Environ. Sci. Technol. 2009, 43, 6769–6775 Solute Coupled Diffusion in Osmotically Driven Membrane Processes NATHAN T. HANCOCK AND TZAHI Y. CATH* Division of Environmental Science and Engineering, Colorado School of Mines, Golden, Colorado 80401 Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x Received April 18, 2009. Revised manuscript received July 10, 2009. Accepted July 13, 2009. Forward osmosis (FO) is an emerging water treatment technology with potential applications in desalination and wastewater reclamation. In FO, water is extracted from a feed solution using the high osmotic pressure of a hypertonic solution that flows on the opposite side of a semipermeable membrane; however, solutes diffuse simultaneously through the membrane in both directions and may jeopardize the process. In this study, we have comprehensively explored the effects of different operating conditions on the forward diffusion of solutes commonly found in brackish water and seawater, and reverse diffusion of common draw solution solutes. Results show that reverse transport of solutes through commercially available FO membranes range between 80 mg to nearly 3,000 mg per liter of water produced. Divalent feed solutes have low permeation rates (less than 1 mmol/m2-hr) while monovalent ions and uncharged solutes exhibit higher permeation. Findings have significant implications on the performance and sustainability of the FO process. Introduction Membrane processes such as reverse osmosis (RO) are widely used in water treatment. RO membranes can reject most constituents present in impaired water, but they can achieve only moderate water recovery (1). RO has additional limitations, including high propensity to fouling and scaling due to the presence of dissolved organic matter and sparingly soluble salts in feed streams (2). Osmotically driven membrane processes, such as forward osmosis (FO), may address these deficiencies by taking advantage of energy and transport characteristics unique to this process (3-5); these include high solute rejection, low-pressure operation, and low fouling propensity (5-7) that can assist RO in achieving higher water recoveries. Studies have identified FO as a potential process for purification of industrial and domestic wastewater (3, 7-9). Novel FO membrane bioreactors have been studied for wastewater reclamation (10), and FO is being investigated for desalination of seawater (11) and brackish groundwater (4). Osmosis is the spontaneous diffusion of water through a semipermeable membrane from a feed solution of higher water chemical potential (lower osmotic pressure) to a draw solution (DS) of lower water chemical potential (higher osmotic pressure) (12). During this process, the feed solution is concentrated and the draw solution is diluted. To produce * Corresponding author phone: 303-273-3402; fax: 303-273-3413; e-mail: tcath@mines.edu. 10.1021/es901132x CCC: $40.75 Published on Web 07/29/2009 2009 American Chemical Society purified water and sustain driving force in the FO process, continuous reconcentration of the DS by RO or distillation is required. Although increasing entropy of the system by dilution is thermodynamically unfavorable, specific applications may realize net benefits in terms of system performance and operating cost (5, 13). While the unique characteristics of FO make it an attractive process for pretreatment of impaired water before treatment with RO or distillation, FO is not without limitations, many of which result from insufficient development and optimization of existing FO membranes and DSs. Low water fluxes despite high driving force is one drawback of currently used FO membranes (14); yet, solute diffusion from the DS into the feed, and vice versa, may pose greater limitation on future implementation of FO and other osmotic processes. Solvent Transport. Intensive research has focused on the effects of concentration polarization (CP) on the nonlinearity of solvent (water) transport in FO (11, 14-16). CP is mostly known in pressure-driven membrane processes where a concentrationboundarylayerdevelopsonthemembrane-feed interface because of the preferential diffusion of water through the membrane (concentrative CP). In FO, both dilutive internal CP in the membrane porous support layer and external CP at the membrane-DS interface have the strongest effect on diminishing water flux (more in Supporting Information S1) (5). Solute diffusion is considered to have negligible effects on water flux; however, its ramifications on system optimization and environmental impact must be considered. Solute Transport: Bidirectional Diffusion in FO. The flux of an individual solute (Js) through semipermeable membranes is governed by chemical potential gradients and is commonly described using Fick’s Law (17): Js ) B∆c (1) where B is the solute permeability coefficient and ∆c is the trans-membrane concentration differential (18). In RO, solutes diffuse from the feed into the permeate; however, in FO, solutes diffuse in two directions: from the feed into the DS (i.e., forward diffusion) and simultaneously from the DS into the feed (i.e., reverse diffusion). Reverse diffusion is different from back diffusion which is a the result of CP phenomena in the feed-membrane boundary layer (17, 19). Reverse and forward diffusion are the focus of the current study. During DS reconcentration cycles, solute diffusion through the FO membrane may jeopardize a combined FO/RO or FO/distillation system. Ions that codiffuse with water into the DS may accumulate in the reconcentration process, and those that have low solubility may precipitate and adversely affect the performance of the hybrid process. Scale inhibitors could be added to the DS, but will increase operating costs. Solutes lost through reverse diffusion require replenishment of the DS, and accumulation of DS solutes in the feed may pose toxicological challenges for sensitive receiving environments or prove detrimental for adjacent treatment processes. To optimize osmotic processes, a closer investigation of solute transport phenomena is essential. To a first order approximation, the literature asserts that salt transport across the membrane is governed by the concentration differential of each individual solute (17). This suggests that while the cumulative concentrations of various solutes affect the osmotic pressure, and therefore the water flux, only the concentration gradient of a particular solute will influence the flux of that solute across the membrane as expected from eq 1. VOL. 43, NO. 17, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6769 Also, eq 1 may be acceptable for dilute solutions; however, early theoretical studies (20, 21) used an irreversible thermodynamic framework to deduce that for RO, diffusion of solutes occurs through both diffusion and convective transport originating from coupled effects associated with solvent diffusion. This understanding provides a second level approximation for a binary system, in which solute flux (Js) through a semipermeable membrane occurs in the same direction as water flux, and can be represented by the convective-diffusion equation: Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x Js ) ω∆π + (1 - σ)Jvjc (2) where the first product represents diffusive mass transport with ω being the membrane’s solute permeability coefficient, and the second term represents convective mass transport, where Jv is the water flux, jc is the average interfacial solute concentration gradient between the feed and permeate sides, and σ is the reflection coefficient defined as the ratio between the negative solute-water phenomenological coefficient divided by the pure water permeability (12). Although eq 2 is limited in its application for modeling of FO processes, especially in mixed electrolyte solutions, it serves to emphasize the complex coupled nature of mass transport in FO. Donnan equilibrium effects are one consequence of coupled diffusion (22), and other unique characteristics of solute diffusion in membrane processes are summarized elsewhere (23). The main objective of this study was to investigate bidirectional mass transport of solutes in forward osmosis of pure and mixed electrolyte solutions, with different draw solutions, under various operating conditions, and with currently available FO membranes. Specifically, the effects of solution chemistry, membrane morphology, and hydrodynamic conditions on mass transport were investigated. Solute coupled diffusion in osmotically driven membrane processes requires further exploration before FO processes can achieve broad acceptance and commercialization for treatment of impaired water. The coupled nature of both solvent-solute and solute-solute mass transport during FO introduces a new level of intricacy that has not been explored before. Materials and Methods FO Membranes. Cellulose acetate membranes are well suited for use in osmotically driven membrane processes (5, 7, 24). Two flat-sheet cellulose triacetate (CTA) membranes, specifically developed for FO by HTI (Albany, OR), were tested in this study. The membranes are entitled CTA-1 and CTA-2; CTA-1 is the less permeable and more selective of the two. During all experiments, the membranes were oriented with their active layer facing the feed. To maintain high quality assurance and control, membrane integrity tests were performed at the beginning and end of every set of experiments. Additional information is available in Supporting Information S2. FO Bench-Scale Apparatus. FO flow cells were constructed with symmetric flow chambers on both sides of the membrane that facilitated parallel, cocurrent flow along the membrane. A supervisory control and data acquisition (SCADA) system was developed to maintain constant experimental conditions (i.e., feed volume, DS conductivity, system temperature of 18 °C) and to collect data during experiments. Feed and DS were continuously circulated between their respective tanks and the membrane cell at 1.6 L/min using gear pumps (Cole-Parmer, Vernon Hills, IL) and feed and DS samples were intermittently drawn for analysis. A flow schematic and description of the system is provided in Supporting Information S3. Solution Chemistries. Salts and other chemicals used in the experiments were all ACS grade and solutions were 6770 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 17, 2009 prepared in deionized water. Complete description is provided in Supporting Information S4. Experimental Procedures NaCl and MgCl2 Reverse Diffusion Characterization. The effects of osmotic driving force on forward diffusion of water and reverse diffusion of DS salts were investigated. Both membranes were tested with deionized feedwater and DS concentrations that ranged from 0.34 M NaCl to saturation. The CTA-2 membrane was also tested with MgCl2 DS having osmotic pressures similar to NaCl experiments (excluding NaCl saturation). Each experiment was terminated after 1-2 h of steady reverse diffusion. See Supporting Information S5 for details on mass transfer calculations. NaCl, MgCl2, and NH4HCO3 Specific Reverse Salt Flux. Specific reverse salt flux (Jspecific) is defined here as the ratio of salt flux (Js (mg/m2-hr)) in the reverse direction and water flux (Jw (L/m2-hr)) in the forward direction, and is related to the water-salt selectivity of the membrane. This quantity is directly related to process efficiency and sustainability. Specific reverse salt flux was quantified through experiments with three DSs at different concentrations, with deionized water feed, and with the two FO membranes. The concentrations of all DS were adjusted to have an osmotic pressure of 4 MPa. NaCl and MgCl2 DSs were also tested at an osmotic pressure of 1.9 MPa (constant concentrations of 0.43 and 0.86 M for NaCl, and 0.29 and 0.57 M for MgCl2). NH4HCO3 DS was tested at one concentration (1.02 M). See Supporting Information S5 for mass transfer calculations and S6 for more details on experimental procedures. Coupled, Multicomponent Diffusion Experiments. Feed solutions composed of a single salt each were employed to elucidate the influence of various chemical potential gradients on the forward and reverse diffusion of individual chemical species. Feed solutions were prepared with concentrations of either 1 or 2 g/L as the feed solution salt (CaSO4, K2SO4, MgSO4, Ba(NO3)2, or NH4HCO3) or as the target ion (H3BO3-B or Na2SiO3-SiO2). Experiments with the CTA-1 membrane were conducted with select feed solution salts, and included 1 or 2 g/L of CaSO4, NH4HCO3, or Na2SiO3-SiO2. Each feed solution was tested with two different DS concentrations of 25 or 50 g/L NaCl. Tests were also conducted with synthetic brackish water feed solution (3.85 g/L MgCl2 · 6H2O, 5.13 g/L CaCl2 · 2H2O, 2.55 g/L Na2SiO3 · 9H2O, 2.7 g/L Na2SO4, and 18 mM HCl) utilizing CTA-2. Experiments with NaCl and MgCl2 DSs were conducted with two different concentrations leading to osmotic pressures of 1.9 and 4 MPa, while an experiment with NH4HCO3 DS was conducted with one concentration, resulting in an average osmotic pressure of 4 MPa. More details are available in Supporting Information S5 and S6. Effect of Crossflow Velocity on the Reverse Diffusion of NaCl. Reverse diffusion of NaCl was investigated under different feed (deionized water) and DS (50 g/L) crossflow velocities (37, 75, and 112 cm/s). Nine experiments were conducted with all combinations of flow velocities on both sides of the membrane. Reverse solute flux was calculated with the same method used in the NaCl and MgCl2 reverse diffusion experiments. Pressures and temperatures were maintained constant. Analytical Procedures. Samples were analyzed for cation concentrations using inductively coupled plasma (Optima 3000, Perkin-Elmer, Fremont, CA) and for anion concentrations using ion chromatography (ICS-90, Dionex, Sunnyvale, CA). Samples containing ammonium were analyzed using Hach spectrophotometer method 10031. Samples containing bicarbonate were analyzed using titration method 8203 (model 16900, Hach, Loveland, CO) with either 0.1 or 0.01 M H2SO4 solution. More details are available in Supporting Information S6. Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x Results and Discussion NaCl and MgCl2 DSs: Water and Reverse Salt Fluxes. Water flux and reverse salt flux as a function of DS osmotic pressure are illustrated in Figure 1a for experiments conducted with NaCl DS and with the CTA-1 and CTA-2 membranes. Results indicate that water flux and NaCl reverse diffusion increase with increasing DS concentration. The higher values observed for the CTA-2 confirm that this membrane is less selective and more permeable compared to the CTA-1; water flux was 37-48% higher and reverse salt flux was 62-72% higher through the CTA-2 than through the CTA-1. As observed in this and past studies (14, 19), a logarithmic relationship exists in FO between water flux and driving force, and because the current experiments were conducted with deionized feedwater, this behavior is primarily the result of internal CP. Data in Figure 1a also indicate that internal CP affects reverse salt flux. When the diffusing water dilutes the DS in the porous support layer, the salt concentration differential is also reduced. Equations 1 and 2 predict that a reduction in concentration differential will result in reduced salt diffusion. Due to charge balance restrictions in the feed solution, we assumed that both anions and cations reverse diffused equally on a molar-equivalent basis. This assumption was further explored and discussed below. A second set of experiments was conducted with the CTA-2 membrane using MgCl2 DS and deionized water feed. MgCl2 concentrations were adjusted to have osmotic pressure (theoretical driving force) similar to the DSs used in the NaCl experiments. Water flux and reverse salt flux as a function of the bulk DS osmotic pressure are shown in Figure 1b. Despite the similar osmotic pressures of the two DS, it can be seen that MgCl2 is a weaker DS; compared to results with NaCl DS, water flux was 25-30% lower when using MgCl2. However, in terms of salt flux, MgCl2 exhibited much slower reverse salt flux: 59-67% lower than NaCl. The lower water flux observed when using the MgCl2 DS is very likely the combined effects of higher viscosity that increased the severity of external CP, and the lower diffusion coefficient of magnesium compared to sodium that increased the severity of internal CP. The reduction in reverse salt flux during MgCl2 experiments may also result from Donnan equilibrium effects, whereby the relatively large magnesium ion diffused slower through the membrane and subsequently limited the diffusion of the counterion (chloride); yet, intensified size exclusion caused by ion shielding and nonspecific side reactions between magnesium ions may also play a role. Specific Reverse Salt Flux. To optimize FO processes it is important to consider bidirectional diffusion; one method to quantitatively evaluate this phenomenon is the ratio between reverse salt flux and (forward) water flux. Larger ratios reflect a decrease in the selectivity of the membrane and lower efficiency of the process. Specific reverse salt fluxes of three DSs through two different FO membranes are illustrated in Figure 1c. For both membranes, MgCl2 had the lowest specific reverse salt flux; when tested with the CTA-2, water flux with the MgCl2 DS was 33% lower compared to results with NaCl, but reverse salt flux with MgCl2 was 57% lower compared to NaCl. Specific reverse salt flux of NaCl was 55% higher for CTA-1 and 64% higher for CTA-2 compared to MgCl2. For both membranes, the smaller monovalent sodium ion has a higher diffusive flux compared to magnesium, which also allows for more chloride ions to diffuse into the feed solution. Chemical analysis of samples drawn during each experiment confirms that DS cations and anions diffuse through the membrane at equal molar-equivalent ratios. More interestingly, both the NaCl and MgCl2 DSs were tested at two different concentrations (driving forces) and they induced similar specific reverse salt flux through each FIGURE 1. Effects of draw solution chemistry on water flux, reverse salt flux, and specific reverse salt flux in (a) NaCl DS experiments with the CTA-1 and CTA-2 FO membranes, (b) NaCl and MgCl2 DSs experiments with the CTA-2 membrane only, and (c) NaCl, MgCl2, and NH4HCO3 DS experiments with the CTA-1 and CTA-2 membranes. In investigation of specific reverse salt fluxes, the NaCl and MgCl2 DS were tested with both the CTA-1 and CTA-2 FO membranes and at two different driving forces (1.9 and 4 MPa), and the NH4HCO3 DS was tested only with an average osmotic pressure driving force of 4 MPa. Error bars indicate variation of results for different DS concentrations. When present, a secondary x-axis correlates osmotic pressure to molar DS concentration. of the membranes (reflected in error bars shown in Figure 1c). The similarity in specific reverse salt flux despite differences in the osmotic driving force indicates that this ratio is related to the membrane’s selectivity. Reverse salt flux of NH4HCO3 was far greater than that of either MgCl2 or NaCl. NH4HCO3 has a higher diffusivity than either NaCl or MgCl2 at the concentration tested (Figure S41b in Supporting Information), which reduces the severity of internal CP. However, the substantial increase in reverse VOL. 43, NO. 17, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6771 Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x solute flux for ammonium and bicarbonate (observed with both membranes) suggests that there may be additional physicochemical effects that enhance the ability of these ions to diffuse through these CTA membranes. The magnitude of specific reverse salt flux of the three DSs provides a unique observation on the sustainable use of these DSs. For example, for each liter of water that permeates through the CTA-2 membrane, 140, 400, or 2,900 mg of MgCl2, NaCl, or NH4HCO3 are lost from the DS, respectively. This mass of DS salts will transfer to the feed solution with possible implications for downstream or adjacent processes, and will require replenishment to maintain operating conditions in the FO system. Coupled, Multi-Component Diffusion. In addition to examining the pure reverse diffusion of parent salts, experiments were conducted to elucidate the competitive nature of simultaneous forward and reverse diffusion of solutes in FO. Single Salt Feed Solution Experiments with CTA-2. A set of four experiments was conducted with each feed solution to determine the mass transport characteristics of water and individual ions. Experiments were conducted with individual feed solutions and with NaCl DS. Solute fluxes in the forward and reverse directions as a function of feed and DS chemistries are summarized in Figure 2. Results indicate that sodium and chloride reverse diffuse at nearly equal-molar proportions for most feed solutions tested (MgSO4, CaSO4, K2SO4, H3BO3, and NH4HCO3). The only exception was experiments conducted with Ba(NO3)2 where chloride reverse diffused substantially faster than sodium. One possible explanation is that the relatively small, polar nitrate ion readily diffuses through the membrane while the larger barium ion diffuses at a substantially lower rate (17, 25, 26). This most likely led to a charge imbalance between the two solutions that was corrected by a faster reverse diffusion of chloride into the feed solution. Data in Figure 2 correlate closely with the anticipated diffusive behavior of solutes based on existing knowledge of the membrane structure of CTA-2 (11). The membrane is dense, nonporous, and negatively charged in the pH range of interest. Diffusion behavior observed in Figure 2 indicates that size exclusion and electrostatic effects have a substantial role in controlling forward solute transport through the membrane. For feed solutes, low molar fluxes for magnesium, calcium, and barium indicate that divalent cations, with large hydration radii, diffuse less readily through the membrane than monovalent ions or polar molecules. Similarly, the large negatively charged divalent sulfate ion had a very low molar flux through the FO membrane. However, when present with monovalent potassium ion, which exhibited a relatively high forward flux, sulfate diffused faster through the membrane, apparently to aid in maintaining solution electroneutrality. Bicarbonate was also observed to diffuse at a relatively high rate through the membrane, which may introduce process limitations for FO desalination of brackish water. Diffusion of salts composed of a common sulfate anion agrees with previous studies (27) and follows the lyotropic series, whereby the hydraulic radii and valence state of the cation is responsible for limiting forward ion diffusion (17). Coinciding with previous observations (28), boric acid diffused significantly faster through the membrane than all other solutes tested; this has important regulatory and technical implications if FO is to be used as a seawater desalination process. Results from these experiments revealed that at an average pH of 6, boric acid rejection by the CTA-2 membrane was 12 and 33% for low and high water flux conditions, respectively. In comparison, commercial RO membranes may achieve 38-95% boron rejection at slightly elevated pH of 8 (29). Silica is the only neutrally charged molecule that was not shown to readily diffuse through the membrane. The Na2SiO3 feed was the only solution that 6772 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 17, 2009 FIGURE 2. Summary of solute flux data from single salt experiments with the CTA-2 membrane and NaCl DS. Forward diffusing solutes are listed on left y-axis with values on the lower x-axis, while reverse diffusing solutes are listed on the right y-axis with values on the upper x-axis. (blue) Represents experiments with 25 g/L NaCl DS and 1 g/L feed solution (FS), (magenta) 25 g/L DS and 2 g/L FS, (yellow) 50 g/L DS and 1 g/L FS, (green) 50 g/L DS and 2 g/L FS concentration. (*) Designates solute flux divided by 10, (†) designates solute flux divided by 104, (X) data unavailable, and (diamonds) designates a net solute flux because the same ion is present in both the feed and DS. became moderately opaque during the course of each experimentsmost likely formation of silica colloids that were well rejected by the FO membrane. The pH of the feed solution shifted during specific experiments. This raises an interesting question regarding the bidirectional diffusion of hydrogen and hydroxide in FO. Findings and discussions are available in Supporting Information S7. Yet, from results in Figure 2 it is difficult to decipher if the reverse diffusion of DS solutes physically inhibits the forward diffusion of feed solutes. Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x FIGURE 3. Solute flux during single salt experiments with the CTA-1 tighter FO membrane and NaCl DS. Symbols and color codes are similar to those defined in Figure 2. Single Salt Feed Solution Experiments with CTA-1. Several single salt feed solution experiments were repeated with the tighter CTA-1 membrane using NaCl DS. Molar fluxes as a function of feed and DS chemistries are summarized in Figure 3. In all cases, the feed and DS solutes diffused slower through the more selective, less permeable CTA-1 than in parallel experiments with the CTA-2. Results indicate that forward solute flux decreased by as high as 88% (in the case of calcium) and by as low as 52% for bicarbonate. Reverse diffusion of sodium and chloride through CTA-1 was observed to decrease by equal rates of 82%, 73%, and 70% for the experiments conducted with CaSO4, Na2SiO3, and NH4HCO3 feed solutions, respectively. It is likely that the higher rejection of calcium ions in the CaSO4 experiment generated an equivalent reduction in reverse diffusion of DS solutes because of Donnan equilibrium effects. Brackish Water Feed Experiments with CTA-2. Tests were conducted with synthetic brackish water feed solution to simulate operation of FO with more complex water chemistries. Forward and reverse solute fluxes as a function of feed and DS constituents are illustrated in Figure 4. Results reemphasize that MgCl2 has slower reverse diffusion for a set driving force; however, there is minimal effect of DS on the forward diffusion of brackish water constituents when compared to results obtained with NaCl DS. When present only in the feed solution, it was observed that sodium diffused through the membrane more rapidly than the other constituents, likely because of its relatively small hydration radius and lower valence charge. Calcium and silica diffused through the membrane at nearly equal rates; yet, the lower rejection of silica compared to calcium (results not shown) was not expected based on prior results. One hypothesis is that silica microcolloids are catalyzed in the presence of magnesium and calcium ions (30), and the FO membrane better rejects colloidal than dissolved silica. Lastly, similar to prior results, FIGURE 4. Forward and reverse solute fluxes through the CTA-2 membrane during experiments with synthetic brackish water feed solution and various DS. Forward diffusing solutes are listed on left y-axis with values on the lower x-axis, while reverse diffusing solutes are listed on the right y-axis with values on the upper x-axis. (yellow) Represents experiments with DS osmotic pressure of 1.9 MPa and (green) represents experiments with DS osmotic pressure of 4 MPa. (*) represents solute flux divided by 10 and (diamonds) designates a net solute flux because ion is present in both the feed and DS. the molar diffusion of sulfate was the lowest of all feed solution species tested. Similar forward diffusion trends were observed in experiments conducted with NaCl DSs. Comparable fluxes of both calcium and sulfate were observed; yet, silica diffused even slower than in the MgCl2 DS experiments. Based on the data, it is possible that higher reverse diffusion of DS solutes may have retarded the forward diffusion of silica. Further evidence of this is provided in results from the NH4HCO3 DS experiment. Substantial reverse diffusion of ammonium and bicarbonate, beyond that of either NaCl or MgCl2, was observed. VOL. 43, NO. 17, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6773 Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x FIGURE 5. Effect of feed (deionized water) and DS (50 g/L) crossflow velocity on (a) water flux, (b) reverse salt flux, and (c) specific reverse salt flux. Blue colors represent lower flux values and red colors represent higher flux values. Remarkably, the faster reverse diffusion of these solutes induced faster forward diffusion of sodium and chloride and reduced forward diffusion of other solutes from the feed solution; beyond levels observed during experiments with the other DSs. In particular, the very slow diffusion of silica may also be attributed to an increase in feed solution pH that stems from reverse diffusion of ammonium and bicarbonate into the feed. Effect of Crossflow Velocity on Reverse Diffusion. DS and feed solution flow velocities affect external CP, and DS flow velocity may indirectly affect internal CP; therefore, both may impact water flux (more details on CP effects are available in Supporting Information S1). Although the selective diffusion of water over solutes is purely related to the membrane’s active layer (not the porous support layer), altering external CP may also affect solute flux in FO. Thus, in another set of experiments with the CTA-2 membrane, both water and reverse solute fluxes were measured under different feed and DS flow velocities. Water flux as a function of feed and DS flow velocities is illustrated in Figure 5a. Water flux reached a maximum value at higher and equal flow velocities on both sides of the membrane. Water flux decreased by ap6774 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 17, 2009 proximately 20% from its maximum value when feed velocity was low and DS velocity was high. In most membrane processes, including FO, lower feed velocity results in higher concentrative external CP effects (11, 19), and specifically in FO, DS solutes that reverse diffuse through the membrane may further concentrate the concentration boundary layer on the feed side of the membrane. Similarly, increasing the feed velocity enables a more rapid dilution of reverse diffusing DS solutes; as the concentration in the feed boundary layer decreases, the chemical potential gradient across the membrane increases and consequently water flux also increases. Higher feed velocity and low DS velocity resulted in a mild decrease in water flux (approximately 11%). At higher feed velocities, DS solutes are being mixed faster into the bulk feed solution; this reduces concentrative external CP and initially promotes faster water diffusion. Simultaneously, lower DS velocity combined with higher water flux creates a more severe dilutive external CP at the interface between the support layer of the membrane and the bulk DS. The combination of changes in external and internal CP on both sides of the membrane is most likely the cause for the ultimate decline in water flux at higher feed velocity and low DS velocity. Reverse salt flux as a function of both feed and DS flow velocities is illustrated in Figure 5b for the same experiments. Higher reverse salt flux was observed during experiments with mutually higher flow velocities, and the lower reverse salt flux occurred during lower flow conditions. These results may be explained through the same mechanistic arguments used in describing the water flux data. At lower and equal flow velocities, the reverse diffusing DS salts are concentrated at the membrane surface on the feed side and the concentration boundary layer on the DS side of the membrane is not well mixed and remains diluted. These two phenomena resulted in diminishing chemical potential gradient between the DS and feed solution and retardation in the net diffusion of salts into the feed solution. Conversely, increased shear flow on both sides of the membrane resulted in rapid replenishment of DS solutes to reduce the effect of both dilutive external and internal CP in the bulk DS and porous support layer, respectively, and provided additional mixing to alleviate concentrative external CP in the feed solution. Optimizing process design may require compromising between water production and reverse salt diffusion. The influence of crossflow velocity on specific reverse salt flux provides additional information that can aid in process design and optimization. Specific reverse salt flux as a function of feed and DS crossflow velocities was calculated from data in Figures 5a and b, and is illustrated in Figure 5c. Based on these results, FO processes should be operated with low feed and DS flow velocities to minimize DS solute loss. Increasing DS flow velocity while maintaining low feed flow velocity is shown to produce the most significant adverse effect, amplifying the specific reverse salt diffusion by close to 20% above the minimum value under conditions tested in this study. On the other hand, mitigating reverse salt diffusion by designing FO processes with low crossflow velocity operation may reduce overall process performance. Feed solution solutes will exhibit increased external CP behavior, which will increase forward diffusion of these constituents, membrane fouling, and further reduce water permeation. CP phenomena, both internal and external, affect water flux in FO, but also have significant effects on reverse solute flux. Results from this study also indicate that membrane structure and DS chemistry play an important role in bidirectional diffusion of solutes. In particular, the reverse diffusion of DS solutes and the forward flux of sparingly soluble salts, and boron, may introduce limitations on seawater or brackish water desalination with FO. These current limitations must be addressed through design of FO membranes with increased water permeability and higher selectivity. Solutes must have both high diffusivity in liquids and low permeability through the membrane selective-layer to be a good DS candidate. Process optimization requires trade-off between driving forces for water flux and hydrodynamic and thermodynamic conditions for solute flux. Acknowledgments We acknowledge the support of California Department of Water Resources (Grant 46-7446-R-08). Special thanks to Dr. Dean Heil and Professor John Dorgan, and to HTI for providing FO membranes. Note Added after ASAP Publication Downloaded by COLORADO SCHOOL OF MINES on August 29, 2009 | http://pubs.acs.org Publication Date (Web): July 29, 2009 | doi: 10.1021/es901132x Due to a production error, the wrong Supporting Information file was linked to the version of this paper published ASAP on July 29, 2009; the correct version published ASAP August 4, 2009. Supporting Information Available Details on solvent transport; membrane structure; bench scale system configuration; experimental procedures; and pH shift during specific experiments. This material is available free of charge via the Internet at http://pubs.acs.org. Literature Cited (1) Van der Bruggen, B.; Lejon, L.; Vandecasteele, C. Reuse, treatment, and discharge of the concentrate of pressure-driven membrane processes. Environ. Sci. Technol. 2003, 37 (17), 3733– 3738. (2) Song, L. F.; Hu, J. Y.; Ong, S. L.; Ng, W. J.; Elimelech, M.; Wilf, M. Performance limitations of full-scale reverse osmosis process. J. Membr. Sci. 2003, 214, 239–224. (3) Cath, T. Y.; Gormly, S.; Beaudry, E. G.; Michael, T. F.; Adams, V. D.; Childress, A. E. Membrane contactor processes for wastewater reclamation in space: Part I. Direct osmosis concentration as pretreatment for reverse osmosis. J. Membr. Sci. 2005, 257 (1-2), 85–98. (4) Martinetti, C. R.; Cath, T. Y.; Childress, A. E. High recovery of concentrated RO brines using forward osmosis and membrane distillation. J. Membr. Sci. 2009, 331, 31–39. (5) Cath, T. Y.; Childress, A. E.; Elimelech, M. Forward osmosis: principles, applications, and recent developments. J. Membr. Sci. 2006, 281 (1-2), 70–87. (6) Mi, B.; Elimelech, M. Chemical and physical aspects of organic fouling of forward osmosis membranes. J. Membr. Sci. 2008, 320, 292–302. (7) Holloway, R. W.; Childress, A. E.; Dennett, K. E.; Cath, T. Y. Forward osmosis for concentration of centrate from anaerobic digester. Water Res. 2007, 41, 4005–4014. (8) Cartinella, J. L.; Cath, T. Y.; Flynn, M. T.; Miller, G. C.; Hunter, K. W.; Childress, A. E. Removal of natural steroid hormones from wastewater using membrane contactor processes. Environ. Sci. Technol. 2006, 40 (23), 7381–7386. (9) Hafez, A.; Kheder, M.; Gadallah, H. Wastewater treatment and water reuse of food processing industries. Part II: Technoeconomic study of membrane separation technique. Desalination 2007, 214, 261–272. (10) Achilli, A.; Cath, T. Y.; Marchand, E. A.; Childress, A. E. The forward osmosis membrane bioreactor: A low fouling alternative to MBR processes. Desalination 2009, 239, 10–21. (11) McCutcheon, J. R.; McGinnis, R. L.; Elimelech, M. Desalination by ammonia-carbon dioxide forward osmosis: influence of draw and feed solution concentration on process performance. J. Membr. Sci. 2006, 278 (1-2), 114–123. (12) Mulder, M. Basic Principles of Membrane Technology, 2nd ed.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1997; p 564. (13) McGinnis, R. L.; Elimelech, M. Energy requirements of ammonia-carbon dioxide forward osmosis desalination. Desalination 2007, 207 (1-3), 370–382. (14) Seppälä, A.; Lampinen, M. J. On the non-linearity of osmotic flow. Exp. Thermal Fluid Sci. 2004, 28 (4), 283–296. (15) Elimelech, M.; Bhattacharjee, S. A novel approach for modeling concentration polarization in crossflow membrane filtration based on the equivalence of osmotic pressure model and filtration theory. J. Membr. Sci. 1998, 145, 223–241. (16) Tan, C. H.; Ng, H. Y. Modified models to predict flux behavior in forward osmosis in consideration of external and internal concentration polarization. J. Membr. Sci. 2008, 324, 209–219. (17) Mallevialle, J.; Odendaal, P. E.; Wiesner, M. R. Water Treatment Membrane Processes; McGraw-Hill: New York, 1996. (18) Merten, U.; Lonsdale, H. K.; Riley, R. L. Boundary-Layer Effects in Reverse Osmosis. Ind. Eng. Chem. Fundam. 1964, 3, 210– 213. (19) Loeb, S.; Titelman, L.; Korngold, E.; Freiman, J. Effect of porous support fabric on osmosis through a Loeb-Sourirajan type asymmetric membrane. J. Membr. Sci. 1997, 129, 243–249. (20) Kedem, O.; Katchalsky, A. Thermodynamic analysis of the permeability of biological membranes on non-electrolytes. Biochim. Biophys. Acta 1958, 28, 229–246. (21) Spiegler, K. S.; Kedem, O. Thermodynamics of hyperfiltration (reverse osmosis): criteria for efficient membranes. Desalination 1966, 1, 311–326. (22) Donnan, F. G. The Theory of Membrane Equilibria. Chem. Rev. 1924, 1 (1), 73–90. (23) Cussler, E. L. Diffusion Mass Transfer in Fluid Systems, 2nd ed.; Cambridge University Press, 1997. (24) McCutcheon, J. R.; McGinnis, R. L.; Elimelech, M. A novel ammonia-carbon dioxide forward (direct) osmosis desalination process. Desalination 2005, 174, 1–11. (25) Paugam, L.; Taha, S.; Dorange, G.; Joauen, P.; Quemeneur, F. Mechanism of nitrate ions transfer in nanofiltration depending on pressure, pH, concentration and medium composition. J. Membr. Sci. 2004, 231, 37–46. (26) Nightingale, E. R. Phenomenological theory of ion solvation. Effective radii of hydrated ions. J. Phys. Chem. 1959, 63 (9), 1381–1387. (27) Ghiu, S. M. S.; Camahan, R. P.; Barger, M. Permeability of electrolytes through a flat RO membrane in a direct osmosis study. Desalination 2002, 144, 387–392. (28) Reid, C. E.; Breton, E. J. Water and ion flow across cellulosic membranes. J. Appl. Polym. Sci. 1959, 1 (2), 133–143. (29) Trussell, R. S.; Trussell, R. R. Boron Removal and Reverse Osmosis; www.trusselltech.com/media/1.pdf (accessed April 9, 2009). (30) Sheikholeslami, R.; Zhou, S. Performance of RO membranes in silica bearing waters. Desalination 2000, 132, 337–344. ES901132X VOL. 43, NO. 17, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 6775