ANALYTICAL CHEMISTRY
1346 in the combustion of methane, a saturated hydrocarbon, to carbon dioxide and water by reaction with excess oxygen. The carbon dioxide is then absorbed. Magnetic Action. In a few cases the separative process involves magnetic action on the desired constituent. (The author suggests the term “magnetition” for this kind of separation.) Two variations of the process seem evident, the seat of the variance being the nature of the entity attracted. In the first variation, old but relatively uncommon, one component of the sample is magnetic and the others are not. Thus, iron particles might be mixed with sand. The metallic component may be separated by mcans of a magnet brought close to the mixture. Obviously, the possibilities are limited. In the second variation, relatively new and increasingly important, ionizable constituents are ionized and then subjected to a magnetic force, as in a mass spectrometer. Because of the difference of the ratios of mass to charge for most constituents, electrical or other differentiation of the presence and the amounts of different constituents is possible. Components SO separahle are either gaseous or readily convertible to this state. EDUCATIOVAL PROBLEMS
In this broad consideration of separations the author, as a teacher, ventures to raise two educational questions-namely, nomenclature and course content. Only brief mention of the nature of the problem is possible. Nomenclature. Because the separative operation is so important in the analytical methods for most multicomponent systems, and because the measuring operation occurs in every quantitative determination, it is suggested once more that the general nature of each of these two unit operations be indicated in naming the over-all method. Anyone having occasion to make much Use of indexing serials and abstracting journals must have been irked by the lack of such functional nomenclature. The author, for example, has spent much time trying to find out whether a given method is coiorimetric and how the desired constituents are separated. Suitable naming and indexing would have made such work unnecessary. A combination term would indicate the general nature of both principal unit operations-namely, separation and measurement. A few examples are volatilization-volumetric, precipitation-gravimetric, electrodeposition-titrimetric, and extraction-absorptiometric. If there is no separation, the measurement term alone would suffice. What to Teach. Of more importance to many, no doubt, is the question of what one should try to include on this unit operation in an elementary quantitative course Because
3rd A n n d Summer Sgmpoelum
isohtion of the desired constituents for measurement encompasses most of the chemistry in a procedure, the question really is what chemistry can and should be taught. Broadly considered, the chemistry involved is of two kinds. The first kind is factual. Examples are the photochemical sensitivity of silver chloride and the hygroscopicity of ignited calcium oxide. If the student does enough analyses and is alert to these facts, hc should emerge as something more than a Lundell determinator. He would know that things are, but not why they are. The second kind is interpretative. Because it involves understanding, as far as this is possible, it is the more important. Such understanding of a procedure concerns the how and why of things, or, as some would say, the underlying theory. Basically this theory comprises the physical chemistry relevant to the various unit operations of a method. Taking precipitation as an example of a separative operation, we find concerned such subjects as the following: choice pf precipitant, solubility of precipitates, colloidal dispersions, ionic equilibria, common and diverse ion effects, activity of ions, p B , buffer solutions, purity of precipitates, oxidation-reduction, complexation, washing precipitate3, and igniting precipitates. To interpret all such phenomena relevant to the analytical process in terms of modern physical chemistry is no small assignment. Teachers of physical chemistry know that extensive problem drill, rather than superficial descriptive treatment, is required to fix such concepts in workable form. Mere mention of the subjects by the analytical instructor will do no harm, but probably little lasting good. Most students now reach quantitative analysis with not more than three semesters of foundation chemistry and often without physics. Then the great majority take one semester of quantitative work. On this background one must treat the rapidly expanding subject of chemical analysis. With this statement of the problem, the serious instructor is left to contemplate the dilemma. LITERATURE CITED
(1) ANAL. CHEM.,21, 2-173, 196-284 (1949); 22, 2-136. 206-71 (1950).
JT. C., “Principles of Quantitative Snalysis,” New York, D. Van Nostrand Co., 1914. (3) Cassidy, H. G.,J. Chem. Education, 23, 427 (1948); 27, 241 (1950). (4) Lundell, G. E. F., and Hoffman, J. I., “Outlines of Methods of Chemical Analysis,” New York, John Wiley & Sons, 1938. ( 5 ) Smith, E. F., “Electro-Analysis,” 6th ed., Philadelphia, P. Blakiston’s Son & Co., 1918. (6) Washington, H. S., Chemical Analysis of Rocks,” New York, John Wiley & Sons, 1930. (2) Blasdale;
RECEIVED July 8, 1950.
- Separathns
Partition Chromatography and Countercurrent Distribution LYMAN C . CRAIG, T h e Rockefeller Institute f o r Medical Research, New York, N . Y .
E
VERYONE interested in separation processes soon learns that he must deal with some form of heterogeneous equilibrium involving the selective transport of solute from one phase to another, and that only under the most favorable circumstance will a single transfer suffice to give an adequate separation. Repeated transfers are required. I t was noted many years ago that if the two phases involved
could be made to flow past each other in sufficiently intimate contact, surprisingly good separations could often be achieved. The effect was broadly interpreted in terms of repeated stepwise contacts or transfers. Many investigators have tried to give a satisfactory theory for the continuous process, and every year brings a fresh supply, undoubtedly a reflection on the yet unsatisfactory nature of our understanding.
V O L U M E 22, NO. 1 1 N O V E M B E R 1 9 5 0
1347
Ip evaluating the efficiency of wntinuous fractionation processes such a s partition chromatography, a common procedure is that of calculating the height of equivalent theoretical column. This involves an experimental determination of the effect of a single stepwise equilibrium separation for’ comparison to the effect of passage through the wlumn. This hv no means irnplies a full understanding of the events taking plaoe in the wlumn. In faaot, Imany of the various faotors bearing on the wlumn separation are so interdepend8ent as to be difficult to study clearly. Because wuntercurrent distribution is a SItepwise equilibrium pmeedure whioh has now been carried to several thousand actual stages, it is interesting in the light of the wntinuous pmeess to review t be factors influenting its successful operation. A slightly different concept of the continuous process is thus reached.
Martin and Synge proposed a theory which probably gave more insight into the nature of chromatography than previous viewpoints, even though i t was not entirely correct. Their theory was developed along the lines proposed by previous workers such as Jantzen (7) and Cornish, Arehibald, Murphy, and Evans ( 4 ) in terms of the “theoretical plate” concept. No one doubts that the effects of the ideal diecontinuous process and the truly continuous become nearly identical when high numbers of stages are in the former. It is the reasoning of ealculus. Furthermore, chromatography presents mechmieal conditions which should give almost the equivalent of infinite numbers of stages, provided there is uniform linear flow of solvent with no channeling or turbulence such as that caused by a packing tihich is not uniform; equilibrium conditions+.g., the partition or adsorption isotherm-are satisfied at any point in the column a t a praotical rate of flow; and the isotherm involved is strictly linear. None of these conditions is ever completely met in practice. A measure of the efficiency of a column is therefore an indication o i the combined effects of the various deviations from ideality. Some of the deviations may be favorable to separation, others unfavorable. Figure 1. Autornatio 220-Cell Glass Distribution Train
In gcnerd, the approach has been to formulate a theory and then attempt to prove it by carrying out separations on an artificial mixture under specified conditions in a given extraction train or column. By correlation of effects the result is interpreted lor proof of the theory. The fundamental weakness of this approach lies in the fact that certain assumptions which may or may riot hold during such a complex interplay of events are dways required and the evidence in the final analysis is only circumstantial at best. Often the theory is known to hold for limiting C R S ~ S only and to deviate widely in actual practice. An example of such an approach is the published theory for the technique called “partition chromatography.” Martin and Syngc first experimented with a stage continuous liquid-liquid extrsction train (9) designed for the separation of amino acids. They iound the procedure laborious +nd evolved the ingenious idea of immobiliaing one of the liquid phases by adsorbing it on Some mechanic$ support known to have weak adsorptive, properties, nuch as silica, starch, or paper (10). This permits a mechanical state of affairs most nearly ideal for the oountercurrent effect from both a practicd and theoretical standpoint, because the operation can be performed as simple percolation. They called the process partition chromatograpby in the thought, that it really was a form of liquid-liquid extraction and not based on adsorption. By now nearly every chemist knows something of the great practical value of the method, not to mention the stimulation it has given to t,he further development o i many other types of countercurrent separation processes.
COUNTERCURRENT DlSTRIBZmON
Several )-cars ago it occurred to the writer that a profitable approach to an understanding of many of the h e r points of frrtetionation theory might be reached by intensively studying 8 strictly discontinuous extraction process ( 5 ) carried to high numbers of stages-i.e., “eountewurrent distribution.” High
Figure 2.
Individual Cell of Distribution Apparatus
resolving power was not expected at fist, but it was thought that because the theory of such a process would be Imown, certain &s sumptions neeessrtry for interpretation of continuous processes could be more critically examined. However, the first separations obtained were encouraging and soon the program necessitated the development of adequate mechanical equipment for performing the thousands of extractions desired. This phase of the study is partly covered in the literature (6). The extractor currently in use in the author’s labarstory (Fig. ure 1) contains 220 glass equilibration cells connected in seriw.
A N . A L Y T IC A L C H E M IS T R Y
1348
,,
Glycine Tyrosine It therefore performs 220 Leucine actual extractions in from 1 Alanine to 3 minutes, depending on &amino butyric. Isoleucine the speed of equilibration Valine Phenylalanine and separation of the phases. Methionine Tryptophane, I t is fitted with electric motors, timers, a filling device, a fraction collector, etc., so that the operation is fully autoniatic. It nil1 run without attention through the 8d night. Twenty-four hours of operation give from 100,OOO to 200,000 actual extractions, which are integrated to give the “column” effect. The design of a single cell 3 2.0 is given in Figure 2. Each .-C cell contains 10 ml. of the lower phase, sufficient to bring the interface to a in 1.0 position C. .4n arbitrary 3n volume no greater than 15 ml. of upper phase is used. 0 200 Equilibrium is established by 0 40 80 120 . rocking from position A to B. The phases separate a t posiT u b e NO. TfJansfei. No. tion B. On tilting to position C the upper phase decants Figure 3. Distribution Pattern of Synthetic Mixture of .Amino Acids through c to chamber d. Now System 0 n-butyl alcohol, 5 % HCI on tilting to Dosition A the 0-0 Experimental rontents of d flow through e 0- Calculated to the adjoining Gll. b is a flat ground-glass stopper. Several methods of operating the apparatus are useful. In each method the sample is placed initially either in one cell or, in order to give greater capacity, in a number of the first cells. Fundamental Procedure. The train is operated until the first upper phase has migrated to cell 219. If the train is examined analytically a t this point, the result is analogous to chromatography in which an effluent has not been obtained but the column is cut into segments. It is also analogous to paper chromatography (3). Single Withdrawal (6). The train is operated further than in the first by permitting the upper phases to flow from the train a t cell 219 into the fraction collector. The upper phases flowing out are analogous to the effluent from a chromatogram. Recycling. The third procedure is used when the mixture consists of two or three closely related components and higher separating power is desired. The apparatus contains two parallel rows of 110 cells, one row above the other, so arranged that the upper phases move in opposite directions. Therefore cell 219 is under cell 0 and the upper phases leaving 219 can be caused to flow back into cell 0 instead of into the fraction collector. The effect of this method of operating is to increase greatly the numher of transfers which ran be applied in the fundamental procedure. It can be considered as accomplishing part, of the result of reflux in a fractional distillation column. Various combinations of the three procedures can be used conveniently in a single separation.
I
cp. 8 8
2
t?
The result of a given experiment can be grasped most readily from a distribution pattern. This is simply a graph of the fraction of the original solute found in each cell at the end of the operation, or a figure proportional to it, plotted as ordinate against the cell number. Fractions in withdrawn phases are plotted against the transfer number on which thev emerge. Figure 3 is an experimental result obtained in the fractionation of an artificial mixture of amino acids, 300 mg. of each. MATHEMATICAL EXPRESSION OF RESULTS
With such large numbers of individual cases to be followed, adequate mathematics must be available for expression of the result. It so happens that the mathematics of probability or statistics are nearly identical with the mathematics needed for this work. Both are based on the binomial series. Previous workers ( 4 , 10) have used the binomial series in developing a theory for the operation of continuous columns. Stene ( 1 4 ) began a theoretical study of extraction from the same viewpoint as the author’s and a t about the same time, but did not implement his views practically with the required apparatus.
i!
Figure 4.
No. of t u b e Relation of Curve Spread and Height with Increasing Transfers
A given solute if it behaves ideally-i.e., complete equilibrium a t each stage and constant partition ratios-will migrate through the series of cells according to the binomial expansion
where K is the partition ratio and n is the number of transfers applied (6). A useful approximation of the binomial for numbers of transfers greater than 25 is Equation 2, which is taken directly from the normal curve of error, the equation so familiar in the mathematics of probability. Y is the fraction of the original solute in a
1349
V O L U M E 2 2 , N O . 11, N O V E M B E R 1 9 5 0 given cell and X is the number of cells that particular cell is removed from the cell of maximum concentration. The degree to which actual experiment and the calculation agree can be seen from Figure 3. - X 2 / 2 n K / ( K C 1)2 1 Y= (2) .\/2=nK/(K 1)2e
n
20
+
The mathematics applicable to the elution series, the bands in the right hand pattern of Figure 3, can be derived from Equation 2 and may be expressed by the approximation of Equation 8 . This serics is most important if one is interested in the analogy to chromatography. 1 -XZ/2n/K y=(3) d 2 7 Ke
Q
3
Q
160
200
240
280
320
350
400
40-
The concentration shifts taking place in a stepwise extraction train for a single solute, K = 1, when increasing transfers are app1ic.d are shown in Figure 4. The bands become broader and lo~ver. This means that a basic part of the process is dilution. ~ v e nafter 200 transfers a small fraction, ( o . ~ ) ~ o remains o,' iii the 0 cell. Obviously, so small an amount is of no practical significance and is neglected. But progressively up the series the amount of solute in each cell increases and a point is finally rcached where a significant amount of solute is present. In t h r present discussion a significant amount will be l.Oyoof that present in the maximum cell. Perhaps a better grasp of the effect of higher and higher numbers of transfers is given by viewing t,he process in a somewhat different light. In a 10-transfer distribution, K = 1, the pattern could be curve A of Figure 5 . Here there is little space for t.hrowing off impurit),, all except cells 1 and 10 having significant amounts of the main solute in them. But if 100 transfers were applied, in the same total volume of solvent, arbitrarily taking sufficiently more solute so that the maximum cell would contain the same amount as in A , and the result were plotted on the same pattern hy renumbering the abscissa, curve B would result'. Now there is significant solute in only 30% of the cells and spare remains for a t least. three other components on each side of the band. Similar consideration of a 1000-transfer distribution gives curve C and more space still for other components.
A higher number of transfers obviously gives a narrower band with respect to the
." .* 8
A B
c
0 0 0
2
20 200
4 40 400
6 60 600
8
80 800
10 100 loo0
No.of t u b e Figure 5.
Comparative Band Widths
200
0
NO.of tube Figure 7.
Distribution Patterns for Separation of Artificial Mixture of Fatty Acids Lower 220 transfers Upper 400 transfers
System.
Heptane/equal velumes of methanol, formamide, and glacial acetic acid
total solvent used. This is the real basis for the increased selectivity of higher transfers, as the comparative 100-transfer and 1000-transfer distribution of a three-component mixture given in Figure 6 shows. As more transfers are applied, the bands appear to become more sharply defined and further removed from each other. That this reasoning holds experimentally is proved by actual experiment. Figure 7 shows the effect of doubling the transfers i n s e p a r a t i n g lauric, myristic, palmitic, and 400 600 800 stearic acids. Lo If an infinite number of transfers could actually be applied, each band would be very narrow indeed with respect to the total volume of solvent and would b e c o m e nearly a line on a pattern no wider than the one above. hloreover, each cell would have vanishingly small capacity unless an infinitely large total volume of solvent wep: used. I n order to overcome the small capacity of a single cell, the solute could be scattered
1350
ANALYTICAL CHEMISTRY
at the-start of the process in a bank of cells, as in Figure 8. r h e it was pIac,ed in seven cells initially and carried to 100 transfers. The end result is the summation of seven separate distributions one cell removed from each othcr. The Bum, curve B , is slightly lower and wider than A , which would be the iesult of starting with the same amount of solute all in a single cell. However, if the solute were scattered in a much higher number oi‘ cells, as in Figure 9, the band would not appear to “erode” to such a marked ttstcnt on migrating through the train and would appear nearly to hold its shape, height, arid width. Thus, the ideal behavior proposed for the chromatograph by Wilson ( 1 5 ) has been developed by the ideal stepwise approach.
such errors. The normal curve of error takes care of such deviations in a most remarkable way. When such errors become f r o quent but are of random occurrence, the effect registered is that of making the curve somewhat too broad. Poor packing and nonuniform flow of solvent are known to have a similar and perhaps much greater effect relatively on the operation of a chromatographic columii ( 8 ) ,perhaps due to channeling.
3ATURE OF PARTITION CHROMATOGRAPHY
\Vith this as a background it may be instructive to inquire into the nature of partition chromatography, beginning with experinicwtal observations. We l a o w that irrespective of the way it operates, the method is remarkably efficient for high dilutions of niany solutes, as the pattern.of Moore and Stein (fS), Figure 10, so beautifully demonstrates. We also know that uniform packing, slow flow rates, and relatively linear adsorption isotherms are required for such a result. These conditions parallel the uniform mechanical transfers, state of equilibrium before transfer, and constant partition ratios required for exact mathematical correlation of countercurrent distribution results. I n spite of this apparent correlation, it is well to question whether or not such a correlation provides an adequate understanding of partiion chromatography. Before attempting an answer to this question, it will be profitable to examine closely the three basic requirements of such a vicw-i.e., perfect mechanical transfer, constant partition ratios, and equilibrium before each transfer.
Tube number, Figure 8.
Effect of Scattering Solute Initially
Experience with the stepwise process has shown that perfect mechanical transfer is relatively not so important. An occasional error of 15 to 20% made in failing t o transfer the upper phase of a single cell (or in transferring some of the lower phase) is not even registered on the shape or position of the h a 1 curve when many transfers are applied, nor id a considerable number of
No.of tube Figure 9,
Hypothetical Band Behavior for Ideal Column
The question of constant partition ratios can be treated most intelligently by expressing the over-all equilibria involved as x partition isotherm, as in Figure 11. This applies for chromatography as well as for countercurrent distribution. The ideal solute approaches curve A over the operational concentration range H itiitl C‘ in the case of liquidliquid extraction are deviations from the ideal caused by the fact that the solute is more or less in a different state of association in the two phases, dependiiig on the concentration. When the stationary phase is a solid (Chromatography), B, the familiar adsorption isotherm, is the type of deviation and C is not encountered. Isotherm A permits a symmetrical distribution curve or withdrawal curve, B permits a steep front and trailing rear, while C permits sloivly rising front and steep rear. Actually, all solutes deviate more or less from A in practical concentrations, but liquid-liquid extraction in general permits concentrations 10- to 100-fold as great as those in partition chromatography before serious deviations occur. Naturally, the limitations imposed by the isotherm do not hold unless complete equilibrium is reached a t each stage throughout the process. There are no stages in a continuous process. Therefore, we can speak only of approach to equilibrium in partition chromatography. But the question can be precisely studied by countercurrent distribution. A cell can be agitated until complete equilibrium is reached and then the transfer made, or the transfer can be made a t a desired level of disequilibrium (1 ). In the first case, the distribution fits the normal curve of error. I n the second case, several forms of the curve are obtained. If equilibrium is approached at the same rate from either phase, a symmetrical distribution curve, but too broad for the calculated one, is obtained. The center of the band is not displaced. If approach to equilibrium from either phase is unequal, a curve skewed either to the right or left may be encountered, and the center of the band may be somewhat displaced from the calculated. Strangely enough, too narrow distribution curves have been encountered on a number of occasions, thus giving much better separations than had been anticipated.
1351
V O L U M E 22, N O . 11, N O V E M B E R 1 9 5 0 ..1syninietrical curve from it continuous column is by no means an indication that the process approaches equilibrium. On the contrary, apparent symmetry may be caused by disequilibrium, as Martin (8) has suggested. 111 the stepwise procedure a theoretical equilibrium curve can be calculated and an experimental one wider or narrower than this may then be an indication of disequilibrium. But with the continuous process there is no such yardstick and exactly what is happening is never known. It is always at some stage of disequilibrium and of deviation from its isotherm. Therefore relative rates of approach to equilibrium :ind small deviations from linearity of the isotherms can be important, if not the deciding factors. Relatively small effrcts not tletcctable by measuring the partition ratio can thereby modify the separation greatly in a critical range.
-Isoleucine
E
Solvents 1:Z:1n-butyl alcohol-n-propyl alcohol-0.lNHC1
When there is the probability of high numbers of plate equivah t s , disequilibrium cannot be dismissed by saying that the volumn operates a t near equilibrium. On this point the stepwise approach is instructive. il50-transfer distribution will agree with the theory, even though the partition ratio varies from 1.1 a t the highest operational ratio to 1.0 a t the lowest. But with a 500transfer distribution, the deviation is apparent as a skewed curve. Obviously, for transfers numbering in the thousands the small tlwiations either in partition isotherm or from equilibrium become greater in importance in relation to the importance of the absolute value of the partition ratio, and may be the basic factors producing separations. This reasoning emphasizes the probable differences between the continuous and the discontinuous process,'even though many transfers are involved in the latter.
8
8d d
c
&= rc
8
."c
3
k
1 9-
Wt.of Solute pes ml in Extract Phase Figure 11.
Often with the continuous process it is possible to make calculations of very high numbers of theoretical plates based on the equilibrium values of the partition ratios of the mixture. Too much confidence should not be placed in such a result, because H.E.T.P. values will not be constadt with a given column used on a mixture. A solute emerging later in the run may give a band of width indicating one half or less the H.E.T.P. value of an earlier solute (II). hri attempt to repeat the process on a different mixture with the same beta values might or might not result in only a small fraction of the efficiency. A narrow band permits the calculation of high numbers of plates but, as has been pointed out, many factors influence the width of the band except in the ideal discontinuous process.
Types of Partition Isotherms
Partition chromatography appears to be generally accepted as a liquid-liquid extraction process, apparently because of the correlation of the rates of migration of bands with liquid-liquid partition ratios (IO). At best this evidence is only circumstantial and probably reflects more the uniform shift in nearly any physical property in a homologous series than the particular property responsible for such an observed effect. The point is analogous to the interpretation of a mixed melting point. Failure to depress may suggest identity, but not necessarily so. On the other hand, a depression offers direct evidence that the two substances are different, just as a migration rate ( R , value) not agreeing with the partition ratio shows that some factor other than liquid-liquid distribution is a t least partly responsible for the effect. Moore and Stein (f2) first reported such a disagreement with the aromatic amino acids on starch columns. Since then many cases have been reported. The author has found the results of partition chromatography of very little use in selecting .systems for countercurrent distribution. Further doubt is thrown on the mechanism by the finding that on starch and paper columns water-miscible solvents give almost the same results as those immiscible with water (2, 18). Observations bearing on this question are numcrous and most of them not really decisive. Actually, partition chromatography always requires an adsorbant active in some degree; otheraise the stationary phase would not be hcld. On the other hand, where two immiscible liquids are in contact on a column, complete ruling out of the possibility of liquid-liquid extraction is equally impossible, even if direct evidence in favor of it is lacking. The point is not very important, because there is so much analogy between countercurrent distribution and chromatography. Certainly, the difference between countercurrent dis tribution and so-called partition chromatography is much greater than that between Tswett chromatography and partition chr+ matography. In fact, there seems little to be gained by assuming
1352
ANALYTICAL CHEMISTRY
that liquid-liquid partition plays any significant role In the latter process. LITERATURE CITED
(1) Barry, G.T., Sato, Y . ,and Craig, L. C., J . B i d . Chem., 174,209 (1948). (2) Bentley, H. R.,and Whitehead, J. K., Biochem. J., 46, 341 (1950). (3) Consden, R.,Gordon, A. H., and Martin, A. J. P., Ibid., 38,224 (1944). (4) Cornish, R.E.,et al., I n d . Eng. Chem., 26,397 (1934). (5) Craig, L.C.,J . B i d . Chem., 155,519 (1944). (6) Craig, L.C., and Post, O., ANAL.CHEM.,21,500 (1949).
3rd Annual Summer S g m p e i u m
(7) Jantzen, “Das fractionierte Distillieren und das fractionierte Verteilen,” Dechema Monographie, Val. V, No. 48, Berlin, Verlag Chemie, 1932. (8) Martin, A. J. P., Endeavour, 6,21 (1946). (9) Martin, A.J. P., and Synge. R. L. M., Biochem. J . , 35,91(1941). (10) Ibid., p. 1358. (11) Moore, S.,private communication. (12) Moore, S., and Stein, W. H., Ann. N . Y . Acad. Sei., 49, 265 (1948). (13) Moore, S., and Stein, W. H., J . Bid. Chem., 178,53 (1949). (14) Stene, S.,Arkiv Kemi,Mineral. Geol., 18A,No. 18 (1944). (15) Wilson, J. N., J . Am. Chem. Soc., 62, 1583 (1940). RECEIVED August 4 . 1950.
- Separatbne
Ion Exchange Separations EDWARD R. TOMPKINS,‘RadiationLaboratory, University of California, Berkeley, Calif. This review discusses the properties and laboratory applications of some commercially available ion exchangers, as well as current theories concerning the mechanism of the separation of substances by ion exchange. A brief description of laboratory procedures for equilibrium experiments and for column separations is given. Several theories designed to predict column separations are evaluated. Results
I
ON exchange resins are solid substances which are insoluble in solvents and solutions usuallj employed in the laboratory but which ionize in polar solvents. The chief usefulness of these substances in the laboratory depends upon their property of entering into exchange reactions with ions in a contacting solution. They are applicable for separating ions because of the differences in the exchange constants for various ionic speries and the eme with which they may be separated from the liquid phase. The general reaction involving the exchange of monovalent ions (for a cation or anion exchanger) may be expressed by the equation AR
+ B*
eBR + A*
of a number of column separations of substances with similar properties are reviewed. This method is unique in its ability to produce extremely pure preparations of members of the lanthanide and actinide rare earth series. It has been used successfully in separating and purifying zirconium and hafnium, the alkaline earth elements, scandium, and many other metallic ions as well as organic substances.
hydroxide form react as bases. As map be seen from the titration curves shown in Figure 1, hydrocarbon cation exchangers with nuclear sulfonic acid groups have acid dissociation constants similar to the constant for sulfuric acid. Likewise, the hydroxide form of the quaternary amine anion exchangers behaves as a strong base. Cation exchangers nith other functional grnupse.g., -COOH, -CHnS03H, -OH, etc.-and the hydroxide form of tertiary amine anion exchangers have much lower dissociation constants. I n fact, the salts of these weakly acidic and basic exchangers are appreciably hydrolyzed in watcr (39). EXCHANGE EQUILIBRIA
(1)
in which R is the eschanger and A* and B‘ are ions competing for the available positions on its matris. For a cation exchanger, R is an anion and -4and B are positively charged; for an anion exchanger, H is positively charged and A and B are anions. PROPERTIES OF EXCHANGE RESINS
I’rioI to the original preparation of synthetic exchange resina, reported by -4dams and Holmes in 1935 (I), exchangers were of limited value for chemical separations. However, because of the rapid development of new types of synthetic exchange resins, a number of very useful products are now available. These resins are cross-linked polymer chains with polar fu,nct,ional groups. Their general structure (6, Figure 1)has been described in detail by Bauman (6,6). The sources of several typical ion exhangers are shown in Table I. Of these, the strongly acidic cation exchangers and the strongly basic anion exchangers have found the widest application for laboratory separations. They are availahle in t,he form of spherical particles, and in some cases, in various mesh sixes. Detailed descriptions of the preparation and properties of these exchangers are given in the literaturc references cited in Table I. In their hydrogen form, cation exchangers behave as acids when in contact with ionic solutions; anion exchangers in the
The value of the equilibrium const,rtnt, -
a -
UBR UAR
X aA*. X UB*
in which a represents the activities of the reactants shown in Equation 1, depends upon the hydrated radii of the two competing
1
I I I I 2 4 6 8 2 N NaOH, MG.15 G. RESIN,
IO
Figure 1. Titration Curves of Several Cation Exchanger (88) Functional Groups A . Phenolic-OH B . Methylene sulfonic-CHBOsH C. Carbo~yl-COOH D , E . Nuclear-SOaH
