Determination of Unsaturated Hydrocarbons by l o w Voltage Mass Spectrometry F. H. FIELD and S. H. HASTINGS Technical and Research Divisions, Humble O i l and Refining Co., Baytown, Tex.
A method of determining unsaturates in petroleum naphthas gives quantitative information concerning the compound types present and the compounds belonging to these types. Interference from paraffins and naphthenes is negligible. The analysis is performed with a mass spectrometer operated with the ionizing voltage adjusted to form only the molecule ions of the unsaturates in the sample analyzed. The method is applicable to the analysis of catalytic cracked naphthas, virgin naphthas and hydroformates, and propylene polymers.
T
HE mass spectrum of a polyatomic compound obtained in
the conventional way-that is, with ionizing voltage in the range 50 to 70 volts-is complex, consisting of ions of many masses ( m / e values). I n general, the mass spectrum may be thought of as consisting of the molecule ion (parent ion) and fragment ions. If, however, the ionizing voltage is adjusted so as to be larger than the ionization potential of the molecule but smaller than the lowest appearance potential, the mass spectrum will contain only one peak, that of the molecule ion. T h e existence of small peaks due t o the carbon-I3 content of the molecule is ignored here as trivial. This simplification of the mass spectrum has, if it could be properly exploited, certain obvious advantages for both qualitative and quantitative analysis. For example, in a mixture of compounds with roughly the same ionization potentials and
Table I. Molecule-Ion Intensities of Various Hydrocarbons at an Ionizing Voltage of 6.90 Volts Compound
Obseryeda Intensity
Aromatics 1050
Toluene o-Xylene
1620 1890
l,B,B-Trimethylbenzene 1,3-Diethylbenzene
2130 1250
1-Heptene 2-Heutene
2-11ethyl-2-hexene I-Octene 2,3-Dimethyl-2hexene 2,4,4-Trimethyl-2pentene I-h-onene 1-Derene Average
3- or 4-Methyl-1-cyclopentene Cyclohexene 2.4-Dimethvl-1-cvclopentene 1-Methyl-1-cyclohexene 4-Methyl-1-cyclohexene ~~
1588
Olefins 2-Methyl-2-pentene I-Hexene 2-Hexene 2.3-Dimethyl-2-butene
Observed Intensity0
Cyclic Olefins
Benzene
Average
Compound
396 868 1330 1380 728 940
Diolefins 1000 220 335 1830 13i 80
800 9.5
1,2-Pentadiene trans-1,3-Pentadiene 1,I-Pentadiene 2-JIethyl-1,3-butadiene 1 ,5-Hexadiene 2.3-Dimethvl-l.3butadiene Average
387 2380 165 1740 0 1880
1089
1480 488 95 79 556
a Expressed as chart divisions (1 chart division current).
-
1 X 10-1' ampere ion
lowest appearance potentials, each peak in the mass spectrum of the mixture obtained under the conditions outlined above would represent the presence in the mixture of isomers of a given compound type and molecular weight. The compound types and molecular weights could then be qualitatively identified by the most cursory inspection of the mass spectrum, and, if appropriate calibration data were available, quantitative calculations of the concentrations of the mixture components could be made quickly with no more equipment than a slide rule. I n particular, the complicated matrix calculations necessitated by fragmentation interference effects would be eliminated. These considerations have long been recognized but not much used. I n 1942, T a y or ( 5 ) applied the method to the determination of simple inorganic gases-e.g., nitrogen, oxygen, carbon monoxide, and carbon dioxide-and Stevenson and Wagner ( 4 ) used i t t o analyze mixtures of isotopically enriched paraffins. Similarly, it has been fairly general practice to measure free radicals formed by pyrolyzing the gas entering the ionization chamber of a mass spectrometer by operating a t ionizing voltages above the ionization potential of the radical but belom the ionization potential of the gas from which the radical is produced. However, in the analysis of complex mixtures of compounds such as those found in petroleum products problems are encountered which are not present in the case of the more simple mixtures referred to above. This paper describes a method for the determination of olefins and aromatics, which overcomes these problems. From an inspection of a table of hydrocarbon ionization potentials, it may be seen that the ionization potentials of olefins and aromatics are generally a t least a volt lower than the ionization potentials of paraffins, if the comparison is made between olefins and paraffins with equal numbers of carbon atoms. For example, Honig (3)finds that the ionization potentials of the normal paraffins range from 13.04 e.v. for methane to 10.19 e.v. for n-CloHp2, while the values for the straight-chain 1-olefins range from 10.62 e.v. for ethylene to 9.51 e.v. for Cl0H20. Furthermore, the variation in ionization potentials of molecules containing five or more carbon atoms (the start of the naphtha range) is relative small. Thus, in going from n-CgH12 to n-CioH22, the ionization potential drops from 10.55 to 10.19 e.v., and in going from l-CsH,o t o 1-CloH20, i t drops from 9.66 to 9.51 e.v. T h e more extensively the olefin )C=C(
group is substituted,
the lower the ionization potential. The ionization potential data on branched paraffins are somewhat sketchy, and the data on naphthenes are for practical purposes nonexistent, but theoretical calculations made in this laboratory by Franklin ( 8 ) indicate that the ionization potentials of paraffins are not dependent upon the structure of the molecule and that naphthene ionization potentials are of about the same magnitude as those of paraffins. The ionization potentiah of aromatics are in general even lower than those of olefins, starting a t 9.4 to 9.5 e.v. for benzene and dropping as the ring hydrogens are replaced by alkyl groups (1). Thp ionization potentials of polynuclear aromatics are lower still. Ijnder these circumstances, one might reasonably expect to find an ionizing voltage which mill ionize only olefins and aromatics, and this expectation is realized in actual practice. The range of ionizing voltages wherein only molecule ions are formed--that is, the voltage difference betn-een the ionization 1248
V O L U M E 2 8 , NO. 8, A U G U S T 1 9 5 6
1249
potential and the lowest appearance potential-also exhibits a dependence upon compound type. T h e data in the literature upon which generalizations must be based are relatively scant\and refer only to compounds of relatively low molecular weight but to the extent that they are representative, they indicate that the range for aromatics is greatest, on the order of 4 e.v., followed by olefins with values from 1 to 3 e.v., which in turn are followed by paraffins with values from 0 to 1 e.v. These facts, taken in conjunction with the above-mentioned relationships betxveen ionization potential and compound type, indicate t h a t it should be possible t o set the ionizing voltage so as to ionize, but not fragment, the unsaturates in a mixture of saturates and unsaturates.
Table 11. Temporal Variations in Intensity of 2,4,4-Tri-methyl-l-penteneIons Date 8/27/54 8/30/54
9/2/54 9,’2/54 9/2/34
2/3/54 917j54 9/8/’54
9/10/54 9/13,’d4 9/13 154 9/14/2 9/15/54 9/16/34
IntensiFy, (Chart Divisions) 487 307 370 378 408 422 456
449 43 1 Filament replaced 630 615
615 5%
EXPERIMENTAL
.411 the n-ork described n-as done with a Westinghouse Type
LV mass spectrometer modified by replacing the original mass spectrometer tube with one manufactured by J. A. Hipple. This
tube, except for engineering details, is similar to the \Vestinghouse tube. T h e ion currents were measured with a Brown recording electrometer, which has a sensitivitj- of 1 X 1O-ls ampere per chart division. Magnetic scanning n-as used. Except for some preliminary n-ork, the ionizing electron current, the ionizing voltage, and the ion dran-out potential (pusher potential) were maintained a t 9.5 pa., 6.90 volts, and 1.9 volts, respectively (hereafter referred to as standard low voltage conditions). The electron current was set a t 9.5 pa. in order to achieve the highest possible ionization sensitivity compatible with the operating characteristics of the instrument, and the ion draTvout potential was set a t 1.9 volts as a continuation of a longstnnding practice which is known from experience in measuring ionization and appearance potentials t o give satisfactory results. I n choosing a value for the ionizing voltage, it was necessary to strike a compromise bet!\-een too high a value, n-hich worilti give a high ionization sensitivity but undesirable amounts of fragmentation and naphthene and paraffin ionization, and too low a value, which would decrease the ionization sensitivity undull-. T h e value of 6.90 volts \vas chosen on the basis of trial and error experience and seems to work w l l , This is the voltage applied betveen the filament and the ionization chamber of the mess spectrometer, but it is not the actual voltage of the ionizing electrons, as they receive additional energy contributions from the ion drawout electrode, the thermal energy of the filament, et,c. K O attempt was made to control the ion source temperature, and, as a consequence, it varied b e t m e n about 175’ and 200” C., depending upon the state of the filament. However, as the only ions formed are molecule ions, such control is not so important as in high voltage mass spectrometry, in xvhich ion fragmentation plays a major role; furthermore, sensitivity variations due t o temperature fluctuations were accounted for by a sensitivity calibration procedure described later. Samples were charged to the instrument by means of the conventional constant-volume pipet technique. Because molecular ionization cross sections decrease sharply with decreasing ionizing voltage below about 50 volts, perhaps the first, problem to be considered in developing a low voltage analytical method is m-hether ion currents of sufficient intensities for the purposes of practical analysis can be achieved. Esperience shows that this is possible and for illustration the moleculeion intensities of various compounds belonging to several compound types are listed in Table I. These compounds were chosen t o include both favorable and unfavorable examples, and thus the table defines rather !vel1 the range of intensities likely to be found with unsaturated hydrocarbons. T h e intensities of t h e aromatics are uniformly high, running between about 1000 and 2000 divisions. As tn-o divisions are clearly distinguishable above instrument noise, aromatic concentrations on the order of 0.1% are detectable. The olefin intensities exhibit a wide variation, the smallest intensities, on the order of 100 divisions, occurring for the 1-olefins, and the largest, about 1500 divisions, for compounds containing the RR’C=CR”R”’ structure. However, even for the less sensitive compounds, concentrations of about 1 to 2% should be detectable. T h e intensities of cyclic olefins and diolefins are intermediate between those of aromatics and acyclic olefins Compounds that produce a vanishingly
.
,
small amount of perent ions even a t high voltages--l,5-hesadiene, for example-are not detectnhle, but, fortunately, the number of these is small. The volume of sample used in obtaining these data vas constant, although this and the corresponding pressure in the mass spectrometer reservoir were never actually measured. The volume used-i.e., the volume of the constant-volume pipetwas as large as possible compatible with the restriction thnt linearity must exist betn-een ion intensity and sample volume. This volume IT-asdetermined empirically. The position of the auxiliary magnet (electron beam collimating magnet) is of critical importance. Small changes in the position of the ausiliery magnet, can effect a tn-0- or threefold change in the magnitude of the ion currents \\-ithont bringing about any perceptible change in the r:itio of collected electron current to emitted electron current. The authors’ practice is t o adjust the position of the auriliary magnet so as to give the largest possible ion currents Compatible with the maintenance of a satisfactory ratio of collected to emitted electron current. The second problem to be considered in the development of spectrometry is that of the long-term stability t y of the apparatus and method. At the standard low voltage operating Conditions, the ionization efficiency curves-that is, the variation in the ion currents with the electron voltage-are very steep, and as a consequence, small changes in the ion source conditions, and particularly voltages, result in relatively large changes in the ion currents. For example, changing the ionizing voltage from 6.90 to 6.80 volts ions from 607 reduced the intensity of 2,4,4-trimethyl-l-pentene to 515 chart divisions-Le., about a 20’% change. By contrast, a t the usual high voltage conditions (50 to 70 volts), the ionization efficiency curves exhibit a broad maximum, and the ion currents are relatively insensitive t o voltage variations. I n the course of the work, it soon became apparent that the instrument sensitivity depends upon many factors, some of which are not controllable, and that rather marked variations in the sensitivity with time can be expected. The sensitivity is greatly affected by the electron voltage, and this includes not only the voltage applied betm-een the filament and the ionization Chamber, but also the ion drawout voltage, as this contributes to the energy of the electrons in the ionization chamber. These voltages must be controlled as closely as possible, but unavoidable errors in making the chosen settings will give rise to small variations in the sensitivity. Among the uncontrollable factors may be listed ion source deposits, variations in the q-ork function of the filament, and, most important, the position of the auxiliary magnet. T o illustrate the day-to-day variations in sensitivity, there are given in Table I1 representative intensities of 2,4,4-trimethgl1-pentene measured over about a 3-week period. The variations were large, particularly when the new filament was installed in the instrument, and it is clear t h a t some means of compensating for the changes must be applied if quantitative Tvork is to bc
ANALYTICAL CHEMISTRY
1250
possible. Consequently, the 2,ij4-trimethyl-1-pentene ion intensities have been used as a measure of the instrument sensitivity a t any given time, and it is assumed that the intensities of ions from other compounds will vary in proportion to the variations intensity. If this assumption in the 2,4,4-trimethyI-l-pentene be valid, multiplication of the observed intensity of an ion of interest by the ratio of the corresponding 2,4,4-trimethyl-lpentene intensity to some arbitrarily chosen constant 2j4,4trimethyl-1-pentene intensity should give a n intensity value for the ion of interest which is independent of the actual instrument sensitivity a t the time the measurement was made. The choice as a sensitivity-calibrating substance of 2,4:4-trimethyI-l-pentene was based on the fact that its ionization intensity at the standard low voltage conditions stands roughly midway between the highest and lowest intensities found with olefins. The intrnsities of ions of interest are corrected to a 2,4,4trimethyl-l-peiitene intensity of 400 chart divisions, which Tvas chopen arbitrarilv for convenience. Intensities thus corrected are referred to as being corrected to standard sensitivity. To determine whether such a one-term correction factor adequately compensates for sensitivity changes, a test mixture consisting of equal volumes of five aromatics (benzene, toluene, o-xylene, 1,2,3-trimethylbenzene, and 1,3-diethylbenzene) and four olefins (2,3-dimethyl-2-butene, 2-methyl-2-hexene, 2,4:4trimethyl-1-pentene, and 3-ethyl-boctene) %vas prepared, and its low voltage mass spectrum measured daily. The average results obtained over a typical 1-month period are listed in Table 111: irom which it can be seen t h a t the ion intensities corrected to standard sensitivity are satisfactorily constant, with an average deviation from average of about 3%. During this period the filament was replaced without materially affecting the ion intensities. Similar results have been obtained in subsequent operations.
distinguished from the molecule ions of interest. d certain amount of such fragmentation has been observed, particularly with compounds containing a group such as tert-butyl, which can form a n ion of relatively low energy. Oi greater importance is fragmentation producing molecule ions of lower molecular weight, both in the parent mass region and at lon-er masses, for in a n analysis of a mixturr Puch fragment ions could not be distinguished from the parent molecule ions of other compounds. I n the course of the work, Ion- voltage c c m s of the parent mass regions have been made for what is probabiy 5 representative number of unsaturated compounds. and no evidence of fragmentation in the parent mass region has lieen o b e r v e d . More n-orrisome is the possibility of fragmentation irivolving the decomposition of an olefin molecule ion to form the molecule ion of an olefin smaller by tn-o or three carbon niimlier>, The energies required for such reactions are not very large. and the reactions conceivably might occur to some estent . Hon-ever, experience indicates t h a t the extent must in actuality ]>e snisll and unimportant for practical purposes.
Table IV. Molecule-Ion Intensities of Saphthene Hydrocarbons at Ionizing Voltage of 6.90 \-olts Intensities expressed as chart divisions (1 chart division = 1 X 10-15 ampere ion current) Obseri-ed Compound Inten3ity
cs
Cyclopentane
C6
Cyclohexane Met hylcyclopentane
7
24 3
C7
Nethylcyclohexane 1,l-Dimethylcyolopentane trans-l,2-Dimethylcyclopentane
3: 11
Table 111. Temporal Yariations in Test Mixture .4v. Compound 2,4.4-TrimethylI-pentene C6 olefins Benzene C; olefins Toluene Cs olefins Cs benzenes Cs benzenes Cia olefins Ciu benzenes Total olefins Total aromatics
Intensity (Chart Divisions) 370 in9 107 59 1i 0 44 205 151 33 143
243 8G9
A v. Deviation (Chart Divisions) 23 4.8 6.4
CQ Deviation, 6.2 4.4 6.0
1.6
" 7
5.1 1.5
,'j.
3 3
5,3
9.6 2 ,0 3.4 3.0
6.2 1.1
4.3 9.6
19.6
70
1,1.3-Trirnethylcyclohexane n-Butylcyclopentane
cia
fert-But ylcyclohexane
1d
n
3.8 2.3
A third problem to he considered is that, while theoretical considerations indicated t h a t unsaturates could be ionized selectively and without significant amounts of fragmentation, the point should be established experimentally. The amounts of fragmentation to be expected a t a given low voltage can best be determined b y measuring the mass spectra of a large number o? unsaturated compounds. I n the initial work, the authors deemed i t excessively time-consuming to make a n extensive survey of the low voltage mass spectra of compounds and proceeded on the basis of satisfactory results of the observations of the behavior of mixtures and a relatively small number of pure compounds. As the work proceeded, i t became clear that satisfactory low voltage quantitative analyses couid be made, with the result t h a t the evtensive pure compound survey became unnecessary for practical purposes. All observations lead to the conclusion t h a t the amount of fragmentation of unsaturates occurring under standard low voltage conditions is in general not significant. Fragmentation producing radical ions-i.e., those of odd mass-is trivial, for such ions can immediatelv be
T o determine the ainonnt of naphthene ionization to be expected under the standard low voltage operating conditions, the molecule-ion intensities of a number of Cc to C l onaphthenes m-ere determined (Table IV). A certain amount of ionization of the naphthenes occurs, the intensities ranging from 2 to 43 chart divisions. If the data for naphthenes and olefins of the same carbon number are compared (see Tables I and IT), the naphthene intensities are as much as 20 to 30% of the olefin intensities for certain C;, CS, and Cl0 compounds. However, perhaps the fairest comparison can be obtained from the average naphthene and olefin intensity. The average intensity of the naphthenes listed in Table IV is 19 divisions, n-hile that of the the naphthene olefins listed in Table I is 556 divisions-i.e., ionization is about 3y0t h a t of the olefine. T o determine whether naphthene and paraffin ionization a t standard lo^ voltage conditions would constitute a problem in a practical analysis, determinations were made of the Cj to CiO mass spectra of three cuts in the paraffin-naphthene fraction of a silica gel percolation of catalytic naphtha. The sum of the intensities of all the ions formed (paraffin, naphthene, and some fragmentation) m-as in each case 32 chart divisions, but the sum of the olefin intensities in a cut taken from the middle of the olefin fraction of t,he percolation was 694 divisions. Thus, the paraffin-naphthene ionization was about 5% that of the olefins and in a sample undergoing analysis one might expect the experimentally observed olefin
1251
V O L U M E 28, NO. 8, A U G U S T 1 9 5 6 concentrations t o be erroneously high by about 5% of the paraffin-naphthene concentration, As t h e olefin analysis will be applied mainly t o samples with relatively high olefin contents, errors from this source will usually be negligible. QUANTITATIVE ANALYSES
Analysis of Synthetic Mixtures. T o gain insight into the accuracy possible with the low voltage method, a number of simple synthetic mixtures were analyzed, and, for illustrative purposes, results of two such analyses made under standard low voltage conditions are given in Tables V7 and VI. T h e samples were prepared by mixing appropriate volumes of the different compounds by means of a 1.0-ml. graduated pipet. It is suspected t h a t some of the errors shown in Table T? for 2,3-dimethyl-2butene should be attributed to an error in volume measurement, for the experimentally determined concentration of this compound is lo^ in all three olefin blends and by approximately the same relative amount. The results of the toluene determination are very good, and those for the olefin determinations are considered to he satisfactor>-, particularly as the experimental total olefin concentrations are correct to within 2% or less. The satisfactor\, results for the olefins tend to support the contention that the olefin-olefin and olefin-naphthene interferences are, at, least in the mixtures investigated, negligihle. These analyses are relatively uncomplicated, bllt even so, the results n-ere considered encouraging.
Table T
.
Quantitative -4nalyses of Toluene-n-Heptane Synthetic Binaries
Tolurnr In Sample, 100
‘>
80 60 40
2u
Table VI.
Peak Height (Chart Divisions) 2450 1940 1490 960 491
Toluene (LIS), -‘c 79 2 60.8 39 2 20 0
Deviation, % -1.0 f1.3 -2 0 0 dv. 1.1
Quantitative Analyses of Olefin Synthetic Blends
Compound Blend 1 2,3-MezC1--2 P-LleCa-;22,3-3Ie:Cs -2
Blend 2,3-.Ve2C4--2 2-SIeCs--2 2,3-MeKa--2 Iso-octane hfethylcyclohexane
% in Sample
R (119)
7, Deviation
33.3 33.3 33.3
30.4
-8.7 +3.9
16.7 16.7 16.7 25.0 25.0
3 4 , ti
34.4 Sum 9 9 . 4
+ E AT..
13.3 19.3 16.8
5,3
-8.4 +9.6 +0.6
... Sum 5 1 . 4
dr.
6.2
Analysis of Catalytic Cracked Naphthas. A promising application of low voltage mass spectrometric (LVAIS) analysis, which was selected for initial study, is t o cracked naphtha fractions for the determination of the olefin and aromatic hydrocarbon types. It was visualized t h a t such an analysis w&ld yield a quantitative molecular weight breakdown of the following compound types: olefins, cyclic olefins plus diolefins, cyclic diolefins, benzenes, indanee, indenes, and naphthalenes. Each of these compound types falls in a separate mass series-i.e., C,H,, (masses io, 84, 98, , ) for olefins; CnH?n-*(masses
68, 82, 96, . . . . ) for cyclic olefins plus diolefins, etc. Other compound types belonging t o the several mass series exist-for example, the acetylenes belong to the C,H2,,- 2 series-but it was thought t h a t the probability of finding them in the cracked naphtha would be negligible. One of the more difficult and time-consuming problems in developing the analysis was the preparation of calibration standards, particularly for the olefin types. Calibration data are readily obtained for compound types that have only one isomer in a given molecular weight range-for example, cyclopentene, benzene, toluene, and naphthalene. I n addition, sufficient information is available concerning the isomer distribution of certain compound types a t certain molecular weights to permit the preparation of synthetic blends for calibratione.g., the distribut,ion of the C j acyclic olefins in catalytically cracked materials is well known. T h e composition of the Cg fraction is also fairly well known, and data are available on the composition of C8 and Cs benzene fractions. Horvever, to obtain calibration data for the olefins above the C j range, it was decided to prepare olefin fractions (preferably of one carbon number) from a representative catalytic naphtha. The naphtha employed was obtained from the plant catalytic cracking unit when an average gas oil feed mixture was cracked under normal operating conditions. This naphtha was percolated over Davison Code 950 silica gel to prepare an olefin concentrate for subsequent distillation. T h e olefin concentrate was distilled in a 40-plate distillation column a t 30 to 1 reflux ratio and 2% cuts Tvere obtained. Low voltage mass spectra were obtained on each of these cuts in order to determine the molecular weight distribution. On the basis of this information the cuts were recombined to produce the narrowest possible carbon-nuniber spread: taking into account the following reservation. It has been shown t h a t the several isomers of a given compound type differ appreciably- in their sensitivities, and therefore it is desirable that the isomer distribution in any calibration hlend correspond as closely as possible to that which will be encountered in an actual sample. I n the catalytic naphtha there conceivably will he several different types of olefins, and a t the higher carbon numbers there will be a number of isomers of the different types. T h e boiling points of these numerous compounds Tvill in general be different, with the result that the boiling range of the olefins of a given carbon number mill be broad, and indeed oftentimes overlap the range of the compounds with one more and one less carbon atom. I n order to minimize discrimination against possibly important isomers in the upper or lover portion of the boiling range, the number of distillation cuts recombined was in some cases large enough to produce a blend containing small amounts of material with one carbon number above and below the carbon number group desired for calibration. These blends were then divided into two portions. One portion was retained for calibration of the 1o~vvoltage mass spectrometer; the second portion was hydrogenated and the product analyzed by high voltage mass spectrometry to determine the relative amounts of paraffins and naphthenes. Ultraviolet spectroscopy showed that in none of the blends was there a significant concentration of conjugated diolefins, and i t was assumed that this indicated also a low concentration of nonconjugated diolefins. Therefore, the saturates found in the hydrogenated blends were considered to come almost exclusively from acyclic mono-olefins. Similarly it could be inferred from the low voltage mass spect,ra of the distillation cuts that’ the concentrations of cyclic diolefins plus dicyclic olefins in the blend mere small, and thus it was assumed that the naphthenes found in the hydrogenated blends came from cyclic olefins. This procedure yields satisfactory calibration standards for acyclic and cyclic olefins when t,he blends can he made u p to contain only one carbon number. However, as the high-voltage analysis gives a measure of only the total concentrations of acyclic
ANALYTICAL CHEMISTRY
1252
and cyclic olefins, for blends containing more than a single carbon number, a correction has to be applied to obtain the concentrations of these compound types as a function of carbon numCompound Type ber. T o do this, any available low voltage calibation data were .Icyclic Cyclic utilized further to analyze the blend. For example, blend 1 olefins olefins Benzenes Indanes h-aphthalenes l\Iaqs contained nothing b u t C j acyclic and cyclic olefins, and these Intensitv. Chart Divisions Yo. c s o . could be determined accurately by the method described above 460 68 5 and the IOTV voltage calibration coefficients obtained. Holyever, 795 70 blend 2 was predominantly Ce acyclic and cyclic olefins but 1010 78 830 6 82 contained a small amount of CSacyclic and cyclic olefins. T h e 945 84 low voltage mass spectrum of this blend was obtained, and the I560 92 1010 91: I concentrations of Cscyclic and acyclic olefins were calculated using 675 98 the sensitivity coefficients obtained from blend 1. The concentra1860 106 tion of these C5 olefins was subtracted from the olefin concentra980 8 110 525 112 tions obtained by the high voltage analysis on the hydrogenated 118 blend, and the difference was taken to be the concentration of 2030 9 120 870 124 the Cg cyclic and acyclic olefins in the blend. 465 126 For blends containing components at three different carbon 1875 128 numbers, a n additional step was required. It was found that 132 2000 10 134 the variation in the sensitivity coefficient for a given compound 720 138 425 140 type as a function of molecular weight is regular, and con142 sequently, the technique described above for blends containing 146 components with two carbon numbers was extended t o the blends 1300a 11 148 550 152 containing components of three carbon numbers, taking as 390 154 approximate l o r voltage sensitivity coefficients for the com156 ponents of the highest carbon number values extrapolated from 160 lOOOa 12 16% the known values a t lower carbon numbers. I n this manner it 400n 166 370a 168 was possible to obtain calibration data for acyclic and cycllc Extrauolated. olefins through the C12 range. The calibration coefficients thus determined are shown in Table VII, along with olefin Table T'III. Analyses of Catalvtic Yaphthas calibration data obtained from pure compounds and synthetic Liquid Volume r/c blends. Also included in this Plant h-aphthas Pilot Unit Naphthas C. Cracker 1 C. Cracker 2 Desulfurized High .1x-. table are data for the various Light Heavy . Light Heary Paraffinica high sulfuru s u l f u r n A sweet" aromatic series of interest, ob15 e tained for the most part from 32.1 10.G 8.8 14.3 0.1 19.13 8.3 7.6 10.1 10.3 pure compounds and synthetic 1.1 8.5 6 .5 6.8 5.6 7.5 3.4 3.0 4.2 a.2 6.1 5,3 blends. 2.0 3.3 2. I 3.4 4.6 3.1 I n application, the method 0.9 3.1 1 7 2.7 3.3 1.2 3.0 1 1 " 1 2.5 0.9 is straightforward Under -.7. - 9 1.4 1 0 0.3 standard operating conditions, 0.9 1.0 1.4 1.4 1.2 1.3 a known amount of the stand3.6 1.6 3.6 3.2 3.2 3.6 ard olefin, 2,4,4-trimethyl-l4.8 0,9 1.9 4.3 4.8 4.3 5.0 2.0 1.9 4.0 4 3 3.8 3.7 0.5 3.1 pentene is charged to the in2.1 1.5 2.1 1.1 2.9 2.5 1.9 2.5 0.4 1.5 0 4 1.7 1.G 0.6 0.9 strument and its ion intensity 0.1 1.5 0.4 1.5 1.4 0.6 2.4 is determined. This is then 1.1 1.0 0.7 0.3 1.8 pumped out of the instrument, 0.5 0.e 0.5 and a charge of the sample to 2.6 1.4 2.6 3.1 4.3 3.8 be investigated is introduced 11.6 3.1 4.8 and scanned over the entire 1.o 10.4 0.7 7.0 0.2 spectral region where peaks 0.5 1.8 are expected to be found. The Indanes 1.0 0.2 0.3 standard compound is then 0.1 Cs 4.7 0.3 0.2 1.3 C!O run again. The intensities of 0.1 4.4 1.4 C11 0.2 1.0 ClZ the standard compound before Naphthalenes and after the scan are read and 1.5 1.3 ... 0.5 0.4 0.4 CiO averaged. The intensities of 0.4 0.6 CI! C!2 the ions in the sample specTotals trum are measured and ad15.8 70.G 40.2 41.5 17.0 38.9 38.8 41.7 Acy. olefins 9.1 8.3 20.6 20.3 16.2 9.3 14.4 16.1 Cyc. olefins justed to standard sensitivity. 24.9 78.9 00.8 61.8 26.3 53.3 5.5.0 57.8 Total olefins By using the calibration data 7 7 . 5 26.0 6 2 , O 61.3 59.5 29.5 5 5 . 9 56.5 F I A (olefins) shown in Table VI1 and the Totals 36.4 G.0 lG.2 17.8 11.0 33.9 11.7 16.0 Benzenes intensities corrected to stand11.1 0.6 9.4 0.5 3.4 3 7 3.2 0.3 Indanes 1.7 0 0 0.5 0.4 0.0 2.1 0.0 0 4 Naph. ard sensitivity, one can calcu47.5 G,5 20.1 21 9 19.0 47.1 12.3 11.3 Total aromatics 47.0 6 3 21.6 24.2 21.4 47.0 12.4 11.6 F I A (aromatics) late the composition of the unknown mixture simply by a Gas oil feed stock. forming ratios. Table VII. Calibration Data for Analysis of Olefin and .4romatic Types by 3Iass Spectrometer at Low Voltage
0
~
4:;
:::
V O L U M E 28, NO. 8, A U G U S T 1 9 5 6
1253
T h e method has been aoTable IX. Precision of Mass Spectrometer and Fluorescent Indicator Adsorption plied to light and heavy cataMethods for Total Olefins and Total Aromatics lytic naphthas produced in the Liquid Volume c/o plant and to catalytic naphTotal Olefins Total Aromatics thas produced in pilot unit MS FIA MS FIA ___-operations. T h e averages of Run 1 2 1 2 3 1 2 1 2 3 triplicate runs on a number of Sainple naphthas are summarized in 60.0 Light naphtha-cat cracker I 59.0 59.7 56.8 57.2 12.2 11.4 11.0 1l.i) 11.4 Heavy naphtha-cat cracker 1 3 0 . 1 28.8 27.6 26.9 24.7 46.5 47.4 45.8 49.4 46.2 Table VIII. Fluorescent indi($atoradsorption ( F I A ) analyLight naphtha-cat cracker 2 55.8 56.0 54.0 51.8 54.3 12.2 12.6 12.3 11.3 13.1 Heavy naphtha-cat cracker 2 26.3 25.7 25.6 25.6 23.7 47.0 47.0 47.0 47.647.4 RPS have also been performed Saphtha from pilot unit on each of the naphthas in Paraffinic feed 78.2 76.9 78.6 80.6 769 6.4 0 1 6.7 0.9'*16.5 Desul. high sulfur 61.3 62.7 62.6 61.9 57.5 23 0 20.3 20.3 20.6; L18.9 duplicate in order to check the High sulfur 62.3 60.3 60.7 63.1 61.5 93.8 24.7 21.9 21.0 22.4 56.0 56.1 55.3 55.0 22.1 20.7 degree of accuracy being obAverage sweet 57.1 19.7 200J&20.4 0, % tained by the mass spectrom0 63 1.39 0 66 0 74 eter analysis. These results are also shown in Table VIII. Excellent agreement b e h e e n Table X. Analysis of Catalytic Naphtha from High Sulfur the mass spectrometer and fluorescent indicator adsorption Gas Oil by Low Voltage hlass Spectrometry malyses is obtained in all cases Liquid Volume % T h e precision of the t x o methods should also be of considerable Component Low voltage h l S composition Detailed s t u d y interest, and the analytical results are presented in a somen-hat Benzene 0.3 0.4 different fashion in Table I X to illustrate this point). The 1 6 1.7 Toluene Ca benzenes 3 3 2.9 standard deviation for the fluorescent indicator adsorption total Cs benzenes 3 9 3.8 Clo benzenes 2 7 2.6 olefin results is 0.63% (16 tests), whereas t h a t for the mass C11 benzenes 2.2 2.9" spectrometer results is 1.39% (24 tests). The standard deviation Indane 0.2 0.3 for total aromatics is 0.66 % by fluorescent indicator adsorption Nethylindanes 1 1 1.4 Dimethylindanes 0 9 ... and 0.74% by mass spectrometer. T h e deviation in the mass Naphthalene ... 0.2 spectrometer total olefin results is perhaps somewhat larger than .$cyclic olefins 15.9 desired and indicates the desirability for more adequate control. C6 12.3 CS 10.2 10.5 This can most readily be achieved by normalizing the mass c7.2 c s 5 . 6 spectrometer result,s to agree with a fluorescent indicator c9 3.6 adsorption analysis on the sample. ClO 8.9 10.5 Cll T h e analyses described above were made shortly after the C12 1.1 Cyclic olefins calibration data were obtained. Therefore, in connection with 1.3 CS 0.5 the problem of the long-term applicability of the calibration 3.9 2.5 c 6 c. 4.4 4.7 data, another catalytic naphtha determination is considered, 3.6 C% 4.3 2 . 3 C B made approximately 8 months after the calibration datu were ClO 7.2 obtained. I n Table X it is compared directly to results Cll C12 0.4, obtained on this same naphtha by a method utilizing the best a Includes diniethylindanes. techniques which were available prior to the development of the low volt,age mass spectrometer method. Included in the anal>-tiTable XI. High Voltage M a s s Spectrometer SensiLiF ities ea! scheme Kere distillation, percolation of numerous distillation of Cs Benzenes fractions, recombination of concentrates, hydrogenation, and Peak IIeight a _ t 3Iasn __ further distillation with subsequent mass spectrometer and Compound B.P., F. 120 106 02 78 infrared analyses on segregated fractions. The results of this Isopropylbenzene 4% 306 147 9 extensive analysis are believed to be accurate and the agreement n-Propylbenzene 464 319 233 7 m-Ethyltoluene 322 524 74 149 between the low voltage mass spectrometer method and the p-Ethyltoluene 324 eo0 144 41 a-Ethyltoluene 329 028 153 50 analysis obtained by the more complex procedure is considered 1.3,5-Trimethylbenzene 329 112 844 32 to be satisfactory, especially in view of the relative amounts of 1,2,4-Trimethylhenzene 757 117 337 24 1,2,3-Trimethylhenzene 720 :38 319 31 time required by the t'wo methods. The rather detailed analysis Weighted average5 661 117 45 61 obtained from the lon- voltage mass spectrometer requires only a According t o infrared analysis of Cs aromatics concentrate. approximately 1.5 hours, about 40 minutes of which is actual instrument scanning time, which could be appreciably reduced with improved instrumentation. At least 500 man-hours would be required to duplicate the more complicated analysis; although the summary results shown in Table X conceal a considerable naphtha or hydroformate. This was essential because the mass spectrometer sensitivities of the individual isomers in a given amount of detailed information actually obtained. molecular weight range vary widely; weighted average sensitivAnalysis of Virgin Naphthas and Hydroformates. Another ities are obtained for actual analyses by preparing synthetic application of the low voltage method is to the determination of Cs, Cg, and C ~ benzene O mixtures corresponding to the known the aromatic molecular weight distribution in hydroformer feeds ratio of occurrence of the isomers. T h e sensitivities of the and products in the higher boiling ranges (about 300" F. and higher), and particularly in distillation cuts thereof. This individual CS benzene isomers and the weighted average are analysis has been made by high voltage mass spectrometry; given in Table X I by way of illustration. From the boiling however, this method has serious limitations, which are easily points shown in this table it is obvious that the average sensitivities as applied to a narrow range distillation fraction would overcome by the low voltage technique. T h e determination of the molecular weight distribution of aromatics by high voltage give results seriously in error, not only for the Cs benzene content, mass spectrometry was originally designed to be performed on but also for Cg, C,, and CS. To emphasize the point, the composition of a 315" to 321' F. fraction from the distillation of an broad cut distillation fractions such as a 200" to 330" F. virgin ~~
3:;
y::)
ANALYTICAL CHEMISTRY
1254
from plant test run hydroformates produced a t different severities. Analyses of fonr Sample 1 2 3 4 distillation cuts from these hyBoiling Range, F. d r o f o r m a t e s a r e shown in 367-373 352-357 362-368 342-363 Table XII. The first column Liquid Volume % by under each sample shows the LV1 LV2 HV LVI LV2 HV LV1 LV2 HV HV LV1 LV2 analysis obtained from conBenzene 0.6 _ . . ... 0.7 ,.. , . . 0.8 ... ... 0.5 ... ... Toluene 2.5 ... ... 3.9 ,.. ,.. 4.2 0 . 3 0.3 1.6 ... ... ventional high voltage mass CSb e n z e n e 3.1 0.4 0.2 6.6 ,.. ,.. 7.5 0.3 0.3 4.1 , . . ... s p e c t r o s c o p y . The second Co benzenes 21.5 1 9 . 8 16.6 13.2 10.7 9.0 1.3 1.4 1.2 0.4 0.8 0.7 CIObenzenes 27.9 22.7 33.5 36.4 3 2 . 7 4 4 . 4 6 2 . 4 6 8 , l 7 2 . 5 74;3 66.3 71.2 column gives the results ob'2 0.0 0.0 a 0.0 0.0 0.5 0.3 2.5 3.6 CIIbenzenes Indanes 5 4.4 4.4 5 0 5.0 4.5 4.5 5 7.7 7.7 tained from low voltage mass Iiaplitlialenes a 0.0 0.0 a 0.0 0.0 5 0.0 0.0 a 0.0 0.0 spectroscopy, where the broadTotal aroniatics 55,Gb 4 6 . 9 5 4 . 7 6 0 . 8 6 4 8 . 4 5 8 . 4 7 6 . 2 6 05,l 7 9 . 1 8 1 . 6 b 7 7 . 3 8 3 . 3 cut weighted average sensitiviTocal olefins 3.06 3.9 3.9 3.96 3.7 .1.7 3.26 2.2 2.2 3.26 2.6 2 0 ties for each molecular weight Total aromatics (FI.1) 5.5.6 60.8 76.2 81.6 Total olefins (FI.4) 3 .0 3.9 3,2 3.2 group are employed. The HV. High voltage 31s. final column shows the results L Y I . Lon. voltage MS. ai-erage sensitivities, obtained from low voltage LT-2. L o a voltage 31s. specific sensitivities for boiling range (see Figure 1). mass spectroscopy when the a S o t determined. 5 Sormalized to I'Id results. effect of boiling range of the sample is taken into consideration. The high voltage analysis did not take into consideration the presence of C11 benzenes and the appreciable concentraaromatic solvent (as determined by infrared) is compared to the tions of indanes. Hoxever, it is not likely that even had they compo>ition which would be indicated by high voltage mass been considered any significant improvement in the erroneoua spec ti ome try. toluene and Csbenzene contents would have been effected. The Actual ( I R ) Calculated (HVIIS) marked improvement in the agreement between the total aroBenzene 0 3 Toluene 0 10 matics by low voltage m a s spectrometry and fluorescent C3 benzenes 0 -4 indicator adsorption when the boiling range effect is taken into C Sbenzenesa 100 74 consideration demonstrates the significance of this refinement. 01$ ,I-propylbenzene. 36$ m-eth>ltoluene Analysis of Propylene Polymers. The low voltage technique has also been applied to the determination of the molecular Two classes of errors are encountered in this caqe: the error weight distribution of olefins in polymers produced from the due to using an average parent ion sensitivity and that due to catalytic polymerization of propylene. As might be anticipated, using average interference coefficients. By carrying out the once again it was necessary to produce concentrates for calibraanalysis a t lon ionizing voltages the interference coefficient error tion of the mass spectronieter for each of the molecular weight is eliminated, because no fragment masses are formed One is groups of olefins. Concentrates of CS, CS, and Ci2 olefins are faced then only with the problem introduced by variations in readily obtained from the distillation of a prop! lene polymer parent ion sensitivities. -4lthough the sensitivity variation is The sensitivities of C7, Cg, Clo,and C11 olefins were then obtained actually greater i n the case of Ion voltage analysis, the problem is by interpolation. To check the validity of the assumption of niore readily solved because of the absence of interferences between groups. The problem is further simplified by the fortunate fact that the sensitivity variation is a relatively smooth function of boiling range, the lowest boiling isomers having the lowest sensitivities and vice versa, as indicated in the following table. Table XII.
Analysis of Hydroformate Distillation Fractions by Mass Spectrometer
0
0
(1
Compound Isoprop ylbenzene n-Proovlbenzene ni-EtLj.itoluene p7-Ethyltoluene a-Ethyltoluene 1 3.5-Trimethylbenzene 1:2,4-Trimethylbenzene 1, ~ . 3 - T r i m e t l i ~ l b e n z e n e
B.P.. F. 306
Observed Intensity 1300
319 322 324 3 29 329 337 349
1835 1870 1725 2510 2400 2300
1110
By employing these data in conjunction with infrared analyses of distillation fractions from a CS aromatics concentrate, it was possible to calculate the points shown in the left-hand curve of Figure 1. Similarly, with pure compound data for individual Cl0 benzenes and infrared analyses of distillation cuts from a C,O aromatics concentrate, the points for the center curve were calculated. Data on individual CI1 benzene isomers are not available; however, in view of the results for CSand CIo isomers the right-hand curve in Figure 1 is taken as a remonable approximation. T h e loxest snd highest boiling C,, isomers are the amylbenzenes and pentamethylbenzene, respectively, and these compounds are available for calibration purposes. One could also expect to prepare a reasonable plot for C I Jbenzenes from data on hevylbenzene and hexamethylbenzene. This technique has been employed in the analysis of distillation cuts from a 300" to 350" F plant hydroformer feed stock and
750
1 300
I
1
325
350
I 375
MID-BOILING POINT,
!OO F.
425
450
Figure 1. Low- voltage mass spectrometer intensities of CS, Clo, and C11 benzenes
1255
V O L U M E 28, NO. 8, A U G U S T 1 9 5 6 linearity between sensitivity and molecular weight the sensitivities derived on the basis of the assumption were applied t o cuts from the distillation falling among the CS,CS, and C12 concentrates. The analyses of these various cuts gave results totaling between 90 and 110% olefin, thus indicating not greater than a 10% error in the sensitivities. Actually, because the distillation would be expected to effect some separation of the various isomers, and average coefficients are applied, these sensitivities when applied to a propylene polymer sample of m-ide boiling range should give good results. Application to High Boiling Materials. It v a s expected t h a t the low voltage technique would have a particularly useful appliration t o the quantitative determination of the compound types present in fractions obtained from the percolation of heating oil :ind gae oil fractions over alumina gel or similar absorbent. Preliminary investigations of this possibility have been successful,
and i t is anticipated t h a t this techniquc will prove useful iii the eventual unraveling of the composition of high boiling aromatics. ACKYOWLEDGMENT
The authors gratefully acknowledge the contributions oi Burl
L. Clark, who performed the greater portion of the e.;pciiniental work. Joe Dzilsky joined the project in its later stages .ind his effortg, too, are greatly appreciated. LITERATURE CITED
(1) Field. F. H., Franklin, J. L., ,J. Chem. Phya. 22, IS05 \ l ! ) S ) .
( 2 ) Franklin, J. L.. Ihid.. 22, 1304 (19.54). (3) Honig, R. E., Ibid., 16, 105 (1948).
(4) Stevenson, D. P., TVagner, C. D., J . A m . Chem. Soc. 72, 5tj1'7 (1950). ( 5 ) Taylor, D. D., U. S. Patent 2,373,151(1900). RECEIVED for review February 10,1956.
Accepted M a l - 3. 1956
Quantitative Infrared Absorption Spectroscopy in Water Solution W. J. POlTS, JR., Dow
and NORMAN WRIGHT
Chemical Co., Midland, Mich.
Quantitative infrared absorption spectroscopy can be carried out in water solution by using a very thin absorption cell with barium fluoride windows. Useful transmittance in the region from 6.5 to 10 microns is obtained on a double-beam spectrometer by insertion of a transmittance screen in the referencebeam; energy is recovered by widening the spectrometer slits by a suitable amount. The method is applicable to many cases where organic materials are soluble in water.
T
HAT water can be used as a solvent for infrared absorption spectroscopy was shown as early as 1905 by Coblentz ( 3 ) . Nore recently Gore, Barnes, and Petersen ( 3 ) and Blout and Lenormant (1) have shown that water, used in conjunction with heavy mater, can have considerable use in this respect, enabling one to obtain an infrared absorption spectrum throughout almost the entire rock salt region. Their results were of a qualitative nature only, however. Plyler and Acquista ( 4 ) have given quantitative absorption spectra of pure water, and have shown that there is a region from ~ 6 . to 5 =10 microns where there is still enough infrared transmittance in reasonable path lengths of water t o suggest its use as a solvent for quantitative analytical purposes. The advent of barium fluoride as an optical material has made possible the construction of a permanent absorption cell. Barium fluoride seems ideally suited for this use, as it is commercially :tvailable, hard, easily polished, and essentially insoluble in uater. With such a cell, a a t e r solutions can be used in much the same way, and m-ith the same accuracy, as carbon tetrachloride or carbon disulfide solutions are used for quantitative absorption spectroscopy a t present.
an amalgam with the brass, which sticks to the fluoridc plate surface. The barium fluoride plates were obtained already cut, ground, and polished from the Perkin-Elmer Corp., Sorwalk, Conn. The cell so constructed has a path length of 0.027 inin.: this distance was determined in the usual way by a fiinge pattern ( s h o m in Figure 2, a ) of the empty cell. The depth of the fringes and their general regularity indicate that, even w t h this short path length, a cell can be made with nearly perfectly pal allel faces if care is used. ,411 spectra were obtained on a double-beam infr arcd spectrometer equip ed with a rock salt prism. The instrument was designed a n z b u i l t in this laboratory; a publication desciibing its construction and features is in preparation. Figure 2, b, is the absor tion spectrum of pure n-ater obtained in the cell just described. 8omparison with Figure 2, n, shoa s t h a t water ceases to transmit a useful amount of radiation at somewhat shorter wave length than the barium fluoride rntoff point; hence, barium fluoride is by no means the limiting factor in the use of water solutions.
APPARATUS AND TECHNIQUES
T h e absorption cell (Figure 1) is constructed in much the same manner as the conventional rock salt cells. The spacer between the barium fluoride plates is made from 0.001-inch shim brass, and is sealed to the plates by coating the brass n i t h mercury to form
Figure 1. Barium fluoride absorption cell for use w-ith water solutions
