7 Cross-Flow Filtration Ramesh R. Bhave 1.0 INTRODUCTION Cross-flow filtration (CFF) also known as tangential flow filtration is not of recent origin. It began with the development of reverse osmosis (RO) more than three decades ago. Industrial RO processes include desalting of sea water and brackish water, and recovery and purification of some fermentation products. The cross-flow membrane filtration technique was next applied to the concentration and fractionation ofmacromolecules commonly recognized as ultrafiltration (UF) in the late 1960's. Major UF applications include electrocoat paint recovery, enzyme and protein recovery and pyrogen re- moval. [11-r31 In the past ten to fifteen years or so, the applications sphere of cross- flow filtration has been extended to include microfiltration (MF) which primarily deals with the filtration of colloidal or particulate suspensions with size ranging from 0.02 to about 10 microns. Microfiltration applications are rapidly developing and range from sterile water production to clarification of beverages and fermentation products and concentration of cell mass, yeast, E-coli and other media in biotechnology related applications .[11-[41 Table 1 shows the types of separations achievable with MF, UF and RO membranes when operated in cross-flow configuration. For MF or UF application, the choice of membrane materials includes ceramics, metals or polymers, whereas for RO at the present only polymer membranes are predominantly used. Although cross-flow filtration is practiced in all the above three types of membrane applications, the description of membrane 2 71 Fermentation and Biochemical Engineering Handbook Cross-Flow Filtration 273 characteristics, operational aspects and applications will be limited to MF and UF, where the cross-flow mode shows the greatest impact on filtration performance compared with dead end filtration. Figure 1 shows the sche- matic of cross-flow filtration including the critical issues and operational modes for clarification or concentration using a semipermeable polymeric or inorganic membrane. Despite the growing use in a broad range of applications, cross-flow filtration still largely remains a semi-empirical science. Mathematical models and correlations are generally unavailable or applicable under very specific and well-defined conditions, owing to the complex combination of hydrodynamic, electrostatic and thermodynamic forces that affect flux and or retention. Membranefouling is not yet fully understood and is perhaps the biggest obstacle to more widespread use of CFF in solid-liquid separations. Membrane cleaning is also not well understood. The success of a membrane- based filtration process depends on its ability to obtain a reproducible performance in conformance with the design specifications over a long period of time with periodic (typically once a day) membrane cleaning. 2.0 CROSS-FLOW VS. DEAD END FILTRATION The distinction between cross-flow and dead end (also known as through-flow) filtration can be better understood if we first analyze the mechanism of retention. The efficiency of cross-flow filtration is largely dependent on the ability of the membrane to perform an effective surface filtration, especially where suspended or colloidal particles are involved. Table 2 shows the advantages and versatility of cross-flow filtration in meeting a broad range of filtration 0bjectives.[']-[~1[~] Figure 2 illustrates the differences in separation mechanisms of CFF versus dead end filtration. High recirculation rates ensure higher cross-flow velocities (and hence Reynold's number) past the membrane surface which promotes turbulence and increases the rate of redispersion of retained solids in the bulk feed. This is helpful in controlling the concentration polarization layer. It may be of interest to note that polarization is controlled essentially by cross-flow velocity and not very much by the average transmembrane pressure (ATP). It should also be noted that higher particle or molecular difisivity under the influence of high shear can enhance the filtration rates. Since difisivity values of rigid particles (MF) under turbulent conditions are typically much higher than those for colloidal particles or dissolved macromolecules (UF) microfiltration rates tend to be much higher than ultrafiltration rates under otherwise similar condition^.[^] 274 Fermentation and Biochemical Engineering Handbook 1 W m m v) Od Y .- - i - K 0 .- Y Y E s - iE m - v) 3 0 3 f I= .- Y ou 75- a W 0, a 75 W 0, L i rn V m Y Cross-Flow Filtration 275 e, 5 4 4 V V m C 3 2 L 0 0 a d c W 0 - - :: w L 0 L e, a 3 - v, C 0 .- E 0 v) r: .d 3 I 276 Fermentation and Biochemical Engineering Handbook DEADEND FlLTnATlON CROSS-FLOW FILTRATION FlLl J r I IT nhTE Figure 2. Cross-flow versus dead end filtration. On the other hand, in dead end filtration the retention is achieved by particle or gel layer buildup on the membrane and in the pores of the medium such as when a depth type filter is used. This condition is analogous to that encountered in packed-bed geometries. In dead end filtration, the applied pressure drives the entire feed through the membrane filter producing a filtrate which is typically particle- free while the separated particles form a filter cake. The feed and filtrate travel concurrently along the length ofthe filter generating one product stream for every feed. In CFF, one feed generates two product streams, retentate and permeate. Per pass recovery in through-flow mode is almost 100% (since only the solids are removed) whereas in the cross-flow mode the per pass recovery typically does not exceed 20% and is often in the 1 to 5% range. Recirculation of retentate is thus necessary to increase the total recovery at the expense of higher energy costs. PERMEAT E Cross-Flow Filtration 277 As the filtration progresses, the filter cake becomes increasingly thicker which results in a reduced filtration rate (at a constant transmembrane pressure). When the flow or transmembrane pressure (depending on the control strategy) approaches a limiting value, the filtration must be inter- rupted in order to clean or replace the membrane filter. This discontinuous mode of operation can be a major disadvantage when handling process streams with a relatively high solid content. Cross-flow filtration can overcome this handicap by efficient fluid management to control the thickness of the concentration-polarization layer. Thus, feed streams with solid loading higher than 1 wt.% may be better suited for CFF whereas feed streams containing less than 0.5 wt.% solids may be adequately served by dead end filtration. However, if the retained solids constitute the product to be recovered or when the nature of solids is the cause of increased fouling, cross-flow filtration should be considered. CFF is also the preferred mode when particle size or molecular weight distribution is an important consideration, such as in the separation of enzymes, antibiotics, proteins and polysaccharides from microbial cell mass, colloidal matter and oily emulsions. Tubular cross-flow filters are being used to continuously concentrate relatively rigid solids up to 70 wt.% and up to 20 wt.% with gelatinous materials. 3.0 COMPARISON OF CROSS-FLOW WITH OTHER COMPETING TECHNOLOGIES Cross-flow filtration as a processing alternative for separation and concentration of soluble or dissolved components competes with traditional equipment such as dead end cartridge filtration, pre-coat filtration and centrifugation. The specific merits and weaknesses of each ofthese filtration alternatives are summarized in Table 3. In addition to the ability to handle wide variations in processing conditions, other considerations may need to be addressed for economical viability of cross-flow filtration. These are briefly discussed below. A more detailed discussion on process design aspects, capital and operating cost considerations is presented in Sec. 6.7. 1. Energy Requirements. Centrifugal devices typically re- quire high maintenance. In contrast, cross-flow filtration requires minimal maintenance with low operating costs in most situations except for large bore (>6 mm) tubular membrane products operating under high recirculation rates. The energy requirements in dead end filtration are typically low. 278 Fermentation and Biochemical Engineering Handbook Cross-Flow Filtration 279 2. Waste Minimization and Disposal. CFF systems mini- mize disposal costs (e.g., when ceramic filters are used) whereas in diatomaceous (DE) pre-coat filtration sub- stantial waste disposal costs may be incurred, particularly if the DE is contaminated with toxic organics. Currently, in many applications, DE is disposed of in landfills. In future, however, this option may become less available forcing the industry to use cross-flow microfiltration technology or adopt other waste minimization measures. 3. Capital Cost. Many dead end and DE based filtration systems can have a relatively low capital cost basis.[2] On the other hand, CFF systems may require relatively higher capital cost. Centrifuges can also be capital intensive especially where large-scale continuous filtration is required. 4.0 GENERAL CHARACTERISTICS OF CROSS-FLOW FILTERS The performance of a cross-flow filter is primarily defined by its efficiency in permeating or retaining desired species and the rate of transport of desired species across the membrane barrier. Microscopic features of the membranes greatly influence the filtration and separation The nature of the membrane material Pore dimensions Pore size distributions Porosity Surface properties such as zeta potential Hydrophobichydrophilic character Membrane thickness From an operational standpoint, the mechanical, thermal and chemical stability ofthe membrane structure is important to ensure long service life and reliability. Table 4 summarizes the influence and significance of these features on the overall performance of a cross-flow filter. The discussion on the general characteristics of polymeric and inorganic membranes is treated separately partly due to their differences in production methods and also due to important differences in their operating characteristics. Fermentation and Biochemical Engineering Handbook I j i i U i I ! 6 I i 2 i G i J C 0 e 0 5 i: P 0 e R m 0 L J - E 3 C V 0 c CI ffi I Cross-Flow Filtration 281 4.1 Polymeric Microfilters and Ultrafilters Symmetric polymeric membranes possess a uniform pore structure over the entire thickness. These membranes can be porous or dense with a constant permeability from one surface to the other. Asymmetric (also sometimes referred to as anisotropic) membranes, on the other hand, typically show a dense (nonporous) structure with a thin (0.1-0.5 pm) surface layer supported on a porous substrate. The thin surface layer maximizes the flux and performs the separation. The microporous support structure provides the mechanical strength. Polymeric membranes are prepared from a variety of materials using several different production techniques. Table 5 summarizes a partial list of the various polymer materials used in the manufacture of cross-flow filters for both MF and UF applications. For microfiltration applications, typically symmetric membranes are used. Examples include polyethylene, polyvinylidene fluoride (PVDF) and polytetrafluoroethylene (PTFE) mem- brane. These can be produced by stretching, molding and sintering fine- grained and partially crystalline polymers. Polyester and polycarbonate membranes are made using irradiation and etching processes and polymers such as polypropylene, polyamide, cellulose acetate and polysulfone mem- branes are produced by the phase inversion proce~s.['1[~1[~] Ultrafiltration membranes are usually asymmetric and are also made from a variety of materials but are primarily made by the phase inversion process. In the phase inversion process, a homogeneous liquid phase consisting of a polymer and a solvent is converted into a two-phase system. The polymer is precipitated as a solid phase (through achange in temperature, solvent evaporation or addition of a precipitant) and the liquid phase forms the pore system. UF membranes currently on the market are also made from a variety of materials, including polyvinylidene fluoride, polyacrylonitrile, polyethersulfone and polysulfone. Microfiltration membranes are characterized by bubble point and pore size distribution whereas the UF membranes are typically described by their molecular weight cutoff (MWCO) value. The bubble point pressure relates to the largest pore opening in the membrane layer. This is measured with the help of a bubble point apparatu~.['l[~I The average pore diameter of a MF membrane is determined by measuring the pressure at which a steady stream of bubbles is observed. For MF membranes, bubble point pressures vary depending on the pore diameter and nature of membrane material (e.g., hydrophobic or hydrophilic). For example, bubble point values for 0.1 to 0.8 pm pore diameter membranes are reported to vary from 1 bar (equals about 282 Fermentation and Biochemical Engineering Handbook 14.5 psi) to 15 bar.[’] However, due to the limited mechanical resistance of some membrane geometries (e.g., tubular and to some extent hollow fiber) such measurements cannot be performed for smaller pore diameter MF and UF membranes. The bubble point apparatus can also be used to determine the pore size distribution of the membrane. Table 5. Polymeric Microfilters and Ultrafilters Material Microfilter Ultrafilter Connvuration Acrylic polymer X X HFF Cellulosic polymer X X FS. PS, SW, HFF Nylon based polyester X X HFF, PS. Fs Polyamide X HFF, FS Polybenzamidazole X Fs. sw Polycarbonate X Fs Polyelhersulfone X SW, T Polye Lliylene X Fs Poly p ro py le ne X HFF, FS. T I’olysulTone X IIFF, SW, T, FS I~olylelrafluorocthylcne X X FS. T Polyvinylidene fluorlde X X SW, T, FS. PF PF - Plate and Frame PS - Pleated Sheet FS - Flat Sheet SW - Spiral Wound T - Tubular (including wide channel) HFF - Hollow Fine Fiber Since the majority of UF membranes have dense surface layers, it is difficult to characterize them with a true pore size distribution. Therefore, polymeric UF membranes are described by their ability to retain or allow passage of certain solutes. The MWCO values for UF membranes can range from as low as 1000 dalton (tight UF) to as high as 200,000 Dalton (loose UF). This roughly corresponds to an “equivalent” pore diameter range from about 1 nanometer (nm) to 100 nm (0.1 pm) as described in Ref. 10. Cross-Flow Filtration 283 Different membrane materials with similar or identical MWCO value may show different solute retention properties under otherwise similar operating conditions. If adsorption effects are negligible, such a result can be attributed primarily to the differences in their pore size distributions. This is illustrated in Fig. 3. It can be seen that, although the two membranes are rated by the same MWCO value, their retention characteristics are distinctly different (sharp versus diffuse). Polymeric cross-flow filters are available in many geometries. These are listed in Table 6. It is obvious that no single geometry can provide the versatility to meet the broad range ofoperating conditions and wide variations in properties. Some cross-flow filters such as cartridge filters have low initial capital cost but high replacement costs and tubular filters may show longer service life but higher operating costs. The optimization of CFF for a specific application may depend on economic and/or environmental factors and is almost impossible to generalize. 1 .o Y C P) U .- g 0 U C 0 U a, aJ a: .- 4.1 '-7 0 ideal Culolf - 1,000 /*- / 1 / / / / /" Dilluse Cutoff '/ / / / 10,000 100,000 Molecular Weight Figure 3. Rejection coefficient as a function ofmolecular weight cutoff of an ultrafiltration membrane. 284 Fermentation and Biochemical Engineering Handbook 0 d E E c'! m 0 rn rn W v) - d Y - L c 00 cdn i 0 Cross-Flow Filtration 285 4.2 Inorganic Microfilters and Ultrafilters Cross-flow membrane filters made from inorganic materials, primarily ceramics and metals, utilize entirely different manufacturing processes compared with their polymeric counterparts, [31 Although carbon membranes do not qualifjl under the inorganic definition, they will be included here due to the similarities with inorganic membranes with regard to their material properties such as thermal, mechanical and chemical resistance as well as similarity in production techniques. Table 7 lists the various commonly used materials and membrane geometries in MF and UF modules. Commercial ceramic membranes are made by the slip-casting process. This consists of two steps and begins with the preparation of a dispersion of fine particles (referred to as slip) followed by the deposition of the particles on a porous A majority of commonly used inorganic membranes are composites consisting of a thin separation barrier on porous support (e.g., MembraloxB zirconia and alumina membrane products). Inorganic MF and UF mem- branes are characterized by their narrow pore size distributions. This allows the description of their separative performance in terms of their true pore diameter rather than MWCO value which can vary with operating conditions. This can be advantageous in comparing the relative separation performance of two different membranes independent of the operating conditions. MF membranes, in addition, can be characterized by their bubble point pressures. Due to their superior mechanical resistance bubble point measurements can be extended to smaller diameter MF membranes (0.1 or 0.2 pm) which may have bubble point pressure in excess of 10 bar with water.L9] Typical pore size distributions of inorganic MF and UF membranes are shown in Fig. 4. The narrow pore size distribution of these membrane layers is evident and is primarily responsible for their superior separation capabili- ties. The manufacturing processes for inorganic membranes have advanced to the point of delivering consistent high quality filters which are essentially defect free. Inorganic MF and UF membranes also display high flux values (see Table 8) which they owe to their composite/asymmetric nature combined with the ability to operate at high temperatures, pressures and shear rates. Two kinds of membrane geometries are predominantly used, the tubular multi-lumen and the multichannel monoliths with circular, hexagonal or honeycomb structures. The number of channels can vary from 1 to 60. Fermentation and Biochemical Engineering Handbook 1 e e 8 m c 0 .- 2 .e N 0 x .- e L m 3 P 3 b 0 % n P) X a x 0 9 e E 0 E .” e P) P) m c) Cross-Flow Filtration 287 1 I PORE MEMBRANE RADIUS CAI - AL-I 19.7 ’ AL-I 19.7 - - - 0.m - - ‘lo I 35 30 x 0 d L Y 0 - L 0 V < E - c L 100 200 300 400 Pore diameter, nm (a) Figure 4. (a) Typical pore size (diameter) and (b) typical pore size (radius) distributions of inorganic h4F and UF membranes. 288 Fermentation and Biochemical Engineering Handbook e 0 Cross-Flow Filtration 289 5.0 OPERATING CONFIGURATIONS There are several operating configurations that are used in industrial practice depending on flow rate of the product, product characteristics and desired final concentrations of the product which is either to be retained by the membrane or recovered in the permeate. 5.1 Batch System Figure 5 shows a simple batch system consisting of a feed tank, a membrane module and a feed pump which also serves as a recirculation pump. The recirculation pump maintains the desired cross-flow velocity over a certain range of transmembrane pressures depending on the type of pump and its characteristics (centrifugal or positive displacement). The filtration continues until the final concentration or desired permeate recovery is achieved, unless the flux drops to an unacceptable level. For the retention of suspended solids (e.g., bacteria, yeast cells, etc.) the final concentration factor can be anywhere from 2 to 40 (and higher in some applications where a recovery of >98% is required). In order to minimize the concentration effects, a ratio of concentrate flow rate to permeate flow rate of about 10 to 1 is maintained (assuming the density differences are not significant). This ensures that at any given time the concentration of solids in the recirculation loop is only about 10% higher than that in the feed loop. Depending on the operating cross-flow velocity and viscosity of retentate, the pressure drop along the length ofthe module can vary from 0.5 bar to more than 2 bar. This often necessitates the use of more than one parallel loop and limits the number of modules in series depending on pump characteristics. The open loop configuration has some advantages in terms of its simplicity, but also has some disadvantages especially when the product is sensitive to heat or shear effects (e.g., intracellular products, some beverages, and enzymes). Furthermore, when higher cross-flow velocities are required (which is typically the case in many applications) the recirculation rates necessary to sustain them may not be achievable in the open loop configura- tion, especially if it is also desirable to maintain a concentrate to permeate ratio of at least 10. This problem can be overcome by placing a feed pump between the feed tank and the recirculation pump, as shown in Fig. 6. The discharge pressure ofthe recirculation pump must be at least greater than the pressure loss along the flow channels in the module or several modules connected in series while maintaining the desired recirculation rate. 290 Fermentation and Biochemical Engineering Handbook 4 @- L c Y B m & 0 I vi Cross-Flow Filtration 291 t 292 Fermentation and Biochemical Engineering Handbook The pipe sizes for feed and return lines for the closed loop operation are much smaller than that for the open loop system which can also reduce the capital and operating cost. The feed tank size can also be much smaller for the closed loop which then allows shorter residence times for heat or shear sensitive products. The average flux (J,) in the batch configuration may be estimated using J, = Jf+ 0.33(Ji - Jr) where Jf = flux at the final concentration Ji = initial flux 5.2 Feed and Bleed Batch Mode. The closed loop operation shown in Fig. 6 may not be suitable in many situations such as when processing large volumes of product and where high product recoveries (>95%) are required. It is well known that flux decreases with an increase in the concentration of retained solids which may be suspended particles or macrosolutes. When high recoveries are required, high retentate solids must be handled by the cross-flow filtration system. For instance, when a 95% recovery is desired, the concentration of solids in the loop must be 20 times higher than the initial feed concentration (assuming almost quantitative retention by the membrane). If the filtration proceeds beyond the 95% recovery, much higher solids concentration in the retentate loop will result which could adversely affect the flux. Figure 7 shows the schematic of a batch feed and bleed system. A constant final concentration in the retentate loop can be maintained by bleeding out a small fraction, either out of the system or to some other location in the process. This operation is described as a batch feed and bleed and is commonly used in the processing of many high value biotechnology products such as batch fermentations to recover vitamins, enzymes and common antibiotics.[l2I The CFF system will require larger surface area since the system must be designed at the flux obtained at the final concentra- tion factor (e.g., 20 for 95% recovery). Cross-Flow Filtration 293 I r 294 Fermentation and Biochemical Engineering Handbook Continuous Mode. When large volumes are processed the batch feed and bleed system is replaced with a continuous system shown in Fig. 8. The size of the feed tank is much smaller compared to that for the batch system. However, since the concentration of solids changes with time, the permeation rate decreases with time. This requires the adjustment of feed flow to the recirculation loop. This value is obtained by adding total permeation rate to the bleed rate. The concentration buildup in the continuous feed and bleed mode of operation is somewhat faster than the batch mode, which translates into a higher surface area requirement due to lower flux at higher solids concentration. Such an operating configuration, however, serves very well in many large scale fermentation broth clarifications (e.g., common antibiot- ics such as penicillin and cephalosporin) and is used when long holding times are not a concern. For continuous processes, the lowest possible system dead volume will enable the operation with low average holding times. This may be important in some applications, especially those involving bacteria-laden liquids. Low system dead volume is also desirable for batch or continuous processes to minimize the volumes of cleaning solutions required during a cleaning cycle. Diafiltration. The product purification or recovery objective in most UF operations is achievable by concentrating the suspended particles or microsolutes retained by the membrane while allowing almost quantitative permeation of soluble products (such as sugars, salts, low molecular weight antibiotics) into the permeate. This approach to concentration of solids obviously has limitations since recoveries are limited by concentration polarization effects. This limitation can be overcome by the use of diafiltrati0n.['1['~1['~] The process involves the selective removal of a low molecular weight species through the membrane by the addition and removal of water. For example, in many antibiotics recovery processes, the broth is concentrated two- to fivefold (depending on the extent of flux reduction with concentration). This corresponds to a recovery of 50 to 80%. Higher recoveries are obtainable by adding diafiltration water or solvent in nonaque- ous medium. The permeate leaving the system is replaced by adding fresh water, usually through a level controller, at the rate which permeate is removed. Diafiltration efficiency can be varied by the mode ofwater addition. Figure 9 shows the schematic for a batch and continuous diafiltration process. Diafiltration can be performed at higher temperatures to facilitate higher permeation rates. A possible disadvantage would be the dilution of the product requiring further concentration (e.g., by evaporation). Cross-Flow Filtration 295 - e, +J +J C e, V C e 8 L Tr 296 Fermentation and Biochemical Engineering Handbook Water Flow, F - Retentate 7 - 7 - t Permeate Flow (PI - - Constant - Volume Batch Diafiltration Water Flow Water Flow Feed Flow CF Filtration - System Stage Q Permeate Flow (= F,) Permeate Flow (-5) Continuous Diafiltration Figure 9. Batch and continuous diafiltration process modes. Cross-Flow Filtration 297 5.3 Single vs. Multistage Continuous System Single stage continuous contiguration may not be economical for many applications since it operates at the highest concentration factor or lowest flux over most of the process duration. Multistage continuous systems on the other hand, can approximate the flux obtained in the true batch mode, depending on the number of stages. The concentrate from each stage becomes the feed to the next stage. The number of stages required will depend on the final recovery or retentate concentration. Figure 10 shows the schematic of a three-stage continuous system. The optimum number of stages will depend on the application, but typically lie between 2 and 4, with the greatest benefit resulting from a single stage to a two stage continuous system. The biggest advantage in using multistage continuous configuration, especially in fermentation and biotech- nology applications, is the minimization of residence time, which may be crucial in preventing excessive bacterial growth or to handle heat labile materials.[15] The other advantageofacontinuous system is the use ofasingle concentrate flow control valve. As membrane fouling andor concentration polarization effect begins to increase over the batch time, flux decreases. This requires the continuous or periodic adjustment ofconcentrate flow which may be accomplished with the ratio controller. One disadvantage with a multistage system is the high capital cost. It is necessary to have one recirculation pump per loop which drives the power requirements and operating costs much higher compared with the batch feed and bleed configuration. 6.0 PROCESS DESIGN ASPECTS 6.1 Minimization of Flux Decline With Backpulse or Backwash Almost all cross-flow filtration processes are inherently susceptible to flux decline due to membrane fouling (a time-dependent phenomenon) and concentration polarization effects which reflect concentration buildup on the membrane surface. This means lower flux (Le., product output) which could drive the capital costs higher due to the requirement of a larger surface area to realize the desired production rate. In some situations, the lower flux could also result in lower selectivity which means reduced recoveries and/or incomplete removal of impurities from the filtrate. For example, removal of inhibitory metabolites such as lactic acid bacteria[l6I or separation of cells from broth while maximizing recovery of soluble products.[2] 298 Fermentation and Biochemical Engineering Handbook v) aJ p- I 0, (0 U i E 0, n T Cross-Flow Filtration 299 300 Fermentation and Biochemical Engineering Handbook Backwashing or backpulsing with permeate can help remove excessive membrane deposits and hence minimize flux decline.C31 Cross-flow micro- and ultrafilters typically operate as surface filtration devices with insignifi- cant pore plugging. If severe pore plugging occurs, backpulse will most likely be ineffective in preventing precipitous flux decline. This type of irreversible fouling may only be corrected by cleaning by chemical andor thermal heat treatment. An essential difference between a backpulse and a backwash is the speed and force utilized to dislodge accumulated matter on the membrane surface. In backpulsing, periodic counter pressure is applied, typically in a fraction of a second (0.1-0.5 seconds), while generating high permeate backpressure (up to 10 bar). Backwash on the other hand is relatively gentle where permeate backpressure values may increase up to 3 bar over a few second duration. Backwash is commonly used with polymer MF/UF filters due to their lower pressure lirnitation~[~1["] compared with inorganic MF/UF filters where backpulsing is used.f3] The maximum benefit of backpulse or backwash is obtained when the retentate pressure during instantaneous reverse filtration is lowest and the applied permeate backpressure is highest. Depending on the operating configuration, a periodic backpulse may be applied on the entire filtration system or when several modules are operating in series, subsequent application will produce more effective results. In the latter case, the retentate pressure may be higher as a result of pressure loss through the interconnected feed channels. It is recommended that when a backpulse or backwash is used, it is applied for the shortest duration possible (to minimize the loss of productivity), it uses minimum permeate volume, and begins simultaneously with the filtration process.[3] Backpulsing is less effective for some smaller pore diameter UF membranes (MWCO <30,000 or pore diameter less than 0.02 pm) and where dense layers are formed or gelatinous products are filtered. It is important to bear in mind that, although backpulsing has the ability to minimize the concentration polarization effects and produce a higher average flux, a certain portion ofthe permeate is consumed (1 to 3% by volume). If permeate is the product of interest, then the net realized flux will be average flux minus permeate volume used during backpulsing. 6.2 Uniform Transmembrane Pressure Filtration In the conventional cross-flow filtration described in previous section, the transmembrane pressure (TMP) along the feed flow channels varies substantially from the feed end ofthe module to the exit or retentate end. This Cross-Flow Filtration 301 occurs due to the pressure loss in the feed channels to maintain the desired flow rate (and hence cross-flow velocity). The shell side or the permeate side is held at a constant pressure. There may be several important consequences which can contribute to a relatively lower flux or loss in separation efficiency. A major consequence is the formation of a nonuniform layer of suspended solids, colloidal matter, andor gel-forming microsolutes retained on the membrane. It is not uncommon to experience a TMP value up to 50% higher at the module inlet compared to that at the outlet, especially at high shear rate or cross-flow velocity. This could result in a substantially lower average flux. In some applications (e.g . , milk or cheese concentration, whey concentration and fermentation broth clarification for product recovery) significant differ- ences in the retention characteristics have also been observed.[18] In many biotechnology related applications, where MF or UF membranes are used, the primary objective is to retain particles (e.g., whole cells or lysed cells, yeast, colloidal matter, andor macrosolutes such as enzymes, pyrogens, proteins, and in some situations oily emulsions). In order to accommodate the wide variations in particle size distributions, a pore diameter is selected that is small enough to retain all the particles or macrosolutes, but large enough to allow the permeation of smaller molecular weight soluble products such as common antibiotics, mono- and disaccharides, organic acids and soluble inorganic salts. Nonuniform TMP values over the filtration surface area may cause substantial (up to 50%) reduction in the product recovery in the permeate. A novel approach to improving the flux andor product recovery utilizes the concept of a uniform transmembrane pres~ure.[~][~~] This is achieved by varying the permeate side pressure with an independent recirculation pump to adjust the TMP to a constant value. A schematic of the UTP and conventional cross-flow configuration is shown in Figs. 11 and 12, respec- tively. The TMP profiles for the two operational modes are shown in Fig. 13. Flux improvements up to 500% have been achieved compared with the conventional cross-flow mode in many important food, beverage and biotech- nology applications. An additional benefit is reduced fouling which means longer duration ofoperation for batch processes andeasier cleaning ofmembrane modules for repeated usage. The only major requirement is the ability of the membrane structure to withstand backpressures up to 5 bar on shell side when filtering high viscosity products such as gelatins, or feed streams with high dissolved solids (20 to 70 wt.%). 302 Fermentation and Biochemical Engineering Handbook c e i p-> I Cross-Flow Filtration 303 304 Fermentation and Biochemical Engineering Handbook Retentate i Feed Retentate t t Feed Permeate -3 Pressure Profiles out 0 1 2 3 K bar Pressure Profiles out Jd- Permeate 7 in 1 2 3 4 bar II 123q bar b) Conventional Cross-flow Operation Figure 13. Transmembrane pressure profiles: (a) uniform transmembrane pressure operation, and (b) conventional cross-flow operation. Cross-Flow Filtration 305 6.3 Effect of Operating Parameters on Filter Performance A number of operating parameters need to be studied to optimize the overall filtration performance. Critical among these are the cross-flow velocity, transmembrane pressure, pore diameter, or MWCO and concentra- tion of the retained species at the end of a batch operation or steady state concentration in continuous filtration. This latter parameter can be related to the recovery of product in the permeate or retentate. Other important operating variables are temperature (and hence viscosity), pH, backpulse or backwash, and pretreatment. Membrane Pore Diameter or Molecular Weight Cutoff. The value of membrane pore diameter will have a major influence on the permeation and separation characteristics for most process filtration applications. The intrinsic membrane permeability is related to the pore diameter for many microfiltration membranes whereas, for ultrafiltration membranes, it is typically indicative of the solute retention properties. Tables 8 and 9 provide typical permeability and retention data for many common MF and UF membranes, re~pectively.[']~[~1[~] The permeability and retention characteristics listed in the tables, however, should only be used as a guide since the actual filter performance may be dependent on a number of other variables and operating conditions. In addition, for many MF/UF membranes, especiallythose made ofpolymeric materials, initial flux and retention properties may significantly alter with repeated use in aggressive conditions or over a longer (6 months to 1 year) period of operation. It is evident that the smaller the pore diameter, the lower the pure solvent (in most cases water) flux, and the higher the ability to retain macrosolutes, colloidal and particulate matter. Users should also be aware that pure solvent flux values are seldom realized in practice and are often at least about an order of magnitude lower in most industrial applications due to effects of fouling and concentration polarization. As a rough rule of thumb, for maximum retention, the pore diameter should be at least about 40 to 50% lower than the smallest particle diameter under the operating conditions. This includes consideration of shear or particle agglomeration/ deagglomeration effects. The nominal MWCO on the other hand should be at least 20 to 30% of the smallest molecular weight of the species to be retained. This is due to the fact that for most membranes, particularly polymeric UF, the MWCO characteristics may be diffused[*] rather than sharp (see Sec. 4.1). Further, secondary layer formation on the membrane surface due to adsorption, fouling and gel polarization will also influence the retention of UF 306 Fermentation and Biochemical Engineering Handbook Table 9. Retention Characteristics of Cross-flow Ultrafilters Molecular welght Membrane Material Manufac:urer Remarks' OltOlT~ loo0 Cellulosic polymen 5Ooo Polysulfone lo.m Zlrconla Polyelliersulfone Polyacrylonl:rlle Cclluloslc polymers 7- Alumina P~IYSUI~OIIC 2o.ooo Polynmltic Zlrconla Cclliilosic polymers I'olysullorle Polyamide m,oOo - 75.000 Cclluinsir pnlynicrs Fliioropolymcr Polysullolie 21 rco ii 1 a 2cQ.m 300.000 Fluoropoiynicrs Polyolcflns Arnlcon Koch Koch Asahl carbosep M5 Sartorliis Flultl Sys:cms/UOP USF Membralox @ 5 nm Dorr-Oliver Tccli Scp Carbosep M4 Sarloriiis Koch. Miillpore lloecllst Tcrli Srp USF KOCll I locrlist-Celanese Memtek Koch Koch, Millipore Tech Scp USF Dorr -0llver Kocl1 Carboscp hll Membralox@ 0.02 wn Celgard 2400 Carbosep M9 Membralox@ 0.05 wm ' May show equlvalcut or Iowcr MWCO depending on solute and may be influenced by operatlng condlllons. Practical considerations, however, require a compromise between the ideal goals and process economics. One major factor is the lack of reliable information and/or molecular weight distribution of macrosolutes. As a result, application specialists or process engineers typically recommend a pore diameter which is about 75% of the smallest particle size or a MWCO value of about 50-60% lower than the smallest macrosolute. The objective is to maximize flux without sacrificing solute retention below the set minimum requirements, Cross-Flow Velocity. The cross-flow velocity, which is also a measure of the shear or turbulence in the flow channels, may have a strong influence on flux. The actual shear or turbulence will depend on several factors such as channel diameter, viscosity and density of retentate and can vary over the duration ofthe filtration (especially for batch operations). This Cross-Flow Filtration 307 can be characterized by the calculation of Reynold’s number on the retentate stream. High Reynold’s numbers (>4000) indicate turbulent flow whereas those below 2000 show laminar flow. The objective is to use a high cross- flow velocity to maximize flux by minimizing the gel polarization layer within the constraints imposed by the allowable pressuredrop or system limitations. It should also be noted that for many applications flux increases with cross- flow velocity. This is illustrated in Fig. 14.[211 The extent of flux improve- ment will depend on process stream, flow regime (laminar or turbulent) and characteristics of the gel polarization layer formed due to concentration buildup at the membranelfeed interface.[3] 1000 [ 800 GOO 400 1 2 3 4 5 Cross-flow velocity, m/s Figure 14. Effect of cross-flow velocity on flux. Yeast concentration, dry-&: (0) 8.5; (0) 30. Blatt et a1.[20] have shown that the mass transfer coefficient can be related to the cross-flow velocity by The value of k can be approximated by Eq. (3) k= D/S 308 Fermentation and Biochemical Engineering Handbook The value of a can vary from 0.3 to 0.8 in laminar flow and 0.8 to 1.3 in turbulent flow. In the absence of particles (e.g., cells): (low particle loading) (high particle loading) (low particle loading) (high particle loading) This behavior has been explained by the so-called “tubular pinch effect,” which enhances movement of particles away from the boundary layer thus reducing concentration polarization effect (see Sec. 3.3). For turbulent flow, the pressure drop along the flow channel may be estimated by using the following empirical approximation: Under laminar flow conditions, This indicates that ahigher cross-flow velocity under turbulent condi- tions can result in more than proportional increase in the pressure drop requiring larger pump discharge pressure to maintain a specified recircula- tion rate. This limits the number of modules that may be placed in series to minimize capital costs. Typical range of cross-flow velocity values is 2 to 7 m/s, The choice of pump is critical to obtain efficient fluid recirculation. It is critical to understand the shear sensitivity of the fluidparticle to be processed to determine the optimal cross-flow velocity in situations where shear-sensitive materials are involved. Concentration of Solute or Particle Loading. It is essential to distinguish or separate the effects of membrane fouling from concentration polarization effects. Cross-Flow Filtration 309 Membrane Fouling. Pretreatment of the membrane or feed solution prior to filtration may be desirable within allowable limits. The various treatment options are discussed in Sec. 6.3. At the start of a filtration run, the solute or solids concentration is relatively small and progressively builds as the permeate is removed from the system. If a substantial flux decline is observed at low solids concentration, membrane fouling aspects are believed to be important. A flux decrease with an increase in solids concentration is largely due to concentration polarization and can be minimized through efficient fluid hydrodynamics andor backpulsing .[31[221[231 Several approaches have been developed to control membrane fouling. They can be grouped into four categories: (a) boundary layer ~ontrol;[~~1[~~]-[*~] (b) turbulence inducers/generator~;[~~I (c) membrane modification~;[~*]-[~~] and (6) use of external field~.[~~]-[~~] In CFF membrane, fouling can be controlled utilizing a combination of the first three approaches (a, b and c). The external field approach has the advantage of being independent of the hydrodynamic factors and type of membrane material.[35] Membrane fouling is primarily a result of membrane-solute interac- ti~n.[~~I These effects can be accentuated or minimized by proper selection of membrane material properties such as hydrophobicityhydrophilicity or surface charge, adjustment of pH, ionic strength and temperature leading to solubilization or precipitation of solutes. Increased solubilization of a foulant will allow its free passage into the permeate. If this is undesirable, precipitation techniques may be used which will enhance the retention of foulants by the membranes. Membrane fouling is generally irreversible and requires chemical cleaning to restore flux. It is important to recognize that fouling in bioprocessing differs from that occurring with chemical foulants. Biofouling originates from microor- ganisms. Microbes are alive and they actively adhere to surfaces to form biofilms. Thus, in addition, to flux decrease there may be significant differences in solute rejection, product purity, irreversible membrane fouling resulting in reduced membrane life. For economic viability of CFF it is imperative that a good and acceptable cleaning procedure is developed to regenerate fouled membranes without sacrificing membrane life. Concentration Polarization. The concentration of the species re- tained on the membrane surface or within its porous structure is one of the most important operating variables limiting flux. Concentration effects in MF/UF can be estimated by using the following most commonly used correlation. [121[371 J = kln[C,/C,] 31 0 Fermentation and Biochemical Engineering Handbook where J = flux k = mass transfer coefficient Cg = gel concentration of at the membrane surface C, = bulk concentration of solute retained by the mem- In membrane filtration, some components (dissolved or particulate) of the feed solution are rejected by the membrane and these components are transported back into the bulk by means of diffusion. The rate of diffusion will depend on the hydrodynamics (laminar or turbulent) and on the concen- tration of solutes. If the concentration of solute at the surface is above saturation (i.e., the solubility limit) a “gel” is formed. This increases the flow resistance with consequential flux decrease. This type of behavior, for example, is typical of UF with protein solutions. In practice, however there could be differences between the observed and estimated flux. The mass transfer coefficient is strongly dependent on diffusion coefficient and boundary layer thickness. Under turbulent flow conditions particle shear effects induce hydrodynamic diffusion of particles. Thus, for microfiltration, shear-induced diffusivity values correlate better with the observed filtration rates compared to Brownian diffusivity calcula- tion~.[~] Further, concentration polarization effects are more reliably pre- dicted for MF than UF due to the fact that macrosolutes diffusivities in gels are much lower than the Brownian diffusivity of micron-sized particles. As a result, the predicted flux for ultrafiltration is much lower than observed, whereas observed flux for microfilters may be closer to the predicted value. Typically MF fluxes are higher than those for UF due to their higher pore diameter values which contribute to higher initial fluxes. However, polarization effects dominate and flux declines with increase in concentration (or % recovery) more sharply in MF than in UF, in general accordance with Eq. (4) under otherwise similar conditions. Figure 15 shows the typical dependence of flux on concentrati~n.[’~I Higher the concentration of the retained species on the membrane compared with its initial value, the higher will be % recovery. However, if the desired product is in the permeate, then % recovery will be dependent on the ratio ofthe batch volume to its final value for batch filtration or the ratio of concentrate in permeate to that in the feed for continuous filtration~.[~~l[~~] brane Cross-Flow Filtration 311 I 0 N 0 Q, c L 0 d u (P U c 0 .- d 2 Y K Q, u c 0 0 31 2 Fermentation and Biochemical Engineering Handbook Transmembrane Pressure. The effect oftransmembrane pressure on flux is often dependent on the influence of concentration polarization at a specified cross-flow velocity and solids loading. For MF or UF at low solids concentration and high cross-flow velocity, flux may increase linearly with TMP up to a certain threshold value (1 to 3 bar), and then remain constant or even decrease at high TMP values. This is illustrated in Fig. 1 6.L2l] At high solids loading, the threshold value may be lower (0.5 to 1.5 bar) and may also require higher cross-flow velocity to offset gel polarization effects. For each application the optimum value may be considerably different and must be empirically determined. 1000 N E L L 2i x' - 500 LL 0 0 1 2 3 4 5 6 Transmembrane Pressure, bar Figure 16. Effect of transmembrane pressure on flux. Yeast concentration, dry-@: (0) 8.5; (0) 30. Cross-Flow Filtration 313 An optimum TMP value is one which maximizes flux without additional energy costs and helps minimize the effects of membrane fouling. In general, higher the solids concentration, the higher the cross-flow velocity and hence TMP to balance the effects of concentration polarization. In most practical situations, the cross-flow velocity and TMP may be interrelated. One useful approach involves performing pressure excursion studies to determine the optimal flux by varying the TMP at a fixed cross-flow velocity until the threshold TMP is attained and then repeat the tests by selecting a lower or higher cross-flow depending, on the observed trend.[40] In many biotechnology applications, such as fermentation broth clari- fications to produce common antibiotics, optimal values of TMP are in the range of 2 to 3 bar (15 to 30 psi) especially at high cell mass concentrations (> 30 wt.%) and cross-flow velocity range of4 to 7 ds.[21[401 In the operation of commercial systems, often several modules (2 to 4) are interconnected to minimize pump costs. This results in significantly higher TMP on the feed end compared to that at the exit (or retentate). Thus, the TMP on the exit end may be closer to the optimal value whereas at the inlet it may be substantially higher (6 or 7 bar), unless the permeate backpressure on each module is controlled independently. Temperature and Viscosity. The operating temperature can have a: beneficial effect on flux primarily as a result of a decrease in visc0sity.[~1[~~] There is an additional benefit for shear thinning viscoelastic fluids, where the viscosity reduces withan increase in shear (i.e., cross-flow velocity). Typical examples are clarification of fermentation broths and concentration of protein ~olutions.[~1[~~] It must be noted that for most fermentation and biotechnology related applications, temperature control is necessary for microbial survival and/or for product stability (e.g., antibiotics, enzymes, proteins and other colloidal materials). For mass transfer controlled operations, such as when concentration polarization is dominant, flux enhancement due to temperature increase will depend on the value of mass transfer coefficient. This is related to the cross- flow velocity, diffusion coefficient and viscosity.[2o] Thus, for example, even though the viscosity may be reduced by a factor of 5, the increase in flux may only be about 50%, due to the nonlinear dependence of flux on viscosity in these situations. For the permeation of a clean liquid (solvent) across a microporous membrane, however, flux increase may be predicted by the Stokes-Einstein relation['] and will be approximately inversely proportional to the viscosity of the permeate. pH, Isoelectric Point and Adsorption. In the filtration of proteins and colloidal substances, the solution pH can have a measurable effect on 314 Fermentation and Biochemical Engineering Handbook flux, especially around the isoelectric point where they tend to destabilize and precipitate. In addition, the surface charge or isoelectric point of the membrane material must also be considered. For example, most inorganic membranes are made out of materials such as silica, zirconia, titania and alumina which have a charge on the surface with isoelectric pH variation from 2 to 9.[31 Similarly polymeric membranes such as cellulose acetate, polyamides and polysulfones also carry surface charges.[’] Surface charge effects may alter the fouling resistance due to changes in the zeta potentials and could have a substantial influence on flux andor separation perfor- mance. For example, proteins with isoelectric point of 8 to 9 will have a positive charge in a neutral or acidic solution and negative charge in alkaline solutions (PH > 9). However, if the above proteins in a neutral medium are filtered through an alumina membrane (isoelectric pH 9) there will be a minimal adsorption on the membrane due to similar charge characteristics. At a solution pH of 8 or 9, which is also the isoelectric pH of protein, however, proteins may precipitate out of solution. This may have a beneficial effect on the flux through a MF or UF membrane. This also illustrates the interactive effects of solution pH, isoelectric pH (of solutes and membrane material) and adsorption. Feed Pretreatment. In most fermentation and biotechnology related applications, feed pretreatment is not a viable option. This is due to the fact that any alterations in the feed properties, especially through the addition of precipitants or flocculants, will likely contaminate the product and or adversely affect its characteristics. Prefiltration is recommended when applicable to remove larger par- ticles and other insoluble matter. However, minor pretreatment chemistries may be allowable, such as pH adjustment to precipitate or solubilize impurities or foulants to maximize flux or retention. For example, protein adsorption and fouling can be reduced by adjusting the pH away from its isoelectric point.[6] The selection ofa suitable pore diameter or MWCO value is done on the basis of the smallest particle size or smallest macrosolute present in the feed. 6.4 Membrane Cleaning Likewise, to the inevitable phenomena of membrane fouling, all membrane based filtration processes require periodic cleaning. Without a safe practical, reproducible, cost effective and efficient cleaning procedure, the viability of cross-flow filtration may be highly questionable. Membrane cleaning process must be capable of removing both external and internal Cross-Flaw Filfrafion 315 deposits. In some special situations, such as strongly adsorbed foulants, recirculation alone may not be adequate and soaking of the membranes in the cleaning solutions for a certain period of time will be necessary. The ultimate success of a membrane process will be largely impacted by the ability of the cleaning procedure to fully regenerate fouled membranes to obtain reproduc- ible initial flux at the start of the next filtration cycle. The ease of finding an effective cleaning process often depends on the thermal and chemical resistance of the membrane material. In other words, the higher the resistance, the easier it is to develop a suitable cleaning procedure. The choice of a cleaning solution depends on several factors such as the nature of the foulants, and material compatibility of the membrane elements, housing and seals. A few general guidelines are available concerning the removal of foulants or membrane deposits during chemical cleaning. [31[411 Common foulants encountered in biotechnology related applications are inorganic salts, proteins, lipids and polysaccharides. In some food or biochemical applications, fouling due to the presence of citrate, tartrate and gluconates may be encountered. Inorganic foulants (e.g., precipitated salts of Cay Mg and Fe) can be removed with acidic cleaners whereas, proteina- ceous and other biological debris can be removed with alkaline cleaners with or without bleaching agents or enzyme cleaners. Many acidic and alkaline cleaners also contain small quantities of detergents, which act as complexing or wetting agents to solubilize or remove insoluble particles, colloidal matter and/or to break emulsions. Oxidizing agents such as peroxide or ozone are also sometimes used to deal with certain type of organic foul ant^.[^^] In addition, organic solvents may be required to solubilize organic foulants that are insoluble in aqueous cleaning solutions. For many polymeric MF/UF membrane modules, material compatibil- ity considerations limit the use of higher cleaning temperatures and strongly acidic/alkaline/oxidizing solutions. Further, with time and repeated cleaning, polymeric filters are susceptible to degradation. The service life of a hydrophobic type is typically a period of 1 to 2 years and up to 4 years for fluoropolymer based membranes ,f6] On the other hand, inorganic membranes can be cleaned at elevated temperatures in strongly alkaline or acidic solutions and can withstand oxidizing solutions or organic solvents. The typical useful service life of inorganic membranes exceeds 5 years and may be used for 10 years or longer with proper cleaning, and good operating and maintenance procedures. i31i6] A careful choice of cleaning solutions and procedures will extend the service life ofthe membrane. In many polymer membrane filtration systems, 316 Fermentation and Biochemical Engineering Handbook membrane replacement costs constitute a major component of the total operating cost. Extending the service life ofthe membrane modules will have a major impact on the return on investment and can be a determining factor for the implementation of a membrane-based filtration technology. Table 10 summarizes the various key parametersthat must be considered in developing a cleaning regimen to regenerate fouled membranes. Product losses during cleaning may be important especially when high recoveries (>95%) are required and the desired product is located in the retentate phase. Additional product loss will occur in the fouled membrane elements. These combined losses may range from 0.5% to 3% which is significant when recovering high value-added product. 6.5 Pilot Scale Data and Scaleup Scaling up membrane filter systems must proceed in a logical and progressive series of steps. It is practically impossible to extrapolate data from a laboratory scale system to design a production scale system.[44] To ensure commercial success, it is often necessary to supplement laboratory data with pilot system capable of demonstrating the viability of the process. This is typically followed-up with extensive testing using demonstration scale or semi-commercial scale filtration system to obtain long-term flux informa- tion and to establish a cleaning procedure to regenerate fouled membrane modules. This exercise is especially important to determine the useful life of membranes. At least a 3 to 6 month testing is recommended regardless of the scale of operation. Pilot scale studies will also allow production of larger quantities of materials for evaluation purposes to ensure that all the separa- tion and purification requirements are adequately met. It is necessary to ensure that the feed stream characteristics are representative of all essential characteristics, such as age of feed sample, temperature, concentration of all components (suspended and soluble), and pH. The filtration time needed to perform a desired final concentration of retained solids or percent recovery of product passing across the mem- brane filter (the permeate) must also be consistent with the actual process requirements. Effects of sample age, duration of exposure to shear and heat, may be very important and must be considered, In the demonstration scale phase, the operating configuration (e.g., batch, feed and bleed, continuous) specified for the production scale system must be used. Careful consideration must be given to the total pressure drop in the flow channels at the desired cross-flow velocity at the final concentration, to ensure proper design of the feed and recirculation pumps. Cross-Flow Filtration 31 7 L 0 Y 0 3 m 5 e E 0 a x W - d E E + 0 3 m z L W i L 0 31 8 Fermentation and Biochemical Engineering Handbook It is important to generate the flux data on a continual basis as illustrated in Fig. 17. This type of information is very vital to identify any inconsistencies in the filtration performance andor to determine ifthere is any irreversible membrane fouling. Reproducible performance will also be helpful to validate the membrane cleaning regimen for the application. 6.6 Troubleshooting Filtration equipment must function in a trouble-free manner and perform in accordance with the design basis. Although most carefully designed, engineered and piloted cross-flow filtration systems will perform to design specifications, occasional failures are not uncommon. For proper troubleshooting of CFF systems, the user must be familiar with the principles of membrane separation, operating and cleaning procedures, influence of operating variables on system performance and equipment limitations. In this chapter, the principles of membrane separations when operating in the cross-flow configuration are discussed in detail along with the influence of operating variables on flux and separation performance. However, proper start-up and shutdown procedures must be followed to maximize the system performance. For instance, the formation or presence of gas or vapor microbubbles can cause severe pore blockage especially for MF and in some UF applications. Therefore, care must be taken to remove air or gas from the feed and recirculation loop at the start of a filtration run to ensure that no air is drawn or retained in the system. This type of operational problem may not only occur during normal filtration but also during backpulsing. When troubleshooting the cleaning operation, a good understanding of the foulants and process chemistry is highly desirable. A thorough under- standing of materials of construction of the seals/gaskets is required for a proper choice of cleaning regimen. The membrane manufacturers guidelines must be properly implemented and combined with the process knowledge and feed characteristics, When working with new or dry membranes, it may be necessary to properly wet the membrane elements. For microporous struc- tures, the use of capillary forces to wet the membrane and fill the porosity is recommended. 6.7 Capital and Operating Cost The manufacture and purification of many biotechnology products derived from fermentation processes involves several separation steps. Up to 90% of the total manufacturing cost may be attributed to various Cross-Flow Filtration 319 320 Fermentation and Biochemical Engineering Handbook separation processes to raise the concentrationofthe product in solution from parts per million or several percent to the final concentrated form. Figure 18 illustrates the contribution of separation costs to the selling price of a biological The most economical system design is achieved by the consideration of both capital and operating costs. When comparing the overall filter perfor- mance against a competing technology, care should betaken to ensure that the total cost or payback is based on the life-cycle rather than solely on the basis of initial capital or operating costs. It is easier to compare two competing technologies or products on the basis of initial capital cost alone. However, this approach may be erroneous unless operating, maintenance and replacement costs are considered along with differences (or savings) in the value of product recovered or lost. For example, clarification of a fermentation broth with a cross-flow filter may cost up to 4 times higher compared to the capital cost of a pre-coat filter, but the operating cost may be only about 50% of that incurred with pre-coat filter.[12] The disposal cost of the filter aid will also add to the savings of CFF over pre-coat filtration.[46] The higher capital cost can be justified through cost savings yielding a reasonable payback (typically in the range of 3 to 4 years). Cross-flow filtration competes with many traditional separation and filtration technologies such as centrifuges, rotary pre-coat filters, cartridge filters, chemical treatment and settling and a filter press. The advantages and disadvantages of some ofthese alternatives were briefly discussed in Sec. 3 .O. This section will highlight key items that make up the major portion of the capital and operating costs in cross-flow filtration. The cross-flow filter accounts for a major portion of the capital cost. The relative percentage contribution to the total capital cost will vary from about 20% for small systems up to 50% for larger systems. Thus, replacement costs, when the CFF has auseful service life of only about a year, can be as much as 50% of the total system cost. Inorganic filters cost more than their polymer counter parts but can last about 5 to 10 years. Capital costs associated with feed pump and recirculation pump@) represents anywhere from 5 to 15% of the total capital costs. The largest contribution to the operating cost in many cross-flow filtration systems is in the energy consump- tion for recirculation.['] For example, in the production of common antibi- otics such as penicillin or cephalosporin, high recirculation rates are main- tained (corresponding to a cross-flow velocity in the range 5 to 8 ds) to minimize concentration polarization.[2] Energy requirements under turbulent flow conditions are also significantly higher than under laminar flow situations, under otherwise similar conditions. In addition, total energy costs Cross-Flow Filtration 321 0) m P 322 Fermentation and Biochemical Engineering Handbook must also consider heating andor cooling requirements as well as to deliver the desired head pressure to overcome hydraulic pressure drop. Operating costs must also consider equipment maintenance, cost of cleaning chemicals and labor costs. CFF systems in general, have substan- tially lower maintenance and labor costs compared with other competing technologies. Cleaning chemical costs are typically low and account for only about 1 to 4% of the total operating costs.[3] 6.8 Safety and Environmental Considerations The proper and efficient operation of a cross-flow filtration system requires a design based on sound engineering principles and must rigorously adhere to safe engineering practices. CFF systems must be equipped with high pressure switches to safely diffise a high pressure situation and must also use materials and design criteria per American Society for Testing and Materials (ASTM) standards. Proper insulation is required in accordance with Occupational Safety andHealth Administration (OSHA) regulations for high surface temperatures or hot-spot when operating at elevated tempera- tures. For corrosive chemicals, proper handling and disposal procedures must be followed for operator safety. Containers approved by OSHA and other regulatory agencies must be used when transporting or transferring hazardous chemicals. In addition, proper procedures must be followed when mixing chemicals, either within the manufacturing process or while handling waste solutions. The majority of CFF processes are operated in a closed configuration which minimizes vapor emissions. Some traditional techniques such as centrifugal processes may generate aerosol foaming in the air (e.g., patho- gens) which is highly undesirable. 7.0 APPLICATIONS OVERVIEW Due to the highly proprietary nature of fermentation of biochemical products, the published descriptions on cross-flow filtration performance are very limited. This section will review some of the more important types of applications where cross-flow filtration is used. The performance descrip- tions are limited by available published information which is often incom- plete. As a result, at best, only qualitative or general comparisons can be made between the various technology alternatives.