798
Anal. Chem. 1981, 53, 798-801
Determination of Halocarbons in Air by Gas Chromatography-High Resolution Mass Spectrometry Fabrlrlo Bruner, * Giancarlo Crescentlni, and Fillppo Manganl
Istituto di Scienze Chimiche, UniversitCi di Urbino, Piazza Rinascimento, 6-61029 Urbino, Italy Enzo Brancaleonl, Achllle Cappiello, and Paolo Ciccioll Istituto sull'lnquinamento Atmosferico del CNR, Area della Ricerca di Roma, 000 16 Monterotondo Scalo, Italy
The qualitative and quantitative analyses of air samples for halocarbons at the pptv level using high-resoiutlon mass spectrometry in the selected Ion monitoring (SIM) mode and very selective packed columns Is descrlbed. Permeation tubes are used as primary standards. Some results obtalned In the determinatlon of CC13F (F-11) and CHCi2F (F-21) in alr samples from Red Sea, Indlan Ocean, and Itallan locations are reported.
identify CHClzF (F-21) in several air samples (12). The advantages of using both high-resolution mass spectrometry and highly selective GC columns convinced us to set up an analytical procedure to overcome most of the difficulties connected to a reliable identification and quantitation of trace compounds in the atmosphere. The importance of exact mass measurement for quantitative analysis is shown in the following, together with examples of application of the method.
The problem of stratospheric ozone depletion by fluorocarbons, raised after the work of Molina and Rowland (1,2), is one of the most challenging themes for increasing efforts of analytical chemists to ascertain the actual concentration of halocarbons in the atmosphere. The rate of increase of halocarbon concentration with time in background air, the search for possible tropospheric sinks for halocarbons, and the determination of halocarbon lifetimes in the troposphere are major topics that require the use of highly accurate and precise analytical methods. There is no doubt that gas chromatography is the most suitable technique to be used for the separation of halocarbons, and good results have been abtained by Guiochon and co-workers using gas-solid chromatography (3). Some of us have developed highly selective GC columns using liquid modified adsorbents ( 4 , 5 )for this type of separation. This technique ensures a much better separation of halocarbons of similar structure with respect to the usual gas-liquid chromatography (6-8).A highly selective GC column is very useful when some halocarbons, present in the air sample in ultratrace amounts, have to be determined in the presence of relatively abundant species of the same kind. The situation is further complicated when two species are not well separated and the detector is sensitive to both compounds. Most measurements of atmospheric concentrations of halocarbons have been carried out by the electron capture detector (ECD), which exhibits high sensitivity and specificity toward halogenated compounds, and outstanding work has been carried out in this field by Lovelock (6). However, in spite of these properties, the ECD has some disadvantages, namely, an insufficient specificity coupled with a low response linearity range in several cases (9). Sampling and sample contamination are also problems one should carefully account for, to ensure compatibility of the GC system with the air sample magnitude and to avoid interferences due to those halocarbons present in every laboratory and in the carrier gas tanks ( 4 ) . Selected ion monitoring (SIM) has been used by Grimsrud and Rasmussen to measure several halocarbons from air samples (10) and by Cronn and Harsch (11)for the same purpose. All these authors used quadrupole mass spectrometers in their work. More recently, some of us have used a medium-resolution mass spectrometer and were able to
Preparation and Calibration of Permeation Tubes. Permeation tubes are prepared from FEP Teflon (DuPont) according to the procedure previously described (13, 14). Their length, internal volume, wall thickness, and permeation surface have been chosen according to the volatility, permeation rate, and range of concentrations required for the analysis of each halocarbon. Graphs showing the loss of weight against time have been recently reported for several halocarbons (9). Linearity is reached within a range of time depending on the dimensions of the permeation tubes and the nature of the compound. After preparation, the tubes are conditioned 24 h at 40 "C and then placed in a double-wallcylindrical glass container similar in shape to the usual condensers for laboratory distillation. A thermostat ensures water circulation in the double-wall interspace at 25 f 0.1 "C. The inner space, where permeation tubes are located, is fed with a flow of extra pure helium further purified by means of a molecular sieves trap kept at -78 "C. The helium stream passes through a coil immersed in the water thermostat before reaching the permeation tubes. A schematic view of the apparatus employed is shown in the top part of Figure 1. Experiments to show whether the simultaneous presence of different permeation tubes might affect the permeation rates gave negative results, so that different permeation tubes are placed in the same container for simultaneous injection of different halocarbons into the GC column during the calibration run. Injection System. The injection system is schematically shown in the lower part of Figure 1. It is possible to inject alternatively into the column either the calibration mixture or the content of the trap where the halocarbons collected from the atmosphere are adsorbed. This can be obtained by switching the two zero volume valves (Valco Instrument Co., Houston, TX). Two calibrated loops are used to sample the calibration from the helium stream. Their volume is 0.1 and 0.2 mL, respectively. The calibration mixture can be injected either directly into the GC column or through a trap identical with those used for air sampling. The latter mode appears to be the most suitable because both calibration mixture and the actual sample undergo the same procedure prior to injection into the GC column. The trap is maintained at about -90 "C by means of dry ice. When the content of the trap has to be injected into the GC column, the flask is replaced with an oven to raise the temperature to about 120 "C, while the trap is closed. By switching the four-port valve the carrier gas is diverted to the trap, and its content is injected into the GC column. The use of Carbopack B (Supelco Co, Bellafonte PA) traps for sampling and desorption of halocarbons has been described and discussed in detail in previous papers ( 4 , 5 ) . Up to 30 L of air
EXPERIMENTAL SECTION
0003-2700/81/0353-0798$01.25/00 1981 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 53. NO. 6, MAY 1981 799
Table I. S. I. M. Programs program 1 reference mass C,H,O‘ (m/e 58.0417) compd ion m/e CHF’W’ 66.9751 CHF,CI CHF37CI’ 68.9721 CF,”Cl+ 84.9647 CF,CI, CF,”Cl+ 86.9627 CHF’TI‘ 66.9751 CHFCI, CHF3’Cl+ 68.9721 CH,CI, CH,”ICI’ 83.9627 C,F,CI* CF,”CI+ 96.9627
E/E,
86 84 68 66 86 84 69 66
program 2 reference m a s C,H,’ (m/e 91.0546) compd ion m/e CFWl,’ 100.9361 CFCI, CF3W3’CI+ 102.9361 CF”CI,’ 100.9361 C,F,CL CF’sCIWl+ 102.9361 CH,C”CI+ 96.9612 CH,CCl, C’~CI,‘ 116.9065 CCI, C3’CI,+ 116.9065 C,HCI, C,H’TICI,‘ 94.9455
E/&
90 88 90 88
93 77 77
95
Permeation Tube
Flow M e t e r
__ direct Inlectlon
~
~~~~
Figure 2. Repemhre scan in the region of mle = 67 at the retentlon time of CHCI,F (F-21).
T0G.C. Column lnisctlon through the trap
V
t
~ o o p ~ ~ , To , , G.C.Colurnn ~
Flgure 1. Schematic view of the calibratlon system. can be analyzed for their halocarbon content using such a sampling technique. Instrumentation. A 3 m long,2 mm i.d. glass column packed with Carhpaek B &lo0 mesh coated with 0.5% SP 1000 is used. The outstanding properties of this column for the separation of halocarbon mixtures have been already discussed (7). A gas Chromatograph, Model 3900 (Dani S.pA, M o m , Italy), is coupled via a jet separator to the ion source of a VG Micromass Model 70-70 F double focusing mass spectrometer (VG Organic Ltd., Altrincham, GB). This instrument shows a maximum resolution of ahout 25000 (1070 valley definition and 0.5% transmission) and is used in the electron impact ionization mode in all this work. All mass chromatopams of the effluents from the CC columns are carried out at a resolving power of about 3000 (10% valley). The two programs u3ed for the multiple ion detection (MID) of chlorofluorocarbonsare shown in Table I. The fint program has rn/e 68.0417 as a reference mms. This ion is rontinously and constantly produced allowing acetone from the liquid injection port of the ion source. For the second program the reference rn, e value chosen was 91.0546,obtained in the same way with toluene. Acetone and toluene were selected as reference hecause their elution does not interfere with the compounds to be detected. The two programs are designed to obtain a sufficiently good sensitivity toward the various ions 10 be monitored according to the ElKO value necessary fur focusing a particular ion.
RESULTS AND DISCUSSION To show the high specificity of the analysis, we show in Figure 2 a repetitive scanning record on the UV paper while the peak of CHC1,F is entering the ion source. It is important to note the complete separation of the two peaks having the
Figure 3. Calibration cuves for several halocarbons. same nominal m/e value 67. The higher one corresponds to the C&* ion, very common in the instrumental background and in the mass spectra of many organic compounds as well. The profile of the CHClzF GC peak is reconstructed by the dotted line. Calibration curves for several halocarbons are reported in Figure 3. Linearity is gocd in the entire range examined and reported for each compound. This includes the range of mass
800
ANALYTICAL CHEMISTRY, VOL. 53, NO. 6, MAY 1981
E.C.D.
-
16 , , h
E
84.96j&
F11
102.93
F113
CH(SCCI~ 98.96
(121p.J)
I
I
0
6 8.07 r
20
10
FlZ
er 8s
i'
30
Ne
(min)TIME
M.I.D.
LU
F F11 F114L
A
68.97
J
fi
A
0.5v9696
m/e
m/e
Flgure 5. Electron capture and multlple ion detection of two samples of analogous concentration, 0
10
20
30
TIME imln)
Flgure 4. Mass fragmentograms of two calibration runs at (A) low and (6)high concentrations.
that is actually injected into the column when air samples are analyzed. The slope of curves is quite different for the various compounds. This depends, of course, not only upon the different ionization efficiency of the molecules but also and to a larger extent on the E/Eo value for the accelerating voltage during the S.I.M. program. For example, CC14 and CF2Clz(F-12) show a lower sensitivity because the accelerating voltage is only 77% for the former and 66% for the latter. This is not a disadvantage because of the relatively high abundance of the two compounds. CHC12F, that is frequently present in small concentrations, shows instead good sensitivity, better than CCl,F, although in the MID program the E/Eovalue is slightly higher for the latter. Comparing the calibration curves made with the mass spectrometer to those obtained with the ECD (9), linearity range of MS appears to be higher than with ECD and the response more uniform. Repetitive measurements made within the same days or different days by different operators and without paying any special care to optimizing and MS conditions yielded a standard deviation of *lo%. We have observed that fluctuations in the MS ion source are mainly responsible for such value of precision. By injection of the calibration mixture immediately before and after the actual air sample, the overall uncertainty is reduced and a standard deviation of &5% for the overall method is achieved. In Figure 4 an example is given of the analysis of two calibration mixtures containing some halocarbons at high and low concentration, respectively. In Figure 5 a comparison of two samples of similar concentrations collected in the same location (Montelibretti, 30 km North East of Rome), analyzed with ECD and the mass spectrometer as detectors, is made. In the ECD chromatogram the CHClzF peak is very small, while the peak from CC13F is out of the linear range of the detector. The two compounds cannot be analyzed in the same sample and this is a disadvantage. Moreover, the retention
Table 11. Concentration Values of Some Halocarbons Collected in a Suburban Area Near Rome compd
concn, PPtV
CCl,F, (F-12) CC1,F (F-11) CHC1,F (F-21) C,Cl,F, (F-113)
223 132 0.9 15.5
compd C,Cl,F, CC1, CH,CCI, C,HCl,
concn, PPtV 15 50
I9 16.5
time is the only parameter for the identification of CHC12F in the ECD chromatogram. In the mass chromatogram both compounds can be quantitatively evaluated. Furthermore, the identification of CHClzF is reliable. In this case the relative intensity of the ions corresponding to mle 66.9751 and 68.9721 together with the retention time excludes any possible confusion with other substances. The peak at mle 66.9751 is well separated from any other peak in the region of mle 67 value originated by other compounds. It should be noted that the CHClzF concentration in this air sample was 0.8 ppt and 20 L of air were trapped. Penkett et al. have recently published a paper (15) reporting the analysis of CHC12F at the sub-parts-per-thousand level by mass fragmentography at mle 67 in the single ion monitoring (SIM) mode using a low-resolution mass spectrometer. They attribute to one of the peaks in a cluster, where the separation is less than 50%, the name of CHC12F. The retention time of the peak is 2 min. Both for the low resolution of the mass spectrometer and for the low resolution of the GC column, the use of SIM in such conditions has negligible practical advantage over the ECD detection, since the identification remains highly doubtful. It is interesting to note that CHCIFz (F-22), which shows a retention time close to that of air and water on our column and is almost completely hidden in the ECD chromatogram, yields a very clean peak at m / e 66.9751 and 68.9721. The sensitivity is not very high for this compound because the fragment ions detected are not the most intense in its mass spectrum.
Anal. Chem. 1981, 53, 801-805
that the method here presented is quite flexible, allowing the simultaneous determination of compounds present in very different concentrations. No correlation apparently exists between the concentrations of CC&Fand CHClzF within the same sample. An evaluation of these data implies a discussion that is out of the scope of this work. The data obtained by using this method in several monitoring campaigns have been presented and discussed in a paper recently published (16).
L0cation:RED S E A Sample Volume: 30 L F12
I
F11
84.96
102.93 116.90
88.86 83.96 68.87 68.97 m/e
1
!*2
1
.
IJ i;d
100.93
JL94.04 96.96 m/e
Flgure 6. Analysis of an air sample collected over the Red Sea.
Table 111. Selected Values of the Concentrations of CHC1,F (F-21) and CCl,,F (F-11) in Some Air Samples
sample
location Central Red Sea Indian Ocean Porto Marghera S. Marco Platform
801
no. 3 8 42 51-53
amtof amtof CC1,F CHC1,F (F-11), (F-21), P,PtV PPtV 135
11
188
36
485
153
0.0 0.7
Figure 6 shows the analysis of a 30-L air sample, collected over the Red Sea where CHClzF is present at higher concentration (36 ppt). In Table I1 a typical concentration pattern for several halocarbons determined by the method here degcribed is shown. Some uncertainty is in the value found for CC14, but it is known that this halocarbon presents difficulties due to possible decomposition during storage or other processes. In Table I11 the concentrations of CC13F and CHClzF determined within the same air sample are reported, showing
ACKNOWLEDGMENT The authors are indebted to A. R. Mastrogiacomo for the preparation of permeation tubes and to the Mass Spectrometry Service for the CNR Rome research area for the use of the instrument.
LITERATURE CITED (1) Moiina, M. J.; Rowiand, F. S. Nafure (London) 1974, 249, 810-812. (2) “Stratospheric Ozone Depletion by Halocarbons: Chemlstry & Transport”; National Academy of Science: Washington DC, 1979. (3) Vidal-Madler, C.; Gonnord, M. F.; Benchah, F.; Guiochon, G. J. Chromafogr. Sci. 1978, 16, 190-194. (4) Bruner, F.; Bertoni, G.; Crescentinl, G. J. Chromafogr. 1978, 167, 399-407. ___ (5) Crescentinl, G.; Bruner, F. Ann. Chim. (Rome) 1978, 66,343-348. (6) Loveiock, J. E.; Maggs, R. J.; Wade, R. J. Nature(Lor0‘on)1973, 241, 194-199. (7) Su, C. W.; Goidberg, E. D. Nafure (London) 1973, 245, 27-29. (8) Westberg, H. H.; Rasmussen, R. A.; Hoidren, M. Anal. Chem. 1974, 46, 1852-1854. (9) Crescentini, G.; Mangani, F.; Mastroglacomo, A. R.; Bruner, F. J. ChrOmatOgr. 1981,204, 445-451. (IO) Grlmsrud, E. P.; Rasmussen, R. A. Atmos. fnvlron. 1975, 9 , 1010-1015. (11) Cronn, D. R.; Harsch, D. E. Anal. Left. 1979, 12, 1489-1492, and references therein.
(12) Crescentini, G.; Bruner, F. Nature (London) 1979,279, 311-312. (13) O’Keeffe, A. E.; Ortman, G. C. Anal. Chem. 1966, 38, 760-785. (14) Slng, H. B.; Salas, L.; Lillian, D.; Arnts, R. R.; Appleby, A. fnviron. Sci. Techno/.1977, 77, 511-513. (15) Penkett, S. A,; Prosser, N. J. D.;Rasmussen, R. A,; Khaiii, H. A. K. Nature (London) 1980, 286, 793-798. (16) Crescentlni, G.;Bruner, F. Ann. Chlm. (Rome) 1980, 70, 631-636.
RECE~VED for review October 20,1980. Accepted January 22, 1981. This work was partially supported by Chemical Manufactures Association under Research Project CF 78-256 R and, in the early stage, by Commission of the European Communities under Contract No. 214-77-1 ENV I.
Effects of Normalization on Feature Selection in Pyrolysis Gas Chromatography of Coal Tar Pitches Matthew S. Klee, Alice M. Harper, and L. B. Rogers” Department of Chemistry, IJniversity of Georgia, Athens, Georgia 30602
Computerized calculationi of variance weights sometimes detected important features that were not obvious from visual Inspections of chromatograms for any two of three classes of coal tar pitches. I n addition, normalizations by one or more peak areas or of peak heights, procedures often resorted to as a means of minimizing the effect of dlfferences In size or sample heterogeneity, generally resulted in decreasing the welghts of all features. F:lnally, a successful means was devised for “correcting” the chromatograms obtained using a column that had gradually degraded durlng a long series of runs.
Correlating the quality of a thermic graphite electrode with
a property of the coal tar pitch from which it was fabricated has been difficult. It has often been necessary to make the actual electrode and assess its quality, rather than to estimate its quality using a physical or chemical measurement of the pitch prior to fabrication. Recently, an attempt at correlating the profiles of the pyrograms of some pitches to their coking values (a measure of pitch quality derived after graphitization), has been moderately successful (I). Those comparisons were based solely on criteria that could be detected easily by eye. The present study had two goals. The first was to determine if pyrolysis gas chromatographic analysis of the coal tar pitches could be optimized to aid in the estimation of the quality of a given coal tar pitch. The second was to determine if algorithms commonly used as preliminary steps in some pattern recognition schemes could significantly aid in the
0003-2700/81/0353-0801$01.25/00 1981 American Chemical Society
