Anal. Chem. 1989, 61, 2118-2121
2118
Norin, H.; Ryhage, R.; Christakopoulos, A.; Sandstrom, M. Chemosphere 1983, 12, 299-315. Edmonds, J. S.;Francesconi, K. A. Experlentia 1987. 43, 553-557. Irgolic, K. J.; Woolson. E. A,; Stockton, R. A,; Newman, R. D.;Bottino, N. R.; Zingaro, R. A.; Kearney, P. C.; Pyies, R. A,; Maeda, S.; McShane, W. J.; Cox, E. R. EHP, Environ. Heallh Perspect. 1977, 19.
61-66. Morita, M.: Shibata, Y. Chemosphere 1988, 17, 1147-1 152. Morita, M.; Uehiro, T.; Fuwa, K. Anal. Chem. 1981,53, 1806-1808. Kurosawa, S.:Yasuda, K.; Taguchi, M.; Yamazaki, S.;Toda, S.; Morita,M.; Uehiro, T.; Fuwa, K. Agric. Biol. Chem. 1980,4 4 , 1993-1994. Morita, M.; Shibata, Y. Anal. Sci. 1987, 3 , 575-577. Beauchemin, D.;Bednas, M. E.: Berman, S.S.; McLaren, J. W.; Siu. K. W. M.; Sturgeon, R. E. Anal. Chem. ?98& 6 0 , 2209-2212. Lawrence, J. F.; Michallk, P.; Tam, G.; Conacher, H. B. S.J. Agric. Food Chem. 1986,3 4 . 315-319. Lau, B. P.-Y.; Michalik, P.; Porter, C. J.; Kroiik, S. Biomed. Environ. Mass Spectrom. 1987, 14, 723-732. Shibata, Y.; Morita, M. Anal. Sci. 1989,5 , 107-109. Berman. S. S.; Sturgeon, R. E. Fresenius' Z . Anal. Chem. 1987, 326, 712-715. Shiomi, K.: Kakehashi, Y.; Yamanaka, H.; Kikuchi, T. Appl. Organomet. Chem. 1987, 1 , 177-183.
Sir: It is gratifying to know that experienced scientists using quite different procedures have largely confirmed our experimental results that arsenobetaine is the major arsenic species in DORM-1 ( I ) . The discrepancy is on the assignment of an approximately 1% fraction that we attributed to arsenocholine. According to the authors, that fraction should have been assigned to tetramethylarsonium ion. We have recently found that tetramethylarsonium ion has retention properties very similar to that of arsenocholine on a Dowex 50W-X8 column. Having been shown the authors' HPLC data, we have to agree that our assignment of arsenocholine is likely to be erroneous. We are grateful to the authors for bringing this to our attention. ACKNOWLEDGMENT We thank W. R. Cullen, Department of Chemistry, Univ-
(16) Shiomi, K.; Aoyama, M.; Yamanaka, H.; Kikuchi, T. Comp. Biochem. Physiol.. C : Comp. Pharmacol. Toxicol. 1988. 9OC, 361-365. (17) Francesconi, K. A,; Edmonds, J. S.;Hatcher, B. G. Comp. Blochem. Physiol., C : Comp. Pharmacol. Toxlcol. 1988, 9OC, 313-316. (16)Norin, H.; Christakopoulos, A.: Sandstrom, M.; Ryhage, R. Chemosphere 1985, 14, 313-323. (19) Edmonds, J. S.;Francesconi, K. A. Sci. Total Environ. 1987, 6 4 , 317-323. (20) Kaise, T.; Hanaoka, K.; Tagawa, S. Chemosphere 1987, 16, 255 1-2558.
Yasuyuki Shibata* Masatoshi Morita Chemistry and Physics Division National Institute for Environmental Studies 16-2 Onogawa, Tsukuba Ibaraki 305, Japan RECEIVED for review February 23, 1989. Accepted June 8, 1989.
ersity of British Columbia for a sample of tetramethylarsonium iodide.
LITERATURE CITED (1) Beauchemin, D.; Bednas, M. E.; Berman, S. S.;McLaren, J. W.; Siu, K. W. M.; Sturgeon, R. E. Anal. Chem. 1988, 60,2209-2212.
K. W. M. Siu* R. E. Sturgeon J. W. McLaren S. S. Berman Division of Chemistry National Research Council of Canada Montreal Road Ottawa, Ontario Canada K1A OR9 RECEIVED for review May 22, 1989. Accepted June 8, 1989.
Complex Mixture Analysis Using Differential Gas Chromatographic Mass Spectrometry Sir: Commercial analytical laboratories routinely use gas chromatographic mass spectrometry (GCMS) to identify and quantitate a relatively small number of well-defined analytes in environmental or clinical samples. The components are most often priority pollutants, drugs, or metabolites whose chromatographic retention times and mass spectra are known. Increasingly, however, analysts are called upon to screen very complex samples and to identify unknown components that may be of environmental or clinical concern. This task is usually straightforward for components that are well separated by gas chromatography and are present in sufficient concentration to provide clean mass spectra. Complications invariably arise when, as is more often the case, the components are incompletely resolved, mass spectra overlap, and interpretation becomes problematic. A host of computer techniques have been proposed (1-8) for the deconvolution of overlapping peaks, providing cleaner mass spectra and allowing interpretation and even quantitation of the unknown analytes. In all of the methods described to date, spectral deconvolution is performed after the data collection is over and the 0003-2700/89/0361-2118$01.50/0
raw mass spectra are stored in the data system. We recently proposed (8) "differential GCMS", a method that improves spectral quality by subtracting successive pain of mass spectra and storing the result. As the concentration of a compound increases, the abundances of ions representing that compound increase, and these ions appear with positive abundances in the differential mass spectrum. As the concentration decreases, ion abundances in the differential mass spectra become negative (8). Several aspects of the method are unusual. Of all the mass spectral deconvolution techniques, differential GCMS is the only one capable of real-time operation. That is, the spectral cleanup could be performed as data are being collected. The processed spectra could be stored along with, or instead of, the raw mass spectra. When the differential mass spectra are summed into a total ion chromatogram, there is an increase in apparent chromatographic resolution (8). This improvement can be quantitatively estimated by using existing theory. Davis and Giddings (9) have developed a theoretical model for a chromatographic system. The model uses Poisson sta0 1989 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 61, NO. 18, SEPTEMBER 15, 1989 100
10
,
2119
I
0
16C
0
2K
m
of theoretical plates required to form a singlet with a probability of 90% as a function of m : (A) standard GCMS experiment, V , / V , = 25, R, = 1.0; (E) differential GCMS experiment. Figure 2. Number
SCAN NUMBER Figure 1. (a) Simulated GCMS data ( 10) showing two Gaussian peaks separated by a distance of 40. (b) Same as a, but with peak separation of 2u. (c) Differential GCMS total ion profile ( 8 )derived from the data set in b. Mass spectral overlap is reduced and apparent
chromatographic resolution is improved. tistics to describe the random elution of compounds in a system with finite peak capacity. For a mixture of m components in a system with a peak capacity of n,, the model predicts that the number of components eluting as singlets is expressed as p1 = me+ where a is the ratio of the number of components to the peak capacity: a = mjn,. The hypothetical peak capacity is determined by dividing the entire chromatographic coordinate space, X , by the width of one chromatographic peak, xo ( x g is, therefore, the minimum distance by which two components may be separated and still be considered chromatographically “resolved”): n, = x/xo The model goes on to predict the number of chromatographic peaks ( p ) eluted from the system to be p = me-a This model can be readily be adapted to the differential GCMS case. If a chromatographic peak is symmetrical, ion abundances in the differential mass spectra will be positive during the first half of the elution and negative during the second half. In terms of chromatographic overlap, the situation can be represented as in Figure 1. The resolution of two Gaussian peaks (Figure la) will be achieved when R, = 1 or xo = 4u,where u represents the average standard deviation of the two peaks (9). In the case of differential GCMS (Figure Ib,c), the peaks achieve the same degree of overlap a t a distance x o = 2u. Because of the differentiation process, the effective “width” of a peak has been halved. Davis and Giddings showed that for a 50-component mix-
ture in a system with a peak capacity of 100, only 30 chromatographic peaks can be expected to elute ( p = me-m/nc= 50e-(50/100) = 30), and only 18 of these will represent pure components ( p l = = 50e-2(50/1m) = 18). Our differentiation has the effect of doubling the peak capacity (n,) of the chromatogrphic system (by halving the peak width) without any change in other variables. This also has the effect of halving a. For the 50-component sample above, differential GCMS would produce 39 distinct chromatographic peaks (p = 50e-w/200= 39), of which 30 would be pure components ( p l = 5Oe-2(50/2W = 30). An alternative perspective for viewing the improvement in chromatographic resolution is to calculate the number of theoretical plates required to provide a high probability (90%) of forming a pure singlet ( p l = 0.9) (9). Figure 2A shows a plot of such a calculation: The number of theoretical plates (N) is plotted as a function of the number of components (m). (A retention volume ratio, V21VI, of 25 and a resolution, R,, of 1are assumed for this plot as in ref 9.) In the differential GCMS data, components are “resolved” when the peak separation is 2a, rather than 4a (see Figure 1). In other words, peak resolution requires only R, = 0.5. The peak capacity (9)
n, =
-“ .I 2
-In
Vv
4Rs Vl is doubled when R, = 0.5 compared to R, = 1.0, consistent with the discussion above. To achieve a given probability of singlet formation, one would require only one-fourth the number of theoretical plates (Figure 2B). This means that an acceptable degree of separation can be achieved on a shorter GC column (lower N )or in a shorter period of time by using differential GCMS than would be necessary for the standard experiment. Martin and Guiochon (11)have proposed a parameter they term y , the extent of separation, to describe the quality of a chromatographic separation of a multicomponent mixture. The model is based on an analogy between the chromatographic separation process and a random depolymerization
2120
ANALYTICAL CHEMISTRY, VOL. 61, NO. 18, SEPTEMBER 15, 1989
100
-100
,
1
I
i
180
300
SCAN NUMBER Flgure 3. (a) Total ion profile and (b) differential GCMS profile from the analysis of the volatile organic compounds in a sample of indoor air. Compounds trapped on Tenax were thermally desorbed into a continuously scanning Hewlett-Packard 5985 (70 eV) mass spectra were collected every 2.2 s from m l z 40 to 400.
170
SCAN Figure 4. (a) Total ion profile and (b) differential GCMS profile for scans 167-173 of the data set in Figure 3.
process and results in mathematical expressions virtually identical with those of Davis and Giddings (9). The y parameter ranges in value from 0 to l, with 0 indicating no separation at all and 1indicating perfect separation, analogous to the extent of chemical reaction (depolymerization). Mathematically, the relationship between the two models may be expressed (11) by = e-a = e - m f n c As before, the effect of differential GCMS is readily observed. For the same 50-component mixture in a 100-peakcapacity system, one calculates the extent of separation as y = c-io/lw = 0.61 before differentiation. After differentiation, -/ becomes e~50’200 = 0.78, closer to 1 and therefore a more complete separation.
B GCMS system.
Electron ionization
The most efficient data collection mode for differential mass spectra would be to collect and store the difference spectra in place of the usual raw mass spectra. Some analysts are understandably uncomfortable with the notion of discarding the raw data. This need not be a cause for concern. (It should be noted, in passing, that the “raw” mass spectra being discussed are already extensively processed digitized, amplified, filtered, centroided, mass assigned, etc., after which real raw data, i.e. the analogue signals from the electron multiplier, are discarded.) The differential mass spectra are the result of a subtraction of successive mass spectra. When the first mass spectrum in a data set is collected, it has no prior spectrum to subtract. If this first spectrum is stored intact (absolute abundances, as measured), and if each subsequent spectrum is stored as the difference between two adjacent spectra, the original mass spectra can be reassembled, postrun, by simply summing the first and second spectra, the second the third, etc. By starting at the first spectrum, stored with its measured abundances, one can “unzip” the differential mass spectra to recover what most analysts now consider to be the raw data. In time, we believe, the differential mass spectra will be shown to provide all of the same information as the raw data, and in a form, after some initial accustomization, that is perhaps more efficient and easier to use. As an illustration of the chromatographic resolution enhancement, we include a data set from the analysis of the organic compounds in an air sample. This sample was collected in a local building after several employees complained of illnesses and unpleasant reactions, thought to be associated with the air inside the building. The organic compounds were trapped on a solid adsorbent and then thermally desorbed into the GCMS system. Figure 3a shows the total ion profile from the data set; the sample obviously represents a very complex mixture. Figure 3b shows the same data set after processing to generate a differential GCMS file. Successive pairs of mass spectra were subtracted, as we have described elsewhere (8). The positive and negative abundances of ions were summed separately and plotted against scan number (of the later scan) to produce the differential GCMS profile in Figure 3b. The improvement in chromatographic resolution is readily apparent, particularly in regions of the chromatogram that are complex, for example, between scans 180 and 250. The
Anal. Chem. 1989, 6 1 , 2121-2124
background from chemical noise is also lower, as evidenced by a decreased base line. In effect, by differentiating the GCMS data set, one is subtracting a “local background” from every scan. A small region of the chromatogram is expanded in Figure 4 (scans 167-173). A partially resolved doublet appears in Figure 4a, the total ion profile. The resolution of these two compounds can be estimated to be 0.6 (R, = ( t 2- t1)/[0.5(w, w J ] ,where t,, t , are retention times and wl,w 2 are peak widths for components 1 and 2). After differentiation (Figure 4b), the two components are resolved almost to base line; R, = 0.9. The peak separation, ( t z- tl), has not changed, but the peak width is substantially narrower, approaching the theoretical prediction discussed above. The mass spectra corresponding to these compounds are also improved. In the original data set, the mass spectra overlapped in many of the scans; in the differential data set, the spectra do not overlap and provide correspondingly easier interpretation. In fact, in this sample, we were able to demonstrate the presence of benzene only after differentiation. In the raw spectra, compounds eluting close to benzene produced many ions in the region where the m / z 78 of benzene was observed, making the identification of benzene somewhat tenuous. After differentiation, a clear mass spectrum of benzene was recovered that exactly matched the spectrum from the reference library (data not shown). In summary, the technique of differential GCMS, although very simple to implement, provides a number of advantages
2121
over conventional GCMS, including better chromatographic resolution, lower background, and cleaner mass spectra. Studies to demonstrate the utility of the method for quantitation of analytes are under way.
ACKNOWLEDGMENT Stimulating discussions with Cliff Carlin, University of Maine, are gratefully acknowledged.
+
LITERATURE CITED (1) Biller, J. E.; Biemann. K. Anal. Lett. 1974, 7,515-528. (2) Dromey, R. G.;Stefik, M. J.; Rindfleisch, T. C.; Duffield. A. M. Anal. Chem. 1976, 48, 1368-1375. (3) Knorr, F. J.; Futrell, J. H. Anal. Chem. 1979, 57, 1236-1241. (4) Sharaf, M. A.; Kowalski, B. R. Anal. Chem. 1982, 5 4 , 1291-1296. (5) Knorr, F. J.; Thorsheim, H. R.; Harris, J. M. Anal. Chem. 1981, 5 3 ,
821-825. (6) King, M. D.; King, G. S. Anal. Chem. 1985, 5 7 , 1049-1056. (7) Lacey, R. F. Anal. Chem. 1988, 58, 1404-1410. (8) Ghosh, A.; Anderegg, R. J. Anal. Chem. 1989, 6 7 , 73-77. (9) Davis, J. M.; Giddings, J. C. Anal. Chem. 1983, 55, 418-424. (10) Ghosh, A.; Morison, D. S.;Anderegg, R . J. J . Chem. Educ. 1988, 65,
A154-Al56. (11) Martin, M.; Guiochon, G. Anal. Chem. 1985, 5 7 , 289-295. * Author to whom correspondence should be addressed
Amit Ghosh Robert J. Anderegg* Department of Chemistry University of Maine Orono, Maine 04469
RECEIVED for review July 3, 1989. Accepted July 3, 1989.
TECHNICAL NOTES Chiral Polysiloxanes Derived from ( R ,R)-Tartramide for the Gas Chromatographic Separation of Enantiomers Kouji Nakamura* Tanabe Seiyaku, 16-89, Kashima 3-chome, Yodogawa-ku, Osaka 532, J a p a n Shoji Hara and Yasuo Dobashi Tokyo College of Pharmacy, 1432-1 Horinouchi, Hachioji, Tokyo 192-03, Japan The first chiral polysiloxane, termed Chirasil-Val, for the gas chromatographic separation of enantiomers was synthesized by coupling L-valine-tert-butylamide as a chiral moiety to a copolymer of (2-carboxypropyl)methylsiloxaneand dimethylsiloxane in 1977 (1). It was found to be extensively applicable to the enantiomeric separation of optical antipodes of amino acid, hydroxy acid, amino alcohol, diol, alcohol, and halocarboxylic acid (2-7). It was possible to synthesize chiral polysiloxanes similar to Chirasil-Val by using commercially available polysiloxanes. Verzele et al. synthesized a chiral polysiloxane by coupling L-valine-tert-butylamideto commercial polysiloxane OV-225 followed by subsequent conversion to acid chloride in two steps (8). Konig et al. also synthesized chiral polysiloxane by coupling L-valine-(S or R)-phenylethylamide to modified commercial polysiloxane XE-60 with dicyclohexyl carbodiimide (DCC) (9). Glass and fused-silica capillary columns coated with these chiral polysiloxanes were used for the gas chromatographic separation of enantiomers. The enantiomeric separation of chiral polysiloxanes is based on differences in the stability of diastereomeric complexes in enantiomeric pairs and the chiral moiety on chiral polysiloxane via hydrogen bonding. We recently reported (R,R)-N,N’-diisopropyltartramide to be broadly applicable as a chiral mobile phase additive in silica
gel chromatography and a chiral stationary phase derived from (R,R)-tartramide to have a considerable scope of application to the liquid chromatographic resolution of enantiomers (10-12). (R,R)-Tartramide was observed to have excellent resolving power in distinguishing enantiomers from each other via hydrogen bonding. The present study was carried out to prepare chiral polysiloxane derived from (R,R)-tartramide for the gas chromatographic separation of enantiomers.
EXPERIMENTAL SECTION Apparatus. Gas chromatographic analysis was performed on a Shimadzu GC-SA equipped with a split injector and a flame ionization detector, using helium as the carrier gas at an inlet pressure of approximately 1.3 kg/cm2. The chromatographic signal was recorded and processed by a Shimadzu C-R3A integrator.
‘H NMR spectra were obtained on a Varian EM-390 spectrometer and chemical shifts were expressed in parts per million ( 6 ) relative to tetramethylsilane as the internal standard. IR spectra were obtained on a Hitachi 260-10 spectrometer. Optical rotation was measured on a Jasco DIP-360 polarimeter. Melting points were determined on a micro hot-plate melting point apparatus. Reagent and Synthetic Preparations. Synthesis of NIsopropyltartaric Acid Monoamide ( 2 ) . N-Isopropyldiacetyltartaric acid monoamide (4.0 g, 14.5 mmol) was dissolved in 15
0003-2700/89/0361-2121$01,50/0 0 1989 American Chemical Society
