Anal. Chem. 2000, 72, 4928-4933
Chromatographic Detection of Nitroaromatic and Nitramine Compounds by Electrochemical Reduction Combined with Photoluminescence following Electron Transfer Steven J. Woltman,† William R. Even,‡ Eskil Sahlin,† and Stephen G. Weber*,†
Chevron Science Center, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, and Sandia National Laboratories, MS 9403, P.O. Box 969, Livermore, California 94550-0969
The oxidizing agent tris(bipyridyl)ruthenium(III), or Ru(bpy)33+, is used as a postcolumn reagent for the detection of nitroaromatic and nitramine explosive compounds. After separation, the explosives are reduced electrochemically to oxidizable products such as hydroxlamines and nitrosamines, and these products react readily with Ru(bpy)33+ and Ru(bpy)32+. The photolumininescence from the latter is used for detection. A porous carbon electrode was used for on-line analyte reduction following chromatography. Another porous carbon electrode was used to generate the nonluminescent Ru(bpy)33+ from Ru(bpy)33+ on-line at high efficiency. The two streams were combined, and the Ru(bpy)32+ produced by oxidation of the reduced analytes was detected by laser illumination and light detection. Reductive hydrodynamic voltammograms of nitrobenzene, 2,4,6-trinitrotoluene, and hexahydro-1,3,5trinitro-1,3,5-triazine indicated that a potential of -1500 mV vs Ag/AgCl was sufficient to achieve a maximum signal from the reduced analytes. HPLC with a water/acetonitrile gradient on a C-18 reversed-phase column was then used to determine these three compounds plus the four additional examples, 1,3,5,7-tetrazocine, 2,4-dinitrotoluene; 2,6-dinitrotoluene, and 4-nitrotoluene. For both hydrodynamic voltammetry and HPLC detection, the photoluminescence following electron-transfer signal was calibrated using the one-electron standards ferrocene and ferrocenecarboxylic acid. Detection limits were in the lownanomolar range for 20-µL injections of nonpreconcentrated nitro compounds. Recently, a novel detection scheme for chromatography, “photoluminescence following electron transfer” (PFET), was described1 in which the photoluminescence of a dye was determined following its reduction by the analyte. We are pursuing this course in hopes of eliminating some of the drawbacks of amperometric electrochemical detection, including electrode fouling, sluggish electron-transfer kinetics, the electrochemical background signal’s enormous sensitivity to the solution environment, †
University of Pittsburgh. Sandia National Laboratories. (1) Woltman, S. J.; Even, W. R.; Weber, S. G. Anal. Chem. 1999, 71, 1504. ‡
4928 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000
and the limitations inherent in a mass transport-limited detector such as flow rate dependence, dependence of signal on molecular size, and the obvious fact that most electrochemical detectors oxidize/reduce less than 100% of the analyte.2 Some of these problems are eliminated or minimized by using porous electrodes as detectors. As the efficiency approaches 100%, the mass transport limitations, e.g., dependence of signal on solute size, disappear. Highly efficient coulometric detectors are also less susceptible to fouling because the signal dependence on surface area is very small for highly efficient electrodes. However, the sensitivity to kinetics, and especially to the solution environment (temperature, pH, ionic strength, ionic constitution, solvent composition), is very real for so-called coulometric detectors. Also, the background current is typically large because of the large surface area. The embodiment of PFET reported here uses as a postcolumn reagent tris(2,2′-bipyridyl)ruthenium(III), denoted hereafter as Ru(bpy)33+, one of a number of similar diimine complexes of ruthenium and osmium.3-7 In the M(II) oxidation state, such complexes luminesce under illumination at visible wavelengths, while in the M(III) oxidation state they do not. The luminescent M(II) form is generated by electron transfer from oxidizable analytes. Ideally, the redox reaction proceeds to completion, so that the M(II) luminescence signal is proportional both to analyte concentration and to the number of electrons transferred per mole of analyte. A single redox standard such as ferrocene can then be used for quantitative calibration of all the analytes that are detected by the system. A limitation of using these Os(III) and Ru(III) complexes for electron-transfer-dependent luminescence detection is that analytes that are not susceptible to oxidation cannot be detected directly. Fortunately, many analytes are susceptible to electrochemical transformation to a form that reacts with M(III). Such a (2) Weber, S. G. Detection Based on Electrical and Electrochemical Measurements. In Detectors for Liquid Chromatography; Yeung, E. S., Ed.; Wiliey: New York, 1986; pp 229-291. (3) Gerardi, R. D.; Barnett, N. W.; Lewis, S. W. Anal. Chim. Acta 1999, 378, 1. (4) Kalyanasundaram, K. Coord. Chem. Rev. 1982, 46, 159. (5) Juris, A.; Balzani, V.; Barigelletti, S.; Campagna, S.; Belser, P.; Von Zelewsky, A. Coord. Chem. Rev. 1988, 84, 85. (6) Kober, E. M.; Sullivan, B. P.; Dressick, W. J.; Caspar, J. V.; Meyer, T. J. J. Am. Chem. Soc. 1980, 102, 7383. (7) Kober, E. M.; Caspar, J. V.; Lumpkin, R. S.; Meyer, T. J. J. Phys. Chem. 1986, 90, 3722. 10.1021/ac000170u CCC: $19.00
© 2000 American Chemical Society Published on Web 09/13/2000
two-step scheme has been exploited in electrochemical detection using dual electrodes in series, with an upstream generator electrode and downstream collector electrode.8-10 The proposed modification of the electron-transfer-dependent luminescence scheme resembles such a dual-electrode approach, using an upstream cathode (generator) while employing the M(III) reagent in place of a downstream anode (collector). Nitroaromatic and nitramine compounds are candidates for detection by this scheme, as they are reducible to oxidizable products.11-13 Nitroaromatic compounds such as 2,4,6-trinitrotoluene (TNT) are reduced successively to hydroxylamines and then to amines, and nitramines such as hexahydro-1,3,5-trinitro-1,3,5triazine (RDX) first to nitrosamines and then to amines.11,12 Trace analysis of these compounds is of considerable interest: these compounds are used in military weapons (e.g., land mines, plastic explosives) and are highly toxic, existing as environmental contaminants at a number of sites. Dinitro- and mononitrotoluenes may be present as degradation products in contaminated soil. Also, dinitrotoluene is present as an impurity in TNT and serves as a volatile marker for the presence of that explosive. This paper examines several aspects of this novel approach. Briefly, the potential required to generate a signal (Eappl) is in the -1.0 to -1.5 V (vs Ag/AgCl) range for nitrobenzene consistent with nitroaromatic behavior. Hydrodynamic voltammograms with photoluminescence detection were generated in different acetonitrile/water compositions in order to evaluate the effect of solvent on Eappl and sensitivity. We have also studied “blank” hydrodynamic voltammograms, which reveal the interference of O2. Following the establishment of useful conditions, flow injection experiments examined linearity and detection limit. Gradient elution conditions were established for the chromatographic separation of several explosives and related compounds. The detection system was shown to provide lower detection limits than other similar separation/detection conditions. EXPERIMENTAL SECTION Chemicals. All chemicals were used as received, except for sodium perchlorate which was recrystallized once from methanol. Nitrobenzene was obtained from Fisher Scientific. (RDX, 1,3,5,7tetrazocine (HMX), TNT, 2,4-dinitrotoluene (2,4-DNT), 2,6-dinitrotoluene (2,6-DNT), and 4-nitrotoluene (4-NT) were obtained from Supelco (Bellefonte, PA) as 1 mg/mL acetonitrile solutions in sealed glass ampules. Ferrocene and ferrocenecarboxylic acid, used as one-electron standards, were obtained from Aldrich. Sources for all other chemicals were as in ref 1. Solutions for Flow Injection. Ru(bpy)32+Cl2, 100 µM, was dissolved in 0.1% trifluoroacetic acid/0.1 M NaClO4/acetonitrile. The solvent/mobile phase for all analytes and standards was 0.1 M NaClO4 in unbuffered acetonitrile/water. The electrolyte filling solutions for the porous electrodes were 0.1% trifluoroacetic acid/ 0.1 M NaClO4/acetonitrile for the Ru(bpy)32+ synthetic electrode (8) MacCrehan, W. A.; Durst, R. A. Anal. Chem. 1981, 53, 1700. (9) Rosten, D. A.; Kissinger, P. T. Anal. Chem. 1982, 54, 429. (10) Mayer, G. S.; Shoup, R. E. J. Chromatogr. 1983, 255, 533. (11) Bard, A. J.; Lund, H. Encyclopedia of Electrochemistry of the Elements; Marcel Dekker: New York, 1979; , Vol. XIII, pp 90-91, 92-93. (12) Bratin, K.; Kissinger, P. T.; Briner, R. C.; Bruntlett, C. S. Anal. Chim. Acta 1981, 130, 295. (13) Wang, J.; Lu, F.; MacDonald, D.; Lu, J.; Ozsoz, M. E. S.; Rogers, K. P. Talanta 1998, 46, 1405.
Figure 1. (A) Flow injection apparatus. The loop of injector 1 was filled with Ru(bpy)32+ solution, which was oxidized to Ru(bpy)33+ in electrode 1, Eappl +1500 mV vs Ag/AgCl. For hydrodynamic voltammograms, the loop of injector 2 was filled with analyte dissolved in mobile phase and electrolyzed at a series of potentials in electrode 2. For standard curves, the loop of injector 2 was filled with mobile phase and the loop of injector 3 was filled with analyte dissolved in mobile phase. Each porous working electrode was used in conjunction with reference and auxiliary electrodes (not shown). (B) Gradient HPLC apparatus. The mobile phase for the column was supplied directly from a nonmetallic gradient pump, while the large loop for Ru(bpy)32+ reagent was retained.
(electrode 1) and 0.1 M NaClO4/50% acetonitrile/50% water for the analyte reduction electrode (electrode 2). Flow Injection Apparatus. Construction of porous electrodes (carbon powder packed in 330-µm-i.d. Nafion tubing), flow injection methods, and the Varian Fluorichrom detector modified for argon ion laser LIF axial illumination have been described previously.1 An additional porous electrode has been added to the flow injection apparatus (Figure 1) for electrochemical conditioning of analytes. In all experiments, low laser power (∼3 mW) was used, unfiltered. The measured power actually delivered through the optical fiber was 2 mW. Luminescent emissions were filtered with a 600-nm long-pass interference filter. It was necessary to be able to alter the solvent environment rapidly. Therefore, in many cases the flow of acetonitrile from two HPLC pumps (50 µL/min) was used to propel the contents of injection loops connected to Rheodyne 7125 injectors. The porous carbon working electrodes were placed in electrolyte solution with auxiliary electrodes of reticulated vitreous carbon and Ag/AgCl reference electrodes, with applied potential Analytical Chemistry, Vol. 72, No. 20, October 15, 2000
4929
controlled by potentiostats (models CV-1A and CV-1B, BAS Inc., West Lafayette, IN). Electrode 1 was packed with Sandia RF2000 carbon powder. Electrode 2 was packed with type 1 spherical glassy carbon powder (Alfa Aesar/Johnson Matthey, Ward Hill, MA, originally 0.4-12 µm). The glassy carbon powder was depleted of fines by repeated agitation, settling, and decantation of a dilute 2-propanol slurry. Packed electrode beds were 330 µm i.d. and 1 mm in length, with an approximate volume of 100 nL. Hydrodynamic Voltammograms. Loop 1 (Figure 1A) was used for Ru(bpy)32+Cl2 solution, loop 2 was used for analyte solution, and loop 3 was bypassed. At the flow rates used, the volumes of loops 1 and 2 were sufficient for 1 h of constantcomposition flow. The potential of electrode 1 was held at +1500 mV vs Ag/AgCl to oxidize Ru(bpy)32+ to Ru(bpy)33+, and the potential of electrode 2 was varied stepwise. Each potential was maintained until a constant luminescence signal was observed. The potential changes were applied in random order, so that any time-dependent phenomena that occurred would not be mistaken for potential-dependent phenomena. Flow Injection of RDX and TNT. Ru(bpy)33+ was synthesized on-line as for hydrodynamic voltammograms, above. An analyte mobile phase containing 75% acetonitrile was used. The potential of electrode 2 was held constant at -1500 mV vs Ag/AgCl. The small 10-µL loop (3) was used to inject RDX and TNT dissolved in mobile phase, at concentrations from 0 (blank) to 20.0 µM. Additional electrochemical blanks of RDX and TNT were obtained by injecting 20.0 µM concentrations of each compound with the analyte reduction electrode disconnected from the potentiostat. HPLC. The apparatus is shown in Figure 1B. A 1 × 100 mm BetaBasic C-18 column (Keystone Scientific, Bellefonte, PA) was used. Mobile phase for the column was supplied by a Dionex GP50 metal-free gradient pump at 50 µL/min. The mobile phases all contained 7% (v/v) 2-propanol. An acetonitrile/water gradient was used for elution. The detector used for HPLC was different from that used in FIA experiments. The axial illumination luminescence configuration was retained, with a short length of 0.01-in.-i.d/0.03-in.-o.d. Teflon AF2400 tubing (Random Technologies, San Francisco, CA) serving as the flow cell. Tubing made of this low-refractive index fluoropolymer tubing (n ) 1.29), creates a liquid core optical waveguide.14 Unlike reference, the excitation light is introduced axially and the emission is monitored transversely. A Hamamatsu R2496 photomultiplier tube with C3830 power supply was used for light intensity measurements. Two convex lenses, between which was a 600-nm long-pass filter, were used to collect/focus the light onto the PMT photocathode. The PMT and the flow cell were enclosed in a light-tight box. A picoammeter (model 427, Keithley Instruments, Cleveland, OH) was used to convert the PMT current to a voltage signal. All other aspects of the system were the same as for flow injection, including the use of the large injection loop to supply Ru(bpy)32+ reagent. RESULTS AND DISCUSSION We anticipate that electrode 2 (Figure 1) will create species capable of reducing Ru(bpy)33+, which will then produce a photoluminescent signal. As a consequence, a plot of luminescence (14) Dasgupta, P. K.; Zhang, G. F.; Boring, C. B.; Jambunathan, S.; Al-Horr, R. Anal. Chem. 1999, 71, 1400.
4930 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000
Figure 2. Hydrodynamic voltammograms of 20.0 µM ferrocene standard (]) and 20.0 µM nitrobenezene in mobile phases with acetonitrile contents of 25 (∆), 50 (9), and 75% (0) by volume. The loop of injector 2 (Figure 1A) was used for ferrocene and nitrobenzene injections. All solutions contained 0.1 M NaClO4 as electrolyte. The acetonitrile content of the ferrocene solution was 90%. Inset: Parameters for transformation of ferrocene hydrodynamic voltammogram (see Results and Discussion section).
intensity as a function of potential applied to electrode 2 is a hydrodynamic voltammogram. Several such voltammagrams are shown in Figure 2. The ferrocence voltammogram acts as a oneelectron standard, allowing for the determination of the apparent number of electrons transferred for analytes. The shape of this voltammogram can be used to assess the adequacy of the potential control within electrode 2. Eappl is plotted against log[(F - Fmin)/ (Fmax - F)] (Figure 2, inset) to give an “Eappl” intercept equal to the half-wave potential and a slope of 59.1 mV for a reversible one-electron transfer. Plotting log[(F - Fmin)/(Fmax - F)] vs Eappl for ferrocene produced a slope of 60.5 mV, indicating that the porous electrode was operating nearly reversibly. The half-wave potential obtained was 422 mV vs Ag/AgCl. An efficiency of 98 ( 0.63% (95% confidence internval) was calculated for ferrocene oxidation, as (Fmin/Fmax) × 100%. Nitrobenzene in various acetonitrile water concentration yielded a signal equivalent to 2.5, 2.6, and 2.7 electrons versus the ferrocene standard (Figure 2); a yield of 2-4 electrons would be expected for oxidation of phenylhydroxylamine, a likely product of the upstream cathode reaction of nitrobenzene.11 As the acetonitrile (AN) content was increased, the half-wave potential of nitrobenzene shifted to more negative potentials and the onset of the wave became sharper. There is a small increase in quantum efficiency of photoluminescence as the AN concentration increases.15 If the flow injection/chromatographic flow stream is 100% aqueous, then the solvent composition in the detector is 50% AN. If the flow injection/chromotographic flow stream is 100% AN, the measurement is made in 100% AN. Over this range, (50100% AN) the quantum efficiency increases from about 0.063 to 0.068, or less than 10%. Thus, we do not see, nor do we expect, a large solvent dependence on the photoluminescence intensity. (15) Sun, H.; Hoffman, M. Z. J. Phys. Chem. 1993, 97, 11956-11959.
Figure 3. Blank voltammograms obtained for nitrobenzene hydrodynamic voltammograms at acetonitrile contents of 25 (∆), 50 (9), and 75% (Ο). Inset: Voltammogram of 20 µM nitrobenzene, 75% acetonitrile, plotted with corresponding blank.
Figure 4. Hydrodynamic voltammograms of 10.1 µM ferrocenecarboxylic acid (]), 10.0 µM TNT (0), and two successive loops of 10.0 µM RDX (9) and (0).
Blank voltammograms (Figure 3) are reproducible and subtracted well from the corresponding nitrobenzene signals. The blank signal exhibits a maximum at approximately -700 mV, which declines to near zero at potentials more positive and negative than this. We hypothesize that at intermediate potentials residual oxygen in the mobile phase was reduced to hydrogen peroxide by the porous electrode and that the peroxide was oxidized by the M(III) reagent, creating M(II). The disappearance of the background at more negative potentials is consistent with the peroxide being reduced to water. We therefore have the unique and useful situation that the background actually decreases as the potential becomes more extreme. Hydrodynamic Voltammograms of Ferrocenecarboxylic Acid, RDX, and TNT. A transformed plot of the luminescence signal for 10.1 µM ferrocenecarboxylic acid standard gave a slope (71.6 mV) indicating quasi-reversible behavior. The half-wave potential was 496 mV vs Ag/AgCl. Porous electrode efficiency was calculated to be 95 ( 1.1% (95% confidence interval). The voltammetric waves (Figure 4) obtained by electrontransfer-induced photoluminescence of RDX and TNT are reasonably consistent with reported reductive voltammograms obtained directly for these compounds on planar electrodes,12,13 although shifted to more negative potentials. The luminescence signal required time to stabilize after each change in potential, in some
Figure 5. Flow injection of RDX (10 µL) for standard curve, with concentrations in micromolar indicated above the peaks. Results for TNT were similar. The loop of injector 2 (3 mL) was used to supply a mobile phase of 0.1 M NaClO4 in 75/25 acetonitrile/water, while the loop of injector 3 (10 µL) was used to inject RDX dissolved in mobile phase. “MP” denotes the disturbance introduced by arrival of the mobile phase from injector 2. Inset: Electrochemical blanks that tested the luminescence signal obtained when RDX and TNT were not electrochemically reduced. RDX (20.0 µM) and TNT (20.0 µM) were injected while the analyte electrolysis electrode (electrode 2 in Figure 1) was at open circuit.
cases a few minutes. This was most pronounced at intermediate potentials. Repeats of the same potential exhibited some variability in the signal obtained, and this is evident in the voltammograms. There was also some evidence of hysteresis, in that the signal obtained for a particular potential was larger if the previous potential had been more negative and smaller if the previous potential had been more positive. A potential of -1500 mV was sufficient to obtain maximum signal for both compounds. The maximum luminescence signal for RDX corresponded to 4.0 ( 0.2 electrons and that for TNT to 6 ( 1 electrons (95% confidence intervals). Standard Curves for TNT and RDX by Flow Injection Analysis. Good linearity was achieved over the range of concentrations used (Figure 5). R2 values for TNT and RDX were close to 1. Blanks obtained by injecting mobile-phase aliquots produced small positive baseline disturbances. Blanks obtained by injecting 20 µM RDX and TNT, with electrode 2 switched off, produced small negative baseline disturbances (Figure 5, inset). HPLC. The gradient separation was compatible with PFET detection, yielding a level baseline (Figure 6A). Hilmi et al.21 mentioned 2-propanol as a possible additive but used SDS ultimately to effect an explosives separation. We found that the addition of 7% 2-propanol improved resolution of the test compounds and also decreased retention time. Changing the reduction potential in increments from -1500 to -2000 mV vs Ag/AgCl did (16) Yinon, J.; Zitrin, S. In Modern Methods and Applications in Analysis of Explosives; John Wiley & Sons: West Sussex, U.K., 1996; pp 212-243. (17) Echols, R. T.; Christensen, M. M.; Krisko, R. M.; Aldstadt, I., J. H. Anal. Chem. 1999, 71, 2739-2744. (18) Narang, U.; Gauger, P. R.; Ligler, F. S. Anal. Chem. 1997, 69, 1961-1964. (19) Narang, U.; Gauger, P. R.; Ligler, F. S. Anal. Chem. 1997, 69, 2779-2785. (20) Hilmi, A.; Luong, J. H. T.; Nguyen, A. Anal. Chem. 1999, 71, 873. (21) Hilmi, A.; Luong, J. H. T.; Nguyen, A. J. Chromatogr., A 1999, 844, 97.
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000
4931
Table 1. Electron Counts and Reproducibility for 20-pmol Injections peak area rsd % (n ) 6)c
electron count
a
peak height rsd % (n ) 6)c
compd
flow injectiona,b
HPLCa
20 µL, 1.0 µM inj
200 µL 100 nM inj
20 µL, 1.0 µM inj
200 µL, 100 nM inj
NB TNT RDX HMX 2,4-DNT 2,6-DNT 4-NT
2.6 6 4
1.8 4.5 2.3 2.3 3.0 3.4 1.5
4.9 4.5 5.6 6.1 4.1 1.5 6.1
7.7 5.8 8.4 1.2 4.0 2.4 6.1
5.0 3.6 3.1 3.7 3.8 2.1 7.0
12.1 2.3 4.4 1.4.2 4.7 1.9 9.3
Relative to ferrocene or ferrocenecarboxylic acid standard. b Continuous analyte flow. c rsd, relative standard derivation.
Figure 6. Chromatograms of nitro compounds. Gradient elution, 50 µL/min, with 1 × 100 mm C-18 column. Mobile phase A, water/ 7% 2-propanol/0.1 M NaClO4. Mobile phase B, acetonitrile/7% 2-propanol/0.1 M NaClO4. Postcolumn reagent, 30 µL/min of 50 µM Ru(III)(bpy)3 in acetontrile/0.1% v/v trifluoroacetic acid/0.1 M NaClO4. (A) Seven nitro compounds at high concentration (5 µM), plus ferrocenecarboxylic acid (10 µM) as one-electron internal standard. Injection loop volume, 20 µL; 10-60% mobile phase B in 35 min. (B) Seven nitro compounds near the detection limit. All compounds 2.5 nM; injection loop volume, 200 µL; 5-55% mobile phase B in 35 min.
not change the peak areas for the nitro compounds. In addition, the peak area for the ferrocenecarboxylic acid (FCA) standard was constant over an applied potential range of 0 to -2000 mV. The apparent numbers of electrons for each of the seven nitro compounds (5 µM), determined relative to FCA (10 µM), are listed in Table 1. The electron counts for NB, RDX, and TNT are 4932 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000
somewhat less than was determined from the hydrodynamic luminescence voltammograms. The relative standard deviation of peak areas and heights for 20-pmol injections are shown in Table 1. They are acceptable, being ∼5% relative to the mean. The concentration detection limits are discussed in comparison to other techniques in the next section. For nonpreconcentrated samples (20-µL loop), they are in the low-nanomolar range. With preconcentration (Figure 6B), using an injection loop 10 times larger, the detection limits are subnanomolar in most cases. This means of preconcentration is made possible by gradient elution and is quite easy, requiring little or no change in the gradient. On the basis of these results, and the flow injection results at higher concentrations, we conclude that the dynamic range of this detection method extends at least from 1 nM to 20 µM. Comparison to Other Methods. Mass Spectrometry. Explosives detection by MS has been reviewed.16 Ion mobility spectrometry was used to detect TNT at 0.6 ppb in air. MS with chemical ionization, using water as the ionizing reagent, had reported mass detection limits of 20 ng (88 pmol) for TNT and 40 ng (180 pmol) for RDX. A mass detection limit of 50 fg (0.2 fmol) for TNT was reported for negative ion MS with atmospheric pressure ionization. Rapid/Sensor-Based Methods for TNT in Aqueous Media. TNT is a good benchmark compound for comparison of rapid or sensorbased analyses in water, as this compound has been determined by a number of methods. A simple flow injection/spectrophotometric method featuring reduction of TNT with sodium sulfite solution gave a detection limit of 0.5 µg/mL (2.2 µM) for aqueous standards.17 A screen-printed fast voltammetric sensor for TNT reported recently by Wang et al.13 had a detection limit of 200 ppb (880 nM). An ion-exchange resin flow-through sensor relied upon suppression of the fluorescence signal of an immobilized dye from formation of a colored TNT adduct. The detection limit was 70 ppb (∼300 nM).18 In cases where a suitable antibody exists, the lowest reported detection limits have been achieved by immunoassay. Various immunoassays have reported detection limits in the range of 0.02-10 ng/mL (0.09-44 nM).18 The most sensitive sensor published was reported by Ligler et al. for a rapid capillary-based displacement flow immunosensor,19 with a detection limit of 0.044 nM (0.44 fmol) for TNT and a dynamic range of approximately 0.044-4400 nM. Separation-Based Methods. When it is desired to determine multiple analytes in a single run, a separation is typically required.
Table 2. Detection Limits method
TNT
RDX
HMX
NB
2,6-DNT 4-NT ref
A. Mass Detection Limits (pmol) for Separation-Based Methods GC/EIMS 0.22 135 GC/MS/NCI 0.55 135 GC/ECD 440 9000 GC/ECD 6.8 × 105 LC/MS/NCI 440 LC/Diode Array 220 LC/UV 4.4 9 LC/UV 10 16 16 20 26 LC/ED 0.90 2.2 2.0 2.4 4.0 LC/PFET 0.094 0.24 0.34 0.26 0.080
32 8.0 0.18
14 14 14 14 14 14 19 22 22 a
B. Concentration Detection Limits (mM) without Preconcentration for Separation-Based Methods LC/UV 500 800 800 1000 1300 1600 22 CE/ED 350 470 470 650 470 21 LC/ED 45 110 100 120 200 100 22 LC/PFET 4.7 12 17 13 4.0 9.1 a a
This work.
Disregarding relatively insensitive TLC methods, this means the use of either capillary electrophoresis, gas chromatography, or HPLC. Most GC and HPLC methods reported in the literature employ a large degree of preconcentration, either by liquid/liquid extraction combined with solvent evaporation or by solid-phase extraction. The mass detection limit provides a suitable basis for comparison in this case. When only the concentration detection limit is reported, the mass detection limit can be roughly estimated if the original sample volume is known (with the understanding that extraction efficiency may vary). Several GC- and LC-based methods are reported in Table 2A. Hilmi et al. have investigated determination of a number of explosives by reductive electrochemical detection combined with separation by micellar capillary electrophoresis20 and HPLC,21 with comparison to UV detection. Preconcentration was not used in these studies. Determination of explosives by capillary electrophoresis required a silver-coated gold electrode,14 while a glassy carbon electrode was used for the HPLC work.13 The advantage of electrochemical detection, besides improving detection limits over UV-based detection, is selectivity.22 This is of particular value for a complex sample matrix. Some difficulties were encountered with system peaks due
to reduction of dissolved oxygen, however. A summary of concentration detection limits for separation-based methods without preconcentration is given in Table 2B. It is evident from the tabulated data that the current technique is 1 order of magnitude more effective (lower detection limit) than the electrochemical detection technique to which it is closely related. Furthermore, the detection limits are in a narrow range of concentrations or quantities. Thus, there are no particularly “good” or “bad” analytes, and the technique seems therefore generally applicable to compounds with these functional groups. The PFET detector employed a low-powered laser (∼3 mW), an uncooled PMT, and two potentiostats. In practice, the potentiostats could be replaced by transformers or batteries since only sufficient potential to perform the electrochemistry is required, rather than a precisely controlled potential needed for amperometric detection. This relatively unsophisticated apparatus can undoubtedly be improved upon with respect to detection limits and “packaging” for ease of use. In addition to nitro-containing explosives, the procedure should be applicable to other compounds with reducible nitro groups such as herbicides (e.g., trifluralin) and drug substances (e.g., chloramphenicol). Nitrate esters such as nitroglycerin can be reduced electrochemically to alcohol plus nitrite ion12 and the nitrite ion may be susceptible to oxidation by the M(III) reagent. Detection of compounds that contain other oxidation-resistant species having chemically reversible electrochemistry, such as metal ions, metal complexes, quinones, and certain cationic nitrogen-containing compounds such as methyl viologen, should also be possible. ACKNOWLEDGMENT We thank NIH for financial support through Grant GM-44842. Parts of this work were supported by U.S. DOE contract AC0494AL85000. E.S. acknowledges financial support from The Swedish Foundation for International Cooperation in Research and Higher Education (STINT).
Received for review February 10, 2000. Accepted July 25, 2000. AC000170U (22) Lewin, U.; Efer, J.; Engewald, W. J. Chromatogr., A 1996, 730, 161.
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000
4933
