Trace explosive vapors of TNT and RDX collected on sorbent-filled thermal desorption tubes were analyzed using a programmed temperature desorption system coupled to GC with an electron capture detector. The instrumental analysis is combined with direct liquid deposition method to reduce sample variability and account for instrumentation drift and losses.
The direct liquid deposition of solution standards onto sorbent-filled thermal desorption tubes is used for the quantitative analysis of trace explosive vapor samples. The direct liquid deposition method yields a higher fidelity between the analysis of vapor samples and the analysis of solution standards than using separate injection methods for vapors and solutions, i.e., samples collected on vapor collection tubes and standards prepared in solution vials. Additionally, the method can account for instrumentation losses, which makes it ideal for minimizing variability and quantitative trace chemical detection. Gas chromatography with an electron capture detector is an instrumentation configuration sensitive to nitro-energetics, such as TNT and RDX, due to their relatively high electron affinity. However, vapor quantitation of these compounds is difficult without viable vapor standards. Thus, we eliminate the requirement for vapor standards by combining the sensitivity of the instrumentation with a direct liquid deposition protocol to analyze trace explosive vapor samples.
Gas Chromatography (GC) is a core instrumental analysis technique of Analytical Chemistry and is arguably as ubiquitous as a hot plate or balance in a chemistry laboratory. GC instrumentation can be used for the preparation, identification, and quantitation of a multitude of chemical compounds and can be coupled to a variety of detectors, such as flame ionization detectors (FIDs), photo-ionization detectors (PIDs), thermal conductivity detectors (TCDs), electron capture detectors (ECDs), and mass spectrometers (MS), depending on the analytes, methodology, and application. Samples can be introduced through a standard split/splitless inlet when working with small sample solutions, specialized headspace analysis inlets, solid phase micro-extraction (SPME) syringes, or thermal desorption systems. GC-MS is often the standard technique used in validation and verification applications of alternative or emerging, detection techniques because of its utility, flexibility, and identification power with established chemical databases and libraries.1–7 GC and its related sampling and detecting components is ideal for routine chemical analysis and more specialized, challenging analytical applications.
An analytical application of increasing interest to military, homeland security, and commercial enterprises is trace explosive vapor detection, with detection including identification and quantitation. Trace explosive vapor detection is an unique analytical chemistry challenge because the analytes, such as 2,4,6-trinitrotoluene (TNT) and cyclotrimethylenetrinitramine (RDX) have physical properties that make them especially difficult to handle and separate using broader, more generic chemical analysis methodologies. The relatively low vapor pressures and sub parts-per-million by volume (ppmv) saturated vapor concentration, combined with relatively high sticking coefficients, necessitate special sampling protocols, instrumentation, and quantitation methods.8–12 A GC coupled to an Electron Capture Detector (ECD) or mass spectrometer (MS) is an effective method for quantitating explosive analytes, specifically dinitrotoluene (DNT), TNT, and RDX.6,13–17 GC-ECD is particularly useful for nitro-energetic compounds because of their relatively high electron affinity. The U.S. Environmental Protection Agency (EPA) has created standard methods for explosive analyte detection using GC-ECD and GC-MS, but these methods have focused on samples in solution, such as ground water, and not samples collected in the vapor phase.2,18–23 In order to detect explosive vapors, alternative sampling protocols must be used, such as vapor collection with sorbent-filled thermal desorption sample tubes, but quantitative detection remains difficult due to lack of vapor standards and calibration methods that do not account for sample tube and instrumentation losses.
Recently, quantitation methods using thermal desorption systems with a cooled inlet system (TDS-CIS), coupled to a GC-ECD have been developed for TNT and RDX vapors.24,25 The losses associated with the TDS-CIS-GC-ECD instrumentation for trace explosive vapors were characterized and accounted for in example calibration curves using a direct liquid deposition method onto sorbent-filled thermal desorption sample tubes. However, the literature focused on instrumentation characterization and method development but never actually sampled, analyzed, or quantitated explosive vapors, only solution standards. Herein, the focus is on the protocol for sampling and quantitating explosive vapors. The protocol and methodology can be expanded to other analytes and trace explosive vapors, such as Pentaerythritol tetranitrate (PETN).
1. Instrument Preparation
2. Preparation of Standards
3. Sample Collection
4. Calibration Curve Generation
5. Sample Analysis
Obtaining quantitative results for trace explosive vapor samples begins with establishing a calibration curve for the TDS-CIS-GC-ECD instrumentation using the direct liquid deposition method of solution standards onto sample tubes to account for instrument losses and differences between solution standards and vapor samples. The TDS-CIS-GC-ECD instrumentation and method for TNT and RDX trace analysis has been previously described in detail elsewhere, but the instrument parameters are summarized in Table 1.24,25 Here, Figure 1 shows a series of chromatograms obtained using the published method and parameters in Table 1. Peaks for 3,4-DNT, TNT, and RDX are observed at 4.16, 4.49, and 4.95 min, respectively. The peak height and area for the internal standard is constant for all masses of TNT and RDX, while the peak height and area increases with mass of the analyte. The peak areas for TNT and RDX for each mass are normalized by the peak area for 3,4-DNT to account for irreproducibility and losses associated with sample tube injection. The normalized peak areas for each analyte are then plotted versus mass on sample tube to establish a calibration curve. A linear regression is conducted to obtain the slope, intercept, and coefficient of determination (R2). The slope and intercept are used for converting the normalized peak area for vapor sample to mass, or ultimately concentration. Figure 2 shows an example calibration curve generated from the chromatograms shown in Figure 1. The error bars indicate one standard deviation with three replicate measurements per mass of analyte (N=3). An ideal calibration curve with no instrument or sampling losses and linear detector response would have an R2 value near unity. An R2 value that significantly deviates from unity, approximately less than 0.98, is typically an indicator the instrument needs servicing, the solution standards were not properly prepared, or solution standards and the internal standard were not properly deposited onto the glass frit of the sample tubes.
The calibration curve, the plot, and related raw data, are saved in the same spreadsheet as the sample information so the calibration used for quantitation is easily accessed and tracked with the analyzed samples. The calibration curve and peak areas from a sample can be used to calculate a vapor concentration using the following set of equations:
(1)
(2)
(3)
(4)
where Aa is the analyte peak area, As is the internal standard peak area, b is the calibration curve Y-intercept for the analyte, S is the calibration curve slope for the analyte, C is a conversion factor to parts-per-billion by volume (ppbv, 109), M is the molecular weight for the analyte (g mol-1), Qs is the sample flow rate (ml min-1), L is a conversion factor from milliliters to liters (103), R is the ideal gas constant (8.314 L kPa K-1 mol-1), T is the temperature (K), t is the sample time (min), and P is the pressure (kPa). These series of equations can be embedded into a spreadsheet for automatic calculation of quantitation values. Importantly, these equations assume an ideal gas, so the concentrations have a reduction in accuracy because none of the analytes are ideal gases.
Figure 3 shows an example of a chromatogram that indicates the instrument is in need of service or new solution standards should be prepared. Additional peaks other than those identified as 3,4-DNT, TNT, and RDX appear in the chromatogram. Additional peaks are always present when using sorbent-filled thermal desorption sample tubes because the sorbent material degrades over time with repeated usage and does not selectively adsorb just DNT, TNT, and RDX. However, the degradation products do not co-elute with 3,4-DNT, TNT, and RDX with a properly maintained instrument.26 A blank tube should be run before and after each calibration series to identify peaks that are present from either sorbent material degradation or impurities captured during the vapor sampling collection. This is easily achieved with the use of a twenty sample tube autosampler, where eighteen calibration standard sample tubes are used for the calibration curve and two additional positions are free for blank tubes at the start and end of the sequence. Additional peaks not observed in the blank, but observed in sample tubes deposited with solution standards to generate a calibration curve, typically indicates solution analyte degradation and new solution standards should be prepared and deposited on a new set of sample tubes. Additional peaks have also been observed if sample tubes are left in the tube conditioner for greater than 3 hr.
Furthermore, the peak shapes deviate greatly from a Gaussian shape, specifically for the peaks at approximately 4.6 and 4.825 min. Some instrumentation and data analysis software packages provide a “Symmetry” calculation for each peak in a chromatogram that attempts to quantitate the deviation from a Gaussian shape. This value can be used as an indicator to replace the column and inlet liner of the instrument when it significantly deviates from unity, where unity indicates a perfect Gaussian peak shape. The ECD is very sensitive to nitro-aromatics such as DNT and TNT, but has a limited dynamic range. This results in peaks becoming clipped at the upper bounds of the dynamic range, as seen for the peak at approximately 4.825 min in Figure 3. If peaks become clipped, then it might be necessary to reduce the sampling time for vapor samples during sample collection. Running a new calibration curve before each sample collection series or on a repeated schedule, such as every other night, is a good way to catalog instrument performance and determine when an instrument requires maintenance or service before analyzing valuable samples.
Figure 1. An example chromatogram of the separation of 3,4-DNT (internal standard), TNT, and RDX using the TDS-CIS-GC-ECD instrumentation with the direct liquid deposition method for generating calibration curves of vapor samples. The chromatogram has been trimmed to the relevant portion, but the total run is 8 min long. The 3,4-DNT peak area is relatively constant (1.5 ng) while the TNT and RDX peak areas, and heights, increase with mass of analyte on sample tube: (black) 0.1 ng, (red) 0.5 ng, (green) 1.0 ng, (blue) 2.0 ng, and (orange) 3.0 ng. Please click here to view a larger version of this figure.
Figure 2. An example calibration curve for (■) TNT and () RDX using the direct liquid deposition method with solution standards and a TDS-CIS-GC-ECD instrumentation. The normalized peak area on the Y-axis is obtained from dividing the TNT and RDX peak areas in a chromatogram by the peak area for 3,4-DNT, the internal standard. The error bars represent one standard deviation for the average of three replicate sample tubes per solution standard, or analyte mass.
Figure 3. A chromatogram resulting from poor instrument maintenance, column degradation, and sample tube sorbent material degradation. Additional peaks other than the 3,4-DNT, TNT, and RDX peaks are observed. The peak at approximately 4.825 min is clipped because the analyte mass is at the upper limit of the dynamic range of the detector. A shoulder appears at 4.850 min, indicating poor separation. The baseline, or lower limit, response is elevated causing baseline drift and increase in noise.
Figure 4. A conceptual diagram illustrating the correct procedure for depositing solution onto a sorbent-filled thermal desorption sample tube for the direct liquid deposition method. The micropipette tip should be touching the glass frit and not the side walls of the sample tube. A new tip should also be used for each deposition between analytes and sample tubes.
Figure 5. An alternative to using personal sample pumps for collecting explosive vapors on sorbent-filled thermal desorption sample tubes. Flexible tubing is used to connect the sample tubes to a mass flow controller (MFC) that allows for electronic input of a desired flow rate. The mass flow controllers, when combined with a pump, automatically adjust for flow rate through the sample tubes to a desired set point regardless of variations between sample tubes. A six MFC configuration is shown with a manifold to connect all MFCs to a common pump, but configurations with different numbers of MFCs are possible.
Parameter Name | Value | Units |
TDS Initial Temperature | 25 | ºC |
TDS Final Temperature | 250 | ºC |
TDS Temperature Ramp | 40 | ºC min-1 |
TDS Hold Time | 2 | min |
TDS Flow Rate | 455 | ml min-1 |
TDS Mode | PTV Solvent Vent | N/A |
TDS Transfer Line Temperature | 300 | °C |
CIS Initial Temperature | 0 | °C |
CIS Final Temperature | 250 | °C |
CIS Temperature Ramp | 12 | °C sec-1 |
CIS Hold Time | 3 | min |
CIS Flow Rate | 108 | ml min-1 |
CIS Mode | PTV Solvent Vent | N/A |
Oven Initial Temperature | 30 | °C |
Oven Initial Hold Time | 0.5 | min |
Oven Final Temperature | 250 | °C |
Oven Temperature Ramp 1 | 40 | °C min-1 |
Oven Temperature Hold 1 | 210 | °C |
Oven Temperature Ramp 2 | 40 | °C min-1 |
Oven Temperature Hold 1 | 250 | °C |
Oven Hold Time | 1 | min |
Column Carrier Gas | Helium | N/A |
Column Flow Rate | 5.6 | ml min-1 |
Column Pressure | 23.642 | psi |
Column Coating | 5% polysilioxane (DB5-MS) | N/A |
Column Length | 15 | m |
Column Inner Diameter (ID) | 0.25 | mm |
Column Outer Diameter (OD) | 250 | mm |
ECD Temperature | 275 | °C |
ECD Flow Rate | 60 | ml min-1 |
ECD Carrier Gas | Nitrogen | N/A |
Table 1. The TDS-CIS-GC-ECD instrumentation parameters for quantitation of TNT and RDX vapors using the direct liquid deposition method.
Solution TNT and RDX Concentration (ng μl-1) |
Approximate Vapor Concentration (ppbv) |
Sampling Time (min) |
0.1 TNT/0.25 RDX | 0.050 TNT/0.125 RDX | 120 |
0.4 TNT/1.0 RDX | 0.200 TNT/0.500 RDX | 30 |
2.0 TNT/5.0 RDX | 1.00 TNT/2.50 RDX | 6 |
Table 2. The approximate sampling time for collecting explosive vapors for three solution concentrations of TNT and RDX. Actual sampling times may need to be adjusted to yield peaks in a chromatogram suitable for quantitation.
Reproducibility is a critical attribute for the quantitation of trace explosive vapors using the direct liquid deposition method with TDS-CIS-GC-ECD instrumentation, and Relative Standard Deviation (RSD) is often used as a metric for reproducibility. We have experienced RSDs for inter- and intra-sample reproducibility of approximately 5% for TNT and 10% for RDX. Any RSD above 15% is used as an indicator to check common sources of variation that reduce the effectiveness of the protocol. Sources of variation that have led to unacceptable RSDs in the past are highlighted in the following discussion.
A common source of variation that can lead to relatively large standard deviations for replicate measurements of solution standards and significant deviation from unity for R2 is consistent deposition of solution standards and the internal standard onto sample tubes. We have found an electronic micropipette is ideal for minimizing variation during deposition, as opposed to a manual micropipette. During several recent projects, where multiple personnel were involved in quantitating explosive vapors over several days of sample collections, the source of variation in the results was largely dependent on the individual and his/her usage of the manual micropipette. At first glance, the usage of a manual micropipette appears relatively straightforward, but small variations in plunger depression and release between users yielded a significant source of variation in the quantitative analysis of explosive vapors. When the manual micropipette was switched for an electronic micropipette, the variation between users could not be distinguished between instrument variation and sampling noise. It is also important to hold the micropipette upright during solution uptake and deposition. The solution should be deposited directly onto the glass frit of the sample tube, i.e., the micropipette tip should contact the glass frit and not the glass side walls. A new micropipette tip should also be used for each deposition and sample tube. Figure 4 shows a conceptual drawing of the procedure for depositing the internal standard or a solution standard onto the glass frit of a sample tube.
Another source of variation that can diminish reproducibility with the quantitation of explosive vapors is the vapor sampling procedure. In the protocol, a commercial sample pump is connected to the sample tube and calibrated with a piston flow meter using a small set screw and screwdriver. The flow rate must be adjusted for every sample tube to account for packing differences of the sorbent material between sample tubes and pump performance. This procedure can be cumbersome and error prone, especially when attempting to collect multiple sample tubes in parallel. Similar to the replacement of a manual micropipette with an electronic micropipette to reduce variation, we have also implemented an electronic sample tube system that uses a vacuum pump and mass flow controllers (MFC). Figure 5 shows a conceptual diagram of a six sample tube vapor collection manifold. The MFCs automatically correct for variations in the packing between sample tubes and automatically adjust the flow rate to 100 ml min-1 without user intervention. The flow rates should still be routinely checked and calibrated with a piston flow meter, but the flow rate can be adjusted electronically rather than manually with a screwdriver. It is possible to create a single MFC sample tube collection manifold, but the MFC-based configuration seen in Figure 5 is meant to demonstrate the scalability of the alternative method. Notably, the individual commercial sample pumps are less expensive than the MFC-based configuration and the MFC-based configuration is a custom assembly, but the MFC-based configuration can reduce variation, improve reproducibility, and be easier to use.
Variation is also present in the TDS-CIS-GC-ECD instrumentation. Over time, as the various internal components of the instrument heat and cool during analysis, parts will expand and contract causing fatigue of consumables, such as ferrules, nuts, columns, and liners. The gradual fatigue of components is unavoidable and a source of variation over time. When performing quantitation of trace (sub parts-per-million by volume, ppmv) explosive vapors, the gradual variation in instrument performance becomes amplified. Thus, it is important to establish calibration curves for quantitation in a timely fashion, typically before analysis of samples. If possible, calibration curves should be generated the same day as the sample analysis to be conducted. This is not always possible due to time constraints and instrument workload. Furthermore, typically at least five replicates are used per mass, or concentration, for a calibration curve because more replicates yield a more robust calibration curve for quantitation. However, the calibration curve in Figure 2 was established with three replicates. The number of replicates was reduced so a full calibration curve over the entire dynamic range of the detector could be established O/N in a single autosampler tray (two blanks, eighteen solution standard sample tubes, and twenty sample tube capacity). To compensate for the reduced number of replicates per analyte mass, a new calibration curve was established O/N with samples run immediately the following day to account for variation associated with instrument drift and prevent a back log of sample tubes for quantitation of explosives vapors with TDS-CIS-GC-ECD, which is approximately 20 min per sample tube.
Determining the peak areas for 3,4-DNT, TNT, and RDX in a chromatogram, such as the example seen in Figure 1, can be a subjective process that can introduce irreproducibility to the quantitation of explosive vapors with direct liquid deposition method and TDS-CIS-GC-ECD instrumentation. Many data analysis software packages supplied with GC-ECD instrumentation include manual and automatic peak detection and integration methods. The field of chromatographic data analysis and its related techniques for automatic peak detection and integration is long and extensive,27–31 with many of the algorithms provided in data analysis software. A complete review of the various characteristics and procedures for integrating peak areas is beyond the scope of this work. It is more important for a research group to standardize, document, and use the same procedure for the calibration curve as the samples to minimize variation in quantitation of explosive vapors from more subjective peak area integration processes.
Finally, degradation of sample tubes and solution standards can affect quantitation of trace explosive vapors. Similar to component fatigue from use and thermal cycling of the TDS-CIS-GC-ECD instrumentation, the sorbent material in the sample tubes can degrade over time with repeated sampling and thermal desorption. New sample tubes are tightly packed and white in color. Over time as the sample air is flowed in one direction and carrier gas, typically helium, flowed in the opposite direction during thermal desorption, the packing of the sample tubes becomes loose and yellow in color. The yellow color indicates degradation of the sorbent material from repeated thermal cycling within the instrumentation and tube conditioner. Additional peaks in blanks and sample tubes are also indicators of sorbent material degradation products.26 After each analysis, sample vapors are conditioned in a tube conditioner for a maximum of 3 hr. This is to desorb any remaining material from the sorbent material and effectively clean the sample tubes. However, leaving a sample tube in the conditioner at relatively high temperature (300 °C) longer than 3 hr can significantly shorten the lifetime of the sample tube and introduce variation in quantitation. Similarly, solution standards will degrade over time, which will artificially reduce the mass, or concentration, of analyte in each standard for a calibration curve. To minimize solution degradation, the solution standards should be stored in an amber glass vial in a freezer or refrigerator and the solution standards should be periodically analyzed using the GC-ECD without the TDS-CIS and a standard split/splitless inlet to identify additional peaks or degradation products. A complementary quantitation method, such as gas chromatography mass spectrometry or high performance liquid chromatography, can also be used to ensure the solution standards have not degraded and are suitable for the direct liquid deposition method for quantitating trace explosive vapors.
The authors have nothing to disclose.
Financial support was provided by the Department of Homeland Security Science and Technology Directorate.
Name of Material/ Equipment | Company | Catalog Number | Comments/Description |
2,4,6-Trinitrotoluene (TNT) | Accu-Standard | M-8330-11-A-10X | 10,000 ng μL-1 |
Cyclotrimethylenetrinitramine (RDX) | Accu-Standard | M-8330-05-A-10X | 10,000 ng μL-1 |
3,4-Dinitrotoluene (3,4-DNT) | Accu-Standard | S-22988-01 | 1000 ng μL-1 |
Tenax® TA Vapor Sample Tubes | Gerstel | 009947-000-00 | Tenax® 60/80 |
CIS4 Liner | Gerstel | 014652-005-00 | |
Transfer Line Ferrule | Gerstel | 001805-008-00 | |
Inlet Liner Ferrule | Gerstel | 001805-040-00 | |
CIS4 Ferrule | Gerstel | 007541-010-00 | |
ECD Detector Ferrule | Aglient | 5181-3323 | |
DB5-MS Column | Res-Tek | 12620 |