Modeling salt accumulation in osmotic membrane bioreactors: Implications for FO membrane selection and system operation

Journal of Membrane Science 366 (2011) 314–324

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Modeling salt accumulation in osmotic membrane bioreactors: Implications for FO membrane selection and system operation Dezhong Xiao a , Chuyang Y. Tang a,b,∗ , Jinsong Zhang b , Winson C.L. Lay a,b , Rong Wang a,b , Anthony G. Fane a,b a b

School of Civil and Environmental Engineering, Nanyang Technological University, Singapore 639798, Singapore Singapore Membrane Technology Centre, Nanyang Technological University, Singapore 639798, Singapore

a r t i c l e

i n f o

Article history: Received 21 July 2010 Received in revised form 7 October 2010 Accepted 9 October 2010 Available online 15 October 2010 Keywords: Salt accumulation Forward osmosis (FO) Osmotic membrane bioreactor (OMBR) Solute reverse diffusion Volumetric concentration factor

a b s t r a c t Novel osmotic membrane bioreactors (OMBRs) have been recently reported in the literature. An OMBR uses a dense salt-rejecting forward osmosis (FO) membrane, which exhibits high retention of organic matter and various other contaminants. Meanwhile, the high rejection nature also leads to the accumulation of salts in the bioreactor, which can adversely affect the biological activities as well as the FO water flux. A salt accumulation model is developed in the current study. Our model suggests that both the bioreactor salt concentration and the FO water flux are controlled by membrane properties (water permeability A, salt permeability B, mass transfer coefficient Km , and membrane orientation relative to the draw solution) and the OMBR operational conditions (salt concentration of the influent wastewater, draw solution concentration, hydraulic retention time (HRT), and sludge retention time (SRT)). The salt accumulation is contributed by both the influent wastewater and the reverse diffusion of solutes from the draw solution, and is directly proportional to the volumetric concentration factor (i.e., the SRT/HRT ratio). The relative importance of reverse diffusion over contribution from influent solutes is governed by the membrane selectivity. For a relatively selective membrane (B/A  the osmotic pressure of the influent water), solute reverse diffusion has negligible effect on OMBR performance. In contrast, the salt accumulation and FO water flux reduction are governed by reverse diffusion for B/A greater than the osmotic pressure of the influent water. The current study reveals the critical importance of the B/A ratio and HRT/SRT ratio for optimized OMBR operation. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Forward osmosis (FO) has emerged as an alternative membrane process to the conventional pressure-driven membrane processes in the recent years [1,2]. Unlike the pressure-driven processes that rely on a hydraulic pressure to “push” water through a membrane, the FO process is driven by an osmotic pressure difference across a semi-permeable dense membrane that retains solutes but allows water to pass through the membrane [2,3]. Pure water permeates through the FO membrane spontaneously from the low-osmotic-pressure solution (i.e., the feed water (FW)) to the high-osmotic-pressure solution (i.e., the draw solution (DS)) without the need of an externally applied pressure [2,3]. The elimination of the hydraulic pressure in FO means that a significantly lower amount of prime energy (i.e., electricity) is required to run an

∗ Corresponding author at: School of Civil and Environmental Engineering, Nanyang Technological University, Blk N1, Rm #1b-35, Singapore 639798, Singapore. Tel.: +65 6790 5267; fax: +65 6791 0676. E-mail address: [email protected] (C.Y. Tang). 0376-7388/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2010.10.023

FO process compared to that used for a typical reverse osmosis (RO) process. Thus, FO can be highly attractive where a high pressure osmotic DS is naturally available (e.g., seawater) or if the DS can be regenerated cost-effectively. On the other hand, like RO membranes, dense FO membranes have nearly complete rejections against a wide range of contaminants including dissolved salts and organic matter [4]. Consequently, FO has found many potential applications in water purification [5], wastewater treatment [6,7], seawater desalination [8] and osmotic power generation (using a derivative process know as pressure retarded osmosis, see Refs. [3,9,10]). A novel type of FO-based membrane bioreactor (MBR) has been recently reported in the literature [7,11–13]. Unlike the conventional MBRs that typically use porous microfiltration (MF) or ultrafiltration (UF) membranes to retain biomass, the osmotic MBR (OMBR) system uses an FO membrane in the bioreactor [12]. With a high-concentration draw solution, pure water is extracted from the mixed liquor and diffuses across the membrane into the DS, while contaminants from the feed wastewater are nearly completely rejected. If wastewater reclamation is needed, pure water can be obtained while the draw solution is also regenerated

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

via a downstream process such as thermal distillation [8]. Where a low grade heat source is available, this treatment scheme can be potentially attractive [14]. Compared to the conventional MBRs, the OMBR offers much higher rejection (dense FO membranes versus porous MF/UF membranes) at a lower hydraulic pressure [4,7]. The better retention of organic matter by FO may lead to 1) improved biological degradation of recalcitrant organics due to the prolonged retention time for organic matter, 2) reduced dissolved organic content in treated effluent, and 3) a likely enhanced removal for micropollutants [4,7]. Several studies have also suggested that FO is a low-fouling alternative to pressuredriven membranes due to the lack of hydraulic pressure [4,12], though more systematic studies are still required to investigate the underlying fouling mechanism(s) [15,16]. Despite the many advantages offered by OMBRs compared to conventional MF- or UF-based MBRs, the high salt rejection property of the FO membrane also introduces a unique new challenge – the dissolved salts rejected by the FO membrane will accumulate inside the bioreactor. In addition, solute will also diffuse from the high concentration DS into the bioreactor under the concentration gradient across the FO membrane [15,17]. This will further enhance the salt accumulation in the bioreactor. The elevated salt concentration in the bioreactor not only reduces the FO permeate flux (due to reduced the effective osmotic driving force and increased internal concentration polarization (see Refs. [15,18] and Section 2)) but also may adversely impact the physical and biological activities in the bioreactor [13]. Thus, a systematic study on the salt accumulation in the novel OMBR system is warranted. This study aims to investigate the salt accumulation behavior in OMBR and its effect on FO performance. A theoretical model is developed based on the internal concentration polarization theory as well as the solute mass balance for the bioreactor. Model verification was performed via independent bench scale FO experiments. The salt accumulation model was then applied to simulate the OMBR performance under various conditions, and the effects of membrane properties and operational conditions were systematically investigated to reveal the consequent constrains on membrane selection as well as reactor design and operation. This study focuses on the effect of salt accumulation on the FO permeate flux behavior. On the other hand, membrane fouling [15] and biological activities [13] are beyond the scope of the current study. To the authors’ best knowledge, this is the first systematic study to model the salt accumulation in OMBR and its implications for OMBR design and operation. While the model is developed primarily for OMBR, the basic principles applied in the model can also be extended to other osmotic reactors. 2. Model development 2.1. Internal concentration polarization and modeling FO water flux and solute reverse transport

inside the support layer to be much greater than the bulk feed water concentration (concentrative ICP, refer to Fig. A1 in Appendix A), which results in reduced effective driving force (the osmotic pressure difference across the active rejection layer) as well as lower permeate flux. Similar reduction in FO water flux also happens for the active-layer-facing-FW (AL-FW) orientation due to the dilution of draw solution inside of the porous support (dilutive ICP, refer to Fig. A1 in Appendix A). The ICP phenomenon was first modeled by Lee et al. [19] in the context of pressure retarded osmosis. A simplified version of ICP model was later developed by Loeb et al. [1] for forward osmosis: Jv = Km ln and Jv = Km ln

 A

− Jv + B Afeed + B draw



 A  draw + B Afeed + Jv + B

(AL-DS, concentrative ICP)

(AL-FW, diluted ICP)

(1)

(2)

where Jv is the volumetric flux of water; Km is the mass transfer coefficient for the membrane porous support layer; A and B are the water permeability and solute permeability of the rejection layer, respectively; and draw and feed are the osmotic pressure of the draw solution and that of the feed water, respectively. It has been demonstrated that the AL-FW orientation suffers greater ICP level and consequently its water flux is lower than that of the AL-DS orientation under identical testing conditions [15,22,23]. In addition, ICP is more severe for higher draw solution concentrations, higher feed water concentrations, as well as lower mass transfer coefficient values [15,22,23]. The mass transfer coefficient Km in Eqs. (1) and (2) can be related to the solute diffusion coefficient D and the membrane structure parameter S by [1,15]: Km =

D S

(3)

where the membrane structural parameter S is a property of the FO support layer, and it is defined as the product of membrane support layer thickness (l) and tortuosity () over its porosity (ε) [1,15]: S=

l ε

(4)

Loeb’s ICP model (Eqs. (1) and (2)) has been widely adopted by other research groups and has been verified experimentally for a wide range of membranes including both cellulose acetate asymmetric membranes and thin film composite membranes [15,18,20,22,23]. For high draw solution concentration and high feed water concentration, the osmotic pressure terms (Adraw and Afeed ) in Eqs. (1) and (2) can be orders of magnitude higher than Jv and B. Under such special conditions, the water flux in the two orientations (AL-DS and AL-FW) become nearly identical and it is determined by [1,22]: Jv = Km ln

Numerous studies have reported on the experimental testing and modeling of FO water flux [15,19–21]. These studies revealed that the typical FO flux is significantly lower than the ideal flux determined from the apparent driving force (osmotic pressure difference between the DS and FW) using the solution-diffusion model alone. Such non-ideal flux behavior is caused by concentration polarization of solutes inside the porous support layer (also known as the internal concentration polarization or ICP) [15,18,19]. For example, when the active rejection layer is oriented towards the draw solution (the AL-DS membrane orientation), solutes from the feed water is accumulated in the porous support layer due to the high retention nature of the FO rejection layer. Meanwhile, some solutes also diffuse through the rejection layer from the high concentration DS into the support. This causes the solute concentration

315



draw

feed



(both AL-DS and AL-FW)

(5)

Similar to the water flux Jv , the solute flux Js is also strongly affected by ICP [15,17]. Tang and coworkers [15,22,23] demonstrated that Js is directly proportional to Jv according to the following equation: B Js = Jv AˇRg T

(both AL-DS and AL-FW)

(6)

where ˇ is the van’t Hoff coefficient, Rg is the universal gas constant, and T is the absolute temperature. In Eq. (6), the Js /Jv ratio is largely determined by B/A (in unit of Pa). The B/A ratio is an important property of the membrane rejection layer. This ratio can be considered as a constant for a given rejection layer and solute type at fixed temperature. In general, a lower B/A ratio (i.e., a more

316

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

tively by: HRT =

Vr Jv Am + Qs

(10)

Vr Qs

(11)

and SRT =

Thus, Eq. (9) can be re-written as:

 1

dCml C B = in + dt HRT AˇRg T or dCml 1 = dt HRT Fig. 1. Mass balance in an OMBR system.

selective rejection layer) is preferred in order to reduce the solute reverse transport across the FO membrane [22,23]. The term Js /Jv or (B/A)/(ˇRg T) has a concentration unit, and it can be treated as an effective concentration arising from the solute reverse diffusion (note that Js /Jv for an RO membrane is actually the permeate concentration). Alternatively, the B/A ratio has a unit in Pa, and it can be viewed as an effective osmotic pressure term arising from the solute reverse diffusion. 2.2. Modeling OMBR performance 2.2.1. Modeling time-dependent OMBR performance According to the FO water and solute flux models presented in Section 2.1, both Jv and Js are strongly dependent on 1) the FO membrane properties (A and B for the rejection layer; S for the support layer), 2) the draw solution concentration (Cdraw ) and type, and 3) feed water concentration (Cfeed ) and type. Consider a typical submerged OMBR (Fig. 1), the FO feed water solute concentration is identical to the solute concentration in the mixed liquor (Cml ). Due to the use of FO membranes in OMBR, salt accumulation will occur in the bioreactor due to two mechanisms: 1) retention of solutes from the feed water, and 2) reverse diffusion of solutes from the high concentration draw solution. As a result, Cml can be significantly higher than the solute concentration in the influent sewage Cin . The high salt concentration in the bioreactor may adversely affect both biological activities [13] and the FO flux performance (Eqs. (1) and (2)). Thus, it is important to understand the dependence of Cml on OMBR operation and FO membrane properties. The mixed liquor salt concentration Cml can be determined as a function of time t via solute mass balance in the OMBR (Fig. 1): dC Vr ml dt accumulation in OMBR

=

(water mass balance)

(8)

Substituting Eqs. (6) and (8) into Eq. (7), we have: Vr

dCml B Jv Am − Qs Cml = (Jv Am + Qs )Cin + dt AˇRg T

(9)

For the OMBR system shown in Fig. 1, the hydraulic retention time ( HRT ) and sludge retention time ( SRT ) are defined respec-



1

−

Cml SRT



SRT

Cml +

(12)

B AˇRg T

 (13)



in +

B A



−



1 SRT

ml +

B A



(14)

2.2.2. Modeling steady state OMBR performance At steady state, dml /dt term in Eq. (14) becomes 0, and the steady-state salt concentration and osmotic pressure in the OMBR can be determined by: steady

Cml

ml

where Vr is the bioreactor volume; Qin and Qs are the respective volumetric flow of the influent sewage and the waste sludge; and Am is the FO membrane area. Similarly, a mass balance equation can be written for water in the OMBR:

B AˇRg T

−

In this case, the B/A is the effective osmotic pressure related to salt reverse diffusion. The time-dependent salt accumulation in the OMBR can be determined by coupling Eq. (14) to the FO water flux model (Eq. (1) or Eq. (2)), where feed = ml and the hydraulic retention time given by Eq. (10). It is worthwhile to note that the draw solution concentration is assumed to be constant for a typical OMBR operation, which can be achieved through a separate draw solution regeneration loop. Where draw solution regeneration is not performed, Cdraw will decrease continuously due to 1) the volumetric dilution effect (i.e., the DS is diluted by the permeate water) and 2) the solute reverse diffusion from DS into the bioreactor [15]. The modeling of Cdraw is presented in Appendix B for the case of no draw solution regeneration. In the current study, it is assumed that the DS concentration is maintained constant throughout the OMBR operation.

−

solutes to the sludge

Qin = Jv Am + Qs

dml 1 = dt HRT

and

(solute mass balance) (7)

Cin +

SRT



In Eq. (13), the term B/AˇRg T is the equivalent concentration (with a unit of M) due to the reverse diffusion of solutes from the DS into the OMBR. Clearly, the salt accumulation in the OMBR is contributed by both the solute inflow from the influent and that due to FO reverse diffusion. Eq. (13) can also be presented in terms of osmotic pressure, recognizing that the solute concentration C and osmotic pressure  are related via  = ˇRg TC. Thus,

Qin Cin + Js Am solutes from back diffusion solutes from influent Qs Cml



HRT

1

−

steady

=

SRT Cin + HRT

=

SRT in + HRT





SRT −1 HRT





SRT −1 HRT

B AˇRg T

(15)

B A

(16)

From Eqs. (15) and (16), it is clear that the steady state salt concentration is superimposed by two independent sources: 1) accumulation of solutes from the influent sewage, and 2) the reverse diffused solutes from the DS. The  SRT / HRT ratio is in fact a volumetric concentration factor, as this ratio is identical to Qin /Qs (Fig. 1). Eq. (16) can be substituted into Eq. (1) (or Eq. (2)) to steady determine the steady-state FO water flux Jv . For typical OMBR operation, both Jv /A and B/A are negligible with respect to draw or steady ml , Jv can be approximated by: steady

Jv

= Km ln



draw (SRT /HRT )(in + B/A)



(17)

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

or steady

Jv

4. Model verification

   draw

= Km ln

in + B/A

 − ln

SRT HRT

 (18)

Thus, the steady-state FO water flux is governed by both the membrane properties (Km and B/A ratio) and the OMBR operational conditions (concentrations of sewage and draw solution) as well as the  SRT / HRT ratio. The ideal FO membrane for OMBR applications shall have a high mass transfer coefficient and low solute reverse transport (low B/A ratio). It is also important to optimize the  SRT / HRT ratio to avoid severe salt accumulation in the bioreactor but still maintain satisfactory treatment efficiency. It is also worthwhile to note that the HRT is not a constant during a typical OMBR operation. As the FO water flux decreases over time, the HRT becomes longer according to Eq. (10). Substituting Eqs. (10) and (11) into Eq. (18), one can obtain an approximation for the steady state water flux using the following equation:

steady

Jv

317

   draw

= Km ln

in + B/A

 − ln

steady

Jv

Qs

Am

 (19)

This water flux value can then be used to determine the HRT and OMBR reactor concentration at steady state according to Eqs. (10) and (15), respectively.

Fig. 2 shows the experimental FO water flux as a function of filtration time for different draw solution concentrations. For both membrane orientations (AL-DS and AL-FW), greater initial water flux was observed at higher draw solution concentrations. However, the AL-DS orientation generally gave a higher water flux compared to the AL-FW orientation at identical Cdraw . For example, the initial flux in AL-FW was ∼28 L/m2 h for a 4 M NaCl draw solution, roughly 50% of that in the alternative orientation. The lower water flux in AL-FW can be attributed to the more severe ICP level in this orientation due to the dilution of draw solution inside the FO support layer (dilutive ICP) [15,20]. The initial flux levels can be modeled by Eq. (1) for AL-DS or Eq. (2) for ALFW. In general, the model predictions agreed very well with the experimental initial flux (Fig. 2), suggesting that the ICP equations (Eqs. (1) and (2)) are adequate for modeling the FO water flux. The experimental water flux declined at longer filtration time. Such flux decline in the absence of membrane fouling can also be modeled by the ICP equations by taking into account the increased feed water concentration as well as the dilution of the draw solution based on mass balance considerations (Appendix B). The FO flux was determined as a function of time by coupling the ICP equations for water flux (Eq. (1) or (2)) with solute reverse diffusion equation for solute flux (Eq. (6)). Once again, the modeled water flux agreed reasonably well with the experimental data, with the flux decline trends correctly predicted under all testing conditions.

3. Experimental methods

a

60 AL-DS

Water Flux (L/m2hr)

50 40 30 20 10 4M 0

0

2M

100

1M 200

0.5M 300

400

Time (min)

b

35 AL-FW

30

Water Flux (L/m2hr)

Bench scale FO experiments were performed to verify the salt accumulation model presented in Section 2. Reagent grade sodium chloride was used as the model solute throughout the current study. Both draw solution and feed water were prepared by dissolving NaCl in MilliQ water (Millipore water system, Billerica, MA) with a resistivity of 18.2 M cm. The flat sheet FO membrane used in the current study was obtained from Hydration Technologies Inc. (Hydrowell filter, HTI, Albany, OR). Its membrane properties have been reported by several groups [15,24]. Briefly, the HTI membrane has a dense cellulose triacetate rejection layer supported by a woven fabric support. The overall membrane thickness is only ∼30–50 ␮m, which is presumably optimized to reduce ICP (see Section 2.1 and Refs. [1,15]). The water and solute transport coefficients A and B were determined by testing this membrane in the reverse osmosis mode [15]. The A and B values obtained in the current study (A = 2.12 × 10−12 m/s Pa, and B = 1.6 × 10−7 m/s) agreed reasonably well with previously reported values [15]. The experimental procedures for FO testing have been fully described by Tang et al. [15] and are briefly summarized here. All the FO tests were performed in a cross-flow FO testing cell (Model C10-T, Nitto Denko, Japan) with an active membrane area of 60 cm2 . Both the feed water and the draw solution were circulated in closed loops. This can be considered as a special case of the OMBR system (in Fig. 1) with no sludge wasting or draw solution regeneration (Appendix B). Both the AL-DS and AL-FW orientations were evaluated, although the AL-DS orientation is unlikely to be practical for OMBR applications due to the high fouling environment [15]. The permeate flux was determined by measuring the weight change of the feed solution using a digital balance connected to a computer data logging system. For all the FO tests, the original feed water contained 10 mM NaCl and the subsequent change in the feed tank concentration was monitored using conductivity measurements. The draw solution concentration used in this study ranged from 0.3 to 5 M.

25 20 15 10 5 0

4M 0

100

2M

1M 200

0.5M 300

400

Time (min) Fig. 2. Experimental FO water flux as a function of filtration time for different membrane orientations. (a) AL-DS; (b) AL-FW. The fitting lines (based on ICP equations and solute accumulation equations) agree reasonably well with the experimental results. Testing conditions: HTI hydrowell membrane; membrane area of 60 cm2 ; 10 mM NaCl as initial FW concentration; 23 ◦ C.

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

5. Model simulations

a

This section applies the salt accumulation model and ICP model presented in Section 2 to simulate the effect of various parameters on OMBR flux performance. The time dependent water flux and reactor concentration are determined by solving the ICP equations (Eq. (1) for AL-DS or Eq. (2) for AL-FW) coupled with OMBR salt concentration equation (Eq. (13) for concentration or Eq. (14) for osmotic pressure) numerically. The effect of membrane properties (A, B, Km , and orientation) and operational conditions (Cdraw , Cin , HRT, and SRT) on the salt accumulation in an OMBR system and the membrane water flux performance is investigated by varying the selected parameter(s) while keeping the others at their respective reference values. Unless otherwise specified, the following reference conditions are applicable:

Concentraon (mM NaCl )

318

Reference membrane: TFC2 (properties listed in Table 1) Membrane orientation: AL-FW Influent wastewater salt concentration: 10 mM NaCl Draw solution concentration: 1 M NaCl Solids retention time (SRT): 10 days Initial hydraulic retention time (HRT0 ): 8 h.

600

HTI TFC2 TFC4

500 400 300 200 100 0

120

240

360

480

600

720

600

720

Time (hr)

b

18

HTI TFC2 TFC4

16

The AL-FW orientation is chosen due to its lower fouling tendency [15], which is more likely to be adopted in actual OMBR implementation (Ref. [12] and also see discussion in Section 5.1). The influent salt concentration is typical for domestic sewage, while the SRT and HRT are selected based on expected typical OMBR operation [13].

TFC1 TFC3 TFC5

0

Water flux (L/m2 hr)

-

700

14

TFC1 TFC3 TFC5

12 10 8 6 4 2 0 0

120

240

360

480

Time (hr) Fig. 3. Effect of B/A ratio on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h; SRT: 10 days. Membrane properties are shown in Table 1.

5.1. Effect of FO membrane properties As discussed in Section 2, membrane properties such as A, B, and Km (or the support structural parameter S) have important effect on OMBR solute accumulation (Eq. (13)) as well as FO water flux (Eqs. (1) and (2)). In particular, Eq. (13) suggests that the B/A ratio plays a critical role in the solute accumulation in OMBR, as this term is related to the effective concentration reverse diffused from the DS into the reactor (Section 2 and also Ref. [15]). To systematically study the influence of the B/A ratio on OMBR performance, two types of FO membranes have been simulated in the current study (Table 1). The first type is the commercially available HTI FO membrane based on cellulose triacetate chemistry (see Section 3). The second type has properties similar to thin film composite (TFC) polyamide FO membranes (see Refs. [22,23,25]). The B/A ratio of the TFC series (TFC-1 to TFC-4) is systematically increased from 0.492 to 492 kPa, compared to the influent sewage osmotic pressure in of 49.2 kPa (for 0.01 M NaCl). The mass transfer coefficient Km is set at 2.5 × 10−6 m/s (50% of the commercial HTI FO membrane), which can be easily achieved with the current TFC FO fabrication technologies [22,23]. In addition, the selectivity properties (A and B values) of the TFC type membranes (TFC-1 to TFC-4) in Table 1 have been chosen to allow these membranes to achieve an identical initial water flux (and thus the same initial HRT) compared to the commercial HTI membrane (15.3 L/m2 h under the reference conditions: 1 M NaCl as DS, 10 mM NaCl FW, and AL-FW orien-

tation). This ensures that the simulated differences in the OMBR performance are solely caused by the B/A ratio, but not due to the difference in the initial HRT (also see Section 5.3). An additional TFC membrane (TFC-5) is also simulated. This membrane has B/A ratio and Km values identical to the respective values for TFC-2. The water permeability of TFC-5 is ∼60% of that for TFC-2. Consequently, TFC-5 has a lower initial water flux compared to the rest of the membranes. The initial HRT for this membrane is set be identical to the other membranes, which can be achieved using a larger membrane area. Fig. 3(a) shows the effect of membrane selectivity (in this case, the B/A ratio) on the solute accumulation in the OMBR. For all the membranes simulated, the solute concentration in the OMBR increased significantly with time. A steady-state reactor concentration is reached when the solute entering into the OMBR is balanced by the salt discharge in the continuous sludge wasting (Eq. (7)). Corresponding to the steady-state solute concentration in the OMBR, a stable water flux is also reached (Fig. 3(b)). It is also worthwhile to mention that the hydraulic retention time is not a constant for the membranes in Fig. 3 as a result of reduced water flux during the startup stage. As water flux decreases over time, HRT will keep increasing until it reaches the steady-state condition (Fig. 4). While the increase in HRT is mild for most cases, TFC-4 membrane expe-

Table 1 Membrane properties used for OMBR simulation. FO type HTI TFC-1 TFC-2 TFC-3 TFC-4 TFC-5

A (m/s Pa) −12

2.12 × 10 5.03 × 10−12 5.05 × 10−12 5.28 × 10−12 9.58 × 10−12 3.50 × 10−12

B (m/s)

B/A (kPa) −7

1.6 × 10 2.48 × 10−9 2.49 × 10−8 2.60 × 10−7 4.72 × 10−6 1.72 × 10−8

75.5 0.492 4.92 49.2 492 4.92

(B/A)/in (−) 1.53 0.01 0.1 1 10 0.1

Km (m/s) −6

5.0 × 10 2.5 × 10−6 2.5 × 10−6 2.5 × 10−6 2.5 × 10−6 2.5 × 10−6

Initial flux (L/m2 h) 15.3 15.3 15.3 15.3 15.3 13.4

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

Steady state water flux (L/m2 hr)

12

35

HRT (hr)

30 25 20

TFC-1

TFC-2

TFC-4

HTI

TFC-3

15 10 5 0 0

120

240

360

480

600

1000 water flux

10

800

8 600 6 400

4 reactor concentraon

200

2 0 0.0001

0.001

Fig. 4. Change of the hydraulic retention time of OMBR system with time. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h; SRT: 10 days. Membrane properties are shown in Table 1.

0.1

1

10

0

(B/A) / πin (-)

720

Time (hr)

0.01

Steady state reactor concentraon (mM NaCl)

40

319

Fig. 5. Effect of B/A ratio on steady state water flux and reactor concentration. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h; SRT: 10 days. The A and B combinations are chosen to keep the initial FO water flux of 15.3 L/m2 h for all simulations.

and riences a drastic change in HRT due to its severe membrane flux reduction. Hence, caution shall be exercised when applying the steady-state equations (Eqs. (15), (16), and (18)), as the steady-state HRT needs to be used instead of the initial HRT of 8 h. Rather, Eq. (19) may be used to determine the steady state water flux, which has already taken into consideration the changes in HRT over time. In Fig. 3(a), it is interesting to note that membranes with greater B/A values tend to suffer more severe salt accumulation. For example, the TFC-4 membrane with the highest B/A value of 492 kPa has the greatest concentration buildup in the reactor with a steady-state reactor concentration of nearly 0.7 M. In comparison, membranes with small B/A values (e.g., TFC-1, TFC-2, and TFC-5) steady have much less solute accumulation problem (Cml ∼0.2 M). As shown in Eqs. (15) and (16), the solute buildup in the OMBR is contributed by two independent source: 1) the volumetric concentration of the solutes from the influent (i.e., the in term) and 2) the solute reverse diffusion from the draw solution (i.e., the B/A term). Thus, a less selective membrane (larger B/A) will lead to a more severe solute accumulation and a more dramatic loss in driving force. Consequently, the least selective membrane (TFC-4) also has the least stable FO water flux, despite that its initial water flux is identical to the HTI membrane and the TFC membranes #1–3 (Fig. 3(b)). In addition to the poorer water flux performance for a less selective membrane, a greater reverse diffusion means more draw solutes need to be replenished if a close-circled draw solution regeneration is required, which can significantly increase the operational cost for OMBR operation. More importantly, the greater solute accumulation in the OMBR can also have detrimental effects on its biological activities [13]. Thus, it is crucial to minimize the effect of solute reverse diffusion on OMBR performance by improving the membrane selectivity (smaller B/A). A closer look at Fig. 3(a) and (b) reveals that TFC-1 and TFC-2 had nearly identical performance in terms of both solute accumulation and FO water flux. For both membranes, it turns out that B/A  in (Table 1). Additional simulation for membranes with identical initial flux (15.3 L/m2 h) reveals that the steady-state water flux and reactor concentration are dependent on B/A only if B/A > 0.1in (Fig. 5). When B/A < 0.1in , further decrease in B/A has no beneficial effects on OMBR performance. This is consistent with Eq. (16), which suggests that the salt accumulation is dominated by the influent concentration only when B/A  in , where the effect of B/A being negligible: steady

Cml

=

SRT Cin HRT

(B/A  in )

(20)

steady

ml

=

SRT in HRT

(B/A  in )

(21)

The above discussion has important implications for FO membrane design/selection for OMBR applications: 1) a more selective membrane is generally preferred over a less selective one to minimize the solute accumulation due to reverse diffusion; 2) however, once B/A is less than 0.1in , further reducing B/A may offer little additional advantage since the salt accumulation in this case is dominated by volumetric concentration of influent solute. From membrane fabrication point of view, making a membrane more selective (lower B/A) typically requires the sacrifice of the water permeability (lower A) [25], which tends to reduce FO water flux (Eqs. (1) and (2)) without any benefit in reducing OMBR solute accumulation if B/A is already small. In reality, TFC-1 in Table 1 is significantly more difficult to fabricate than TFC-2, noting that they have essentially the same A value but the B value is drastically lower for TFC-1. It is also interesting to compare the performance of TFC-5 to the other TFC membranes (Fig. 3(a)). Despite that this membrane has a lower initial water flux (13.4 L/m2 h) compared to that for the other TFC membranes (15.3 L/m2 h), its steady-state flux (9.0 L/m2 h) is higher compared to those for TFC-3 (8.1 L/m2 h) and TFC-4 (3.1 L/m2 h). This can be attributed to its much milder solute accumulation compared to those for the more leaky membranes TFC-3 (B/A = in ) and TFC-4 (B/A = 10in ), respectively (Fig. 3(b)). The solute accumulation of TFC-5 is mainly contributed by solutes from the influent water due to its low B/A ratio (B/A = 0.1in ), similar to those for TFC-1 and TFC-2. In comparison, solute reverse diffusion dominates for the more leaky TFC-3 and TFC-4. It is worthwhile to mention that there has been an over-emphasis in the FO literature on the FO water flux based on membrane coupon tests (note that no severe solute accumulation exists in typical coupon tests). In contrast, the importance of the solute reverse diffusion (or the effect of B/A ratio) has been addressed only in a few publications [15,17,26]. The simulation in Fig. 3(a) clearly demonstrate that a “leaky highflux” membrane (e.g., TFC-4) coupon may perform poorly with severe salt accumulation and low stable flux in a submerged OMBR reactor as a result of solute reverse diffusion. Therefore, it is important to recognize that the reactor water flux performance cannot be adequately represented by small membrane coupon tests. The B/A ratio shall be carefully optimized to limit solute reverse diffusion and to ensure stable performance at reactor/module level. In summary, membranes with lower B/A values are preferred (even if at the cost of lower A value) when B/A ≥ in since the solute reverse diffusion play a dominant role in OMBR solute accumulation in this

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

Concentraon (mM NaCl)

a

300

a

300

250

Concentraon (mM NaCl)

320

250

200 150 100

3Km (S=0.2mm) 1.5Km (S=0.4mm) 1Km (S=0.6mm) (1/1.5)Km (S=0.9mm)

50

200 150 100

AL-FW (HRT0 = 8 hrs) AL-DS (HRT0 = 3.5 hrs)

50

AL-DS (HRT0 = 8 hrs)

0

0 0

120

240

360

480

600

0

720

120

240

Time (hr) 3Km (S=0.2mm) 1.5Km (S=0.4mm) 1Km (S=0.6mm) (1/1.5)Km (S=0.9mm)

Water flux (L/m2 hr)

30 25 20 15 10

b

480

600

720

600

720

40

AL-FW (HRT0 = 8 hrs) AL-DS (HRT0 = 3.5hrs) AL-DS (HRT0 = 8 hrs)

35

Water flux (L/m2 hr)

b

35

360

Time (hr)

30 25 20 15 10 5

5

0

0 0

120

240

360

480

600

720

Time (hr) Fig. 6. Effect of mass transfer coefficient on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h; SRT: 10 days; other membrane properties identical to TFC-2 (Table 1).

case (Fig. 5). In contrast, membranes with higher A values are preferred (even if at the cost of increased B/A) when B/A is already less than 0.1in , since increased water permeability can enhance membrane flux without adversely affecting solute accumulation under such conditions. An optimal B/A is likely ∼0.1in (Fig. 5), which minimizes the negative impact of solute reverse diffusion (Eqs. (16) and (18)) but without over-compromise on the membrane A value. The above discussion has not taken into consideration of transport of trace elements through FO membranes. Where rejection of trace organics/metals are of interest, their transport behavior shall be further considered for the reactor/module optimization. Besides the membrane A and B values, the mass transfer coefficient Km plays an important role in FO performance, since ICP (and thus water and solute fluxes) has an exponential dependence on Km . Fig. 6(a) and (b) show the effect of Km on the salt accumulation and the flux behavior, respectively. All the membranes simulated have A and B values identical to those for the reference membrane TFC-2 (Table 1). The mass transfer coefficient varies from 1/1.5Km (1.67 × 10−6 m/s) to 3Km (7.5 × 10−6 m/s). The corresponding S values (Eq. (3)) ranges over 0.2–0.9 mm, comparable to values of 0.4–0.6 mm reported for commercial FO membranes as well as membranes fabricated in our lab [15,22,23]. In Fig. 6(a), all the membranes experienced similar level of salt accumulation despite the large difference in their mass transfer coefficient. This is consistent with Eq. (15) which predicts nearly identical steady state reactor salt concentration since all the membranes have the same B/A value. Indeed, due to the small B/A value (0.1in ), the increase in the reactor solute concentration is mainly caused by the accumulation of solutes from the influent water. As expected, higher initial water flux is available for membrane with greater mass transfer coefficient (Fig. 6(b)), which can be explained by the

0

120

240

360

480

Time (hr) Fig. 7. Effect of membrane orientation on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h (unless stated otherwise); SRT: 10 days; TFC-2 membrane (Table 1).

reduced level of ICP (refer to Eqs. (1) and (2), and Ref. [24]). Despite such difference in the initial water flux, all the membranes experiences similar levels of relative flux reduction, which is consistent with the nearly identical solute accumulation behavior for these membranes. Thus, from membrane fabrication/selection point of view, membranes with larger Km values (i.e., smaller S values) are preferred to maintain a higher steady-state flux without suffering greater solute accumulation in the OMBR. The effect of membrane orientation on OMBR performance is simulated for the TFC-2 membrane (Table 1). Fig. 7(a) shows the salt accumulation behavior, and Fig. 7(b) presents the water flux over time. Once again, the initial HRT in both membrane orientations are kept identical (8 h). Consistent with reports in the literature, the AL-DS orientation has higher initial water flux compared to that for the baseline AL-FW (Fig. 7(b)). Over time, the OMBR solute concentration increases as salt accumulates in the reactor (Fig. 7(a)). This leads to the decline of water flux for both membrane orientations. However, the AL-DS orientation seems to be much more sensitive to the increase in reactor concentration, a phenomenon previously reported by Chou et al. [22]. Since AL-DS has a higher initial flux, a more dramatic ICP loss is expected as a result of the increase in reactor concentration, noting the exponential dependence of ICP on the water flux. When the steady state flux is achieved, the ALDS orientation experiences a more than 50% flux loss, significantly higher than that for the alternative orientation. Correspondingly, the HRT at steady-state condition (18.6 h) is more than doubled of the initial HRT for the AL-DS orientation. Such long HRT may be undesirable in practice, as it suggests a low treatment throughput. To compensate for greater flux decline in the AL-DS orientation, one may use shorter initial HRT upfront. For example, using the same amount of membrane area in AL-DS and AL-FW, a short initial HRT

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

a

450

TFC-2 membrane

400

Concentraon (mM NaCl)

Concentraon (mM NaCl)

a

350 300 250 200 150 100 50

3 mM

10 mM

30 mM

120

240

360

480

600

800

600 500 400 300 200

720

18

b

12 10 8 6 4

3 mM

120

240

360

480

30 mM 600

720

18 TFC -4 membrane

16

14

Water flux (L/m2 hr)

Water flux (L/m2 hr)

0

10 mM

Time (hr)

TFC-2 membrane

16

3 mM

0

Time (hr)

b

TFC -4 membrane

700

100

0 0

321

10 mM

30 mM

2

14 3 mM

12

10 mM

30 mM

10 8 6 4 2

0 0

120

240

360

480

600

720

Time (hr) Fig. 8. Effect of influent salt concentration on OMBR performance for a tight membrane (TFC-2). (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; DS: 1 M NaCl; HRT0 : 8 h; SRT: 10 days.

of 3.5 h can be achieved. Unfortunately, the shorter initial HRT also leads to faster and more severe accumulation of salt in the OMBR (Fig. 7(a)), an effect that will be further discussed in Section 5.3. Correspondingly, the water flux declines even faster compared to that for an initial HRT of 8 h. At steady state conditions, nearly identical flux is achieved in both membrane orientations, as expected from Eq. (19). The above discussion implies that the AL-FW configuration may be favored for a more stable OMBR reactor operation. Despite that the existing literature has convincingly demonstrated that the AL-DS orientation can achieve greater water flux based on membrane coupon tests [2,18,22,23,27], such advantage can be easily lost as a result of solute accumulation due to its inherently less stable flux behavior (see Fig. 7(b) and Refs. [15,22]). Tang et al. [15] has further reported that the AL-DS orientation is also more sensitive to membrane fouling, as the entrapment of the foulants in the porous support can lead to reduced support layer porosity and thus enhanced ICP of the membrane. Therefore, it is recommended that the AL-FW is the preferred membrane orientation for typical OMBR applications. 5.2. Effect of influent salt concentration and draw solution concentration The effect of influent solute concentration on the OMBR water flux and salt accumulation are presented in Fig. 8(a) and (b), respectively, for the reference membrane TFC-2 (Table 1). Clearly, higher influent salt concentration Cin results in faster salt accumulation and higher steady state OMBR solute concentration (Fig. 8(a)). The more severe solute accumulation at higher Cin can be readily explained Eq. (15), which suggests that the solute accumulation is contributed by both the retention of salt in the influent sewage and

0

0

120

240

360

480

600

720

Time (hr) Fig. 9. Effect of influent salt concentration on OMBR performance for a leaky membrane (TFC-4). (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; DS: 1 M NaCl; HRT0 : 8 h; SRT: 10 days.

the reverse diffusion from DS. The initial FO water flux for TFC-2 does not have a strong dependence on Cin for the range of concentrations (3–30 mM NaCl) simulated. On the other hand, significantly lower steady state water flux is attained at higher Cin , as a result of the more severe solute accumulation. Additional simulations have also been performed for a more leaky membrane TFC-4 (Fig. 9(a) and (b)), which has a B/A ratio of 492 kPa (Table 1) or an equivalent concentration of 100 mM due to the reverse diffusion (Eq. (6)). Interestingly, the influent solute concentration (3–30 mM) has little effect on the OMBR performance for the leaky membrane TFC-4. In this case, B/A  in , such that the solute accumulation behavior is dominated by the reverse diffusion. The contribution by salt in the influent is less important. Due to the huge salt reverse diffusion, severe solute accumulation and flux reduction are expected regardless of the influent solute concentration. In contrast, solute accumulation behavior is dominated by the influent contribution for the more selective TFC-2 membrane (B/A  in ). Once again, the current study reveals the critical importance of the relative magnitude of B/A ratio with respect to in . As discussed in Section 5.1, FO membranes with B/A ∼ 0.1in may be preferred for OMBR applications. The effect of DS concentration on OMBR reactor salt concentration and water flux is shown in Fig. 10(a) and (b). As expected, greater DS concentration results in higher water flux due to the greater osmotic driving force available (Fig. 10(b)). As before, the solute accumulation in the OMBR leads to significant reduction in water flux. The solute accumulation is more severe for greater DS concentration based on our simulation (Fig. 10(a)). It is worthwhile to mention that smaller membrane areas have been used for membranes at greater DS concentrations in order to achieve an identical HRT value of 8 h. If the same amount

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

300 250

a Concentration (mM NaCl)

a Concentraon (mM NaCl)

322

200 150 100 50 4M 0

0

120

2M 240

1M

360

480

HRT24 hrs HRT 16 hrs HRT 8 hrs HRT 4 hrs

350 300 250 200 150 100 50

0.5M 600

400

0

720

0

120

240

Time (hr)

b

30

b Water flux (L/m2 hr)

Water flux (L/m2 hr)

480

600

720

18 16

25 20 15 10 5 4M 0

360

Time (hr)

0

120

2M 240

360

1M 480

12 10 8 6

HRT 24 hrs HRT 16 hrs HRT 8 hrs HRT 4 hrs

4 2

0.5M 600

14

0

720

0

120

240

5.3. Effect of hydraulic retention time and sludge retention time Hydraulic retention time and sludge retention time are two important operational parameters in an OMBR system. To study the effects of these two parameters on the salt accumulation and water flux performance, simulations based on the TFC-2 membrane have been performed using different HRTs and SRTs. The simulation results are shown in Figs. 11 and 12. Neither HRT nor SRT has any effect on the initial water flux since the initial reactor concentration is the same as the influent concentration. In Fig. 11(a), system operated under shorter HRT suffers more severe solute accumulation. This is accompanied by a more dramatic water flux decline (Fig. 11(b)). Similarly, greater OMBR concentration and lower water flux are expected at longer SRT (Fig. 12(a) and (b)). Such effect can be readily explained by Eq. (15), where the SRT/HRT ratio represents the volumetric concentration factor. Indeed, further simulation indicates that nearly identical steady state reactor concentration and FO water flux are obtained at identical SRT/HRT ratio (simulation results not shown). In a well designed OMBR, the solute reverse diffusion through the FO membrane shall be relatively low (B/A  in ), and the steady state reactor concentration is given by ( SRT / HRT )Cin . Since the value of Cin is largely determined by the given application, the major parameter controlling solute accumulation is the SRT/HRT ratio. An excessive SRT/HRT ratio can lead to very severe salt accumulation in the OMBR, thus such situation shall be strictly avoided.

a Concentration (mM NaCl)

membrane area is to be used, the solute accumulation and flux reduction will be even more drastic for the greater DS concentrations (results not shown), as a result of the shorter HRT used (Eqs. (15) and (18) and Section 5.3).

480

600

720

Fig. 11. Effect of hydraulic retention time on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; SRT: 10 days. TFC-2 membrane (Table 1).

400 SRT 2.5 days SRT 5 days SRT 10 days SRT 20 days

350 300 250 200 150 100 50 0

b

0

120

240

360

Time (hr)

480

600

720

18 16

Water flux (L/m2 hr)

Fig. 10. Effect of draw solution concentration on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; HRT0 : 8 h; SRT: 10 days. TFC-2 membrane (Table 1).

360

Time (hr)

Time (hr)

14 12 10 8 SRT 2.5 days SRT 5 days SRT 10 days SRT 20 days

6 4 2 0

0

120

240

360

Time (hr)

480

600

720

Fig. 12. Effect of sludge retention time on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h. TFC-2 membrane (Table 1).

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

323

(e.g., 50% every 5 days), such gain in average flux is even greater. However, less frequent intermittent operation requires discharging larger volume of mixed liquor from the reactor, which may adversely affect the OMBR biological stability. Such constraint shall be taken into consideration for optimizing the intermittent wasting strategy. 6. Conclusions A salt accumulation model is developed in the study, which captures the solute contribution from the influent wastewater and the reverse diffusion of solutes from the draw solution. The effects of membrane properties (A, B, Km , and membrane orientation) and the OMBR operational conditions (Cin , CDS , HRT, and SRT) on OMBR performance are simulated using this model coupled with ICP equations. Based on our simulation: • The salt accumulation is more severe at greater in and B/A. • The relative importance of reverse diffusion over contribution from influent solutes is governed by the membrane selectivity. When B/A  in , solute reverse diffusion has negligible effect on OMBR performance. In contrast, the salt accumulation and FO water flux reduction are governed by reverse diffusion for B/A > in . • The optimal B/A ratio for OMBR operation is ∼0.1in . • The salt accumulation is directly proportional to the SRT/HRT ratio. Both shorter HRT and longer SRT have the effect of promoting salt accumulation that can adversely affect FO water flux.

Fig. 13. Effect of intermittent sludge wasting on OMBR performance. (a) Salt accumulation; (b) FO water flux. Simulation conditions: AL-FW; initial FW concentration: 10 mM NaCl; DS: 1 M NaCl; HRT0 : 8 h; equivalent SRT: 10 days (infinite for the case of no wasting). TFC-2 membrane (Table 1).

It is also interesting to note that the duration to reach the steady state seems directly dependent on SRT (Fig. 12(a) and (b)). Longer SRT typically requires longer time to reach steady state. Indeed, the steady state is reached when the solutes entering into the OMBR is balanced by the solutes discharge via waste sludge. A lower sludge waste rate (i.e., longer SRT) thus requires longer time to reach a stable condition. The current study suggests that SRT can be used as an appropriate time scale for the OMBR system to reach the steady state condition. Finally, different wasting strategies are investigated using the solute accumulation and water flux model (Fig. 13(a) and (b)). An SRT of 10 days is used as a baseline case for comparison. One option considered in our simulation is “zero waste discharge” (i.e., an SRT of infinite). In this case, all the solutes entering into the OMBR will accumulate in the reactor, and the mixed liquor solute concentration shall asymptotically approach the draw solution concentration, at which the water flux approaches zero. Thus, “zero liquid discharge” shall be avoided, as it will inevitably leads to excessive solute accumulation and flux loss for OMBR applications. To avoid such situation, another option is to have intermittent sludge discharge. For example, an equivalent SRT of 10 days is achieved by discharging 25% of the reactor volume every 2.5 days. In general, such intermittent wasting strategy results in some sort of zigzag flux and reactor concentration curves. The lowest water flux and the highest reactor concentration for the intermittent wasting are comparable to the respective values in the baseline case (continuous sludge wasting, SRT = 10 days). On the other hand, the average water flux may be slightly higher for the intermittent case compared to the baseline. For less frequent intermittent discharge

The current study reveals the critical importance of the B/A ratio and HRT/SRT ratio for optimized OMBR operation. Acknowledgements This research was funded by Environment & Water Industry Development Council of Singapore (EWI) through Project #MEWR C651/06/173 (EWI0901-02-01). We also wish to acknowledge the funding support for Dezhong Xiao on this project from Nanyang Technological University under the Undergraduate Research Experience on CAmpus (URECA) programme. Hydration Technology Inc. is thanked for providing us free FO membrane material. Appendix A. Fig. A1 shows the typical salt concentration profiles across an FO membrane for both membrane orientations. In the AL-DS orientation, the salt concentration inside the support layer is higher than the bulk feed water salt concentration. In the AL-FW orientation, the salt concentration inside the support layer is lower than the bulk salt concentration of the DS. Appendix B. The flux behavior and solution concentration change over time can be simulated for FO reactors without maintaining a constant draw solution concentration. Consider a FO process with the initial draw solution salt concentration Cd0 , initial feed water concentration Cf0 , initial DS volume Vd0 and initial FW volume Vf0 . The total membrane area is Am . After time t, the amount of salt remained in draw solution tank can be determined via the following mass balance equation:



Cdt Vdt = Cd0 Vd0 −

t

Js Am dt

(A1)

0

where Cd0 Vd0 and Cdt Vdt are the total mass of solute in the draw

t

solute at time 0 and time t, respectively. The term 0 Js Am dt is the accumulative mass of solute lost from the draw solution due to

324

D. Xiao et al. / Journal of Membrane Science 366 (2011) 314–324

Fig. A1. Salt concentration profiles in AL-DS and AL-FW orientations. Adapted from Refs. [2,3,18].

reverse diffusion. By replacing Js using Eq. (6), the draw solution concentration can be determined by Cdt =

Cd0 Vd0 − (B/AˇRT ) Vd0 +

t

t

J A dt 0 v m

J A dt 0 v m

(A2)

Eq. (A2) can be used together with ICP equations (Eqs. (1) and (2)) and FW concentration equation to simulate the FO performance where draw solution dilution occurs. References [1] S. Loeb, L. Titelman, E. Korngold, J. Freiman, Effect of porous support fabric on osmosis through a Loeb-Sourirajan type asymmetric membrane, Journal of Membrane Science 129 (1997) 243–249. [2] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications, and recent developments, Journal of Membrane Science 281 (2006) 70–87. [3] Y. Xu, X. Peng, C.Y. Tang, Q.S. Fu, S. Nie, Effect of draw solution concentration and operating conditions on forward osmosis and pressure retarded osmosis performance in a spiral wound module, Journal of Membrane Science 348 (2010) 298–309. [4] R.W. Holloway, A.E. Childress, K.E. Dennett, T.Y. Cath, Forward osmosis for concentration of anaerobic digester centrate, Water Research 41 (2007) 4005–4014. [5] R.J. Salter, Forward osmosis, Water Conditioning and Purification 48 (2005) 36–38. [6] A. Seppala, M.J. Lampinen, W. Kotiaho, A new concept for an osmotic energy converter, International Journal of Energy Research 25 (2001) 1359–1379. [7] E.R. Cornelissen, D. Harmsen, K.F. de Korte, C.J. Ruiken, J.J. Qin, H. Oo, L.P. Wessels, Membrane fouling and process performance of forward osmosis membranes on activated sludge, Journal of Membrane Science 319 (2008) 158–168. [8] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, Desalination by ammonia– carbon dioxide forward osmosis: influence of draw and feed solution concentrations on process performance, Journal of Membrane Science 278 (2006) 114–123. [9] S. Loeb, Large-scale power production by pressure-retarded osmosis, using river water and sea water passing through spiral modules, Desalination 143 (2002) 115–122. [10] K. Gerstandt, K.V. Peinemann, S.E. Skilhagen, T. Thorsen, T. Holt, Membrane processes in energy supply for an osmotic power plant, Desalination 224 (2008) 64–70. [11] L.P. Wessels, E.R. Cornelissen, Werkwije en inrichting voor het in een membraan-filtratie-eenheid behandelen van een waterige afvalstroom afkom-

stig van een bioreactor (Dutch Patent NL1028484), (2005), Netherlands. [12] A. Achilli, T.Y. Cath, E.A. Marchand, A.E. Childress, The forward osmosis membrane bioreactor: a low fouling alternative to MBR processes, Desalination 238 (2009) 10–21. [13] W.C.L. Lay, Y. Liu, A.G. Fane, Impacts of salinity on the performance of high retention membrane bioreactors for water reclamation: a review, Water Research 44 (2010) 21–40. [14] R.L. McGinnis, M. Elimelech, Energy requirements of ammonia–carbon dioxide forward osmosis desalination, Desalination 207 (2007) 370–382. [15] 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, Journal of Membrane Science 354 (2010) 123–133. [16] W.C.L. Lay, T.H. Chong, C.Y. Tang, A.G. Fane, J. Zhang, Y. Liu, Fouling propensity of forward osmosis: investigation of the slower flux decline phenomenon, Water Science and Technology 61 (2010) 927–936. [17] N.T. Hancock, T.Y. Cath, Solute coupled diffusion in osmotically driven membrane processes, Environmental Science and Technology 43 (2009) 6769–6775. [18] J.R. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal concentration polarization on flux behavior in forward osmosis, Journal of Membrane Science 284 (2006) 237–247. [19] K.L. Lee, R.W. Baker, H.K. Lonsdale, Membranes for power generation by pressure-retarded osmosis, Journal of Membrane Science 8 (1981) 141–171. [20] G.T. Gray, J.R. McCutcheon, M. Elimelech, Internal concentration polarization in forward osmosis: role of membrane orientation, Desalination 197 (2006) 1–8. [21] J.R. McCutcheon, M. Elimelech, Modeling water flux in forward osmosis: implications for improved membrane design, AIChE Journal 53 (2007) 1736–1744. [22] S. Chou, L. Shi, R. Wang, C.Y. Tang, C. Qiu, A.G. Fane, Characteristics and potential applications of a novel forward osmosis hollow fiber membrane, Desalination 261 (2010) 365–372. [23] R. Wang, L. Shi, C.Y. Tang, S. Chou, C. Qiu, A.G. Fane, Characterization of inhouse made composite forward osmosis hollow fiber membranes, Journal of Membrane Science 355 (2010) 158–167. [24] J.R. McCutcheon, M. Elimelech, Influence of membrane support layer hydrophobicity on water flux in osmotically driven membrane processes, Journal of Membrane Science 318 (2008) 458–466. [25] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Effect of membrane chemistry and coating layer on physiochemical properties of thin film composite polyamide RO and NF membranes. II. Membrane physiochemical properties and their dependence on polyamide and coating layers, Desalination 242 (2009) 168–182. [26] W.A. Phillip, J.S. Yong, M. Elimelech, Reverse draw solute permeation in forward osmosis: modeling and experiments, Environmental Science and Technology 44 (2010) 5170–5176. [27] Y. Wang, F. Wicaksana, C.Y. Tang, A.G. Fane, Direct microscopic observation of forward osmosis membrane fouling, Environmental Science and Technology 44 (2010) 7102–7109.