Removal of fluorinated surfactants by reverse osmosis – Roleof surfactants in membrane fouling

Clement Baudequin a, Zhaohuan Mai a, Mohammed Rakib a, Isabelle Deguerry b,Romain Severac b, Martial Pabon c, Estelle Couallier a,n

a Laboratoire de Génie des Procédés et Matériaux, École Centrale Paris, Grande Voie des Vignes, 92295 Châtenay-Malabry, Franceb DuPont de Nemours France SAS, allée de Chantereine, F-78711 Mantes laVille, Francec DuPont de Nemours International, 2 Chemin du Pavillon, 1218 Geneva, Switzerland

a r t i c l e i n f o

Article history:Received 27 June 2013Received in revised form13 January 2014Accepted 25 January 2014Available online 3 February 2014

Keywords:Reverse osmosisSurfactantFluorinated surfactantFoulingAdsorption

a b s t r a c t

The current treatment for firefighting waters containing fluorinated surfactants is incineration. A processfor purifying firefighting water was proposed in a previous paper: electrocoagulation and filtrationfollowed by reverse osmosis. Here the objective was to demonstrate the efficiency of reverse osmosis toreduce the fluorinated surfactant concentration from 30 to 0.1 mg L�1 at large scale and over severaldays. A polyamide thin film membrane was selected. Retention rate, obtained at laboratory and industrialscale for fluorinated surfactant (RR¼0.994–0.999), was sufficient to reach the limit concentration inpermeate defined in the project. Permeability (0.5 L h�1 m�2 bar�1) and retention rates were stable forseveral days. The flux decline observed was governed by the accumulation of surfactants. The fouling dueto adsorption of the surfactant mixture was successfully described by a combination of the resistance-in-series and Langmuir adsorption models.

& 2014 Elsevier B.V. All rights reserved.

1. Introduction

Fluorinated surfactants are a particular class of surfactantswhich can drastically lower the water surface tension. This abilityplays a key role in the efficiency of firefighting waters [1]. Inaddition, fluorinated surfactants are highly resistant to thermaland chemical attacks. The current treatment for firefighting watersis therefore incineration, which is very expensive.

A previous paper proposed a process for purifying firefightingwater containing fluorinated surfactants [2]. Experimental resultsindicated that electrocoagulation and filtration followed byreverse osmosis can potentially treat the water from fire extin-guishment. In the present paper we have focused on the reverseosmosis step.

Based on both membrane and molecular properties, Bellona [3]extensively reviewed the factors governing the rejection of organicmolecules by nanofiltration and reverse osmosis membranes, witha focus on organic micropollutants: sieving, electrostatic repulsion,and molecular interaction between solute and membrane. Highlyapolar molecules can adsorb on the membrane, giving high

apparent retention rates. However, the membrane may reachsaturation and the solute may leak, leading to lower equilibriumretention rates. Hydrogen bonding and acidity properties in thepolar solutes decrease retention rates (competition between waterpermeation and similar mechanisms). This point was confirmed bySotto [4].

In water, surfactants can interact via hydrophobic interactions,with apolar tails being excluded fromwater by apolar Lifshitz–Vander Waals (LW) interactions; electrostatic interactions, withcharged entities in case of ionic surfactants; Polar Lewis acid–baseinteractions (AB), if the molecule contains electron donor and/orelectron acceptor groups as well as acidic protons or basic groups[5,6].

Most of the reverse osmosis or nanofiltration membranes haveactive layers made of aromatic thin film polyamide or celluloseesters. The intermolecular forces able to arise are basically thesame as for surfactants. Zeta potential of various reverse osmosisand nanofiltration membranes were measured by Elimelech et al.[7]. Most of the nanofiltration and reverse osmosis membranes arenegatively charged at pH above 5 to 6. However, the resultingmembrane charge also depends strongly on the solution chemistryand on the adsorption of charged compounds that can modify themembrane surface. Childress measured zeta potential of polya-mide and cellulose acetate reverse osmosis membranes before and

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/memsci

Journal of Membrane Science

http://dx.doi.org/10.1016/j.memsci.2014.01.0630376-7388 & 2014 Elsevier B.V. All rights reserved.

n Corresponding author. Tel.: +33 240172667.E-mail address: estelle.couallier@univ-nantes.fr (E. Couallier).

Journal of Membrane Science 458 (2014) 111–119

after sodium dodecyl sulfate adsorption and the noted membranesbecame more negatively charged [8].

According to literature, surfactant–membrane interaction foranionic [9] and non-ionic surfactants [10] is related to themembrane's hydrophobicity. Negative membrane charges increaseanionic surfactant rejection [9]. Boussu described the filtration ofanionic, cationic and non-ionic surfactants with various nanofil-tration membranes [11]. The main factor governing flux declinewas adsorption of surfactants and pore blocking when the adequ-ateness between pore size and surfactant size permitted it.The non-ionic surfactant was better retained and caused lessfouling with the most hydrophilic membranes with the lowestcut-off. These compounds could lead to a rise of flux. For theanionic surfactant, electrostatic repulsion increased the rejectionand lowered the flux decline for each membrane compared to thenon-ionic surfactant. These results were confirmed by Li [12] whostudied fatty acid fouling of reverse osmosis membranes. pHbelow pKa and rising ionic strength lower electrostatic repulsionand improve fouling and flux decline.

Solid surface tensions of polyamide [13,14] and celluloseacetate membranes [15,16] have been measured, showing anapolar surface tension γLW in the range of 32 to 39 mJ m�2 forboth membranes. The polar components of the solid surfacetension were different, but the electron donor component, γ� ,was predominant for most of the membranes (Table 1). Therefore,these results indicate that substantial apolar interactions can beexpected with these kind of membranes, as well as polar interac-tions preferably with electron acceptor molecules.

The hydrophobic character of the membranes can give rise toapolar (LW) interactions. We may consider that polyethylenecorresponds to hydrocarbonated tail of surfactant, polytetrafluor-oethylene to fluorinated one and octane to hydrophobic part of themembrane. The cohesion energy between hydrocarbonated orfluorinated surfactant and membrane immersed in water (respec-tively, bodies 1, 2 and 3) if water is the only polar compound in thesystem is [6]

ΔG132 ¼ þ2½√ðγLW1 γLW3 Þþ√ðγLW2 γLW3 Þ�√ðγLW1 γLW2 Þ�γLW3 �2√ðγþ3 γ

�3 Þ�ð1Þ

After calculations with the values given in Table 2, the follow-ing values are obtained:

ΔGoctane water polyethylene ¼ �101:95 mJ m�2

ΔGoctane waterpolytetraf luoroethylene ¼ �102:02 mJ m�2

According to these results, the difference of membrane affinitybetween hydrocarbon and fluorinated tail of surfactant is negli-gible and interaction difference is expected to depend mainly onthe polar heads.

Fouling of membrane can result from several mechanismssuch as pore blocking, adsorption, cake or gel layer formation.Flux decline is strongly dependent on the feed solution and on the

membranes. The resistance-in-series model is commonlyemployed to describe flux decline [18]. In this model the flux J(L h�1 m�2 or m s�1) is a function of the driving force ΔP, theviscosity η and the total resistance Rtot:

J ¼ ðΔPÞ=ðηRtotÞ ð2Þ

Rtot includes the intrinsic resistance of the membrane to purewater, Rm and every additional resistance Rcp for concentrationpolarization, Ra for adsorption, Rg for a gel layer, Rc for a cake andRp for pore blocking.

Rtot ¼ RmþRcpþRaþRgþRcþRp ð3Þ

The resistance-in-series model was used by Kaya et al. [19] toevaluate the flux decline for solutions containing anionic, non-ionic surfactants, dye and salts. The authors attributed a greatcontribution to flux decline to the deposition of surfactant aggre-gates on the membrane surface.

Langmuir and Freundlich models were used to describe adsorp-tion of organic pollutants on reverse osmosis membrane [20,21]and experimentally linked to flux decline [4,22–24].

The objective of this work was to demonstrate the capacity ofreverse osmosis to treat at industrial scale the firefighting water, byreducing the fluorinated surfactant concentration to 0.1 mg L�1.Specific goals included identification of a relevant membrane,evaluation of the retentate concentration impact on the perme-ability, retention rate and flux stability assessment. The stronglink between adsorption of surfactants and flux decline was alsostudied.

2. Material and methods

The experimental study of the reverse osmosis was organized inthree steps: first, available membrane materials were screened in aflat sheet Osmonics Sepa CFII reverse osmosis and nanofiltrationcell. Then real and model pretreated pilot firefighting waters weretested in a Millipore pilot with a laboratory scale spiral-woundmodule, and finally the observed properties were checked duringseveral days' experiments in a semi-industrial Polymem pilot.

Table 2Surface tensions of octane, polyethylene, polytetrafluoroethylene and water(mJ m�2) [17].

Compound γLWS γþS γ�S

Octane 21.6 0 0Polyethylene 33.0 0 0Polytetrafluoroethylene 17.9 0 0Water 21.8 25.5 25.5

Apolar surface tension: γLWS , electron acceptor component of the surface tension:γþS , and electron donor component of the surface tension: γ�S .

Table 1Solid surface tension of several polyamide and cellulose membranes, in mJ m�2, calculated after contact angle measurements.

Membrane material γLWS γþS γ�S Reference

Aromatic polyamide (XLE) (interfacially partially crosslinked polymerizedpolyamide composite membrane, DOW-FilmTec, Edina, MN, USA)

35 E1 10–20 Sessile drop [13]

Thin film composite polyamide (FT-30) (Film Tec, Minneapolis, MN) 32.2 1.82 28.68 Captive buble [15]Cellulose acetate (CD) heat-treated cellulose triacetate/diacetate blend membrane (Osmonics Desal, Vista, CA) 38.1 0.01 37.43 Captive buble [15]Cellulose acetate (CE) heat-treated cellulose triacetate/diacetate blend membrane (Osmonics Desal, Vista, CA) 37.5 1.69 28.51 Captive buble [15]

Apolar surface tension: γLWS , electron acceptor component of the surface tension: γþS , and electron donor component of the surface tension: γ�S .

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119112

2.1. Devices used for reverse osmosis experiments

2.1.1. Osmonics Sepa CFII cellThe screening of different available membrane materials was

performed with a Sepa CFII cell from Osmonics. The cell acceptsflat sheet membranes, with an effective area of 140 cm2. Themembranes retained for the screening were: Osmonics nanofiltra-tion CK (cellulose triacetate/diacetate, for softening) and DK (thinfilm polyamide with approximate molecular weight cut-off of150–300 Da for uncharged organic molecules) and reverse osmo-sis SG (thin film polyamide, with high sodium chloride rejectionand a smooth, fouling-resistant membrane surface), AD (thin filmpolyamide for seawater desalination) and CE (cellulose triacetate/diacetate for brackish water desalination). The setup comprised ahigh pressure pump, a by-pass, a pressure valve and a tank.Pressure and flow were adjusted by the interplay of the two valvesand superficial velocities were 0.1 and 0.5 m s�1. Each experimentwas carried out with a 31 mil spacer (0.7874 mm), in recyclemode, at 30 1C, at the pressures recommended by the supplierfor each membrane during 1 h until permeate flux was constant.Before use, membranes were washed for 20 min with water, thenwith NaOH, pH 10–10.5, for 30 min, and then with water againuntil neutral pH. Integrity was verified with NaCl.

2.1.2. Millipore ProScale pilotA Millipore nanofiltration/reverse osmosis pilot was used with

an SG1812C-28D reverse osmosis spiral membrane supplied byOsmonics, with a thin film polyamide active phase of 0.37 m2

surface area and feed spacers of 28 mil (0.7112 mm). The pilot hada dead volume of 0.8 L, superficial velocity was 84.10�3 m s�1 andoperating pressure was 20 bar. All permeability measurementswere scaled to 25 1C. According to the manufacturer, SG mem-brane is a smooth fouling-resistant thin film polyamide membranefor brackish water.

2.1.3. Polymem pilotTwo longer experiments were performed with a reverse

osmosis pilot from Polymem. The pilot had a 100 L tank, a deadvolume of 9.9 L, a high pressure pump, a recirculation loop andwas equipped with a fouling-resistant thin film polyamide CSMRE2540-FE reverse osmosis membrane with an area of 2.5 m2. Therecirculation rate was set to 50%, the pressure to 30 bar andthe superficial velocity to 0.09 m s�1 and the temperature insidethe loop was maintained at 30 1C by an external thermocryostat.

2.2. Products

Real firefighting waters can be of a great variety, and it isdifficult to obtain samples with a stable composition in sufficientamounts to run experiments. Hence it was decided to usesynthetic solutions and also to produce pilot firefighting waters(PFW) by the extinguishment of artificial solvent fires with aspecific firefighting foam. The products used to generate suchsolutions are described in Table 3. Non-ionic, anionic and zwitter-ionic carbonated and fluorinated surfactants were used in water.

The objective of this work is to predict the retention offluorinated surfactant in industrial conditions. In industrial condi-tions, the filtration will be led in a continuous mode at highvolume reduction ratio and concentration factor (CF) in order tomaximize the treated permeate volume. That is why the syntheticsolutions used were nearly 20 times the concentration measuredafter pretreatment, i.e., with fluorinated surfactant concentrationbetween 500 and 850 mg L�1.

2.2.1. Solutions used for membrane screeningThe solutions used for screening are summarized in Table 4.

2.2.2. Solutions for test on Millipore pilotDuring the first experiment on Millipore pilot, the reverse

osmosis of 5 L of pilot firefighting water (PFW) was done to checkthe behavior of the membrane with real pretreated PFW. PFWswere produced by the extinction of 0.25 m2 (2 L) n-heptane fires.The firefighting foam contained the components described inTable 3 and diethanolamine to reach pH 7.5. The n-heptane andthe eventual emulsion at the n-heptane/water interface werediscarded by decantation. PFW was pretreated by electrocoagula-tion and filtration in the laboratory. It contained 133 mg L�1 offluorinated surfactant before electrocoagulation and approxi-mately 30 mg L�1 after pretreatment (analytical method 1) [2].

In a second experiment on Millipore pilot, a model feedsolution was used to measure flux decline versus concentrationof surfactants. The composition of the global feed solution is givenin Table 5 and consisted of a solution with increasing concentra-tion of all compounds. The concentration was artificially increasedby introducing concentrates in the feed solution (addings).Another experiment was done with a model solution with thesame proportions as the previous model solution, but on the basisof a fluorinated surfactant concentration of 470 mg L�1.

Table 3Components used to generate synthetic solutions.

Chemical name Developed formula Name of commercialproduct

CAS Supplier Critical micellarconcentration

M(g mol�1)

Dipropylene glycol methyl ether Dowanols DPM 34590-94-8

DowChemicals

133 mmol L�1 14831 g L�1

Sodium octylsulfate Disponils SOS 842a 142-31-4 Cognis 460 mmol L�1

(sodium nonyl sulfateCMC)

232

Sodium caprylamphopropionate Tegotenss AM VSFa 64265-45-8

EvonikGmbH

309

D-Glucopyranose oligomers (C8–C10 alkylpolyglucosides)

Simulsols SL8a 68515-73-1

Seppic 354

Amphoteric fluorinated surfactant fraction Non-communicated Fluorinated surfactant DuPont 0.10 g L�1 570

a Commercial products.

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119 113

2.2.3. Solutions for large-scale experimentsThe compositions of solutions used during the large-scale

experiments on Polymem pilot are given in Table 6.Fluorinated surfactant used for pilot firefighting water is more

hydrophobic than the fluorinated surfactant used for synthetic solu-tions. The polar head remains unchanged (amphoteric polar head).

2.3. Analytical methods

Surfactant quantification was performed by the HPLC method.

2.3.1. Method 1Sample homogenization was achieved in 60 mL glass vials, using a

Heidolph Topmix 94323 vortex. High performance liquid chromato-graphy (HPLC) measurements were performed with an analyticalsystem composed of a Knauer K-501 HPLC pump (Eurosep Instru-ments), a Rheodyne valve with a 40 mL injection loop, an EclipseZorbax XDB-C8 analytical column (Agilent Technologies, 4.6 mmdiameter, 150 mm length, 5 mm particle size), a column oven at35 1C and the mobile phase was methanol:water 70:30 v-v at a0.5 mLmin�1 isocratic flow rate. The detector was an ESA EvaporativeLight-Scattering Detector (ELSD, Chromachem, Eurosep Instrument),attenuation was 2, nitrogen pressure was 1.5 bar, nebulization andevaporation temperatures were 50 and 70 1C, respectively; dataacquisition and processing was done with Azurs software. Simulta-neous resolution of the anionic, non-ionic, amphoteric hydrocarbonsurfactants and fluorinated surfactant could not be achieved with theaforementioned mobile phase, which was intended to resolve thefluorinated surfactant only, with a limit of detection of 1.4 mg L�1. Thecalibration curve was established from 4 to 1000 mg L�1, the errorwas below 5%, every analysis was done twice and out of range

concentrated samples were diluted. Samples with a turbidity superiorto 2 NTU were filtered on Roth 0.45 mm polyvinylidene fluoride(PVDF)-made syringe. After PVDF filtration the recovery rate offluorinated surfactant was 90%.

2.3.2. Method 2HPLC–ELSD analyses were performed according to the previous

method but the isocratic pump was replaced by a Hitachi L-2130gradient pump. The only volatile compound that could not bedetected by ELSD was Dowanols DPM, the cosolvent. The foursurfactants, namely, Disponils SOS 842, Tegotenss AM VSF,Simulsols SL8, and fluorinated surfactant, could be separated withthe eluent gradient given in Table 7.

To detect the surfactant at low concentration (o2 mg L�1) asample concentration method was set up. This method consisted of(1) volume reduction by evaporation and (2) surfactants redissolu-tion by a 70% methanol recovery solution containing 30 mg L�1 ofNaCl. Accurate volumes near 60 mL of dilute solutions were evapo-rated to dry at 90 1C in glass vials. 3 mL of recovery solution wereadded to the vials before vortex agitation. The resulting solutionscontained concentrated surfactants that allowed analysis down to0.2 mg L�1, with an average recovery rate of 90%. Samples of lowerconcentration were analyzed by solid phase extraction (SPE) andliquid chromatography coupled with mass spectrometry (LC–MS) bythe Norwegian Institute for Air Research (NILU). NILU's laboratoriesare accredited according to NS-EN ISO/IEC 17025.

The concentration in free monomers of surfactants contained inconcentrated solution was estimated by surfactant quantificationin an ultrafiltration permeate, as ultrafiltration is known to retainsurfactant micelles [25,26]. 2 L concentrated solutions were ultra-filtred on a Rayflow X100 ultrafiltration pilot with a WatsonMarlow 624U peristaltic pump, a LAUDA RM6 thermocryostat,and 200 cm2 5 kDa polyethersulfone membranes from Novasep at2 bar. According to appearing rejections (Table 8), Disponils SOS842 was mainly dispersed as isolated compound, whereas thefluorinated surfactant was mostly in micelles. Tegotenss AM VSFand Simulsols SL8 distributions were intermediate. The eventual-ity of mixed micellization remained out of the scope of this work.

3. Results

3.1. Screening of flat reverse osmosis and nanofiltration membranesin the Osmonics cell

The screening of different membrane materials aimed atidentifying the most appropriate membrane material for pre-treated firefighting water, considering fluorinated surfactantretention properties and flux decline. For this purpose, polyamideand cellulose acetate reverse osmosis and nanofiltration mem-branes were tested in the Osmonics cell.

The results were similar for the two velocities and only results at0.1 m s�1 are presented. The permeabilities at 25 1C were calculatedto facilitate the comparison between the different experiments.

Table 4Compositions in mg L�1 of solutions used for the screening of different availablemembranes on Sepa CFII cell.

Surfactants SG AD CE CK DK

Disponil SOS 842 12,082 13,144 11,651 11,386 9638Tegotens AM VSF 4036 4062 3472 3493 3187Simulsol SL8 1328 1544 1296 1616 1201Fluorinated surfactant 782 816 628 656 527

Table 5Artificially concentrated model solution composition after addings (Millipore pilot).

Number of addings 0 1 2 3 4

Total volume (L) 3.8 4.3 4.8 5.3 5.8Dowanols DPM (mg L�1) 270 562 1122 2258 4561Disponils SOS 842 (mg L�1) 87 175 357 734 1476Tegotenss AM VSF (mg L�1) 20 43 85 127 220Simulsols SL8 (mg L�1) 20 43 90 185 355Fluorinated surfactant (mg L�1) 23 46 100 205 417

Table 6Compositions in mg L�1 of concentrated solutions in the Polymem pilot. Thevolumes were 40 L, including the dead volume.

Name of commercial product Large-scale exp. 1 Large-scale exp. 2

Dowanols DPM 51,650 43,551Disponils SOS 842 16,000 14,777Tegotenss AM VSF 4001 2289Simulsols SL8 2008 1868Fluorinated surfactant 804 745

Table 7Composition of the eluent gradient given to the Hitachi L-2130 gradient pump,for the simultaneous analysis of surfactants from pilot firefighting waters (%v).

Time (min) Methanol Water Ammoniumacetate, 1mol L�1

0 60 39 111 99 0 114 99 0 116 60 39 118 60 39 1

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119114

The choice of the appropriate membrane is a matter of retention toperform the separation efficiently, especially the retention rate ofthe fluorinated surfactant. The membrane achieving the highestretention rate for the fluorinated surfactant is the SG membrane,with 99.4% (Table 9). This membrane is unfortunately the one thathas the lowest rate of pure water flux available with the concen-trated solution, 0.22 L h�1 m�2 bar�1 (Table 10).

3.2. Reverse osmosis of model and pilot pretreated firefightingwaters with a spiral-wound SG module

3.2.1. Reverse osmosis of pretreated pilot firefighting waterReverse osmosis was used to treat 5 L of pretreated pilot

firefighting water in the Millipore pilot. The experiment consisted

of, successively, 15 min of full recycle, extraction of 2 L of perme-ate, 20 min of full recycle and extraction of 1.7 L of permeate.Results are presented in Table 11. According to the analysisperformed by the Norwegian Institute for Air Research, fluorinatedsurfactant concentration in whole extracted volumes of permeatewere 10.47 and 16.39 μg L�1, giving apparent retention rates of99.97% and 99.96% for the first and second permeates, respectively.These permeate concentrations matched the aimed concentrationof o0.1 mg L�1, however, these experiments were done with lowvolume reduction ratios of 1.5 and 2.8. Mass balance showed somemissing fluorinated surfactant in the retentate, assumed to beadsorbed on the membrane with an adsorption density between73 and 162 mg m�2.

3.2.2. Reverse osmosis of a model pretreated firefighting waterReverse osmosis was performed on pretreated pilot firefighting

water until a volume reduction ratio of 2.8, but a high volumereduction ratio could not be reached because of the low available

Table 8Monomer content in concentrated solutions, estimated on the basis of ultrafiltrationpermeate concentration (given) and retentate concentration after equilibrium (notgiven).

Surfactant Alimentationconcentration (mg L�1)

Permeateconcentration(mg L�1)

Apparentrejection (%)

Disponils

SOS 84211,649 10,135 13

Tegotenss

AM VSF3390 2000 41

Simulsols

SL81948 643 67

Fluorinatedsurfactant

825 66 92

Table 9Concentration in permeates and rejections of the fluorinated surfactant duringmembrane screening (analytical method 2).

Membrane CP (mg L�1) R (%)

SG 4.8 99.4AD 17.7 97.8CE 13.7 97.7CK 44.1 93.2DK 40.2 93.6

Fig. 1. Permeability versus total fluorinated surfactant concentration for threedifferent experiments, compared to litterature. ○: Reverse osmosis of modelpretreated pilot firefighting water in the Millipore pilot with a spiral-wound SGmembrane at 20 bar, concentration measured in retentate. ⎕: The membrane usedby Tang et al. [27,28] was an ESPA membrane, the pressure was 13.8 bar and thefluorinated surfactant was perfluorooctane sulfonate. ∇: The data from flat sheetSG membrane are from part 3.1, Δ: those for pretreated pilot fire fighting waterfrom Section 3.2.

Table 11Fluorinated surfactant concentration and membrane permeability during pretreated pilot firefighting water reverse osmosis, 20 bar (analytical method 1). aadditional resultsfrom the Norwegian Institute for Air Research (NILU).

Step L at 25 1C(L h�1 m�2 bar�1)

VR (L) VRR Expected CR (mg L�1) Measured CR (mg L�1) Mean CP (μg L�1) Fluorinated surfactantmass balance (mg)

Full recycle 1.033 5.8 1 2775 2071 – �40710Concentration 0.914 – – – – 10.47a –

Full recycle 0.921 3.8 1.5 41711 3472 – �27710Concentration 0.779 – – – – 16.39a –

Full recycle 0.760 2.1 2.8 74733 4672 – �60738

L: permeability, VR: retentate volume, VRR: volume reduction ratio, CR: retentate concentration, and CP: permeate concentration.

Table 10Flows, permeabilities and flux decline during membrane screening with concentrated model solution in the Osmonics cell for cross-flow velocities of 0.1 m s�1.

Membrane Pressure(bar)

Water flow(L h�1)

Water permeability(L h�1 m�2 bar�1)

Model solution permeateflow (L h�1)

Model solution permeability(L h�1 m�2 bar�1)

Available water flux with themodel solution (%)

SG 30 0.792 1.89 0.140 0.22 12AD 55 0.912 1.19 0.413 0.58 49CE 30 0.492 1.17 0.231 0.54 46CK 15 0.408 1.94 0.144 0.69 36DK 7 0.300 3.06 0.042 0.43 14

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119 115

volume of pilot firefighting water. Hence the study of the pre-treated pilot firefighting water concentration with reverse osmosiswas done by artificial concentration of model pretreated fire-fighting water by successive addings (Table 5), in full recyclemode. Measurements were done 1 h after each adding. In thispart, fluorinated surfactant is taken as an indicator of the wholecomposition of the solution, the other compounds being supposedto stay in the same proportions as the feed solution. During theexperiment, membrane permeability decreased with increasingconcentrations (Fig. 1). Though permeability decrease was strong,higher concentrations than 200 mg L�1 of fluorinated surfactantled to small permeability reductions. Considering the fluorinatedsurfactant concentrations, permeabilities were the same magni-tude as the previous experiments, in the same range as the resultsdescribed by Tang et al. [27,28] obtained with perfluorooctanesulfonate solutions and thin film composite polyamide ESPA ROmembrane from Hydranautics, and to the data obtained with theflat sheet SG membrane during the screening despite the differentcomposition of the feed solution.

No surfactants were detected in all permeates with the con-ventional HPLC analytical method. An additional reverse osmosisexperiment with model pretreated firefighting water, containing470 mg L�1 of fluorinated surfactant, was performed to quantifyfluorinated surfactant concentration in permeate with the eva-poration method. After 2 h of equilibrium, permeate concentrationwas 0.271 mg L�1, corresponding to a 99.94% retention rate, closeto the results obtained in the preceding experiment.

The SG membrane in spiral-wound module showed highretention rates, even higher than those measured with flat sheetSG membrane in the Osmonics cell. The SG membrane foulingwith increasing concentration showed that most of the fluxdecline occurred at relatively low concentration, between 0 and200 mg L�1 of fluorinated surfactant, corresponding to a volumereduction ratio of approximately 7. The flux decline between aconcentration of 400 mg L�1 and 800 mg L�1 remain rather low(Fig. 1).

Flux decline versus the sum of initial concentration for allintroduced molecules presented in Table 12 can be compared toliterature: the filtration of octanoic acid at 2 mmol L�1 or sodiumdodecylsulfate 0.5 mmol L�1 at pH 7 on aromatic polyamide thin-film membrane generated a flux decline of 30%, comparable withour results [8,12,29]. Flux decline of a RO UTC-20 (Toray Ind. Inc.,Japan) as a function of concentration of different organic com-pounds was presented by Van der Bruggen et al. and wascomparable to values presented here, despite difference in thechemical formula [29]. Considering the major compounds present(Dowanols and disponils) have not reached their critical micellarconcentration, these results are consistent with those described byArcher et al. [30].

In this section, the retentate concentration of the fluorinatedsurfactant was taken as an indicator of the other surfactants beingalso present in proportional concentrations. It has to be kept inmind that these compounds participate to the flux decline.

According to the results presented above, surfactant adsorptionon the membrane occurred. Therefore, an adsorption-based model

was considered. If the flux decline is described on the basis of theresistance-in-series model, the permeability is L:

L¼ 1ηRtot

ð4Þ

Assuming that adsorption alone is responsible for the fluxdecline, the total resistance is

Rtot ¼ RmþRa ð5ÞRm, the intrinsic membrane resistance can be obtained on the basisof the membrane water permeability when CR¼0:

Rm ¼ ð1=ηLwaterÞ ¼ 1:82� 1014 m�1

Therefore,

Ra ¼1ηL

�Rm ð6Þ

On the basis of experimental data, Ra versus CR was plotted(Fig. 2). Though the amount of all the surfactants adsorbed on themembrane was not determined during the experiment withthe Millipore pilot, it was assumed that Ra could be described by

Table 12Reverse osmosis of model pretreated pilot firefighting water in the Millipore pilotwith a spiral-wound SG membrane at 20 bar:initial concentration (mmol L�1) of allintroduced compounds (Dowanols, Disponils, Tegotenss, simulsols, fluorinatedsurfactant) and related flux decline (%).

Adding 0 1 2 3 4

Flux decline (%) 38 62 72 75 79Total concentration (mmol L�1) 2.3 4.9 9.8 19.6 39.6

Fig. 2. Ra, the resistance to mass transfer due to adsorption, versus CR the retentateconcentration of the fluorinated surfactant, during reverse osmosis of modelpretreated firefighting water in the Millipore pilot, at 20 bar, with a spiral-woundSG membrane. The concentration of the fluorinated surfactant is an indicator of theglobal concentration of other surfactants also present.

Fig. 3. Model and experimental data for the permeability versus retentateconcentration of the fluorinated surfactant, during reverse osmosis of modelpretreated firefighting water in the Millipore pilot, at 20 bar, with a spiral-woundSG membrane. The concentration of the fluorinated surfactant is an indicator of theglobal concentration of other surfactants also present.

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119116

the following equation, analog to Langmuir:

Ra ¼ Rmaxa

ðKCRÞð1þKCRÞ

ð7Þ

The plot of 1/Ra versus 1/CR was almost linear. The linearregression enabled a calculation of Rmax

a and K:

Rmaxa ¼ 7:7� 1014 m�1; K ¼ 13:1 mg m�2

The permeability L as function of CR could be modeled (Eq. (8))and this model was confronted to experimental data in Fig. 3.

L¼ 1ηðRmþððRmax

a ðKCRÞÞ=ð1þKCRÞÞÞð8Þ

On the basis of the fluorinated surfactant concentration, themodel described very well the experimental data. The resistancedue to adsorption seemed to be directly related to the retentateconcentration, in the same way as the adsorption density is relatedto the equilibrium concentration in Langmuir isotherms. Severalassumptions were made for this model: the equilibrium perme-ability was assumed to be reached for each concentration, the bulkconcentration of the fluorinated surfactant was taken as a refer-ence for the global retentate concentrations in surfactant mixture,Ra was assumed to depend on a Langmuir-type expression.

According to our model of flux decline, the limit in flux declinewould be reached at saturation of the membrane, and there wouldstill remain a minimum permeability. After the value obtained forRmaxa , the permeability after maximum adsorption would be

Lmin¼0.378 L h�1 m�2 bar�1. Therefore, from an industrial pointof view, with an SG membrane the volume of pretreated pilotfirefighting water could be highly reduced, without additionalhigh fouling due to the surfactant mixture. Some additional workwould be required to improve the model by considering theconcentration of each surfactant, higher concentrations and per-meabilities over long periods. The assessment of its validity onother reverse osmosis membranes would also be interesting.

These results were obtained during the filtrations of pilot andmodel firefighting water. However, the experiments were doneover short periods and the stability of the membrane's behaviorregarding flux decline and retention has to be examined overlonger periods. In the literature, some fluorinated surfactants haveshown high initial rejection, and progressive leaking over timeduring nanofiltration experiments [31]. This eventual undesiredleaking had to be checked in the case of firefighting water reverseosmosis. In addition, the concentrations of the other surfactantshave not been measured yet and could provide useful information.

3.3. Study of the reverse osmosis on longer periods in an industrialpilot

Two experiments were performed with 40 L concentratedsolutions: the first (test 1) during 8 h and the second (test 2)during 10 days.

During both experiments, the flux decline was strong, theavailable ratio of initial permeability was 20–25%. However, inboth cases, the flux declined very quickly, within the first minutes,and no transition between pure water flux and flux with concen-trated solutions could be observed. Although quick, the fluxdecline seemed to stabilize over 10 days, near a permeability of0.5 L h�1 m�2 bar�1. The concentration difference between thetwo solutions employed explained the slight difference in perme-ability. The membrane was washed between the two tests with awater solution containing 5% v-v ethanol and 0.5 g L�1 NaCl. Thiscleaning did not seem to be sufficient to recover the initial water

permeability, but the remaining fouling did not generate more fluxdecline in the second experiment. The flux decline caused by thesurfactant solutions, though high, appeared to be fast, stable andreversible.

Retention rates were very high, i.e., 499.75% for all surfactants(Fig. 4). The values for each surfactant seemed to correlatebetween both experiments, especially for the retention of Dis-ponils SOS 842. The anionic surfactant had the lowest retentionrate in spite of the eventual electrostatic repulsion and theretention rates of the other surfactants were above 99.9% but norejection tendency according to the chemistry of surfactants couldbe extracted from the experimental data. The rejection of thefluorinated surfactant was very high, above 99.95% (Table 13).

Fig. 4. Large-scale RO experiments in the Polymem pilot with 40 L of concentratedsolution, at 30 bar, and cross-flow velocity of 0.095 m s�1. Retention rates of foursurfactants: Test 1 (8 h) ○: Disponils SOS 842; ∇: Tegotenss AM VSF; ✧: Simulsols

SL8; ✩: fluorinated surfactant. Test 2 (10 days): �: Disponils SOS 842; ▼: Tegotenss

AM VSF; ✦: Simulsols SL8; ★: fluorinated surfactant.

Table 13Average rejection rates during experiments of RO inthe industrial pilot.

Surfactants Rejection (%)

Disponils SOS 842 99.86Tegotenss AM VSF 99.98Simulsols SL8 99.96Fluorinated surfactant 99.95

Fig. 5. Surfactants adsorption density during reverse osmosis in the Polymem pilotof 40 L of concentrated solution, at 30 bar, and cross-flow velocity of 0.095 m s�1.Test 1: ○: Disponils SOS 842; ∇: Tegotenss AM VSF; ✧: Simulsols SL8,✩: fluorinated surfactant. Test 2: �: Disponils SOS 842; ▼: Tegotenss AM VSF;✦: Simulsols SL8; ★: fluorinated surfactant.

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119 117

The mass balance was estimated for each surfactant, based oninitial composition of the solutions, on retentate concentrationand on an estimation of the loop volume of 5 L. Adsorptiondensities depicted in Fig. 5 and summarized in Table 14 assumedthe membrane was the only interface where significant adsorptiontook place. The largest adsorption density (more than ten timesthe others) was the Disponils SOS 842's one, in spite of itsnegative charge and the membrane's negative charge. This surfac-tant was also the one that had the highest concentration withoutbeing involved in micelles in the feed solution. Therefore, assum-ing that the fouling was due to the adsorption of the surfactants, itcould be mainly attributed in the mixture to the adsorption of theanionic surfactant. The adsorption of the fluorinated surfactant ismore important than in previous experiments in the Milliporepilot. The adsorption of Tegotens s, simulsol s and fluorinatedsurfactant is of the same magnitude as the adsorption describedby Sotto et al. for 2-chlorophenol and 2 nitrophenol, in the samerange of bulk concentration, on BW30 reverse osmosis membrane[4] and non-ionic surfactant (2.4 mmol m�2 at 0.1 mmol L�1 [10]).

The stability of the permeability and rejection was confirmedduring reverse osmosis experiments in the Polymem pilot. Theflux decline was high and quick, and was mostly attributed to theadsorption of the anionic surfactant, in concentrated solutions. Atfirst sight, this compound was expected to show repulsive electro-static interactions. But considering its high adsorption density(Fig. 5), the anionic surfactant seemed to adsorb via its hydro-phobic tail, interacting with hydrophobic parts of the membrane,hiding them from water, regardless of electrostatic interactions,despite pH 7 and no additive salts. Rather than a surface having auniform surface charge, the membrane could be a patchwork ofanionic sites on a hydrophobic support. This would conform toboth polar and apolar characters of reverse osmosis membranes(Table 1), and would account for the observed adsorption of theanionic surfactant. The other surfactants also showed someadsorption density, which opened the question of mixed adsorp-tion of surfactants at the membrane surface. Synergistic effectswere already mentioned for nanofiltration of anionic and non-

ionic surfactants [32]. This field is very broad and the datacollected was not sufficient to discuss this point.

However, the higher adsorption density of the anionic surfac-tant compared with other surfactants could be related to its highermonomer content (Table 8). In concentrated solutions, adsorptionseemed to be mostly due to the free anionic surfactant while othersurfactants were significantly involved in micelles, thus lessavailable for adsorption.

The anionic surfactant was presumably the main compoundinvolved in flux decline. Fouling minimization by selective anionicsurfactant removal, if desired, should be carefully studied withparticular attention to the behavior of the other surfactants, andthe eventual changes in their adsorption and rejection.

Results of all experiments are summed up in Table 15.

4. Concluding remarks

During screening of membranes for the treatment of firefight-ing water, a polyamide thin film was selected. It was verified thatflux and retention rates were stable for several days at large scale.According to the high adsorption density of the anionic surfactantduring reverse osmosis experiments, the membrane–surfactantsinteractions seemed to occur mainly via hydrophobic interactions.The surfactant adsorption at the membrane surface appeared to bemixed, and the adsorption density was linked to the free monomerconcentrations. The flux decline observed during reverse osmosiswas governed by the adsorption of surfactants. The fouling wassuccessfully described by a combination of the resistance-in-seriesand Langmuir adsorption models. For further membrane optimi-zation, hydrophilic reverse osmosis membrane materials with γþ

and γLW components as low as possible, as well as high electrondonor character should be tested.

Nomenclature

CF concentration factorΔG cohesion energy (mJ m�2)J flux (L h�1 m�2 or m s�1)K Langmuir type constant (g m�2)L permeability (L h�1 m�2 bar�1 or L h�1 m�2 Pa�1)ΔP transmembrane pressure (bar or Pa)PFW pilot firefighting watersR resistance (m�1)Ra resitance due to adsorption of surfactants on mem-

brane (m�1)Rmaxa maximum resistance due to adsorption which can

be reached (m�1)γLW apolar surface tension (mJ m�2)γ� electron donor component of surface tension

(mJ m�2)γþ electron acceptor component of surface tension

(mJ m�2)η viscosity (Pa s)

References

[1] M.J. Rosen, Surfactants and Interfacial Phenomena, third edition, John Wileyand Sons, Inc., 2004.

[2] C.m. Baudequin, E. Couallier, M. Rakib, I. Deguerry, R. Severac, M. Pabon,Purification of firefighting water containing a fluorinated surfactant by reverseosmosis coupled to electrocoagulation-filtration, Sep. Purif. Technol. 76 (2011)275–282.

Table 15Summary of the experimental results obtained during reverse osmosis experi-ments. The only surfactant considered is the fluorinated surfactant.

Devices Osmonicscell

Milliporepilot

Polymempilot

Membranes SG, flat sheet SG, spiral-wound

RE2540-FE,spiral-wound

CR (mg L�1) 774 23–417 804CP (mg L�1) 4.8 10–271�10�–3 110–

665�10�3

R (%) 99.38 99.94–99.97 99.95–99.99Flow (L h�1) 0.14 3–11 37.5Permeability(L h�1 m�2 bar�1)

0.22 0.4–1.5 0.5

Available waterpermeability (%)

12 20 21

Table 14Surfactants adsorption density during reverse osmosis in the Polymem pilot of 40 Lof concentrated solution, at 30 bar, and cross-flow velocity of 0.095 m s�1.

Surfactants Adsorption (g m�2) (mmol m�2)

Disponils SOS 842 49–61 210–263Tegotenss AM VSF 0.65–6.75 2.1–22Simulsols SL8 1.08–4.6 3–13Fluorinated surfactant 0.37–1.6 0.6–2.8

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119118

[3] C. Bellona, J.r.E. Drewes, P. Xu, G. Amy, Factors affecting the rejection of organicsolutes during NF/RO treatment, literature review, Water Res. 38 (2004)2795–2809.

[4] A. Sotto, J.s.M. Arsuaga, B. Van der Bruggen, Sorption of phenolic compoundson NF/RO membrane surfaces: influence on membrane performance, Desali-nation 309 (2013) 64–73.

[5] C. Baudequin, Conception d'une unité mobile pour le post-traitement d'eauutilisée pendant des opérations d'extinction d'incendie, in: EA 4038 Labor-atoire de Génie des Procédés et Matériaux, Ecole Centrale Paris, Châtenay-Malabry, 2011.

[6] C.J. van Oss, Development and applications of the interfacial tension betweenwater and organic or biological surfaces, Colloids Surf. B: Biointerfaces 54(2007) 2–9.

[7] M. Elimelech, W.H. Chen, J.J. Waypa, Measuring the zeta (electrokinetic)potential of reverse osmosis membranes by a streaming potential analyzer,Desalination 95 (1994) 269–286.

[8] A.E. Childress, S.S. Deshmukh, Effect of humic acid substances and anionicsurfactants on the surface charge and performance of reverse osmosismembranes, Desalination (1998) 167–174.

[9] C.K. Yeom, S.H. Lee, J.M. Lee, Effect of the ionic characteristics of chargedmembranes on the permeation of anionic solutes in reverse osmosis, J. Membr.Sci. 169 (2000) 237–247.

[10] G. Cornelis, K. Boussu, B. Van der Bruggen, I. Devreese, C. Vandecasteele,Nanofiltration of nonionic surfactants – effect of the molecular weight cutoffand contact angle on flux behavior, Ind. Eng. Chem. Res. 44 (2005) 7652–7658.

[11] K. Boussu, Influence of Membrane Characteristics on Flux Decline andRetention in Nanofiltration, Katholieke Universiteit Leuven, 2007.

[12] H. Li, Y. Lin, P. Yu, Y. Luo, L. Hou, FTIR study of fatty acid fouling of reverseosmosis membranes: effects of pH, ionic strength, calcium, magnesium andtemperature, Sep. Purif. Technol. 77 (2011) 171–178.

[13] G. Hurwitz, G.R. Guillen, E.M.V. Hoek, Probing polyamide membrane surfacecharge, zeta potential, wettability, and hydrophilicity with contact anglemeasurements, J. Membr. Sci. 349 (2010) 349–357.

[14] A. Subramani, E.M.V. Hoek, Direct observation of initial microbial depositiononto reverse osmosis and nanofiltration membranes, J. Membr. Sci. 319 (2008)111–125.

[15] J.A. Brant, A.E. Childress, Assessing short-range membrane–colloid interac-tions using surface energetics, J. Membr. Sci. 203 (2002) 257–273.

[16] T. Bialopiotrowicz, B. Janczuk, The wettability of a cellulose acetate membranein the presence of bovine serum albumin, Appl. Surf. Sci. 201 (2002) 146–153.

[17] C.J. van Oss, Interfacial Forces in Aqueous Media, Taylor and Francis, 2006.[18] M. Mulder, Basic Principles of Membrane Technology, LLC, 2nd edition,

Springer-Verlag, New York, 1996.

[19] Y. Kaya, H. Barlas, S. Arayici, Evaluation of fouling mechanisms in thenanofiltration of solutions with high anionic and non ionic surfactant contentsusing a resistance-in-series model, J. Membr. Sci. 367 (2011) 45–54.

[20] C. Fargues, C. Sagne, A. Szymczyk, P. Fievet, M.-L. Lameloise, Adsorption ofsmall organic solutes from beet distillery condensates on reverse-osmosismembranes: consequences on the process performances, J. Membr. Sci. 446(2013) 132–144.

[21] J. Heo, L.K. Boateng, J.R.V. Flora, H. Lee, N. Her, Y.-G. Park, Y. Yoon, Comparisonof flux behavior and synthetic organic compound removal by forward osmosisand reverse osmosis membranes, J. Membr. Sci. 443 (2013) 69–82.

[22] B. Van der Bruggen, L. Braeken, C. Vandecasteele, Evaluation of parametersdescribing flux decline in nanofiltration of aqueous solutions containingorganic compounds, Desalination 147 (2002) 281–288.

[23] B. Van der Bruggen, L. Braeken, C. Vandecasteele, Flux decline in nanofiltrationdue to adsorption of organic compounds, Sep. Purif. Technol. 29 (2002) 23–31.

[24] H. Li, Y. Lin, Y. Luo, P. Yu, L. Hou, Relating organic fouling of reverse osmosismembranes to adsorption during the reclamation of secondary effluentscontaining methylene blue and rhodamine B, J. Hazard. Mater. 192 (2011)490–499.

[25] I. Escudero, M.O. Ruiz, J.M. Benito, J.L. Cabezas, D. Dominguez, J. Coca, Recoveryof [alpha]-Phenylglycine by Micellar Extractive Ultrafiltration, In Honour ofProfessor Ryszard Pohorecki on the Occasion of his 70th Birthday, vol. 84,2006, pp. 610–616.

[26] R. Urbanski, E. Goralska, H.-J. Bart, J. Szymanowski, Ultrafiltration of surfactantsolutions, J. Colloid Interface Sci. 253 (2002) 419–426.

[27] C.Y. Tang, Q.S. Fu, A.P. Robertson, C.S. Criddle, J.O. Leckie, Use of reverseosmosis membranes to remove perfluorooctane sulfonate (PFOS) from semi-conductor wastewater, Environ. Sci. Technol. 40 (2006) 7343–7349.

[28] C.Y. Tang, Q.S. Fu, C.S. Criddle, J.O. Leckie, Effect of flux (transmembranepressure) and membrane properties on fouling and rejection of reverseosmosis and nanofiltration membranes treating perfluorooctane sulfonatecontaining wastewater, Environ. Sci. Technol. 41 (2007) 2008–2014.

[29] B.V.d. Bruggen, L. Braeken, C. Vandecasteele, Flux decline in nanofiltration dueto adsorption of organic compounds, Sep. Purif. Technol. 29 (2002) 23–31.

[30] A.C. Archer, A.M. Mendes, R.A.R. Boaventura, Separation of an anionicsurfactant by nanofiltration, Environ. Sci. Technol. 33 (1999) 2758–2764.

[31] E. Steinle-Darling, M. Reinhard, Nanofiltration for trace organic contaminantremoval: structure, solution, and membrane fouling effects on the rejection ofperfluorochemicals, Environ. Sci. Technol. 42 (2008) 5292–5297.

[32] Y. Kaya, C. Aydiner, H. Barlas, B. Keskinler, Nanofiltration of single and mixturesolutions containing anionics and nonionic surfactants below their criticalmicelle concentrations (CMCs), J. Membr. Sci. 282 (2006) 401–412.

C. Baudequin et al. / Journal of Membrane Science 458 (2014) 111–119 119