.
plex in which chlorine atoms have replaced one or more pyrogallol molecules. The result's of these experiments indicat'e that for maximum color development the chloride ion concentration should be kept at' a constant low level and this can be accomplished by evaporating the sample solut,ion in the presence of a measured amount of sodium (25 mg.) to drive off free hydrochloric acid and by accurately dispensing the chloride-containing buffer solution (1 ml.). Some Properties of OsmiumPyrogallol Complexes. EXISTENCE OF L O W E RCOMPLEXES.Studies of the osmium-pyrogallol system have shown t h a t at' a p H of approximately 1 a blue complex is formed that' exhibits maximum absorbance near 665 mp and has a low solubility in amy1 alcohol; a brown complex, also wit,h a low solubility in amyl alcohol, and with an absorbance maximum at' about 475 mp, is formed by osmium in the p H range 4.5 to a t least' 6.
Absorption spectra of solutions containing the above-mentioned complexes are shown in Figure 1 (curves 3 and 4). Because of the lower maxima of these curves it is assumed that the pyrogallol to reagent ratio of the corresponding complexes is lower than that of the complex formed at a p H of approximately 3 in the recommended procedure for color development. EXTRACTABILITY OF OSMIUM-PYROGALLOL COMPLEXESIN T.B.P. As judged visually, one 10-ml. portion of T.B.P. will extract nearly completely the osmium-pyrogallol complexes formed in the p H range 1 to 5.5, ANIONICNATUREOF OSMIUM-PYROGALLOL COMPLEXES.T h a t the osmium-pyrogallol complexes are anionic is indicated by their strong retention by the anion exchange resin Amberlite IRA-400. COMBINING RATIOS. Results of experiments using the method of continuous variations ( 2 ) and the mole ratio method ( S ) , suggest the possibility that
the analytically useful complex (585 mp) formed a t a pH of approximately 3 is a 3 to 1 complex of osmium to pyrogallol, and that the complex formed at approximately p H 1 is a 2 to 1 complex. EXAMINATION OF PHLOROGLUCINOL AND RESORCINOL AS REAGENTS.Both phloroglucinol (1,3,5-benzenetriol) and resorcinol (1,3-benzenediol) were ineffective as reagents for osmium under conditions similar to those of the recommended procedure for color development; this is probably because the functional groups are not located on adjacent carbon atoms in these compound. LITERATURE CITED
(1) Jasim, F., Magee, R. J., Wilson, C. L., Mikrochim. Acta 1-2. 11 11962). ( 2 ) Job, P., Ann. Chi&. (I&~me)'(lO)9, 113 (1928). (3) Yoe, J. H., Jones, A. L., IND. ENG. CHEM.,ANAL. ED. 16, 111 (1944). RECEIVEDfor review July 27, 1964. Accepted November 23, 1964.
Correlation and Prediction of Solvent on Paper Chromatographic R, Values KENNETH A. CONNORS School o f Pharmacy, University of Wisconsin, Madison, Wis. An equation is given that relates the R, values for a compound in two paper chromatographic solvent systems. This equation appears to describe with reasonable accuracy many experimental R, values for several classes of compounds and types of solvent systems. By combining the theoretical equation with an experimentally determined reference quantity, fairly accurate predictions can b e made of R, values for a series of compounds in one solvent if the corresponding R, values in a second solvent are known.
T
HE goal of the chromatographer is thp capability of predicting accurately the extent of zone migration in a defined chromatographic system. Many factors operate to control the migration of the zone; the most important of these are the structure of the sample compound, the nature of the solvent system, and the nature of the support. Much attention has been directed to the relationship between molecular structure of solutes and their chromatographic behavior-see, for example, the reviews of Cassidy ( 2 ) and Giddings and Keller ( 4 ) , and the exten-
sive investigations of Green and Marcinkiewicz (b)-and some quantitative correlations have been discovered. These results are of greatest value in structure determination and identification problems. The practicing analyst usually is presented with a mixture of closely related substances to be separated. The most effective way to influence the chromatographic bphavior of such a mixture (since the molecular structures usually cannot readily be altered) is by suitable selection of the solvent system. Many qualitative guide rules for the selection of chromatographic solvents are used; an example is the useful observation that a polar developer will cause faster zone migration of a polar solute than will a nonpolar developer. Recently Soczewinski and Wachtmeister (8,9) have presented some quantitative relationships between paper chromatographic R values and fractional composition for certain types of solvent mixtures. The effect of the support is, of course, critical in adsorption chromatographic procedures. I n this paper only partition chromatographic systems are considered, so the solid support is (ideally) not of great importance, since it serves
mainly to support the statioiiary phase. Actually the support probably also acts in a minor way as an adsorbent for the solute, thus complicating the problem. I n the case of paper chromatography the support is believed to influence the nature of the internal phase markedly. I n seeking correlations between zone migration and solvent systems fewer complications may be expected with column partition chromatographic results than with paper chromatographic data. However, few R values are recorded for column partition studies, while a great many Rf values are available from paper chromatographic separations. I n this paper a semi-empirical correlation of paper chromatographic Rfvalues in different solvent systems is presented. THEORY
The R f value of a compound is related to its partition coefficient, K , and the relative phase volume fraction, U , by Equation 1,
Rj
=
KC KC+1 ~
which expresses the idea that only that fraction of solute that is in the mobile VOL. 37, NO. 2, FEBRUARY 1965
*. 261
+F
-RjX(dF/dRfX) (1 - RfX) [RfX F ( l - RjX)]*
+
R: Figure 1 . Plots of Equation 4 for several fixed values of F
phase can migrate. The partition coefficient is defined as the ratio of the concentration in the mobile phase to the concentration in the stationary phase, and the volume fraction is similarly the ratio of mobile to stationary phase volumes in any element of the chromatographic system. Equation 1 is not entirely accurate for several reasons: A true distribution equilibrium may not be operative; the partition isotherm may not be linear; the composition of the phases may be different at different points in the chromatogram; and the velocity of flow and the effective phase volume of the mobile phase may not be the same at all points. For convenience the Rf value is written with a superscript indicating the solvent system to which the R, refers. Thus the R, values in the different solvent systems X and Y may be written
which shows that as R f X approaches 1.0, the slope approaches F , while as R f X approaches 0.0, the slope approaches 1/F. If a straight line is drawn tangent to the curve a t point (1,1),the sum of the slope and intercept of this extrapolated line will be unity. There is no reason a priori to expect that F will be a constant for a series of compounds, though such constancy seems possible. Figure 1 shows several plots according to Equation 4 with various assigned constant values of F. Even with constant F within a series, a considerable variety of correlations is possible. Because F may vary, more complex curves may be expected in practice. I n order for a smooth curve to be observed, it is necessary (except for fortuitous coincidences) that the quantity F be a continuous function of RfX to which the individual members of the set assign particular values. (Though really just a convenient fiction, this point of view is always adopted in rationalizing rate and equilibrium correlations.) Thus, for example, one may expect to correlate Rf values between two solvent systems for a homologous series of acids only if the degree of ionization of each acid is the same within each solvent (the degree of ionization may be different in the two solvents), or if the degree of ionization within the set of acids changes smoothly. F can be calculated for any compound if RfX and R f Y are known; Equation 4 may be rearranged to
F =
RfY =
GY
(3)
where GX = KX CX, etc. Equations 2 and 3 may be combined to give Equation 4, (4)
where
Equation 4 relates the R, value for a compound in one solvent system to the R, value in another solvent. I n a practical sense it is desirable to be able to apply this equation to a series of compounds, but it seems reasonable to expect that only a set of closely related compounds could be treated so simply. From Equation 4 some of the features of a plot of RfYvs. R,Xmay be determined. The curve will pass through points (0,O) and (1)l). The slope of the curve is given by Equation 6, 262
ANALYTICAL CHEMISTRY
(6)
RjX(1 - RfY) R f Y ( l - RyX)
(7)
COMPARISON OF THEORY W I T H EXPERIMENTAL CORRELATIONS
To the extent that Equations 2 and 3 are reasonably accurate representations of RfX and RIY, Equation 4 must be an equally reliable statement of the connection between R f X and R f Y ; in fact, Equation 4 may be more accurate than Equations 2 and 3 because of the possibility of error cancellation. If parameter F is constant within a series of compounds, it should be possible to fit quantitatively the set of experimental R, values with a curve calculated from Equation 4. If such a simple correspondence cannot be made, the equation may still be valid, but F may itself be a function of RfX, and this dependence must then be taken into account. Inapplicability of Equations 2 or 3 to the systems would also result in a poor fit, and this situation might be difficult to distinguish from the case of a variable F.
Figure 2. Plot of some barbiturates
Rf
data (7) for
Smooth line calcd. with Equation A, taking F = 3.35. See Table I for solvent compositions
R, values for nine barbituric acid derivatives in two solvent systems ( 7 ) gave F values ranging from 2.53 to 4.20, with a mean value of 3.35 + 0.40. The data are plotted in Figure 2 ; the smooth curve was drawn with Equation 4, using F = 3.35. (The solvent systems are described in Table I.) The agreement between the points and the calculated line is very satisfactory, even though F is not truly a constant. Part of the variation in F is due to experimental error in the R, values (assuming that all nine compounds truly belong to the set). I n Figure 3, R, data for some inorganic phosphates are shown (6). The F values for this series of compounds vary smoothly from 0.34 to 0.82. The smooth lines in the figure were calculated with constant F values of (from top to bottom) 0.4, 0.6, and 0.8. Evidently no entirely satisfactory compromise is obtainable with a constant F value, though the absolute error in RID incurred by the use of F = 0.6 is nowhere larger than 0.06. It would be possible to fit the curve accurately by using a variable F . Forty-five uracil derivatives (including some uridine, thymine, and thymidine derivatives) are represented in the correlation shown in Figure 4 (3). The line in the figure was calculated with a constant value of 0.75 for F ; this was the mean of five values selected a t equal distances along the set of points. (Practically the same line would have been obtained if the F for uracil alone had been taken for the calculation.) F varies by a factor of about 2 a t the extremes of the set of experimental points. I n terms of absolute deviations from the calculated line, this correlation is fairly satisfactory over most of its course, though it is clear that use of a variable F would yield a superior correlation. The results shown in Figures 2, 3 , and 4 are characteristic of those systems that yield fairly smooth plots of R, values. It must be admitted that many data
If F is very large or very small for a combination of solvent systems, the correlation may appear to fail to pass through points (0,O) and (1,l)-Le., for such systems a substance could, for example, show R, = 0 in one solvent and have a high R , in the second solvent. Though accurate predictions would not be successful in this case, there seems to be no compelling reason to doubt that the curve must pass through (0,O) and
RP
(111). PREDICTION O F RI VALUES
R,C Figure 3. Rf data for some condensed phosphates in two solvent showing effect of variable systems (6),
F Smooth curves calcd. with Equation 4. From top to bottom, F = 0.4, 0.6, 0.8. See Table I for solvent compositions
cannot be plotted in this way. There are several causes of scatter in these plots. Probably the most serious of these has already been mentioned: a (discontinuously) variable degree of ionization from member to member of the series. If the two solvent systems are very similar, this may not lead to serious scatter. If the compounds contain ionizable groups, the best chance to obtain a smooth correlation is by suppressing the ionization completely or by permitting it to occur to its fullest extent. Sometimes one of these situations may obtain in one solvent and the other in the second solvent. This seems the most likely explanation for the reasonable correlation ( F = 7.2 + 1.7) observed between R f values for nine aliphatic monocarboxylic acids (3) in an acidic solvent (A) and a basic solvent (H) (see Table I). Part of the difficulty in locating useful correlations is in determining just what compounds constitute the set. Many sets are very difficult to identify, and failure to locate good correlations may reflect an inadequate awareness of all of the factors operating in the systems, rather than an inherent absence of correlation. The same uracil derivatives whose Rf* values are plotted against RfB in Figure 4 gave a scatter diagram when RfA was plotted against the R f H values (see Table I). However, a fair correlation was noted between the R f values of the ten dihydrouracil derivatives in this group of compounds. A solvent system obviously must be capable of effecting a reasonable separation of the members of a set if meaningful correlations are sought. A bunching of R values, especially in the ranges 0.0 to 0.1 or 0.9 to 1.0, is common for many combinations of solutes and solvents. Such data are of little use in establishing relationships of the kind developed here.
Equation 4 provides a theoretical basis for estimating R f values in one solvent system from the corresponding values in another solvent. Although it may sometimes be possible to predict whether F will be greater or less than unity for a specified system (of course the F value can be converted into its reciprocal merely by exchanging the axes, but this is a trivial point), a quantitative prediction of F from fundamental data (such as equilibrium solubilities, pK, values, heats of solution, etc.) is not yet possible. To make use of Equation 4 in estimating R f values, it is therefore necessary to utilize chromatographic data to estimate F. The RIY values for a set of compounds can be calculated if the corresponding R,X values are known and if the R f Y value for one typical, well-behaved, member of the set is known. The R f values for this reference compound, denoted R*X and R*Y, are used to obtain an estimate of F (symbolized F,) for the entire set:
If F is a constant for the set, F* should permit accurate calculation of all RIY values for the set. If F is a variable quantity, the reference compound should be selected in the middle of the R f range of interest. It may be advisable in some instances to employ more than one reference compound in order to reduce error due to a variable F . If more than one pair of R f values are available, they should be utilized to obtain as reliable an estimate of F as possible. A typical calculation is shown in Table I1 for several barbiturates ( 7 ) . The agreement is within experimental error for most of the results. Another example is given in Table 111; again the agreement is good, indicating that F is substantially constant over the range investigated. DISCUSSION
The theoretical relationship between R f values in two solvent systems is simply expressed in Equations 4 and 5. I t appears that for many chromatographic systems this relationship pro-
Figure 4. Rr data for uracil derivatives in two solvent systems (3) Smooth line calcd. with Equation 4 and F = 0.7.C. See Table I for solvent compositions
Table 1. Solvent Systems Referred to in Text and Figure Captions
Code Refletter Solvent system erence A tert-Butyl alcohol, methyl (3) ethyl ketone, formic acid, water (40:30: 15: 15) B 1-Butanol, glacial acetic (3) acid, water (50:25:25) C 75 ml. 2-propanol, 25 ml. (6) water, 5 grams trichloroacetic acid, 0.3 ml. 20% ammonia D 70 ml. 2-propanol, 10 ml. (6) water, 20 ml. 20% trichloroacetic acid, 0.3 ml. 25Yc ammonia E Chloroform-50yc aqueous ( 7 ) formamide F Chloroform, benzene (1 : 1)/ (7) 50y0 aq. formamide G Benzene-50%. - aa.- form- (. 7.) amide H tert-butyl alcohol, methyl (3) ethyl ketone, water, ammonium hydroxide (40:30:20: 10) I Ethanol, water, 15N am- ( 1 ) monium hvdroxide (40: 5: 1) J 1-Propanol, water, 15N am- ( 1 ) monium hydroxide (40 : 5: 1)
Table 11. Calculation of Barbiturate R, Values in Solvent System G from Values in Solvent E 0 . b
RfQ Compound RS Calcd. Exptl. Pentobarbital 0.57 0.15 0.10 Amobarbital 0.57 0.15 0.10 Baytinal 0.72 0.25 0.25 Kemithal 0.76 0.29 0.27 Hexobarbital 0.87 0.47 0.48 Thiogenal 0.87 0.47 0.45 Prominal 0.89 0.52 0.52 Eunarcon 0.92 0.60 0.68 Reference compound was thiopental: R , E = 0.81, R,Q = 0.36, F , = 7.6. a Data from ( 7 ) . b See Table I for solvent compositions.
VOL. 37, NO. 2, FEBRUARY 1965
263
Table 111.
Calculation of RI Values for Some Acids.,&
RI’ Calcd. Exptl. Acetic 0.22 0.37 0.38 Propionic 0 30 0 48 0 45 Acrylic 0 28 0 45 0 43 Dihydracrylic 0 06 0 12 0 12 Phloretic 0 31 0 49 0 42 3-Iodophloretic 0.31 0 , 4 9 0.42 3,5-Diiodophloretic 0.10 0 , 1 9 0.21 Reference compound was hydracrylic = 0.16, R,‘ = 0.29, F , = 0.47. acid: a Data from ( 1 ) . b See Table I for solvent compositions. Acid
Rf
vides an adequate quantitative description of the correlation. The quantity F may be essentially constant or may vary within a set of compounds; in order for a number of compounds to be considered members of the same set it is necessary that F be a continuous function of R, for the set. It seems that sound corrections could be applied to bring some deviant values into the setfor example, a series of acids might constitute a smoothly correlated set if their pK’s and the pH of the solvents were available to permit adjustment of the experimental R I values in accordance
with their different degrees of ionization. By combining the theoretical relationship with an empirical estimate of F , it is possible to predict R, values in one solvent system from the corresponding values in another solvent. A very great number of RI values can thus be represented by Equation 4 and appropriate values for F,, the estimate of F . Consider, for example, the case of the barbituric acid derivatives in Table 11. If the R, values for the reference compound were determined over the entire range of solvent composition from pure chloroform to pure benzene [reference R,’s could be interpolated graphically, or perhaps by means of the equation of Soczewinski and Wachtmeister ( 9 ) ] ,then with Equations 4 and 8 the R, value of any other barbiturate in any chloroform-benzene solvent mixture could be predicted if its R, is known in a single solvent. Table I1 shows the calculation of R f in pure benzene (mobile phase) from the R, in pure chloroform. Any intermediate as in Figure 2composition-e.g., could be treated in the same way if only the R, for the reference compound were available. The method is not limited to relating Rr’s between similar solvents, though it is most accurate in such cases, but can correlate data and lead to pre-
dictions of R, for solvent systems that are chemically very dissimilar. The capability of predicting R/ values from a minimal amount of prior information should permit the rational choice of the optimum solvent system for a given separation with a very small expenditure of experimental effort. LITERATURE CITED
(1) Cahnmann, H. J., Jlatsura, T., J . A m . Chem. SOC.82, 2050 (1960). (2) Cassidy, H. G., “Fundamentals of Chromatography,” Vol. X, in “Technique of Organic Chemistry,” Chap. XIII, A. Weissberger, ed., Interscience, New York, 1957. (3) Fink, K., Cline, R. E., Fink, R. M., ANAL.CHEY.35, 389 (1963). (4) Giddings, J. C., Keller, R. A., Chap. 6 in “Chromatography,” E. Heftmann, ed., Reinhold, New York, 1961. (5) Green, J., Marcinkiewicz, S., in “Chromatographic Reviews,” z’ol. 5, M. Lederer, ed., Elsevier, Amsterdam, 1963. (6) Hettler, H., in “Chromatographic Reviews,]’ Vol. 1, R l . Lederer, ed., Elsevier, Amsterdam, 1959. ( 7 ) Macek, K., Arch. Pharm. 293, 545 (1960). (8) Soczewinski, E., J . Chromatog. 8 , 119 (1962). (9) Soczewinski, E., Wachtmeister, C. A., Zbzd., 7, 311 (1962). RECEIVED for review September 17, 1964. Arrepted December 16, 1964.
Reactions of Heteropoly Acids with Organic Compounds Detection of Aldehydes and Cyclic Ketones JOHN H. BILLMAN, DONALD B. BORDERS,’ JOHN A. BUEHLER,* and ALFRED W. SEILINGS Department of Chemistry, Indiana University, Bloomington, Ind.
b 12-Molybdosilicic acid and its sodium salt are highly specific and sensitive color reagents for the detection of aldehydes and cyclic ketones containing at least one a-hydrogen atom. Approximately 150 carbonyl-containing compounds have been tested along with a large assortment of compounds with other functional groups. A positive test is the appearance of a blue color in basic solution. Noncyclic ketones and carbohydrates do not give a positive test. In general, compounds containing simple readily oxidizable functional groups, such as the aromatic amino or phenolic group failed to give a positive test. Ascorbic acid and a few poly aromatic amines and polyphenols react to produce the blue color. An explanation is given for the behavior of the exceptional results. Other heteropoly acids and their salts were found to b e less specific color reagents.
264
ANALYTICAL CHEMISTRY
H
ETEROPOLY ACIDS have been widely
used as analytical reagents and their analytical chemistry has been the subject of several reviews (10, 11, 18-20). Many of the analytical applications are based on the colorimetric estimation of the heteropoly acid in its deep blue reduced form. Elements capable of acting as the heteroatoms of a heteropoly acid may be determined quantitatively by reaction with molybdate, tungstate, or vanadate ions, reduction of this complex, and measurement of the resulting blue color. A variety of reducing agents may be determined in an analogous manner. I n the latter case the reducing agent is allowed to react with a solution of a heteropoly acid to produce the heteropoly blue whose absorbance may be measured. Heteropoly acids have also been applied as color reagents for the qualitative detection of a variety of organic reducing agents. Among the types of organic compounds which have been reported to reduce heteropoly acids are
aldehydes, phenols, and amines (5) , sugars ( l a ) ,polyolefins (S), and certain steroidal ketones and alcohols (S,8). From this list, it can be seen that the application of heteropoly acids as qualitative color reagents has been limited because of their lack of specificity. h few years ago, preliminary data on the color reactions of organic compounds with 12-molybdosilicic acid in basic solution were reported (1). Under the conditions investigated, 12-molybdosilicic acid appeared to be a selective color reagent for the detection of aldehydes containing at least one a-hydrogen atom. These reactions have been reinvestigated and the study has been extended to include other heteropoly acids. In addition a more extensive series of compounds was tested and the procedure was modified. 1 Present address, Lederle Laboratories, Pearl River, N. Y. 2 Present address, LaMoyne College, Memphis, Tenn. 3 Present address, Morton Chemical Co., Woodstock. Ill.