Solvent Extraction David B. Todd 1.0 EXTRACTION CONCEPTS Liquid-liquid extraction is a unit operation frequently employed in the pharmaceutical industry, as in many others, for recovery and purification of a desired ingredient from the solution in which it was prepared. Extraction may also be used to remove impurities from a feed stream. Extraction is the removal of a soluble constituent from one liquid into another. By convention, the first liquid is the feed (F) which contains the solute at an initial concentrationXf The second liquid is the solvent (S) which is at least partially immiscible with the feed. The solvent may also have some solute present at an initial concentration of x, but usually < is essentially zero. The solvent does the extraction, so the solvent-rich liquid leaving the extractor is the extract (E). With the solute partially or completely removed from the feed, the feed has become rejned so the feed-rich liquid leaving the extractor is the raflnate (R). When the feed and solvent are brought together, the solute (A) will distribute itselfbetween the two liquid phases. At equilibrium, the ratio ofthis distribution is called the distribution coeficient (m): 348 Solvent Exlraction 349 m=yA= concentration of A inextract phase X, concentration of A in raflinate phase The distributioncoefficient, m, is a measure ofthe affinity ofthe solute (A) for one phase (E, 5') over the other phase (E R). The concentration ofA may be expressed in various units, but for ease of subsequent calculations, it is preferable to express the concentration on a solute-free basis for both phases. For example, in the extraction of acetone from water with toluene: weight acetone weight acetonefree water X= weight acetone weight acetonefree toluene Y= Although the units of m appear to be dimensionless, they actually are If more than one solute is present, the preference, or selectivity, of the (weight acetone-free water)/(weight acetone-free toluene). solvent for one (A) over the other (B) is the separation factor (a). The separation factor (arn) must be greater than unity in order to separate A from B by solvent extraction, just as the relative volatility must be greater than unity to separate A from B by distillation. The analogy with distillation can be carried a step further. The extract phase is like the vapor distillate, a second phase wherein the equilibrium distribution ofA with respect to B is higher than it is in the feed liquid (liquid bottoms). Extraction requires that the solvent and feed liquor be at least partially immiscible (two liquid phases), just as distillation requires both a vapor and a liquid phase. 350 Fermentation and Biochemical Engineering Handbook Extraction requires that the solvent and feed phases be of different densities, Even though extraction may successfilly remove the solute from the feed, a fbrther separation is required in order to recover the solute from the solvent, and to make the solvent suitable for reuse in the extractor. This recovery may be by any other unit operation, such as distillation, evaporation, crystallization and filtration, or by krther extraction. Extraction is frequently chosen as the desired primary mode of separation or purification for one or more of the following reasons: 1. Where the heat of distillation is undesirable or the tem- perature would be damaging to the product (for example, in the recovery of penicillin from filtered broth). 2. Where the solute is present in low concentration and the bulk feed liquor would have to be taken overhead (most fermentation products). 3. Where extraction selectivity is favorable because ofchemi- cal differences, but where relative volatilities overlap. 4. Where extraction selectivity is favorable in ionic form, but not in the natural state (such as citric acid). 5. Where a lower form or less energy can be used. The latent heat of most organic solvents is less than 20% that of water, so recovery of solute from an organic extract may require far less energy than recovery from an aqueous feed. 1.1 Theoretical Stage The combinations of mixing both feed and solvent until the equilibrium distribution of the solute has occurred, and the subsequent complete separa- tion of the two phases is defined as one theoretical stage (Fig. 1). The two functions may be carried out sequentially in the same vessel, simultaneously in two different zones of the same vessel, or in separate vessels (mixers and settlers). Extraction may also be performed in a continuous differential fashion (Fig. 2), or in a sequential contact and separation where the solvent and feed phases flow countercurrently to each other between stages (Fig. 3). Solvent Extraction 351 STAGE MIXING TO PHlSE SEPARATION ECUlLlBRlW TO EWlLlBRlUl Figure 1. Theoretical stage. Extract r-+ Feed L Raffinate Figure 2. Differential extraction. Feed Figure 3. Sequential contact and separation. 352 Fermentation and Biochemical Engineering Handbook 2.0 DISTRIBUTION DATA Although data for many systems are available in the literature,['] in many cases it will be necessary for the engineer to obtain the distribution information for his own specific application. The simplest method is to mix solvent and feed liquors containing varying quantities of solute in a separatory bel, and analyze each phase for solute after settling. Where feed and solvent are essentially immiscible, the binary plot, such as shown in Fig. 4, is usefbl. For later ease of calculation, it is desirable to express concentrations on a solute-free basis. If there is extensive miscibility, a ternary plot (Fig. 5) would be preferable. Tie lines represent the equilibrium between the coexisting phases. &I-& gw X Figure 4. Binary plot of distribution data. Figure 5. Ternary plot of distribution data. Plotting the data on log-log graphs may be helphl in understanding some of the underlying phenomena and interpolating or extrapolating meager data. An example is shown in Fig. 6 for the distribution of phenol between water and various chlorinated methanes. In the dilute region, the limiting Solvent Extraction 353 slope is generally always unity. However, as the solute becomes more concentrated, there may be a tendency for solute molecules to associate with each other in one of the phases. Thus, the equilibrium data in Fig. 6 suggest that the phenol molecules form a dimer in the organic phase, probably by hydrogen bonding, leading to a slope of 2 in the distribution plot. The possibility of complex formation in one of the phases illustrates the concern that many industrial extraction processes involve not only the physical transfer of molecules across an interface but, also, that there may be a sequence of chemical steps which have to occur before the physical transfer can take place, and which may be rate limiting. I IO' Y 8 phenol gorganlc 10-1 10- g phenol IO" 8 ualrr Figure 6. Distribution of phenol between water and chlorinated methanes. Whenever the distribution coefficient is greatly different than unity, there is an implication that there exists an uflnity ofthe solute for that specific solvent, and this affinity may involve some loose chemical bonding. Examples of computer programs for predicting and correlating equi- librium data are described by Lo, Baird, and Hamon.[*] 354 Fermentation and Biochemical Engineering Handbook 3.0 SOLVENT SELECTION The molecular formula of the solute may suggest the type of solvent which may be selective for its extraction, based on probable affinities between related fbnctional groups. Thus, to extract organic acids or alcohols from water, an ester, ether, or ketone (of sufficient molecular weight to have very limited solubility in the aqueous phase) might be chosen as the solvent. The pH of aqueous phase feeds may also be very important. The sodium or potassium salts of an organic salt may well prefer the aqueous media at pH > 10, but in the acidulated form may readily extract into the organic phase if the pH is low. Specific factors taken into consideration in the selection of a solvent include: 1. Selectivity-the ability to remove and concentrate the solute from the other components likely present in the feed liquor. 2. Availability-the inventory of solvent in the extraction system can represent a significant capital investment. 3. Immiscibility withthe feed-otherwise there will need to be recovery of the solvent from the raffinate, or a continual and costly replacement of solvent as make up. 4. Density diflerential-too low a density difference between the phases will result in separation problems, lower capacity, and larger equipment. Too large a density difference may make it difficult to obtain the drop sizes desired for best extraction. 5. Reasonable physical properties-too viscous a solvent will impede both mass transfer and capacity. Too low an interfacial tension may lead to emulsion problems. The boiling point should be sufficiently different from that of the solute if recovery of the latter is to be by distillation. 6. Toxicity-must be considered for health considerations of the plant employees and for purity of the product. 7. Corrosiveness-may require use of more expensive mate- rials of construction for the extraction process equipment. Solvent Exiraction 355 8. Ease ofrecovepas transfer of the solute from the feed still entails the further separation of solute from the solvent, solvent recovery will need to be as complete and pure as possible to permit recycle to the extractor as well as minimizing losses and potential pollution problems. 4.0 CALCULATION PROCEDURES Sizing the equipment required for a given separation will depend upon both the flow rates involved and the number of stages that will be required. With a binary equilibrium plot, Fig. 7, the distribution of extract and raffinate following one stage of contact is readily determined. Representing a mass balance of the solute transferred: Thus, a line can be drawn from X,, with a slope of F/S to the intersection with the equilibrium line, thus establishing YE and X,. Y X Figure 7. Graphical solution for single contact. 356 Fermentation and Biochemical Engineering Handbook For multiple contact, Fig. 8, the operating line can be written around some point in the column between stage “n” and (n +l): X Figure 8. Graphical solution for multiple contact. Since liquid-liquid extraction frequently involves only a few stages, the above equation can be used for an analytical solution. The desired concentration of extract YE is set equal to &, and the ramate in equilibrium with the first stage, XI, is determined from the equilibrium curve. With this value ofX, , 4 is calculated from the above operating equation; then X, is determined from the equilibrium line and the calculation procedure is continued until X, 5 X,. Agraphical solution is also readily obtainable. The operating line, with slope F/S, is drawn from the inlet and outlet concentrations. The number of stages is then stepped off in the same fashion as with a McCabe Thiele diagram in distillation, as shown in Fig. 8. With a ternary equilibrium diagram, such as Fig. 5, the process result can be determined graphically, In Fig. 9, the addition of solvent to a feed containing X, solute will be along the straight line connecting S with XF. From an overall mass balance, the compositionMof the mixture of feed and Solvent Extraction 357 solvent is determined. WithMin the two-phase zone, the overall mixturekt separates along a tie line to end points YE andXR on the equilibrium curve. The relative quantities ofeach phase can be calculated using the inverse lever- arm rule. Figure 9. Graphical soh Solvent Fad Liqwr ion for single contact with ternary equilibrium & 8. With more than one contact, an operating point Q is located outside the ternary diagram, as shown in Fig. 10. With a specified solvent/feed ratio and a desired rafiate purity, XI, with the given feed, X, the composition of the final extract, Y,,, is fixed by material balance. Point Q is formed by the intersection of the line drawn from Y,, through X,, with the line drawn from the fresh solvent Y, through XI , Figure 10. Graphical solution for multiple contact. Point M in Fig. 9 represented the material balance: F+S=E+R=M 358 Fermentation and Biochemical Engineering Handbook Point Q in Fig. 10 represents a hypothetical quantity obtained by rearrangement of the above equation: F- E= R-S= Q The material balance for each stage is: Thus, a line through Q represents the operating line between stages. The number of stages is obtained by sequentially stepping off first the equilibrium distribution along a tie line, and then to the next stage by a line drawn from point Q through the rafiate to locate the next extract. 4.1 Simplified Solution If the distribution coeficient is constant, and if there is essentially no mutual solubility, the fraction not extracted, Y, can be calculated directly as a function of the extraction factor, E, and the number of stages, n. X,-Y,Im . mS Y= , E=- X, -Y, lm F Tre~bal[~l discusses the derivation ofthese equations and presents agraphical solution reproduced here as Fig. 1 1. Even when the two limitations of immiscibility and constant distribu- tion coefficient do not quite hold, Fig. 11 does allow a quick estimate of the trade-offs between solvent/feed ratio and number of stages required to obtain a desired degree of extraction (raffinate purity). The above solutions are all based on ideal or theoretical stages. Even in discrete stage systems, like mixer-settlers, equilibrium may not be attained because of insufficient time for diffusion of solute across the phase boundary or insufficient time for complete clarification of each stage. Solvent Extraction 359 Figure 11. Countercurrent multistage extraction with immiscible solvents and constant distribution coeffrcient. (From: Liquid Extraction by R. E. Treybal. Copyright0 1963, McGraw-Hill. Used with the permission.) In continuous differential extractors (columns) it has been convenient to think in terms of a height equivalent to a theoretical stage (HETS), and to correlate HETS as a function of system and equipment variables. Alter- nately, correlations may be obtained on the basis of the height of a transfer unit (HTU), which is more amenable to calculations which separately include the effects of ba~kmixing.[*1[~] 360 Fermentation and Biochemical Engineering Handbook 4.2 Sample Stage Calculation An aqueous waste stream containing 3.25% by weight phenol is to be extracted with one-third its volume of methylene chloride to produce a rafiate without more than 0.2% phenol. How many stages are required? Graphical Solution. Figure 12 is constructed using the equilibrium data forthe distribution of phenol between methylene chloride and water from Fig. 6. Figure 12. Stages for sample calculation. The operating line is determined on a solute free basis as follows: 3.25(100) - 3.36 g phenol 96.75 100 g water X, = - Solvent Exlraction 361 0.2(100) - 0.20 g phenol 99.8 100 g water x, = - Thus, per 100 g of water feed, the amount of phenol removed is: 3.36 - 0.20 = 3.16 g At a volumetric feed rate of solvent equal to one-third the feed, and a specific gravity of 1.3 1 for methylene chloride, the weight ratio of solvent to feed is: = 0.451 ws - 1 (1.31) W, 3 (1.0) (0.9675) -_- The phenol removed from the 100 g of water (3.16 g) must be in the extract, which contains 45.1 g of methylene chloride: 3.1 6 g phenol 45.1 100 g MeCl, YE =-(loo) = 7.01 The operating line is drawn from (3.36, 7.01) to (0.20, 0.00) in Fig. 12 and the stages stepped off. The stages are counted at the intersections with the equilibrium line. It is seen that the fourth stage produces a rafiinate with a value less than required. Thus, the number of theoretical stages is interpo- lated to be 3.8. Analytical Solution. The equation for the operating line is determined from the inlet and outlet concentrations. The operating line equation relates the extract concentration of one stage to the rafiinate concentrate from the previous stage. Y,,+l = 2.22 X,, - 0.444 Starting with YE, which is Yl for the first extraction stage, the raffinate Xl in equilibrium is determined from the distribution curve Fig. 6: 362 Fermentation and Biochemical Engineering Handbook at Yi = 7.01,XI = 1.43 Yz = 2.22 (1.43) - 0.444 = 2.73 X. = 0.784 from Fig. 6 X, = 0.42 from Fig. 6 X, = 0.150 from Fig. 6 Y, = 2.22 (0.784) - 0.444 = 1.30 Y4 = 2.22 (0.42) - 0.444 = 0.488 Since X4 is less than the observed X,. = 0.20, the fractional stage is estimated as follows: X3 - X,. - 0.42-0.20 0.22 X3 - X4 0.41 - 0.15 0.26 - - - -= 0.85 So the total number of stages is calculated to be 3.85. Short-Cut Solution. The curved equilibrium relationship means that the Treybal plot, Fig. 11, perhaps cannot be used. The required stages can be bracketed by calculating the extraction factor at each end ofthe extraction. At the dilute end: - 3.15 0.63 0.20 D=- E = ED = (0.45 1)(3.15) = 1.42 WF 0.2 3.36 Y = -= 0.060 n = 4.9 from Fig. 11 At the concentrated end: - 4.90 7.0 1 1.43 D=- Solvent Exfraction 363 E =(0.451)(4.90) = 2.21 n = 3.5 from Fig. 11 Using an average extraction factor of E = 1.8 1 , the number of stages from the Treybal plot is 4.1. The Treybal plot can be used to provide estimates for other require- ments as well. For example, ifit were desired to increase the amount ofphenol extracted from 94 to 99%, what increase in solvent flow or number of stages would be required? At E = 1.81, Y = 0.01, n = 6.3 stages At n = 4.1, Y = 0.01, E = 2.8 required Thus, the solvent flow would have to be increased by a factor of L& = 1.55 1.8 1 Thus, to increase extraction from 94 to 99% would require 57% more stages or 55% more solvent, or some lesser combination of both. 5.0 DROP MECHANICS An understanding of the performance of extraction equipment is furthered by an understanding of what may be going on inside individual drops. With the assumption of transfer of a solute A from a dispersed feed phase into a continuous solvent, as shown in Fig. 13, a concentration profile across the interface would appear to have a discontinuity (Fig. 14). The discontinuity is a consequence ofthe distribution coefficient, and reflects the general practice of choosing a solvent which has a greater preference for the solute than the feed phase has. If activities instead of concentrations were used, there would be no discontinuity at the interface. Transfer of solute across the interface can be assumed to be controlled by what happens through the immobilized films on both sides ofthe interface. Handles and BaronL5] have presented generalized correlations for the calcu- lation of the individual inside and outside coefficients for mass transfer across these films. 364 Fermentation and Biochemical Engineering Handbook 0 transfers from the drop into the continuous phase. &A ‘A A May k circulating. The drop may k oreillafk~. Figure 13. Drop mechanics. Interface Figure 14. Solute concentration at the interface. Small drops lead to more transfer area and better extraction, but to slower settling and less capacity. Thus, selection of extraction equipment fiquently involves a compromise choice balancing efficiency against capacity. The terminal velocity of liquid drops is the same as solid spheres when the diameter is small. The drag coefficient versus Reynold’s number can be recalculated to provide a diameter-free ordinate versus a velocity-free abscissa to facilitate direct solution, as shown in Fig. 15. With drops, a maximum velocity is attained, and this maximum has been correlated with a parameter based on physical properties of the system. The practical sequence of this phenomenon in column extraction is illustrated in Fig. 16. Drops larger than ds won’t travel any faster, so there is no capacity gain, and they have less specific area, so there will be an efficiency loss. Drops smaller than ds will result in more extraction by providing more transfer area and a longer contact time, but at the potential expense of lower capacity. Solvent Extraction 365 C, drag coefficient d drop diameter g gravitational constant Re Reynolds number Sd drop parameter u drop velocity pc continuous phase density 4 density difference & continuous phase viscosity 0 interfacial tension Figure 15. Dimensionless drop velocity vs. dimensionless drop diameter. Diameter Figure 16. Drop velocity vs. drop diameter. 366 Fermentation and Biochemical Engineering Handbook It is generally desirable to provide as uniform a drop size as possible. A wide range in drop sizes may allow the smaller drops to attain equilibrium, but they are en route longer, while the larger drops zip through, not attaining equilibrium. It is also considered desirable to allow drops to coalesce and be redispersed, as mass transfer from a forming drop is always higher than it is from a stagnant drop. Backmixing caused by flow patterns induced in the equipment can also deleteriously affect performance by reducing the driving force gradient, as illustrated in Fig. 17. Sleicherr61 presents procedures for calculating the consequences of backmixing on overall extraction results, Figure 17. Effect of backmixing on extraction driving force. 6.0 TYPES OF EXTRACTION EQUIPMENT Extraction equipment can be classified by function as providing discrete stages or continuous differential contact. Separation may be by gravity alone or by centrifugal force. Additional energy may be applied to control drop size, either by mechanical agitation or pulsation. This classification is shown in Table 1, along with major examples of available equipment. 6.1 Non-Agitated Gravity Flow Extractors Spray Column. The simplest differential extractor is the spray column (Fig. 1 Sa.), which depends upon the initial dispersion ofthe dispersed phase to create favorably sized droplets. There is no means provided to redisperse this phase if any coalescence occurs. Although the equipment is simple and inexpensive, it is difficult to obtain more than one stage extraction. The passage of the dispersed phase induces considerable backmixing of the continuous phase, particularly in larger diameter columns. Solvent Extraction 367 Table 1. Classification of Industrial Extraction Equipment Flow by Gravity alone Centrifugal Force Drop Sue control by Gravity alone Mechanical rotation Mechanical reciprocation Flow through baffles Stagewise Perforated Place Column Mixer-Settler Westfalia Extractor Robatel Extractor Continuous Differential Spray Column Packed Column RM= Oldshue-Rushton Column ARD Column Kuhni Column Raining Bucket Contactor Karr Column Pulsed Packed Column Pulsed Perforated Plate Column Podbielniak Extractor Alfa Lava1 Extractor a b C E F S R Figure 18. Non-agitated gravity flow extractors. (a) Spray, (b) packed, and (e) perforated plate. 368 Fermentation and Biochemical Engineering Handbook Packed Column. (Fig. 18b.) Interphasecontact can be improved in the spray column by providing extensive surface for coalescence and redispersion. This surface is provided with packing which provides surface while maintain- ing a large open area for flow, such as Raschig rings, Berl saddles, and variants thereof. There is some loss in capacity because of the cross section occupied by the packing, but this is more than offset by the gain in improved mass transfer and lessening of continuous phase backmixing. Packing should be chosen that preferentially is wetted by the continu- ous phase to discourage formation of rivulets of the dispersed phase bypassing through the column. In large diameter columns, redistribution trays should be installed to overcome potential channeling. Smaller packing size is generally more efficient, but restricts flow more, and is more prone to fouling by trapping solids. Eckertr7I summarizes design criteria for the selection of packing for packed columns. Perforated Plate Column. (Fig. 18c.) Sieve trays can be placed in the spray column to cause coalescence and redispersion of the dispersed phase. The trays can be designed to permit flow of both phases through the same perforations, but such trays generally have a quite narrow operating range. Generally, some sort of downcomer (or upcomer) is provided to allow a separate path for the continuous phase and one-way flow of the dispersed phase through the perforations. The density difference between the two phases and the height of coalesced phase provide the driving force for redispersion through the orifices. In contrast with vapor-liquid columns, tray efficiencies are very low (5 to 30%) in liquid-liquid systems. The trays do limit continuous phase backmixing as well as provide drop redispersion, but at the expense of reduced capacity. 6.2 Stirred Gravity Flow Extractors Provision of a shaft through the extraction column allows for repeated redispersion of the drops via various impellers located along the shaft. A variety of industrial equipment is available, with the differences being in the design of the impellers on the shaft for dispersion, and stators in the column for baffling and coalescence. Stirred columns offer the operator increased flexibility in operation by independent control over the dispersion process. RDC Column. The rotating disc contactor (Fig. 19) provides for redispersion by a series of discs along the shaft, combined with a series of fixed stators. Vortices are formed in each compartment, and the shear of the fluid against the rotor or stator causes the drop breakup. In many instances, Solvent Edraction 369 performance can be predicted from first principles, relating drop size to the energy input, and calculating slip velocity and mass transfer coefficients based on that diameter and the physical properties of the system (see Strand, Olney & Ackerman[*]). E S e R Figure 19. Rotating disc contactor (RDC). With increasing rotational speed, efficiency improves as drops become smaller, but maximum capacity is lessened. Increased rotational speed also increases continuous phase backmixing, and causes some segregation of the phases as the lighter phase accumulates around the shaft while the denser phase hugs the wall. At the same energy input, dispersingthe light phase leads to smaller drops because all of the light phase must pass over the tips of the spinning discs; whereas dispersion of a denser phase is brought about primarily by fluid motion over the stationary ring baffles. Oldshue-Rushton Column. This column is similar to the RDC, except that the flat rotor discs have been replaced with turbine type agitators (Fig. 20). As with the RDC, the diameter ofthe agitators can be varied along the shaft to compensate for the progressive change in the physical properties of the system as extraction occurs. Other variations of stirred columns which are available include the asymmetric rotating disc (ARD) contactor, the Kiihni column, and two types of Scheibel columns. The rotor of the ARD is located off center, which permits more elaborate baffling for the necessary transport of flows with less backmixing. 3 70 Fermentation and Biochemical Engineering Handbook n E S K Figure 20. Oldshue-Rushton column. The Kuhni column employs radial flow impellers located between perforated plates for compartmentalization. The first Scheibel column used wire mesh zones to promote coalescence and limit backmixing between turbine-agitated mixing zones. A later Scheibel column used a shrouded radial impeller and multiple ring baffles to direct most of the rotor’s energy towards dispersion and away from axial mixing. Raining Bucket Contactor. This contactor consists of a series of scoops located on a slowly rotating, baffled rotor within a horizontal cylindrical vessel (Fig. 2 1). An interface is maintained near the middle, and the scoops capture and then allow one phase to rain through the other, and vice versa, once each revolution. Figure 21. Raining bucket contactor. Solvent Extraction 371 There is little, if any, control of droplet size, but the raining bucket contactor is the only one that disperses each phase in the other. If the flow ratio differs greatly from unity, backmixing of the low flow phase can be serious, and line out with changed operating conditions can take a long time. 6.3 Pulsed Gravity Flow Extractors Liquid Pulsed Columns. The liquid in a packed or perforated plate column may be pulsed to promote better mass transfer (Fig. 22). If a sieve plate column is pulsed, downcomers are no longer required. Pulsing can be caused by a piston pump or by air pulsing external tothe column. Frequencies are generally 1 to 3 Hz and amplitude up to 20 mm. Drop size is dependent upon the product of amplitude times frequency. As this product is increased, the smaller diameter drops so produced lead to more holdup and better mass transfer, but to a fall off in capacity. Eventually, at a high enough amplitude x frequency product, backmixing increases to the extent that efficiency also begins to diminish. E Perforated Plater Figure 22. Liquid pulsed columns. Mechanically Pulsed Column. The Kurr column (Fig. 23) consists of perforated plates ganged on a common shaft which is oscillated by an external drive. The perforated area and hole size are much larger than in typical sieve plate operation. At high amplitudex frequency product in larger columns, the tendency for excessive backmixing can be curtailed by instal- lation of some fixed baffles. 372 Fermentation and Biochemical Engineering Handbook Figure 23. Karr reciprocating plate column. Mixer-Settlers. The extraction fhction of bringing feed and solvent intimately together, and then allowing them to separate is frequently done in mixer-settlers. The fimctions may be done in separate vessels, or in different portions of the same vessels, or sequentially in the same vessel on a batch basis. As noted earlier in Fig. 3, the flows of feed and solvent can be countercurrent to each other through a series of mixer-settlers. Sizing of the mixer is based upon providing sufficient agitation and sufficient residence time to allow equilibrium to be approached, and thus will depend upon the flows to be processed as well as the physical properties of the two liquids. Since some extractions actually involve a chemical reaction, the time of contact can be very important. If, for reasons of improved mass transfer, it is desired to disperse the high flow phase, it may be necessary to recycle some of the low flow phase to keep an appropriate phase ratio in the mixer different than the feed flow ratio. The settler must provide a long enough quiescent residence time for the emulsion which is produced in the mixer to break, and a low enough lineal velocity for the two phases to become essentially free of entrainment. In some instances, coalescing material, such as wire mesh, may be installed to lessen entrainment, however, such material should be used with some caution because of the tendency for fouling by accumulation of foreign material. It is frequently possible to introduce one of the phases into the eye of the impeller, and thus be able to pump one entering fluid while the other flows by gravity from the next upstream and downstream stages, without the need for separate interstage pumps. Solvent Extraction 3 73 6.4 Centrifugal Extractors Many of the commercial extraction processes encountered in the pharmaceutical industry involve systems which emulsifjr readily and are exceedingly difficult to separate cleanly. Stability of the solute may also be a factor, and rapid separation may be required to prevent degradation and loss of the product. Centrifigal extractors fill an important niche for just such problems. The most common centrifigal extractor is the Podbielniak@ Contactor, as shown in cutaway view in Fig. 24. Essentially it is a sieve plate column that has been wrapped around a shaft and spun to create a multigravitational force to do both the redispersion and the separation. All fluids enter and leave through shaft passageways and mechanical seals. The performance of centrifugal extractors has been described by Todd and Davies in general detail[g] and specifically for pharmaceutical use.[10] The primary benefits of centrifigal extractors accrue from their compactness and superior clarifying capabilities. Solvent inventory can be held to a minimum. Centrifigal extractors are also particularly appropriate handling high phase ratios, as the low flow phase can be kept continuous without much backmixing, thereby allowing the large flow fluid to be dispersed to provide more mass transfer area. '\DRAIN Figure 24. Podbielniak" centrihgal extractor. 374 Fermentation and Biochemical Engineering Handbook The Podbielniak and Alfa Lava1 centrifugal extractors are essentially continuous differential contactors. The Westfalia and Robatel centrifugal extractors contain discrete mechanical stages, and flow from one to another is effected by spill over discs and skimmers according to usual centrifbgal clarifier practice. As the number of discrete stages is increased, the allowable flow rates are proportionately decreased. 6.5 Equipment Size Calculation Agitated Columns. The size of an extraction column frequently can be estimated from a knowledge of the flow rates and physical properties, combined with some empirical generalizations. 1. The maximum capacity (at zero stirrer speed or pulsation) is directly related to the terminal velocity of the dispersed phase through the minimum physical constriction in the column. 2. The terminal velocity of the dispersed phase droplets is related to the physical properties of the system by the correlation shown in Fig. 15. 3. For many systems, the effect of hindered settling can be approximated by: yt=- (5.5) 1-h h 1-h where 6, V, are the superficial lineal velocities of the drop, dispersed phase, and continuous phase, and h is the holdup. 4. Agitated columns are frequently operated so that the capacity is half what it would be at no agitation (zero rpm or pulsation). Agitation is used to reduce droplet diameter to this equivalent point to increase mass transfer rate and mass transfer area. 5. For sizing purposes, the diameter of the column will be chosen so that the column is operating at 75% ofthe flood point. Solvent Exfraction 3 75 6. The holdup at flooding can be determined by differentiat- ing the equation in criterion #3. Combining this relation- ship with all the constants leads to the following equation: 0.138 0.88 D=-Q:5 0 09 ($) [+L] B Pc A, (with D in meters, Qdin m3h, ,uc in poise, ain dyneslcm, and p in g/cc). The factor B is related to holdup and dependent upon phase ratio, as shown in Fig. 25. 0.1 1 10 adad Figure 25. Empirical constant B for determining agitated column diameter. The countercurrent contact zone height will depend primarily upon the number of stages required (n) and the column characteristics. The effect of backmixing also increases the column diameter. A reasonable first approxi- mation of extraction height (L) required for agitated columns is: L = 0.g4~fi (L'and D in meters) Most columns also require clarifying zones at each end to provide for coalescence and to minimize entrainment. These zones also are dependent upon column diameter. The combined height required (Z) for the clarifjing zones can be approximated by: Z=3G (2 and D in meters) 3 76 Fermentation and Biochemical Engineering Handbook Estimates of column size required for three different cases are tabu- lated in Table 2. Case A involves the removal of dioxane from a benzene stream with water as the extracting solvent. Case B involves the recovery of methyl ethyl ketone from a heptane stream with water. Case C is for the removal of phenol from an aqueous stream with methylene chloride. In addition to the calculated heights and diameters, the total traffic flow (the sum of both flows divided by the column cross-section) is listed. Typical traffic flows for agitated columns are in the 25 to 100 m3/m2 hr range. Table 2. Examples of Column Sizing Calculations Cm Remove solute from feed with solvent Flow QC Ratio Qc/Qd Constant E Int. tens. (I Spec. grav. pc Viscosity PC Pd AP QJ'S Diameter D Theo. stages, n Ht, contact L clarif. z Total H Traffic flow d/hr d/hr poise dynelcm rn m m m/hr A Dioxane Benzene (c) Water (d) 20.6 13.3 1.54 0.356 0.0065 30 0.884 1 .oo 0.1 16 3.65 0.476 2.61 1.146 4 4.03 3.21 7.24 32.9 B MEK Heptane (d) Water (cl 30.9 16.4 1 .Ea 0.344 0.010 1 .oo 0.312 4.06 0.477 45 0.688 2.22 1.122 6 5.97 3.17 9.14 47.9 C Phenol Water (d) MeCh (c) 7.6 22.7 0.33 0.447 0.007 1.31 1.00 0.31 2.75 0.462 45 2.51 0.642 4 3.01 2.40 5.41 93.6 Note: (c) = continuous phase; (d) =dispersed phase. Solvent Exfraction 3 77 The manufacturer of the extraction column will likely select the next larger diameter size for which he has standardized components. He may also insist upon some pilot plant test to confirm the capacity and efficiency requirements. The manufacturers of other proprietary extraction devices, such as centrifugal extractors, will be able to provide estimates of the probable size equipment required, based on comparisons with similar systems and their own accumulated design experience. Many pharmaceutical extractions do not lend themselves to simple straightforward analytical solutions. Rarely is there a case of simple extraction of a single solute from a clean feed with pure solvent. There may well be solids present which can stabilize emulsions and cause excessive entrainment. Usually, more than one solute is present, so selectivity as well as extent of extraction becomes important. Also, the solvent may contain residual solute from the solvent recovery section. Again, suppliers of extraction equipment should be contacted for their help in solving real industrial extraction problems. Packed Columns. Capacity of packed columns is strongly dependent upon the packing being used. As the surface area of the packing is increased to improve efficiency, in general, both the hydraulic radius and the fraction void decrease, thereby increasing resistance to flow and lowering capacity. For a given extraction, the maximum capacity (flooding rate) generally follows the form: where Kis a function of packing characteristics and physical properties of the system. Compared to agitated columns, both diameter and height will have to be larger. Flow redistributers are advisable at periodic intervals to offset the tendency for channeling and bypassing frequently encountered in packed columns. Characteristics of various packings and correlations for capacity and stage height are given by Treybalr31 and E~kert.[~] Mixer-Settlers. The mixing required for adequate dispersion can be determined and scaled-up by the methods outlined by Oldshue.["] Sizing of settlers poses some uncertainty in that solvent recycle within the process may lead to accumulation of an interfacial vag, which tends to stabilize emulsions at the interface. For a first approximation, an arbitrary residence time, like 20 minutes, might be assumed unless bench shake-outs indicate an even longer time required for adequate clarification. 378 Fermentation and Biochemical Engineering Handbook Proprietary Extractors. Manufacturers or proprietary design ex- traction equipment (such as the Podbielniak Centrifkgal Extractor or the RTL (raining bucket) Contactor) provide catalogs listing the relative capaci- ties of the various sizes of equipment which are offered. Pilot equipment is usually available for determining extraction performance, and the manufac- turer utilizes both the pilot data and experience with similar systems to provide assured commercial designs. 7.0 SELECTION OF EQUIPMENT The choice of extraction equipment should be based on the minimum annual cost for the complete package of extractor and accessory equipment, including operating and solvent loss costs. In addition to the requirements of processing so much feed and solvent with a required number of theoretical stages, there are the practical consid- erations concerning contamination, entrainment, emulsification, floor space, height requirements, cleanability, and versatility to handle other than design rates. The suitability of various type extractors with respect to each of these considerations is listed in Table 3. Not all of the features compared in the table can be equated. The tabulation is provided to show comparisons to aid in the selection of suitable equipment. Table 3. Extractor Selection Chart . LOW Cost . High Oper- Effi- Capital ating ciency Mixer settler 324 Spray 451 Perf. plate 4 5 2 Packed 442 Pulse 334 Agitated 3 4 4 Centrifugal 2 3 4 5 is outstanding 4 is good 3 is adequate Total Through- put High Volu. Lowest metric . Space, Flex- Effi- Verti- lbility ciancy cal Floor 4 251 2 115 2 214 2 214 4 434 4 434 4 555 2 is fair 1 is poor 0 is unsuitable Ability to Cope with Systems Which with Emulsify Solids 1 3 2 4 3 2 3 1 3 3 3 3 5 2 Solvent Extraction 3 79 Other criteria for the selection of an extractor are the ease of separation of the two phases and the difficulty of extraction. For example, if the two phases have a large density difference, or at least one is quite viscous, the energy required to get a good enough dispersion for good extraction may lead to excessive backmixing of the continuous phase. The extractor selection map depicted in Fig. 26 reflects the above considerations plus the number of stages required. Where the degree of extraction exceeds the probable maximum staging achievable in one extrac- tor, the extractors can be used in series. 7 6 NUMBER 5 OF STAGES REQUIRED 4 3 2 1 EASY DIFFICULTY DIFFICULT OF EXTRACTION Figure 26. Extractor selection map 8.0 PROCEDURE SUMMARY Liquid-liquid extraction should be considered as a desirable route for product recovery and purification along with fractional crystallization and distillation. The ability to make separations according to chemical type, rather than according to physical properties such as freezing point or vapor pressure, is one of extraction’s major attractions. Energy frequently can be 380 Fermentation and Biochemical Engineering Handbook saved in the recovery of valuable products from dilute broth solution since a small quantity of a selective solvent can be used, and recovery from the concentrated extract is then facilitated. Selectivity of potentially attractive solvents can frequently be deter- mined from simple shake-outs over the desired concentration range. From these distribution data, the combinations of amount of solvent and number of theoretical stages can be calculated. Suppliers ofextraction equipment will likely wish to participate in pilot testing to confirm the correlations for capacity and efficiency of the specific equipment being considered. After installation, the equipment suppliers can also provide technical assistance in bringing the extraction equipment on line and solving problems which may arise from the commercial plant operation with its potential variation in feed and solvent quality and accumulation of impurities. 9.0 ADDITIONAL INFORMATION With Treybal's book['] essentially out of print, the Handbook of Solvent Extraction by Lo, Baird, and HansonL2] provides a most comprehen- sive reference. In addition to the previously cited Perry's Handbook chapter on liquidextraction byRobbins,['] The EssentialsofExtraction by Humphrey, Rocha, and Fair,[12] and a three part A Fresh Look at Liquid-Liquid Extraction, [131 provide briefer, but very useful guidelines. Details of extraction processes specifically involved in pharmaceutical production have been described by King et al.,[14] and by Kroner, Hustedt, and K~la.['~] REFERENCES 1. Robbins, L. A,, Sec. 15, Liquid-Liquid Extraction, Perry's Chemical EngineersHandbook, 6thEd. (R. H. Peny, D. W. Green, J. 0. Maloney, eds.), McGraw Hill, New York (1984) Lo, T., Baird, M. H. I.,Hanson, C.,HandbookofSolventExtraction, John Wiley & Sons, New York (1983) Treybal, R. E. LiquidExtraction, 2nd Edition, McGraw Hill, New York (1963) Hanson, C. (ed), Recent Advances in Liquid-Liquid Extraction, Perga- mon Press, Oxford ( 197 1) 2. 3. 4. Solvent Exiraction 381 5. 6. 7. 8. 9. 10. 11. 12. 13. 14 15. Handlos, A. E. andBaron, T., Mass and heat transfer from drops in liquid- liquid extraction, AIChE. Journal, 3: 127-135 (1957) Sleicher, C. A., Axial Mixing and Extraction Efficiency, AIChE Journal, Eckert, J. S., Extraction Variables defined, Hydrocarbon Processing, Strand, C. P., Olney, R. B., and Ackerman, G. H., Fundamental Aspects of Rotating Disk Contactor Performance, AIChE Journal, 8:252-261 (1962) Todd, D. B. and Davies, G. R., Performance of Centrifugal Extractors, Proceedings ISEC, 74, 3:2379-2398 Todd, D. B. and Davies, G. R., Centrifugal Pharmaceutical Extractions, Filtration & Separation, 10(6)663-666 (1973) Oldshue, J. Y., Mixing, Handbook of Fermentation Engineering. Humphrey, J. L., Rocha, J. A., and Fair, J. R., The Essentials of Extraction, Chem. Engr., 91(18):76-95 (Sept. 17, 1984) Cusack, R. W., Fremeaux, P., Glatz, D., and Karr, A., A Fresh Look at Liquid-Liquid Extraction, Chem. Engr., 98(2):66-76 (Feb. 1991); 98(3):132-138 (Mar. 1991); 98(4):112-120 (Apr. 1991) King, M. L., Forman, A. L., Orella, C., and Pines, S. H., Extractive Hydrolysis for Pharmaceuticals, CEP, 81(5):36-39 (May 1985) Kroner, K. H., Hustedt, H., and Kula, M. R., Extractive Enzyme Recov- ery: Economic Considerations, Process Biochemistry, 19: 170-179 (Oct. 1984) 5:145-149 (1959) 55(3):117-124 (1976)