Anal. Chem. 1982, 5 4 , 967-971
967
Determination of Field Ionization Relative Ion Sensitivities in Mass Spectrometry of Coal-Derived Oil Components Tadashl Yoshida and Yosuke Maekawa Government Industrial Development Laboratory, Hokkaido 2- 17 Tsukisamu-Higashi, Toyohira, Sapporo 06 1-0 1, Japan
Toshiaki Shlmada Faculty of Engineering, Hokkaido University, Sapporo 060, Japan
The determination of fleid ionlzation (FI) relative ion sensitlvity was carried out an components separated from coaiderived oil by liquid chromatography and gel permeation chromatography. The effects of molecular welght and structure of component on the relative ion sensltivity were studied, and the relationship between molecular weight and relative ion sensitivity was obtalned for various z series. Relative ion sensltivttlas ( f l ) of components in various z series are exponentially proportional to their molecular weights (MI),that is, the composite length of alkyl side chalns. The slope of the in I, vs. Ml plot for each x series Increases with the increase in the size of the aromatic ring and decreases with the decrease in the z value. The molecular weight dependence is much smaller in saturated hydrocarbons than in aromatic ones.
A detailed description of coal-derived oil is essential for efficient chemical utilization and development of economical coal liquefaction processes. However, coal-derived oil is a multicomponent mixture consisting mainly of aromatic and hydroaromatic hydrocarbons and an analysis of its composition is highly complex. Coal-derived oil has generally been described by molecular weight distribution and average molecular weight. Field ionization (FI) mass spectrometry introduced by Beckey (1)gives only molecular ions, free from fragment ions, for hydrocarbons (2,3)4mdtherefore should be ideally suited not only to the determinations of the average value and distribution of molecular weights (4)but also to a compositional analysis of coal-derivecl oil. In a previous paper (5),we investigated reproducibility and quantitative aspects of FI mass spectra from coal-derived oil. The accuracy of spectral data in the quantitative FI mass analysis of a mixture is in part governed by the availability of data for relative ion sensitivities. Relative ion sensitivities depend on molecular structure and weight and have already been reported for alkanes, cycloalkanes, and low molecular weight aromatic hydrocarbons (6, 7). However, coal-derived oil components have more complicated molecular structures and higher molecular weights than the model compounds, and so the available relative ion sensitivity data cannot be used. It is therefore necessary to determine the relative ion sensitivities of the components in coal-derived oil. In this paper, determination of FI relative ion sensitivity was carried out on components separated from coal-derived oil by liquid Chromatography (LC) and gel permeation chromatography (GPC), and the distribution of the components in the mixture was obtained by correcting the peak intensities. Furthermore, the effects of molecular weight and structure of component on the relative ion sensitivity were investigated. The determination of sensitivities of heteroatom-containing 0003-2700/62/0354-0967$01.25/0
aromatic components in oil is not included in this paper. EXPERIMENTAL SECTION Preparation and Fractionation of Coal-Derived Oil. Soya-koishi brown coal (ultimate analysis: C, 72.0%; H, 5.5%; 0, 21.1%) produced at Hokkaido in Japan was used. The coal was hydrogenated over a red-mud catalyst at 450 "C for 60 min under an initial hydrogen pressure of 9.81 MPa. The hydrogenated products were extracted with hexane in a Soxhlet extractor. The extracted oil was separated into five ring-type fractions on a liquid chromatograph by the modified Bureau of Mines API 60 method (8): saturates (Fr-P), monocyclic aromatics (Fr-M), bicyclic aromatics (Fr-D), tri- and tetracyclic aromatics (Fr-T),and polycyclic aromatics and polar aromatic compounds (Fr-PP). This modified method is the same as the Bureau of Mines API 60 method (9) except for the separation of F r T fraction and the use of cyclohexane instead of n-pentane. The Fr-P, -M, and -D fractions were further separated into 24-48 fractions on a GPC column (GlOOOHG), 25 mm X 600 min, Toyo-soda kogyo Ltd). About 300-400 mg of the ring-type fractions were injected into the column, and chloroform was used as the mobile phase with a flow rate of 3.5 mL/min. The exact weights of the GPC fractions separated were determined after completely vaporizing the solvent. The recovery of injected sample was approximately 92 wt % . Measurements of FI Mass Spectra. All FI mass spectra were measured by a JMS-D300 double-focusing mass spectrometer equipped with an electron impact (EI)/FI/field desorption (FD) combination ion source (JEOL Ltd). Ions were detected by an electrical detector connected to a data-analysis system, JMA-2ooO. The emitter current was kept at 10 mA to prevent the sample from condensing on the emitter. The scanning speed for the mass range of m / z 100-550 was 5 s/cycle, and scanning was repeated 150-200 times. The chamber temperature was kept at 100 "C. The ring-type and GPC fractions were subjected to FI mass analysis by the direct-inlet method: they were charged into a glass tube with active alumina powder to avoid rapid evaporation of volatile components,and the measurements were performed while the direct-inlet probe was slowly heated to 360 OC. RESULTS AND DISCUSSION The relative ion sensitivity of a coal-derived oil component can be calculated from the ratio of the actual weight of the isolated component and the apparent weight of the component estimated from the peak intensity in the FI mass spectrum of the mixture. However, as isolation of components in coal-derived oil is practically impossible, LC and GPC techniques were used to fractionate the mixture into aromatic ring-type fractions and subsequently molecular size fractions. The purposes of the fractionation are (1)to divide coal-derived oil into as fine fractions as possible in order to determine the actual weights of the components, (2) to avoid overlap of components with the same mass but different molecu1,ar structures, (3) to fractionate coal-derived oil containing various types of molecular structures and a wide range of molecu1,w weights into homologous series and groups with similar molecular sizes to minimize the differences in ion sensitivity, as 0 1962 American Chemlcal Soclety
ANALYTICAL CHEMISTRY, VOL. 54, NO. 6, MAY 1982
968 1000-
L
" f z
iy
0
100
160
200
250
300 MI2
360
400
450
1
150
200
/L
0100
150
200
0 00
%
, 100
500
Fr-M
1000-
- 4 50
,
250
300
350
400
450
260
300
360
400
450
-&QO
1
5 r 500 0 00
M /Z
Flgure 1. F I mass spectra of ring-type Fr-P, -M, and -D fractions.
ion sensitivity in low-voltage E1 and FI mass spectrometry greatly depends on the molecular weight and structure of a component. The finer the fractionation of the coal-derived oil is, the higher the accuracy of relative ion sensitivities obtained for the components will be. Figure 1 shows the FI mass spectra of ring-type Fr-P, -M, and -D fractions separated from coal-derived oil by liquid chromatography. The molecular weight distribution covers the mass range of m / z 150-400 for the Fr-P fraction, and m/z 130-350 for the Fr-M and -D fractions. The dominant components of the Fr-M and -D fractions are distributed in the low-mass range, m / z 150-250. In the present investigation the relative ion sensitivities of components in these mass ranges were determined. The apparent weights (m,(app))of components in each ring-type fraction are calculated from the spectral data in Figure 1 by eq 1, where W is the weight of the ring-type
mi(app)= W
X
MiIi/CMiIi
(1)
fraction, M, is the mass of the ith peak, and Ii is the intensity of the ith peak. The peak intensity of the isotopic ion (M,+J was added to its parent ion (Mi). The molecular weight of hydrocarbon is expressed by the following equation: M = 12 X C, HP,+z,and the components in coal-derived oil can be assigned to several structural types by the use of the z value. From this relative ion sensitivities of the components can be summarized for various z series. Figure 2 shows the GPC curves of ring-type Fr-P, -M, and -D fractions. As the elution curves of ring-type fractions are not proportional to the RI curves, due to the differences in the refractive index of the eluted component, the weights of GPC fractions were determined. The elution curves of ring-type fractions are characterized by the existence of two maxima, as shown clearly in the curve for the Fr-M fraction. The chainlike molecules, such as paraffins and aromatic rings with long alkyl side chains, are eluted with smaller elution volumes, while the ringlike molecules, such as cycloparaffins and aromatic hydrocarbons, are eluted with larger elution volumes. This is also confirmed by the results in Figure 3 where Mn curves shift to larger elution volume in the order Fr-P, -M, and -D. The fact that the Fr-D fraction is eluted with a larger elution volume than Fr-M, in spite of the analogous molecular weight distributions, indicates that the lengths of alkyl side chains in Fr-D are shorter than those in
+
Fraction number
Flgure 2. Elution curves of ring-type Fr-P, -M, and -D fractions.
,on I 0
I L L L 4
8
U.L,A-.L.12 16 20
24
Fraction number
Flgure 3. Number average molecular weights of GPC fractions separated from ring-type Fr-P, -M, and -D fractions.
Fr-M. The inflection point in the Mn curve for the Fr-P fraction is considered to be caused by the structural change of the eluted component, from paraffinic to cycloparaffinic hydrocarbons with increasing fraction number. Thus, it can be assumed that the separation of ring-type fractions is effected mainly depending on the length of the alkyl side chains. However, as can be seen from the FI mass spectra of the GPC fractions of Fr-D in Figure 4, the separation of components is better in the lower molecular weight region. Some components with higher molecular weights are eluted through several fractions. This molecular sieve effect may be improved by connecting the GPC column for higher molecular weights in series. The actual weights (m,(ad)) of components in GPC fractions are calculated from their spectral data by eq 2, where w is the
weight of the GPC fraction. The weight of a component which has been eluted through several fractions was determined by a summation of the amounts present in each GPC fraction. It should however be noted that the FI mass spectra of GPC fractions do not represent their actual compositions, due to
ANALYTICAL CHEMISTRY, VOL. 54, NO. 6 , MAY 1982
969
-
Table I. Determination of Relative Ion Sensitivities of Components Assigned t o Various z Series in Ring-Type Fractionsa Fr-P z
series +2 0 -2
a
Fr-D
Fr-M
________
_
eq
z series
In y = (2.6 x 1 0 - 3-~0.807 l n y = (1.7 x 10-3)3c- 0.612 In y =: (3.1 X - 1.091
-6
eq lny lny In y lny In y
-8
-10 -12 -14
I
_
z series
= (9.3 X 10-3)x- 1.451 = (10.5 x l o w 3 & - 2.005 = (9.9 X lO-)x - 2.031
= (7.8 x 10-3)x- 1.596 = (5.5 x 10-3)x- 1.135
-12
-14 -16 -18 -20
-22
eq
In y lny In y In y In y lny
= (20.9 X W 3 ) X - 3.998 = (15.8 X = (12.5 X
10T3)X- 3.456 10-3)x- 2.970 = (10.8 x 10-3)x- 2.598 = (7.6 x 10-3)x- 1.799 = (5.4 x 10-3)x- 1.080
y is the relative ion sensitivity and x is the molecular weight.
-
I_____
E
LJl..l---
L
0-
100
200
150
260
O 00 300
350
300
350
M/Z
100
150
200
250 M/Z
2 5
Fr-D-34
1000-
[r
0 5.--L--150
,
~200
250
300
I
i
d
350
M /Z
I"
'
I
' c L . A , ' , 100
150
200
M /Z
250
300
350
0 00
I
150
200
250
-_Lb
300
350
M /Z
Figure 5. Correlation of relative ion sensitivity with molecular weight for each z series.
Flgure 4. Representative F I mass spectra of GPC fractions separated from ring-type Fr-D fraction.
differences in the ion sensitivities of components. Considering that the components in GPC fractions are distributed in a relatively small mass range and are structurally similar, the ion sensitivities of components in the GPC fractions in this study were assumed to be equal to one another, in order to estimate the actual weights. The relative ion sensitivity (ti)of a component can be calculated by eq 3. Figure 5 shows the semilogarithmical plots
fz
=:
mr(app)/mi(act)
(3)
of relative ion sensitivities of components calculated by the above equation against their molecular weights for each z series. The relative ion sensitivity increased exponentially with molecular weight, and a linear relationship was obtained in the In f, vs. Mi plot for every z series. The slope was the smallest for the ring-type Fr-P fraction, and increased with
increases in the size of the aromatic ring. I t decreased with a decrease in the z value in the ring-type Fr-M and -D fructions. From the slope for the z series, the relationship between the relative ion sensitivity and the molecular weight of a component is derived and it is summarized in Table I. From these equations, we can estimate the relative ion sensitivity of any component assigned to a z series. Figure 6 shows the correlation of the slopes of the plots with the z values. In the ring-type Fr-P fraction, the slopes for cycloparaffins ( z = 0, -2) increase slightly with the decrease in z values. The slopes for the Fr-M and -D fractions decreased significantly with the decrease in the z values, and the slopes for Fr-D were higher than those for Fr-M. As the molecular structure of the aromatic ring system of components assigned to a z series for a ring-type fraction is basically identical, the molecular weight differences in a z series are caused by differences in the length of alkyl side chains. The results in Figure 5 and 6 lead to the following conclusions: (1) the relative ion sensitivities of components in a z series me exponentially proportional to their molecular weights (the composite length of alkyl side chains), (2) the effect of alkyl side chain on relative ion sensitivity increases with the increase in the size of the aromatic ring, and this effect is much smaller for saturated hydrocarbons than for aromatic ones, (3) the effect of alkyl side chains on the relative ion sensitivity decreases with the decrease in z value, as the aromatic ring
970
ANALYTICAL CHEMISTRY, VOL. 54, NO. 6, MAY 1982
x10-3
g
20
15
Table 111. Apparent and Corrected Mole Percentages for Various z Series in FI Mass Spectra for Fr-M and D
I
d
mol %
i\
I
51
O’
i
0 -2
-;- 6 -A
-1’0-12
-14-16
-1’8-60-22-214
Z serles
Flgure 6. Correlation of slopes of relative ion sensitivity plots with z series. Table 11. Apparent and Corrected Mole Percentages for Various Mass Ranges in FI Mass Spectra for Fr-M and D mass range 100-149 150-199 200-249 250-299 300-349 350-399
Mn
Mn
100-149 150-199 200-249 250-299 300-349 350-399
mol % corrected apparent Fr-M 9.0 42.1 30.1 14.5 3.7 0.6 206 Fr-D 4.6 36.3 28.1 19.8 9.4 1.9 224
z series
apparent
corrected
-6 -8 -10 -12 -14
Fr-M a 10.3 32.5 26.4 19.4 11.4
7.1 37.6 29.0 17.1 9.3
Fr-D -12 23.8 30.1 -14 30.9 35.1 -16 19.1 17.2 -18 10.3 7.0 -20 8.7 6.2 -22 7.3 4.5 a Excludes peaks assigned to z numbers beyond Z = -16. Excludes peaks assigned to z numbers beyond 2 = -24.
ri-
R
-
I O 0-
-
-
15.0 52.0 24.8 6.9 1.1 0.2 3.89
I
50
10.2 51.7 24.2 10.6 3.1 0.2 198
system approaches hydroaromatic structures. In the case of the Fr-P fraction, the slopes increase with the decrease in the z value as the molecular structure changes from cycloparaffinic to hydroaromatic. In summary, it becomes clear that the effect of molecular weight (alkyl side chain) on relative ion sensitivity is small in paraffinic and hydroaromatic molecular structures and increases with the size of the aromatic ring. Therefore, the actual molecular weight distribution profiles of the ringtype fractions, compared with those of the spectra in Figure 1, are unchanged for the Fr-P fraction and shift to a lower mass range for the Fr-M and -D fractions. Table I1 summarizes mole percentages for components, calculated from the apparent and corrected peak intensities in the mass spectra of the ring-type Fr-M and -D fractions. The correction increases the amounts of lower-mass componenta, below m/z 200,1.3to 1.5 times. The number average of molecular weights for the Fr-M and -D fractions also decreases and the actual molecular weight distribution generally shifts to the lower mass ranges. Similarly Table I11 summarizes the mole percentages for the z series in the Fr-M and -D fractions. The amounts of the components assigned to z = -8 (1,2,3,4-tetrahydronaphthalene type) and z = -10 (1,2,3,4,4a,g,9a,lO-octahydroanthracene type) for the Fr-M fraction and z = -12 (naphthalene type) and z = -14 (1,2,3,4-tetrahydrophenanthrene and biphenyl types) for the Fr-D fraction increase, and the structural distribution generally shifts to smaller ring struc-
I
100
I
l
l
,
l
,
l
150
l
200
’
,
l
l
l
250
rn/z
Flgure 7. Comparison of our data with published data in FI and E1 (electron impact ionization) relative ion sensitivities of aromatic and hydroaromatic model hydrocarbons: (A)author’s data, ( 0 )published E1 data, (0)published FI data. tures by the correction for relative ion sensitivities. Compared with the apparent distribution in the mass spectrum, the correction indicates that the coal-derived oil has an abundance of components with smaller molecular structures and weights. In Figure 7 , the FI relative ion sensitivity data and the EI/FI published data (7) obtained for model hydrocarbons are plotted against the molecular weights. As the low mass model hydrocarbons in Figure 7 were not found in our samples, we estimated the relative ion sensitivities of these model hydrocarbons by using the equations in Table I. The resulting relative ion sensitivities were recalculated to obtain values relative to published data for benzene and naphthalene. As shown in Figure 7, the E1 relative ion sensitivity is highly dependent on both molecular weight and structure, but this dependence in the FI relative ion sensitivity is very small while showing the same tendency as the E1 data. Similar tendencies were observed in our results where the molecular weight dependence was greater than that in the reference data. The relative ion sensitivity data obtained here is based on actual components separated from coal-derived oil, and not on lowmass model compounds, and consequently our results would be more useful for a quantitative analysis of coal-derived oil.
LITERATURE CITED (1) Beckey. H. D. Int. J . Mass Spectrom. Ion Phys. 1989, 2 , 500. (2) Beckey, H. D. “Field Ionization Mass Spectrometry”; Pergamon Press: Oxford, England. 197 1. (3) Beckey, H. D. “Principlesof Field Ionization and Field Desorption Mass Spectrometry”; Pergamon Press: Oxford, England, 1977. (4) Anbar, M.; St. John, G. A. Fuel 1978, 57, 105.
Anal. Chem. 1982, 5 4 , 971-974 (5) Yoshlda, T.; Maekawa, Y.; Hlguchl, T.; Kubota, E.; Itagaki, Y.; Yokoyama, S. Bull. Chem. SOC.Jpn. 1881, 54, 1171. (6) Mead, W. L. Anal. Chetm. 1968, 4 0 , 743. (7) . . SCheDDele, S. E.; Grizzle, P. L.; Greenwood, G. J.; Marriott, T. D.; Perrelra,'N. 0. Anal. Chem. 1976, 48, 2105. (8) Yokoyama, S.; Tsuzukl, N.; Kato, T.; Sanada, Y. J. Fuel SOC.Jpn. 1978, 57, 74a.
971
(9) Hlrsch, D. E.; Hopklns, R. L.; Coleman, H. J.; Cotton, F. 0.;Thompson, C. J. Anal. Chem. 1972, 4 4 , 915.
for review November 2, lggl. Accepted January 25, 1982.
Determination of Sulfur-Containing Species in Solids by Molecular Emission Cavity Analysis Jau-Hwan Tzeng and Qulntus Fernando* Department of Chernlstty, Unlversify of Arizona, Tucson, Arizona 8572 1
Two types of flames, the nltrogen-cooled and the argoncooled hydrogen flame, have been used for the determination of sulfur-contalnlng specles In solids by molecular emlsslon cavity analysis (MECA). The argon-cooled flame has a much greater sensltlvlty than the nltrogen-cooled flame for the determination of Sod2-. I n a solid mlxture containing Sa, s*-, SOa2-, and SO,2", the presence of one or more of these sulfur-contalnlng species can be determlned with the ald of the argon-cooled flame. The nltrogen-cooled flame Is useful in speclal cases, for example, In the determination of the comwhich are present In a ponents of a mixture of Sa and SO-: solld matrix. All these sulfur-contalnlng species can be quantltatlvely deterrnlned In the argon-cooled flame In the concentration range from about 10 to 5000 ppm. The varlatlon from 10% to 20% In the reproduclbillty of these measurements has been attrlbuted to the nonhomogeneltyof the solid materials and the mall sample sizes that had to be used In these determlnatlonr.
A method based on MECA (1) was proposed for the determination of s", Sg, SO3%,and SO?- in solids (2,3). In this method, a solid inample containing one or more of these sulfur-containing species was weighed in a small aluminum cup which wasi placed in a cluartz-lined cavity that was fitted on the end of' a stainless steel rod. A solution containing phosphoric acid and a wetting agent was added to the solid sample and the end of the steel rod was positioned in a relatively cool hydrogen--nitrogen flame. The emission from the molecular sulfur that was produced in the cavity was recorded a t 384 nm as a function of time. The peaks in the emission spectrum were identified with each of the sulfur-containing species in the solid sample, and the peak areas or peak heights were used in conjunction with calibration curve to determine the concentration of each of the sulfur species. In this method, it is evident that there are several variables that must be carefully controlled to obtain the maximum sensitivity and an acceptable reproducibility. The sensitivity of the method is governed by the intensity of the flame emission, which depends on the population of the molecular sulfur species, Sz, in the excited state, and their residence time in the flame. Allhough the mechanism of formation of molecular sulfur from the various sulfur-containing species is not fully understood, i t has been established that several characteristics of the flame, in particular its temperature, affect the emission intensity. The excited state Sz molecules that
emit are almost exclusively produced by radical interactions rather than by thermal excitation. If monatomic argon gas is substituted for the diatomic nitrogen gas in the MECA flame, the number of radicals produced by collisional processes in the flame should increase. As a consequence, the number of Sz molecules in the excited state should also increase. We have observed that, under comparable conditions, the emission intensity of Sz molecules produced in a hydrogen-argon h e is greater than that in a hydrogen-nitrogen flame. In this work we have compared the use of these two types of flames in the determination of sulfur-containing compounds by MECA. Additional variables that affect the emission intensity are the concentrations of the components of the wetting solution (H3P04and Triton X-100) and the surface composition of the material that was used in the fabrication of the sample cup. In all our previous work that was reported ( 2 , 3 ) ,small aluminum foil cups, 3 mm in diameter and 2 mm deep, were used, and no special attention was paid to the nature of the aluminum surface. In subsequent experiments, however, we have observed that the emission intensity produced in the cavity is dependent on the previous history of the aluminum foil surface. In the work that is reported below, we have adopted an empirical approach in our attempts to maximize the signa1:noise ratio and to separate the emission peaks that are obtained from solid samples containing sulfur in several oxidation states. The effects of all the variables mentioned above have been studied, and an optimum set of operating conditions has been found for the identification and determination of sulfur-containing molecules and anions in a solid matrix. Attempts were made to use this MECA technique for the determination of sulfur-containing species in coal. EXPERIMENTAL SECTION Equipment. A modified flame emission system which was described in detail in an earlier publication (2) was used for all the emission measurement. The burner assembly included an inlet for premixed hydrogen, the coolant gas (Nzor Ar), and air. The modular flame emission system consisted of a scanning monochromator coupled to a photomultiplier detector with the output displayed on a fast response strip chart recorder. The MECA sample introduction device was the same as that used in all previous experiments (2, 3). Reagents. Solid standards containing one or more sulfur species were prepared from Analytical Reagent grade chemicals: elemental sulfur as sublimed sulfur (Ss); sulfite as NazSO,; sulfate as Na$304;sulfide as CuS (Ultrapure 99.998%)or PbS (Ultrapure 99.999+%). A range of concentration of the solid sulfur containing species was obtained by the laborious process of serial dilution
0003-2700/82/0354-0971$01.25/00 1982 American Chemical Society