to 12 hours of elapsed time. Operator times are 30 to 40% of the above. The analytical scheme is given in Figure 10. The above times are all based on the use of one apparatus and 30- t o 48-mesh C-22 firebrick. The elapsed times could probably be reduced 10 to 20% by using a coarser support. If two units are available, the complete analysis of Cb and Cg olefins can be made in 6 or 8 hours, or about half as much operator time. The methods developed are applirable
to higher boiling hydrocarbons, particularly with regard to type separation, although this will become more difficult a t higher carbon numbers because of the rapidly increasing number of isomers and the decreaqing effect of structure on activity coefficients. ACKNOWLEDGMENT
The author is indebted to Sigurd Groennings for his interest, advice, and encouragement, and for assistance m-ith the manuscript.
LITERATURE CITED
(1) Griddle, D, IT.,L~ T ~K. L,, ~ ANAL.CHEM.23, 1620 (1951). (2) Dimbat, RZ., Porter, P. E., Stross, F. H., Zbid.,28, 290 (1956). (3) Eggertsen, F. T., Knight, H. S., Zbid., 30, 15 (1958'1. (4) Eggertsen, F. T., Knight, H. S., Groennings, S., Ibzd., 28, 303 (1956). (5) Porter, P. E., Deal, C. H., Stross, F. H., J. Am. Chem. SOC.78, 2999 (1956). R~~~~~~~ for review 31arch 1, 1957. Accepted July 22, 1957.
Gas Chroma tog raphy Effect of Type and Amount of Solvent on Analysis of Saturated Hydrocarbons F. T. EGGERTSEN
and H. S. KNIGHT Shell Development Co., Emeryville, Calif.
b In the gas chromatography of saturated hydrocarbons, adsorption on the surface of the support may be controlled to provide a wide naphtheneparaffin selectivity range b y adjusting the amount of solvent. From this standpoint a study of the effects of type and amount of solvent on selectivity and separating efficiency for Cj to C? saturates i s presented. With the bare support paraffins are retarded, but the peaks are broadened b y tailing. Small amounts of supported solvent prevent tailing without much effect on selectivity. By gradually increasing the amount of solvent, the packing i s transformed continuously into the liquid partition type, where the role of adsorption i s negligible and naphthenes are retarded. The separation of certain Cg paraffins is also affected b y the amount of solvent. The most retentive solvents for naphthenes vs. paraffins were ethylene glycol and 2,2'-oxydipropionitrile. The separating efficiency, determined at temperatures to give comparable emergence times, was affected very little b y variations in the amount of supported solvent in the range from 1 to 20%. Larger or smaller amounts of solvent tend to broaden the peaks, decreasing the efficiency.
G
is the general name for a chromatographic separation effected by a moving gaseous phase. The classical chromatographic techniques, adsorption and partition, may be employed, depending on whether a n active surface or a supported liquid is used in the column. When the separation is due to different degrees A S CHROMATOGRAPHY
of physical adsorption the process will be called gas-solid chromatography, and when due to partition it will be called gas-liquid chromatography. With small amounts of liquid the separation depends on both adsorption and partitioning, and this will be referred to by the general term gas chromatographv, or, more specifically, gas chromatography over a liquid-niotiified solid adsorbent. A recent publication on the analysis of saturated hydrocarbons ( 3 ) stated that paraffins emerge later than naphtheiies of the same boiling point in gas-solid chromatography, while in gasliquid chromatography the naphthenes are invariably retarded. Tailing of the peaks, normally encountered when dry surfaces are employed, was shown to be virtually eliminated by adding small amounts of liquid. This did not affect the naphthene-paraffin sequence. The usefulness of one such liquid-modified column consisting of 1.5% squalane (hydrogenated squalene) on Pelletex (a furnace black) was demonstrated for the analysis of Cs and Cs saturates. The liquid-modified type in conjunction with the liquid type has marked potentialities for saturates analysis in general; because their naphthene-paraffin selectivities are in reverse directions, they supply complementary information. The aim or the present study was to develop columns of both types to give the greatest possible resolution and naphthene-paraffin selectivity. Various solid adsorbent. and liquids were tested. Particular emphasis was directed toward the effect of amount of liquid on the solid carrier, because this factor determines the direction and degree of naphthene-
paraffin selectivity and also affects the resolution or separating efficiency. THEORY
The theory of gas-liquid chromatography, introduced by James and Martin (4) and Martin and Synge (6), has been recently elaborated by Pierotti and coworkers ( 7 ) and by Porter, Deal, and Stross (8). Although some of the columns described here are not the partition type, the results obtained are calculated in terms of the above partition theory for purposes of comparison, and data that are not strictlj applicable are so noted. According to the theory, Vi;
-
V, =
HOT',
(1)
and
n-here Vp IT,
H" V, yo
Po
retention volume corrected for pressure drop ( 4 ) = void volume in column = ratio of solute per unit volume of solvent to that in gas, measured at infinite dilution = volume occupied by stationary solvent in column = activity coefficient at infinite dilution = vapor pressure of the solute =
The term, V i - V,, is proportional to the emergence time measured from the air peak, designated as t . Upon making this substitution in Equation 1 and combining n i t h Equation 2, we get (3) Thus, the emergence timeu of two coniponents of equal vapor pressure are inversely proportional to their activity VOL. 30, NO. 1, JANUARY 1958
15
~
coefficients. The ratio of activity coefficients, calculated from a knowledge of emergence times and vapor pressures, is a measure of solvent selectivity for the solutes concerned. As a simple selectivity test for naphthenes us. paraffins, the emergence times of cyclohexane (boiling point, 80.7" C.) and P,~-dirnethylpentane (boiling point, 80.5" C.) were determined. Because they have equal vapor pressures a t the temperatures of interest, the ratio of the naphthene to paraffin emergence times, tN/tp, is equal to the inverse activity coefficient ratio. y P / y N . As a measure of separating efficiency, apparent theoretical plates mere calculated from the formula, (4V/A)* where, in the chromatogram, V is the distance from the time of sample addition to the peak and A is the extrapolated width of the base of the peak(1, 5 ) . EQUIPMENT AND MATERIALS
Apparatus. T h e apparatus used was essentially as described by Dimbat, Porter, and Stross ( 2 ) . The sensing elements were thermal conductivity cells (Gow-Mac Instrument Co., Madison, K. J.) of the convection-diffusion type with a 100-ma. bridge current. The recording potentiometer was operated a t 5 to 25 mv. for full-scale deflection. The conductivity cell block was contained in a thermostated oil bath at 100" C. The column was heated separately in a second liquid bath or in a furnace (Hevi-Duty Multiple Unit ?io. 2, with a chamber 2l/2 X 5 inches), and connected to the conductivity cell block by means of l/rinch copper tubing. As in previous work (S), samples were injected directly into the cell block by means of a syringe connected to the carrier gas inlet line. The carrier gas was helium, flowing usually at 20 ml. per minute. Preparation of Columns. T h e supported liquids were added t o t h e solid carrier as solutions in petroleum ether (for squalane and other hydrocarbons) or in ethyl alcohol (for oxygenated compounds). T h e low boiling solvent was evaporated slowly on a hot plate or steam bath while stirring, and finally in an oven for 1 hour at 110" C. The material was then packed with the aid of an electric vibrator into the desired length of l/a-inch (outside diameter) copper tubing, the ends were plugged with glass wool, and the tubing was wound into a coil on a 13/&ch tube as a mandrel, All columns were dried a t least 0.5 hour at 110" C. in a flow of helium before testing. All percentages of liquid are expressed as grams per 100 grams of dry solid support. Solid Supports. Solid supports and adsorbents are listed in Table I together with their sources and surface areas. I n addition t o these, C-22 firebrick (Johns-Manville Co.), a Celite-base material having a surface 16
ANALYTICAL CHEMISTRY
Table 1.
Emergence Time Ratio, t N / t P I of Cyclohexane to 2,4-Dimethylpentane with Various Solid Adsorbents ('/r-inch column packed &h granules in 14- to 100-mesh range; helium, about 20 ml. per
minute; sample, about 2 mg.) Surface Area, Sq. Length, bleters/Gram" Feet
Temp., Solid (2.6 Silica gel, Grade 70 340 2 80 (Davison Chemical Co.) Celite VI11 3 4 22 (Johns-Slanville Co.) Porous glass 200 4 50 (Corning Glass Works) Alumina, Alorco F-20 180 2 150 (Aluminum Co. of America) Magnesia 10 180 100 to 180 (Westvaco Alineral Prod. Div.) Carbon, Columbia Grade L 1500 1.5 280 (Union Carbide Chemicals Div.) 20 10 120 Carbon, Pelletex (Godfrey Cabot Co.) Low temperature nitrogen adsorption method. * Temperature selected to give emergence times of a few minutes to 0.5 hour.
area of 3 square meters per gram, was used as a support in some of the tests. The granule size of all of the solids was in the range from 14 to 100 mesh. Most of the work was done with Celite VIII, Pelletex, and C-22 firebrick, using 14to 48-mesh granules. Supported Liquids. These are listed together with their sources in Table 111, b u t two of them may need further description: Squalane is a CSoHe2paraffin (2,6,10, 15,19,23-hexarnethyltetracosane; molecular weight,423; boilingpoint,210°C. at 1 mm.) obt'ained by hydrogenation of squalene (an acyclic isoprenoid obtained from shark liver oil) over a platinum catalyst at 200 to 1000 pounds per square inch. Its high boiling point, good thermal stability, and low viscosity made it an especially attractive liquid of the paraffin type. Triol 230, formula, Hb-CH2CHa
x""
CHIOH
bH2OH
I
-CH2-O-CH2-CH2-OH
It is obtainable from Union Carbide Chemicals Co., 90% pure. Hydrocarbon Test Samples. Purity of all of these was 99% or better. They were obtained from Phillips Petroleum Co., Bartlesville, Okla., or t h e American Petroleum Institute. EFFECT
OF
AMOUNT OF SUPPORTED LIQUID
As the first step in developing effective columns of the solid adsorbent or paraffin-retarding type, a number of adsorbents were tested for naphtheneparaffin selectivity. Fine granular solids were employed in a column of such length and temperature to cause emergence of the cyclohexane or 2,4dimethylpentane in a few minutes to 0.5 hour. The t N / t P ratios obtained are shown in Table I, where surface areas of the solids are also given.
O
t v / IP
0.il 0 67 0 60
0 55
0.57 0 3i 0.39
With all solid adsorbents the t N / t p ratios were less than unity. These values, ranging from 0.4 to 0.7, are much lower than for liquid-type columns, where the ratio was always greater than unity (as shown below). Tailing was encountered with all solid adsorbents to a n extent which appeared to increase roughly with the surface area. As reported previously, the tailing can be reduced by adding a strongly adsorbed or high boiling liquid to the adsorbent (3). For example, it was prevented with silica gel of 340 square meters per gram by 25% of water or ethylene glycol. With the lower area Celite VIII, on the other hand, less than 1% of water, squalane, ethylene glycol, or Triol 230 was sufficient to eliminate tailing. Tailing with Pelletex was prevented by 1.5 to 5% of various liquids, including cetane, squalane, Ondina 133 (a mineral oil), Triol 230, or a polyaromatics fraction; however, the separating properties of these lowliquid packings varied somewhat with the particular liquid employed. When the amount of supported liquid was just sufficient to prevent tailing the tN/tP ratio still remained less than unity, paraffins being retarded as with the bare solid. I n all cases, however, the liquid-modified adsorbents were less retentive than the bare solid. I n general, better efficiency in terms of total apparent theoretical plates could be achieved with supports of low surface area, such as Celite VIII, C-22 firebrick, and Pelletex; these were therefore used. The poor efficiency observed with high area solids is probably due t o restricted diffusion in these highly porous materials. The two solvents, Triol 230 and squalane, representing polar and nonpolar solvent types, supported on Celite VI11 or Pelletex, were selected for detailed investigation. By studying the
effect of aiiiount of solvent with these systems it was hoped to develop efficient columns of both the p a r a f i retarding (liquid-modified solid) and naphthene-retarding (liquid) types. Figure 1 illustrates the effect of the amount of Triol and squalane, supported on Pelletex, on the relative emergence times of cyclohexane and 2,4-dimethylpentane. It shows that with the bare solid adsorbent the paraffin is retained more strongly than the naphthene. This situation prevails also with small amounts of supported liquids on the adsorbent, while the times decrease for both hydrocarbons, reaching a minimum (or nearly constant value) a t 2.5% Triol and 1.5y0squalane. Then, with more supported liquid the emergence time increases and so does the emergence time of the naphthene relative to the paraffin, t.v/tp. It is apparent that a t the intersection of the curves the column is nonselective because tAV/tP= 1. Beyond this point there is a reversal of the sequence of emergence of naphthenes and paraffins. Figure 2 shows the change in tAy/tp ratio with amount of Triol or squalane supported on Celite or Pelletex. With both solids squalane was the more effective liquid for reducing adsorption by the surface, as indicated by the steeper initial slope of the squalane curves. This can be attributed to its better solvent power for hydrocarbons and possibly also to differences in surface orientation. With large amounts of supported liquids the t N / t p ratio rises to 1.3 to 1.4 for squalane and 2.0 or more for Triol, the more polar solvent. Higher ratios were attained
with Triol on Celite than on Pelletex, probably because ac!sorption effects are more difficult to overcome with the higher surface area Pelletex. The effect of amount of liquid on the resolution of individual paraffins was also studied using the following mixture of hexanes as a test sample: Peak No. in Figure 3 1 2
3 4 5
solid, component 3 (2-methyl pentane) emerges with component 4 (3-methylpentane) ; then, as the solvent increases, component 3 moves toward component 2 (2,3-dimethylbutane) and finally emerges with it. The effects observed may not be entirely attributable to the percentage of stationary phase; difBoiling Point, O C.
Compound 2,2-Dimethylbutane 2,3-Dimethylbutane 2-hlethylpentane 3-RIethylpentane n-Hexane
In all cases the test columns were inch; relatively short, only 10 feet by the helium sweeping rate was about 20 ml. per minute and the sample charge, about 4 mg. Temperatures were selected to cause emergence of n-hexane in 20 to 50 minutes. For the case of squalane 011 Celite VI11 the effect of amount of liquid on the resolution is shown graphically in Figure 3. Tailing is severe with the bare solid and resolution is poor. The presence of 0.5% solvent virtually eliminates tailing, and a partial separation of all components is achieved. With 3% solvent components 2 and 3 appear as one broad peak and they merge even more completely when 20% squalane is used. With 40% solvent the resolution is so poor that no welldefined peaks are obtained. The amount of solvent also affects the relative emergence times of components 2, 3, and 4. With the bare
Relative Amounts by Volume
49.7 58.0
1 2
60.3 63.2 68 8
3 4 10
ferences in column temperature may also be a factor. Similar tailing effects and change in relative positions of components 2, 3, and 4 were observed with other solventr carrier combinations; these chromatograms are not presented. The effect of amount of solvent is further indicated by the data in Table 11, where the apparent number of theoretical plates is given for component 1 (2,sdimethylbutane) and the occurrence of tailing is noted. As mentioned before, the temperature for each test was selected to cause total emergence of the sample in 20 t o 50 minutes. Data for diisodecyl phthalate also are included for comparison with Triol and squalane. The number of calculated plates was 900 to 1400 for columns containing amounts of solvent from about 1 to 20%. With less solvent or with none, the number is often lower, presumably
TRIOL ON CELITE Vill. 23.C.
I
2 5c
TRIOL ON P E L L E T E X . 12C’C
5 2’oy TRIOL O S PELLETEX. - 1
1
II
7 6 c
SUPP0f:TED LIQUID. % w
Figure 1 . Effect of amount of solvent on emergence times of cyclohexane and 2,4-dimethylpentane
0
3QUALA:iE
O S PELLETEX.
,
30
70
C l i 7 3 R i 3 D LIQL:D
,
IZC’C.
40
%*
Figure 2. Effect of amount of supported liquid or1 f . v / t p ratio VOL. 30, NO. 1, JANUARY 1958
17
20 ml./rnm
40% G L Y C O L ON C - 2 1 FIREBRICK, 25.C..
9 1 I5
CYCLOHEXANE AND
~
A L L Cy S A P H T H E A € S
(20001
(10001
EMERGENCE TIME, rnm 30
Figure 3. Effect of amount of solvent on separation of five Ca paraffins on squalane-Celite Vlll column
due to tailing. The precision of these measurements is i 1 0 0 to 200. Plate numbers for peaks which, tailed badly are not included because of the uncertainty in measuring the widths of such peaks. \F7ith 40CT, liquid on Pelletex or Celite, the plate number was very low. However, this large amount of squalane was accommodated on C-22 firebrick witliout causing a decrease in plate number. 'The superiority of C-22 firebrick over other solids as a support for large amounts of liquid (40% diisodecyl phthalate) has already been reported by Dimbat and Stross ( 2 ) . SELECTIVITY OF VARIOUS LIQUIDS NAPHTHENES AND PARAFFINS
18
ANALYTICAL CHEMISTRY
A0 70 BOILING POINT. 'C
80
90
100
Figure 4. Emergence times and partition coefficients for Ca to Cg saturates
0
A
Naphthenes lsoparafftns n-Paraffins
paraffins. Consequently, three of the best columns were selected for a more thorough study in which the emergence times of all the common Ca to C?naphthenes and paraffins were measured individually, using 3-mg. samples in every
selective for naphthenes us. paraffins. It was desired to determine whether the degree of naphthene-paraffin selectivity indicated in the tests with cyclohexane and 2,4-dimethylpentane holds generally for other naphthenes and
~
Table II.
(Columns, 10 feet
FOR
It is evident that for maximum retention of naphthenes-that is, for high tN/tP ratio-relatively large amounts of supported liquid are desirable, and Triol 230, a polar solvent, gives a higher ratio than does squalane, a nonpolar solvent (Figure 2). Also, higher ratios mere attained with Celite VI11 as a support for Triol than 1Tith Pelletex. I n a search for other solvents which retard naphthenes even more strongly than does Triol, several solvents were tested as 40% liquid on C-22 firebrick (14 to 48 mesh). With about 2 mg. each of cyclohexane and 2,4dimethylpentane as test samples, the t N / t p ratios were determined as shown in Table 111. The temperatures employed were 25" t o 80' C. and the test columns varied in length from 5 to 50 feet. All t H / t P ratios were greater than 1, the highest being exhibited by the strongly polar ethylene glycol and 2,2'oxydipropionitrile. Silicone SF-96 gave the lowest ratio and thus was the least
10
40
Column Squalane/Pelletex
x
Apparent Theoretical Plates
inch; test peak, 0.2 to 0.5 mg. of 2,2-dimethylbutane) Temp.,
% Liquid 0 0.5 1.0
1.5 3 10
20 40
Triol/Pelletex
1.5
3 10
20 Squalane/Celite VI11 9
40 0 0.1 0.3 0.5 1 3 10
20
Squalane/C-22 firebrick
c.
95
TO
25 25 25 40 GO 80
25 25 25 25 25 0 0 0 0 0
25
70
io
40 3 10
120 25
40
70 25 25 25
20
Diisodecyl plitlialate/C-22 firebrick
O
5
10
20
70
io
900
900
1000 900 1000
200
800 Tailing 1100 1300
900 200
1
' ' Tailing 700 800 1200 1100 1100 1000 100 (1 mg. cyclohexane) 1100 900 1400
1200 900 1200 1300
case t o permit a precise comparison, Long columns were employed as follows:
squalane on C-22 firebrick, but in the liquid-modlfied column (1.5% squalane
Column and Length Temp., 25 407' ethylene glycol/C-22 firebrick, 100 feet 25 37, squalane/C-22 firebrick, 50 feet 1.5y0squalane/Pelletex, 50 feet 40
These particular packingswereselected because they represent, respectively: (1) one of the most strongly retentive columns for naphthenes ( t N / t p = 2.7); (2) a n essentially liquid-type column ( t N / t P = 1.3) which gave the greatest number of individual peaks in tests with a mixture of all 26 of the commonly encountered Cs to C7 saturates; and (3) the most useful column found of the liquid-modified solid adsorbent type ( t N / t p = 0.67). The emergence times and partition coefficients (computed using Equation 1) are plotted against the boiling points in Figure 4. This plot provides a comparison of the emergence times of naphthenes and paraffins a t equal boiling point and thus gives a measure of column selectivity. The values given on the partition coefficient scale for 1.5y0 squalane on Pelletex are indicated in parentheses because they are not true partition coefficients for the supported liquid; in this case the behavior of the packing is governed not only by liquid partition but also by adsorption effects a t the solid int'erface. Figure 4 shows that all three columns exhibit naphthene-paraffin selectivity. The column containing 40% glycol on C-22 firebrick is the most selective, naphthenes being greatly retarded, as expected from t'he high tq,/tp ratio. With this column cyclohexane (boiling point, 81' C.) and all the dimethylcyclopentanes (88" to 99' C.) emerge well after n-heptane (98" C.). As expected, the naphthenes are also retarded in the essentially liquid-type column of 37',
O
C.
Inlet Pressure, Mm. 430 280 280
Helium,
hf 1. /Minute 20 60 27
on Pelletex) the situation is reversed. Thus, conclusions drawn from the cyclohexane and 2,4-dimethylpentane pair are generally applicable in the range studied. The possibility of obtaining separate carbon number cuts with the column of 1.5% squalane on Pelletex is also indicated in Figure 4. For example, the highest boiling Cs compound, c) clohevane (81" C,), emerges well ahead of all the C7 compounds. On the other hand, with the 3% squalane and 40% glycol on C-22 firebrick columns, naphthenes are retarded relative to paraffins of the same boiling point, and emerge with paraffins of higher carbon number With all columns the group of four dimethylcyclopentanes (88"t o 92" C.) tends to approach the paraffin curves, indicating relatively poor selectivity for these particular naphthenes. I n place of squalane other hydrocarbon liquids may be used in the above packings with little or no change in the resolution of Csto C7 saturates. With Pelletex as the support, hexadecane was about equal t o squalane. The mineral oil Ondina 133 (about C17 to CI9 saturates), supported on either Pelletex or firebrick, retards naphthenes slightly more than does squalane, but is otherwise equivalent. However, syualane has a lower volatility than these other liquids and therefore would tie expected to give more stable columns. The emergence times for paraffins and iiaphtl~rneqdo not approach each other as the molecular weight increases, the absolute time difference being roughly maintained. This suggests that
Table 111. Naphthene-Paraffin Selectivity of Various Liquids (40yoliquid on C-22 firebrick, test samples, 2 mg. of cyclohexane and 2,4-dimethylpentane)
Liquid and Source Silicone, SF-96 (General Electric Co.) Glycerol (Shell Chemical Corp.) lliglycerol (Eastman Kodak Co.) Squalane (Shell Development Co.) Mineral oil, Ondina 133 (Shell Oil Co.) Bicyclohexvl (Monsanto Chemical Co.) Diisodecvl phthalate (Rlonsanto Chemical Co.) Triethylene glycol (Union Carbide Chemicals Co.) Dimethylsulfolane (Shell Chemical Corp.) Triol 230 (Union Carbide Chemicals Co.) Ethylene glycol (F. 11.Speekman Co.) 2,2'-Oxydipropionitrile (American Cyanamid Co.)
Column C.
-is
t\/k
25 25
1 3 1 1 1 -1 1 4 14 1 4 1 8 10
Temp,
so SO
SO
so
25 25 25 25
25
1 2
2 5
2 7 2 8
with the glycol column a type separation can be achieved through Cs and possibly higher, a t least for narrow cuts. For type separation above about Cg and for samples of wide boiling range, a still more selective solvent would probably he required. Solvent selectivity is best expressed in terms of activity coefficients, which are inversely proportional to emergence times a t equal vapor pressures (Equation 3). At the column temperatures employed, the vapor pressures of the hydrocarbons investigated follow closely the sequence of the boiling points. Hence, the activitl- coefficient ratio of naphthenes to paraffins as classes can be estimated from the curves in Figure 4. For example, with the ethylene glycol column, it may be concluded that the activity coefficients of naphthenes are about half those of paraffins. It is apparent also that the dimethylcyclopentanes group (85" to 92' C.) has relatively higher activity coefficients than the other naphthenes. Thus, as methyl groups are added to the ring, the activity coefficient approaches that for the paraffins, as might he eupected. W'ith the paraffinic solvent, squalane, supported on C-22 firebrick, the naphthenes still have lower activity coefficients (greater emergence times) than the paraffins, but the difference is only 15 to 25%. I n order to achieve the reverse situation-that is, lower activity coefficients for paraffins than for naphthenes-it was necessary t o employ active solid surfaces, as with the 1.5y0squalane on Pelletex column. CONCLUSIONS
The results illustrate the use of adsorption effects of the support as a variable in gas chromatographic separations. Because of these effect3, the separations obtainable with liquidmodified solids (up to a few per cent liquid) depend upon the amount as well as the type of supported liquid. By altering the amount of liquid, the positions of certain saturates in the chromatogram, particularly naphthenes, can be changed a t mill. It follows that, in order to resolve a given saturates mixture, there is an optimum amount of an>- given liquid which will yield a maximum number of separate peaha. Based on the information obtained in thiq study, a very satisfactory detailed analysis of the Cg to C; saturates in a ieformer charge stock n-as accomplished. This work is reported separately in this issue. ACKNOWLEDGMENT
The authors are indebted to Sigurd Groennings for helpful advice and encouragement and for valuable assistance in preparing the manuscript. VOL. 30, NO. 1, JANUARY 1958
19
Groennings, S., Ibid., 28,303 (1956).
LITERATURE CITED
(1) Deemter, J. J. van, Zuiderweg, F. J., Klinkenberg, A., Chem. Eng. Sci., 5 , 271 (1956). (2) Dimbat, M., Porter, P. E., Stross, F. H., ANAL.CHEX 28, 290 (1956). (3) Eggertsen, F. T., Knight, H. S.,
(4) James, A. T., Martin, A. J. P., Biochem. J . 50, 679 (1952).
(5) Keulemans, A. I. hf., "Gas Chromatography", p. 113, Reinhold, Sew York, 1957. (6) Martin, A. J. P., Synge, R. L. hl., Ibid., 35, 532 (1943).
(7) Pierotti, G. J., Deal, C. H., Derr, E. L., Porter, P. E., J . Am. Chem. SOC.78, 2989 (1956). (8) Porter, P. E., Deal, C. H., Stross, F. H., Ibid., 78, 2999 (1956). RECEIVEDfor review April 4, 1957. Accepted July 22, 1957.
Determination of Five- to Seven-Carbon Saturates by Gas Chromatography F. T.
EGGERTSEN and SIGURD GROENNINGS
Shell Development Co., Emeryville, Calif,
'
b The saturates portion o f a reformer charge stock, containing chiefly Ce and C, compounds, can be analyzed for all but one of its 25 Cs to C7 components b y gas chromatography employing three separate columns. In general the results agree with those obtained b y an American Petroleum Institute cooperative testing of the same sample by other means (mass and infrared spectrometry, refractivity intercept, and catalytic dehydrogenation); the average deviation is about 0.3%. The average accuracy i s about 0.170 as established from analyses of a synthetic blend, simulating the CS to C7's in the actual sample. Certain C i s present in the latter cause some interference, but this can be corrected for b y mass spectrometry of a few cuts. The procedure, with or without spectrometry, i s relatively fast, requiring 12 to 16 hours.
S
hydrocarbons in a petroleum distillate can be determined by various methods, including mass and infrared spectrometry, refractivity intercept, and catalytic dehydrogenation. To ascertain just what these methods will do and how much effort is involved in obtaining various degrees of information, the American Petroleum Institute's Committee on Analytical Research recently carried out an extensive cooperative test program with a reformer charge stock containing essentially C6 and C7 saturates (9). The results from the individual laboratories in some cases shorv appreciable deviations; yet the average values must be considered good, eren though the accuracy is not stated. However, the procedures for the detailed analysis are rather time-consuming, principally because the physical measurements must be carried out on numerous sharply distilled fractions. I n view of the rapid strides made with gas chromatography (GC), it seemed that a simpler method could be devised ATURATED
20
ANALYTICAL CHEMISTRY
gram, obtained from Johns-Manville; 14- t o 48- or 30- t o 60-mesh granules. PELLETEX.A furnace black (Godfrey Cabot Co., Boston, Mass.) of about 24 square meters per gram surface area; 14- to 48-mesh granules. Supported Liquids. SQUALANE.A Cs&z paraffin (2,6,1OJ15,19,23-hexamethyltetracosane) of molecular weight 423 and boiling point 210' C, a t 1 mm., obtained by hydrogenation of squalene (an acyclic isoprenoid from shark liver oil) over a platinum catalyst a t 200 to 1000 pounds per square inch (11). Its high boiling point, good thermal stability, and low viscosity make it an especially attractive liquid of the paraffin type. 2,2'-OXYDIPROPIONITRILE, obtained from American Cyanamid Co.; boiling point 120" C. at 1 mm. Hydrocarbon T e s t Samples. A P I COOPERATIVE TESTSAMPLE. A 170" to 220' F. (72' to 124' C. b y A S T M D-86 distillation) straight run catalytic reformer charge stock from California crude, was supplied by the California Research Corp., San Francisco, EXPERIMENTAL Calif. The saturates, 91.3y0 by weight, as determined by ASTM D-939 (9), were Apperatus. T h e assembly deisolated by liquid phase chromatography scribed previously, with helium as in a specially designed large scale FIA carrier gas and thermal conductivity (fluorescent indicator adsorption) colcells for detection, was employed (6). umn ('7). From a charge of 10 ml., Except in a few experiments when there was obtained 99 =t 1% of the fractions were t o be collected for saturates, which must, therefore, be mass spectrometric analysis, t h e eluted considered representative of the satuhydrocarbons were oxidized by passage rates in the sample. over hot copper oxide as they emerged SYNTHETICBLEEDS. These were from the column and were thus deprepared from Cs to C7 saturated hydrotected as carbon dioxide. The comcarbons of minimum 997, purity, obbustion train consisted of an 8 X 1/4 inch tained from Phillips Petroleum Co., column of copper oxide (J. T. Baker Bartlesrille, Okla., or from the American Chemical Co., wire form) at 650' C. and Petroleum Institute. The hydrocara 1 foot X, inch column of calcium bons concerned are listed according to sulfate (Drierite) at room temperature boiling points in Table I with abbrevito remove water of combustion. The ations used in the chromatograms. thermal conductivity cells were kept in Column Preparation. T h e supa liquid bath a t ambient temperature. ported liquids were added t o t h e Samples were added directly to the solid carrier as solutions in petroleum column inlet through a serum cap. ether (for squalane) or in ethyl alcohol The inlet section, n-hich was packed (for oxydipropionitrile) . T h e solvent with glass wool, was heated electrically was evaporated slorvly on a hot plate to about 100' C. to achieve rapid or steam bath while stirring, and finally evaporation of the sample during chargin an oven for 1 hour a t 110' C. The ing. Solid Supports. C-22 FIREBRICK.d r y material was packed with the aid of an electric vibrator into 50 feet of I/,A Celite-base material having a surinch outside diameter copper tubing by face area of about 3 square meters per
to furnish the same amount of information. A gas chromatographic method for complete analysis of Cg and Ce saturates, employing a liquid-modified solid adsorbent as the column packing, has been reported by the authors (6); and further study of naphthene-paraffin selectivity yielded two other distinctive types of column packing (4). By combining information obtained from separate determinations with these columns a satisfactory detailed saturates analysis through 0, appeared feasible. T h e test sample chosen was the same material used in the cooperative program mentioned above. Mass spectrometry was utilized t o study interferences by Cs hydrocarbons, which overlap the C?'s to some extent. Also, a combined gas chromatographicmass spectrometric procedure was developed as an alternative to one employing gas chromatography alone.
