Anal.
Chem. 1981, 5 3 , 113-117
use of a matrix: Standing, K. G., paper presented at University of Munster, Oct 1980.
LITERATURE CITED Daves, G. D., Jr. Acc. Chem. Res. 1070, 12, 359-365. Day R. J.; Unger, S. E.; Cooks, R. G. Anal. Chem. 1980, 52, 557A. Schuken. H A . In!. J . Mass Spectrom. ron phys. 1979, 32, 97-283. Posthumw, M. A.: Kistemaker, P. G.; Meuzelaar, H. L. C.; Ten Noever de Brauw, M. C. Anal. Chem. 1078. 50, 985-991. MacFarhne, R. D.; Torgerson. D. F. Science 1976, 191, 920-925. Stirnpson, B. P.; Evans, C. A., Jr. Blomed. Mass Spectrom. 1978, 5 , 52-63. Benninghoven, A.; Sichterrnann, W. K. Anal. Chem. 1978, 50, 1180-1 184. Sichtermann, W.; Junack, M.; Eicke, A,; Benninghoven, A. Fresenlus' Z . Anal. Chem. 1980, 301, 115-116. Gaffney, J. S.; Pierce, R. C.; Friedman, L. J . Am. Chem. Soc. 1077, 99, 4293-4298. Gaffney, J. S.; Friedman, L. Annual Conference on Mass Spectrometry and Allied Topics, 1977, paper WD-6. Keough, T.; Destefano, A. J. Annual Conference on Mass Spectrometry and Allied Topics. 1980, paper MAMOC6. Jonkman, H. T.; Michl, J.; King, R. N.; Andrade, J. D. Anal. Chem. 1078, 50, 2078-2082. Lancaster, G. M.; Honda, F.; Fukuda, Y.; Rabahis, J. W. J. Am. Chem. Soc.1979, 101, 1951-1958. (14) Barber, M.; Vickermn, J. C.; Wolstenholme, J. J . Chem. Sot. Faraday Trans. 11980, 76, 549-559.
113
(15) Moor, J.; Waight, E. S. Org. Mass Spectrom. 1974, 9 , 903-912. (16) Schuken. H A . Methods Blochem. Anal. 1977, 24, 313-446. (17) Prome. J. C.: Puzo. G. Ora. Mass SDectrom. 1977, 12. 28-32. (l8j Chait, B. T.;'Field,.F. H. Annual Conference on Mass Spectrometry and Allied Topks, 1980, paper FAMOBIO. (19) Cdton, R. J.; Murday, J. S.; Wyatt, J. R.; DeCorpo, J. J. Swf. Scl. 1070, 84, 235-248. (20) Winkier. H. U.; Beckey, H. D. Org. Mass Spectrom. 1972, 8 , 655-660. 50 1-502. Day, R. J.; Unger, S. E.; Cooks, R. G. J. Am. Chem. Soc. 1970, 101, Sarnmons, M. C.; Bursey, M. M.; Brent, D. A. Biomed. Mess Specfrom. 1974, 1 , 169-171. Becker, 0.; FiKstenau, N.; Knippelberg, W.; Krueger, F. R. Org. Mass Spectrom. 1977, 12, 461-64. Krueger, F. R.; Wien, K. A&. Mass Spectrom. 1078, 1429. Schuken, H A . ; Beckey, H. D. Org. Mess Spectrom. 1973, 7 , 881-887. (26) &ade,H.; Cooks, R. 0. J. Am. Chem. Soc. 1078, 100, 5615-5621. (27) E M e , A.; Sichtermann, W. K.; Benninghoven, A. Org. Mass Spectrom. 1080. 15, 289-294.
RECEIVED for review August 6, 1980. Accepted October 27, 1980. This work was supported by the National Science Foundation CHE-78-08728 and MRL Program DMR-7723798. L.K.L. thanks the National Science-Council of the Republic of China for a fellowship.
Direct-Linked Gas Chromatography-Fourier Transform Infrared-Mass Spectrometer System Charles L. Wilklns," Gary N. G I s , Gregory M. Brissey, and Steven Stelner Department of
Chemistry, Universify of Nebraska-Lincoln, Lincoln, Nebraska 68588
Inttiai results of the joint use of complementary infrared and mass spectral information obtained from a dlrectly linked gas chromatograph-FourIer transform infrared-mass spectnnneter combinatlon are presented. It is demonstrated that the efficacious and unambiguous qualitative anaiysls of model mlxtures can be accompilshed in situations where mass spectral or infrared Information alone would not be adequate.
Reports over the past decade have documented the successful development of rapid and sensitive Fourier transform (FT-IR) spectrometers capable of making on-line measurements of infrared spectra of eluents from gas chromatograph columns (1-8). Recently, Kuehl and Griffiths have observed (9) that the sensitivity of GC/FT-IR systems has reached the point that they are viable as an alternative or complement to mass spectrometry for qualitative mixture analysis of mixtures containing components sufficiently stable and volatile for separation by GC. This reiterates the early suggestion by Low and Freeman (IO),substantiated by the later results of others (I1,12),that a direct linkage of GC/FT-IR and MS might yield an invaluable tool for mixture analysis. The use of multisource data of the type such a chemical information system can provide has been shown to be a powerful identification and structure elucidation tool (12-15). A pyrolysis GC system using a combination of trapping, m a s chromatographic separation, elemental analysis, and on-the-fly infrared spectrometry has shown the validity of the multiinstrument system concept (16). More recently, we have presented a preliminary account of our investigation of the first general direct-linked GC/FT-IR/MS system (17). 0003-2700/8 1/0353-0 1 13%01.OO/O
Subsequently, Hirschfeld has discussed preliminary experiments along these lines (18). In the present paper we describe more fully the results of our initial studies of a general-purpose on-line GC/FT-IR/MS chemical information system.
EXPERIMENTAL SECTION Instrumentation. Figure 1 is a block diagram of the system components. A Kratos MS-5076 high-resolution mass spectrometer, operating under control of a NOVA 4X computer using Kratus DS-55 software, was linked via a heated glass-lined stainlesa steel transfer line to a Nicolet 7199 GC/ET-IR system controlled by a Nicolet 1180 computer and using Nicolet GC/FT-IR software. A Varian 3700 gas chromatograph with a 6 ft X l/q in. 0.d. copper column, packed with 5% Bentone 34 and 5% didecyl phthalate supported on 60-80 mesh acid washed firebrick was used for all separations. The column was operated between 50 and 100 OC with He as carrier gas at a flow rate of 40-70 cm3/min. The GC effluent was split with an adjustable microvalve (a "T" configuration Scientific Glass Engineering Inc. MNVTU microvalve) in order to properly match the sample requirements of both the mass spectrometer and the FT-IR. This valve, which could be adjusted to route between 0 and 100% of the effluent to either the mass spectrometer or the infrared instrument (20 full turns of valve control to cover this range), was adjusted by first closing it to the mass spectrometer (Le., 100% effluent routed to the FT-IR). The valve was then carefully opened during elution of test samples to obtain an acceptable total ion current at the mass spectrometer detector. Less than half a turn proved necessary in all cases. Transfer lines were 1.6 mm o.d., 0.7 mm i.d. stainless steel glass-lined tubing. The line to the MS-5076 had a volume of 0.4 cm3and the line to the FT-IR light pipe had a volume of 0.2 cm3, Transfer lines and the FT-IRlight pipe were heated with heating tape (450 "C limit) to approximately 200 OC. The FT-IR transfer line and light pipe were controlled with a Chromel/Alumel thermocouple as a sensor. The MS-5076 transfer line was con0 1980 American Chemical Society
114
ANALYTICAL CHEMISTRY, VOL. 53, NO. 1, JANUARY 1981
YS-SO76
I
I
dL i ,------..-__.-
NICOLET 7199 GC-R
Molt& ZETA
R o m
I
NlCCLET lleo
;
.J
I
A B
Rs-eJ2
-----_.._ i
WKWOROS
Table I. Composition of Test Mixtures and Gas Chromatograph Conditions colinjec- flow rate, umn tion vol, cm3/ tepp, minHe C sample components PL
OMWORD
ROTTER
C
DISK
S M W
Flgure 1. Block diagram of the dlrect-linked GC/FT-IR/MS system.
D
trolled manually with a Variac and a copper/constantan thermocouple. The infrared light pipe was the standard Nicolet design, a 42 cm long gold-plated Pyrex tube, with a path length to volume ratio of 14 cm/cm3 (volume 3 cm3). It was held at 180 OC. The IT-IR detector was a Hg/Cd-Te liquid-nitrogen-cooled detector. GC effluent was introduced to the mass spectrometer via a jet separator with a flow of 12.5 cm3/min of helium as makeup gas. Spectral Parameters. Electron impact mass spectra were obtained by using a source voltage of 70 eV and scan times of 7.2 s / s m (range m/z 30-606 at resolution of ca. 3000) or 16.4 s/scan (range m / z 50-606 at resolution of ca. 15000). For all of the experiments reported, the lower resolution, higher speed scan proved satisfactory. Higher resolution measurements were made during a few runs to verify the capability of operating at enhanced resolution where necessary. Fourier transform infrared spectra were obtained by using a light pipe temperature of 180 "C and dry nitrogen as purge gas. Spectra were collected with resolution of 4 cm-' over the range 750-4000 cm-l and four spectra coadded (requiring a total measurement time of 4.4 s) for each spectral file stored. For continuous real-time monitoring, the Nicolet software "chemigram", which transforms 1K of each 4K word data file during the progress of the chromatography, was used. Samples. Mixtures were prepared by mixing approximately equal volumes of reagent grade chemicals. Mixture compositions, GC injection volumes, and separation conditions are summarized in Table I. Spectral Searches. Infrared spectra searches of a compressed version of the 2300 vapor-phase infrared spectra compiled by Leo Azarraga of the U S . Environmental Protection Agency, Athens, GA, laboratory were carried out with search software supplied by Nicolet Instrument Corp. For searching, spectral fdes were first reduced from 4 to 16 cm-' by the following procedure. First, a 17-point Savitsky-Golay smooth is performed on the 4096 point spectra file. Then, after the first 228 pointa are discarded,every fourth point is taken until 460 data points representing a 4000-450-cm-' spectrum are obtained. Intensities are normalized so that the largest peak has an absorbance of 1 and then truncated to 10 bits. An absolute difference metric is used for comparison. For this metric, the match factor M is defined by eq 1. Where Xi and Yi are the
E
4ooocm-1
M =
Xi
IXi- Yil
(1)
absorbance intensities of the sample and the library spectrum at wavenumber i. Thus a perfect match would yield a value of M of zero. Once all library spectra have been compared, the five closest matches are listed, along with the calculated values of M. Mass spectral library searches utilized a compressed form of the NIH/EPA/MSDC data base (ca. 32000 E1 maw spectra) and Kratos DS-55 search software. The algorithm employed is based upon that described by Clerc (14). In its present version, the spectrum to be compared with the library is first reduced to a 128-bit binary code by encoding peaks with each resolution interval as "1" if they exceed a specified threshold. Each bit representa
o-, m-, p-xylene anisole acetophenone methyl o-toluate methyl salicylate benzene toluene p-xylene 3-methylcyclopentene cyclohexene
1.0 1.0
65 70
170
1.0
40
120
1.0
40
60
trans,trans-
2.0
30
50
1.0
40 40
100 100
2,4-hexadiene
90
cis, trans-
F
2,4-hexadiene n-pentane n-hexane n-heptane
1.0
a single m / z value. The search itself then proceeds as described (14). When the library search is complete, the five (or more) most closely matching the unknown are listed, together with their mrea, which are given as percentages of an ideal (i.e., perfect) match. System Operation. Instrument setup, which involved movement of the GC/FT-IR system to an adjacent laboratory, connection of the transfer lines, alignment of the FT-IRsystem, and adjustment of both the FT-IRand mass spectrometer required approximately 3 h. In operation, a mixture sample is injected and both the NOVA 4X and Nicolet 1180 computer monitoring begun. The FT-IR software plots a reconstructed gas chromatogram using total integrated IR intensity of low resolution spectra over previously specified spectral windows. As many as five such reconstructed gas chromatograms can be generated simultaneously in this manner. A simple threshold algorithm is used to reject nonsample (Le., base line) spectra. At the same time, a mass spectral reconstructed gas chromatogram is produced by the DS-55 system. These were used to verify lack of chromatographic degradation, as well as to estimate relative transfer line delays.
RESULTS AND DISCUSSION
It is well-known and widely accepted that mass spectrometric analysis of gas chromatographic effluents (GC/MS) is an extraordinarily valuable tool for mixture analysis. Nonetheless, even when materials possess the stability and volatility necessary for GC separation, there are a significant number of cases where mass spectra alone are inadequate for identification of unknowns. The analysis of isomeric substances (which, unfortunately are often found together) is a case in point. Furthermore, when computer library searches are performed, it is generally the case that numerous library spectra of different substances fit the unknown well. Thus, positive identification often requires manual consultation of reference spectra and visual comparison. Although this may not be onerous for a single unknown, the very realistic possibility of needing to do this for as many as five spectra each for 15,20, or more effluents whose spectra are obtained in a 0.5-h GC-MS run, quickly renders the approach impractical or, at least, the slowest step in the analysis. It has been known for years, and repeatedly demonstrated (12-15), that the use of complementary information can greatly constrain the identification problem and simplify structure elucidation, when unknown compounds are encountered. As mentioned at the outset, FT-IR spectrometry has reached the point where it is technically feasible to assemble a linked GC/FT-IR/MS system and to evaluate its use for
ANALYTICAL CHEMISTRY, VOL. 53, NO. 1, JANUARY 1981
problems of this sort. Significantly, infrared spectra are often best for distinguishing substances which pose the greatest difficulty for mass spectrometry; the converse is also true. This then, was the motivation for the research discussed here. Sample Composition. For our initial studies of a linked system, we chose to examine a series of model mixtures of known composition. The components were selected so as to contain a number of representative substances which also provide examples in several cases of situations where either mass spectral or infrared analysis alone would yield, at best, ambiguous results. For example, sample A was a mixture of the three isomeric xylenes which yield essentially identical mass spectra, although their gas-phase infrared spectra (Figure 2) show significant differences. The low signal to noise ratio for these spectra was due to peak broadening during the GC separation, which caused relatively low sample concentration in the light pipe. In contrast, sample F, the mixture of homologous alkanes, is simple to analyze by mass spectrometry (although a mass spectral library search program might not find it so), but the infrared spectra are almost indistinguishable. Figures 3 and 4 contain the mass and infrared spectra obtained for this sample. The other samples were comprised of a variety of representative simple compounds (e.g., sample B was a mixture of materials having aromatic rings and various functionality and sample E was a mixture of two very similar isomers, which we felt would be difficult for either technique to identify unambiguously). For all of these analyses, it was desired, in addition to testing the practicality of the direct-linked instrument combination, to evaluate the hypothesis that the results of joint use of separate spectral library searches would lessen the ambiguity of identification and obviate the need for manual consultation of more than one or two library spectra per “unknown”. Sensitivity was not being tested in these studies. Rather, the present focus is on the capabilities for unambiguous and accurate analysis in situations where sample quantity is not limiting. Sample C was used initially, to determine the proper split ratio, as described in the Experimental Section. That ratio was not changed throughout the subsequent analyses. Because we have not yet fully developed the supervisory software necessary to combine and interpret the results of on-line library searches, all data were obtained and stored (the infrared spectra as interferograms on magnetic tape, the mass spectra on magnetic disk) for subsequent library search and analysis. Evaluation of Search Results. In evaluating the library searches, which were carried out with the NOVA 4X for mass spectra and the Nicolet 1180 for infrared spectra, it is convenient to consider the three possible outcomes of each search, prior to comparing the results of the IR and MS searches. Match thresholds were chosen to produce the five best matches for every search, regardless of type (Le., IR or MS). The outcome of any particular search could be the absence of the sought-for material among the best matches, its presence, or (for the mass spectral searches, since the library contains duplicate spectra) its presence more than once. The search results are summarized in Table 11. In that table, an asterisk in either the IR or MS columns indicates the component listed adjacent appeared among the five best matches (although not necessarily as the best match). The numbers in the MS column indicate the number of duplicate spectra of the component which appear among the best five matches. A minus indicates no library spectrum corresponding to the component appears. A double asterisk indicates all components of the mixture appear in the search result. As might be expected, for sample A all mass spectral library searches yielded all mixture components among the best five matches. Similarly, sample F yielded all mixture components
1
115
p-XILLK
I;
9
1
a-XYLEM
81
for each of the IR searches. In neither of these cases did the complementary measurement (IR or MS, respectively) yield such an ambiguous answer. Neither search yielded isomer identification for the components of sample E. In only one case did neither search include the correct identification among the best matches. That was sample D (where, however, the mass spectral search yielded the isomer l-methylcyclo-
116
ANALYTICAL CHEMISTRY, VOL. 53, NO. 1, JANUARY 1981
"ii
n
7
"1
,
N-HEXWE
83
m
m e 0
50
Kx,
iio
mh
N-tEPTIWE
'i" w
90
100
Fhwe 4. Gas-phase FT-IR spectra of pentane, hexane, and heptane at 4-cm-' resolution. 110
m Ir
Flgure 3. Low-resolutionGC-mass spectra of pentane, hexane, and
heptane.
pentene). The IR library did not contain 3-methylcyclopentene. In several casea, multiple spectra matches in the MS search would strongly suggest the correct identification,but
ANALYTICAL CHEMISTRY, VOL. 53, NO. 1, JANUARY 1981
Table 11. Infrared and Mass Spectral Library Search Results" sample
components
A
p-xylene m-xylene o-xylene anisole acetophenone methyl o-toluate methyl salicylate benzene toluene p-xylene 3-methylcy clopentene cyclohexene trans, trans2,4-hexadiene cis, trans2,4-hexadiene n-pentane n-hexane n-heptane
B
C D E
IR
MS combinedb
* * *
** ** **
*
4 2
*
+ + + N + +
* * * *
4
+
5 5
+ + +
*
**
N
* ** ** ** ** **
+
2
** **
N N
4 + 3 + 2 + " Asterisk indicates compound among five best matches; double asterisk indicates all mixture components were among five best matches; minus indicates compound not among five best matches; a number indicates the number of duplicate spectra of the compound among five best N indicates ambiguous or erroneous identifimatches. cation. Plus sim indicates correct identification. F
, 100
7
,075
3242
- 3230
I
,050 ,025
,000 1748
-
1736
,100 ,075 1706
-
1694 ,050
.ooo 'OZ5 1606
-
1594
u n :rij 11, ,100
,
,
,~
,~
,
600
72c
,025
.ooo
im
240
360
480
840
960
TIME, SECONDS
Figure 5. A chemigram of sample B during analysis, using functional group specific windows.
for a slightly larger number this would not be so. Consideration of the joint application of IR and MS search results (listedin the column labeled "Combined") is revealing. When the criterion applied is that the identification of a component is made when one and only one of the members of the respective sets of best matches is common, 12 of the 16 compounds are correctly and unambiguously identified.
117
The four ambiguous cases (marked N in Table II) are, in every case,quite close. For example, although anisole did not appear among the IR search results, because it was not contained in the library, ethoxybenzene was the closest match. Further, in the only case where IR and MS searches agreed on two compounds (sample E), the only error was the failure to distinguish the cis from the trans isomer. As mentioned before, sample D was the worst result, in that neither search included the correct component (even here, the MS search yielded a very similar isomer). Finally, the results from sample B, consisting of compounds of various functionality, illustrate the usefulness of the reconstructed gas chromatogram employing various windows for functional group assignment. The windows chosen (Figure 5) were selected to show higher specificity for alcohols (3230-3242cm-'), esters (1736-1748cm-'), ketones (1694-1706 cm-I), or ethers (1594-1606cm-I). The major peaks clearly reveal the elution order of the four components, reemphasizing the complementary aspect of the information obtained (in this case being of value in determining gas chromatographic elution order).
ACKNOWLEDGMENT We wish to thank Rick Lapp and Robert L. White for helpful discussions and Frank Crow for advice and technical assistance. We also wish to acknowledge the use of the facilities of the Midwest Center for Mass Spectrometry, supported under the National Science Foundation Regional Instrumentation Facilities program. LITERATURE CITED Low, M. J. D.; Freeman, S. K. Anal. Chem. 1987, 39, 194. Low, M. J. D. Anal. Left.1988, 1 , 819. Kker, K. L. Am. Lab. (F8lrfkM, Conn.) 1973, 5(6), 40. Azanaga, L. V. "GCIIR with Submicrogram Sensitlvity", Presented at 5th Annual Symposium on Recent Advances in Analytical Chemistry of Pollutants, Jekyil Island, OA, 1975. Wail, D.; Mantz, A. W. Appl. Specfrosc. 1977, 31, 525. Coney, P.; Mattson. D. R.; Wright, J. C. Am. Lab. (FairfjsM, Conn.) 1978, 70(5), 126. Krishnan, K.; Curbelo, R.; Chiha, P.; Noonan, R. C. J . Chromafogr. Scl. 1979, 17, 413. Mattson, D. R.; Juilan, R. L. J. Chromafogr. Scl. 1979, 77, 416. Kuehl, D. T.; Griffiths, P. R. And. Chem. 1980, 52, 1344. Low, M. J. D.; Freeman, S. K. J. Agric. FoodChem. 1988, 16, 525. Macnaughton, D., Jr.; Rogers, L. E.; Wemimont, G. Anal. Chem.
-.
1972. 44. ... . . , 1421. . .- . .
Uden, P. C.; Carpenter, A. P.. Jr.; Heckett, H. M.; Hendersen, D. E.; Slggla, S. Anel. Chem. 1979. 51, 38. JWS. P. C.: Kowalski. B. R.: Isenhow. T. L.: Reillev. C. N. Anal. C h m . 1989, 41, 1949. Naegell, P. R.; C k c , J. T. Anal. Chem. 1974, 46, 739A. Zupan, J. Anal. Chlm. Act8 1978, 103, 273. Uden, P. C.; Hendersen, D. E.;Uoyd, R. J. J. Chromfogr. 1976, 126, 225. Wliklns, C. L.; Oiss, 0. N.; Stelner. S.; Brissey, G. M.; White, R. L. Presented at the 22nd Rocky Mountain Conference on Analytical Chemistry; Denver, CO, Aug 1980 paper 28. Hkschfeld, T.; Crawford, R.; Wong, C.; Sanborn, R. Presented at the Second Chemical Congress of the North American Continent, Las Vegas, NV; Aug 1980; Symposium on Analy4cai Chemistry in the ' ~ O ' S , Paper 113.
RECEIVED for review September 10,1980. Accepted October 27, 1980. The support of the National Science Foundation under Grant CHE-79-10263is gratefully acknowledged.
