Separation and Purification Technology 141 (2015) 339–353

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Nanofiltration of oil sands boiler feed water: Effect of pH on water fluxand organic and dissolved solid rejection

http://dx.doi.org/10.1016/j.seppur.2014.12.0111383-5866/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Tel.: +1 (780) 492 9099; fax: +1 (780) 492 2200.E-mail address: sadrzade@ualberta.ca (M. Sadrzadeh).

1 Tel.: +1 (780) 492 9099; fax: +1 (780) 492 2200.

Mohtada Sadrzadeh a,⇑, Javad Hajinasiri a,1, Subir Bhattacharjee b, David Pernitsky c

a Department of Mechanical Engineering, University of Alberta, 6-074 NINT Building, Edmonton, AB T6G 2G8, Canadab Water Planet Engineering, 721 S Glasgow Ave, Inglewood, CA 90301, United Statesc Suncor Energy Inc., P.O. Box 2844, 150-6th Ave. SW, Calgary, AB T2P 3E3, Canada

a r t i c l e i n f o

Article history:Received 17 July 2014Received in revised form 30 October 2014Accepted 2 December 2014Available online 23 December 2014

Keywords:Oil sandsSAGDProduced water treatmentNanofiltrationMembrane processes

a b s t r a c t

Nanofiltration (NF) is a promising advanced treatment technique for oilfield wastewaters. In this study,we performed a crossflow NF on model boiler feed water (BFW) obtained from a thermal in-situ bitumenrecovery process called steam assisted gravity drainage (SAGD) with the intent to remove dissolvedorganic matter and silica. Nanofiltration with a tight NF membrane was selected as the treatment tech-nique since it provided reasonable product quality in a single-stage process and was found to be moreenergy efficient than reverse osmosis (RO). Two different feed pH values, 8.5 and 10.5, were investigated.Total organic carbon (TOC) and Si rejection of approximately 98% and divalent cation rejection greaterthan 99% was achieved at both pH values. Higher total dissolved solids (TDS) rejection, but lower fluxrates were seen at pH 8.5 compared to pH 10.5. Surface characterization of the fouled membranes indi-cated the presence of organics (primarily carbon and oxygen) and inorganics (mainly silicon and iron) inthe fouling deposits. Larger amounts of deposit were seen at pH 8.5 compared to pH 10.5. Increasing feedpH from 8.5 to 10.5 recovered the water flux more than 20% which demonstrated the critical role of pH asa pulsation technique to reduce membrane fouling. Fluorescence excitation emission matrix spectros-copy (FEEMs) on the feed and permeate showed that the organic matter that passed through the mem-brane was mainly hydrophilic compounds. Overall, this research indicated that NF is a viable treatmentprocess for silica and TOC removal for SAGD BFW.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

Thermally enhanced oil recovery methods such as steamassisted gravity drainage (SAGD) and cyclic steam stimulation(CSS) are widely practiced for bitumen extraction from oil sandsin Alberta, Canada. In this process, steam is injected underground,which condenses to heat up the bitumen, thereby reducing itsviscosity. The bitumen and condensed water are pumped out tothe surface where the bitumen and water are separated and thewater is treated for reuse as boiler feed water (BFW).

A typical conventional SAGD water treatment plant is shown inFig. 1. In the conventional diluted SAGD process, organic diluentsare added to the produced fluids to reduce the viscosity ofbitumen. The diluted bitumen and the produced water are firstseparated using a series of gravity and flotation vessels, namely,the free water knock out (FWKO) vessel to remove the bitumen,

followed by gravity skim tanks and induced static flotation (ISF)to separate residual oil from the produced water, and walnut shellfilters to bring the free oil content in the produced water below20 mg/L. The produced water is then treated in a warm lime soft-ener (WLS) to remove silica, following which, the suspended solidsare removed in after-filters, and the residual multivalent cationslikes Ca2+ and Mg2+ are removed using weak-acid cation exchanger(WAC). The treated water is then used as BFW in once throughsteam generators (OTSGs). A portion of the blowdown from theOTSGs is typically recycled to the warm lime softener or the BFWtank; the remainder is sent to disposal. It is important to note thatthe entire water treatment process train operates at a water tem-perature of between 80 and 90 �C, a challenging temperature forpolymeric membranes.

No treatment is provided for TDS removal or for the removal ofdissolved organic matter in the conventional process. OTSGs aretherefore used in conventional SAGD applications as they are ableto operate with BFW containing higher levels of TDS and Ca/Mg/Sithan conventional drum boilers. To compensate for the use of lowerquality BFW, steam quality in SAGD OTSGs has historically been

Table 1Typical SAGD OTSG BFW specifications.

Parameter Units Specification

Dissolved and/or particulate hardness mg/L as CaCO3 <0.5pH – 9.8–10.5Silica mg/L as SiO2 <75Free oxygen lg/L as O2 <7Conductivity mS <12Bitumen in water (hexane extraction) mg/L <0.5Turbidity NTU <7.5Iron lg/L as Fe <250

340 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

limited to 80% to provide sufficient water volume to cool the innersurface of the tubes in the radiant section, and to prevent super-sat-uration and formation of inorganic scales. OTSGs BFW specificationsvary somewhat between in situ operators, based on the manufac-turers’ recommendations and by specific operating experience inthe field. Typical SAGD BFW specifications are summarized inTable 1. It should be noted that an alternative process configurationusing evaporators for desalination in conjunction with drum boilershas been employed in some SAGD installation [1].

Deviations from these BFW specifications have resulted in heatexchanger fouling and boiler tube failures in the field [2–4], andsignificant research efforts are underway to understand the mech-anisms of fouling. Analysis of deposits from failed tubes has foundhigh levels of organic carbon in addition to Ca/Mg/Si. It is not clearwhether the high levels of carbon are due to deposition and cokingof free and emulsified oil, or whether they are due to temperature-related precipitation of dissolved organic material. In order to pre-vent tube failures, and to allow operation at higher steam qualities,nanofiltration (NF) was investigated as a polishing technology tofurther reduce the concentrations of Ca, Mg, and Si in the BFW,and to remove dissolved organic matter.

Membrane separation processes have been widely applied inproduced water treatment due to their distinct advantages overtraditional processes, primarily lower operating expenses andenergy consumption. Numerous previous investigators have con-sidered the use of loose membranes, MF and UF, for oil removalfrom oily produced water [5–10]. For separation of silica, dissolvedorganic matter (DOM) and salt from produced water, tighter NFand RO membranes are required. Although, NF and RO are widelyapplied in water and wastewater treatment, there are a fewrecords in peer-reviewed literature for their application in desali-nation and organic removal of oilfield produced water.

A brief overview of published efforts on oilfield produced watertreatment using NF and RO membranes is summarized in Table 2.Although several studies have been published on the use of NF andRO for treatment of oil sands process affected water (OSPW) asso-ciated with the surface mining and extraction of bitumen [11,12],no published studies were found on the use of NF membrane pro-cesses for SAGD produced water treatment. Produced water fromin situ oil sands operations is higher in dissolved organic carbonthan OSPW [11–13], and the nature of the dissolved organics havebeen shown to be different [4], suggesting that organic fouling maybe more severe in SAGD applications.

Fig. 1. Process flow diagram of a typical SAGD-based

Every DOM has specific physicochemical properties, such ascharge and molecular conformation, which controls the rate offouling and subsequently performance of membrane process[14]. Meanwhile, the interaction of DOM functional groups withsolution chemistry, e.g. monovalent and divalent concentrations,pH and presence of silica, induces different effects on propertiesof DOM macromolecules and thus the structure and hydraulicresistance of fouling layer [15–18]. Many of earlier studies indi-cated that the DOM in mining OSPW consists primarily of naph-thenic acid-like compounds [19–22]. However, DOM in SAGDproduced water was demonstrated to be more representative ofhumic acids than naphthenic acids [4] which is predicted to resultin significant change in membrane performance.

The performance of membranes is proven to be significantlyaffected by the feed pH through changing DOM surface charge[23–27]. Dyke and Bartels [28] showed that, DOM rejection mayincrease or remain constant by increasing the pH, depending onthe type of utilized produced water. In some waters they observedDOM reduction by decreasing the pH from their original value.However, Mehrotra and Banerjee [13] showed that by decreasingthe feed pH from 8 to 4, the permeate TOC content decreased morethan 90%. They attributed this to the precipitation of inorganic car-bonates and bicarbonates on the membranes surface at lower pHand subsequently enhancing the permselective properties of mem-branes. Tao et al. [29] increased their feed pH form 8.0 to 11 to pre-vent oil precipitation on the membrane surface by reaching tobeyond its solubility limit. Doran et al. [30] suggested an approachfor increasing the boron removal by increasing the pH of the ROfeed water to ionize the boron. These findings suggest that theeffect of pH on membrane fouling (flux decline) and rejection of

in situ bitumen extraction water treatment plant.

Table 2Overview of earlier studies on oilfield produced water using NF and RO membranes.

Refs. Producedwatertype

Membranetype

Membrane saltrejection

Feed characteristics Major findings

pH Temp.(�C)

TDSa (mg/L) TOCb

(mg/L)Silica(mg/L)

Ca2+ & Mg2+

(mg/L)

Kim et al. [25] OSPW NF & ROg

(PA TFC)NF 96–98%RO 99–99.5%i

8–9 22 ± 1 2477 48.3 n/a 73 The average efficiency of salt removal was about 87.3% forRO membranes and 68.9% for NF membranes

Peng et al. [26] OSPWc NFd (PAe

TFCf)95% 7.3–8.5 20 1549–4920 lS/cm 46–85 n/a 30–80 The applied NF membranes rejected more than 95% of TOC

(mainly naphthenic acid) and divalent ionsMehrotra & Banerjee

[27]Oil sandsin situproducedwater

RO (PA TFC) n/a 7.3 24 1770 270 n/a 70 Up to 98% and 78% of TDS and TOC were removed.Decreasing feed pH to 4, increased TOC rejection from 78%to 98%

Dyke & Bartels [28] Offshoreproducedwater

NF(polymerich)

n/a 6.5 40 58,000 176 19 1400 31–90% organic matter rejection and 6–11% inorganicrejection were observed. Rejection of organic matterincreased 60% by increasing pH from 6.6 to 8.2

Tao et al. [29] Oilfieldbrine

RO(polymerich)

n/a 7.8–8.0 40 6862 170 250 80 TDS rejection of 95% and TOC rejection of 96% observed. pHof feed increased to 10.6–11 to avoid oil precipitation onmembrane

Doran et al. [30] Oilfieldproducedwater

RO(polymerich)

n/a 8.2–10.8

35 6000 120 10 1–5 TOC and TDS reduced by 98%. As the pH increased to 10.8 alarger faction of the boron become ionized and rejected

Cakmakce et al. [31] Oilfieldproducedwater

NF & RO (PATFC)

NF 90% RO 98–99.5%i

7.1–7.8 25 7000–30,000 780–1680

n/a n/a Different pretreatment alternatives to RO and NF wereinvestigated and TDS and organic matter removal up to 98%obtained

Murray-Gulde et al.[32]

Oilfieldproducedwater

RO (PA TFC) 99% 7.7 Roomtemp.

6554 77 n/a 65 95% and 77% of TDS and TOC were rejected, respectively

Mondal et al. [33] Oilfieldproducedwater

NF & RO (PATFC)

NF 80–96% RO99.4

8.4–8.7 Roomtemp.

722–2090 69–136 7.4–14.4

1.7–12.3 RO membrane rejected 67% and 47% and NF membranesremoved 28–34% and 15–47% of TOC and TDS, respectively

Xu et al. [34] Gas fieldproducedwater

NF & RO (PATFC)

NF 90–96% RO 98–99.5%

8.5 ± 0.2 22.5 5520 ± 718 1.8 ± 0.2 2.7 ± 0.6 40 ± 5 Up to 98% of salt and 82% of organic matter rejected by NFand RO membranes

a Total dissolved solids.b Total organic carbon.c Oil sands process affected water.d Nanofiltration.e Polyamide.f Thin film composite.g Reverse osmosis.h Type of the polymer is unknown.i The corresponding value is MgSO4 rejection.

M.Sadrzadeh

etal./Separation

andPurification

Technology141

(2015)339–

353341

342 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

organic and inorganic materials is entirely related to the character-istics of treated water.

A closer look to Table 2 reveals interesting differences for TOCand TDS rejection values. In some studies NF and RO resulted insimilar TOC and TDS rejections [29–31], while higher TOC rejectionthan TDS or vice versa, were also observed [13,28,32–34]. This isattributed to different molecular weights and physicochemicalproperties of organic matter as well as different types and concen-trations of monovalent and divalent ions in various producedwaters.

In the present work, the performance of a commercial NF mem-brane for Ca/Mg/Si polishing and DOM removal from a modelSAGD BFW was tested. A tight NF membrane was chosen in prefer-ence to a loose NF and a RO membrane as it offered significantlyhigher removal of organic matter and salt than loose NF as wellas less energy consumption than RO. Crossflow NF experimentswere conducted using a model BFW, prepared by diluting SAGDboiler blowdown (BBD) water obtained from an operating conven-tional SAGD plant in northern Alberta. Tests were performed at atemperature of 50 �C and at pH values of 10.5 and 8.5. pH 10.5 cor-responds to the typical BFW pH used in operating plants. pH 8.5was tested to investigate the effect of pH reduction on membraneperformance; pH reduction from 10.5 to 8.5 has been shown toincrease ultrafiltration membrane fouling when treating SAGDblowdown streams [35]. Based on a parametric study, it is pro-posed here that a pH pulsation technique can be used to reducethe membrane fouling and recover the water flux. A suitablydesigned crossflow NF process is demonstrated be a superior alter-native technique to current SAGD produced water treatment meth-ods, especially in terms of producing higher quality water byconsuming lower amount of chemicals and energy.

2. Materials and methods

2.1. Water sample and reagents

A model SAGD BFW was prepared by dilution of BBD waterfrom a SAGD water treatment plant located in the Athabasca oilsands region of Alberta, Canada. The samples were collected hot,shipped in sealed containers, and kept in a nitrogen blanket untilthey were opened for treatment [35]. BBD water was diluted five-fold to acquire the properties of BFW (Table 3). Concentrations ofdissolved silica and other inorganics were measured by inductivelycoupled plasma-optical emission spectrometry (ICP-OES). 1 M HCland NaOH (Sigma Aldrich) were used to adjust the model BFW pH.

2.2. Membrane

Nanofiltration (NF) was performed with a FilmTec NF90 mem-brane, which is a thin film composite membrane comprises threelayers: (i) a polyester support web, (ii) a microporous PSf interlayer and (iii) an ultrathin aromatic polyamide active layer. The

Table 3Properties of BBD water diluted to model BFW.

Parameters BBD water Model BFW

pH 12.0 10.5TOC (mg/L) 2450 500TDS (mg/L) 10,500 1800Conductivity (mS/cm) 17.8 3.5Na+ (mg/L) 4280 880Cl� (mg/L) 2700 510Mg2+ (mg/L) 0.37 0.18Ca2+ (mg/L) 3.30 0.66Iron, total (mg/L) 2.12 0.48SiO2, dissolved (mg/L) 92 21

average roughness of NF90 membrane is in the range of 60–70 nm [36–38] and its zeta potential from streaming potentialmeasurement is �18 mV within the pH range of 7–10 in 10 mMNaCl solution [38]. The average hydraulic resistance (Rm), realrejection (Rr) and observed rejection (Ro) of the membrane are4.0 ± 0.1 � 1013 m�1, 97% and 90 ± 1%, respectively [39]. A looseNF (NF270) and a RO (BW30) membrane with observed rejectionof �50 and �99.5, respectively, were also supplied from FilmTecand tested to provide sufficient evidence about the advantages ofNF90 for treatment of BFW. Based on manufacturer data, maxi-mum operating pressure and temperature and tolerable pH rangefor all membranes are 4135 kP, 45 �C and 3–10, respectively. Inthe present work, the feed water temperature and the pH weredeliberately adjusted at 50 �C and 10.5, which are outside the man-ufacturer’s normal recommended operating envelope, to performfiltration under the harsh conditions expected for the full-scaleindustrial application. Hence, the experimental results are validfor the short-term use of polymeric membranes for treatment ofBFW. Other researchers have also demonstrated successful NFand RO treatment of oilfield produced waters at polymeric mem-brane threshold temperatures [28–30].

2.3. Crossflow membrane filtration setup

The schematic view of the crossflow membrane filtration setupis shown in Fig. 2. The maximum allowed operating pressure of thesetup is 6895 kPa which is provided by a diaphragm pump (Hydra-Cell) of maximum flow 6.8 LPM. The effective filtration and flowchannel area are 140 cm2 and 1.62 cm2, respectively. In all experi-ments crossflow rate was set at 1 LPM which gives a 0.1 m/s cross-flow velocity and a laminar Reynolds number of 344. The modelBFW was supplied from a 19 L atmospheric pressure tank. A bypassvalve and a back pressure regulator (Swagelok) were used to adjustthe feed flow rate and the trans-membrane pressures. The feedtemperature was increased to 50 �C by a recirculating chiller (Iso-temp 3013, Fisher Scientific). A weighing balance (Mettler ToledoEL4001) was used to measure the permeate flow rate. The conduc-tivity of the feed water was obtained by an Accumet Research(AR60) conductivity meter. The pH and conductivity of the perme-ate were measured by a Mettler Toledo (SevenMulti S47) pH andconductivity meter. All measurements were directly collected intoa computer using LabVIEW (National Instruments) data acquisitionsoftware [39].

2.4. Experimental methodology

Four experiments were conducted using NF90 membrane to findthe effect of pH on water flux, TOC and TDS rejection and depositionof organic and inorganic matter on the membrane surface (Table 4).Two extra experiments were also conducted similar to run 1(Table 4) using NF270 loose NF and BW30 RO membranes to com-pare their performances with NF90. During each 6 h experiment,temperature, feed flow rate and permeate flux were maintained at50 �C, 1.6 � 10�5 m3/s and 1.8 � 10�5 m3/m2s, respectively. Thetrans-membrane pressure was adjusted on 350, 550 and 825 kPafor NF270, NF90 and BW30, respectively, to acquire constant initialpermeate flux of 1.8 � 10�5 m3/m2s. All membrane samples weresoaked in de-ionized water for 24 h prior to use. Before each exper-iment the membranes were compacted with de-ionized water at1400 kPa until the permeate flux stabilized.

2.5. Characterization techniques

2.5.1. Total organic carbon (TOC) analysisTotal organic carbon (TOC) is the amount of carbon bound in

organic compounds present in the water and is used as an indicator

Fig. 2. Schematic view of crossflow filtration setup.

Table 4Variation of pH with time in conducted experiments.

Run Experiment time (h)

1 2 3 4 5 6

1 10.5 10.5 10.5 10.5 10.5 10.52 8.5 8.5 8.5 8.5 8.5 8.53 8.5 8.5 8.5 8.5 10.5 10.54 10.5 10.5 8.5 8.5 10.5 10.5

M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353 343

of DOM in the present work. It was measured using a combustion-type TOC analyzer (Shimadzu, model TOC-V; detection range 3–25,000 mg/L). Since organics in the SAGD produced water aremainly water soluble [4], the unfiltered model BFW was analyzedby the TOC analyzer.

2.5.2. Inductively coupled plasma-optical emission spectroscopy (ICP-OES)

Emission spectroscopy using ICP is a rapid, sensitive and conve-nient method for the determination of metal ions in aqueous solu-tions. The concentrations of dissolved silica and other inorganicspresented in Table 3 were measured by an ICP-OES instrument(Agilent 735 ICP-OES) using EPA method 200.7. In this methodthe water sample is nebulized and the resulting aerosol is trans-ported into inductively coupled argon plasma generated by radiofrequency power. The high temperature (6000–10,000 K) of theplasma leads to almost complete dissociation of molecules andefficient atomization and ionization in the sample. Emission spec-tra are produced when the excited atoms and ions return to lowerenergy states. The spectra are dispersed by a high resolutionechelle polychromator and the intensities of the lines are moni-tored by a charged coupled device (CCD). In OES, the power ofthe radiation emitted by a constituent after excitation is directlyproportional to its concentration.

2.5.3. Fluorescence excitation emission matrix spectroscopy (FEEMs)Fluorescence excitation emission matrix (FEEM) spectroscopy

has been shown in previous studies to be a useful tool for the clas-sification of various dissolved organic fractions in present in SAGDproduced water [4]. A wavelength range of 200–500 nm with 5 and

10 nm intervals for excitation and emission wavelengths, respec-tively was used in this research. The FEEM of de-ionized waterwas subtracted from the FEEM of the samples to remove most ofthe Raman peaks. The model BFW samples were diluted using DIwater to a TOC level of around 12 mg/L to avoid inner filtration(quenching) effects on the fluorescence analysis. The pH valuesof feed and permeate were similar in all experiments whichremoves the pH induced effects on fluorescence for comparing feedand permeate FEEM results. For membrane feed samples adjustedto pH = 8.5, the permeate pH was increased to 10.5 prior to FEEManalysis to find the effect of pH on FEEM results.

2.5.4. Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy

Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy provides information on the type of functionalgroups present to the depths less than 1 lm. NF90 membranesbefore and after WLS inlet filtration were examined using ATR-FTIRmicroscope (Thermo Nicolet Nexus 670 FTIR, USA). This instru-ment is equipped with a mercury–cadmium–tellurium (MCT)detector and has a resolution of 4 cm�1. A total of 512 scans wereaveraged for each spectral measurement. The internal reflectionelement was a zinc selenide (ZnSe) ATR plate with an apertureangle of 45�. All membrane samples were scanned over the rangeof 600–4000 cm�1.

2.5.5. Field emission scanning electron microscope-energy dispersiveX-ray (FESEM-EDX)

The membranes were sputter coated with a thin film of chro-mium. Surface images of the membranes were obtained using JEOL6301F FESEM. All membranes were imaged at a magnification of20,000 times. FESEM provides qualitative information on the depo-sition of foulants on the membrane. Semi-quantitative elementalanalysis was done via a PGT IMIX EDX system with 135 eVresolution.

2.5.6. X-ray photoelectron spectroscopy (XPS)X-ray photoelectron spectroscopy (XPS) is a surface sensitive

technique widely used for analyzing elemental composition(except for H and He). Chemical binding information for the top

344 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

1–10 nm of the surface can be provided by XPS. Original and usedmembranes analyzed using a Kratos AXIS ULTRA spectrometerequipped with a monochromatic Al Ka X-ray source. In the presentwork the source was run at a power of 210 W (14 mA, 15 kV) and ahybrid lens with a spot size of 700 lm � 400 lm. Survey spectrawere collected with a pass energy of 160 eV, step size of 0.4 eV,and sweep time of 100 s in the range of 0–1100 eV. High resolutionspectra (collected with pass energy of 20 eV, step size of 0.1 eV,and sweep time of 200 s) were measured in the appropriate bind-ing energy (BE) ranges as determined from the survey scan for theNa 1s, Ca 2p, Fe 2p, Si 2p, O 1s, N 1s, C 1s, S 2p and Cl 2p core lines.The number of scans varied from 10 to 25 to obtain good signal/noise ratios. High resolution spectra were analyzed with the aidof the CasaXPS software.

Fig. 3. Water flux and rejection for model BFW filtration using loose NF (NF270),tight NF (NF90) and RO (BW30) membranes at pH = 10.5.

3. Results and discussion

When salt and DOM need to be removed from water, as in thepresent work, NF and RO are applied. Nanofiltration and RO arepressure-driven processes where the transmembrane pressuremust overcome the osmotic pressure of the feed water to forcewater through a dense membrane. Generating this pressurerequires certain amount of energy which is directly related to thedensity of the membrane. Generally, for denser membranes rejec-tion will be increased but so will the required pressure and thusthe energy demand [40]. Hence, the type of the applied membranefor a specific application is typically identified by a trade-off rela-tionship between energy consumption and product quality. Lowerenergy consumption is always favored and the product quality isidentified by environmental standards (for water disposal) andtechnical constraints (for water recycle and reuse).

In this section, energy requirement and product quality of atight NF membrane (NF90), a loose NF membrane (NF270), and aRO membrane (BW30) are compared to select the winner and con-duct further experiments on that. Flux (normalized with respect toinitial flux) and TOC and TDS rejection of these membranes at 50 �Cand pH of 10.5 are shown in Fig. 3. All experiments were conductedat a constant feed flow rate (1.6 � 10�5 m3/s) and permeation flux(1.8 � 10�5 m3/m2s) on a same model BFW to minimize the effectof feed chemistry and hydrodynamics. As can be observed, fluxdecline was similar for NF90 and BW30 membranes (�10%) after6 h run. NF90 and BW30 membranes have almost the same surfaceproperties (roughness, contact angle and zeta potential values[10,33,41]) and are expected to show the same fouling behavior,thus flux decline. A slight improvement in NF270 flux decline isattributed to its very smooth surface and stronger hydrophilicproperties than NF90 and BW30 as indicated by their surfaceroughness and contact angle values [10]. According to Fig. 3,NF90 and BW30 rejected more than 98% of DOM which is themajor responsible material for fouling of SAGD equipment [2–4].The maximum TDS rejections obtained by NF90 and BW30 were96% and 98%, respectively. NF270 rejected a maximum of 72% ofsalt and silica and 80% of DOM. Flux and rejection results demon-strated that the tight NF can produce a very high quality productwhich is comparable with the RO one. Compared with BW30,NF90 was also found to be less energy intensive, thus associatedwith lower operating cost. The applied trans-membrane pressurefor NF270, NF90 and BW30 membranes were 350, 550 and825 kPa, respectively. This shows that NF90 is more energy effi-cient than BW30 while providing almost the same water quality.Hence, a single stage NF with NF90 membrane is suggested for fur-ther experimental investigations. In what follows, permeationproperties of NF90 membrane at various pH as well as character-ization of feed, permeate and fouling deposits on the membraneare presented.

3.1. Water flux and TDS and TOC rejections of NF90 membrane

The performance of a membrane process is generally monitoredby measuring the permeation of the desired material (water) andthe removal of unwanted material (salt and DOM). The rate ofwater flux decline over time gives an indication of the fouling ten-dency of the membrane and the fouling potential of the wastewa-ter. Fouling deteriorates membrane performance by decreasing thewater flux and its quality and ultimately shortens membrane life[14]. Hence, vast efforts have been devoted to reduce membranefouling before operation (e.g. improving membrane propertiesand pretreatment of feed water), during operation (e.g. optimiza-tion of operating condition and pulsation techniques) and afteroperation (physical and chemical cleaning). In the present work,the effect of pH on water flux recovery and TOC and TDS rejectionduring crossflow NF of SAGD produced water was investigated.

3.1.1. Continuous operation at fixed pHFlux and TOC and TDS rejection of the NF membrane at 50 �C

and a constant pH of 10.5 are shown in Fig. 4. The normalized fluxdeclined due to the combined fouling of silica, organic matter anddivalent ions present in the model BFW (Table 3). However,according to the data presented in Table 3, the concentration of sil-ica and divalent ions in the model BFW (�20 mg/L) is very low ascompared to the concentration of organic matter (500 mg/L).Hence, DOM fouling is expected to be the dominant fouling mech-anism in the present work. A rapid initial flux decline (seen afterthe first 6 h of filtration), followed by a slower flux decrease is evi-dence of DOM fouling [42–44]. This behavior can be explained byconsidering the mechanism governing organic fouling. Initialadsorption of organic foulants on the membrane surface decreasespermeate flux sharply due to DOM gel formation (plugging of hotspots), pore blocking and constriction as well as induced hydro-phobic properties. Slower flux decline, in the next fouling stage,is attributed to the compaction and thickening of the fouling layer,which induces an additional resistance as well as a reduction ofpermeation drag as permeation flux decreases [42,45–47].

A study by Nghiem and Hawkes [48] showed that a sharp fluxdecline due to pore blocking and pore constriction would be more

Fig. 4. Water flux and rejection for model BFW filtration using NF90 membrane atpH = 10.5.

M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353 345

severe with membranes having a larger pore size. For salt rejectingNF and RO membranes, plugging of hot spots by organic matter hasbeen found to be the key reason for the rapid initial flux decline[43,49–52]. Hot spots are the valleys on the membrane surfacewith the lowest thickness and highest water flux. Rapid cloggingof valleys results in significant loss of permeate flux [50]. Themembrane used in the present study (NF90), with the averageroughness of 60–70 nm [36–38] and molecular weight cut off(MWCO) of 150–250 Da [53,54], is certainly susceptible to valleyblocking by particles larger than 60 nm as well as pore blockingby DOM with molecular weight less than 250 Da. As will be dis-cussed later in this paper, FESEM images found the presence of par-ticles with 100 nm diameter on the fouled membrane.Furthermore, our previous study showed that approximately 40%of the organic fraction in the BBD had molecular weight less than500 Da [4] which increases the possibility of pore blocking byDOM.

Increasing hydrophobicity of membrane generally leads tomore vulnerability to fouling due to hydrophobic interactionsbetween the membrane surface and the hydrophobic foulants[43]. The variation of the hydrophilic/hydrophobic properties of amembrane during filtration depends on the hydrophilic/hydropho-bic properties of both the virgin (unfouled) membrane and the feedwater constituents [43,55]. Violleau et al. [56] observed that theirinitially hydrophilic PA membrane became more hydrophobic afterthe sorption of organic foulants. Cho et al. [57] showed that organicfouling decreased the hydrophobicity of hydrophobic membraneswhereas increased the hydrophobicity of hydrophilic membranes,regardless of type of organic foulants. Tu et al. [43] indicated thatmembranes fouled by humid acid, sodium alginate, colloidal silicaand CaSO4 all became more hydrophobic than virgin membranes.The induced hydrophobicity was more severe in the more hydro-philic membrane. Yuan and Zydney [58] observed an increase inthe hydrophobicity of hydrophilic polyethersulfone membranes(contact angle �44�) fouled by two types of humic acids; morehydrophobic soil-based Aldrich and less hydrophobic aquaticSuwannee River. They reported more induced hydrophobicity bya soil based NOM than an aquatic one. In this study, the majororganic components in the model BFW are hydrophobic acids(mainly humic type �40% [4]) and the active layer of the NF90membrane is made of moderately hydrophilic PA (contact angle�60� [59]). According to the literature, the specific UV absorbanceat 254 nm (SUVA254, defined as the UV absorbance divided by the

concentration of dissolved organic carbon) is a good criterion toclassify organic materials as humic or non-humic type [60,61].SUVA254 values greater than 4 L mg�1 m�1, as in the model BFW,indicate the presence of primarily humic type organics with higharomacity and hydrophobicity [4]. Hence, hydrophobicisation ofthe NF90 membrane surface after fouling becomes inevitable.

It must be noted that the flux decline observed in Fig. 4 mightbe intensified (10% flux decline within 6 h) by the interactionbetween organic matter and salt. Taking a closer look at Table 3,it is found that the ionic strength of the model BFW is relativelyhigh with a Na+ concentration of 880 ppm. As a general rule,organic fouling becomes more severe as the ionic strength of thefeed water increases [16,18,38]. Higher ionic strength leads to asignificant reduction of membrane and organic macromoleculessurface charges due to double layer compression and chargescreening. This causes a decrease in electrostatic repulsionbetween the membrane surface and DOM. Therefore, the deposi-tion rate of DOM onto the membrane surface increases, whichresults in a thicker deposit layer. Furthermore, organic macromol-ecules become coiled due to reduced interchain electrostatic repul-sion at high ionic strength, which leads to the formation of aclosely packed fouling layer. The resulting fouling layer providesa higher hydraulic resistance to water flow; this leads to a severepermeate flux decline [18].

Fig. 4 shows that the NF90 membrane rejected 80% of the TDSduring the initial stage of filtration, which is less than the realrejection of the membrane (90–95%). TDS rejection increased overtime and reached the real rejection value after 6 h of flow. Lee et al.[62] demonstrated that for the cake enhance concentration polar-ization (CECP) governed fouling mechanisms, like colloidal fouling,salt rejection decreases with time. However, the enhancement ofTDS rejection with time in the present work again confirms thedominance of organic fouling in SAGD produced water filtration.Plugging of hot spots by organic matter resulted in a decrease inwater flux as demonstrated in the literature [43,49–52]. TOC rejec-tion was consistently �98% during the experiment. This is attrib-uted to the larger size of DOM compared to the dissolved solids,mainly Na+ and Cl� monovalent ions, in the water. As a matter offact, the NF90 membrane was able to reject 98% of the organic mat-ter at a feed pH of 10.5, regardless of the fouling progress on themembrane surface. DOM/membrane interaction during the initialstage of filtration and DOM/DOM interaction on the fouled mem-brane had the same effect on the removal of organic matter. Itcan be predicted that the portion of DOM which passed throughthe membrane was small hydrophilic organic compounds. Thishypothesis can be tested by conducting FEEMs analysis on perme-ate and feed samples, as will be presented later.

By decreasing the feed pH, water flux and TDS rejection behav-ior with time changed significantly. As can be observed in Fig. 5,water flux decreased to 93% of the initial flux after 2 h, thenremained constant. TDS and TOC rejections remained constantfor the whole range of experiment at 98%. According to our earlierstudy, acidification of feed water causes precipitation of silicananoparticles and also coprecipitation of organic compounds byadsorption on the surface of the silica nanoparticles [35]. This isdue to the decrease in electrostatic repulsion of these foulants afterprotonation at lower pH values. Detailed explanation of electro-static double layer interactions between the solutes is presentedin the next section. Deposition of both silica particles and organiccompounds on the membrane surface results in the formation ofa cake layer with a high resistance to permeate flow, and subse-quently low water flux. It must be noted that the vertical axis inFig. 5 is the ratio of water flux to the initial flux. This dimensionlessvalue is helpful to find the rate of flux decline irrespective to theexact flux values. Rapid formation of a cake layer on top of themembrane at lower feed pH switches the membrane-foulant

Fig. 5. Water flux and rejection for model BFW filtration using NF90 membrane atpH = 8.5.

Fig. 6. Water flux and rejection for model BFW filtration using NF90 membrane atpH = 10.5–8.5.

346 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

interaction phenomenon to foulant–foulant interaction in less than2 h. Consequently the steady state condition is reached in a shortertime. The cake filtration increased the TDS rejection from 95% (atpH = 10.5 after 6 h run) to 98% (at pH = 8.5) which was constantduring the experiment.

Comparing filtration results at pH 8.5 and 10.5, it was foundthat the lower pH value led to more stable results (in terms ofwater flux and rejection). However, coprecipitation of silica andorganic matter on the membrane surface at lower pH is predictedto decrease the permeate water flux. In order to determine theeffect of pH on water flux and flux recovery, fouling behaviorwas studied in a dynamic pH experiment. As will be discussedlater, an instantaneous change in water flux by a sudden changein feed water pH helps to understand the key role of pH for SAGDNF treatment.

Fig. 7. Water flux and rejection for model BFW filtration using NF90 membrane atpH = 10.5–8.5–10.5.

3.1.2. Membrane operation with varying pHFigs. 6 and 7 show the effect of a step change in feed pH during

constant operation on flux and rejection. As shown in Fig. 6,increasing the pH from 8.5 to 10.5 increased the flux by 20%, butdecreased the TDS rejection. Fig. 7 shows that pH reduction from10.5 to 8.5 decreased the flux suddenly and increased the TDSrejection to 98%. Returning the pH back to 10.5 quickly returnedthe flux and rejection to previous trend. For all pH values, morethan 98% of the organic matter was rejected by NF. TOC reducedfrom 500 ppm to less than 10 ppm in all experiments. DynamicpH experiments confirmed that a steadier water flux and higherTDS rejection occurred at lower pH. However, the overall waterflux was higher at higher pH values. Instantaneous change of fluxand TDS rejection by injecting acid or alkaline into the feed solu-tion demonstrates the significant role of pH on fouling, especiallyin presence of both silica and organic matter. The reason behindthis behavior is believed to be the rapid change in foulant/foulantand foulant/membrane interactions by changing the pH.

The permeate flux under various pH values is affected not onlyby the characteristics of the membrane but also by the propertiesof the solution. The effect of feed pH on the size and zeta potentialof organic matter has been broadly reported in the literature[18,23–27]. These investigations found that no obvious variationexisted in the average size of organic matter under various pH val-ues, however their surface charges were shown to become morepositive with decreasing pH. The higher negative charge at higherpH values causes the cake layer to become more ‘‘open’’ due to the

inter-foulant repulsion, and this increases the permeation flux. Inthe present work, DOM is the major constituents responsible forfouling. Protonation of DOM functional groups (mainly carboxylicacid) at lower pH reduces the charge and subsequently electro-static repulsion of organic matter [18,23–27]. In addition, macro-molecular conformation of organic matter varies with pH, so thatsmaller configuration is created at lower pH values [16,18]. Thiscauses the formation of a denser fouling layer and subsequently,a flux decline.

In addition to inter-particle repulsion, there is also particle-membrane repulsion due to a larger negative charge on the mem-brane surface at higher pH [63,64]. Elimelech and Childress [44,64]conducted a comprehensive work on the effect of pH on zetapotential for thin film polyamide membranes at NaCl, Na2SO4,CaCl2 solutions of various concentrations. In all cases, the zetapotential of the membranes became more negative as pHincreased. Polyamide membranes contain carboxylic (R�COO�)and amine (R�NH3

+) ionizable functional groups which are respon-sible for development of surface charge [64,65].

The inter-particle repulsion as well as particle-membranerepulsion prevent the particles from depositing, and lead to the

Fig. 8. Fluorescence excitation–emission matrices (FEEMs) of model BFW atpH = 8.5, permeate at pH = 8.5 and permeate at pH increased to 10.5. Excitationat 5 nm intervals from 200 to 500 nm and emission data collected at an interval of10 nm. Permeate has the effective TOC of 9 and the feed was diluted to the TOC of12 mg/L. The color scale representing the fluorescence intensity is logarithmic in allparts of the figure with the range varying from 0.1 (blue) to 500 (red). (Forinterpretation of the references to color in this figure legend, the reader is referredto the web version of this article.)

Fig. 9. Fluorescence excitation–emission matrices (FEEMs) of model BFW andpermeate at pH = 8.5. Excitation at 5 nm intervals from 200 to 500 nm and emissiondata collected at an interval of 10 nm. Permeate has the effective TOC of 9 and thefeed was diluted to the TOC of 12 mg/L. The color scale representing thefluorescence intensity is logarithmic in all parts of the figure with the rangevarying from 0.1 (blue) to 500 (red). (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.)

Table 5Fluorescence peak intensities of fluorophores in SAGD BFW (data related to signatureof organic acid fractions was taken from Thakurta et al. [4]).

DOM fractions Ex/Em range

Model BFW 200–375/300–500Hydrophobic acid (HPoA) 310–340/400–500Hydrophobic neutral (HPoN) 225–250/325–380Hydrophobic base (HPoB) 260–290/280–320Hydrophilic acid (HPiA) 320–375/375–500Hydrophilic neutral (HPiN) 210–225 and 250–275/280–320Hydrophilic base (HPiB) 210–225/275–310 and 325–400

M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353 347

reduction of the thickness of cake layer. These phenomena canexplain the higher permeation fluxes observed at higher pH values[16,18,38].

3.2. Rejection of organics

Fluorescence excitation emission matrix spectrophotometry(FEEMs) has been used in prior investigations of SAGD waters to

characterize the nature of the dissolved organics [4]. The FEEM sig-natures of DOM in the model BFW and the permeate at two pH val-ues, 8.5 and 10.5, are shown in Figs. 8 and 9. The fluorescenceresponse for the feed occurs over a wide range of wavelengths,with a dominant peak at an excitation/emission (Ex/Em) wave-length range of 210–350/325–450 nm. This wide range of wave-lengths demonstrates that a wide variety of organic compoundsexists in the model BFW. The Ex/Em wavelength ranges for peaksassociated with these organic compounds are presented in Table 5.The fluorescence emission intensity peaks of DOM are generallyobserved due to the presence of high aromaticity, hydroxyl, andamine groups in the organic fluorophores. The fluorescence inten-sity of any fluorophore can also be reduced by the interfering

Table 6Inorganic rejection by NF90 membrane obtained by ICP-OES.

Elements (mg/L) RDLa Model BFW NF90 permeate Rejection (%)

Na+ 0.5 880 53 94Cl� 1.0 510 15 97Mg2+ 0.02 0.18 <0.02 >99Ca2+ 0.03 0.66 <0.03 >99Iron, total 0.03 0.48 <0.03 >99SiO2, dissolved 0.1 21 0.4 98

a Reportable detection limit.

Fig. 10. ATR-FTIR spectra of NF90 membrane before and after filtration.

348 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

effects of other molecules present in a system containing a mixtureof fluorophores. In the present work, in presence of high concentra-tion of hydrophobic acid (HPoA), hydrophilic acid (HPiA),

Fig. 11. FESEM-EDX of (a) original membrane, (b) used membrane at pH = 10.5, (

hydrophilic base (HPiB) and hydrophobic neutral (HPoN), the fluo-rescence intensity of hydrophilic neutral (HPiN) and hydrophobicbase (HPoB) fractions are quenched in the feed fluorescence EEM,especially at pH = 10.5.

The FEEMs contour maps for the membrane permeates showthat some signatures have been totally removed after NF. As canbe seen, the major organic compounds in the permeate werehydrophilic compounds (HPiN, HPiA and HPiB). Lower molecularweight hydrophilic DOM could easily pass through the membrane,especially at the initial stage of filtration (see time 30 min inFigs. 4–7). It is worth noting that all the FEEMs were obtained foridentical TOC of 9–12 mg/L. Hence, the fluorescence intensity inthe feed compared to the permeate was not a simple concentrationeffect.

It must be noted that the NF90 membrane rejected 98% oforganic matter in the model BFW; from 500 mg/L to around10 mg/L. Since hydrophilic matter makes up almost 45% of modelBFW [4], more than 95% of these organic compounds were alsorejected and deposited on the membrane surface. This observationsuggests that all types of organic compounds in the model BFWwere responsible for the fouling of the NF membranes.

3.3. Rejection of inorganics

Rejection of inorganic materials from model BFW was measuredby ICP-OES analysis and the results are presented in Table 6.Greater than 99% rejection was seen for divalent ions such asFe2+, Ca2+ and Mg2+. 98% rejection was observed for the dissolvedsilica. These low levels of inorganic, scale-forming species in theNF permeate, would greatly reduce the fouling propensity of theBFW if NF was applied as a polishing step in the conventional SAGDprocess train.

c) used membrane at pH = 8.5 and (c) used membrane at pH = 8.5 then 10.5.

Fig. 12. FESEM-EDX of beads, small crystals and large crystals deposited on the membrane surface fouled at pH = 8.5.

M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353 349

3.4. Fouling characterization

Virgin and fouled membranes were analyzed by ATR-FTIR,FESEM-EDX and XPS to identify the elements deposited on themembrane surface and to quantify the amount of foulant.

Fig. 13. XPS survey spectra of NF90 membrane before and after filtration.

3.4.1. ATR-FTIR resultsAll samples were scanned from 650 to 4000 cm�1. Over the

wave number range 2000–4000 cm�1 little change in the spectraof the base membrane was observed after filtration. Thus Fig. 11gives results only for wave numbers between 650 and2000 cm�1. This figure indicates that the peak heights associatedwith the base membrane are reduced after filtration.

The ATR-FTIR spectra contain peaks at 694, 1151, 1487, 1503and 1584 cm�1 that indicate the presence of the PSf interlayer.Peaks at 1650 and 1541 cm�1 indicate amide I and amide II forthe NF90 membrane. NF90 is made from m-phenylene diamine, aprimary amine. Though filtration leads to a decrease in PA andPSf associated peak heights, no new peaks are detected. Theabsence of peaks representing organic foulants is most likely dueto the fact that the peaks associated with PSf and PA are much lar-ger than the small peaks that represent adsorbed organic species.

3.4.2. FESEM-EDX resultsFig. 11 shows FESEM images of the NF90 membranes before and

after filtration. After NF, a coating of rejected solutes on the mem-brane surface was formed. Larger amounts of foulant wereobserved when the pH of the feed water was decreased to 8.5, asshown in Fig. 11c and d. It is hypothesized that decreasing thepH caused precipitation of silica particles and coprecipitation oforganic compounds, which are adsorbed on the surface of the silicananoparticles. EDX elemental analysis indicated the presence ofsilica and iron in the foulant material for all used membranes.The sulfur and oxygen peaks in Fig. 11 are likely related to theNF90 PSf microporous interlayer as seen in the ATR-FTIR spectra(Fig. 10). The iron peak became larger and the sulfur peak becameshorter as the pH decreased from 10.5 (Fig. 11c) to 8.5 (b). Thisresult shows that more solutes precipitated on the membrane

surface at lower pH. The most interesting result was obtainedwhen the pH of the feed solution increased from 8.5 to 10.5 duringfiltration. As can be observed both silica and iron peaks shortenedconsiderably which indicates re-dissolving of these materials athigher feed pH. This observation suggests that fouling is partiallyreversible by increasing the pH.

Taking a closer look to Fig. 11, it was found that deposited fou-lants on the membrane surface were in three forms: beads, smallcrystals and large crystals. These three types of deposited solutesand their relevant EDX analysis are shown in Fig. 12. Accordingto this figure it can be concluded that silica/iron complexes arepresent in all forms and are homogenously distributed on themembrane surface. Needle-like crystals are believed to be mineralaegerine crystals (a sodium iron silicate complex has the chemicalformula NaFeSi2O6 in which the iron is present as Fe3+). Beads areassumed to be mainly silica particles capped with organic matterand inorganic precipitates (mainly metal oxides like Fe2O3). EDXresults of beads, small crystals and large crystals suggest thatsulfur and oxygen also exist in the foulant material.

Table 7XPS surface elemental analysis (atomic%).

Membrane C O N C:O:N ratio S Na Ca Fe Si Cl

Original membrane 74.25 12.97 11.78 6.30:1.10:1.00 0.36 0.29 0.00 0.00 0.00 0.35NF at pH = 8.5 70.09 20.36 2.83 24.76:7.19:1.00 1.22 2.13 0.53 1.18 1.96 0.10NF at pH = 10.5 68.27 18.57 7.76 8.79:2.39:1.00 0.30 1.97 0.49 0.80 1.62 0.22NF at pH = 8.5–10.5 65.09 23.20 4.26 15.27:5.44:1.00 0.88 1.26 0.51 1.70 1.87 0.13

Fig. 14. High resolution XPS spectra of NF90 membrane before and after filtration.

350 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

3.4.3. XPS resultsXPS spectra for original and used membranes are shown in

Fig. 13. The original membrane shows peaks corresponding to

elements present in the PA thin film: C 1s, N 1s and O 1s. Filtrationof model BFW led to a change in the peak heights for C, N and O.The C and O peaks increased in height while the peak for N

Fig. 15. Detailed XPS deconvoluted C 1s, O 1s and N 1s scans of NF90 membrane before (a, b, c) and after filtration at pH = 8.5 (d, e, f).

M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353 351

decreased. Inorganic deposition (Fe, Si, Ca) was also observed onthe used membranes. Further peaks corresponding to Na and Oappeared as NaKLL and OKLL, representing Auger emission.

Table 7 gives elemental compositions for the original and usedmembranes. The NF90 membrane is a fully aromatic PA compositemembrane [66]. The theoretical O/N ratio is expected to be 1.0when the PA layer is fully cross-linked. Table 7 indicates that theapplied NF90 membrane in this study has an O/N ratio of 1.1 whichis similar to that of a pure PA. A slight excess of oxygen withrespect to the corresponding theoretical ratio for PA indicates alow content of impurities on the surface of this membrane. Thesurface of the NF90 membrane contains S, Na and Cl (<0.5%) inaddition to the elements characteristic of the PA. This indicatesthat solvents and washing agents were not completely eliminatedduring the process of manufacturing the membrane [67].

Filtration led to a change in the C:O:N ratio. The amount of Cincreased significantly, indicating adsorption of organic foulants.Table 7 indicates that the greatest increases in the ratios of C/Nand O/N occurred for filtration at pH = 8.5. Higher amounts of inor-ganic matter were deposited on the membrane surface at pH = 8.5.

High resolution XPS spectra give more information about theorganic and inorganic components and bonds on the membranesurface. Based on Fig. 14, it can be concluded that:

1. Na on the original membrane is NaCl remained after washing bymanufacturer. However Na added after filtration is possiblyrelated to either remnants of the feed water that have evapo-rated and left NaCl behind or Na2O precipitates.

2. Ca, Fe and Si on the membrane surface correspond to oxides ofthese elements.

3. The position of the binding energy for SiO2 is 101 eV. This bind-ing energy is around 103 eV for pure SiO2. The binding energyobtained in the present work could be related to a sodium ironsilicate complex, e.g. NaFeSi2O6 called mineral aegirine.

4. Sulfate on the original membrane is related to sulfuric acidwhich is usually used for post-treatment of membranes. Kulkar-ni et al. [68] indicated that sulfuric acid may be used tohydrophilize PA thin film composite membranes. By carefullycontrolling the hydrophilization conditions, the permeate fluxmay be increased and fouling decreased without loss in therejection behavior of the membrane. Used membranes showedsulfide spectra instead of sulfate which may correspond tometal and salt sulfides like FeS and Na2S.

The multi-region spectra of C 1s, O 1s, and N 1s for both mem-branes are shown in Fig. 15. According to this figure, all spectrarelated to membrane chemical bonds were weakened. Aromatic/aliphatic bonds in the used membrane increased as compared tothe original membrane (Fig. 15a and d). This may show the higharomacity of the organics in the model BFW. Carboxylic bonds alsointensified in used membranes (Fig. 15b and e). A small peakrelated to Si–OH was observed in O 1s spectrum. Fig. 15c and fshows the presence of amino acids in the deposited matter onthe membrane surface.

4. Conclusion

Model SAGD BFW was treated with a crossflow NF process.First, a tight NF membrane was selected in preference to a looseNF and a RO membrane based on its superior rejection properties

352 M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353

and energy efficiency. Then, experiments were conducted on thismembrane at constant feed temperature (50 �C) and two pH values(10.5 and 8.5). The NF membrane rejected up to 98% of salt, DOMand silica, and >99% of divalent ions. TOC rejection was �98%regardless of the pH. At higher pH (10.5), the initial TDS rejectionwas about 80% and increased up to 96% after a 6 h run. However,at lower pH (8.5), TDS rejection was steady at 98% during the entireexperiment. The higher TDS rejection at pH 8.5 was attributed tothe formation of a cake-layer formed from the co-precipitation ofsilica and organic matter on the membrane. A step-change in pHfrom 8.5 to 10.5 increased the water flux by almost 20% immedi-ately. Surface characterization of the fouled membranes usingATR-FTIR, FESEM-EDX and XPS analyses provided valuable infor-mation about the constituents in the model BFW, which weredeposited on the membrane. Energy dispersive X-ray (EDX) indi-cated the presence of silica and iron in fouling deposits. X-ray pho-toelectron spectroscopy (XPS) showed the presence of inorganics(mainly metal oxides) and organics (having carboxylic and aminoacid functional groups) on the membrane surface. Fluorescenceexcitation emission matrix spectroscopy (FEEMs) showed thathydrophilic compounds made up the major fraction of the organicsthat passed through the membrane.

This study shows the feasibility of performing NF at a high pH asa polishing step in the conventional SAGD water treatment processtrain for the production of a higher-quality BFW. The results can beinterpreted to provide two possible process configurations forSAGD produced water treatment and BBD recycle. First, the NF pro-cess can be seen as an alternative to the current WLS-WAC processconfiguration, completely replacing the conventional treatmentprocess, and providing reliable removal of TDS, silica, divalent ionsand TOC from the produced water. Although more studies areneeded to clearly indicate whether such a process will be econom-ically viable, and what is the maximum recovery achievable fromsuch a plant, the process seems to be technically feasible. Second,the BBD and makeup water can be combined and treated througha separate NF system, yielding a stream containing low silica, TDS,DOM, and divalent ions that can be mixed with a conventionally(WLS-WAC) treated BFW. This prevents the risks associated withmixing the BBD and makeup water with the entire produced waterstream before the WLS in the current conventional process config-uration, but rather conditioning the recycle and makeup streamsfrom the produced water in a segregated manner.

Acknowledgement

Financial support for this work through the Natural Sciencesand Engineering Research Council of Canada (NSERC) IndustrialResearch Chair in Water Quality Management for Oil Sands Extrac-tion is gratefully acknowledged. The authors would like to thankDr. Fleck (Professor, Chair of Mechanical Engineering Department,University of Alberta) for kindly supporting water treatment pro-jects and his helpful suggestions.

References

[1] W.F. Heins, Technical advancements in SAGD evaporative produced watertreatment, J. Can. Pet. Technol. 48 (2009) 27–32.

[2] D.W. Jennings, A. Shaikh, Heat-exchanger deposition in an inverted steam-assisted gravity drainage operation. Part 1. Inorganic and organic analyses ofdeposit samples, Energy Fuels 21 (2007) 176–184.

[3] S. Wang, E. Axcell, R. Bosch, V. Little, Effects of chemical application onantifouling in steam-assisted gravity drainage operations, Energy Fuels 19(2005) 1425–1429.

[4] S. Guha Thakurta, A. Maiti, D.J. Pernitsky, S. Bhattacharjee, Dissolved organicmatter in steam assisted gravity drainage boiler blow-down water, EnergyFuels 27 (2013) 3883–3890.

[5] A. Zaidi, K. Simms, S. Kak, The use of micro/ultrafiltration for the removal of oiland suspended solids from oilfield brines, Water Sci. Technol. 25 (1992) 163–176.

[6] B.A. Farnand, T.A. Krug, Oil removal from oilfield produced water by cross flowultrafiltration, J. Can. Pet. Technol. 28 (2000) 18–24.

[7] I.W. Cumming, R.G. Holdich, I.D. Smith, The rejection of oil using anasymmetric metal microfilter to separate an oil in water dispersion, WaterRes. 33 (1999) 3587–3594.

[8] E. Gorouhi, M. Sadrzadeh, T. Mohammadi, Microfiltration of oily wastewaterusing PP hydrophobic membrane, Desalination 200 (2006) 319–321.

[9] R.S. Faibish, Y. Cohen, Fouling and rejection behavior of ceramic and polymer-modified ceramic membranes for ultrafiltration of oil-in-water emulsions andmicroemulsions, Colloid Surf. A – Physicochem. Eng. Asp. 191 (2001) 27–40.

[10] T. Bilstad, E. Espedal, Membrane separation of produced water, Water Res. 34(1996) 239–246.

[11] E.-S.S. Kim, Y. Liu, M. Gamal El-Din, The effects of pretreatment onnanofiltration and reverse osmosis membrane filtration for desalination ofoil sands process-affected water, Sep. Purif. Technol. 81 (2011) 418–428.

[12] H. Peng, K. Volchek, M. MacKinnon, W.P. Wong, C.E. Brown, Application ofnanofiltration to water management options for oil sands operation,Desalination 170 (2004) 137–150.

[13] A.K. Mehrotra, A. Banerjee, Evaluation of reverse osmosis for the treatment ofoil sands produced water, Water Pollut. Res. J. Canada 21 (1986) 141–152.

[14] Q. Li, M. Elimelech, Organic fouling and chemical cleaning of nanofiltrationmembranes: measurements and mechanisms, Environ. Sci. Technol. 38 (2004)4683–4693.

[15] C. Bellona, M. Marts, J.E. Drewes, The effect of organic membrane fouling onthe properties and rejection characteristics of nanofiltration membranes, Sep.Purif. Technol. 74 (2010) 44–54.

[16] B.A. Braghetta, F.A. Digiano, W.P. Ball, NOM accumulation at NF membranesurface. Impact of chemistry and shear, J. Environ. Eng. 124 (1998) 1087–1098.

[17] C. Jucker, M.M. Clark, Adsorption of aquatic humic substances on hydrophobicultrafiltration membranes, J. Memb. Sci. 97 (1994) 37–52.

[18] S. Hong, M. Elimelech, Chemical and physical aspects of natural organic matter(NOM) fouling of nanofiltration membranes, J. Memb. Sci. 132 (1997) 159–181.

[19] J. Masliyah, Z.J. Zhou, Z. Xu, J. Czarnecki, H. Hamza, Understanding water-based bitumen extraction from Athabasca oil sands, Can. J. Chem. Eng. 82(2004) 628–654.

[20] L.A. Pérez-Estrada, X. Han, P. Drzewicz, M. Gamal El-Din, P.M. Fedorak, J.W.Martin, Structure-reactivity of naphthenic acids in the ozonation process,Environ. Sci. Technol. 45 (2011) 7431–7437.

[21] M. Gamal El-Din, H. Fu, N. Wang, P. Chelme-Ayala, L. Pérez-Estrada, P.Drzewicz, et al., Naphthenic acids speciation and removal during petroleum-coke adsorption and ozonation of oil sands process-affected water, Sci. TotalEnviron. 409 (2011) 5119–5125.

[22] A.C. Scott, W. Zubot, M.D. MacKinnon, D.W. Smith, P.M. Fedorak, Ozonation ofoil sands process water removes naphthenic acids and toxicity, Chemosphere71 (2008) 156–160.

[23] M. Kabsch-Korbutowicz, K. Majewska-Nowak, T. Winnicki, Analysis ofmembrane fouling in the treatment of water solutions containing humicacids and mineral salts, Desalination 126 (1999) 179–185.

[24] W. Yuan, A.L. Zydney, Effects of solution environment on humic acid foulingduring microfiltration, Desalination 122 (1999) 63–76.

[25] Z. Wang, Y. Zhao, J. Wang, S. Wang, Studies on nanofiltration membranefouling in the treatment of water solutions containing humic acids,Desalination 178 (2005) 171–178.

[26] B.A. Braghetta, F.A. Digiano, W.P. Ball, Nanofiltration of natural organic matter:pH and ionic strength effects, J. Environ. Eng. 123 (1997) 628–641.

[27] K.L. Jones, C.R. O’Melia, Protein and humic acid adsorption onto hydrophilicmembrane surfaces: effects of pH and ionic strength, J. Memb. Sci. 165 (2000)31–46.

[28] C.A. Dyke, C.R. Bartels, Removal of organics from offshore produced watersusing nanofiltration membrane technology, Environ. Prog. 9 (1990) 183–186.

[29] F.T. Tao, S. Curtice, R.D. Hobbs, J.L. Sides, J.D. Wieser, C.A. Dyke, et al., Reverseosmosis process successfully converts oil field brine into freshwater, Oil Gas J.38 (1993) 88–91.

[30] G.F. Doran, K.L. Williams, J.A. Drago, S.S. Huang, L.Y.C. Leong, Pilot study resultsto convert oil field produced water to drinking water or reuse quality, in: SPEAnnual Tech. Conf. Exhib., Society of Petroleum Engineers Inc, New Orleans,1998, pp. 403–417.

[31] M. Cakmakce, N. Kayaalp, I. Koyuncu, Desalination of produced water from oilproduction fields by membrane processes, Desalination 222 (2008) 176–186.

[32] C. Murray-Gulde, J.E. Heatley, T. Karanfil, J.H. Rodgers, J.E. Myers, Performanceof a hybrid reverse osmosis-constructed wetland treatment system forbrackish oil field produced water, Water Res. 37 (2003) 705–713.

[33] S. Mondal, S.R. Wickramasinghe, Produced water treatment by nanofiltrationand reverse osmosis membranes, J. Memb. Sci. 322 (2008) 162–170.

[34] P. Xu, J.E. Drewes, D. Heil, Beneficial use of co-produced water throughmembrane treatment: technical-economic assessment, Desalination 225(2008) 139–155.

[35] A. Maiti, M. Sadrezadeh, S. Guha Thakurta, D.J. Pernitsky, S. Bhattacharjee,Characterization of boiler blowdown water from steam-assisted gravitydrainage and silica–organic coprecipitation during acidification andultrafiltration, Energy Fuels 26 (2012) 5604–5612.

[36] Q. Li, M. Elimelech, Natural organic matter fouling and chemical cleaning ofnanofiltration membranes, Water Sci. Technol. Water Supply. 4 (2004) 245–251.

[37] T. Rizwan, S. Bhattacharjee, Initial deposition of colloidal particles on a roughnanofiltration membrane, Can. J. Chem. Eng. 85 (2007) 570–579.

M. Sadrzadeh et al. / Separation and Purification Technology 141 (2015) 339–353 353

[38] C. Bellona, J.E. Drewes, The role of membrane surface charge and solutephysico-chemical properties in the rejection of organic acids by NFmembranes, J. Memb. Sci. 249 (2005) 227–234.

[39] M.A.-A. Mamun, Colloidal Fouling of Salt Rejecting Nanofiltration Membranes:Transient Electrokinetic Model and Experimental Study, University of Alberta,2012.

[40] N.C. Mueller, B. van der Bruggen, V. Keuter, P. Luis, T. Melin, W. Pronk, et al.,Nanofiltration and nanostructured membranes – should they be considerednanotechnology or not?, J Hazard. Mater. 211–212 (2012) 275–280.

[41] L.D. Nghiem, S. Hawkes, Effects of membrane fouling on the nanofiltration ofpharmaceutically active compounds (PhACs): mechanisms and role ofmembrane pore size, Sep. Purif. Technol. 57 (2007) 176–184.

[42] L.D. Nghiem, P.J. Coleman, NF/RO filtration of the hydrophobic ionogeniccompound triclosan: transport mechanisms and the influence of membranefouling, Sep. Purif. Technol. 62 (2008) 709–716.

[43] K.L. Tu, A.R. Chivas, L.D. Nghiem, Effects of membrane fouling and scaling onboron rejection by nanofiltration and reverse osmosis membranes,Desalination 279 (2011) 269–277.

[44] A.E. Childress, M. Elimelech, Effect of solution chemistry on the surface chargeof polymeric reverse osmosis and nanofiltration membranes, J. Memb. Sci. 119(1996) 253–268.

[45] A. Seidel, M. Elimelech, Coupling between chemical and physical interactionsin natural organic matter (NOM) fouling of nanofiltration membranes:implications for fouling control, J. Memb. Sci. 203 (2002) 245–255.

[46] X. Zhu, M. Elimelech, Colloidal fouling of reverse osmosis membranes:measurements and fouling mechanisms, Environ. Sci. Technol. 31 (1997)3654–3662.

[47] L. Song, M. Elimelech, Particle deposition onto a permeable surface in laminarflow, J. Colloid Interface Sci. 173 (1995) 165–180.

[48] L.D. Nghiem, S. Hawkes, Effects of membrane fouling on the nanofiltration oftrace organic contaminants, Desalination 236 (2009) 273–281.

[49] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Fouling of reverse osmosis and nanofiltrationmembranes by humic acid – effects of solution composition andhydrodynamic conditions, J. Memb. Sci. 290 (2007) 86–94.

[50] E.M. Vrijenhoek, S. Hong, M. Elimelech, Influence of membrane surfaceproperties on initial rate of colloidal fouling of reverse osmosis andnanofiltration membranes, J. Memb. Sci. 188 (2001) 115–128.

[51] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Characterization of humic acid fouledreverse osmosis and nanofiltration membranes by transmission electronmicroscopy and streaming potential measurements, Environ. Sci. Technol. 41(2007) 942–949.

[52] Y.-N. Wang, C.Y. Tang, Nanofiltration membrane fouling by oppositely chargedmacromolecules: investigation on flux behavior, foulant mass deposition, andsolute rejection, Environ. Sci. Technol. 45 (2011) 8941–8947.

[53] M.J. López-Muñoz, A. Sotto, J.M. Arsuaga, B. Van der Bruggen, Influence ofmembrane, solute and solution properties on the retention of phenolic

compounds in aqueous solution by nanofiltration membranes, Sep. Purif.Technol. 66 (2009) 194–201.

[54] H.M. Krieg, S.J. Modise, K. Keizer, H.W.J.P. Neomagus, Salt rejection innanofiltration for single and binary salt mixtures in view of sulphateremoval, Desalination 171 (2005) 205–215.

[55] P. Xu, J.E. Drewes, T.-U. Kim, C. Bellona, G. Amy, Effect of membrane fouling ontransport of organic contaminants in NF/RO membrane applications, J. Memb.Sci. 279 (2006) 165–175.

[56] D. Violleau, H. Essis-tome, H. Habarou, J.P. Croud, M. Pontid, Fouling studies ofa polyamide nanofiltration membrane by selected natural organic matter: ananalytical approach, Desalination 173 (2005) 223–238.

[57] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic matter:comparison of flux decline, NOM rejection, and foulants during filtration withthree UF membranes, Desalination 127 (2000) 283–298.

[58] W.E.I. Yuan, A.L. Zydney, Humic acid fouling during ultrafiltration, Environ. Sci.Technol. 34 (2000) 5043–5050.

[59] A.R.S. Teixeira, J.L.C. Santos, J.G. Crespo, Sustainable membrane-based processfor valorisation of cork boiling wastewaters, Sep. Purif. Technol. 66 (2009) 35–44.

[60] J.L. Weishaar, G.R. Aiken, B.a. Bergamaschi, M.S. Fram, R. Fujii, K. Mopper,Evaluation of specific ultraviolet absorbance as an indicator of the chemicalcomposition and reactivity of dissolved organic carbon, Environ. Sci. Technol.37 (2003) 4702–4708.

[61] H. Yamamura, K. Okimoto, K. Kimura, Y. Watanabe, Influence of calcium on theevolution of irreversible fouling in microfiltration/ultrafiltration membranes, J.Water Supply Res. Technol. 56 (2007) 425–434.

[62] S. Lee, J. Cho, M. Elimelech, Combined influence of natural organic matter(NOM) and colloidal particles on nanofiltration membrane fouling, J. Memb.Sci. 262 (2005) 27–41.

[63] D.B. Burns, A.L. Zydney, Buffer effects on the zeta potential of ultrafiltrationmembranes, J. Memb. Sci. 172 (2000) 39–48.

[64] 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.

[65] H. Jacobasch, J. Schurz, Characterization of polymer surfaces by means ofelectrokinetic measurements, Prog. Colloid Polym. Sci. 77 (1988) 40–48.

[66] C. Tang, Y. Kwon, J. Leckie, Probing the nano- and micro-scales of reverseosmosis membranes – a comprehensive characterization of physiochemicalproperties of uncoated and coated membranes by XPS, TEM, ATR-FTIR, andstreaming potential measurements, J. Memb. Sci. 287 (2007) 146–156.

[67] M.J. Ariza, J. Benavente, E. Rodríguez-Castellón, L. Palacio, Effect of hydration ofpolyamide membranes on the surface electrokinetic parameters: surfacecharacterization by x-ray photoelectronic spectroscopy and atomic forcemicroscopy, J. Colloid Interface Sci. 247 (2002) 149–158.

[68] A. Kulkarni, D. Mukherjee, W.N. Gill, Flux enhancement by hydrophilization ofthin film composite reverse osmosis membranes, J. Memb. Sci. 114 (1996) 39–50.