ANALYTICAL CHEMISTRY
732 Fteric effects, etc. In a few instances, where calculated values differ greatly from the experimental values, salt formation is an obvious explanation as illustrated by the failure of phenol to move on calcium hydroxide. .4n example of the calculation of the R value for ethylaniline on Florisil ( a synthetic magnesium silicate, Floriden Co., 200/300 mesh) is given below.
I ~ s H ? = 1. Fraction shonine electron pair on nitrogen in resonance structures = 1/4; for ethylaniline use D, = 0.25 H , for one hydrogen = 1 -vat= C ~ H C~ ~ =H 77~ 29 = io6 l ) d for petroleum ether is hy definition 1 H Afor uetroleum ether is bv definition 1 -4;for blorisil was detcrmiGed as 3000 DH for Florid was determined as 260
+
+
2000 X 0.25
->
+ 260 X
1
relatively unimportant in this r o r k since Austin and Shipton ( I ) have shown R to be independent of flow. The R values were determined with 0.01 M solutions of the adsorptives, using an initial volume sufficient to form a layer of Polution 1 cm. thick above the adsorbent in the chromatographic tube. In this region the rate of moyement is nearly independent of concentration and initial volume ( 3 ) . The solvents were thiophene-free reagent grade benzene and specially purified Skellysolve B. Substances used as adsorbed compounds were of the highest purity grade and were repurified where necessary. The adsorbents are listed in Table I. They were used as obtained from the manufacturer. 4CKYOW LEDGMEYT
This work was done under a contract with the Office of Naval Research. The authors wish to express their appreciation for this assistance. =
7.15
EXPERIMENTAL
Extensive data concerning the flow rate, etc., of the adsorbents used here will be published elsewhere in connection with a survey of adsorbents. In general, the flow rates of the columns (75 X 9 nim.) were between 3 and 25 nim. per minute; however, this is
LITER4TURE CITED
(1)Austin, C. R , and Shipton, J., J . Council Sca. Ind. Research, 17, 115 (1944). (2) Consden, R.,Gordon, A. H., and Martin, A. J. P., Biochem. J . , 38.224 . (1944). ~ ~ ~ . (3) LeRosen, A.L.,’J. Am. Chem. SOC.,67,1683(1945). (4)Ibid., 69,87(1947). (5) Weil-Malherbe, H., J . Chem. SOC.,1943,303. RECEIVED.4ugust 9, 1950. Presented before the Divisions of Analytical, Industrial and Engineering, Petroleum, and Physical and Inorganic Chemistry, .4dsorption Symposium, a t the 116th Meeting of the AMERICAN CHEMICAL SOCIETY, Atlantic City, K.J.
Effect of Side Chain on the Chromatographic Adsorption of Some Ketones EDGAR D. SMITH1 AND ARTHUR L. LEROSEN Louisiana State University, Baton Rouge, La.
A
COMPLETE series of straight-chain methyl ketones from Cs to was used to determine the effect of the side chain on chromatographic behavior. Certain periodic increases in adcorption affinity were observed with increasing chain length, alihough the general trend n-as toward decreasing adsorption affinity. Earlier work in these laboratories has indicated that adsorption on the usual adsorbents, excluding charcoal ( I ) , is generally due to chemical-type interactions between the adsorbed compound and the adsorbent. These interactions appear to be explainable on the basis of the relative electron-donating and accepting tendencies of the adsorbentandof the functional group of the molecule adsorbed, and they have been represented by an equation of the type (6):
niolecular weight of the members of a given homologous series ( 4 ) . There was, however, little data on which to base this assumption and so it was decided to prepare and determine R values of several homologous series in order to provide data for determining the proper mathematical relationship between the R value and side chain molecular weight. Data will be presented for a series of aliphatic straight-chain methyl ketones from C3 through C19, and certain conclusions from these data will be pointed out. While, in general, the data support the type of equation postulated above, it was found that the molecular weight relationship is somewhat more complex than was assumed a t first. These facts can probably best be brought out by a brief discussion of each of the accompanying graphs and tables of the experimental data. DISCUSSION OF GRAPHS AND TABLES
j’ = (1
- R),R =
( 1 / M a e ) ( k ) = k(Mac)-’
where f is the adsorption affinity; R, the ratio between the distance moved down the adsorbent column by the adsorbate zone and solvent, respectively; M,,, the molecular weight of the side chain attached to a functional group such as hydroxyl, amino, or hetone; and k , the summation of the interaction tendencies between adsorbent and adsorbed compound. .4s indicated by this equation it was a t first assumed that there was a simple reciprocal relationship between the R value and the 1
Present address, Buckeye Cotton Oil Co., Memphis, Tenn.
Figure 1 is a plot of RL (leading edge) values rn ordinates versus the total number of carbon atoms in the various ketones as abscissa when calcium carbonate (Merck heavy powder) was used as the adsorbent and petroleum ether as the developing solvent. Only weak and apparently general adsorption results so that no conclusions can be drawn other than that this adsorbent and, p r e sumably, others of its type such as lime, magnesium oxide, or alumina are of little value in effecting separations of the ketones tested even when a very weak developer is used. When a stronger developer such as benzene was used the R values were, of course, 1.00, indicating no adsorption.
V O L U M E 2 3 , NO. 5, M A Y 1 9 5 1
733
In connection with work in these laboratories on the specificity of chromatographic adsorbents it became desirable to determine the effect of the organic side chain on the behavior of homologous series of compounds. The first series of compounds selected for this study were the methyl ketones with straightchain side chains. The complete series, from three i o twenty carbon atoms, was studied and a few other miscellaneous ketones were included to indicate other trends. A n unexpected result of this study was the observation of periodic increases in adsorption affinity with increasing length of side chain. No clear interpretation of the observed phenomena seems obvious at present, but it is suggested that the periodic increases in adsorption affinity may be due either to fit-patterns or to molecular configuration of the adsorbed molecule.
When using silicic acid (Llerck heavy powder reagent) as the adsorbent and benzene as developing solvent a steady rise in R value was found with increasing side-chain length in the homologous series, resulting in an almost smooth curve (Figure 2, upper). Theie are slight irregularities in this curve, however, a t COand Cln,which are believed significant as will be discussed later. The Eo-called miscellaneous ketones were included in these plots to phon- that structural relationships must also be considered in predicting RL values. While space will not permit a discussion of these ketones in detail, it. will be seen that branched chain compoun& are adsorbed less and cyclic aliphatic ketones more than the corresponding straight-chain methyl ketones. Both of these facts appear to point out the importance of steric effects in chromatographic adsorption.
The miscellaneous ketones fall about the curve essentially as wa9 noted for silicic acid. Figure 3 shows a plot of the trailing edge R values ( R T )(uncorrected for the distance moved by the solvent during the time that the initial volume of solution was added to the column) for all three adsorbents studied. This graph is included here, although, in general, RT values appear to be more sensitive to the "dips" in the R value curves than are the leading edge values and, moreover, appear to point to a periodicity of these dips-for example, silicic acid and Florisil both show depressions in RT values a t Cs, CI1, CI4, and Cl,. In addition, work with an a p parently impure sample of Czo (melting point, 46" to 49" C.; literature, 58' to 59" C.) indicated that another large dip probably occurs a t this point but these data were not included here because of uncertainty as to the purity of this material. It appears, however, that there are rather large depressions in R value at Ca, Cir, and C20, with smaller ones a t C11 and 17.
g b 0 5 1
0.
!
f
r
'I 0
I
;1
I
1 4
,
5
6
7
B
9
10
1 /I
I2
13
14
I5
16
I7
16
19
CARBON ATOMS
R L OS VALUES
Figure 2. RL 7-alues of Ketones Form 0.01 M Benzene Solutions on Silicic Acid ( u p p e r ) and on Florisil (lower)
t
0. straight-chain methyl ketones;
$, miscellaneous ketones: 1, methyl isopropyl; 2, diethyl; 3, cyclohexanone; 4, methyl isobutyl; 5, methyl phenyl: 6, ethyl phenyl; 7, dibenzyl
I
4
5
6
7
8
9
10
I1
12
I3
14
15
16
17
16
19
CARBON ATOMS
Figure 1. R L Values of Straight Chain Methyl Ketones from 0.01 M Petroleum Ether Solutions on Calcium Carbonate
figure 2, lower, shows that on Florisil (a synthetic magnesium silicate, Floridin Co., 200/300 mesh) there is again a steady, though slower, rise in R value as the side-chain molecular weight is increased. The deviations from a smooth curve which were noticed in the preceding figure are now seen to be much greater a t C14 and, as will be shown in the succeeding figures, cannot be ignored as being within the limit of experimental error. The RL value depressions found with Florisil are in about the same places as with silicic acid. On florisil the RL values tend to a p proach a maximum value of about 0.45 so that for the higher members of the ketone series Florisil is a stronger adsorbent than Filicic acid, while for the lower members the reverse is true.
Figure 4 brings out the effect of drying ( 3 , 6) the adsorbents (130" C. for 72 hours) before using for RL value determinations. The dips are accentuated by this treatment and dried silicic acid is a stronger adsorbent than dried Florisil throughout the homologous series. It seems apparent from these curves that water adsorbed on these materials greatly decreases their general strength as well as their ability to separate the homologous ketones. In work of this sort the question naturally arises as to whether the small deviations from a smooth curve obeerved are significant, or whether they are simply due to experimental error. Therefore, statistical data were taken, Table I, which showed that the probable error of a single RL value determination was about 3%. and that of an R L determination about 6%. Since a t least two values were taken for all points shown on the preceding curves and from 6 to 10 determinations made about the points where dips were noted, it will be seen that the data must be considered reproducible t o at least 0.01 R value. Even so, however, only the large depressions in R L value noted a t CU may be regarded as "statistically reliable" ( d ) , but since these repeated depressions have also been observed in other homologous series now under in-
734
ANALYTICAL CHEMISTRY
Table I.
Statistical Treatment Showing Reliability of R Value Determinations
(0.01 M benzene solution of methyl dodecyl ketone) R Values Run Zone Column KO. Limits’ Length, h l m . RL RT 1 19-30 80 0.38 0.24 2 20-31 79 0.39 0.2: 3 24-36 83 0.41 0.28 4 24-33 0 . 3 9 0.29 24 a 19-33 (9 0.42 0.24 6 18-32 80 0.40 0.23 23-34 82 0.42 0.28 24-35 86 n. 41 0.28 9 26-37 96 0.40 0.28 in 22-31 84 0.38 0.27 Average 0.40 0.26 Probableerror (B), r 0 . 0 1 1 0.016 2.8 6.2 Probable % error
Further work is necessary to establish the significance of the R value depressions revealed in this work. Preliminary work in these laboratories has indicated that the location of these dips is dependent only on the nature of the adsorbed molecule, although the magnitude of the dips is affected by the adsorbent. The authors are presently continuing work along these lines. EXPERIME3TAL
-b
All of the chemicals used as adsorbed compounds in this work were fairly well known compounds of which almost half were commercial samples. These were purified by standard techniques where necessary. Table I1 summarizes the sources of all of these materials. As indicated in this tabulation it was found necessary t o synthesize most of the straight-chain methyl ketones used and
Measured in millimeters from top of column.
vestigation, the authors have felt compelled to include these smaller R value depressions in their discussion. As about half of the ketones studied in this work were synthetic samples specially prepared for this study, and the rest were conimercial samples which were purified where necessary, it seemed unlikely that concentration effects could be the cause of the RL value depressions noted. The data plotted in Figure 5 were taken, nevertheless, to be certain that small errors in making up the 0.01 M solutions used could not cause variations in RL value of the order observed. The flatness of these curves would seem to eliminate this possibility. These curves remain fairlv flat even at concentrations of 0.03 and 0.04 .%?as in future work it may be found desirable, or even necessary, to work in this range of concentrations.
Q
FLOWS L
I
4
5
6
7
8
3
IC
I
CARBOY
2
I3
14
I5
16
17
18 19
ATOMS
Figure 4. RL Values of Ketones from 0.01 N Benzene Solutions on Dried Adsorbents 0 , s t r a i g h t - c h a i n m e t h y l ketones:
$, miscellaneous ketones silicic acid: 1, methFl p h e n y l ; 2, e t h y l phenyl
Table 11. 0.1
on
Source of Pure Ketones Used in This Work
SO. I
,
I
I
I
I
I
I
I
I
I
I
j
4
5
6
7
8
9
10
/I
12
13
14
15
16
l
17
I
1
18
19
CARBON ATOMS
Figure 3. RT Values for Methyl Ketones from 0.01 M Benzene Solutions on Florisil and Silicic Acid,
Carbon Atoms 3 4 5 5
5
and from 0.01 ,MPetroleum Ether Solutions on Calcium Carbonate
6
The main object of this study was to provide a mathematical formulation of the effect of side chain on chromatographic adsorption. The data in Figure 6 show the approximate linearity of a log-log plot of “adsorption affinity” [(l - R ) / R ] versus the molecular weight of the side chain of the straight-chain methyl ketones. The general equation of these straight lines are of the is type predicted but it was found that the exponent of the M8c not - 1 as nas assumed in this first approximation. Instead, the value of this exponent depends on the adsorbent used and perhaps also upon the particular homologous series under study and other factors. The k values determined by this type of plot represent the summation of all of the interaction tendencies causing adsorption and should prove useful in assigning values of donor and acceptor strengths to adsorbents and adsorbed compounds in future work. The fact that there is a constant value for the homologous series seems to indicate the general correctness of the assumption that the functional group of a molecule is of primary importance in causing its adsorption.
8 9
6 6 7 8
Ketone
Source of Pure Chemicals
Methyl methyl Methyl ethyl Methyl propyl Methyl isopropyl Diethyl Methyl butyl Methyl isobutyl Cyclohexanone Methyl pentyl Methyl hexyl Methyl phenyl Methyl heptyl
9 10
Ethyl phenyl Methyl octyl
11
Methyl nonyl
.
12
Methyl decyl
13
Methyl undecyl
14
Methyl dodecyl
15
Methyl tridecyl
15
Dibenzyl
16
Methyl tetradecyl
17
Methyl pentadecyl
18
Methyl hexadecyl
19
Methyl heptadecyl
20
Methyl octadecyl
Eastman Kodak White Label Eastman Kodak White Label Eastman Kodak practical, redistilled Eastman Kodak White Label Eastman Kodak White Label calcium Synthetic from calcium valerate acetate Eaatman Kodak White Label Eastman Kodak practical, redistilled Eastman Kodak White Label Eastman Kodak practical, redistilled Eastman Kodak White Label Synthetic from heptyl magnesium bromide acetyl chloride Eastman Kodak White Label Synthetic from octyl cadmium bromide acetyl chloride Synthetic from calcium caproate calcium acetate Synthetic from decyl cadmium bromide acetyl chloride Synthetic from calcium laurate calcium acetate Synthetic from dodecyl cadmium bromide acetyl chloride Synthetic from calcium myristate oalcium acetate Synthetic from phenylacetic acid acetic anhydride Synthetic from tetradecyl cadmium bromide acetyl chloride Synthetic from methyl magnesium iodide palmityl chloride Synthetic from hexadecyl cadmium iodide acetyl chloride Synthetic from calcium stearate calcium acetate Synthetic from octadecyl cadmium bromide acetyl chloride
+
+
+
+
+
+
+ +
+
+
+
+
+ +
V O L U M E 23, NO. 5, M A Y 1 9 5 1
735
of about i 5 mm. with the dry adsorbent under full wat'er pump vacuum. -4s the last of this initial volume of 0.01 IIf solution was entering the top of the adsorbent column, additional pure solvent was added until the leading edge of the solvent just reached the bottom of the packed tube. Development was then stopped by releasing the vacuum, the packed column extruded, and t,lie leading and trailing edges of the zone located by streaking either with dilute alkaline permanganate solution or with a saturated solution of 2,4-dinit,rophenyIliydrazine in 2 Jf hydrochloric acid. The permanganate test AoLved up as DRIED SILICIC ACID a green zone against a purple backgrouiiii color, aiicl the 2,4--dinitrophenylhydrazine gave an orange zone on a yellom background. The R values were calculsted as the ratio 1 , I 1 , , I I 5 10 15 20 25 30 35 40 45 between the distance moved by the leading or trailing edge MOLARITY x 10) of t,he zone from the top of the adsorbent column, to that distance moved by the solvent-i.e., the column length. Figure 5. Effect of Concentration on RLYalues of 3Iethjl Dodecyl Ketone Developed with Benzene Since the trailing edge of these zones could not begin to move from the top of the adsorbent until development with pure solvent was commenced, the apparent X T values obtained in this way were, of course, lower than their actual rite of iiiovenient which approximated that of the leading edges. -4s all of the important data obt,ained in thi8 study are shown in the accompanying graphs with sufficient' accuracy for anyone desiring the numerical values, no tabular data will be included here.
I
SU.\IhIARY
0.0-
c
D\@
6-
SILICIC
c4-
I
30
60
100
I
200
1
,
I
300
SIDE OHAIM M W
Adsorption -4ffinities of StraightChain Rlethyl Ketones General equation. ( 1 - R ) / R = k (.M,?c)n Fluorisil. n = - 0.45, k = 13.9 Silicic acid. n = - 1.10, k = 170
Figure 6.
these materials mere prepared by standard methods found in thc literature. -4s only very small quantities of these materials were required most of them 1% ere prepared by the relatively inefficient Grignard method from the commercially obtainable straight-chain halides. Chromatographic methods were found very valuable in removing the by-product materials from the desired ketone. These by-products were generally the corresponding hydrocarbons which were removed by chromatographing on silicic acid using petroleum ether as solvent and developer. By this method the hydrocarbons were washed through the adsorbent into the filtrate leaving the almost pure ketone near the top of the chromatographic column. The ketones were then eluted and recovered in the pure state by recrystallization where practicable. In other instances the liquid ketones were developed farther down the adfiorbent with benzene and obtained pure by elution and evaporation of the eluate. I n all cases the physical properties, either boiling point or melting point, checked closely with the values given in the literature except as noted in the case of methyl octadecyl ketone (GO). The 0.01 M solutions used in the R value determinations were made up either by pipetting or by weighing. The gieatest error possible by these methods was estimated t o be about 5%. T h e R values were determined by pipetting r r 2 ml. ( T = inside radius of the chromatographic tube in centimeters) of these standard solutions on t o the top of the adsorbent in a KO.1 chromatographic tube (inside diameter, 9 mm.; length, 90 mm.) packed to a height
The, effect of side chain o n the strength of chromatographic adsorption of some ketones on calcium carbonate. silicic acid, and Florid has been studied. Several types of side chains were investigated as well as a complete series of straight-chain methyl ketones from C8to G o . T h e rates of movement oi the adsorbate zones down the adsorbent columns were used a3 a measure of adsorption strengths. It was found that, in general, the ketones appeared to be adsorbed due to interactions between their carboriyl oxygen and the adsorbent, and that the heavier the side chain the smaller was t.he adsorption strength. Certain exceptions were noted, however, where the addition of a -CH2 group to the side chain decreased t.he rate of zone movement appreciably. These decreases i n R value seemed to be periodic within the honiologous series and preliminary work has indicated that the location of these Rvalue depressions is dependent only on the nature of the adsorbed molecule, although their magnitude is affected by the adsorbent. Not only the mass of the side chain but also it3 nature is of import,ance in determining adsorption strength. The effect of concentration on R value was studied and found t o be slight up to 0.05 M) and data n-ere given to show the statistical reproducibility of measuring the rate of zone movement ( R values) as a measure of adsorption strengths. ACKNOR LEDGMENT
T h e authors wish to thank H. T. Jackson for his asaistance in obtaining much of the dat.a presented in this paper. They ale0 wish t o acknowledge their indebtedness t o the Office of Saval Research whose generous sponsorship has made this work possible. LITERATURE CITED
( 1 ) Claesson, S., Arkiv. Kerni, Mineral. Geol., 23A (1) (1946).
(2) Crumpler and Yoe, "Chemical Computations and Errors," pp. 194-5, New York, John Wiley & Sons, Inc., 1940. (3) Elder, A. L., and Springer, R. A,, J. Phys. Chem., 44, 943 (1940). (4) Holmes, H. N., and NcKelvey, J. B., Ibid., 32, 1522 (1928). (5) LeRosen, A. L., Monaghan, P. H., Rivet, C. A . , and Smith, E. D., ANAL.CHEM.,23, 730 (1951). (6) Trueblood, K. K., and Maimberg, E. W., Ibid., 21, 1055 (1949). RECEIVED August 9, 19.50. Presented before the Divisions of Analytical, Industrial, and Engineering, Petroleum, and Physical and Inorganic Chemistry, Adsorption Symposium, at the 116th Xeetinp of t h e AMERICAN CHEMICAL SOCIETY, -4tlantio City, ?J J.
'
