Rejection of nine haloacetic acids and coupled reverse ... · Rejection of nine haloacetic acids and coupled reverse draw solute permeation in forward osmosis Fan-xin Kong a,Hong-weiYanga,

Desalination 341 (2014) 1–9

Contents lists available at ScienceDirect

Desalination

j ourna l homepage: www.e lsev ie r .com/ locate /desa l

Rejection of nine haloacetic acids and coupled reverse draw solutepermeation in forward osmosis

Fan-xin Kong a, Hong-wei Yang a, Xiao-mao Wang a,⁎, Yuefeng F. Xie a,b

a State Key Joint Laboratory of Environmental Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing 100084, Chinab Environmental Engineering Programs, The Pennsylvania State University, Capital College, PA 17057-4898, USA

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Forward osmosis rejects haloaceticacids (HAAs) well.

• The active-layer facing feed water(AL-FW) orientation is preferred.

• The solution-diffusion model predictsHAA rejection better for AL-FW.

• The solution-diffusion model accuratelypredicts reverse draw solutionpermeation.

⁎ Corresponding author. Tel.: +86 10 6278 1386.E-mail address: [email protected] (X. W

http://dx.doi.org/10.1016/j.desal.2014.02.0190011-9164/© 2014 Elsevier B.V. All rights reserved.

The behaviors of haloacetic acid (HAA) rejection by forward osmosis are obtained by experiments and predictedby the solution-diffusion model combined with concentration polarization for both the AL-FW (active layer fac-ing the feed water) and the AL-DS (active layer facing the draw solution) orientations.

a b s t r a c t

Article history:Received 22 September 2013Received in revised form 15 February 2014Accepted 17 February 2014Available online 13 March 2014

Keywords:Haloacetic acids (HAAs)Forward osmosisConcentration polarizationSolution-diffusion modelReverse draw solute permeation

The rejection of haloacetic acids (HAAs) by forward osmosis (FO) and the coupled reverse draw solute perme-ation were experimentally determined and mathematically modeled by using the solution-diffusion model forboth the AL-FW (active layer facing the feed water) and the AL-DS (active layer facing the draw solution) orien-tations. The rejection ratio for each HAA increased with the increase of draw solute concentration for the AL-FWorientation. In contrast, the HAA rejection ratio could reach its maximum under a medium osmotic pressure dif-ference for the AL-DS orientation. The rejection ratios for all HAAs were higher than 94.6% for the AL-FW orien-tation and ranged from 73.8% to 89.1% for the AL-DS orientation under a draw solute concentration of 1 mol/LNaCl. The reverse draw solute flux for the AL-FW orientation was lower than that for the AL-DS orientation.The model-predicted HAA rejection results matched well with the experimental rejection ratios for the AL-FWorientation. However, the model over-estimated the rejection ratios for the AL-DS mode, likely due to the adop-tion of inaccurate mass transfer coefficient for internal concentration polarization. Regarding the reverse drawsolute permeation, a general agreement between the model prediction and experimental data was observedfor both orientations.

© 2014 Elsevier B.V. All rights reserved.

ang).

1. Introduction

Forward osmosis (FO) is a process driven by the difference in chem-ical potentials across a semi-permeable membrane with little hydraulicpressure applied. In recent years, the potential application of FO to

Fig. 1. Schematic diagrams of the bench-scale forward osmosis system (top) and the dia-phragm cell test unit (bottom).

2 F. Kong et al. / Desalination 341 (2014) 1–9

wastewater reclamation has been extensively studied [1–3], which wasnormally fulfilled by combining FO with other technologies as a hybridprocess [4]. The hybrid processes may compete favorably with the sim-ple pressure-driven processes such as reverse osmosis (RO) owing tothe fact that the former may have a lower demand on electrical energyand its membranes may suffer less membrane fouling [2,5–7]. One ofthe biggest concerns about wastewater reclamation is the removalefficiencies for the large number of trace organic compounds at con-centrations in the magnitudes of ng/L to μg/L. Common trace organiccompounds include endocrine disruptors (EDCs), pharmaceutical andpersonal care products (PPCPs) and disinfection by-products (DBPs)[8–11]. The reclaimed water is better to be free of these trace organiccompounds due to their potentially high health risks to humans andaquatic organisms.

Previous studies [2,12–17] showed that the rejection of PPCPs andEDCs by FO process varied among the compounds and the operationalconditions. Results by Hancock revealed that rejection efficiencies byFO depended primarily on molecular size and charge. Rejection of pos-itively and negatively charged trace organic compounds was greaterthan 80%, while rejection of the nonionic trace organic compounds var-ied, between approximately 40% and 90% [18]. Xie et al. [13] foundthat the membrane orientation (AL-FW (active layer facing feedwater) andAL-DS (active layer facingdrawsolution)) affected the rejec-tion performance.

Development of mathematical models capable of predicting the re-jection of the trace organic compounds by the FO process will improvethe economic efficiency of the process and expand its applications. Byfar, there has been only a few work conducted on the modeling of therejection performance by FO, among which the model based on the so-lution-diffusion model developed by Jin et al. [19] was found able to ac-curately predict the FO rejection of inorganic solutes. It is also desirableto model the permeate water flux and the reverse draw solute fluxthrough the membrane. The limited water flux and the substantial re-verse draw soluteflux are among the biggest impediments to the viabil-ity of the FO processes [20,21]. In this regard, a few studies investigatedthe reverse draw solute permeation in the FO process and proposeda relationship with the water flux [22,23]. Better understandingand prediction of feed contaminant and coupled reverse draw solutepermeation in FO membrane processes can facilitate to optimize theoperational conditions and benefit to fabricate FOmembranes thatmin-imize the loss of drawsolute into the feed solution, thereby reducing op-erating costs.

Haloacetic acids (HAAs) are among the DBPs with the highest con-centrations in chlorinated or chloraminated sewage treatment plant ef-fluent. The trichloroacetic acid (TCAA) concentration can be as high as471 μg/L in the chlorinated wastewater effluent [24]. This study mainlyinvestigated the rejection of nine HAAs and the reverse draw solute per-meation in the FO process for both the AL-FW and the AL-DS orienta-tions. Substantial attention was paid to the mathematical descriptionof the water, HAA and draw solute fluxes. This study also evaluatedthe feasibility of the solution-diffusionmodel in predicting the rejectionof organic contaminants and coupled draw solute permeation in the FOprocess.

2. Materials and methods

2.1. The FO system and operations

The schematical diagram of the bench-scale FO system is shown inFig. 1. The membrane unit was custom built with two flow channels of2mm in height both,whichwere separated by an FOmembrane couponthat had an effective area of 40.5 cm2. The commercial FO membranewas provided byHydration Technologies, Inc. (Albany, OR) and, accord-ing to the manufacturer, was made of cellulose triacetate supported byembedded polyester screen mesh. The recirculation flows of the feedwater (FW) and the draw solution (DS) on the two sides of the

membrane were counter-current with a flow velocity of 20.4 cm/sboth, each regulated by a variable-speed gear pump (Longer, USA).The relatively high cross-flow velocity can substantially suppress theexternal concentration polarization (ECP) that will be described in de-tail later.

Ultrapure water (Milli-Q, Millipore, USA) was used for the FW andthe DS preparation. The FW was prepared by dissolving a mixture ofthe nine haloacetic acids (HAAs) (Sigma-Aldrich, Germany) into a10 mmol/L NaCl (as background solute) solution with the final concen-tration for each HAA at 200 μg/L. The FW ionic strength was similar tothe typical values for freshwater including sewage treatment plant efflu-ent, although our preliminary study showed that the FW ionic strength(0–20 mmol/L) had negligible influence on the rejection of HAAs. Thenine HAAs were monochloroacetic acid (MCAA), monobromoaceticacid (MBAA), dichloroacetic acid (DCAA), bromochloroacetic acid(BCAA), dibromoacetic acid (DBAA), trichloroacetic acid (TCAA),bromodichloroacetic acid (BDCAA), dibromochloroacetic acid (CDBAA)and tribromoacetic acid (TBAA). Physicochemical properties for eachHAA including molecular weight (MW) and dissociation constant(pKa) were summarized in Table 1. Sodium bicarbonate was addedinto the FWto adjust the pH to neutral (pH=7). Nonatural organicmat-ter (NOM) or soluble microbial products (SMPs) were contained in theFW, though both NOM and SMP could not only adsorb certain trace or-ganic compounds but also lead to membrane fouling, which wouldgreatly affect the rejection performance of the FO process [12]. The DSwas a NaCl solution at a concentration varying from 0.1 to 3 mol/L.

The FO experiments were conducted in an air-conditioned roomwith the temperature set at 25 ± 1 °C. Each FO test lasted for about4 h. The duration was long enough to obtain accurate water, HAA andreverse draw solute permeation fluxes. On the conclusion of each FOtest, the volume variation of the DS (or the FW) was less than 5%, andas such the draw solute (in the DS) and the HAA (in the FW) concentra-tion variationwere negligible. The FW tankwas placed on a digital scale(Mettler Toledo, Germany) connected to a computer data logging

Table 1Physicochemical properties of the nine haloacetic acids.

Haloacetic acids (HAAs) Structure MWa (g/mol) pKaa DBb (m2/s)

Chloroacetic acid (MCAA) 93.98 2.65 1.20E−09

Bromoacetic acid (MBAA) 138.95 2.73 1.16E−09

Dibromoacetic acid (DCAA) 128.94 1.37 1.07E−09

Bromochloroacetic acid (BCAA) 173.39 1.39 1.04E−10

Dibromoacetic acid (DBAA) 217.84 1.47 1.00E−09

Trichloroacetic acid (TCAA) 163.39 0.093 9.68E−10

Bromodichloroacetic acid (BDCAA) 207.83 0.051 9.42E−10

Chlorodibromoacetic acid (CDBAA) 252.29 0.13 9.18E−10

Tribromoacetic acid (TBAA) 296.74 0.22 8.96E−10

a Obtained by using the ACD Lab software.b Calculated using the Wilke–Chang equation.

3F. Kong et al. / Desalination 341 (2014) 1–9

system. The weight change as a function of time was utilized to deter-mine the permeate water flux,

Jw ¼ ΔmΔtAm

ð1Þ

where Δm/Δt is the weight change rate and Am is the effective area ofthe membrane coupon.

At specified time intervals, samples were taken from either the FWtank or the DS tank for HAA concentration, chloride ion and pH valuedetermination. The variation of the HAA concentrations in the DS wasused to determine the HAA flux through the FO membrane, while thevariation of the NaCl concentration in the FWwas used for the determi-nation of the reverse solute flux through the FO membrane,

Jw ¼ Δ Vcð ÞΔtAm

ð2Þ

where Δ(Vc) is the variation of target compound in the FW or DS tanksduring the specified time interval Δt.

Given the permeatewater flux (Jw) and theHAAfluxes (JHAA), the re-jection ratio of each HAA compound (RHAA) in the FO process can be ob-tained by

RHAA ¼ 1− JHAAJwcF;HAA

ð3Þ

where cF,HAA is the HAA concentrations in the FW, which did not changesignificantly in each FO test.

2.2. Mathematical expressions for the water, HAA and draw solute fluxes

The mathematical expressions for the water, HAA and draw solutefluxes can be referred to the model developed by Elimelech, Tang andtheir colleagues [25–29]. Most of the previous models made use of thesolution-diffusion model coupled with concentration polarization. Forthe AL-FW orientation, the draw solute experiences dilutive internalconcentration polarization (ICP) on the DS side and concentrative ECPon the FW side. The permeate water flux Jw is expressed by

Jw ¼ A πD;NaCl exp − Jw=Km;NaCl

� �−π F;NaCl exp Jw=Km;NaCl

� �� �: ð4Þ

The reverse draw solute flux JNaCl can be expressed as

JNaCl ¼ BNaCl CD;NaCl exp − Jw=Km;NaCl

� �−C F;NaCl exp Jw=Km;NaCl

� �� �: ð5Þ

The HAA flux through the FO membrane can be expressed as

JHAA ¼BHAA exp Jw=Km;HAA

� �1þ BHAA= Jw

cF;HAA ð6Þ

where A and B are the water permeability coefficient and the solute(draw solute or HAA) permeability coefficient, respectively, and Km

and km are the mass transfer coefficients for ICP and ECP, respectively.Km is also called the solute resistivity for diffusion within the porouslayer, and is related with the S value by Km = D/S. The S parameter isthe intrinsic length scale of ICP in the FO membrane support layer,which is analogous to the boundary layer thickness for ECP. Mathemat-ically, S is related to the thickness (t), tortuosity (τ) and porosity (ε) ofthe support layer by S= tτ / ε. The S value is usually obtained by fittingthe water flux vs. draw solute osmotic pressure data under known hy-draulic conditions (Eq. (4) or Eq. (7) below). (Care however must betaken in that it is not a direct method and the inaccuracy of a numberof parameters would lead to the inaccurate determination of the Svalue.) Given the S value, the Km value for any solute can be easily ob-tained only if the diffusivity (D) is known. Km is hardly affected by thehydrodynamic conditions. In contrast, the km value is controlled by thehydrodynamics and can be calculated by km = ShD / dh, where Sh anddh are the Sherwood number and the hydraulic diameter of the FO cellchannel, respectively.

For the AL-DS orientation, the draw solute experiences dilutive ECPon the DS side and concentrative ICP on the FW side. The permeatewater flux can be expressed as

Jw ¼ A πD;NaCl exp − Jw=Km;NaCl

� �−π F;NaCl exp Jw=Km;NaCl

� �� �: ð7Þ

The reverse draw solute flux can be expressed as

JNaCl ¼ BNaCl CD;NaCl exp − Jw=Km;NaCl

� �−C F;NaCl exp Jw=Km;NaCl

� �� �: ð8Þ

Table 2HAA permeability coefficients using diaphragm cell and the pressurized filtration testunits.

HAAs Diaphragm celltest (m/s)

Pressurized filtrationtest (m/s)

Average values(m/s)

MCAA 3.00E−07 2.05E−07 2.53E−07MBAA 2.18E−07 1.59E−07 1.89E−07DCAA 8.62E−08 6.13E−08 7.37E−08TCAA 4.34E−08 3.26E−08 3.80E−08BCAA 4.38E−08 3.30E−08 3.84E−08BDCAA 2.03E−08 1.78E−08 1.91E−08DBAA 2.50E−08 2.13E−08 2.32E−08CDBAA 1.73E−08 1.20E−08 1.46E−08TBAA NAa 1.49E−08 1.49E−08

a Not available due to substantial hydrolysis [8].

4 F. Kong et al. / Desalination 341 (2014) 1–9

The HAA flux from the FW to the DS can be expressed as

JHAA ¼BHAA exp Jw=Km;HAA

� �

1þBHAA exp Jw=Km;HAA

� �exp − Jw=Km;HAA

� �Jw

cF;HAA: ð9Þ

A rough calculation revealed that the Sherwood number and the kmvalue were approximately 150 and 4 × 10−5 m/s, the latter of whichwas much larger than the water permeation flux (Jw) (see below).Therefore, ECP was sufficiently suppressed so that it did not affect theflux of water, NaCl or HAA in this study (i.e. exp(Jw / km) ≈ 1).

2.3. Determination of the permeability coefficients

A pressurized cross-flow filtration test unit was used for the deter-mination of thewater, NaCl andHAApermeability coefficients pertinentto the FOmembrane [30]. In the filtration unit, the active layer of the FOmembrane faced the high pressure side (i.e. under RO testing mode) toeliminate the influence of ICP. The effective membrane area was 42 cm2

and the water temperature was maintained at 25 ± 1 °C.The water permeability coefficient (A) was derived from the water

flux versus the applied pressure (0–12 bar) obtained by filtering high-purity water (Milli-Q, Millipore, USA), i.e. Jw = A · ΔP. The NaCl perme-ability coefficient (BNaCl) was derived from the rejection ratio versus theapplied pressure (Eq. (10)) by filtering a 10 mmol/L NaCl solution.

R ¼ 1

1þ BA ΔP−Δπð Þ

ð10Þ

where Δπ is the osmotic pressure difference across the membrane. TheHAA permeability coefficients (BHAA) were also derived from the rejec-tion ratio versus the applied pressure (Eq. (10)) by filtering the mixedsolution of the nine HAAs with 10 mmol/L NaCl as background. It wasassumed that the rejection of each HAA was independent when thebackground salt concentration was much higher than that of any HAA.

The HAA permeability coefficients can also be determined by usingthe diaphragm cell test [31]. The cylindrical diffusion cell (Fig. 1) wasmade of glass and consisted of two compartments which were separat-ed by an FO membrane with an effective surface area of 20.4 cm2. Eachcompartment had a volume of 1 L. For the diaphragm cell test, one com-partment contained a solution of NaCl at 10 mmol/L, and the othercontained a solution of both NaCl at 10 mmol/L and the target HAAs at500 μg/L each. The pH of the HAA solution was adjusted to around 7by using aminimum volume of the NaHCO3 solution. The ionic strengthof the two compartments was almost the same (approximately10 mmol/L), thus avoiding water permeation driven by osmotic pres-sure difference. The solutions in both compartments were continuouslymagnetically stirred and thermostatically stabilized at a temperature of25 ± 1 °C. Samples from the two compartments were taken for HAAmeasurement every day that lasted for 18 days. The permeability coef-ficient can be calculated by using Eq. (11) [31].

B ¼ 1Amt 1=vSoð Þ þ 1=vReð Þð Þ1n

c0So−c0RectSo−ctRe

!ð11Þ

where c0So and c0Re are theHAA concentrations in the source and receiv-er of the diffusion process at t=0, ctSo and ctRe are the concentrations ofthe solute in the source and receiver at time t, Am is themembrane area,and vSo and vRe are the volumes of the source and receiver of the dia-phragm cell, respectively.

2.4. Analytical methods

The HAAs were analyzed by a gas chromatograph equipped with anelectron capture detector (Agilent 7890A, Santa Clara, USA) after liquid–

liquid extraction according to the U.S. EPA Standard Method 552.3 (U.S.EPA, 2003). The NaCl concentration was determined by monitoring thechloride ion concentration, which was detected by an ion chromato-graph (Metrohm, Switzerland).

3. Results and discussion

3.1. Permeability coefficients of the FO membrane

By conducting the pressurized cross-flow filtration tests, the A andBNaCl values were determined as 2.38 (±0.04) × 10−7 m/(s bar) and2.90 (±0.5) × 10−7 m/s, respectively, which are very close to that re-ported by others [19].

The determined permeability coefficients for each HAA by both thepressurized cross-flow filtration test and the diaphragm cell test wereshown in Table 2 (the determination of the TBAA permeability coeffi-cient was not attainable using the diaphragm cell test due to the sub-stantial hydrolysis of TBAA [8] during the test). The values obtainedfrom the diaphragm unit were slightly higher than that from the pres-surized filtration (Table 2). This might lie in the fact that the permeabil-ity coefficient of the solute is a concentration-dependent parameter.Studies showed that both the diffusion coefficient in the dense polymerand the partitioning coefficient at the solution–polymer (both of whichdetermine the permeability coefficient) are weakly dependent onthe solution concentration [32,33]. However, the reliance of the perme-ability coefficient on the solute concentration can usually be ignored[34,35].

3.2. Water permeation, HAA and reverse NaCl fluxes

An increase in water permeation fluxwas observedwith an increasein the DS NaCl concentration (i.e. apparent osmotic pressure differencebetween the DS and FW) for both the AL-FW and AL-DS orientations(Fig. 2). Under the same osmotic pressure difference, the water perme-ation flux for the AL-FW orientation was always lower than that for theAL-DS orientation. For example, when the DS NaCl concentration was1 mol/L, the water permeation fluxes for the AL-FW and AL-DS orienta-tionswere 12.8 and 20.6 L/(m2h), respectively.When the FOmembranewas operated in the AL-FW orientation, severe dilutive ICP of thedraw solute occurred which greatly reduced the effective osmoticpressure difference across the FO membrane. In contrast, mild concen-trative ICP was mainly responsible for the decrease of the effective os-motic pressure for the AL-DS orientation. Given the permeabilitycoefficients of water (A) and NaCl (BNaCl), the mass transfer coefficientof NaCl (Km,NaCl) can be obtained byfitting thewaterflux and theosmot-ic pressure data using either Eq. (4) (for the AL-FW orientation) orEq. (7) (for the AL-DS orientation). The Km,NaCl values obtained fromthe AL-FW orientation data and the AL-DS orientation data were almost

Fig. 3. HAA fluxes for the AL-FS orientation (top) and the AL-DS orientation (bottom).

Fig. 4. Experimental results of the NaCl flux as a function of DS NaCl concentration forboth the AL-DS and AL-FW orientations. The feed water contained 10 mmol/L NaCl and200 μg/L HAAs each.

0 0.5 1 1.5 2 2.5 30

5

10

15

20

25

30

35

40

NaCl Concentration (M)

Wat

er F

lux

(L/(

m2 h

)

AL-FWFittedAL-DSFitted

Fig. 2. Experimental results of the water flux and fitted results using either Eq. (4) orEq. (7) as a function of DS NaCl concentration for both the AL-DS and AL-FW orientations.The feed water contained 10 mmol/L NaCl and 200 μg/L HAAs each.

5F. Kong et al. / Desalination 341 (2014) 1–9

identical, which were 2.78 × 10−6 and 2.73 × 10−6 m/s respectively.These values matched well with those reported by Jin et al. [19](2.73 × 10−6 m/s), but somewhat lower than those reported by Tanget al. [27] (4.2 × 10−6 and 3.3 × 10−6 m/s for the AL-FW and AL-DSorientations, respectively). With both Km,NaCl (2.76 × 10−6 m/s) andthe diffusion coefficient of NaCl (DNaCl, 1.5 × 10−9 m2/s [36]) known,the S value (i.e. the intrinsic parameter of the FO membrane whichdetermines the ICPwithin the supporting layer bymembrane thickness,tortuosity and porosity) was calculated to be 5.43 × 10−4 m. Notehowever that the osmotic pressure for the draw solution might beunder-estimated due to use of the Morse equation (i.e. linear relation-ship of the osmotic pressurewith the concentration) for the calculation.It indicates that the Km,NaCl value might be somewhat over-estimatedand in turn the S value was somewhat under-estimated. However, theunder-estimation was estimated within 10%.

For both orientations, the HAA fluxes also increased with the in-crease of the DS NaCl concentration (Fig. 3). A general trend could befound that the smaller the molecular weight of the HAA species, thehigher the HAA flux through the FO membrane. In other words, theHAA flux mainly followed the declining order of MCAA, MBAA, DCAA,TCAA/BCAA, DBAA/BDCAA, CDBAA and TBAA. It was assumed that FOfollows the solution-diffusion model, and an HAA of smaller weightmight have a higher “diffusion” rate within the active layer of the FOmembrane. In addition, the stability (i.e. hydrolysis) and the waterbonding potential of HAAs would probably also influence the perme-ation of HAAs through the FO membrane [37,38].

The main difference between the two orientations is that under thesame osmotic difference, the HAA flux for the AL-FW orientation wasmuch lower than that for the AL-DS orientation (Fig. 3). Severe concen-trative ICP of the HAA species occurred for the AL-DS orientation, whichgreatly increased the HAA concentration at the active layer–supportinglayer interface. In contrast, no ICP of theHAA specieswould occur for theAL-FW orientation. Another difference between the two orientations isthat the HAA fluxes increased acceleratedly for the AL-DS orientationwhile it increased deceleratedly for the AL-FW orientation when theDS NaCl concentration was increased. The higher water flux for theAL-DS orientation than that for the AL-FW orientation and the criticalrole of ICP of HAA are responsible for the difference. The extent ofHAA concentration polarization is exponentially dependent on thewater permeation flux which results from the severe concentrativeICP caused by high water flux for the AL-DS orientation [39].

For reverse draw solute permeation, as the DS NaCl concentrationincreased, themeasured reverse draw solute flux increased for both ori-entations (Fig. 4). However, the reverse draw solute flux for the AL-FW

6 F. Kong et al. / Desalination 341 (2014) 1–9

orientation is much lower than that for the AL-DS orientation. For in-stance, when the DS NaCl concentration was 1 mol/L, the reversedraw solute fluxes were 0.21 and 0.55 mol/(m2h) for the AL-FW andAL-DS orientations, respectively (Fig. 4).

3.3. HAA rejection ratios

With both the water permeation flux and the HAA fluxes obtained,the rejection ratios of the various HAAs by FO could be calculatedusing Eq. (3). The results demonstrated that the rejection ratio foreach HAA showed a monotonous increase with the increase of DSNaCl concentration for the AL-FW orientation (Fig. 5). This is due tothe more enhancement of the water permeation flux than the HAAfluxwith the increase of the DSNaCl concentration (Eq. (3)). In contrast,the HAA rejection ratios could reach their maximum under a mediumosmotic pressure difference for the AL-DS orientation. The increase inwater permeation flux was faster than the HAA flux when the DS NaClconcentration was relatively low, while the increase in water perme-ation flux was slower when the DS NaCl concentration was relativelyhigh. Therefore, HAA rejection increased initially with the increase ofwater flux due to the enhanced dilution effect and a further increasein water flux resulted in a reduced rejection due to very severe ICP of

Fig. 5. The HAA rejection ratios obtained by experiments and predicted by the model as afunction of the DSNaCl concentration for the AL-FS orientation (top) and the AL-DS orien-tation (bottom).

HAAs. This convex profile was also observed in other studies for boricacid removal by FO [19].

The rejection ratios for MCAA and MBAA were 94.6% and 95.1%, re-spectively, and that for the rest of HAAs were higher than 97.0% forthe AL-FW orientation, with a DS NaCl concentration at 1 mol/L. Thisperformance is comparable to that of nanofiltration (NF) and reverseosmosis (RO). Agus and Sedlak reported that the HAA rejection ratioswere between56% and 100% byNF, and between 86% and 94%by ROde-pending on the applied pressures [38]. If all FO, NF and RO reject soluteby the samemechanism, the FOmembrane should have a surface “pore”size that is comparable to that of an NF or RO membrane. In addition,similar to the NF/RO membranes, another mechanism contributing totheHAA removal in FOmight be charge repulsion [19]. HAAswere high-ly rejected by FO, NF or RO because HAAs dissociate and are negativelycharged unless the pH is very low (≤3) (Table 1). Most FO, NF and ROmembrane surfaces are also negatively charged at circum-neutral pHranges. Under the DS NaCl concentration of 1 mol/L, the rejection ratiosfor all the respective HAAs ranged from73.8% to 89.1% for the AL-DS ori-entation, which were high but lower than those for the AL-FW orienta-tion. It is because of the much higher HAA flux, albeit relatively higherwater permeation flux for the AL-DS orientation. Based on current FOmembrane technology, the AL-DS orientation may not be preferreddue to its lower removal efficiency of pollutants and higher membranefouling potential [14,19].

3.4. HAA rejection and reverse draw solute flux prediction

With the S value obtained, the mass transfer coefficient for eachHAA (Km,HAA) can be calculated by the ratio of respective diffusioncoefficient listed in Table 1 to the S value (Km = D/S). In thisstudy, the average of the two values of the HAA permeability coeffi-cient (BHAA) was used for the prediction of the rejection ratios of theHAAs (Table 2).

Given the water permeability coefficient (A), the HAA permeabilitycoefficient (BHAA) and the HAA mass transfer coefficient (Km,HAA), therejection ratio for each HAA can be predicted by using Eq. (3) in com-bination with Eq. (6) for the AL-FW orientation or Eq. (9) for the AL-DS orientation. The modeling results matched well with theexperimentally-obtained rejection ratios when the FO was operated inthe AL-FW orientation mode (Fig. 5). However the modeling resultsover-estimated the rejection ratios when the FO was operated in theAL-DS mode (for which Km,HAA took an important role). There aresome possible reasons for the over-estimation. As described earlier,the under-estimation of the S value would be partly responsible forthe over-estimation of theKm,HAA value and in turn theHAA rejection ra-tios. If it were the case, the rejection ratio over-estimation would bemore significant forMCAA andMBAA, because both of themhave small-er molecular weight and thus larger diffusion coefficient than others.Recently, studies [40] showed that the permeability coefficient mightbe better quantified byKm= δD/S, where δ is the so-called constrictivityfactor. The constrictivity factors for sodium ion, chloride ion andthe HAAs may not be identical, which was also partly responsible forthe over-estimation of the HAA rejection ratios. However, becauseNa+, Cl− and the HAA species are all small molecules, such effect ofδ would be very limited. Another probable reason for the over-estimation of the rejection ratio was that the actual diffusivity for eachHAA was lower than that shown in Table 1. In this study, the molecularsize of each HAA was calculated by using the Wilke–Chang equation,which was then utilized for the calculation of the diffusivity by usingthe Einstein–Stokes equation.More accurate determination of the diffu-sivity for each HAA in the FO membrane support layer may to some ex-tent reduce the difference between the predicted and the experimentalresults for the HAA rejection ratios. Moreover, FO membranes were re-cently found containing pores of sub-nm sizes in the active layer[41–43]. As such, somewhat in contradictory to the assumption of thesolution-diffusion model that the solute (both NaCl and HAA) was

Fig. 6. The reverse selectivity of the draw solute as a function of DSNaCl concentrations forboth the AL-FW and AL-DS orientations.

Fig. 7.A comparisonof the experimental results andmodel predictions of the reverse drawsolute flux for the AL-FS orientation (top) and the AL-DS orientation mode (bottom). Thefitting line (slope = 1) represents a perfect agreement.

7F. Kong et al. / Desalination 341 (2014) 1–9

transferred through the FO membrane active layer by diffusion only,mass transfer of HAA through the active layer by convection was alsopossible especially when the water permeation flux was high. The con-tribution by this mechanism however needs more investigation.

Given the water permeability coefficient (A) and the NaCl perme-ability coefficient (BNaCl), the reverse drawsolute (NaCl) flux can be pre-dicted by using Eq. (12) [23,27].

JNaClJw

¼ BNaCl

AnRgTð12Þ

where n is the number of dissolved species of the draw solute (2 forNaCl), Rg the universal gas constant, and T the absolute temperature.An assumption that underlies Eq. (12) is that the calculation of the os-motic pressure follows the van't Hoff equation [22,23,27]. Eq. (12)shows that the ratio of reverse draw solute flux to thewater flux is inde-pendent of the DS concentration or the extent of the concentration po-larization [22,23,27]. As shown in Fig. 6, the experimental resultsconfirmed that the ratio of reverse draw solute to the water flux was aconstant for both the AL-FW and AL-DS orientations. The desire forhigher water flux and the concurrent higher reverse draw solute fluxare in contradiction for the current FO process. Compared with the AL-FW orientation, both water flux and reverse draw solute flux for theAL-DS orientation are higher because of the absence of dilutive ICP ofthe draw solute. According to Eq. (12), the best way to reduce the re-verse flux of the draw solute is to either decrease the NaCl permeabilitycoefficient (BNaCl) or increase the water permeability coefficient (A). Byfar, a number of studies were devoted primarily to the increase of thewater permeability coefficient by incorporation into the FO membraneactive layer of highly water-permeable aquaporin-based proteolipo-somes [44] or nanomaterials (e.g. zeolite particles or nanotube) [45,46].

The reverse draw solute flux can be also predicted by the solution-diffusion model (Eq. (5) for the AL-FW orientation and Eq. (8) for theAL-DS orientation) by assuming that the NaCl concentrations in thefeed and draw solutions are constant, given the experimentally deter-mined parameters of the membrane Km,NaCl (2.76 × 10−6 m/s), DNaCl

(1.5 × 10−9 m2/s), S (5.43 × 10−4 m), and k (1.96 × 10−5 m/s). Theagreement between the model prediction and the experimental datais evident (Fig. 7), with the slope of the correlation line equal to 1. Itdemonstrated that the solution-diffusion model combined with con-centration polarization can be used for the prediction of the reversedraw solute permeation in a multicomponent system.

4. Conclusions

In this paper, thewater flux, solute flux and reverse draw solute fluxwere systematically investigated:

(1) For both membrane orientations, the HAA fluxes and reversedraw solution permeation flux increased with the increase ofthe DS concentration. The smaller the molecular weight of anHAA, the higher the HAA flux through the FO membrane. TheHAA fluxes increased acceleratedly for the AL-DS orientationwhile it increased deceleratedly for the AL-FW orientation withthe increase of the DS concentration. The reverse draw soluteflux for the AL-FW orientation was much lower than that forthe AL-DS orientation.

(2) The rejection ratio for each HAA showed a monotonous increasewith the increase of DS concentration for the AL-FW orientation.In contrast, the HAA rejection ratios could reach their maximumunder amediumosmotic pressure difference for theAL-DS orien-tation. The rejection ratios for HAAsweremore than 94.6% for theAL-FW orientation and ranged from 73.8% to 89.1% for the AL-DSorientation under the DS NaCl concentration of 1 mol/L.

(3) The solution-diffusion model combined with the concentrationpolarization model was able to predict the bi-directional masstransfer in FO systems. However the modeling results over-estimated the rejection ratios for HAAs for the AL-DS orientation

8 F. Kong et al. / Desalination 341 (2014) 1–9

probably due to the adoption of inaccurate mass transfer coeffi-cient for internal concentration polarization. More precise deter-mination of the HAA diffusivity in the FO membrane supportlayer and quantification of the contribution by convection arerequired.

NomenclatureA water permeability of the membrane (L/(m2 h bar))Am effective area of the membrane coupon (cm2)B solute permeability of the membrane (μg/(m2 h bar))c solute concentrations (μg/L)c0 concentrations of the solute at t = 0 (μg/L)ct concentrations of the solute at time t (μg/L)dh hydraulic diameter of the channel (m)D diffusion coefficient (m2/s)Jw permeate water flux (L/(m2 h))JHAA HAA fluxes (μg/(m2 h))JNaCl draw solute (NaCl) flux (mol/(m2 h))km mass transfer coefficient accounting for ECP (m/s), km ¼ ShD

dhKm resistivity for diffusionwithin the porous support layer (m/s),

Km ¼ DS

MB molecular weight of the solvent (water) (g/mol)n number of dissolved species of the draw solute (2 for NaCl)R solute rejection ratioRg universal gas constant (J/(K mol))S intrinsic parameter related to support layer structure (m),

S = tτ / εSh the Sherwood number of the appropriate flow regime in the

rectangular flow channelT absolute temperature (K)t thickness of the support layer (m)v volume of the solution in the diaphragm cell (L)VA solute molar volume at the boiling point (m3/(kg mol))

Greek symbolsΔt specified time interval (h)Δ(Vc) variation of target compound content (μg)Δm/Δt weight change rate (μg/(L h))ΔP hydraulic pressure difference (bar)Δπ osmotic pressure difference across the membrane (bar)Δc solute concentration difference across the active layer

(mol/L)μB solvent viscosity (Pa s)ε porosity of the support layerτ tortuosity of the support layerφ association parameter of the solvent (2.6 for water)

SubscriptsD draw solutionF feed waterHAA each of the haloacetic acidsNaCl draw soluteRe receiver of the diffusion processSo source of the diffusion processw water

Acknowledgments

The authors acknowledge the financial support provided by the Chi-neseNatural Science Foundation (No. 51278268, No. 51290284) and theBeijing Municipal Natural Science Foundation (No. 8132043).

References

[1] L.A. Hoover, W.A. Phillip, A. Tiraferri, N.Y. Yip, M. Elimelech, Forward with osmosis:emerging applications for greater sustainability, Environ. Sci. Technol. 45 (2011)9824–9830.

[2] S. Zhao, L. Zou, C.Y. Tang, D. Mulcahy, Recent developments in forward osmosis: op-portunities and challenges, J. Membr. Sci. 396 (2012) 1–21.

[3] T.S. Chung, S. Zhang, K.Y. Wang, J.C. Su, M.M. Ling, Forward osmosis processes: yes-terday, today and tomorrow, Desalination 287 (2012) 78–81.

[4] Y.J. Choi, J.S. Choi, H.J. Oh, S. Lee, D.R. Yang, J.H. Kim, Toward a combined system offorward osmosis and reverse osmosis for seawater desalination, Desalination 247(2009) 239–246.

[5] B. Qi, Y. Wang, S. Xu, Z. Wang, S. Wang, Operating energy consumption analysis ofRO desalting system: effect of membrane process and energy recovery device(ERD) performance variables, Ind. Eng. Chem. Res. 51 (2012) 14135–14144.

[6] P. Patel, Water desalination takes a step forward, Environ. Sci. Technol. 40 (2006)3454–3455.

[7] A. Achilli, T.Y. Cath, E.A.Marchand, A.E. Childress, The forward osmosismembranebio-reactor: a low fouling alternative to MBR processes, Desalination 239 (2009) 10–21.

[8] Y.F. Xie, Disinfection Byproducts in Drinking Water: Formation, Analysis, and Con-trol, Lewis Publishers, Boca Raton, Florida, USA, 2004.

[9] C. Ort, M.G. Lawrence, J. Rieckermann, A. Joss, Sampling for pharmaceuticals andpersonal care products (PPCPs) and illicit drugs in wastewater systems: are yourconclusions valid? A critical review, Environ. Sci. Technol. 44 (2010) 6024–6035.

[10] G.A. Loraine, M.E. Pettigrove, Seasonal variations in concentrations of pharmaceuti-cals and personal care products in drinking water and reclaimed wastewater insouthern California, Environ. Sci. Technol. 40 (2005) 687–695.

[11] C. Carrara, C.J. Ptacek, W.D. Robertson, D.W. Blowes, M.C. Moncur, E. Sverko, S.Backus, Fate of pharmaceutical and trace organic compounds in three septic systemplumes, Ontario, Canada, Environ. Sci. Technol. 42 (2008) 2805–2811.

[12] R. Valladares Linares, V. Yangali-Quintanilla, Z. Li, G. Amy, Rejection ofmicropollutants by clean and fouled forward osmosis membrane, Water Res. 45(2011) 6737–6744.

[13] M. Xie, L.D. Nghiem, W.E. Price, M. Elimelech, Comparison of the removal of hydro-phobic trace organic contaminants by forward osmosis and reverse osmosis, WaterRes. 46 (2012) 2683–2692.

[14] M. Xie, W.E. Price, L.D. Nghiem, Rejection of pharmaceutically active compounds byforward osmosis: role of solution pH andmembrane orientation, Sep. Purif. Technol.93 (2012) 107–114.

[15] A.A. Alturki, J.A. McDonald, S.J. Khan, W.E. Price, L.D. Nghiem, M. Elimelech, Removalof trace organic contaminants by the forward osmosis process, Sep. Purif. Technol.103 (2013) 258–266.

[16] J.L. Cartinella, T.Y. Cath, M.T. Flynn, G.C. Miller, K.W. Hunter, A.E. Childress, Removalof natural steroid hormones fromwastewater using membrane contactor processes,Environ. Sci. Technol. 40 (2006) 7381–7386.

[17] T.Y. Cath, N.T. Hancock, C.D. Lundin, C. Hoppe-Jones, J.E. Drewes, A multi-barrier os-motic dilution process for simultaneous desalination and purification of impairedwater, J. Membr. Sci. 362 (2010) 417–426.

[18] N.T. Hancock, P. Xu, D.M. Heil, C. Bellona, T.Y. Cath, Comprehensive bench- andpilot-scale investigation of trace organic compounds rejection by forward osmosis,Environ. Sci. Technol. 45 (2011) 8483–8490.

[19] X. Jin, C.Y. Tang, Y. Gu, Q. She, S. Qi, Boric acid permeation in forward osmosis mem-brane processes: modeling, experiments, and implications, Environ. Sci. Technol. 45(2011) 2323–2330.

[20] N.T. Hancock, T.Y. Cath, Solute coupled diffusion in osmotically driven membraneprocesses, Environ. Sci. Technol. 43 (2009) 6769–6775.

[21] M. Park, J. Lee, C. Boo, S. Hong, S.A. Snyder, J.H. Kim, Modeling of colloidal fouling inforward osmosis membrane: effects of reverse draw solution permeation, Desalina-tion 314 (2013) 115–123.

[22] W.A. Phillip, J.S. Yong, M. Elimelech, Reverse draw solute permeation in forward os-mosis: modeling and experiments, Environ. Sci. Technol. 44 (2010) 5170–5176.

[23] J.S. Yong, W.A. Phillip, M. Elimelech, Coupled reverse draw solute permeation andwater flux in forward osmosis with neutral draw solutes, J. Membr. Sci. 392–393(2012) 9–17.

[24] J.M. Duan, W. Li, P.P. Sun, Q. Lai, D. Mulcahy, S.N. Guo, Rapid determination of ninehaloacetic acids in wastewater effluents using ultra-performance liquid chromatog-raphy–tandem mass spectrometry, Anal. Lett. 46 (2013) 569–588.

[25] J.G. Wijmans, R.W. Baker, The solution-diffusion model: a review, J. Membr. Sci. 107(1995) 1–21.

[26] J.R. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal con-centration polarization on flux behavior in forward osmosis, J. Membr. Sci. 284(2006) 237–247.

[27] C.Y. Tang, Q. She, W.C.L. Lay, R. Wang, A.G. Fane, Coupled effects of internal concen-tration polarization and fouling on flux behavior of forward osmosis membranesduring humic acid filtration, J. Membr. Sci. 354 (2010) 123–133.

[28] B.S. Chanukya, S. Patil, N.K. Rastogi, Influence of concentration polarization on fluxbehavior in forward osmosis during desalination using ammonium bicarbonate, De-salination 312 (2013) 39–44.

[29] C.Y. Tang, Q. She, W.C.L. Lay, R. Wang, R. Field, A.G. Fane, Modeling double-skinnedFO membranes, Desalination 283 (2011) 178–186.

[30] T.Y. Cath, M. Elimelech, J.R. McCutcheon, R.L. McGinnis, A. Achilli, D. Anastasio, A.R.Brady, A.E. Childress, I.V. Farr, N.T. Hancock, J. Lampi, L.D. Nghiem, M. Xie, N.Y. Yip,Standardmethodology for evaluating membrane performance in osmotically drivenmembrane processes, Desalination 312 (2013) 31–38.

[31] E.L. Cussler, Diffusion: Mass Transfer in Fluid Systems, Cambridge University Press,Cambridge, UK, 2009.

9F. Kong et al. / Desalination 341 (2014) 1–9

[32] E. Drazevic, S. Bason, K. Kosutic, V. Freger, Enhanced partitioning and transport ofphenolic micropollutants within polyamide composite membranes, Environ. Sci.Technol. 46 (2012) 3377–3383.

[33] A. Ben-David, Y. Oren, V. Freger, Thermodynamic factors in partitioning and rejec-tion of organic compounds by polyamide composite membranes, Environ. Sci.Technol. 40 (2006) 7023–7028.

[34] N. Nestle, R. Kimmich, Concentration-dependent diffusion coefficients and sorptionisotherms. Application to ion exchange processes as an example, J. Phys. Chem. 100(1996) 12569–12573.

[35] M. Park, J.J. Lee, S. Lee, J.H. Kim, Determination of a constantmembrane structure pa-rameter in forward osmosis processes, J. Membr. Sci. 375 (2011) 241–248.

[36] V. Vitagliano, P.A. Lyons, Diffusion coefficients for aqueous solutions of sodium chlo-ride and barium chloride, J. Am. Chem. Soc. 78 (1956) 1549–1552.

[37] R. Chalatip, R. Chawalit, R. Nopawan, Removal of haloacetic acids by nanofiltration,J. Environ. Sci. China 21 (2009) 96–100.

[38] E. Agus, D.L. Sedlak, Formation and fate of chlorination by-products in reverse osmo-sis desalination systems, Water Res. 44 (2010) 1616–1626.

[39] D.H. Jung, J. Lee, D.Y. Kim, Y.G. Lee, M. Park, S. Lee, D.R. Yang, J.H. Kim, Simulation offorward osmosis membrane process: effect of membrane orientation and flow di-rection of feed and draw solutions, Desalination 277 (2011) 83–91.

[40] S. Zhao, L. Zou, Relating solution physicochemical properties to internal concentra-tion polarization in forward osmosis, J. Membr. Sci. 379 (2011) 459–467.

[41] A. D'Haese, P. Le-Clech, S. Van Nevel, K. Verbeken, E.R. Cornelissen, S.J. Khan, A.R.Verliefde, Trace organic solutes in closed-loop forward osmosis applications: influ-ence of membrane fouling and modeling of solute build-up, Water Res. 47 (2013)5232–5244.

[42] Y.N. Wang, J. Wei, Q. She, F. Pacheco, C.Y. Tang, Microscopic characterization ofFO/PRO membranes — a comparative study of CLSM, TEM and SEM, Environ. Sci.Technol. 46 (2012) 9995–10003.

[43] Y. Fang, L. Bian, Q. Bi, Q. Li, X. Wang, Evaluation of the pore size distribution of a for-ward osmosis membrane in three different ways, J. Membr. Sci. 454 (2014)390–397.

[44] Y. Zhao, A. Vararattanavech, X.S. Li, C. HelixNielsen, T. Vissing, J. Torres, R. Wang, A.G.Fane, C.Y. Tang, Effects of proteoliposome composition and draw solution types onseparation performance of aquaporin-based proteoliposomes: implications for sea-water desalination using aquaporin-based biomimetic membranes, Environ. Sci.Technol. 47 (2013) 1496–1503.

[45] N. Ma, J. Wei, R. Liao, C.Y. Tang, Zeolite–polyamide thin film nanocomposite mem-branes: towards enhanced performance for forward osmosis, J. Membr. Sci.405–406 (2012) 149–157.

[46] K. Goh, L. Setiawan, L. Wei, W.C. Jiang, R. Wang, Y. Chen, Fabrication of novel func-tionalized multi-walled carbon nanotube immobilized hollow fiber membranes forenhanced performance in forward osmosis process, J. Membr. Sci. 446 (2013)244–254.