Anal. Chem. 2005, 77, 3365-3371
Method for Analysis of Polar Volatile Trace Components in Aqueous Samples by Gas Chromatography Johan Pettersson and Johan Roeraade*
Royal Institute of Technology, Department of Analytical Chemistry, SE-100 44 Stockholm, Sweden
A new method has been developed for direct analysis of volatile polar trace compounds in aqueous samples by gas chromatography. Water samples are injected onto a short packed precolumn containing anhydrous lithium chloride. A capillary column is coupled in series with the prefractionation column for final separation of the analytes. The enrichment principle of the salt precolumn is reverse to the principles employed in conventional methods such as SPE or SPME in which a sorbent or adsorbent is utilized to trap or concentrate the analytes. Such methods are not efficient for highly polar compounds. In the LiCl precolumn concept, the water matrix is strongly retained on the hygroscopic salt, whereas polar as well as nonpolar volatile organic compounds show very low retention and are eluted ahead of the water. After transfer of the analytes to the capillary column, the retained bulk water is removed by backflushing the precolumn at elevated temperature. For direct injections of 120 µL of aqueous samples, the combined time for injection and preseparation is only 3.5 min. With this procedure, direct repetitive automated analyses of highly volatile polar compounds such as methanol or tetrahydrofuran can be performed, and a limit of quantification in the low parts-per-billion region utilizing a flame ionization detector is demonstrated. Trace analysis of volatile compounds in aqueous matrixes has been and still is a considerable challenge for the gas chromatographer. Large efforts have been invested in this area, and today, numerous methods are available. Most of these methods include an intermediate step in which the solutes of interest are isolated from the water matrix and concentrated prior to GC analysis. A comprehensive review on the subject has been made by Vreuls et al.1 Liquid extraction is one of the classical approaches. Although basically a laborious operation, this procedure has also been automated and performed on-line with GC.2,3 Another commonly employed strategy is the purge-and-trap method.4 The benefit of this approach is that nonvolatile materials are not collected, while the volatile solutes of interest are concentrated, (1) Vreuls, J. J.; Louter, A. J. H.; Brinkman, U. A. Th. J. Chromatogr., A 1999, 856, 279-314. (2) Roeraade, J. J. Chromatogr. 1985, 330, 263-274. (3) Fogelqvist, E.; Krysell, M.; Danielsson, L. G. Anal. Chem. 1986, 58, 15161520. (4) Snow, N. H.; Slack, G. C. Trends Anal. Chem. 2002, 21, 608-617. 10.1021/ac040170k CCC: $30.25 Published on Web 04/19/2005
© 2005 American Chemical Society
enabling low detection limits. Another widely used concentration principle is based on solid-phase extraction (SPE),5 including solidphase microextraction (SPME)6 and open tubular traps.7,8 For nonpolar analytes, these concentration methods can be quite straightforward and reliable; however, for analysis of volatile polar trace components, these techniques are not particularly suitable. The fundamental problem is that the affinity of the analytes for water is much stronger than their affinity for extraction solvents or adsorbents, thus leading to very poor extraction efficiencies. Several attempts have been made to improve the situation, for example, in SPME, in which the use of polar sorptive or adsorptive fiber coatings has been suggested, as well as an addition of salt to the water matrix or adjusting the pH.9,10 However, recoveries are still low and are critically dependent on the composition of the matrix. Also, SPE is an unsuitable strategy for determination of volatile polar solutes, since retention of such solutes is very poor in the presence of water. In fact, this method is difficult to employ even for enrichment of volatile nonpolar compounds. Residual water, stripped from the SPE cartridge disturbs the subsequent analysis.11 Attempts to improve this situation have been made12-14 in which the extract from the SPE column was dried in a separate drying cartridge. However, to achieve low limits of detection, a large-volume injection of the extract is required, and thereby, the most volatile solutes risk coelution with the solvent. In view of the fundamental problems described above, a more attractive approach would be a direct, large-volume injection of the aqueous sample onto an empty capillary precolumn (retention gap), followed by analyte concentration using the solvent effect.15 In this way, potential sources of errors related to sample pretreatment would be eliminated. Although many attempts have been (5) Lisˇka, I. J. Chromatogr., A 2000, 885, 3-16. (6) Arthur, C. L.; Pawliszyn, J. Anal. Chem. 1990, 62, 2145-2148. (7) Grob, K.; Habich, A. J. Chromatogr. 1985, 321, 45-58. (8) Blomberg, S.; Roeraade, J. HRC CC 1988, 11, 457-461. (9) Buchholz, K. D.; Pawliszyn, J. Environ. Sci. Technol. 1993, 27, 2844-2848. (10) Matisova´, E.; Sedla´kova´, J.; Sleza´ckova, M.; Welsch, T. J. High Resolut. Chromatogr. 1999, 22, 109-115. (11) Poole, C. F.; Poole, S. K. Chromatography Today; Elsevier Science Publications B. V.: Amsterdam, 1991; p 783. (12) Vreuls, J. J.; Ghijsen, R. T.; de Jong, G. J.; Brinkman, U. A. Th. J. Chromatogr. 1992, 625, 237-245. (13) Pico´, Y.; Vreuls, J. J.; Ghijsen, R. T.; Brinkman, U. A. Th. Chromatographia 1994, 38, 461-469. (14) Hankemeier, T.; Louter, A. J. H.; Dalluege, J.; Vreuls, R. J. J.; Brinkman, U. A. Th. J. High Resolut. Chromatogr. 1998, 21, 450-456. (15) Grob, K., Jr. J. Chromatogr. 1983, 279, 225-232.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005 3365
made, this approach has not proved useful. Volatile solutes are lost during the water elution step. These losses are related to the relatively high boiling point of water, the poor wettability of the precolumn, and fractionation effects due to azeotrope formation.16,17 Attempts to improve the situation by adding cosolvents to the sample18,19 were not successful for refocusing the most volatile solutes. Such compounds elute as broadened or split peaks or coelute with the cosolvent. A different strategy was adopted by Schomburg et al.,20 who employed a double-oven system in which the water was eluted on a Tenax column prior to a transfer of the adsorbed analytes to a capillary column. Although large-volume injections could be performed, the setup required a 1:30 split injection. Thus, the analysis was dealing with only parts-per-million concentrations. In addition, methods based on the use of a programmed temperature vaporizer (PTV) filled with an adsorbent have been used in which the water vapor is foreflushed via a large split flow.21-24 However, due to the limited selectivity of the adsorbent, a loss of the more volatile polar components is unavoidable. In addition, artifacts stemming from a thermal breakdown of the Tenax adsorbent have been observed, resulting in a chromatographic interference with the analytes.20,21 The present approach is fundamentally different. We propose a concept in which a dry salt is utilized in a precolumn kept at optimized temperature. Lithium chloride has particularly interesting characteristics, as has been reported by Kolb et al.,25 who used this material for removing water vapor from headspace samples. Since the hygroscopic salt has a greater affinity for water than for organic solutes, including polar components, the solutes are eluted ahead of the bulk of the water. Subsequently, the water, which is retained in the precolumn, is backflushed while the analytical separation of the organic trace components is performed. The capacity of the precolumn is adequate for large-volume, direct injections of aqueous samples. Since the interference of the bulk water is eliminated, even polar volatile trace components, such as methanol, can be directly quantified down to very low concentrations. EXPERIMENTAL SECTION Instrumentation. A HP6890 gas chromatograph (Agilent Technologies, Palo Alto, CA) equipped with a split/splitless injector (180 °C), a flame ionization detector (FID, 300 °C), and a thermal conductivity detector (TCD, 170 °C) was used in the experiments. A thermally isolated precolumn backflush unit having a separate heating control was integrated with the original Agilent injector. The basic instrumental configuration has previ(16) Grob, K., Jr.; Neukom, H.; Li, Z. J. Chromatogr. 1989, 473, 401-409. (17) Grob, K., Jr.; Li, Z. J. Chromatogr. 1989, 473, 381-390. (18) Grob, K., Jr.; Li, Z. J. Chromatogr. 1989, 473, 391-400. (19) Goosens, E. C.; de Jong, D.; de Jong, G. J.; Brinkman, U. A. Th. J. Microcolumn Sep. 1994, 6, 207-215. (20) Schomburg, G.; Bastian, E.; Behlau, H.; Husmann, H.; Weeke, F.; Oreans, M.; Mu ¨ ller, F. HRC CC 1984, 7, 4-12. (21) Sen ˜ora´ns, F. J.; Tabera, J.; Ville´n, J.; Herraiz, M.; Reglero, G. J. Chromatogr. 1993, 648, 407-414. (22) Mu ¨ ller, S.; Efer, J.; Engewald, W. Chromatographia 1994, 38, 694-700. (23) Mol, H. G. J.; Janssen, H.-G. M.; Cramers, C. A.; Brinkman, U. A. Th. J. High Resolut. Chromatogr. 1993, 16, 459-463. (24) Pocurull, E.; Biedermann, M.; Grob, K. J. Chromatogr., A 2000, 876, 135145. (25) Kolb, B.; Zwick, G.; Auer, M. J. High Resolut. Chromatogr. 1996, 19, 3742.
3366 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
Figure 1. Schematic drawing of the precolumn backflush system operated in the splitless injection mode: (A) injection port; (B) injector heating block, kept at 190 °C; (C) precolumn extension; (D) precolumn; (E) capillary column; (F) carrier gas supply; (G) flow meter/ regulator; (H) 10-port switching valve, kept at 120 °C; (I) pressure gauge; (J) backflush line with shutoff valve and pressure regulator; and (K) purge line with a needle valve. A parallel FID/TCD is shown in the figure, which was used in the experiments to monitor water breakthrough.
ously been outlined by Hagman et al.26 A schematic of the setup is shown in Figure 1. A packed precolumn (total length, 20 cm, including an empty section for sample injection, 8 cm) was mounted in the injector/precolumn unit. The packing material containing the LiCl was positioned in the heated precolumn zone, whereas the upper, empty part was located in the injector zone. A 10-port switching valve (Valco Instruments Co. Inc., Houston, TX, P/N C10UWP) was used to direct the route of the carrier gas flow. In the injection mode, the carrier gas was guided directly to the inlet of the precolumn, which was connected in series with the capillary column. After injection and subsequent prefractionation of the sample, the valve was switched to engage the backflush mode. The carrier gas was then routed to the lower part of the precolumn unit, thus backflushing the precolumn while maintaining the carrier gas flow through the capillary column. The system could be operated in either split or splitless mode. In the split mode, the backflush route (J in Figure 1) was opened during the injection and the sample transfer stage, thus serving as a split line. A septum purge line (3 mL/min) was included. In the backflush mode, this purge line is utilized to sweep out any dead volumes around the lower part of the precolumn. After sample transfer, the precolumn was regenerated in the backflush mode during each analysis by raising the temperature of the precolumn zone to 160 °C and increasing the flow rate to 85 mL/ min. Sample Introduction. A syringe-based injection device was constructed to enable large-volume injections at controlled and low injection speeds. This device consisted of a motor-driven micropositioner (MM-3M-EX-2, P/N 25100-07 and 25100-21, Fine Science Tools GmbH, Heidelberg, Germany) and a precision controller (MC-3B-II-S, P/N 25100-22, Fine Science Tools GmbH) assembled to a 50-µL gastight injection syringe (Agilent Technologies, P/N 5182-9667) or a 250-µL syringe (Hamilton Company, Reno, NV, P/N 1725LNT). For the injection of small volumes (1 µL), a standard automatic liquid sampler (HP7673, Agilent Technologies) was employed. (26) Hagman, G.; Roeraade, J. J. High Resolut. Chromatogr. 1990, 13, 99-103.
Columns and Connectors. The packed precolumn (borosilicate glass) had an internal diameter of 4 mm. The packing material consisted of Chromosorb W 60/80 mesh AW impregnated with lithium chloride (7447-41-8, 99.99+%, Aldrich Chemical Co, Milwaukee, WI, P/N 203637) in a weight ratio of 1:1. The packing material was prepared by suspending/dissolving both components in water, followed by drying in a rotary evaporator at 50 °C. Two silanized glass wool plugs (Supelco, Bellefonte, PA, P/N 2-0409) were used to keep the packing material (0.79 g) in position. An initial conditioning of the precolumn can be performed at very high temperatures due to the absence of organic material; however, the best results were obtained with a flow of humid nitrogen at 170 °C for 4 h, which virtually eliminated all background peaks. A DB-624 capillary column (30 m × 0.32 mm i.d. × 1.8 µm film thickness, Agilent Technologies) was used as analytical column. When cryofocusing was employed, a short DB-WAX capillary column (0.9 m × 0.32 mm i.d. × 0.5 µm film thickness, Agilent Technologies) was connected with a press-fit glass union ahead of the DB-624 and the cryotrap. The purpose of this short piece of capillary column was to avoid a plugging of the capillary column by preventing condensed water from directly reaching the cryotrap. The effluent at the exit of the capillary column was divided into two equal parts, using a 1/16-in. tee piece (Valco Instruments Co. Inc., P/N ZT1), which were directed into two uncoated, deactivated fused-silica capillaries (0.3 m × 0.32 mm i.d., Agilent Technologies, P/N 19091-60600), each connected to one of the detectors to perform parallel detection. Helium was used as carrier gas. Sample Preparation and Calibration. Samples were prepared by diluting test compounds in water of chromatographic quality (Lichrosolv, Merck, Darmstadt, Germany). All test compounds were of analytical grade. Calibration of the TCD was performed by injecting 1.0 µL of pure water while operating the precolumn in a continuous foreflush mode. This procedure results in a complete elution of water from the precolumn. Safety Considerations. Lithium chloride is harmful if swallowed. Lithium chloride is also irritating to eyes, the respiratory system, and skin. In case of eye or dermal exposure, rinse with copious amounts of water. In any case of exposure, contact a physician. RESULTS AND DISCUSSION Water is often the overwhelming matrix component in connection with analysis of trace organic volatiles, and its presence in a capillary GC column usually leads to peak distortions and other disorders. The removal of water prior to GC separations is, therefore, required. For gaseous samples, several methods for water removal have been utilized, such as dynamic or static headspace, including cold trapping,27 desiccants,25,28-30 or Nafion permeation tubes.27,31 These methods may be more or less successful, depending on the particular application; however, for (27) Noij, T.; Van Es, A.; Cramers, C.; Rijks, J.; Dooper, R. HRC CC 1987, 10, 60-66. (28) Schmidbauer, N.; Oehme, M. HRC CC 1985, 8, 404-406. (29) Matusˇka, P.; Koval, M.; Seiler, W. HRC CC 1986, 9, 577-583. (30) Doskey, P. V. J. High Resolut. Chromatogr. 1991, 14, 724-728. (31) Burns, W. F.; Tingey, D. T.; Evans, R. C.; Bates, E. H. J. Chromatogr. 1983, 269, 1-9.
Figure 2. The vapor pressure of pure water and five saturated salt solutions vs the temperature of the solution. Data were obtained from literature tables.39
aqueous samples, there is still a considerable need for new methodology, as pointed out in the introduction of this report. In the current concept, the entire aqueous sample is first vaporized in the upper empty section of a precolumn and subsequently passed through a section containing a hygroscopic salt. Thereby, most of the water vapor is retained by the salt, while the volatile organic components are eluted ahead of the water. In this way, even large sample volumes can be injected without excessive water breakthrough. The choice of salt is important from several aspects. It must have a good water retaining capacity and be easy to regenerate. Furthermore, the packed bed should retain its morphology as much as possible. Many salts liquefy very quickly, and a caking occurs after reconditioning, leading to pressure drop variations or, in the worst case, a blockage of the precolumn. Another important aspect is the inertness of the salt. In earlier work, Kolb et al.25 successfully utilized lithium chloride for drying headspace samples. This particular salt seems to be most suitable. An additional important advantage of LiCl is that it strongly reduces the vapor pressure of water. This can be seen from the graphs in Figure 2, where the vapor pressure of water over saturated solutions of some common salts is shown as a function of temperature. Although other desiccants may have better characteristics in terms of reduction of vapor pressure, it is crucial to consider reactivity, catalytic behavior, and adsorptivity, particularly at elevated temperature. Thus, after summarizing these different aspects, LiCl was chosen as the water adsorbent in our present work. Sample Introduction. Direct splitless injection of large amounts of an aqueous sample into a hot injector or precolumn creates a serious risk for overflow of the injector and a subsequent backflow into the system, leading to sample losses and detrimental memory effects.32 The evaporation of l µL of water generates ∼1 cm3 of vapor (at a typical injector pressure of 1 bar), which is roughly similar to the internal volume of commercial injector liners. This would suggest that the sample has to be injected at a very slow speed. Assuming an inlet flow of 4 mL/min through (32) Grob, K. Classical Split and Splitless Injection in Capillary Gas Chromatography; Dr. Alfred Hu ¨ thig Verlag: Heidelberg, 1986.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3367
the capillary column, injection of a sample volume of 120 µL would require at least 30 min to maintain stable foreflush conditions through the column. Such an extended injection time is not particularly attractive. In our experiments, we found that we could inject the sample at remarkably high rates (120 µL/min, splitless) without leading to any signs of sample backflow or system contamination. We concluded that the reason for this beneficial behavior is that the water vapor is almost instantaneously adsorbed by the top layer of the anhydrous lithium chloride. A semiliquid phase in the top layer of the LiCl seems to be created, since we noted some changes of the morphology of this layer. However, no caking of the salt was experienced, even after more than 50 large-volume injections. The recondensation of water in the entrance zone strongly enhances the overall transport of sample into the precolumn. The mechanism is similar to the situation in regular splitless injection if the solvent is recondensed in the first part of the capillary column. Grob32 has shown that under such conditions, the sample vapor transport from the evaporation zone into the capillary column is strongly accelerated. The condensation capacity of the packed salt bed is far greater than that of the head of a capillary column, and therefore, a very high sample injection speed is possible. Optimization of the Operating Parameters. For practical applications, a number of parameters need to be optimized. An essential issue is to minimize the amount of water which enters the capillary column, since large-volume injections require the analytes to be reconcentrated by cryofocusing. It is obvious that the water retained by the LiCl is not permanently absorbed. For a given temperature, an equilibrium will be established in the precolumn, where the water has its specific vapor pressure (Figure 2). Thus, the water is gradually stripped from the precolumn; however, as soon as the analytes have been transferred to the capillary column, the remaining water is backflushed. Therefore, the amount of water which will enter the capillary column is limited and is dependent on the temperature of the LiCl and the effluent volume which is needed for quantitative transfer of the analytes from the precolumn to the capillary column. The breakthrough behavior of a polar model substance (nbutanol) eluting from the precolumn was studied by performing a series of splitless injections (1 µL of a 220 ppm water solution of the analyte) onto the precolumn. The capillary column was kept at 50 °C (k′ for the butanol ∼10), while no analyte reconcentration by cryofocusing was included. Under these conditions, peak broadening of the butanol in the capillary column is small compared to peak broadening in the precolumn. The different elution profiles of the butanol peak were studied by gradually shortening the period between sample injection and backflush activation of the precolumn. In Figure 3, the results of these experiments are shown with the precolumn kept at 110 °C. As can be seen, an ∼1.6-min foreflush is needed for a total transfer of the butanol. The k′ of the butanol on the precolumn was 1.15, as calculated from the retention time of methane (tM). The anomalous retention time shift of the butanol peak in Figure 3 is due to pressure drop effects in the system during backflush. In another series of experiments, the dependence of the breakthrough volume on the precolumn temperature was studied, again with butanol as a model sample. Apart from the precolumn 3368 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
Figure 3. Five chromatograms showing the gradual elution of n-butanol at a precolumn temperature of 110 °C. These curves were obtained by varying the time for engaging the backflush mode after injection. The time in foreflush of each analysis is shown next to the corresponding chromatogram. Sample: 1 µL of water containing n-butanol (220 ppm). Pulsed splitless injection with a splitless flow rate of 9.3 mL/min (2 bar).
Figure 4. The peak area of n-butanol vs the effluent volume of the precolumn shown for several temperatures of the precolumn. In addition, the breakthrough curve for methane is shown. Apart from the temperature of the precolumn, the experimental conditions are given in Figure 3. Table 1. Retention Factor (k′) of n-Butanol on the Precolumn at Different Temperatures precolumn temp (°C)
retention factor (k′)
140 130 120 110 100
0.14 0.31 0.59 1.15 2.04
temperature, the same setup conditions as shown in Figure 3 were used for the experiments. Figure 4 shows the results. The points in the graphs represent the integrated area response for the amount of eluted butanol versus the effluent volume. In addition, a response graph for methane is included, which allows a direct calculation of k′ for the butanol at different temperatures The calculated values of k′ are shown in Table 1. Note that the LiCl/ water acts as a stationary phase. At the higher temperatures, butanol is mainly present in the gas phase of the precolumn, and fast elution is possible. Apart from reducing the retention volume of the analytes in the precolumn, another important objective is to find conditions
Figure 5. The peak area of water versus the effluent volume from the precolumn shown for several temperatures of the precolumn. See Figure 3 for experimental conditions (except for the precolumn temperature).
Figure 6. The peak area of water observed at a 90% elution of n-butanol. See Figure 3 for experimental conditions (except for the temperature of the precolumn).
under which the amount of water which is coeluted with the analyte is minimal. Using the data from the TCD, which was connected in parallel with the FID, the rate of water breakthrough was evaluated. The peak areas of water versus the effluent volume (same experiments as in Figure 4) for a number of precolumn temperatures are presented in Figure 5, which shows that the water breakthrough per effluent volume is nearly constant for a given precolumn temperature. To find the optimal value for the precolumn temperature, the elution volume of butanol was first determined for each temperature using the results compiled in Figure 4. The elution volume was defined as the effluent volume at which 90% of the total peak area for the butanol was obtained. Thereafter, the data shown in Figure 5 were interpolated to determine the level of water transferred to the capillary column at the corresponding temperature and effluent volume. In Figure 6, the calculated levels of water in terms of peak area versus the temperature are shown. Apparently, there is an optimum value at 110 °C. The physical processes that interact to cause the shape of the curve are currently not clear to us. Performance of the System. Using the optimized system, the sample capacity was investigated by performing a stepwise increment of the injection volume of a test sample, containing four low-boiling polar compounds in water. Cryofocusing on the capillary column was employed. The injection speed was 120 µL/ min, and the time before engaging the backflush mode was 3.5 min for all runs. The results in terms of peak area versus sample volume are shown in Figure 7. As can be seen, injections of up to
Figure 7. Absolute area response as a function of the injection volume for four polar volatile test solutes. Sample: 1 ppm of each of the component in water. Pulsed splitless injection at a flow rate of 9.3 mL/min. The effluent volume prior to engaging the backflush was 33 mL.
at least 120 µL can be performed without leading to any significant deviation from the straight lines, which indicates a complete transfer of the compounds to the main column. When still larger sample volumes are injected, the most polar solute, n-butanol, is no longer quantitatively transferred. The incomplete elution is caused by the prolonged sample injection time, leading to solute band broadening and the increased retention on the water phase that is injected into the precolumn. An interesting observation was that the magnitude of water breakthrough was independent of the injected sample volume. In fact, the results compiled in Figure 7, which were obtained under identical instrumental conditions, were from sample volumes between 10 and 160 µL, but the average amount of breakthrough of water was 0.61 µL, with a relative standard deviation of only 1.0% (n ) 10). This indicates that the vapor equilibrium condition established in the precolumn is constant, regardless of the amount of water present. To understand this mechanism, it is useful to consider the vapor pressure of water over a saturated salt solution, which contains an excess of salt. In such a system, adding more water will have no influence on the vapor pressure as long as the solution remains saturated. The analytical performance was further evaluated using a water sample containing nine polar, low-boiling analytes and a sample injection volume of 100 µL. The repeatability of the system was determined for two different analyte concentrations, 20 and 5 ppb. There were occasional signs of peak splitting due to the presence of the recondensed (0.61 µL) water. Using a 1:1 split ratio eliminated these problems. The repeatability data obtained (shown in Table 2) were very good, considering the low concentration of the analytes. The recovery of the analytes was determined by measuring area response data obtained from 100-µL injections (n ) 7) of the 20 ppb solution and comparison of these data with the results from 1-µL injections (n ) 3) of a 2 ppm solution in which the salt in the precolumn was replaced by a short plug of silanized glass wool. The comparative results are shown in Table 2, and it can be concluded that very little analyte is lost on the salt precolumn. The almost quantitative elution of these polar analytes at trace levels is noteworthy, particularly since 50% of the precolumn Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3369
Table 2. Performance of the LiCl Precolumn, Based on Absolute Peak Areasa RSD (%) component methanol acetone isopropyl alcohol tetrahydrofuran ethyl acetate n-butanol 1,3-dioxane methyl isobutyl ketone mesityl oxide
at 20 ppb
at 5 ppb
recovery (%) ((95% confidence limit)
R2 2.5-100 ppb
8.8 6.3 1.9 2.7 1.6 2.5 1.3 2.5
8.1 15.1 11.6 7.2 4.8 8.2 3.3 5.8
104.6 ( 13.5 175.7b ( 15.9 97.8 ( 5.6 91.8 ( 6.1 107.1 ( 5.6 106.7 ( 6.4 90.0 ( 5.4 93.2 ( 6.5
0.9920 0.9649b 0.9962 0.9976 0.9981 0.9982 0.9982 0.9982
2.8
13.7
93.1 ( 5.2
0.9980
a The table shows the recoveries, coefficients of determination (R2) and the relative standard deviations (RSD), based on data from seven replicate runs. Each calibration standard was used for four replicate measurements. GC operating parameters: pulsed split injection (1:1) at a precolumn flow rate of 18.6 mL/min and using a foreflush time of 2 min. Cryofocusing of the column inlet was applied to reduce band dispersion. Chromatographic run conditions: the oven was set to 60 °C for 7 min, then at 15 °C/min to 130 °C; the column flow was 2.0 mL/min (after the cryofocusing, which was discontinued 1.0 min after engagement of the backflush). b The high recovery value as well as the corresponding low R2 value is due to an artifact caused by an interfering background peak.
consists of nondeactivated Chromosorb W. It is commonly known that the use of bare Celite-based support materials, such as Chromosorb W, leads to a strong adsorption or total loss of polar trace compounds. The excellent results obtained can be explained by the extraordinary deactivation power of water present in the sample.33,34 The deactivation effect of water has also been exploited by other workers who used wetted carrier gas or steam as the mobile phase.35-38 In this way, highly polar compounds, such as 2,4-butanediol and glycerol, have been chromatographed using unsilanized Celite-type support material.38 To evaluate the linearity of calibration, standards with 2.5, 5, 10, 20, 50, and 100 ppb solutions of the nine analytes were analyzed. Excellent results were obtained for all compounds except acetone, due to coelution with a background peak. The coefficients of determination for the nine analytes are shown in Table 2, and most values are above 0.996. The limit of detection, calculated at a signal-to-noise level of 3:1, was found to be ∼1 ppb for all analytes. Finally, the robustness of the system was evaluated by repetitive injection of a sample originating from a sewage treatment plant. This sample was collected from the sedimentation compartment directly after aerobic degradation of organic material. Thus, this sample still contained a significant amount of impurities, including nonvolatile material. Figure 8a shows the first chromatogram of a series of 20 injections of 100 µL of the sample; Figure 8b shows the last chromatogram of the series. No deterioration of the chromatography or additional peaks originat(33) Knight, H. S. Anal. Chem. 1958, 30, 2030-2032. (34) Grob, K.; Vorburger, T. J. High Resolut. Chromatogr. 1996, 19, 27-31. (35) Nonaka, A. Anal. Chem. 1972, 44, 271-276. (36) Davis, A.; Roaldi, A.; Tufts, L. E. J. Gas Chromatogr. 1964, 2, 306-308. (37) Berezkin, V. G.; Viktorova, E. N.; Gavrichev, V. S. J. Chromatogr. 1988, 456, 551-556. (38) Viktorova, E. N.; Berezkina, L. G. J. High Resolut. Chromatogr. 1996, 19, 59-61. (39) Timmermans, J. The Physico-Chemical Constants of Binary Systems in Concentrated Solutions;Interscience Publications: New York, 1960; Vol. III.
3370 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
Figure 8. Chromatogram of water (direct injection of 100 µL) from the sedimentation compartment in a sewage treatment plant directly after aerobic degradation of organic material. (a) The first chromatogram; a new precolumn was used. (b) Chromatogram of the water after 20 consecutive analyses. GC operating parameters were the same as in Table 2 but with the oven temperature program ramp extended to 145 °C.
ing from residual sample buildup in the precolumn can be observed. The robustness of the system in terms of absence of memory effects is due to the fact that the precolumn is active in the backflush mode at an elevated temperature during most of the analysis period. This results in the removal of water from the injected sample, but in addition, components which are not transferred to the capillary column as well as possible degradation products stemming from residual material are backflushed out of the system. After a prolonged usage with heavily contaminated samples, the nonvolatile residues will start to affect the system performance due to adsorption and other adverse effects. At this instance, the precolumn needs to be replaced, but this is a simple operation, similar to replacing a conventional inlet liner in a GC injector. CONCLUSIONS A new approach for the analysis of polar volatile trace analytes in aqueous samples has been developed and evaluated. The concept is based on the use of an integrated precolumn containing lithium chloride, which provides a large difference in affinity between the water matrix and the analytes. The analytes are eluted ahead of the bulk amount of the water, which is removed from the system by backflushing at elevated temperature during the analysis. The injection and prefractionation of sample volumes of at least 100 µL is accomplished within only a few minutes, while low-boiling polar compounds are quantitatively transferred to the main separation column. A repeatability test with nine polar volatiles showed coefficient of variation values between 3.3 and 13.7% at the 5 ppb level. The correlation of determination (R2) was between 0.9920 and 0.9982. A series of 20 large-volume injections of water from a sewage plant did not lead to any deterioration of the resulting chromatograms. We conclude that the method provides a new useful platform for automated analysis of polar components in aqueous samples present in the low partsper-billion region.
ACKNOWLEDGMENT We gratefully acknowledge financial support from the Swedish Research Council.
Received for review October 13, 2004. Accepted March 15, 2005. AC040170K
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3371
