Determination of Liquid-Liquid Distribution Coefficients by Gas-liquid Chromatography Richard
J. Sheehan' and Stanley H. Langer
Department of Chemical Engineering, University of Wisconsin, Madison, Wis. 53706
Liquid-liquid distribution coefficients of o-xylene between two immiscible and unreactive liquids suitable as stationary phases were predicted from the chromatographic retention volumes of o-xylene in these liquids. Retention volumes and infinite dilution activity coefficients are given for several hydrocarbons in squalane, Carbowax 750, triethanolamine, oxydipropionitrile, a n d tris(2,4-dimethylphenyl) phosphate stationary phases
80"C. Distribution coefficients are predicted where the phases are immiscible.
at
Procedures for tesing are given. This rapid method is useful for screening potential liquid extraction solvents or solvent pairs for liquid-liquid chromatography.
T h e wide commercial use of solvent-assisted separations has led to many suggestions for choosing solvents with a minimum of experimental work. A number of the approaches and concepts are reviewed by Kyle and Leng (1965), Tassios (1969), and Treybal (1963). Most methods involve predicting activity coefficients in the two phases for the purpose of calculating distribution coefficients. While gas chromatography has long been recognized as a means of obtaining activity coefficients a t infinite dilution, y-, of eluted materials in stationary phases (Porter et al., 1956; Kwantes and Rijnder, 1958; Langer and Purnell, 1963), such readily experimentally determined data have rarely, if ever, been applied t o prediction of liquid-liquid distribution coefficients. Furthermore, y e data may be useful over a relatively wide concentration range (Treybal, 1963). Gas chromatography has been used in screening extractive distillation solvents (Doring, 1961; Porter and Johnson, 1960; Rock, 1956; Sheets and Marchello, 1963; Sideman, 1964; Warren et al., 1959), and good agreement with accepted data was found. However, we are aware of only the attempt by Lesteva et al. (1966) to relate gas chromatography to liquid-liquid extraction processes, and then in a way different from that described here, This study is directed toward examining the potential usefulness and applicability of the gas chromatographic data available in the literature to liquid-liquid extraction processes. Theory and Background
A solute will be distributed between two immiscible liquids, subscripted 1 and 2, so that its activity will be the same in both phases, or XlYl XIIX2
= X?Y2 = Y2lYl
(1) (2)
Also, y = can be related to V i by
Y = = RT/MV;PO
(3)
The derivation of this equation and the experimental ' Present address, Amoco Chemical Corp., 2500 New York Ave., Whiting, Ind. 46394
44
Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 1, 1971
conditions needed to insure a state of infinite dilution in the column have been discussed by Martire and Pollara (1965), Porter et al. (1956), Langer and Purnell (1963), and Sheehan (1969). From Equations 2 and 3
Changing mole fractions into weight fractions and making appropriate simplifications allowed by the infinite dilution condition yields W l l W? =
V $ Jvi,
(5)
This retention volume ratio measured a t column temperature T should equal the weight fraction distribution coefficient of solute a t infinite dilution between the two immiscible liquids maintained a t temperature T previously employed as stationary phases. A knowledge of the molecular weights and densities of the stationary phases allows calculation of mole fraction, volume fraction, or molar concentration ratios; the values of the solute vapor pressure allow calculation of individual activity coefficients from Equation 3. The speed and simplicity of data determination by gas chromatography makes compilation of y = and/or V l values convenient. Distribution data for any pair of potentially useful stationary phases could be obtained by taking or a simple ratio. With such a ratio, any errors in V i resulting from the nonideality of the solute in the carrier gas phase will cancel if measurements are made under approximately the same conditions (Conder and Langer, 1967). When static measurements are difficult owing to very low solubility, the chromatographic method may be preferred. The miscibility, reactivity, and stability of the two phases in contact a t the temperature of operation is the other information needed. Since small amounts of extraneous material in a liquid can alter y e for a solute dissolved in that solvent (Baumgarten and Gerster, 1954), more accurate data could be obtained by allowing slightly miscible liquids of interest t o equilibrate in contact with each other before using them as stationary phases. However, this would require the analysis of each system pair separately, and the time savings which can be realized by testing each stationary phase only once would be lost. The chromatographic method is best suited for screening
purposes, and attractive systems should be tested in actual operation until essentially immiscible systems are identified. Also, caution must be exercised with some stationary phases since solutes can sometimes adsorb on the liquid surface (Martire, 1968) resulting in erroneous values of
SERUM CAPS
TEFLON FILM
Y .
Lesteva et al. (1966) have reported what is apparently the only attempt to obtain liquid extraction data by gas chromatography. Since they were interested in extracting higher boiling compounds from volatile solvents, Equations 4 and 5 were not directly applicable. By limiting their study to compounds which show similar activity coefficient us. concentration behavior, they assumed from the Redlich and Kister (1948) equation that the activity coefficient ratio of a solute in two volatile extracting agents was the same as the ratio for the extracting agents in the solute. Therefore, only distribution coefficient ratios for any two of a series of extracting agents with a common solvent and solute could be determined. The solute was used as the stationary phase. Good agreement was found between static and chromatographic relative distribution coefficients for systems such as the acetonitrile distribution between water, a common solvent, and several Cshydrocarbon extracting agents. Experimental
Infinite dilution activity coefficients were determined a t 80° C for 10 solutes (C7 to C9 n-alkanes and 1-alkenes, benzene, toluene, o-xylene and methyl toluene) in five stationary phases [squalane, Carbowax 750, triethanolamine (TEA), 3,3'-oxydipropionitrile (ODPN), and tris(2,4-dimethylphenyl) phosphate ( T D M P P )] using previously described procedures (Langer and Purnell, 1963; Conder, 1968). The distribution coefficient of o-xylene was determined statically in the U cell shown in Figure 1. A connector at the top allowed both liquids to be in contact with the vapor. The cell was fabricated from 10-mm tubing and contains about 5 ml of the heavier phase, 2 ml of the lighter phase, and 20 p1 of solute to approximate infinite dilution. The solute is injected into the phase in which it is least soluble to speed equilibration. The cell allows sampling by syringe through the septum and protective Teflon liner. Each system was equilibrated in a constant temperature bath a t 8 P C for one week. The phases were sampled with a Hamilton 10-pl syringe and analyzed by gas chromatography using the injector shown in Figure 2. The immiscible liquids, stationary phase type materials, would not vaporize to any significant extent a t any reasonable injector temperature. A heated stainless steel block (cartridge heater) with a horizontal vaporization chamber lower than the carrier gas connections served as the injector. The solute vaporized from pooled liquid phase. The resulting peaks tailed somewhat, but areas could be measured by planimeter with adequate accuracy. Each peak area was divided by the volume of the phase injected and multiplied by the molar volume to give the area per mole of solvent. Division of the two corrected areas results in the mole fraction ratio. Results
The densities and molecular weights of the stationary phases are listed in Table I. These permit calculation of the concentrations in the static systems. Where densities
LIGHTER LIQUID
HEAVIER LIQUID
IO MM TUBING Figure 1. Liquid-liquid extraction cell
-r
-n
SEPTUM
1
I
k2.5 CMA
L
-
5 CM,-I
Figure 2. Gas chromatographic injector
Table 1. Densities and Molecular Weights of Stationary Phases at 80' C Stationary Phase
Density
Molecular Weight
Squalane Carbowax 750 TEA
0.770 1.058 1.089 0.999 1.094
422.8 750 149.2 124.2 410.5
ODPN TDMPP
are almost equal, phase separation may take an excessive amount of time. Table I1 lists V i and y*. The relative affinity of the polyglycol (Carbowax 750), TEA, and ODPN for aromatics is illustrated by the retention of benzene relative to the much higher boiling nonane and nonene. The opposite is true for the aliphatic squalane where the specific attraction between squalane and the saturated hydrocarbons is considerably greater. The high molecular weight of some of the phases results in small activity coefficients when Equation 3 is used. Size difference effects have been discussed by us earlier (Langer and Sheehan, 1968). Activity coefficients emphasize again the specific affinity for aromatics relative to aliphatics for some of the liquid phases. Table I11 lists mole fraction ratios for four immiscible pairs of liquids with hydrocarbon solutes distributed between them. These values are the inverse of the activity coefficient ratios. The o-xylene distribution was checked statically, and good agreement was found for the squalaneCarbowax 750 and squalane-ODPN systems. The high viscosity of TEA limited the amount which could be drawn into the syringe. The resulting small peak was difficult to measure, and so the static value could be in error Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 1, 1971
45
Table II. Specific Retention Volumes and Activity Coefficients at Infinite Dilution of Solutes at 80’ C Corbowox 750
Squolone Solute
n-Heptane n-Octane n-Nonane 1-Heptane 1-Octane 1-Nonene Benzene Toluene o-Xylene rn-Ethyltoluene
V
A
187.6 433.6 996.0 161.6 372.6 857.7 116.0 287.7 819.2 1415.0
Y 0.649 0.687 0.723 0.647 0.686 0.719 0.592 0.622 0.670 0.708
V:,
Y
19.0 36.3 69.1 25.1 47.7 90.4 96.9 182.0 444.9 573.1
~~
3.62 4.63 5.88 2.35 3.02 3.84 0.401 0.554 0.695 0.985
Corbowox 750/ Squolone
TEA/ Squolone
ODPN/ Squolone
TEA/ TDMPP
n-Heptane n-Octane n-Nonane
0.180 0.148 0.123
0.00959 0.00716 0.00538
0.0102 0.00738 0.00537
0.0345 0.0274 0.0220
1-Heptene 1-Octene 1-Nonene
0.275 0.227 0.187
0.0146 0.0108 0,00802
0.0196 0.0141 0.0101
0.0403 0.0318 0.0252
Benzene Toluene o-Xylene (Static) rn-Ethyltoluene
1.478 1.122 0.963 (0.966) 0.718
0.0980 0.0672 0.0535 (0.0456) 0.0351
0.200 0.139 0.112 (0.113) 0.0713
0.107 0.0791 0.0638 (0.0989) 0.0481
by a considerable amount. Considering this, the agreement between the static and chromatographic data is quite good for these systems. The T E A - T D M P P system apparently reacted as evidenced by the formation of a precipitate and lack of agreement between the two methods. The chromatographic values are probably the values for the unreacted system. The alkanes and alkenes have low solubility in TEA and ODPN. The solubility of aromatics is greater, but while the mole fraction ratios of Table I11 are of interest from the interaction point of view, the retention ratios which can be obtained from Table I1 may be of greater interest in terms of practical extraction by weight. Data of this table show the strong selectivity of ODPN and Carbowax 750 for aromatics relative to alkanes and alkenes. The distribution coefficients decrease as the solute size increases. The larger solutes tend to go to the squalane and T D M P P phases more than the smaller solutes. The alkanes and alkenes probably show this decrease because the smaller members can fit into the polyether or other phases with less disruption than can the larger molecules. The Hildebrand-Scatchard expression qualitatively takes this into account (Langer and Sheehan, 1968). The aromatic solutes show a sharper drop upon increase in size, but here the aliphatic nature of the solute increases as well as size. This relation between distribution coefficients and retention volumes could be used for purposes other than screening extraction solvents. The data might suggest potential liquid-liquid chromatographic systems. “Theoretical” distribution coefficients could be found for liquid pairs where miscibility or reactivity preclude the determination of static values. Agreement between static and 46
ODPN
V;
Y
V,’
5.1 8.8 15.2 6.7 11.4 19.5 32.3 54.8 124.3 140.7
67.7 96.0 134.3 44.2 63.5 89.6 6.04 9.25 12.51 20.2
6.5 10.9 18.2 10.8 17.9 29.6 79.4 136.1 313.3 343.8
Ye 63.8 93.1 134.7 33.0 48.6 70.9 2.95 4.48 5.96 9.92
TDMPP V,‘
53.8 116.7 250.7 60.4 130.5 280.7 110.1 251.9 708.5 1064.0
Y? 2.33 2.63 2.96 1.78 2.02 2.26 0.644 0.731 0.798 0.970
~
Table 111. Mole Fraction Ratios at 80’ C
Solute
TEA
Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 1, 1971
chromatographic distribution determinations probably indicates that the liquids involved are immiscible and unreactive. Such agreement is also further confirmation that equilibrium vapor-liquid partition coefficients can be determined chromatographically (Porter et al., 1956). A knowledge of a mole fraction distribution coefficient and one y r gives the other yl and therefore a prediction of chromatographic behavior of a solute or pair of solutes in a volatile or untested stationary phase. This might be particularly helpful for the study of complex, thermally unstable solutes such as steroids which may not chromatograph easily. Chromatographic data could be used to check the accuracy of data determined by other means. A V: us. temperature determination would allow prediction of distribution coefficients a t various temperatures. The major disadvantages of this chromatographic screening method are the need to use nonvolatile stationary phases together with the solutes a t infinite dilution only. The state of infinite dilution in which the solute is completely surrounded by the extracting agent is of interest, however, because it indicates the limiting separation for the solvents. Also, y z reflects the distribution of the system during the final stages of extraction when product purity is determined. Circular column techniques to permit the use of volatile “stationary” phases are being developed (Porter and Johnson, 1960; Sideman, 1964) as is the employment of finite dilution conditions in the column (Conder, 1968). These could greatly expand the application of the gas chromatographic solvent screening method to liquid-liquid extraction studies. Nomenclature
M p“
R T VRT u: X
Y Y‘
molecular weight of the stationary phase vapor pressure of the solute, mm Hg gas constant absolute temperature of the column, ’ K corrected retention volume of solute per gram of stationary phase a t column temperature T , mlig weight fraction of solute mole fraction of solute activity coefficient activity coefficient of solute a t infinite dilution in the stationary phase
Literature Cited
Baumgarten, P. K., Gerster, J. A., Ind. Eng. Chem. 46, 2396 (1954). Conder, J. R., “Progress in Gas Chromatography,” pp. 209-70, J. H. Purnell, Ed., Interscience-Wiley, New York, 1968.
Conder, . R., Langer, S.H., Anal. Chem. 39, 1461 (1967). Doring, C. E., 2. Chem. 1, 347 (1961). Kwantes, A., Rijnders, G. W. A., “Gas Chromatography,” pp. 125-35, D. H. Desty, Ed., Academic Press, London, 1958. Kyle, B. G., Leng, D. E., Ind. Eng. Chem. 57 (2), 43 (1965). Langer, S. H., Purnell, J. H., J . Phys. Chem. 67, 263 (1963). Langer, S. H., Sheehan, R . J., “Progress in Gas Chromatography,” pp. 289-323, J. H. Purnell, Ed., Interscience-Wiley, New York, 1968. Lesteva, T. M., Gorodnikov, S. K., Zheleznyak, A. S., Z h . Prik. Khim. 39, 1628 (1966). Martire, D. E., “Progress in Gas Chromatography,” pp. 93-120, J. H. Purnell, Ed., Interscience-Wiley, New York, 1968. Martire, D. E., Pollara, L. Z., “Advances in Gas Chromatography,” pp. 335-61, J. C. Giddings, Ed., Marcel Dekker, New York, 1965. Porter, P. E., Deal, C. H., Stross, F., J . Amer. Chem. Soc. 78, 2999 (1956).
Porter, R . S., Johnson, J. F., Ind. Eng. Chem. 52, 691 (1960). Redlich, O., Kister, A. T., Ind. Eng. Chem. 40, 345 (1948). Rock, H., Chemie Ing. Tech. 28, 489 (1956). Sheehan, R . J., Doctoral thesis, University of Wisconsin, 1969. Sheets, M. R., Marchello, J. M., Petrol. Refiner 42 (12), 99 (1963). Sideman, S., Bull. Chem. SOC.Japan 37, 1565 (1964). Tassios, D., Chem. Eng. J . 76 ( 3 ) , 118 (1969). Treybal, R. E., “Liquid Extraction,” 2nd ed., McGrawHill, New York, 1963. Warren, G. W., Warren, R. R., Yarborough, V. A., Ind. Eng. Chem. 51, 1475 (1959).
RECEIVED for review August 13, 1969 ACCEPTED August 3, 1970 Supported in part by the Petroleum Research Fund, the E. I. duPont de Nemours and Co. (Fellowship for R . J. S.),and the Wisconsin Alumni Research Foundation.
Behavior of a Chromatographic Reactor Chieh Chu’ and louis C. Tsang2 University of California, Los A ngeles, Calif, 90024 The behavior of a chromatographic reactor was studied by use of the LangmuirHinshelwood kinetic model to account for the competitive adsorption on the catalyst surface. The effects of various parameters such as input wave form, reverse reaction rate constant, average reactant concentration in the feed, adsorption equilibrium constants, and active center concentration were investigated. Some limited study of the effect of longitudinal dispersion was also included.
M a n y chemical reactions cannot go to completion because of the existence of reverse reactions. In a chromatographic reactor, the reactants are introduced in pulses. If the products are adsorbed on the catalyst in the reactor to different extents and for different lengths of time, these products can be separated from one another, thus diminishing the rate of the reverse reaction. In this way, an ordinarily equilibrium-limited reaction can be carried t o completion or much nearer completion. That this is feasible has been shown experimentally by Roginskii et al. (1962) on the dehydrogenation of cyclohexane to benzene and by Semenenko et al. (1964) on the dehydrogenation of n-butene to divinyl. As for theoretical analyses, Roginskii and Rozental (1964) studied reaction kinetics under chromatographic conditions. Gaziev et al. (1963) investigated the effect of the reactant pulse shape and the order of chemical reaction on conversion. Magee (1963) used a mathematical model in terms of gas phase concentration. Gore (1967) improved on this by using the concentrations of adsorbed species instead, but the power law model was still retained. ’ T o whom correspondence should be addressed.
‘Present address. Monsanto Co., St. Louis, Mo. 63166
Since chemical reactions in a chromatographic reactor take place on the surface of a catalyst with a limited number of active centers, it is important in analyzing the performance of a chromatographic reactor to account for the effect of competitive adsorption. This work uses the Langmuir-Hinshelwood kinetic model instead of the power law model. Furthermore, longitudinal dispersion is important in the chromatographic phenomenon (Gore, 1967; Magee, 1963; Roginskii and Rozental, 1964). This work also includes this phenomenon. With these two factors included, this work investigates the effects of various pertinent parameters on the behavior of a chromatographic reactor. These parameters are: input wave form, reverse reaction rate constant, average reactant concentration in the feed, adsorption equilibrium constants, active center concentration, and longitudinal dispersion. I t should be mentioned however, that, because of the particular numerical procedure being used, the effect of longitudinal dispersion was studied only t o a limited extent. Reaction Kinetics
The following reaction is considered: When reactant A enters the reactor, it is adsorbed on the catalyst surface inside the reactor, and dissociates into two products, one Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 1, 1971
47
