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

Desalination 341 (2014) 1–9

Contents lists available at ScienceDirect

Desalination journal homepage: www.elsevier.com/locate/desal

Rejection of nine haloacetic acids and coupled reverse draw solute permeation in forward osmosis Fan-xin Kong a, Hong-wei Yang a, Xiao-mao Wang a,⁎, Yuefeng F. Xie a,b a b

State Key Joint Laboratory of Environmental Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing 100084, China 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

• Forward osmosis rejects haloacetic acids (HAAs) well. • The active-layer facing feed water (AL-FW) orientation is preferred. • The solution-diffusion model predicts HAA rejection better for AL-FW. • The solution-diffusion model accurately predicts reverse draw solution permeation.

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

a r t i c l e

a b s t r a c t

i n f o

Article history: Received 22 September 2013 Received in revised form 15 February 2014 Accepted 17 February 2014 Available online 13 March 2014 Keywords: Haloacetic acids (HAAs) Forward osmosis Concentration polarization Solution-diffusion model Reverse draw solute permeation

A B S T R A C T

The rejection of haloacetic acids (HAAs) by forward osmosis (FO) and the coupled reverse draw solute permeation were experimentally determined and mathematically modeled by using the solution-diffusion model for both the AL-FW (active layer facing the feed water) and the AL-DS (active layer facing the draw solution) orientations. The rejection ratio for each HAA increased with the increase of draw solute concentration for the AL-FW orientation. In contrast, the HAA rejection ratio could reach its maximum under a medium osmotic pressure difference for the AL-DS orientation. The rejection ratios for all HAAs were higher than 94.6% for the AL-FW orientation and ranged from 73.8% to 89.1% for the AL-DS orientation under a draw solute concentration of 1 mol/L NaCl. 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-FW orientation. However, the model over-estimated the rejection ratios for the AL-DS mode, likely due to the adoption of inaccurate mass transfer coefficient for internal concentration polarization. Regarding the reverse draw solute permeation, a general agreement between the model prediction and experimental data was observed for both orientations. © 2014 Elsevier B.V. All rights reserved.

1. Introduction

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

http://dx.doi.org/10.1016/j.desal.2014.02.019 0011-9164/© 2014 Elsevier B.V. All rights reserved.

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

2

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

wastewater reclamation has been extensively studied [1–3], which was normally fulfilled by combining FO with other technologies as a hybrid process [4]. The hybrid processes may compete favorably with the simple pressure-driven processes such as reverse osmosis (RO) owing to the fact that the former may have a lower demand on electrical energy and its membranes may suffer less membrane fouling [2,5–7]. One of the biggest concerns about wastewater reclamation is the removal efficiencies for the large number of trace organic compounds at concentrations in the magnitudes of ng/L to μg/L. Common trace organic compounds include endocrine disruptors (EDCs), pharmaceutical and personal care products (PPCPs) and disinfection by-products (DBPs) [8–11]. The reclaimed water is better to be free of these trace organic compounds due to their potentially high health risks to humans and aquatic organisms. Previous studies [2,12–17] showed that the rejection of PPCPs and EDCs by FO process varied among the compounds and the operational conditions. Results by Hancock revealed that rejection efficiencies by FO depended primarily on molecular size and charge. Rejection of positively and negatively charged trace organic compounds was greater than 80%, while rejection of the nonionic trace organic compounds varied, between approximately 40% and 90% [18]. Xie et al. [13] found that the membrane orientation (AL-FW (active layer facing feed water) and AL-DS (active layer facing draw solution)) affected the rejection performance. Development of mathematical models capable of predicting the rejection of the trace organic compounds by the FO process will improve the economic efficiency of the process and expand its applications. By far, there has been only a few work conducted on the modeling of the rejection performance by FO, among which the model based on the solution-diffusion model developed by Jin et al. [19] was found able to accurately predict the FO rejection of inorganic solutes. It is also desirable to model the permeate water flux and the reverse draw solute flux through the membrane. The limited water flux and the substantial reverse draw solute flux are among the biggest impediments to the viability of the FO processes [20,21]. In this regard, a few studies investigated the reverse draw solute permeation in the FO process and proposed a relationship with the water flux [22,23]. Better understanding and prediction of feed contaminant and coupled reverse draw solute permeation in FO membrane processes can facilitate to optimize the operational conditions and benefit to fabricate FO membranes that minimize the loss of draw solute into the feed solution, thereby reducing operating costs. Haloacetic acids (HAAs) are among the DBPs with the highest concentrations in chlorinated or chloraminated sewage treatment plant effluent. The trichloroacetic acid (TCAA) concentration can be as high as 471 μg/L in the chlorinated wastewater effluent [24]. This study mainly investigated the rejection of nine HAAs and the reverse draw solute permeation in the FO process for both the AL-FW and the AL-DS orientations. Substantial attention was paid to the mathematical description of the water, HAA and draw solute fluxes. This study also evaluated the feasibility of the solution-diffusion model in predicting the rejection of organic contaminants and coupled draw solute permeation in the FO process. 2. Materials and methods 2.1. The FO system and operations The schematical diagram of the bench-scale FO system is shown in Fig. 1. The membrane unit was custom built with two flow channels of 2 mm in height both, which were separated by an FO membrane coupon that had an effective area of 40.5 cm2. The commercial FO membrane was provided by Hydration Technologies, Inc. (Albany, OR) and, according to the manufacturer, was made of cellulose triacetate supported by embedded polyester screen mesh. The recirculation flows of the feed water (FW) and the draw solution (DS) on the two sides of the

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

membrane were counter-current with a flow velocity of 20.4 cm/s both, each regulated by a variable-speed gear pump (Longer, USA). The relatively high cross-flow velocity can substantially suppress the external concentration polarization (ECP) that will be described in detail later. Ultrapure water (Milli-Q, Millipore, USA) was used for the FW and the DS preparation. The FW was prepared by dissolving a mixture of the nine haloacetic acids (HAAs) (Sigma-Aldrich, Germany) into a 10 mmol/L NaCl (as background solute) solution with the final concentration for each HAA at 200 μg/L. The FW ionic strength was similar to the typical values for freshwater including sewage treatment plant effluent, although our preliminary study showed that the FW ionic strength (0–20 mmol/L) had negligible influence on the rejection of HAAs. The nine HAAs were monochloroacetic acid (MCAA), monobromoacetic acid (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 each HAA including molecular weight (MW) and dissociation constant (pKa) were summarized in Table 1. Sodium bicarbonate was added into the FW to adjust the pH to neutral (pH = 7). No natural organic matter (NOM) or soluble microbial products (SMPs) were contained in the FW, though both NOM and SMP could not only adsorb certain trace organic compounds but also lead to membrane fouling, which would greatly affect the rejection performance of the FO process [12]. The DS was a NaCl solution at a concentration varying from 0.1 to 3 mol/L. The FO experiments were conducted in an air-conditioned room with the temperature set at 25 ± 1 °C. Each FO test lasted for about 4 h. The duration was long enough to obtain accurate water, HAA and reverse draw solute permeation fluxes. On the conclusion of each FO test, the volume variation of the DS (or the FW) was less than 5%, and as such the draw solute (in the DS) and the HAA (in the FW) concentration variation were negligible. The FW tank was placed on a digital scale (Mettler Toledo, Germany) connected to a computer data logging

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

3

Table 1 Physicochemical properties of the nine haloacetic acids. Haloacetic acids (HAAs)

MWa (g/mol)

Structure

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 b

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

system. The weight change as a function of time was utilized to determine the permeate water flux, Δm Jw ¼ ΔtAm

ð1Þ

where Δm/Δt is the weight change rate and Am is the effective area of the membrane coupon. At specified time intervals, samples were taken from either the FW tank or the DS tank for HAA concentration, chloride ion and pH value determination. The variation of the HAA concentrations in the DS was used to determine the HAA flux through the FO membrane, while the variation of the NaCl concentration in the FW was used for the determination of the reverse solute flux through the FO membrane, Jw ¼

ΔðVcÞ ΔtAm

ð2Þ

where Δ(Vc) is the variation of target compound in the FW or DS tanks during the specified time interval Δt. Given the permeate water flux (Jw) and the HAA fluxes (JHAA), the rejection ratio of each HAA compound (RHAA) in the FO process can be obtained by RHAA ¼ 1−

J HAA J w c F;HAA

ð3Þ

where cF,HAA is the HAA concentrations in the FW, which did not change significantly 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 solute fluxes can be referred to the model developed by Elimelech, Tang and their colleagues [25–29]. Most of the previous models made use of the solution-diffusion model coupled with concentration polarization. For the AL-FW orientation, the draw solute experiences dilutive internal concentration polarization (ICP) on the DS side and concentrative ECP on the FW side. The permeate water flux Jw is expressed by


 J w ¼ A π D;NaCl exp − J w =K m;NaCl −π F;NaCl exp J w =K m;NaCl :

ð4Þ

The reverse draw solute flux JNaCl can be expressed as


 J NaCl ¼ BNaCl C D;NaCl exp −J w =K m;NaCl −C F;NaCl exp J w =K m;NaCl :

ð5Þ

The HAA flux through the FO membrane can be expressed as

J HAA ¼


 BHAA exp J w =K m;HAA 1 þ BHAA =J w

c F;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 porous layer, and is related with the S value by Km = D/S. The S parameter is the intrinsic length scale of ICP in the FO membrane support layer, which is analogous to the boundary layer thickness for ECP. Mathematically, S is related to the thickness (t), tortuosity (τ) and porosity (ε) of the support layer by S = tτ / ε. The S value is usually obtained by fitting the water flux vs. draw solute osmotic pressure data under known hydraulic conditions (Eq. (4) or Eq. (7) below). (Care however must be taken in that it is not a direct method and the inaccuracy of a number of parameters would lead to the inaccurate determination of the S value.) Given the S value, the Km value for any solute can be easily obtained only if the diffusivity (D) is known. Km is hardly affected by the hydrodynamic conditions. In contrast, the km value is controlled by the hydrodynamics and can be calculated by km = ShD / dh, where Sh and dh are the Sherwood number and the hydraulic diameter of the FO cell channel, respectively. For the AL-DS orientation, the draw solute experiences dilutive ECP on the DS side and concentrative ICP on the FW side. The permeate water flux can be expressed as


 J w ¼ A πD;NaCl exp − J w =K m;NaCl −π F;NaCl exp J w =K m;NaCl :

ð7Þ

The reverse draw solute flux can be expressed as


 J NaCl ¼ BNaCl C D;NaCl exp −J w =K m;NaCl −C F;NaCl exp J w =K m;NaCl :

ð8Þ

4

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

The HAA flux from the FW to the DS can be expressed as J HAA ¼ 1þ


 BHAA exp J w =K m;HAA

 c F;HAA : BHAA exp J w =K m;HAA exp −J w =K m;HAA

ð9Þ

Jw

3. Results and discussion

A rough calculation revealed that the Sherwood number and the km value were approximately 150 and 4 × 10−5 m/s, the latter of which was much larger than the water permeation flux (Jw) (see below). Therefore, ECP was sufficiently suppressed so that it did not affect the flux 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 determination of the water, NaCl and HAA permeability coefficients pertinent to the FO membrane [30]. In the filtration unit, the active layer of the FO membrane faced the high pressure side (i.e. under RO testing mode) to eliminate 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 highpurity water (Milli-Q, Millipore, USA), i.e. Jw = A · ΔP. The NaCl permeability coefficient (BNaCl) was derived from the rejection ratio versus the applied pressure (Eq. (10)) by filtering a 10 mmol/L NaCl solution. 1

R¼ 1þ

B AðΔP−ΔπÞ

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

ð10Þ

where Δπ is the osmotic pressure difference across the membrane. The HAA permeability coefficients (BHAA) were also derived from the rejection ratio versus the applied pressure (Eq. (10)) by filtering the mixed solution of the nine HAAs with 10 mmol/L NaCl as background. It was assumed that the rejection of each HAA was independent when the background salt concentration was much higher than that of any HAA. The HAA permeability coefficients can also be determined by using the diaphragm cell test [31]. The cylindrical diffusion cell (Fig. 1) was made of glass and consisted of two compartments which were separated by an FO membrane with an effective surface area of 20.4 cm2. Each compartment had a volume of 1 L. For the diaphragm cell test, one compartment contained a solution of NaCl at 10 mmol/L, and the other contained a solution of both NaCl at 10 mmol/L and the target HAAs at 500 μg/L each. The pH of the HAA solution was adjusted to around 7 by using a minimum volume of the NaHCO3 solution. The ionic strength of the two compartments was almost the same (approximately 10 mmol/L), thus avoiding water permeation driven by osmotic pressure difference. The solutions in both compartments were continuously magnetically stirred and thermostatically stabilized at a temperature of 25 ± 1 °C. Samples from the two compartments were taken for HAA measurement every day that lasted for 18 days. The permeability coefficient can be calculated by using Eq. (11) [31]. ! 1 c0So −c0Re B¼ 1n t ð11Þ Am t ðð1=vSo Þ þ ð1=vRe ÞÞ cSo −ctRe where c0So and c0Re are the HAA concentrations in the source and receiver of the diffusion process at t = 0, ctSo and ctRe are the concentrations of the solute in the source and receiver at time t, Am is the membrane area, and vSo and vRe are the volumes of the source and receiver of the diaphragm cell, respectively. 2.4. Analytical methods The HAAs were analyzed by a gas chromatograph equipped with an electron capture detector (Agilent 7890A, Santa Clara, USA) after liquid–

3.1. Permeability coefficients of the FO membrane By conducting the pressurized cross-flow filtration tests, the A and BNaCl values were determined as 2.38 (± 0.04) × 10−7 m/(s bar) and 2.90 (±0.5) × 10−7 m/s, respectively, which are very close to that reported by others [19]. The determined permeability coefficients for each HAA by both the pressurized cross-flow filtration test and the diaphragm cell test were shown in Table 2 (the determination of the TBAA permeability coefficient was not attainable using the diaphragm cell test due to the substantial hydrolysis of TBAA [8] during the test). The values obtained from the diaphragm unit were slightly higher than that from the pressurized filtration (Table 2). This might lie in the fact that the permeability coefficient of the solute is a concentration-dependent parameter. Studies showed that both the diffusion coefficient in the dense polymer and the partitioning coefficient at the solution–polymer (both of which determine the permeability coefficient) are weakly dependent on the solution concentration [32,33]. However, the reliance of the permeability 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 flux was observed with an increase in the DS NaCl concentration (i.e. apparent osmotic pressure difference between the DS and FW) for both the AL-FW and AL-DS orientations (Fig. 2). Under the same osmotic pressure difference, the water permeation flux for the AL-FW orientation was always lower than that for the AL-DS orientation. For example, when the DS NaCl concentration was 1 mol/L, the water permeation fluxes for the AL-FW and AL-DS orientations were 12.8 and 20.6 L/(m2h), respectively. When the FO membrane was operated in the AL-FW orientation, severe dilutive ICP of the draw solute occurred which greatly reduced the effective osmotic pressure difference across the FO membrane. In contrast, mild concentrative ICP was mainly responsible for the decrease of the effective osmotic pressure for the AL-DS orientation. Given the permeability coefficients of water (A) and NaCl (BNaCl), the mass transfer coefficient of NaCl (Km,NaCl) can be obtained by fitting the water flux and the osmotic pressure data using either Eq. (4) (for the AL-FW orientation) or Eq. (7) (for the AL-DS orientation). The Km,NaCl values obtained from the AL-FW orientation data and the AL-DS orientation data were almost

Table 2 HAA permeability coefficients using diaphragm cell and the pressurized filtration test units. HAAs

Diaphragm cell test (m/s)

Pressurized filtration test (m/s)

Average values (m/s)

MCAA MBAA DCAA TCAA BCAA BDCAA DBAA CDBAA TBAA

3.00E−07 2.18E−07 8.62E−08 4.34E−08 4.38E−08 2.03E−08 2.50E−08 1.73E−08 NAa

2.05E−07 1.59E−07 6.13E−08 3.26E−08 3.30E−08 1.78E−08 2.13E−08 1.20E−08 1.49E−08

2.53E−07 1.89E−07 7.37E−08 3.80E−08 3.84E−08 1.91E−08 2.32E−08 1.46E−08 1.49E−08

a

Not available due to substantial hydrolysis [8].

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

5

40

AL-FW Fitted AL-DS Fitted

35

Water Flux (L/(m2h)

30 25 20 15 10 5 0

0

0.5

1

1.5

2

2.5

3

NaCl Concentration (M) Fig. 2. Experimental results of the water flux and fitted results using either Eq. (4) or Eq. (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.

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 Tang et al. [27] (4.2 × 10−6 and 3.3 × 10− 6 m/s for the AL-FW and AL-DS orientations, respectively). With both Km,NaCl (2.76 × 10− 6 m/s) and the 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 which determines the ICP within the supporting layer by membrane thickness, tortuosity and porosity) was calculated to be 5.43 × 10− 4 m. Note however that the osmotic pressure for the draw solution might be under-estimated due to use of the Morse equation (i.e. linear relationship of the osmotic pressure with the concentration) for the calculation. It indicates that the Km,NaCl value might be somewhat over-estimated and in turn the S value was somewhat under-estimated. However, the under-estimation was estimated within 10%. For both orientations, the HAA fluxes also increased with the increase of the DS NaCl concentration (Fig. 3). A general trend could be found that the smaller the molecular weight of the HAA species, the higher the HAA flux through the FO membrane. In other words, the HAA flux mainly followed the declining order of MCAA, MBAA, DCAA, TCAA/BCAA, DBAA/BDCAA, CDBAA and TBAA. It was assumed that FO follows the solution-diffusion model, and an HAA of smaller weight might have a higher “diffusion” rate within the active layer of the FO membrane. In addition, the stability (i.e. hydrolysis) and the water bonding potential of HAAs would probably also influence the permeation of HAAs through the FO membrane [37,38]. The main difference between the two orientations is that under the same osmotic difference, the HAA flux for the AL-FW orientation was much lower than that for the AL-DS orientation (Fig. 3). Severe concentrative ICP of the HAA species occurred for the AL-DS orientation, which greatly increased the HAA concentration at the active layer–supporting layer interface. In contrast, no ICP of the HAA species would occur for the AL-FW orientation. Another difference between the two orientations is that the HAA fluxes increased acceleratedly for the AL-DS orientation while it increased deceleratedly for the AL-FW orientation when the DS NaCl concentration was increased. The higher water flux for the AL-DS orientation than that for the AL-FW orientation and the critical role of ICP of HAA are responsible for the difference. The extent of HAA concentration polarization is exponentially dependent on the water permeation flux which results from the severe concentrative ICP caused by high water flux for the AL-DS orientation [39].

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

For reverse draw solute permeation, as the DS NaCl concentration increased, the measured reverse draw solute flux increased for both orientations (Fig. 4). However, the reverse draw solute flux for the AL-FW

Fig. 4. Experimental results of the NaCl flux 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.

6

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

orientation is much lower than that for the AL-DS orientation. For instance, when the DS NaCl concentration was 1 mol/L, the reverse draw solute fluxes were 0.21 and 0.55 mol/(m2h) for the AL-FW and AL-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 calculated using Eq. (3). The results demonstrated that the rejection ratio for each HAA showed a monotonous increase with the increase of DS NaCl concentration for the AL-FW orientation (Fig. 5). This is due to the more enhancement of the water permeation flux than the HAA flux with the increase of the DS NaCl concentration (Eq. (3)). In contrast, the HAA rejection ratios could reach their maximum under a medium osmotic pressure difference for the AL-DS orientation. The increase in water permeation flux was faster than the HAA flux when the DS NaCl concentration was relatively low, while the increase in water permeation flux was slower when the DS NaCl concentration was relatively high. Therefore, HAA rejection increased initially with the increase of water flux due to the enhanced dilution effect and a further increase in water flux resulted in a reduced rejection due to very severe ICP of

HAAs. This convex profile was also observed in other studies for boric acid removal by FO [19]. The rejection ratios for MCAA and MBAA were 94.6% and 95.1%, respectively, and that for the rest of HAAs were higher than 97.0% for the AL-FW orientation, with a DS NaCl concentration at 1 mol/L. This performance is comparable to that of nanofiltration (NF) and reverse osmosis (RO). Agus and Sedlak reported that the HAA rejection ratios were between 56% and 100% by NF, and between 86% and 94% by RO depending on the applied pressures [38]. If all FO, NF and RO reject solute by the same mechanism, the FO membrane 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 to the HAA removal in FO might be charge repulsion [19]. HAAs were highly rejected by FO, NF or RO because HAAs dissociate and are negatively charged unless the pH is very low (≤3) (Table 1). Most FO, NF and RO membrane surfaces are also negatively charged at circum-neutral pH ranges. Under the DS NaCl concentration of 1 mol/L, the rejection ratios for all the respective HAAs ranged from 73.8% to 89.1% for the AL-DS orientation, which were high but lower than those for the AL-FW orientation. It is because of the much higher HAA flux, albeit relatively higher water permeation flux for the AL-DS orientation. Based on current FO membrane technology, the AL-DS orientation may not be preferred due to its lower removal efficiency of pollutants and higher membrane fouling potential [14,19]. 3.4. HAA rejection and reverse draw solute flux prediction

Fig. 5. The HAA rejection ratios obtained by experiments and predicted by the model as a function of the DS NaCl concentration for the AL-FS orientation (top) and the AL-DS orientation (bottom).

With the S value obtained, the mass transfer coefficient for each HAA (Km,HAA) can be calculated by the ratio of respective diffusion coefficient listed in Table 1 to the S value (Km = D/S). In this study, the average of the two values of the HAA permeability coefficient (BHAA) was used for the prediction of the rejection ratios of the HAAs (Table 2). Given the water permeability coefficient (A), the HAA permeability coefficient (BHAA) and the HAA mass transfer coefficient (Km,HAA), the rejection ratio for each HAA can be predicted by using Eq. (3) in combination with Eq. (6) for the AL-FW orientation or Eq. (9) for the ALDS orientation. The modeling results matched well with the experimentally-obtained rejection ratios when the FO was operated in the AL-FW orientation mode (Fig. 5). However the modeling results over-estimated the rejection ratios when the FO was operated in the AL-DS mode (for which Km,HAA took an important role). There are some possible reasons for the over-estimation. As described earlier, the under-estimation of the S value would be partly responsible for the over-estimation of the Km,HAA value and in turn the HAA rejection ratios. If it were the case, the rejection ratio over-estimation would be more significant for MCAA and MBAA, because both of them have smaller molecular weight and thus larger diffusion coefficient than others. Recently, studies [40] showed that the permeability coefficient might be better quantified by Km = δD/S, where δ is the so-called constrictivity factor. The constrictivity factors for sodium ion, chloride ion and the HAAs may not be identical, which was also partly responsible for the over-estimation of the HAA rejection ratios. However, because Na+, Cl− and the HAA species are all small molecules, such effect of δ would be very limited. Another probable reason for the overestimation of the rejection ratio was that the actual diffusivity for each HAA was lower than that shown in Table 1. In this study, the molecular size of each HAA was calculated by using the Wilke–Chang equation, which was then utilized for the calculation of the diffusivity by using the Einstein–Stokes equation. More accurate determination of the diffusivity for each HAA in the FO membrane support layer may to some extent reduce the difference between the predicted and the experimental results for the HAA rejection ratios. Moreover, FO membranes were recently found containing pores of sub-nm sizes in the active layer [41–43]. As such, somewhat in contradictory to the assumption of the solution-diffusion model that the solute (both NaCl and HAA) was

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

7

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

transferred through the FO membrane active layer by diffusion only, mass transfer of HAA through the active layer by convection was also possible especially when the water permeation flux was high. The contribution by this mechanism however needs more investigation. Given the water permeability coefficient (A) and the NaCl permeability coefficient (BNaCl), the reverse draw solute (NaCl) flux can be predicted by using Eq. (12) [23,27].

J NaCl B ¼ NaCl Jw AnRg T

ð12Þ

where n is the number of dissolved species of the draw solute (2 for NaCl), Rg the universal gas constant, and T the absolute temperature. An assumption that underlies Eq. (12) is that the calculation of the osmotic pressure follows the van't Hoff equation [22,23,27]. Eq. (12) shows that the ratio of reverse draw solute flux to the water flux is independent of the DS concentration or the extent of the concentration polarization [22,23,27]. As shown in Fig. 6, the experimental results confirmed that the ratio of reverse draw solute to the water flux was a constant for both the AL-FW and AL-DS orientations. The desire for higher water flux and the concurrent higher reverse draw solute flux are in contradiction for the current FO process. Compared with the ALFW orientation, both water flux and reverse draw solute flux for the AL-DS orientation are higher because of the absence of dilutive ICP of the draw solute. According to Eq. (12), the best way to reduce the reverse flux of the draw solute is to either decrease the NaCl permeability coefficient (BNaCl) or increase the water permeability coefficient (A). By far, a number of studies were devoted primarily to the increase of the water permeability coefficient by incorporation into the FO membrane active layer of highly water-permeable aquaporin-based proteoliposomes [44] or nanomaterials (e.g. zeolite particles or nanotube) [45,46]. The reverse draw solute flux can be also predicted by the solutiondiffusion model (Eq. (5) for the AL-FW orientation and Eq. (8) for the AL-DS orientation) by assuming that the NaCl concentrations in the feed and draw solutions are constant, given the experimentally determined 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). The agreement between the model prediction and the experimental data is evident (Fig. 7), with the slope of the correlation line equal to 1. It demonstrated that the solution-diffusion model combined with concentration polarization can be used for the prediction of the reverse draw solute permeation in a multicomponent system.

Fig. 7. A comparison of the experimental results and model predictions of the reverse draw solute flux for the AL-FS orientation (top) and the AL-DS orientation mode (bottom). The fitting line (slope = 1) represents a perfect agreement.

4. Conclusions In this paper, the water flux, solute flux and reverse draw solute flux were systematically investigated: (1) For both membrane orientations, the HAA fluxes and reverse draw solution permeation flux increased with the increase of the DS concentration. The smaller the molecular weight of an HAA, the higher the HAA flux through the FO membrane. The HAA fluxes increased acceleratedly for the AL-DS orientation while it increased deceleratedly for the AL-FW orientation with the increase of the DS concentration. The reverse draw solute flux for the AL-FW orientation was much lower than that for the AL-DS orientation. (2) The rejection ratio for each HAA showed a monotonous increase with the increase of DS concentration for the AL-FW orientation. In contrast, the HAA rejection ratios could reach their maximum under a medium osmotic pressure difference for the AL-DS orientation. The rejection ratios for HAAs were more than 94.6% for the AL-FW orientation and ranged from 73.8% to 89.1% for the AL-DS orientation under the DS NaCl concentration of 1 mol/L. (3) The solution-diffusion model combined with the concentration polarization model was able to predict the bi-directional mass transfer in FO systems. However the modeling results overestimated 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 coefficient for internal concentration polarization. More precise determination of the HAA diffusivity in the FO membrane support layer and quantification of the contribution by convection are required.

Nomenclature A 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 dh Km resistivity for diffusion within the porous support layer (m/s), K m 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 ratio Rg universal gas constant (J/(K mol)) S intrinsic parameter related to support layer structure (m), S = tτ / ε Sh the Sherwood number of the appropriate flow regime in the rectangular flow channel T 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 symbols Δt 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 layer τ tortuosity of the support layer φ association parameter of the solvent (2.6 for water)

Subscripts D draw solution F feed water HAA each of the haloacetic acids NaCl draw solute Re receiver of the diffusion process So source of the diffusion process w water

Acknowledgments The authors acknowledge the financial support provided by the Chinese Natural Science Foundation (No. 51278268, No. 51290284) and the Beijing 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: opportunities 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: yesterday, 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 of forward 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 of RO 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 osmosis membrane bioreactor: a low fouling alternative to MBR processes, Desalination 239 (2009) 10–21. [8] Y.F. Xie, Disinfection Byproducts in Drinking Water: Formation, Analysis, and Control, Lewis Publishers, Boca Raton, Florida, USA, 2004. [9] C. Ort, M.G. Lawrence, J. Rieckermann, A. Joss, Sampling for pharmaceuticals and personal care products (PPCPs) and illicit drugs in wastewater systems: are your conclusions valid? A critical review, Environ. Sci. Technol. 44 (2010) 6024–6035. [10] G.A. Loraine, M.E. Pettigrove, Seasonal variations in concentrations of pharmaceuticals and personal care products in drinking water and reclaimed wastewater in southern 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 system plumes, Ontario, Canada, Environ. Sci. Technol. 42 (2008) 2805–2811. [12] R. Valladares Linares, V. Yangali-Quintanilla, Z. Li, G. Amy, Rejection of micropollutants 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 hydrophobic trace organic contaminants by forward osmosis and reverse osmosis, Water Res. 46 (2012) 2683–2692. [14] M. Xie, W.E. Price, L.D. Nghiem, Rejection of pharmaceutically active compounds by forward osmosis: role of solution pH and membrane 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, Removal of 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, Removal of natural steroid hormones from wastewater 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 osmotic dilution process for simultaneous desalination and purification of impaired water, J. Membr. Sci. 362 (2010) 417–426. [18] N.T. Hancock, P. Xu, D.M. Heil, C. Bellona, T.Y. Cath, Comprehensive bench- and pilot-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 membrane 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 membrane processes, 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 in forward osmosis membrane: effects of reverse draw solution permeation, Desalination 314 (2013) 115–123. [22] W.A. Phillip, J.S. Yong, M. Elimelech, Reverse draw solute permeation in forward osmosis: modeling and experiments, Environ. Sci. Technol. 44 (2010) 5170–5176. [23] J.S. Yong, W.A. Phillip, M. Elimelech, Coupled reverse draw solute permeation and water 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 nine haloacetic acids in wastewater effluents using ultra-performance liquid chromatography–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 concentration 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 concentration polarization and fouling on flux behavior of forward osmosis membranes during humic acid filtration, J. Membr. Sci. 354 (2010) 123–133. [28] B.S. Chanukya, S. Patil, N.K. Rastogi, Influence of concentration polarization on flux behavior in forward osmosis during desalination using ammonium bicarbonate, Desalination 312 (2013) 39–44. [29] C.Y. Tang, Q. She, W.C.L. Lay, R. Wang, R. Field, A.G. Fane, Modeling double-skinned FO 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, Standard methodology for evaluating membrane performance in osmotically driven membrane processes, Desalination 312 (2013) 31–38. [31] E.L. Cussler, Diffusion: Mass Transfer in Fluid Systems, Cambridge University Press, Cambridge, UK, 2009.

F. Kong et al. / Desalination 341 (2014) 1–9 [32] E. Drazevic, S. Bason, K. Kosutic, V. Freger, Enhanced partitioning and transport of phenolic 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 rejection of organic compounds by polyamide composite membranes, Environ. Sci. Technol. 40 (2006) 7023–7028. [34] N. Nestle, R. Kimmich, Concentration-dependent diffusion coefficients and sorption isotherms. 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 constant membrane structure parameter in forward osmosis processes, J. Membr. Sci. 375 (2011) 241–248. [36] V. Vitagliano, P.A. Lyons, Diffusion coefficients for aqueous solutions of sodium chloride 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 osmosis 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 of forward osmosis membrane process: effect of membrane orientation and flow direction of feed and draw solutions, Desalination 277 (2011) 83–91. [40] S. Zhao, L. Zou, Relating solution physicochemical properties to internal concentration polarization in forward osmosis, J. Membr. Sci. 379 (2011) 459–467.

9

[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: influence 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 of FO/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 forward 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 on separation performance of aquaporin-based proteoliposomes: implications for seawater 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 membranes: 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 functionalized multi-walled carbon nanotube immobilized hollow fiber membranes for enhanced performance in forward osmosis process, J. Membr. Sci. 446 (2013) 244–254.