecular weight cut-offs (MWCO) of 20 and0 kDa, respectively, were used. The membranes were supplieds flat sheets and were stored dry under ambient conditions. TheES polymer comprises phenylene rings connected by alter-ate sulfonyl (SO2) functional groups and ether (–O–) linkagesFig. 1 ). The sulfonyl groups confer rigidity (with a high glassransition temperature) and, together with the ring structures,hemical resistance and relative hydrophobicity. The ether link-ges make the polymer less hydrophobic and more flexible,ence more amenable to processing.

The membranes were characterized by conventional tech-iques – measurements of flux, contact angle, streamingotential, tensile strength at break, and pore size distributionand by XPS, ATR-FTIR, and AFM. Each piece of membraneas thoroughly rinsed with deionized water at 50 ◦C by shaking

t at 1.66 Hz (100 rpm) on a mechanical shaker for 1 h beforehe experiment. The fouling experiments were performed asreviously described a 150-mL stirred cell [14]. In brief, theeed comprised a solution of 0.3 g of bovine serum albuminBSA) per liter of phosphate buffer solution (pH 7.2). The foul-ng experiments were performed at 25 ± 1 ◦C and 1 × 105 Pa1 bar) nitrogen pressure for 30 min, i.e., sufficient time for the

Fig. 1. Molecular structure of polyethersulfone.

1 Membrane Science 305 (2007) 176–184

l1TdsaHrif

2

2

t(sTdt(

2

(aatrtntwutw

2

pfl(ituabTbbpT0etv

avhs

eo51ecMtia

bKS

ξ

wtto

2

w

78 E. Arkhangelsky et al. / Journal of

ids. Commercially available bleach (Unilever Best Foods Israel,5 g/L free chlorine) was used as the hypochlorite source.he concentration of free chlorine in the stock solution wasetermined by the 4500-Cl B iodometric method I [16]. Fresholutions were prepared daily. Chlorination was performed atconstant pH of 7.2, corrected if necessary with concentratedCl and NaOH solutions. The membranes were immediately

insed with deionized water after they had been soaked for var-ous times to give doses (concentration C × exposure time t) ofree chlorine of 10 or 100 g h/L.

.2. Methods of characterization

.2.1. X-ray photoelectron spectroscopyXPS analyses were performed with a PHI-5000C ESCA sys-

em (Perkin-Elmer) with Al K� radiation as the X-ray source1486 eV). The electron flood gun was operated at an energyetting of 3 eV to compensate for membrane surface charging.he membrane sample was directly pressed to a self-supportedisk (10 mm × 10 mm), mounted on a sample holder, and thenransferred into the analyzer chamber. The whole spectrum0–1200 eV) was recorded four times for each sample.

.2.2. ATR-FTIR spectroscopyATR-FTIR spectra were recorded on a Nicolet spectrometer

model 5PC, Thermo Electron, Waltham, MA, USA). The ATRccessory contained a ZnSe crystal (25 mm × 5 mm × 2 mm) atnominal incident angle of 45◦, yielding about 12 internal reflec-

ions at the sample surface. All spectra (100 scans at 4.0 cm−1

esolution and subtracted from the appropriate background spec-ra) were recorded at 25 ◦C. The instrument was purged with dryitrogen to prevent atmospheric moisture from interfering withhe spectra. Samples taken from the membranes and stored inater were blotted with clean filter paper to remove excess liq-id and then dried in air. Finally, to complete the drying process,hey were placed in a desiccator over P2O5 for 2 h. The samplesere then clamped to the ATR crystal.

.2.3. Streaming potentialThe streaming potential was measured in the stirred cell by

umping a KCl solution through the membrane pores, i.e., theow was directed perpendicular to the membrane active layerFig. 2 ). To this end, the membrane cell was remodeled tonclude a salt bridge between the electrodes on the feed andhe permeate sides of a membrane cell, in a setup similar to thatsed by Pontie et al. [17]. The membrane sample was placedbove a PTFE spacer (0.2 mm in thickness) with the mem-rane skin-layer facing the flow channel (200 mm × 74 mm).he electrolyte solution was forced through the membraney static pressure given by nitrogen. Gas pressure was sety a precision regulator equipped with a digital pressure dis-lay (IR2000-FO2 and ISE40, respectively, SMC Corporation,okyo Japan) at a range of 0–2 × 105 Pa (0–2 bar) and with a

.01 × 105 Pa (0.01 bar) accuracy. The electrical potential differ-nce was measured between the feed and permeate by openinghe back-pressure valve, thus allowing the feed electrolyte toacuum soak into a vial containing an electrode. Such a setup

lobv

Fig. 2. Experimental device used for streaming potential measurements.

llows the electrodes to be kept separate from the cell and pre-ents the electrokinetic effects that arise from the convectiveydrodynamic movement of the electrolyte near the electrodeurface.

For streaming potential measurements, the membrane wasquilibrated by soaking it in a 0.001 mol/L KCl solutionvernight. Measurements were performed for 0.5, 1, 10 and0 mM KCl at pressures of 0.5 × 105 and 1 × 105 Pa (0.5 andbar) in the pH range 3–10.5 (at pH values higher than 10.5, thelectrode is unstable; at pH measurements lower than 3, solutiononductivity will be too large and will affect the measurement).easurements were conducted with Ag/AgCl reference elec-

rodes (model 723/733, Metrohm Ltd., Switzerland) and a highmpedance digital multimeter (EDM 2347, Escort, Taiwan). Thesymmetry potential of the electrode pair was less than 1 mV.

The streaming potential values for pristine and fouled mem-ranes were calculated from four measurements for eachCl strength/pressure combination by applying the Helmholtz-moluchowski equation (Eq. (1)):

= �E

�P

ηk

ε0ε(1)

here �E is the potential difference across the membrane, �Phe applied pressure, η the dynamic viscosity of the electrolyte, khe conductivity of bulk electrolyte, ε the dielectric permittivityf water, and ε0 is the permittivity of vacuum.

.2.4. Sieving testsThe molecular determination of UF pore size distribution

as performed by measuring membrane rejection of polyethy-

ene glycols (PEG) with different molecular weights. Solutionsf PEG were preferred to solutions of dextrans or polystyrenesecause of the low susceptibility of PEG to conformationalariations in response to changes in temperature and pH. The

Memb

mmi(pm

P6tGdcco(f

R

wio

eb

r

w

a

f

wtswpa

2

Cmiamfmm(f

2

(nswimtfiTaoo

2p

Tt

3

3

casta[as

pristine and BSA-fouled membranes, but the peak was absent inthe NaOCl-treated membrane. The origin of this peak is prob-ably the nitrogen-containing, nonionic, water-soluble polymerPVP, which is added to the original membrane blend as a spacer

E. Arkhangelsky et al. / Journal of

onodispersivity (the ratio of the weight-average molecularass to the number-average molecular mass) of the PEGs

s reported to be in the 1.1–1.2 range by the manufacturerSigma–Aldrich Co.). Additionally, polydispersed syntheticolymers such as PEG show negligible interactions with PESembranes.A separate rejection test was performed for each of eight

EG molecules (molecular masses of 300, 600, 2000, 3350,000, 10,000, 20,000, and 35,000 Da, Sigma–Aldrich) andhree polyethylene oxide (100, 200 and 600 kDa, Fluka GmbH,ermany) polymers. The influent polymer solution (3 g/L ofeionized water) was transferred through a PES membrane at aonstant pressure of 0.506625 × 105 Pa (0.5 atm). The effluentoncentration was determined by a combustion-infrared methodn a total organic carbon analyzer Apollo 9000 TOC analyzerTekmar Company, Cincinnati, OH). The rejection coefficientor each polymer was calculated according to Eq. (2):

=(

1 − Ci,p

Ci,0

)× 100% (2)

here Ci,p is the concentration of polymer of molecular mass in the permeate (mg/L), and Ci,0 is the concentration of polymerf molecular mass i in the feed (mg/L).

The Stokes radius of macromolecules that served for sievingxperiments was determined from their diffusivities in a solutiony using the following approximation [18]:

(nm) = 0.078M0.33 (3)

here M is the molecular weight of the polymer molecule.Pore size distribution was obtained using a log-normal prob-

bility density function [19]:

(dp) = 1

dp ln σp√

2πexp

[−1

2

(ln(dp/μp)

lnσp

)2]

(4)

here dp is the pore diameter (m), f(dp) the solute separation, μphe geometric mean pore diameter (m), and σp is the geometrictandard deviation of the membrane. The function parametersere determined from the rejection coefficient-hydraulic radiuslot, where μp was calculated as ds corresponding to f = 50%nd σp = ds90/ds50.

.2.5. Contact angle, permeability and flux measurementsContact angle was measured with a goniometer (Rame-Hart,

A) using the sessile drop method to determine an index ofembrane hydrophobicity. In this test a ∼20 �L drop of deion-

zed water was placed onto the dried membrane surface with

microsyringe, and the air–water–surface contact angle waseasured within 10 s. Contact angle measurements were per-

ormed in triplicate using separate pieces of membrane. As aeasure of membrane permeability, pure water flux was deter-ined by filtration of deionized water for 30 min at 1 × 105 Pa

1 bar) N2 pressure. Then the deionized water was shifted to theeed solution to carry out the experiment.

Fb

rane Science 305 (2007) 176–184 179

.2.6. Atomic force microscopyAFM images were obtained on a Dimension 3100 instrument

Digital Instruments-Veeko, Santa Barbara, CA) using siliconitride DNP tips (Digital Instruments-Veeko) with nominalpring constants of 0.32 N/m. The contact mode was employed,hich makes use of “hard core” repulsive (but very short range)

nteractions between the tip and the surface. A strip severalicrons wide was first scanned in the dry state. The tip was

hen disengaged without moving the sample, the fluid cell waslled with water, and the same area was imaged under liquid.o obtain the closest possible correspondence between the drynd wet images, it was typically necessary to correct the lateralffset (X–Y) a few times until satisfactory correspondence wasbtained. Each image represents a 0.5 �m × 0.5 �m scan area.

.2.7. Tensile strength at break measurement and scanningrobe microscopy studies

The experiments were conducted with Universal Mechanicalester UMT-2 (CETR, Campbell, CA, USA) at the manufac-

urer’s laboratories, as described in a separate publication [20].

. Results and discussion

.1. X-ray photoelectron spectroscopy

XPS experiments were conducted to determine the surfaceomposition of membrane samples and demonstrate their rel-tive changes upon chemical treatment (Fig. 3 ). The XPSpectrum of the pristine PES membrane (upper curve) showedhat the membrane is composed of sulfur [3.7 at.% (2p at 170 eV)nd (2s at 234 eV)], carbon [50.4 at.% (1s at 284.6 eV)], nitrogen5.9 at.% (1s at 403 eV)], and oxygen [37.2 at.% (1s at 537 eV)nd 1.5 at.% (2s at 28 eV)]. Hydrogen was not included in thepectrum since XPS cannot detect hydrogen.

A well-pronounced N peak was present in the spectra of both

ig. 3. XPS spectra of pristine, BSA-fouled and hypochlorite-treated PES mem-rane.

180 E. Arkhangelsky et al. / Journal of Membrane Science 305 (2007) 176–184

Table 1Elemental content concentrations in pristine and hypochlorite-treated PESmembranes

Sample Element concentration (at.%)

C S O

Pristine 50.4 6.7 38.710 g h/L 51.3 8.5 38.6100 g h/L 52.2 9.3 38.5

aibmibcb

ahdcpfi.pt

3

PaTm

TA

I(

11111

Fc

avtaPap

3

tomsoGd1TiCoo

from 28.5 nm at 0.1 mM KCl to 0.9 nm at 0.1 M KCl (Table 3). The 4-nm limit corresponds to a 5 mM KCl concentration,

Fig. 4. Proposed mechanism of NaOCl attack on PES membrane.

nd hydrophilicity promoter. In the course of the cleaning exper-ments this polymer is washed as the filtration progresses, as maye seen from the absence of the N peak in the NaOCl-cleanedembrane. The slight increase in nitrogen concentration (to 7%)

n the BSA-fouled membrane may be attributed to the two amineonds in the protein structure. The spectrum of the NaOCl-leaned membrane shows that a small quantity of chlorine wasound to the membrane surface [0.4 at.% (2p at 1.6 eV)].

An examination of the concentrations of elemental C, S,nd O on the membrane surfaces indicates that in response toypochlorite cleaning carbon and sulfur increased and oxygenecreased (Table 1 ). The elevated concentrations of sulfur andarbon in the upper part of the PES membrane may indicateartial scission of the sulfonyl Ph–S bond. In the presence of suf-cient hypochlorite the possible mechanism is shown in Fig. 4The reaction occurs through an intermediate step in whichhenyl sulfinate PhSO2 dissociates into PhSO2

− followed byhe formation of PhSO3

− [21].

.2. ATR-FTIR spectroscopy

The ATR-FTIR spectra of the pristine PES membrane and of

ES membranes treated with doses of 10 and 100 g h/L chlorinere shown in Fig. 5 , and the peak assignments [22] are given inable 2 . The strong bands at 1650 and 1578 cm−1 represent aro-atic bands characteristic of the PES membrane [8]. The sharp

able 2ssignment of the relevant IR bands in the range of 1800–800 cm−1

R bandcm−1)

Range given in theliterature [22]

(cm−1) Assignment

650 1620–1680 Aromatic stretching vibrations578 About 1580 Aromatic systems485 1460–1550 C–S stretch241 1275–1200 C–O–C150 1160–1120 SO2 stretching

m

ig. 5. ATR-FTIR spectra of pristine PES membranes and PES membranesleaned with 10 and 100 g h/L NaOCl.

bsorption peaks at 1150 cm−1 were ascribed to the symmetricibration of the SO2 group, and the distinct peaks at 1241 cm−1

o C–O–C vibrations. The intensity of the peak at 1485 cm−1,ttributed to C–S vibration, was lower in the chlorine-treatedES membranes than in the pristine membrane, which indicatedweakening of the C–S bond, as required by the formation ofhenyl sulfonate.

.3. Streaming potential and sieving

Measurements of transmembrane streaming potential requirehat conditions for the Helmholtz-Smoluchowski approximationf the Poisson-Boltzmann equation for strong electrolytes areet, i.e., that the pore size is larger than the Debye length k. Fig. 6

hows the pore size distribution function calculated on the basisf the retention values obtained from sieving experiments. Aaussian distribution was obtained with a hydrodynamic meaniameter of 3.2 nm and a tail for pore size values larger than2 nm, a size usually not encountered in normal distributions.he 8 nm value was approximated as a MWCO of 20 kDa, which

s the MWCO value claimed by the manufacturer (Sterlitechorp.). The distribution displayed in Fig. 6 suggests that tobtain meaningful streaming potential values, the Debye lengthf the membrane pore walls should not exceed 4 nm.

For our system, the Debye length decreased exponentially

eaning that the transmembrane streaming potential should be

Fig. 6. Pore size distribution function of PES-20 membranes.

E. Arkhangelsky et al. / Journal of Memb

Table 3Calculated values of the Debye length

KCl (mM) Debye length (nm)

0.1 28.460.5 12.71 95 4.02

10 2.8450 1.27

1

T

moiapftidssa

FN1

i1htti

ulsoebgo

i7ii

00 0.9

he relative dielectric permittivity of KCl was taken as 70 [25].

easured with an electrolyte having an ionic strength of 10 mMr higher. Experimental confirmation was obtained from stream-ng potential measurements for KCl concentrations of 0.5, 1, 10nd 50 mM (Fig. 7A ). Streaming potential measurements wereerformed at pH values ranging from 3 to 10.5. None of theour curves crossed the pH axis, probably because the isoelec-ric point of PES is located at superacid conditions. However, its not possible to determine the isoelectric point under such con-

itions, since they do not fall within the pH range for chemicaltability provided by the manufacturer. All the PES membraneamples were negatively charged, and as the pH became morelkaline, the absolute values of the streaming potential measured

ig. 7. Plots of streaming potential of PES-20 membrane vs. (A) pH and (B)aOCl concentration for a pristine membrane (�) and at KCl concentrations ofmM (©) and 10 mM (�).

−riPmiw7dm

tFwiromcmop

p

rane Science 305 (2007) 176–184 181

n 10 mM KCl increased from −5 mV at pH 3 to −12 mV at pH0.5, in keeping with the findings of Wang et al. [23]. Since PESas no dissociated functional groups, specific ion adsorption ishe only possible process for the formation of surface charge, i.e.,he initial charge can be attributed to the adsorption of hydroxylons on the PES membrane surface.

In the pH range of 3–10.5, streaming potential absolute val-es obtained for 10 and 50 mM KCl solution were significantlyower than those for 1 and 0.5 mM solutions. With Debye lengthmaller than pore size, the membrane allows partial transferf both chloride and potassium and ceases to function as anlectronegative barrier. For 1 and 0.5 mM solutions the UF mem-ranes exhibited charge-selective behavior and contributed to theap in KCl concentrations between the feed and permeate sidesf the membrane.

A similar set of measurements were performed after soak-ng the membrane samples in aqueous chlorite solutions (pH.2) at dosages of 20–120 g h/L (Fig. 7B). Chemical treatmentncreased the negative charge on the membrane. The stream-ng potential dropped from −11 mV for a pristine membrane to

22 mV for a membrane treated with 120 g h/L of free chlo-ine. A possible explanation for the increased electronegativitys the partial scission of the Ph–S bond and formation of chargedhSO3

− compounds as explained in Section 3.1. The measure-ents performed with 1 and 10 mM of KCl produced almost

dentical results, in contrast to the results shown in Fig. 7A,here the streaming potential measured with 1 mM KCl at pH.2 was much more negative. We therefore conclude that oxi-ation with hypochlorite resulted in a gradual enlargement ofembrane pore size.The enlargement was verified with solute tests and virus fil-

ration experiments. The results of solute tests are plotted inig. 8 as percentage polymer retention of versus its moleculareight. A significant increase in polymer retention after chem-

cal cleaning was observed for all the solutes starting with 3%etention of PEG with molecular weight of 200 Da. Applicationf 120 g/L h free chlorine resulted in 100% retention of poly-ers with molecular weight of 20 kDa and higher. The straight

onclusion of the observed tendency is that hypochlorite treat-ent results in significant narrowing of membrane pores. The

bservation is supported by previous reports [3,12] of a tight

ore size distribution after the hypochlorite treatment.

Arylsulfonate groups however are known to react witholyethylene glycols to form amorphous matrix [24]. It is pos-

Fig. 8. Solute separation of pristine and chlorine-treated membranes.

1 Membrane Science 305 (2007) 176–184

sarpwiwtgdtcoos

3

etTciAmPtcbic

56mtd

Fe

82 E. Arkhangelsky et al. / Journal of

ible that Ph-SO3− end groups formed after the cleaning have

n increased affinity to PEGs and the performed solute testsesulted in adsorption of PEGs on the membrane surface. Thatossibility was verified with filtration experiments performedith bacteriophages MS2, phi X 174 and T4 [18]. A signif-

cant reduction in viral retention after hypochlorite cleaningas observed. Based on viral experiments and streaming poten-

ial measurements we conclude that oxidation resulted in aradual enlargement of membrane pore size beyond the 9 nmiameter value. As a result, we propose that streaming poten-ial measurements with electrolytes of different ionic strengthsould constitute an additional method for the characterizationf membrane pore size distribution, particularly in the absencef a reliable technique for the direct measurement of poreize.

.4. Contact angle and permeability measurements

Hydrophilicity or wettability of the PES membrane wasvaluated from measurements of the contact angle betweenhe membrane surface and the air/water interface (Fig. 9 ).he finding that the contact angle decreased with increasinghlorination contradicts previous assumptions [3] that dislodg-ng the PVP would lead to increased contact angle values.gain, one possible explanation is the partial ionization of theembrane surface due to the formation of phenyl sulfonate,hSO3

− (charged compounds are usually more hydrophilichan uncharged compounds). The increased hydrophilicity ofleaned membrane can also be explained by formation ofigger pores due to membrane degradation. Increased poros-ty could be responsible of spreading out of the drop byapillarity.

The initial permeability of the PES-20 membrane was0 ± 1 × 10−5 L/m2 h Pa (50 ± 1 L/m2 h bar) and increased to

−5 2 2

0 × 10 L/m h Pa (60 L/m h bar) after each chlorine treat-ent [14] (Fig. 10 ). The observed increase may be related to

wo parallel phenomena, ionization of the membrane surface,ue to the partial scission of the Ph–S bond, and dislodging of

ig. 9. Contact angles of pristine PES membrane and of PES membranesxposed to various doses of hypochlorite.

F(

PcBdmpp

3

ciR(dfsp

ig. 10. Fouling of PES membranes cleaned by free chlorine at doses of 5 g h/L1), 24 g h/L (2), 72 g h/L (3), and 120 g h/L (4). Adapted from Ref. [14].

VP, leading to increased pore diameter. Thus, each consecutiveleaning leads to higher pure water flux and a higher degree ofSA fouling. Treated membranes exhibited an increased ten-ency to fouling with rising hypochlorite concentration. Theoderate fouling of 10% for 1 g/L free chlorine was hardly com-

arable to the 30% fouling observed when the clean-in-place waserformed with 5 g/L.

.5. Atomic force microscopy

AFM images of the active surface layers of pristine andhlorine-treated PES membranes are shown in Fig. 11 . Bothmages were taken of the same scan area of 0.5 �m × 0.5 �m.oughness, which is calculated as the peak-to-valley distance

the distance between the highest data point and the lowest

ata point of the surface), was 14.5 nm for pristine and 9.8 nmor NaOCl-cleaned membranes. We assumed a direct relation-hip between the size of the sink leading to the membraneore and the pore size. A bigger sink, then, probably indi-

E. Arkhangelsky et al. / Journal of Membrane Science 305 (2007) 176–184 183

(A) an

ctilv(

3

omgmta

3

sc2ct

Fr

dsmrcmt

oAmiotmmlni

m

Fig. 11. AFM images of the surface of the active layer of pristine

ates a larger pore size. Although qualitative, the scans suggesthat the average size of sinks in chlorine-treated membraness larger which, in turn, may be indicative of correspondinglyarger pore sizes in chlorine-treated membranes, as was observedia streaming potential measurements and virus filtration testsSection 3.3).

.6. Tensile strength at break measurement

As the treatment dose increased from 0 (pristine) to 50 g h/Lf free chlorine, the tensile strength at break of the PES-20embrane decreased from 16.4 to 11.6 MPa (Fig. 12 ); elon-

ation at break decreased from 4.55% to 3.85%, and Young’sodulus values declined from 130 to 90 MPa [20]. The results

hus indicate that membrane mechanical strength deteriorateds treatment dose increased.

.7. Discussion

UF membranes made of PSf, PES, PVDF and CA are con-idered highly chlorine-tolerate. The total concentration of free

hlorine that can be applied for membrane cleaning is as high as50 g, and up to that level the membrane is considered chemi-ally and mechanically stable. Recent reports however indicatehat the membrane’s polymer structure changes upon the oxi-

ig. 12. Tensile strength at break for PES-20 membrane cleaned with hypochlo-ite.

cfpbf0irompwbbmmrtfm

d chlorite-treated (B) PES membranes. Images are 5 �m × 5 �m.

ation with free chlorine, and the alteration occurs at levelsufficiently lower than the 250 g. Two recent studies on PSfembranes [9,10] point on structural changes as total chlo-

ine concentration reaches 1 g threshold. In one study [10] thehanges of the polymer structure were linked to membraneechanical strength, and it was shown that after 1 g free chlorine

hreshold the membrane mechanical strength deteriorates.The current research indicates that the chlorine treatment

f PES membranes lead to enlarged pore sizes as depicted byFM, transmembrane streaming potential and contact angleethods; and more charged membrane as depicted by stream-

ng potential and contact angle measurements. Our explanationf the observed is a partial scission of C–S bond that leadedo loss of polymer integrity. The microscopic changes in poly-

er structure were well pronounced at macroscopic levels. Theechanical strength of chlorine-cleaned membranes became

ower as the total chlorine concentration increased. More sig-ificant flux drop was observed after chemical cleaning withncreased hypochlorite dose.

The reported scission of C–S bond and alteration ofacroscopic membrane characteristics occurred at hypochlorite

oncentrations falling within the safety margins recommendedor UF membrane operation. Noticeably the current study waserformed at rather accelerated conditions where the PES mem-ranes were challenged with 150 mg/L HOCl, and membraneouling was performed by filtration of solution containing.3 g/L BSA protein only. At water treatment plants the chlorinen the cleaning solution is primarily used for disinfection and theesidual chlorine may attack the membrane material. The totalf 250 g is probably addressing both disinfection and possibleembrane oxidation side effect, when in our study a significant

ortion of HOCl was devoted toward membrane degradation. Itill be desirable to determine the precise characteristics of mem-rane integrity and the extent of chlorine concentration that cane used for UF membrane cleaning. Without having that infor-ation the water treatment sites should periodically check forembrane integrity, and not to assume that the membrane will

emain intact within the given chlorine safety margins. After all,he first signs of membrane disintegration were observed at 1 gree chlorine total [10], and at levels of 5 g chlorine total theembrane ceases to serve an effective virus removal barrier [2].

1 Memb

4

saopsifacf

ttsdmtttwmo

A

SEi

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

84 E. Arkhangelsky et al. / Journal of

. Conclusions

A systematic study of PES membrane degradation in bleacholutions was undertaken by comparing the properties of cleanednd pristine membranes. The results of conventional testing andf advanced spectroscopy studies indicated that the cleaningrocess substantially affects the subsequent performance andurface properties of the PES membrane. The chemical clean-ng of PES membranes appeared to dislodge the PVP componentrom the membrane matrix, causing decreases in contact anglend flux values and an increase in pore size. Thus, each suc-essive cleaning of the PES surface led to increasingly severeouling and more significant flux drops.

Other testing methods revealed additional shortcomings ofhe chemical cleaning process. XPS and ATR-FTIR spec-roscopy findings suggest possible chain scission of etherulfone, with the formation of phenyl sulfonate. The resultanteterioration in the mechanical strength of chemically cleanedembranes may cause a much earlier loss in membrane integrity

han that stated by the manufacturer. Although chain scission inhe PES molecules affected the mechanical and textural proper-ies of the membrane at the microscopic scale, the phenomenonas not reflected in permeability measurements. Additionalicroscopic tools should therefore be used to assess the degree

f integrity loss.

cknowledgements

This project was partially supported by a grant from Thetephen and Nancy Grand Water Research Institute (GWRI)..A. thanks the Rieger Foundation for a Rieger-JNF fellowship

n environmental studies.

eferences

[1] C. Liu, S. Caothien, J. Hayes, T. Caothuy, T. Otoyo, T. Ogawa, Mem-brane Chemical Cleaning: From Art to Science (2006), Pall Corporation,Port Washington, NY 11050, USA, http://www.pall.com/water 19565.asp(accessed November 30, 2006).

[2] V. Gitis, J. Gun, R.C. Haught, R.M. Clark, O. Lev, Application of nanoscaleprobes for the evaluation of the integrity of ultrafiltration membranes, J.Membr. Sci. 276 (2006) 185–192.

[3] I.M. Wienk, E.E.B. Meuleman, Z. Borneman, T. Van Den Boomgaard,C.A. Smolders, Chemical treatment of membranes of a polymer blend-mechanism of the reaction of hypochlorite with poly(vinyl pyrrolidone), J.Polym. Sci. Polym. Chem. 33 (1) (1995) 49–54.

[4] M. Nystrom, H. Zhu, Characterization of cleaning results using combinedflux and streaming potential methods, J. Membr. Sci. 131 (1–2) (1997)195–205.

[5] S.H. Wolff, A.L. Zydney, Effect of bleach on the transport character-istics of polysulfone hemodialyzers, J. Membr. Sci. 243 (1–2) (2004)389–399.

[6] H. Zhu, M. Nystrom, Cleaning results characterized by flux, streamingpotential and FTIR measurements, J. Membr. Sci. 138 (1998) 309–321.

[

[

rane Science 305 (2007) 176–184

[7] Y.N. Kwon, J.O. Leckie, Hypochlorite degradation of crosslinkedpolyamide membranes I Changes in chemical/morphological properties,J. Membr. Sci. 283 (1–2) (2006) 21–26.

[8] S. Belfer, R. Fainchtain, Y. Purinson, O. Kedem, Surface characterizationby FTIR-ATR spectroscopy of polyethersulfone membranes—unmodified,modified and protein fouled, J. Membr. Sci. 172 (1–2) (2000) 113–124.

[9] S. Rouaix, C. Causserand, P. Aimar, Experimental study of the effectsof hypochlorite on polysulfone membrane properties, J. Membr. Sci. 277(1–2) (2006) 137–147.

10] E. Gaudichet-Maurin, F. Thominette, Ageing of polysulfone ultrafiltrationmembranes in contact with bleach solutions, J. Membr. Sci. 282 (1–2)(2006) 198–204.

11] V. Gitis, R.C. Haught, R.M. Clark, J. Gun, O. Lev, Nanoscale probes forthe evaluation of the integrity of ultrafiltration membranes, J. Membr. Sci.276 (1–2) (2006) 199–207.

12] J.J. Qin, Y.M. Cao, Y. Li, Effect of hypochlorite concentration on proper-ties of posttreated outer-skin ultrafiltration membranes spun from celluloseacetate/poly(vinyl pyrrolidone) blends, J. Appl. Polym. Sci. 97 (2005)227–231.

13] F. Thominette, O. Farnault, E. Gaudichet-Maurin, C. Machinal, J.C. Schrot-ter, Ageing of polyethersulfone ultrafiltration membranes in hypochloritetreatment, Desalination 200 (1–3) (2006) 7–8.

14] D. Kuzmenko, E. Arkhangelsky, S. Belfer, V. Freger, V. Gitis, Chemicalcleaning of UF membranes fouled by BSA, Desalination 179 (1–3) (2005)323–333.

15] C.J. Gabelich, J.C. Frankin, F.W. Gerringer, K.P. Ishida, I.H. Suffet,Enhanced oxidation of polyamide membranes using monochloramine andferrous iron, J. Membr. Sci. 258 (2005) 64–70.

16] L.S. Clescerl, A.E. Greenberg, A.D. Eaton (Eds.), Standard Methods forExamination of Water and Wastewater, 20th ed., American Public HealthAssociation, 1999.

17] M. Pontie, X. Chasseray, D. Lemordant, J.M. Laine, The streaming poten-tial method for the characterization of ultrafiltration organic membranesand the control of cleaning treatments, J. Membr. Sci. 129 (1997) 125–133.

18] E. Arkhangelsky, V. Gitis, Effect of transmembrane pressure on rejec-tion of viruses by ultrafiltration membranes, J. Membr. Sci., submittedfor publication.

19] J. Ren, Z. Li, F.S. Wong, A new method for the prediction of pore sizedistribution and MWCO of ultrafiltration membranes, J. Membr. Sci. 279(2006) 558–569.

20] E. Arkhangelsky, D. Kuzmenko, N.V. Gitis, M. Vinogradov, S. Kuiry, V.Gitis, Hypochlorite cleaning causes degradation of polymer membranes,Tribol. Lett., in press.

21] J.L. Kice, A.R. Puls, The reaction of hypochlorite with various oxidizedderivatives of disulfides and with sulfinate ions, J. Am. Chem. Soc. 99 (10)(1977) 3455–3460.

22] J. Coates, Interpretation of infrared spectra, a practical approach, in: R.A.Meyers (Ed.), Encyclopedia of Analytical Chemistry, Chichester, JohnWiley and Sons Ltd., UK, 2000.

23] M. Wang, L.G. Wu, X.C. Zheng, J.X. Mo, C.J. Gao, Surface modifica-tion of phenolphthalein poly(ether sulfone) ultrafiltration membranes byblending with acrylonitrile-based copolymer containing ionic groups forimparting surface electrical properties, J. Colloid Interface Sci. 300 (1)(2006) 286–292.

24] K. Ito, Y. Tominaga, H. Ohno, Polyether/salt hybrid (IV). Effect of ben-zenesulfonate group(s) and PEO molecular weight on the bulk ionicconductivity, Electrochim. Acta 42 (10) (1997) 1561–1570.

25] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 87th ed., CRCPublishing, NY, U.S.A., 2006.