8 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 concentration X. 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 I, but usually y is essentially 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 refined so the feed-rich liquid leaving the extractor is the raffinate(r) When the feed and solvent are brought together, the solute (a)will distribute itself between thetwoliquid phases. Atequilibrium, the ratio ofthis distribution is called the distribution coefficient(m)
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 extraction 349 Xa concentration of A in raffinate phase The distribution coefficient, m, is a measure of the affinity of the solute (A) for one phase(E, S)over the other phase(F, R). The concentration of A 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-free wate ight aceto Although the units of m appear to be dimensionless, they actually (weight acetone-free water /weight acetone-free toluene) If more than one solute is present, the preference, or selectivity, of the solvent for one(A)over the other( B)is the separation factor(a) aar The separation factor(aaB)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 of A 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
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 de Even though extraction may successfully remove the solute from the feed, a further separation is required in order to recover the solute fro 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 further 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 of chemi- cal differences, but where relative volatilities overlap 4. Where extraction selectivity is favorable inionic 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 distr The combinations ofmixing both feed and solvent until the equilibrium ibution 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)
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 temperature 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 ofchemical 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 separation 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 =ITHEORETICAL ING TO PHASE SEPARAT ION EQU IL1B Figure 1. Theoretical stage Feed Figure 2. Differential extraction Feed 心 Raffinal Figure 3. Sequential contact and separation
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 arying quantities of solute in a separatory funnel, 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 useful For later ease of calculation, it is desirable to express concentrations on a solute-free basis. If there is extensive miscibility, a termary plot(Fig. 5)would be preferable. Tie lines represent the equilibrium between the coexisting phases HAc Figure 4. Binary plot of distribution data Figure 5. Ternary plot of distribution data Plotting the data on log-log graphs may be helpful 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
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 concern that many industrial extraction processes involve not only 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 Limin L imati Figure 6. Distribution of phenol between water and chlorinated methanes Whenever the distribution coefficient is greatly different than unity thereis an implication that there exists an affinity of the 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 Hanson. [2
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 equilibrium 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 which may beselective for its extraction, based on probable affinities between related functional 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 oH 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 solver Selectiviry-the ability to remove and concentrate the olute from the other components likely present in the feed 2. Availability-the inventory of solvent in the extraction system can represent a significant capital investment 3. Immiscibility with the feed-otherwise there will need to be ecovery of the solvent from the raffinate, or a continual and costly replacement of solvent as make 4. Density differential-too low a density difference between the phases will result in separation problems, lower capacity, and larger equipment. Too large a density ifference 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
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 materials of construction for the extraction process equipment
Solvent Extraction 355 8. Ease of recovery-as 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 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 (Ys -YE S=(XF-XRF ( (X x Thus, a line can be drawn from XF, with a slope of F/s to the intersection with the equilibrium line, thus establishing YE and XR Figure 7. Graphical solution for single contact
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 aroune ome point in the column between stage"n"and (n+l) S(Yn 1 -s)=F(X,-XR) (x;-x)=5(xn)-5(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 Y affinate in equilibrium with the first stage, Xi, is determined from the equilibrium curve. With this value of i, I, is calculated from the above operating equation; then X2 is determined from the equilibrium line and the calculation procedure is continued until Xn sX A graphical 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 termary equilibrium diagram, such as Fig. 5, the process result can be determined graphically. In Fig. 9, the addition of solvent to a feed containing XF solute will be along the straight line connecting S with XF From an overall mass balance, the composition m of the mixture of feed and
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. with M in the two-phase zone, the overall mixture M separates along a tie line to end points Ye and Xr on the equilibrium curve The relative quantities ofeach phase can be calculated using the inverse lever- arm rule Feed Liquor Figure 9 Graphical solution for single contact with temary equilibrium data with more than one contact, an operating point Q is located outside the temary diagram, as shown in Fig. 10. With a specified solvent/feed ratio and a desired raffinate purity, X,, with the given feed, Xp the composition of the final extract, In, is fixed by material balance. Point Q is formed by the intersection of the line drawn from I, through XF, with the line drawn from the fresh solvent Is through X, Figure 10. Graphical solution for multiple contact Point M in Fig 9 represented the material balance F+S=E+R=M
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 leverarm 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