Separation and Detection of Explosives on a Microchip Using Micellar

Mar 11, 2000 - Combinatorial and Evolution-Based Methods in the Creation of Enantioselective Catalysts. Manfred T. Reetz. Angewandte Chemie Internatio...
1 downloads 12 Views 134KB Size
Download PDF
Anal. Chem. 2000, 72, 1872-1878

Separation and Detection of Explosives on a Microchip Using Micellar Electrokinetic Chromatography and Indirect Laser-Induced Fluorescence Susanne R. Wallenborg and Christopher G. Bailey*

Sandia National Laboratories, P.O. Box 969/ms9671, Livermore, California 94551

A new approach for sensitive detection on a microfabricated chip is presented. Indirect laser-induced-fluorescence (IDLIF) was used to detect explosive compounds after separation by micellar electrokinetic chromatography (MEKC). The detection setup was used in an epifluorescence configuration with excitation provided by a nearIR diode laser operating at 750 nm. To achieve indirect detection, a low concentration of a dye (5 µM Cy7) was added to the running buffer as a visualizing agent. Using this methodology, a sample containing 14 explosives (EPA 8330 mixture) was examined. Concentrations of 1 ppm of trinitrobenzene (TNB), trinitrotoluene (TNT), dinitrobenzene (DNB), tetryl, and 2,4-dinitrotoluene (2,4DNT) could be detected with S/N ratios between 3 and 10. Analyses showing 10 peaks, with plate numbers on the order of 60 000, were completed within 60 s using a 65 mm long separation channel. The three isomers of nitrotoluene (2-, 3-, and 4-nitrotoluene) were not resolved. Additionally, the two nitramines (HMX and RDX) could only be detected at much higher concentrations, likely due to the low fluorescence quenching efficiencies of these compounds. The analysis method was also used to separate and detect nitroaromatic compounds in extracts from spiked soil samples. The presence of 1 ppm (1 µg of analyte/1 g of soil) of TNB, DNB, TNT, tetryl, 2,4-DNT, 2,6-DNT, 2-NH2-4,6-DNT, and 4-NH2-2,6-DNT could readily be detected. In the interest of increasing the sensitivity of the analysis, various on-chip injection schemes were evaluated. It was found that a 250 µm double-T injector gave a 35% increase in peak signal compared to a straight-cross injector, which is less than expected based on injected volume. Over the last century, large quantities of ammunition and explosive compounds have been buried at military bases and other sites throughout the world. As part of the remediation of these sites, there is a need for fast and sensitive methods to screen large numbers of environmental samples for the presence of explosive compounds. Separation of explosives has traditionally been performed using liquid chromatography (LC) with ultraviolet (UV) absorption1,2 or electrochemical detection.3,4 Recently, capillary * Corresponding author. Phone: (925) 294-1351. E-mail: [email protected].

1872 Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

electrokinetic techniques, such as MEKC5-11 and capillary electrochromatography (CEC)12 have proven useful for the separation of explosives. The separation method specified by the U.S. Environmental Protection Agency (EPA) is based on two separate LC separations employing columns with octadodecyl (C18) and cyano (CN) functionalities. Typically, the analysis time for a mixture of 14 nitroaromatics and nitramines (EPA 8330 mixture) using only the C18 column is 30-40 min.1 After this, a confirmatory analysis employing the CN column must be performed. Separation of the EPA 8330 mixture and other nitroaromatic compounds has been achieved within 15 min, using MEKC and CEC.5-12 In the CEC study,12 a separation of the EPA mixture was demonstrated in 2 min using a 12 cm packed capillary column. However, due to the short length of column protruding from the detector, the manipulation required for sample introduction was awkward. To take full advantage of the available efficiency and speed of miniaturized separation systems the injection and detection procedures must be compatible with short separation lengths. This issue is elegantly addressed in microfabricated systems (microchips), as the fluid flow paths can easily be manipulated and controlled by voltage switching rather than by column manipu(1) Weisberg, C. A.; Ellickson, M. L. Am. Lab. 1998, 32N-32V. (2) Ko ¨hne, A. P.; Dornberger, U.; Welsch, T. Chromatographia 1998, 48, 9-16. (3) Bratin, K.; Kissinger, P. T.; Briner, R. C.; Bruntlett, C. S. Anal. Chim. Acta 1981, 130, 295-311. (4) Lewin, U.; Efer, J.; Engewald, W. J. Chromatogr., A 1996, 730, 161-167. (5) Wallenborg, S. R.; Bailey, C. G. Electrophoresis, submitted for publication. (6) Application Update 136; Dionex Corp., Sunnyvale, CA. (7) Luong, J. H. T.; Guo, Y. J. Chromatogr., A 1998, 811, 225-232. (8) Hilmi, A.; Luong, J. H. T.; Nguyen, A.-L. Anal. Chem. 1999, 71, 873-878. (9) Oehrle, S. A. Electrophoresis 1997, 18, 300-302. (10) Oehrle, S. A. J. Chromatogr., A 1996, 745, 233-237. (11) Yik, Y. F.; Lee, H. K.; Li, S. F. Y. J. High Resolut. Chromatogr. 1992, 15, 198-200. (12) Bailey, C. G.; Yan, C. Anal. Chem. 1998, 70, 3275-3279. (13) Shultz-Lockyear, L. L.; Colyer, C. L.; Fan, Z. H.; Roy, K. I.; Harrison, D. J. Electrophoresis 1999, 20, 529-538. (14) Haswell, S. J. Analyst 1997, 122, 1R-10R. (15) Seiler, K.; Fan, Z. H.; Fluri, K.; Harrison, D. J. Anal. Chem. 1994, 66, 34853491. (16) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113. (17) Kutter, J. P.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1997, 69, 51655171. (18) von Heeren, F.; Verpoorte, E.; Manz, A.; Thormann, W. Anal. Chem. 1996, 68, 2044-2053. (19) Jacobson, S. C.; Ermakov, S. V.; Ramsey, J. M. Anal. Chem. 1999, 71, 3273. 10.1021/ac991382y CCC: $19.00

© 2000 American Chemical Society Published on Web 03/11/2000

lation.13-20 A variety of electrokinetic techniques have been implemented on microchip devices including capillary electrophoresis (CE),16,21-23 capillary gel electrophoresis,18,24,25 MEKC,17,18,26 and open-channel electrochromatography.27,28 However, only the last two methods can be used to separate uncharged species. While open-channel electrochromatography requires coating of the microfabricated channels, MEKC is performed by adding a surfactant in excess of the critical micelle concentration (cmc) to the running buffer.29 The micelles act as a pseudostationary phase, which separates analytes on the basis of their affinity for the hydrophobic interior of the micelle. However, addition of the surfactant to the running buffer increases the separation current, which can result in excessive Joule heating, leading to irreproducible migration times and analyte band broadening.30 A microfabricated channel in a glass substrate has more effective heat dissipation than a fused-silica capillary and is therefore less sensitive to high separation currents.21,23 Another important attribute of microfabricated systems is their ability to perform fast separations.25,31,32 This was recently demonstrated by Jacobson et al.,32 who showed the separation of two oppositely charged dyes in less than 1 ms using chip-based CE. Most separation methods used to determine explosives have employed UV absorbance detection at a wavelength of 254 nm.1,2,6,7,9-12 An inherent problem with UV detection is the linear dependence of absorbance on optical path length. To address this, many commercial CE instruments use a UV absorbance detection cell with an increased path length (e.g., a bubble cell or a Z- or U-cell). In a chip configuration, this is not as easily achieved, although Liang et al.33 have described the use of a microfabricated U-cell for integrated absorbance and fluorescence detection. An additional difficulty with UV absorbance detection in chip-based separations is that borosilicate glass absorbs light with wavelengths shorter than 380 nm. Fused-silica substrates can be used to overcome this problem but have not been generally investigated due to the higher cost. The most common detection method for chip-based separations to date has been laser-induced fluorescence (LIF), since the sensitivity does not suffer with reduced path length. However, LIF detection requires that the analyte possess a fluorescence band that can be excited at a wavelength provided (20) Harrison, D. J.; Fluri, K.; Seiler, K.; Fan, Z.; Effenhauser, C. S.; Manz, A. Science 1993, 261, 895-897. (21) Fan, Z. H.; Harrison, D. J. Anal. Chem. 1994, 66, 177-184. (22) Harrison, D. J.; Manz, A.; Fan, Z.; Lu ¨ di, H.; Widmer, H. M. Anal. Chem. 1992, 64, 1926-1932. (23) Seiler, K.; Harrison, D. J.; Manz, A. Anal. Chem. 1993, 65, 1481-1488. (24) Yao, S.; Anex, D. S.; Caldwell, W. B.; Arnold, D. W.; Smith, K. B.; Schultz, P. G. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 5372-5377. (25) Liu, S.; Shi, Y.; Ja, W. W.; Mathies, R. A. Anal. Chem. 1999, 71, 566-573. (26) Moore, A. W.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1995, 67, 41844189. (27) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 2369-2373. (28) Kutter, J. P.; Jacobson, S. C.; Matsubara, N.; Ramsey, J. M. Anal. Chem. 1998, 70, 3291-3297. (29) Khaledi, M. G. Handbook of Capillary Electrophoresis; CRC Press: Boca Raton, FL, 1994; Chapter 3. (30) Nelson, R. J.; Burgi, D. S. Handbook of Capillary Electrophoresis; CRC Press: Boca Raton, FL, 1994; Chapter 21. (31) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 1114-1118. (32) Jacobson, S. C.; Culbertson, C. T.; Daler, J. E.; Ramsey, J. M. Anal. Chem. 1998, 70, 3476-3480. (33) Liang, Z.; Chiem, N.; Ocvirk, G.; Tang, T.; Fluri, K.; Harrison, D. J. Anal. Chem. 1996, 68, 1040-1046.

by a laser source. Alternatively, the analyte can be tagged with a fluorescent group to fulfill these requirements. This procedure, however, makes the analysis more complex because an additional reaction step is included. Another alternative is to employ indirect laser-induced-fluorescence (IDLIF) detection.34-37 IDLIF has been used by Jacobson et al.27 to determine the void volume but has, to our knowledge, not been used for analyte detection on a chip. We recently showed the use of capillary-based CEC and MEKC with IDLIF detection for the analysis of explosives.5 In that work a compact and efficient diode laser was employed in combination with a low concentration of a visualizing agent added to the running buffer. In this work, we present the MEKC separation of a mixture of explosive compounds with IDLIF detection using a microfabricated glass structure. EXPERIMENTAL SECTION Instrumentation. The separations were performed in microfabricated borofloat glass chips obtained from Alberta Microelectronic Corp. (Edmonton, Alberta, Canada). The cross section of the channels was approximately semicircular with a width of 50 µm at the top and a depth of 20 µm. The distances from the buffer, analyte, and analyte waste reservoirs to the intersection were 10 mm while the length of the separation channel was 76 mm. Three different injector designs, a straight-cross, a 100 µm offset doubleT, and a 250 µm offset double-T design, were evaluated. The chips were mounted on an x, y, z micropositioner to facilitate alignment with the IDLIF detector. The separations were electrokinetically driven using voltages of 1-4 kV. The high voltage power supply (Stanford Research Systems, Sunnyvale, CA) was connected to a locally built switch box, which was used to alternate between injection and separation modes. This device provided countervoltages to the analyte and analyte waste reservoirs when used in the separation mode and to the buffer and buffer waste when used in the injection mode.13,16,27 The countervoltages were necessary to create sharp injection plugs16 and to avoid leaking from the analyte and analyte waste channels during the separation.20,22,38 During the injection, 100%, 90%, 90%, and 0% of the high voltage (1 kV) were applied at the analyte, buffer, buffer waste, and analyte waste reservoirs, respectively, for 60 s. In the separation mode, 100%, 60%, 60%, and 0% of the high voltage were applied at the buffer, analyte, analyte waste, and buffer waste reservoirs, respectively. The junction voltage, i.e., the potential at the intersection of the channels was calculated using Kirchhoff’s rules, and the field strength in each channel was determined from the voltage at the intersection and the length of each channel. A home-built epifluorescence IDLIF system was used for detection. Laser excitation was provided by a 750 nm diode laser (Power Technology, Little Rock, AR). A dichroic filter (Omega Optical Inc., Brattleboro, VT) was used to reflect the laser beam 90° into a microscope objective (Mitutoyo MTI Corp., Aurora, IL) with a 20× magnification. The objective focused the laser beam into the center of the separation channel and was also used to (34) Amankwa, L. N.; Kuhr, W. G. Anal. Chem. 1991, 63, 1733-1737. (35) Yeung, E. S.; Kuhr, W. G. Anal. Chem. 1991, 63, 275A-282A. (36) Kennedy, S.; Caddy, B.; Douse, J. M. F. J. Chromatogr., A 1996, 726, 211222. (37) Kaneta, T.; Imasaka, T. Anal. Chem. 1995, 67, 829-834. (38) Manz, A.; Effenhauser, C. S.; Burggraf, N.; Harrison, D. J.; Seiler, K.; Fluri, K. J. Micromech. Microeng. 1994, 4, 257-265.

Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

1873

collect and refocus the resulting fluorescence. The fluorescence was optically filtered through a 750 nm blocking filter (Omega) and a 780 nm long-pass filter (Omega) and was detected with a model HC120-05 photomultiplier tube (Hamamatsu, Bridgewater, NJ). The signal from the PMT was amplified and filtered using a preamplifier (Stanford Research Systems Inc.). Injections were visually inspected using an inverted fluorescence microscope (Olympus IX70, Olympus America Inc., Melville, NY). The injection and separation voltages were applied using the power supply and switch box described above. The injections were done using the same buffer as the explosive separations, with 0.4 mM fluorescein added to the running buffer as a marker. Chemicals and Reagents. 1,3,5-Trinitrobenzene (TNB), 1,3dinitrobenzene (DNB), nitrobenzene (NB), 2,4,6-trinitrotoluene (TNT), tetryl, 2,4-dinitrotoluene (2,4-DNT), 2,6-dinitrotoluene (2,6DNT), 2-nitrotoluene (2-NT), 3-nitrotoluene (3-NT), 4-nitrotoluene (4-NT), 2-amino-4,6-dinitrotoluene (2-Am-4,6-DNT), 4-amino-2,6dinitrotoluene (4-Am-2,6-DNT), and the 14-explosive mixture (EPA 8330) were all purchased from Radian International (Austin, TX). Boric acid (Sigma, St. Louis, MO), fluorescein (Molecular Probe Inc., Eugene, OR), and sodium dodecyl sulfate (SDS) (Life technologies Inc., Gaithersburg, MD) were used as received. The visualizing agent, Cy7, was obtained from Amersham Pharmacia Biotech (Piscataway, NJ). Stock solutions of 5 mM Cy7 were prepared in DMSO and stored at -70 °C. All buffers were prepared using deionized water (Labconco Water Pro PS, Kansas City, MO). The borate buffer was pH-adjusted to 8.5 prior to the addition of the SDS. All buffers were stored at 4 °C. Prior to the addition of Cy7, the buffer was filtered through a 0.02 µm syringe filter (Whatman Inc., Haverhill, MA) and degassed by sonication. Each day, the channels on the microchip were rinsed with 0.1 M sodium hydroxide followed by water and running buffer using a vacuum to pull the solutions through the channels. Soil samples. Soil samples obtained locally were spiked to contain 1 or 5 ppm (1 or 5 µg of analyte/g of soil) of explosives and left over a period of 5 days. The extractions were carried out by agitating 2 g of the soil with 2 mL of acetonitrile for 60 min. The samples were then centrifuged at 2500 rpm for 10 min. After centrifugation, the supernatant was filtered through a 0.02 µm syringe filter (Whatman). A 1.0 mL portion of the extract was evaporated to dryness using a slow flow of nitrogen and reconstituted in 50 µL of acetonitrile and 150 µL of separation buffer. Prior to injection, the sample was filtered through a 0.5 µm stainless steel frit-in-a-ferrule filter (Upchurch Scientific, Oak Harbor, WA). Assuming 100% extraction efficiency, the analytes present in the soil were concentrated five times using the described procedure. Data Handling. All electropherograms were recorded on a laptop computer using a program written in Labview (National Instruments, Austin, TX). Peak height and peak area measurements were done with Igor (Wave Metrics, Lake Oswego, OR). For calculation of peak efficiency, Gaussian fits were performed and the widths at half-height were measured. The plate numbers were calculated using N ) 5.54(tr/wh)2, where N is the peak efficiency, tr is the migration time, and wh is the width at halfpeak height. The resolution between the peaks was calculated using the following equation: Rs ) (2 ln 2)1/2((tr2 - tr1)/(wh1 + wh2)). 1874 Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

Figure 1. Chip-based MEKC-IDLIF electropherogram of the EPA 8330 mixture of nitroaromatics and nitramines. Analytes: 20 ppm of each TNB (1), DNB (2), NB (3), TNT (4), tetryl (5), 2,4-DNT (6), 2,6DNT (7), 2-, 3-, and 4-NT (8), 2-Am-4,6-DNT (9), and 4-Am-2,6-DNT (10). Conditions: MEKC buffer, 50 mM borate, pH 8.5, 50 mM SDS, 5 µM Cy7, separation voltage 4 kV, separation distance 65 mm.

Safety Considerations. The instrumentation uses high voltage and should be handled carefully. The explosive compounds are toxic and should be handled and disposed in a safe manner. Gloves should be used in handling these compounds, since the explosives may also be absorbed through the skin. RESULTS AND DISCUSSION Recently we demonstrated the use of capillary electrochromatography and capillary-based MEKC with IDLIF detection for the analysis of 14 explosive compounds (EPA 8330).5 In the interest of exploring and developing portable systems for rapid field analysis, we transferred the MEKC-IDLIF method to a microchip-based system. Figure 1 shows the separation and detection of 14 explosives in the EPA 8330 mixture using a microfabricated chip with a separation length of 65 mm and a separation voltage of 4 kV. As can be seen, 10 peaks were detected. The three isomers of nitrotoluene (2-, 3-, and 4-NT) were not resolved. Additionally, the two nitramines (HMX and RDX) could only be detected at much higher concentrations (>2000 ppm), likely due to the low fluorescence quenching efficiencies of these compounds compared to the nitroaromatics. Since the quantum yield of Cy7 is higher in aqueous environments than in hydrophobic environments, it is believed that the depletion of fluorescence is due to a quenching of the dye, rather than a displacement of the fluorophore from the interior of the micelles.5 Kennedy et al.36 also reported higher detection limits for the nitramines compared to the nitroaromatic compounds using MEKC-IDLIF. Nevertheless, the separation shown in Figure 1 demonstrates that IDLIF is a suitable detection method for chip-based systems and that a number of important explosive compounds can be separated and detected rapidly. Experiments were also performed at lower separation voltages (2, 2.5, 3, and 3.5 kV) to see the effect on background noise, peak efficiency, and resolution. While the migration times were a factor of 2 longer using a separation voltage of 2 kV compared to 4 kV, neither the resolution between two closely eluting peaks (DNB

Figure 2. Electropherograms obtained after the second, seventh, and twelfth injection of 20 ppm of the EPA 8330 mixture. Peak identification and conditions were as for Figure 1, except that the separation distance was 60 mm.

and NB) nor the peak efficiency was significantly changed. The peak efficiency observed using separation voltages of 2-4 kV was 6 × 104 plates (8 × 105 plates m-1), while the resolution between DNB and NB was 1.2. At this point, our instrumentation is limited to a maximum of 4 kV, but it is of interest to examine the use of higher voltages for faster separations and possibly higher efficiencies. Using the microchip, no increase in background noise was observed upon increasing the separation voltage from 2 to 4 kV (185 and 370 V cm-1). This is in contrast to a capillary-based MEKC-IDLIF system, where an increase in field strength from 100 to 400 V cm-1 resulted in an unstable background fluorescence.5 This was attributed to an increase in Joule heating across the capillary. It has been observed that even small thermal fluctuations can cause significant distortions of the background signal in an indirect detection system.39 These effects can be reduced by cooling the separation system. When the separations are performed using microfabricated channels rather than capillaries, the problems associated with excessive Joule heating are reduced as a result of the more effective heat removal of the glass surrounding the separation channels.21,23 In addition to the use of higher separation voltages, decreasing the separation distance can reduce the analysis time. This was readily done by moving the detection system closer to the injection region. Using a separation distance of 30 mm instead of 60 mm, the separation time was decreased from 57 to 30 s. As a result of the shorter separation length, however, the resolution between DNB and NB in the EPA 8330 mixture decreased from 1.38 to 0.74. Using a separation voltage of 4 kV and a separation distance of 60 mm, the reproducibility of the separation was studied. The relative standard deviation (RSD) in migration time for the (39) Xu, X.; Kok, W. Th.; Poppe, H. J. Chromatogr., A 1997, 786, 333-345.

Figure 3. Normalized background fluorescence (a), normalized peak heights for TNB (O), DNB (2), TNT (]), tetryl (4), and 2,4-DNT (×) (b), and peak signal/background fluorescence (c) versus the injection number. Conditions were as for Figure 2.

Figure 4. Plot of peak area versus injector configuration. Analytes: TNB (]), DNB (O), TNT (2), and tetryl (b). Conditions were as for Figure 1.

different analytes in the 8330 mixture was e1% (n ) 12). While the migration time showed excellent reproducibility, the peak height decreased with time after approximately the fifth to tenth injection. This can also be seen in Figure 2, which shows the second, seventh, and twelfth injections of the EPA 8330 mixture. The RSD in peak height for six injections (2-7) was 1.7-3.8% for TNB, DNB, TNT, tetryl, 2,4-DNT, 2,6-DNT, and 2-Am-4,6-DNT. Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

1875

Figure 5. Fluorescence microscope images of microchip injections, using different injector configurations. Marker: 0.4 mM fluorescein. Injections were done at 1 kV (a-c) and separations at 1 kV (d-f). Injector configurations: straight cross (a, d), 100 µm offset double-T (b, e), and 250 µm offset double-T (c, f). The illustrations to the left of the photographs show the injector configurations and applied voltages during injection and separation modes. Arrows indicate flow directions. The microfabricated channels on the chip have been outlined in (a) and (d) for clarification.

The peak height then decreased approximately 35% from the seventh to the twelfth injection. In general, the first separation obtained after placing new separation buffer and new sample in the reservoirs showed distorted background fluorescence, likely due to old buffer/sample still present in the channels. The decrease in peak height with time is likely a result of buffer and/ or sample degradation caused by electrolysis at the high voltage electrodes.40,41 Figure 3 shows (a) the normalized background fluorescence measured just before the TNB peak, (b) the normalized peak height for TNB, DNB, TNT, tetryl, and 2,4-DNT, and (c) the ratio of peak height and background fluorescence versus the injection number. In Figure 3a,b, the background fluorescence and peak signal have been normalized to the respective values obtained for injection number 2. From Figure 3, it is clearly seen that the decrease in peak height is correlated with a lower background fluorescence. If sample degradation (i.e., oxidation or reduction of the analytes) had been the main cause of the decreased peak heights, a fast decrease in peak height compared to background signal would have been expected. As seen in Figure 3c, however, the ratio of the peak height and the background fluorescence signal is constant or only slightly decreasing with (40) Bello, M. S. J. Chromatogr., A 1996, 744, 81-91. (41) Macka, M.; Andersson, P.; Haddad, P. R. Anal. Chem. 1998, 70, 743-749.

1876

Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

injection number. Therefore, most of the decrease in sensitivity can be attributed to degradation of the dye. This is seen after only a few injections because of the small volumes of the vials on the chip. These effects can be ameliorated by using larger volume vials or by changing the running solutions frequently. However, neither of these approaches is satisfactory for a miniaturized, fieldportable system, and we are currently pursuing alternative approaches to minimize buffer degradation using a separate electrode vial arrangement.42,43 In this work, three different injector configurations were studied. Figure 4 presents the peak areas for TNB, DNB, TNT, and tetryl using a straight-cross, a 100 µm offset double-T, and a 250 µm offset double-T injector. An increased peak area was observed for the double-T injectors compared to the straight cross, while the resolution between DNB and NB was constant at 1.4. The average increase in peak area for TNB, DNB, TNT, and tetryl using the 250 µm double-T injector compared to the straight cross was approximately 35%. This is consistent with recent results from Liu et al.,25 who observed an increase in peak area of 20-25% when comparing a 250 µm offset double-T and a straight-cross (42) Desiderio, C.; Fanali, S.; Bocek, P. Electrophoresis 1999, 20, 525-528. (43) Park, S. Dissertation, The University of Kansas, Lawrence, KS, 1996; pp 136-144.

Figure 6. Separation of 1 ppm each of six nitroaromatic compounds. Peak identifications and conditions were as for Figure 1.

injector. On the basis of the relative volumes present in the cross and double-T injectors, an increase in peak area of approximately 5 times would be expected when the 250 µm offset double-T design is compared to the straight cross. As recently pointed out by Shultz-Lockyear et al.,13 the size and shape of the injection plug are dependent on several factors such as applied voltages and ionic strengths in the sample and buffer. Nevertheless, a larger increase than 20-35% in peak signal would be expected for the 250 µm offset injector relative to the straight cross. To reconcile these results, studies of the injected plug were conducted by imaging an injection of 0.4 mM fluorescein using a fluorescence microscope. It was evident that the size of the injected plug was not significantly changed for the different injector configurations. Figure 5a-c shows the shapes and sizes of the injection plugs during the injection mode, with 1, 0.9, 0.9, and 0 kV applied to the analyte, buffer, buffer waste, and analyte waste reservoir, respectively. Figure 5d-f shows the actual sizes of the plugs injected into the separation channel for the three different injector configurations. The photographs in Figure 5d-f were taken after switching to the separation mode (1, 0.6, 0.6, and 0 kV applied to the buffer, analyte, analyte waste, and buffer waste). In addition to taking the photographs shown above, we visually examined the injection plugs while switching from the injection mode to the separation mode. It was obvious that parts of the injection plugs present in the double-T injectors were driven back into the analyte waste channel as a result of the countervoltages applied to this reservoir. This explained the relatively small increase in peak area observed for the 250 µm offset double-T injector compared to the straight-cross injector. A stable fluorescence background could be reliably obtained using countervoltages of 60% of the separation voltage applied to the analyte and analyte waste reservoirs during the separation step. Using a separation voltage of 4 kV and countervoltages of 2.4 kV, the junction voltage is calculated to be 2.8 kV. As the potential is lower in the analyte and analyte waste reservoirs than at the junction, there will be a significant back-flow toward these reservoirs during the separation step. To increase the size of the injected plug in an offset double-T design, higher countervoltages should be applied to the analyte and analyte waste reservoir to reduce the magnitude of the back-flow. If, for example, 80% of the high voltage is applied to the analyte and analyte waste reservoirs during the separation step, the potential difference between the junction and the analyte and analyte waste reservoir

Figure 7. MEKC-IDLIF analysis of extracts from spiked soil samples: soil blank (a), soil containing 1 ppm of each analyte (b), and soil containing 5 ppm of each analyte (c). Peak identifications and conditions were as for Figure 1.

is only 0.12 kV. This smaller difference should result in a smaller back-flow and a larger portion of the sample plug being injected. However, the small potential difference may not be sufficient to prevent leaking from the analyte and analyte waste channels due to hydrodynamic effects. This is also in agreement with our observations of a more unstable background fluorescence at countervoltages higher than 60%. Another approach is to keep the analyte and analyte waste reservoirs floating for a short period of time during the switch from the injection mode to the separation mode, after which the countervoltage is applied. Keeping the analyte and analyte reservoirs floating would minimize the backflow and result in a larger portion of the plug present in the offset T being injected. Applying a sufficient countervoltage shortly after switching from the injection to the separation mode would then prevent leaking. To be able to use this approach in a reproducible way, a switch box capable of applying fixed or floating potentials to individual reservoirs for a given period of time must be employed. During this work our switch box had a fixed countervoltage of 60% and was manually controlled. As the 250 µm offset double-T injector geometry gave the highest peak signal without a loss in resolution, this design was used to determine the detection limits. Figure 6 shows the injection of 1 ppm (each compound) of TNB, DNB, TNT, tetryl, 2,4-DNT, and 2,6-DNT. This concentration gave signal-to-noise ratios of 3-10, with the lowest value from 2,6-DNT and the highest from tetryl. Calibration curves for TNB, DNB, TNT, and tetryl (plotted as peak height vs injected concentration) showed that the linear range (r > 0.997, n ) 9; three concentrations injected) was only between 1 and 5 ppm. These values are similar to those measured in the capillary IDLIF study.5 Analyte concentrations of 10 ppm showed severe negative deviations from linearity. The narrow linear range is in agreement with previous findings for indirect LIF detection where fluorescence quenching plays a role.34,44 As the chip-based MEKC-IDLIF system primarily is meant to be used as a quick screening tool for the presence of (44) Takeuchi, T.; Yeung, E. S. J. Chromatogr. 1986, 366, 145-152.

Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

1877

explosives rather than a tool for quantifying the amounts, the linear range of the method is at this point of secondary importance. As our future goal is to implement the microchip in a portable device for field screening of contaminated samples, it is important that this technique can be used to analyze environmental samples. Figure 7 shows the separation of nine nitroaromatic compounds in extracts from spiked soil samples. As can be seen, 1 ppm amounts (1 µg of analyte/g of soil) of TNB, DNB, TNT, tetryl, 2,4-DNT, 2,6-DNT, 2-Am-4,6-DNT, and 4-Am-2,6-DNT present in the soil are readily separated and detected using the microchip MEKC-IDLIF system. The peak from NB could not be detected at this level but was clearly seen in the extract from the soil containing 5 ppm of explosives. In conclusion, it has been shown that microchip MEKC-IDLIF can be used as a rapid, selective, and sensitive method to separate and detect nitroaromatic compounds in environmental samples. This is an important tool for rapid screening of samples from sites

1878

Analytical Chemistry, Vol. 72, No. 8, April 15, 2000

suspected to be contaminated with explosive material. In addition, IDLIF as a detection method for chip-based separations will be a valuable complement to LIF for species that do not have an accessible fluorescent band or lack a functional group that can be fluorescently tagged. ACKNOWLEDGMENT The Swedish Natural Science Research Council (NFR) is gratefully acknowledged for a postdoctoral fellowship for S.R.W. Gary Hux is acknowledged for constructing the high-voltage switch box. This work was supported by the Sandia Laboratory Directed Research and Development Program.

Received for review November 30, 1999. Accepted January 20, 2000. AC991382Y