Thin-film composite P84 co-polyimide hollow fiber membranes for osmotic power generation

Thin-film composite P84 co-polyimide hollow fiber membranes for osmotic power generation

Applied Energy 114 (2014) 600–610 Contents lists available at ScienceDirect Applied Energy journal homepage: www.elsevier.com/locate/apenergy Thin-...

3MB Sizes 1 Downloads 174 Views

Applied Energy 114 (2014) 600–610

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Thin-film composite P84 co-polyimide hollow fiber membranes for osmotic power generation Xue Li a, Tai-Shung Chung a,b,⇑ a b

NUS Graduate School for Integrative Science and Engineering, National University of Singapore, Singapore 117456, Singapore Department of Chemical & Biomolecular Engineering, National University of Singapore, Singapore 117576, Singapore

h i g h l i g h t s

g r a p h i c a l a b s t r a c t

 Novel thin-film composite (TFC)

hollow fiber membranes are molecularly designed and developed.  The membrane supports achieve superior tolerance to high pressures up to 23 bar.  The TFC membranes perform well in pressure retarded osmosis (PRO) process.  The effect of flow rate on PRO was investigated for the first time.

Inner surface

Cross-section

Outer surface

Hollow fiber membrane 14 12

0< Δ P< Δ π

1st run 2nd run

Power density 10 (W m-2) 8 6 4

Fresh water Salty water

Feed: DI water Draw: 1 M NaCl 24 ± 1 ˚C

2

water

0

0

4

8

12

16

20

Δ P (Bar)

water water

Turbine

a r t i c l e

i n f o

Article history: Received 21 May 2013 Received in revised form 11 October 2013 Accepted 14 October 2013 Available online 6 November 2013 Keywords: Hollow fibers Osmotic power generation Thin-film composite membrane Pressure retarded osmosis

a b s t r a c t A series of well-designed thin-film composite (TFC) hollow fiber membranes via dual-layer co-extrusion technology for pressure retarded osmosis (PRO) applications is reported in this work. By controlling the phase inversion process during spinning, we have molecularly engineered hollow fiber membranes with various structures, dimensions, pore characteristics, and mechanical properties as supports for the synthesis of TFC membranes. Under hydraulic tests, these hollow fiber membrane supports possess high burst pressures from 13 to 24 bar. The TFC membranes fabricated by interfacial polymerization on the inner surface of the hollow fiber supports not only exhibit relatively high power densities of 5– 12 W m2 but also display a superior tolerance to high pressures up to 21 bar. The TFC membrane synthesized on a small dimensional hollow fiber support, which was spun from a P84 co-polyimide/ethylene glycol (EG)/N-methyl-2-pyrrolidinone (NMP) dope solution with a bore fluid of a water/EG/NMP mixture, shows the most impressive PRO performance (i.e., 12 W m2 at 21 bar using water and 1 M NaCl as feeds). Experimental results also suggest that inner-selective TFC hollow fiber membranes made from small dimensional fiber supports by means of delayed demixing during the fiber spinning are preferential for high pressure PRO processes. In addition, it was found that the flow rate of brine solutions plays a crucial effect on TFC membrane performance for osmotic power generation. By investigating the pressure drop as a function of flow rate, one may be able to choose appropriate PRO operation conditions to further ensure the sustainability of hollow fiber membranes for power generation. Ó 2013 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Department of Chemical & Biomolecular Engineering, National University of Singapore, Singapore 117576, Singapore. Tel.: +65 6516 6645; fax: +65 6779 1936. E-mail address: [email protected] (T.-S. Chung). 0306-2619/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apenergy.2013.10.037

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

1. Introduction The scarcity and vulnerability of current fossil fuel energy motivate other alternative, renewable, and environmentally friendly energy sources. Currently, salinity gradient energy, a clean energy from ocean and various sources of salty water, has received worldwide attention [1,2]. To effectively convert such a promising energy into electricity, pressure retarded osmosis is getting more and more interest [3,4]. Pressure retarded osmosis (PRO) is a membrane-based technique that was already proposed in the 1970s [5– 7]. Although generating salinity gradient energy by PRO is theoretically feasible, the ineffective membranes used at that time made it economically infeasible. Most of early works on PRO was discontinued due to the absence of suitable PRO membranes [8,9]. Things change when forward osmosis (FO) membranes have been rapidly developed [2,10,11]. FO membranes share a similar osmotic mechanism with PRO membranes, targeting for high selectivity with high water flux and low salt flux. Thus, the recent improvement in FO membranes revives research and development for PRO processes. A Norwegian power company, Statkraft, has conducted research on salinity gradient energy since 1997 and built the first prototype plant in 2009 by mixing river water and seawater under the PRO mode. By doing so, the potential of PRO has been practically demonstrated. To commercialize the production of osmotic energy via PRO process, a membrane power density of more than 5 W m2 is crucial [3,12–14]. Besides the power density, another challenge is to have a membrane that could withstand high hydraulic pressures, since the operating pressure of the salty water compartment in the PRO process may increase up to 13 bar (for seawater) and even higher (for other highly concentrated salty water, such as RO retentate). Several studies have attempted to utilize conventional FO membranes for PRO applications [15–20]. However, without chemical or physical reinforcements, they suffer severe performance deterioration due to irreversible deformation, densification and damage when subjected to high hydraulic pressures because most FO membranes were originally designed to operate only at no or low pressures [2,17–20]. To meet the requirements of high water permeability and salt rejection, most osmotic membranes are designed to have a thin selective layer on top of a fully porous substrate. The fabrication of thin-film composite (TFC) membranes via interfacial polymerization is one of the preferred methods because it provides advantages such as availability of various moieties for interfacial polymerization, diverse tunability of supporting layer, and ease of fabrication [15–25]. Although TFC membranes with a highly porous support layer have been designed for FO applications with minimal internal concentration polarization and enhanced water permeability, they are structurally vulnerable under high pressure PRO operations. Hence, one of the major scientific challenges for material and membrane scientists is how to design a robust TFC membrane for osmotic power generation without sacrificing its high water permeability and low salt permeability. Hollow fiber membranes have advantages of (1) high membrane area per unit membrane module volume, leading to low operating costs and high productivity, (2) self-mechanical support, preventing adverse effects of spacers used in flat-sheet membrane modules, and (3) ease of handling during fabrication and operation [26,27]. As a result, there are growing interests in using hollow fiber membranes for PRO applications [16,28]. However, the knowledge to fabricate high performance PRO hollow fiber membranes is very limited. Therefore, this work aims at revealing the science and engineering of developing a TFC hollow fiber membrane with a high water flux and good tolerance to high pressure PRO operations. The effects of spinning conditions on membrane morphology and PRO performance will also be investigated. In addition,

601

operation condition such as flow rate also inevitably affects the mass transfer near the membrane surface and pressure drop along the flowing channel. Thus, the effects of flow rate will be studied. P84 co-polyimide (as shown in Fig. 1) was chosen in this study as the material for the fabrication of hollow fiber membrane support because of its balanced physiochemical properties and good chemical resistance. The successful development of P84 flat-sheet membranes with good PRO performance in our previously study [29] also stimulated us to explore its potential in the hollow fiber configuration. In this work, we firstly examined the feasibility of fabricating P84 hollow fiber supports with desirable morphologies and properties via dual-layer co-extrusion technology. We then prepared TFC PRO membranes from these hollow fiber supports. Robust TFC hollow fiber membranes have been developed with superior tolerance to high pressures without compromising PRO performance. 2. Experimental 2.1. Materials and chemicals A P84 co-polyimide (BTDA-TDI/MDI, co-polyimide of 3,30 ,4,40 benzophenone tetra-carboxylic dianhydride and 80% methylphenylene diamine + 20% methylene diamine) was purchased from HP Polymer, Austria. Fig. 1 shows its chemical structure. The solvent N-methyl-2-pyrrolidinone (NMP, >99.5%) and non-solvent ethylene glycol (EG, 99.9%) were ordered from Merck and VWR, respectively, and were used to prepare the spinning solutions. The deionized water used in experiments was produced by a Milli-Q ultrapure water system (Millipore, USA). A 50/50 wt% mixture of glycerol (Industrial grade, Aik Moh Pains & Chemicals Pte. Ltd., Singapore) and de-ionized water was prepared to post-treat as-spun hollow fiber supports before drying. Polyethylene glycol 4000, 6000, 10,000, 20,000, and 35,000 (PEG, Mw = 4000 g mol1, 6000 g mol1, 10,000 g mol1, 20,000 g mol1, and 35,000 g mol1, respectively, Sigma–Aldrich) were employed to characterize the molecular weight cut-off (MWCO), mean pore and pore size distribution of hollow fiber supports. m-Phenylenediamine (MPD, >99%) and 1,3,5-benzenetricarbonyl trichloride (TMC, 98%) were bought from Sigma–Aldrich. Hexane and sodium chloride (NaCl) were procured from Merck. All chemicals were used as received. 2.2. Fabrication of P84 hollow fiber supports and post-treatment Prior to preparing polymer solutions, the P84 polymer powder was dried overnight at 90 ± 5 °C in a vacuum oven (2 mbar) to remove moisture content. The dehydrated P84 polymer was dissolved in a NMP/EG solution and Table 1 states the solution composition. After the spinning dope solution was prepared, it was degassed for several hours and then stored in a 500 mL syringe pump (ISCO Inc.) overnight before spinning. The P84 hollow fiber supports were prepared by a dry-jet wet spinning process as described in our previous publication [24]. By changing the spinning conditions, such as bore fluid composition, bore fluid flow rate, take-up speed, and spinneret dimension, four different hollow fiber membranes were spun from the same polymer solution with different morphologies and dimensions. All nascent fibers did not experience additional extra drawing (i.e., no extension) after leaving the spinneret, which means that the take-up speed of the hollow fibers was almost the same as the falling speed into the coagulation bath. Two dual-layer spinnerets illustrated in Fig. 2 were employed in this work and Table 1 tabulates the spinning conditions. The spinning dope was conveyed through the middle channel, while NMP was transported via the outer channel in order to induce delayed demixing at the outer membrane surface. After

602

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

O

O

O

N

N

O

O

80%

20%

n

P84 co-polyimide Fig. 1. Chemical structure of P84 co-polyimide.

spinning, the as-spun hollow fiber supports were rinsed with tap water for two days to remove residual solvents and EG. A glycerol/water mixture (50/50 wt%) was used to soak hollow fiber supports for another two days, and then they were dried in ambient air at room temperature. Soaking membranes in the glycerol/water mixture may diminish membrane shrinkages and prevent pores from collapse during drying. Hollow fiber modules were fabricated by using Swagelok stainless fittings [24]. For each hollow fiber sample, three or four modules with an outer diameter of 3/8 in. were prepared and tested to ensure good repeatability. Each module consisted of five fibers with an effective length of around 12.5 cm. A slow curing epoxy resin (EP 231, Kuo Sen, Taiwan) was used for module potting. The length of the potting portion for each end was around 1.5 cm. 2.3. Fabrication of P84-TFC membranes The TFC polyamide membrane was synthesized on the inner surface of P84 hollow fiber supports by interfacial polymerization using the following procedures: (1) the membrane module was held vertically and purged by compressed air to sweep impurities on the inner surface of hollow fibers; (2) a MPD solution of 1.1 wt% was introduced to the inner surface of hollow fiber supports vertically from bottom to top at a flow rate of 2.2 mL min1 for 3 min; (3) the excess MPD aqueous solution was removed by purging with compressed air for 3 min; and (4) a hexane solution containing 0.2 wt% TMC was fed to the inner face of the fibers with a flow rate of 2.2 mL min1 for 3 min followed by a 15-min heat treatment in an oven at 65 °C. Finally, the resultant TFC membrane was washed thoroughly with de-ionized water and stored in de-ionized water before tests. 2.4. Characterizations 2.4.1. Morphology, topology, porosity, MWCO, pore size, pore size distribution and mechanical properties of P84 hollow fiber supports The membrane morphology was examined by field emission scanning electronic microscopy (FESEM, JEOL JSM-6700F). Before FESEM tests, samples were prepared in liquid nitrogen followed by platinum coating using a Jeol JFC-1100E Ion Sputtering device. Membrane surface topology was studied using a Nanoscope IIIa atomic force microscope (AFM) from Digital Instruments. Each membrane sample was scanned at a rate of 1.00 Hz using the tapping mode. Mean roughness (Ra) was determined from the averages of at least 10 samples. The characteristics of P84 porous substrates such as pore size, molecular weight cut-off (MWCO) and pore size distribution were measured by the solute transport method as described elsewhere [18,30]. PEG solutions with a concentration of 200 ppm, each comprising different molecular weights, were used to measure the solute rejection under a hydraulic pressure difference of 1 bar. The

concentrations of the neutral solutes were measured by a total organic carbon analyzer (TOC ASI-5000A, Shimadzu, Japan). The measured feed (Cf) and permeate (Cp) concentrations were used for the calculation of the effective solute rejection coefficient Rs (%):

  Cp  100% Rs ¼ 1  Cf

ð1Þ

According to the Einstein equation, a Stokes radius which is used to describe the dimension of solutes could be expressed as:

DAB ¼

kT 6r pg

ð2Þ

where r is the Stokes radius, DAB is the diffusivity, k is the Boltzmann constant, and g is the solvent viscosity. On the other hand, the diffusivity is also a function of the intrinsic viscosity [g] [31]:

DAB ¼

2:5  106 kT

gðMw ½gÞ1=3

ð3Þ

where Mw is the molecular weight. Combining Eqs. (2) and (3) yields the following equation for the Stokes radius of the solute as a function of Mw and [g] [32]:

r ¼ 2:12  108  ðMw ½gÞ1=3

ð4Þ

As a result, the solute diameters ds (nm) of PEG (Eq. (5)) can be determined as follows [33,34]:

ds ¼ 3:35  102  M 0:557 w

ð5Þ

The mean effective pore size and the pore size distribution were then obtained according to the traditional solute transport approach by ignoring the influence of the steric and hydrodynamic interaction between solute and membrane pores, the mean effective pore size (lp) and the geometric standard deviation (rp) can be assumed to be the same as ls (the geometric mean size of solute at Rs = 50%) and rg (the geometric standard deviation defined as the ratio of the ds at Rs = 84.13% over that at Rs = 50%). The MWCO was determined at Rs = 90%. Therefore, based on lp and rp, the pore size distribution of a membrane can be expressed as the following probability density function [30,35]:

" # 2 ðln dp  ln lp Þ dRðdp Þ 1 pffiffiffiffiffiffiffi exp  ¼ 2 ddp dp ln rp 2p 2ðln rp Þ

ð6Þ

Mechanical properties of membranes including the elongation at break, maximum tensile strength, and Young’s modulus were measured at the constant elongation rate of 10 mm min1 with a starting gauge length of 50 mm by an Instron tensiometer (Model

603

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610 Table 1 Spinning conditions of P84 hollow fiber membrane supports. Spinning parameter

Hollow fiber code I

Polymer solution (wt%) Polymer solution flow rate (mL min1) Bore fluid composition (wt%) Bore fluid flow rate (mL min1) Outer fluid composition (wt%) Outer fluid flow rate (mL min1) Air gap (cm) External coagulant Take up speed (m min1) Spinneret dimension (mm)

II

P84/EG/NMP: 21/10/69 1.2 DI water DI water/NMP (50/50) 1 1 Pure NMP 0.2 1 Tap water (room temperature) 1.65 (Free falling) 1.65 (Free falling) 2.0 mm dual layer 2.0 mm dual layer

5542, Instron Corp.). Ten samples were tested to minimize the experimental error and ensure the accuracy. The estimated burst pressure was calculated by the Barlow’s equation based on tensile stress and fiber dimension [36]:



2vT OD  SF

ð7Þ

where v is the wall thickness, T is the maximum tensile stress, OD is the outer diameter and SF is the safety factor (usually SF = 1). 2.4.2. Determination of pure water permeability and burst pressure of the hollow fiber supports The pure water permeability (PWP) of the membrane supports were determined by testing the membranes using a lab-scale circulating filtration unit that has been described elsewhere [24,34]. The feed solution with a particular pressure was applied into the lumen side of the hollow fiber. Prior to tests, the hollow fiber supports were conditioned at 1 bar for 1 h. The PWP experiments were carried out by flowing de-ionized water at a constant pressure of 1 bar at 24 ± 1 °C. For burst pressure tests, the hollow fiber supports were applied with a series of pressures. The pressure at which hollow fiber breaks is defined as the burst pressure of the support. 2.4.3. Determination of mass transport characteristics of the TFC membranes The water permeability A, salt rejection Rs, and salt permeability B, of the TFC membranes were also determined by testing the membranes using a lab-scale circulating filtration unit, described elsewhere [24,37]. The water permeability, A, was obtained from the pure water permeation flux under an applied hydraulic pressure DP of 5 bar at 24 ± 1 °C. The salt rejection Rs was tested under a trans-membrane pressure of 5 bar using a 1000 ppm NaCl feed solution with salt concentrations measured by a conductivity meter (Schott Instruments Gmbh, Germany). The salt permeability B was derived from the solution–diffusion theory as follows [7,8]:

B¼A

ð1  Rs ÞðDP  DpÞ Rs

ð8Þ

where Dp is the osmotic pressure difference across the membrane. 2.4.4. Determination of TFC membranes performance in PRO tests The PRO tests for osmotic power generation were conducted on a lab-scale PRO set-up using membrane modules as described in previous publications [17,18]. Similar to FO tests, model RO retentate (1 M NaCl) and de-ionized water were used as draw and feed solutions, respectively. TFC membranes were oriented in the PRO mode for all tests. Counter-current flows at the same flow rates Q (0.1, 0.2, or 0.3 L min1) were applied to both the draw solution and the feed solution, and a constant temperature of 24 ± 1 °C was maintained. A variable-speed gear pump (Cole-

III

IV

DI water/EG/NMP (32/40/28) 1

DI water/EG/NMP (32/40/28) 0.6

1.65 (Free falling) 2.0 mm dual layer

2.20 (Free falling) 1.6 mm dual layer

Palmer, Vernon Hills, IL) was utilized to recirculate the feed solution (DI water) through the shell side of the hollow fibers, and a high-pressure hydra cell pump was employed to recirculate the draw solution through the lumen side. Prior to tests, the TFC hollow fiber membranes were conditioned at a particular pressure for 1 h (i.e., 10, 10, 12, or 21 bar, for TFC-I, II, II, and IV, respectively). Then the TFC hollow fiber modules were tested in the PRO process at no hydraulic pressure. After that, their PRO performance was conducted at various hydraulic pressures from 0 bar to their particular pressures in series. For each TFC hollow fiber, three or four modules were prepared and tested to ensure good repeatability. The power density is calculated by the following equation:

W ¼ J w DP

ð9Þ

where DP is the hydraulic pressure difference across the membrane and Jw is the water permeation flux. Jw can be determined from either experimental measurements in PRO tests or the following theoretical equation [17]:

" Jw ¼ A



pD;b exp 

#  Jw 1 D P  k 1 þ JB ½expðJ w K m Þ  1

ð10Þ

w

where pD,b is the osmotic pressure of the bulk draw solution, Km is the solute resistivity, k refers to the mass transfer coefficient which is related to external concentration polarization, as described elsewhere [38,39]. k can be calculated by Eq. (11) using solute diffusion coefficient, Ds, hydraulic diameter of the flow channel, dh, and Sherwood number, Sh [40]:



Sh  Ds dh

ð11Þ

where Sh is determined according to different flow types as follows:

 0:33 dh ðLaminar flowÞ Sh ¼ 1:85 Re Sc L

ð12Þ

ðTurbulent flowÞ Sh ¼ 0:04Re0:75 Sc0:33

ð13Þ

where Re is the Reynolds number, Sc is the Schmidt number, and L is the length of the flow channel. The osmotic pressure of the draw solution at the membrane surface pD,m which represents the dilutive external concentration polarization (ECP) can be described as:



pD;m ¼ pD;b exp 

Jw k



ð14Þ

Thus, the relationship between water flux and driving force in a FO process under the PRO mode can be described as Eq. (15) if the external concentration polarization is taken into account [8,40]:

Jw ¼

1 ApD;m  J w þ B ln Km ApF;b þ B

ð15Þ

604

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

Middle channel (polymer solution) Inner channel (bore fluid) 1.60 mm

2.00 mm

1.30

1.74

1.14

1.58

0.60

1.00

0.44

0.84

1.6 mm dual layer spinneret

Outer channel (NMP solvent)

2.0 mm dual layer spinneret Fig. 2. Schematic designs of the two dual-layer spinnerets.

where pF,b is the osmotic pressure of the bulk feed solution. From the solute resistivity Km, membrane structural parameter, St, can be determined:

St ¼ K m Ds

ð16Þ

where Ds values have been summarized by researchers [41]. 3. Results and discussion 3.1. Morphology, surface roughness, and pore characteristics of hollow fiber supports Fig. 3 displays morphologies of P84 hollow fiber supports (from I to IV) spun with different spinning conditions (as shown in Table 1) via dual-layer spinnerets. Due to the delayed demixing induced by the outer channel fluid NMP, all hollow fiber supports exhibit a highly porous structure on outer surfaces [24,42]. Compared to outer surfaces, all inner surfaces possess relatively smaller pores because water mixtures (non-solvent) in the bore channel cause more rapider demixing. The addition of EG, a weak non-solvent additive and a pore former, into the spinning solution not only increases viscosity but also promotes uniformly solvent exchange during the phase inversion process [29]. As a result, the membrane supports are full of interconnected open-cell pores. According to the cross-section images, two layers of sponge-like structure are formed beneath the top and bottom membrane surfaces. There are some macrovoids between them. Interestingly, the number of macrovoids decreases from fiber I to fiber IV. Apparently, water, the strongest non-solvent bore fluid, induces fast phase inversion and results in macrovoids in the hollow fiber supports. The addition of the NMP solvent into the bore fluid retards the phase inversion rate and reduces the number of macrovoids. A further decrease in water percentage by adding a weak non-solvent (i.e., EG) into the bore fluid leads to a slower precipitation rate and then enhances the delayed demixing. As a result, the number of macrovoids is further reduced. These hollow fiber supports possess similar wall thicknesses since other spinning parameters were

kept the same (e.g., spinning dope composition, dope flow and bore fluid rates, air gap, and take-up speed). By changing the 2.0 mm dual-layer spinneret with a 1.6 mm dual-layer spinneret, the resultant hollow fiber support has a smaller dimension as shown by the fiber IV in Fig. 3. This fiber has a very similar microstructure to that of the fiber III. Fig. 4 shows the AFM images and surface roughnesses of the inner surfaces of these hollow fiber supports. All fiber supports show similar surface topologies with a small roughness around 4–6 nm over an area of 5  5 lm2. Among them, fiber II using water/NMP as the bore fluid has the smoothest inner surface probably resulting from the effect of delayed demixing [43]. However, fibers III and IV spun using water/EG/NMP as the bore fluid exhibit slightly rougher inner surfaces. This is possibly due to the hydrophilic nature of EG that enhances pore evolution at the membrane surface and results in a rough structure. In addition, EG in the polymer solution introduces high affinity with the bore fluid of water/EG/ NMP. This may further delay the solvent exchange, leading to a rough surface [44]. Fig. 5 displays the probability density function curves of pore characteristics of P84 hollow fiber supports measured by the solute transport method with a series of organic solutes (i.e., PEGs), while Table 2 summarizes the mean effective pore diameter, molecular weight cutoff (MWCO), and geometric standard deviation. It can be seen clearly that the composition of bore fluid has a significant impact on the nanostructure of the inner surface. Fiber I spun using water as the bore fluid has the sharpest pore size distribution and smallest geometric standard deviation of 1.20. It has a small pore diameter of 4.46 nm and the lowest MWCO of 13.4 kDa. As NMP is added into the bore fluid while keeping other spinning conditions the same, both the pore dimension and distribution broaden. The mean pore diameter and MWCO of fiber II greatly increase to 7.49 nm and 41.4 kDa, respectively. Meanwhile, the standard deviation slightly increases to 1.37. After introducing EG into the bore fluid, it complicates the solvent exchange and phase inversion processes during membrane formation. Both fibers III and IV have small pore diameters of 3.5 nm but have large MWCOs of 25– 30 kDa and high geometric standard deviations. Their probability density function curves overlap with each other, indicating two

605

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

100 nm Inner surface

100 µm

100 nm Outer Surface

Overall cross-section

100 nm Cross-section ( ) (inner)

10 µm

100 nm

Cross-section

Cross-section ((outer))

I

(Water)

II

(W t /NMP) (Water

III

(Water/EG/NMP)

IV

(Water/EG/NMP) Bore fluid composition Fig. 3. SEM morphologies of as-spun P84 hollow fiber supports as functions of bore fluid composition and other spinning conditions.

fibers share similar nanostructures due to the use of the same bore fluid regardless of different spinneret dimensions. 3.2. Burst pressure and mechanical properties of P84 hollow fiber supports The burst pressure of hollow fiber supports is defined as the pressure at which a hollow fiber support bursts, which is evidenced by a sudden increase in water permeability as a result of structure failure. Fig. 6 depicts the burst pressures of these hollow fiber supports in terms of pure water permeability (PWP) as a

function of pressure. Generally, all PWPs initially decline with an increase in pressure as a result of membrane compaction under pressure. Fiber I has the lowest initial PWP of 130 L m2 h1 bar1 at 2 bar. Its PWP decreases from 2 to 12 bar but suddenly increases at 13 bar, which is over the maximum pressure that this hollow fiber support can endure. Fiber II shows an improved initial water permeability at 2 bar because it has bigger and broader pore sizes. However, comparing to Fiber I, it has a similar burst pressure of 13 bar. With the addition of EG into the bore fluid, Fiber III has a slightly enhanced burst pressure of 14 bar because EG not only reduces the number of macrovoids but also induces a slower but

Spinning direction

I

II

III

Raa(nm)=5.42±0.36 (nm)=5 42 0 36

(nm)=4.68±0.19 Ra(nm)=4 68 0 19

Ra(nm)=5 (nm)=5.38±0.19 38 0 19

IV

Ra(nm)=5 (nm)=5.58±0.07 58 0 07

Fig. 4. AFM images of inner surfaces of as-spun PI hollow fiber supports. aRa is the mean roughness.

606

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

-1

Probability density function (mm )

0.6 0.5

Fiber I (Water) (W t /NMP) Fiber II (Water/NMP) Fib Fiber III ((Water/EG/NMP)) Fiber IV (Water/EG/NMP)

0.4 0.3 0.2 0.1 0.0 0

5

10

15

20

Pore diameter (nm) Fig. 5. Probability density function curves of P84 hollow fiber supports.

Table 2 Pore characteristics in P84 hollow fiber supports. Pore characteristics

Hollow fiber code II

III

IV

4.46 13.4 1.20

7.49 41.4 1.37

3.61 29.3 2.18

3.51 26.1 2.08

300

14 bar 250

-2

-1

-1

Pure water permeability (l m h bar )

Mean pore diameter (nm) MWCO (kDa) Geometric standard deviation

I

13 bar

200

24 bar

150

more uniform phase inversion process [24]. Fiber IV was made by a small dual-layer spinneret and exhibits a greatly improved burst pressure of 24 bar. To understand the variation of burst pressure, the mechanical properties of these hollow fiber supports were investigated. The elongation at break, tensile strength, Young’s modulus, and toughness are listed in Table 3. For all hollow fiber supports, the elongation at break increases in the order of fiber I < fiber II < fiber III < fiber IV. The weak ductility of fiber I was resulted from the rapid phase inversion because water is a strong non-solvent. The nascent P84 hollow fiber was quickly precipitated that instantly frozen the inner structure with high asymmetry, thus reduced the ductility of the as-spun hollow fiber supports. The delayed demixing induced by bore fluids comprising NMP and/or EG leads to more relaxed polymeric chains that can absorb the elongating force. Moreover, Krok and Pamula and Ali et al. reported that the incorporation of glycol additives may decrease the rigidity of polymer chains [45,46] and thus toughen the hollow fiber supports. By integrating the stress–strain curve, Table 3 also summarizes the calculated toughness. Fiber I exhibits the smallest toughness of 0.56  106 J m3 because it has a high asymmetric inner skin structure due to rapid phase inversion and the most macrovoids among all fibers. As anticipated, fiber II with fewer macrovoids than that of fiber I displays a much better toughness (1.08  106 J m3). Further diminishes in phase inversion rate and the quantity of macrovoids significantly improves the toughness of fibers III and IV (both about 1.5  106 J m3). The improved toughness greatly contributes to the enhanced burst pressure of fibers III and IV. The Barlow’s formula is commonly used in the industry to predict the burst pressure of isotropic tubes [36]. Since all four hollow fiber supports have similar wall thicknesses, their tensile strengths and ODs are critical to the estimation of burst pressure. As shown in Table 3, fiber IV with the smallest dimension possesses the highest estimated burst pressure, followed by fiber III, fibers II and I. This result agrees quite well with the experimental results, indicating that hollow fiber supports fabricated from a delayed demixing during the phase inversion process and with a small dimension might be preferential for the high pressure PRO process.

13 bar 3.3. Morphology, transport properties and PRO performance of TFC hollow fiber membranes

100

Fiber I (Water) Fiber II (Water/NMP) Fiber III (Water/EG/NMP) Fiber IV (Water/EG/NMP)

50 0 0

4

8

12

20

16

24

Applied pressure (bar) Fig. 6. Burst pressures of P84 hollow fiber membranes (supports). A pure water stream pressurized under a particular pressure was running through the lumen of the hollow fiber supports. The down-stream pressure was recorded at each point.

Fig. 7 illustrates the active polyamide layers (i.e., TFC layers) of TFC hollow fiber membranes after interfacial polymerization on their inner surfaces. The TFC layer is fully covered by many small globules and few worm-like domains. The globular sizes are smaller than our previously reported for polyamide-imide thin-film composite (PAI-TFC) membranes where large surface pores of 10–20 nm existed on the substrate and resulted in large sizes of globular and worm-like domains [18]. Since the current hollow fiber supports have small pore sizes of 3–8 nm, the monomer

Table 3 Mechanical properties, hollow fiber dimensions, and burst pressures of as-spun P84 hollow fiber membranes. Properties

Elongation at break (%) Tensile strength (MPa) Young’s modulus (MPa) Toughnessa (106 J m3) OD/ID (lm/lm) Thickness (lm) Burst pressure (experimental, bar) Burst pressure (estimated, bar) a

Hollow fiber code I

II

III

IV

16.0 ± 1.8 4.1 ± 0.6 196.3 ± 11.4 0.56 ± 0.09 1175/782 196.6 13 13.7

28.8 ± 3.0 4.4 ± 0.5 215.8 ± 5.7 1.08 ± 0.21 1244/863 190.8 13 13.5

35.1 ± 4.6 4.7 ± 0.8 211.4 ± 11.3 1.53 ± 0.29 1343/928 206.5 14 14.5

38.0 ± 1.9 4.4 ± 1.4 222.8 ± 6.1 1.52 ± 0.31 1011/630 190.3 24 16.6

Toughness was calculated by taking the integral underneath the stress–strain curve.

607

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

100 nm

100 nm

100 nm

TFCTFC II

TFC-II

100 nm

TFC-III

TFC-IV

Cross-section (inner surface)

I Inner surface f

Fig. 7. SEM morphologies of active polyamide layers (TFC layers) of P84-TFC hollow fiber membranes.

(0.075–0.138 L m2 h1 bar1). As the intrinsic properties of TFC membranes, these permeability values indicate moderate water fluxes and low reverse salt fluxes in the subsequent osmotic operation. Fig. 8 shows the membrane performance in high pressure PRO tests in terms of water flux and power density as a function of pressure. Based on the burst pressure data, the maximum pressures that these TFC hollow fiber membranes can endure are 12, 12, 13, or 23 bar for TFC-I, II, II, and IV, respectively. Therefore, conditioning under high pressures of 10, 10, 12, and 21 bar were conducted respectively to these four membranes prior to PRO tests. The PRO tests were conducted in a safe pressure range to avoid the ‘‘bursting’’ phenomenon where the water flux reversely flows across the membrane from the draw solution side into the fresh water side. As shown in Fig. 8A–D, all four TFC membranes demonstrate similar moderate initial water fluxes of 24–27 LMH at 0 bar after conditioning. Their water fluxes decline with increasing DP because of the reduction in effective driving force, while reverse salt fluxes increase. TFC-I and TFC-II possess higher resistance to salt because of their initial salt fluxes of 3 g m2 h1 at 0 bar and slower increase trends with increasing pressure, while

Table 4 Transport properties of the four TFC membranes. Hollow fiber code

Water permeability, A (L m2 h1 bar1) Salt permeabilitya, B (L m2 h1) Salt rejectiona (%)

TFC-I

TFC-II

TFC-III

TFC-IV

0.863 0.078 97.9

0.904 0.138 96.5

0.918 0.075 98.1

0.906 0.086 97.8

a 1000 ppm NaCl as the feed solution in RO tests under an applied pressure of 5 bar.

migration during interfacial polymerization tends to be confined and results in a smoother surface [47]. The resultant TFC layers have a uniform morphology with a thickness of 100 nm regardless of hollow fiber supports. Table 4 summarizes the transport properties of the four TFC hollow fiber membranes. The water permeability was obtained from RO tests and the salt permeability was calculated from Eq. (8) which is related to the salt rejection in RO tests. All TFC membranes exhibit moderate water permeabilities (0.863–0.918 L m2 h1 bar1) and low salt permeabilities

Water flux (LMH)

(A)

W Water (2nd run)

Salt

Salt (2nd run)

20

40

20

30

10 0

0

14 12

4

8

12

16

20

0 0

14

(E)

1st run 2nd run

10

4

8

12

16

20

20

30 20

20

10

10 0

0 0

4

8

12

16

20

8

8

8

8

6

6

6

6

4

4

4

4

2

2

2

2

0

0

0

0

12

16

20

0

4

8

12

16

20

0

4

8

12

16

20

4

8

12

16

20

8

12

16

20

(H)

12 10

8

0 0

14

(G)

12

0

10

4

30 20

10

10

0

50 40

40

14

(F)

12

0

60

(D)

50 30

20 10 10

10

TFC-IV

60

(C)

50 30

40

20

10

TFC-III

60

(B)

50 30

30

0

Power density (W m-2)

TFCTFC II

60

Water

Reverse salt flux (gMH)

TFC-II 30

0

4

Hydraulic pressure difference (bar) Fig. 8. Membrane performance in PRO tests using de-ionized water as the feed solution and 1 M NaCl as the draw solution. Top (A–D): water flux and reverse salt flux, and bottom (E–H): power density as a function of hydraulic pressure difference. (The empty dots are repeated data from the 2nd run using the same membrane after PRO tests. The flow rates for both feed and draw solutions are 0.1 L min1.

608

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

TFC-III and TFC-IV have smaller salt resistance because of their initial salt fluxes of 10–20 g m2 h1 at 0 bar and higher increase trends with increasing pressure. Fig. 8E–H shows the power density of these TFC membranes. The empty dots shown in the figure are the repeated data from the 2nd run using the same membranes after the 1st PRO tests. These results confirm the performance stability. TFC-I, -II, and -III exhibit the maximum power density of around 5–6 W m2. On the other hand, the TFC-IV, which was spun with a small dimensional spinneret, shows not only superior mechanical properties and burst pressure, but also an excellent maximum power density of 12 W m2 at 21 bar. In addition, the smaller dimension leads to a higher linear lumen velocity under the same flow rate, which can subsequently lower the external concentration polarization in the lumen and produce a better PRO performance. Clearly, the TFC-IV hollow fiber membrane is very promising for commercially osmotic power generation. Based on Eqs. (15) and (16), the structural parameter St of the TFC-IV hollow fiber membrane is calculated to be 685 lm due to its thick wall (190 lm) and the relatively not wettable nature of P84. This value is slightly higher than other TFC hollow fiber membranes (595 lm) [22], but comparable with some TFC flatsheet membranes (670–710 lm) and orders of magnitude smaller than RO membranes (14–38 mm) [48]. The peak power density mathematically calculated from Eq. (10) is 11.1 W m2 at the hydraulic pressure of 21 bar, which is slightly lower than the experimental value of 12 W m2. It has been reported that the experimental results in high pressure PRO processes usually deviate from the predicted data based on membrane transport properties determined in FO and RO tests [28,36]. The slightly higher power density is possibly due to the minimization of defects and membrane deformation such as wall thinning when subjected to high pressure operations.

a plateau at DP in the range of 12–18 bar when Q is doubled to 0.2 L min1. A similar plateau was also observed at Q of 0.3 L min1. In addition, slight increases in water flux were noticed at DP of 21 bar under Q of 0.2 and 0.3 L min1. The empty dots shown in Fig. 9 are also the repeated data from the 2nd run using the same membranes after the 1st PRO tests. The 2nd run membrane has a much higher water flux and salt flux than the 1st run membrane when Q are equal to 0.2 and 0.3 L min1. Clearly, a high Q deteriorates the stability of membrane performance. Fig. 9D–F shows the variation of power density as a function of Q. The power density in the 2nd run deviates from the 1st run at Q of 0.2 and 0.3 L min1 and surges sharply when DP > 10 bar. The major effect of the variation in PRO performance is probably attributed to the micro-deformation of the TFC layer under high pressures (e.g., 18–21 bar for the current case). It has been brought to our attention that the high hydraulic pressure in PRO tests may have adverse effects on the TFC layer of the membrane. Although no apparent damage or delamination was observed, one may expect minor structural deformation, defect formation, and chain distortion in the TFC layer [29]. Such structural deformation may be reversible in the low flow rate operation (i.e., 0.1 L min1), but become serious in the high flow rate operation (i.e., 0.2–0.3 L min1). The actually local pressure on the inner wall of hollow fibers is higher than the one recorded at the down-stream pressure gauge. Table 5 compares the estimated pressure drop with the experimental pressure drop of the draw solution in the PRO operation under different conditions. Theoretically, the pressure drop is proportional to flow rate determined by the Hagen–Poiseuille equation [49]. The estimated pressure drops are 0.151, 0.302, 0.453 bar, respectively with different Q. In reality, the rough inner surface and the micro-deformation of the TFC layer may induce higher resistance to flow and further increase the pressure drop. Therefore, the reduction in pressure may be amplified by a higher Q because of the irreversible structural deformation and flow friction. Consistent with our hypothesis, the brine solution flowing through the lumen channel of the membrane module shows a small experimental pressure drop of <0.2 bar under a low flow rate of 0.1 L min1, while a high flow rate of 0.2 L min1 exhibits a large pressure drop of >0.4 bar. Clearly, the experimental pressure drop with a high flow rate significantly deviates from the estimated value. As a result, the downstream pressure measured by an external

3.4. Effect of flow rate on the PRO performance of TFC hollow fiber membranes

Power density (W m-2)

0.1 L min -1

40

(A)

Water Salt

30

Water (2nd run) Salt

(2nd

run)

50 40 30

20

20

10 0

60

10 0 0

24

4

(D)

20

8

12

16

0.2 L min -1

40

(B)

50

30

40 30

20

20

10

10 0

0

20

0 24

1st run 2nd run

60

4

8

12

16

12

8

8

4

4

4

0

0

20

0

4

8

12

16

20

0

4

8

12

16

20

8

12

16

20

(F)

20

8

16

0

24

16

12

20 10 0

12

8

30

0

16

4

40

10

12

60 50

20

16

0

(C)

30

20

(E)

20

0.3 L min -1

40

Reverse salt flux (gMH)

Water flux (LMH)

The TFC-IV fiber was chosen to further investigate the effect of counter-current flow rate Q on PRO performance. Fig. 9A–C illustrates the variations of water flux and reverse salt flux as a function of Q. As aforementioned, the water flux decreases with increasing DP when Q is 0.1 L min1. However, the water flux tends to reach

0

4

Hydraulic pressure difference (bar) Fig. 9. TFC-IV membrane performance in PRO tests using de-ionized water as the feed solution and 1 M NaCl as the draw solution. Top (A–C): water flux and reverse salt flux, and bottom (D–F): power density as a function of hydraulic pressure difference. The empty dots are repeated data from the 2nd run using the same membrane after PRO tests.

609

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610 Table 5 Pressure drops of draw solutions in high pressure PRO operations. Flow rate of draw solution

1

Linear velocity (m s ) Pressure drop (estimateda, bar) Pressure drop (experimentalb, bar) Rec

0.1 L min1

0.2 L min1

0.3 L min1

1.1 0.151 0.182 ± 0.023 680

2.1 0.302 0.492 ± 0.024 1361

3.2 0.453 0.818 ± 0.021 2041

lL a Pressure drop estimated by the Hagen–Poiseuille equation: DP ¼ 8Q pr4 where Q is the volumetric flow rate, l is the dynamic viscosity, L is the length of module, r is the radius of the flowing channel. b Real pressure drop when the draw solution flows through the hollow fiber module at a particular flow rate. c Re is the Reynolds number.

gauge is able to represent the actual pressure inside the membrane module when flow rate is 0.1 L min1. However, in the cases of flow rate P0.2 L min1, the pressure drop increases significantly, thus underestimates the actual pressure applied on the inner fiber wall. The recorded pressure at the down-stream gauge is essentially lower than the actually local pressure inside the fiber. This overpowering local pressure may not only damage the inner membrane skin but also deteriorate the PRO performance. In summary, choosing an appropriate flow rate of the brine solution is critical to the optimal TFC membrane performance for power generation. Future works should be focused on the complicated science and relationship among membrane mechanical properties, local pressure and membrane deformation, and flow rate of draw solutions, to ensure the membrane sustainability for osmotic power generation under a certain magnitude of flow rate Q. 4. Conclusions We have strategically designed a series of novel TFC-PRO hollow fiber membranes via interfacial polymerization. Firstly, hollow fiber membrane supports with a high degree of concentricity and minimal defects were fabricated by employing dual-layer spinnerets during spinning. Secondly, a thin-film dense polyamide layer was synthesized on top of the inner surface of the hollow fiber support. By adjusting the spinning conditions, hollow fiber supports with different morphologies and characteristics as well as TFC membranes with different hollow fiber supports and performance were prepared. The following conclusions can be drawn from this work: (1) The TFC hollow fiber membranes synthesized on small dimensional hollow fiber supports, which were made by delayed demixing, not only show superior mechanical properties and burst pressure, but also exhibit an excellent maximum power density of 12 W m2 at 21 bar using water and 1 M NaCl as feeds (24 °C). This type of hollow fiber supports has much better mechanical properties comparing with other hollow fiber supports in terms of elongation at break, tensile strength, Young’s modulus, and toughness. (2) Hollow fiber supports spun via a dual-layer spinneret by using NMP in the outer channel as an external coagulant and non-solvents (i.e., water, water/NMP, and water/EG/ NMP) in the bore channel as internal coagulants have a fully porous outer surface and a relatively dense inner surface. The addition of NMP and EG into the bore fluid promotes delayed demixing, increases the interconnectivity of pores, broadens the pore size distribution, and enlarges the inner diameter of hollow fiber supports. (3) The flow rate (Q) of draw solutions greatly affects the PRO performance. A high Q may lead to an irreversible structural deformation and high friction on the TFC membrane. In this

study, the suitable flow rate for the PRO operation is 0.1 L min1 (i.e., velocity of 1.1 m s1), which not only exhibits the lowest pressure drop but also delivers consistent PRO performance.

Acknowledgments This research was funded under the project entitled ‘‘Membrane development for osmotic power generation, Part 1. Materials development and membrane fabrication’’ (1102-IRIS-11-01) and NUS Grant Number of R-279-000-381-279. This research grant is supported by the Singapore National Research Foundation under its Environmental & Water Technologies Strategic Research Programme and administered by the Environment & Water Industry Programme Office (EWI) of the PUB. Special thanks are due to Dr. D.W. Mangindaan, Dr. P. Sukitpaneenit, Dr. P. Wang, Dr. H. Wang, Dr. S. Zhang, and Miss. J. Gao for their useful comments and suggestions. References [1] Ramon GZ, Feinberg BJ, Hoek EMV. Membrane-based production of salinitygradient power. Energy Environ Sci 2011;4:4423–34. [2] Chung TS, Li X, Ong RC, Ge QC, Wang HL, Han G. Emerging forward osmosis (FO) technologies and challenges ahead for clean water and clean energy applications. Curr Opin Chem Eng 2012;1:246–57. [3] Gerstandt K, Peinemann K-V, Skilhagen SE, Thorsen T, Holt T. Membrane processes in energy supply for an osmotic power plant. Desalination 2008;224:64–70. [4] Achilli A, Childress AE. Pressure retarded osmosis: from the vision of Sidney Loeb to the first prototype installation — review. Desalination 2010;261:205–11. [5] Loeb S. Production of energy from concentrated brines by pressure-retarded osmosis. I. Preliminary technical and economic correlations. J Membr Sci 1976;1:49–63. [6] Loeb S, Van Hessen F, Shahaf D. Production of energy from concentrated brines by pressure-retarded osmosis, II. Experimental results and projected energy costs. J Membr Sci 1976;1:249–69. [7] Loeb S, Mehta GD. A two-coefficient water transport equation for pressureretarded osmosis. J Membr Sci 1979;4:351–62. [8] Lee K, Baker R, Lonsdale H. Membranes for power generation by pressureretarded osmosis. J Membr Sci 1981;8:141–71. [9] Mehta GD, Loeb S. Performance of permasep b-9 and b-10 membranes in various osmotic regions and at high osmotic pressures. J Membr Sci 1979;4:335–49. [10] Qin JJ, Liberman B, Kekre KA. Direct osmosis for reverse osmosis fouling control: principles, applications and recent developments. Open Chem Eng J 2009;3:8–16. [11] Phuntsho S, Shon HK, Hong S, Lee S, Vigneswaran S. A novel low energy fertilizer driven forward osmosis desalination for direct fertigation: evaluating the performance of fertilizer draw solutions. J Membr Sci 2011;375:172–81. [12] Skilhagen SE, Dugstad JE, Aaberg RJ. Osmotic power–power production based on the osmotic pressure difference between waters with varying salt gradients. Desalination 2008;220:476–82. [13] Thorsen T, Holt T. The potential for power production from salinity gradients by pressure retarded osmosis. J Membr Sci 2009;335:103–10. [14] Sivertsen E, Holt T, Thelin W, Brekke G. Modelling mass transport in hollow fibre membranes used for pressure retarded osmosis. J Membr Sci 2012;417– 418:69–79.

610

X. Li, T.-S. Chung / Applied Energy 114 (2014) 600–610

[15] Yip NY, Tiraferri A, Phillip WA, Schiffman JD, Elimelech M. High performance thin-film composite forward osmosis membrane. Environ Sci Technol 2010;44:3812–8. [16] Chou S, Wang R, Shi L, She Q, Tang CY, Fane AG. Thin-film composite hollow fiber membranes for pressure retarded osmosis (PRO) process with high power density. J Membr Sci 2012;389:25–33. [17] Zhang S, Fu FJ, Chung TS. Substrate modifications and alcohol treatment on thin film composite membranes for osmotic power. Chem Eng Sci 2013;87:40–50. [18] Li X, Zhang S, Fu FJ, Chung TS. Deformation and reinforcement of thin-film composite (TFC) polyamide–imide (PAI) membranes for osmotic power generation. J Membr Sci 2013;434:204–17. [19] Han G, Zhang S, Li X, Chung TS. High performance thin film composite pressure retarded osmosis (PRO) membranes for renewable salinity-gradient energy generation. J Membr Sci 2013;440:108–21. [20] Fu FJ, Zhang S, Sun SP, Wang KY, Chung TS. POSS-containing delamination-free dual-layer hollow fiber membranes for forward osmosis and osmotic power generation. J Membr Sci 2013;443:144–55. [21] Han G, Zhang S, Li X, Widjojo N, Chung TS. Thin film composite forward osmosis membranes based on polydopamine modified polysulfone substrates with enhancements in both water flux and salt rejection. Chem Eng Sci 2012;80:219–31. [22] Wang R, Shi L, Tang CY, Chou S, Qiu C, Fane AG. Characterization of novel forward osmosis hollow fiber membranes. J Membr Sci 2010;355:158–67. [23] Widjojo N, Chung TS, Weber M, Maletzko C, Warzelhan V. The role of sulphonated polymer and macrovoid-free structure in the support layer for thin-film composite (TFC) forward osmosis (FO) membranes. J Membr Sci 2011;383:214–23. [24] Sukitpaneenit P, Chung TS. High performance thin-film composite forward osmosis hollow fiber membranes with macrovoid-free and highly porous structure for sustainable water production. Environ Sci Technol 2012;46:7358–65. [25] La YH, Sooriyakumaran R, Miller DC, Fujiwara M, Terui Y, Yamanaka K, et al. Novel thin film composite membrane containing ionizable hydrophobes: pH dependent reverse osmosis behavior and improved chlorine resistance. J Mater Chem 2010;20:4615–20. [26] Baker RW. Membrane technology and applications. 2nd ed. England: John Wiley & Sons Ltd.; 2004. [27] Nunes SP, Peinemann K-V. Membrane technology in the chemical industry. Germany: Wiley-VCH Verlag GmbH; 2001. [28] Zhang S, Chung TS. Minimizing the instant and accumulative effects of salt permeability to sustain ultrahigh osmotic power density. Environ Sci Technol 2013;47:10085–92. [29] X. Li, T.S. Chung, Effects of free volume in TFC membranes on osmotic power generation. AIChE J. doi:10.1002/aic.14217, (in press). [30] Su JC, Zhang S, Chen HZ, Chen HM, Jean YC, Chung TS. Effects of annealing on the microstructure and performance of cellulose acetatemembranes for pressure-retarded osmosis processes. J Membr Sci 2010;364:344–53. [31] Hsieh FH, Matsuura T, Sourirajan S. Reverse osmosis separations of polyethylene glycols in dilute aqueous solutions using porous cellulose acetate membranes. J Appl Polym Sci 1979;23:561–73. [32] Singh S, Khulbe K, Matsuura T, Ramamurthy P. Membrane characterization by solute transport and atomic force microscopy. J Membr Sci 1998;142:111–27.

[33] Zhang S, Wang KY, Chung TS, Chen H, Jean YC, Amy G. Well-constructed cellulose acetate membranes for forward osmosis: minimized internal concentration polarization with an ultra-thin selective layer. J Membr Sci 2010;360:522–35. [34] Wang KY, Matsuura T, Chung TS, Guo WF. The effects of flow angle and shear rate within the spinneret on the separation performance of poly(ethersulfone) (PES) ultrafiltration hollow fiber membranes. J Membr Sci 2004;240:67–79. [35] Youm KH, Kim WS. Prediction of intrinsic pore properties of ultrafiltration membrane by solute rejection curves: effects of operating conditions on pore properties. J Chem Eng Jpn 1991;24:1–7. [36] Zhang S, Sukitpaneenit P, Chung TS. Strategies and illustrations of ultrahigh power density membranes for osmotic energy production. Chem. Eng. J. (accepted). [37] Wang KY, Yang Q, Chung TS, Rajagopalan R. Enhanced forward osmosis from chemically modified polybenzimidazole (PBI) nanofiltration hollow fiber membranes with a thin wall. Chem Eng Sci 2009;64:1577–84. [38] Achilli A, Cath TY, Childress AE. Power generation with pressure retarded osmosis: an experimental and theoretical investigation. J Membr Sci 2009;343:42–52. [39] Yip NY, Tiraferri A, Phillip WA, Schiffman JD, Hoover LA, Kim YC, et al. Thinfilm composite pressure retarded osmosis membranes for sustainable power generation from salinity gradients. Environ Sci Technol 2011;45:4360–9. [40] McCutcheon JR, McGinnis RL, Elimelech M. Desalination by ammonia-carbon dioxide forward osmosis: influence of draw and feed solution concentrations on process performance. J Membr Sci 2006;278:114–23. [41] Lobo VMM. Mutual diffusion coefficients in aqueous electrolyte solutions (technical report). Pure Appl Chem 1993;65:2613–40. [42] Peng N, Widjojo N, Sukitpaneenit P, Teoh MM, Lipscomb GG, Chung TS, et al. Evolution of polymeric hollow fibers as sustainable technologies: past, present, and future. Prog Polym Sci 2012;37:1401–24. [43] Sukitpaneenit P, Chung TS. Molecular design of the morphology and pore size of PVDF hollow fiber membranes for ethanol–water separation employing the modified pore-flow concept. J Membr Sci 2011;374:67–82. [44] Liu Y, Koops GH, Strathmann H. Characterization of morphology controlled polyethersulfone hollow fiber membranes by the addition of polyethylene glycol to the dope and bore liquid solution. J Membr Sci 2003;223:187–99. [45] Krok M, Pamula E. Poly(L-lactide-co-glycolide) microporous membranes for medical applications produced with the use of polyethylene glycol as a pore former. J Appl Polym Sci 2012;125:E187–99. [46] Ali M, Zafar M, Jamil T, Butt MTZ. Influence of glycol additives on the structure and performance of cellulose acetate/zinc oxide blend membranes. Desalination 2011;270:98–104. [47] Li X, Wang KY, Helmer B, Chung TS. Thin-film composite membranes and formation mechanism of thin-film layers on hydrophilic cellulose acetate propionate substrates for forward osmosis processes. Ind Eng Chem Res 2012;51:10039–50. [48] Wei J, Qiu C, Tang CY, Wang R, Fane AG. Synthesis and characterization of flatsheet thin-film composite forward osmosis membranes. J Membr Sci 2011;372:292–302. [49] Welty J, Wicks CE, Wilson RE, Rorrer GL. Fundamentals of momentum, heat and mass transfer. 5th ed. USA: John Wiley & Sons, Inc.; 2008.