- Email: [email protected]
Forward osmosis using draw solutions manifesting liquid-liquid phase separation
Desalination xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Desalination journal homepage: www.elsevier.com/locate/desal
Forward osmosis using draw solutions manifesting liquid-liquid phase separation Siavash Darvishmanesh, Brian A. Pethica, Sankaran Sundaresan⁎ Department of Chemical and Biological Engineering, Princeton University, Princeton, NJ 08544, USA
A R T I C L E I N F O
A B S T R A C T
Keywords: Forward osmosis Lower critical solution temperature Glycol ethers Liquid-liquid phase separation
Desalination via forward osmosis using draw agents whose regeneration is aided via liquid-liquid phase separation has gained much attention in recent years. In the present study, mixtures of two different glycol ethers, tripropylene glycol methyl ether and tripropylene glycol n-butyl ether, have been studied as potential draw agents. Water activity, viscosity and diffusion coefficient of draw solutions have been measured at different mixture compositions, concentrations and temperatures. Osmotic pressures of these draw solutions decreases strongly with increasing temperature. Forward osmosis experiments performed with these draw solutions reveal appreciable initial loss of trans-membrane water flux, reverse solute flux and severe concentration polarization.
1. Introduction In forward (or direct) osmosis (FO), water from an aqueous solution selectively passes through a membrane to a second solution (referred to as the draw solution) at the same pressure based solely on the difference in the water activity (osmotic pressure) of the two solutions [1]. A number of studies have explored the feasibility and benefits of FO as an alternative to reverse osmosis (RO) for seawater desalination [2–6]. The overall cost of producing desalinated water from seawater via FO process is affected by the type of draw solution employed and its regeneration. Regeneration processes using waste heat or geothermal sources are being explored as routes to cost-effective desalination [7–9]. For example, a recyclable salt solution (a mixture of ammonium bicarbonate and ammonium hydroxide dissolved in water) has been employed as a draw agent in FO, in which low-grade heat can be employed to remove this thermolytic salt from diluted draw solution [10]. In another proposed process, FO is coupled with membrane distillation (MD) [11], where FO is applied to reduce membrane fouling and scaling which is detected in pressure-driven membrane processes such as RO and nanofiltration (NF); MD process separates the water and regenerates the draw solution for FO using waste heat or geothermal sources. Aqueous solutions of thermo-responsive organics (chemicals or polymers) showing lower critical solution temperature (LCST) have also been explored as draw solutes [12–17]; in this case, draw solution diluted via water extraction in an FO process step is heated to a temperature above LCST to induce spontaneous phase separation into organic-rich and water-rich phases. The water-rich phase is then further upgraded through second-stage RO or NF. The use of thermo-responsive ⁎
organics as draw solutions for FO was first patented in 1968 [18]; however lack of suitable FO membranes thwarted its application. Recently, modifications of this concept for a draw solution manifesting LCST have been patented for integrated FO-NF systems [12,13]. Two types of FO processes using thermo-responsive materials, whose aqueous solutions manifest LCST, have been reported in the literature: (i) Direct use of the thermoresponsive material as a draw agent [14,17]; (ii) Indirect use by formation of aqueous two-phase system with an inorganic draw agent [19]. Thermo-responsive draw agents include block copolymers of polyethylene oxides and polypropylene oxides [12] and fatty acid or fatty alcohol polyethylene glycols polymers [13]. Concentrated aqueous solutions of these molecules are much more viscous than aqueous solutions of inorganic salts (such as MgSO4), leading to severe internal concentration polarization (ICP) in the membrane's support layer (in the so-called FO mode where the support layer faces the draw solution) as well as external concentration polarization (ECP) in the flowing draw solution (in both FO mode of operation and the so-called Pressure Reduced Osmosis (PRO) mode). Furthermore, these organic molecules tend to foul the membrane [20]. Indirect use of thermo-responsive inorganics for desalination was introduced by Rajagopalan et al. [19] as the Aquapod© desalination process. In this method, an aqueous solution of an inorganic salt (specifically MgSO4) is employed as draw agent in the FO step. The diluted draw solution is sent to an aqueous two-phase contactor, where it is concentrated by extracting water using a concentrated aqueous solution of UCON660©, which is a polyethylene oxide-polypropylene oxide block copolymer. The polymer-rich phase is then separated and heated
Corresponding author. E-mail address: [email protected] (S. Sundaresan).
http://dx.doi.org/10.1016/j.desal.2017.05.036 Received 16 January 2017; Received in revised form 30 May 2017; Accepted 30 May 2017 0011-9164/ © 2017 Elsevier B.V. All rights reserved.
Please cite this article as: Darvishmanesh, S., Desalination (2017), http://dx.doi.org/10.1016/j.desal.2017.05.036
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
at which the draw agent turned cloudy (to the eye). This experiment was repeated several times, after allowing the sample to cool down yielding a transparent single-phase. For a given composite draw agent, Tcl changed with wt% GE, and the minimum was recorded as its LCST. Two instruments were tested for the osmotic pressure measurement: Knauer K-7000 Vapor Pressure Osmometer (VPO) [22] and the AquaLab Tunable Diode Laser (TDL) infrared vapor pressure meter [23]. The VPO measures the difference in temperature between a drop of the test solution on a thermistor and a reference solution (which is pure water) in a closely thermostatted cell ( ± 0.001 °C) which results from condensation of water vapor on the test drop [22]. The TDL instrument measures water vapor concentration from the absorption of an infrared laser beam passing through (an equilibrated) vapor phase in contact with the test liquid in a sealed thermostatted cell. The temperature in the chamber is accurate to ± 0.2 °C [23]. The water activity can be estimated with an accuracy of ± 0.005. Both methods were calibrated using as reference standard solutions of NaCl [24] or pure water (unit activity). Measurements with the VPO instrument manifested slow long-term drift with GE solutions, particularly at higher concentrations (but not with solutions of inorganic salts). In contrast, the TDL instrument consistently yielded reproducible measurements. A key test of the reliability of an instrument was made as follows. By heating a draw solution to a chosen temperature above its cloud point the solution was allowed to separate into two phases. After allowing ample time for the two phases to equilibrate (typically overnight), the two phases were separated and the water activities of the two phases were measured at the chosen phase separation temperature. The TDL instrument yielded nearly identical water activities for both phases, which is what one would expect for phases in equilibrium; in contrast, the VPO yielded very different results. (See Table 1 summarizing results from TDL meter.) Therefore, in what follows, we present only the data obtained from the TDL instrument. The relationship between the osmotic pressure and water activity of sodium chloride solutions at various temperature can be found in the literature [24]. The water activities of various draw solutions measured in our study are converted to equivalent osmotic pressures of sodium chloride solution possessing the same water activity as the draw solution at the temperature of interest. This equivalent osmotic pressure (or more directly the water activity) is the most relevant metric to assess the driving force afforded by a draw solution for water extraction from seawater (which is mostly a sodium chloride solution) in an FO process. Viscosities were measured using an SI-Analytics Ubbelohde viscometer, where the temperature was controlled to within ± 1 °C. Self-diffusion coefficient measurements were made on 1H NMR using a Bruker Avance III 500 MHz NMR spectrometer. The 1D diffusion ordered spectroscopy (DOSY) experiments were run with 50 G/cm magnetic field gradient, and the data were analyzed using advanced Bayesian DOSY transformation method. Samples were mixtures of the two glycol ethers in D2O. The performance of GE mixtures as draw solutes in the FO system was tested on a lab-scale circulating filtration unit, as described by Cath et al. [25]. Cellulose acetate FO membrane was purchased from Fluid Technology Solutions, Inc. (USA). DI water and NaCl solution were employed as a feed solution. The Sepa CF042 solvent-stable cross-flow permeation cell was purchased from Sterlitech™ (USA); it has an active
to a temperature above its cloud point temperature (Tcl) to form two phases; a polymer-rich phase which is cooled and returned to the aqueous two-phase contactor, and a water-rich phase which is sent to a further upgrading step. Potential application of low molecular weight polymers and nonionic surfactants as FO draw agents has been examined by several other researchers. Polypropylene glycol (PPG425) as a draw agent for FO was assessed by Jørgensen [17]. The largest osmotic pressure recorded in this study was ~50 atm at 288 K, well below the magnitudes cited by patents [12,13] for their draw agents. Polyethylenimine (PEI) derivatives and glycol ethers (GEs), whose aqueous solutions manifest LCST and have lower viscosities (than aqueous solutions of polymers mentioned above), have been patented recently [21]. It appears that these draw solutions have rather low osmotic pressures and can barely draw water from synthetic seawater feed solution (3.5 wt% NaCl). Nakayama et al. [14] recently demonstrated extraction of water from seawater via FO using a draw solution consisting of an aqueous solution of diethylene glycol n-hexyl ether (DEH; 146.23 Da; LCST ~10 °C). It is likely that reverse solute diffusion would be more pronounced for this solute because of its low molecular weight, but reverse solute flux was not reported. These authors have also suggested an integrated FO-FO process [14,15]. Strong temperature (T) and composition dependencies of the osmotic pressure of aqueous solutions of thermo-responsive draw agents play central roles in the success of FO process. To the best of our knowledge, there is only one published study on the temperature dependence of osmotic pressure of aqueous solutions of thermoresponsive draw agents at conditions relevant to desalination via FO [17]. In this study, Jørgensen [17] found that the osmotic pressure of aqueous solutions of PPG425 increases with concentration at constant T. Furthermore, the osmotic pressure at fixed concentration increases appreciably with decreasing T, without any sign of reaching a plateau (in the temperature range studied). In other words, one could increase the driving force significantly by lowering the operating T; however, lowering T increases the viscosity of the draw solution and decreases the diffusivity appreciably, thereby lowering FO-based desalination process efficiency. Higher viscosity implies higher pumping cost and lower diffusivities increase the severity of concentration polarization (CP). On the other hand, lowering T could decrease reverse solute flux of the draw agent (as a result of lower diffusivity). Thus, experimental data and a good understanding of the concentration and temperature dependencies of osmotic pressure, viscosity and diffusivity would be valuable for FO process optimization. This consideration motivated the present study, where we have studied the temperature and concentration dependence of the phase behavior, osmotic pressure, kinematic viscosity and self-diffusivities of a model thermo-responsive material as potential draw agent. Specifically, we have studied mixtures of two GEs: Tripropylene glycol methyl ether (TPM) (206.27 Da) and Tripropylene glycol mono-n-butyl ether (TPnB) (248.35 Da). Aqueous solutions of each of these GEs manifest a LCST, but they are widely separated; by mixing them, one can tune the LCST. FO experiments were also performed at three different temperatures with a draw solution at one particular composition in order to identify achievable water flux, fouling and reverse flux characteristics. 2. Materials and methods
Table 1 Water activities of two phases obtained through phase separation of 50 wt% aqueous solution of a composite glycol ether mixture (50%TPM-50%TPnB). Measurements were made using TDL water activity meter (with accuracy ± 0.005).
DOWANOL™ TPM Glycol Ether, DOWANOL™ TPnB Glycol Ether and Deuterium oxide (D2O) were purchased from Sigma-Aldrich (USA). The two GEs were mixed at various weight ratios to prepare composite draw agents. The composite draw agent was then mixed with deionized water (DI) to make draw solutions. Each draw solution was thus characterized by wt% of GE and TPM-TPnB composition ratio. A test tube containing the draw solution was immersed and heated in a jacketed beaker connected to a water bath (Fisher Scientific Model 3013S Heating-Cooling Recirculator, USA) to identify the temperature 2
Temperature (K)
Water-rich phase water activity
Organic-rich phase water activity
293 303 313
0.971 0.974 0.977
0.974 0.973 0.981
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
membrane area of 42 cm2. The flow rates of both draw and feed solutions, which passed through the permeation cell channel in a cocurrent fashion, were controlled by flow meters. Cross flow velocity which (CFV) was calculated by dividing the volumetric flow rate [l pm] in the flow channel by the cross sectional area (0.0023 × 0.0392 m2) of the flow channel. The temperature of the whole FO system was kept at 295 ± 0.5 K (unless mentioned otherwise). In the FO experiments presented in this report, 50 wt% aqueous solution of a composite draw agent (87.5 wt% TPM-12.5 wt% TPnB) was used as draw solution. (The LCST of this composite agent is ~ 313 K.) A reservoir containing the draw solution was placed on an electronic balance (Ohaus Pioneer PA2202, USA), connected to a computer to record its mass as a function of time, from which one can readily extract water permeation flux, Jv (lit/m2·h, commonly abbreviated as LMH) as a function of time. Reverse solute flux Js, (g/m2·h, commonly abbreviated as gMH) refers to the rate at which draw solute is transported across the membrane from the draw side to the feed side. Average reverse flux over the course of an experiment was found from the initial and final volumes of the feed stream, the initial and final organic content in feed stream samples (measured using Total Organic Carbon (TOC) Analyzer (Shimadzu Scientific Instruments, USA)), and the duration of the FO experiment. Forward osmosis experiments were conducted using both possible orientations of the asymmetric membranes. In the so-called FO mode, the draw solution was on the support layer side and the feed was exposed directly to the active layer; this is the typical orientation in FO [26,27]. In the PRO mode, the draw solution was exposed to the active layer and the feed was on the support layer side. In an effort to understand challenges in the second-stage RO (finishing step), water flux and solute rejection were determined at 295 K in a high-pressure chemical-resistant HP4750 stirred Cell, Sterlitech™ (USA) for five different RO membranes (BW30, 73 U, ACM4, ACM3, GE-AK). The effective RO membrane area was 13.85 cm2. Pure water flux was first measured, after the membranes were compressed for 3 h at 100 psi. Subsequently, experiments were carried out with 200 ml of aqueous solutions of composite GE mixture (87.5 wt%TPM-12.5 wt% TPnB) with GE concentrations of 1 and 5 wt%. Solute rejection was determined from GE concentrations (measured through TOC analyzer) of the initial and final feed, and the accumulated permeate after filtration of 100 ml of the feed solution. Four different samples from each RO membrane were evaluated to determine average water permeability and solute rejection.
Fig. 1. Cloud point temperature loci of aqueous solutions of composite glycol ether mixtures with different TPM-TPnB weight percent ratios: (×) 80–20; (■) 70–30; (◊) 50–50.
based on the temperatures of the feed saline water to be desalinated and the available heating medium. Addition of TPM to TPnB affects the osmotic pressure significantly. This effect is presented in Fig. 2(a-c), where the equivalent osmotic pressure is plotted against temperature. Each panel shows the results for a particular GE loading level, and the different data sets in each panel correspond to different TPM:TPnB weight ratios. Under all conditions tested, the osmotic pressure decreases upon increasing T. Of the compositions tested, TPM afforded the highest osmotic pressure and addition of TPnB lowered the osmotic pressure. Thus, at a given temperature and total GE concentration level, increasing TPnB level lowers the driving force for FO afforded by the draw agent; on the other hand, as seen in Fig. 1, increasing TPnB level allows phase separation at a lower T and hence the process could operate with waste heat available at a lower T. Thus, an optimal composition would clearly be a compromise between an FO step (high osmotic pressure) and the subsequent phase separation step (low Tcl). Huang et al. [29] have observed that phase separation in these mixtures is accompanied by aggregate formation; the degree of aggregation typically increases as T increases, leading to a decrease in the osmotic pressure. In the results presented in Fig. 3, the rate of change in osmotic pressure with temperature does not manifest discernible temperature dependence at the lower temperatures studied, but this dependence can be seen more clearly at higher temperatures at least for the GE mixtures with no or small TPnB content. As the higher temperatures, the osmotic pressure becomes less sensitive to temperature, indicating a more gradual change in the degree of aggregation. The FO experiments described below were performed using aqueous solutions of glycol ether mixture with specific composition of 87.5 wt% TPM - 12.5 wt% TPnB. The lowest cloud point temperature of aqueous solutions of this composite GE mixture was found to be ~ 313 K. Effect of temperature on osmotic pressure of a 50 wt% aqueous solution of this composite mixture is presented in Fig. 3a. Fig. 3b shows the variation of equivalent osmotic pressure with GE wt% at 295 K. The inflection seen at intermediate concentrations is typical of thermo-responsive systems that manifest LCST [14]. Fig. 3c shows the same osmotic pressure data in Fig. 3b, except that it is now plotted against (g GE/g water), following Wilson and Stewart [30]; remarkably, the data now collapse into two nearly linear segments. Wilson [31] has suggested that these linear segments are indicative of aggregates of glycol ether in water matrix at low GE wt% and the inverse at high GE loading levels.
3. Results and discussion 3.1. Temperature dependence of equivalent osmotic pressure of glycol ether mixtures Fig. 1 shows the cloud point temperatures of aqueous GE solutions at different concentrations for composite draw agents containing different TPM/TPnB ratios. The water-TPM system has Tcl values in excess of 373 K, and could not be measured in our laboratory. This figure shows that addition of TPM to TPnB raises Tcl. For water-GE mixtures, whether hydrophobic or hydrophilic interactions dominate depends strongly upon T and the GE molecular structure [28]. Above LCST, the water–GE system manifests phase separation over a range of compositions. The LCST increases with a decrease in the carbon number of the end group because of decreasing hydrophobic character; for example, changing the end group from n-butyl (in TPnB) to methyl (in TMP) increases LCST from 253 K to a value in excess of 373 K. The Tcl locus of a mixture of TMP and TPnB lies between those of water-TMP and waterTPnB systems. Analogously, the LCST of water–GE system can also be manipulated by mixing glycol ethers, as one can readily surmise from Fig. 1. This is consistent with data reported by Christensen et al. [28]. Fig. 1 clearly shows that one can tune the phase separation temperature by adjusting the TMP-TPnB ratio in an effort to optimize the FO process 3
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
(a)
(b)
(c)
Fig. 3. a: Effect of temperature on the osmotic pressure of 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. b: Effect of GE concentration on the osmotic pressure of aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. T = 295 K. c: Effect of GE concentration on the osmotic pressure of aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. T = 295 K. The results in Fig. 3b have been replotted using a different scale for the GE concentration.
Fig. 2. Effect of temperature on osmotic pressure of Glycol ethers–water mixtures at three different concentrations and several composition ratios. TPM:TPnB weight ratio: (●) 100:0; ( ) 90:10; (×) 80:20; (Δ) 70:30; (□) 60:40; (◊) 50:50. Panel (a): 20 wt% aqueous solution; panel (b) 50 wt%; panel (c): 80 wt%.
associated with low viscosity (as viscosity is inversely proportional to diffusivity [33]) helps reduce the effects of CP. In the present study we sought to examine the temperature dependence of viscosity and diffusivity of glycol ether draw solutions. Fig. 4 shows the variation of kinematic viscosity and self-diffusion coefficient of a model glycol ether draw solution with temperature. The inverse relationship between viscosity and diffusivity is readily apparent. Increasing the temperature
3.2. Effect of temperature on viscosity and diffusivity of draw solution Ideal draw solutions should exhibit low viscosity, which would lower the pumping cost [32]; more importantly, the higher diffusivity 4
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
Fig. 4. The effect of temperature on kinematic viscosity and self-diffusion coefficient of aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. Open symbols: Diffusivity; Filled symbols: Kinematic viscosity. (■) 30 wt%; (●) 50 wt%; (▲) 70 wt%. Fig. 5. evolution of water permeation flux over time under FO mode using water as feed stream and 50 wt% aqueous solution of 87.5% TPM - TPnB 12.5% as draw stream. Attempted regeneration of membrane shows the flux reduction is irreversible.
leads to an appreciable decrease in the kinematic viscosity of the draw solution and an increase in the diffusivity, both of which are beneficial from transport considerations. (Increasing the temperature from 293 K to 313 K lowers the kinematic viscosity by 50% and increases the diffusion coefficient by factor of two.) However, this beneficial effect of increasing temperature is accompanied by a sharp decrease in the osmotic pressure of the draw agent (Fig. 3), which lowers the effectiveness of the draw agent in an FO process. These opposing effects of T imply that there would be an optimum temperature representing a compromise between these two considerations. Fig. 4 also shows the variation of kinematic viscosity of 30 wt% and 70 wt% aqueous solutions. As concentration of GE increases, the viscosity increases (and correspondingly, the diffusivity would decrease).
irreversible loss in trans-membrane water flux caused by GE draw solution, which is discussed later. In what follows, we only report the results obtained in the plateau region following irreversible performance loss. The performance of fresh FO membranes was tested with sodium chloride as draw solute and freshwater on the feed side, and the results are presented in Table 2. The water flux in the FO mode (i.e., the draw solution is on the support layer side) is less than that in the PRO mode (i.e., the draw solution is on the dense selective layer side), which can be attributed to ICP in the FO mode. Increasing the cross-flow velocity of the NaCl solution along the membrane surface increases the membrane flux, by lowering the severity of ECP; the sharp increase in transmembrane water flux even in the FO mode shows that in this mode, in addition to ICP, there must have been appreciable ECP outside the support layer that got weakened by increased flow rate. To assess the extent of loss in performance due to ICP and ECP, experiments were performed with freshwater on both sides and gentle applied pressure (of the order of a few atm) on one side to determine membrane permeability (see Fig. 6); such experiments yielded permeability of 0.62 LMH/atm for fresh membranes at 295 K, which is consistent with the findings of Jorgensen [17]. The osmotic pressure of 1 M NaCl solution is 47 atm and so the theoretical water flux in the absence of any CP (ICP and ECP) would be ~29 LMH, which is significantly larger than fluxes observed, revealing that appreciable concentration polarization persisted even at the high circulation rates [35] in both FO and PRO modes. The Reynolds numbers determined on the basis of cross-stream
3.3. FO process performance In an FO process, concentration polarization can affect performance on both sides of the membrane. On the feed side, the polarized layer becomes more concentrated than the bulk (if a solute that is rejected by the membrane is present), and on the draw side, the polarized layer is more dilute than the bulk; both lead to a decrease in the actual osmotic driving force across the membrane. The severity of these effects can be controlled to some extent with cross-flow and well-designed hydrodynamics [34]. Appreciable initial decrease in trans-membrane water flux was observed with glycol ether draw solutions. Fig. 5 illustrates performance loss, where forward osmosis experiments were performed in the FO mode. First, FO experiments were performed using 1 M NaCl solution as draw solution (step 1). No fouling was observed. The membrane was then washed by flushing fresh water through the cell and reused in an experiment where 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 was used as draw solution. The membrane flux decreased appreciably with time and appeared to level off after an hour (step 2). (As the draw solution is used in a circulation mode, the draw solution is constantly undergoing dilution; however, the extent of dilution over this time period is small. As a result, the change in flux cannot be attributed to this dilution, which can readily be inferred from the results for step 1.) The membrane was then washed and reused with the 1 M NaCl solution as draw agent (step 3). The water flux was measurably smaller than that in step 1, but no further loss was observed. Finally, the membrane was washed again and reused with the 50 wt% GE solution (step 4), recovering the water flux plateau observed at the end of step 2. These results demonstrate
Table 2 Water permeation flux under FO and PRO modes using water as feed stream and 1 M aqueous solution of NaCl as draw stream. Experiments were performed at 295 K. Pump flow rate#
1000 ml/minb (360/ 360)
200 ml/minb (72/ 72)
50 ml/minb (18/ 18)
FO mode (LMH)a PRO mode (LMH)
8.4 16.8
5.8 11.9
5.3 11.3
a
LMH = lit/m2/h. # same pump rate was used on both sides. The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides. The cross-sectional area of the channel available for cross-flow of the fluid is 0.23 × 3.92 cm. Circulation rates of 50, 200 and 1000 ml/min correspond to cross flow velocity of 0.009, 0.036 and 0.180 m/s, respectively; this conversion between pump flow rate and cross-flow velocity applies for Tables 3-6 as well. b
5
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
Table 4 Water permeation flux under FO and PRO modes using water as feed stream and 1 M aqueous solution of NaCl as draw stream. Experiments were performed at 305 K and 315 K. 305 K Pump flow rate# FO mode (LMH) PRO mode (LMH)
315 K
200 ml/mina (92/92) 7.1
50 ml/mina (23/23) 6.4
200 ml/mina (112/112) 7.5
50 ml/mina (28/28) 6.9
13.1
12.5
13.6
13.1
# Same pump rate was used on both sides. a The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides.
consequences of CP. The osmotic pressure of the GE draw solution employed in these experiments is 76 atm, which is higher than that of the 1 M NaCl solution (see Fig. 5); yet, lower water flux was achieved with GE draw solution (compare Tables 2 and 3), clearly indicating a greater overall resistance to trans-membrane water flux with GE draw solution. As expected, the water flux with GE draw solution is greater in the PRO mode than in FO mode [36]; this clearly shows that ICP plays an important role in the FO mode. With the GE draw solution, the water flux increased upon increasing the cross-flow rate of the draw solution (Table 3) as in the case of the 1 M NaCl draw solution (Table 2), but this increase is relatively weaker for the GE draw solution. Forward osmosis experiments were performed at two additional temperatures to supplement the results in Tables 2 and 3. These are presented in Table 4 for NaCl draw solution and Table 5 for GE solution. Temperature affects the performance of most membrane processes; for example, it affects both the solution viscosity and diffusivity of solutes, thereby influencing both the concentration and hydrodynamic boundary layers [37]. In the case of NaCl draw solution, the transmembrane flux increased with increasing T (compare Tables 2 and 4). In contrast, increasing the temperature to 305 K from 295 K led to a very small increase in water flux when GE solution was used as draw solution (compare Tables 3 and 5). This suggests that the loss in osmotic driving force based on bulk composition resulting from increase in T was compensated by a decrease in various resistances. However, further increase in temperature (to 315 K, see Table 5) led to a significant reduction in water flux, which can be attributed to acute loss of osmotic driving force. It is clear from Tables 4 and 5 that thermo-responsive materials manifesting LCST manifest, when used as draw solutions In FO process, manifest more complex dependence on T than draw solutions based on inorganic salts. Table 6 summarizes results obtained (after the initial period of performance loss) in experiments with a GE draw solution and synthetic seawater (0.6 M) in the feed side. The table presents data on membrane water flux, salt rejection and reverses draw solute flux. Experiments were run in the FO, PRO, pressure-assisted FO (AFO) and pressure-assisted PRO (APRO). Assisted FO and PRO experiments were performed with an applied pressure of 400 kPa. Visual observation of the
Fig. 6. Membrane water fluxes at different applied pressures. Membranes were precompacted at 400 kPa. Pure water flux was measured using fresh membranes and membranes exposed to glycol ether draw solution. Pressure was applied on: (●) dense layer of a fresh membrane; (□) support layer of a fresh membrane; (Δ) dense layer of a membrane which had been operated in a PRO mode with GE draw solution; (×) support layer of a membrane which had been operated in a FO mode with GE draw solution.
velocity of the draw solution, the draw solution kinematic viscosity and the smaller dimension of the cross-flow cross section are shown in Table 2 for the three different pump speeds. It is clear that the flow is in the laminar regime. As shown in Fig. 6, the permeability of membrane exposed to GE draw solution is lower (even after washing) than that of the fresh membrane, clearly indicating irreversible increase in membrane resistance. For fresh membranes, the orientation of the membrane does not affect the water permeability. In water permeability measurements with membranes that have been used in forward osmosis experiments using GE draw solution, pressure was applied to the side that was exposed to GE draw solution; if the forward osmosis was done in FO (PRO) mode, then pressurization was done on the support (dense layer) side during water permeability test. The water permeability of a membrane exposed to GE draw solution is less than that of a fresh membrane. Membrane used with GE draw solution in PRO mode had a lower pure water flux than that used in FO mode (discussed later). Effects of membrane orientation and draw solution cross-flow velocity on water flux achieved with a 50 wt% aqueous solution of a composite GE mixture with a TPM:TPnB weight ratio of 87.5:12.5 (whose osmotic pressure, viscosity and diffusivity were presented in Figs. 3 and 4 above) are presented in Table 3. As a result of the high viscosity of the GE draw agent, high flow rates were difficult to establish. The higher the draw solution flow rate, the higher the pressure buildup in the draw side of the membrane, which opposed the water flux through the membrane. In order to compensate for this pressure buildup (which occurred at lower flow rates as well), the feed side of the membrane was also pressurized in our experiments so that the mechanical pressure was balanced on either side of the membrane. In this manner, the water flux across the membrane was measured for different flow rates of the draw solution, allowing us to isolate the
Table 5 Water permeation flux under FO and PRO modes using water as feed stream and 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 as draw stream. Experiments were performed at 305 K and 315 K.
Table 3 water permeation flux under FO and PRO modes using water as feed stream and 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 as draw stream. Experiments were performed at 295 K. Pump flow rate#
1000 ml/mina (360/ 63)
200 ml/mina (72/ 12)
50 ml/mina (18/ 3)
FO mode (LMH) PRO mode (LMH)
2.6 7.1
2.2 6.0
2.2 5.6
305 K Pump flow rate# FO mode (LMH) PRO mode (LMH)
315 K
200 ml/mina (92/20) 2.40
50 ml/mina (23/5) 2.39
200 ml/mina (112/28) 0.65
50 ml/mina (28/7) 0.43
6.16
6.01
2.74
2.71
# Same pump rate was used on both sides. a The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides.
# Same pump rate was used on both sides a The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides.
6
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
operated in PRO mode, which could be due to the fact that the dense layer is more likely to see a greater concentration of GE in the PRO mode (and hence a greater extent of GE binding) than in FO mode. Membranes exposed to GE require a threshold pressure difference to initiate pure water flux, which suggests presence of adsorbed (or absorbed and bound) GE complex even after washing (see Fig. 6). In actual FO operation, one would expect a larger amount of GE adsorbed onto or absorbed into the dense layer and the extra resistance for transmembrane water flow because of GE uptake could be greater than what one can infer from Fig. 6 (for washed membranes). The fact that the performance stabilizes (as opposed to continuing to decline) is also suggestive of the existence of a limit to how much GE could be taken up by the membrane; for example, if it is adsorption on the surface, the limit would be a monolayer. The resistance for water transport offered by GE adhered in or on the dense layer is an added complexity in the case of FO with GE draw solution, which is not observed with draw solutions consisting of common inorganic salts. This argument is also consistent with the weaker effect of draw solution circulation rate seen in the case of GE draw solution; it simply implies that the added membrane resistance coming from adhered GE, which is unaffected by circulation rate, is a bigger concern than the ECP in the case of GE draw solution. Based on the phase diagram of GE solutions (Fig. 1), we estimate that water-rich phase obtained by heating diluted draw solution to a temperature modestly above Tcl would contain between 1 and 5 wt% of glycol ethers. The higher the phase separation temperature, the lower the concentration of GE mixture in the water-rich phase would be. After phase separation, an additional step is needed to remove the remaining GE from the water-rich phase. The performance characteristics of 5 different RO membranes in a simulated polishing step were tested; Fig. 7 summarizes the trans-membrane water flux of various RO membranes with feed samples having initial GE loading of 0, 1 and 5 wt %. During the course of an RO experiment, the loading level of GE roughly doubled. Loss of trans-membrane water flux was evident in the presence of GE mixtures. Water flux decreased rapidly as soon as the feed solution containing glycol ether was introduced to the cell. The water fluxes reported in Fig. 7 for GE solutions with initial loadings of 1 and 5 wt% differ approximately by a factor of two. Based on the osmotic pressure data, the theoretical driving force (applied pressure minus the osmotic pressure) is estimated to vary from ~112 psi to ~104 psi for the solution with an initial loading of 1 wt% and from ~80 psi to ~40 psi for that with 5 wt% initial loading. These translate to log-average effective driving forces [41] of ~108 psi and ~58 psi, respectively, for the two solutions. These values differ by a factor of ~2,
Table 6 Water permeation flux, NaCl rejection and reverse draw solute flux under FO, PRO and Pressure-assisted FO and PRO modes, using 0.6 M aqueous solution of NaCl as feed stream and 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 as draw stream. Pump speed = 200 ml/min (on both sides). Experiments were run at 295 K.
Water Flux (LMH) NaCl rejection % Reverse draw solute flux (gMH) a
FO mode
PRO mode
FO assisted modea
PRO assisted modea
1.3 ± 0.1 98.4 ± 0.4 4.9 ± 0.4
3.0 ± 0.4 98.5 ± 0.6 2.9 ± 0.2
1.6 ± 0.01 98.0 ± 0.8 3.8 ± 0.3
4.2 ± 0.23 98.8 ± 0.3 3.0 ± 0.2
Pressure assist at 400 kPa.
membranes after the experiments revealed no sign of membrane deformation against the spacers when exposed to a pressure of 400 kPa. Deformation was observed by Blandin et al. [38] at 600 kPa, who also reported a beneficial effect of pressure assist in FO experiments using NaCl as draw solute. In our experiments, upon pressure assist, water flux increased (Table 6) for FO mode from 1.3 LMH (0 Pa) to 1.6 LMH (400 kPa) and for PRO from 3.0 LMH (0 Pa) to 4.1 LMH (400 kPa). These improvements are a lot smaller than theoretically possible enhancement (estimated based on permeability data from Fig. 6), which is readily understood as being due to various transport resistances (that goes beyond the resistance of a fresh membrane). The flux obtained in PRO mode is higher than what we got with FO and AFO experiments; this observation points to the strong adverse effect of the ICP in the FO mode. In every operational mode (FO, PRO, AFO or APRO), the reverse flux of the draw solute was quite large (> 2.9 gMH). Comparison of FO and AFO results shows a decrease in reverse draw solute flux with increasing trans-membrane water flux; however, this is not borne out in PRO and APRO experiments where the reverse solute fluxes were roughly independent of trans-membrane water flux. The salt rejection was roughly comparable in all the cases. No additional membrane fouling or water flux decrease was observed during FO test which was in line with previously reported literature [38]. Water fluxes in PRO mode, achieved in this study are higher than the one reported by Nakayama et al. [14] for DEH (0.6 LMH) and Jørgensen for PPG425 (1.1 LMH) [17], which is attributed to the higher osmotic pressure of the draw solution used in our study. As the FO membrane is dense and nonporous, the reverse flux was indicative of the solubility and diffusivity of the draw solute in the membrane layer. It is worth reiterating the water flux obtained with GE draw solution is much lower than what was obtained with inorganic draw solutions. In FO experiments with 1.6 M ammonium-carbon dioxide draw solution and 0.5 M NaCl feed solution, for which the difference in osmotic pressure is 48.4 atm, McCutcheon and Elimelech [34] obtained water flux of 10.08 LMH. Clearly, thermo-responsive draw solutions fall short significantly. The results presented above may be rationalized qualitatively as follows. Fig. 5 shows that high initial water fluxes, comparable to those achievable with salt solutions, can be obtained with GE draw solutions. However, the flux rapidly declines (Fig. 5) and this is accompanied by an irreversible increase in membrane resistance (Fig. 6). Such rapid decline of water flux was also reported for fatty acids as draw agents [39]. (Organic molecules are known to cause membrane fouling [40].) It is clear from Table 6 that there is appreciable reverse flux of GE, which shows that GE does dissolve into and diffuse through the dense layer of the membrane; it seems reasonable to hypothesize that some of this dissolved GE is bound essentially irreversibly (and is responsible for the irreversible loss in water flux). The more strongly bound GE could be inside the dense layer and/or on the surface as adsorbed layer. Fig. 6 reveals a greater overall resistance for a membrane that had been
Fig. 7. Water permeation flux using RO membranes. Experiment were conducted in stirred- dead end cell at 120 psi and 295 K. Water flux was first measured using pure water as feed solution prior to RO experiments using glycol ether solutions. Experiments were then performed by loading 1 wt% and 5 wt% aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. The loading level of GE in the retentate nearly doubled during the course of each experiment; correspondingly the osmotic pressure to be overcome increased as well. See text for further discussion.
7
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
4. Summary
Table 7 Percent rejection and flux normalized rejection (FNR) of glycol ethers by various RO membranes. 1 wt% and 5 wt% aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 were used as feed solution. Relevant membrane flux of 20 LMH was used to calculate FNR for different RO membrane.
GE-AK ACM3 ACM4 73 U BW30
1% GE solution
5% GE solution
FNR for 1% GE solution
FNR for 5% GE solution
98.1 96.5 97.8 98.4 98.1
96.2 95.3 96.8 96.7 95.9
98.7 98.3 99.4 99.5 99.5
98.6 99.1 99.6 99.4 99.5
In this study the applicability of mixtures of two glycol ethers, tripropylene glycol methyl ether and tripropylene glycol n-butyl ether, as draw solutions was assessed. These glycol ethers are larger in size than the glycol ethers used in previous studies [14]. Cloud point temperatures of aqueous solutions of mixtures of TPM and TPnB at different total concentrations and composition ratios were determined. Water activities and kinematic viscosities of draw solutions as well as selfdiffusion coefficients of the GEs in the draw solutions were determined at different temperatures. Trans-Membrane water flux and reverse flux of GE during forward osmosis were measured at several different operating conditions. Removal of residual glycol ethers from the water–rich phase that would be obtained upon phase separation of diluted draw solution through a second-stage RO step was also studied. One can tune the cloud point temperature locus by adjusting the TPM:TPnB ratio, which is attractive for process optimization. Addition of TPnB to TPM lowers the cloud point temperature and the osmotic pressure (at a specified glycol ether wt%). The osmotic pressure and kinematic viscosity of draw solution were found to increase with glycol ether concentration, but decrease sharply with increase in temperature. The self-diffusion coefficient for the glycol ethers increases with increasing temperature. These led to a complex temperature dependence of trans-membrane water flux in FO operation. In forward osmosis experiments, PRO mode operation yielded greater flux than the FO mode, which is well known in the literature. PRO mode operation with a small pressure assist was the most favorable mode in our studies. Similar salt rejection was observed for all tested conditions. Although these draw agents do afford high osmotic pressures, they lead to appreciable initial loss of trans-membrane water flux and reverse solute flux. The water flux is hindered by concentration polarization resulting from high viscosities of these draw solutions as well as adherence of GE to the dense layer. In addition, the water produced using such draw agents would require one or two finishing RO steps to recover most of the draw solutes and this RO step is also prone to initial loss of trans-membrane water flux. In a single RO step operating at an applied pressure of 120 psi, > 95% rejection of the glycol ether could be achieved, but depending on the concentration of glycol ethers in the feed, 400–1000 ppm of glycol would present in the permeate. Additional RO steps or operation of the RO step at higher applied pressures could recover most of these glycol ethers. On the whole, the present study shows that there are significant challenges to using these types of draw agents. Membrane modifications to alleviate the acute initial loss of trans-membrane water flux are perhaps the most critical need if such draw agents are to be used. Most studies on FO tend to impose high circulation rate of the draw agent to minimize ECP. However, this is impractical for high-viscosity draw agents; the energy spent on circulation lowers energy efficiency and the higher pressures that build up on the draw solution side at higher circulation rates adversely affect the trans-membrane water flux. Water produced in processes using such draw agents will have to be subjected to multiple finishing RO steps to recover most of the draw solutes. Even then, the water will likely require additional adsorption step prior to being suitable for human consumption. Indirect use of thermo-responsive material as in the process described by Rajagopalan [19], which uses an ionic salt as a draw agent appears to be more promising.
which is roughly consistent with the observed difference in the water fluxes. Membrane fouling has been investigated extensively by many researchers [42,43]. A combination of relatively smooth, hydrophilic, and negatively charged film layer typically produces better water permeability, salt rejection, and fouling resistance in water purification applications. Loss of water flux and need for frequent cleaning are mitigated in practical field conditions by including an effective pretreatment to remove foulants. Various strategies have been suggested for RO applications including chemical or physicochemical pretreatments [44,45]. In the present context, as the loss of performance arises due to the draw agent, the usual pretreatment options do not appear viable. Membrane alteration to inhibit adsorption of the draw agent on the dense membrane later is likely to be a more fruitful avenue of further development. Rejection of glycol ether mixture by RO membranes is presented in Table 7. None of the RO membranes tested was able to reject the glycol ethers completely. Thus, in order to produce chemical-free water through a second-stage polishing step in a liquid-liquid phase separation based desalination process, one must use draw agents with molecular weight higher than those used in the present study. However, increasing the molecular weight lowers the osmotic pressure (at given wt% of draw solution) and increases the viscosity, both of which have adverse effects on membrane flux in the FO step. With higher molecular weight draw agents, one could deploy NF instead of RO in the second stage. Lower percent rejection of the solute is observed in the RO step for the feed solution with 5 wt% GE mixture. This is likely due to difference between effective driving forces for RO in our experiments where we held the applied pressure constant in all the experiments. It is known that for RO process, solute rejection can vary with applied pressure [41]. Higher the applied pressure, higher the rejection is. Feed solution with 5 wt% GE has higher osmotic pressure and needs higher applied pressure in order to reach the effective driving force experienced by the 1 wt% solution. To probe this effect further, we have calculated flux normalized rejection (FNR), using the procedure described by Adhikari et al. [41], and an industrially relevant membrane flux of 20 LMH for all the RO membranes. The FNR values are reported in Table 7. The FNR values 1 and 5 wt% cases are closer to each other than the observed percent rejection, which suggests that applied effective pressure is an important consideration in GE rejection by RO membranes. Before closing, we note that the shape of the phase diagram (see Fig. 1) is very important in simplification of the second stage RO (or NF) step. Specifically, it would be much more attractive if the water-rich phase in the two-phase mixture is naturally very low in the concentration of the draw agent. DEH used in Nakayama et al. [14] and PPG425 used by Jorgensen [17] have this character. It would also be interesting to study draw solutes with well-controlled extents of branching to understand how branching affects reverse solute flux.
Acknowledgment This work was supported by Princeton University through the Helen Shipley Fund for Innovation. References [1] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications,
8
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
osmotically driven membrane processes, Desalination 312 (2013) 31–38. [26] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, A novel ammonia—carbon dioxide forward (direct) osmosis desalination process, Desalination 174 (1) (2005) 1–11. [27] 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, J. Membr. Sci. 278 (1) (2006) 114–123. [28] S.P. Christensen, et al., Mutual solubility and lower critical solution temperature for water + glycol ether systems, J. Chem. Eng. Data 50 (3) (2005) 869–877. [29] H. Huang, R. Verrall, B. Skalski, Study of the aggregation behavior of mixed micelles of anionic surfactants and ethoxylated alcohols by time-resolved fluorescence quenching, Langmuir 13 (18) (1997) 4821–4828. [30] A.D. Wilson, F.F. Stewart, Deriving osmotic pressures of draw solutes used in osmotically driven membrane processes, J. Membr. Sci. 431 (2013) 205–211. [31] A.D. Wilson, Design of the Next-Generation FO Draw Solution, Rational Design of Next-generation Nanomaterials and Nanodevices for Water Applications, International Water Association (IWA), 2016, pp. 103–130. [32] M. Xie, et al., Effects of feed and draw solution temperature and transmembrane temperature difference on the rejection of trace organic contaminants by forward osmosis, J. Membr. Sci. 438 (2013) 57–64. [33] A. Sagiv, et al., Analysis of forward osmosis desalination via two-dimensional FEM model, J. Membr. Sci. 464 (2014) 161–172. [34] J.R. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal concentration polarization on flux behavior in forward osmosis, J. Membr. Sci. 284 (1) (2006) 237–247. [35] G.T. Gray, J.R. McCutcheon, M. Elimelech, Internal concentration polarization in forward osmosis: role of membrane orientation, Desalination 197 (1) (2006) 1–8. [36] Y. Xu, et al., Effect of draw solution concentration and operating conditions on forward osmosis and pressure retarded osmosis performance in a spiral wound module, J. Membr. Sci. 348 (1) (2010) 298–309. [37] S. Phuntsho, et al., Influence of temperature and temperature difference in the performance of forward osmosis desalination process, J. Membr. Sci. 415 (2012) 734–744. [38] G. Blandin, et al., Validation of assisted forward osmosis (AFO) process: impact of hydraulic pressure, J. Membr. Sci. 447 (2013) 1–11. [39] P. Zhao, et al., Fatty Acid Fouling of Forward Osmosis Membrane: Effects of pH, Calcium, Membrane Orientation, Initial Permeate Flux and Foulant Composition, Journal of Environmental Sciences (2016). [40] B. Mi, M. Elimelech, Chemical and physical aspects of organic fouling of forward osmosis membranes, J. Membr. Sci. 320 (1) (2008) 292–302. [41] B. Adhikari, et al., Compatibility study of nanofiltration and reverse osmosis membranes with 1-cyclohexylpiperidenium bicarbonate solutions, J. Membr. Sci. 527 (2017) 228–235. [42] M. Goosen, et al., Fouling of reverse osmosis and ultrafiltration membranes: a critical review, Sep. Sci. Technol. 39 (10) (2005) 2261–2297. [43] K.O. Agenson, T. Urase, Change in membrane performance due to organic fouling in nanofiltration (NF)/reverse osmosis (RO) applications, Sep. Purif. Technol. 55 (2) (2007) 147–156. [44] Y. Soffer, R.B. Aim, A. Adin, Membrane for water reuse: effect of pre-coagulation on fouling and selectivity, Water Sci. Technol. 42 (1–2) (2000) 367–372. [45] A. Schäfer, A.G. Fane, T. Waite, Cost factors and chemical pretreatment effects in the membrane filtration of waters containing natural organic matter, Water Res. 35 (6) (2001) 1509–1517.
and recent developments, J. Membr. Sci. 281 (1) (2006) 70–87. [2] S. Lee, et al., Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO), J. Membr. Sci. 365 (1) (2010) 34–39. [3] Y.-J. Choi, et al., Toward a combined system of forward osmosis and reverse osmosis for seawater desalination, Desalination 247 (1) (2009) 239–246. [4] V. Yangali-Quintanilla, et al., Indirect desalination of Red Sea water with forward osmosis and low pressure reverse osmosis for water reuse, Desalination 280 (1) (2011) 160–166. [5] T.-S. Chung, et al., Forward osmosis processes: yesterday, today and tomorrow, Desalination 287 (2012) 78–81. [6] J.E. Miller, L.R. Evans, Forward osmosis: a new approach to water purification and desalination, Report Prepared by Sandia National Laboratories, Albuquerque, New Mexico 87185 and Livermore, California 94550, 2006. [7] B. Van der Bruggen, P. Luis, Forward osmosis: understanding the hype, Rev. Chem. Eng. 31 (1) (2015) 1–12. [8] D.L. Shaffer, et al., Forward osmosis: where are we now? Desalination 356 (2015) 271–284. [9] R.L. McGinnis, M. Elimelech, Global challenges in energy and water supply: the promise of engineered osmosis, Environ. Sci. Technol. 42 (23) (2008) 8625–8629. [10] McGinnis, R.L., Osmotic desalination process. 2010, U.S. Patent No. US20100108587. [11] Cath, T.Y., A.E. Childress, and C.R. Martinetti, Combined membrane-distillationforward-osmosis systems and methods of use. 2011, U.S. Patent No. 8029671. [12] Carmignani, G., S. Sitkiewitz, and J.W. Webley, Recovery of retrograde soluble solute for forward osmosis water treatment. 2015, U.S. Patent No. 20120267308. [13] Iyer, S., Systems and methods for forward osmosis fluid purification using cloud point extraction. 2011, U.S. Patent No. 20100155329. [14] D. Nakayama, et al., Lower critical solution temperature (LCST) phase separation of glycol ethers for forward osmotic control, Phys. Chem. Chem. Phys. 16 (11) (2014) 5319–5325. [15] Y. Mok, et al., Circulatory osmotic desalination driven by a mild temperature gradient based on lower critical solution temperature (LCST) phase transition materials, Phys. Chem. Chem. Phys. 15 (44) (2013) 19510–19517. [16] Y. Cai, et al., Towards temperature driven forward osmosis desalination using semiIPN hydrogels as reversible draw agents, Water Res. 47 (11) (2013) 3773–3781. [17] M.K. Jørgensen, Investigation of Polypropylene Glycol 425 as Possible Draw Solution for Forward Osmosis, Aalborg University, Aalborg, 2009. [18] Lazare, L., Desalination of sea water. 1968, U.S. Patent No. 3386912. [19] Rajagopalan, N. and V.A. Patel, Solvent removal and recovery from inorganic and organic solutions. 2014, Patent US8852436. [20] R. Wei, et al., Highly permeable forward osmosis (FO) membranes for high osmotic pressure but viscous draw solutes, J. Membr. Sci. 496 (2015) 132–141. [21] Lee, Y., et al., Thermo-responsive draw solute for forward osmosis and method for water desalination and purification using the same. 2012, Google Patents. [22] Osmometers - KNAUER, (Available from:), http://www.knauer.net/en/products/ osmometers.html, (2017). [23] AquaLab: Water Activity Meters and Moisture Analyzers, (Available from:), http:// www.aqualab.com, (2017). [24] E.C.W. Clarke, D.N. Glew, Evaluation of the thermodynamic functions for aqueous sodium chloride from equilibrium and calorimetric measurements below 154 C, J. Phys. Chem. Ref. Data 14 (2) (1985) 489–610. [25] T.Y. Cath, et al., Standard methodology for evaluating membrane performance in
9
Contents lists available at ScienceDirect
Desalination journal homepage: www.elsevier.com/locate/desal
Forward osmosis using draw solutions manifesting liquid-liquid phase separation Siavash Darvishmanesh, Brian A. Pethica, Sankaran Sundaresan⁎ Department of Chemical and Biological Engineering, Princeton University, Princeton, NJ 08544, USA
A R T I C L E I N F O
A B S T R A C T
Keywords: Forward osmosis Lower critical solution temperature Glycol ethers Liquid-liquid phase separation
Desalination via forward osmosis using draw agents whose regeneration is aided via liquid-liquid phase separation has gained much attention in recent years. In the present study, mixtures of two different glycol ethers, tripropylene glycol methyl ether and tripropylene glycol n-butyl ether, have been studied as potential draw agents. Water activity, viscosity and diffusion coefficient of draw solutions have been measured at different mixture compositions, concentrations and temperatures. Osmotic pressures of these draw solutions decreases strongly with increasing temperature. Forward osmosis experiments performed with these draw solutions reveal appreciable initial loss of trans-membrane water flux, reverse solute flux and severe concentration polarization.
1. Introduction In forward (or direct) osmosis (FO), water from an aqueous solution selectively passes through a membrane to a second solution (referred to as the draw solution) at the same pressure based solely on the difference in the water activity (osmotic pressure) of the two solutions [1]. A number of studies have explored the feasibility and benefits of FO as an alternative to reverse osmosis (RO) for seawater desalination [2–6]. The overall cost of producing desalinated water from seawater via FO process is affected by the type of draw solution employed and its regeneration. Regeneration processes using waste heat or geothermal sources are being explored as routes to cost-effective desalination [7–9]. For example, a recyclable salt solution (a mixture of ammonium bicarbonate and ammonium hydroxide dissolved in water) has been employed as a draw agent in FO, in which low-grade heat can be employed to remove this thermolytic salt from diluted draw solution [10]. In another proposed process, FO is coupled with membrane distillation (MD) [11], where FO is applied to reduce membrane fouling and scaling which is detected in pressure-driven membrane processes such as RO and nanofiltration (NF); MD process separates the water and regenerates the draw solution for FO using waste heat or geothermal sources. Aqueous solutions of thermo-responsive organics (chemicals or polymers) showing lower critical solution temperature (LCST) have also been explored as draw solutes [12–17]; in this case, draw solution diluted via water extraction in an FO process step is heated to a temperature above LCST to induce spontaneous phase separation into organic-rich and water-rich phases. The water-rich phase is then further upgraded through second-stage RO or NF. The use of thermo-responsive ⁎
organics as draw solutions for FO was first patented in 1968 [18]; however lack of suitable FO membranes thwarted its application. Recently, modifications of this concept for a draw solution manifesting LCST have been patented for integrated FO-NF systems [12,13]. Two types of FO processes using thermo-responsive materials, whose aqueous solutions manifest LCST, have been reported in the literature: (i) Direct use of the thermoresponsive material as a draw agent [14,17]; (ii) Indirect use by formation of aqueous two-phase system with an inorganic draw agent [19]. Thermo-responsive draw agents include block copolymers of polyethylene oxides and polypropylene oxides [12] and fatty acid or fatty alcohol polyethylene glycols polymers [13]. Concentrated aqueous solutions of these molecules are much more viscous than aqueous solutions of inorganic salts (such as MgSO4), leading to severe internal concentration polarization (ICP) in the membrane's support layer (in the so-called FO mode where the support layer faces the draw solution) as well as external concentration polarization (ECP) in the flowing draw solution (in both FO mode of operation and the so-called Pressure Reduced Osmosis (PRO) mode). Furthermore, these organic molecules tend to foul the membrane [20]. Indirect use of thermo-responsive inorganics for desalination was introduced by Rajagopalan et al. [19] as the Aquapod© desalination process. In this method, an aqueous solution of an inorganic salt (specifically MgSO4) is employed as draw agent in the FO step. The diluted draw solution is sent to an aqueous two-phase contactor, where it is concentrated by extracting water using a concentrated aqueous solution of UCON660©, which is a polyethylene oxide-polypropylene oxide block copolymer. The polymer-rich phase is then separated and heated
Corresponding author. E-mail address: [email protected] (S. Sundaresan).
http://dx.doi.org/10.1016/j.desal.2017.05.036 Received 16 January 2017; Received in revised form 30 May 2017; Accepted 30 May 2017 0011-9164/ © 2017 Elsevier B.V. All rights reserved.
Please cite this article as: Darvishmanesh, S., Desalination (2017), http://dx.doi.org/10.1016/j.desal.2017.05.036
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
at which the draw agent turned cloudy (to the eye). This experiment was repeated several times, after allowing the sample to cool down yielding a transparent single-phase. For a given composite draw agent, Tcl changed with wt% GE, and the minimum was recorded as its LCST. Two instruments were tested for the osmotic pressure measurement: Knauer K-7000 Vapor Pressure Osmometer (VPO) [22] and the AquaLab Tunable Diode Laser (TDL) infrared vapor pressure meter [23]. The VPO measures the difference in temperature between a drop of the test solution on a thermistor and a reference solution (which is pure water) in a closely thermostatted cell ( ± 0.001 °C) which results from condensation of water vapor on the test drop [22]. The TDL instrument measures water vapor concentration from the absorption of an infrared laser beam passing through (an equilibrated) vapor phase in contact with the test liquid in a sealed thermostatted cell. The temperature in the chamber is accurate to ± 0.2 °C [23]. The water activity can be estimated with an accuracy of ± 0.005. Both methods were calibrated using as reference standard solutions of NaCl [24] or pure water (unit activity). Measurements with the VPO instrument manifested slow long-term drift with GE solutions, particularly at higher concentrations (but not with solutions of inorganic salts). In contrast, the TDL instrument consistently yielded reproducible measurements. A key test of the reliability of an instrument was made as follows. By heating a draw solution to a chosen temperature above its cloud point the solution was allowed to separate into two phases. After allowing ample time for the two phases to equilibrate (typically overnight), the two phases were separated and the water activities of the two phases were measured at the chosen phase separation temperature. The TDL instrument yielded nearly identical water activities for both phases, which is what one would expect for phases in equilibrium; in contrast, the VPO yielded very different results. (See Table 1 summarizing results from TDL meter.) Therefore, in what follows, we present only the data obtained from the TDL instrument. The relationship between the osmotic pressure and water activity of sodium chloride solutions at various temperature can be found in the literature [24]. The water activities of various draw solutions measured in our study are converted to equivalent osmotic pressures of sodium chloride solution possessing the same water activity as the draw solution at the temperature of interest. This equivalent osmotic pressure (or more directly the water activity) is the most relevant metric to assess the driving force afforded by a draw solution for water extraction from seawater (which is mostly a sodium chloride solution) in an FO process. Viscosities were measured using an SI-Analytics Ubbelohde viscometer, where the temperature was controlled to within ± 1 °C. Self-diffusion coefficient measurements were made on 1H NMR using a Bruker Avance III 500 MHz NMR spectrometer. The 1D diffusion ordered spectroscopy (DOSY) experiments were run with 50 G/cm magnetic field gradient, and the data were analyzed using advanced Bayesian DOSY transformation method. Samples were mixtures of the two glycol ethers in D2O. The performance of GE mixtures as draw solutes in the FO system was tested on a lab-scale circulating filtration unit, as described by Cath et al. [25]. Cellulose acetate FO membrane was purchased from Fluid Technology Solutions, Inc. (USA). DI water and NaCl solution were employed as a feed solution. The Sepa CF042 solvent-stable cross-flow permeation cell was purchased from Sterlitech™ (USA); it has an active
to a temperature above its cloud point temperature (Tcl) to form two phases; a polymer-rich phase which is cooled and returned to the aqueous two-phase contactor, and a water-rich phase which is sent to a further upgrading step. Potential application of low molecular weight polymers and nonionic surfactants as FO draw agents has been examined by several other researchers. Polypropylene glycol (PPG425) as a draw agent for FO was assessed by Jørgensen [17]. The largest osmotic pressure recorded in this study was ~50 atm at 288 K, well below the magnitudes cited by patents [12,13] for their draw agents. Polyethylenimine (PEI) derivatives and glycol ethers (GEs), whose aqueous solutions manifest LCST and have lower viscosities (than aqueous solutions of polymers mentioned above), have been patented recently [21]. It appears that these draw solutions have rather low osmotic pressures and can barely draw water from synthetic seawater feed solution (3.5 wt% NaCl). Nakayama et al. [14] recently demonstrated extraction of water from seawater via FO using a draw solution consisting of an aqueous solution of diethylene glycol n-hexyl ether (DEH; 146.23 Da; LCST ~10 °C). It is likely that reverse solute diffusion would be more pronounced for this solute because of its low molecular weight, but reverse solute flux was not reported. These authors have also suggested an integrated FO-FO process [14,15]. Strong temperature (T) and composition dependencies of the osmotic pressure of aqueous solutions of thermo-responsive draw agents play central roles in the success of FO process. To the best of our knowledge, there is only one published study on the temperature dependence of osmotic pressure of aqueous solutions of thermoresponsive draw agents at conditions relevant to desalination via FO [17]. In this study, Jørgensen [17] found that the osmotic pressure of aqueous solutions of PPG425 increases with concentration at constant T. Furthermore, the osmotic pressure at fixed concentration increases appreciably with decreasing T, without any sign of reaching a plateau (in the temperature range studied). In other words, one could increase the driving force significantly by lowering the operating T; however, lowering T increases the viscosity of the draw solution and decreases the diffusivity appreciably, thereby lowering FO-based desalination process efficiency. Higher viscosity implies higher pumping cost and lower diffusivities increase the severity of concentration polarization (CP). On the other hand, lowering T could decrease reverse solute flux of the draw agent (as a result of lower diffusivity). Thus, experimental data and a good understanding of the concentration and temperature dependencies of osmotic pressure, viscosity and diffusivity would be valuable for FO process optimization. This consideration motivated the present study, where we have studied the temperature and concentration dependence of the phase behavior, osmotic pressure, kinematic viscosity and self-diffusivities of a model thermo-responsive material as potential draw agent. Specifically, we have studied mixtures of two GEs: Tripropylene glycol methyl ether (TPM) (206.27 Da) and Tripropylene glycol mono-n-butyl ether (TPnB) (248.35 Da). Aqueous solutions of each of these GEs manifest a LCST, but they are widely separated; by mixing them, one can tune the LCST. FO experiments were also performed at three different temperatures with a draw solution at one particular composition in order to identify achievable water flux, fouling and reverse flux characteristics. 2. Materials and methods
Table 1 Water activities of two phases obtained through phase separation of 50 wt% aqueous solution of a composite glycol ether mixture (50%TPM-50%TPnB). Measurements were made using TDL water activity meter (with accuracy ± 0.005).
DOWANOL™ TPM Glycol Ether, DOWANOL™ TPnB Glycol Ether and Deuterium oxide (D2O) were purchased from Sigma-Aldrich (USA). The two GEs were mixed at various weight ratios to prepare composite draw agents. The composite draw agent was then mixed with deionized water (DI) to make draw solutions. Each draw solution was thus characterized by wt% of GE and TPM-TPnB composition ratio. A test tube containing the draw solution was immersed and heated in a jacketed beaker connected to a water bath (Fisher Scientific Model 3013S Heating-Cooling Recirculator, USA) to identify the temperature 2
Temperature (K)
Water-rich phase water activity
Organic-rich phase water activity
293 303 313
0.971 0.974 0.977
0.974 0.973 0.981
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
membrane area of 42 cm2. The flow rates of both draw and feed solutions, which passed through the permeation cell channel in a cocurrent fashion, were controlled by flow meters. Cross flow velocity which (CFV) was calculated by dividing the volumetric flow rate [l pm] in the flow channel by the cross sectional area (0.0023 × 0.0392 m2) of the flow channel. The temperature of the whole FO system was kept at 295 ± 0.5 K (unless mentioned otherwise). In the FO experiments presented in this report, 50 wt% aqueous solution of a composite draw agent (87.5 wt% TPM-12.5 wt% TPnB) was used as draw solution. (The LCST of this composite agent is ~ 313 K.) A reservoir containing the draw solution was placed on an electronic balance (Ohaus Pioneer PA2202, USA), connected to a computer to record its mass as a function of time, from which one can readily extract water permeation flux, Jv (lit/m2·h, commonly abbreviated as LMH) as a function of time. Reverse solute flux Js, (g/m2·h, commonly abbreviated as gMH) refers to the rate at which draw solute is transported across the membrane from the draw side to the feed side. Average reverse flux over the course of an experiment was found from the initial and final volumes of the feed stream, the initial and final organic content in feed stream samples (measured using Total Organic Carbon (TOC) Analyzer (Shimadzu Scientific Instruments, USA)), and the duration of the FO experiment. Forward osmosis experiments were conducted using both possible orientations of the asymmetric membranes. In the so-called FO mode, the draw solution was on the support layer side and the feed was exposed directly to the active layer; this is the typical orientation in FO [26,27]. In the PRO mode, the draw solution was exposed to the active layer and the feed was on the support layer side. In an effort to understand challenges in the second-stage RO (finishing step), water flux and solute rejection were determined at 295 K in a high-pressure chemical-resistant HP4750 stirred Cell, Sterlitech™ (USA) for five different RO membranes (BW30, 73 U, ACM4, ACM3, GE-AK). The effective RO membrane area was 13.85 cm2. Pure water flux was first measured, after the membranes were compressed for 3 h at 100 psi. Subsequently, experiments were carried out with 200 ml of aqueous solutions of composite GE mixture (87.5 wt%TPM-12.5 wt% TPnB) with GE concentrations of 1 and 5 wt%. Solute rejection was determined from GE concentrations (measured through TOC analyzer) of the initial and final feed, and the accumulated permeate after filtration of 100 ml of the feed solution. Four different samples from each RO membrane were evaluated to determine average water permeability and solute rejection.
Fig. 1. Cloud point temperature loci of aqueous solutions of composite glycol ether mixtures with different TPM-TPnB weight percent ratios: (×) 80–20; (■) 70–30; (◊) 50–50.
based on the temperatures of the feed saline water to be desalinated and the available heating medium. Addition of TPM to TPnB affects the osmotic pressure significantly. This effect is presented in Fig. 2(a-c), where the equivalent osmotic pressure is plotted against temperature. Each panel shows the results for a particular GE loading level, and the different data sets in each panel correspond to different TPM:TPnB weight ratios. Under all conditions tested, the osmotic pressure decreases upon increasing T. Of the compositions tested, TPM afforded the highest osmotic pressure and addition of TPnB lowered the osmotic pressure. Thus, at a given temperature and total GE concentration level, increasing TPnB level lowers the driving force for FO afforded by the draw agent; on the other hand, as seen in Fig. 1, increasing TPnB level allows phase separation at a lower T and hence the process could operate with waste heat available at a lower T. Thus, an optimal composition would clearly be a compromise between an FO step (high osmotic pressure) and the subsequent phase separation step (low Tcl). Huang et al. [29] have observed that phase separation in these mixtures is accompanied by aggregate formation; the degree of aggregation typically increases as T increases, leading to a decrease in the osmotic pressure. In the results presented in Fig. 3, the rate of change in osmotic pressure with temperature does not manifest discernible temperature dependence at the lower temperatures studied, but this dependence can be seen more clearly at higher temperatures at least for the GE mixtures with no or small TPnB content. As the higher temperatures, the osmotic pressure becomes less sensitive to temperature, indicating a more gradual change in the degree of aggregation. The FO experiments described below were performed using aqueous solutions of glycol ether mixture with specific composition of 87.5 wt% TPM - 12.5 wt% TPnB. The lowest cloud point temperature of aqueous solutions of this composite GE mixture was found to be ~ 313 K. Effect of temperature on osmotic pressure of a 50 wt% aqueous solution of this composite mixture is presented in Fig. 3a. Fig. 3b shows the variation of equivalent osmotic pressure with GE wt% at 295 K. The inflection seen at intermediate concentrations is typical of thermo-responsive systems that manifest LCST [14]. Fig. 3c shows the same osmotic pressure data in Fig. 3b, except that it is now plotted against (g GE/g water), following Wilson and Stewart [30]; remarkably, the data now collapse into two nearly linear segments. Wilson [31] has suggested that these linear segments are indicative of aggregates of glycol ether in water matrix at low GE wt% and the inverse at high GE loading levels.
3. Results and discussion 3.1. Temperature dependence of equivalent osmotic pressure of glycol ether mixtures Fig. 1 shows the cloud point temperatures of aqueous GE solutions at different concentrations for composite draw agents containing different TPM/TPnB ratios. The water-TPM system has Tcl values in excess of 373 K, and could not be measured in our laboratory. This figure shows that addition of TPM to TPnB raises Tcl. For water-GE mixtures, whether hydrophobic or hydrophilic interactions dominate depends strongly upon T and the GE molecular structure [28]. Above LCST, the water–GE system manifests phase separation over a range of compositions. The LCST increases with a decrease in the carbon number of the end group because of decreasing hydrophobic character; for example, changing the end group from n-butyl (in TPnB) to methyl (in TMP) increases LCST from 253 K to a value in excess of 373 K. The Tcl locus of a mixture of TMP and TPnB lies between those of water-TMP and waterTPnB systems. Analogously, the LCST of water–GE system can also be manipulated by mixing glycol ethers, as one can readily surmise from Fig. 1. This is consistent with data reported by Christensen et al. [28]. Fig. 1 clearly shows that one can tune the phase separation temperature by adjusting the TMP-TPnB ratio in an effort to optimize the FO process 3
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
(a)
(b)
(c)
Fig. 3. a: Effect of temperature on the osmotic pressure of 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. b: Effect of GE concentration on the osmotic pressure of aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. T = 295 K. c: Effect of GE concentration on the osmotic pressure of aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. T = 295 K. The results in Fig. 3b have been replotted using a different scale for the GE concentration.
Fig. 2. Effect of temperature on osmotic pressure of Glycol ethers–water mixtures at three different concentrations and several composition ratios. TPM:TPnB weight ratio: (●) 100:0; ( ) 90:10; (×) 80:20; (Δ) 70:30; (□) 60:40; (◊) 50:50. Panel (a): 20 wt% aqueous solution; panel (b) 50 wt%; panel (c): 80 wt%.
associated with low viscosity (as viscosity is inversely proportional to diffusivity [33]) helps reduce the effects of CP. In the present study we sought to examine the temperature dependence of viscosity and diffusivity of glycol ether draw solutions. Fig. 4 shows the variation of kinematic viscosity and self-diffusion coefficient of a model glycol ether draw solution with temperature. The inverse relationship between viscosity and diffusivity is readily apparent. Increasing the temperature
3.2. Effect of temperature on viscosity and diffusivity of draw solution Ideal draw solutions should exhibit low viscosity, which would lower the pumping cost [32]; more importantly, the higher diffusivity 4
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
Fig. 4. The effect of temperature on kinematic viscosity and self-diffusion coefficient of aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. Open symbols: Diffusivity; Filled symbols: Kinematic viscosity. (■) 30 wt%; (●) 50 wt%; (▲) 70 wt%. Fig. 5. evolution of water permeation flux over time under FO mode using water as feed stream and 50 wt% aqueous solution of 87.5% TPM - TPnB 12.5% as draw stream. Attempted regeneration of membrane shows the flux reduction is irreversible.
leads to an appreciable decrease in the kinematic viscosity of the draw solution and an increase in the diffusivity, both of which are beneficial from transport considerations. (Increasing the temperature from 293 K to 313 K lowers the kinematic viscosity by 50% and increases the diffusion coefficient by factor of two.) However, this beneficial effect of increasing temperature is accompanied by a sharp decrease in the osmotic pressure of the draw agent (Fig. 3), which lowers the effectiveness of the draw agent in an FO process. These opposing effects of T imply that there would be an optimum temperature representing a compromise between these two considerations. Fig. 4 also shows the variation of kinematic viscosity of 30 wt% and 70 wt% aqueous solutions. As concentration of GE increases, the viscosity increases (and correspondingly, the diffusivity would decrease).
irreversible loss in trans-membrane water flux caused by GE draw solution, which is discussed later. In what follows, we only report the results obtained in the plateau region following irreversible performance loss. The performance of fresh FO membranes was tested with sodium chloride as draw solute and freshwater on the feed side, and the results are presented in Table 2. The water flux in the FO mode (i.e., the draw solution is on the support layer side) is less than that in the PRO mode (i.e., the draw solution is on the dense selective layer side), which can be attributed to ICP in the FO mode. Increasing the cross-flow velocity of the NaCl solution along the membrane surface increases the membrane flux, by lowering the severity of ECP; the sharp increase in transmembrane water flux even in the FO mode shows that in this mode, in addition to ICP, there must have been appreciable ECP outside the support layer that got weakened by increased flow rate. To assess the extent of loss in performance due to ICP and ECP, experiments were performed with freshwater on both sides and gentle applied pressure (of the order of a few atm) on one side to determine membrane permeability (see Fig. 6); such experiments yielded permeability of 0.62 LMH/atm for fresh membranes at 295 K, which is consistent with the findings of Jorgensen [17]. The osmotic pressure of 1 M NaCl solution is 47 atm and so the theoretical water flux in the absence of any CP (ICP and ECP) would be ~29 LMH, which is significantly larger than fluxes observed, revealing that appreciable concentration polarization persisted even at the high circulation rates [35] in both FO and PRO modes. The Reynolds numbers determined on the basis of cross-stream
3.3. FO process performance In an FO process, concentration polarization can affect performance on both sides of the membrane. On the feed side, the polarized layer becomes more concentrated than the bulk (if a solute that is rejected by the membrane is present), and on the draw side, the polarized layer is more dilute than the bulk; both lead to a decrease in the actual osmotic driving force across the membrane. The severity of these effects can be controlled to some extent with cross-flow and well-designed hydrodynamics [34]. Appreciable initial decrease in trans-membrane water flux was observed with glycol ether draw solutions. Fig. 5 illustrates performance loss, where forward osmosis experiments were performed in the FO mode. First, FO experiments were performed using 1 M NaCl solution as draw solution (step 1). No fouling was observed. The membrane was then washed by flushing fresh water through the cell and reused in an experiment where 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 was used as draw solution. The membrane flux decreased appreciably with time and appeared to level off after an hour (step 2). (As the draw solution is used in a circulation mode, the draw solution is constantly undergoing dilution; however, the extent of dilution over this time period is small. As a result, the change in flux cannot be attributed to this dilution, which can readily be inferred from the results for step 1.) The membrane was then washed and reused with the 1 M NaCl solution as draw agent (step 3). The water flux was measurably smaller than that in step 1, but no further loss was observed. Finally, the membrane was washed again and reused with the 50 wt% GE solution (step 4), recovering the water flux plateau observed at the end of step 2. These results demonstrate
Table 2 Water permeation flux under FO and PRO modes using water as feed stream and 1 M aqueous solution of NaCl as draw stream. Experiments were performed at 295 K. Pump flow rate#
1000 ml/minb (360/ 360)
200 ml/minb (72/ 72)
50 ml/minb (18/ 18)
FO mode (LMH)a PRO mode (LMH)
8.4 16.8
5.8 11.9
5.3 11.3
a
LMH = lit/m2/h. # same pump rate was used on both sides. The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides. The cross-sectional area of the channel available for cross-flow of the fluid is 0.23 × 3.92 cm. Circulation rates of 50, 200 and 1000 ml/min correspond to cross flow velocity of 0.009, 0.036 and 0.180 m/s, respectively; this conversion between pump flow rate and cross-flow velocity applies for Tables 3-6 as well. b
5
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
Table 4 Water permeation flux under FO and PRO modes using water as feed stream and 1 M aqueous solution of NaCl as draw stream. Experiments were performed at 305 K and 315 K. 305 K Pump flow rate# FO mode (LMH) PRO mode (LMH)
315 K
200 ml/mina (92/92) 7.1
50 ml/mina (23/23) 6.4
200 ml/mina (112/112) 7.5
50 ml/mina (28/28) 6.9
13.1
12.5
13.6
13.1
# Same pump rate was used on both sides. a The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides.
consequences of CP. The osmotic pressure of the GE draw solution employed in these experiments is 76 atm, which is higher than that of the 1 M NaCl solution (see Fig. 5); yet, lower water flux was achieved with GE draw solution (compare Tables 2 and 3), clearly indicating a greater overall resistance to trans-membrane water flux with GE draw solution. As expected, the water flux with GE draw solution is greater in the PRO mode than in FO mode [36]; this clearly shows that ICP plays an important role in the FO mode. With the GE draw solution, the water flux increased upon increasing the cross-flow rate of the draw solution (Table 3) as in the case of the 1 M NaCl draw solution (Table 2), but this increase is relatively weaker for the GE draw solution. Forward osmosis experiments were performed at two additional temperatures to supplement the results in Tables 2 and 3. These are presented in Table 4 for NaCl draw solution and Table 5 for GE solution. Temperature affects the performance of most membrane processes; for example, it affects both the solution viscosity and diffusivity of solutes, thereby influencing both the concentration and hydrodynamic boundary layers [37]. In the case of NaCl draw solution, the transmembrane flux increased with increasing T (compare Tables 2 and 4). In contrast, increasing the temperature to 305 K from 295 K led to a very small increase in water flux when GE solution was used as draw solution (compare Tables 3 and 5). This suggests that the loss in osmotic driving force based on bulk composition resulting from increase in T was compensated by a decrease in various resistances. However, further increase in temperature (to 315 K, see Table 5) led to a significant reduction in water flux, which can be attributed to acute loss of osmotic driving force. It is clear from Tables 4 and 5 that thermo-responsive materials manifesting LCST manifest, when used as draw solutions In FO process, manifest more complex dependence on T than draw solutions based on inorganic salts. Table 6 summarizes results obtained (after the initial period of performance loss) in experiments with a GE draw solution and synthetic seawater (0.6 M) in the feed side. The table presents data on membrane water flux, salt rejection and reverses draw solute flux. Experiments were run in the FO, PRO, pressure-assisted FO (AFO) and pressure-assisted PRO (APRO). Assisted FO and PRO experiments were performed with an applied pressure of 400 kPa. Visual observation of the
Fig. 6. Membrane water fluxes at different applied pressures. Membranes were precompacted at 400 kPa. Pure water flux was measured using fresh membranes and membranes exposed to glycol ether draw solution. Pressure was applied on: (●) dense layer of a fresh membrane; (□) support layer of a fresh membrane; (Δ) dense layer of a membrane which had been operated in a PRO mode with GE draw solution; (×) support layer of a membrane which had been operated in a FO mode with GE draw solution.
velocity of the draw solution, the draw solution kinematic viscosity and the smaller dimension of the cross-flow cross section are shown in Table 2 for the three different pump speeds. It is clear that the flow is in the laminar regime. As shown in Fig. 6, the permeability of membrane exposed to GE draw solution is lower (even after washing) than that of the fresh membrane, clearly indicating irreversible increase in membrane resistance. For fresh membranes, the orientation of the membrane does not affect the water permeability. In water permeability measurements with membranes that have been used in forward osmosis experiments using GE draw solution, pressure was applied to the side that was exposed to GE draw solution; if the forward osmosis was done in FO (PRO) mode, then pressurization was done on the support (dense layer) side during water permeability test. The water permeability of a membrane exposed to GE draw solution is less than that of a fresh membrane. Membrane used with GE draw solution in PRO mode had a lower pure water flux than that used in FO mode (discussed later). Effects of membrane orientation and draw solution cross-flow velocity on water flux achieved with a 50 wt% aqueous solution of a composite GE mixture with a TPM:TPnB weight ratio of 87.5:12.5 (whose osmotic pressure, viscosity and diffusivity were presented in Figs. 3 and 4 above) are presented in Table 3. As a result of the high viscosity of the GE draw agent, high flow rates were difficult to establish. The higher the draw solution flow rate, the higher the pressure buildup in the draw side of the membrane, which opposed the water flux through the membrane. In order to compensate for this pressure buildup (which occurred at lower flow rates as well), the feed side of the membrane was also pressurized in our experiments so that the mechanical pressure was balanced on either side of the membrane. In this manner, the water flux across the membrane was measured for different flow rates of the draw solution, allowing us to isolate the
Table 5 Water permeation flux under FO and PRO modes using water as feed stream and 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 as draw stream. Experiments were performed at 305 K and 315 K.
Table 3 water permeation flux under FO and PRO modes using water as feed stream and 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 as draw stream. Experiments were performed at 295 K. Pump flow rate#
1000 ml/mina (360/ 63)
200 ml/mina (72/ 12)
50 ml/mina (18/ 3)
FO mode (LMH) PRO mode (LMH)
2.6 7.1
2.2 6.0
2.2 5.6
305 K Pump flow rate# FO mode (LMH) PRO mode (LMH)
315 K
200 ml/mina (92/20) 2.40
50 ml/mina (23/5) 2.39
200 ml/mina (112/28) 0.65
50 ml/mina (28/7) 0.43
6.16
6.01
2.74
2.71
# Same pump rate was used on both sides. a The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides.
# Same pump rate was used on both sides a The numbers in parenthesis indicate Reynolds numbers on the feed and draw sides.
6
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
operated in PRO mode, which could be due to the fact that the dense layer is more likely to see a greater concentration of GE in the PRO mode (and hence a greater extent of GE binding) than in FO mode. Membranes exposed to GE require a threshold pressure difference to initiate pure water flux, which suggests presence of adsorbed (or absorbed and bound) GE complex even after washing (see Fig. 6). In actual FO operation, one would expect a larger amount of GE adsorbed onto or absorbed into the dense layer and the extra resistance for transmembrane water flow because of GE uptake could be greater than what one can infer from Fig. 6 (for washed membranes). The fact that the performance stabilizes (as opposed to continuing to decline) is also suggestive of the existence of a limit to how much GE could be taken up by the membrane; for example, if it is adsorption on the surface, the limit would be a monolayer. The resistance for water transport offered by GE adhered in or on the dense layer is an added complexity in the case of FO with GE draw solution, which is not observed with draw solutions consisting of common inorganic salts. This argument is also consistent with the weaker effect of draw solution circulation rate seen in the case of GE draw solution; it simply implies that the added membrane resistance coming from adhered GE, which is unaffected by circulation rate, is a bigger concern than the ECP in the case of GE draw solution. Based on the phase diagram of GE solutions (Fig. 1), we estimate that water-rich phase obtained by heating diluted draw solution to a temperature modestly above Tcl would contain between 1 and 5 wt% of glycol ethers. The higher the phase separation temperature, the lower the concentration of GE mixture in the water-rich phase would be. After phase separation, an additional step is needed to remove the remaining GE from the water-rich phase. The performance characteristics of 5 different RO membranes in a simulated polishing step were tested; Fig. 7 summarizes the trans-membrane water flux of various RO membranes with feed samples having initial GE loading of 0, 1 and 5 wt %. During the course of an RO experiment, the loading level of GE roughly doubled. Loss of trans-membrane water flux was evident in the presence of GE mixtures. Water flux decreased rapidly as soon as the feed solution containing glycol ether was introduced to the cell. The water fluxes reported in Fig. 7 for GE solutions with initial loadings of 1 and 5 wt% differ approximately by a factor of two. Based on the osmotic pressure data, the theoretical driving force (applied pressure minus the osmotic pressure) is estimated to vary from ~112 psi to ~104 psi for the solution with an initial loading of 1 wt% and from ~80 psi to ~40 psi for that with 5 wt% initial loading. These translate to log-average effective driving forces [41] of ~108 psi and ~58 psi, respectively, for the two solutions. These values differ by a factor of ~2,
Table 6 Water permeation flux, NaCl rejection and reverse draw solute flux under FO, PRO and Pressure-assisted FO and PRO modes, using 0.6 M aqueous solution of NaCl as feed stream and 50 wt% aqueous solution of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 as draw stream. Pump speed = 200 ml/min (on both sides). Experiments were run at 295 K.
Water Flux (LMH) NaCl rejection % Reverse draw solute flux (gMH) a
FO mode
PRO mode
FO assisted modea
PRO assisted modea
1.3 ± 0.1 98.4 ± 0.4 4.9 ± 0.4
3.0 ± 0.4 98.5 ± 0.6 2.9 ± 0.2
1.6 ± 0.01 98.0 ± 0.8 3.8 ± 0.3
4.2 ± 0.23 98.8 ± 0.3 3.0 ± 0.2
Pressure assist at 400 kPa.
membranes after the experiments revealed no sign of membrane deformation against the spacers when exposed to a pressure of 400 kPa. Deformation was observed by Blandin et al. [38] at 600 kPa, who also reported a beneficial effect of pressure assist in FO experiments using NaCl as draw solute. In our experiments, upon pressure assist, water flux increased (Table 6) for FO mode from 1.3 LMH (0 Pa) to 1.6 LMH (400 kPa) and for PRO from 3.0 LMH (0 Pa) to 4.1 LMH (400 kPa). These improvements are a lot smaller than theoretically possible enhancement (estimated based on permeability data from Fig. 6), which is readily understood as being due to various transport resistances (that goes beyond the resistance of a fresh membrane). The flux obtained in PRO mode is higher than what we got with FO and AFO experiments; this observation points to the strong adverse effect of the ICP in the FO mode. In every operational mode (FO, PRO, AFO or APRO), the reverse flux of the draw solute was quite large (> 2.9 gMH). Comparison of FO and AFO results shows a decrease in reverse draw solute flux with increasing trans-membrane water flux; however, this is not borne out in PRO and APRO experiments where the reverse solute fluxes were roughly independent of trans-membrane water flux. The salt rejection was roughly comparable in all the cases. No additional membrane fouling or water flux decrease was observed during FO test which was in line with previously reported literature [38]. Water fluxes in PRO mode, achieved in this study are higher than the one reported by Nakayama et al. [14] for DEH (0.6 LMH) and Jørgensen for PPG425 (1.1 LMH) [17], which is attributed to the higher osmotic pressure of the draw solution used in our study. As the FO membrane is dense and nonporous, the reverse flux was indicative of the solubility and diffusivity of the draw solute in the membrane layer. It is worth reiterating the water flux obtained with GE draw solution is much lower than what was obtained with inorganic draw solutions. In FO experiments with 1.6 M ammonium-carbon dioxide draw solution and 0.5 M NaCl feed solution, for which the difference in osmotic pressure is 48.4 atm, McCutcheon and Elimelech [34] obtained water flux of 10.08 LMH. Clearly, thermo-responsive draw solutions fall short significantly. The results presented above may be rationalized qualitatively as follows. Fig. 5 shows that high initial water fluxes, comparable to those achievable with salt solutions, can be obtained with GE draw solutions. However, the flux rapidly declines (Fig. 5) and this is accompanied by an irreversible increase in membrane resistance (Fig. 6). Such rapid decline of water flux was also reported for fatty acids as draw agents [39]. (Organic molecules are known to cause membrane fouling [40].) It is clear from Table 6 that there is appreciable reverse flux of GE, which shows that GE does dissolve into and diffuse through the dense layer of the membrane; it seems reasonable to hypothesize that some of this dissolved GE is bound essentially irreversibly (and is responsible for the irreversible loss in water flux). The more strongly bound GE could be inside the dense layer and/or on the surface as adsorbed layer. Fig. 6 reveals a greater overall resistance for a membrane that had been
Fig. 7. Water permeation flux using RO membranes. Experiment were conducted in stirred- dead end cell at 120 psi and 295 K. Water flux was first measured using pure water as feed solution prior to RO experiments using glycol ether solutions. Experiments were then performed by loading 1 wt% and 5 wt% aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5. The loading level of GE in the retentate nearly doubled during the course of each experiment; correspondingly the osmotic pressure to be overcome increased as well. See text for further discussion.
7
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
4. Summary
Table 7 Percent rejection and flux normalized rejection (FNR) of glycol ethers by various RO membranes. 1 wt% and 5 wt% aqueous solutions of a composite glycol ether mixture with a TPM:TPnB weight ratio of 87.5:12.5 were used as feed solution. Relevant membrane flux of 20 LMH was used to calculate FNR for different RO membrane.
GE-AK ACM3 ACM4 73 U BW30
1% GE solution
5% GE solution
FNR for 1% GE solution
FNR for 5% GE solution
98.1 96.5 97.8 98.4 98.1
96.2 95.3 96.8 96.7 95.9
98.7 98.3 99.4 99.5 99.5
98.6 99.1 99.6 99.4 99.5
In this study the applicability of mixtures of two glycol ethers, tripropylene glycol methyl ether and tripropylene glycol n-butyl ether, as draw solutions was assessed. These glycol ethers are larger in size than the glycol ethers used in previous studies [14]. Cloud point temperatures of aqueous solutions of mixtures of TPM and TPnB at different total concentrations and composition ratios were determined. Water activities and kinematic viscosities of draw solutions as well as selfdiffusion coefficients of the GEs in the draw solutions were determined at different temperatures. Trans-Membrane water flux and reverse flux of GE during forward osmosis were measured at several different operating conditions. Removal of residual glycol ethers from the water–rich phase that would be obtained upon phase separation of diluted draw solution through a second-stage RO step was also studied. One can tune the cloud point temperature locus by adjusting the TPM:TPnB ratio, which is attractive for process optimization. Addition of TPnB to TPM lowers the cloud point temperature and the osmotic pressure (at a specified glycol ether wt%). The osmotic pressure and kinematic viscosity of draw solution were found to increase with glycol ether concentration, but decrease sharply with increase in temperature. The self-diffusion coefficient for the glycol ethers increases with increasing temperature. These led to a complex temperature dependence of trans-membrane water flux in FO operation. In forward osmosis experiments, PRO mode operation yielded greater flux than the FO mode, which is well known in the literature. PRO mode operation with a small pressure assist was the most favorable mode in our studies. Similar salt rejection was observed for all tested conditions. Although these draw agents do afford high osmotic pressures, they lead to appreciable initial loss of trans-membrane water flux and reverse solute flux. The water flux is hindered by concentration polarization resulting from high viscosities of these draw solutions as well as adherence of GE to the dense layer. In addition, the water produced using such draw agents would require one or two finishing RO steps to recover most of the draw solutes and this RO step is also prone to initial loss of trans-membrane water flux. In a single RO step operating at an applied pressure of 120 psi, > 95% rejection of the glycol ether could be achieved, but depending on the concentration of glycol ethers in the feed, 400–1000 ppm of glycol would present in the permeate. Additional RO steps or operation of the RO step at higher applied pressures could recover most of these glycol ethers. On the whole, the present study shows that there are significant challenges to using these types of draw agents. Membrane modifications to alleviate the acute initial loss of trans-membrane water flux are perhaps the most critical need if such draw agents are to be used. Most studies on FO tend to impose high circulation rate of the draw agent to minimize ECP. However, this is impractical for high-viscosity draw agents; the energy spent on circulation lowers energy efficiency and the higher pressures that build up on the draw solution side at higher circulation rates adversely affect the trans-membrane water flux. Water produced in processes using such draw agents will have to be subjected to multiple finishing RO steps to recover most of the draw solutes. Even then, the water will likely require additional adsorption step prior to being suitable for human consumption. Indirect use of thermo-responsive material as in the process described by Rajagopalan [19], which uses an ionic salt as a draw agent appears to be more promising.
which is roughly consistent with the observed difference in the water fluxes. Membrane fouling has been investigated extensively by many researchers [42,43]. A combination of relatively smooth, hydrophilic, and negatively charged film layer typically produces better water permeability, salt rejection, and fouling resistance in water purification applications. Loss of water flux and need for frequent cleaning are mitigated in practical field conditions by including an effective pretreatment to remove foulants. Various strategies have been suggested for RO applications including chemical or physicochemical pretreatments [44,45]. In the present context, as the loss of performance arises due to the draw agent, the usual pretreatment options do not appear viable. Membrane alteration to inhibit adsorption of the draw agent on the dense membrane later is likely to be a more fruitful avenue of further development. Rejection of glycol ether mixture by RO membranes is presented in Table 7. None of the RO membranes tested was able to reject the glycol ethers completely. Thus, in order to produce chemical-free water through a second-stage polishing step in a liquid-liquid phase separation based desalination process, one must use draw agents with molecular weight higher than those used in the present study. However, increasing the molecular weight lowers the osmotic pressure (at given wt% of draw solution) and increases the viscosity, both of which have adverse effects on membrane flux in the FO step. With higher molecular weight draw agents, one could deploy NF instead of RO in the second stage. Lower percent rejection of the solute is observed in the RO step for the feed solution with 5 wt% GE mixture. This is likely due to difference between effective driving forces for RO in our experiments where we held the applied pressure constant in all the experiments. It is known that for RO process, solute rejection can vary with applied pressure [41]. Higher the applied pressure, higher the rejection is. Feed solution with 5 wt% GE has higher osmotic pressure and needs higher applied pressure in order to reach the effective driving force experienced by the 1 wt% solution. To probe this effect further, we have calculated flux normalized rejection (FNR), using the procedure described by Adhikari et al. [41], and an industrially relevant membrane flux of 20 LMH for all the RO membranes. The FNR values are reported in Table 7. The FNR values 1 and 5 wt% cases are closer to each other than the observed percent rejection, which suggests that applied effective pressure is an important consideration in GE rejection by RO membranes. Before closing, we note that the shape of the phase diagram (see Fig. 1) is very important in simplification of the second stage RO (or NF) step. Specifically, it would be much more attractive if the water-rich phase in the two-phase mixture is naturally very low in the concentration of the draw agent. DEH used in Nakayama et al. [14] and PPG425 used by Jorgensen [17] have this character. It would also be interesting to study draw solutes with well-controlled extents of branching to understand how branching affects reverse solute flux.
Acknowledgment This work was supported by Princeton University through the Helen Shipley Fund for Innovation. References [1] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications,
8
Desalination xxx (xxxx) xxx–xxx
S. Darvishmanesh et al.
osmotically driven membrane processes, Desalination 312 (2013) 31–38. [26] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, A novel ammonia—carbon dioxide forward (direct) osmosis desalination process, Desalination 174 (1) (2005) 1–11. [27] 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, J. Membr. Sci. 278 (1) (2006) 114–123. [28] S.P. Christensen, et al., Mutual solubility and lower critical solution temperature for water + glycol ether systems, J. Chem. Eng. Data 50 (3) (2005) 869–877. [29] H. Huang, R. Verrall, B. Skalski, Study of the aggregation behavior of mixed micelles of anionic surfactants and ethoxylated alcohols by time-resolved fluorescence quenching, Langmuir 13 (18) (1997) 4821–4828. [30] A.D. Wilson, F.F. Stewart, Deriving osmotic pressures of draw solutes used in osmotically driven membrane processes, J. Membr. Sci. 431 (2013) 205–211. [31] A.D. Wilson, Design of the Next-Generation FO Draw Solution, Rational Design of Next-generation Nanomaterials and Nanodevices for Water Applications, International Water Association (IWA), 2016, pp. 103–130. [32] M. Xie, et al., Effects of feed and draw solution temperature and transmembrane temperature difference on the rejection of trace organic contaminants by forward osmosis, J. Membr. Sci. 438 (2013) 57–64. [33] A. Sagiv, et al., Analysis of forward osmosis desalination via two-dimensional FEM model, J. Membr. Sci. 464 (2014) 161–172. [34] J.R. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal concentration polarization on flux behavior in forward osmosis, J. Membr. Sci. 284 (1) (2006) 237–247. [35] G.T. Gray, J.R. McCutcheon, M. Elimelech, Internal concentration polarization in forward osmosis: role of membrane orientation, Desalination 197 (1) (2006) 1–8. [36] Y. Xu, et al., Effect of draw solution concentration and operating conditions on forward osmosis and pressure retarded osmosis performance in a spiral wound module, J. Membr. Sci. 348 (1) (2010) 298–309. [37] S. Phuntsho, et al., Influence of temperature and temperature difference in the performance of forward osmosis desalination process, J. Membr. Sci. 415 (2012) 734–744. [38] G. Blandin, et al., Validation of assisted forward osmosis (AFO) process: impact of hydraulic pressure, J. Membr. Sci. 447 (2013) 1–11. [39] P. Zhao, et al., Fatty Acid Fouling of Forward Osmosis Membrane: Effects of pH, Calcium, Membrane Orientation, Initial Permeate Flux and Foulant Composition, Journal of Environmental Sciences (2016). [40] B. Mi, M. Elimelech, Chemical and physical aspects of organic fouling of forward osmosis membranes, J. Membr. Sci. 320 (1) (2008) 292–302. [41] B. Adhikari, et al., Compatibility study of nanofiltration and reverse osmosis membranes with 1-cyclohexylpiperidenium bicarbonate solutions, J. Membr. Sci. 527 (2017) 228–235. [42] M. Goosen, et al., Fouling of reverse osmosis and ultrafiltration membranes: a critical review, Sep. Sci. Technol. 39 (10) (2005) 2261–2297. [43] K.O. Agenson, T. Urase, Change in membrane performance due to organic fouling in nanofiltration (NF)/reverse osmosis (RO) applications, Sep. Purif. Technol. 55 (2) (2007) 147–156. [44] Y. Soffer, R.B. Aim, A. Adin, Membrane for water reuse: effect of pre-coagulation on fouling and selectivity, Water Sci. Technol. 42 (1–2) (2000) 367–372. [45] A. Schäfer, A.G. Fane, T. Waite, Cost factors and chemical pretreatment effects in the membrane filtration of waters containing natural organic matter, Water Res. 35 (6) (2001) 1509–1517.
and recent developments, J. Membr. Sci. 281 (1) (2006) 70–87. [2] S. Lee, et al., Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO), J. Membr. Sci. 365 (1) (2010) 34–39. [3] Y.-J. Choi, et al., Toward a combined system of forward osmosis and reverse osmosis for seawater desalination, Desalination 247 (1) (2009) 239–246. [4] V. Yangali-Quintanilla, et al., Indirect desalination of Red Sea water with forward osmosis and low pressure reverse osmosis for water reuse, Desalination 280 (1) (2011) 160–166. [5] T.-S. Chung, et al., Forward osmosis processes: yesterday, today and tomorrow, Desalination 287 (2012) 78–81. [6] J.E. Miller, L.R. Evans, Forward osmosis: a new approach to water purification and desalination, Report Prepared by Sandia National Laboratories, Albuquerque, New Mexico 87185 and Livermore, California 94550, 2006. [7] B. Van der Bruggen, P. Luis, Forward osmosis: understanding the hype, Rev. Chem. Eng. 31 (1) (2015) 1–12. [8] D.L. Shaffer, et al., Forward osmosis: where are we now? Desalination 356 (2015) 271–284. [9] R.L. McGinnis, M. Elimelech, Global challenges in energy and water supply: the promise of engineered osmosis, Environ. Sci. Technol. 42 (23) (2008) 8625–8629. [10] McGinnis, R.L., Osmotic desalination process. 2010, U.S. Patent No. US20100108587. [11] Cath, T.Y., A.E. Childress, and C.R. Martinetti, Combined membrane-distillationforward-osmosis systems and methods of use. 2011, U.S. Patent No. 8029671. [12] Carmignani, G., S. Sitkiewitz, and J.W. Webley, Recovery of retrograde soluble solute for forward osmosis water treatment. 2015, U.S. Patent No. 20120267308. [13] Iyer, S., Systems and methods for forward osmosis fluid purification using cloud point extraction. 2011, U.S. Patent No. 20100155329. [14] D. Nakayama, et al., Lower critical solution temperature (LCST) phase separation of glycol ethers for forward osmotic control, Phys. Chem. Chem. Phys. 16 (11) (2014) 5319–5325. [15] Y. Mok, et al., Circulatory osmotic desalination driven by a mild temperature gradient based on lower critical solution temperature (LCST) phase transition materials, Phys. Chem. Chem. Phys. 15 (44) (2013) 19510–19517. [16] Y. Cai, et al., Towards temperature driven forward osmosis desalination using semiIPN hydrogels as reversible draw agents, Water Res. 47 (11) (2013) 3773–3781. [17] M.K. Jørgensen, Investigation of Polypropylene Glycol 425 as Possible Draw Solution for Forward Osmosis, Aalborg University, Aalborg, 2009. [18] Lazare, L., Desalination of sea water. 1968, U.S. Patent No. 3386912. [19] Rajagopalan, N. and V.A. Patel, Solvent removal and recovery from inorganic and organic solutions. 2014, Patent US8852436. [20] R. Wei, et al., Highly permeable forward osmosis (FO) membranes for high osmotic pressure but viscous draw solutes, J. Membr. Sci. 496 (2015) 132–141. [21] Lee, Y., et al., Thermo-responsive draw solute for forward osmosis and method for water desalination and purification using the same. 2012, Google Patents. [22] Osmometers - KNAUER, (Available from:), http://www.knauer.net/en/products/ osmometers.html, (2017). [23] AquaLab: Water Activity Meters and Moisture Analyzers, (Available from:), http:// www.aqualab.com, (2017). [24] E.C.W. Clarke, D.N. Glew, Evaluation of the thermodynamic functions for aqueous sodium chloride from equilibrium and calorimetric measurements below 154 C, J. Phys. Chem. Ref. Data 14 (2) (1985) 489–610. [25] T.Y. Cath, et al., Standard methodology for evaluating membrane performance in
9