Anal. Chem. 2004, 76, 1696-1701
Inability of Unpacked Gooseneck Liners To Stop the Sample Liquid after Injection with Band Formation (Fast Autosampler) into Hot GC Injectors Stefan Bieri,† Philippe Christen,*,† Maurus Biedermann,‡ and Koni Grob‡
Laboratory of Pharmaceutical Analytical Chemistry, School of Pharmacy, University of Geneva, 20 Bd d’Yvoy, CH-1211 Geneva 4, Switzerland, and Official Food Control Authority of the Canton of Zurich, P.O. Box, CH-8030 Zurich, Switzerland
With the introduction of the fast autosampler, the rules on how to perform splitless injection should have been revisited: the fast autosampler releases the sample liquid as a band that is no longer nebulized by solvent vapors as in previous injection techniques. In empty gooseneck liners, the sample liquid is shot to the bottom of the injector and jumps around in a largely uncontrolled manner. Visual experiments showed that the liquid is partially rejected. Another portion passes by the column inlet into the column attachment region, while the last part may directly enter the column. The impact on chromatography was investigated by using a mixture of n-alkanes: the higher boiling components passing by the column entrance into the zone of the column attachment were largely lost, i.e., were discriminated against the volatile components. It was concluded that empty gooseneck liners are not suitable for injection by fast autosamplers. Discrimination due to Incomplete Elution in the Syringe Needle. Thermospray. Classical split and splitless injections into hot vaporizing injectors often suffer from a distortion of the sample composition, commonly termed “discrimination”. This may be a serious problem for samples composed of compounds covering a wide range of volatilities. Usually there is a loss toward the lateeluted components, i.e., a discrimination against the high-boiling compounds. Several mechanisms may alter the sample composition in the injector:1 One is the partial sample evaporation inside the syringe needle, which occurs when a sample in a volatile solvent is injected through a needle heated in the hot injector. While the volatile components evaporate together with the solvent, high-boiling compounds tend to be deposited onto the internal wall and remain in the needle.2 Two opposed strategies have been developed to cope with this problem: optimization of the process achieving * To whom correspondence should be addressed. E-mail: philippe.christen@ pharm.unige.ch. Fax +41 (0)22 379 68 08. † University of Geneva. ‡ Official Food Control Authority of the Canton of Zurich. (1) Grob, K. Split and Splitless Injection for Quantitative GC; Wiley-VCH: Weinheim, 2001. (2) Grob, K.; Neukom, H. P. J. High Resolut. Chromatogr., Chromatogr. Commun. 1979, 2, 15-21.
1696 Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
almost complete elution from the needle and fast autosampler injection suppressing evaporation inside the needle. Around 1980, the injection process was optimized by the first strategy, concluding that the “hot-needle technique” provided the least discrimination and the best reproducibility. The sample liquid is withdrawn into the barrel of the syringe, the empty needle inserted in the injector, allowing it to heat up to the injector temperature by waiting 3-5 s, and then the plunger is depressed as rapidly as possible.3 Since the losses of high-boiling compounds primarily occur in the rear part of the needle, the temperature of which is determined by the top part of the injector, it is essential that the injector is well heated to the septum.4 With these measures, discrimination against the high-boiling sample components is strongly reduced, but not completely overcome. Evaporation of solvent in the needle induces the liquid leaving the needle to nebulize, which results in so-called thermospray.5,6 The sample is split into microdroplets, which are suspended in the gas phase and evaporate to a large extent without contact with surfaces. This allows use of empty injector liners, which largely avoid contacts with possibly active sites in the injector and, therefore, provide a gentle sample evaporation. Fast Autosamplers and Injection with Band Formation. In 1985, Hewlett-Packard introduced the “fast autosampler”,7-9 capable of suppressing evaporation inside the needle by combining a short residence time of the needle in the injector (100-ms needle dwell time), a low temperature of the top part of the injector, and a rather short syringe needle. The fast autosampler eliminated problems related to evaporation inside the needle but also created a new one, which was overlooked for a long time. In 1992, two groups pointed out that during injection by a fast autosampler the sample liquid leaves the needle as a band that rushes through the hot vaporizer chamber at a high velocity. In an empty chamber, the band of liquid reaches the column entrance (3) Grob, K.; Rennhard, S. J. High Resolut. Chromatogr., Chromatogr. Commun. 1980, 3, 627-633. (4) Grob, K.; Neukom, H. P. J. Chromatogr. 1980, 198, 64-69. (5) Grob, K.; Biedermann, M. J. Chromatogr., A 2000, 897, 237-246. (6) Grob, K.; Biedermann, M. Anal. Chem. 2002, 74, 10-16. (7) Snyder, W. D. Technical Paper 108, Hewlett-Packard, Palo Alto, CA, 1985. (8) Grob, K.; De Martin, M. J. High Resolut. Chromatogr. 1992, 15, 335-340. (9) Qian, J.; Polymeropoulos, C. E.; Ulisse, R. J. Chromatogr. 1992, 609, 269276. 10.1021/ac035161a CCC: $27.50
© 2004 American Chemical Society Published on Web 02/13/2004
at such a high speed that the transfer of heat from the liner wall to the liquid required for evaporation (primarily the solvent) remains negligible and, hence, not even the solvent is evaporated.10 Visual experiments showed that the liquid hits the bottom of the vaporizing chamber and performs violent movements, jumping backward, sometimes nearly up to the septum. The droplets fall back again and largely evaporate from the metallic surface when straight, empty inserts are used. Since suppression of solvent evaporation inside the needle also suppresses thermospray, the starting point for sample evaporation in the injector fundamentally changes. When a straight, empty liner is used for injection with band formation, most of the sample material evaporates from the metallic surface at the bottom of the vaporizing chamber.11 As long as the droplets contain solvent, they “dance” above this surface, but when solvent evaporation is completed, the boiling point rises and most of the sample material is deposited onto it (or on to the nonevaporating byproducts deposited there by previous injections). Apparently some users noticed problems, and thus, the gold disk was introduced to protect the sample from the stainless steel surface at the bottom of the injector. The success of this measure is poorly documented. Later, the gooseneck liner (liner with taper at the bottom) was introduced: it was originally used with the restriction at the top, advertised to avoid loss of sample vapor by expansion backward out of the vaporizing chamber, particularly when overloading the liner with sample vapor, for which it was not very successful.12 Turned upside down, i.e., with the constriction at the bottom, it could allow the sample to be protected from the metal surface at the bottom. Despite frequent use, no data on the performance of this liner have been reported, nor has the question of the position (height in the vaporizing chamber) of the column entrance been answered. Stopping a liquid released by a fast autosampler as a band until it is evaporated is a demanding task.13,14 Liquids with a low-boiling matrix (solvent) are repelled from hot surfaces by a cushion of vapor. To keep them in place in a hot environment by deposition on a surface presupposes that this surface is cooled to the boiling point of the liquid. Packings of low thermal mass, such as glass, quartz, or fused-silica wool, are rapidly cooled, reliably stop the liquid, and keep it in the network of fibers until evaporation is completed. However, owing to insufficient inertness, they often cause problems for adsorptive, labile, and high-boiling sample components.15 Liners with built-in obstacles, such as the laminar or minilaminar liner (Restek), may trap the liquid, but they are not commonly used for splitless injection, nor has their effectiveness been demonstrated in routine application. Empty Gooseneck Liners for Splitless Injection with Band Formation. For splitless injection, classical teaching recommends the use of empty, straight liners. This concept originates from the time before the fast autosampler was introduced, i.e., before arresting of the sample liquid above the column entrance became (10) Grob, K. J. High Resolut. Chromatogr. 1992, 15, 190-194. (11) Biedermann, M. Visualization of the Evaporation Process during Classical Split and Splitless Injection in GC, CD-ROM; Restek Corp.; Bellefonte, PA, 2000. (12) Grob, K. Split and Splitless Injection for Quantitative GC; Wiley-VCH: Weinheim, 2001; p 277. (13) Grob, K.; Biedermann, M. J. Chromatogr., A 2000, 897, 247-258. (14) Grob, K.; De Martin, M. J. High Resolut. Chromatogr. 1992, 15, 399-403. (15) Grob, K.; Wagner, C. J. High Resolut. Chromatogr. 1993, 16, 464-468.
an issue.16 The empty liner minimized adsorptive and retentive surfaces in the vaporizing chamber, which often hampered the already critical transfer of the solute material into the column during the splitless period.17 The column entrance is installed 5 mm above the bottom of the chamber in order to accumulate septum particles and other dust below the column entrance and out of the way of the sample. This teaching is not valid for injection with the fast autosampler. Surprisingly, it seems that even 18 years after the introduction of this injection technique, nobody has offered a convincing solution or has even adequately addressed the problem. This paper does not present a solution to the problem, but it reports on the evaluation of the gooseneck liner as the probably most widely used means for halting the band of liquid released by a fast autosampler, using both chromatographic and visual methods. EXPERIMENTAL SECTION Instrumentation. A HP 5890 series II gas chromatograph (Agilent Technologies, Palo Alto, CA) was used with a flame ionization detector (FID). The instrument was equipped with its standard hot split/splitless injector as well as with a PTV injector (CIS-3, Gerstel, Mu¨lheim an der Ruhr, Germany) used as an oncolumn injector (on-column kit from Gerstel). Both injectors made use of electronic pressure control. A Graphpack column fitting was mounted in the split/splitless injector as an alternative to the classical gold-plated seal used to connect the column to the vaporizer. Splitless injections were carried out with a HP 6890 series fast automatic liquid sampler. Standard commercial gooseneck liners (4 mm i.d., 6.5 mm o.d., 78.5 mm long) from Restek (Bellefonte, PA) were used. Splitless injection involved an injector temperature of 300 °C, a splitless period of 60 s, and purge and septum purge flow rates of 50 and 3 mL/min, respectively. The 1-µL injections were performed with a 10-µL syringe equipped with a 42 mm × 0.63 mm o.d. needle with cone tip (Agilent). Helium was used as carrier gas in the constant flow mode (1 mL/min, 6.6 psi measured at 40 °C). The 1-µL on-column injections were accomplished with a 10µL syringe equipped with a 42 mm × 0.47 mm o.d. needle with a cone tip (Gerstel) through the septumless injector head of the CIS-3 injector. The injector temperature was held at 20 °C for 0.1 min and then increased at 5 °C/s to 40 °C and at 0.5 °C/s to 300 °C. During injection (0.05 min), the carrier gas was supplied at a constant pressure of 6.6 psi and then at a constant flow of 1 mL/ min. A 60 cm × 0.53 mm i.d. uncoated precolumn, deactivated in the laboratory with diphenyltetramethyldisilazane, was connected to the column by means of a glass press-fit connector. A 15 m × 0.25 mm i.d. HP5-MS column, 0.1-µm film thickness, was used with an initial oven temperature of 40 °C (1 min) and a program from 40 to 290 °C at 20 °C/min and hold 1.5 min. For accurate positioning of the column entrance in the vaporizing chamber, the distance between the column entrance and the bottom of the chamber (metallic injector body) was measured. A 6-mm-o.d. rod was introduced into the cool injector from which the septum part and the liner had been removed. Then the height of the capillary inlet inside the injector was indirectly (16) Grob, K.; Grob, K. J. High Resolut. Chromatogr., Chromatogr. Commun. 1978, 1, 57-64. (17) Grob, K.; Romann, A. J. Chromatogr. 1981, 214, 118-121.
Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
1697
Figure 1. Videoframes recording a 3-µL injection with band formation of a perylene solution in chloroform; empty gooseneck liner (a1) with a piece of 0.25-mm-i.d. capillary (a2); glass wool (a3). (A) situation before injection; (B) sample liquid leaving the needle as a band (b1) and hitting the bottom of the liner. The following frames (representing 40-ms duration) from the video show the situation at the times indicated in milliseconds at the bottom of each picture. For further explanations, see text.
determined by the displacement of the rod when the capillary was installed from the oven through the column attachment. Test Samples. Chromatographic Experiments. Liner performance was chromatographically evaluated with test mixtures containing n-alkanes ranging from C9 to C30 in hexane or dichloromethane at 10 µg/mL. Since the FID provides equal response per mass unit for the alkanes, peak areas are expected to be equal for all components. Absolute peak areas obtained by splitless injections were compared with those obtained by on-column injection of the same volume. Absence of losses in splitless injection resulted in area ratios of 1. Repeatability of absolute peak areas was expressed by relative standard deviation (RSD; n ) 6). Videotaped Observations. Visual experiments and their videotaping were largely performed as described in ref 5. Briefly, the gooseneck liner was inserted in a glass U-tube (gift from Restek, shown in Figure 1 of ref 13), imitating the conditions of a hot splitless injector. The device was thermostated at 200 °C in a silicone oil bath on a heating plate. The gas flow rate through the liner was adjusted to 2 mL/min, simulating the column flow during the splitless period. The injected solvents were rendered fluorescent by addition of 0.3 or 0.03% perylene to chloroform and hexane, respectively. Fluorescence of perylene is strong when in solution, but weak when dry; thus, the fluorescence of perylene can be used as a marker of nonevaporated solvent. The experiments were performed in a dark room using a strong (120 W) UV lamp at 366 nm. The videos were taped using a Sony DCRPC9 digital camera taking individual frames of 40-ms duration. RESULTS AND DISCUSSION Visual Experiments. Perylene solutions were injected into a transparent imitation of an injection chamber and videotaped to observe the behavior of a band of liquid after hitting the funnelshaped bottom of a gooseneck liner. Since no fast autosampler could be used for this installation, liquids were injected manually and band formation was forced by inserting the syringe needle into the hot zone by merely ∼5 mm, preventing relevant evaporation inside the needle. Empty Gooseneck Liner. (a) 0.25-mm-i.d. Capillary. Figure 1A shows an experimental setup used for testing the empty 1698 Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
gooseneck liner (a1). A short piece of 0.25 mm i.d. × 0.36 mm o.d. fused-silica capillary (a2) was mounted, the entrance positioned 5 mm above the base of the chamber and level to the orifice of the gooseneck. The liner was sealed against the U-tube by means of two O-rings, forcing the gas flow through the gooseneck. A dense plug of glass wool (a3) was placed closely below the liner in order to collect sample liquid and facilitate visual observation of a nonevaporated sample passing the gooseneck beside the column inlet. The videoframe taken at the moment of the injection (first 40 ms, B) shows the band of liquid shot to the bottom of the chamber (b1; tip of syringe needle not visible). When reaching the base of the chamber, some liquid was redirected upward at least 4 cm high (b2). Another portion of liquid left the chamber through the gooseneck, i.e., through the narrow space between the outer wall of the capillary and the restriction of the liner (b3). It is visible as a fluorescent spot (b4) in the glass wool. This fluorescence is more pronounced after 160 ms (c3). Frames C-F show the subsequent evaporation process. In the vaporizing chamber, the liquid contracted to some droplets hanging on the liner wall (c1, d1, and e1). Basically, sample liquid containing volatile solvent is unable to contact a hot liner wall (repulsion by vapors), but in this particular case, some liquid adhered to residual glass wool fibers remaining on the wall from previous experiments. The spot c2 shows some sample liquid adhering to the outer surface of the fused-silica capillary within the region of the gooseneck. It remained there for another 100 ms. Then the solvent was evaporated, leaving dry perylene (or, in real samples, the sample components) on the polyimide coating of the column wall. The strong fluorescence of spot c3 indicates that a large part of the sample liquid was shot beside the column through the gooseneck to the bottom, where it evaporated during ∼2 s (videoframe F). This experiment showed that the gooseneck is not appropriate for stopping the sample liquid above the column entrance. The round shape of the bottom surface redirects a part of the liquid upward, and the funnel shape guides another substantial portion to the bottom of the injector. In another experiment, the piece of 0.25-mm-i.d. capillary was sealed against the outlet of the gooseneck with epoxy glue, the
Figure 2. Videoframes (G-L) of an injection as in Figure 1, but using a piece of 0.53-mm-i.d. capillary with removed polyimide (g2). Gooseneck liner (g1); glass wool (g3); U-tube (g4). For further explanations, see text.
entrance being positioned at the same level as before, so that the entire gas flow was forced through the capillary. The portion of liquid shot past the column entrance hit the glue. Part of it was rejected, while the rest remained there. Only occasionally, when the band of liquid precisely hit the column entrance, did some liquid enter the column. (b) 0.53-mm-i.d. Capillary. A 0.53-mm-i.d. capillary has a roughly 4 times larger cross section than a 0.25-mm-i.d. capillary and correspondingly reduces the space between the column inlet and the constriction of the gooseneck, but also enables the liquid to enter the capillary more easily. The experimental setup is shown in Figure 2G. The capillary (g2), from which the polyimide was removed (burned off and dissolved in methanol), passed through the glass wool plug (g3). In this way, liquid leaving the liner (g1) through the gap between the gooseneck and the outer capillary wall was visible by fluorescence in the glass wool (g3), well distinguished from liquid entering and passing through the transparent capillary. Videoframe H shows the initial band of liquid (h1) released by the syringe needle. Frames H and I show that some liquid (h2 and i1) was again reverted into the upper part of the vaporizing chamber. The major portion of the liquid (h3 and i2) passed through the piece of fused silica (g2) and became visible as a band (i3) in the bend of the U-tube (g4). Only little fluorescence was visible in the glass wool (not visible in Figure 2), suggesting that the liquid entered the capillary rather than leaving through the space between the gooseneck and its outer wall as observed for the 0.25-mm-i.d. capillary. Videoframes H-L show strong fluorescence of the capillary section in the gooseneck. They do not enable us to distinguish whether this perylene was outside or inside the capillary. Therefore, the column was removed and its outer and inner walls were rinsed with dichloromethane. Both contained perylene. Geometrical Consideration. The bore of the restriction of the standard commercial gooseneck liner is 0.8 mm. With the 0.25-mm-i.d. capillary (0.36 mm o.d.) installed, this left some 0.22 mm of room between the liner and the capillary wall. Using the 0.53-mm-i.d. capillary (without polyimide), the gap between the column and the liner was hardly 0.1 mm. The inner diameter of the syringe needle (standard 26S gauge) is 0.13 mm. The visual experiment shows that the band of liquid has a width of ∼0.3 mm, which suggests that it consists of aligned droplets that cannot be resolved by the video because of the high speed. When the band is perfectly directed to the column entrance, it seems plausible that the liquid enters the 0.53-mm-i.d. capillary, but at best a fraction of it only should enter a column of 0.25-mm
Figure 3. Bottom region of injector with restriction part of gooseneck liner. Three scenarios for the band of liquid released by a fast autosampler: (1) the liquid hits the bottom of the liner and is rejected; (2) the funnel-shaped bottom guides the liquid through the restriction of the gooseneck, sometimes down to the column attachment; (3) the sample liquid directly enters the column.
i.d.. The rapid formation of vapors upon approaching the hot surfaces of the capillary is expected to repel the liquid, but the low thermal mass of the fused silica might enable a rapid cooling of the column wall, so that the solution could come into contact with it. Three Scenarios for the Injection into Empty Gooseneck Liners. The visual observations can be summarized by the three scenarios shown in Figure 3. In reality, often two or all three are combined. (1) The liquid is rejected into the upper part of the chamber. Small droplets (e.g., of exploded liquid) remain suspended in the gas until evaporated, whereas larger droplets fall back to the bottom.11 Solute material vaporized in the center of the chamber will be carried into the column by the carrier gas. Vapors formed at the bottom, just above the gooseneck, also reach the column, provided the column entrance is positioned within the orifice of the gooseneck, but they do not reach it if the entrance is higher. (2) The band of liquid leaves the chamber through the narrow space between the column inlet and the liner restriction. It reaches the metallic bottom part of the injector body and may proceed further down toward the ferrule of the column attachment. The consequences on quantitative analysis are shown below. (3) The band of liquid directly enters the column. This corresponds to an on-column injection: the sample enters the column as a whole, i.e., including the involatile material (contaminants). Chromatographic Performance. (a) Packed Gooseneck Liner. As known from previous visual experiments, a plug of glass wool keeps the liquid in place.13 Therefore, splitless injection into Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
1699
Figure 4. Ratio of peak area obtained by splitless (gooseneck liner) and on-column injection (RSL/OC): (A) test mixture of n-alkanes in hexane; (B) test mixture in dichloromethane (average values of six consecutive injections). With GW, with glass wool; w/o GW, without glass wool.
the gooseneck liner packed with a plug of glass wool was used as a reference to calibrate correct peak areas. The plug was ∼8 mm high, weighed some 10 mg, and was positioned 10 mm above the column entrance (the latter mounted some 5 mm above the bottom of the injector body). The absolute peak areas of the n-alkanes in the test mixture were compared to those obtained by on-column injection of the same volume. An area ratio RSL/OC ) 1 corresponds to the expectation in a perfect result. As shown in Figure 4, the packed (“with GW”) gooseneck liner with the column at position 5 mm (Figure 3) resulted in area ratios between 0.90 and 0.96 for the alkanes dissolved in hexane and between 0.82 and 0.88 for those in dichloromethane. The area ratios increased to 0.97-1.01 and 0.92-0.98, respectively, by applying a pressure pulse doubling the flow rate during the splitless period (not shown). The RSDs (n ) 6) of the absolute peak areas were 0.23-0.56 and 0.55-0.82% for the solutions in hexane and dichloromethane, respectively (Figure 5). (b) Empty Gooseneck Liner. The empty gooseneck liner (“w/o GW”) performed far worse than the packed liner. With the column entrance at the standard position (5 mm over the injector bottom), the area ratios for the hexane solution were between 0.31 and 0.84 (Figure 4); i.e., in splitless injection up to 69% of the solute material was lost. For the dichloromethane solution, area ratios were between 0.22 and 0.68; i.e., losses reached 78%. Losses increased with a higher solute boiling point, i.e., resulted in discrimination against the high-boiling compounds. They were poorly reproducible, as shown by the relative standard deviations increasing parallelly to the losses and reaching 6 and 11% for the hexane and the dichloromethane solution, respectively (Figure 5). A lower position of the column entrance improved the results. At the position “0 mm” (see Figure 3), losses became negligible 1700 Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
Figure 5. RSDs of absolute peak areas from six consecutive injections using a gooseneck liner: (A) test mixture in hexane; (B) test mixture in dichloromethane.
up to C15 but reached the same high percentage for the highboiling components. With the column installed at the orifice of the column ferrule attachment, 15 mm below the bottom of the chamber, the losses of the high-boiling compounds dropped to some 20%. This confirms that at least a large part of the losses observed with higher column positions are due to sample liquid shot down into the region of the column attachment. The higher losses of high-boiling components (discrimination) can be explained by the solutes returning back up to the column entrance when the latter is installed at a normal height: the extremely small volume of solute vapors must be carried by a flow of vapors generated by solvent evaporation near the column attachment. Solutes evaporating together with the solvent catch this carrier, but vapors of high-boiling compounds, formed after the solvent is fully evaporated, miss it. CONCLUSION Rules on how to perform splitless injection originate from the time before the fast autosampler was introduced and, as we know today, assume injection by thermospray using a hot syringe needle. The fast autosampler introduced injection with a band formation, shooting the liquid sample through empty liners to the bottom, which initiates processes largely out of control. The idea of using an empty, straight liner and having the column inserted 5 mm into the vaporizing chamber, positioned above the deposit of injector particles, must be therefore revisited. In split injection, the fast autosampler is commonly used with liners containing a plug of deactivated glass wool. This proved to be a perfect means to arrest the band of liquid, but for splitless injection, such packing frequently exhibits excessive adsorptivity and chemical activity (since far smaller amounts of solute are
injected, these effects are more apparent than in split injection). This favored the use of the unpacked gooseneck liner for trace analysis with splitless injection. The results of the present work cast doubt on empty gooseneck liners for splitless injection with fast autosamplers. The round shape at the bottom rejects part of the liquid toward the top of the chamber. Another part slips out of the liner through the bottom orifice, guided by the funnel-type shape of the gooseneck, i.e., passing outside of the column inlet into the column attachment area. Finally, some liquid is shot directly into the column when
reproducibility. This is exactly the problem the fast autosampler intended to eliminate! These findings merely specify a problem. The solutions may point into two directions: either alternative liners are designed, which enable stopping the shooting liquid without adverse effects on labile and high-boiling sample components, or thermospray injection should be preferred when the use of empty liners is a prerequisite. It is disturbing to think that gas chromatographers of today are routinely confronted with a technique for which such basic problems have not been solved.
its entrance is sufficiently wide. The chromatographic data showed
Received for review October 2, 2003. Accepted January 14, 2004.
drastic discrimination against the high boilers and poor result
AC035161A
Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
1701
