Chem. Eng. Technol.

10 (1987) 291 -296

29 1

Calculation of Reactive Extraction in Countercurrent Columns

Hans Jorg Bart, Andreas Bauer and Rolf Man*
The kinetics of extraction processes with chemical reaction was studied experimentally on the system copper/Acorga PT 5050 and quantified according to meaningful kinetic models. These were coupled with other models describing the behaviour of dispersions in columns in order to permit a computer aided simulation of reactive extraction. Thus, a better prediction of column performance is achieved.

1 Introduction
Reactive extraction is a unit operation which plays an increasingly important role in the recovery of metal salts from leaching solutions and waste waters. The separation is effected by contacting an aqueous phase containing the material to be extracted with an organic phase. The organic phase consists of an extractant which promotes the transfer of the solute from the aqueous to the organic phase, an aliphatic or aromatic hydrocarbon as diluent and, in most cases, also a modifier which improves the solubility of the complex Ill. The extractant may act as a liquid ion exchanger or a chelating agent, reacting with the substances to be extracted. The formed complex is transferred to the organic phase and may be reextracted. It is decomposed by a suitable stripping agent such as an acid. Fig. 1 shows the chemical structure of a complex formed from two extractant molecules chelating a copper ion. The extraction/reextraction process results in an enrichment and/or selective recovery of the desired material.

namic properties restrict their use to the conditions for which they have been designed. Reaction kinetics complicates the process design. In many cases of reactive extraction, e.g. with copper, the kinetics of the reaction is in fact the rate determining step with respect to mass transfer. If reaction kinetics is neglected in mixer-settler design, a slow reaction can be accounted for by running the mixer at lower throughput rates and thus longer residence times. In columns, a reduction of throughput of the dispersed phase results in decreased hold-up which causes a decline in interfacial area per unit volume and hence in mass transfer rate. The latter is valid since, in kinetically controlled systems, mass transfer is directly proportional to interfacial area. Therefore, column height must be increased which often does cause problems. To overcome these disadvantages, it is desirable to consider not only chemical kinetics but also other physical phenomena such as axial mixing and drop size distribution in the calculation of columns. This can be very helpful in the performance of pilot plant experiments. The simulation may be divided into three steps:
-

Quantification of phenomena occurring at or near the interface, i.e. mass transfer and chemical reaction. Quantification of hydrodynamic behaviour of the dispersion inside the column, whose major aspect is axial mixing. Combination of steps 1 and 2 in a computer simulation programme.

i9

Fig. 1. Chemical structure of copper- Acorga P 50 complex.

2 Mass Transfer and Chemical Reaction

Selection of a mass transfer model including chemical reaction involves a decision on the location of the chemical reaction, i.e. whether it occurs in the aqueous phase, the organic phase or at the interface. Location at the interface is plausible since, as a rule, components dissolved in the aqueous phase are practically insoluble in the organic phase and vice versa. Therefore, mass transfer is assumed to proceed in the following steps:
-

Whereas conventional liquid-liquid extraction is a physical mass transfer process, in reactive extraction, chemical reactions also occur which may be the rate determining step of the overall process. Reactive extraction is nowadays usually carried out in mixer-settlers on account of easy scale-up and good adaptability to varying process parameters. Calculations of reactive extraction in mixer-settlers are published elsewhere [2]. Columns provide the advantage of smaller space requirements but fluid dy-

Dipl.-Ing. Dr. H.J. Bart, Dipl.-Ing. A. Bauer, and Prof. Dr.-Ing. R . Marr, Abteilung Thermische Verfahrenstechnik, Technische Universitat Graz, Inffeldgasse 25, A-8010 Graz.

Metal ion and extractant diffuse to the interface from aqueous and organic phases respectively. Metal ion and extractant react according to Eq. (1) forming the complex. The complex and released H ions diffuse from the interface into organic and aqueous phases respectively'.
+

0 VCH Verlagsgeaellschaft mbH. D-6940 Weinheim, 1987

0930-7516/87/0510-0291 $02.50/0

292
Me2+

Chem. Eng. Technol. 10 (1987) 291 -296

+ 2 RH

t )

MeR,

+ 2 H+

(1)

3 Phase Equilibria
Equilibrium distribution of the extracted material between organic and aqueous phase can be described by a simple equation. Setting the chemical reaction rate equal to zero, to fulfil the condition of equilibrium, Eq. (6) can be transformed into: [MeR2I [Hf12 [Me2 + [RHy

Eq. (1) explains why the extraction Process is accompanied by a change in PH. Concentration profiles from interface to bulk, as shown in Fig. 2 may, according to the two-film theory of Whitman [3], be assumed as linear. Mass transfer rates per unit area are proportional to the concentration differences between the bulk and the interface and to the partial mass transfer coeffcients. These, in turn, are proportional to the diffusivities of the species (according to the two-film theory). On the basis of

Kex =

(8)

+ pH 2.0
0.
1.

2.

3.

4.

Aqueous phase

Interface

Organic phase

Concentration (aqueous phase1 Fig. 3. Equilibrium curves for copper/5 vol-% Acorga PT 5050.

Fig. 2. Concentration profiles around the interface.

Assuming continuity (no storage effects) the reaction rate must be equal to the transfer rates. Reaction rate is calculated according to Eq. (6) or (7). Eq. (6) represents a first order reaction, proposed by Nitsch 141 and Teramoto [5], including a term for the reverse reaction. This equation should be valid for both extraction and reextraction since it accounts for the reverse reaction. Nonetheless, Albery [6] proposed a different rate equation for the stripping process (7). Combination of Eqs (2) - (5) and (6) or (7) yields concentration versus time profiles. The diffusional resistance of H ions is neglected because their diffusivity is much higher than that of the other species, i.e. the interfacial H concentration can be set equal to the bulk concentration and Eq. (5) is no longer taken into account:
+
+

This implies that D is not constant but decreases with decreasing concentration of the free extractant [RH] and with increasing H concentration. K,, is the individual extraction constant of the system. Equilibrium distribution curves shown in Fig. 3 fall into two regions. In the first, the distribution coefficient is practically constant while, in the second one, the organic phase is saturated. The curve reaches a plateau corresponding to maximum capacity of the organic phase which depends on pH.
+

4 Column Hydrodynamics
Hydrodynamic features of a column are e.g. convection and axial mixing. The latter implies that, with respect to mixing, column behaviour deviates from the ideal plug flow. Axial mixing reduces the concentration differences which are the driving force of mass transfer and, thus, determine performance. Axial mixing can be described by various models, such as the tank cascade, backflow and dispersion models. In the present work, the dispersion model was chosen which describes axial mixing by a single parameter D,,, used in analogy to the usual diffusion coefficient and measured via residence time distribution of the

i'lR = k R

[Me2+] [RHI2 -~ kR [H+I Kex

[MeR,] [ H + ]

(7)

1) List of symbols at the end of the paper.

Chem. Eng. Technol. 10 (1987) 291 -296 phases in the column. D,, is correlated with the Bodenstein number (Bo) via a characteristic length and the axial velocity of the liquid. The overall mass balance of each species around any column section yields the usual mass transfer equation for continuous operation:

293
Table 1. Concentration and pH' Ranges of Stirring Cell Experiments.
Aqueous phase initial copper conc. initial pH Organic phase extractant active extractant (P 50) conc. diluent CuSO,, H,SO, 0.05 - 12.40 [g/l] 1.0 -4.0

5 vol-% Acorga PT 5050

0.090 [mol/l] Shellsol T

for the organic phase and tionally, the aqueous phase was analyzed by atomic absorption spectroscopy before and after the experiment. Furthermore, the pH was continuously recorded. Each experiment was terminated after 5 - 6 h on attaining 30 - 50% of equilibrium. Table l shows the concentration and pH ranges of the experiments:

for the aqueous phase. The first term accounts for axial mixing, the second for convection and the third represents the overall mass transfer including diffusional transport to and from the interface as well as the rate of chemical reaction. Naturally, the p-values in columns are not the same as those measured in the stirring cell on account of different hydrodynamic conditions. These values must be calculated by suitable equations such as that of Linton-Sutherland [7] which employs diffusivities via the dimensionless numbers Re and Sc. Mass transfer rate n per unit area is calculated by simultaneous solution of Eqs (2)-(4) and (6) or (7). Eqs (10) and ( 1 1) are solved by numerical integration with the following boundary conditions which are approximations since the dispersion model is valid only for an infinitely long column. At the bottom of the column:

5.2 Stirring Cell

There are two principal techniques for the measurement of interfacial mass transfer processes:
- rising/falling drops and
-

stirring cell/constant interface cell.

dC,q
dh

A stirring cell shown in Fig. 4 was chosen for mass transfer experiments aiming at the determination of kinetic parameters because of known interfacial area and hydrodynamics. The cell contains 0.5 1 each of aqueous and organic phase. The relatively small interfacial area per unit volume of 3.7 m2/m3is one of the drawbacks of the stirring cell because it restricts the measurements to the initial stage of the reaction and the proximity of equilibrium cannot be examined.

and

At the top of the column: dc01 ~- 0 and

dh

5 Experimental
5. I Chemicals

Aqueous acidic copper sulphate solutions were contacted with the organic phase both in the stirring cell and in the column. In addition, equilibrium data were obtained in shaking funnel experiments. The copper-Acorga PT 5050 complex is brown in colour and, therefore, its concentration in the organic phase could be continuously monitored by a spectrophotometer. Addi-

Fig. 4. Stirring cell. 1 stirrer, 2 PP flange, 3 glass tube, 4 steel draft tube with baffles, 5 organic phase, 6 interface, 7 aqueous phase.

294 Each phase was stirred by a four-blade impeller, located within a draft tube in order to achieve a well-defined phase circulation. The aim of this arrangement is to obtain a smooth interface and a radial flow parallel to it [4].
1.5
I

Chem. Eng. Technol. I0 (1987) 291 -296

,
I

. I

, , , ,-

5.3 Column
Experiments with a countercurrent extraction column (SHE 80 mm in diameter and 2 m in active height [8]) were performed in order to verify the results of computer simulations. After each column experiment, the loaded organic phase was stripped by 2 M sulphuric acid and re-used.

+
v1

m VI

E
m ._ +
.w

- 0

6 Results
6.1 Stirring Cell Experiments There are two methods of evaluating k,:
- From the initial mass transfer rate or - by integration of the mass transfer and reaction rate equa-

0.0 0.

'

'

'

'

'

'

'

'

'

'

'

5.

10.

15.

I n i t i a l copper concentration 1g/1]

Fig. 5. Initial mass transfer rates as functions of initial pH and initial copper concentration.
3.0

tions (Eqs (2)-(7)) over the whole period of each experiment.

2.5
c

As a rule, the first method is employed since, at a considerable distance from equilibrium, the concentration profiles are almost linear and the reverse reaction can be neglected. This is particularly true for rising/falling drop experiments where contact times are significantly shorter than those in the stirring cell. In this case, equilibrium had been achieved to the extent of 50% and, consequently, the reverse reaction must be taken into account.
Models, which do not consider the reverse reaction, are far less capable of fitting the experimental data, even at the start of the experiment when the reverse reaction is least important. Therefore, k, and fl were determined by numerical integration of Eqs (2)-(7) and fitting the resulting concentration versus time curves to the measured ones. The problem is to determine whether diffusion or chemical reaction is rate controlling. The objective of the stirring cell experiments is to quantify chemical kinetics and to eliminate diffusional resistances as far as possible since diffusional parameters measured in the stirring cell would not be valid in mixer-settlers or columns on account of different fluid dynamics. One method of separating diffusion and reaction influences is to vary the impeller speed. If the overall mass transfer rate becomes independent of stirrer speed, it can be assumed that diffusion effect had been eliminated. A variation of the overall mass transfer rate with pH implies that chemical reaction is rate limiting because resistance caused by H + diffusion can almost certainly be neglected. k, is determined in the kinetic regime, i.e. at low metal concentrations and low pH, where the influence of transport phenomena is minimal and chemical reaction is assumed to be rate determining. Fig. 5 represents the stirring cell experiments. The initial mass transfer rates depend not only on the initial aqueous metal con-

,z 1 . 5 t
\ m

2.0

. ,'
'\

-.

1
-1
6.

+
W c

m L

g
U

1.0

0.

2.
Time [ h l

4.

Fig. 6. Stirring cell experiment; initial pH 3.1.

3.0

2.5

.o +
+ L c W
m

1.5

1.0

0.5

0.

20.
Time \hl

40.

rn

1
60.

Fig. 7. Stirring cell experiment; initial pH 3.1 extrapolated to 50 h.

Chem. Eng. Technol. I0 (1987) 291 -296

295

centration but also on pH, which suggests a kinetic limitation throughout the whole concentration range. Fig. 6 shows a plot of concentration versus time for an initial aqueous copper concentration of 0.02 mol/l and an initial pH of 3.1. The interfacial concentration of the complex is comparatively high at the beginning of the experiment which indicates that the interface is saturated by complex molecules. This would confirm the suggestion of very fast adsorption kinetics and a subsequent slow diffusion of the complex into the organic bulk.

,.s
1 .O

m ._
W

k , values were found to be scattered in the range of 0.0006 to 0.002 cm/s which indicates that the model does not describe the chemical reaction correctly. Fitting of the reaction orders to experimental values showed that the discrepancies are due to the incorrect order of reaction with respect to H + which is between -0.2 and -0.3, in contrast to - 1 assumed in the model. The order of reaction with respect to the metal ion was determined as 2, the extractant but, in fact, it is with respect to the metal ion and 1 with respect to the extractant. Orders of reaction, determined by Lorbach [9] for this system, are in good agreement.
Fig. 7 shows the simulation of a stirring cell experiment, extrapolated to a period of 50 h. The interface concentrations approach those in the bulk as equilibrium is attained and mass transfer comes to a standstill. This relatively long delay in reaching equilibrium is due to the small interfacial area per unit volume of the stirring cell, compared to the conventional extraction equipment.

0.s -

0.0 0.00

0.01

0.02

lclsel

Concentration [rnol/ll

Fig. 8. Concentration profiles in the countercurrent column; comparison of

simulation and experiment.

Table 2. Results of Column Experiments, System CopperiAcorga PT 5050.

[ M ~ ' + I , PH, [gill
uaq

"or

xd

Recovery HETS

[-I
1.00 1.00 1.02 3.00 3.00
.~

[l/hl

Whl

[%I
23.1 24.0 30.2 24.0 30.2

[%I
81.4 63.5 42.6 99.8 93.8
~ ~~

[ml
1.62 2.89 5.34 1.47 3.20

1 .o 1.o
1 .o 1 .o 1 .o

10 20 50 20 50

10 20 50 20 50

6.2 Column Experiments

Eqs (10) and (1 1) are solved numerically by the method of finite differences, simultaneously solving Eqs (2) - (7) represented by the third term in Eq. (10) or (11). Mass transfer coefficients were calculated from Sh numbers. The drops were approximately 1 mm in diameter which was determined from photographs. With respect to this diameter, a constant Sh number of 6.58, valid for rigid spheres, was assumed. The Sh number for the aqueous phase was calculated according to the LintonSutherland equation [7], resulting in the following @values:

Axial mixing of aqueous phase was estimated from the initial concentration step at the top of the column. Eqs (15) and (16) yield Eq. (17), where Bo is the Bodenstein number based on a characteristic length of 1 m. The experiments showed a relatively high axial mixing depending on throughput (Boa, = 2 for 10 m3/m2hor Boa, = 4 for 40 m3/m2h). The respective dispersion coefficient D,, is 0.001 m2/s: dcaq dh
'

Boa,

"a,
~

, Boa, =
',,in

(161, (17)

Daxaq

- %h=Hk

PMe = 0.012 (cm/s), /3MleR2 PRH = 0.00015 ( a d s ) .

0.00014 (cm/s), Experimental concentration profiles could be simulated within a scatter range of f lo%, the errors resulting mainly from the kinetic model. The influence of axial mixing lies within the scatter range.

Computer simulation yields concentration profiles of all species in the bulk and in the interface (Fig. 8). By using the equilibrium function (Eqs (8), (9)), it is possible to calculate the number of ideal transfer stages NSTachieved in the column. The respective HETS values which denote the column height corresponding to one theoretical stage, found in column experiments, are listed in Table 2. Copper/Acorga PT 5050, known to be a very slow reaction, gave Ns, values mostly less than unity, depending on phase throughput. In solvent extraction, however, one theoretical stage can achieve as much as 99% recovery. Earlier work [8] on the system copper/LIX 64N also produced very low HETS values.

7 Conclusions
Parameters of a model for interface mass transfer and chemical reaction in reactive extraction were determined by measuring concentration-time profiles in a stirring cell. These parameters were used in a column simulation programme together with hydrodynamic parameters. The resulting concentration profiles along the column permitted the evaluation of column perfor-

296
mance. Measured and calculated concentration profiles show a satisfactory agreement. It was found necessary to account for the reverse reaction and the influence of pH. Although axial mixing was comparatively high in the employed column, the influence of chemical kinetics proved to be much stronger than that of axial mixing. The high HETS values indicate that a mixer-settler is more suitable for the investigated process, on account of lower capital costs. With kinetically controlled systems, long residence times are required and the main advantage of columns, namely maintenance of high concentration gradients, is less important.
v
v
'd

Chem. Eng. Technol. 10 (1987) 291 -296

[m/sl [l/hl effective convection velocity phase throughput hold-up

1-1

Subscripts
0 aq ax bulk ex F H in int Me MeR, or R RH

Acknowledgement
This investigation was supported by the Oesterreichische Nationalbank, Project 2467.

initial aqueous phase axial bulk extraction forward proton inflow interface metal ion metal complex organic phase reverse extractant

Received: November 7, 1986 [CET 181

References
[I] Marr, R., Bart, H.J., Chem.-Ing.-Tech. 54 (1982) No. 2, pp. 119- 129. [2] Bauer, A , , Bart, H.J., M a r , R., Chem.-Ing.-Tech. 58 (1986) No. 2, pp. 154- 155. [3] Whitman, W.G., Chem. Metal. Eng 29 (1923) p. 147. [4] Nitsch, W., Ber. Bunsenges. Phys. Chem. 83 (1979) pp. 1171 - 1177. [5] Teramoto, M., et al., Extraction of Copper with SME 529 J. Chem. Eng Jpn 16 (1983) No. 3, p. 203. [6] Alhery, J.W., Fisk, P.R., The Kinetics of Extraction of Copper with Acorga P 50 studied by a diffusion cell. Proceedings of Hydrometallurgical Symposium, Manchester 1981. 171 Linton, M., Sutherland, K.L., Chem. Eng Sci. 12 (1960) pp. 2 14 - 229. (81 Gauhinger, W., Max, R.,Calculation of Counter-Current Extraction Columns Using Mass Transfer Propenies of Single Drops and the f the Dispersion, AIChE Summer meeting Hydrodynamic Behaviour o 1984, 19.122. 8. 1984, Philadelphia, Pa. Paper 47a. [9] Lorbach, D., Bart, H.J., Marr, R., Ger. Chem. Eng 9 (1986) No. 5, pp. 321-327. [lo] Bauer, A,, Bart, H.J., Marr, R., Numerical Calculation of Extraction with Chemical Reaction 5th Italian-Yugoslavian-Austrian CEC, 16. - 18. 9. 1986, Portoroz (YU). [ l l ] Miiller, H., Diplomarbeit, TU Graz 1985. [I21 Martinak, A,, Diplomarbeit, TU Graz 1986.

Symbols used
interfacial area per unit volume Bodenstein number (based on 1 m) partial mass transfer coefficient (metal ion) partial mass transfer coefficient (complex) partial mass transfer coefficient (extractant) partial mass transfer coefficient (proton) concentration equilibrium distribution coefficient axial dispersion coefficient column height coordinate hydrogen ion concentration (aq. phase) height of a theoretical stage active column height reaction rate constant equilibrium constant metal ion concentration (aq. phase) complex concentration (org. phase) mass transfer rate, reaction rate number of ideal transfer stages extractant concentration (org. phase)