998
Anal. Chem. 1983, 55, 998-1001
plasma and most of the active sites on the glass are perhaps covered. Consequently, there is low probability for small amounts of tricyclic drugs to be selectively adsorbed on the glassware. During the second step, the residue obtained after evaporating organic solvent is reconstituted with mobile phase containing n-butylamine. The amine concentration is at least 1 order of magnitude higher than that of tricyclic drugs, and perhaps the amine modifier saturates the active sites on the glass. These conditions then allow for nearly quantitative recovery of all drugs, which other methods have not achieved. In the drying process, some of the organic amines are more susceptible to degradation in free base form than in their salt form. Therefore, we do not suggest that the extracts remain at dryness for more than a few minutes after the organic solvent has been evaporated. We used an air or nitrogen stream to aid evaporation, and we have not found drug loss at the room temperature evaporation.
ACKNOWLEDGMENT We thank all of the drug manufacturers who provided the pure drugs as gifts. We also thank the People’s Republic of China for the opportunity of Shude Yang, a visiting scholar on leave from the Chinese Academy of Medical Sciences, Peking, China, to come to the University of Wisconsin, Madison. Registry No. Amoxapine, 14028-44-5;8-hydroxyamoxapine,
61443-78-5;doxepin, 1668-19-5;desmethyldoxepin, 1225-56-5; imipramine, 50-49-7;desipramine, 50-47-5;amitriptyline, 50-48-6; nortriptyline, 72-69-5.
LITERATURE CITED (1) Scogglns, B. A.; Maguire, K. P.; Norman, T. R.; Burrows, G. D. Ciin. Cbem. (Winston-Salem, N . C . ) 1980, 2 6 , 5-17. (2) Gapta, R. N.; Milnar, G. Biopharm. Drug Dispos. 1980, I , 259-278. (3) Vandernark, F. L.; Adarns, R. F.; Schrnldt, G. J. Clin. Chem. (WinstonSalem, 1978,, 2 4 , 87-91. (4) Wallace, J. E.; Shlrnek, E. L., Jr.; Harris, S. C. J . Anal. Toxicoi. 1981, 5 , 20-23. (5) Proeles, H. F.; Lohman, H. H.; Miles, D. G. Ciin. Cbem. (Winston-Saiem, N.C.)1978, 2 4 , 1948-1953. (6) Thorna, J. J.; Bondo, P. B.; Kozak, C. M. Tber. Drug Monit. 1979, 7 , 335-338. (7) Bannlster, S. J.; VanderWal, S.J.; Dolan, J. W.; Snyder, L. R. Clin. Cbem. (Winston-Salem, N . C . ) 1981, 2 7 , 849-855. (8) Sonsalla, P. J.; Jennison, T. A,; Flnkle, 8. S. Clin. Chem. (WinstonSalem, N . C ) 1982, 28, 457-481. (9) Koteel, P.; Mulllns, R. E.; Gadsden, R. H. Ciin. Cbem. (Winston-Salem, N.C.)1982, 28, 462-466. (10) Kabra, P. M.; Mar, N. A.; Marton, L. J. Ciin. Cbim. Acta 1981, I l l , 123- 132. (1 1) Smyly, D. S.; Woodward, B. B.; Conrad, E. C. J . Assoc. Off. Anal. Cbem. 1976, 59, 14. (12) Snyder, L. R.; Kirkland, J. J. “Introduction to Modern Liquid Chromatography”, 2nd ed.; Wlley: 1979; p 270. (13) Tasset, J. J.; Hasson, F. M. Ciin. Chem. (Winston-Salem, N . C . ) 1982, 28, 2154-2157.
RECEIVEDOctober 5 , 1982. Resubmitted March 2, 1983. Accepted March 2, 1983.
Effects of Interferogram Sampling of Gas Chromatography/Fourier Transform Infrared Data on Gram-Schmidt Chromatogram Reconstruction Robert L. Whlte, Gary
N. Glss, Gregory M. Brlssey, and Charles L.
Wllklns”
Department of Chemistry, University of California -Riverside, Riverside, California 9252 1
A theoretlcal and experlmental study of the effect of changlng Interferometric sampling of gas chromatography/Fourler transform Infrared (GC/FT-I R) Gram-Schmldt reconstructed gas chromatograms (GC) on peak signal-to-noise ( S I N ) rallo and relative peak heights Is presented. I t Is determlned that no generally optimum vector dlsplacernent exists because identity of the GC eluents and instrumental instablllty determlne whlch portion of an Interferogram will produce the greatest GC/FT-IR sensitlvlty.
In 1977, de Haseth and Isenhour reported the first application of the Gram-Schmidt vector orthogonalization procedure to reconstruction of gas chromatography/Fourier transform infrared (GC/FT-IR) chromatograms (1). The major advantage of the technique is that chromatograms can be calculated by using time-domain interferometric data and the time-consuming process of Fourier transformation of each acquired data file can be avoided. Because only a small portion of the interferometric data is used in Gram-Schmidt reconstructions, de Haseth and Isenhour attempted to determine which portion of the raw data produced the best chromatograms. Using data acquired during a single GC/ FT-IR separation, they empirically determined an optimum
vector sampling for 2048-point interferograms to be a region starting 60 data points to the right of the centerburst position and extending right for 100 data points. Subsequently, GC/FT-IR manufacturers provided users with Gram-Schmidt reconstruction software employing this vector sampling. In addition, Isenhour and co-workers have shown that quantitative analysis is possible by using Gram-Schmidt GC/FT-IR reconstructions and that the empirically determined “optimum” interferogram region can be useful for inteferogram-based library searches (2, 3 ) . In 1981, we reported that simplex optimization of four test mixture Gram-Schmidt GC/FT-IR reconstructions resulted in a different optimum vector displacement for each separation when average SIN for the chromatograms was maximized (4). In each instance, the optimum vector sampling included data in the vicinity of the centerburst. This was inconsistent with the earlier report by de Haseth and Isenhour which suggested the optimum sampling was 60 data points away from the centerburst. Comparison of simplex-derived optimum chromatograms with chromatograms generated by using data 60 points away from the centerburst position revealed that relative chromatographic peak heights for mixture component eluents varied when the vector sampling region was changed. These results indicate that a generally applicable optimum vector sampling for Gram-Schmidt reconstructions may not
0003-2700/83/0355-0998$01.50/00 1983 American Chernlcal Soclety
ANALYTICAL CHEMISTRY, VOL. 55, NO. 7, JUNE 1983
a
999
relative to the centerburst position. For the purpose of discussion, it will be assumed that each of the n peaks comprising the absorbance spectrum has a Lorentzian line shape. With this assumption, a particularly simple time domain function can be written to represent the contribution from each absorbance peak to the absorption interferogram
P(6), = Ace-@ cos (27rq8)
(3)
where A, is the amplitude of the time domain oscillation for the ith peak at zero mirror retardation (centerburst position) and p, is a time domain damping constant for the ith peak. Summation of the contributions from all n peaks yields an expression for the absorption interferogram n
I(@ = CAce-@cos ( 2 ~ 2 ~ 6 ) L=l
(4)
Fourier transformation and magnitude calculation of eq 4 yield (to a close approximation) -0324
0
0.324
MIRROR RETARDATION (mm) Flgure 1. Representative interferograms obtained during a typical GC/FT-IR experiment: (a) background interferogram; (b) interferogram containing ethyl acetate absorption information; (c) the result of subtracting file b from file a.
exist. Instead, an optimized vector displacement may depend upon both the iiample mixture separated and the stability of the GC/FT-IR instrument used. As with all GC/FT-IR chromatogram-generating algorithms, the object of the Gram-Schmidt process is to distinguish between background data files and data files containing sample absorption information. Ideally, the difference between two background data files should be random noise and the difference lbetween an interferogram containing sample absorption information and a background data file should be an absorption interferogram. Figure 1 shows a background interferogram, a sample interferogram, and a difference interferogram formed by subtracting the sample file from the background file. Because the difference interferogram theoretically contains only sample absorption information (referred to as the absorption interferogram in this paper), the degree to which a chromatogram reconstruction method is successful is dependent upon how readily the absorption interogram can be detected. I n this paper, the Gram--Schmidt GC/FT-IR reconstruction method is evaluated with respect to chromatogram S I N and relative eluent peak height as a function of interferogram vector displacement. Theoretical and experimental arguments are presented to establish that a universal optimum vector sampling does not exist. THEORY A low-resolution (4cm-l) infrared absorbance spectrum can be resolved into a sum of frequency domain peak shapes of various amplitudes and widths, spanning the frequencies of sample absorption. A mathematical function of frequency, BJ), can be derived which represents a typical gas-phase absorbance spectrum as the summation of n peaks n
B(v) = c P ( J ) c c=1
(1)
where P ( J ) ,is an absorbance peak centered at frequency nC. A time domain absorption interferogram (analogous to Figure IC)can be representeld by
I(6) = k P ( 6 ) ; i=1
where 6 is the interferometer moving mirror retardation
The intensity of each line shape expression maximizes at J = vi to give
and the full-width-at-half-height of each peak is given by
-
(7) It is evident from eq 7 that narrow frequency domain spectral components produce lightly damped time-domain oscillations (small p,) and wide band spectral features are represented by more heavily damped oscillations (large p,). When a combination of wide and narrow spectral features are observed in the time domain simultaneously (such as in FTIR), the relative contribution from each type of feature will change as a function of time (or distance for an interferogram). For example, near the centerburst position, both narrow and wide band information can be sampled. Far from the centerburst, primarily narrow spectral features are represented (Figure 2). Therefore, it is expected that relative reconstructed chromatographic peak heights for eluents with different IR spectra should vary when the Gram-Schmidt vector sampling is changed. Also, because spectral information content of an absorption interferogram is theoretically greatest in the vicinity of the centerburst position, Gram-Schmidt reconstructions using this region would be expected to yield the highest chromatogram SIN. In practice, interferometer and source instability may introduce so much noise in the reconstructed chromatogram that another region of the interferogram becomes the optimum vector sampling. Instrumental differences are probably the cause of the conflict between our results (4)and those of de Haseth and Isenhour (1).
EXPERIMENTAL SECTION The GC/FT-IR system used for this study was a Nicolet 7199 FT-IR spectrometer coupled to a Varian 3700 gas chromatograph. A 35 m, 0.44 mm i.d. glass Carbowax 20M support-coated open tubular (SCOT) capillary column was used in series with a 0.159 m X 2 mm i.d. gold-lined glass lightpipe interface for the separation in Figure 4. The GC separation was carried out at 45 "C isothermal and column flow rate was 6 mL/min. The sample consisted of 0.I-pL splitless injection of a mixture of pentane, heptane, and ethyl acetate in a volumetric ratio of 7:2:1, respectively. Data acquisition consisted of digitizing 4096 data point
1000
ANALYTICAL CHEMISTRY, VOL. 55, NO. 7,JUNE 1983
a
bl
d
C
!
b
,I -0
215
4300
215
4300
2d
430 0
U5
430
TIME (sed
I
0
TIME
-
Figure 2. Illustration of the time domain sum of frequency peaks having different line widths: (a) lightly damped cosine function; (b) heavily damped cosine function; (c)the sum of a and b. The contrlbutlon of the heavily damped function to c diminishes as time Increases.
Flgure 4. Comparison of GC/FT-IR chromatogram reconstructions of a separation of pentane, heptane, and ethyl acetate: (a)total integrated absorbance; (b) Gram-Schmidt reconstruction based on a 100 data point vector sampling symmetrical about the centerburst; (c) GramSchmidt reconstruction with a 100 dimensional vector displaced 60 data points; (d) Gram-Schmidt reconstruction wlth a 100 dimensional vector displaced 120 data points. Table I. Comparison of Chromatogram SIN for Reconstructions in Figure 3
‘“1
501
reconstruction integrated absorbance Gram-Schmidt centerburst symmetrical sampling Gram-Schmidt 60 data point centerburst displacement Gram-Schmidt 120 data point centerburst displacement
I(
1
0 7 , 4000 2900 I800
700
WAVENUMBERS
Flgure 3. Frequency domain spectral components derived from Fourier transformation and magnitude calculation of an absorption interferogram derived from 2-methyl-2ethyl-3-hydroxybutyrlc acid, ethyl ester, using: (a) a portion of the interferogram located 120 data points to the right of the centerburst and extending right for 200 data points, (b) a 200 data polnt sectlon sampled symmetrically about the centerburst, and (c) the entire 4096 data polnt absorptlon interferogram.
interferograms which could later be transformed to yield 4 cm-l IR spectral resolution. All calculations were performed with the Nicolet 1280 data system supplied with the Nicolet 7199 FT-IR.
RESULTS AND DISCUSSION One of the observations reported earlier (4)was that the relative chromatogram peak heights generated by using Gram-Schmidt reconstructions with data including the centerburst position were consistently similar to peak heights derived from total integrated absorbance reconstructions. In contrast, reconstructions using interferometric data 60 points displaced from the centerburst position produced chromatogram peak heights which were much different from total integrated absorbance peak heights. These observations can be attributed to the interferogram property that all spectral features are represented in the vicinity of the centerburst whereas narrow spectral features predominate in interferogram
signal to noise ratio ethyl pentane heptane acetate 65.9 118.0
14.4 35.3
10.8 24.7
58.3
25.3
58.7
41.5
14.8
43.8
regions displaced from the centerburst. To illustrate this with an example, frequency domain spectral representations were compared from interferometric data extracted from a centerburst symmetrical sampling and a sampling displaced from the centerburst position by 120 data points (Figure 3). The absorption interferogram contained both broad and narrow spectral features and was generated by subtracting an interferogram containing sample information from a background interferogram. Clearly, the centerburst symmetrical sampling in Figure 3b more closely follows the outline of the spectral features in Figure 3c than does the plot in Figure 3a. The spectrum obtained from the centerburst-displaced sampling contains exaggerated contributions from narrow spectral features (Figure 3a). Figure 4 shows the effect of the choice of interferogram vector sampling region on relative Gram-Schmidt reconstructed peak heights. The separated components in order of elution were pentane (solvent), heptane, and ethyl acetate. As expected, the Gram-Schmidt reconstruction using centerburst data resembles the total integrated absorbance chromatogram. However, the relative peak height for ethyl acetate is greatly enhanced when a displaced vector sampling is used for chromatogram reconstruction (Figure 4c,d). The S I N of the ethyl acetate peak in Figure 4c,d is greater than the corresponding peak in Figure 4b while the S I N values of the alkane peaks are considerably lower in Figure 4c and Figure 4d than in Figure 4b (see Table I). This behavior was expected since the infrared spectrum of ethyl acetate contains many more narrow features than does an alkane IR spectrum (Figure 5). Furthermore, the increase in SIN for ethyl acetate
Anal. Chem. 1983, 5 5 , 1001-1004
:‘11“”
Lo
CONCLUSION
la
ACETATE
li
:-I _
m
I
’HEPTANE
1001
b
m
The optimum portion of an interferogram used as a vector for Gram-Schmidt orthogonalization reconstructions is heavily dependent upon the identity of the mixture componc?nts (particularly their spectral peak widths) as well as GC/FT’-IR instrument stability. When substances having similar IR absorbance spectra are detected (i.e., isomers), little difference in relative chromatogram peak height is observed when different vector displacements are used (4). However, when GC eluents have different absorbing properties, relative peak heights are dependent upon the interferogram vector sampling. Therefore, if one intends to maximize Gram-Schmidt reconstruction sensitivity for a given substance, an optimum interferogram sampling must be determined. A universally applicable optimum Gram-Schmidt vector displacement simply does not exist. However, a best “compromise” value based upon the considerations delineated here might well exist. From this and previous results (4) it does not appear that the 60-point displaced choice routinely used in commerical software is that hest compromise. In a collaborative research effort, we are currently investigating interferometer stability of various commerical GC/FT-IR instruments and its effect on optimum Gram-Schmidt interferogram vector sampling. The results of this study will be published at a later time.
LITERATURE CITED (1) de I
g=----J 4000
--’3200
2400
1600
800
W AVENUMBERS
Figure 5. (a) Gas-phase infrared spectrurn of ethyl acetate. (b) Gas-phase infrared spectrum of heptane. with increasing vector displacement clearly suggests a significant instrumental noise contribution since, in principle, S I N should decrease as data increasingly further from the centerburst are sampled.
Haseth, J. A.; Isenhour, T.
L. Anal. Chem. 1977, 4 9 , 1977-1981.
(2) Sparks, D. T.; Lam, R. 8.; Isenhour, T. L. Anal. Chem. 1982. 5 4 , 1922-1926. (3) Small, G. W.; Rasmussen, G. T.; Isenhour, T. L. Appl. Spectrosc. I W g , 33, 444-449. (4) White, R. l..; Giss, G. N.; Brissey, G. M.; Wilkins, C. L. Anal. Chem. I981, 53, 1778-1782.
RECEIVED for review January 6, 1983. Accepted March 17, 1983. The support of the National Science Foundation under Grant CHE-82-08073 is gratefully acknowledged.
Nonsuppressor Ion Chromatography of Inorganic and Organic Anions with Potassium Hydroxide as Eluent Tetsuo Okadia and ‘Tooru Kuwamoto” Department of Chemistry, Faculty of Science, Kyoto University, Sakyo-ku, Kyoto 606, Japan
The elution behaviors of lnorganlc anlans and weak organic acids were investlgatced witH nonsuppressed Ion chromatography. As an eluent, a potasslum hydroxlde solution was remarkably effective for the separatlon and determlnatlon of F-, CI-, Br-, NO,-, and NOs-. For SO?- It was necessary to use a higher concentration of potassium hydroxide solution. The detection Uimlts of F-, CI-, Br-, NO,-, and NO,- were 1.5 ppb, 2.5 ppb, 15 pplb, 15 ppb, and 15 ppb, respectively. Moreover, it was found that the potasslum hydroxide eluent was also appilcable to the separatlon and determination of weak organlc trclds, such as derivatlves of phenol or benzolc acid, carboxylic acids, and dicarboxylic acids.
The separation and determination of anions, tedious and intricate tasks in the past, have been easily carried out by using ion chromatography. For example, in several reports, the
determination of CN- in air ( I ) , the blow-down water from a boiler (2),aqueous solution after the collection of a trace level ion with concentrator columns (3) and alkali metal or alkaline-earth metal (4), transition-metal ions (5, 6), and amines (4, 7) have been described. Ion chromatography as developed by Small et al. (4) usually uses a separation column or membrane-exchanger or cation-exchanger column to remove most of the background conductance of the eluent with a carbonate buffer solution as an eluent. Gjerde et al. described a means to determine common inorganic anions with nonsuppressor type ion chromatography by using a phthalate or a benzoate solution as the low conductivity eluent (8-10). They also investigated the separation of alkali and alkalineearth metal ions with nitric acid as the eluent ( 1 1 ) . Similarly, the authors considered the disadvantages oE the use of suppressor type ion chromatography. In it, the solute ions are partially lost by the suppressor column owing to adsorption, and organic acids having a low dissociation con-
0003-;!700/83/0355-1001$01.50/00 1983 American Chemical Society
