|
Membranes,
the Finest Filtration INTRODUCTION The notion that salt could be removed from seawater without a phase change took hold of Dr. Sourirajan's imagination in the late 1950's. The actual invention of the reverse osmosis (RO) membrane took place in his laboratory at UCLA around 1960. (Thus Sourirajan could be called not only the "Father of Reverse Osmosis," but the initiator of all crossflow membrane technology.) The full commercialization of RO and its sister crossflow technology, ultrafiltration (UF), occurred in the early 1970's. About a dozen U.S. companies and a couple in Europe and Japan found they had to educate their customer base on both the technical reality and the potential uses of their products. In the 1980's, crossflow membrane processes became well accepted in industry and medicine, and the inevitable industry shakeouts and price competition characteristic of a maturing technology followed. Today hundreds of manufacturing processes, waste treatment and water purification applications rely on crossflow membrane for cost-effective separations. This paper overviews crossflow membrane technology and its current applications, highlights some recent innovations and speculates on emerging trends. The topic is very broad so will not be covered in detail here. Issues important to some users and potential users will be missed; especially regarding applications, which cannot be covered in less than a book. The reader is invited to read further in the large body of literature available on membrane, and contact the author with specific questions on applications. CROSSFLOW MEMBRANE TECHNOLOGY DEFINED Membrane filtration is the separation of the components of a pressurized fluid, effected by polymeric or inorganic membranes (generally man-made). The openings in the membrane material (pores) are so small that a significant fluid pressure is required to drive the liquid through them; the pressure required varies inversely with the size of the pores (basically classical orifice theory). There are now four commonly accepted categories or "classes" of membrane, defined based on the size of the material they will remove from the carrier liquid. Moving from the smallest to largest pore size, these are Reverse Osmosis (RO), Nanofiltration (NF), Ultrafiltration (UF), and Microfiltration (MF). These membranes have pores so small they will plug and blind off instantly, unless they are run in the crossflow mode. (Some MF applications are run in the traditional normal flow mode.) Unlike traditional filtration, all the influent does not pass through the media. Rather it is split into the permeate (filtrate) and concentrate streams, the latter of which flows parallel to the membrane. Hence the term "crossflow." (Somehow this process has also picked up the inaccurate term "tangential flow," which is not an accurate description of the geometry of the process.) Likewise, traditional filtration is often called by the clumsy terms "dead-end" or "straight-through" flow filtration." The term "normal-flow" would work better for two reasons: it is a geometric term which describes how the flow approaches the media, and it is what most people normally think of as filtration. Briefly, crossflow membrane is a pressure driven process wherein a semipermeable media acts as a surface filter to split the feed stream into two effluents: a purified stream and a stream more concentrated in solutes too large to pass through the pores of the particular membrane. There are four basic classes of membrane, based on relative pore size. THE CLASSES OF CROSSFLOW MEMBRANE Ultrafiltration membranes can separate medium to large size dissolved molecules from the solvent, due largely to the simple sieving mechanism. While the size of the pores allows a significant transport of the solvent (usually water) through the membrane when under pressure, solutes in the 5,000 to 500,000 molecular weight range are excluded from transport based mainly on their physical size. To a small degree, the ionic, dipole moment and Van der Waal's forces associated with the molecules play a role in separation as well. Ultrafiltration class membranes are defined to include only those membranes with pores too large to reject or remove salt ions, but small enough to reject larger dissolved and colloidal species. The pores are generally accepted as ranging in size from 20 to 500 angstroms (Å) diameter. The overwhelming majority of UF membranes in commercial use are polymeric in nature, although new ceramic and metal oxide aggregate UF membranes are emerging and may grow to comprise as much as 10% of the market. Reverse osmosis affects separation of even smaller solutes, including salts with ionic radii in the Å range. The mechanism of rejection is more complex than in UF and MF and is not based solely on a sieving mechanism, but on electrochemical interactions as well. The exact mechanism is both theoretical and disputed. There are at least two different schools of thought regarding this phenomenon: The most scientific and highly developed is Dr. Sourirajan's "surface-force-pore flow" approach. Simply put, the preferential sorption of water molecules and subsequent desorption of ions (repulsion by dielectric forces) causes exclusion of charged solutes even smaller than the membrane pores from movement through the membrane. The alternate "solution-diffusion" theory, stated simply, holds that RO membranes are like porous films in which both solvent and solute "dissolve" into the membrane. The solute moves through the membrane mainly under concentration gradient forces, while the solvent transport is dependent on the hydraulic pressure gradient. Pores in reverse osmosis membranes are so small they have not yet been resolved, even by the most advanced microscopic techniques. They are generally regarded to be in the 4 to 8 Å range, four orders of magnitude smaller than the finest of the normal-flow particle filters. Nanofiltration is a term coined about six years ago to define membranes, which were already in use, referred to then as "loose RO." They have pores close to one nanometer diameter (1-Å) and affect partial salt rejection. Typical NF membranes pass a higher percentage of monovalent salt ions than diva-lent and trivalent ions. Most NF membrane polymers carry formal charges which exclude higher valence ions more than monovalents from passing through the membrane with the solvent water. Nanofiltration membranes span the gap between RO and UF classes. While UF has been formally defined by the ASTM consensus standards group (Standard D-1 129), no such official definition yet exists for RO or NF. Microfiltration is well understood as the "fine" end of particle filtration, with pores of from 0.1 to 1 micron diameter (perhaps up to 3 or even 10 microns). As such, MF membranes have pores two to five orders of magnitude larger than the other classes. What is not as well known is that when the MF media is a membrane, it also can be run in the crossflow as well as normal-flow mode. This may provide lower cost operation and much longer media life. MEMBRANE MATERIALS Since RO membranes require both extremely small pores and significant water sorption tendency, only two materials are in common use: cellulose acetate and polyamide polymers. The CA membranes tolerate chlorine at levels used for microbial control, while PA membranes will be destroyed by even low levels of chlorine. However, the PA membranes produce both higher rejection and flux, and tolerate a wider pH range on a continuous basis and a higher continuous temperature than CA membranes (pH 2-8 for CA, 2-11 for PA, 40°C [104°F] for CA, 65°C [149°F] for PA). There is a sulfonated polysulfone RO membrane which reportedly achieves salt rejection via a formal, induced surface charge. This membrane does not reject salt to the same high degree as PA, and the feed must be softened to remove all divalent ions. Nanofiltration membranes are generally PA and CA based, with a few notable proprietary specialty polymers emerging in the market as well. Ultrafiltration membranes are made from a wider range of more rugged polymers, since their larger pore size and "sieve" mechanism of separation allow more material choices. Cellulose acetate, polyvinylidene fluoride, and especially polysulfone are the most common. Polysulfone UF membranes can withstand a pH range of 0.5 to 13, temperatures to 85°C(185°F), and 25 mg/L of free chlorine on a continuous basis. Materials
for microfiltration membranes include everything used for RO and UF membranes,
as well as polycarbonate, polypropylene, polyethylene and even PTFE (Teflon*),
and therefore can have quite rugged physical and chemical characteristics. Figure 1 shows commonly accepted guidelines for RO, NF and UF membranes. MF membranes have such a range of parameters that their properties are not readily reduced to a table. 1Values assume a one-year minimum target life. In most cases, operation outside the recommended values is feasible but will result in shortened life. 2Only for spiral-wound element configuration. 3Values assume ambient temperature operation. Crossflow MF, and more recently UF membranes, are also formed from ceramics. These are generally alumina based and have about the same chemical resistance of the polysulfone polymer but are more heat resistant. MEMBRANE ELEMENTS Membranes themselves are either formed as flat sheets, hollow fibers, or coated tubes. They must be configured into elements (devices) to manage the flow streams in the membrane machines and support the membrane under the required hydraulic pressures. Flat sheet configurations include the plate-and-frame and spiral-wound design; the latter predominates all forms including hollow fiber and tubular. The plate-and-frame design allows a variety of feed and permeate channel designs, but is a high cost approach and finds only niche applications. Spiral-wound design affords the best "all-around" characteristics of high packing density, low cost, and rugged, high-pressure operation. With the recent advent of specialized feed channel spacer materials, a wider range of applications now employs the spiral design. Hollow "fine" fibers - the diameter of human hair - are used mainly for RO purification of seawater where extensive pretreatment protects them from fouling in the norm. Hollow fibers of greater lumen diameter (0.5 to 2 mm) are used in UF processes. They can handle high solids loading without plugging and can be back-flushed to remove foulant layers. Since they are self-supporting homogeneous fibers, they are limited by the tensile, compressive and flexural strengths of the membrane material, which is porous. This limits the operating pressure and flow rates to less than those of spiral wounds. Tubular elements are large-scale versions of hollow fibers (0.6-2.5 cm diameter lumens) with the membrane coated on the inside wall of another porous material. This support material gives the tube its strength, so higher operating pressures are possible. However, since the forces act on the tensile strength of the tube, this design usually has lower pressure limits than spiral-wound and plate-and-frame, where the compressive strength of the membrane is the limiting factor. As with the plate-and-frame design, the much higher cost per unit area of membrane limits the tubular configuration's use. Elements are placed in pressure-bearing housings and arrayed in machines in a manner to maximize flows and minimize space and cost. Figure 2 shows a rough comparison of configurations. MACHINES AND SYSTEMS A simple machine design includes a pump to provide the driving pressure and crossflow velocity required, the elements in housings, connecting plumbing, control valve(s) and instruments. Pressure gauges and motor controls are required, but the degree of instrumentation and other controls can and does vary tremendously. Roughing prefilter cartridges are usually included to reduce membrane fouling. Everything from simple on/off conductivity sensor lights to complex, sophisticated PLC and computer-controlled systems are employed today for crossflow membrane machines. Machines can be placed in a facility in separate pieces, but are generally assembled on a frame or skid by the manufacturer. They can be tested at the factory and shipped ready to operate as soon as they are plumbed in and wired. The various possibilities of machine design are beyond the scope of this paper. As part of integrated systems, crossflow membrane machines often follow pretreatment equipment such as dual media filters and chemical injection. They can be preceded or followed by other unit processes such as degasification or activated carbon adsorption. For ultrapure water applications, ion exchange polishers (DI) are sometimes used to remove that small percentage of ions which passes through an RO membrane. With increased acceptance and reduced cost of the technology, two-pass RO systems (where the permeate is further treated by another RO machine) have replaced many RO-DI systems. DESIGN CONSIDERATIONS The most important design considerations are understood by most basic equipment manufacturers, and are touched on only briefly here. Higher-pressure causes higher permeate throughput (membrane flux) and would thus increase efficiency, except that it also causes more severe fouling by retained substances. Higher crossflow velocity reduces fouling, so a balance of flow and pressure must be achieved. The optimum balance varies by membrane type and especially by feed solution characteristics. The strength of the membrane, element and hardware determines the maximum hydraulic pressure and crossflow rate (which cause AP forces to act on the element) that can practically be applied. The ratio of permeate to feed volume also affects the fouling rate and is known as "recovery." Recovery is measured on both the individual element basis and for the entire machine. Feedwater applications usually run at 75-80% machine recovery, with 90% the practical upper limit. Some UF and RO applications have relatively high fouling rates, and these are run at lower recoveries (50-75%). Seawater Desal™ting via RO is typically run as low as 40% due to the very high osmotic pressure generated as the salt in the feed stream is concentrated. Flux (permeate rate per unit area of membrane) is directly proportional to effective pressure, absent the flux reducing fouling phenomenon. Thus with a pure water feed stream, increasing pressure proportionally increases permeate output, and the pressure limit would be based on the membrane/element/hardware strength and pump limitations. In actual practice, the increased fouling which results from increased flux and the practical limit of the crossflow, which mitigates this fouling, determine the optimum pressure. Osmotic pressure is the natural force which seeks to pull water from the permeate side of the membrane to equilibrate the solute concentrations on each side. Since this reverses the work desired of the system, an applied hydraulic pressure must first overcome this natural energy (hence the term reverse osmosis). For any solute, the greater the concentration, the higher the osmotic pressure. Osmotic pressure potential varies among ions; the higher the specific ion's activity, the greater the osmotic pressure. For organic solutes, it is roughly the rule that the smaller the molecule, the higher the osmotic pressure. While overcoming osmotic pressure is rarely a design concern for UF or MF systems, it can be a major consideration for NF and RO systems since they concentrate salts and smaller organics. The effective driving pressure in an NF or RO system is the gauge hydraulic pressure less the osmotic pressure differential. For optimum separation performance, about 200 psi (13.8 bar) effective pressure is required and performance falls off markedly below 100 psi (6.9 bar). Nevertheless, undersink RO is successfully applied in the home at typical line pressures of 40-60 psig. Temperature also influences flux due to viscosity reduction; the warmer the feed stream the higher the throughput. Of course the higher the flux, the higher the fouling rate. Arriving at the optimum balance of pressure, recovery, temperature and crossflow rate is an engineering art, and will vary with each feed source as long as proper design of crossflow velocity and recovery is considered; however, increasing temperatures up to the limits of the membrane and system will tend to increase system efficiency. Cold feed sources require larger systems (increased membrane). Decreasing solution viscosity has the same effect as increasing temperature (viscosity is the determinant for the temperature effect on flux). APPLICATIONS For those new to crossflow membrane, what can be accomplished with the technology is probably of most interest. Hundreds of applications have been commercialized, falling in three broad categories: water purification, manufacturing process separations, and waste treatment. Commercially proven examples are listed by these categories for a quick reference. Water Purification ·
Boiler feed PROCESS ·
Juice and milk concentration WASTE TREATMENT ·
Tertiary sewage water recovery These are just typical applications and are not an exhaustive list. New applications are being developed constantly and require only initiative and basic understanding of membrane technology. Potential users should work with an experienced vendor or consulting engineer to explore the feasibility of membrane for their applications. RECENT ADVANCES As a commercial process, crossflow membrane is relatively new. With the exception of a constant stream of new applications, only a few major advances for aqueous based treatment have occurred since the 1970's. Highlights of these follow: The technique of making composite membranes (as opposed to a homogenous material structure) which improved both flux and separation, was patented in 1981. During the 1980's, this type RO membrane grew to become the dominant type used. It also spurred the development of composite NF and UF membranes more recently. These "thin-layer-composite" membranes increase the possible polymer chemistry choices available, and the membrane manufacturers' industry is now focused on this technique. Surface treatment techniques to add unique separation characteristics to existing membrane "substrates" is a more recent area of interest. Adding formal charges via basic chemical bond changes (i.e., sulfonation), grafting ligand groups, plasma deposition, etc., are all potential techniques to change separation ability and reduce fouling tendency. Together with the composite structure approach, altering surface characteristics is the wave of the future for polymeric membrane formation. Ceramic and metallic membranes have been commercially available for about 10 years and, due to their substantial cost, their acceptance is emerging slowly. A few crossflow MF ceramics applications have been established. Composite layer techniques also allow large pore UF class ceramics to be made. Adding formal surface charges to ceramics is claimed to show some NF and RO properties, although this technique is not yet proven. The much higher cost of all such inorganic membranes will limit the applications to a few niche applications for the foreseeable future. "Elements" or sometimes "devices" are terms for the products engineered to contain the membranes for use in the physically rigorous environments where they are employed. In the area of element design, replacing engineering thermoplastics with stainless steel for structural parts has allowed higher temperature operation and hot water sanitization of membrane systems. This has expanded the application of the technology in the food, beverage, medical and pharMace™utical industries. A related advance was the "Full-Fit™™" design, which eliminated outer covers and seals on the predominant spiral-wound design element. This has shown reduced fouling in waste and process applications due to improved fluid dynamics, as well as increased cleanliness and decreased microbial growth for pure water systems. The range of applications has also been increased by the design of diffrent feed spacers for the spiral-wound configuration. Increased mesh size reduced the channel plugging sometimes experienced with high particulate solids feed streams. Mesh design improvements have increased turbulence and reduced surface fouling. A "clear channel" spacer was developed, and this combined the hollow fiber design's high viscosity fluid capability and lack of channel plugging from suspended solids with the high-pressure capability and inherent ruggedness of the spiral-wound element design. Enhanced systems controls have improved the operational efficiency of liquid treatment systems employing crossflow membrane. Microelectronics technology allows greater control of the overall process. PLC controllers are now used for remote monitoring of a membrane system, automatic change of valve settings to adjust system recovery to obtain either the desired concentration or permeate purity. They electronically record crucial operating parameters, and automatic variability of shutdown and start-up, cleaning and flushing cycles, etc. All these have improved cost effectiveness dramatically. Even more important than improved system control is the industry's evolving realization that treatment systems are often most efficient if they combine several unit processes. While UF and RO machines were once considered exotic polishing units for "roughing" filters, they now function as the roughing filter for processes such as ozonation or combining adsorption processes, which remove trace contaminants. For example, integrated systems combining several unit operations, with crossflow membrane as the "heart," routinely produce ultrapure water of 18.3 megohm resistivity, 10 µg/L (ppb) TOC, and parts per billion concentrations of salt ions, VOC's etc. Even when single unit operations, including the crossflow processes, could alone achieve the separations target, the most cost-effective design often includes a combination system. Exploiting the synergy of separate unit processes is the trend for large, industrial treatment systems. Going one step further, true wisdom dictates going farther upstream in industrial processes, targeting membrane uses that result in less material use, more material recovery and reuse, and subsequently less pollution and energy use. This is, of course, a fundamental challenge to all of mankind's industrial processes. For the smaller point-of-use systems, such as for laboratory use or consumer use in the home, companies experienced in that marketplace are designing complete packages for simplified use by consumers. Thus, home RO units are now available at retail appliance outlets, and hospitals and labs now order packaged systems from scientific supply catalogs. The important trends for the users of crossflow membrane are lower cost, more sophisticated systems, more user friendly hardware, and consequently wider application.
Fig. 1 Operating parameters for widely used polymeric RO and UF membranes. GENERIC RO AND UF MEMBRANE CONTINUOUS OPERATION PARAMETERS'
Fig.2: CROSSFLOW MEMBRANE CONFIGURATION COMPARISON
Fig.3: Typical Operating Pressures - psig (bar*)
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||