Anal. Chem. 2000, 72, 3063-3069
High-Speed GC/MS of Gasoline-Range Hydrocarbon Compounds Using a Pressure-Tunable Column Ensemble and Time-of-Flight Detection Tincuta Veriotti and Richard Sacks*
Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109
A pressure-tunable series-coupled ensemble of two capillary GC columns is combined with a time-of-flight MS detector for the high-speed characterization of mixtures containing hydrocarbon compounds. The column ensemble consists of a nonpolar 5% phenyl poly(dimethylsiloxane) column and a very polar poly(ethylene glycol) column. The TOFMS instrument uses time-array detection to obtain up to 500 complete electron mass spectra per second. Instrument software allows for automated peak finding and the spectral deconvolution of severely overlapping unknown chromatographic peaks, if their fragmentation patterns are significantly different and if at least two spectra can be recorded between the peak apexes. By adjusting the carrier-gas pressure at the columnjunction point, the separations between adjacent peak pairs can be adjusted to enhance the capabilities of the TOFMS detector. The sensitivity of peak-pair separation to changes in junction-point pressure is studied for combinations of alkanes, olefins, and aromatic compounds. When complete separation is required, the use of pressure-tunable column ensembles cannot always provide sufficient control of peak-pair separation for structurally similar compounds. However, complete chromatographic separation typically is not required with the TOFMS detection, and a pressure-tunable column ensemble is very useful for the high-speed characterization of hydrocarbon mixtures. Gas chromatography (GC) and GC with mass spectrometric detection (GC/MS) are frequently used for the analysis and characterization of mixtures containing gasoline-range hydrocarbon compounds.1-5 These compounds are ubiquitous in the environment and in many industrial settings. Petroleum distillate samples are very complex and may contain hundreds of compounds. Many of these compounds have similar molecular structures and thus are difficult to separate and to characterize. Columns, often 100-m-long, have been developed for these (1) George, C.; Rood, D. J. Chromatogr. Sci. 1999, 37, 448. (2) Teng, S.; Williams, A.; Urdal, K. J. High Resolut. Chromatogr. 1994, 17, 469. (3) Story, M.; Laser, B.; Habfast, K. Int. J. Mass Spectrom. Ion Processes 1983, 48, 47. (4) Berger, T. Chromatographia 1996, 42, 63. (5) Durand, J.; Gautier, S.; Robert, E.; Guilhem, M.; Phan TanLuu, R. J. HighResolut. Chromatogr. 1997, 20, 289. 10.1021/ac000081h CCC: $19.00 Published on Web 06/06/2000
© 2000 American Chemical Society
applications.6 Despite this, complete separations often cannot be achieved, even with analysis times of several hours. Samples from cracking plants are less complex, but often contain 20-30 components. Production unit operators often rely on compositional analysis by GC for monitoring unit operation. The combination of GC with MS detection is very useful for these applications since it provides more complete compositional analysis than any other method.7,8 Comprehensive two-dimensional GC has been used recently for the separation of complex hydrocarbon mixtures.9,10 The high peak capacity obtained from two-dimensional separations is beneficial for complex mixture analysis. In addition, chromatograms show structured patterns in the two-dimensional retention plane that can be used to determine compound class. While the information content in two-dimensional chromatograms can be large, the acquisition of two-dimensional chromatograms of complex mixtures is time-consuming. Several technologies have been developed for increasing GC analysis speed. These technologies include microbore columns,11 high-speed temperature programming,12,13 inlet systems which produce very narrow sample vapor plugs,14,15 and column ensembles with adjustable selectivity.16-19 For high-speed separations, relatively short columns are operated at unusually high carrier-gas flow rates. While very fast separations have been achieved, the use of shorter columns results in significantly lower peak capacity than can be achieved with longer columns. Singlechannel flame ionization detection has been used for most highspeed GC (HSGC) studies. Little work has been reported on (6) Lipari, F. J. Chromatogr. 1990, 503, 51. (7) McFadden, W. Techniques of Combined GC/MS: Applications in Organic Analysis, Wiley-Interscience: New York, 1973. (8) Karasek, F.; Clement, R. Basic Gas Chromatograpy-Mass Spectrometry, Elsevier: Amsterdam, 1988. (9) Venkatramani, C.; Phillips, J. J. Microcolumn Sep. 1993, 5, 511. (10) Frysinger, G.; Gaines, R. J. High Resolut. Chromatogr. 1999, 22, 251. (11) van Es, A. High-Speed Narrow Bore Capillary Gas Chromatography; Huthig Buch Verlag: Heidelberg, Germany, 1992. (12) Ehrmann, E. U.; Dharmasena, H. P.; Carney, K.; Overton, E. B. J. Chromatogr. Sci. 1996, 34, 533. (13) Grall, A.; Leonard, C.; Sacks, R. Anal. Chem 2000, 72, in press. (14) van Es, A.; Janssen, J.; Bally, R.; Cramers, C.; Rijks, J. J. High Resolut. Chromatogr. Chromatogr. Commun. 1987, 10, 273. (15) Klemp, M.; Peters, A.; Sacks, R. Environ. Sci. Technol. 1994, 28, 369A. (16) Hinshaw, J.; Ettre, L. Chromatographia 1986, 21, 561. (17) Sacks, R.; Akard, M. Environ. Sci. Technol. 1994, 28, 428A. (18) Akard, M.; Sacks, R. Anal.Chem. 1994, 66, 3036. (19) Smith, H.; Sacks, R. Anal. Chem. 1997, 69, 5159.
Analytical Chemistry, Vol. 72, No. 14, July 15, 2000 3063
HSGC of more complex hydrocarbon mixtures since peak capacity is inadequate for these applications. The recent development of time-of-flight (TOF) MS with time array detection20-22 provides a means for the high-speed analysis and characterization of complex hydrocarbon mixtures. Most lowresolution MS detectors for GC/MS are either a quadrupole or ion trap, and chromatographic peaks are sequentially identified by comparison with standard spectral libraries.7,8 These scanning instruments are relatively slow, and maximum full-scan spectral acquisition rates for quadrupole instruments are a few spectra per second or less. At these low scan rates, concentration biasing of mass spectra occurs because of the change in sample partial pressure in the ion source as the chromatographic peak evolves. Thus, complete separation and relatively broad peaks may be required for reliable characterization. This is incompatible with the relatively narrow peaks usually obtained with HSGC. With TOFMS using time array detection, complete mass spectra without concentration biasing can be obtained at rates as high as 500 spectra per second.23,24 These high spectral acquisition rates allow for accurate tracking of the narrow chromatographic peaks produced by HSGC. In addition, spectral deconvolution of severely overlapped chromatographic peaks can be obtained for unknown mixtures if the fragmentation patterns from the overlapping components are sufficiently different. The end result is greatly reduced resolution requirements and therefore increased peak capacity. The automated peak finding and spectral deconvolution software available with some commercial TOFMS instruments is very convenient and reduces analysis time.24 However, software algorithms require some peak apex separation to recognize the presence of two or more components in a single chromatographic feature from an unknown mixture. Higher spectral acquisition rates provide a reduction in the minimum required peak apex separation, but result in larger data files. For complex mixtures, some chromatographic peaks may overlap sufficiently that automated peak finding and spectral deconvolution may not be possible even with the highest available spectral acquisition rates. Pressure-tunable ensembles of capillary GC columns using different stationary phases have been used to structure chromatograms for HSGC to utilize the available peak capacity more efficiently.16-19 This can result in significant reductions in analysis times. By changing the pressure at the junction point between two series-coupled columns, the selectivity of the ensemble can be adjusted to affect the overall elution pattern. While structuring chromatograms is very effective for mixture components having significantly different functionalities, peak separations for more chemically similar component pairs show less sensitivity to the pressure changes used to adjust the ensemble selectivity. The use of TOFMS detection can substantially reduce the resolution requirements for the analysis and characterization of unknown mixtures, and only a small peak apex separation may (20) Erickson, E.; Enke, C.; Holland, J.; Watson, J. Anal. Chem. 1990, 62, 1079. (21) Holland, J.; Newcombe, B.; Tecklenburg, R.; Davenport, M.; Allison, J.; Watson, J.; Enke, C. Rev. Sci. Instrum. 1991, 62, 69. (22) Watson, J.; Schultz, G.; Tecklenburg, R.; Allison, J. J. Chromatogr. 1990, 518, 283. (23) van Deursen, M.; Janssen, H.-G.; Cramers, C. Proceedings of the 21st International Symposium on Capillary Chromatography and Electrophoresis, Park City, UT, June 20-24, 1999; p 186. (24) Leonard, C.; Sacks, R. Anal. Chem. 1999, 71, 5177.
3064
Analytical Chemistry, Vol. 72, No. 14, July 15, 2000
be required for successful automated peak finding and the spectral deconvolution of significantly overlapped chromatographic peaks. A pressure-tunable ensemble of capillary GC columns using different stationary phases has been used to structure chromatograms in order to enhance the peak finding and spectraldeconvolution capabilities of TOFMS using time array detection.24,25 This paper considers the high-speed characterization of gasolinerange hydrocarbon compounds including alkanes, olefins, and aromatics. The sensitivity of peak-pair separation to changes in tuning pressure is examined for several peak pairs representing different combinations of structural features. The utility of the combination of a pressure-tunable ensemble of capillary columns and TOFMS detection is demonstrated for the high-speed characterization of a 25-component hydrocarbon mixture. EXPERIMENTAL SECTION Apparatus. The experimental system used for these studies has been described in detail.24 The pressure-tunable column ensemble consists of a 10-m-long, 0.18-mm-i.d. poly(ethylene glycol) (DB-Wax, J&W Scientific, Inc., Folsom, CA) column followed by a 10-meter-long, 0.18-mm i.d. 5% phenyl poly(dimethylsiloxane) (DB-5, J&W Scientific, Inc.) column. Both columns have 0.18-µm-thick bonded stationary phases. The column ensemble is operated in a Hewlett-Packard 6890 GC equipped with electronic inlet pressure control. Samples were injected by a Hewlett-Packard 7683 autoinjector. The pressure at the column junction point is controlled by an electronic pressure controller (MKS model 640A, MKS Instruments, Inc., Andover, MA). This device uses an absolute-pressure capacitance manometer to control the pressure anywhere in the range 0-100 psia in 0.1 psi steps. The 0-5.0 V control signal is provided by a Gateway 2000 P5-75 computer equipped with a 12bit A/D board (DT-2801, Data Translation, Inc., Marlboro, MA). The outlet from the pressure controller, which is located outside of the GC oven, is connected to the column junction point by a 8-cm-long segment of 0.53-mm fused silica tubing and a stainless steel “T” splitter (MT1C56, Valco Instruments, Houston, TX). A vent line, consisting of a 1.25-m-long, 0.25-mm diameter uncoated fused silica tube, also is connected to the outlet of the pressure controller by means of a second T splitter. This line vents to atmospheric pressure and is used to reduce the pressure equilibration time when the controller set-point pressure is changed to lower values. This has been discussed in detail in a prior publication.26 A model Pegasus II time-of-flight mass spectrometer (LECO Corp., St Joseph, MI) was used for mixture characterization. This instrument uses reflectron technology and time-array detection20-22 to obtain complete mass spectra (5-1000 Da) with spectral acquisition rates as high as 500 Hz. Electron ionization is used with the electron energy set at 70 eV. Ion packets are formed by orthogonal beam deflection techniques.27 A microchannel plate is used for ion detection. Completely automated peak finding and spectral deconvolution software is provided with the instrument. The outlet from the column ensemble is connected to the TOFMS (25) Sacks, R.; Coutant, C.; Veriotti, T.; Grall, A. J. High Resolut. Chromatogr., in press. (26) Leonard, C.; Sacks, R. Anal. Chem. 1999, 71, 5501. (27) Pinkston, J.; Rabb, M.; Watson, J. Allison, J. Anal. Chem. 1990, 62, 1079.
Table 1. Compound Names, Boiling Points, and Unique Masses (mass/charge) for Test Mixture peak no.
compound
bp (°C)
unique mass
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
2-octene cycloheptane toluene heptane 2,3-dimethyl1,2,4-trimethylcyclohexane(R) 2,5-dimethyl-2,4-hexadiene 1,2,4-trimethylcyclohexane(S) 1-nonene nonane 7-methyl-1,6-octadiene ethylbenzene p-xylene m-xylene cyclooctene o-xylene cumene styrene 1-ethyl-3-methylbenzene decane 1,3,5-trimethylbenzene 1,2,4-trimethylbenzene isobutylbenzene bicyclopentadiene 4-methylstyrene 1,2,3-trimethylbenzene
123 118.5 110.6 140 141-143 132-134 141-143 146 151 143-144 136 138 138-139 145-146 143-145 152-154 145-146 158-159 174 162-164 168 170 170 170-175 175-176
55 41 91 43 69 67 69 41 43 41 91 91 91 67 91 105 104 105 43 105 105 91 66 117 105
by a heated interface, also provided with the TOFMS instrument. The interface was heated to 250 °C, and the MS flight tube was operated at a pressure of 1 × 10-6 Torr. Materials and Procedures. Table 1 lists the compounds in the test mixture. Boiling points and masses used for spectral deconvolution (unique masses) are also found in the table. Components were selected such that different classes of hydrocarbon compounds including normal, branched, and cyclic alkanes; mono- and diolefins (both conjugated and nonconjugated); and aromatic compounds are all represented in the mixture. Equal volumes of components 1-17 and double this volume of the lower vapor pressure components 18-25 were mixed, and 5.0-µL headspace samples above this mixture were injected using a split ratio of 20:1. The nitrogen ion at m/z 28, from air in the headspace samples, was used for carrier-gas holdup time measurements for the column ensemble. Holdup times for the individual columns were calculated from standard equations for gas flow in capillary tubes.28 Hydrogen carrier gas, after purification with filters for oxygen, water vapor, and hydrocarbons, was provided for the GC inlet and for the inlet to the electronic pressure controller used at the column junction point. For all studies, the GC oven temperature was 80 °C (isothermal) and the inlet pressure was 49.7 psia. Data acquisition and processing were accomplished by use of the Pegasus II software, version 1.21 (LECO Corp.). Peak characterization was based on comparison with the NIST mass spectral database. RESULTS AND DISCUSSION The automated peak finding and spectral deconvolution algorithms used in this study require that chromatographic peak (28) Grant, D. W. Capillary Gas Chromatography; Wiley: New York, 1996; pp 16-20.
Figure 1. Analytical-ion chromatograms (a and c) and extractedion chromatograms (b and d) for the early-eluting portion of the 25component hydrocarbon mixture. Chromatograms a and b were obtained with a junction-point pressure of 36.4 psia and c and d with a junction-point pressure of 39.5 psia. See Table 1 for correlation of peak apex numbers to numbered compounds and unique-ion mass per charge.
apex separation be large enough to obtain at least two mass spectra between the peak apexes. With a spectral acquisition rate of 500 spectra/s, a minimum peak separation of only 6 ms is required. This gives a theoretical maximum peak capacity of about 167 peaks/s. This large value allows for the rapid characterization of complex mixtures using HSGC. Larger peak separations may be required for mixture components having similar fragmentation patterns and for compounds present in lesser abundance, where signal-to-noise ratios may be inadequate for reliable peak finding of very closely spaced peaks. The work reported here used a spectral acquisition rate of 200 Hz (5 ms/spectra), equating to a minimum peak apex separation of 15 ms. Signal-to-noise ratios decrease steadily with increasing spectral acquisition rate; however, a peak-smoothing algorithm provided with the instrument software is effective in enhancing signal-tonoise ratios. A 20-point smoothing routine was used on all data presented here. Use of this routine made the resulting signal-tonoise ratios equivalent to those obtained with a spectral acquisition rate of 10-20 Hz with no smoothing. Retention Characteristics with Pressure-Tunable Selectivity. When the carrier gas pressure at the junction point between two columns with different stationary phases is changed, retention times for all mixture components change. This results from a change in the ensemble selectivity as well as a change in the ensemble holdup time. Figure 1 shows analytical-ion chromatograms, a and c, and extracted-ion chromatograms for all of the unique ions used for peak finding, b and d. Unique-ion mass per charge values are listed in Table 1. The junction-point pressures were 36.4 psia (250.1 kPa) for chromatograms a and b and 39.5 psia (271.4 kPa) for chromatograms c and d. The chromatograms show the first 17 components to elute. Vertical lines correspond to the peaks found with the automated peak-finding software. The numbers by the lines correspond to the component numbers in Table 1. The analytical ion chromatograms show the sum of the extracted ion chromatograms for all the masses, which exceed a software-selectable signal-to-noise ratio, for all found peaks. The Analytical Chemistry, Vol. 72, No. 14, July 15, 2000
3065
Figure 2. Plots of ensemble retention time vs junction-point pressure for the 25-component hydrocarbon mixture. The plot labeled U is for an unretained component. Plot numbers correspond to the component numbers in Table 1.
information contained in the analytical-ion chromatogram is similar to the information in the total-ion chromatogram but with substantially increased signal-to-noise ratios. At a junction-point pressure of 36.4 psia in Figure 1b, peak pairs 3/9 and 7/8 coelute and only a single peak was found for each pair. Note that components 3, 7, 8, and 9 appear as a single, relatively broad feature in the single-channel presentation of Figure 1a. When the junction-point pressure is increased to 39.5 psia, significant changes in the peak pattern are observed. Note that in Figure 1b, component 19 (decane) elutes after the aromatic compounds 11,12, and 13; while in Figure 1d, component 19 elutes before the aromatic compounds. A large shift in component 3 (toluene) relative to the nonaromatic compounds 7, 8, and 9 also is observed for the two junction-point pressure values. For the higher junction-point pressure value, component pairs 3/9 and 7/8 are sufficiently well separated that peak finding and spectral deconvolution were successful. The separation between the apexes of peaks 7 and 8 is about 160 ms. However, at this junction-point pressure, components 15 and 16 coelute, and peak finding was not successful. Figure 2 shows measured retention times for all 25 mixture components as the junction-point pressure is changed over the range from 16.0 to 40.0 psia (110-275 kPa). Measurements were made at 1.0 psi intervals. Numbers by the plots correspond to the component identification numbers in Table 1. The plot labeled U is for nitrogen ion and shows the ensemble holdup time variation with junction-point pressure. Note that the holdup time for a pressure-tunable column ensemble is a minimum when the column junction-point pressure is set to the value that would occur at the junction point in the absence of any external connections (natural junction-point pressure). The ensemble holdup time is equal to the sum of the hold-up times of the individual columns, and values increase rapidly as the junction-point pressure approaches the inlet pressure or the detector pressure. The shapes of the plots in Figure 2 are dependent on the retention of the compounds on the individual columns and on the changes in the holdup times for the individual columns with changes in the junction-point pressure. Note that many crossings of pairs of these plots occur over the useful junction-point pressure 3066 Analytical Chemistry, Vol. 72, No. 14, July 15, 2000
Figure 3. Plot of retention time values at a junction-point pressure of 40 psia vs values at 16 psia for the 25-component mixture. The numbered points correspond to the component numbers in Table 1. The inset shows the first 10 components on an expanded time scale.
range shown in the figure. For any pressure where a pair of plots cross, the corresponding components have the same retention time and a coelution will occur. Note that plots 17 (styrene) and 19 (n-decane) cross at a junction-point pressure of about 25.5 psia. These compounds represent the extremes of polarity for the compounds in the test mixture. Their separation (difference in retention times) is very sensitive to changes in junction-point pressure, and over most of the useful pressure range, these components are well-separated. The separations of many peak pairs are very insensitive to junction-point pressure and, over significant ranges of junction-point pressure, may be inadequate for automated peak finding and spectral deconvolution. For a pressure-tunable column ensemble, the relative contributions that the individual columns make to the ensemble selectivity is equal to the relative contributions that the corresponding columns make to the ensemble holdup time.16,18 The junctionpoint pressure values used in this study ranged from 16 to 40 psia. At 16 psia, the first (polar) column contributes 27.9% to the ensemble holdup time. At 40 psia, the first column contributes 75.6% to the ensemble holdup time. Thus, increasing the junctionpoint pressure increases the relative contribution of the polar column, and the ensemble selectivity becomes more characteristic of the polar column. Therefore, polar solutes are retained to a greater degree than nonpolar compounds. In Figure 3, the retention times of the 25 mixture components for a junction-point pressure of 16 psia are plotted vs the corresponding retention time values for 40 psia. This represents the extreme ends of the plots in Figure 2. Each component then occupies a single point in the resulting retention-time plane. The numbers near the points correspond to the component numbers in Table 1. Retention times for junction-point pressures of 16 psia or 40 psia are found by projection of the points onto the respective axis. Any component, which has the same retention-factor value on the two columns in the ensemble, will appear as a point on the extrapolation of a line connecting the plot origin to the point for a component which is unretained on either column (point U). This
isopolarity line is shown by a broken line in the figure. The degree of scatter of the points about this line indicates the degree of orthogonality of the selectivities obtained for the 16 and 40 psia junction-point pressures. The 16 psia axis corresponds to a 72.1% holdup time contribution from the nonpolar column. Since most of the target compounds are relatively nonpolar, they have larger ensemble retention factor values for the 16 psia junction pressure value. These points lie below the isopolarity line. Points 9 and 19 represent n-nonane and n-decane, respectively. These are the least polar components in the test mixture, and their points are located furthest below the line. Points 17 and 24 are for styrene and 4-methylstyrene, respectively. These are the most polar components in the test mixture, and their points are located furthest above the line. The ease of separation of any pair of components is related to the distance between the corresponding points in Figure 3. For any pair of components, if the retention time for one of them is larger at both of the two extreme pressure values (16 and 40 psia), coelution of the components is impossible and the corresponding plots in Figure 2 do not cross. If the retention time for one of the components is larger at 16 psia and the retention time for the other component is larger at 40 psia, the corresponding plots in Figure 2 will cross and coelution will occur for some junctionpoint pressure value in the available range. Component pair 17/ 19 is the most extreme example of this case, and as seen by the plots in Figure 2, their peak separation is very sensitive to changes in junction-point pressure. The most congested portion of Figure 3 is shown on an expanded time scale in the Figure 3 inset. In particular, note that the points for components 7, 8, and 9 are very close together. The corresponding regions in the chromatograms will be very congested at all available junction-point pressures. This represents the most challenging aspect of this characterization problem. Optimization Techniques. Optimization involves the selection of the junction-point pressure values that will provide for a complete, automated characterization of the target mixture. The mixture of components corresponding to the cluster of points contained in the box of Figure 3 produces the most congested region of the ensemble chromatograms. Retention time vs junction-point pressure for this subgroup is examined more closely in Figure 4a for the junction-point pressure range from 28 to 37 psia. The numbers by the plots correspond to the component numbers in Table 1. The curved lines are quadratic regression fits to the measured retention-time data. The correlation coefficients (R2 from tR) for the fitted lines are given in Table 2. In Figure 4b, ensemble retention factors for the six components are plotted vs the ensemble junction-point pressure. For the junction-point pressure range considered in Figure 4, the relative contribution of the first column to the ensemble selectivity increases linearly with junction-point pressure. For every mixture component, the retention values for the two columns in the ensemble are additive and the plots in Figure 4b should be linear as observed in the linear regression fits shown in the figure. Table 2 gives the slopes and correlation coefficients (R2 from ko) for the six plots. The slopes of the plots in Figure 4b provide direct information on the polarity of the respective compounds with respect to the columns used in the tandem ensemble. Since the fractional
Figure 4. Plots of ensemble retention times (a), ensemble retention factors (b), and peak separations (c) vs junction-point pressure for six of the mixture components. Plot numbers correspond to the component numbers in Table 1. Table 2. Compound Names, Regression Correlation Coeffients, and Slopes for the Compounds in Figure 4 peak
compound
3 6 7
toluene 2,5-dimethyl-2,4-hexadiene 1,2,4-trimethylcyclohexane(S) 1-nonene nonane 7-methyl-1,6-octadiene
8 9 10
R2 (from tR)
R2 (from ko)
slope (from ko)
0.9984 0.9958 0.9927
0.9690 0.9899 0.9920
0.0041 -0.0098 -0.0181
0.9853 0.9920 0.9964
0.9863 0.9948 0.9963
-0.0195 -0.0234 -0.0161
contribution of the first (polar) column increases with increasing junction-point pressure, a positive slope indicates greater retention on the polar column. Only toluene (plot 3) shows an increase in ensemble retention factor with increasing junction-point pressure. The plot for n-nonane (plot 9) has the largest negative slope, indicating the least polar character of the six compounds. The slope for 1-nonene (plot 8) has a smaller negative slope than that for n-nonane, illustrating the increased polarity resulting from the double bond. The presence of two nonconjugated double bonds in 7-methyl-1,6-octadiene (plot 10) results in greater polarity (less negative slope) than that of 1-nonene. Still greater polarity is observed in the conjugated double bonds of 2,5-dimethyl-2,4hexadiene (plot 6). Note that the polarity of 1,2,4-trimethylcyclohexane (plot 7) is very similar to that of 1-nonene, and this is the most difficult component pair to separate in the mixture. From these plots, the separation window diagram shown in Figure 4c was obtained. In the window diagram, the peak separations of all peak pairs which have separations less than 1.0 s are plotted vs the junction-point pressure. Window diagrams are often used for the optimization of separation quality for tunable Analytical Chemistry, Vol. 72, No. 14, July 15, 2000
3067
Figure 5. Extracted-ion chromatograms for the six components from Figure 4 using junction-point pressure values of 28.0 psia (a), 32.0 psia (b), 35.0 psia (c) and 37.0 psia (d). Peak numbers correspond to the component numbers in Table 1.
column ensembles, but some measure of chromatographic resolution is typically the dependent variable. With the TOFMS instrument used here, peak finding and spectral deconvolution are based on peak apex separation, not on chromatographic resolution. Thus, peak-pair separation is a more useful dependent variable for window-diagram optimization of junction-point pressure. Component pair 7/8 shows the lowest sensitivity to changes in the junction-point pressure. The greatest separation observed for these components for the pressure range considered in Figure 4 is about 120 ms. Figure 5 shows extracted-ion chromatograms for the six components described by the plots in Figure 4. Chromatograms a, b, c, and d were obtained at junction-point pressure values of 28.0, 32.0, 35.0, and 37.0 psia, respectively. At 28.0 psia, peaks 6, 7, 8, and 9 show significant overlap, but the peak finding and spectral deconvolution algorithms were successful and the characterization of these components was complete. Note that little useful information regarding these components could be obtained under these conditions using single-channel detection. In chromatogram b, components 7, 8, and 9 show no significant separation of the peak apexes and only a single peak was found. This is consistent with the very small peak apex separations indicated in the window diagram in Figure 4c for a junction-point pressure value of 32.0 psia. When the junction-point pressure was increased to 35.0 psia (chromatogram c), component 9 was found, but the separation of components 7 and 8 was inadequate and only a single component is indicated. In addition, only a single peak was found for components 6 and 10. When the pressure is increased to 37.0 psia as in Figure 5d, all components are found and spectral deconvolution was successful. The elution patterns and peak separations observed in all four chromatograms in Figure 5 are consistent with the plots in Figure 4. Note that despite the very similar response of components 7 and 8 to changes in ensemble selectivity, the use of electronic control of the junction-point pressure is useful in establishing conditions suitable for the complete high-speed characterization of these six mixture components. Without the unique capabilities of TOFMS with time-array detection, it would 3068
Analytical Chemistry, Vol. 72, No. 14, July 15, 2000
not be possible to obtain reliable quantitation or characterization of components 7 and 8 in unknown mixtures in the time frame of these experiments. In the window diagram as well as in the chromatograms, it is shown that components 7 and 8 cannot be adequately separated for successful automated peak finding over a considerable range of junction-point pressures. This applies to every peak pair that coelutes in the available junction-point pressure range, as each will define a range of junction-point pressure values for which that peak pair will not be adequately separated. The size of each of these unfavorable pressure intervals depends on the sensitivity of the peak-pair separation to changes in the junction-point pressure (slopes in Figure 4c). As the number of mixture components that elute in a specified retention interval increases, the number of unfavorable junction-point pressure intervals increases and the probability of there existing a favorable junctionpoint pressure for complete automated mixture characterization decreases. The use of on-the-fly selectivity changes during an analysis, which has been demonstrated26,29,30 (programmable selectivity), will be useful for the high-speed characterization of more complex mixtures. The ability of the instrument software to find and spectrally deconvolute overlapping component peaks is based on the number of spectra that can be recorded between the peak apexes. Decreasing the spectral acquisition rate will decrease this number, and greater peak separation is required for automated peak finding. Thus, with a pressure-tunable column ensemble, the size of each unfavorable pressure interval increases inversely with spectral acquisition rate. Other factors also affect the peak apex separation required for successful peak finding. Notable examples include the signal-to-noise ratios of the individual peaks and the ratio of peak amplitudes. These factors are not considered in this report. Complete Mixture Characterization. For the 200 Hz spectral acquisition rate used in these studies, a peak apex separation of 15-20 ms is adequate for automated peak finding if the fragmentation patterns for the two components are significantly different. Two of the compounds in the test mixture, p-xylene and m-xylene (components 12 and 13), have very similar fragmentation patterns. For the conditions used here, a peak apex separation of at least 300 ms is required for successful automated peak finding of these two components. This could not be achieved for junction-point pressure values less than 32 psia. Figure 6 shows a separation window diagram for the pressure range 33-40 psia. The peak-apex separation of the peak pair having the smallest separation (critical pair) is shown as the bold line in the figure. The predicted separation of peak pair 7/8 is less than 100 ms over this entire pressure range. For pressure values less than 35 psia, the separation of this peak pair is inadequate for automated peak finding. Note that there are a number of very sharp V-shaped features in the diagram. The sharpest of these features occur in the junction-point pressure ranges where an aromatic and a nonaromatic component change elution order. Figure 7 shows the analytical-ion chromatogram a and extractedion chromatograms b for the complete mixture using a junction(29) Smith, H.; Sacks, R. Anal. Chem. 1998, 70, 4960. (30) Grall, A.; Sacks, R. Anal. Chem., submitted for publication.
Figure 6. Peak separation window diagram for the 25-component mixture and a junction-point pressure range 33.0-40.0 psia. Numbers by the plots correspond to the component numbers in Table 1.
Figure 7. Analytical-ion chromatogram (top) and extracted-ion chromatograms (bottom) showing successful automated peak finding for the complete 25-component mixture. Peak numbers correspond to the component numbers in Table 1. The junction-point pressure value was 38.2 psia.
point pressure of 38.2 psia. This pressure is indicated by the broken vertical line in the window diagram of Figure 6. From the window diagram, this pressure should result in adequate peakpair separations for successful peak finding and complete characterization of the mixture. This is confirmed by examination of the extracted-ion chromatograms where all 25 component peaks have been found. This is remarkable considering that visual inspection of the analytical-ion chromatogram shows fewer than half of the 25 component peaks are adequately separated for successful characterization with scanning MS instruments. CONCLUSIONS The degree of retention pattern control that can be achieved with a pressure-tunable column ensemble using a nonpolar silicon
gum phase and a polar poly(ethylene glycol) phase is useful for the characterization of hydrocarbon mixtures using TOFMS with time-array detection. While large changes in peak separation for similar compounds cannot be achieved, large changes are not required in order to enhance the automated peak finding and spectral deconvolution features of the instrument software. From the window diagram analysis, it appears that more than adequate critical-pair separation can be achieved for the mixture described here, and thus, further reductions in analysis/characterization time may be obtained by the use of higher isothermal temperature or shorter columns. For GC/MS using scanning MS instruments, peak apex separations of about one peak width are required when unknown mixtures are characterized by the use of spectral libraries for electron ionization. For many of the hydrocarbon mixture components in this study, the sensitivity of peak-apex separation to changes in junction-point pressure is inadequate to achieve large reductions in analysis time without the use of spectral deconvolution. A peak apex separation of one average peak width results in a chromatographic resolution of 1.0. With automated peak finding and spectral deconvolution, the minimum required peakapex separation of 6 ms (500 spectra/s) suggests an increase in available peak capacity of more than 2 orders of magnitude can be obtained for comparable retention time intervals. A relatively narrow boiling-point range for hydrocarbon compounds was used for these isothermal studies. Since more than adequate separation is obtained for the later-eluting components 18-25, temperature programming could be used to obtain a further reduction in analysis time. For wider boiling-point range mixtures, temperature-programmed operation will be required. The use of a pressure-tunable column ensemble with columnoven temperature programming for HSGC has not been studied extensively. The application of pressure-tunable column ensembles to diesel and kerosene-range hydrocarbon mixtures will require the use of a polar column with greater thermal stability than the poly(ethylene glycol) column used in this study. Recent work has shown that trifluoropropyl polysiloxane is a useful polar phase for these applications.30 However, this phase is less polar than poly(ethylene glycol), and thus, the sensitivity of peak pair separations to changes in junction-point pressure will be reduced for some of the peak pairs. ACKNOWLEDGMENT The authors gratefully acknowledge the LECO Corp., St Joseph, MI, for use of the Pegasus II TOFMS instrument.
Received for review January 22, 2000. Accepted April 26, 2000. AC000081H
Analytical Chemistry, Vol. 72, No. 14, July 15, 2000
3069
