Improved Performance and Maintenance in Gas Chromatography

May 4, 2007 - flushing or flow diversion is employed for this purpose. Both techniques necessitate the use of numerous com- ponents such as unions, te...
0 downloads 0 Views 279KB Size
Anal. Chem. 2007, 79, 4162-4168

Improved Performance and Maintenance in Gas Chromatography/Isotope Ratio Mass Spectrometry by Precolumn Solvent Removal Ulrich Flenker,*, † Moritz Hebestreit,† Thomas Piper,‡ Frank Hu 1 lsemann,‡ and Wilhelm Scha 1 nzer‡

Manfred Donike Institute and Institute of Biochemistry, German Sport University Cologne (DSHS), Carl-Diem-Weg 6, 50933 Cologne, Germany

The crucial step in current concepts to interface isotope ratio mass spectrometry (IRMS) to gas chromatography (GC) is efficient solvent removal. This is due to the essential postcolumn conversion of the analytes into simple gases, which is performed by either combustion or pyrolysis. The capacity of this step merely suffices to convert the analytes. Already small amounts of solvent present in the respective furnace can cause severe damage. In conventional GC/IRMS interfaces, the solvent is removed after passage of the GC column. Either backflushing or flow diversion is employed for this purpose. Both techniques necessitate the use of numerous components such as unions, tee pieces, valves, and capillary connections. Often this results in significant deterioration of the chromatographic resolution. In contrast, accurate GC/IRMS measurements require baseline separation of adjacent peaks. Moreover, maintenance of conventional interfaces may be tedious and time consuming, mostly because the numerous connections are prone to leakage. In order to avoid these drawbacks, we propose a concept to efficiently remove the solvent before passage of the GC column. It is based on the use of a cooled injection system operated in solvent vent mode, where the solvent elimination is supported by an auxiliary pump. Most unions and tee pieces thus can be removed. The chromatographic resolution is considerably enhanced. In particular, analysis of high-boiling and polar compounds can be improved. At the same time, the maintenance of the system is significantly facilitated. Under the chosen conditions, partial losses of low-boiling analytes during solvent elimination were not associated with significant isotope fractionation. The combination of gas chromatography and isotope ratio mass spectrometry (IRMS) has become a widely used analytical method in recent years. By integration of online combustion (gas chromatography/combustion/isotope ratio mass spectrometry, GC/ C/IRMS), several techniques have been developed that allow for compound-specific isotope analysis (CSIA) of complex matrixes. * To whom correspondence should be addressed. biochem.dshs-koeln.de. Phone: +49(0)22149825060. † Manfred Donike Institute. ‡ Institute of Biochemistry.

E-mail:

4162 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

uli@

In particular, the measurement of the stable isotope ratios of carbon (13C/12C) and nitrogen (15N/14N) is routinely performed by this analytical design. The possibility to perform CSIA by coupling gas chromatography to IRMS on principle has been known since 1976.1 This development represents a major breakthrough in the field of stable isotope analysis, especially regarding the measurement of the bioelements at natural isotopic abundances. Commercial instrumentation based on the design proposed by Brand has been available since 1988.2 This has facilitated enormous progress in scientific disciplines as diverse as geology,3 ecology,4 and consumer protection,5,6 to name but some. In doping control for instance, it provides the possibility to successfully discriminate physiological steroid production from the illicit administration of the corresponding synthetic hormone.7,8 A relatively recent development is the use of high temperatures of ∼1400 °C under reductive conditions (thermal conversion, GC/ TC/IRMS) to quantitatively convert the analytes into elemental hydrogen and carbon monoxide. This has facilitated CSIA of hydrogen9 (2H/1H) and oxygen10 (18O/16O). Unfortunately, these powerful and sophisticated techniques on principle lack specificity because accurate isotope measurements can only be performed on simple gases. Unlike organic mass spectrometry, IRMS does not yield any structural information. To the analyst not familiar with GC/IRMS, the outcome of such a measurement resembles a GC analysis with flame ionization (1) Sano, M.; Yotsui, Y.; Abe, H.; Sasaki, S. Biomed. Mass Spectrom. 1976, 1-3. (2) Brand, W. A. J. Mass Spectrom. 1996, 31, 225-235. (3) Sessions, A. L.; Burgoyne, T. W.; Schimmelmann, A.; Hayes, J. M. Org. Geochem. 1999, 30, 1193-1200. (4) McClelland, J. W.; Montoya, J. Ecology 2002, 83, 2173-2180. (5) Buisson, C.; Hebestreit, M.; Weigert, A. P.; Heinrich, K.; Fry, H.; Flenker, U.; Banneke, S.; Prevost, S.; Andre, F.; Schaenzer, W.; Houghton, E.; Le Bizec, B. J. Chromatogr., A 2005, 1093, 69-80. (6) Hebestreit, M.; Flenker, U.; Buisson, C.; Andre, F.; Le Bizec, B.; Fry, H.; Lang, M.; Weigert, A. P.; Heinrich, K.; Hird, S.; Scha¨nzer, W. J. Agric. Food Chem. 2006, 54, 2850-2858. (7) Becchi, M.; Aguilera, R.; Farizon, Y.; Flament, M.-M.; Casabianca, H.; James, P. Rapid Commun. Mass Spectrom. 1994, 8, 304-308. (8) Hebestreit, M.; Flenker, U.; Fussho ¨ller, G.; Geyer, H.; Gu ¨ ntner, U.; Mareck, U.; Piper, T.; Thevis, M.; Ayotte, C.; Scha¨nzer, W. Analyst 2006, 131, 10211026. (9) Hilkert, A. W.; Douthitt, C. B.; Schlu ¨ ter, H. J.; Brand, W. A. Rapid Commun. Mass Spectrom. 1999, 13, 1226-1230. (10) Hener, U.; Brand, W. A.; Hilkert, A.; Juchelka, D.; Mosandl, A.; Podebrad, F. Z. Lebensm. Unters. Forsch. A 1998, 206, 230-232. 10.1021/ac0621468 CCC: $37.00

© 2007 American Chemical Society Published on Web 05/04/2007

detection. Unless an additional specific detector is mounted, the only useful parameter for the identification of interesting compounds is the retention time. Therefore, extraordinary good GC resolution is required. But even more important, any overlap of adjacent peaks can induce dramatic errors when calculating the isotope ratios from the chromatogram. The bias introduced by nonresolved signals by far can exceed possible errors arising from the mere contribution of isotopically differing compounds.11,12 These circumstances necessitate the demand for true baseline separation of all relevant signals. This requirement, however, is quite difficult to be fulfilled. A typical hardware setup of a GC/ IRMS system involves numerous capillary connections and unions, which all represent small dead volumes. This often results in significant peak broadening and tailing. The popular Thermo Electron Combustion Interface III for instance possesses a backflush design and makes use of three tee pieces and six unions. Moreover, four of the screws involved reside in the GC cabinet where thermal stress makes these parts especially susceptible to leakage. Additionally, each connection represents a potential source of error, which makes maintenance and troubleshooting of a GC/ IRMS system awkward and time-consuming. It can take several hours and occasionally even several days until sufficient leak tightness is reached again following routinely required interventions. Change of the GC column, for instance, does require a lot more time as compared to conventional GC systems. The delicate design of GC/IRMS interfaces foremost results from the necessity to prevent solvents from entering the combustion or pyrolysis reactor. In GC/C/IRMS, mostly copper(II) oxide is used as the oxidizing agent. It is usually housed in a ceramic or quartz tube in the form of wires or pellets. Already small amounts of solvent present in such a furnace can completely reduce the copper oxide, which in turn will result in full degradation of this part of the system. This is due to the restricted oxygen pool, which is designed to suffice for the conversion of the analytes only. The situation is less problematic in GC/TC/ IRMS, where no reagents are required to convert the analytes and where the presence of some elemental carbon is even helpful to shift the equilibrium to the product side.10,13,14 However, a nearly complete solvent removal is necessary likewise. Otherwise carbon originating from pyrolysis of the solvent will plug up the reactor tube. Presumably, most of these problems arise because conventional GC/IRMS interfaces do not remove the solvent until it has passed the GC column. It therefore can be hypothesized that employment of a cooled injection system (CIS)soperated in solvent vent modeswould result in numerous advantages. In this operating mode, the solvent is vaporized in the injection port at relatively low temperatures while the analytes are trapped there. Simultaneously, most of the vapor is flushed through the GC’s split line. First, use of this procedure can significantly improve (11) Goodman, K. J.; Brenna, J. T. Anal. Chem. 1994, 66, 1294-1301. (12) Brenna, T. Effects of Chromatographic Overlap on Uncertainty. In Application of Gas Chromatography-Combustion-Isotope Ratio Mass Spectrometry to Doping Control; Bowers, L. D., Hildebrand, R. L., Symanski, E. J., Eds.; Colorado Springs, CO, 2003; 2nd Annual USADA Symposium on Anti-Doping Science, Los Angeles, CA, August 21-24, 2003. (13) Duff, R. E.; Bauer, S. H. J. Chem. Phys. 1962, 36, 1754-1767. (14) Koziet, J. J. Mass Spectrom. 1997, 32, 103-108.

chromatography because the solvent-induced peak broadening innate to split/splitless injectors can be avoided. Second, it can be assumed that careful selection of the solvent and appropriate adjustment of temperature, pressure, and carrier gas flow would produce sufficient solvent elimination to permit complete disassembly of the back-flush system. This should result in further improvement of chromatography and in much easier and much less time-consuming maintenance. EXPERIMENTAL SECTION Hardware. A CIS 4 cooled injection system (Gerstel, Mu¨lheim, Germany) was mounted into a GC 6890 (Agilent Technologies, Bo¨blingen, Germany). The Peltier option was used as cooling. The GC was coupled to a Delta plus XP gas isotope ratio mass spectrometer (Thermo Electron, Bremen, Germany) by a GC Combustion Interface III (Thermo Electron). The combustion interface, however, was widely modified: The back-flush system including all the tee unions was unmounted. Instead, a makeup gas adapter (TCEF1.5.5T, Valco Instruments, Houston, TX) was used to connect the GC column to the combustion furnace. At this position, a constant helium flush of ∼0.3 mL/min was added to the gas flow resulting from the GC. A tee piece was mounted outside the GC, which allowed addition of O2 to the makeup helium in order to enable regular recharge of the reactor. The reduction furnace was unmounted. The Nafion water trap was replaced by a Dewar vessel containing an acetone-dry ice mixture (∼ -80 °C). The open split connection to the mass spectrometer and the reference gas inlet were left in their original states. The GC column used for all experiments was a M&N Optimaδ3 (Macherey & Nagel, Du¨ren, Germany). The length was 30 m, the inner diameter was 0.25 mm, and the film thickness was 0.25 µm. It was placed immediately in front of the wires of the combustion tube in order to facilitate rapid oxidation of the analytes. By default the oxidation reactor consists of a ceramic tube, which is loosely inserted into a heater. The whole device is mounted vertically on the GC housing, and therefore, the tube sinks until its upper connection touches the heater. This effects that the lower part of the tube extends into the GC where temperatures are relatively cool in comparison to the 940 °C in the reaction zone. In addition, the lower connection represents a thermal capacity, independent from its design. In order to prevent the emergence of cold spots in the analytical gas flow and to promote rapid combustion of the analytes, the tube was elevated by ∼1.5 cm and fixed in this position. The makeup gas adapter thus had contact to the bottom of the combustion oven, assuring that the makeup gas adapter was kept at relatively high temperatures during all operation intervals. The CIS 4 was operated in solvent vent mode. Standard conditions resulted in a residual pressure of 0.08-0.15 bar in the injector. Consequently, a small gas flow into the GC column remained which carried small but significant amounts of solvent into the reactor. Therefore, an auxiliary membrane pump (N 86 KT.18, KNF, Freiburg, Germany) was connected to the split vent. The vacuum in the split line could be regulated via a controllable bypass. The absolute value of the vacuum could be read off from a mechanical vacuum meter. Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4163

Figure 1. Schematic drawing of the new GC/IRMS interface.

The pump was switched on via a relay during solvent vent. The relay was controlled by the time event list of vendor-provided software (ISODAT NT, version 2.0). A ball valve with a fixed opening pressure of 0.06 bar (series C, Swagelok, Solon, OH) was incorporated into the split line. Therefore, outside the operation intervals of the auxiliary pump, it was guaranteed that the split vent did not represent an appreciable resistivity. Figure 1 shows a schematic drawing of the new interface. An A200SE automatic liquid sampler (CTC Analytics, Zwingen, Switzerland) was employed to inject the samples. Operation Mode. Samples were injected at a speed of 1 µL/ s. The set point for the solvent vent gas flow was 100 mL/min, and the set point for the pressure was 0 bar. The pressure in the split line was adjusted to -0.3 bar during solvent vent. Solvent vent was performed at varying temperatures (see Experiments), and the corresponding interval lasted 90 s. The auxiliary pump was operated during this time period only. Following solvent vent, the split valve was closed and the temperature of the CIS was raised to 300 °C at a rate of 10 °C/s. This procedure effects evaporation of the analytes and subsequent recondensation on the GC column (“sample transfer”). The temperature of the CIS was maintained for 5 min. The temperature control was then switched off. The initial temperature of the GC likewise was set to the corresponding value of the CIS, but it was kept constant for a total period of 180 s. After solvent vent and sample transfer, a GC temperature program appropriate for the respective class of analytes was run. Figure 2 gives an example for the operational scheme of the system. The GC was operated in constant flow mode. A value of 1.2 mL was adjusted during all investigations. Experiments. The effectiveness of solvent elimination and the suitability of the device for analysis of high-boiling compounds was tested by injection of 1-7 µL of a standard mixture. The mixture contained five different steroids: 5R-androstane-3β-ol, etiocholanolone, 11-ketoetiocholanolone, 5β-pregnane-3R, 20R-diol, and 11β-hydroxyandrosterone. The steroids were purchased from Sigma-Aldrich (Steinheim, Germany). Acetone (GC grade, Merck, Darmstadt, Germany) served as solvent in this case. The mixture was adjusted to different concentrations, so that ∼100 ng of each compound was injected independent of the solvent volume. This mixture also served to investigate the suitability of the system for analysis of high-boiling compounds. 4164 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

Figure 2. Example of an operational scheme of the GC/IRMS interface proposed by the authors. Only the first minutes of a measurement are shown.

In order to assess and compare the quality of the chromatography, the peak width (PW) and the tailing factor (TF) at 5% peak height were calculated. For this purpose, the analyses were exported to ASCII files. These data then were analyzed by own functions written in the R-language.15 The latter is a programming environment developed for vectorized and statistical computation. Solely the signal m/z 44 was processed. The PW and TF values were compared to measurements of the same standard mixture that were performed on a system equipped with a conventional GC/IRMS interface. This device encompassed a GC 5890 (Agilent Technologies) and a Delta C mass spectrometer connected by a GC Combustion Interface II (Thermo Electron). The conventional system, however, already was modified and optimized for the analysis of high-boiling compounds, especially steroids. Metal tee pieces were substituted by glass pressfit connectors. The water trap and the placement of the furnace and of the capillaries were modified as described above. Analyses were performed in splitless mode, and a deactivated fused-silica capillary (i.d. ) 0.32 mm, l ) 2 m) served as retention gap, which facilitated proper splitless injections. The temperature programs on the two systems inevitably differed in their initial parameters. On the new system, solvent vent was performed at 40 °C for 90 s after which the GC was kept at this temperature for another 90 s. The splitless interval on the conventional system lasted 90 s and was performed at 60 °C. Subsequently, the temperature programs were identical. Temperature was raised to 245 °C at a rate of 30 °C/min. It then was raised to a final value of 300 °C at a rate of 3 °C/min. The two systems also differed in the carrier gas control. The conventional system did not possess a constant flow option and therefore had to be operated at constant pressure, which was adjusted to 30 psi. The net flow of this system was measured at the open split connection to the MS using a bubble meter. It was measured during the second temperature ramp (245-300 °C) and roughly varied from 1.3 to 1.0 mL/min. In order to test the suitability of the new interface for analysis of low-boiling compounds, a mixture of diisopropylamines (DIPAs, (15) R Development Core Team. R: A language and environment for statistical computing; R Foundation for Statistical Computing: Vienna, Austria, 2004.

Figure 3. Intensity of residual solvent peaks with and without operation of the auxiliary pump. The injection volume was 1 µL. The ordinate has been scaled to a maximum intensity of 1500 mV. A signal from an injection volume of 7 µL with pump operation has been added.

synthesized in the laboratory of the authors) with different chain lengths (6-10, 12-18, and 20 carbons; DIPA-6-DIPA-20) was analyzed at different initial temperatures of the injection port. The concentrations were ∼100 ng/µL for each compound. The solvent was n-hexane (analytical grade, Merck). The initial temperatures were set to 5, 10, 15, 20, 30, 35, and 40 °C. This should induce different conditions for the trapping of the DIPAs, where an efficient trapping is limiting for the analysis of low-boiling compounds. Starting from the abovementioned temperatures, a first ramp of 30 °C/min followed until 120 °C was reached. Subsequently, the temperature was raised to a final value of 300 °C at a rate of 15 °C/min. All other parameters were as described above. The dependence of the response from the respective compound and from the initial temperature was evaluated by descriptive statistics. A loess16 scatterplot smoother was employed to allow for visual inspection of these data. The loess algorithm serves explorative rather than explanatory purposes and does not presume any specific relationship between the variables. The influence of different initial temperatures on the δ13C values of the various DIPAs was investigated by fitting a linear mixed effects model17 to the data set. The number of the analysis was regarded as random effect in order to account for the runto-run analytical error. Initial temperature and compound identity were regarded as fixed effects. RESULTS AND DISCUSSION Solvent Elimination. Figure 3 shows the effects of the auxiliary pump operation on the residual amount of solvent. When 1 µL is injected without pumping, a signal with an intensity of ∼35 V appears at 200 s. Although the width of this signal is less than 7 s, its area sums up to more than 175 V‚s. This peak can be removed completely when the auxiliary pump is operated during (16) Cleveland, W.; Grosse, E.; Shyu, W. Local regression models. In Statistical Models in S; Chambers, J., Hastie, T., Eds.; Wadsworth & Brooks/Cole: Pacific Grove, CA, 1992. (17) Pinheiro, J.; Bates, D.; DebRoy, S.; Sarkar, D. nlme: Linear and nonlinear mixed effects models; 2004 R package version 3.1-50.

Figure 4. Influence of different temperatures during solvent vent on peak areas of different DIPAs.

solvent vent. A signal from an injection of 7 µL of acetone with pump operation has been added. It is confirmed that complete removal of the solvent can be achieved by operation of the auxiliary pump. No signal can be found at 200 s in case the pump is switched on. However, a broad signal can be found at 250 s, which obviously shows systematic variation with injected solvent volume and pump operation. A solvent volume of 1 µL results in a signal of 380 mV, whereas this is reduced to an intensity of 240 mV by operation of the pump. This phenomenon probably can be attributed to impurities in the acetone. A reduced pressure in the injection port would give rise to evaporation and hence to increased loss of this contaminant fraction. Experiments conducted later that employed tert-butyl methyl ether freshly distilled over calcium hydride as solvent resulted in total disappearance of any peak in the mentioned time interval. This supports the abovementioned hypothesis. In addition, it suggests the use of solvents of very high purity. Low-Boiling Compounds. Figure 4 shows the relationship of the responses (peak area m/z 44, V‚s) and initial temperatures grouped by the number of carbons of the respective DIPA. DIPAs with less than 12 carbons exhibit a negative relationship between temperature and response. This trend clearly inverts from C-14 on. Moreover, the trend becomes nonlinear. Independent from the chosen temperature, the compounds with shorter chains exhibit smaller responses. However, this relationship is nonlinear and DIPA-18 and DIPA-20, for instance, tend to give smaller responses than DIPA-17. The vapor pressure of the short-chain DIPAs certainly is still too large to facilitate efficient trapping of these molecules even at relatively low temperatures. Consequently, a significant portion is lost during solvent vent and this portion must increase with temperature. In contrast, long-chain DIPAs seem to be trapped efficiently at elevated temperature. The loss at low temperatures probably can be observed because they are carried out of the insert by solvent droplets. This phenomenon, however, will also be present for the shorter chain molecules. Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4165

Figure 5. Influence of different temperatures during solvent vent on δ13C values of different DIPAs.

Compared to 5 °C, the vapor pressure of acetone is roughly five times higher at 40 °C (∼0.1 and ∼0.5 bar, respectively). Efficient trapping of long-chain DIPAs therefore might be supported by relatively high temperatures, which effect rapid evaporation of the solvent. Independent from the underlying mechanism, the question arises whether the observed incomplete sample transfer of lowboiling compounds is accompanied by significant isotope fractionation. Figure 5 shows the relationship of initial temperature and measured δ13CVPDB values grouped by component. Regression lines from the fitted linear mixed effects model have been added. With the exception of DIPA-6 and DIPA-7, the regression lines exhibit a positive slope. However, none of the slopes is significantly different from zero at p < 0.05. The total standard deviations of the δ13CVPDB values of these compounds range from (0.14 (DIPA10) to (0.40‰ (DIPA-15). The other standard deviations fall between (0.2 and (0.3‰. These values are slightly larger than the empirical precision of 13C/12C measurements performed by GC/C/IRMS, which is usually close to (0.2‰. The slopes of DIPA-6 and DIPA-7 are negative and significantly different from zero (p < 0.05). It has to be mentioned that the peak heights of these compounds generally were much smaller than 0.5 V. As can be taken from Figure 4, the corresponding peak areas usually fell well below 1 V‚s. At such low signal intensities, isotope ratio measurements usually are not signal independent. One reason for this phenomenon is the presence of H• donors in the ion source, most importantly water. In the case of CO2 analysis, the reaction 12C16O+ + RH f H12C16O+ + R• will result in a bias toward 2 2 apparent 13C enrichment. This systematic error increases with smaller relative CO2 concentration.18 On the other hand, at low intensities the relative contribution of column bleeding inevitably (18) Leckrone, K. J.; Hayes, J. M. Anal. Chem. 1998, 70, 2737-2744.

4166

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

increases. Toward lower intensities, 13C/12C ratios calculated from a GC signals therefore will be considerably influenced by the 13C/12C ratio of the GC column. With respect to DIPA-6 and DIPA-7, it can be assumed that the observed relation between temperature and calculated 13C/ 12C ratio is a result of different signal intensities rather than of significant isotope fractionation induced by sample losses. In the latter case incidentally, positive slopes are to be expected because of the preferential loss of light isotopologues. High-Boiling Compounds. Figure 6 shows a comparison of the peak shapes obtained from the steroid standard mixture measured on the conventional and on the new interface. The peaks were centered to identical retention times and are drawn to the same scale. They are presented according to the order of elution. The employment of the new interface generally results in largely improved chromatography. Foremost, much narrower peaks can be observed with less tailing and less fronting. However, the improvement appears to be more pronounced with respect to the tailing. A possible explanation is that the fronting is not caused by the properties of the interface itself but that it is due to some overload of the column. A look at Figure 7 supports this hypothesis. The tailing factors calculated at 5% peak height obtained by measurements on the conventional interface are only slightly larger than unity. In contrast, the use of the new interface results in significantly smaller tailing factors. They are consistently smaller than unity, indicating some fronting. It can be concluded that the tailing caused by the properties of the conventional interface has been widely removed. Under the usual conditions, the peaks only appear to be relatively symmetrical because the fronting masks the tailing. The true dimension of the improvements achieved by the new interface is illustrated by Figure 8. In comparison to the conventional interface, the peak widths at 5% height are considerably reduced. Moreover, increasing retention times cause only small increases in peak width. The earliest eluting compound (5R-androstane-3β-ol) exhibits a peak width of 6.2 and 4.9 s on the conventional and the new interface, respectively. In contrast, this difference increases to 15.4 versus 7.9 s for 11β-hydroxyandrosterone, the investigated compound with the greatest retention time. The improvements undoubtedly can be attributed to the design of the new interface. The conventional system as well as the new one did not possess a Nafion dryer, so this can be ruled out as a significant source of peak broadening. The combustion furnaces were placed identically, and avoidance of cold spots at the union therefore cannot be responsible for the improvement of the chromatography. Although constant pressure conditions as present in the conventional system are inferior to constant flow, this cannot explain the pronounced differences. The net flow of the carrier gas in the conventional system did not change significantly during elution of the steroids. It was similar to the value of the new system. Moreover, the shape of GC peaks obtained from underivatized steroids appears to be insensitive to different carrier gas flow. During method development, the authors never observed significant influence of this factor on the chromatographic resolution, where flow rates from 0.8 to 2.5 mL/min were tested. Only two factors remain to account for the observed improvements: (1) the performance charactersistics of the CIS itself and (2) the discontinuation of the back-flush principle.

Figure 6. Peaks of the steroid standard mixture measured on a conventional and on the new interface. The peaks are drawn at the same scale and centered to identical retention times. The numbers correspond to the order of elution. 1, 5R -androstane-3β-ol; 2, etiocholanolone; 3, 11-ketoetiocholanolone; 4, 5β -pregnane-3R-20R-diol; 5, 11β-hydroxyandrosterone.

Figure 7. Tailing factors (5% peak height) of the steroid standard mixture measured on a conventional and on the new interface (xj ( s, n ) 5). Compound numbers identical to Figure 6.

It was previously recognized by Goodman19 that chromatography is adversely affected by the back-flush design of GC/IRMS systems. While maintaining the injection mode, Goodman likewise observed significant improvements of the chromatography by introduction of an interface without back-flush mode. Therefore, the performance of the design proposed in the present study presumably can be attributed to this feature. All steroids were analyzed in free form. With respect to the performance of the new GC/IRMS interface, derivatization is by no means necessary. This at least applies to steroids with up to three oxygens. On conventional GC/IRMS interfaces, some improvement in resolution and peak shape can be achieved by derivatization, however, at considerable costs. In CSIA, derivatization is always problematic. On the one hand, it necessitates the correction of the calculated isotope ratios; on the other, it can be associated with isotope fractionation giving rise to systematic errors.20 Moreover, some common derivatization moieties can be harmful for the reactor, at least during prolonged operation periods. This especially applies to fluorinated reagents.21 But also the common introduction of trimethylsilyl groups will degrade the reactor sooner or later due to the formation and deposition of (19) Goodman, K. J. Anal. Chem. 1998, 70, 833-837. (20) Docherty, G.; Jones, V.; Evershed, R. P. Rapid Commun. Mass Spectrom. 2001, 15, 730-738.

Figure 8. Peak widths (5% peak height) of the steroid standard mixture measured on a conventional and on the new interface (xj ( s, n ) 5). Compound numbers identical to Figure 6.

silicon dioxide. When analyzing silylated compounds, many analysts favor the immediate injection of the reaction mixture into the GC. However, the substitution of traditional solvents by such reagents can cause considerable problems in GC/IRMS devices. This is due to possible blockage of the back-flush vent by formation of siloxanes or by the mere presence of viscous reagents such as the widespread N-methyl-N-trimethylsilyltrifluoroacetamide. Regarding these factors, the possibility to perform the analysis of underivatized compounds can be considered a big advance. This new option largely extends the range of molecules amenable to CSIA. As mentioned before, the stationary phase generally seems to be overloaded by the injected amounts of substance. Therefore, use of GC columns with higher capacities would probably result in even narrower peaks. Similarly, additional improvements can be expected by the intoduction of more polar phases, which are better suited for free compounds. (21) Meier-Augenstein, W. GC and IRMS Technology for 13C and 15N Analysis on Organic Compounds and Related Gases. In Handbook of Stable Isotope Analytical Techniques; de Groot, P. A., Ed.; Elsevier: Amsterdam, 2004; Vol. 1.

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

4167

CONCLUSION A new method to couple gas chromatography to isotope ratio mass spectrometry has been described. The main characteristic is employment of a cooled injection system and auxiliary vacuum to remove the solvent before passing the GC column. This facilitates extensive simplification of the device. As a result, maintenance workload is largely reduced. Even more important, chromatographic resolution improves significantly, which is obviously due to omission of tee pieces and unions. The benefit is most pronounced for high-boiling and relatively polar analytes. In many cases, derivatization now can be omitted, which excludes

4168

Analytical Chemistry, Vol. 79, No. 11, June 1, 2007

some important sources of systematic error in compound-specific isotope analysis. ACKNOWLEDGMENT Preliminary results from this study were presented at the 24th Cologne Workshop on Doping Analysis (June 4-9, 2006) and at the 29th annual meeting of the German Association for Stable Isotope Research (October 4-6, 2006). Received for review November 14, 2006. Accepted March 22, 2007. AC0621468