1858
Anal. Chem. 1983, 55, 1858-1862
Multichannel Imaging Spectrophotometer for Direct Analysis of Mixtures on Thin-Layer Chromatography Plates M. L. Gianelli,’ D. H. Burns, J. B. Callis,* C. D. Christian, and N. H. Andersen Department of Chemistry, BG-10,University of Washington, Seattle, Washington 98195
We descrlbe a slngle beam Imaging spectrophotometer whlch Is capable of recordlng the reflectance or transmlttance spectrum of up to 64 000 posltlons on an object, wlthln a few mlnutes. The instrument Is well suited to the analysis of mlxtures directly on thln-layer chromatography plates. The thin-layer plate Is lllumlnated unlformly wlth monochromatlc light and the resulting transmission Image is recorded by means of an SIT vidlcon Interfaced to a dlgltal computer. For complex mlxtures, a sequence of Images may be recorded at varlous wavelengths of Illumlnatlon. For ease of use, real-tlme displays of the thln-layer Image and absorptlon spectra at selected reglons on the plate are provlded. We show that the data from this Instrument can be represented as a three-dlmenslonal array whlch consists of the outer product of two elutlon vectors and one spectroscoplc vector. Processlng of these data can be accornpllshed In several ways, lncludlng quantltatlve analysls of known components via least squares or rank annlhllatlon, as well as qualltatlve analysls of unknown components via “factorlratlon”, derlvlng the spectrum and R, value of each. Prellmlnary results show that selected components of complex mixtures can be readily quantlfied at the submlcrogram level, even in the case where chromatographlc resolution Is Incomplete.
Touchstone and Sherma (1) have provided an excellent account of scanning densitometry as a means for quantitation of materials directly on thin-layer plates. In the past, densitometric scanning has been performed by mechanically displacing the plate past a fixed measuring aperature. A number of commercial densitometers based upon mechanical scanners are avaiable and perform quite effectively on onedimensional thin-layer chromatographs and gels. However, this approach can become exceedingly tedious for scanning two-dimensional plates and it is therefore worthwhile to consider alternative schemes. Four recent reports (2-5) illustrate the capabilities of vidicon-type image detectors. Aside from the elimination of mechanical scanning, a large multichannel advantage in sensitivity is also obtained (2). Furthermore, as Lemkin and Lipkin have shown (6), many algorithms developed for image processing are readily and efficaciously applied to analysis of these types of data. In many applications it is sufficient to scan the plate at one wavelength. However, for more complex systems, sensitivity for some components may suffer considerably. Moreover, in some cases it may be desired to obtain a complete spectrum for each component to improve capabilities for qualitative analysis. In this paper, we describe an instrument which allows us to rapidly obtain, in digital form, a complete absorption spectrum for every point on a thin-layer plate or gel in a few minutes. Furthermore, we show that qualitative and quantitative analyses of the data can be performed. ‘Present address: E. I. du Pont de Nemours, Inc., Wilmington,
DE
19898.
THEORY The data collected by our instrument consist of an ordered sequence of transmission images of the thin-layer plate, each at a specific wavelength. Thus, the resultant data may be represented as a three-dimensional array, T:
where I t i l k is the array element which corresponds to the intensity of the light transmitted through the plate at the plate location, x i , y,, at the kth wavelength. Provided that (a) the amount of light absorbed is very small so that Beer’s law may be assumed to be obeyed, (b) one has previously measured the light transmitted through a blank plate, and (c) the density of sampling is sufficiently high so that the distributional error is avoided (7), it is possible to obtain the absorbance array, A, according to
where I o i j h is the intensity of light transmitted through the blank plate at location x;,yj at wavelength k, and I t , , is the corresponding element for the sample plate. Under the above circumstances, eq 2 can be transformed into an expression which relates the observed absorbances to the number of molecules at each location on the plate, nip For a plate containing r noninteracting components it is possible to write
(3) where the index 1 enumerates the r components, d k is the optical cross-section a t wavelength k, and a is the area of the plate encompassed by the spatial coordinates i j . For one-dimensional chromatography, it is appropriate to integrate the absorbance along the y direction perpendicular to the migration axis. Furthermore, it is reasonable to assume that the relative distribution of the lth component along the migration axis is independent of the concentration of that component, as well as all others. Under these circumstances
where A* is a matrix formed from the array A by collapsing the j index by integration. Each matrix element A*ik is composed of three factors: a’, the number dependence of the integrated absorbance for the lth species; x l j , the position dependence of the relative number of molecules of the lth species along the x axis; and s ’ k , the relative absorbance of the Ith species. Thus, A* can be expressed compactly as 1=1
(5)
where the spatially sequenced set f1 ( x j } may be thought of as the normalized elution vector and the wavelength se( d k ) may be thought of as the normalized quenced set 8’ spectral vector; the symbol * denotes the outer product operation. The bilinear form of eq 5 lends itself well to analysis
0003-2700/83/0355-1858$01.50/00 1983 American Chemical Soclety
ANALYTICAL CHEMISTRY, VOL. 55, NO. 12, OCTOBER 1983
by powerful algorithms developed in this laboratory (8-12) for similar types of data sets, such as the emission-excitation matrix obtained by the video fluorometer (13), as well as GC/MS and LC/UV data. For quantitative analysis, when all of the components are knowns, for which standard chromatograms are available, the method of least squares is most appropriate (9). Here, A* is assumed represented by
A* =
@'[A*]'
(7) then we require that
where the Tjk values represent the weighting factors. Differentiation of eq 8 gives the least-squares condition
In the case where the mixture consists of unknowns, it is still possible to obtain sorne qualitative information about the number of components and their spectra by eigenanalysis (factor analysis) (8). Here, a set of "left" and "right" eigenvalues and eigenvectors are found from the following equations:
[A,*]TA* =
A = [Ajjk]
1 1
Aijk =
QX
1.1
$,S k
(12)
1=1
or more compactly t:hat (13)
1=1
where [A*]L is the standard matrix produced by species 1 and p1 is its amount divided by its standard amount. Then the best approximation to A:E in the least-squares sense requires that the weighted sum of the squares of the elements of the error matrix E be a minimum. That is, if
-k v
expressions which are formally similar to those of eq 4 and 5, Le., from eq 2 we can write that
A = Cg'al*9L*sl
1
1859
&jk
(loa)
where [A*ITis the transpose of A*, tkis the lzth eigenvalue, and vk and ah are the corresponding eigenvectors. As is well-known, the number of eigenvalues which are larger than the noise threshold is a reasonable approximation to the rank of A*. This quantity is in turn a sensible lower bound to the number of components in the mixture. Moreover, the 9 and are linearly related to the sought for elution and spectral vectors. In the case where there are only two components in the mixture, it is readily possible to find a transformation matrix which produces linear combinations of the ok and ah vectors which are reasonable approximations to the elution and spectral vectors. A more frequent analytical situation is the case where it is desired to extract a few knowns from a variable background of unknowns. Here, the method of rank annihilation is suitable (10-12). Again, we assume that A* is represented by eq 5 and that the rank of A* is equal it0 r, the number of components in the mixture. Then we try to find a coefficient y m such that the quantity [A*]Iemdefined by
[A*]rem= A* - r"[A*lI"
where the spatially (sequenced set 7' 3 (y\] expresses the normalized elution vector along the y axis, and u Lis a scalar containing all of the number dependence for the lth component. Appellof and Davidson (14, 15) have considered the analysis of data sets of the form of eq 12 and 13. In the caFie where all of the components are knowns, for which standard chromatograms are available, it is quite straightforward to m e the method of least squares to obtain estimates for the amounts of each component. The generalization of eq 6- 9 to the case of two-dimensional chromatography is trivial and will not be done here. In the case where none of the components are knowns, it is nevertheless possible to estimate the number of components, as well as the spectra and retention times of each (14). In this case, a linear decomposition of the array cannot be accomplished by means of eigenanalysis, as before, and a nonlinear least-squares method must be used. However, if a decomposition can be found, it can be shown to be totally uniqule, unlike the case of a two-dimensional matrix (8). Essentialby, we attempt to minimize the quantity R , defined by
(11)
is of rank r - 1 (has r - 1 finite size eigenvalues). [A*Irem is the matrix remaining after subtraction of the standard matrix, [A*]" for the mth component multiplied by ym,its estimated contribution to the observed matrix. At first, it might appear that rank annihilation would be a very computationally intense procedure involving repeated diagonalization of large matrices to search iteratively for the optimal value of ym. Fortunately it is possible to work in a basis set of the eigen spaces OS [A*ITA*and A*[A*IT,so that the algorithm is tractable (IO). For two-dimensional chromatography, if the relative distribution of the lth component along both axes is independent of the concentration of all the other components, we obtain
where the quantity 6, which has no meaning in the absenve of calibration, has been absorbed into the definition of x, .y, and s. Unfortunately, the parameters x, y , s, and r must tie obtained by iterative nonlinear methods which are not only computationally explensive but require a good starting estimate. However, as Appellof and Davidson have shown, it is possible to obtain a reasonable basis set for each of the three dimensions ( x , y , s) of the array from the eigenvectors of thle covariance matrix of' each dimension. We keep only that number of eigenvectors which are associated with sufficiently large eigenvalues. U&eof these basis sets greatly reduces the computational difficulties of the problem, as well as providos an excellent approximation to the number of components in the mixture. It is also possible to perform rank annihilation in three dimensions ((15).Again, the basic idea is the same as for one-dimensional chromatography, expressed by eq 11. As above, we work in a basis set consisting of the eigenvectors of the covariance matrix for each of the three dimensions of A. This renders the analysis computationally feasible. The interested reader is referred to the original exposition for further details (15).
EXPERIMENTAL SECTION Apparatus. A schematic of our single beam imaging spectrophotometer is given in Figure 1. A 20-W tungsten halogen lamp operated from a highly stable dc power supply served ats the light source. Wavelength selection was accomplished with a Bausch and Lomb 1/4 m monochromator (No. 33-2098) equipped with a 600 lines/mm grating blazed for 300 nm. Slits were set to provide a spectral bland-pass of 7 nm. Wavelength scanning was accomplished by means of a variable speed gear motor (Holtzer-Cabot). In order to uniformly illuminate the thin-laycbr plate, an image of the diffraction grating was projected on to the plate by means of a lens placed at the exit slit (field stop). Light transmitted through the plate was collected by an f/0.78 CCTV lens and focused onto the photosensitive surface of the SIT vidicon camera (Quantex QX. 10). Our digital television photometry
1860
-
ANALYTICAL CHEMISTRY. VOL. 55. NO. 12. OCTOBER 1983
u TUNGSTEN LAMP
V MONOCHROMATOR
1 7 1
TLC PLATE SHUTTER
T V MONITOR
POP 11/04
GRAPHICS TERMINAL
Figure 1. Block diagram of imaging spectrophotometer
system has been previously described (13). Software. The camera was operated to collect a sequence of images synchronously with the scanning of the wavelength drive of the monochromator. The software for this operation was developed hy Hershbsrger et al. (16) for another purpose hut was readily modified to serve the present need. Real-time displays of the monochromatic plate image and the absorption spectrum at two locations on the plate are available. Data consisting of a series of images over a selected wavelength region are stored on disk for later postrun processing, either by the PDP 11104 or by a remote VAX 11/780. Reagents. The chromatographically pure free tetraphenylporphine (H,TPP), zinc tetraphenylporphine (ZnTPP), and palladium etioporphine (PdEP) were the kind gift of Martin Gouterman. For spotting on the thin-layer plates, they were dissolved in spectrcquality dichloromethane (Matheson, Coleman and Bell). Solvents used for chromatography were spectrograde quality. Chromatography Procedure. All one-dimensional chromatography was carried out by developing the plate in a dichloromethaneJcyclohexane (5050) mixture for 3.5 min. Two-dimensional chromatography was carried out by first developing the plate for 3.5 min with ethyl acetate and then 4.0 min with the dichloromethane/cyclohexanemixture. Whatman LHP-K plates, high-performance quality, were used for all separations. RESULTS AND DISCUSSION Our first series of studies was designed to explore the reproducibility of the data. First, an image of a blank plate illuminated a t 420 nm was obtained and stored. Then the plate was spotted with 50 ng of zinc tetraphenylporphine (ZnTPP) in several places, and a second image was collected. From these two images, the absorbance image was computed according to eq 2. This experiment was repeated six times with six thin-layer plates. For a single spot, repeated measurements of the integrated absorbance were performed with a relative standard deviation of less than 1%.The spotto-spot variation for spots on the same plate exhibited a relative standard deviation of less than 2%, while the spotto-spot variation for spots on different plates were less than 5%. We next explored the dynamic range of the system and the capability to analyze several one-dimensional chromatograms simultaneously. The data for this are presented in Figure 2. A plate was spotted with various amounts, in the low microgram range, of H,TPP porphine in five separate lanes (1.1, 2.2,3.3,4.4, and 5.5 pg, respectively, right to left in 2A). In lane 3, a mixture of H,TPP along with ZnTPP and PdEP was spotted. Figure 2A shows the raw transmission image of the sample plate and a blank plate, respectively, a t 420 nm, and Figure 2 8 the absorbance image of the sample as an isometric projection, which is useful for visualizing the quantitative
Flgure 2. Use of the imaging spectrophotometer for anab'zing several onedimensional chromatograms simultaneously.
relationships among the absorbances of the spots. Figure 2C compares the spectrum of H,TPP from the spot in lane 1with the spectrum of the same spot obtained on a Cary 219 spectrophotometer, illustrating good agreement. Finally, Figure 2D is a plot of the integrated absorbance vs. the (known) amount of H,TPP in each lane. At 580 nm there is an excellent linear correlation between observed absorbance and quantity of porphyrin, whereas at 420 nm we observe a leveling off of observed absorbance at high concentrations. This latter phenomenon is attributed to the higher absorptivity of the porphines a t 420 nm which leads to high absorbances; consequently greater photometric errors are expected. Sources of photometric error include stray light in the ancient single monochromator we used, as well as "glare" from scattered light propagating through the highly opaque thin-layer plate. Both of these sources would tend to increase the apparent transmission of the spots and thus lead to underestimation of the absorbance. Indeed, independent experiments showed that good linearity could he obtained a t 420 nm when smaller amounb of porphyrins were applied to the plate. For example, a linear calibration plot was obtained for 5-30 ng of ZnTPP. In the case of transmission measurements on thin-layer plates then, we can he sure that the photometric system does not limit performance. Previous studies showed that with signal averaging, each pixel of the SIT camera could achieve a signal-to-noise ratio of over 500 to 1,as well as a dynamic range of 10000 to 1 (13). The multicomponent chromatogram presented in Figure 2 (lane 3) was also analyzed qualitatively and quantitatively according to the procedures outlined in the theoretical section. For this analysis, the z axis information was collapsed by integration, leaving the matrix A* according to eq 4 and 5. Rank analysis of these data shows that there are three independently absorbing components. We next attempted to estimate the amount of tetraphenylporphine in lane 3 by the methods of least squares and rank annihilation. As the standard we use the data from lane 4. Least squares, assuming one component, gave 2.8 pg and rank annihilation gave 3.0 pg, while the correct result was 3.3 pg. The fact that rank annihilation gives a more satisfactory answer is hardly sur-
ANALYTICAL CHEMISTRY, VOL. 55. NO. 12. OCTOBER 1983 * 1861
-
+- % 554
%
5 9 1 n.
"l
Figwe 3. Two-dimensionalchranatogaphy contour plots of a mixblre of H,TPP. ZnTPP. and PdEP at selected absorption wavelengths: (A) 420 nm. (E) 523 nm, (C) 554 nm. end (0) 594 nm.
Table I. TweDimensional Chromatography Data Analyzed by Rank Annihilation, RA, and Least Squares, LS present H,TPP ZnTPP PdEP
1.40 0.91
1.31
analysis, Irg RA 1.51 0.86 1.44
LS 1.69 0.13 1.46
prising in view of the high degree of overlap of the spots. Our final data set is shown in Figure 3 and was designed to test the instrument's capabilities for two-dimensional analysis. Shown are several absorbance images of the plate taken a t various wavelengths for a mixture of three porphines. Especially striking is the degree of differential enhancement of the selected (compound) spots as a function of wavelength of illumination. In Figure 3A, the plate is imaged a t 420 nm where all of the porphyrins absorb. The next three images were taken at 523 nm, 554 nm, and 594 nm; HzTPP absorbs strongly a t 523 nm, PdEP a t 554 nm, and ZnTPP a t 594 nm. In Figure 3B,then, the main contour peak is due to HzTPP, while there is a weaker peak to the left of it due to weak ahsorption by PdEP. Likewise in Figure 3C the left P d E P peak is strong and the right H z T P P peak somewhat diminished. In Figure 3D,only the ZnTPP is seen. Quantitative analysis of this mixture is summarized in Table I using rank annihilation and least squares. As with the one-dimensional chromatograms, two-dimensional chromatograms can he analyzed quantitatively even when the spots are severely overlapping. Figure 4 illustrates the result of qualitative analysis of the two-dimensional chromatogram hy decomposing it according to the algorithm detailed in ref 15. The input data, assumed to he of the form of eq 6,are shown as a "stack" of images in Figure 4, typical examples of which are displayed. The decomposition of these data into their three components is illustrated in Figure 4B-D. The absorption spectrum and the retention profiles along the x and y axes are all shown for each of the three components. The spectra shown superimposed nearly exactly on those of the pure components. Imaging spectrophotometry of thin-layer plates is yet another example of the benefit of using a "hyphenated" method (17)for multicomponent analysis, i.e., a separation technique coupled with a spectroscopic technique. Our instrument
Flgure 4. Threedimensional rank analysis: (A) stack of all t w d i mensional chromatography images as a function of wavelength: (E-0) individual spatial-spectral components (E = P d E P C = H,TPP D =
ZnTPP). Each three-dimensionalplot is a spatial representation of the associated compound (spechum) obtained by factorization of lhe stack of chromatography images. compares favorably with the GC/MS in that the operator has great flexibility for display and analysis of the data: plotting the absorbance spectrum of a selected spot is analogous to plotting the mass spectrum of a selected GC peak, while displaying the transmission image of the plate a t a selected wavelength is analogous to displaying a selected ion chromatogram. Although the resolution of capillary GC is much higher than that of conventional one-dimensional TLC, the great flexibility inherent in and the ease of implementation of two-dimensional Chromatography may in some cases provide a compelling reason to prefer TLC with ahsorhance detection, especially when the low cost per analysis, parallel processing advantage, and reliability of the instrumentation are taken into account. We believe that multichannel detectors deserve careful consideration for use as scanning devices for thin-layer plates. With our SIT vidicon based system, the entire plate may be scanned in a time as short as 17 ms, although typically we use several seconds to coadd several hundred frames for S I N improvement. The present SIT vidicon detector could easily he replaced by a far less expensive silicon target vidicon or photodiode array, especially under circumstances where the reflection imaging mode is to he employed. Thus, a system based upon the use of an electronic imaging device need not he overwhelmingly expensive, especially with the advent of low cost read/write mass storage devices for the video images. Clearly, the ability to quickly acquire absorption data for each spot will greatly increase the power and versatility of thin-layer chromatography. ACKNOWLEDGMENT We gratefully acknowledge useful discussion with M. Gouterman and Bruce Skoropinski. LITERATURE C I T E D C.: Sherma, J. "Denshometty
(1) Touchstone. J.
In Thin Layer Chromatography": W h y : New York. 1979. (2) Gianelli, M. L.: Callis, J. E.: Andersen. N. H.: Christian, G. D. Anal. Chem. 1981. 53. 1357-1361.
1862
Anal. Chem. 1983, 55, 1862-1869
(3) Kramer. J.; Gusev, N. B.: Friedrich. P. Anal. Biochem. 1980. 708. 295-298. (4) Curtis, T. G.; Seitz, W. R. J . Chromatogr. 1977, 734,513-516. (5) Lester, E. P.;Lemkin, P. F.; Lipkin, L. E. Anal. Chem. 1981, 53, 391A-404A. Lemkin,b.; Lipkin, L. Comput. Biomed. Res. 1981, 776, 441-445. Ornstein, L. Lab. Invest. 18, 7, 250-265. Warner, I. M.;Christian, G. D.; Davidson, E. R.; Caiiis, J. B. Anal. Chem. 1977, 49,564-573. (9) Warner, I. M.;Christian, G. D.; Davidson, E. R. Anal. Chem. 1877, 49,2155-2159. (10) Ho, C. N.; Christian, G. D.; Davidson, E. R. Anal. Chem. 1978, 5 0 , 1108-1 113. (11) Ho, C. N.; Christian, G. D.; Davidson, E. R. Anal. Chem. 1980, 52, 107 1- 1079.
(12) Ho, C. N.; Christian, G. D.; Davidson, E. R. Anal. Chem. 1981, 53, 92-98. (13) Johnson, D. W.; Gladden, J. A.; Caiiis, J. B.; Christian, G. D. Rev. Scl. Instrum. 1878, 5 0 , 118-126. (14) Appellof, C. J.; Davidson, E. R . Anal. Chem. 1981, 53, 2053-2056. (15) Appeliof, C. J.; Davidson, E. R. Anal. Chlm. Acta 1983, 746,9-14. (16) Hershberger, L. W.; Caiiis, J. B.; Christian, G. D. Anal. Chem. 1981, 53, 971-975. (17) Hirschfeid, T. A. Anal. Chem. 1980, 52, 297A-312A.
RECEIVED for review February 22, 1983. Accepted June 29, 1983. This research was supported by NIH Grants GM-22311 and GM-26935.
Determination of Aniline and Substituted Derivatives in Wastewater by Gas and Liquid Chromatography Ralph M. Riggin,* Thomas F. Cole, and Stephen Billets’ Battelle, Columbus Laboratories, 505 King Avenue, Columbus, Ohio 43201
A detalled evaluation of gas chromatography (GC) and highperformance liquid chromatography (HPLC) methods for the determlnatlon of anilines In aqueous media has been conducted. An optlmlred analytical approach based on GC with thermionic nitrogen-phosphorus selective detection (NPD) Is described. Thls method is capable of determining a wlde variety of anilines at the low part-per-billlon level In industrial aqueous discharges, as well as effluents from publlcly owned treatment works (POTW). Method performance data for authentic envlronmental samples are presented. Anaiytlcal preclslon was generally 5-15% RSD and recoveries were generally 75% or better.
Aniline and its substituted derivatives (referred to as “anilines” in this paper) are widely produced for a variety of industrial and commerical purposes, including dyestuff and pesticide manufacturing. Since these compounds have a significant water solubility, they are often present in wastewater discharges from such manufacturing facilities. The toxic nature of anilines dictates that their discharge concentrations be controlled, hence requiring the use of a reliable analytical method for determining anilines in such aqueous effluents. A variety of analytical methods have been reported for the determination of selected anilines. Such analytical techniques have included: GC (1-6), HPLC (7, 8 ) , colorimetry (9), thin-layer chromatography (TLC) (lo),and spectrophotometry (11). In general only the GC and HPLC approaches have been sufficiently sensitive and selective to detect anilines in complex environmental media. A variety of GC detectors have also been studied, including electron capture detection (ECD) ( 4 , 5 ) ,nitrogen-phosphorus thermionic detection (NPD) (22),and Hall electrolytic conductivity detection in the nitrogen selective mode (HECD) (6). The HPLC detection systems which have been investigated include ultraviolet (UV) and electrochemical (EC) detection (7, 8). While these methods have been successfully employed for a few selected anilines, none of these procedures is directly Present address: United States Environmental Protection Agency, Environmental Monitoring and Support Laboratory, $AD, P.O. Box 15027, Las Vegas, NV 89114.
useful for the determination of a broad spectrum of aniline compounds. This limitation is the result of chromatographic separation, detection, and/or sample preparation schemes which were streamlined for the few compounds of interest in a particular study. The United States Environmental Protection Agency (EPA) has as part of its mission the responsibility for providing test procedures for organic pollutants which would be monitored under such regulations as the Clean Water Act or Toxic Substances Control Act (TSCA). Since many of the anilines have been placed on the priority list for testing under TSCA (23),the need arose for an analytical procedure for determining a wide variety of anilines in industrial wastewater effluents. Compounds of interest included aniline and its bromo-, chloro-, and nitro-substituted derivatives. Compounds containing functional groups other than these (e.g., aminophenols, phenylenediamines) were not included. This report describes the development and evaluation of GC and HPLC approaches with selective detectors for analyzing trace concentrations in selected industrial wastewaters.
EXPERIMENTAL SECTION Reagents. All organic solvents were “distilled in glass” grade from Burdick and Jackson. Reagent water was obtained from a Millipore Milli-Q water purification system consisting of reverse osmosis, mixed bed ion exchange, and activated charcoal purification modules. Chemical standards of the various anilines were purchased from Aldrich Chemical or Eastman Chemicals and were the highest purity available. Compound identities and purities were confirmed by GC/MS with 70-eV electron impact ionization, HPLC, and melting point determination. HPLC mobile phases were prepared by mixing the appropriate volumes of organic solvent and aqueous buffer. The mixture was filtered through a Nucleopore 0.2-ym polyester filter, degassed by heating t o near boiling in a covered Erlenmeyer flask, and placed in a Teflon-capped glass bottle to cool. Alumina (Fisher Adsorption Grade, 80-200 mesh) and basic alumina (Brockman Activity I, 8C-200 mesh) were purchased from Fisher Scientific. Florisil (PR Grade, 60-100 mesh) was purchased from Sigma Chemical Co. Apparatus. Most gas chromatographic studies were performed by using a Hewlett-Packard 5730 system equipped with both packed and capillary column capabilities as well as flame ionization (FID) and thermionic nitrogen-phosphorus (NPD) detectors. A Tracor Model 560 capillary GC system equipped with a Hall electrolytic conductivity detector (HECD) in the nitrogen selective mode and a Carlo Erba Model 2150 capillary GC system equipped
0003-2700/83/0355-1862$01.50/00 1983 American Chemical Society
