position of the CH3 and COOH group is more influential than that of the two COOH groups. Loss of the two carboxylic groups with rearrangement of a hydrogen atom gives rise to alkylbenzenic-type peaks (m/e 77, 91 . . .), but no clear structural correlation can be deduced. POLYCARBOXYLIC ACIDS. Spectra of aromatic acids with more than two carboxylic groups substituted on the benzenic ring are very difficult to obtain. They vaporize very slowly and have a great tendency t o decompose. We examined the 1,3,5-, 1,2,4-, and 1,2,3benzenetricarboxylic acids and obtained only partial spectra, which are not tabulated. Nevertheless, some indication of the behavior of these compounds can be deduced. The parent peak is presented only by the 1,3,5,-tricarboxylic acid. The 1 , 2 , 4 and 1,2,3- acids apparently decompose to the corresponding carboxy anhydrides. The most abundant ion for these latter compounds is a t m/e 148[P - 62(H20 COZ)]. The large pc.ak a t m / e 149 for 1,3,5-tricarboxylic acid could be derived from loss of a COz group by thermal decomposition thus forming isophthalic arid, and subsequent loss of an OH group. The m/e 166'149 ratio is about the same for isoplithalic acid and for the 1,3,5-
+
tricarboxylic acid, thus sustaining this view. Aromatic Aldehydes. Table IV includes the most significant peaks found in the spectra of the six aromatic aldehydes we examined. The molecule ion in aromatic aldehydes is a very stable one; therefore, the parent peaks are very important. KO structural correlations have been noted save for a slight decrease in abundance with increasing molecular weight. 31/e 106, 120 . , , etc., are also parent peaks of the alkylbenzenes, but identification in a mixture is possible due to the size of the P - 1 ion, which is much more important in aldehydes than in alkylbenzenes. The corresponding ions result from P OH for acids, P - OCH3 from methyl esters, and, analogously, decrease in abundance nith the proximity of methyl substituents to the functional group. The P - 29(CHO) peak derives from the loss, by a cleavage, of the functional group. It is often the base peak and its abundance diminishes rrith the number of methyl substituents. As for the alkylbenzenic-type peaks (m/e 77, 9 1 . . .), only those derived from direct 01 cleavage of the functional group are important. Spectra of terephthalaldehyde, mol. wt. 134, and 1,Cbenzenecarboxy al-
dehyde also have been obtained. Terephthalaldehyde presents a spectrum analogous to that of benzaldehyde. The major difference is that P - 29 is much less important in the former. However, P - 29 in benzaldehyde (m/e 77) is exactly equal to the sum of P 29 and m/e 77 in terephthalaldehyde. 1,4-Benzenecarboxy aldehyde presents a composite spectrum with both acidic and aldehydic characteristics. In fact, both the P - 1, P - 29 (aldehydic), and P - 17, P - 45 (acidic) peaks are present. Aldehydic-type peaks seem t o be, however, more abundant and important. ACKNOWLEDGMENT
R e thank Harold Kail, G. R. Taylor, and J. L. Taylor for obtaining and tabulating the spectra. LITERATURE CITED
(1) Aczel, Thomas, Lumpkin, H. E., ANAL. CHEM.32, 1819 (1960). (2) Lumpkin, H. E., Nicholson, D. E.,
(3) . , Lui
33,476 (1961). (4) McLafferty, F. W., Gohlke, R. S., Ibzd., 31,2076(1959). RECEIVEDfor review July 20, 1960. Accepted November 21, 1960. 8th Meeting of ASTM E 1 4 Committee, Atlantic City, N. J., June 27-July 1, 1960.
Total Analysis of Olefinic Naphthas by Mass Spectrometry A. J. FRISQUE, H. M. GRUBB, C. H. EHRHARDT, and R. W. VANDER HAAR Research and Development Department, Standard Oil Co. (Indiana), Whiting, Ind.
b Fragment-ion and low-voltage parent-ion mass spectra, used separately for analysis, suffer from serious disadvantages. Fragment-ion spectra are reproducible but nonselective; low-voltage parent-ion spectra are selective but poorly reproducible. In a new method to determine hydrocarbon types in olefinic naphthas, the two spectra are combined so that the strong feature of each 'offsets the weak feature of the other. A faster and more detailed total analysis of the naphtha, therefore, results. The CEC 21-103 mass spectrometer was modified simply and inexpensively to provide the needed range of ionizing and repeller voltages.
K
and detailed methods for analyzing naphthas are important to the petroleum refiner because composition is related to product quality APID
and to the process and operating variables used in refining. For a material as complex as a naphtha bracketing the boiling range of gasoline, hydrocarbon types, not individual components, must be determined. I n general, methods of analysis classify hydrocarbon types according to degree of either hydrogen saturation or cyclization, but no single method classifies in both ways. Fluorescent indicator adsorption (4) determines saturates, olefins, and aromatics but if the parafKnnaphthene ratio measuring cyclization is required, refractive index and density must be measured on the saturates alone. Mass spectrometry determines either saturation or cyclization but does not distinguish between the two directly. Olefin and naphthene isomers with the same value of z in their CnH2,+ formulas are not resolved through their fragment spectra. When both are
present, olefins must be determined in dependently (8). I n theory, supplementary chemical techniques should not be needed to distinguish olefins from naphthenes by mass spectrometry. Low-voltage parent-ion techniques selective for ole6ns and aromatics have been reported (5-7) and fragment-ion methods that give the sum of olefin and naphthene isomers (6,3) have been tested by Research Division IV of ASTM Committee D-2. From olefins measured at lowvoltage and the olefins-plus-naphthenes measured a t normal voltage, it should be possible to distinguish between the two without supplementary techniques. Unfortunately, the precautions which have been necessary to compensate for the poor reproducibility of low-voltage spectra have discouraged their application as a substitute for sample separation. Olefin determinations based on VOL. 33,
NO.
3, MARCH 1961
389
low-voltage spectra have required either instrument sensitivity calibrations before and after each spectrum was taken (6) or the removal of noncontributing saturates to permit normalization of the data (6, 7). These time-consuming precautions, however, need not apply to a total analysis method. Here, the normalvoltage spectra that are necessary to distinguish between paraffins and naphthenes, can serve the additional function of internally standardizing the lowvoltage spectra and thus compensate for poor reproducibility. A new method has been developed that combines low- and normal-voltage spectra so that the strong feature of each offsets the weak feature of the other. The low-voltage spectrum determines only olefin-alkylbenzene ratios, which are negligibly affected by the poor reproducibility of absolute peak heights. Olefins are then calculated from the more accurate alkylbenzene content as determined from the normal-voltage spectrum. The normal-voltage spectrum thus internally standardizes the low-voltage spectrum, which in turn provides the selectivity otherwise lacking. Separations are required only for the initial low-voltage calibration. EXPERIMENTAL
il CEC Model 21-103 mass spectrometer was modified to obtain the required low ionizing and fixed repeller voltages, according to the schematic diagram shown in Figure 1. Major advantages of this circuitry over a previous modification ( 7 ) are the absence of dry cells and provision for a wide range of ionization conditions. Ionizing voltage can be varied from 2 to 100 volts in two ranges: 2 to 50 and 30 to 100 volts. Repeller voltages from 0 to 10 volts are obtainable with a directreading ten-turn potentiometer. Satisfactory low-voltage conditions were established by scanning the saturates from a catalytic naphtha a t successively lower ionizing voltages, with 3 volts applied to the repellers, until
Table 1.
CarbonNo. 5 6 7 8 9 10 11 12b 0
b
I
Figure 1.
Modification of CEC lsatron for low-voltage use
saturate peaks disappeared. Although the nominal voltage a t disappearance depends on filament conditian as well as on repeller and filament geometry, the effective voltage as measured by the peak-height ratio of two compounds of different ionization potentials is little affected by these factors. This effective voltage, approached from the high voltage side, gave a toluenebenzene parent peak ratio of 2 : l for an equal volume mixture of the two compounds, which corresponds to a nominal voltage of 7.5 with the instrument used. To obtain materials for calibrating a t low-voltages, where sensitivity coefficients are required for each carbon number, a catalytic naphtha covering the
Low-Voltage Coefficients for Catalytic Naphthas
Cyclodiolefins6 m/e Coeff.
Cycloolefins m/e Coeff.
66 80 94 108 122 136 150 164
68 82 96 110 124 138 152 166
0.43 0.36 0.28 0.23 0.37 0.51 0.64 0.78
Olefins headings denote (C,H,, Extrapolated values.
390
OD 3 VU-I50
ANALYTICAL CHEMISTRY
0.98 0.81 0.47 0.57 0.63 0.69 0.76 0.82
+ =) class only.
Monoolefins m/e Coeff.
Alkylbenzenea m/e Coeff.
70 84 98 112 126 140 154 168
78 92 106 120 134 148 162
0.49 0.59 0.89 1 .oo 1.15 1.29 1.44 1.58
1.oo
0.50 0.32 0.28 0.38 0.47 0.57
range of 5 through 12 carbon atoms was fractionally distilled into 10% cuts which were analyzed. To determine the olefin and alkylbenzene contents of these cuts, the supplementary procedure for olefinic naphthas that has been circulated to members of Section M, R.D. IV, ASTM Committee D-2 which is currently awaiting finalized form, was used. This combination mass-spectrometricchemical method. which is a modification of Brown's original fragment-ion method (d), uses relative rather than absolute intensities in the characteristic fragment-peak summations (3) and lists calculated in\ ewes to determine ten hydrocarbon types. These inverses depend somewhat on the cycloparaEnolefin ratio that is initially determined by trial and error. Two fragment-ion spectra are used for an analysis, that of the total cut, and that of the saturate fraction from the total cut separated, in this case, by acid treatment (1). A disadvantage of the method, the need to separate chemically the saturates routinely, is not a disadvantage in nonroutine calibrations. Here, the separation provides an alternative to hydrogenation (5) for obtaining low-voltage calibrations. To minimize the errors that might result from analyzing narrow boiling cuts, their spectra were first combined to calculate the over-all composition of the unfractionated naphtha. Then, the olefin and alkylbenzene contents of each cut were calculated from the contributions of each cut to these total values. These
known olefin and alkylbenzene values and their low-voltage peak heights were used to calculate the low-voltage coefficients shown in Table I. Here, since relative sensitivities were used, benzene was arbitrarily assigned the value of unity. The product of these coefficients and the corresponding peak heights gave relative olefin and alkylbenzene concentrations for each carbon number. With these relative values and the total alkylbenzene content determined from the normal-voltage spectrum, olefins were determined.
To determine the day-to-day stability of the low-voltage sensitivity ratios, a catalytic naphtha was analyzed on three successive days. The mono-olefin results, 27.1, 26.8, and 25.6%, from the low-voltage spectra corresponded to mono-olefin plus monocycloparafi results of 35.5, 34.3, and 34.Q%, respectively, obtained from the fragment spectra. The 1.5y0 spread in both sets of results suggests that low-voltage precision approaches that for fragment spectra under the conditions used.
Table II. Effect of Small Voltage Changes on Apparent Concentrations
(Liquid volume yo) Volts Cyclo- Cyclo- Mono- Alkyloleben(Ioniz- diole- olefins zenes fins fins ing) Absolute Peak Heights 7.0 7.5 8.0
0.28 0.39 0.51
Absolute peak heights are poorly reproducible under low-voltage conditions because, in this steeply rising portion of the ionization-efficiency curve, small changes in voltage cause large changes in peak heights. Relative peak heights are more reproducible because all peaks are similarly affected when such voltage changes occur. Table I1 shows olefin and alkylbenzene low-voltage results for a catalytic naphtha obtained a t three ionizing voltages: the standard voltage of 7.5 and a t 0.5 volt on either side of this value. Results are compared for both absolute and relative peak heights. Although the over-all 1-volt change causes an apparent two- to threefold change in concentration with absolute peak heights, olefin concentrations are little affected by this change with relative peak heights. Under the normal ionizing conditions of 50 to 70 volts, a similar I-volt change would not perceptibly affect the peak heights, inasmuch as these higher voltages lie in the flat portion of the ionization-efficiency curve. Three naphthas-ne catalytically cracked, one thermally cracked, and one reformed-were each analyzed by the present method and the combination method used for calibration. Table I11 compares the olefin and naphthene portions of the total analyses by the two methods. The catalytic naphtha coefficients of Table I were used for all naphthas to determine the general applicability of a single set of coefficients. Results between the two methods agree sufficiently well to suggest that the shorter direct rnethod can rep’ace combination methods for many applications, and. that a single set of low-voltage coefficients can have wide applicability, extending even to different naphthas.
The present method, which has been in routine use for more than a year, has reduced analysis times for olefinic
Table 111.
0.49 0.38 0.31
7.0 7.5 8.0
b
0
a
9.2 8.1 7.6
..
10.5 9.2 10.0
.. ..
Naphthene-Olefin Results by Two Methods
(Liquid volume %) Catalytic Thermal Naphtha Naphtha Monocycloparaffins Dicycloparailins TricycloparafEns Mono-olefins Cyclo-olefins Cyclodiolefins Parailins Alkylbenzenes Indanea Naphthalenes
29.3 53.8 85.0
Relative Peak Heights
CONCLUSION DISCUSSION
5.9 9.5 16.3
5.2 8.4 12.5
14.6 2.2 0.4 14.0 8.7 1.0
14.7 2.3 0.8 13.9 8.6 0.6 22.6 33.5 2.6 0.4 100.0
-
b
(i
26.9 3.7 0.2 8.3 10.3 3.4
28.1 5.1 1.1 7.1 8.9 1.5 30.9 15.3 0.7 0.3 100.0
-
Reformed Naphtha b
(1
7.2 0.4 0.0 1.6 0.4 0.2
7.9 0.6 0 0 0.8 0.2 0.3 26.2 63.5 0.4 0.1
100.0
Combination method.
* New method.
naphthas by removing the need for routine chemical separation of hydrocarbon types. Furthermore, contrary to what might be expected of a shortened procedure, more compositional details result from using it. Olefin and alkylbenzene portions of the total analysis are frequently reported by carbon number, which would not be possible if fragment spectra alone were used. Use of a normal-voltage spectrum to standardize a low-voltage spectrum internally should also apply to other materials. Although alkylbenzenes are the “built-in” internal standard with naphthas, any compound or group of compounds of similarly low ionization potential in some other material could serve a similar purpose. The sole requirement of such a compound or group
of compounds is a unique mass spectrum for both normal- and low-ionizing voltages. LITERATURE CITED
(1) Am. SOC.Testing Materials, “Stand-
ards on Petroleum Products and Lubricants,” p. 515, 1959. ( 2 ) Brown, R. A., ANAL. CHEM.23, 430 (1951). (3) Clerc, R. J., Hood, A., O’Neal, M. J., Jr., Ibid., 27,868 (1955). (4) Criddle, D. W., LeTourneau, R. L., Ibid., 23,1620 (1951). (5) Field, F. H., Hastings, S. H., Ibid., 28, 1248 (1956). (6) Keerns, G. L., Maronowski, N. C., Crable, G. F., Ibid., 31, 1646 (1959). (7) Lumpkin, H. E., Ibid, 30, 321 (1958). (8) Mikkelsen, L., H o p b s , R. L., Yee, D. Y., Ibzd., 30, 317 (1958).
RECEIVED for review September 8, 1960. Accepted November 23, 1960.
VOL. 33, NO. 3, MARCH 1961
391