On-Site Detection of Explosives in Soil and Water Samples Using an Electronic Nose

The detection of explosives in soil and water is a substantial problem for the U.S. government and the populous at large. Huge quantities of explosives have been dumped onto military bases for the past century both as waste from the manufacture of military ordinance and as expended ammunition.

The cleanup job is massive and any new technology which can accelerate the remediation or characterization of these sites will have a big impact on the quality of these sites and the adjacent communities in the future. The U.S.

EPA has commenced the task of assessing new technology for use in environmental areas of interest to both the state, federal and local governments through a program called the Environmental Technology Verification (ETV) Program. Electronic Sensor Technology (EST) has taken part in two of these programs in the last few years. The first was the discovery of polychlorinated biphenyls (PCBs) [1] in soils and the second was the wellhead monitoring of volatile organic compounds (VOCs) in ground water [2].

This multipurpose technology is concisely explained, and the approaches for detection of explosives in water and soil are presented in this article.

GC/SAW nozzle interface showing interaction of column and acoustic cavity

Figure 1. GC/SAW nozzle interface showing interaction of column and acoustic cavity

The SAW Detector

Crucial research has been conducted with chemical coatings applied to Surface Acoustic Wave (SAW) crystals. A typical approach is to expose a collection of SAW crystals with varied polymer coatings to the vapor to be characterized.

In theory each polymer coating will adsorb the vapors differently and by comparing reaction patterns from the collection of sensing crystals, identification can be realized. However, polymer coatings decrease the sensitivity of the SAW crystal and restrict detection to nanogram levels. Additional loss in sensitivity happens because the collected vapor sample has to be split between several sensing crystals.

The lack of specificity of polymer coatings means that in general each coated crystal reaction overlaps the reaction of other crystals to a certain extent and in this case pattern recognition with over-lapping reactions is very hard. Coated crystals also suffer from extended analysis times caused by the need for the analyte to diffuse into and out of the coating.

A unique type of SAW vapor detector with picogram sensitivity and which does not apply polymer coatings was created [3] to resolve the issues of low sensitivity and slow response. As a detector for use in a high speed gas chromatograph, the SAW detector is perfect. The detector has zero dead volume that increases its reaction to low levels of transient vapors.

The detection area is about the same size as the inner diameter of a capillary column so that all the column effluent can be collected onto the sensor. The sensing crystal contains a very high Q SAW resonator positioned close to a small thermoelectric cooling element. A thermoelectric element offers the precise control of cooling required for vapor adsorption and concurrently the ability to clean the crystal using thermal desorption when required.

The focused SAW resonator sensing element offers part-per-trillion sensitivity for semi-volatile compounds and part-per-billion sensitivity for volatile organics. The crystal functions by maintaining very focused and resonant surface acoustic waves at 500 MHz on the face of a single crystal quartz chip. By focusing the vapor through a micro-nozzle as illustrated in Figure 1, picogram sensitivity can be realized.

This result [4] is 1000 times lesser than SAW crystals coated with polymers. Since the crystal is produced from single crystal quartz without polymer coatings, both precision and long term stability are attained over a wide temperature range.

GC/SAW Fast Chromatography System

By integrating SAW detectors with high speed temperature programmed chromatographic columns, specificity over a broad range of vapors at the part-per-billion level in near real time (<1 minute) has been realized [5]. The GC/SAW provides the benefits of a cost-effective solid state detector and the specificity of a temperature programmed GC column.

Figure 2 shows the key elements of a GC/SAW vapor detection system. The analysis is done in two steps corresponding to the two positions a GC valve. In the sample position vapor to be analyzed passes into a heated inlet where it is adsorbed onto sample preconcentrator loop trap. The trap holds absorbent specific to the chosen analyte (e.g. Tenax) or for higher molecular weight compounds may be an open metal tube. Choice of sample time and flow rate fixes the total quantity of airborne vapors collected in the loop trap.

The GC valve is rotated to an inject position and the loop trap is quickly heated by a capacitive discharge which causes captured vapors to be moved to the GC column via helium carrier gas. These vapors recondense on the inlet of a chromatographic column contained primarily at low temperature. A microprocessor then applies a linear temperature ramped heating profile to the GC column.

The column divides the injected compounds so they are eluted at various times where they then condense on the SAW crystal and are detected as frequency alters. The frequency is sampled at high speed and digitized as separate time slices. The display of the derivative of these slices versus time produces a traditional chromatogram. However, each time slice is autonomous from those in proximity to it and can be seen as a separate sensing element.

These elements gathered together make up a virtual array. When the system is used in this way, it is same as an Electronic Nose with numerous sensors. It produces an output that is a visual representation of the sum of all the sensors known as a Vaporprint®.

GC/SAW Detection System

Figure 2. GC/SAW Detection System

The Model 7100 GC/SAW system is used in this research (Figure 3). A laptop computer (not shown) offers a completely integrated user interface in a Windows 95 operating environment. The system was used to assess samples of water and soil as a prescreening assessment to participate in a U.S. EPA sponsored ETV program.

The positive results shown here resulted in an invitation to take part in a field study at Tennessee. Former field testing of similar systems has shown the ability to detect a broad range of compounds including explosives, drugs, polychlorinated biphenyls, volatile organics and dioxins [6,7].

The new technology’s part-per-billion (picogram) sensitivity and field performance has been validated by the Office of National Drug Control, the Department of Energy and the U.S. Environmental Protection Agency (EPA-ETV).

The 7100 Vapor Analysis System

Figure 3. The 7100 Vapor Analysis System


Testing was conducted on four explosives in water and soil. The explosives were shipped, stored on ice, straight from the ETV program sponsors. Six samples each in both matrices were assessed. The samples comprised of a performance evaluation sample, a blank, and four real world samples from two government owned facilities.

The water samples were collected from the Volunteer Army Ammunition Plant in Chattanooga, Tennessee and the soil samples were taken from the Louisiana Army Ammunition Plant in Shreveport, Louisiana. Calibration materials were acquired from AccuStandard (New Haven, Connecticut) as low concentration of individual compounds in acetone. The rest of the solvents were reagent grade.

Testing Methods


Calibration was done on the tool to regulate the response factor for each of the expected explosive materials. Dilutions were made by injecting a variety of volumes of the standards into 1 ml of acetone. The compounds were then injected as 1 ml aliquots straight into the instrument inlet during a 30-second sampling phase. A typical chromatogram of a mixture of the four explosives injected directly into the inlet of the 7100 is shown in Figure 4.

The injected levels ranged from 300 to 1200 pg. The column was ramped from 50 to 170 °C at 18 °C/sec and the detector was maintained at 30 °C during the analysis. Sampling time was 30 seconds and the same conditions were sustained for the whole study. The data analysis for each attempt was finished in <1 minute with the last peak of interest eluting in less than 5 seconds. Data (Table 1) for the direct injection of the explosives TNT, RDX, 2,6 DNT and 2,4 DNT into the 7100.

The plotted results of the calibration of the system are illustrated in Figure 5. TNT and 2,6 DNT display linear response over the range predicted for the unidentified samples. 2,4-DNT displays saturation of the trap at higher concentrations. This is because of its higher volatility and the very small breakthrough volume of this compound in the trap used in the tool.

The decreasing reaction of RDX with lower concentration is perhaps due to loss of material on cold surfaces at the low concentrations.

Soil Analysis

Soil was tested by weighing 1 gm of soil into a 2 ml vial. One ml of acetone was put into the vial and the mixture was strongly shaken for three minutes and then centrifuged. An established amount of liquid was then extracted and measured into a new vial where it was diluted up to 1000:1 with acetone.

Then one micro liter of diluted liquid was injected into the inlet of the 7100. Inlet temperature was kept at 200 °C. Analyte recovery for the soil technique was examined by spiking 1 gm of soil with 1 micro gram of TNT. The TNT was extracted as explained above and then 1 micro liter (1 ng TNT) was injected. Average recovery over three samples was 109% with an RSD of 16% which was considered satisfactory.

The method detection limits (MDLs) for the soil technique were 488 mg/kg (ppbm) for 2,4-DNT, 15 mg/kg for TNT, 463 mg/ kg for 2,6-DNT, and 200 mg/kg for RDX. These values were well within the expected requirements of the ETV program. The MDLs of the two DNT isomers could be enhanced by using a different trap. The RDX MDL is restricted by the loss of the compound at low levels of concentration.

Typical chromatogram of injection of explosives. Upper race is the derivative of the frequency data shown on the lower trace

Figure 4. Typical chromatogram of injection of explosives. Upper race is the derivative of the frequency data shown on the lower trace

Table 1. Summary of the calibration data for the direct injection of explosives

Reference 24-DNT 26-DNT TNT RDX
'99061610.445 1321 1784 2561 1675 TNT/others
'99061610.481 1131 1512 1910 1372 TNT/others
'99061610.514 1304 1668 2123 936 TNT/others
'99061610.550 1309 1680 1983 1354 TNT/others
'99061610.585 1233 1476 1588 1452 TNT/others
average 1259.6 1624 2033 1357.8
rsd 6.3 7.9 17.4 19.7
'99061611.020 1871 2496 3882 3481 TNT/others
'99061611.053 2011 3010 4355 3039 TNT/others
'99061611.091 1820 2447 3743 3303 TNT/others
'99061611.324 1676 2595 4283 2894 TNT/others
'99061611.355 1765 2562 3296 2469 TNT/others
average 1828.6 2622 3911.8 3037.2
rsd 6.8 8.6 11.0 12.9
'99061611.422 2719 5949 9998 10310 TNT/others
'99061611.455 2543 6450 9496 10648 TNT/others
'99061611.485 2976 8882 11212 13547 TNT/others
'99061611.523 2761 5708 9178 8060 TNT/others
'99061611.555 2536 6092 11292 11636 TNT/others
average 2707 6616.2 10235.2 10840.2
rsd 6.7 19.6 9.5 18.5

Calibration Curves for four explosives. The x-axis is the number of pg injected, the y-axis is the signal from the instrument in Hz.

Figure 5. Calibration Curves for four explosives. The x-axis is the number of pg injected, the y-axis is the signal from the instrument in Hz.

Water Analysis

Water was examined using Supelco (Bellefonte, Pennsylvania) Sep-Pak RDX SPE cartridges. Each cartridge was conditioned by running 15 ml of acetonitrile followed by 30 ml reagent grade water through it at 10 ml/minute. 100 ml of the water sample was then drawn via the cartridge at the same rate using a modifiable valve and a vacuum system to sustain flow. Then the cartridge was flushed with 5 ml of acetonitrile resulting in a 20:1 concentration step.

One micro liter of the extract was injected into the 7100. Analyte recovery for the water technique was analyzed by spiking 100 ml of H2O with 40 mg of TNT and then extracting the explosive using the above technique. Recovery surpassed 100% in two out of three trials with the failure due to an error in the extraction procedure. These results were again considered satisfactory to examine the unknown samples.

The MDLs for the water technique in mg/l (ppbv) were 23 for 2,6 DNT 0.75 for TNT, 24.4 for 2,4-DNT, and 10 for RDX. Method detection limits for the water were restricted by the 100 ml initial sample size. Using 500 ml samples as recommended in the Supelco literature would have provided a larger concentration and improved the MDLs by a factor of 5.


The ETV program team submitted six samples of water and six samples of soil and for analysis. The samples were measured twice and the results then returned to the EPA for scoring.

The returned scores were matched against a reference laboratory result and an anticipated result. The expected result was known either because of spiking the sample with a known quantity of explosive (PE samples) or by former measurement of the levels of explosives in these or similar soils. Each set of samples was transferred to a reference laboratory for testing using HPLC Method 8030.

Table II illustrates the data returned from the ETV Program team after accumulating the submitted results. The table is split into four areas, RDX and TNT in both water and soil. No DNT isomers were discovered in EST’s data or the reference lab results.

The acceptance range is bracketed to display a range of acceptable results. In a number of cases the results attained by the reference laboratory were not within the predicted range, signifying that there is some issue with the comparison methodology or the reference lab. The results do reveal that for RDX and TNT in soil, the 7100 analysis was either within or very near to the predicted result. The results for water were also thought to be relatively good for the TNT data.

The RDX results were consistently low and reported as below the detection limit. Some of the samples were well above the MDL of 10 mg/liter for RDX and yet were not noticed. The extraction of RDX from Sep-Pak cartridges will be assessed additionally during future testing.

Soil Results for RDX (mg/kg)
Reference laboratory 3.2 110 3300 4400 4160 3700
EST <DL 68 3413 3190 4321 4129
Soil Results for TNT (mg/kg)
Reference laboratory <0.5 40 94 100 178 220
EST <DL 31 201 100 133 138
Water Results for RDX (µ g/l)
Spike #1
Spike #2
No Info
No Info
No Info
Reference laboratory <0.5 10 54 DNA DNA 640
Water Results for TNT (µ g/l)
Spike #1
Spike #2
Reference laboratory <DL 18 73 DNA DNA 14900
EST <DL 13 50 100 3075 16225

DL = Detection Limit
DNA = Did not analyze


The 7100 model is a quantitative tool for the detection of vapors in many different matrices. However, the qualitative nature of an electronic nose is at times beneficial when a measurement of the difference between two samples is required but the precise nature of the difference is not completely understood.

The VaporPrint is a visual way of showcasing chromatographic information so that the human eye can more easily partake in the “pattern recognition” process. Figure 6 is a derivative VaporPrint of the explosive mix illustrated in Figure 4. The unique nature of this display is exposed to the relative concentrations of the numerous components constituting the mix. However, for certain compounds, the relative distribution of the components is fixed and the resulting VaporPrint is unique. As a comparison, a frequency VaporPrint of gasoline is illustrated in Figure 7.

The dual nature of the 7100 as both a precision high speed gas chromatograph and a multisensor electronic nose makes it a unique tool for the study of vapors.

Derivative VaporPrint of the explosive mix shown in Figure 4

Figure 6. Derivative VaporPrint of the explosive mix shown in Figure 4.

Frequency VaporPrint of a headspace sample of gasoline vapor

Figure 7. Frequency VaporPrint of a headspace sample of gasoline vapor.

Summary and Conclusions

The features of a semi-portable tool for the detection of vapors in water, soil, and air have been explained. The use of the tool for the detection of explosive has revealed that the Model 7100 Vapor Analysis System can be used for quick screening of water and soil for explosives at the mg/kg and water at the mg/l concentration level. Moreover, the tool was used in a blind test sponsored by the U.S. EPA ETV program and was qualified as a suitable technology to partake in a full field study of explosives in water and soil.

The tool can achieve an analysis in <1 minute after injection of the sample. The rapid response of the tool also qualifies it as an electronic nose with the benefit that it has stable sensors that do not drift and do not need continuous calibration.


1 Dindal, Amy B., Bayne, Charles K. and Jenkins, Roger, A. - Oak Ridge National Laboratory and Billets, Steven and Coglin, Eric N. – U.S. EPA, “Environmental Technology Verification Report - Measurement of PCBs in Soils and Solvent Extracts,” August 1998.

2 Einfeld, Wayne, Sandia National laboratory and Billets, Steven and Coglin, Eric N. – U.S. EPA, “Environmental Technology Verification Report - Measurement of Chlorinated Volatile Organic Compounds in Water,” November 1998.

3 United States Patent No. 5,289,715, Vapor Detection Apparatus and Method Using an Acoustic Interferometer.

4 E. Staples, G. Watson, and W. Horton, “Spectral Density of Frequency Fluctuations in SAW Sensors,” 186th Meeting of the ElectroChemical Society, Miami Beach, Florida, October 9-14, 1994.

5 Edward J. Staples and Gary W. Watson , “GC/SAW Non-Intrusive Inspection System”, White House Conference, Office of National Drug Control Policy, New Hampshire, October 1995.

6 G.W. Watson and E.J. Staples, “SAW Resonators as Vapor Sensors,” Proceedings of the 1990 Ultrasonics Symposium, pp.311-314, 90CH2938-9

7 G.W. Watson, W. Horton, and E.J. Staples, “GAS Chromatography Utilizing SAW Sensors,” Proceedings of the 1991 Ultrasonics Symposium, pp.305-309.

This information has been sourced, reviewed and adapted from materials provided by Electronic Sensor Technology.

For more information on this source, please visit Electronic Sensor Technology.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this article?

Leave your feedback