Ha me Response in the Quantitative Determination of High Molecular Weight Paraffins and Alcohols by Gas Chromatography GERALD PERKINS, Jr., R. E. LARAMY, and L. D. LIVELY' Confinental Oil Co., Ponca City, Okla.
b The flame response of high molecular weight alcohols and paraffins is found to b e additive throughout the molecular range investigated. The slope for a homologous series of paraffins when relative response is plotted vs. carbon number is identical, within experimental error, to that obtained for a homologous series of alcohols. The response for the alcohols is also additive with carbon number, but appears to have an absolute response equivalent to one-half carbon less than the corresponding paraffin. The flame detector in conjunction with a packed column of glass microbeads coated with 0.1% liquid phase and column temperature programming gives quantitative results for homologous series of paraffins. A homologous series containing even-number carbon atoms from tetradecane through dopentacosane is resolved in a single analysis.
T
HE quantitative characteristics of the flame ionization detector have been studied for a variety of compounds (1, 3, 4, 10, 11). I n general, all of this work has been limited to compounds of relatively low molecular weight. Good agreement is shown in the results, which indicate a very orderly and predictable response for each molecular species. Some data, however (2, 5 ) , have been interpreted to give an entirely different relative response for paraffins and alcohols and this would lead t o serious disagreement for compounds of higher molecular weight. The purpose of this paper is to resolve possible disagreement by elaborating on previous work ( I O ) and extending the study to include relatively high molecular weights for the two molecular types, paraffins and alcohols. Data are incidentally presented which illustrate the use of the flame ionization detector to obtain quantitative data when coupled with packed columns and programmed column temperature. The addition of programmed column temperature permits the quantitative analysis of homologous
* Present address, Fisher Scientific Co., Pittsburgh, Pa. 360
0
ANALYTICAL CHEMISTRY
series of rather wide molecular weight range. REAGENTS AND APPARATUS
The alcohols were obtained from t h e Applied Science Laboratories. The alcohol purity was assumed t o be in excess of t h e precision of t h e experiment. Paraffins. T h e paraffins were Phillips pure grade obtained from t h e Phillips Petroleum Co. For t h e response study their purity was assumed t o be in excess of t h e precision of t h e experiment. For t h e quantitative analysis study each compound used in t h e preparation of blends was chromatographed separately and its concentration corrected for a n y impurities present. Apparatus. RESPONSE STUDY.T h e basic chromatograph was built in our laboratory. A Keithley, Model 415, electrometer a n d a Leeds & S o r t h r u p recorder were used as t h e amplifier a n d recording systems, respectively. No flow splitter or column was used. I n place of the column a n open tube connected t h e injection port t o the detector. This should eliminate a n y effect inherent in a column. A Hamilton 1-pl. syringe, used to inject the samples, was calibrated to determine the volume of sample held by the needle. A Disc integrator was employed to obtain the area measurements of the detector response. QUANTITATIVE ANALYSISSTUDY.For this study, a different detector and injection block, also built in our laboratory, were employed. To obtain column temperature programming, the temperature programmer and air oven of an F & M Model 500 programmed temperature chromatograph were utilized. A Keithley, Model 415, electrometer and a Bristol I-mv. recorder were used for the amplifier and recording system, respectively. The 6-foot column consisted of 0.25-inch aluminum tube packed with 0.10 weight % Apiezon L coated on a solid support of 60- 80-mesh glass microbeads. Alcohols.
PROCEDURE
As in previous work ( I O ) relative response has been measured as a function of carbon number for a homologous series under specified conditions. These data have Response Factor
Study.
then been related to corresponding data for hydrocarbons. Certain conditions observed during the measurements are: The preheater and open tube leading from the injection point to the detector were kept at a temperature sufficient to prevent condensation. The p a r a f h and alcohol samples, as well as the sgringe, were heated to prevent solidification in the syringe before injection. The area of the detector response was integrated with a Disc integrator; three or four replicate runs were made for each compound and the average was taken for the final value. Areas did not deviate more than 5 to 7y0 in any instance. The absence of a column required that a known volume of pure material be injected. To realize sufficiently accurate volumes, calibration of a 1-pl. syringe was necessary. This was done in accordance with the procedure previously outlined (IO). Response measurements were made with several volumes of a given compound, in our case ranging from 0.20 to 1.00 p l . The resultant area was plotted as a function of the injected volume of sample as measured by the syringe graduations. Extrapolation of this line gave a n area for zero volume, or more precisely, for the residual volume of the syringe needle. From this, the actual volume of the syringe, including the volume of residual sample in the needle, can be calculated. For our particular syringe the zero volume was 0.22 pl, I n practice, the entire sample was introduced into the injection chamber, by allowing time for the residual material t o evaporate from the needle. Quantitative Analysis Study. The chromatographic conditions used were as follows: Flow rate
Starting temperature Programming rate Air pressure Column
Hydrogen pressure
30 ml./min. 40" C. 5 6"/min. 15 p.s.i. 6-foot x l/c-inch
aluminum tube packed with 0.1 % Apiezon L on 60/80-mesh glass microbeads 3 p s.i.
To determine the influence of hydrogen pressure on response, hydrogen pressure n-as plotted us. response. AS expected, a maximum was observed,
in this case a t a hydrogen partial pressure of 3 p.s.i. This region of maximum sensitivity and minimum change of response was used throughout the study for minimizing errors due to fluctuation in gas flow rate. RESULTS AND DISCUSSION
Response Factor Study. T h e present d a t a for paraffins, combined with the earlier data of Perkins et al. (IO), were fitted with a single least-squares straight line with which all points coincided within experimental error. The slope of the fitted line was 0.663. A similar combination of present and past data for alcohols was also fitted by a single line, coinciding with all data within experimental error. The slope of this line was 0.628. The alcohol response appears to be within the range of one-half carbon number less than the corresponding paraffins. However, this is not fully conclusive for the higher molecular weights, since experimental error has become larger than the expected difference. I t is this factor which limits the molecular range through which experimental verification is possible. The slope of the two response curves may be considered identical within experimental error. To make these data more easily comparable to the work of others, carbon number has been plotted vs. “effective carbon number’’ (Figure 1). The effective carbon number of the paraffins was determined from values taken from the least-squares line. The response of hexane was set equal to 6 and all other values were adjusted accordingly. The effective carbon number for the alcohols was obtained by calculating the average value for the response per mole per carbon atom of the paraffins, and then dividing the experimentally determined response per mole of the alcohols by this figure. The solid line represents the calculated line expected for the paraffins. The dashed line represents the curve expected for the alcohols assuming a response of one-half carbon number less. Considering the experimental
O
1
--A 0
A 0
, , , , ,
1
I I I I I I I I I I I I 1 ~ 3 ~ 5 ~ 7 ~ 9 1 o i 1 1 ~ 1 ~ 1 4 1 ~ 1 ~ 1 7 1 CARBON NUMBER
I
Effective C No. of paraffins from least squares plot Effective C No. of alcohols based on respanse of 1/2 C No. less than respective paraffin Paraffin data from previous work (9) Paraffin data from present work
the extension of its conclusions to compounds of higher molecular weight. As can be observed in Figure 1,the linearity continues through the higher molecular weights studied, alcohol response being less than the corresponding paraffin. All the experimental points fall along the straight line within the error of the experiment. Although, because of the experimental error, we cannot conclusively state that the response is equal to exactly one-half carbon number less, no major deviation in slope is observed as the molecular weight increases. To conclude that there is any significant difference in the slope of the paraffin curve and alcohol curve seems unjustified on the basis of our results. The present work, coupled with excellent agreement of earlier results (10, 11), seems to substantiate the unique and additive contributions of functional groups to the response of the hydrogen a 1 9 2 0 flame ionization detector.
Table I. Compound
Alcohol data from previous work ( 9 ) Alcohol data from present work
c7
CS
, ’r 3r’1 ;i”
38H
T H
%
Area
Absolute
%
error
Blend I 2.50 1.72 1.85 i.82 4.11 3.63 5.i7 5.60 i.35 6.98 10.41 10.39 17.28 17.01 15.91 16.72 9.88 10.88 8.01 9.02 6.38 7.10 4.59 4.72 5.96 5.50
C-A”
C9 CI0 Cli
error, the points coincide well with these lines. The theories which attempt to explain the difference in response of the flame ionization detector to different functional groups have not been discussed, since agreement as to why this occurs is by no means unanimous. The two prominent theories of flame response are discussed in detail in the literature (IO,1 2 ) . Later work (7) gives considerable support to the point of view of Sternberg, Galloway, and Jones (11). The present work was entirely experimental and was designed to substantiate the published work and to demonstrate
Quantitative Results
Weight
CI2
Cu Cl4 ClS CIS ClS c 2 0
0.78 0.03 0.48 0.17 0.37 0.02 0.27 0.S1 1.00 1.01 0.72 0.13 0.46
Blend I1 2.71 5.84 9.44 24.59 25.33 16.46 15.62
CS
Cb ClO C~I Cis Cis C,l
2.58 6.06 9.54 25.40 25.66 16.76 15.99
0.13 0.22
0.10 0.81 0.33 0.36 0.3i
26H
I
4pn
24H 22H
2OH
4BH 52H -c-2__
_ I
55--
-50
45
40
35
30 TIME
Figure 2.
20
25 I N
15
10
5
0
MINUTES
Programmed temperature chromatogram column Glass microbeads (60- 80-mesh.)
VOL 35, NO. 3, MARCH 1963
361
\vas decided to investigate t h e quantitative characteristics of the flame detector used i n conjunction 11-ith a packed column. As a further extension, temperature programming was incorporated. With the increased sensitivity of t h e flanie ionization detector allowing use of less sample than the thermal conductivity detector, a glass microbead column n as used to take advantage of its favorable characteristics. One of the major advantages in employing glass beads is the elution of components much below their boiling points, thus permitting the analysis of wider molecular ranges (6, 8, 9). The quantitative results obtained are given in Table I for two blends of homologous series of paraffins. The maximum deviation in all cases is 1% or less. Figure 2 illustrates the advantage of employing a glass microbead column, a flame ionization detector, and column temperature programming. This is a Chromatogram of a sample containing the even numbered paraffins from tetradecane through dopentacosane. Nikelly (9) has recently reported the excellent results obtained for such systems with glass microbead columns and the present work indicates that, with proper choice of conditions, the possibilities can be even more startling than originally described. Since it n a s shown that
quantitative results could be obtained with a homologous series of paraffins up through eicosane, there is no apparent reason to suspect this chromatogram would not be quantitative in its entirety. However, standard samples in the niolecular weight range of dopenta. cosane 17-ere not readily available for synthesizing a standard blend. The sample size used in obtaining this chromatogram n-as 0.4 pl., illustrating the need for the added sensitivity of the flame ionization detector. Such an analysis would be difficult with conventional detectors and conventional columns. The quantitative results given in Table I were obtained iyithout the use of calibration factors. It follows from the nature of the response that the use of factors is unnecessary within molecular classes, except for the very lorn members. When samples contain members of different molecular species, factors based upon hydrocarbon response may be calculated in advance from knowledge of the subtractive contributions of the various functional groups involved (10). The glass microbeads as reported (9) make possible the analysis of high molecular weight compounds, since they are eluted at a temperature far below their boiling point. The previous use of glass microbeads was limited by the ease
with which the column overloads. K i t h the advent of the flame detector the increased sensitivity permits the use of a much smaller sample charge and column overloading is virtually eliminated. LITERATURE CITED
(1) ilndreatch, A. J., Feinland, R., ANAL. CHEY.32, 1021-4 (1960).
(2) Ettre, L. S., “Gas Chromatography,” p. 546, Academic Press, S e w York, London, 1962. (3) Ettre, L. S.,Informal Symposium on Capillary Chromatography, R. P. I\’, ASTM D-2, Section L, Atlantic City, S. J., June 27, 1961. ( 4 ) Ettre, L. S., Claudy, H. N , C h e m Can. 12, SO. 9,32-6 (1960). (5) Ettre, L. S., Coates, F. J., Cieplinski, E. W., European Convention of Chemical Engineering, Frankfurt am Main, Western Germany, June 12, 1961. (6) Frederick, D. H., e k e , W. D., “Gas Chromatography, Chap. 3, Academic Press, NeF York, 1962. (i) Fueno, T., Mukherjee, K. R., Ree, T., Eyring, H., Eighth Symposium (International) on Combustion, Baltimore, Md., 1962. (8) Hishta, C., Nesserly, J. P., Reschke, R. F., 4 x . 4 ~ .CHEM.32, 880 (1960). (9) Xkelly, J. F., Ibzd., 34, 4i2 (1962). (10) Perkins, G., Jr., Rouayheb, G. hl., Lively, L. D., Hamilton, W. C., “Gas Chromatography,” Chap. 19, Academic Press, Sew York, London, 1962. (11) Sternberg, J. C., Galloway, W. S., Jones, D. T. L., Ibid., Chap. 18. RECEIVEDfor review October 1, 1962. Accepted January 9, 1963.
Use of Differential Reaction Rates to Analyze Mixtures of Organic Materials Containing Same Functional Group Application to Mixtures of Unsaturated Compounds SIDNEY SIGGIA, J. G O R D O N HANNA, and NICHOLAS M. SERENCHA
O h Research Center, Olin Mathieson Chemical Corp., New Haven 4, Conn.
b The rate approach for the analysis of mixtures of organic materials containing the same functional group has been applied to unsaturated compounds. Both bromination and hydrogenation are used. The usual second-order rate plots are used in the case of brominations. Hydrogenation is treated as a pseudo-first-order reaction b y maintaining a relatively large excess of hydrogen present during the reaction. Standard reactionrate plots show linear portions for each component in mixtures,
R
of addition to olefinic bonds have been used in special cases for the analysis of mixtures of unsaturated compounds. The rates of
362
ATES
ANALYTICAL CHEMISTRY
reaction of perbenzoic acid with olefinic double bonds have been used as a basis of analysis (6, 6). The amount of reaction is measured after a specified time, and is related to the original concentration by reference to a calibration curve prepared from the analysis of known mixtures under the same conditions. A procedure for the determination of internal and external double bonds in polymers also uses the rates of reaction with perbenzoic acid (8). The composition of the original mixture is found by extrapolation t o zero time of the flat portion of the curve obtained by a plot of per cent reacted us. time. Mixtures of ethyl elaidinate and ethyl oleate have been analyzed, using the difference in rates
of addition of mercuric acetate ( 2 ) . Again, reference to a calibration curve relates the amount of reaction a t a specified time t o the concentrations of compounds in the original mixture. The separation of the amount! of substitution obtained while adding bromine t o unsubstituted linkages can be measured, based on differences in rates of bromine conwnption ( 7 ) . Several additional determinations are made at increased reaction times in excess of that necessary for complete saturation of the double bonds. The results are plotted against time and extrapolated to zero time to obtain the correct value for the addition reaction. One component can sometimes be determined in a mixture by a selective hydrogenation
