Liquid-liquid (or solvent) extraction is a countercurrent separation process for isolating the constituents of a liquid mixture. In its simplest form, this involves the extraction of a solute from a binary solution by bringing it into contact with a second immiscible solvent in which the solute is soluble. In practical terms, however, many solutes may be present in the initial solution and die extracting ‘solvent’ may be a mixture of solvents designed to be selective for one or more solutes, depending upon their chemical type.
Solvent extraction is an old, established process and together with distillation constitute the two most important industrial separation procedures. The first commercially-successful liquid-liquid extraction operation was developed for the petroleum industry in 1909 when Edeleanu’s process was employed for the removal of aromatic hydrocarbons from kerosene, using liquid sulfur dioxide as solvent. Since then many other processes have been developed by the petroleum, chemical, metallurgical, nuclear, pharmaceutical and food processing industries.
Whereas distillation affects a separation by utilizing the differing volatilities of the components of a mixture, liquid-liquid extraction makes use of the different extent to which the components can partition into a second immiscible solvent. This property is frequently characteristic of the chemical type so that entire classes of compounds may be extracted if desired. The petroleum industry takes advantage of this characteristic of the process and has used extraction to separate, for example, aromatic hydrocarbons from paraffin hydrocarbons of the same boiling range using solvents such as liquified sulfur dioxide, furfural and diethylene glycol. In general, extraction is applied when the materials to be extracted are heat-sensitive or nonvolatile and when distillation would be inappropriate because components are close-boiling, have poor relative volatilities or form azeotropes.
The simplest extraction operation is single-contact batch extraction in which the initial feed solution is agitated with a suitable solvent, allowed to separate into two phases after which the solvent containing the extracted solute is decanted. This is analagous to the laboratory procedure employing a separating funnel. On an industrial scale, the extraction operation more usually involves more than one extraction stage and is normally carried out on a continuous basis. The equipment may be comprised of either discrete mixers and settlers or some form of column contactor in which the feed and solvent phases flow countercurrently by virtue of the density difference between the phases.
Final settling or phase separation is achieved under gravity at one end of the column by allowing an adequate settling volume for complete phase disengagement.
Any one extraction operation gives rise to two product streams: the extracted feed solution, more usually termed the raffinate phase, and the solvent containing extracted solute termed the extract phase. This nomenclature is unique to liquid-liquid extraction processes and will be used from hereon.
No single criterion can be used to assess the suitability of a solvent for a particular application and the final choice is invariably a compromise between competing requirements. Thus not only should the solvent be selective for the solute being extracted but it should also possess other desirable features such as low cost, low solubility in the feed-phase and good recoverability as well as being noncorrosive and noninflammable. Furthermore, interfacial tension between the two phases should not be so low that subsequent phase disengagement becomes difficult and the density difference between the phases should be large enough to maintain countercurrent flow of the phases under the influence of gravity.
Of these factors, the first to be considered is the selectivity of the solvent, or the ease with which it extracts the desired solute from the feed stream. This is most readily understood by considering a simple ternary system consisting of a solution of solute C in a solvent A (the feed solution) and an extracting solvent B, which is designed to extract C from A. A simple single-stage extraction is shown on conventional triangular coordinates in Figure 1a. Here, a mixture of A and C of composition F is mixed with a pure solvent B in such proportions as to give an overall composition M. This lies inside the miscibility curve and so the mixture will separate into two separate phases, R and E, joined by an equilibrium tie line RE. If the solvent B is now stripped from each phase, the solvent-free composition of R is given by D and the solvent-free composition of E by the point G. Both solvent-free compositions lie, of course, on the side of the triangle AC and it will be seen that the initial feed solution of composition F has been separated into two solutions D and G, which have low and high concentrations of C, respectively.
If this operation is now repeated using another solvent B' , the corresponding concentrations may be as shown in Figure 1b. In this instance, the solvent-free concentrations D and G are closer to the initial feed concentration F, and the separation of C is not as good as in the first case. It will be noted that the two solvents B and B' are associated with equilibrium tie lines of very different slopes, and effects of this nature may be quantified by defining a solvent selectivity βCA, analogous to relative volatility in distillation, such that:
Since X/XCA is the partition coefficient of the system, m,
In most instances, β varies widely with concentrations; and for practical purposes, a solvent should be selected that gives high values of β in excess of unity and satisfies the other criteria listed above.
The extraction of aqueous solutions is usually carried out using organic solvents or mixtures thereof. In recent years, interest has developed in the possibilities of using a second aqueous phase loaded with a suitable polymer so the extracted solute does not come into contact with organic solvents. This is of particular interest to the pharmaceutical and foodstuffs industries [see Verrall (1992) and Hamm (1992) for details of such aqueous-aqueous systems].
The first step in the design of any extraction process is the determination of the equilibrium relationships between the feed solution and the proposed solvent. This enables the suitability of the solvent to be assessed in terms of its selectivity, as well as the calculation of the numbers of extraction stages required for any set of flow conditions and degree of separation.
Equilibrium data are usually determined directly in the laboratory since such measurements are more accurate than values calculated from predictive equations. Equilibria may be represented graphically on either triangular or rectangular coordinates, and a full discussion of the determination and representation of liquid-liquid equilibria has been presented by Treybal (1963). The correlation of equilibrium data is best achieved in terms of activity coefficients calculated from laboratory equilibrium measurements. A large number of semi-empirical equations are available for this purpose, but two models have found wide acceptance: the NRTL and the UNIQUAC equations for nonelectrolytes. In the absence of experimental data, it is not possible to determine the parameters of these equations and one must turn to purely predictive models, such as the regular solution and the UNIFAC models. All these procedures, as well as correlations for electrolyte solutions, have been discussed in detail by Newsham (1992) and this source should be consulted for further information.
Contactors or extractors are specialized items of equipment designed to bring the feed and solvent phases together in such a manner that rapid transfer of the solute takes place from one phase to the other, followed by subsequent phase separation. In practice, efficient extraction involves four separate requirements:
The initial dispersion of one phase into the other in the form of droplets.
The maintenance of a fine dispersion in order to provide a large interfacial area for diffusion from one phase to the other.
The provision of an adequate holding or retention time for an acceptable level of diffusion to take place.
Final separation of the dispersion into raffinate and extract phases.
Numerous contactors have been described in the literature and the characteristics of the principal types have been summarized by Pratt and Stevens (1992). These authors also discussed the selection, design and scale-up of industrially-relevant contactors.
In its simplest form, a contactor merely consists of a stirred tank in series with a settling chamber through which the two phases flow (Figure 2a). Such arrangements are termed mixer-settlers and a number of units may be assembled in cascade to give the required degree of extraction. A typical assembly with countercurrent phase flows is shown in Figure 2b. Such devices can become uneconomical when a high level of extraction is called for because of the multiplicity of units employed. Each calls for separate stirrers and instrumentation for interface control in each settler, and it is more usual to employ some form of ‘column’ contactor in which only one settling chamber is involved, irrespective of the degree of extraction required.
Figure 2. Contactor arrangements. (a) Single-stage mixer-settler. (b) Countercurrent multiple contact using mixer-settlers. (c) Spray column. (d) Packed column. (e) Rotating disc column. (f) Air-pulsed plate column. (g) Electrostatic column. (F = feedstream; S = solvent; R = raffinate; E = extract. The feedstream is assumed to be the heavier phase throughout.)
The simplest column contactor is the spray tower (Figure 2c) in which one phase is dispersed into the other and overall flows are countercurrent through the column. Such units are inefficient and are of little interest outside the laboratory. If however some form of ordered or random packing, such as raschig rings, is introduced into the tower, extraction efficiency is increased several-fold. Such packed columns (Figure 2d) are an important item of industrial equipment. The packing not only reduces the gross back-mixing evident in the spray tower but also serves to establish a controllable droplet size distribution, as well as inducing additional turbulence inside and outside the droplets so diffusion from one phase into the other proceeds more rapidly [Batey and Thornton (1989)].
In column contactors described so far, energy available for droplet dispersion (and hence, the interfacial area available for solute transfer) is derived solely from the density difference between the phases; so the physical properties of the system set a limit to achievable extraction efficiency. This limitation may be overcome by supplying additional energy to the contactor and this concept has given rise to a large variety of so-called mechanical columns. Columns with coaxial rotating members of various designs have been described in the literature and the so-called rotating disc contactor illustrated in Figure 2e is a good example. This is basically a spray column with a central rotating shaft bearing a series of flat discs that rotate between fixed annular baffles. The shear forces set up produce very small droplets of dispersed phase and a correspondingly large interfacial area for mass transfer. Whilst such a unit gives good extraction efficiencies, the rotating shaft involves seals or bearings within the column and is therefore unsuitable for processing toxic or corrosive liquids.
This limitation relative to corrosive liquids may be overcome if mechanical energy is introduced in the form of reciprocatory, rather than rotary, motion. Such contactors are known as pulsed columns, and the reciprocatory motion may be applied either to the plates in the column or to the process fluids themselves. The latter procedure is now virtually universal. A typical pulsed column is illustrated in Figure 2f and consists, in essence, of a column shell fitted with a number of fixed perforated plates or sieve trays. The pulse may be imparted to the process fluids by attaching a cylinder closed by a reciprocating piston to the base of the column, or more usually by applying a sinusoidally-varying air pressure to a vertical standpipe connected to the base of the column (Figure 2f). Such a device is known as an air-pulsed column [Thornton (1954)] and has the advantage that the process fluids are isolated from the pulsing mechanism by a pocket of air or inert gas. Such contactors have found wide application in the nuclear reprocessing industries. Column contents are usually pulsed sinusoidally at a frequency within the range 1–3 cycles per second and with an amplitude, measured in the column, of 12 mm or less. The perforated plates typically have a free area of 25% and are drilled with 3 mm diameter holes on a triangular pitch; the spacing between successive plates is 50 mm. Such a plate geometry does not allow the dispersed phase droplets to pass through readily except under the influence of the pulse, and very small droplets giving rise to large interfacial areas are readily obtained by varying the pulse frequency and/or amplitude. The extraction efficiency of the unit is therefore easily varied by changing the pulse characteristics and very high rates of extraction may be obtained with these columns. From an industrial viewpoint, mixer-settlers, rotating disc and pulsed columns have been employed successfully in a wide range of situations and extensive performance data are available in published literature.
Mechanical energy is not the only method of producing small droplets and thereby large interfacial areas for solute transfer. On the laboratory scale, both sonic and electrical energy have been employed successfully. Thus a sonic generator in contact with process fluids in a spray-type column gives rise to good droplet dispersions and high extraction rates [Thornton (1953)].
A promising procedure developed in recent years is electrostatic extraction, wherein electrical energy is employed to effect dispersion of one phase into the other by charging the dispersed phase entry nozzle to a high potential relative to a secondary electrode downstream in the column [Thornton and Brown (1966); Stewart and Thornton (1967)]. The droplets formed at the entry nozzle are now very small and carry an electrical charge so that they are accelerated at high velocity towards the secondary electrode. Furthermore, since the droplets carry a charge they oscillate rapidly due to the lowered interfacial tension, and contactors can be designed with very small contact times coupled with high extraction rates comparable with those usually associated with pulsed-plate columns. Rapid extraction coupled with low retention times in the contactor are particularly appropriate to processes in which the solute is of biological origin, or is otherwise unstable during extraction operation. In practice, such electrostatic contactors comprise a column equipped with a number of insulated nozzle trays with a potential gradient between successive trays (Figure 2g). Designs for large-scale units have not yet been investigated in any detail, but the use of radial nozzle trays for larger diameter columns has been proposed [Thornton (1989)]. For a general discussion of electrostatic extraction, see Thornton (1976) and Weatherley (1992).
The design of a column-type contactor basically involves the calculation of two geometrical parameters, viz., the height of column necessary to give the required degree of extraction and the diameter to handle the necessary flow rates under prescribed operating conditions. In the design of stage-wise units, such as mixer-settlers, the corresponding parameters are power input and residence time of the mixing chamber and the residence time of the settler necessary for satisfactory phase setting.
The fundamental quantities usually employed to describe the hydrodynamic behavior of column-type contactors are the phase flow rates, fractional holdup of the dispersed phase and the characteristic mean velocity of the droplets. If the fractional voidage of the column is denoted by ε, the quantities are related by the holdup equation:
The term () on the right hand side of Equation (1) takes into account the reduction in mean velocity of a multiplicity of droplets by comparison with the velocity of a single droplet in infinite media. This so-called hindered rising term can frequently be represented by a function of holdup (1 – x) so long as the droplet size is small and is independent of the phase flow rates—conditions frequently met in mechanical contactors. On this basis, the holdup equation takes the form:
A plot of [Vd + (x/1 – x)Vc] versus x(1 – x) is linear with a slope of and enables characteristic velocity to be determined from flow rate and holdup measurements. Progressive increases in either or both phase flow rates finally result in flooding of the contactor, which is manifested by the appearance of a second interface at the opposite end of the column to the main interface. This condition corresponds to maximum values of the flow rates beyond which holdup remains constant, and can be found by differentiating Equation (2) [Thornton and Pratt (1953)].
where L represents the ratio Vdf/Vcf and subscript f denotes values at the flood point. Thus the phase superficial velocities at the flood point, together with the associated dispersed phase holdup, may be determined from Equations (3)– (5) once the value of the droplet characteristic velocity is known; the latter is readily obtained by plotting experimental holdup measurements in accordance with Equation (2) and measuring the slope of the linear plot. By this means flood point data may be obtained from holdup measurements and vice versa [Thornton and Pratt (1953); Thornton (1956)]. It is important to note that Equations (2)–(5) are only appropriate for situations where mean droplet size is small and is constant up to the flood point. Furthermore the assumption is made that the effective buoyancy force between the phases is proportional to the difference in densities of the mixture and the dispersed phase (ρm – ρd). By the mixture law, this is equal to (ρc – ρd)(1 – x), thus accounting for the (1 – x) term on the right hand side of Equation (2). At higher Reynolds numbers, it is possible, in principle, to derive equations analogous to Equations (2)–(5) provided that satisfactory equations can be formulated to describe droplet motion in the column. This is not always possible at present and numerous semi-empirical expressions have been proposed in lieu of Equation (2) [Pratt and Stevens (1992)]. It should be noted that Equations (3)–(5) cannot be used to predict flooding rates in packed columns since droplet coalescence sets in prior to the flooding point.
Relationships can be established between the Sauter mean droplet size, dvs, and the characteristic velocity by introducing the concept of column impedance, I. Thus in the case of a spray tower of height H, let the time taken for a droplet to move through this distance be t. The terminal velocity of the droplet, U, is then equal to H/t. In a contactor such as a perforated plate-column, the same size droplets will take longer than time t to pass through a height H because they will suffer a small but finite delay, Δt, at each plate. In a column of N plates, the total delay will amount to NΔt and the velocity of the droplets relative to the continuous phase will be given by H/(t + NΔt). In the limit as holdup tends to zero, this velocity approaches the characteristic velocity , and taking the ratio of spray and plate column velocities yields the expression
where I is the column impedance and is defined as the fractional increase in time of passage of a single droplet with respect to an empty spray column. The terminal velocity U can be written in terms of mean droplet size and the physical properties of the systems using published drag coefficient data, thereby establishing the link between droplet size and for a known value of I via Equation (6). Values of I range from zero to some finite value, depending upon the characteristics of the contactor and the properties of the system. For the application of this concept to pulsed plate-columns and the use of laboratory holdup measurements to characterize the behavior of such contactors, see Batey et al. (1987).
Diffusion of a solute from one liquid phase to another is a complex process governed by molecular and/or eddy diffusional mechanisms. Mass transfer flux is proportional to the instantaneous concentration driving force, and the ratio of these quantities is called the mass transfer coefficient and may be defined in terms of either the continuous or the dispersed phase driving force. Thus if solute A is transferring from phase c to phase d, the flux NA is given by:
In practice, it is not feasible to measure the two interfacial concentrations cci and cdi in a practical extraction system and so, since the interfacial concentrations are assumed to be in equilibrium, diffusion from phase c to d can be considered to be diffusion of solute through two phases or resistances in series. No resistance will be offered by the interface itself because the chemical potentials in each phase will be equal. On this basis, the so-called overall coefficients Ko can be defined:
where all the concentrations are now known. For a linear equilibrium curve, the relationships between the individual phase or film coefficients and the overall coefficients are readily shown to be [Treybal (1963)]
Numerous mathematical models have been proposed for calculating values of the individual phase coefficients k, but these make assumptions regarding the hydrodynamic characteristics of the phase in question [Skelland (1992); Javed (1992)]. Knowledge of turbulence levels and flow patterns in the vicinity of the interface is still far from complete and this limits the use of models for predicting k values. Contactor design is therefore based upon experimental measurements of the mass transfer coefficients, using the actual type of contactor and extraction system in question.
For a detailed consideration of design procedures for column contactors and mixer-settler devices, see the comprehensive account by Pratt and Stevens (1992).
Reference must be made, however, to the complications arising when the Marangoni Effect or spontaneous interfacial turbulence is present. Many solutes promote intense turbulence at the interface as they diffuse from one phase to the other [Perez de Ortiz (1992)]. The level of turbulence cannot be predicted from first principles and the consequences can only be assessed from extensive pilot plant studies. Thus, for example, extraction efficiency of countercurrent column contactors is usually expressed in terms of the height of a transfer unit or HTU [Treybal (1963)]. The overall HTU based upon the continuous phase driving force is defined as:
where the specific surface area a is equal to 6 εx/dvs, so that
Marangoni effects can influence mean droplet size, dvs, through enhanced interdroplet coalescence rates; the holdup x, by virtue of a correspondingly increased value of in Equation (2) and the overall mass transfer coefficient Koc through enhanced turbulence at the droplet interface [Thornton et al. (1985); Javed et al. (1989)]. Moreover, values of Koc become time-dependent and decrease as the droplet interface ages. (See Thornton (1987) for a further discussion of this problem.)
The terms in square brackets with subscript M in Equation (12) are therefore all dependent upon the level of interfacial turbulence induced by the solute. The extent to which these quantities are modified by Marangoni phenomena is not yet quantifiable. It is therefore always desirable to study the characteristics of any proposed extraction system in the laboratory before proceeding to the design stage.
Details of relevant process chemistry and extraction operations in the hydrometallurgical, nuclear, pharmaceutical and food industries are provided by Thornton (1992).
a Specific interfacial area
c Solute concentration
dvs Sauter mean droplet diameter
H Height of column
(HTU)oc Height of an overall transfer unit based on the continuous phase
I Column impedance defined by Eq. (6)
k Single-phase or film mass transfer coefficient
Ko Overall mass transfer coefficient
m Equilibrium distribution or partition coefficient
NA Transport flux of solute A
U Droplet terminal velocity in infinite media
V Superficial phase velocity
Droplet characteristic velocity, i.e., mean droplet velocity when Vc = o and Vd→ o
x Fractional holdup of dispersed phase. With subscripts concentration (wt fraction)
βCA Selectivity of solvent for solute C from an A–C solution
ε Fractional voidage of column
p Phase density
c Continuous phase
d Dispersed phase
i At interface
AA A in A-rich solution
AB A in B-rich solution
CA C in A-rich solution
CB C in B-rich solution
* At equilibrium
Batey, W., Thompson, P. J., and Thornton, J. D. (1987) Column impedance and the hydrodynamic characterization of pulsed plate-columns. Extraction ’87. Instn. Chem. Engrs. Symposium Series. 103: 133.
Batey, W. and Thornton, J. D. (1989) Partial mass transfer coefficients and packing performance in liquid-liquid extraction. I & EC Research. 28: 1096.
Hamm, W. (1992) Science and Practice of Liquid-liquid Extraction (Ed. J. D. Thornton). Clarendon Press. Oxford. 2, Ch 4.
Javed, K. H., Thornton, J. D., and Anderson, T. J. (1989) Surface phenomena and mass transfer rates in liquid-liquid systems. AIChE Journal. 35(7): 1125.
Javed, K. H. (1992) Science and Practice of Liquid-liquid Extraction (Ed. J. D. Thornton). Clarendon Press, Oxford. 1, Ch 4.
Newsham, D. M. T. (1992) Ibid. 1, Ch 1.
Perez de Ortiz, E. S. (1992) Ibid. 1, Ch 3.
Pratt, H. R. C. and Stevens, G. W. (1992) Ibid. 1, Ch 8.
Skelland, A. H. P. (1992) Ibid. 1, Ch 2.
Stewart, G. and Thornton, J. D. (1967) Charge and velocity characteristics of electrically charged droplets. Instn. Chem. Engrs. Symposium Series. 26: 29, 37.
Thornton, J. D. (1953) Improvements in or relating to columns for liquid-liquid extraction processes. Brit. Pat. 737, 789 (filed 1953, published 1955).
Thornton, J. D. and Pratt, H. R. C. (1953) Flooding rates and mass transfer data for rotary annular columns. Trans. Instn. Chem. Engrs. 31: 289.
Thornton, J. D. (1954) Recent developments in pulsed column techniques. Chem. Eng. Prop. Symposium Series. 50 (No. 13): 39. See also Thornton, J. D. Improvements in or relating to liquid-liquid extraction columns. Brit. Pat. 756, 049 (filed 1954, published 1956).
Thornton, J. D. (1956) Spray liquid-liquid extraction columns. Prediction of limiting holdup and flooding rates. Chem. Eng. Sci. 5: 201. DOI: 10.1016/0009-2509(56)80031-6
Thornton, J. D. and Brown, B. A. (1966) Liquid-fluid extraction process. Brit. Pat. 1, 205, 562 (filed 1966, published 1970).
Thornton, J. D. (1976) Electrically enhanced liquid-liquid extraction. Birmingham University Chemical Engineer. 27 (No. l): 6.
Thornton, J. D., Anderson, T. J., Javed, K. H., and Achwal, S. K. (1985) Surface phenomena and mass transfer interactions in liquid-liquid systems. AIChE Journal. 31 (7): 1069.
Thornton, J. D. (1987) Interracial phenomena and mass transfer in liquid-liquid extraction. Chemistry and Industry (SCI London). 16 March: 193.
Thornton, J. D. (1989) Euro. Pat. Specn. 0356030 AZ (filed 1989).
Thornton, J. D. (1992) Science and Practice of Liquid-liquid Extraction (Ed. J. D. Thornton). Clarendon Press, Oxford. 2.
Treybal, R. E. (1963) Liquid Extraction. McGraw Hill, NY.
Verrall, M. S. (1992) Science and practice of liquid-liquid extraction (Ed. J. D. Thornton). Clarendon Press, Oxford. 2, Ch 3.
Weatherley, L. R. (1992) Ibid. 2, Ch 5.