hurst Works, private communication,
1960. (19) Jones, E. A., Parkinson, T. F., T*G., J . Chem* PhW 235 ( l Y 3 Uj .
(20) Klinkenberg, A., Chem. Eng. Sci. 15, 255 (1961). (21) Lysyj, I., Newton, P. R., ANAL. CHEM.35,90 (1963). (22) Melior, J. M., “Comprehensive Treatise on Inorganic and Theoretical
Chemistry,” Supplement 11, Part I, Longmans, Green, London, 1956. (23) Muetterties, E. L., phiup, w.D., J . Am. Chem. SOC.79, 332 (1957). . . (24) Phillips, T. R., geylan, D., “Gas Chromatography 19G2,” M. van sway, ed.1 PP. 247-59, Butterworthh London, 1962. (25) Phillips, T. R., Owens, D. R., “Gas Chromatography 1960,” R. P. W. Scott,
ed., pp. 308-17, Butterworths, London,
1960. (26) Reed, C. D., Harrison, J. A., J. Sci. Instr. 36, 240 (1959).
RECEIVEDfor review June 27, 1963. Accepted September 13, 1963. Presented in part a t Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pa., 1961.
Separation of Stereoisomers by Column Fractional Precipitation WOLFGANG W. SCHULZ’ and WILLIAM C. PURDY Department of Chemistry, University o f Maryland, College Park, Md.
b Column fractional precipitation was successfully employed in separating 1 ,3,5-tri-(4-bromophenyl)benzene from 1 ,3,5-tri-(3-bromophenyl)benzenet
0,-
p-quaterphenyl from m,p-quaterphenyl, and in determining the purity of a tribromodecaphenyl, 1,3-di-(2-chlorophenyl)-l,3-cyclohexadiene, and 1,3di-(3-bromophenyl)benzene. Isomeric steroid mixtures of 5/3-androstan-3a01-1 7-one and 5a-androstan-3a-0117-one and of 3a-hydroxycholestane and 3/3-hydroxycholestane were seporated, but a mixture of Sa-androstan3,17-dione and 5/3-androstan-3,17dione could not be resolved.
B
WILLIAMS(1) described a column fractional precipitation
AKER AND
technique which has been successfully applied to the fractionation of industrial types of polystyrene of large molecularweight distribution and of polymers of narrow distribution. Separation is achieved by subjecting the polymer, supported on an inert packing of micro glass beads, to a simultaneous temperature and solvent gradient. A linear temperature gradient is maintained along the length of the column and the percentage of “good” solvent in the initial nonsolvent increases exponentially, At a particular temperature and solvent composition, the most soluble molecular-weight species will dissolve and move down the column to a lower temperature region where it will again precipitate. When the solvent composition has been sufficiently enriched with “good” solvent, the precipitate will redissolve and proceed down the column in a series of precipitation and dissolution steps. The material will leave the column as a saturated solution. I n a previous paper, Schulz and 1 Present address, Esso Research and Engineering Co., Linden, N. J.
2044
ANALYTICAL CHEMISTRY
tion was transferred to a 250-ml. Erlenmeyer flask and then equilibrated in a constant-temperature bath. The solution was titrated with nonsolvent from a 50-ml. buret to a visual turbidimetric end point. In many cases, the precipitate was redissolved by an additional 20 ml. of good solvent and titrated again. The titrations were run a t temperatures varying by 10-degree intervals from 10” to 60” C. Plots of the solvent composition a t the turbidimetric end point us. temperature were EXPERIMENTAL constructed. I n Figure 1 are shown Reagents and Apparatus. Reagent the solubility curves of 3a-hydroxygrade chemicals were used without cholestane and 3p-hydroxycholestane as further purification. All polyphenyl obtained from these turbidimetric titrasamples were kindly made available tions. FRACTIONATION PROCEDURES. ilfter by G. F. Woods of this laboratory. This steroid samples were provided preliminary solubility studies in various by the Division of Biochemistry, solvents and solvent pairs, the pdioxane-water system was chosen to Walter Reed Army Institute of Reseparate 1,3,5-tri- (3-bromo phenyl) search. Commercial p-dioxane was benzene from 1,3,5-tri-(4-bromophenyl)purified by distillation from all-glass benzene, water being the nonsolvent. apparatus and was stored in the presence A 1 : l mixture of the two isomers was of sodium (3). The fractional precipitation appafinely ground and samples ranging from ratus was built in this laboratory and 22 to 57 mg. were mixed with glass has been described elsewhere (14). beads and transferred to the column. The temperature gradient of the appaThe 250-ml. mixing vessel was initially ratus was determined to be linear from filled with either 33 or 45% p-dioxane 0” to 65” C. The measurement of the in water and was surrounded by a solvent gradient has been described in a heating mantle. Solvent, which was previous paper (1.4). preheated to 50” C., was delivered to All absorbance measurements were the column a t a rate ranging from 6 to 10 ml./hour. The top of the column made either with a Beckman DK-1 or was heated t o 63” C. Thirty to 39 Model B spectrophotometer. Procedure. SELECTION O F SOLVENT fractions were automatically collected. The fractions were analyzed spectroSYSTEMS. The choice of the solvent photometrically. system depended on preliminary soluThe purity of a tribromodecaphenyl bility studies and turbidimetric titrawas investigated in two stages. A tions. Approximately 0.1 gram of the substance under investigation w~ts 45.7-mg. sample was placed on the column and treated initially with a dissolved in 5 ml. of good solvent. solvent mixture of 48% p-dioxane in Nonsolvent was added with a pipet water. Forty fractions were collected. until turbidity was observed. The A second sample of equal size was fracappearance of the precipitate and the reversibility of the precipitation process tionated with solvent of a starting composition of 807, p-dioxane in mater. were noted. Promising solvent sysThirty-six fractions were collected. tems were further investigated by Ultraviolet spectra were obtained on turbidimetric titrations. 1-mi. aliquots of each of the fractions. Approximately 100 mg. of a single Similar purity analyses were made for substance was weighed into a 100-ml. 1,3 - di - (2 - chlorophenyl) - 1,3- cyclovolumetric flask and diluted with good hexadiene and 1,3-di-(3-brornophenyl)solvent. A 10-ml. aliquot of the solu-
Purdy (14) derived equations for the average solvent composition and the interstitial column volume for a fractional precipitation column of the Baker and Williams type. This work shows that column fractional precipitation is not limited t o the separation of polymers, but can be extended to the separation of other high molecularweight organic substances.
Tube Number
Figure 2.
Separation of two isomeric polyphenyls A.
B. 0
ti)
20
30
40
50
60
Teinpercture ( " C )
Figure 1. Solubility curves from turbidimetric titrations A. B.
1,3,5-Tri-(3-bromophenyI)benzene 1,3,5-Tri-(4- bromopheny1)benzene
3a-Hydroxy;holestane 36-Hydroxy~:holertane
benzene. Since these latter two substances are highly viscous liquids, they could be mixed directly with the glass beads. A water-n-propanol solvent system was used in the latter two cases, with pure water initially in the 250-ml. mixing vessel. Fractions were analyzed spectrophotometrically. Fractionation of a mixture of three quaterphenyls, o,p (I), m,p (11), and p,p (111) was investigated with an initial 10% n-propanol in water in the 250-ml. mixing vessel and a sample size of 26.4 mg. The three isomers were present in a weight rat o of o,p:m,p:p,p = 1:1:0.25.
initially 40% dioxane, was used for system (ii), a mixing volume of 500 ml. and pure water were used for systems (i) and [iii). THE ZIMMERMAKN METHOD.The ketonic steroid fractions were analyzed by the colorimetric Zimmermann method (16). This method was modified for nonketonic saturated steroids, such as system (ii) (8). Five milliliters of the fractions were evaporated in 50-ml. test tubes and heated for 30 seconds in boiling water with 1 ml. of a freshlyprepared solution of 1.250 grams of recrystallized 3,5-dinitrobenzoyl chloride in 25 ml. of reagent-grade pyridine. Each mixture was quickly transferred with 25 ml. of benzene to a 60-ml. separatory funnel and washed successively-twice with 10 ml. of 1N hydrochloric acid, twice with 10 ml. of 1N sodium hydroxide, and twice with water. The benzene was evaporated on a steam bath. The color was developed by adding 16 drops of technical-grade acetone and 4 drops of a 0.1% ethanolic potassium hydroxide solution. The mixtures were allowed to stand for 5 minutes and then were diluted with ethanol. The absorbance was measured a t 550 mp and plotted against tube number.
Because of the rapid discoloration of the 3,5-dinitrobenzoyl chloride solution, 50 mg. of the acid chloride was added directly to the evaporated steroid fractions, followed by 1 ml. of pyridine. This variation was applied to fractions from the fractional precipitation of 3ahydroxycholestane. RESULTS AND DISCUSSION
Fractionation of Polyphenyl Systems. I n contrast t o the detailed spectrum of benzene, the ultraviolet absorption spectrum of polyphenyls and halogenated polyphenyls consists of broad peaks in the range of 210 to 280 mp. The difference between the spectra of stereoisomeric polyphenyls often consists only in small wavelength shifts. However, these shifts are generally sufficient to study the degree of separation of mixtures. The separation of 1,3,5-tri-(3-bromopheny1)benzene from 1,3,5-tri-(4-bromophenyl) benzene and the distribution curves of these compounds are shown in Figure 2. The three fractions of greatest overlap were analyzed from the absorbance values a t 260.2 and
For the fractionstic 11 of hormones, three diastereoisomeric bystems were chosen: (i) 5p-androstan-3a-oI-17-one 5ay-androstan-3a-ol-17-o~e (ii)3P-hydroxycholestane 3 a-hydroxych slestane (izi) 5p-androstan-3,17-dione
5a-androstan-3,17-dione
The water-p-dioxane solvent system was applicable in each case. While the 250-ml. mixing vessel, contaiiiing
Tube Number
Figure 3. A. B.
Purity analysis of a tribromodecaphenyl
Plot of absorbance at 281.0 mp
Plot of absorbance at 260.0
V.I
tube number
mp va. tube number
VOL. 35, NO. 13, DECEMBER 1963
2045
0.1
12 14 16 10
PO PP 94 26
20 30 32 34 36
TUBE NUMBER
15
Figure 4. Purity analysis of b rornopheny I) benzene
17
19
21
23 Tube Number
1,3-di-(3Figure 5.
Separation of two quaterphenyls A.
253.0 mp using simultaneous equations. The absorptivities of the pure substances at these wavelengths were determined from spectra of the pure compounds t o be: ~260.2
&so.o
for and
= =
52,360liters/mole-em. 61,370liters/mole-cm.
1,3,8-tri-(3-bromophenyl)benzene u 2 M )= . ~ 77,130liters/mole-cm. = 65,310 liters/mole-cm.
a263.0
for lJ3,5-tri-(4-bromophenyl)benzene. The calculations were not carried t o fractions of smaller overlap, as the error became so large that the calculated concentrations would have been without significance. The purity analysis of a tribromodecaphenyl showed the presence of four major impurities (see Figure 3). The structure of the tribromodecaphenyl was believed t o be:
Br
.-. I L’
2046
ANALYTICAL CHEMISTRY
8.
o,p-Quoterphenyl m,p-Quaterphenyl
This structure gives rise t o a spectrum mith a broad maximum at 281.0 mfi. The impurities were discovered from the spectra of individual fractions and the relative absorbance values a t particular wavelengths. The broad peak of the distribution curve (Figure 3) is due to compound IV, its fractions having the same spectrum a\ the unfractionated material. The relative height of a peak is not a measure of the quantity of impurity present, since one cannot assume equal abiorptivity values for all substances of the mixture. This has been brought out in the change of relative peak heights upon plotting the absorbance a t different wavelengths. The broad peak of the distribution curie was extrapolated to fractions 17 and 32. The absorbance due to compound IV in these fractions was estimated to be equal t o one-half of the absorbance of fraction 25, which contained no impurities. By recording the ultraviolet spectra of fractions 17 and 32 against fraction 25 as a reference solution, the absorbance due to compound IV in fractions 17 and 32 was subtracted out. The resulting spectra proved to be identical, having two maxima at 257.4 and 2284 mp and were, as such, completely different from previously recorded spectra. Although the structures of the impurities in fractions 17 and 32 are not known, one ran assume that both compounds contain the iame chromophoric groupl giving rise to identical spectra. They
must be different compounds because of their differing solubilities. The purity analysis of 1,3-di-(3bromopheny1)benzene resulted in a very sharp distribution curve, indicating the absence of impurities (Figure 4). The curve represents the typical distribution of a substance which rapidly establishes equilibrium. A sharp skewed distribution was generally found with amorphous substances, such as the dibromoterphenyl under discuqsion, which is a viscous liquid. The distribution curve for 1,3-di(2- chlorophenyl) - 1,3 - cyclohesadiene, which is also a liquid and which was fractionated under the same conditions as the dibromoterphenyl, was slightly broader, ranging from fraction 15 to 34, with an initial small shoulder which may have been due to the presence of traceq of the monochloro compound, l-phenyl3 - (2 - chlorophenyl) - 1,3 - cyclohexadiene. Mixtures of three terphenyls and of some quaterphenyls have been partially separated by adsorption chromatography on various common adsorbants (‘7). The separation of three isomeric quaterphenyls (I, 11, 111) was investigated by column fractional precipitation. From turbidimetric titrations it n a s elident that the o,p-isomer (I) was slightly more soluble than the m p isomer (11) and that the p,p-i.omer wai very inqoluble, in fact so inqohible that no definite turbidirnt-tric end points could he o b e r r r d from t h r very dilute
sollitions which were einployed (0.00067 gram/30 ml). The separation of opquaterphenyl from m,p-quaterphenyl i\ shown in Figure 5 . The peak? :IIC broad and
