Anal. Chem. 2003, 75, 4211-4216
Characterization and Quantitative Analysis with GC/TOFMS Comparing Enhanced Separation with Tandem-Column Stop-Flow GC and Spectral Deconvolution of Overlapping Peaks Tincuta Veriotti and Richard Sacks*
Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109
Time-of-flight mass spectrometry is unique in that ion abundance ratios are constant over the chromatographic peak profile provided that the peak contains only one component. This provides the means for the automated finding and spectral deconvolution of overlapping chromatographic peaks from completely unknown mixtures if the mass spectra for the overlapping components are sufficiently unique. This can greatly reduce the chromatographic resolution requirements, which allows for very rapid quantitative analysis as well as for high-speed mixture characterization. High-speed GC with stop-flow operation of a series-coupled column ensemble can be used to completely separate some component pairs that coelute from the column ensemble, thus eliminating the need for spectral deconvolution of those mixture components. This provides two options for high-speed qualitative and quantitative analysis, using either the mass spectra from deconvoluted overlapping peaks or the mass spectra from the completely separated peaks obtained with stopflow operation of the tandem column ensemble. These options are compared with respect to the similarity for spectral matching with a library and to peak area linearity with concentration, calibration plot correlation coefficients, and shot-to-shot reproducibility. High-speed GC (HSGC) with tandem-column tunable selectivity has been combined with time-of-flight (TOF) MS detection for the high-speed characterization of complex mixtures including pesticides1 and essential oils.2-4 With TOFMS, unlike scanning (nonbatch) MS techniques, ion-abundance ratios are constant across a chromatographic peak profile for peaks containing a single component. Changes in these ratios across a peak profile indicate overlapping chromatographic peaks,5-6 and this provides a means for the automated finding and deconvolution of the overlapping peaks from completely unknown mixtures. This can drastically reduce the GC resolution requirements, thus permitting (1) Veriotti, T.; Sacks, R. Anal. Chem. 2001, 73, 3045. (2) Veriotti, T.; Sacks, R. Anal. Chem. 2001, 73, 4395. (3) Veriotti, T.; Sacks, R. Perfum. Flavor. 2002, accepted. (4) Veriotti, T.; Sacks, R. Anal. Chem. 2002, accepted. (5) Erickson, E.; Enke, C.; Holland, J.; Watson, J. Anal. Chem. 1990, 62, 1079. (6) Schlag, E. Time-of-Flight Mass Spectrometry and its Applications; Elsevier: New York, 1994. 10.1021/ac020522s CCC: $25.00 Published on Web 07/11/2003
© 2003 American Chemical Society
very fast mixture characterization. Limitations of this technology include the need for some peak-apex separation in order for the instrument software to recognize the presence of more than one component in a chromatographic peak and the requirement that the mass spectra from the overlapping peaks be unique, as determined by instrument software.5,7 An additional issue is the reliability of spectral deconvolution procedures needed for the generation of the mass spectra of the individual components. The use of a tandem-column ensemble with programmable selectivity for obtaining more complete separations in HSGC has been demonstrated for a variety of samples.8-11 The ensemble consists of a polar and a nonpolar column connected in series. A particularly useful embodiment of tandem-column programmable selectivity is a first-column stop-flow system, in which the carriergas pressure at the junction point between the columns in the ensemble can be increased to the GC inlet pressure for short intervals.1-4,12 During these intervals, carrier gas flow stops in the first column and increases in the second column of the ensemble. This is very useful for obtaining enhanced separation of peak pairs that are completely separated by the first column in the ensemble but coelute from the ensemble. The combination of TOFMS and tandem-column HSGC with stop-flow operation is useful for increasing separation in cases in which the TOFMS software fails to find and deconvolute overlapping GC peaks that have inadequate peak apex separation or inadequate spectral uniqueness.2,4 This combination also provides a means for the direct comparison of peak-area quantitation and spectral matching with a library for mass spectra from completely separated peaks and from the spectral deconvolution of overlapping peaks. This study uses a 9-component mixture from a 70-component spectral library of mostly mono terpenes and oxygen-containing terpene structures. These compounds are of interest in the flavor and fragrance industries. This mixture produces several coelutions in the chromatogram from the tandem ensemble of a nonpolar and a polar capillary column. All coelutions can be eliminated with stop-flow operation of the column ensemble, thus allowing the (7) Leonard, C.; Sacks, R. Anal. Chem. 1999, 71, 5177. (8) Hinshaw, J. V.; Ettre, L. S. Chromatographia 1986, 21, 561. (9) Laub, R. J.; Purnell, J. H. J. Chromatogr. 1975, 112, 71. (10) Smith, H.; Sacks, R. Anal. Chem. 1997, 69, 5159. (11) Coutant, C.; Sacks, R. Anal. Chem. 2000, 72, 5450. (12) Whiting, J.; Sacks, R. Anal. Chem. 2002, 74, 246.
Analytical Chemistry, Vol. 75, No. 16, August 15, 2003 4211
comparison of mass spectra from the normally coeluting components when they are completely separated by enhanced separation methods and when they are deconvoluted by TOFMS software. EXPERIMENTAL SECTION Apparatus. The experimental platform for these studies uses an HP6890 GC (Hewlett-Packard, Atlanta GA) with a time-of-flight mass spectrometer (TOFMS) detector (Pegasus II, LECO Corp., St. Joseph, MI). The GC is equipped with an HP7683 autoinjector (HP, Atlanta, GA), electronic inlet-pressure control and a split/ splitless inlet. An HP flame ionization detector (FID) is used as a second detector to monitor a portion of the effluent from the first column in the ensemble. The pressure-programmable column ensemble consists of 7.0 m of 0.18-mm-i.d. fused-silica capillary with a 0.2-µm-thick film of trifluoropropylmethyl polysiloxane (Rtx-200, Restek, Bellefonte, PA), followed by 7.0 m of 0.18-mm-i.d. capillary with a 0.2-µmthick film of 5% phenyl dimethyl polysiloxane (Rtx-5, Restek). The junction point between the columns is connected to a 350-mL ballast chamber via a pneumatically driven, low-dead-volume valve (MOPV-1/50, SGE, Austin, TX). The valve is connected to the column junction point with a low-dead-volume all-glass splitter and a 5.0-cm-long, 0.25-mm-i.d. segment of deactivated fusedsilica tubing. The pneumatic valve is actuated by a computercontrolled solenoid valve (GH3412, Precision Dynamics, Phoenix, AZ) connected to a 50-55-psig compressed-air source. The ballast chamber and plumbing have been described.1 The carriergas pressure in the ballast chamber is controlled with a pressure controller (model 640A, MKS Instruments, Andover, MA) having a set-point resolution of 0.1 psi. The junction point between the columns is also connected to the FID by means of a 0.5-m-long, 0.05-mm-i.d. deactivated fused-silica tube. About 10% of the effluent from the first column is split to the FID, and the remainder passes through the second column and is detected by the TOFMS. Materials and Procedures. Hydrogen purified with filters for water vapor, oxygen, and hydrocarbons was used as carrier gas. The inlet pressure was 35.0 psig (49.3 psia). Split injection was used with a split ratio of 150:1. Injection size was 0.1 µL. All experiments used an initial oven temperature of 80 °C and a programming rate of 50 °C/min. starting at the time of injection. Mass spectra were recorded for the mass range (m/z) of 35-350. The user-defined signal-to-noise ratio threshold for automated peak finding was 25. The test mixture used in this study is described in Table 1. Retention-factor values for the nine components on the individual columns at the starting temperature of 80 °C also are presented. The nine compounds are part of a 70-compound spectral library created for the characterization and analysis of essential oils. The library was created by injecting the pure substances and saving the mass spectra by Pegasus II software. Normal alkanes through C12 were also injected so that retention-index data could be obtained on both columns in the ensemble. For quantitative analysis, the mixture was diluted with acetone to obtain concentrations in the range 0.133 to 0.667 vol %. Data acquisition and instrument control were provided by a 330-MHz PC (Dell, OptiPlex GX1). For FID measurements and stop-flow valve control, a 16-bit A/D board (PCI-DAS1602/16, 4212 Analytical Chemistry, Vol. 75, No. 16, August 15, 2003
Table 1. Retention Factors on the Rtx-200 and Rtx-5 Columns and Boiling Points for Mixture Components peak no.
compd name
kRtx-200
kRtx-5
bp (°C)
1 2 3 4 5 6 7 8 9
camphene furfural eucalyptol terpinolene benzaldehyde octanal β-caryophyllene geranyl acetate eugenol
0.724 1.117 1.327 1.327 1.898 2.001 6.333 7.233 7.506
1.411 0.432 2.218 3.250 1.220 1.442 7.684 5.323 4.709 254
159-160 162 176-177 175-178 66-68 137-139
Figure 1. Chromatograms of the nine-component mixture from the FID monitoring the column junction point (a), from the column ensemble without stop-flow operation (b), and from the column ensemble with stop-flow operation (c). Peak numbers correspond to the compound number in Table 1. The vertical arrows in chromatogram (a) indicate the points in the chromatogram where stop-flow operation is used to enhance the ensemble separation.
Computer Boards, Middleboro, MA) was used with LabVIEW software (National Instruments, Austin, TX). Chromatograms from the FID were processed by Grams/32 software (Galactic Industries, Salem, NH). For most experiments, the TOFMS was used with a spectral acquisition rate of 25 spectra/s. Rates of 100 and 500 spectra/s were used for some experiments. Automated peak finding and spectral deconvolution were accomplished with manufacturer’s software. RESULTS AND DISCUSSION Figure 1 shows chromatograms of the nine-component mixture using the dual-column ensemble. Chromatogram (a) is from the FID monitoring the effluent from the first (polar) column.
Figure 2. Extracted ion chromatograms and difference mass spectra for peaks 1 and 2 without stop-flow operation (a) and with stop-flow operation (b). The chromatograms display m/z values of 96 (solid lines) and 136 (dashed lines). The vertical lines in the chromatograms indicate the apex locations of the found peaks. The mass spectra were obtained by subtracting the library spectra from the real peak spectra.
Chromatograms (b) and (c) display the total-ion current (tic) using the TOFMS to monitor the effluent from the column ensemble. Note that chromatogram (a) is displayed on a different time axis. Chromatogram (b) was obtained without the use of stop-flow operation, and the nine components produce only 5 distinct peaks with severe overlaps of peaks 1/2, 4/6, and 7/8/9. From chromatogram (a), the first column in the ensemble results in the complete separation of all these component groups that co-eleute from the column ensemble. The coelutions of peaks 3/4 and the excessive overlap of 5/6 in chromatogram (a) are not a problem, since they are completely separated by the second (nonpolar) column. The four vertical arrows in chromatogram (a) indicate the points in the chromatogram where stop-flow operation of the first column can be used to enhance the resolution of the different component groups that coelute from the column ensemble. If the flow in the first column is stopped at the time indicated by the left-most arrow, component 1 is in the second column but component 2 is still in the first column. In chromatogram (c), the flow in the first column was stopped for 5.0 s, with the result that peaks 1 and 2 are completely separated. The first-column flow was stopped a second time to separate components 4 and 6, and two additional times to separate components 7, 8, and 9. Using stop-flow operation, a pair of target compounds must be completely separated by the first column. The resolution of components 8 and 9 is 1.5 in chromatogram (a), which is adequate for this application. Chromatograms (b) and (c) were processed by the TOFMS software, and the peaks were identified with the 70-component essential-oil library generated with the Pegasus II TOFMS instrument. Note that extensive use of the automated peak finding and spectral deconvolution software are required for chromatogram (b), but no deconvolution is needed for chromatogram (c).
Spectra Comparison. A requirement for successful automated peak-finding and spectral deconvolution of unknown mixtures is that the presence of all peak apexes can be detected. This, in turn, requires that at least two complete mass spectra are recorded between the peak apexes. Initial studies using a spectral acquisition rate of 25 spectra/s were successful in finding and identifying peaks 1, 3, 5, 7, and 9. However, the peak apex separations of 1/2, 4/6, and 7/8 are too small for automated peakfinding. With a spectral acquisition rate of 50 spectra/s, peaks 4 and 6 are correctly identified, and at 100 spectra/s, peak 2 is found and correctly identified. Even 500 spectra/s, which is the maximum spectral acquisition rate for the instrument, is inadequate for finding peak 8 because of the very small peak-apex separation. Note that deconvolution is very successful if the minimum separation-criterion for automated peak finding is satisfied, and the eight found peaks were all correctly identified as the first hits (greatest similarity) in the user library. For chromatogram (c), all components were correctly identified. Figure 2 shows extracted-ion chromatograms for peaks 1 and 2 with m/z values of 96 (solid lines) and 136 (dashed lines) for the case without stop flow (a) and with stop flow (b). Vertical lines indicate the peak apexes determined by automated peakfinding software. Also shown are the corresponding mass spectra presented as the differences (subtraction) between the real peak spectrum and the library spectrum for peaks 1 (left) and 2 (right). For a perfect spectral match with the library, the corresponding difference spectrum will have 0 amplitude for all m/z. For peak 1, a unique ion (m/z that is not detected in peak 2) exists, and thus, the peak shape can be precisely defined. Here, the difference spectra are nearly the same for (a) and (b), indicating that spectral deconvolution was very successful. Peak 2 has no unique ion, and although the peak is successfully identified from the user library, the difference spectra for the Analytical Chemistry, Vol. 75, No. 16, August 15, 2003
4213
Figure 3. Extracted ion chromatograms and difference mass spectra for peaks 4 and 6 without stop-flow operation (a) and with stop-flow operation (b). The chromatograms display m/z values of 121 (solid lines) and 57 (dashed lines). The vertical lines in the chromatograms indicate the apex locations of the found peaks. The mass spectra were obtained by subtracting the library spectra from the real peak spectra.
deconvolution case (a) indicates a poorer match than for the stopflow case, in which the peaks are completely separated. Figure 3 shows similar data for peaks 4 and 6. Here, a situation similar to Figure 2 occurs with peak 4 having a unique ion and peak 6 having no unique ion with adequate signal-to-noise ratio. The solid-line chromatograms are for m/z of 121, and dashedline chromatograms are for m/z of 57. Again, vertical lines indicate the peak apexes determined by the peak-finding software. In this case, the difference spectra indicate a somewhat better spectral match for both peaks when complete separation is achieved with stop-flow operation. Figure 4 shows extracted-ion chromatograms for peaks 7, 8, and 9 using m/z values of 161 (solid lines), 69 (dashed lines), and 164 (dotted lines) for the case without stop-flow (a) and with stop-flow (b). Also shown are the corresponding mass spectra presented as the difference spectra from the user library. Note that m/z values 161 and 164 are unique to peaks 7 and 9, respectively. The peak apex separation of peaks 7 and 8 in chromatogram (a) is 95% confidence) larger for chromatogram (b) than for than for chromatogram (c). For cases in which deconvolution is required in chromatogram (b), the similarities are comparable for some compounds in the two chromatograms but are substantially smaller for other peaks in chromatogram (b). For components 1 and 2, adequate peak apex separation is achieved, and automated peak-finding and deconvolution are successful, but the similarity for peak 1 with the user library is substantially larger than for peak 2. This is consistent with Figure 2 and is the result of the lack of a unique ion for peak 2. For the coelution of peak 4, which has a unique ion, and peak 6, which does not have a unique ion, the similarity of peak 4 to the user library is slightly smaller with deconvolution than with complete separation using stop-flow operation, but the similarity of peak 6 is substantially smaller with spectral deconvolution. For the coelution of peaks 7, 8, and 9, complete separation with stop-flow operation yields high similarity values for all three compounds. With spectral deconvolution, the similary of peak 7 with the user library is only slightly lower than with complete separation. For peak 9, however, substantially lower similarity is obtained with spectral deconvolution. Since peak 8 was not found by automated peak finding software, no similarity is reported.
Figure 4. Extracted ion chromatograms and difference mass spectra for peaks 7, 8, and 9 without stop-flow operation (a) and with stop-flow operation (b). The chromatograms display m/z values of 161 (solid lines), 69 (dashed lines), and 164 (dotted lines). The vertical lines in the chromatograms indicate the apex locations of the found peaks. The mass spectra were obtained by subtracting the library spectra from the real peak spectra. The difference spectrum for component 8 is shown only for the stop-flow case, since this peak was not found by the peak-finding software. Table 2. Average Similarities for Mixture Components Obtained with a User Library and the NIST Library Using Stop-Flow Operation and Deconvolution Software similarity user library peak no.
compd
1 2 3 4 5 6 7 8 9
camphene furfural eucalyptol terpinolene benzaldehyde octanal β-caryophyllene geranyl acetate eugenol
NIST library
stopstopdeconvolution flow deconvolution flow 903 768 953 906 952 760 929 821
895 871 946 923 946 917 945 943 974
882 795 854 860 694 800 759
870 797 841 826 898 853 825 710 890
As expected, use of the NIST library results in lower similarities in all cases, and for peaks 6 and 9, substantially larger similarities are obtained with complete separation (stop-flow) than for the case with deconvolution. With the relatively small user library, all found peaks were correctly identified as the first hits (largest similarities). With the large NIST library, peaks 1, 4, 5, and 6 were identified as first hits for the case without stop-flow (overlapping peaks), and peaks 1, 2, 3, 4, 5, and 6 were identified as first hits for the case with complete separation using stop flow. All other peaks were identified in the first 10 hits, except for peak 2 with deconvolution. No similarity value is reported for this case. Peak-Area Quantitative Analysis. Figure 5 shows analytical curves (log peak area vs log concentration) for the nine mixture components. Each point represents the average from five replicate injections. The plots have been displaced vertically for presentation
Figure 5. Analytical curves (log peak area vs log concentration) for the nine-component mixture without stop-flow (a) and with stop flow (b). Each log peak area value is an average from five replicate injections.
clarity. The plots on the left were obtained without stop-flow operation (see chromatogram (b), Figure 1), and the plots on the right were obtained from complete separation using stop-flow operation. Table 3 summarizes data from the plots in Figure 5. In the Table, correlation coefficients (R2) and log-log slopes of the plots are presented for the nine mixture components. For the case with no stop-flow, overlapping peaks were processed with instrument software, except for component 8, which was not found in the chromatogram. All correlation coefficients are g0.97, indicating good linearity of the log-log plots. For components 3 and 5, which are adequately separated without stop-flow, R2 values are g0.992. In Analytical Chemistry, Vol. 75, No. 16, August 15, 2003
4215
Table 3. Correlation Coefficients and Log-Log Slopes for the Nine-Component Mixture Comparing Stop-Flow and Deconvolution Methods R2 peak no.
compd
1 2 3 4 5 6 7 8 9
camphene furfural eucalyptol terpinolene benzaldehyde octanal β-caryophyllene geranyl acetate eugenol
slope
stopstopdeconvolution flow deconvolution flow 0.993 0.997 0.992 0.987 0.993 0.987 0.973 0.989
0.994 0.999 0.998 0.998 0.998 0.999 0.999 0.997 0.999
0.92 0.94 1.05 1.02 1.02 1.00 0.91 0.95
0.88 1.24 1.12 1.09 1.35 1.17 1.09 0.98 1.07
all cases in which complete separation occurred, R2 values are g0.992. For peaks 1 and 2, R2 values are only slightly greater with complete separation. For peaks 4 and 6, 7 and 9, R2 valves are substantially larger for the case with complete separation using stop-flow operation. The log-log slopes range from 0.91 to 1.08 for the case without stop-flow operation. These values are all reasonably close to the expected slope value of 1.0 for completely linear response with concentration. When stop-flow operation is used to achieve a complete separation, the log-log slopes range from 0.88 to 1.35, with five of the values outside the range 0.9 to 1.1. The reason for the larger deviations from linear response with stop-flow operation is unclear. Relative standard deviations for five replicate injections are in the range 1-10% for all nine peaks and for all concentrations, both with complete separation and when deconvolution is used to process overlapping peaks. CONCLUSIONS Spectral deconvolution of overlapping peaks works very well when all peak apexes can be located. This requires that at least two complete mass spectra are recorded between the peak apexes. When the peaks have unique ions to accurately define the peak
4216 Analytical Chemistry, Vol. 75, No. 16, August 15, 2003
shapes, similarities are comparable to values obtained with complete separation. However, similarities may be degraded for both peaks when one of them has no unique ion. When the automated peak-finding software fails to find a peak because of inadequate peak-apex separation, as was the case for peak 8 in the work reported here, the found peak may still be identified with reasonable reliability despite the unaccounted contribution to the mass spectrum from the unfound peak if a unique ion is present in the found peak. Further work is required to determine the limitations of this aspect of the deconvolution software. The creation of a user library for a specific application or a specific class of compounds is very useful and can dramatically improve mixture characterization. With a large general library, such as the NIST database, the library spectra may be obtained under conditions significantly different from the unknown mixture, and it may be difficult to identify some peaks without prior knowledge of the sample. The use of a tandem column ensemble with stop-flow operation is an attractive complement to GC/TOFMS, since the need for spectral deconvolution often is eliminated. This can result in greater similarities and, in some cases, more reliable peak characterization. In addition, complete separation may result in larger correlation coefficients for analytical curves, especially for cases in which unique ions are not present in one or more of the overlapping peaks. Finally, for cases in which there is inadequate peak apex separation for automated peak-finding, stop-flow operation often can be used to obtain adequate separation. The enhanced separation with stop-flow operation also may allow for the use of lower spectral acquisition rates, which can substantially reduce computer processing time and data file size. ACKNOWLEDGMENT The authors gratefully acknowledge LECO Corp., St Joseph, MI, for use of the Pegasus II TOFMS. Received for review August 12, 2002. Accepted May 20, 2003. AC020522S
