Microplasma Mass Spectrometric Detection in Capillary Gas

University of Oslo, P.O. Box 1068 Blindern, 0316 Oslo, Norway. A simple and miniaturized 350-kHz helium discharge for plasma mass spectrometric detect...
0 downloads 0 Views 105KB Size
Anal. Chem. 1998, 70, 513-518

Microplasma Mass Spectrometric Detection in Capillary Gas Chromatography Cato Brede,*,† Stig Pedersen-Bjergaard,‡ Elsa Lundanes,† and Tyge Greibrokk†

Department of Chemistry, University of Oslo, P.O. Box 1033 Blindern, 0315 Oslo, Norway, and School of Pharmacy, University of Oslo, P.O. Box 1068 Blindern, 0316 Oslo, Norway

A simple and miniaturized 350-kHz helium discharge for plasma mass spectrometric detection in gas chromatography (GC) has been developed. The plasma was sustained at low pressure within the end of the capillary GC column (0.32-mm i.d.) inside the ion source housing of a quadrupole mass spectrometer. This allowed direct introduction of ions from the plasma to the mass analyzer using only a repeller and electrostatic lenses to focus the ions. The plasma was sustained in only 25 mL min-1 of helium, which was accepted by the mass spectrometer vacuum system. This low gas flow also served to enhance the energy density of the discharge and to produce a narrow spray of ions toward the mass analyzer. Due to the miniaturized nature of the plasma, it was operated at a low power level (2.0 W), and traces of oxygen were added to avoid deposition of carbon. With this new concept for GC plasma mass spectrometric detection, chlorine was successfully monitored down to the 2.2 pg s-1 level without interference from elements like C, S, P, O, F, and N. Element-selective detection in capillary gas chromatography (GC) can be performed with several types of detectors, including modified flame ionization detectors (i.e., O-FID), the nitrogen phosphorus detector (NPD), the flame photometric detector (FPD), the sulfur chemiluminescence detector (SCD), and the Hall electrolytic conductivity detector (HECD). For most applications, these classical detectors offer acceptable sensitivity and selectivity at a relatively low price and, therefore, are the most popular. Nevertheless, each detector responds only to one or a few elements, which obviously limits their versatility. Atomic emission detectors (AED), on the other hand, are able to simultaneously detect a wide variety of elements and may, therefore, replace several of the classical element-selective GC detectors mentioned above. With the commercially available instrument,1,2 which applies a helium microwave-induced plasma (MIP) for atomization and excitation3,4 and a photodiode array for detection, even hard-to-excite atoms like F and Cl can be detected at low-picogram levels. However, the cost of this detector, which is similar to that of conventional quadrupole mass * To whom correspondence should be addressed. Fax: +47 22855441. (1) Quimby, B. D.; Sullivan, J. J. Anal. Chem. 1990, 62, 1027-1034. (2) Sullivan, J. J.; Quimby, B. D. Anal. Chem. 1990, 62, 1034-1043. (3) Bulska, E. J. Anal. At. Spectrom. 1992, 7, 201-210. (4) de Wit, A.; Beens, J. J. Chromatogr. Libr. Ser. 1995, 56, 159-200. S0003-2700(97)00259-X CCC: $15.00 Published on Web 01/01/1998

© 1998 American Chemical Society

spectrometers (MS), seems to limit the propagation of the GC/ AED technique. In addition to AED, multielement-selective detection can be achieved by plasma mass spectrometry.5 Although inductively coupled plasma mass spectrometry6 (ICP-MS) is commercially available and combines the outstanding atomization and ionization properties of ICP with the high sensitivity of mass spectrometry, this instrument is even more expensive than the AED. Also, ICP-MS is an expensive chromatographic detector because it consumes more than 500 times as much gas as the GC. In order to reduce operational costs, the flow of plasma gas has to be reduced using discharges of smaller dimensions. This was realized with the Beenaker cavity which allowed microwaveinduced plasmas (MIPs) to be operated at flow rates down to 50 mL min-1. With atmospheric pressure7-9 and low pressure7,10,11 MIPs as ionization sources, chlorinated compounds have been detected at the low-picogram level following gas chromatography. Recently, also radio frequency glow discharge mass spectrometry (RF-GD-MS) has been used for element-selective GC detection of some tetraalkyltins12 at a plasma gas (He) flow rate of 0.6 L min-1. Both GC/MIP-MS and GC/RF-GD-MS utilize ion extraction similar to that used in ICP-MS; atomic ions formed in the plasma outside the mass spectrometer are extracted through two narrow cones into the vacuum region.13-15 Unfortunately, during this process, most of the ions are lost, and less than 1% of the ions are focused into the mass analyzer.16 In addition, with both GC/ MIP-MS and GC/RF-GD-MS, which utilize external ionization sources, expensive and dedicated mass spectrometers have to be (5) Vela, N. P.; Olson, L. K.; Caruso, J. A. Anal. Chem. 1993, 65, 585-597. (6) Jarvis, K. E.; Gray, A. L.; Houk, R. S. Handbook of ICP-MS; Blackie Academic & Professional, Blackie & Son Ltd.: London, UK, 1992. (7) Creed, J. T.; Mohamad, A. H.; Davidson, T. M.; Ataman, G.; Caruso, J. A. J. Anal. At. Spectrom. 1988, 3, 923-926. (8) Mohamad, A. H.; Creed, J. T.; Davidson, T. M.; Caruso, J. A. Appl. Spectrosc. 1989, 43, 1127-1131. (9) Suyani, H.; Creed, J.; Caruso, J.; Satzger, R. D. J. Anal. At. Spectrom. 1988, 4, 777-782. (10) Creed, J. T.; Davidson, T. M.; Shen, W.-L.; Caruso, J. A. J. Anal. At. Spectrom. 1990, 5, 109-113. (11) Story, W. C.; Olson, L. K.; Shen, W.-L.; Creed, J.; Caruso, J. A. J. Anal. At. Spectrom. 1990, 5, 467-470. (12) Olson, L. K.; Belkin, M.; Caruso, J. A. J. Anal. At. Spectrom. 1996, 11, 491496. (13) Campargue, R. J. Phys. Chem. 1984, 88, 4466-4474. (14) Olivares, J. A.; Houk, R. S. Anal. Chem. 1985, 57, 2674-2679. (15) Douglas, D. J.; French, J. B. J. Anal. At. Spectrom. 1988, 3, 743-747. (16) Houk, R. S.; Shum, S. C. K.; Wiederin, D. R. Anal. Chim. Acta 1991, 250, 61-70.

Analytical Chemistry, Vol. 70, No. 3, February 1, 1998 513

Figure 1. Diagram of GC/MP-MS setup with (A) gas chromatograph, (B) tee for mixing with oxygen-doped external helium flow, (C) heated transfer line, (D) union, (E) 1/16-in.-o.d., 0.5-mm-i.d. grounded steel tubing containing the capillary column, (F) steel rod carrying the rf voltage from the generator, (G) 50-mm-diameter transparent acrylic disk with Teflon plugs, (H) rf electrode containing the capillary column, (I) second grounded electrode containing the capillary column, (J) repeller, (K) ion lenses, and (L) quadrupole mass spectrometer. (Not to scale).

used. Thus, neither technique may be carried out with commercial and low-cost bench-top mass spectrometers. In the present work, microplasma mass spectrometry (MP-MS)17 has been developed as a new concept for elementselective detection in capillary gas chromatography. The MP-MS system was based on a 350-kHz radio frequency helium plasma which has previously been used for GC/AED18-20 and further miniaturized in our laboratory.21,22 The plasma was sustained at low pressure within the ion source housing of a conventional quadrupole mass spectrometer. In order to be compatible with the internal dimensions and the vacuum considerations of mass spectrometric instrumentation, the plasma was miniaturized and sustained in 25 mL min-1 of helium inside the end of the fused silica GC column (0.32-mm i.d.). Compared with ICP-MS and MIP-MS, microplasma mass spectrometry resulted in very low operational costs. In addition, high sensitivity was obtained for chlorine-selective detection. In the present work, attention is focused on instrumental development of MP-MS and optimization for chlorine-selective detection. EXPERIMENTAL SECTION Gas Chromatography. The gas chromatograph, the microplasma mass spectrometer, and the interface are illustrated in Figure 1. A HP model 5730A GC (Hewlett-Packard, Wilmington, DE) with a 18740B capillary column control unit was used for the chromatography (Figure 1, part A). All separations were (17) Brede, C.; Pedersen-Bjergaard, S.; Lundanes, E.; Greibrokk, T. Norway Patent Application 970707, 1997. (18) Skelton, R. J., Jr.; Chang, H.-C. K.; Farnsworth, P. B.; Markides, K. E.; Lee, M. L. Anal. Chem. 1989, 61, 2292-2298. (19) Skelton, R. J., Jr.; Markides, K. E.; Lee, M. L.; Farnsworth, P. B. Appl. Spectrosc. 1990, 44, 853-857. (20) Skelton, R. J., Jr.; Markides, K. E.; Farnsworth, P. B.; Lee, M. L.; Yang, F. J. J. High Resolut. Chromatogr. Chromatogr. Commun. 1988, 11, 75-81. (21) Pedersen-Bjergaard, S.; Greibrokk, T. Anal. Chem. 1993, 65, 1998-2002. (22) Pedersen-Bjergaard, S.; Greibrokk, T. J. Microcolumn. Sep. 1994, 6, 1118.

514 Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

performed on a 25-m × 0.25-mm-i.d. CP-Sil 8 CB-MS fused silica capillary column (Chrompack, Middelburg, The Netherlands) coated with 0.25 µm of 5% phenyl and 95% dimethylpolysiloxane. Samples of 1.0 µL were injected splitless with 45-s split delay at an injector temperature of 250 °C. The carrier gas was 99.9999% He (Aga, Oslo, Norway) at a flow rate of 1-2 mL min-1. Inside the GC oven, the end of the GC column was connected to a 1/ -in. tee (VICI, Houston, TX), where the GC effluent was mixed 16 with 20-30 mL min-1 of 99.9999% He, which served as plasma gas (Figure 1, part B). The plasma gas flow was measured by a model 820 Top ) Trak flow monitor (Sierra Instruments, Monterey, CA) and controlled using a VCD-1000 flow controller (Porter Instrument Co., Hatfield, PA). Traces of 99.999% oxygen (Aga) were added as scavenger gas to the plasma gas through a 2.15-m × 20-µm-i.d. fused silica restrictor (Polymicro Technologies, Phoenix, AZ). The flow of oxygen was measured as the inlet pressure (0-7 bar) prior to the restrictor. A 1.0-m × 0.32-mmi.d. HP1 fused silica capillary column (Hewlett-Packard) coated with 0.17 µm of dimethylpolysiloxane was used as transfer tubing from the tee at the column outlet to the MS. This capillary was contained in a laboratory-built heated transfer line (Figure 1, part C), which was made of 0.5-mm-i.d. and 1/16-in.-o.d. steel tubing and heated to 300 °C with a flexible heating tape (Cole Parmer, Niles, IL) using a model 020-086-06-026-13-00 temperature controller (Eurotherm, Reston, VA). Mass Spectrometry. The mass spectrometer (Figure 1, part L) was a modified model 201 Vestec dedicated thermospray LC/ MS (Vestec Corp., Houston, TX). This system consisted of a quadrupole mass analyzer (Hewlett-Packard), a model 342 channeltron electron multiplier (Detector Technology, Sturbridge, MA), control electronics from the Hewlett-Packard model 5970 MSD, and a Hewlett-Packard 59970C ChemStation. The mass range was 3-800 amu, with both positive and negative ion detection included. In this work, positive ion detection was used unless otherwise noted.

The mass spectrometer was modified in several ways prior to plasma detection. While the original system was operated with one mechanical pump, each of the two oil diffusion pumps was backed separately in this work by double-stage rotary vacuum pumps. The 150 L s-1 analyzer diffusion pump was supported by an EDM12, and the 300 L s-1 ion source diffusion pump was supported by an EDM20 (Edwards, West Sussex, UK). At 25 mL min-1 of helium to the plasma, the pressure at the analyzer Penning gauge was kept below 4 × 10-6 Torr. The vapor pumping outlet of the ion source was plugged, and the thermospray probe, tip heater, sampling cone, and discharge electrodes were removed. This left the bare ion source with an open 10-mm channel in the direction of the probe entrance. The microplasma developed in the present work was constructed as a probe and placed directly on the probe entrance, originally intended for a thermospray interface, using a rubber 19- × 4-mm O-ring (Per-Kr. Askim, Oslo, Norway) to seal off the vacuum. Thus, the plasma was sustained under low-pressure conditions inside the ion source of the mass spectrometer. Plasma Probe. The microplasma was sustained on the last 3.5 cm of the 0.32-mm-i.d. fused silica capillary GC column, where the stationary phase and polyimide coating were removed. The plasma part of the GC column was located in a small plasma probe (Figure 1, parts H and I). The plasma probe was mounted on a transparent acrylic disk of 50-mm diameter and 10-mm thickness, which fitted the MS entrance and allowed easy handling and service of the microplasma (Figure 1, part G). The transparent character of the disk enabled visual inspection of the plasma inside the mass spectrometer. In addition, the isolating properties of the disk prevented leakage of the high voltages used for plasma operation. Two 4-mm holes were drilled in the middle of the disk 8 mm apart, center to center. Two Teflon plugs were fitted to the holes, and a 10-cm × 0.5-mm-i.d. (1/16-in. o.d.) steel tubing was placed in one of them. This tubing was electrically grounded and served to guide the fused silica GC capillary column into the mass spectrometer (Figure 1, part E). A 1/16-in. ZU1T union (VICI) was attached to the GC capillary by a graphite ferrule and to the steel tubing by a Vespel ferrule (Figure 1, part D). The union was placed directly at the outlet of the transfer line and wrapped in heating tape in order to avoid cold spots and potential analyte condensation. A three-electrode system was connected to the GC capillary column situated inside the mass spectrometer. At both ends of the plasma region, the uncoated GC capillary was surrounded by small grounded steel electrodes, while a highvoltage radio frequency electrode was placed approximately at the middle of the discharge (Figure 1, parts E, H, and I). The rf power, which was supplied by an HPG-2 power supply (ENI, Rochester, NY), was introduced through a 10-cm × 1.6-mm-o.d. steel rod inserted in the second Teflon plug of the disk (Figure 1, part F). With this arrangement, plasma conditions were obtained in the entire 5-cm region between the two grounded electrodes. The first part of the plasma served to effectively heat the sample, while atomic ions were collected from the final part. Since the latter (15 mm) was made slightly shorter than the first (16 mm), a brighter and more energetic plasma was produced to stimulate atomic ionization. In order to ensure electrical isolation and to improve the mechanical stability of the plasma probe, two pieces of 1.6-mm-i.d., 6.0-mm-o.d. quartz tubes (Heigar, Oslo,

Norway) were inserted between the three electrodes. The internal diameter of the quartz tubes exactly fit the small electrodes, whose lengths were 28 and 24 mm, respectively. With these quartz tubes, the probe tip ended only 5 mm from the axis of the ion lenses and the quadrupole analyzer. This was done in order to ensure an efficient transfer of ions from the capillary outlet to the mass analyzer by the voltages set on the repeller and the lenses. The focusing of atomic ions was performed using the repeller and electrostatic lenses located near the thermospray ion source block (Figure 1, parts J and K). After departure from the plasma, the ions were drawn into the lens stack, which was located in a plane 90° to the plasma probe. An advantage of the 90° configuration was the elimination of photons in the direction of the quadrupole and detector. This, in combination with the off-axis electron multiplier detector, completely eliminated the background problem of photon detection. Before insertion of the plasma probe, the fused silica was pulled through the electrodes for removal of the polyimide coating by burning and wiping. Finally, the capillary was withdrawn and positioned with its end 0.5 mm behind the front electrode tip. This procedure was repeated when required, usually once a week or when the capillary was damaged by the plasma. Chemicals and Sample Preparation. Analytical grade 1,2dichlorobenzene, dibenzothiophene, dibenzofuran, and 1-fluoronaphthalene were all purchased from Aldrich (Steinheim, Germany). To prove the utility of the technique, a commercial sparkling white wine was selected as a sample. A 100-mL sample of wine was extracted with 3 × 25 mL of cyclohexane (Rathburn, Walkerburn, UK), followed by evaporation to dryness and resolution in 20 µL of cyclohexane. MS Tuning. For MS tuning and plasma optimization, a constant amount of chlorine was introduced to the plasma by a small flow of chlorodifluoromethane (Hydro, Oslo, Norway) added through the GC injector using a 5-µm-i.d. fused silica capillary. Estimation of Selectivity, Detection Limits, and Linearity. Selectivity, detection limits, and linearity were determined using 1-µL manual splitless injections of 1,2-dichlorobenzene solved in cyclohexane. Selectivity of chlorine to other elements was defined as the ratio of response per mole of chlorine to the response per mole of another element.1 Detection limits were defined as the amount of element required to produce a signal 2 times the noise level divided by the peak width at half-height.1 The linear range was defined as the concentration range where the response factor (area per unit mass) deviated by less than 20%.1 RESULTS AND DISCUSSION Operational Considerations. During the initial experiments, the plasma was sustained in 2.5 mL min-1 of helium GC carrier gas. Although this plasma was easily ignited, an additional flow of helium was added in order to effectively promote analyte ionization and plasma stabilization. At least 24 mL min-1 of helium was required to obtain a stable plasma. Flow rates in the range 24-40 mL min-1 were accepted by the mass spectrometer, owing to the enhanced pumping capacity, as discussed earlier. With a neat helium plasma, a very low analyte loadability was observed, owing to the formation of carbon deposits from eluting material. This problem was effectively suppressed by the introduction of traces of oxygen to the plasma. Nevertheless, a substantial amount of carbon from the solvent easily overloaded Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

515

Figure 2. Effect of power on chlorine ion abundance at a total helium flow of 25 mL min-1 with traces of oxygen.

the plasma, and consequently the plasma generator was turned off during solvent elution. The plasma was easily ignited using the ignite function of the generator, which produced a short pulse of power. Optimization of Applied Power and Capillary Position for Chlorine Detection. Figure 2 illustrates the effect of applied power on the chlorine abundance measured at m/z 35. Maximum signal was obtained at 1.9 W. This is an extremely small power consumption compared to 50-100 W used in low-pressure He MIP-MS,5,10 30 W used in He GC/GD-MS,12 or 20 W used in oncolumn He GC/AED.21,22 Extinction of the first plasma, combined with a large decrease in signal, occurred at a power level slightly below 1.5 W, while the fused silica capillary was damaged above 2.5 W. In this work, the total length of the probe was kept constant. However, the length of the fused silica capillary column inside the probe was varied. Initially, the column outlet was aligned with the front electrode outlet. It was possible to pull the column back into the grounded front electrode with the probe inserted and without evacuation of the MS. The optimum chlorine signal was found when the column ended approximately 0.6 mm inside the front electrode. This observation probably arose from an extended contact between the plasma gas and the grounded metal electrode. This enhanced the energy transfer to the plasma region close to the probe outlet, and more chlorine ions were produced. Upon pulling the capillary farther back from the probe tip, the signal decreased again. This was probably due to ions colliding with the electrode walls, either because of space charge effects or simply because of the diverging nature of the plasma spray. Optimization of Helium Flow and Reagent Gas for Chlorine Detection. As discussed above, both helium and oxygen were added to the plasma in order to promote analyte ionization. To obtain maximum sensitivity for chlorine-selective detection, the combined effect of helium and oxygen flow on the chlorine signal (m/z 35) was studied (Figure 3). Maximum signal was found at approximately 25 mL min-1 of He, with an oxygen level set by 2.25 bar (see Experimental Section). Below 24 mL min-1 of He, a fast decline in signal was observed. At higher flow rates, more oxygen could be added to reach local vertices, but higher levels of oxygen partly quenched the plasma. An oxygen-deficient plasma resulted in a decreased signal, which is opposite to observations in GC/AED.22 One other important consideration of oxygen doping was the measured effect on background signal 516 Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

Figure 3. Effect of helium flow and amount of oxygen added on chlorine ion abundance. Plasma power was 2.0 W.

and noise (deviation of background signal). Both decreased with an increasing oxygen level, and the optimum S/N ratio for chlorine was found at the same oxygen level that gave the highest chlorine signal. The advantages of using a higher oxygen level than optimum were the tolerance of larger amounts of carbon and quicker plasma recoveries in case of overload. This was experienced during chromatography of real samples, which required an oxygen level set by 3.0 bar at a He flow of 25 mL min-1. Background Spectra in Positive Mode. Since selectivity depends on well-separated mass peaks and the absence of interfering species, it is a necessity that the background spectrum contain as few mass peaks as possible. Figure 4 shows the background mass spectrum of an oxygen-doped He plasma. Signals were generated at m/z 4 (He+), 5 (HeH+), 8 (He2+), 14 (N+ or 28Si2+), 16 (O+), 17 (OH+), 18 (OH2+ or HeN+), 28 (28Si+, N2+ or CO+), 30 (NO+), 32 (O2+ or HeSi+), 40 (Ar+), 44 (HeAr+ or CO2+), and 45 (HeArH+ or CO2H+). The presence of these fragments was in accordance with results previously reported for low-pressure He MIP-MS.10 Metal and metal oxide ions originating from sampler and skimmer are usually found in background spectra of both ICP-MS and MIP-MS. In this work, the plasma was kept in a fused silica capillary and just briefly touched the outlet tip steel electrode. Thus, metal and metal oxide ions were absent in the present microplasma, which produced a simpler background spectrum as compared with MIP-MS and ICP-MS. After some time of He purging through the GC and gas supply lines, the peaks at m/z 14 and 30 were barely visible in the spectrum. This indicated a decreased nitrogen contamination from air. Therefore, air leakage through fittings, GC septum, and venting lines was avoided by keeping the system pressurized also when not in operation. Figure 5 shows the spectrum of the oxygen-doped helium plasma with a small amount of chlorodifluoromethane added. The typical isotope pattern of Cl+ was observed at m/z 35 and 37. The relatively large amount of HCl+ at m/z 36 and 38 was probably due to reaction with hydrogen from chlorodifluoromethane. When hydrogen was used as a scavenger gas instead of oxygen, the abundance of HCl+ increased at the expense of Cl+. This indicated a potential advantage of using hydrogen as a scavenger gas, which is currently under further investigation.

Figure 4. Mass spectrum of oxygen-doped helium plasma. Helium flow of 25 mL min-1 and 2.0 bar of O2 were used.

Figure 5. Mass spectrum of oxygen-doped helium plasma with a small amount of chlorodifluoromethane. Helium flow of 25 mL min-1 and 2.0 bar of O2 were used. Table 1. Selectivity of Chlorine to Other Elements elementa

selectivityb

S P C O F N

1.5 × 103 2.4 × 103 8.9 × 104 >6.6 × 103 3.5 × 103 2.3 × 104

a The following compounds were used: dibenzothiophene (S), triethyl phosphate (P), naphthalene (C), dibenzofuran (O), 1-fluoronaphthalene (F) and pyridazine (N). b Defined as the ratio of response per mole of chlorine to the response per mole of another element.

Background Spectra in Negative Mode. Negative ion detection was only briefly investigated. Using neat He, only one peak at m/z 16 (O-) was observed in the background spectrum. Such a simple spectrum is in accordance with similar reports of negative ion detection in ICP-MS,23 and represents an interesting approach for future detection of the most electronegative elements. The negative ion mass spectrum of an oxygen-doped helium plasma with a small amount of chlorodifluoromethane was very simple and contained peaks only at m/z 16 (O-), 19 (F-), and 35/37 (Cl-). (23) Vickers, G. H.; Wilson, D. A.; Hieftje, G. M. Anal. Chem. 1988, 60, 18081812.

Selectivity, Detection Limits, and Linearity for Chlorine Detection. The most important feature of this detection system was the high chlorine selectivity. The possible interferences at m/z 35 caused by polyatomic species of other atoms have been discussed for He GC/MIP-MS,7 although no numbers on selectivity were mentioned. In the present work, compounds (above the 100-ng level of each) containing C, S, P, O, F, and N were injected while observing the response on the chlorine channel at m/z 35 (Table 1). A selectivity above 103 was found for all of the elements introduced, with the highest selectivity found relative to carbon. A large amount of carbon (200 ng) produced a slightly negative peak on the chlorine trace, which indicated temporary removal of a background species by reaction with carbon or quenching of the plasma. A somewhat lower amount of carbon (100 ng) did not produce an observable peak. Slightly positive peaks were found on the chlorine chromatograms of the compounds containing P (19 ng), F (21 ng), N (41 ng), and S (17 ng). This was probably due to formation of low-abundant species at m/z 35. For oxygen, no signal was observed for the amount of compound introduced (10 ng). Therefore, the selectivity of chlorine to oxygen was higher than the value estimated from this amount. The high selectivity toward oxygen probably arose from the fact that oxygen already was present at a relatively high level as scavenger gas. Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

517

the extract. However, the selectivity for chlorine was excellent, as evidenced by the Cl trace. Three chlorinated compounds were detected at the 5-30-pg level (corresponding to 1-6 pg mL-1 of wine). The identities of these are currently unknown, but GC/ MS work is in progress for compound characterization.

Figure 6. (a) C- and (b) Cl-selective chromatograms of an extract from a sparkling white wine recorded in SIM mode. One microliter was injected splitless at an injector temperature of 250 °C. The oven temperature was held at 50 °C for 4 min and then programmed at 10 °Cmin-1 to 200 °C.

For chlorine, the detection limit was 2.2 pg s-1. This is comparable to the lowest values obtained with atomic emission detection (1.1 pg of Cl s-1 at S/N ) 2)21 and microwave-induced plasma mass spectrometric detection (2 pg of Cl s-1 based on 3σ).8 The chlorine response was linear over at least 3 orders of magnitude, which is also comparable to the works mentioned above. The repeatability of the signal (RSD ) 5.6% for four injections) was limited by the manual injection technique when all other parameters were kept constant. The detection limit for chlorine in negative mode was estimated to be 5.5 pg s-1, which is somewhat higher than that obtained in positive mode. Oxygen was used as a scavenger gas at the conditions that gave the lowest detection limit in positive mode. Further work is in progress to optimize Cl-selective detection in the negative mode. Applications. To demonstrate the applicability and selectivity of the system, a cyclohexane extract of sparkling white wine was analyzed (Figure 6). Large amounts of carbon were detected in

518 Analytical Chemistry, Vol. 70, No. 3, February 1, 1998

CONCLUSIONS In this work, microplasma mass spectrometry (MP-MS) has been developed as a new concept for element-selective detection in capillary gas chromatography. The MP-MS system was based on a low-pressure 350-kHz helium plasma sustained within the vacuum region of the mass spectrometer (MS). In order to be compatible with the internal dimensions and the vacuum considerations of MS instrumentation, the plasma was miniaturized and sustained inside the end of a fused silica GC column (0.32-mm i.d.). Owing to the small dimensions of the discharge, the plasma was operated at low power levels (2 W) and required only 25 mL min-1 of helium for the purpose of stabilization. In contrast to the existing techniques of inductively coupled plasma mass spectrometry (ICP-MS) and microwave-induced plasma mass spectrometry (MIP-MS), the MP-MS concept emerged as a highly attractive detector for capillary gas chromatography. Owing to the simplicity and the miniaturized nature of the plasma, the instrumental costs were relatively low. In addition, the operational costs were marginal, owing to the low consumption of helium (25 mL min-1). With the MP-MS system, chlorine was selectively detected down to 2.2 pg s-1 practically without interference from C, S, P, O, F, and N. Work is currently in progress to further optimize the MP-MS system. The efforts are principally focused on reduction of the flow of external helium in order to further improve the compatibility of microplasma ionization with small mass spectrometers. Hopefully, plasma ionization may be supplied as a low-cost option to conventional bench-top mass spectrometers within the near future, and the application field of these instruments will be expanded even to multielement-selective detection. ACKNOWLEDGMENT The authors thank Dr. Einar Solheim at the University of Bergen, Norway, for the donation of the Vestec LC-MS and for helpful advice regarding this instrument. We thank Mr. Kjell Arne Sulutvedt at Hydro, Norway, for the donation of the GC. Received for review March 6, 1997. Accepted November 11, 1997. AC9702599