Simultaneous Temperature and Holdup Time Ratio Optimization for

Michael Akard† and Richard Sacks*. Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109. Analysis time for high-speed, capilla...
0 downloads 0 Views 285KB Size
Articles Anal. Chem. 1996, 68, 1474-1479

Simultaneous Temperature and Holdup Time Ratio Optimization for Tandem Column Tunable Selectivity with High-Speed GC Michael Akard† and Richard Sacks*

Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109

Analysis time for high-speed, capillary column GC is reduced by the use of a pressure-tunable tandem combination of a nonpolar poly(methylsiloxane) (DB-5) column and a polar trifluoropropyl poly(methylsiloxane) (Rtx-200) column operated isothermally at an optimized oven temperature. By adjusting the pressure at the midpoint between the two columns, the residence times of all sample components are adjusted to give the maximum resolution of the critical component pair. The result is a two-dimensional optimization where the column temperature and the midpoint pressure were adjusted to give the shortest possible analysis time. A previously defined relative resolution function, which requires only empirical capacity factor data for the target compounds, is used as the dependent variable in the optimization algorithm. The resulting three-dimensional resolution map is projected parallel to the relative resolution axis in order to obtain a useful two-dimensional display from which the optimal operating conditions can be determined. High-speed gas chromatography (HSGC) is gaining popularity in a variety of applications where short analysis time or high sample throughput is of critical importance.1-10 Special inlet systems,1-3 which produce very narrow vapor plugs, and short (2-10 M) capillary columns are used to separate relatively simple mixtures on a time scale from a few seconds to a few tens of seconds. However, for more complex mixtures, the reduced peak capacity of HSGC relative to conventional GC results in a greater † Present address; Chromatofast, Inc., 912 N. Main St., Suite 14, Ann Arbor, MI 48104. (1) Klemp, M.; Akard, M.; Sacks, R. Anal. Chem. 1993, 65, 2516-2521. (2) Liu, Z.; Zhang, M.; Phillips, J. J. Chromatogr. Sci. 1990, 28, 567. (3) Van Es, A.; Janssen, J.; Bally, R.; Cramers, C.; Rijks, J. J. High Resolut. Chromatogr. 1987, 10, 273. (4) Klemp, M.; Peters, A.; Sacks, R. Environ. Sci. Technol. 1994, 28, 369A. (5) Akard, M.; Sacks, R. Environ. Sci. Technol. 1994, 28, 428A. (6) Van Es, A.; Cramers, C.; Rijks, J. J. High Resolut. Chromatogr. 1989, 12, 303-307. (7) Van Es, A.; Jansen, J.; Bally, R.; Cramers, C.; Rijks, J. J. High Resolut. Chromatogr. 1987, 10, 273-279. (8) Ysacker, P.; Janssen, H-G.; Snijders, H.; Leclercq, P.; Cramers, Cl. J. Microcolumn Sep. 1993, 5, 413-419. (9) Peters, A.; Sacks, R. J. Chromatogr. Sci. 1991, 29, 403. (10) Rankin, C.; Sacks, R. LC-GC 1991, 9, 428.

1474

Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

probability of peak overlap.11 Tunable selectivity, where two columns of different selectivity are combined in tandem, can result in the much more efficient use of available peak capacity.10-15 This can result in drastic reductions in analysis time for more complex mixtures.5,12-15 Optimization strategies for tuning the selectivity of a tandem column ensemble often involve the adjustment of the midpoint pressure12-15 between the two columns or the relative lengths of the two columns.16 A resolution variable for the most poorly separated pair of components (critical pair) is then plotted versus either the midpoint tuning pressure, the carrier gas holdup time fraction of either column, or the equivalent length fraction of either column. This produces a series of peaks or windows of high resolution separated by valleys of zero resolution which correspond to coelution of the critical pair components. The window of greatest amplitude (greatest resolution of the critical pair) defines the optimal pressure, holdup time fraction, or column length fraction which will give the shortest analysis time. Column temperature is also an important variable in GC. Capacity factors of all components are dependent on column temperature. In conventional GC practice, isothermal or linearramped temperature programming is usually employed. Temperature programming is essential for the analysis of mixtures containing compounds with a wide range of volatilities. To use temperature programming in HSGC, the temperature programming rate should be greater than that in conventional GC. Potential problems associated with high-speed programming include irreproducibility in the temperature ramp, variability of temperature along the column (cold spots), and inadequate temperature sensing rates.17 As relatively small variations in temperature result in significant retention time changes, component identification is more difficult and less reliable. Reliable window diagrams for tunable selectivity also require accurate (11) Davis, J. M.; Giddings, J. C. Anal. Chem. 1983, 55, 418. (12) Akard, M.; Sacks, R. J. Chromatogr. Sci. 1993, 31, 297. (13) Deans, D.; Scott, I. Anal. Chem. 1973, 45, 1137. (14) Benicka´, E.; Krupcı´k, J; Kuljovsky, P.; Re´pka, D.; Garaj, J. Mikrochim. Acta 1990, 3, 1. (15) Matisova, E.; Kovacicova, E.; Garaj, J.; Kraus, G. Chromatographia 1989, 27, 494. (16) Sandra, P.; David, F.; Proot, M.; Diricks, G.; Verstappe, M.; Verzele, M. J. High Resolut. Chromatogr. Chromatogr. Commun. 1985, 8, 782. (17) Jain, V.; Phillips, J. J. Chromatogr. Sci. 1995, 33, 55. 0003-2700/96/0368-1474$12.00/0

© 1996 American Chemical Society

measurement of capacity factor values.18,19 For these reasons, most HSGC work has been done isothermally and has considered mixtures containing compounds with a rather limited range of volatilities. The choice of the column temperature is crucial to the separation quality. Computer calculations based on an extended form of the Golay-Giddings equation have shown that capacity factor values in the range 1-2 result in the greatest rate of generation of theoretical plates.20 Lower temperatures will increase the capacity factors of earliereluting components, generally improving their separation and increasing the peak capacity of the chromatogram. However, separation time increases rapidly with decreasing temperature. The role of temperature on column selectivity is not always adequately appreciated. The capacity factors change for all components as the temperature is varied, but they typically change to different extents. The result of this differential change in capacity factor values for the different components is that selectivity will be temperature dependent. Window diagrams have been used to predict the optimal temperature for an isothermal separation.21,22 The work described in this report considers the simultaneous optimization of isothermal column temperature and carrier gas holdup time fraction for a two-column tandem ensemble using a nonpolar poly(methylsiloxane) column followed by a moderately polar trifluoropropyl poly(methylsiloxane) column. The goal of this two-dimensional optimization is the determination of the coordinates in polarity-temperature space of the greatest critical pair resolution. EXPERIMENTAL SECTION Apparatus. A high-speed cryofocusing sample collection and inlet system was used to generate sample vapor plugs with widths in the 5-10 ms range. This system has been described in detail.1,4 The inlet system was mounted on the side wall of a Varian 3700 GC equipped with two flame ionization detectors. The nonpolar (DB-5) column was 2.5 m long, 0.25 mm i.d., and had a 0.25-µmthick stationary phase film. The polar (Rtx-200) column was 2.8 m long, 0.25 mm i.d., and also had a 0.25-µm stationary phase film. The midpoint between the two columns was connected to an H2 carrier gas source by means of a two-stage pressure regulator (Model 18A, Scott Specialty Gases, Troy, MI) and a fused silica capillary pneumatic restrictor. A PX304-050AV pressure transducer and a DP2000 digital meter (Omega Engineering, Stamford, CT) were used for pressure measurements. Column oven temperature was measured by the thermocouple supplied with the Varian 3700 GC. Since the inlet pressure and the outlet pressure at the detector were constant, an increase in the midpoint pressure resulted in a decrease in the pressure drop along the first column (nonpolar) and an increase in the pressure drop along the second column. This results in an increase in the carrier gas holdup time for the first column and a decrease in the holdup time for the second column. This increases the influence of the first column on the overall separation selectivity, with the result that the overall (18) Akard, M.; Sacks, R. Anal. Chem. 1994, 66, 3036. (19) Akard, M.; Sacks, R. Anal. Chem. 1995, 67, 2733. (20) Rankin, C. Ph.D. Thesis, University of Michigan, Ann Arbor, MI, 1991. (21) Matisova, E.; Kovacicova, E.; Garaj, J.; Kraus, G. Chromatographia 1989, 27, 494. (22) Repka, D.; Krupcik, J.; Bencika, E.; Maurer, T.; Engewald, W. J. High Resolut. Chromatogr. 1990, 13, 333.

polarity of the tandem column combination decreases. A decrease in the midpoint pressure has the opposite effect. For sample collection and cryofocusing, the capillary metal tube in the inlet system was cooled to about -100 °C by a continuous flow of cold nitrogen gas. For sample injection, the metal tube was ballistically heated using a capacitive discharge power supply and reached a peak temperature of about 150 °C. Between sample collection and injection, the gas flow direction through the metal trap tube was reversed so that the sample was injected from the same end of the trap tube from which it was collected. This flow reversal greatly reduces sample decomposition during the injection process and results in a narrower injection plug.1,4 The computer used for instrument control, data collection, and data analysis was a Gateway 2000 4DX2-66V (Gateway 2000, Sioux Falls, SD). The 12-bit A/D board was a CIO-AC16 (Computer Boards, Inc., Mansfield, MA). The software for the A/D interface was LabTech Notebook for Windows. An amplifier built in-house was used to offset the baseline signal. This results in improved utilization of the bipolar signal conversion available with the A/D board. The signals from the Varian 3700 FIDs were amplified by an electrometer amplifier constructed in-house and having a time constant of about 5 ms. Data sampling was performed at 100 Hz. Materials and Procedures. Gas-phase samples were prepared in Tedlar gas sampling bags. Liquid sample mixtures were injected into a bag by means of a microsyringe. Samples were diluted with dry air and allowed to equilibrate for at least 2 h. Sample collection in the cold trap tube was accomplished by inserting the capillary inlet restrictor from the cryofocusing inlet system into the septum of the gas sampling bag. A vacuum pump was used to pull sample into the cold trap tube. Gas sample size typically was about 0.5 mL. To obtain direct gas holdup time measurements for both columns in the tandem ensemble, a gas switching valve was used to route effluent from the first column directly to one of the Varian 3700 FIDs through a low-dead-volume fused silica capillary restrictor. This FID was only used for holdup time measurements. The switching apparatus is described fully in refs 5 and 12. Overall holdup time for the tandem column ensemble was measured using the other Varian FID. Holdup times were determined by sampling a bag partially filled with natural gas and diluted with air. The retention time of the first peak was then taken to be the overall holdup time. At the trapping temperatures used for this study, methane was not collected in the trap tube. The first peak detected from the trapped sample is probably a mixture of propane and butane, both of which are partially trapped. Since these components have very low retention at the column oven temperature used, reasonably reliable holdup time measurements were obtained. The capacity factor dependence on temperature was studied for the individual column types. For these studies, both columns in the ensemble were of the same type. For each component and for both column types, between 3 and 5 different temperatures were used for capacity factor determination. Oven temperatures ranged from 30 °C to 80 °C, and holdup times were determined for each temperature to account for viscosity changes. RESULTS AND DISCUSSION Capacity Factor Variations with Temperature. Analysis time depends directly on the capacity factor, klast, of the last component of interest as shown in eq 1. In eq 1, L is the column Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

1475

tr last ) (L/µj )(klast + 1)

(1)

length, µj is the average linear carrier gas velocity, and tr last is the retention time of the last component of interest. Capacity factors are dependent on the selective interactions of the solutes with the stationary phase and the saturation vapor pressure, P°, of the sample components at the oven temperature as shown in eq 2. In eq 2, P° is a function of the column

k∝

1 γP°(T)

(2)

temperature, T. Selective interaction with the stationary phase is expressed by the activity coefficient, γ. For the purpose of selecting the optimal temperature which will give the greatest resolution of the critical pair, an empirical relationship between the log of the capacity factor and the reciprocal of the column temperature is very convenient:

log k ) (m/T) + b

(23) Benicka´, E.; Krupcı´k, J.; Re´pka, D.; Sandra, P. Chromatographia 1992, 33, 463.

Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

DB-5

Rtx-200

k component (A) acetaldehyde (B) tert-butyl methyl ether (C) chloroform (D) tetrachloromethane (E) benzene (F) fluorobenzene (G) 3-pentanol (H) 3-methyl-1-butanol (I) 1-pentanol (J) 2-pentanone (K) tetrachloroethylene (L) 1,3-dichloropropane (M) chlorobenzene (N) 1-chlorohexane (O) cyclopentanone

k

calcd exptl 0.00 0.29 0.45 0.73 0.72 0.92 1.04 1.63 2.17 1.01 2.86 2.44 3.93 4.44 2.40

0.01 0.28 0.46 0.74 0.73 0.92 1.05 1.64 2.20 1.01 2.86 2.45 3.95 4.45 2.38

R2 0.079 0.992 0.993 0.997 0.998 0.999 0.999 0.999 0.999 0.999 1.000 1.000 1.000 1.000 1.000

calcd exptl 0.07 0.22 0.26 0.34 0.50 0.79 0.91 1.48 1.86 1.97 1.37 2.43 2.72 2.91 5.15

0.08 0.21 0.27 0.34 0.50 0.79 0.91 1.49 1.89 1.98 1.36 2.43 2.72 2.91 5.06

R2 0.887 0.967 0.975 0.982 0.995 0.997 0.999 0.999 0.998 0.999 0.999 0.999 1.000 1.000 1.000

(3)

If athermal effects are neglected, the slope, m, is nearly proportional to the heat of vaporization of the sample component. Values for the slope and intercept, b, were determined experimentally for the set of components listed in Table 1. Figure 1 shows plots of log k vs 1/T and error bars ((2σ) for the components listed in Table 1 using the Rtx-200 column. Note that the error bars are quite small for all components but A (acetaldehyde). The primary reason for the increased uncertainty for acetaldehyde is the very low capacity factors in the temperature range used here. At higher temperatures, this compound is nearly unretained and produces a very narrow peak. The precision of capacity factor determination is limited by the sampling frequency of the data acquisition system (100 Hz) as well as by the very small difference between the retention time for acetaldehyde and the holdup time. Table 1 also gives correlation coefficients (R2), calculated (linear regression) capacity factors at 40 °C, and experimental values for capacity factors at 40 °C for all compounds in the test mixture. Satisfactory results were obtained for all compounds except acetaldehyde. The majority of the correlation coefficients are greater than three nines, and the differences between predicted and actual capacity factors are relatively small. Thus, eq 3 is useful for the reliable prediction of capacity factors at all temperatures in the range studied. The differing slopes in Figure 1 illustrate the potential to tune selectivity with temperature alone. Benicka et al.23 have discussed the potential applications of using a single stationary phase with two different ovens for selectivity tuning. For any temperature in Figure 1 where the plots for different compounds cross, coelution will occur at that temperature, and a complete separation is impossible. Note that in Figure 1, component C has a smaller slope than components D and B, which elute before and after C, respectively, over most of the temperature range studied here. This means that, at higher temperatures, B and C would be well separated, while D and C would be close to coeluting. At lower temperatures, C and D would become better separated while C

1476

Table 1. Calculated and Experimental Capacity Factor Data at 40 °C on the DB-5 and Rtx-200 Columns

Figure 1. Plots of the log of the capacity factor versus the reciprocal of the absolute temperature for an Rtx-200 column with a 0.25-µm film thickness. Components are listed in Table 1. Error bars indicate (2 standard deviations.

and B would become less well separated. Component pairs I-J, H-K, and M-N would also exhibit significantly different resolution at different temperatures. Tandem Column Tuning. In the work described here, tandem (serially coupled) columns using different stationary phases were operated isothermally in the same oven. Twodimensional tuning was achieved by adjusting the oven temperature as well as the relative residence times of components in the two columns. To implement two-dimensional tuning, it is necessary to determine the effect of oven temperature on the two-column ensemble. First, capacity factor values for each component on the DB-5 and the Rtx-200 columns are determined at an arbitrary pair of temperatures using eq 3. Next, these values are used to determine an overall capacity factor, kov, for each component at each of the two specified temperatures using eq 4.18 In eq 4, ka

kov ) ka + (tmb/tm)(kb - ka)

(4)

and kb are the capacity factors for a given component on the two different phases at a specified temperature, and tm and tmb are

Figure 2. Two-dimensional window diagram plotting relative resolution (shading) versus oven temperature and holdup time fraction of Rtx-200. The dark regions are coordinates where components coelute. The brighter areas are peaks, or ridges. The ticks indicate the slices shown in subsequent figures where temperature and holdup time fraction are held constant.

the overall holdup time for the tandem ensemble and the holdup time for column b alone, respectively. This equation has been described previously.18 This is repeated for the second temperature, and the result is two capacity factors, kov1 and kov2, for the two temperatures, T1 and T2. These capacity factors are then used in eq 3 to determine slope and intercept values for the tandem ensemble. The optimization algorithm was based on the calculation of a previously described19 relative resolution parameter, R, given in eq 5. In eq 5, ∆kov is the difference in the overall capacity factors

R ) ∆kov/(kova + 1)

(5)

and kova is the average overall capacity factor for a critical pair using specified values for the holdup time fraction tmb/tm and column ensemble temperature T. The relative resolution is a reliable measure of separation quality over the entire range of capacity factor values and requires only capacity factor measurements. The simultaneous tuning of temperature and holdup time fraction of Rtx-200 is shown in Figure 2. This is a threedimensional window diagram similar to the one shown by Matisova´ et al.21 The horizontal axis is the holdup time fraction from eq 4. The vertical axis is the temperature of the tandem column ensemble. The shading in the figure corresponds to the values of relative resolution, with the light areas giving greater resolution of the critical pair. This corresponds to a projection parallel to the relative resolution axis. The dark regions correspond to the locus of temperature and holdup time fraction values where a specific critical pair coleute. Each dark trough corresponds to a different critical pair. Note that the troughs are not linear because of the nonlinear relationship between capacity factor and temperature. The lighter regions correspond to coordinates of relatively high relative resolution for the critical pair. A time fraction of 0 in Figure 2 corresponds to the separation being performed using the DB-5 column only. Note that the separation is poor at all temperatures when only this nonpolar

Figure 3. Plots of the log of the capacity factor versus the reciprocal of the absolute temperature. The data are derived for holdup time fractions of 0.47 on Rtx-200 and 0.53 on DB-5 with film thickness for both columns of 0.25 µm.

column is used. A time fraction of 1.0 corresponds to the separation being performed using only the Rtx-200 column. The optimal temperature (brightest region) when only this column is used occurs at a temperature of about 43 °C. This corresponds to the best compromise temperature for the B-C and C-D critical pairs. The best overall separation occurs at the coordinates of the brightest region in the figure. This would not necessarily be the shortest separation time, however. In general, the shortest separation time will occur at the highest temperature for which the critical pair is just separated with a specified resolution. In addition, for this mixture, the overall capacity factor of the last component increases with increasing polarity of the tandem column ensemble (larger tmb/tm values). Thus, for this mixture, separation time decreases toward the lower left portion of the resolution map of Figure 2. For mixtures in which the last-eluting component is nonpolar, the overall separation time will decrease with increasing polarity of the tandem column ensemble, and the shortest analysis times will occur in the lower right region of Figure 2. Validation of the Two-Dimensional Tuning Model. The tick marks on the plot at the 0.47 holdup time fraction and the 40 °C temperature indicate the coordinates of high critical pair resolution that were used to validate the tuning model. Figure 3 shows plots of log kov vs 1/T for a holdup time fraction (tmb/tm) of 0.47 on the DB-5/Rtx-200 tandem ensemble. The component labels are the same as in Figure 1. Note that the elution order changes from Figure 1 are the result of the addition of some DB-5 character to the overall phase selectivity. Using the tandem column ensemble, component K coelutes with J at low temperatures. At high temperatures, K and J are separated, but K and L coelute. The best separation appears to be at an intermediate temperature. From the data in Figure 3, a temperature optimization window diagram was constructed. In the window diagram, the value of relative resolution for the most poorly resolved component pair is plotted versus the isothermal column temperature. The resulting window diagram is shown in Figure 4. This window diagram corresponds to a vertical slice through Figure 2 at the location of the vertical tick marks. Since this region corresponds to a single bright zone, the corresponding window diagram contains a single high-resolution window. Note that the two flanks of the trace are not linear but show some curvature. This is Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

1477

Figure 4. Window diagram plotting the relative resolution of the critical pair versus the oven temperature. The data are derived for holdup time fractions of 0.47 on Rtx-200 and 0.53 on DB-5. The circles labeld a-c correspond the three chromatograms in Figure 5.

Figure 6. Overall capacity factors versus the holdup time fraction of Rtx-200 at 40 °C. Components are listed in Table 1. The horizontal lines labeled a-c correspond to similarly labeled chromatograms in Figure 8.

Figure 5. Chromatograms of the mixture listed in Table 1. The oven temperatures are 44, 38, and 33 °C for chromatograms a, b, and c, respectively. The chromatograms were obtained with holdup times fractions of 0.47 on Rtx-200 and 0.53 on DB-5.

Figure 7. Window diagram plotting relative resolution of the critical pair versus the holdup time fraction on Rtx-200. The points labeled a-c correspond to similarly labeled chromatograms in Figure 8.

expected since the capacity factors vary nonlinearly with temperature. Figure 3 indicates that the best separation will occur at a temperature of about 38 °C. The circles around the data points in Figure 4 at temperatures of 33, 38, and 44 °C correspond to the chromatograms c, b, and a, respectively, in Figure 5. Note that the best separation occurs as predicted at a temperature of 38 °C (chromatogram b) and not at the lowest temperature. At both the higher and lower temperatures (chromatograms a and c, respectively), several inadequately resolved components are observed. For chromatogram b, all components are adequately resolved. The utility of isothermal temperature optimization is apparent. The effect of temperature on overall separation time is also significant. It can be seen in Figure 5 that the retention time of the last component drops from almost 22 s at 33 °C to just over 15 s at 44 °C. The primary consideration is the separation of the critical pair, and the highest temperature that provides adequate resolution of the critical pair will give the shortest overall separation time at a given holdup time fraction. Figure 6 shows plots of the overall capacity factor versus the holdup time fraction of Rtx-200 at 40 °C for all components in the test mixture. This corresponds to a horizontal slice at the tick marks in Figure 2. Positive slopes of traces indicate greater retention on the Rtx-200 column (polar compounds), while negative slopes indicate greater retention on the DB-5 column. Holdup time fractions for which the traces cross will result in coelution of the corresponding compounds. Holdup time fractions

between the crossing points will give windows of greater relative resolution. The isothermal window diagram for the mixture is shown in Figure 7. For this window diagram, the relative resolution of the critical pair is plotted as a function of the holdup time fraction, tmb/tm. Zero values of relative resolution give the corresponding holdup time fractions where coelutions occur. Significantly more structure is observed in the window diagram of Figure 7 than in the window diagram of Figure 4. This is because elution order is more sensitive to phase polarity changes than to temperature changes. Also, note that the line segments in Figure 7 are straight, as expected from the linear relationship of overall capacity factor to holdup time fraction (see eq 4). The critical pairs are indicated for each window in the figure. The circles labeled a-c in Figure 7 and the vertical lines in Figure 6 correspond to the chromatograms of Figure 8. From the window diagram, it can be seen that the time fraction of 0.45 for point c is at the best of the time fractions evaluated. As expected, chromatogram c demonstrates the best separation in Figure 8. For chromatogram a, using a holdup time fraction of 0.21, components M and O nearly coelute. For chromatogram b, using a holdup time fraction of 0.30, several inadequately resolved pairs are observed. Using two-dimensional selectivity tuning with a GC system designed for high-speed operation as described in this report, a 15-component mixture was completely separated in less than 20 s using 0.25-mm-i.d. capillary columns. The use of the cryofocusing inlet system results in a very precise zero time reference

1478 Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

Figure 8. Chromatograms for the components listed in Table 1 at a temperature of 40 °C. The chromatograms a, b, and c correspond to holdup time fractions on Rtx-200 of 0.21, 0.30, and 0.45, respectively.

for the reliable measurement of capacity factors. This greatly facilitates the tuning operation. The experimental approach is quite straightforward, and only capacity factor versus temperature data are required for the target compounds using the two columns.

Isothermal column temperature can be used as a tuning variable for a single column or a tandem column ensemble. The overall optimization of a separation must take temperature into account because the total analysis time is a function of the temperature. While temperature adjustment generally results in smaller changes in selectivity than polarity adjustment, the optimization of temperature can significantly reduce analysis time. Use of temperature and holdup time fraction on a tandem column gives two independent variables with which to tune a separation. This added dimension dramatically increases the size of the parameter space for selectivity tuning. Further work is needed in the area of selectivity tuning using two independently controlled ovens. Each column can be operated at a different isothermal temperature. High-speed temperature programming would also be very useful if the inherent problems could be overcome. Received for review October 9, 1995. Accepted February 16, 1996.X AC951021S X

Abstract published in Advance ACS Abstracts, March 15, 1996.

Analytical Chemistry, Vol. 68, No. 9, May 1, 1996

1479