Anal. Chem. 1999, 71, 2123-2129
Temperature Programming for High-Speed GC Carrie Leonard, Andrew Grall, and Richard Sacks*
Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109
Fast temperature programming (20-50 °C/min) is used with relatively short separation columns to achieve highspeed separations of mixtures covering a wide boiling point range. A cryofocusing inlet is used to obtain narrow injection plugs. High-speed temperature-programmed chromatograms are evaluated by considering local peak capacity as a function of carbon number and boiling point for the normal alkanes in the range C8-C19. The peak capacity generation rate (peaks per second) as a function of carbon number and the total cumulative peak capacity as a function of time are also considered for various column lengths and carrier gas flow rates. Column lengths in the range 3.6-25.4 m and average carrier gas velocity values in the range 50-200 cm/s are considered. For a 6.8-m-long, 0.25-mm-i.d. column operated at an average carrier gas velocity of about 100 cm/s and using a nominal programming rate of 50 °C/min, C19 elutes in 178 s with a total peak capacity of 168 peaks. If the programming rate is reduced to 20 °C/min, the C19 elution time more than doubles but the total peak capacity increases by only 20%. For a 25.4-m-long column using a nominal 50 °C/min programming rate, the C19 retention time is 262 s with a peak capacity of 279 peaks. The use of average carrier gas flow rates greater than about 100 cm/s, which is common in isothermal high-speed GC, results in a considerable loss in total peak capacity with remarkably little reduction in analysis time. Pressing needs to increase sample throughput and reduce analysis costs have prompted widespread interest in the development of methods for high-speed speciation and analysis of organic carbon. The positive attributes of gas chromatography (GC) for the analysis of organic carbon have long been recognized, and in recent years the development of high-speed gas chromatography (HSGC) has received considerable attention.1-10 High-speed isothermal separations are achieved by the use of relatively short (1) Sacks, R.; Smith, H.; Nowak, M. Anal. Chem. 1998, 70, A37. (2) Overton, E. B.; Dharmasena, H. P.; Carney, K. R. Field Anal. Chem. Technol. 1997, 1, 87. (3) Borgerding, A. J.; Wilkerson, C. W., Jr. Anal. Chem. 1996, 68, 2874. (4) Nowak, M.; Gorsuch, A.; Smith, H.; Sacks, R. Anal. Chem. 1998, 70, 2481. (5) Klemp, M.; Akard, M.; Sacks, R. Anal. Chem. 1993, 65, 2516. (6) Li, W. C.; Andrews, A. R. J. J. High Resolut. Chromatogr. 1996, 19, 492. (7) van Es, A.; Janssen, J.; Cramers, C.; Rijks, J. J. High Resolut. Chromatogr. Chromatogr. Commun. 1988, 11, 852. (8) Akard, M.; Sacks, R. Anal. Chem. 1995, 67, 2733. (9) Smith, H.; Sacks, R. Anal. Chem. 1997, 69, 5159. (10) Smith, H.; Sacks, R. Anal. Chem., in press. 10.1021/ac9812936 CCC: $18.00 Published on Web 04/28/1999
© 1999 American Chemical Society
columns and high carrier gas flow rates. Microbore columns also are used frequently to achieve higher separation speed.11 Important advancements have involved the development of inlet devices capable of generating the narrow injection plugs needed for HSGC.4-7 Devices are now available which can deliver to the column sample vapor plugs less than 10 ms (σ values) in width. The development of strategies for managing the reduced peak capacity (number of peaks that will fit in the chromatogram at a specified resolution) associated with the use of shorter columns and higher carrier gas flow rates has received considerable attention.8-10 Tunable and programmable column selectivity has been developed in order to structure chromatograms and utilize the available peak capacity more efficiently. In some cases, this has resulted in 1-2-order-of-magnitude reductions in separation time. The development of comprehensive two-dimensional GC has resulted in dramatic increases in peak capacity.12 Until recently, most HSGC work has involved isothermal separations. In part, this was due to the inadequacies of most laboratory GC instruments. High-speed separations covering a wide boiling point range can be achieved only by the use of relatively high rates of temperature programming. The development of instrumentation for high-speed temperature programming is progressing rapidly,13-15 and commercial instruments are now available with programming rates in excess of 10 °C/s. These instruments also achieve more rapid cooldown, which is important for reducing cycle time. However, most of the current generation of laboratory GC instruments have maximum available programming rates limited to about 1-2 °C/s or less. Temperature programming for conventional GC has been studied in detail.16-25 The monograph by Harris and Habgood20 is noteworthy. Several temperature-programmed GC modeling (11) van Es, A. High-Speed Narrow Bore Capillary Gas Chromatography; Huthig Buch Verlag: Heidelberg, Germany, 1992. (12) Liu, Z.; Sirimanne, S. R.; Patterson, D. G., Jr.; Needham, L. L.; Phillips, J. B. Anal. Chem. 1994, 66, 3086. (13) Ehrmann, E. U.; Dharmasema, H. P.; Carney, K.; Overton, E. B. J. Chromatogr. Sci. 1996, 34, 533. (14) Hail, M.; Yost, R. A. Anal. Chem. 1989, 61, 2410. (15) MacDonald, J. J.; Wheeler, D. Int. Lab. 1998, 28, 6. (16) Dose, E. Anal. Chem. 1987, 59, 2414. (17) Curvers, J.; Rijks, J.; Cramers, C.; Knauss, K.; Larson, P. J. High Resolut. Chromatogr. 1998, 8, 607. (18) Dolan, J.; Snyder, L.; Bautz, D. J. Chromatogr. 1991, 541, 1. (19) Snow, N. H.; McNair, H. M. J. Chromatogr. Sci. 1992, 30, 271. (20) Harris, W. E.; Habgood, H. W. Programmed Temperature Gas Chromatography; Wiley: New York, 1966. (21) Bakeas, E. B.; Siskos, P. A. Anal. Chem. 1996, 68, 4468. (22) Panda, S.; Bu, Q.; Huang, B.; Edwards, R. R.; Laio, Q.; Yen, K. S.; Parcher, J. F. Anal. Chem. 1997, 69, 2485. (23) Frame, G. M.; Cochran, J. W.; Bowadt, S. S. J. High Resolut. Chromatogr. 1996, 19, 657. (24) Guiochon, G. J. Anal. Chem. 1977, 50, 1812.
Analytical Chemistry, Vol. 71, No. 11, June 1, 1999 2123
studies have considered the prediction of retention times based on the column holdup time and the thermodynamic properties of the analytes.15-18 Peak capacity in temperature-programmed GC has received considerable attention. Numerous reports of the optimization of initial temperature and programming rate for specified sets of target compounds have appeared.21-23 The effects of column length and flow rate on the rate of production of theoretical plates and peak capacity have been studied in detail for high-speed isothermal GC.24-26 The selection of column length, carrier gas flow rate, and temperature-programming rate specifically for the case of HSGC has received less attention. The situation for high-speed temperature-programmed GC is more complex, since the higher boiling point components typically show minimal migration for most of the duration of the analysis. This study was undertaken to explore the use of relatively high rates of temperature programming for the purpose of extending the domain of HSGC to wide boiling point range mixtures, which would be impractical to separate under isothermal conditions. The objective was the determination of operating conditions, which significantly reduce analysis time with minimal sacrifice in resolution and peak capacity. The empirical study described here considers the selection of column length, carrier gas flow rate, and temperature-programming rate for separations in the time frame 1-5 min using programming rates in the range 20-50 °C/min (0.33-0.83 °C/ s). These rates are available from most laboratory GC instruments. Principal evaluation criteria include the separation number (TZ) and the rate of peak capacity generation. The sacrifice of peak capacity for higher speed and the tradeoffs of boiling point resolution and boiling point range are considered in this report. EXPERIMENTAL SECTION Apparatus. All separations were performed using a Varian 3500 capillary gas chromatograph (Varian, Walnut Creek, CA). A flame ionization detector was used at a temperature of 250 °C. Columns were 0.25-mm-i.d. nonpolar dimethylpolysiloxane (DB1, J&W Scientific, Folsom, CA), with a stationary-phase film thickness of 0.25 µm. The carrier gas was prepurified hydrogen. Initial sample injections were performed using a split inlet heated to 250 °C. Split ratios varied from 370:1 to 480:1 in order to prevent column loading. A Gateway 2000 4DX2-66V PC (Gateway 2000, Sioux Falls, SD) with a 16-bit A/D board (Computer Boards Inc., Mansfield, MA; model CIO-DAS-1600) and Labtech Notebook software (Laboratory Technologies Inc., Wilmington, MA) were used to control data acquisition. The chromatograph was modified for HSGC by the addition of a cryofocusing inlet (Model L Cryointegrator, Chromatofast, Inc., Ann Arbor, MI) and a high-speed electrometer, built in house. Sample is introduced to the high-speed inlet by injection into the conventional split inlet of the Varian gas chromatograph. The highspeed inlet system uses a vacuum pump to pull sample vapor and carrier gas from the split inlet through a capillary metal trap tube, which is cooled to at least -80 °C. After sample collection, the trap tube is pressurized with carrier gas, and the tube is resistively (25) Ogan, K.; Scott, R. P. J. High Resolut. Chromatogr. Commun. 1984, 7, 382. (26) Said, A. S. J. High Resolut. Chromatogr. Commun. 1983, 6, 200.
2124 Analytical Chemistry, Vol. 71, No. 11, June 1, 1999
Figure 1. Temperature-time profiles for nominal programming rates of 50 (B), 45 (C), 40 (D), 35 (E), 30 (F), and 25 (G) °C/min. Trace A is for a linear programming rate of 50 °C/min.
heated by a capacitive-discharge power supply. This injects a sample vapor plug about 10 ms in width into the separation column.5 Procedures. Solutions of n-alkanes from hexane to nonadecane were prepared by combining the pure components and diluting with n-pentane to achieve the desired concentrations. Relatively high concentrations were used (0.35-0.4%) in order to minimize coelution of early eluting alkanes with the solvent. For all temperature-programming experiments, the initial column oven temperature was 50 °C, and temperature-programming rates ranged from 20 to 50 °C/min. The temperature program was initiated simultaneously with injection from the high-speed inlet system. To determine if initial injection plug width from the high-speed inlet system varied with component boiling point, individual n-alkanes were injected onto a 3.8-m column operated with an average hydrogen carrier gas velocity of about 100 cm/s. The isothermal oven temperature for each compound was adjusted to maintain a constant retention time of 10 s. Chromatograph oven performance was evaluated by recording the actual oven temperature profile at varying nominal programming rates. The oven was programmed from 50 to 250 °C. A 0.07mm type J thermocouple wire (Omega Engineering, Stamford, CT) was suspended in the center of the column oven. The analog signal from a thermocouple meter (DP116-JC1, Omega Engineering, Stamford, CT) was recorded by computer. RESULTS AND DISCUSSION Figure 1 shows oven temperature vs time plots obtained from the thermocouple located in the GC oven. Traces labeled B-G are for nominal programming rates of 50, 45, 40, 35, 30, and 25 °C/min, respectively. Line A provides a reference for a linear program rate of 50 °C/min. Good program linearity is achieved for the 25 and 30 °C/min cases over the entire temperature range. For the higher programming rates, significant curvature is observed at higher temperatures. For the 50 °C/min case, large deviations from linearity are observed for temperatures greater than about 170 °C. Similar curvature was reported for a HewlettPackard 5890 gas chromatograph for a nominal programming rate of 60 °C/min.13 The fused-silica columns used here are very thin
Table 1. Peak Widths, Isothermal Temperatures, and Retention Times for C8, C10, C12, C14, C16, and C18 Normal Alkanes compd
temp, C°
peak width, ms
retention time, s
C8 C10 C12 C14 C16 C18
35 72 100 130 155 178
127 136 146 149 135 138
9.74 9.99 9.90 10.14 9.87 9.85
walled and thus should have good thermal accommodation. Thus, it is assumed that, even at the relatively high programming rates used here, the internal column temperature does not deviate significantly from the oven temperature. The compounds in the test mixture cover a boiling point range from 68 (C6) to 330 °C (C19). The effect of compound boiling point on the injection plug width obtained from the cryofocusing inlet system has not been reported previously. While injection plug widths were not measured in the present study, isothermal highspeed chromatograms were obtained for the individual n-alkanes under conditions where significant differences in the injection plug width could be observed. Table 1 shows peak widths expressed as standard deviations, retention times, and isothermal column temperatures for the n-alkanes with even carbon numbers. The column temperature was adjusted to give a retention time of about 10 s for each of the compounds. Since the on-column time for each compound was very short and nearly constant, any significant changes in the bandwidths of the eluted compounds could be attributed to changes of the injection plug width from the highspeed injection device. Over the entire range, peak widths vary by no more than 22 ms, and no trends are observed. Optimization Strategies. Much work has been reported describing the optimization of isothermal separations with respect to analysis time.24-26 Two useful parameters defining the temporal efficiency of a separation are the separation time ts and the peak capacity np. These are given by eqs 1 and 2, respectively, for isothermal separations27 where L is the column length, u is the
ts ) (L/u)(k + 1)
(1)
np ) 1 + [(L/H)1/2/4Rs] ln(ts/tm)
(2)
average carrier gas velocity, k is the retention factor of the last eluting component of interest, H is the plate height for the column, Rs is the required resolution, and tm is the column holdup time. Equation 2 assumes that H is constant over the separation. In general, H shows considerable variation with retention,28 and more accurate expressions for isothermal peak capacity have been reported.28,29 The requirements and tradeoffs for reducing separation time in isothermal separations are clear. Separation time is reduced by using shorter columns and higher carrier gas velocities. The tradeoff is the loss of peak capacity. Note that, to double the peak (27) Giddings, J. C. Anal. Chem. 1969, 39, 1027. (28) Krupcik, J.; Garaj, J.; Cellar, P.; Guiochon, G. J. Chromatogr. 1984, 312, 1. (29) Shen, Y.; Lee, M. L. Anal. Chem. 1998, 70, 737.
Figure 2. Chromatograms using a nominal programming rate of 40 °C/min: (a) 6.8-m column with 96 cm/s gas velocity; (b) 3.6-m column with 96 cm/s gas velocity; (c) 3.6-m column with 197 cm/s gas velocity.
capacity, the column length and thus the separation time must be increased by a factor of 4 (this neglects gas compression effects). If the peak capacity is divided by the separation time, the average number of peaks per second that can be generated (separated) by the column is obtained. This is a useful measure of overall efficiency of the separation with respect to the utilization of time. Optimization strategies for HSGC using temperature programming have not been adequately investigated, and the requirements for obtaining a high rate of peak production (np/ts) are less clear. However, it is widely accepted that high rates of temperature programming are required for high-speed separations and that longer columns may lose resolution at high programming rates. The explanation for this loss is that retention factors may fall too quickly, and thus inadequate interaction with the stationary phase may occur in a significant portion of the column length. Figure 2 shows high-speed chromatograms obtained using a nominal programming rate of 40 °C/min beginning at the time of injection. The starting temperature was 50 °C. The mixture contained C6-C19 normal alkanes in a pentane solvent. Peaks labeled 8, 9, and 19 are for the corresponding alkanes. Chromatogram a was obtained with a 6.8-m-long column operated at an average carrier gas velocity of 96 cm/s. Thus the holdup time is 7.1 s. The separation time is 208 s, and the total peak capacity (C6-C19) is about 188 peaks at a resolution of 1.18. For chromatogram b, the column length was reduced to 3.6 m but the carrier gas velocity was unchanged. Here, the separation time is reduced to 172 s, and the peak capacity is reduced to 136 peaks. For chromatogram c, the average carrier gas velocity was increased to 197 cm/s but the column length was unchanged at 3.6 m. The separation time is 154 s, and the peak capacity has been reduced to 95 peaks. Thus, doubling the carrier gas velocity Analytical Chemistry, Vol. 71, No. 11, June 1, 1999
2125
resulted in only about a 10% reduction in separation time but caused a 30% loss in peak capacity. The small dependence of analysis time with respect to carrier gas velocity illustrated by chromatograms b and c is noteworthy. In isothermal GC using the same column, a similar increase in gas velocity would result in some loss in peak capacity, due to reduced column efficiency. However, the separation time would be cut in half, and the temporal efficiency of the separation (average peak capacity per second) would increase. The opposite is seen in Figure 2. The relatively weak dependence of separation time on column length observed upon comparing chromatograms a and b is also noteworthy. Under isothermal conditions, the column lengths and carrier gas velocities used for chromatograms a and c would result in a reduction of analysis time by a factor of 4. Using temperature programming with a 40 °C/min rate, the analysis time is reduced by only 26%. These results confirm that, for high-speed separations using relatively fast temperature programming, retention times are far less dependent on column length and carrier gas velocity than is the case for isothermal HSGC separations. Total peak capacity and temporal efficiency, however, are strongly dependent on these column parameters. The holdup time for the chromatograms in Figure 2 ranges from about 7.1 s in (a) to about 1.8 s in (c). This is clearly seen in the early portions of the chromatograms where the retention times for C8 and C9 decrease by nearly a factor of 4 from (a) to (c). These early eluting compounds have appreciable migration rates at the starting temperature of 50 °C, and thus their retention times are influenced more by the holdup time (L/u) and less by the temperature-programming rate. Efficiency Criteria. Equation 2 is not useful for predicting peak capacity in temperature-programmed HSGC. Peak capacity in temperature-programmed GC is often measured locally over a restricted region of the chromatogram. A very useful measure is the Trennzahl number TZ,30 which is defined as the number of peaks that will fit between two reference peaks with a resolution of 1.18 (31). Ettre32 has discussed the use of TZ in detail. Grob33 described TZ as the only measure of separation efficiency that is compatible with temperature-programmed GC. Usually, the set of normal alkanes is used as the reference peaks, and TZ is defined as the number of peaks that will fit between an adjacent pair of the alkanes.
TZ ) [∆tR/
∑W ] - 1 h
(3)
Here ∆tR is the difference in retention times for the two reference peaks and ∑Wh is the sum of their half-height peak widths. The number of peaks (np)Rs that will fit between a pair of normal alkanes at any specified resolution Rs can be described by eq 4.
(np)Rs ) (1.18/Rs)TZ
(4)
The cumulative peak capacity (np)i,j of a chromatogram extending (30) Kaiser, R. E. Chromatographie in der Gasphase, 2nd ed.; Bibliografisches Institut: Mannheim, Germany, 1966; Vol. 2. (31) Hurrell, R. A.; Perry, S. G. Nature (London) 1962, 196, 571. (32) Ettre, L. A. Chromatographia 1975, 80, 291. (33) Grob, K., Jr.; Grob, K. J. Chromatogr. 1981, 207, 291.
2126 Analytical Chemistry, Vol. 71, No. 11, June 1, 1999
Figure 3. Plots of TZ values (a) and peak capacity production rates (b) for a 15-m column with carrier gas velocity values of 49.5 (A), 95.7 (B), 148 (C), and 198 (D) cm/s.
from normal alkane i to normal alkane j is found by summing the TZ values and adding the peaks from the reference compounds.
(np)i,j )
∑
i,j(TZ
+ 1)
(5)
Local Peak Capacity. Figure 3a shows plots of TZ values from high-speed chromatograms obtained with a temperature-programming rate of 30 °C/min versus the mean carbon number CH of the reference alkanes. For example, the TZ value for the interval C10-C11 is plotted at a carbon number (CH value) of 10.5. For all plots, the column length was 15.0 m. Plots A-D are for average carrier gas velocity values of 49.5, 95.7, 148, and 198 cm/s, respectively. A programming rate of 30 °C/min was used because good linearity was observed in the measured temperature-time profile (see Figure 1). These local TZ plots provide direct information regarding boiling point resolution on the nonpolar column. Since the boiling point range of each TZ interval is fixed by the boiling points of the reference alkanes, the boiling point difference required for the separation of two compounds that differ only in boiling point can be determined. For example, for CH ) 14.5, the TZ value is about 19.3 for the 49.5 cm/s case (Figure 3a, plot A), and the boiling point difference is 17 °C. Thus the boiling point resolution is equal to about 1.1 peaks/°C. For components C6 to about C9, which have appreciable migration rates at the starting temperature of 50 °C, relatively smaller TZ values are obtained because of relatively small initial retention factors and thus reduced interaction with the stationary phase. For the high-boiling-point components, the TZ values decrease nearly linearly with increasing average carbon number. This decrease has been attributed to the fact that the retention time difference between adjacent homologues decreases with increasing molecular weight and with increasing column temper-
ature.34,35 Jones et al.36 have examined in detail the linearity of this trend. For most of the boiling point range considered here, the TZ values are very similar for the 49.5 and 95.7 cm/s plots. The corresponding analysis times (C19 elution time) are 360 and 315 s, respectively. Thus, the higher gas velocity gives greater boiling point resolution per unit time. For the gas velocities of 148 and 198 cm/s, the TZ values are considerably smaller. In Figure 3b, the TZ values from part a are divided by the corresponding retention time intervals to obtain values for the number of peaks generated per second. For each CH value, this is a direct measure of the overall temporal efficiency of the separation in the corresponding region of the chromatogram. For all gas velocities studied, temporal efficiency is greatest for the low-boiling-point (early eluting) components and decreases gradually with increasing boiling point. Note that, for the higher-boilingpoint compounds, the boiling point resolution decreases fairly rapidly with increasing CH but the rate of peak production decreases very slowly. The early eluting compounds in this study benefit more from the narrow injection plugs produced by the cryofocusing inlet system. This results in a high rate of peak generation with the 95.7 cm/s case being the most efficient. The higher-boiling-point components are essentially frozen at the mouth of the column at the 50 °C starting temperature. For this case, retention is controlled primarily by the temperature-programming rate, and the rate of peak capacity production shows only a weak dependence on boiling point. In this boiling point region, the lower gas velocities, which are closer to the optimum value for this columncarrier gas combination, give the greater temporal efficiencies. For all conditions tested, peak widths increase gradually with increasing carbon number (increasing retention time). For the case of a 15-m-long column operated with an average gas velocity of 96 cm/s, the increase is about 20 ms/carbon in the range C12C19. Jones et al.36 reported comparable relative increases in peak widths for several homologous series using programming rates up to 17 °C/min. Since injection plug widths show no significant change with carbon number for the C8-C19 range, the broader peaks from the higher-boiling-point compounds are the result of on-column band broadening. Some of this broadening may occur during the time interval that the peaks are nearly frozen at the mouth of the column. It is clear that this broadening causes the gradual decrease in the peak capacity generation rate with increasing CH in Figure 3b. Figure 4 shows similar plots but for four different column lengths, with an average carrier gas velocity of about 100 cm/s. Plots A-D are for column lengths of 3.6, 6.8, 15.0, and 25.4 m, respectively. Again, the temperature-programming rate was 30 °C/ min. For the boiling point range above C12, the TZ values for the 25.4-m column are about twice those of the 3.8-m column. The corresponding analysis times are 359 and 218 s. The peaks-per-second plots in Figure 4b show that, for the higher-boiling-point range, where elution times are controlled (34) Rooney, T. A.; Hartigan, M. J. J. High Resolut. Chromatogr. Commun. 1980, 3, 416. (35) Jennings, W.; Yabumoto, K. J. High Resolut. Chromatogr. Commun. 1980, 3, 177. (36) Jones, L. A.; Kirby, S. L.; Garganta, C. L.; Gerig, T. M.; Mulik, J. D. Anal. Chem. 1983, 55, 1354.
Figure 4. Plots of TZ values (a) and peak capacity production rates (b) for a carrier gas velocity of 95.7 cm/s with column lengths of 3.6 (A), 6.8 (B), 15 (C), and 25.4 (D) m.
more by the temperature-programming rate than by the column holdup time, the rate of peak capacity production increases steadily with increasing column length. For the lower-boiling-point compounds, which have appreciable migration rates at the starting temperature, the shortest column gives the greatest number of peaks per second. This is consistent with results for high-speed isothermal separations. Plots similar to those in Figures 3 and 4 were obtained for programming rates of 20, 30, 40, and 50 °C/min. Results for programming rates of 30 and 50 °C/min are summarized in Table 2 for CH ) 14.5. In general, the TZ values show only a weak dependence on programming rate, with the 50 °C/min rate giving the smaller values. The largest differences between the 30 and 50 °C/min rates occur for the longer columns operated at the lower carrier gas velocities. Also note that, for the 3.6- and 6.8-m columns, the TZ values decrease steadily with increasing gas velocity. For the 15- and 25.4-m columns, the largest TZ values occur for the 100 cm/s case. These results are explained by noting that, for a fixed programming rate, elution temperatures increase with increasing column length and with decreasing carrier gas velocity. For the 15- and 25.4-m columns at the lowest gas velocity, the elution temperatures are so high that, over a significant portion of the column length, inadequate interaction with the stationary phase occurs and the TZ values are lower. For these columns, an increase in carrier gas velocity from about 50 to about 100 cm/s reduces the elution temperatures sufficiently to give larger TZ values, despite poorer column efficiency at the higher gas velocity. Cumulative Peak Capacity. Figure 5a shows plots of the cumulative (total) peak capacity versus retention time for the boiling point range C8-C19 using nominal temperature-programming rates of 50 (A), 40 (B), 30 (C), and 20 (D) °C/min. The cumulative peak capacity is found from eq 5. For the interval CnCn+1, the retention time of Cn+1 is used for these plots. Thus, each Analytical Chemistry, Vol. 71, No. 11, June 1, 1999
2127
Table 2. TZ Values for CH ) 14.5 Using Various Column Lengths and Average Carrier Gas Velocities with Nominal Programming Rates of 30 and 50 °C/min 50 cm/s
3.6-m column 6.8-m column 15.0-m column 25.4-m column
100 cm/s
150 cm/s
200 cm/s
50 °C/min
30 °C/min
50 °C/min
30 °C/min
50 °C/min
30 °C/min
50 °C/min
30 °C/min
11.6 15.3 16.4 16.9
12.8 16.0 19.3 20.0
10.0 13.9 17.8 19.4
10.8 14.4 19.9 21.6
8.4 11.8 14.7 16.3
9.2 12.0 15.9 17.6
7.5 9.6 12.4
8.1 10.0 13.0
Cumulative peak capacity versus time plots using a and a carrier gas velocity of 95.7 cm/s for program50 (A), 40 (B), 30 (C), and 20 (D) °C/min. (b) for the 50 °C/min case.
Figure 6. Cumulative peak capacity versus time plots for a 50 °C/ min nominal programming rate: (a) 15-m column with carrier gas velocities of 49.5 (A), 95.7 (B), 148 (C), and 198 (D) cm/s; (b) 95.7 cm/s carrier gas velocity with column lengths of 3.6 (E), 6.8 (F), 15 (G), and 25.4 (H) m.
point represents the retention time for the corresponding alkane and the total peak capacity generated from C8 up to that time in the chromatogram. All plots are for the case of a 25.4-m-long column and nominal-average carrier gas velocity of about 100 cm/ s. The rightmost point in each plot corresponds to the retention time of C19 and thus the analysis time for the mixture. For the 20 °C/min nominal programming rate, a peak capacity of nearly 294 peaks is generated in about 493 s. For a nominal programming rate of 30 °C/min, the peak capacity drops only slightly to 271 peaks but the analysis time is reduced to about 359 s. For a nominal rate of 50 °C/min, the cumulative peak capacity falls to about 238 peaks and the analysis time falls to about 262 s. Thus, as the programming rate is increased from 20 to 50 °C/s, the analysis time is reduced by 47% while the total peak capacity is reduced by only 19%. The slopes of these plots at any values along the time axis give the rates of peak production at the corresponding points in the chromatogram. Over most of the time range, the use of the highest programming rate results in the greatest overall efficiency with respect to the utilization of time. Comparable results were obtained using column lengths of 3.8 and 16 m. For the boiling point range above C12, the plots in Figure 5 are relatively linear.
The linearity decreases as the nominal temperature-programming rate increases. This may be the result of increasing negative deviation from the nominal temperature as the programming rate increases (see Figure 1). Figure 5b shows the chromatogram corresponding to plot A in Figure 5a. Note that considerable peak capacity is generated prior to C8. When this is added to the total, a peak capacity of 279 peaks is obtained in 262 s. This corresponds to an average peakcapacity-generation rate of nearly 1.1 peaks/s. Figure 6 shows plots of cumulative peak capacity versus time for several average carrier gas velocity values using a 15-m-long column (part a) and for several column lengths using an average carrier gas velocity of 100 cm/s (part b). All plots are for a 50 °C/min nominal programming rate. For Figure 6a, plots A-D correspond to carrier gas velocities of 49.5, 95,7, 148, and 198 cm/s, respectively. Plot B has the steepest slope over the entire separation time, indicating the greatest rate of peak generation. The tradeoff of speed for peak capacity is clearly illustrated. For plot D (198 cm/s), the entire boiling point range from C8 to C19 is eluted in under 190 s but the total peak capacity is only 158 peaks. For relatively simple mixtures spanning a wide boiling point range, large peak capacity may not be needed and the use of
Figure 5. (a) 25.4-m column ming rates of Chromatogram
2128 Analytical Chemistry, Vol. 71, No. 11, June 1, 1999
Figure 7. Plots of cumulative peak capacity (a) and analysis time (b) versus carrier gas velocity for a 50 °C/min nominal programming rate and column lengths of 3.6 (A), 6.8 (B), 15 (C), and 25.4 (D) m.
higher carrier gas velocities can result in significantly faster separations. For plots E-G in Figure 6b, column lengths of 3.6, 6.8, 15.0, and 25.4 m, respectively, were used. The longest column gives the greatest rate of peak capacity production over the entire boiling point range. However, the total analysis time is more than twice that obtained with the shortest column. Thus, shorter columns are attractive for relatively simple mixtures where high peak capacity may be less important. For relatively simple mixtures spanning a more restricted boiling point range, for example C8C12, the speed advantage of the shorter columns is even more apparent. This is due to the fact that, for a more restricted boiling point range, the holdup time is a larger fraction of the total analysis time. Figure 7 summarizes total peak capacity (C8-C19) values (part a) and the corresponding separation times (part b) for the various column length and carrier gas velocity combinations using a nominal programming rate of 50 °C/min. Plots A-D are for column lengths of 3.6, 6.8, 15.0, and 25.4 m, respectively. The advantage of using relatively long columns at a gas velocity of about 100 cm/s is clear. For the shorter columns, the decrease in separation time with increasing carrier gas velocity is remarkably small, and the use of higher velocities is not justified for temperature-programmed HSGC using the parameter ranges investigated in this study.
CONCLUSIONS Many analysis procedures using temperature-programmed GC could be made significantly faster with minimal loss in resolution by the use of higher programming rates. Most laboratory GC instruments can be programmed at rates of 50 °C/min or more, but significant departures from program linearity may occur. Much higher programming rates with good program linearity can be obtained with instruments using unconventional heating technologies.13,15 Fast temperature programming is needed for HSGC separations of wide boiling point range mixtures. Fast temperature programming for separations in the 1-5 min time frame can be achieved with conventional laboratory GC instruments. The use of still higher programming rates is attractive for obtaining further reductions in separation times. Temperature programming rates at least as high as 50 °C/min can be used with 0.25-mm-i.d. columns as long as 25 m with acceptable losses in boiling point resolution and peak-capacity-generation rate. When longer columns are used with high programming rates, higher than optimal carrier gas velocities result in improved column performance as well as somewhat shorter separation times. Using a 25-m column length, a total peak capacity of nearly 300 peaks (C6-C19) can be obtained for a 5-min analysis time. Shorter columns and higher average carrier gas velocities result in only modest reductions in separation times but result in more significant losses in boiling point resolution and cumulative peak capacity. However, these conditions are attractive for relatively simple mixtures, which do not require a large peak capacity for a complete separation. With the availability of commercial instrumentation, more applications are appearing, and it is likely that high-temperature programming rates (1-10 °C/s) will find increasing utilization. Many existing procedures using conventional temperature programming are based on adequate separation of a critical component pair (most difficult pair to separate). In these cases, even modest reductions in peak capacity associated with increases in programming rate may result in unacceptable critical-pair overlap. To utilize more effectively fast temperature programming, greater attention to column selectivity will be required. ACKNOWLEDGMENT The authors gratefully acknowledge the Varian Instrument Co. for the gift of the gas chromatograph and the Centers for Disease Control and Prevention, National Environmental Laboratory, for financial support. Received for review November 20, 1998. Accepted March 14, 1999. AC9812936
Analytical Chemistry, Vol. 71, No. 11, June 1, 1999
2129
