Anal. Chem. 2008, 80, 760-768
Dual-Injection System with Multiple Injections for Determining Bidimensional Retention Indexes in Comprehensive Two-Dimensional Gas Chromatography Stefan Bieri† and Philip J. Marriott*
School of Applied Sciences, RMIT University, G.P.O. Box 2476V, Melbourne, Victoria, 3001, Australia
A new instrumental approach for collection of retention index data in the first (1D) and second (2D) dimensions of a comprehensive two-dimensional (2D) gas chromatography (GC×GC) experiment has been developed. Firstdimension indexes were determined under conventional linear programmed temperature conditions (Van den Dool indexes). To remove the effect that the short secondary column imposes on derived 1D indexes, as well as to avoid handling of pulsed GC×GC peaks, the proposed approach uses a flow splitter to divert part of the primary column flow to a supplementary detector to simultaneously generate a conventional 1D chromatogram, along with the GC×GC chromatogram. The critical 2D indexes (Kova´ ts indexes) are based upon isovolatility curves of normal alkanes in 2D space, providing a reference scale against which to correlate each individual target peak throughout the entire GC×GC run. This requires the alkanes to bracket the analytes in order to allow retention interpolation. Exponential curves produced in the 2D separation space require a novel approach for delivery of alkane standards into the 2D column by using careful solventfree solid-phase microextraction (SPME) sampling. Sequential introduction of alkane mixtures during GC×GC runs was performed by thermal desorption in a second injector which was directly coupled through a short transfer line to the entrance of the secondary column, just prior to the modulator so that they do not have to travel through the 1D column. Thus, each alkane mixture injection was quantitatively focused by the cryogenic trap, then launched at predetermined times onto the 2D column. The system permitted construction of an alkane retention map upon which bidimensional indexes of a 25perfume ingredient mixture could be derived. Comparison of results with indexes determined in temperature-variable one-dimensional (1D) GC showed good correlation. Plotting of the separation power in the second dimension was possible by mapping Trennzahl values throughout the 2D space. The methodology was applied to the separation * Corresponding author. Phone: +61 3 99252632. Fax: +61 3 99253747. E-mail:
[email protected]. † Present address: Official Food Control Authority of Geneva, 20 Quai Ernest Ansermet, 1211 Geneva, Switzerland.
760 Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
of a standard mixture composed of 25 analytes (very diverse in polarity and structure) suspected to be allergens in perfume samples. The method will allow straightforward determination of temperature-variable retention indexes of target analytes. Although uptake of comprehensive two-dimensional (2D) gas chromatography (GC×GC) was originally slow, improvements in both cryogenic modulation technology and data-processing methods has led to a greater acceptance and a better reliability of GC×GC. Thus, this emerging technique and its corresponding knowledge base have expanded with great rapidity1-7 during the past few years. However, it is apparent that many established methods in one-dimensional (1D) GC have yet to be tested or applied to the GC×GC experiment, and so the full potential of GC×GC may not yet be entirely exploited or understood. The initial impetus for GC×GC, since the early work of Liu and Phillips8 in 1991, derived from the considerably enhanced separation power compared to conventional 1D approaches, the increased sensitivity through refocusing during the modulation process, and the generation of an often well-organized 2D pattern (i.e., group or class type separation). On the one hand, all these benefits are accessible without increasing analysis time, notwithstanding the needs for new data-processing techniques. On the other hand, the very fast 2D separation and 2D presentation format demands that several chromatographic and detection parameters must be revisited. The reproducibility of absolute retention time (tR) is an essential parameter to define the property of a solute which may be used for (tentative) identification purposes, and retention finds its greatest value when incorporated into relative retention calculations such as retention indexes (I), so it seems worthwhile to reconsider the concept of I values in the case of GC×GC. Indeed, as components are accurately located within the contour (1) Phillips, J. B.; Xu, J. J. Chromatogr., A 1995, 703, 327-334. (2) Phillips, J. B.; Beens, J. J. Chromatogr., A 1999, 856, 331-347. (3) Bertsch, W. J. High Resolut. Chromatogr. 2000, 23, 167-181. (4) Marriott, P.; Shellie, R. Trends Anal. Chem. 2002, 21, 573-583. (5) Marriott, P. J.; Morrison, P. D.; Shellie, R. A.; Dunn, M. S.; Sari, E.; Ryan, D. LC‚GC Eur. 2003, 16, 23-31. (6) Dalluge, J.; Beens, J.; Brinkman, U. A. T. J. Chromatogr., A 2003, 1000, 69-108. (7) Beens, J.; Brinkman, U. A. T. Anal. Bioanal. Chem. 2004, 378, 1939-1943. (8) Liu, Z.; Phillips, J. B. J. Chromatogr. Sci. 1991, 29, 227-231. 10.1021/ac071367q CCC: $40.75
© 2008 American Chemical Society Published on Web 01/05/2008
plot9 by two separate retention ordinates, the retention index system should provide complementary identification information in GC×GC. Only a few papers have dealt with the potential of bidimensional retention indexes, and thus, it is still an emerging topic of investigation in GC×GC.10-14 The retention index system as introduced by Kova´ts in 195815 was designed to be applicable to chromatograms generated under isothermal conditions. By choosing a series of homologues as the reference system which calibrates the time axis of the chromatogram in units of retention indexes, the position of a substance is expressed relative to two consecutive homologues that elute before and after it with (z) and (z + 1) carbon atoms, respectively. n-Alkanes are commonly applied as reference solutes, but other suites may be employed.16-18 Van den Dool and Kratz19 extended the I notion to temperature gradient conditions by demonstrating that during linear temperature-programmed operations the temperature coefficient may be neglected in most cases. Guiochon20 confirmed this finding, whereas Habgood and Harris21 argued that the temperature dependence on retention indexes must be taken into account. The reproducibility of linear retention indexes determination in routine analysis as showed by Bicchi et al.22 demonstrated that the slightly reduced precision compared to the isothermal mode is largely compensated by the advantage of temperature-programmed operations. In the particular case of GC×GC, the first temperatureprogrammed dimension can be characterized with the usual Van den Dool approach, whereas transposition of the isothermal Kova´ts index to the fast second dimension requires a new paradigm. Indeed, to linearly interpolate each analyte to the reference series in the first dimension a single reference series chromatogram is adequate to calibrate the entire first dimension. The discrete nature of the second-dimension separation space, which essentially arises because each modulation event is a unique chromatogram at its own pseudoisothermal temperature, means theoretically the reference data set must be applied to each individual 2D chromatogram. This inter alia requires each reference series compound to have its retention time defined for each 2D chromatogram. Experimentally, this can only be achieved if the reference compounds are introduced into the 2D column over a range of temperatures that allows the relationship between temperature and retention to be established. In turn, this requires multiple injections of the reference mixture into the GC×GC system, to construct a second-dimension retention map that will allow a logarithm interpolation of any target solute between two (9) Shellie, R.; Marriott, P.; Leus, M.; Dufour, J.-P.; Mondello, L.; Dugo, G.; Sun, K.; Winniford, B.; Griffith, J.; Luong, J. J. Chromatogr., A 2003, 1019, 273-278. (10) Western, R. J.; Marriott, P. J. J. Chromatogr. Sci. 2002, 25, 832-838. (11) Western, R. J.; Marriott, P. J. J. Chromatogr., A 2003, 1019, 3-14. (12) Arey, J. S.; Nelson, R. K.; Xu, L.; Reddy, C. M. Anal. Chem. 2005, 77, 71727182. (13) Bieri, S.; Marriott, P. J. Anal. Chem. 2006, 78, 8089-8097. (14) Zhu, S.; Lu, X.; Qiu, Y.; Pang, T.; Kong, H.; Wu, C.; Xu, G. J. Chromatogr., A 2007, 1150, 28-36. (15) Kova´ts, E. Helv. Chim. Acta 1958, 41, 1915-1932. (16) Castello, G. J. Chromatogr., A 1999, 842, 51-64. (17) Peng, C. T. J. Chromatogr., A 2000, 903, 117-143. (18) Mjøs, S. A.; Meier, S.; Boitsov, S. J. Chromatogr., A 2006, 1123, 98-105. (19) Van den Dool, H.; Kratz, P. D. J. Chromatogr. 1963, 11, 463-471. (20) Guiochon, G. Anal. Chem. 1964, 36, 661-663. (21) Habgood, H. W.; Harris, W. E. Anal. Chem. 1964, 36, 663-665. (22) Bicchi, C.; Binello, A.; D’Amato, A.; Rubiolo, P. J. Chromatogr. Sci. 1999, 37, 288-294.
consecutive reference solutes. The procedure generates so-called isovolatile reference curves, which trace the continuously decreasing second-dimension retention time with increasing oven temperature of a (constantly) introduced solute onto the 2D column during a complete GC×GC run. Such a decay curve was observed by Beens et al.23 with an injector that continuously introduced its bleed into the serially coupled columns of a GC×GC system. Western and Marriott10,11 proposed timed sequential split solution injections of members of a series of reference homologues to build isovolatile curves. Recently Bieri and Marriott13 demonstrated that introduction of preselected reference mixtures by solid-phase microextraction (SPME) enabled production of an entire retention correlation map. Furthermore, they developed a method capable of generating three independent indexes by using a dual secondary column arrangement (GC×2GC). These authors pointed out, however, that the resulting precision of retention indexes in GC×GC is not as good as that obtained in conventional singlecolumn analysis, primarily because poorly retained compounds on the 2D column provide little retention space between consecutive alkanes in which to interpolate the retention of the compound of interest. Other investigators24,25 exploited retention index data obtained from conventional GC operating conditions to predict and/or optimize GC×GC separations. The present investigation proposes a novel approach which achieves higher 2D separation numbers and, thus, a much more acceptable Kova´ts retention index basis for the critical second dimension. For this, an innovative way for direct delivery of reference alkane standards into the 2D column inlet by using a dual-injection system was developed. This may also allow improved access to alkane-based indexes for polar 2D columns. An effluent split arrangement allows first-dimension retention indexes to be accurately and easily estimated. The introduction procedure can likewise be employed to generate isovolatility curves for target solutes, as opposed to reference compounds, to permit temperature-variable indexes for each analyte of interest. This suggests that regeneration of databases of retention indexes for new columns can be now undertaken relatively readily. EXPERIMENTAL METHODS Instrumentation. One-dimensional GC and GC×GC analyses were conducted on an Agilent 6890 GC (Agilent Technologies, Burwood, Australia) equipped with two conventional split/splitless injectors, both with electronic pressure control, and two highspeed flame ionization detectors (FID). To perform the GC×GC analysis, the GC was fitted with a longitudinally modulated cryogenic system (LMCS) from Chromatography Concepts (Doncaster, Australia). Both detectors were operated at 250 °C, and data acquisition rates were fixed at 100 and 20 Hz for FID 1 and 2, respectively. Injector 1 (INJ 1) was fitted with a conventional GC column, whereas INJ 2 was connected by means of a short transfer capillary (deactivated fused-silica 0.5 m × 0.1 mm i.d.) and a three-way splitter to the junction between the 1D and 2D columns (Figure 1). Hydrogen carrier gas was used at a constant pressure of 25 psi for INJ 1 set at 250 °C and 20 psi for INJ 2 set (23) Beens, J.; Tijssen, R.; Blomberg, J. J. High Resolut. Chromatogr. 1998, 21, 63-64. (24) Beens, J.; Tijssen, R.; Blomberg, J. J. Chromatogr., A 1998, 822, 233-251. (25) Vendeuvre, C.; Bertoncini, F.; Thie´baut, D.; Martin, M.; Hennion, M. C. J. Sep. Sci. 2005, 28, 1128-1136.
Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
761
Table 1. Linear Retention Indexes of the Target Components and Confidence Intervals for Three Consecutive Replicate Injectionsa 1D
column phase column phase initial oven temp temp program; °C/min 2D
Figure 1. Schematic of the GC×GC system incorporating a primary column flow splitter (A) and an additional injection port (INJ 2) connected to the midpoint between the two dimensions (B). The LMCS device is shown in both its upper trapping position and its lower release or reinjection position.
at 280 °C. Split ratios were 10:1 and 60:1 for INJ 1 and 2, respectively. All liquid samples were injected with a fast autosampler (Agilent model 7683). The 1D column was a low-polarity 5% phenyl-polysilphenylene-siloxane phase (BPX5, 30 m × 0.25 mm i.d., 0.25 µm film thickness (df), SGE International, Ringwood, Australia). The 2D column was either the recently introduced midpolar biphenyl-polysiloxane column (Optima δ-6, 2 m × 0.1 mm i.d., 0.1 µm df, Macherey-Nagel, Oensingen, Switzerland) or a polar poly(ethylene glycol)-coated column (BP20, 2 m × 0.1 mm i.d., 0.1 µm df, SGE). By means of a three-way splitter, the primary column flow was equally split to the 2D column that was connected to FID 1 and to an empty deactivated capillary (2 m × 0.1 mm i.d.) attached to FID 2. Standard and Sample Preparation. A standard solution made up of 25 compounds (see Table 1) was prepared daily in hexane at a concentration of 100 mg/L each (from individual stock solutions of 2500 mg/L). All compounds were purchased from Sigma-Aldrich in purity exceeding 97%, where available. Firstdimension linear temperature-programmed retention indexes were determined by injection of the standard solution augmented with a mixture of normal alkanes ranging from C10 to C23 at 50 mg/L each, in dichloromethane. Delivery of n-alkane reference mixtures directly to the 2D column was achieved by successively sampling the headspace (HS) of selected alkanes with a 30 µm film thickness (100% poly(dimethylsiloxane)) SPME fiber (Supelco, Sigma Aldrich Pty Ltd, Bellefonte, PA). Alkanes were sampled individually. To adapt sampling conditions for the wide volatility range of these alkanes, the SPME fiber was exposed at various temperatures as follows: at room temperature in 20 mL vials in the case of C9-C12, at 40 °C in 5 mL vials for C13 and C14, at 80 °C in 5 mL vials for C15-C17, and at 140 °C in 5 mL vials for C18-C21. Exposure times varied from 1 to 20 s, and thus, any required composite mixture of alkanes could be sampled within less than 2 min. Sampling order started from the less volatile alkane to the most volatile ones, so as to avoid partial desorption of low boiling point hydrocarbons when sampling heavier alkanes at higher sampling temperatures. Selected alkanes series were then thermally desorbed at predetermined times by inserting the fiber into INJ 2. A split ratio of 60:1 and a straight glass liner with 0.75 mm 762 Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
BPX5 BPX5 BPX5 BPX5 BPX5 BPX5 FS δ-6 FS BP20 FS FS 60 °C 60 °C 60 °C 60 °C 60 °C 100 °C 3 3 3 3 5 3
no.
compd
1 2 3 4 5 6 7a 8 7b 9 10 11 12 13 14 15 16 17 18 19a 20a 20b 19b 21a 21b 22 23 24 25
limonene benzyl alcohol phenylacetaldehydeb linalool methyl 2-octynoate citronellol citral geraniol citral isomer cinnamaldehyde hydroxycitronellal anisyl alcohol methyl 2-nonynoate cinnamic alcohol eugenol coumarin isoeugenol R-isomethylionone amyl cinnamaldehyde amylcinnamic alcohol impurity of 20b lyral impurity of 19a farnesol farnesol isomer hexylcinnamaldehyde benzyl benzoate benzyl salicylate benzyl cinnamate
lowest confidence intervalc highest confidence interval
I1D-1 I1D-2 I1D-3 I1D-4 I1D-5 I1D-6 1035 1049 1060 1107 1212 1235 1250 1259 1280 1296 1300 1304 1311 1327 1367 1466 1466 1481 1663 1674 1678 1686 1700 1701 1725 1764 1794 1898 2135
1035 1050 1061 1107 1213 1235 1251 1260 1281 1297 1302 1305 1312 1329 1368 1467 1468 1482 1665 1675 1680 1688 1702 1704 1726 1765 1796 1900 2138
1034 1048 1059 1106 1211 1234 1250 1259 1280 1295 1300 1303 1310 1327 1367 1465 1465 1481 1663 1674 1678 1686 1701 1702 1725 1763 1794 1898 2134
1035 1065 1064 1108 1213 1237 1251 1262 1281 1301 1305 1312 1321 1342 1374 1465 1474 1476 1665 1677 1683 1690 1706 1704 1728 1766 1799 1903 2141
1037 1051 1063 1107 1212 1235 1252 1260 1282 1300 1302 1307 1311 1331 1371 1475 1471 1486 1669 1679 1684 1692 1706 1706 1728 1770 1803 1909 2146
1041 1053 1067 1108 1210 1233 1251 1258 1281 1303 1302 1308 1310 1330 1370 1471 1468 1484 1664 1675 1679 1687 1702 1702 1726 1764 1795 1899 2135
0.1 0.6
0.1 0.5
0.1 0.5
0.1 0.7
0.1 0.5
0.1 0.4
a Both citral and farnesol are purchased as mixtures of Z and E isomers. FS ) fused-silica (deactivated). b Not in the list of the SCCNDFP of 24 allergens (ref 32). c n ) 3.
i.d. (6.5 mm o.d., 78.5 mm long) was employed to ensure a fast transfer time of the reference mixtures. GC×GC Experiments. The thermal modulator was operated with a trap temperature of -10 °C using expanding liquid CO2 as coolant and a modulation period of either 8 or 12 s. The moving trap equipped with a fast actuating arm and external clock controller was set to spend 4 s (or 6 s), both in the top and the down position (accounting for the time to move from the top to the down position or vice versa). Method 1 was used to build the retention map which included Trennzahl (TZ) values in the second-dimension separation space (presented in Figure 2) and for determining the isovolatility curves of geraniol (8) and cinnamaldehyde (9) with their corresponding Kova´ts retention index as a function of temperature (presented in Figure 7). A modulation period of 12 s that was initiated at 4.95 min (this accounts for the first 6 s, i.e., 0.05 min, during which the cryotrap stays in the top position) and stopped at 59.5 min was used. The oven was initially set at 60 °C and then linearly increased to 240 °C at 3 °C/min. Method 2 was applied to determine 1D and 2D retention indexes with a modulation period of 8 s initiated at 5 min and
Figure 2. Second-dimension retention and Trennzahl (TZ) map obtained with multiple SPME introductions of selected n-alkane series through INJ 2. (A) Exponential isovolatility curves of alkanes C10-C21 (dotted lines) and areas of iso-TZ values in the second dimension (shaded zones where TZ ) 10, 8, 6, or 4). The solid line links the alkane series (C10-C21) injected at t ) 0 min through INJ 1. (B) Second-dimension chromatogram after injection of C12-C18 into INJ 2 and released by the LMCS into the 2D column at 36.2 min.
stopped at 59.5 min. The oven was initially set either at 60 or 100 °C and linearly increased to 240 °C at 3 or 5 °C/min. The 25 standards mixture with added alkane series was injected at time t ) 0 (INJ 1), and multiple alkane injections (INJ 2) were introduced during the same GC×GC experiment. The selected alkane series were introduced by SPME through INJ 2 at 7, 9, 11, 14, 16, 19, 21, 24, 28, 32, 36, 40, 44, 48, 51, and 54 min. The holdup time in the second-dimension separation column was determined by using second-dimension retention times of three consecutive alkanes (tR1, tR2, and tR3) according to the equation introduced by Ettre and Hinshaw26 (e.g., in Figure 3 at 32.2 min (169 °C) the retention times of C15 (tR1), C16 (tR2), and C17 (tR3) were 5.37, 7.77, and 11.55 s, respectively. Thus tM ) (tR1tR2 - tR22)/ (tR3 - tR2) - (tR2 - tR1) ) 1.20 s. This is in accordance with the retention time of the column bleed (1.28 s) measured at the same elution temperature and which can be used as a good approximation for tM). Method 3 was employed as a fast alternative to generate alkane isovolatility curves. For this, a modulation period of 12 s initiated at t ) 0 min and stopped at 29 min was used. The oven was initially set at 81 °C and then ballistically heated to the next isothermal plateau and stabilized for about 30 s. In total 16 steps, at 81, 89, 93, 105, 108, 119, 123, 132, 146, 159, 170, 180, 192, 204, 213, and 222 °C were utilized to build the 2D alkane map. Selected alkane mixtures (presented in Figure 6) were directly introduced onto the 2D column by SPME when the trap was resting in the upper position. The oven was then quickly heated to the next step 12 s after elution of the last alkane. In all cases, GC×GC chromatogram files were exported from the ChemStation in ASCII format for 2D conversion using 2DGC Converter (Chromatography Concepts). Origin V7.5 plotting (26) Ettre, L. S.; Hinshaw, J. V. Basic Relationships of Gas Chromatography, 2nd ed.; Advanstar Communication: Cleveland, OH, 1993; pp 142-143.
Figure 3. Contour plot of target compounds (1-25) coinjected (at t ) 0) with a mixture of n-alkanes (C10-C23, which follows the dotted line forming the lower retention boundary). This line delimits the grayed zone enclosing wrapped around analytes (15 and 25) or weakly retained alkane of series injected through INJ 2. The upper grayed zone is the extended region including eight additional retention estimates (circles) enabling extrapolation of the isovolatility curves.
program was used for presentation of 2D separation contour plots and for all curve fitting calculations. One-Dimensional Analysis. Single-column isothermal indexes were determined at different temperatures between 40 and 220 °C at 10 °C intervals (method 4) using the GC×GC setup depicted in Figure 1, however, without initiating the LMCS Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
763
modulator. The 0.1 µL samples of the individual target substances together with their bracketing alkane series were injected directly onto the 2D column (Optima δ-6, 2 m × 0.1 mm i.d., 0.1 µm df) with the fast autosampler through INJ 2. The injector was set at 250 °C, operated at a split ratio of 250:1, and used a 1 mm i.d. liner with a plug of deactivated glass wool. Alternatively, these Kova´ts retention indexes were confirmed by conventional 1D GC experiments on a 5 m × 0.1 mm i.d., 0.1 µm film df of identical stationary phase (method 5). Hydrogen carrier gas was used at a constant flow of 0.5 mL/min. Injection conditions were the same as those for method 4. For each analyte the retention index was measured at three different temperatures. RESULTS AND DISCUSSION It has been shown that the 2D column of a serially coupled column set, even if very short, may significantly affect estimation of first-dimension linear retention indexes.13,27 To circumvent this shortcoming, the GC×GC setup proposed here equally splits the primary column flow to two capillaries of identical geometry (Figure 1). One column is the second-dimension column (2 m × 0.1 mm i.d. × 0.1 µm df) mounted through the longitudinally modulating cryogenic trap and connected to FID 1, which monitors the signal of the modulated chromatographic peaks used to generate the GC×GC contour plot. The second column was a deactivated transfer line (2 m × 0.1 mm i.d.) linking the firstdimension column directly to FID 2 and presents the 1D separation independently of the second column stationary phase. The flow splitting reduces the sample signal on both detectors and the flow volume through each channel compared with a conventional nonsplitting arrangement. The latter reduction in column flow is counterbalanced (and can be compensated and adjusted) by the secondary pneumatic flow control of the second injector. Table 1 gives the calculated values of 1D retention indexes for 25 target compounds according to the method developed by Van den Dool and Kratz19 for linear temperature program conditions. Two column sets were evaluated, namely, the combination of an apolar BPX5 primary column with either a midpolar Optima δ-6 or a polar BP20 second dimension column. An uncoated transfer line was coupled in parallel to the 2D separation column in both cases (Figure 1). First-dimension indexes were also measured with a slightly faster oven ramp (i.e., 5 instead of 3 °C/min) and with a higher initial oven temperature (i.e., 100 instead of 60 °C). In each case, three replicate injections were carried out. For the temperature range and program conditions under consideration overall confidence intervals of the calculated retention indexes varied between 0.1 and 0.7 retention units. Results show that 1D retention indexes did not vary by more than 1 I unit if they were measured at the outlet of the transfer line (i.e., Table 1: IT1D-1 and IT1D-3). Conversely, when data were calculated across the total of a dual coupled column arrangement the combination of BPX5-δ-6 (IT1D-2) led occasionally to retention shifts of 3 units, whereas a shift of up to 17 units was measured for the combination BPX5BP20 (IT1D-4) compared to a pure BPX5 phase. The influence of the short secondary column stationary phase on retention indexes is more pronounced for large polarity difference between the columns. The initial oven temperature and the program rate (27) Mondello, L.; Casilli, A.; Tranchida, P. Q.; Dugo, G.; Dugo, P. J. Chromatogr., A 2005, 1067, 235-243.
764
Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
during GC runs must be kept the same for retention indexes to be comparable. This is outlined in Table 1 where data show that shifts of up to 11 units were observed when the oven rate (IT1D-5) or the initial oven temperature (IT1D-6) were increased. The developed GC×GC setup allowed direct introduction of selected reference standards to the inlet of the 2D column and construction of an alkane-based retention map in the 2D separation space of conventional GC×GC experiments. From earlier work, it became apparent that reference substances injected during the chromatographic analysis do not have to elute through the 1D column, and this led to the modified system depicted in Figure 1. Thus, the compounds can be almost instantaneously delivered into the 2D column with the dual-injection system; INJ 2 was connected to the midpoint (i.e., point where the 1D and 2D columns are interfaced) by means of a short deactivated narrow bore capillary (0.75 m × 0.1 mm i.d.). Selected alkane series can now be introduced to the fast 2D column at precise moments during the GC×GC run, which then formed a reference scale upon which second-dimension pseudoisothermal retention indexes (I) can be derived. The indexes are termed pseudoisothermal because despite a relatively long modulation period of 8 or 12 s and a slow temperature rate of 3 °C/min, the increase in temperature for each second-dimension analysis was only 0.4 or 0.6 °C, respectively. Thus, for the short duration of each individual second-dimension chromatogram there was only a slight increase in oven temperature, and separations can be considered as isothermal. Therefore, 2D I values were calculated according to Kova ´ ts’ relationship.15 In order to avoid repeated introduction of interfering solvent injection bands, alkanes were individually and sequentially sampled by SPME in the HS and desorbed in INJ 2. Figure 2 depicts the contour plot of a multiple injection run which permitted construction of a retention map based on isovolatile curves of alkanes. To generate these curves an initial liquid injection (via INJ 1) of an alkane mixture ranging from C10 to C23, together with 16 consecutive SPME desorptions (via INJ 2) from alkane series C10-C11 (first injection) to C15-C21 (last injection) were performed throughout the chromatographic run. Times of injection were chosen so as to position the less volatile alkane of the series close to the upper time limit of the modulation period (12 s). As a result, a suite of consecutive alkane spots (C12-C21) eluted in the very upper part of the contour plot. By linking the peak apexes of each particular alkane C10-C21, it was possible to plot their corresponding exponential functions (generic equation: y ) A1 exp(-x/t1) + A2 exp(-x/t2) + A3 exp(-x/t3) + y0) calculated using the plotting program Origin V7.5. The correlation coefficient (R2) varied between 0.9995 and 0.9999, except for C21 (four retention data points) which gave an R2 of 0.996. Figure 2B illustrates a discrete 2D chromatogram obtained by delivering C12-C18 into the 2D column at 36.2 min (at point marked B, Figure 2A). Multiple alkane introductions allows some new possibilities for quantifying the separation power of the 2D space by determining TZ28 values between two consecutive alkane homologues eluting from the 2D column (i.e., values which determine the number of chromatographic peaks that can be placed between two selected homologues and separated from each other with a resolution of 1.117). Zones of identical separation values were determined by measuring the decrease of TZ between two consecutive alkanes (28) Kaiser, R. Z. Anal. Chem. 1962, 189, 1-14.
as a function of higher elution temperature on the 2D column. Thus, Figure 2A shows that the lines for C12/C13 converge toward higher elution temperature, and so TZ decreases from 10, through 8, to 6, then 4. Figure 2 shows that, as expected, the best TZ values (TZ ) 10) are obtained at lower elution temperature and higher 2D retention factor values. TZ not only gives a good idea of how many component peaks can be resolved in a given segment of the chromatogram, but it can also be directly linked to the retention index notion. As demonstrated by Ettre,29 according to the relationship ∆I ) 100/(TZ + 1), if TZ ) 10 then two adjacent components can be separated with a resolution of 1.177 if their retention index values differ by 9.1 index units. However, if TZ values are only 4 units, as in the lower part of the contour plot in Figure 2, then only compounds with retention index difference higher than 20 may be separated. As pointed out previously,13 a consequence of this is that accuracy of 2D I values will decrease with reduced TZ values. Thus, for greatest precision analytes should be distributed throughout zones of highest possible separation number. Therefore, the present approach demonstrates preferred results with 2 m long 2D column of intermediate polarity. With conventional apolar-polar column combinations, the lower 2D retention limit is defined by the nearly horizontal “line” formed by the retention times of an alkane suite injected at t ) 0 (Figure 2A). Such column combinations ensure that alkanes are comparatively well retained on the 1D column and thus, elute at higher temperature, so are the less retained components on the polar 2D column (i.e., they elute closely to the 2D column holdup time, 2t ). Therefore, all other components (more polar) are dispersed M over the retention plane above the limiting alkane line. Alternatively, when the 2D column polarity approaches that of the 1D, alkanes start to be distributed along a diagonal. In such a case, column orthogonality is lost.30,31 In the present case, the intermediate polarity of the 2D column in conjunction with the constant pressure mode (decreasing linear velocity with increasing temperature) led to a high percent of inaccessible separation space (2D space located under the alkane line injected at t ) 0, solid line in Figure 2A). Thus, except for solutes that undergo wraparound, no peak elutes within the first 4 s of the 2D time for any modulation event (which corresponds to the separation region of low TZ values). The consequence of this is that it permits the time window of the second dimension to be arbitrarily shifted down or corrected by 4 s (i.e., to commence the 2D separation time at 4 s) and to subtract these 4 s from the overall modulation period. Thus, a time scale of 8 s was finally chosen and the early part of 2D zone utilized for locating solutes that potentially wrap around. Figure 3 depicts the 2D separation of the test mixture together with the initially injected alkane mixture. In addition, 16 successive reduced alkane series, composed mostly of three alkanes, were delivered onto the 2D column by SPME via INJ 2, to allow calculation of the isovolatility curves. Because some analytes were not located between two bracketing alkane curves within the effective separation space (4-12 s), the curves were extrapolated up to 2tR ) 21 s in order to include an upper reference (29) Ettre, L. S. Chromatographia 1975, 8, 291-299. (30) Venkatramani, C. J.; Xu, J.; Phillips, J. B. Anal. Chem. 1996, 68, 14861492. (31) Ryan, D.; Morrison, P.; Marriott, P. J. Chromatogr., A 2005, 1071, 47-53. (32) Shellie, R.; Marriott, P.; Chaintreau, A. Flavour Fragrance J. 2004, 19, 9198.
Figure 4. Linear plots of the logarithm of 2tRz′ vs z, injected under pseudoisothermal conditions (same injection time on the seconddimension column). A represents the extrapolated point which permits calculation of the retention time of C14 (2tR14′) at the same launch temperature measured for alkanes introduced at t ) 14.8 min. Points B-H are also extrapolated data points.
retention value for each alkane series. Because the reference series eluted under pseudoisothermal conditions, the theoretical retention estimate was calculated according to the log 2tRz′ relationship of a homologous series versus the number of carbon atoms (z) of their members. The results of these calculations are illustrated in Figure 4 where linear plots of log 2tRz′ versus z were determined for each alkane series (based upon three retention data, R2 fluctuated between 0.9997 and 0.9999). For instance, at 14.8 min, C11, C12, and C13 were introduced onto the 2D column which allowed prediction of the retention time of C14, which in turn was then used for determination of the isovolatile curve of tetradecane. So, all target compounds could now be encompassed between two isovolatility curves for which two bracketing alkane retention time were estimated. Consequently, by use of the exponential function formulas, the retention time of two bracketing alkanes was accessible at any temperature, for all analytes except 25 (refer to Figure 3). 2D retention indexes (2I δ6-1) were then compared with those obtained in 1D GC runs. Isothermal I values (Ical-1) were determined at various temperatures at 10 °C intervals on the 2 m long second-dimension column (Optima δ-6 phase) by injecting liquid samples of individual target compounds along with appropriate alkanes through INJ 2 of the GC×GC set up. Second, I data were collected (Ical-2) during conventional GC measurements on a 5 m column of the same stationary phase at three different temperatures. In both cases, I versus T plots were drawn for each analyte and gave to a good approximation linear curves. Typical plots are shown in Figure 5 for compounds 1-6, 8, 9 and confirmed that similar results were obtained with both data collection approaches. I comparisons were then made by assessing the elution temperature of the target compounds from the 1D column during GC×GC experiments and calculating their retention indexes at the corresponding temperature by means of the linear equations of I versus T. For instance, cinnamaldehyde (9), which eluted at a temperature of 129 °C in the GC×GC run and had a calculated 2D retention index of 1429 (method 2) or 1428 (method 3), had an index determined in 1D conditions of 1430 Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
765
Table 2. Second-Dimension Retention Indexes Obtained by Using Various Methodsa
Figure 5. Kova´ ts retention indexes (I) as a function of elution temperature Te at 10 °C intervals for target compounds 1-6, 8, and 9 determined by three methods: (i) using the 2D column in the arrangement shown in Figure 1 (Optima δ-6, 2 m × 100 µm × 0.1 µm) (solid lines), (ii) using a 5 m × 0.1 mm × 0.1 µm column operated in conventional 1D GC mode (dashed lines), and (iii) based upon data obtained in Figure 7 (two dotted lines marked with an asterisk). (Linear equations of I vs T can be obtained upon request.)
(Table 2 and Figure 5). Alternatively, 2D retention indexes (2Iδ6-2) were estimated by using an external isovolatility map generation approach (method 3). Instead of injecting the reference alkanes during the GC×GC run, isovolatile curves were calculated after data collection of second-dimension retention times of selected alkanes at various launch temperatures. The flow conditions were the same as those in GC×GC experiments. A GC run comprising 16 isothermal plateaus was conducted with total analysis time less than 30 min (Figure 6). At each temperature alkanes were introduced by SPME through INJ 2, trapped by the LMCS, and launched on the 2D column. The retention data obtained by this way also allowed calculation of the individual isovolatile curves. 2D I values (2I 2 δ6-2) were then calculated on the basis of tR values of three consecutive GC×GC separations (Table 2). Figure 7 points out another feature of this GC×GC arrangement, namely, that isothermal retention indexes of a compound can be determined at any temperature by sampling and injecting the compound of interest, either as a separate experiment or together with an adequate alkane series. This allows determination of retention indexes as a function of temperature of an analyte on the second-dimension column phase or location of the compound within the 2D separation space as a function of 1D elution temperature. The example of geraniol (8) and cinnamaldehyde (9) is given in Figure 7 where these compounds are systematically introduced by SPME during a GC×GC run. After an initial injection of an alkane series, they were thermally desorbed 16 and 12 times each, respectively, in the course of the GC×GC run. From these values, the linear plot I versus T was drawn (Figure 5, asterisked dotted lines for 8 and 9) and used to calculate the equivalent second dimension I (2Ical-3) at the elution temperature corresponding to the GC×GC run (Table 2). It can be seen that the five methods for calculating the retention indexes for 8 and 766 Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
2I -6 δ
compd no.
method 2
1 2 3 4 5 6 7a 7b 8 9 10 11 12 13 14 15 16 17 18 19a 19b 20a 20b 21a 21b 22 23 24
1072 ( 4 1135 ( 5 1153 ( 3 1149 ( 4 1289 ( 7 1284 ( 2 1334 ( 1 1366 ( 1 1318 ( 1 1429 ( 1 1390 ( 4 1423 ( 1 1389 ( 5 1442 ( 4 1464 ( 3 1636 ( 2 1577 ( 2 1555 ( 3 1778 ( 6 1770 ( 5 1806 ( 5 1804 ( 3 1811 ( 2 1770 ( 6 1796 ( 2 1878 ( 3 1940 ( 3 2060 ( 7
method 3
2I cal-1 2D config method 4
2I cal-2 1D config method 5
1070 ( 4 1133 ( 3 1153 ( 2 1147 ( 3 1287 ( 7 1283 ( 2 1333 ( 1 1365 ( 1 1317 ( 1 1428 ( 1 1389 ( 3 1422 ( 2 1387 ( 6 1441 ( 4 1463 ( 2 1638 ( 1 1576 ( 4 1552 ( 2 1775 ( 5 1767 ( 5 1804 ( 3 1802 ( 2 1810 ( 2 1768 ( 6 1793 ( 3 1879 ( 5 1944 ( 2 2057 ( 7
1073 1136 1156 1149 1287 1286 1334 1365 1318 1429 1389 1423 1388 1442 1464 1642 1578 1556 1778 1771 1809 1803 1813 1773 1797 1879 1945 2059
1074 1136 1157 1151 1288 1286 1334 1366 1319 1430 1389 1424 1389 1444 1465 1642 1579 1556 1778 1772 1810 1800 1810 1774 1798 1880 1947 2059
2I
δ-6
2I cal-3 method 1
1319 1429
a Confidence interval are calculated for three replicate injections of target compounds plus the alkanes series.
9 using either 1D or various GC×GC methods (2I values) were highly consistent across the methods. Since the 1D method is the classical way to determine I values, this confirms the method proposed here for collection of 2I data generates equivalent data to the conventional approach. It also confirms that the assumption of isothermal conditions for the 2D column during the GC×GC experiment is acceptable, within a reasonable uncertainty range of I values. Examination of Figure 7 shows the temperature dependency of the isovolatility lines for 8 and 9, compared to the reference alkane lines. This can be considered an indication of 2I variation with temperature. Ideally, the trends in data observed in Figure 7 should correlate with those in the classical temperature-variable method (Figure 5). Thus, from Figure 5 it can be seen that the I value for 9 has a rather strong dependence on temperatureshigh slopeswith an increase in I as T increases. Likewise 8 has a more gentle slope, so its I value has little temperature dependency. In Figure 7, the line for 9 starts just above the isovolatility line for C14 and trends much closer to the line for C15 as its elution temperature increases; this suggests its 2I value will be about 1410 at its lower elution temperature and increase to about 1470 at higher Te in agreement with Figure 5. Conversely, the line for 8 is located just above the isovolatility line for C13 over the total temperature range, so it increases slightly from about 1310 to about 1320 over this range. Note that this GC×GC experiment is conducted within 1 h and can be completed simultaneously for as many compounds as may be present in a chosen mixture (that give identifiable,
Figure 6. Contour plot of a multistep isothermal run (the oven track is represented by the dotted line and includes 16 brief isothermal plateaus, with temperatures indicated on the right-hand scale expressed in °C) and 16 separate alkane injections. At each plateau, preselected alkane series (1-16) were directly introduced by SPME through INJ 2. Injection 1 comprised C10 + C11; injection 8 comprised C11-C15. The 2D separation of the alkane suite C12-C18 (injection 11) is depicted as the vertical chromatogram at the launch time 17 min (i.e., the time that the LMCS unit was moved to its inject position).
Figure 7. Variation of 2D retention as a function of elution from 1D, using successive injections of 8, 9, and alkanes through INJ 2. (A) Variation of the second-dimension retention of compounds 8 and 9 (solid linessisovolatility curves) relative to alkanes (dashed linessisovolatility curves). An alkane mixture (C10-C23) is injected at t ) 0. (B) Discrete second-dimension chromatogram obtained after SPME introduction of 8, 9, C12C15 at 23.0 min which is close to the elution temperature of 9 in the GC×GC sample separation (i.e., if injected at time t ) 0).
resolvable peaks in the GC×GC experiment) that can be sampled by using SPME. It also covers the full range of elution temperatures of all the compounds over the temperature-programmed analysis with the single temperature-programmed condition. The alternativesclassicalsmethod requires individual analysis at each temperature, with alkanes and analyte (mixtures) chosen to suit the different isothermal conditions. If one assumes a mixture of, for example, 20 compounds for one GC×GC experiment (1 h), then in a single day (10 experiments), it should be possible to generate original temperature-variable retention index values on the 2D column for about 200 compounds. This has significant implications for re-evaluation of retention indexes for researchers who require I value databases to support identification. Often databases are developed with specific columns (and usually were done so some years ago), and so these columns must always be used, at the expense of choosing newer phases that might offer other advantages to the analyst. Here, a simple opportunity arises
that allows database generation and renewal, and new columns to be incorporated into the database, with less tedious experimentation. CONCLUSION The opportunities that arise from the described system are manyfold. A completely new implementation of delivery of reference compounds to the 2D column of a GC×GC system with dual injection has been developed. It has been demonstrated in the context of experimental determination of retention indexes during GC×GC separation, allowing facile generation of I values in both dimensions of the GC×GC setup. This approach allowed direct sample introduction of reference compounds to the seconddimension column, bypassing the primary column. This permitted isovolatility curves to be derived and acquisition of retention indexes of target analytes as a function of temperature. Databases of retention indexes should now be acquired much faster than Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
767
has been possible in the past. In the case of compounds that were not bracketed between two isovolatile reference curves an alternative approach to extending the retention correlation map through a stepwise isothermal temperature-programmed analysis proved to be successful. By utilizing comparative single-column reference data, the present work confirms that 2D indexes can be reliably estimated by using the GC×GC approach. If a suitable database can be developed, it is probable that an indication of the potential separation that might be obtained with a given column set can be attained, although the precision of retention indexes for early eluting compounds in the GC×GC experiment is not particularly good. Nevertheless, 2I values should assist the analyst in the identification of compounds, particularly when combined with 1I values and other spectral data. Consequently, development of a 1I/2I/mass spectrometry procedure will be extremely beneficial for unsupervised identification capabilities. With the proposed instrumental setup, reference series of homologues can be chosen independently of column polarity. That is, alkane-based retention indexes may also be determined on polar 2D columns, since selected alkanes series can readily be distributed throughout the second-dimension separation space by the direct delivery of the reference compounds to the 2D column. This overcomes the need to introduce the compounds through the 1D column, as had been reported by this group previously. Alternatively, it is possible to inject suitably selected polar reference series to calibrate the 2D axis of a polar second-dimension column. We also suggest the use of TZ values based upon alkane series as a meaningful expression of separation efficiency in the 2D separation space. As an aside, the dual-injection system via the second supplemental gas supply offers the potential to modify the seconddimension gas flow, which may assist in retention optimization during GC×GC analysis.
768
Analytical Chemistry, Vol. 80, No. 3, February 1, 2008
The first-dimension Van den Dool retention indexes were measured by diverting part of the primary column flow to a second FID, which generated a chromatogram of the first separation dimension. From a more theoretical point of view, bidimensional retention indexes convey valuable information about the physicochemical properties of a solute as the retention phenomena translates the complex interaction forces existing between the stationary phase and the solute (i.e., dispersion, dipole-dipole, dipole-induced dipole, electron donor-acceptor, and hydrogen-bonding forces). Thus, a general feature of multidimensional indexes is that it has the potential to gather information capable of identifying a solute with a high degree of discriminating value (retention data produced on different stationary phases). The unique nature of bidimensional retention index information offers the basis for novel retention-structure correlation and forms the framework for predicting peak position in the 2D plot. Or, by choosing adequate probe molecules, bidimensional indexes might be capable of revealing exquisite information about the column set or column orthogonality. ACKNOWLEDGMENT This work was supported by the Swiss National Science Foundation (SNF) under Grant PBGE2-108827. We thank Paul Morrison for his assistance in this project and SGE International for providing capillary columns.
Received for review June 27, 2007. Accepted September 26, 2007. AC071367Q
