WO2008137082A1

WO2008137082A1 - Method for designing membranes for osmotically driven membrane processes

Info

Publication number: WO2008137082A1
Application number: PCT/US2008/005696
Authority: WO
Inventors: Menachem Elimelech; Jeffrey Mc Cutcheon
Original assignee: Yale University
Current assignee: Yale University
Priority date: 2007-05-02
Filing date: 2008-05-02
Publication date: 2008-11-13
Anticipated expiration: 2009-11-02

Abstract

Osmotically driven membrane processes, such as forward osmosis (FO) and pressure retarded osmosis (PRO), rely on the utilization of large osmotic pressure differentials across semi-permeable membranes to generate water flux. The present invention relates to improved membranes for use in such osmotically driven membrane processes. Current generation polymeric membranes used in liquid separations are typically comprised of a selective barrier supported by a porous structure. This structure is not ideal for osmotically driven membrane processes unless certain membrane characteristics are tailored appropriately. The support layer porosity, thickness, tortuosity, and hydrophilicity all play a crucial role in water flux performance across asymmetric semi-permeable membranes. The membrane support layers must be thin, highly porous, non-tortuous, and/or hydrophilic if they are to be used in FO and PRO processes. These goals are to be accomplished without sacrificing water permeability and salt rejection. Various methods for making these new membranes are described herein.

Description

METHOD FOR DESIGNING MEMBRANES FOR OSMOTICALLY DRIVEN

MEMBRANE PROCESSES

CROSS-REFERENCE TO RELATED APPLICATIONS

The present application claims the benefit of U.S. Provisional Application Serial No. 60/927,222, filed May 2, 2007, the subject matter of which is herein incorporated by reference in its entirety.

FIELD OF THE INVENTION

The present invention relates generally to improved membranes for use in osmotically driven membrane processes including forward osmosis and pressure retarded osmosis processes.

BACKGROUND OF THE INVENTION

Membrane technology has revolutionized the separations industry by providing a highly selective low cost alternative to standard separations processes. Pressure driven membrane processes, however, require the use of hydraulic pressure which uses electricity, a high quality and increasingly expensive form of energy. This has led to a growth in the field of osmotically driven membrane processes. These processes rely on transmembrane osmotic pressure which is created by a concentrated draw solution or osmotic agent. Various configurations of osmotically driven membrane processes may be used for separation, concentration/dewatering, and power generation. These, for example, include the ammonia-carbon dioxide forward osmosis desalination process, the ammonia- carbon dioxide osmotic heat engine, pressure retarded osmosis for generation of energy from salinity gradients (e.g., between river water and seawater), and treatment of various liquid wastes via forward osmosis utilizing brines as draw solutions.

Reverse osmosis (RO) is a pressure driven process in which the hydraulic resistance of the membrane is the primary resistance to water flux after the osmotic pressure has been overcome. In RO membranes, the hydraulic resistance is almost entirely limited to the thin dense surface layer (i.e., "rejection layer") as the microporous support layer and fabric backing layer have only minor or insignificant resistance. In contrast, forward osmosis (FO) is a diffusion (osmosis) driven process instead of a pressure driven process, so the factors affecting water flux are dramatically different than in reverse osmosis. Because of this, a high performance forward osmosis membrane requires a dramatically different structure than a reverse osmosis membrane.

Forward osmosis (FO) and other osmotically driven membrane processes have become more popular as a means for the sustainable production of water and power. These processes rely on the osmotic pressure difference between two solutions: a concentrated draw solution (also referred to as the osmotic agent) and a relatively dilute feed solution. Solvent, which is typically water, will move from the dilute feed into the concentrated draw. The osmotic pressure difference takes the place of hydraulic pressure generated by pumps, thereby reducing the electrical power requirements for certain membrane separation processes and in some cases, as with pressure retarded osmosis

(PRO), can generate power.

Water production with FO and power production with PRO have been investigated for a number of years. These processes may also apply to various concentration practices such as those involved with food production and waste treatment and minimization. Further development of these various osmotically driven technologies, while spurred on by economics and energy cost, has been hindered by poor membrane performance.

Most of today's semi-permeable membranes are designed specifically for pressure driven membrane processes, not osmotically driven membrane processes. Poor membrane performance in osmotically driven membrane processes is attributed to mass transfer limitations and interfacial phenomena associated with the internal structure of current generation asymmetric membranes. The chemical compatibility of the membrane in the presence of various s feed and draw solutions must also be considered.

For water to transport across a membrane, an osmotic pressure gradient must be established across a selective portion of the membrane. The osmotic pressure gradient is established between a relatively dilute feed solution and a far more concentrated draw solution. Various configurations of osmotic processes have used a variety of draw solutions, including for example ammonia-carbon dioxide, seawater, magnesium chloride, calcium chloride, sodium chloride, potassium chloride, sucrose, magnesium sulfate, potassium nitrate, ammonium carbonate, dextrose, and sucrose, by way of example and not limitation. The selective portion of the membrane must not allow the passage of solutes from the feed to the draw solution or from the draw solution to the feed solution. High rejection by the membrane, high osmotic efficiency, low toxicity, high diffusivity, and low reactivity with the membrane and system components are all important criteria when selecting the draw solution.

There are two primary obstacles hindering further development of osmotically driven membrane processes. First, as mentioned above, the selection of an appropriate draw solution, or the solute which generates the osmotic driving force, is critical for these processes to be economically viable. These solutions must be selected based on their physical and chemical properties as well as their intended end use (either removed and recovered or consumed). The second and most problematic impediment to further development of these technologies is the lack of suitable membranes designed specifically for osmotically driven membrane processes, as will be discussed below.

During water flux across the membrane, solutes in the feed solution are dragged towards the membrane interface, where they are subsequently rejected. This causes an increase is solute concentration at the membrane surface on the feed side of the membrane. This phenomenon is called "concentration polarization" (CP) and is well understood and characterized in pressure driven membrane processes.

In osmotically driven membrane processes, the same phenomenon will occur on the feed side of the membrane. However, as water flux permeates the membrane, the pure water will dilute the draw solution at the membrane interface on the permeate side (downstream), reducing the concentration of the draw solution at this interface. The osmotic pressure gradient, reasonably synonymous with concentration gradient, is established only over this selective membrane portion and thus one must consider the actual osmotic pressure at the interfaces, not in the bulk. Figure 1 illustrates this phenomenon by showing the osmotic pressure profile across the membrane. The bulk draw solution (on the left of the figure) has a higher osmotic pressure than the bulk feed solution. Near the membrane interfaces (both feed and permeate sides), the effective osmotic pressures are induced by concentration polarization phenomena (concentrative on the feed side, dilutive on the permeate side).

On the feed side of the membrane, convection will continue to drag solute towards the membrane surface while diffusion will cause the motion of the solute back in to the bulk due to the concentration gradient. On the draw side of the membrane, convection will drag solute away from the permeate interface, inducing diffusion from the more concentrate bulk solution toward the membrane. These phenomena are also illustrated in Figure 1. Improving mass transfer on either side of the membrane, through increased turbulence or crossflow velocity, will reduce the severity of these phenomena. Similar methods are often employed in pressure driven membrane processes to limit the effects of concentration polarization. Figure 1 illustrates an ideal, symmetric, homogeneous, dense membrane. Most of today's membranes, however, are asymmetric and have a thin selective layer (also known as the active layer) supported by one or more porous supporting layers. These porous supporting layers provide mechanical strength to the fragile active layer, sometimes only a few hundred nanometers thick, when the membrane is under pressure. In pressure driven membrane processes, this support layer plays little role in membrane flux performance since the vast majority of the hydraulic resistance is in the active layer. When the active layer is facing the feed solution, as is typical of pressure driven membrane filtration processes, and is referred to as the "FO mode" in the present method, water permeating the selective layer simply percolates through the porous layers and out of the membrane to the permeate side.

In osmotic processes for membranes oriented in the same direction (active layer against the feed solution), the draw solution must diffuse into the support layer in order to establish the osmotic pressure gradient across the active layer. This is due to the fact that the porous layers (which have pores in the range of tens of nanometers to tens of microns) do not reject salt to any appreciable degree and therefore no osmotic pressure gradient is established. As water permeates the membrane, the solutes within the porous layer are diluted, creating an internal concentration polarization effect. As illustrated in Figure 2, solute from the bulk draw solution diffuses into the support layer to compensate for the dilution, but the diffusion is hindered by the porous layer. Increasing crossflow velocity or turbulence does not mitigate this phenomenon because these system parameters do no improve mass transfer within the protective confines of the support layer.

A similar but concentrative CP phenomenon will occur when the support layer is facing the feed solution, as depicted in Figure 3. This orientation of the membrane is typically used in PRO processes since the permeate stream is pressurized and the support layer is typically oriented away from hydraulic pressure. In this 'PRO mode', solutes from the feed solution are dragged into the support layer by convection. Here, the solutes build up in concentration as the selective layer rejects them. The solutes will attempt to diffuse back into the bulk feed solution but the support layer hinders this back diffusion and also negates the effects of turbulence and crossflow which occur on the external surface of the membrane.

These internal concentration polarization phenomena significantly reduce the effective osmotic driving force across the membrane, thereby reducing permeate water flux.

Current generation membranes, which have thick porous supporting layers, are not ideal for use in osmotically driven processes and thus further development of these types of processes will depend on the design of a membrane that is tailored for use in such osmotically driven processes.

Previous membrane design has focused on creating highly selective and permeable membranes for use in pressure-driven membrane processes. These current generation membranes are designed with highly permeable "active" layers which have high selectivity for water. Though somewhat resistant to water flow, the active layer's relative thinness (i.e., less than 1 μm for desalination membranes) makes them highly permeable, but their fragility requires them to be supported by one or more porous supporting layers. These porous layers provide integrity to the membrane for high pressure applications while not significantly increasing the overall hydraulic resistance.

However, the presence of these porous layers has led to poor performance when used in osmotically driven membrane processes, most notably pressure retarded osmosis

(PRO) and forward osmosis (FO). These layers cause internal concentration polarization which reduces the effective osmotic driving force. Internal concentration polarization has been accurately modeled and found to be related to the structure of the support layer, most notably its thickness and porosity. Furthermore, no studies have determined the importance of support layer chemical properties, such as hydrophobicity, on osmotic flux performance.

Support layer hydrophobicity plays little role in pressure-driven membrane processes. Water permeates the active layer by a solution-diffusion mechanism and then simply percolates through the pores and cavities within the support layer. Thus, the support layer does not need to fully wet in order to ensure adequate permeate water flux.

However, when these asymmetric membranes are used in osmotically driven membrane processes, the support layer must fully wet to ensure effective water transport. If the support layer does not wet, vapor or air trapped in the pores not only blocks the passage of water, but it may also exacerbate internal concentration polarization by reducing the continuity of the water within the layer, and thereby reducing the effective porosity. The inventors of the present invention have investigated the effects of membrane support layer wetting on water flux in osmotically driven membrane processes. Using various techniques, relative hydrophobicities of various support layer polymers were compared and correlated to water flux performance. The inventors of the present invention determined that support layer hydrophobic ity hindered osmotic flux through asymmetric membranes designed for pressure driven membrane processes. It was further demonstrated that improving the wetting of the membrane support layer results in a significant increase in water flux for osmotically driven membrane processes.

SUMMARY OF THE INVENTION It is an object of the present invention to provide improved membranes that are designed for use in osmotically driven membrane processes, including forward osmotic (FO) and pressure retarded osmotic (PRO) processes.

It is an object of the present invention to provide composite membranes for use in osmotically driven membrane processes with support layers exhibiting at least one of high porosity, low tortuosity, and minimized thickness.

It is another object of the present invention to improve the wettability of the support layer in membranes designed for use in osmotically driven membrane processes.

It is another object of the present invention to suggest the usage of selected polymers, polymer blends, block co-polymers, or additives to increase the hydrophilicity of the support structure.

It is still another object of the present invention to suggest methods of making the above said membranes using phase inversion casting of hydrophilic polymers to create porous supports for membranes usable in osmotically driven membrane processes. It is still another object of the present invention to suggest methods for designing membrane substrates using novel nonwoven polymer nanofiber webs to vastly increase porosity and interconnectivity while reducing the thickness of the porous support layer.

It is another object of the present invention to improve the wettability and structure of the support layer in membranes designed for use in osmotically driven membrane processes without sacrificing chemical compatibility of the membrane with regards to the feed or draw solutions.

It is another object of the present invention to use nonwoven polyester support layers, when deemed necessary, of decreased thickness and tortuosity combined with improved wettability.

BRIEF DESCRIPTION OF THE FIGURES

For a fuller understanding of the invention, reference is made to the following description taken in connection with the accompanying figures, in which: Figure 1 depicts an osmotic pressure profile in a forward osmosis process across a dense, symmetric membrane.

Figure 2 depicts an osmotic pressure profile in a forward osmosis process across an asymmetric membrane oriented in the forward osmosis mode and illustrates dilutive internal concentration polarization. Figure 3 depicts an osmotic pressure profile in a forward osmosis process across an asymmetric membrane oriented in the pressure retarded osmosis mode and illustrates concentrative internal concentration polarization.

Figure 4 depicts cross-sectional scanning electron microscope (SEM) images for three different membranes: (a) Osmonics CE; (b) Filmtec SW30 XE; and (c) Hydration Technologies CA membranes. The bar in each SEM image represents 100 μm.

Figure 5 depicts cross-sectional SEM images for two different RO membranes with the fabric layer removed: (a) Filmtec SW30 XE; and (b) Osmonics CE. The bar in each SEM image represents 100 μm.

Figure 6 depicts a top view of the polyethylene terephthalate (PET) fabric layer (right) and the porous cellulosic support layer (left) of the Osmonics CE membrane. The porous material between the two layers is carbon tape. Figure 7 depicts steady state forward osmosis flux measurements for the Filmtec SW30 XLE membrane averaged over at least 1 hour.

Figure 8 depicts s teady state forward osmosis flux measurements for the CE membrane averaged over at least 1 hour. Figure 9 depicts forward osmosis flux measurements for the CE membrane with the fabric layer removed compared to flux measurements for the CA membrane over a range of NaCl draw solution concentrations.

Figure 10 shows a diagram depicting the osmotic pressure profile across a membrane oriented in the PRO mode (draw solution on the active layer). Figure 11 depicts water flux data for the SW30 XLE membrane under 3 different conditions.

Figure 12 depicts water flux data for the SW30 XLE with the fabric layer removed with and without RO pretreatment.

Figure 13 depicts water flux data for the CE membrane under different conditions. Figure 14 depicts forward osmosis flux measurements for the CE membrane with the fabric layer removed compared to flux measurements for the CA membrane over a range of NaCl draw solution concentrations.

Figure 15 depicts forward osmosis flux measurements for the CE and SW30 XLE membrane. Also, while not all elements are labeled in each figure, all elements with the same reference number indicate similar or identical parts.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The present invention relates generally to improved semi-permeable membranes for use in osmotically driven membrane processes including forward osmosis and pressure retarded osmosis processes. The improvements of these new membranes over existing current generation reverse osmosis membranes leads to a reduced severity of internal concentration polarization and an improved wettability by using polymer materials of greater hydrophilicity. In one embodiment, the present invention relates generally to a semi-permeable membrane for use in a forward osmosis (FO) process, wherein said semi-permeable membrane is used to separate a draw solution and a feed solution, the semi-permeable membrane comprising:

{W159767₂} a) an active layer having a high degree of selectivity for water and a very low resistance to water transport; b) a hydrophilic support layer exhibiting at least one of high porosity, low tortuosity and minimal thickness; and c) optionally, a backing layer, wherein the support layer is chemically resistant to conditions presented by the draw solution (i.e., basic, acidic or oxidative environments) and the active layer is chemically resistant to conditions presented by the feed solution (such as fouling, cleaning agents, etc.).

In another embodiment, the present invention relates generally to a semi-permeable membrane for use in a pressure retarded osmosis (PRO) process, the semi-permeable membrane comprising: a) an active layer having a high degree of selectivity for water and a very low resistance to water transport; b) a hydrophilic support layer exhibiting at least one of high porosity, low tortuosity and minimal thickness; and c) optionally, but preferably, a backing layer; wherein the semi-permeable membrane is capable of withstanding pressures generated during the PRO process.

Reducing the severity of internal concentration polarization can be accomplished by changing the characteristics of the support layer. The ideal membrane would have no support layer, but this is impractical because the membrane must have some structural strength. Since a support layer of some kind will nearly always be present, the following issues must be taken into consideration when designing the support layer in membranes for osmotically driven membrane processes:

1) Minimizing the thickness of the support layer - this will reduce the distance that solute must diffuse to get into/out of the support layer;

2) Maximizing the porosity and pore connectivity of the support layer - this will reduce the diffusional resistance within the support layer and increase the effective diffusion coefficient; and 3) Minimizing the tortuosity of the support layer - this will reduce the diffusional resistance within the support layer and increase the effective diffusion coefficient.

While internal CP has been found to have a significant impact on flux performance, other phenomena play a role as well. Water interactions with the support layer affect the wetting of the support layer. For instance, support layers which are extremely hydrophobic do not wet readily when passively exposed to water. In pressure driven membrane processes, the wettability, or hydrophobicity, of the support layer is relatively unimportant, as the permeate water simply drains through the hydrophobic porous layer. In osmotically driven processes however, these process layers are passively exposed to water and must wet spontaneously. Typical support layers of salt rejecting membranes are made of polyester non-woven fabrics and porous polysulfone membranes. These polymers are somewhat hydrophobic and thus do not wet spontaneously. If the support layer does not wet out, then water cannot travel in any appreciable amount across the membrane in the absence of a hydraulic driving force. If the membrane only partially wets, then the internal concentration phenomena described above are enhanced in their severity as pathways for diffusion of solutes are limited by the presence of air or vapor in the support layer. The inventors of the present invention have determined that in order to develop membranes tailored for osmotically driven membrane processes, the membrane support layers must have an increased degree of hydrophilicity throughout the entire layer so that they will wet spontaneously in the presence of aqueous solutions.

To that end, the inventors of the present invention have determined that there are several polymers that can be considered for the membrane support layer. In addition, the type of polymer used for an actual membrane will greatly depend on the method of making the support layer. In a preferred embodiment, the membrane support must be made with a polymer that has the following properties:

1) Hydrophilicity;

2) Castable/formable in very thin layers but still structurally sound;

3) During casting/formation the film is highly porous and non-tortuous; 4) Once set, the polymer can withstand the chemical environment around the membrane during operation and cleaning; and

5) Once set, the polymer can withstand the hydraulic pressure of PRO processes, and the handling stress associated with module construction for either PRO or FO.

Polymers that may be suitable for use in membranes in accordance with the present invention include polyacrylonitrile; polyamide; poly (vinyl alcohol) after extensive cross linking; sulfonated polysulfone and sulfonated polyethersulfone (as described for example in U.S. publication No. 2007/0163951 to McGrath, the subject matter of which is herein incorporated by reference in its entirety); standard polysulfone/polyethersulfone polymers after treatment in oxidative environment (plasma or peroxide treatment); polysulfone/polyethersulfone polymers with hydrophilic additives (such as surfactants or hydrophilic nanoparticles); given by way of example and not limitation. For use with an ammonia-carbon dioxide osmotic process, the material used will depend on the process. For ammonia-carbon dioxide forward osmosis processes, the membrane support layer is subject to the ammonium salt solution and must be therefore very resistant to alkaline environments. For the ammonia-carbon dioxide osmotic heat engine (i.e., PRO mode), the selective layer is exposed to concentrated alkaline environment, so its resistance to such an environment is a prime concern. Fortunately, today's selective interfaces in membrane applications are very chemically resistant to high salinity alkaline environments. Still, support layers with minimum hydrolytic capabilities should be considered in all cases. For these reasons, the following polymers are preferred for membranes used with the ammonia-carbon dioxide forward osmosis and osmotic heat engines: polyamides, sulfonated polysulfone and sulfonated polyethersulfone (as described for example in U.S. publication No. 2007/0163951 to McGrath); standard polysulfone/polyethersulfone polymers after treatment in oxidative environment (plasma or peroxide treatment); polysulfone/polyethersulfone polymers with hydrophilic additives (such as surfactants or hydrophilic nanoparticles); given by way of example and not limitation.

These support layers can be cast by phase inversion as is typically done in fabricating pressure driven membranes. In addition, electrospun nanofiber supports may also be usable in the practice of the invention. Synthesis techniques are critical to the use of some of these specific polymers as osmotic membrane support layers as not all may be castable by phase inversion to yield highly porous and stable structures.

In one embodiment, membranes are designed for use in either FO or PRO processes. The support layers composed of hydrophilic polymers cast into porous films onto nonwoven hydrophilic polyester supports. These two-tiered support structures would then be coated with a highly selective active layer (typically polyamide based cast through interfacial polymerization).

The general process is as follows:

Using a polyester nonwoven support layer, with a thickness of about 50 to about 125 micrometers, after it was modified to be more hydrophilic as a substrate, a polymer solution containing a hydrophilic polymer is cast by phase inversion to a thickness of between about 30 and about 70 micrometers. In one embodiment, a thickness of about 50 micrometers is used for the cast polymer support layer. An interfacially polymerized polyamide layer is then cast on top of the porous polymer film to a thickness of between about 50 and about 600 nanometers. In one embodiment the polyamide layer is cast to a thickness of about 200-400 nanometers. The method is similar to the making of a reverse osmosis membrane, but yields a selective membrane with a relatively thick, albeit hydrophilic support layer, which would be useful in certain configurations of the PRO process, most notably the ammonia-carbon dioxide osmotic heat engine with deionized water as a working fluid.

In another embodiment, the present invention relates to a membrane that is suitable for use in an osmotically driven process that incorporates a porous layer formed by electrospun nanofibers instead of the porous polymer support layers used previously. While nanofiber electrospinning is generally well known, it has not been used previously to provide a support layer for a semi-permeable membrane usable in an osmotically driven process.

Typical water filtration membranes are based on a composite structure which includes a nonwoven polyester-like support (100-125 micrometers thick), a thinner porous polymer support structure which is cast upon this nonwoven film (about 50 micrometers thick), and a very thin selective barrier layer which conducts the actual separation (perhaps 200-400 nanometers thick). The relatively thick support layers support the thin and relatively fragile selective barrier while operating under pressure. As discussed previously, in order to develop a semi-permeable membrane that is usable in osmotically driven processes such as forward osmosis and pressure retarded osmosis, there is a need for ultra thin, highly porous, hydrophilic support layers beneath the thin selective barrier. This would include optimization of both the thick polyester (PET) nonwoven and the polymer layer. The inventors have discovered that the use of nanofibers in place of the polymer layer would help achieve this with use of a thinner polyester support layer and three general procedures were considered:

1) General procedure for making nanofiber supported membranes for PRO applications:

An ultra-thin, porous polyester nonwoven support film is used as a substrate, and polymer nanofibers that are electrospun into a nonwoven mat are deposited onto the polyester nonwoven film. The polyester support film is custom designed to have a highly porous structure, be more hydrophilic, and be as thin as 80 micrometers for membranes intended for PRO applications (though thicker is reasonable for higher pressure applications). The electrospun nanofiber mat will typically have a thickness of about 20 micrometers but can be thicker for higher pressure applications. A thin barrier layer is cast upon this nanofiber mat which must be compatible with the draw solution chemistry. This three tiered structure (polyester nonwoven, nanofibers, selective layer) is capable of providing a good result in PRO applications.

2) First procedure for making nanofiber supported membrane for FO applications: An ultra-thin, porous polyester nonwoven support film is used as a substrate, and polymer nanofibers that are electrospun into a nonwoven mat are deposited onto the polyester nonwoven film. For membranes to be used in FO applications (with the draw solution against the supporting layers), the membrane must be made as thin as possible. Standard polyester supports used in RO membrane fabrication are not viable. Ultra-thin, highly porous, and hydrophilic polyester nonwovens (as thin as 50 micrometers) are used to support the electrospun fiber mat (about 20 micrometers thick). After deposition onto the polyester, a thin active layer is cast upon this nanofiber mat which must be compatible with the feed solution chemistry (perhaps 200-400 nm thick). This active layer is typically a polyamide based polymer layer cast by interfacial polymerization as is typical with salt rejecting RO membranes. Both the nanofiber and polyester backing layer must have chemical resistance to the various draw solutions that are of significant alkalinity, acidity, or oxidative potential.

3) Second procedure for making nanofiber supported membranes for FO applications:

Electrospun nanofiber mats are themselves quite strong for a given thickness. These mats may serve to provide all of the necessary support to the selective layer during operation since there is no hydraulic pressure in FO. The nanofiber mats are spun onto a substrate to which there is little adhesion (for example, a woven netting or a Teflon sheet). After deposition onto the substrate, a thin active layer is cast upon this nanofiber mat which must be compatible with the feed solution chemistry (perhaps 200-400 nm). This active layer is typically a polyamide based polymer layer cast by interfacial polymerization as is typical with salt rejecting RO membranes. Once the active layer is set, the nanofibers are removed from the substrates, resulting in a nanofiber supported membranes. For membranes supported entirely by nanofibers, the thickness of the nanofiber mat may be at least 30 micrometers. The use of additives to create nanocomposite nanofibers with greater mechanical strength than the polymer alone may reduce the required thickness. The nanofiber backing layer must have chemical resistance to the various draw solutions that are of significant alkalinity, acidity, or oxidative potential.

In all of these procedures mentioned, these nanofiber webs can create supporting structures in membranes with incredibly high porosity (in excess of 80%) and pore interconnectivity (tortuosity approaching 1). Because of the high degree of fiber interconnectivity, the mechanical strength of these webs exceeds that of typical porous polymer films, allowing them to be made thinner, typically in the range of about 20 micrometers. As discussed previously, porosity, pore interconnectivity, and thinness are all critical aspects of support structures for FO and PRO membranes.

Hydrophilicity of the nanofiber web can be tailored by selecting polymers or miscible polymer blends of polymers that are hydrophilic or by chemically treating hydrophobic polymers. Hydrophilic polymers, by way of example and not limitation, that can be used for phase inversion casting or electrospinning nanofiber webs include polyacrylonitrile, crosslinked polyvinyl alcohol, sulfonated polysulfone, sulfonated polyethersulfone, block copolymers that incorporate hydrophilic functional groups such as hydroxyl, carbonyl, carboxyl, amino, sulfhydryl, and phosphate groups, and combinations of one or more of the foregoing. To increase the hydrophilicity of hydrophobic polymers, such as polysulfone and polyethersulfone, oxidative treatment with plasma or chemicals (e.g., peroxide, acid) may be used to increase hydrophilicity. In addition, incorporation of surfactant, such as sodium dodecyl sulfate, or hydrophilic nanomaterials, such as hydrophilic inorganic nanoparticles, into the electrospun fiber or polymer film is also an option.

Thus, it can be seen that for membranes to perform well over long periods of time, they must be compatible with their environment and must be designed with consideration to process specifications. Thermal and chemical conditions need to be optimized for enhanced performance of various osmotically driven membrane processes and the membranes must be able to withstand these conditions. The membranes must be compatible with both the feed and the draw solution and not degrade or be physically or chemically altered in the presence of either the feed or the draw solution. Current generation reverse osmosis membranes consisting of polyamide active layers, for example, are highly resistant to acidic and basic conditions and high temperatures yet will degrade in the presence of chlorine. Cellulosic membranes are chlorine resistant, yet degrade outside of ambient pH ranges due to hydrolysis reactions and will also break down at higher temperatures. This particular aspect of compatibility is critical to draw solutions such as ammonia-carbon dioxide draw solutions.

Current generation membranes which reject salt (reverse osmosis) are highly permeable and have a high degree of salt rejection. For membranes to work in osmotically driven membrane processes, they must also have a high degree of selectivity for water and a very low resistance to water transport (i.e., thin selective layer). Therefore, it is expected that membranes can be designed with different structural and chemical properties and would still have a high degree of permeability and selectivity. Existing reverse osmosis membrane technology can therefore be linked to the improved membrane structural and chemical design characteristics to produce novel forward osmosis membranes.

In another embodiment, the present invention relates generally to a method of increasing the degree of saturation of a semi-permeable membrane for use in an osmotically driven process, wherein the semi-permeable membrane comprises a backing layer, a porous support layer, and an active layer having a high degree of selectivity for water and a very low resistance to water transport, the method comprising the step treating the semi-permeable membrane to increase its degree of water saturation by at least one of:

a) pretreating the membrane in a reverse osmosis mode for a period of time and at a suitable pressure and flux to purge air out of the porous support layer; and

b) degassing the feed solution to remove dissolved gas and air bubbles prior to introducing the feed solution to the semi-permeable membrane.

Data demonstrating importance of having a hydrophilic layer in osmotically driven processes:

A. Materials and Methods

1) Forward osmosis crossflow setup A custom built crossflow filtration cell was used and the flux into the draw solution was measured gravimetrically with a scale. Additional modifications and various experimental protocols are described in further detail below.

2) Membrane selection Three commercially available membranes were reviewed. The cross sections of each membrane were imaged using SEM. The first, shown in Figure 4(a), is a cellulosic brackish water reverse osmosis (RO) membrane, available from GE Osmonics (Fairfield, CT) (designated CE). The next, shown in Figure 4(b), is a thin film composite RO membrane, available from Dow Filmtec (Midland, MI) used in seawater desalination (designated SW30 XLE). Finally, the membrane shown in Figure 4(c) is a membrane available from Hydration Technologies, Inc. (Albany, OR) (designated CA). This membrane is specifically tailored for forward osmosis (FO) water purification. In each SEM micrograph, the various layers of the membranes are indicated. These layers are discussed in greater detail below. 3) Removal of fabric backing layers

The CE and SW30 XLE membranes are both fabricated using a polyester (PET) non-woven fabric support. This fabric is used as a substrate to cast the cellulosic CE membrane and the SW30 XLE polysulfone (PS) support layer by phase inversion. In order to study the hydrophobicity of the support layer, the fabric layers were carefully removed by peeling them away from the other layer (or layers) of each membrane. SEM micrographs in Figure 5(a) and (b) show the cross sections of the CE and SW30 XLE membrane, respectively, with their fabric layers removed. The membrane selective layers are facing up with the support layers facing down. Note that there is significant roughness on the support layer side on both membranes, which remains due to the strong binding between the two layers at the interfacial region between the polymer and the PET non- woven fabric. Even with the removal of the fabric layer, the structural integrity of both membranes was observed to be fairly well intact.

B. Analysis

The effects of support layer hydrophobicity on water flux are described by considering contact angle analysis of the various membrane layers, membrane orientation, membrane pre-wetting, PET fabric removal, feed solution gas content, and the presence of surfactants. Flux data were taken in both the FO and PRO modes for the SW30 XLE and

CE membranes.

1) Contact angle measurement and analysis A traditional method of determining a material's hydrophobicity is through water contact angle analysis and each support layer of the membranes described herein was tested to determine its relative hydrophobicity. Contact angles of the membrane active layers were also measured for comparison. Temperature during the measurements was 21-22°C. Results are summarized in Table 1. Water contact angle measurements were accomplished with the sessile drop method using a VCA video contact angle system from AST products (Billerica, MA). Water contact angles were taken of the membrane active layers, the polyester non-woven fabric layers, and the backing layers of the membrane after removal of the fabric. Water drop size was 1 μL. Room temperature was maintained at 21-22°C during the measurements. Roughness of the surfaces was not taken into account. To account for significant variations between different measurements on the same substrate (caused by surface roughness and/or chemical heterogeneities), five locations on two independent samples were tested. A computer program was used (VCAOptimaXE) to determine the contact angle on both sides of a sessile drop and the results were averaged for all measurements.

Table 1: Measured contact angles of the various layers of the membranes used.

Contact Standard

Membrane Layer ^An9^|e Deviation

(degrees) (degrees)

CA active 62.0 7.2

CA backing 63.6 13.0

SW30 XLE active 37.5 2.8

SW30 XLE fabric 67.5 5.0

SW30 XLE polysulfone 95.2 4.9

CE active 59.1 4.0

CE fabric 61.8 8.0

CE porous support 19.4 5.1

Included in this table are results from the CA membrane (active and backing layers), the SW30 XLE membrane (active, PET fabric, and PS support after the fabric had been removed), and CE membrane (active layer, PET fabric, and cellulosic support after the fabric had been removed). The measurements of these contact angles can help qualitatively assess each layer's wettability, with higher angles indicating greater hydrophobicity. While no strict definition relating contact angle to subjective terms like hydrophobic and hydrophilic exists, generally contact angles near or above 90° indicate significant hydrophobicity while lower contact angles indicate increasing hydrophilicity. Multiple locations on two independent samples were tested in order to average out errors that occur due to roughness and/or chemical heterogeneity.

The first membrane listed in Table 1 is the CA membrane, which has no fabric support, although it does have a self-supporting porous structure. This membrane, which has been studied in previous investigations on forward osmosis, shows a moderate degree of hydrophilicity (62.0° for active layer, 63.6° for backing layer). This is expected due to its cellulosic composition in which the cellulose polymer backbone contains hydrophilic hydroxyl groups. The similarity between the active and backing layer is not surprising, since the membrane is made through the phase inversion process and the polymer material is the same throughout the membrane. The standard deviation of the measurement is reasonably high given the macroscale roughness caused by the embedded polyester mesh. Roughness increases the variability in the contact angle measurement due to pinning and other surface effects which create metastability within the sessile droplet. This also explains the increased standard deviation of the measurements on the backing layer, which has an increased roughness due to surface porosity.

The SW30 XLE membrane is an RO thin film composite membrane with a polyamide active layer supported by a polysulfone (PS) support layer. The SW30 XLE porous support is a polysulfone UF membrane, which is cast upon a PET non-woven fabric layer. Each of these layers will play a role in water transport across the membrane during osmotically driven membrane processes, so each needs to be characterized for hydrophobicity. The active layer is hydrophilic, which is expected since the water must diffuse through this part of the membrane and increased hydrophilicity decreases the active layer fouling propensity. The support layers are more hydrophobic. The PET fabric is moderately hydrophilic (67.5°) and the PS support is hydrophobic (95.2°).

The CE membrane is comprised of an asymmetric cellulosic layer which has an active layer and a supporting porous structure. This self-supported cellulosic layer is cast upon a PET fabric support. The active layer, or the dense portion of the cellulosic layer, has nearly the same hydrophilicity as the CA membrane (59.1°). This is not surprising considering both membranes are cellulosic. The lower standard deviation is due to the smoothness of the CE active layer. The non-woven PET fabric layer has a moderate degree of hydrophilicity (61.8°), similar to that of the SW30 XLE PET fabric layer. The cellulosic support of the CE membrane was measured as being very hydrophilic (19.4°), though this is unexpected due to the fact that the material is the same as the active layer. This low contact angle measurement is likely due to the rough surface morphology, which is formed by contact with the fabric layer during membrane casting. When the cellulosic membranes are cast onto the PET fabric, the fabric fibers will imprint on the surface, leaving a mold of the fibers in the polymer, as shown in Figure 6. This roughness, which can also be seen in Figure 5(b), allows for a drop of fixed volume to spread by capillary action into the crevices on the surface. The spreading is facilitated by the polymer's hydrophilicity and improves the capillary movement of water not only into these crevices, but also into the pores within the membrane. Hence, this will give the appearance of a lower contact angle. A similar effect was not noticed on the SW30 XLE PS support layer because the material is intrinsically hydrophobic, so the presence of roughness would not necessarily facilitate capillary motion of water. In fact, on many rough, hydrophobic surfaces, water will not rest entirely on the solid surface, but will also be supported by air trapped within the crevices on the surface.

2) Flux measurement and analysis a) Experimental protocol Temperature and stability equilibration:

Water flux measurements were taken using a gravimetric method. The draw solution reservoir mass was constantly monitored on a scale which outputs to a computer. In this particular system configuration, the draw solution was run in closed loop. Therefore, trapped air in the system or vibrations may cause erroneous flux measurements. To circumvent this problem, deionized water was circulated on both sides of the membrane until the baseline "flux" was measured at zero. During this equilibration procedure, both deionized water reservoirs were also circulated through inline heat exchangers which sat in a constant temperature water bath. The solutions were circulated in a closed loop mode until they reached a stable temperature of 20⁰C. After the draw solution reservoir mass and the temperature had stabilized, an appropriate amount of 5 M NaCl stock solution was added to bring one of the solutions, in this case the draw solution, up to a concentration of 1.5 M.

Reverse osmosis pretreatment: In some of the experiments, the membrane was "pretreated" in a reverse osmosis mode prior to running a flux test in the forward osmosis cell. This pretreatment under hydraulic pressure was used to purge air and/or vapor out of the support layers prior to forward osmosis. It was thought that moderately high hydraulic pressures might induce liquid water to displace air or vapor within the porous support layers, partially or perhaps fully wetting the membrane. An RO crossflow unit with identical channel dimensions to the FO crossflow unit was used for this pretreatment. Deionized feedwater was used at a

{Wl 597672} hydraulic pressure of 450 psi (31.0 bar) with a temperature between 25 and 27°C. Tighter temperature control was difficult due to the heat generated by the high pressure pump. The membrane active layer was facing a deionized water feed solution as the support layers provided mechanical strength to the membrane while under pressure. Pressure and flux were maintained for one hour. It was assumed that this amount of time would be suitable to achieve a steady state degree of wetting, and additional pretreatment times were not considered as part of this study. After this time, the membrane was removed, submerged quickly in water to avoid any degree of drying or drainage, and transferred to the forward osmosis cell.

Feed solution degassing:

Next, feed solution degassing was done to eliminate the possibility of air bubbles moving into the support layer from the feed during FO tests. Degassing can also improve the wetting of the support layer, as air or vapor trapped in the support layer might be induced to diffuse out into the degassed bulk solution. This treatment was only done with the feed solution when the membrane was oriented with the feed against the support layer (i.e., in the PRO mode).

A hollow fiber membrane contactor, with polypropylene fibers, was used for the degassing (Celgard MiniModule®). This contactor was placed inline with the recirculating feed solution. The hydrophobic polypropylene hollow fibers are typically used in membrane vacuum distillation processes for degassing and this particular module is used for degassing small scale deionized water systems. The feed solution (deionized water) flowed through the lumen of the fibers while a vacuum was drawn on the shell side. A vacuum of at least 29 inches Hg was drawn by a direct drive vacuum pump (Edwards, Murray Hill, NJ). The feed solution was run in closed loop and dissolved oxygen levels were measured with a dissolved oxygen probe (Fisher Scientific, Waltham, MA). After 1 hour of degassing, dissolved oxygen levels were below 0.1 mg/L and steady state had been achieved. Using oxygen as a surrogate for all dissolved gases, we assumed that the other gases in solution had likewise been removed to a significant extent. Degassing was maintained throughout the duration of the experiment.

Surfactant additions to feed solution To observe how changes in wettability of the support layer of the various membranes affect water flux, sodium dodecyl sulfate (SDS) was added to the deionized feed solutions which were against the membrane support layers (PRO mode). The membranes tested in this fashion were the CE and the SW30 XLE membranes. The SDS was added to the feed solution after the equilibration procedures described earlier and after the addition of NaCl to the draw solution. Steady state flux data was taken for 50 minutes followed by addition of an appropriate amount of 100 mM SDS stock solution to bring the feed solution concentration to 1 mM SDS.

b) Water flux in FO mode

For water flux measurements, the first membrane orientation that was considered is the FO mode. In this orientation, the membrane active layer is facing the feed solution and the support layer is facing the draw solution. This is the typical orientation for most membrane separation processes, such as reverse osmosis. However, in the tests described herein, flux is osmotically driven and no hydraulic pressure is used.

i) SW30 XLE membrane

The first tests were conducted with the SW30 XLE membranes. This membrane is typically used for seawater desalination and therefore has a very high salt rejection and thick porous support layers intended to withstand the high hydraulic pressures of seawater RO (up to 1200 psi or 82.7 bars). Osmotically driven water flux was measured using a deionized water feed and a 1.5 M NaCl draw solution, corresponding to a transmembrane osmotic pressure of about 70 atm (1029 psi, 70.9 bar), after various pretreatments to the membrane. Conductivity of the deionized feedwater was monitored to measure salt transport from the draw solution to the feed. For all tests, salt transport from the draw solution was found to be small and did not appreciably impact the transmembrane osmotic pressure difference.

Figure 7 presents the averaged steady state water flux data. The water flux through this membrane is very low when operated in the FO mode: less than 0.5 gallons/ft² membrane area/day (gfd).

The RO pretreatment conditions were as follows: deionized feedwater against the active layer, 450 psi (31.0 bar) pressure, 25±2 ⁰C, run for at least 1 hour. The Forward

{W159767₂} osmosis experimental conditions were as follows: membranes oriented in the FO mode, with the feed (DI water) against the active layer and the draw solution (1.5 M NaCl) against the support layer. Crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^"2 h^"1. Note also that typical seawater RO systems operate between 9-11 gfd, and those are operated at lower transmembrane driving forces.

The PET fabric layer of the membrane could be removed and flux slightly increased. This is likely due to a decreased internal concentration polarization effect due to the decreased overall thickness of the support layer. When the membrane undergoes reverse osmosis pretreatment, the flux increases for both the full membrane and the membrane with no fabric, though the effect is far more noticeable for the membrane with the fabric removed. Pretreating the membrane with hydraulic pressure (i.e., reverse osmosis mode), removes some of the air trapped in either the polyester or polysulfone support layer. This air is effectively impassible by water and its presence reduces the effective porosity of the support layer. By purging the air from the layer, a greater degree of water continuity can exist within the support layer, reducing the prevalence of internal CP by increasing the number of solute diffusion pathways while simultaneously increasing the available pathways for water transport.

It is interesting to note that the RO pretreatment seems to have more impact on the polysulfone (PS) support layer. This may be attributed to the higher degree of hydrophobicity of this layer relative to the PET layer (a contact angle of 95.2° vs. 67.5°, respectively). This layer will likely have a higher percentage of its pores filled with air or vapor prior to RO pretreatment and therefore plays a more significant role in hindering water transport through the membrane. Also, since the relative porosity of the PS layer is low compared to the PET fabric, continuity of water in the PS support layer will also be more of an issue, and lack of complete wetting will exacerbate this problem. Therefore, pre-wetting the membrane by purging the air/vapor from this layer improved water flux. When the fabric layer is still attached to the membrane the effect of pre-wetting is countered by the additional internal CP effects caused by the increased overall support layer thickness.

U) CE membrane Similar experiments were carried out with the CE membrane in the FO mode. Again, salt transport from the draw solution was found to be very small for all tests. The water flux performance data are shown in Figure 8. The RO pretreatment conditions were as follows: deionized feedwater against the active layer, 450 psi (31.0 bar) pressure, 25±2 ⁰C, run for at least 1 hour. Forward osmosis experimental conditions were as follows: membranes oriented in the FO mode, with the feed (DI water) against the active layer and the draw solution (1.5 M NaCl) against the support layer. Crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^"2 IT¹.Flux was measured through the full membrane before and after RO pretreatment. RO pretreatment had little effect on flux performance and, in fact, caused a slightly reduced flux. When the PET fabric is removed, the flux shows a dramatic improvement.

This water flux data suggests that the PET layer may play two roles in controlling osmotic flux across the membrane. First, by removing the PET support layer, internal CP is reduced due to the decreased support thickness, as occurred before with the S W30 XLE. However, the removal of the moderately hydrophilic PET support layer also exposes the more hydrophilic cellulosic portion of the membrane. The new interface is also highly indented which may improve the uptake of water into the support structure. This part of the membrane, shown in Figure 5(b), is asymmetric, but hydrophilic, so water easily permeates the pores. However, wetting of the PET support layer may not be an issue since no improved flux was seen after the RO pretreatment. The slight decrease in flux after RO pretreatment may be due to compaction of the PET fabric or the cellulosic porous support, causing a decreased porosity which may increase the hydraulic resistance within these layers or increase the severity of internal CP. After noting the exceptional flux performance of the CE membrane when the fabric layer was removed, the data set was expanded to include a range of draw solution concentrations up to 1.5 M NaCl. The data are shown in Figure 9 and are compared to the previously studied CA membrane in the same orientation. The membranes are oriented in the FO mode, with the feed (DI water) against the active layer and the draw solution (NaCl) against the support layer. Experimental conditions: crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^"2 h^"1. The similarity of the flux data is notable, being that the modified CE membrane is a commercial RO membrane and the CA is a commercial FO membrane. While the membrane support layers differ in thickness (slightly), porosity, and tortuosity, support layer characteristics which play a role in determining the severity of internal CP, as well as other structural characteristics (the embedded mesh support in the CA membrane), their materials are similar and both are cast through the phase inversion process. Thus, a commercial RO membrane was shown to compare favorably with commercial FO membranes after simple modification.

The above data suggest that hydrophilicity of the support layer is important when considering the mechanisms of resistance to osmotic flux through asymmetric membranes in the FO mode. Removal of the non-woven support layers resulted in higher flux for both the SW30 XLE and CE membranes, with the effect being much more pronounced with the

CE membranes once the relatively hydrophilic cellulosic membrane material was exposed to the draw solution. RO pretreatment was found to at least partially wet the PS structure of the SW30 XLE membrane, but its effect on the PET non-woven fabric was inconclusive for both membranes.

c) Water flux in the PRO mode

When the draw solution is placed against the active layer and the feed solution is against the support layer, the membrane is oriented in the PRO mode. This mode is typically used for PRO applications where the draw solution is pressurized, and hence, requires the mechanical support of the porous layer on the feed side of the membrane. For typical PRO applications, salinity in the feedwater will create concentrative internal concentration polarization effects. However, the flux tests described below were designed to solely test the impact of membrane hydrophobicity on water flux and therefore were conducted with a deionized water feed solution. In this case, no internal CP effects will occur. Note, however, that salt may leak across the membrane active layer from the draw solution and enter in the support layer. If this occurs, diffusion out of this layer is hindered by the porous structure and internal CP would form. This form of internal CP is illustrated in Figure 10. The feed (the right side) is deionized water. A dilutive external concentration polarization layer exists on the permeate side of the membrane. No internal concentration polarization exists in the support layer since the feed is deionized water. Small amounts of internal concentration polarization may occur due to salt leakage from the draw solution, but the amount of leakage is expected to be small due to the high salt rejection characteristics of the membrane.

For most high rejection RO membranes, however, it is reasonable to assume that very little salt traverses the membrane from the draw solution and any unexpected water flux behavior can be attributed to support layer chemistry rather than structure. This negligible salt transport was confirmed by conductivity measurements of the deionized feedwater during each test.

i) SW30 XLE membrane

The first tests in this mode were conducted with the SW30 XLE membrane. As above, a 1.5 M NaCl draw solution was used along with a deionized water feed solution, but the membrane was oriented in the PRO mode. The first test was done with the intact membrane. RO pretreatment conditions: deionized feedwater against the active layer, 450 psi (31.0 bar) pressure, 25±2 ⁰C, run for at least 1 hour. Forward osmosis experimental conditions: membranes oriented in the PRO mode, with the feed (DI water) against the support layer and the draw solution (1.5 M NaCl) against the active layer. Crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^"2 h^"1. As shown in Figure 11, water flux over time was very low (less than 1 gfd).

Membrane is oriented in the PRO mode. This was surprising considering the feed is deionized water and therefore no internal CP should be occurring. Previous investigations on water flux through commercial membranes (including the PA-300, NS-101, NS-200, BM-I-C and the Permasep B-IO hollow fiber) oriented in the PRO mode concluded that salt transport from the draw solution was the primary cause of low flux due to internal concentration polarization effects. However, with high rejection desalination membranes, such as the SW30 XLE, salt leakage from the draw solution should be minimal, even for higher concentration salt solutions. Salt leakage should be further mitigated by the water flux itself, which is in the opposite direction of the salt flux. Therefore, we would expect to see far greater fluxes given the high transmembrane osmotic pressure (70 atm)

Support layer hydrophobicity was therefore considered as a possible hindrance to osmotic flux across the membrane. If the membrane support does not wet passively when

{W159767₂} exposed to the deionized feedwater, then water transport is hindered through the support layer. Also, should any salt traverse the membrane from the draw solution, this salt would be less able to diffuse out of the support layer because of the lack of water continuity which reduces diffusion pathways. A series of experiments were conducted to verify this hypothesis.

If support layer wetting was causing reduced flux, then it would be expected that wetting the membrane would improve flux. Using the same RO pretreatment as described above, deionized water was forced through the membrane in order to displace the air or water vapor present in the support layer. After this pretreatment, the membrane was placed in the FO cell in the PRO mode. As shown in Figure 11, flux was initially higher, around 6 gfd, but fell slowly over time to around 2 gfd.

There are several possible reasons for the transient flux behavior shown in Figure 11. Convective flow of water may have transported air or vapor bubbles from the bulk solution into the support layer, slowly displacing water within the membrane causing a slowly decreasing water flux. To verify this, the feedwater was degassed before use. Degassing removes the dissolved gas and air bubbles from the bulk solution and has been shown to improve the wetting of hydrophobic materials by water. Another flux test was run after the membrane was pretreated with RO and after the feedwater was degassed. Degassing also took place throughout the experimental run. The data in Figure 11 indicates that degassing the feed solution does not improve flux, suggesting that air and vapor are not entering the support layer from the bulk feed solution and that some other mechanism is governing the degree of saturation in the support layer. It is likely that the wetting mechanism driven by RO and the drainage mechanism driven by FO are different, creating a 'pseudo' hysteresis of wetting behavior for this support layer. This behavior is not a true hysteresis, which would require the drainage and imbibition to be accomplished in the same direction and by similar mechanisms, but we use the term here for explanation as it is commonly used to discuss wetting and drainage mechanisms in porous media. Osmotically induced drainage may be occurring at such a rate that the local pressure within the pores drops, causing the spontaneous formation of water vapor. To further elucidate the wetting mechanism, the PET fabric layer was removed.

With both no treatment and RO pretreatment, the SW30 XLE membrane with the fabric layer removed was tested in the PRO mode and compared to the intact membrane as seen in Figure 12. The membrane is oriented in the PRO mode. Also shown is the membrane flux performance with the fabric layer (data from Figure 8, open squares). RO pretreatment conditions: Deionized feedwater against the active layer, 450 psi (31.0 bar) pressure, 25±2 °C, run for at least 1 hour. Forward osmosis experimental conditions: membranes oriented in the PRO mode, with the feed (DI water) against the support layer and the draw solution (1.5 M NaCl) against the active layer. Crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^~2 h^~\

The membrane performed similarly with and without the fabric support layer. However, after RO pretreatment, the flux was steady at nearly 5 gfd (about 6 times higher than before the pretreatment). What is interesting to note, though, is that the flux does not decrease with time as it did with both support layers present. This suggests that the wetting and drainage mechanisms are different for the PET and the PS support layers. It would appear that after partial wetting by RO, the PS support does not drain which would cause a reduction in flux. However, when the PET fabric is attached, the overall wetting and drainage mechanisms of the entire support layer change.

U) CE membrane

Identical tests were run with the CE membrane and the results are shown in Figure 13. The membrane is oriented in the PRO mode. The support fabric was removed for the indicated test. RO pretreatment conditions: deionized feedwater against the active layer,

450 psi (31.0 bar) pressure, 25±2 ⁰C, run for at least 1 hour. Forward osmosis experimental conditions: membranes oriented in the PRO mode, with the feed (DI water) against the support layer and the draw solution (1.5 M NaCl) against the active layer. Crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or

36.0 I m-² IT¹.

When testing the untreated membrane, water flux was initially high, around 6 gfd, but quickly dropped to about 2 gfd. When the membrane was pretreated with RO, the initial fluxes were higher (over 10 gfd), but a similar drop in flux occurred. Degassing the feed solution after RO pretreatment of the membrane resulted in no change in water flux behavior. These results were similar to the SW30 XLE membrane except for the fact that the initial flux of the untreated membrane was higher than the steady state flux. The steady state flux of the CE membrane also was higher than that of the SW30 XLE membrane.

The reasoning behind the transient water flux, as described above, is likely due to differing drainage and imbibition mechanisms of the PET fabric layer. The PET fabric is initially wet to some degree which fully wets the cellulosic membrane. Upon osmosis, water is drawn out of the cellulosic portion of the membrane which consequently drains the PET fabric. If water continuity is broken (i.e., drying) before water can enter the fabric, then the fabric may have difficulty rehydrating. If this is the case, the drainage mechanism by osmosis is faster than the wetting mechanism by water cohesiveness and capillarity. The same "hysteresis" apparently occurs in the SW30 XLE membrane as well and can now be attributed to the PET fabric layer. By observing this behavior in the CE membrane, we have found that this imbibition/drainage hysteresis likely occurs in the PET fabric layer for both the CE and SW30 XLE membranes. The differences between the initial and steady state flux of the CE and SW30 XLE membranes in the PRO mode after RO pretreatment still must be explained. For the SW30 XLE membrane, water must transport through the PET fabric followed by the PS support layer and then the polyamide active layer. In the CE membrane, water must pass through the PET fabric followed by the porous portion of the cellulosic layer until reaching the dense portion of this layer. As seen above, the hydrophobicity of both PET fabric layers is similar. The porous portion of the cellulosic membrane, however, is very hydrophilic and absorbent and therefore will spontaneously wet even when in contact with a partially hydrated PET layer regardless of RO pretreatment. This explains the higher steady state flux of the CE membrane when compared to the SW30 XLE membrane. In the CE membrane, a slightly higher initial water flux after RO pretreatment is likely due to a marginal increase in saturation of the PET layer, which improves water continuity for a short time at the beginning of a test (initial fluxes of over 10 gfd vs. 8 gfd without pretreatment). In the SW30 XLE membrane, however, the RO pretreatment will improve wetting of both layers. During a test, however, the PET layer drains while the PS layer does not (as indicated in Figure 12). Therefore, in the SW30 XLE membrane, the overall wettability of the intact support layer is limited by the PS support at first which is very likely less saturated than the more hydrophilic PET. As the PET fabric drains, flux is limited by that layer. Therefore, in the CE membrane, the limiting factor to water flux is the wettability of the fabric layer while in the SW30 XLE membrane, the limiting factor is the wettability of the PS support at first and then the PET fabric after it dries. In other words, the drier layer will control the overall wettability of the composite support layer, not necessarily the more hydrophobic layer.

As with the FO mode, upon removal of the polyester fabric from the CE membrane, the flux increases dramatically. Once the fabric layer is removed, the feed solution fully wets the self-supported cellulosic membrane and hence easily transports through the support layer structure. The slight decrease in flux over the course of the experiment is due to the dilution of the draw solution which occurred due to the high volume flux. As indicated above with the FO mode, since we saw excellent water flux performance with the CE membrane once its fabric layer had been removed, we expanded the data set to include lower concentrations of draw solution and compared the results to the performance of the CA membrane from a previous investigation. The results are shown in Figure 14. The membranes are oriented in the PRO mode, with the feed (DI water) against the support layer and the draw solution (NaCl) against the active layer. Experimental conditions: crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^~2 h^~ '. Again, the fluxes of the modified CE membrane were nearly equal to the commercial CA membrane for tests in the PRO mode.

c) SDS addition to the feed solution

The above data shows differing evidence of the impact of wettability of the polyester and polysulfone supports on water flux. With the SW30 XLE membrane, the wettability of the polysulfone has been shown to play a significant role in water flux for both the PRO and FO mode, but the PET fabric has little effect until the membrane is pretreated with RO. Conversely, the PET fabric in the CE membrane impacts flux far more dramatically in both cases. These previous tests only utilized RO to purge air and/or vapor from these various porous layers, but other experiments exist to further elucidate the effects of wetting on support layer water transport.

Sodium dodecyl sulfate (SDS), an anionic surfactant, can be added to the feedwater to improve the wetting of the support layer during PRO oriented tests. This was done for the membranes which experienced poor flux behavior (the SW30 XLE with and without PET fabric, and the CE membrane with PET fabric). The flux tests involved a brief equilibration time (50 minutes) to obtain a baseline flux. After 50 minutes, SDS was added to the feed solution, raising its concentration to 1 mM. This concentration was chosen because it is high enough to significantly reduce the surface tension of water (from 70.9 to 53.9 mN/m as determined by using a Wilhelmy plate tensiometer), yet low enough not to create foam or bubbles. At the critical micelle concentration (about 8 mM), the solution will get somewhat cloudy and suds begin to form in the feed tank, and this was not desired. The flux performance data are shown in Figure 15, which depicts forward osmosis flux measurements for the CE and SW30 XLE membrane. Also included is data for the SW30 XLE membrane with no fabric. The membranes are oriented in the PRO mode. SDS stock solution was added to the deionized feedwater (final SDS concentration after addition was 1 mM) at time equal to 50 minutes. Experimental conditions: deionized water feed solution, 1.5 M NaCl draw solution, crossflow rate and temperature of both feed and draw solutions are 21.3 cm/s and 20 ⁰C, respectively. Note that a water flux of 10 μm/s corresponds to 21.2 gal ft^"2 d^"1 (gfd) or 36.0 1 m^"2 h^"1.

Note that for the SW30 XLE membrane, there is a sharp increase in flux upon the addition of SDS to the feed solution. This is due to an immediate wetting of the support layer which subsequently improves the continuity of water in this layer, facilitating water flow. However, we do not know how each various layer of the support layer behaves in the presence of SDS. In order to differentiate the effect of SDS on the PET fabric and the PS support, we can look at the data with the CE membrane. When SDS is added to the feed solution during the tests with the CE membrane, the flux does not change. This either suggests that the wettability of the PET fabric plays no role in water transport (unlikely based on the results above) or that the SDS does not adequately wet this fabric more than it already is. If we assume that the PET layers of the CE and SW30 XLE are similar, than we hypothesize that the SDS is improving the wetting of the PS support layer, not the PET fabric.

The data also shows that after the increase in flux caused by SDS for the SW30 XLE membrane, there is a subsequent flux decline. This decrease is due to the accumulation of SDS molecules within the porous support layer. SDS is rejected by the active layer of the membrane and thus will be retained within the support layer. The SDS

{WI597672} amasses near the active layer and effectively fouls its pores. The increase in concentration of the SDS within the support layer, caused by internal CP, may also have another effect. SDS concentration likely increases to beyond that of the CMC, causing micelles to form within the support layer. These larger SDS aggregates would likely cause even more severe fouling, further decreasing water flux.

Membrane support layer hydrophobicity significantly hinders water flux in osmotically driven membrane processes. Lack of sufficient support layer wetting not only exacerbates internal concentration polarization phenomena, but also disrupts water continuity within the membrane, thereby reducing the pathways for water transport. Improved wetting of the support layer has been shown to increase water flux, especially for pressure retarded osmosis applications with dilute feed solutions. This improved wetting was achieved by purging the air and vapor out of the support layer with RO pretreatment prior to testing, or with the use of a surfactant to improve wetting within the layer. It was shown, though, that these treatments impacted the various layers of each membrane differently depending on the structure and hydrophobicity of the layers. It was also demonstrated that the most unsaturated layer, not the most hydrophobic layer, controlled water continuity and hence overall support layer wetting within the composite support layer. These results have shown that improving membrane design for osmotically driven membrane processes should not only focus on the structural characteristics of the support layer but on the chemistry as well. Hydrophilic membrane support layers are critical to improving flux performance in future process technologies that rely on osmotic transport across polymeric asymmetric membranes.

Finally, while the invention has been particularly shown and described with respect to preferred embodiments thereof, it will be understood that changes in form and details may be made therein without departing from the scope and spirit of the invention.

Claims

WHAT IS CLAIMED IS:

1. A semi-permeable membrane for use in a forward osmosis (FO) process, wherein said semi-permeable membrane is used to separate a draw solution and a feed solution, the semi-permeable membrane comprising:

a) an active layer having a high degree of selectivity for water and a very low resistance to water transport; b) a hydrophilic support layer exhibiting at least one of high porosity, low tortuosity and minimal thickness; and c) optionally, a backing layer, wherein the support layer is chemically resistant to conditions presented by the draw solution and the active layer is chemically resistant to conditions presented by the feed solution.

2. The semi-permeable membrane according to claim 1, wherein the support layer comprises a hydrophilic material selected from the group consisting of polyacrylonitrile, polyamide, crosslinked polyvinyl alcohol, sulfonated polysulfone, sulfonated polyethersulfone, standard polysulfone/polyethersulfone polymers after treatment in an oxidative environment, polysulfone/polyethersulfone polymers with hydrophilic additives or nanocomposites, and combinations of one or more of the foregoing.

3. The semi-permeable membrane according to claim 1, wherein the active layer comprises a material selected from the group consisting of polyamides, polysulfonamides, cellulose polymers and combinations of one or more of the foregoing.

4. A semi-permeable membrane for use in a pressure retarded osmosis (PRO) process, the semi-permeable membrane comprising: a) an active layer having a high degree of selectivity for water and a very low resistance to water transport; b) a hydrophilic support layer exhibiting at least one of high porosity, low tortuosity and minimal thickness; and c) optionally, a backing layer; wherein the semi-permeable membrane is capable of withstanding pressures generated during the PRO process.

5. The semi-permeable membrane according to claim 4, wherein the porous support layer comprises a hydrophilic material selected from the group consisting of but not exclusive to polyacrylonitrile, polyamide, crosslinked polyvinyl alcohol, sulfonated polysulfone, sulfonated polyethersulfone, standard polysulfone/polyethersulfone polymers after treatment in an oxidative environment, polysulfone/polyethersulfone polymers with hydrophilic additives or nanocomposites, and combinations of one or more of the foregoing.

6. The semi-permeable membrane according to claim 4, wherein the active layer comprises a material selected from the group consisting of polyamides, polysulfonamides, cellulose polymers and combinations of one or more of the foregoing.

7. The semi-permeable membrane according to claim 4, wherein the semi-permeable membrane is capable of tolerating a high pH of an ammonia carbon dioxide draw solution and has a high solute rejection.

8. A method of increasing the degree of saturation of a semi-permeable membrane for use in an osmotically driven process, wherein the semi-permeable membrane comprises a backing layer, a porous support layer, and an active layer having a high degree of selectivity for water and a very low resistance to water transport, the method comprising the step treating the semi-permeable membrane to increase its degree of water saturation by at least one of:

9. A semi-permeable membrane for use in a PRO process, the membrane comprising: a) a supportive hydrophilic polyester backing layer; b) an electrospun nanofiber mat layer on the support layer; and c) a selective barrier layer on the electrospun nanofiber mat layer, said selective barrier layer having a high degree of selectivity for water and a very low resistance to water transport.

10. The semi-permeable membrane according to claim 9, wherein the electrospun nanofiber mat layer comprises a hydrophilic material selected from the group consisting of polyacrylonitrile, polyamide, crosslinked polyvinyl alcohol, sulfonated polysulfone, sulfonated polyethersulfone, standard polysulfone/polyethersulfone polymers after treatment in an oxidative environment, polysulfone/polyethersulfone polymers with hydrophilic additives or nanocomposites, and combinations of one or more of the foregoing.

11. The semi-permeable membrane according to claim 9, wherein the supportive backing layer comprises a hydrophilic polyester support.

12. The semi-permeable membrane according to claim 9, wherein the selective barrier layer comprises a hydrophilic polyamide layer cast by interfacial polymerization.

13. The semi-permeable membrane according to claim 9, wherein the porosity of the nanofiber layer is greater than about 80%.

14. The semi-permeable membrane according to claim 9, wherein the tortuosity of the membrane is about 1.

15. The semi-permeable membrane according to claim 10, wherein the nanofibers farther comprise a surfactant or hydrophilic nanomaterials incorporated therein.

16. A method of making a semi-permeable membrane for use in an osmotically driven process, the method comprising the steps of: a) electrospinning polymer nanofibers into a non-woven mat on a backing layer; and b) casting a barrier layer on top of the electrospun nanofiber mat layer,

wherein the barrier layer has a high degree of selectivity for water and a very low resistance to water transport; and

wherein a semi-permeable membrane support structure is formed having high porosity and high tortuosity.

17. The method according to claim 16, wherein the membrane nanofiber layer has a porosity of at least about 80%.

18. The method according to claim 16, wherein the membrane has a tortuosity of about 1.

19. The method according to claim 16, wherein the electrospun nanofiber mat layer comprises nanofibers selected from the group consisting of polyacrylonitrile, crosslinked polyvinyl alcohol, sulfonated polysulfone, sulfonated polyethersulfone, and combinations of one or more of the foregoing.

20. The method according to claim 19, wherein the nanofibers further comprise a surfactant or hydrophilic nanomaterials incorporated therein.

21. The method according to claim 16, wherein the backing layer is a hydrophilic polyester support.

22. The method according to claim 16, wherein the active layer comprises a material selected from the group consisting of polyamides, polysulfonamides, cellulose polymers and combinations of one or more of the foregoing, and wherein the active layer is cast by interfacial polymerization or phase inversion.

23. The method according to claim 22, wherein the thickness of the active layer is between about 200 and about 400 nanometers.

24. A method of making a semi-permeable membrane for use in an osmotically driven process, the membrane comprising: a) an electrospun nanofiber mat layer as a support layer; and b) a selective barrier layer on the electrospun nanofiber mat layer, said selective barrier layer having a high degree of selectivity for water and a very low resistance to water transport; wherein the electrospun nanofiber mat layer functions as substantially the only support layer.

25. The method according to claim 24, wherein the membrane nanofiber layer has a porosity of at least about 80%.

26. The method according to claim 24, wherein the membrane has a tortuosity of about 1.

27. The method according to claim 24, wherein the electrospun nanofiber mat layer comprises a hydrophilic material selected from the group consisting of polyacrylonitrile, polyamide, crosslinked polyvinyl alcohol, sulfonated polysulfone, sulfonated polyethersulfone, standard polysulfone/polyethersulfone polymers after treatment in an oxidative environment, and combinations of one or more of the foregoing.

28. The method according to claim 27, wherein the nanofibers further comprise a surfactant or hydrophilic nanomaterials incorporated therein.

29. The method according to claim 24, wherein the active layer comprises a material selected from the group consisting of polyamides, polysulfonamides, cellulose polymers and combinations of one or more of the foregoing and is cast by interfacial polymerization or phase inversion.

30. The method according to claim 29, wherein the thickness of the active layer is between about 200 and about 400 nanometers

31. The method, according to claim 24, wherein the nanofibers are electrospun onto a removable substrate and the removable substrate is removed following interfacial polymerization of the active layer, thereby resulting in a two tiered membrane structure comprising only the nanofiber mat layer and the selective barrier layer.