0263 8762/99/$10.00+0.

00

Institution of Chemical Engineers

Trans IChem E, Vol 77 Part A, October 1999

MULTICOMPONENT EFFECTS IN LIQUID-LIQUID EXTRACTION

M. HAEBERL and E. BLASS*
Merck KGaA, T-Invest, Darmstadt, Germany *Lehrstuhl fur Fluidverfahrenstechnik, Technical University Munich, Munich, Germany

ne of the reasons problem s arise in the design of extraction equipment is an unsatisfactory knowledge of the mechanisms responsible for the transport of components from the bulk of one phase across the interfacial area into the bulk of the other contacting phase. An investigation was carried out on mechanisms relevant for mass transfer, the importance of multicomponent effects in liquid-liquid extraction and also which models are suitable for describing mass transfer. For this, experiments were performed in a countercurrent spray column with the system ethylacetate-acetone-water. The saturation of the feed phases, the extraction factor and the direction of mass transfer were varied. When phases fed into the column are saturated, both phases change their concentration along the binodal line. When the phases are initially unsaturated, both extraction trajectories approach the binodal line. Comparing the experimental concentration pro les with simulation results, it is shown, that in the area of undersaturation, extraction trajectories can be described only if the complete matrix of multicomponent diffusion coef cients is known and the effects of diffusive coupling are taken into account. When mass transfer coef cients are calculated at bulk concentrations, the achieved results are closer to the measured extraction trajectories than in the case where they are calculated depending on interfacial concentrations.

Keywords: liquid-liquid extraction; mass transfer; diffusion; extraction paths; saturation

INTRODUCTION The calculation of mass transfer is one of the key problems in the design of apparatuses for liquid-liquid extraction. The number of publications on the subject is correspondingly large. Measurements with single droplets in stagnant continuous phases as well as with swarms of droplets in countercurrent columns were performed in various systems. Sherwood et al.1 analysed previous work and gave their own data for mass transfer in single droplets both in packed columns and a spray column. Johnson and Bliss2 made multiple measurements in a spray column and gave some empirical rules of design and operation of spray columns. Steiner 3, Steiner et al.4 collected a large number of experimental mass transfer coef cients and compared the results with calculated data. Mass transfer in various countercurrent extractors was investigated by Blass and co-workers at Munich Technical University (Hirschmann 5, Goldmann 6, Otillinger 7, Hufnagl 8, Rauscher9, Zamponi 10). Blass et al.11 recommend the determination of mass transfer coef cients in the droplet phase using the Handlos and Baron 12 correlation and in the continuous phase with the help of the penetration theory.

Concentration Trajectories in Ternary Extraction Only a few authors have investigated extraction paths in ternary systems with partially miscible main components (Table 1). In a modi ed Lewis cell, Sethy and Cullinan 13
647

observed straight extraction paths in the n-heptane phase in the system acetonitrile-benzene-n-heptane. The concentrations follow the straight line connecting the initial composition and the nal equilibrium composition, which means that no diffusive coupling occurs between the component uxes. Krishna et al.14 show that the measured concentration trajectories cannot be described even qualitatively by binary mass transfer calculations and it is absolutely necessary to take into account the in uences of multicomponent effects. 18 33 Schermuly , Schermuly and Blass used the system glycerol-water-acetone in a countercurrent spray column for extraction experiments and determined the concentration pro les of both the phases. While the concentration trajectories measured in the stirred cell show some timedepending change of concentration, i.e. they are paths of equilibration, the extraction paths determined by Schermuly and Blass33 give local, but, as the column was run at steadystate, time-invariant concentration pro les. Mass transfer of water was directed from the continuous acetone phase into the dispersed glycerol phase in all the experiments. When phases are saturated, i.e. with binodal composition, the extraction paths follow the binodal line all along the length. Schermuly and Blass33 successfully calculated mass transfer in saturated phases, using a uctuation model, that presumes 19 20 vanishing Fick diffusion (Hampe et al. ) and exclusively uid-dynam ically induced mass transfer. When feed-phases are unsaturated, the extraction trajectory does not reach the binodal line and can be described only by taking into account diffusive coupling because of multicomponent

648

HAEBERL and BLASS

Table 1. Extraction paths during equilibration in a stirred cell.

Author Sethy and Cullinan Standart et al.

15 13

System Acetonitrile / benzene / n-heptane Glycerol / acetone / water

16

Experimental conditions Heavy (acetonitrile) phase is recirculated Phases are not circulated Heavy (glycerol) phase is recirculated Phases are not circulated Phases are not circulated

Concentration trajectories

n-Heptane phase: all straight

Both phases: distinctly curvilinear, straight only in 9tie-line 9 experiments Acetone phase: distinctly curvilinear, straight only in 9tie-line 9 experiments The same results as given by 15 Standart et al. No data published

Cullinan and Ram Krishna et al.14

Glycerol / acetone / water Glycerol / acetone / water MIBK / acetic acid / water

17 Bulic ka and Procha zka 3

mass transfer. Hence, for calculation of mass transfer coef cients a complete matrix of diffusion coef cients is necessary. Berger21 performed experiments in a countercurrent spray column, using the system water-succinic acid-nbutanol. Though he did not give extraction trajectories in his work, they may be obtained easily, by plotting his data into 22 the equilibrium diagram by Mis) ek et al. . Independent of the direction of mass transfer, the water phase as well as the n-butanol phase approaches the binodal line asymptotically and reaches it after only a short distance. In this work the phenomena and mechanisms of multicomponent mass transfer are discussed. Using saturated and unsaturated feeds, the way multicomponent effects in uence the extraction trajectories were investigated, and from this the dominant mechanisms of mass transfer were determined. It will be shown, whether it is necessary to differentiate between saturated and unsaturated phases for calculation of mass transfer.

velocity (Taylor and Krishna 23) are determined in a ternary system by:

with J1 + J2 + J3 = 0 The molar uxes Ni relative to a stationary co-ordinate reference frame are, however, of technical importance. The total ux Nt, which is given by the sum of all component uxes, is needed for calculation of these uxes.

( )
J1 J2 Ni

= 2ct

k11 k21

k12 k22

](

xI1 2 xB 1 xI2 2 xB 2

( 1)

= Ji + xi Nt
.

( 2)

In liquid-liquid extraction, the total ux Nt is generally nonzero as is shown in this work. This total ux is also responsible for variation of mass ows of both phases in extraction columns and in uences the concentration pro le between bulk and interface.

FUNDAMENTALS OF LIQUID EXTRACTION MASS TRANSFER The following factors must be taken into account for calculation of mass transfer in liquid extraction: The presence of two coexistent phases, which may vary considerably in their physical properties and their molecular interactions. There is an interfacial area in between these two phases, which may in uence mass transfer, depending on its width, physical properties and molecular structure, e.g. Marangoni effects are originated in the area. Fluid dynam ics are extremely complex especially in two phase dispersed systems and there is a strong interdependency of uid dynam ics and mass transfer. Thus, mass transfer is in uenced by mixing inside the droplet and by the ow of the continuous phase in the vicinity of the droplet, as well as by breakage and coalescence of droplets.
The molar ux Ji between the two phases in contact is described analogously to a diffusive ux as the product of the transport coef cient and driving force. The difference of molar concentrations at the interface and in the bulk phase is used as a driving force. The mass transfer coef cients kij must take into account molecular and uid-dynam ical in uences. Hence they mainly depend on the transport mechanisms. Molar uxes relative to the mean molar

MECHANISMS AND MODELS FOR MASS TRANSFER Ignoring interfacial effects, there are only the diffusive and convective mechanisms of mass transfer left under discussion. With increasing mobility of droplets, an increased in uence of convection is usually assumed. The various in uences of both the mechanisms is re ected by the dependency of the mass transfer coef cient on the diffusion coef cient. k ~ Dn ( 3)
with the exponent n varying from zero to one. The various models for calculating mass transfer between droplets and a continuous phase were described in detail by 24 25 26 Slater , Temos , and Javed . To investigate the in uence of diffusion and convection only the following basic types of models are considered: 1. The two equations by Schermuly and Blass for saturated, respectively unsaturated phases which are applicable for both continuous and dispersed phases. 27 2. The models of Kronig and Brink and of Handlos and 12 Baron for dispersed phases. 28 3. The penetration model of Higbie for continuous phases. A detailed description of the models is given in 29 Haeberl . Diffusion effects on mass transfer are taken into
33

Trans IChemE, Vol 77, Part A, October 1999

MULTICOMPONENT EFFECTS IN LIQUID-LIQUID EXTRACTION

Table 2. Investigated models for diffusive and convective mass transfer.
Continuous phase Diffusive Higbie
28

649

Convective
33

Schermuly and Blass (Unsaturated phases)

Schermuly and Blass (Saturated phases / Fluctuation model)

33

Dispersed phase

Diffusive

Kronig and Brink

27

Schermuly and Blass (Unsaturated phases) Convective Handlos and Baron

33

Diffusive/ diffusive Diffusive/diffusive Diffusive/ convective Convective/ convective

12

Schermuly and Blass (Saturated phases / Fluctuation model)

33

account most strongly using the models by Higbie 28 and Kronig and Brink 27. The results calculated with the equation of Schermuly and Blass 33 for unsaturated phases are very similar, as the direct proportional ity between mass transfer coef cient and diffusion coef cient is just replaced by a square root proportionali ty. Based on his theory of transport mechanisms at phase boundaries, Hampe 30,31 concluded that in the vicinity of the phase boundary, molecular diffusion vanishes and is replaced by another transport mechanism. Schermuly 18 and Schermuly and Blass32,33 developed a model for an exclusively uid-dynam ically determined transport mechanism on this basis. A mixing process between continuous and dispersed phases takes place in a circulating droplet by the uctuation of uid elements around their stream lines. It can be described by effective diffusivity as de ned by Handlos and Baron 12.

between both eigenvectors of the matrix of the mass transfer coef cients (Gupta and Cooper) 35. The matrix of mass transfer coef cients [k] is diagonal if:

The matrix of diffusion coef cients [D] is diagonal or there is purely convective mass transfer.
Thus conclusions can be drawn about the relevant transport mechanism from the course of the extraction paths: if the extraction path is directed towards equilibrium , nevertheless cross coef cients D12 and D21 are distinctly different from zero, diffusion is not a dominant transport mechanism and uid dynam ic effects prevail. If the matrix of diffusion coef cients is diagonal, differentiation of both the mechanisms is certainly no longer possible. Then both diffusive and convective transport result in decoupled uxes and the extraction path is a straight line towards the equilibrium state.

Some General Remarks on Fluid Dynamical Models for Mass Transfer

Formally, a convective ux can be calculated in the same way as a diffusive ux. The concentration difference between bulk and interface is the driving force for mass transfer. If this becomes zero, there will be no mass transfer. The mass transfer coef cient however does not depend on diffusion coef cients, but on the relative velocity between sedimenting droplets and the continuous phase. Thus mass transfer affects all components according to their concentration in the exchanged uid element. With purely convective mass transfer there can be no coupling effects between the uxes of components, and the matrix of mass transfer coef cients will always be diagonal.

EXPERIMENTAL RESULTS
The system ethylacetate-acetone-water has one miscibility gap: both ethylacetate and water are completely miscible with acetone and partially miscible with each other. The experimental values of density, viscosity, interfacial tension, and equilibrium concentrations can be found in Haeberl29. While density and viscosity are linearly depende nt on acetone concentration in the ethylacetate phase, there is a distinct maximum of viscosity at about 10 mole % of acetone in the aqueous phase. The interfacial tension between saturated ethylacetate and aqueous phases at 20C, was measured as 6.20 mN/ m, which decreases with increasing concentrations of acetone.

STRUCTURE OF MATRIX [k] AND DIRECTION OF EXTRACTION TRAJECTORIES Sethy and Cullinan 34, as well as Schermuly and Blass 33, show that the extraction path is directed towards equilibrium , in the case where the matrix of mass transfer coef cients [k] is diagonal, i.e. both the cross coef cients k12 and k21 are zero. In all the other cases a curved extraction path is obtained, the tangent of which does not coincide with connecting straight lines between bulk concentration and interface concentration, while the area of possible directions, which the extraction path can take, is stretched Trans IChemE, Vol 77, Part A, October 1999

Ternary Diffusion Coef cients The ternary diffusion coef cients were measured in the system ethylacetate-acetone-water with holographic interferometry by Pertler36. He determined the matrices of diffusion coef cients in the aqueous phase at 10 mass % of acetone at ve different concentrations of ethylacetate and in the ethylacetate phase at 20 mass % of acetone at ve concentrations of water at 20C. The diffusion coef cients were approximated for calculation of mass transfer by polynomial equations. It was assumed that the changes in the structure of the matrix do

650

HAEBERL and BLASS

not depend qualitatively on acetone concentrations, but the binodal composition is decisive. The binary diffusion coef cients were assumed to be constant within the range of concentrations under discussion. 36 Pertler found the following characteristic behaviour of diffusion coef cients in both phases:

The cross-coef cient D21 strongly depends on concentration with a distinct minimum. At medium concentrations its absolute value is much higher than both the main coef cients. The cross-coef cient D12 approaches zero at the binary line. Both cross coef cients simultaneously approach zero and the main coef cients have identical values at the binodal line: D12

EXPERIMENTAL PROGRAM The experiments accomplished can be divided into 4 groups. Despite the fact that because of wide range of solubility all the three components take part in mass transfer, the direction of mass transfer is described here in a classical style by the direction of transfer of acetone. The initial concentration of the extracted phase, i.e. ethylacetate phase in Group I and II and aqueous phase in Group III and IV, as well as the extraction factor E, varied in every of the four experiment groups. xac,EA MEA E= ( 4) xac,W MW
The extracting phase was fed into the column virtually free of acetone in all the experiments. The variation of the extraction factor was achieved by changing of the ow rate of the continuous phase. The ow rate of the aqueous phase was kept constant to avoid the in uence of changing droplet size distribution on mass transfer. Next to no interdrople t coalescence occurs during the experiments with mass transfer from the continuous to the dispersed phases (Groups I and II). The Sauter mean droplet diameter was about 2 mm in all the experiments, with a maximum in the drop size distribution between 1.5 mm and 2.0 mm. Interdroplet coalescence, however, was very intensive with mass transfer from the dispersed water phase to the continuous ethylacetate phase (Gr. III and IV). This effect can be explained by enrichment of the transferred component in the drainage lm between two colliding droplets, which results in a local decrease of interfacial tension. The experiments of Saboni et al.37 con rm this mechanism. To exclude the in uence of coalescence on mass transfer, experiments with this direction of mass transfer were done at a lower volumetric ow rate. Thus, hold-up of the dispersed phase in the column is so low that interdroplet coalescence occurs only at a very low scale and the droplet size distribution is analogous to the one in experiments with groups I and II. Performance of these single droplet experiments was, of course, rather problematic because of the extremely long time of up to 19 hours before a steady state was achieved, and in any case, sampling of the dispersed phase is very dif cult.

= D21 = 0 and D11 = D22 0

The matrix of the diffusion coef cients is diagonal. Hence the diffusive uxes are not coupled there. However, this does not mean that there is no diffusive transport.

Extraction Column The counter-current spray column described by Schermuly and Blass33 was used for the extraction experiments. The column has an active height of 4.3 m, an inner diameter of 25 mm, and is thermostated by a water- lled double jacket. The heavier aqueous phase is fed at the top of the column through a 1.0 mm PTFE nozzle, the lighter ethylacetate phase enters the column at the bottom end. To prevent droplets of aqueous phase from wetting the glass wall, the column was silanized. Five sampling points for continuous and dispersed phases, where probes can be positioned in the centre of the column, were built in along the column. Droplets of the dispersed phase are collected in a small cup where they coalesce, while a sample of the continuous phase is withdrawn through a coaxial outer tube. The upperm ost probe is positioned in a distance of 170 mm from the nozzle, where droplet formation is nished. Along with sampling from the feed and drain pipes, concentrations can be determined at seven positions in each phase. Droplets can be photographed through at windows built in at the sampling points. The photos of droplets were read through a CCD camera (having processing facilities allowing the contrast to be raised between droplet and background) into a PC system and interpreted using the picture analysing software Optimas. The droplets appeared as spheres and rotation ellipsoids and length of radius and of both the main axes respectively, was determined automatically. The system was calibrated by a picture of one of the probes. The reliability of the method was proved by comparing with the results of manual measuring.

EXTRACTION TRAJECTORIES Mass Transfer from Ethylacetate Phase to Aqueous Phase Examples of extraction paths for experiments with saturated and unsaturated feed phases are given in Figure 1. The ratio of volumetric ow rate between dispersed and continuous phases was 0.3, the extraction factor had the value E = 3 .8 during both the experiments. The ethylacetate phase was introduced into the column in

Table 3. Program of extraction experiments.

Group of experiments I II III IV Direction of mass transfer ethylacetate (c)Nacetone Nwater ethylacetate (c)Nacetone Nwater water (d)Nacetone Nethylacetate water (d)Nacetone Nethylacetate (d) (d) (c) (c) Saturation state of feed phases Both unsaturated At least one unsaturated Both unsaturated Both saturated

Trans IChemE, Vol 77, Part A, October 1999

MULTICOMPONENT EFFECTS IN LIQUID-LIQUID EXTRACTION

651

Figure 2. Concentration pro les in the aqueous phase for the experiments
shown in Figure 1.

Figure 1. Extraction trajectories in ethylacetate-acet one-water system. Extraction factor E = 3.8; top: feeds nearly saturated; bottom: feeds
unsaturated.

the experiment in Figure 1 (top) in a saturated state with 30 mass % of acetone. Its composition changed along the binodal line and it left the column with 20 mass % of acetone. As a consequence of the position of the binodal curve, water concentration along with acetone concentration must also decrease in the ethylacetate phase . The aqueous phase entered the column in a nearly saturated state and with a content of acetone of 1 mass %. In the beginning, its extraction path seems to go parallel to the binodal line, but reaches it at the third measuring point and follows it up to the exit with 21 mass % acetone. As well as in the ethylacetate phase, changes of concentrations in the aqueous phase are determined mainly by the binodal line, hence ethylacetate concentration must also increase from 7 mass % to 12 mass % while accepting acetone. The experiment plotted in Figure 1 (bottom ) differs from the above described only because both of the phases were introduced into the column in a strongly unsaturated state. The ethylacetate trajectory immediately strives for the binodal line with hardly any decrease of acetone concentration. It is saturated and already follows the binodal line at the lowest sampling point in the column. Thus it is also forced, together with acetone, to deliver the water it initially accepted. The water trajectory is nearly a straight line until it reaches the binodal line. Comparing the extraction trajectories in Figure 1, it is obvious that the nal concentrations are almost identical.

They are determined by the equilibrium in the water phase, and by the mass balance in the ethylacetate phase. To investigate the in uence of the increased, due to undersaturation of the feeds, mass transfer of water and ethylacetate, the concentrations versus the column length were plotted (Figure 2). The heavier aqueous phase enters the column at the top at HK = 0. It is obvious that in the upper part of the column the unsaturated phase accepts considerably more acetone and delivers more water than the saturated one. The high negative ux of water may cause an increased transfer of acetone due to diffusive coupling. It becomes clear from Figure 3, that the extraction trajectories strongly depend on the extraction factor. The feeds were unsaturated in both the experiments with extraction factors of E = 3 .8, E = 0.9, respectively. If at a constant ow of dispersed phase the ow rate of the continuous phase is reduced, its extraction trajectory becomes longer, linked to the mass balance. The aqueous phase is enriched in Figure 3 (top) up to the maximum concentration of acetone, i.e. equilibrium concentration to the feed of the ethylacetate phase. Thus there is a very low driving force for extraction of acetone from the ethylacetate phase. This explains why the ethylacetate phase rst gets saturated with water and only then delivers acetone. For the entering water phase there is a high driving force for acceptance of acetone. Thus its extraction trajectory is nearly a straight line. In Figure 3 (bottom) none of the phases reaches equilibrium . Compared to the previous example, the driving force for the transfer of acetone is higher for the ethylacetate phase and lower for the aqueous phase, and the water trajectory is now asymptotically curved towards the binodal line. It becomes clear from the examples shown in Figures 1 and 3 that extraction trajectories are in uenced by local driving forces in the column as well as by the matrix of diffusion coef cients.

Mass Transfer from Aqueous Phase into Ethylacetate Phase The results of an experiment with unsaturated feed phases are shown in Figure 4. As was discovered in the previous experiments, with opposite direction of mass transfer, extraction trajectories approach the binodal line asymptotically and both the phases are already saturated after a short

Trans IChemE, Vol 77, Part A, October 1999

652

HAEBERL and BLASS

Column Model The cell circulation model of Wasowski and Blass38 was used for calculation of extraction paths in the countercurrent spray column. The wake ow, which contains some amount of the continuous phase carried behind droplets, was taken into account in this model as well as ow rates of the dispersed and continuous phases. The wake going from the top towards the bottom through the column in an exactly opposite direction to the ow of the continuous phase, is separated from the droplets at the bottom by coalescence of droplets, mixes with continuous phase there and raises with the ow in the direction of the head of the column. The composition of the wake ow corresponds at every level of the column to the concentration of the outgoing continuous phase. As sedimenting droplets cause the back mixing in the continuous phase, the wake ow is calculated depending on the ow rate of the dispersed phase, using the equations given by Letan and Kehat39 for wake volum e in liquid-liquid spray columns. The wake ow rates must be taken into account in the mass balances of top and bottom stages as out ow and in ow respectively. Equations for Droplet Sedimentation Rate
Both circulating and oscillating droplets were observed during the extraction experiments. The equations of Klee 40 41 and Treybal and of Kumar et al. were applied for calculation of sedimentation velocities, both of the equations were obtained by adopting to measured data of circulating and oscillating droplets, and are applicable for a system with interfacial tension between 0 .33 10 23 N m 21 # s # 42.43 10 23 N m 21 . The equations developed for spray columns by 41 42 were applied for Kumar et al. and by Harmathy calculation of the relative droplet velocities in the column.

Figure 3. In uence of the extraction factor on the extraction trajectories. Top: E = 3 .80; bottom: E = 0 .90.

time. The entering aqueous phase has a remarkably strong tendency to become saturated with ethylacetate, though there is a high driving force for extraction of acetone. It is also obvious that the aqueous phase becomes very quickly and completely depleted of acetone with this direction of mass transfer. Thus the entering ethylacetate phase cannot be enriched with acetone in the lower part of the extraction column.

Simulation Program
The goal of the simulation was to prove how the application of various transport mechanisms affects the conformity between experimental and calculated data. To demonstrate this in uence as clearly as possible, no tted parameters were used. The calculation takes into account the following effects:

Figure 4. Extraction trajectory with mass transfer from aqueous phase into ethylacetate phase; ow of dispersed phase Vd = 0.08 l h21 , E = 1 .8.

Back mixing in the continuous phase is caused by wake effects. The wake ow rate is calculated depending on the ow rate of the dispersed phase. The physical properties depend on the concentration of the phases. Viscosity and density are determined from 3 dimensional surface ts, interfacial tension and ternary diffusion coef cients from polynom ial ts. Mass transfer cannot be assumed to be equimolar. Thus mass transfer coef cients depend on the total ux and must be calculated in an iterative procedure. As mass transfer is not equimolar, the ow rates of both the phases change along the column. Because of multicomponent effects, the component uxes are not independent. They are calculated using the linearized theory of Toor 43 . The ternary mass transfer coef cients are calculated by decoupling the transport equations and subsequent inverse transformation. The driving force for mass transfer, i.e. the difference of Trans IChemE, Vol 77, Part A, October 1999

MULTICOMPONENT EFFECTS IN LIQUID-LIQUID EXTRACTION concentration between bulk and interface, is determined at each stage by solving the polynomial equation for the binodal line. As the drop size distribution at the upper and lower end of the column differ only slightly, break-up and coalescence of droplets were not taken into account, nor were the change of droplet sizes by mass transfer. Convergence was very fast in most cases and constant values for mass transfer coef cients were achieved already after the 3rd to 7th iteration. There were problem s with convergence only with the Kronig and Brink model27 . Several routines from the NAG-FORTRAN library, version mark 15, together with several algorithms of Numerical Recipes (Press et al.)44 were used for the calculation. It was helpful for program ming to follow the IK-Cape 45 guideline.

653

INTERPRETATION OF RESULTS Figure 5 (top) shows the results of an experiment with saturated feed phases and with extraction paths calculated according to the mass transfer models by Higbie 28 for continuous phases, as well as by Handlos and Baron 12 for dispersed phases. The path goes all its length by the binodal line and thus is constrained as to its direction. In addition, all the equilibrium states lie on the binodal line and the path must always go towards equilibrium . A diagonal matrix of mass transfer coef cients satis es in this case to describe the mass transfer. As the model of Handlos and Baron 12

presumes purely convective mass transfer, it will, in any case, give a diagonal matrix. However, a transport matrix calculated with diffusive models such as the penetration theory by Higbie 28, will certainly also be diagonal along the binodal lines, since cross coef cients in the matrix of ternary diffusion coef cients will approach zero. That is why it is no longer surprising that in this case all the models come to rather similar results, except for the equation by Kronig and Brink 27 which gives too small mass transfer coef cients for long contact times. If feed is introduced into the column unsaturated, all the directions in the concentration diagram are in general open for the extraction path. The experiment has nevertheless shown that the paths of both the phases in the system ethylacetate-acetone-water strive for the binodal line and only stay in the undersaturation area for a short segment. As Figure 5 (bottom) for the experiment with unsaturated feed phases and extraction factor of E = 1 .8 shows, the course of the extraction path in this case can also be described with the help of the model by Higbie 28 and Handlos and Baron 12 recommended by Blass et al.11 with quite satisfactory precision. As regards to the dominating transport mechanisms, it is interesting to examine extraction paths in the undersaturated area more precisely. An experimental extraction path of the aqueous phase together with concentration curves, calculated according to the equations by Higbie 28 and Handlos and Baron 12, and to both the models by Schermuly 18, 33 Schermuly and Blass , are plotted in Figure 6. Starting with pure water the experimental trajectory is curved towards the binodal line. It is best described with the mass transfer model for unsaturated phases and diffusive transport by Schermuly and Blass33. While applying the model for continuous phases by Higbie 28 and for dispersed phases by Handlos and Baron 12, the calculated path is curved towards the binodal line obviously more strongly than the curve of the measured data. If the calculation is based on the 33 uctuation model by Schermuly and Blass , which assumes purely uid-dynam ical transport, the calculated trajectory is directed straight towards the binodal line and the difference between experimental and calculated data becomes distinctly bigger. Plotting the concentration pro le versus the height of the column con rms these conclusions (Figure 7). The concentrations of all the three components are best

Figure 5. Extraction trajectories calculated by Higbie and Handlos and Baron. Top: saturated feed phases, E = 3 .8; bottom: unsaturated feed phases, E = 1 .8.

Figure 6. Experimental trajectory of aqueous phase with unsaturated feed,

compared to the results of different mass transfer models.

Trans IChemE, Vol 77, Part A, October 1999

654

HAEBERL and BLASS coef cients must be diagonal there as well. Extraction trajectories calculated with the diffusive model of Schermuly and Blass33 using different sets of mass transfer coef cients are shown in Figure 8. It becomes obvious that in the case where mass transfer coef cients are calculated at bulk concentrations, the experimental results can be described better. As the matrix of mass transfer coef cients [k] is diagonal at interfacial concentrations, diffusive coupling is not taken into account and the calculated trajectory is curved more strongly towards the binodal line, than the experimental one.

Figure 7. Concentration pro les in the aqueous phase versus column height. Calculation performed with diffusive model by Schermuly for unsaturated phases.

described by Schermuly 9s model for multicomponent diffusion, especially near the entrance point of the dispersed phase.

In uence of Diffusion
In general, diffusion coef cients can be calculated either at the bulk or at interfacial concentration. Cullinan and Ram 10 concluded from the experiments in a stirred cell, that mass transfer is controlled by the situation at the interface and that diffusion coef cients should be determined for this concentration. Calculating ternary mass transfer coef cients in the aqueous phase at a position of H = 2180 mm in the column for the experiment shown in Figure 5 (bottom ) with the help 33 of the diffusive model of Schermuly and Blass , produced the following results: At bulk phase concentration:

=
=

(
(

1 .533 10 24

8 .903 10 27 1 .733 10 24
0 1.463 10
24

27 .723 10 25
1 .463 1024 0

)
m s

m s

At interfacial concentration:

As the matrix of the Fick 19 diffusion coef cients becomes diagonal at the binodal line, the matrix of mass transfer

CONCLUSIONS As shown by Pertler 36, Fick 19 diffusion does not vanish at the binodal line, but the matrix of ternary diffusion coef cients becomes diagonal. Thus the matrix of mass transfer coef cients is a diagonal matrix in saturated phases, independent of whether it is calculated with a purely uiddynam ical model or with a diffusive model. It is thus impossible to distinguish from the extraction trajectory, running along the binodal line, between the diffusive and convective mechanisms of mass transfer. The results show that using the diffusive model for the calculation of mass transfer, the extraction trajectories can be well described in the area of miscibility and along the binodal line. From our results it does not become clear whether molecular clusters are formed in saturated phases. 19 However, following Ficks phenom enological de nition of diffusion, mass transfer can surely be named diffusive in extraction with saturated phases. The presence of diffusion, and resulting from it the in uence of diffusive coupling on multicomponent mass transfer, does not mean that a complete matrix of diffusion coef cients is always necessary for design of liquid-liquid extraction columns. As these results for the system ethylacetate-acetone-water show, extraction trajectories usually reach the binodal line shortly after the feed point and then follow the binodal line. If, for an extraction process, the part of the trajectory in the area of miscibility is not important or both main components are nearly immiscible, suf cient results can be obtained even with purely uid-dynam ical models. However, if, for the process or apparatus design, it is necessary to calculate extraction trajectories for unsaturated phases, it is de nitely necessary to take into account all the set of multicom ponent effects in mass transfer. REFERENCES
1. Sherwood, T. K., Evans, J. E. and Longcor, J. V. A., 1939, Extraction in spray and packed columns, Ind Eng Chem Res, 31: 1144 1150. 2. Johnson, H. F. Jr. and Bliss, H., 1946, Liquid-liquid extraction in spray towers, Trans AIChE, 42: 331 358. 3. Steiner, L., 1988, Rechnerische Erfassung der Arbeitsweise von Flu ssig-Flu ssig-Extraktions kolonnen, Fortschr-Ber VDI Reihe 3, No 154 (VDI-Verlag, Du sseldorf). 4. Steiner, L., Oezdemir, G. and Hartland, S., 1990, Single drop mass transfer in the water-toluene- acetone system, Ind Eng Chem Res, 29: 1313 1318. 5. Hirschmann, K., 1984, Beitrag zur Auswahl unpulsierter Siebbodenex traktionskolonnen, Diss, (TU Mu nchen). 6. Goldmann, G., 1986, Ermittlung und Interpretation von Kennlinien feldern einer geru hrten Extraktionskolonn e, Diss, (TU Mu nchen). 7. Otillinger, F., 1988, Fluiddynamik und Stoffubergang in Zentrifugalextraktoren, Diss (TU Mu nchen).

Figure 8. Extraction trajectory in aqueous phase, E = 1 .80; Two calculated trajectories based on mass transfer coef cients at bulk concentration, and on interfacial concentration respectively.

Trans IChemE, Vol 77, Part A, October 1999

MULTICOMPONENT EFFECTS IN LIQUID-LIQUID EXTRACTION

8. Hufnagl, H., 1992, Dynamisches Verhalten einer Flu ssig- ussig Extraktionskolonn e, Diss (TU Mu nchen). 9. Rauscher, H., 1992, Untersuchung einer pulsierten Siebboden-Extr a ltnissen, Diss (TU Mu nchen). ktionskolonne bei extremen Phasenverha 10. Zamponi, G., 1996, Das dynamische Verhalten einer geru hrten Solventextraktion skolonne, Diss (TU Mu nchen). 11. Blass, E., Goldmann, G., Hirschmann, K., Mihailowitsch, P. and Pietzsch, W., 1985, Fortschritte auf dem Gebiet der Flu ssig-Flu ssig Extraktion, Chem-Ing-Tech, 57: 565 581. 12. Handlos, A.E. and Baron, T., 1957, Mass and Heat Transfer from Drops in Liquid-liquid Extraction, AIChE J, 3: 127 136. 13. Sethy, A. and Cullinan, H. T., 1975, Transport of mass in ternary liquid-liquid systems, AIChE J, 21: 3, 571 582. 14. Krishna, R., Low, C. Y., Newsham, D. M. T., Olivera-Fuente s, C. G. and Standart, G. L., 1985, Ternary mass transfer in liquid-liquid extraction, Chem Eng Sci, 40(6): 893 903. 15. Standart, G. L., Cullinan, H. T., Paybarah, A. and Louizos, N., 1975, Ternary mass transfer in liquid-liquid extraction, AIChE J, 21(3): 554 559. 16. Cullinan, H. T. and Ram, S. K., 1976, Mass transfer in a ternary liquidliquid system, Can J Chem Eng, 54: 156 159. 17. Bulicka, J. and Prochazka, J., 1979, Ternary liquid-liquid mass transfer in turbulent ow, Chem Eng Commun, 3: 325 337. bergang bei der 18. Schermuly, W., 1990, Dreikomponente n-Stoffu Extraktion mit dem Stoffsystem Glycerin-Aceton-Wasser, Diss (TU Mu nchen). 19. Fick, A., 1855, On liquid diffusion, Phil Mag, 10: 30 39. 20. Hampe, M. J., Schermuly, W. and Blass, E., 1991, Decrease of diffusion coef cients near binodal states of liquid-liquid systems, Chem Eng Technol, 14: 219 225. 21. Berger, J., 1981, Grundlagen des Stoffaustausch es in einer Flussig ussig Spruhkolonne, Diss (ETH Zu rich). 22. Misek, T., Berger, R. and Schro ter, J. (Eds.), 1985, Standard Test Systems for Liquid Extraction Studies, (IChemE, Rugby). 23. Taylor, R. and Krishna, R., 1993, Multicomponent Mass Transfer, (John Wiley & Sons, New York). 24. Slater, M. J., 1994, Rate coef cients in liquid-liquid extraction systems, in: Godfrey, J. C. and Slater, M. J. (eds.), Liquid-Liquid Extraction Equipment, (John Wiley & Sons, Chichester) 45 94. 25. Temos, J., 1992, Mass transfer to freely moving droplets, Ph D Thesis, (The University of Melbourne, Australia). 26. Javed, K. H., 1992, Mass transfer coef cients, interfacial effects and surface renewal, in: Thornton J.D.(ed.), Science and Practice of Liquid-Liquid Extraction, Vol. 1, (Clarendon Press, Oxford) 210 243. 27. Kronig, R. and Brink, J.C., 1950, On the theory of extraction from falling droplets, Appl Sci Res, A2: 142 154. 28. Higbie R., 1935, The rate of absorption of a pure gas into a still liquid during short periods of exposure, Trans AICheE, 31: 365 383. 29. Haeberl, M., 1996, Mehrkomponente n-Stoffubergang bei der Flussigussig-Extraktion, Diss (TU Mu nchen).

655

30. Hampe, M. J., 1980, Zur Thermodynamik der Transportprozes se in Grenz a chensystemen, Diss (TU Mu nchen). 31. Hampe, M. J., 1981, Entropieerzeuge nde Prozesse an Phasengrenz a chen, Ber. Bunsenges, Phys Chem, 85: 834 842. 32. Schermuly, W. and Blass, E., 1991, Three-compon ent mass transfer in liquid-liqud extraction with the system glycerol-acetone- water, Part 1: The in uence of vanishing diffusion coef cients in binodal states on the mechanism of mass transfer, Chem Eng Technol, 14: 186 191. 33. Schermuly, W. and Blass, E., 1991, Three-compon ent mass transfer in liquid-liqud extraction with the system glycerol-acetone- water, Part 2: Calculation and experimental investigation of three-componen t mass transfer in a countercurrent extraction column, Chem Eng Technol, 14: 253 263. 34. Sethy, A. and Cullinan, H. T., 1975, Transport of mass in ternary liquid-liquid systems, AIChE J, 21(3): 571 582. 35. Gupta, P.K. and Cooper, A.R. Jr., 1971, The [D] matrix for multicomponent diffusion, Physica, 54: 39 59. ren und terna ren Flu ssig- ussig 36. Pertler, M., 1996, Diffusion in bina Systemen, Diss (TU Mu nchen). 37. Saboni, A., Gourdon, C. and Chesters, A., 1993, The in uence of mass transfer on coalescence in liquid-liquid systems, in: Logsdail and Slater, Solvent Extraction in the Process Industries, Proc of ISEC 93, (Elsevier Applied Science, London, New York), 506 513. 38. Wasowski, T. and Blass, E., 1989, Neue Aspekte zum Ein uss des Blasenwakes auf die Hydrodynamik und den Stoffaustausch in einer, Blasensa ule, Chem-Ing-Tech, 61: 519 530. 39. Letan, R. and Kehat, E., 1968, The mechanism of heat transfer in a spray column heat exchanger, AIChE J, 14: 398 405. 40. Klee, A. J. and Treybal, R. E., 1956, Rate of rise or fall of liquid drops, AIChE J, 2(4): 444 447. 41. Kumar, A., Vohra, D. K. and Hartland, S., 1980, Sedimentation of droplet dispersions in counter-curren t spray columns, Can J Chem Eng, 58: 154 159. 42. Harmathy, T., 1955, Theory of troughput in extraction spray towers, Acta Technica (Acad Sci Hung) 12: 209 242. 43. Toor, H.L., 1964, Solution of the linearized equations of multicomponent mass transfer: II. Matrix methods, AIChE J, 10: 460 465. 44. Press, W.H., Teukolsky, S.A., Vetterling W.T. and Flannery, B.P., 1992, Numerical Recipes in FORTRAN, 2nd ed., (Cambridge University Press). 45. IK-CAPE, 1993, Portables Programmieren in Fortran, (Programmierrichtlinien, Frankfurt).

ADDRESS
Correspondence concerning this paper should be addressed to Dr Ing. M. Haeberl, Frankfurter Strasse 19A, D-64807 Dieburg, Germany.

The manuscript was received 3 March 1999 and accepted for publication after revision 1 June 1999.

Trans IChemE, Vol 77, Part A, October 1999