Talanta 54 (2001) 427– 438

www.elsevier.com/locate/talanta

Determination of nitroaromatic, nitramine, and nitrate ester

explosives in soil by gas chromatography and an electron
capture detector
Marianne E. Walsh *
US Army Cold Regions Research and Engineering Laboratory, 72 Lyme Road, Hano6er, NH 03755 -1290, USA

Abstract

Hazardous waste site characterization, forensic investigations, and land mine detection are scenarios where soils
may be collected and analyzed for traces of nitroaromatic, nitramine, and nitrate ester explosives. These thermally
labile analytes are traditionally determined by high-performance liquid chromatography (HPLC); however, commer-
cially available deactivated injection port liners and wide-bore capillary columns have made routine analysis by gas
chromatography (GC) possible. The electron-withdrawing nitro group common to each of these explosives makes the
electron capture detector (ECD) suitable for determination of low concentrations of explosives in soil, water, and air.
GC-ECD and HPLC-UV concentration estimates of explosives residues in field-contaminated soils from hazardous
waste sites were compared, and correlation (r \0.97) was excellent between the two methods of analysis for each of
the compounds most frequently detected: 2,4,6-trinitrotoluene (TNT), hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX),
2,4-dinitrotoluene (2,4-DNT), 1,3-dinitrobenzene (1,3-DNB), 1,3,5-trinitrobenzene (TNB), and octahydro-1,3,5,7-te-
tranitro-1,3,5,7-tetrazocine (HMX). The analytes were extracted from soils with acetonitrile by 18 h of sonication in
a cooled ultrasonic bath. Two soil-to-solvent ratios were evaluated: 2.00 g:10.00 ml and 25.0 g:50.0 ml. GC-ECD
method detection limits were similar for the two soil-to-solvent ratios and were about 1 mg kg − 1 for the di- and
trinitroaromatics, about 10 mg kg − 1 for the mono-nitroaromatics, 3 mg kg − 1 for RDX, 25 mg kg − 1 for HMX, and
between 10 and 40 mg kg − 1 for the nitrate esters (nitroglycerine [NG] and pentaerythritol tetranitrate [PETN]). Spike
recovery studies revealed artifacts introduced by the spiking procedure. Recoveries were low in some soils if the
amount of soil spiked was large (25.0 g) compared to the volume of spike solution added (1.00 ml). Recoveries were
close to 100% when 2.00-g soil samples were spiked with 1.00 ml of solution. Analytes most frequently found in soils
collected near buried land mines were the microbial transformation products of TNT (2-amino-4,6-dinitrotoluene
[2-Am-DNT] and 4-amino-2,6-dinitrotoluene [4-Am-DNT]), manufacturing impurities of TNT (2,4-DNT, 2,6-DNT,
and 1,3-DNB), and TNT. The microbial reduction products of the isomers of DNT and of 1,3-DNB were also
detected, but the ECD response to these compounds is poor. © 2001 Elsevier Science B.V. All rights reserved.

Keywords: Explosives; Soil; Compounds

* Tel.: +1-603-646-4666; fax: + 1-603-646-4785.

E-mail address: marianne@crrel.usace.army.mil (M.E. Walsh).

0039-9140/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 3 9 - 9 1 4 0 ( 0 0 ) 0 0 5 4 1 - 5
428 M.E. Walsh / Talanta 54 (2001) 427–438

1. Introduction are collected at the detector anode and produce a

small background or reference current. When
Traditionally, determination of explosives in electron-capturing analytes (halogenated, nitro-
soil served either forensic or hazardous waste genated, or oxygenated compounds) elute from
investigations. More recently, there is interest in the chromatographic column, they form nega-
ultrasensitive methods to determine the chemical tively charged ions, which are swept out of the
signature associated with buried land mines [1]. detector. Concurrently, the voltage across the cell
Gas chromatography (GC), a valuable analytical electrodes is pulsed to collect the remaining free
technique for most organic environmental con- electrons. The pulse rate changes to maintain a
taminants, has not been used routinely for quanti- constant cell current; the change in pulse fre-
tative determination of explosives residues in soil. quency is proportional to the analyte concentra-
However, these thermally labile analytes may be tion [10]. For di- and trinitroaromatic explosives,
accurately determined by gas chromatography by ECD detection limits of 1 pg have been reported
using direct injection into a deactivated liner and [11].
a short (6-m) wide-bore capillary column [2]. Re- To evaluate the performance of the ECD for
cently an analytical method for determination of the analysis of explosives residues in soils, concen-
explosives in drinking water was developed that tration estimates obtained by GC-ECD were com-
was based on solid phase extraction (SPE) and pared to those obtained by the standard method.
determination by GC with electron capture detec- Jenkins et al. [12] developed the standard method
tion (ECD) [3]. This paper describes the analysis for explosives in soil (SW-846 Method 8330 [13])
of soil extracts using gas chromatographic condi- to characterize military sites contaminated with
tions similar to those used to analyze water explosives residues from the production or use of
extracts. high-explosives munitions. For this standard ana-
GC detectors suitable for determination of ex- lytical method, a 2.00-g soil sample is extracted by
plosives were reviewed by Yinon and Zitrin [4] 18 h of sonication with 10.00 ml of acetonitrile
and include the flame-ionization detector (FID), (AcN). The AcN extract is mixed 1:1 (v/v) with
mass spectrometer (MS), nitrogen-phosphorus de- aqueous calcium chloride to flocculate fines prior
tector (NPD), electron capture detector (ECD), to filtration and analysis by high-performance
and thermal energy analyzer (TEA). Selectivity liquid chromatography with an ultraviolet detec-
and sensitivity are achieved with those detectors tor (HPLC-UV). Explosives concentrations of 1.0
responsive to the nitro group common to ni- mg kg − 1 (1.0 ppm) or higher may be determined
troaromatic, nitramine, and nitrate ester explo- using this procedure, and detection limits are
sives. The most selective detector for explosives is sufficiently low for human health or ecological
the TEA, which detects only compounds that risk assessments. Jenkins et al. [12] chose HPLC-
produce NO or NO2. The NPD is slightly less UV rather then GC for several reasons: compati-
selective than the TEA, but insensitive to nitrate bility of the thermally labile analytes with room
esters [5]. The ECD, widely used for the determi- temperature chromatographic separation, large
nation of halogenated compounds, is less selec- linear range of the detector, ruggedness of the
tive, but is more sensitive for nitroaromatics than method, ability to analyze high concentration (\
the TEA [6] or NPD [7,8]. 40 mg l − 1) water samples by direct injection, and
The ECD was chosen because of its routine use compatibility of the solvent (acetonitrile) used to
in environmental laboratories and its compatibil- extract soils with reversed-phase HPLC.
ity with acetonitrile, the preferred solvent for ex- In the 1970s, Jenkins, Leggett, and Murrmann
traction of explosives from soils [9]. The ECD used GC-ECD when they characterized the va-
contains a source of beta particles (63Ni) that pors from military-grade TNT in conjunction
collide with the detector make-up gas (nitrogen or with efforts to detect buried land mines by sam-
argon/methane) to produce many low energy elec- pling the atmosphere [14 –16]. Some solvent (ben-
trons per beta particle. These low energy electrons zene) extracts of soil were analyzed as well.
 M.E. Walsh / Talanta 54 (2001) 427–438 429

Instrumentation at that time was not conducive to concentrations in soils from antitank firing ranges
quantitative determination of explosives in soil, where octol-filled (70% HNIX:30% TNT) rockets
especially on a routine basis. Improvements in have been fired [18,19]. For land mine detection,
injection port liners, GC columns, and most re- the analytes of interest were predicted to be the
cently the ECD detector [11] have led us to reex- constituents of TNT vapor, principally the iso-
amine the utility of the GC-ECD for mers of DNT, DNB and TNT [20].
determination of explosives in soil for both haz-
ardous waste site characterization and land mine
detection. 2. Experimental methods
Analytes of interest differ somewhat for haz-
ardous waste characterization and land mine de- 2.1. Matrices
tection (Table 1). Soils that were contaminated by
the manufacture or use of explosives most likely Field-contaminated soils were from Iowa AAP
contain TNT and RDX [17], the most common (Army Ammunition Plant), Milan AAP (TN),
found of the military-grade explosives. Co-con- Nebraska Ordnance Plant, Monite (NV), Eagle
taminants such as manufacturing by-products and River Flats OB/OD (Open Burning/Open Detona-
biodegradation products may also be present. tion) (AK), Raritan Arsenal (NJ), Savanna Army
More recently, HMX has been found at high Depot (IL), Chickasaw Ordnance Works (TN),

Table 1
Analytes of interest for two applications of analytical method for explosives in soil: hazardous waste characterization and mine
detection

Analyte Class Abbreviation CAS numbera Hazardous waste Mine detection

characterization

Octahydro-1,3,5,7-tetranitro-1,3,5,7- Nitramine HMX 2691-41-0

tetrazocine
Hexahydro-1,3,5-trinitro-1,3,5-triazi Nitramine RDX 121-82-4

ne
1,3,5-Trinitrobenzene Nitroaromatic TNB 99-35-4

1,4-Dinitrobenzene Nitroaromatic 1,4-DNB 100-25-4

1,3-Dinitrobenzene Nitroaromatic 1,3-DNB 99-65-0

1,2-Dinitrobenzene Nitroaromatic 1,2-DNB 528-29-0

2,4,6-Trinitrophenylmethylnitramine Nitroaromatic/nit Tetryl 479-45-8

ramine
Nitrobenzene Nitroaromatic NB 98-95-3

2,4,6-Trinitrotoluene Nitroaromatic 2,4,6-TNT 118-96-7

4-Amino-2,6-dinitrotoluene Amino-nitroarom 4-Am-DNT 1946-51-0

atic
2-Amino-4,6-dinitrotoluene Amino-nitroarom 2-Am-DNT 355-72-78-2

atic
2,4-Dinitrotoluene Nitroaromatic 2,4-DNT 121-14-2

2,6-Dinitrotoluene Nitroaromatic 2,6-DNT 606-20-2

2-Nitrotoluene Nitroaromatic 2-NT 88-72-2

3-Nitrotoluene Nitroaromatic 3-NT 99-08-1

4-Nitrotoluene Nitroaromatic 4-NT 99-99-0

3,5-Dinitroaniline Amino-nitroarom 3,5-DNA 618-87-1

atic
Nitroglycerine Nitrate ester NG 55-63-0

Pentaerythritoltetranitrate Nitrate ester PETN 78-11-5

a
Chemical abstract service registry number.
430 M.E. Walsh / Talanta 54 (2001) 427–438

US Naval Ammunition Depot (GA), Camp higher soil-to-solvent ratio (25.0 g soil:50.0 ml
Shelby (MS), Fort Ord (CA), CFB-Valcartier acetonitrile). For each soil, duplicate 25.0-g sub-
(Québec), and Fort Leonard Wood (MO). samples were extracted with 50.0 ml acetonitrile in
Blank matrices were Ottawa sand, an Army a cooled sonic bath for 18 h. If enough soil was
Environmental Center standard soil obtained available, matrix spikes (MS) and matrix spike
from Rocky Mountain Arsenal (CO), a soil from duplicates (MSD) were also prepared and ex-
Fort Leonard Wood, and a silt obtained locally in tracted. Soils (25.0-g) were spiked with 5.00 ml of
Hanover, NH. a spike solution (50.0 mg l − 1 nitroaromatics and
200 mg l − 1 RDX) and left to air-dry for 24 h in a
2.2. Calibration fume hood. The spike solution contained the ana-
lytes of interest for mine detection (Table 1). The
Standards were prepared from standard analyt- spiked concentration in soil was 10.0 mg kg − 1 for
ical reference material (SARM) from the US nitroaromatics and 40.0 mg kg − 1 for RDX. All
Army Environmental Center, Aberdeen Proving samples were extracted with 50.0 ml acetonitrile
Ground, MD, or obtained commercially from containing 3,4-DNT (25.0 mg l − 1) as an internal
Supelco (Bellefonte, PA) and Restek (Bellefonte, standard. These samples were also extracted for
PA). All solutions were prepared in acetonitrile 18 h in a cooled sonic bath. Prior to GC-ECD
(Burdick and Jackson, Muskegon, MI). Calibra- analysis, extracts were filtered through Millex SR
tion standards were prepared fresh each day over filter units. Prior to HPLC-UV analysis, 0.50 ml
the range 0.40 to 100 mg l − 1 from 10.0-mg l − 1 of each filtered acetonitrile extract was mixed with
combined stock solutions that were stored at 2.00 ml reagent-grade water (MilliQ).
− 22°C in the dark. Soils collected from an experimental mine field
at Fort Leonard Wood were extracted without
2.3. Extraction air-drying using 2.00 g soil:5.00 ml acetonitrile
and/or 20.0 g soil:50.0 ml acetonitrile.
Archived field-contaminated soils were chosen
based on previous HPLC analysis that indicated 2.4. Method detection limits and spike reco6ery
the presence of several of the Method 8330 ana-
lytes over a wide concentration range (B 1.0 to Method detection limits and spike recoveries
50000 mg kg − 1). Following the soil extraction were determined using the two soil:solvent ratios
procedure specified by Method 8330, 2.00-g soil used for the extraction of field samples. Following
subsamples were extracted with 10.00 ml acetoni- the Method 8330 protocol, seven replicate 2.00-g
trile (with no internal standard) for 18 h in a soil samples (either Ottawa sand or AEC standard
cooled sonic bath. To compare concentration esti- soil) were spiked with 1.00 ml of 10.0- or 100-mg
mates obtained by GC to those obtained by l − 1 spike solutions to yield 5.00- and 50.0-mg
HPLC-UV, the extracts were split. For GC-ECD kg − 1 samples containing the analytes of interest
analysis, a portion of each acetonitrile extract was for hazardous waste site characterization (Table
filtered through a Millex SR filter unit (Millipore, 1). After 1 h, 9.00 ml of acetonitrile was added
Bedford, MA) prior to injection into the GC. For and the samples extracted for 18 h in a cooled
HPLC-UV analysis, another aliquot of each ace- sonic bath. At 2 h into the sonication period, a
tonitrile extract was mixed with an equal volume small aliquot of the extract was taken for analysis
of aqueous calcium chloride to flocculate fine soil to determine the stability of NG and PETN in a
particles, and the mixture filtered through a cooled sonic bath.
Millex SR filter unit prior to injection into the A second set of Ottawa sand samples was
HPLC. spiked by adding either 1.00 or 5.00 ml of a
Additional archived soils that had trace ana- 50.0-mg l − 1 solution to 25.0-g soil samples to yield
lytes (less than detection limit as determined by 2.00- and 10.0-mg kg − 1 samples for the nitroaro-
previous HPLC analyses) were extracted using a matics of interest for mine detection (Table 1).
 M.E. Walsh / Talanta 54 (2001) 427–438 431

The spike solution also contained RDX at a LC-CN) column eluted with either 1.5 ml per min
concentration four times greater than the ni- 1:1 methanol:water or 1.2 ml per min 12:13:62
troaromatics. The samples spiked with 1.00 ml methanol:acetonitrile:water. Injection volume for
were aged uncapped for 1 h prior to extraction. each separation was 100 ml. Following these
The samples spiked with 5.00 ml were aged 24 h HPLC separations, absorbance was recorded at
uncapped. All samples were extracted with 50.0 254 nm on a Spectra Physics Spectra 100 variable
ml, acetonitrile containing 3,4-DNT (25.0 mg l − 1) wavelength UV detector.
as an internal standard. These samples were also
extracted for 18 h in a cooled sonic bath.
3. Results
2.5. Instrumentation
3.1. Field-contaminated soils: concentration
2.5.1. GC-ECD estimates by GC-ECD and HPLC-UV
Initially, an HP 5890 equipped with an
Ni63ECD was used. Later, an HP 6890 equipped 3.1.1. Wide concentration range
with a micro cell Ni63ECD was used. For both To test the feasibility of using GC-ECD for the
GCs, direct 1.0-ml, injections at 250°C were used. analysis of soil extracts, 24 archived field-contam-
The injection port liner was a deactivated Restek inated soils were chosen that contained several of
Uniliner. The analytical columns were 6.0-m× the analytes of interest over a wide range of
0.53-mm ID fused-silica, 1.5-mm film thickness of concentration (based on previous HPLC-UV
either 100% polydimethylsiloxane (J and W DB-1) analysis) and a variety of sites across North
or (5% phenyl)methylsiloxane (HP-5). The GC America. Because some of these soils were ana-
oven was temperature programmed as follows: lyzed by HPLC-UV over a decade ago, another
100°C for 2 min, 10°C per min ramp to 200°C, HPLC-UV analysis was performed at the same
20°C per min ramp to 250°C, hold 5 min. The time as the GC-ECD analysis, and extract splits
carrier gas was hydrogen or helium at 12–15 ml were used from 2.00-g soil aliquots extracted with
per min. The makeup was nitrogen (30 – 40 ml per 10.0 ml acetonitrile. The standard HPLC separa-
min). Confirmation columns were Restek tion was used initially, then for those soils con-
RTX-200 (Crossbond trifluoropropyl methyl- taining isomers of DNT and Am-DNT, extracts
polysiloxane) and Restek RTX-225 (50% were reanalyzed using one of the alternative
cyanopropylmethyl – 50% phenyl methylpolysilox- HPLC separations described above.
ane). A pre-column was not used because extra For GC, extracts were analyzed on the HP5890
column length results in degradation of HMX [2]. where the ECD had a very limited linear range.
All extracts were diluted by at least of factor of 10
2.5.2. HPLC-UV and some up to 106 to be within the calibration
Initial studies used the HPLC separation spe- range. These large dilutions probably minimized
cified in Method 8330. A 25-cm by 4.6-mm (5-mm) the deposition of high boiling point residues in the
octadecyl (Supelco LC-18) column was eluted injection port liner and on the head of the GC
with 1.5 ml per min 1:1 methanol:water. Two column, a potential problem that concerns us
alternative separations to achieve resolution of when GC is used routinely for the analysis of soil
DNB, DNT, and Am-DNT isomers used either a extracts.
25-cm by 4.6-mm (5-mm) octadecyl (Burdick and Details of the results of these analyses are re-
Jackson OD5) column eluted with 1.4 ml min − 1 ported in [21]. The GC-ECD concentration esti-
33:13:54 methanol:acetonitrile:water or a 15-cm mates were correlated with those obtained by
by 3.9-mm (4-mm) Nova Pak C8 (Waters Mil- HPLC-UV for those analytes that were detected
lipore) column eluted with 1.4 ml per min 15:85 ten or more times out of the 24 samples. The most
isopropanol:water. The confirmation separation frequently detected analytes were TNT (18 times),
was on a 25-cm by 4.6-mm (5-mm) cyano (Supelco RDX (11 times), 2,4-DNT (15 times), TNB (19
432 M.E. Walsh / Talanta 54 (2001) 427–438

Table 2
Correlation of concentration estimates obtained by GC-ECD to those obtained by the standard HPLC-UV procedure

Analyte Site type Concentration range (mg kg−1) n Slope r

TNT Hazardous waste 0.28–27100 18 1.35 0.992

RDX Hazardous waste 0.28–13200 11 1.24 1.00
2,4-DNT Hazardous waste 0.10–95100 15 1.25 0.999
1,3,5-TNB Hazardous waste 0.15–570 19 0.81 0.976
1,3-DNB Hazardous waste 0.05–325 12 1.04 0.998
HMX Hazardous waste 3.5–2020 10 0.65 0.978
TNT Hazardous waste 0.02–0.27 10 1.14 0.997
2,4-DNT Hazardous waste 0.02–8.59 16 1.12 0.997
2,4-DNT Mine field 0.027–14.8 36 1.04 0.996
4-Am-DNT Mine field 0.10–1.51 26 1.05 0.956
2-Am-DNT Mine field 0.11–1.83 26 0.99 0.951

times), DNB (12 times), and HMX (ten times). ous HPLC analyses were thought to show a peak
All correlation coefficients were greater than 0.97. for TNT, but TNT was not detected by GC.
However, with the exception of DNB, slopes of Rather the GC showed peaks for several isomers
the least squares regression models were signifi- of DNT, one of which (3,4-DNT) co-elutes with
cantly different from the expected value of 1.00. TNT using the standard HPLC analytical separa-
For TNT, RDX, and 2,4-DNT, slopes were all tion and another (3,5-DNT) coelutes with TNT
greater than 1.00 (Table 2). This difference in using the standard confirmation HPLC
slope may be an artifact of the experimental error separation.
associated with the large dilutions required for
GC-ECD and the dominance of high values on 3.1.2. Trace concentrations
the slope obtained from a least-squares model. Another series of archived soils was selected for
Concentrations for each of these analytes spanned extraction and analysis. For land mine detection,
over six orders of magnitude. Two other problems very low (1 mg kg − 1) concentrations may need to
were the considerable scatter in TNB data and the be detected, and soils were selected for which
underestimation of HMX by GC-ECD. The TNB previous HPLC-UV analysis had either shown
scatter is most likely due to TNB’s instability in trace (less than the HPLC-UV detection limit of
solution, and the underestimation of HMX is 200 mg kg − 1) concentrations of either TNT, 2,4-
likely due to thermal degradation during the GC DNT, or RDX, or the soils were collected near
analysis. Although accurate GC analysis of HMX samples that had trace concentrations. For the
is possible, the first peak to degrade in shape extraction of these soils, 25.0-mg soil subsamples
following multiple injections of water or soil ex- and 50.0 ml of acetonitrile containing 25.0-mg l − 1
tracts is the HMX peak. Despite these problems, 3,4-DNT as an internal standard were used. If
GC-ECD analysis of soil extracts was judged enough soil was available, matrix spikes/matrix
feasible. The GC-ECD offered two significant ad- spike duplicates (MS/MSDs) were also prepared.
vantages over the standard HPLC method: lower The GC analysis was conducted on an HP6890
detection levels and improved chromatographic equipped with a micro-ECD. Example chro-
resolution of the isomers of DNT and Am-DNT. matograms of a real soil extract are shown in Fig.
One important advantage of using both HPLC 1. No dilutions were performed for the GC-ECD
and GC analysis is the ability to independently analysis because of the expanded linear range of
confirm analyte identities in complex chro- the micro-ECD [11]. For the HPLC analysis, the
matograms. This advantage was apparent for one Nova Pak C8 (Waters Millipore) separation de-
of the Monite soils. Chromatograms from previ- scribed previously was used.
 M.E. Walsh / Talanta 54 (2001) 427–438 433

TNT was detected by GC-ECD in all 13 ex- duplicates, the median relative percent difference
tracts ranging from 1.3 to 273 mg kg − 1 [21]. For (RPD%) was 11%. One sample (Camp Shelby)
showed very poor agreement between replicates.
This sample had 2,4-DNT as the highest concen-
tration analyte, and the TNT, in this case, was
probably present as in impurity in 2,4-DNT, not
as a high-explosives residue. 2,4-DNT was found
in 11 of the 13 extracts, over the concentration
range 0.94 –8600 mg kg − 1. Two of the soils
showed extreme heterogeneity for this analyte
with concentration estimates of 36 and 8600 mg
kg − 1 in duplicates for one sample, and 145 and
4100 mg kg − 1 in duplicates for another sample. At
the sites where these soils (Eagle River Flats
OB/OD and Camp Shelby NE Quad) were col-
lected, propellant grains were scattered across the
soil surfaces. 2,4-DNT is a propellant ingredient,
and such extreme heterogeneity is consistent with
particulate contamination, such as from a propel-
lant grain fragment. In the samples with relatively
high 2,4-DNT concentrations, 2,6-DNT and 3,4-
DNT were detected as well by GC-ECD. These
two isomers co-eluted on the HPLC separation.
Thus, 3,4-DNT is not suitable as an internal
standard for extracts from soils contaminated
with high concentrations of 2,4-DNT.
The Am-DNT isomers were detected by GC in
ten of the 13 soils. For the samples from Fort
Ord, several unidentified peaks eluted near the
Am-DNTs on the GC analysis and made quantifi-
cation difficult. Similar interferences were not ob-
served in other soils, although a large unidentified
peak did elute after 2-AmDNT from four other
soils. Subsequent analyses led us to believe the
peak was a phthalate ester.
Fig. 1. Example GC-ECD chromatograms of a field-contami- The only DNB isomer detected was 1,3-DNB,
nated soil using (a) HP-5; (b) RTX-225; and (c) RTX-200
columns. The soil was collected from the Nebraska Ordnance which was found in two of the soils at around 5.0
Plant and contains analytes most frequently found at haz- mg kg − 1.
ardous waste sites contaminated with explosives residues. (a) Recoveries from the MS/MSD samples were
HP-5 ([5%-phenyl]methylsiloxane), injector 250°C, 100°C for 2 excellent in some samples and low in others.
min, 10°C per min ramp to 200°C, 20°C per min ramp to
250°C, hold 5 min, detector 280°C. (b) RTX-225 ([50% Matrix effects are discussed later.
cyanopropylmethyl–50% phenyl]methylpolysiloxane), injector TNT and 2,4-DNT concentrations above 20 mg
200°C, 100°C for 2 min, 10°C per min ramp to 250°C, hold 6 kg − 1 were quantifiable by HPLC, and the GC
min, detector 220°C. (c) RTX-200 (Crossbond trifluoropropyl concentrations estimates were correlated with
methylpolysiloxane), injector 250°C, 100°C for 1 min, 5°C per
min ramp to 190, 1°C per min ramp to 200°C, 20°C per min those newly obtained by HPLC using splits of the
ramp to 250°C. same extract (Table 2). Again, good correlation
434 M.E. Walsh / Talanta 54 (2001) 427–438

(r =0.997 for both analytes) was found between Table 3

Method detection limits (mg kg−1) of nitroaromatics,
the two methods of determination, and again the
nitramines, and nitrate esters in spiked soils determined by
slopes of the least-squares models were greater GC-ECD
than 1.00 (1.14 and 1.12 for TNT and 2,4-DNT,
respectively), indicating slightly higher estimates AEC soil Ottawa sand
by GC.
Method 1a Method 1a Method 2b
There was no degradation in the GC peak
shapes or heights despite an extended run with 1,3-DNB 0.93 0.73 1.2
over 70 injections of undiluted soil extracts and 1,4-DNB 0.86
standards. 1,2-DNB 0.64
2,6-DNT 0.81 0.69
2,4-DNT 0.88 0.69 0.86
3.2. Confirmation columns
TNB 4.91 1.6
TNT 0.45 2.4c 11c
Because analyte identity is based solely on reten- RDX 3.1 3.4 2.6
tion time when using an ECD, confirmation is 4-Am-DNT 1.6 1.5 0.70
important. If concentrations are very high, a mass 3,5-DNA 2.0 2.1
2-Am-DNT 3.1 2.0 0.84
spectrometry or photodiode array detector can
NB 14 17
yield confirmation. For lower concentrations, o-NT 12 12
analysis by GC-ECD and HPLC-UV provides m-NT 11 11
confirmation based on different physical proper- p-NT 9.5 10
ties of the analytes (vapor pressure and polarity for NG 31 13
PETN 35 16
separation, and electronegativity and UV absorp-
Tetryl 66d 20
tion for detection). However, when concentrations HMX 26 25
are very low (less than 50 mg kg − 1), secondary
column confirmation by GC-ECD must be used. a
Method 1: 2.00 g soil extracted with 10.0 ml acetonitrile.
b
Two confirmation columns that are more polar Method 2: 25.0 g soil extracted with 50.0 ml acetonitrile.
c
Ottawa sand contained an interference that co-eluted with
than the HP-5 analytical column were used. They
TNT.
were a Restek RTX-200 (Crossbond trifluoro- d
AEC soil contained an interference that co-eluted with
propyl methylpolysiloxane) and Restek RTX-225 tetryl.
(50% cyanopropylmethyl – 50% phenyl methyl-
polysiloxane). Example chromatograms are shown hazardous waste characterization. Samples were
in Fig. 1. spiked at 5.00 and 50.0 mg kg − 1 because the ECD
Of these two columns, the RTX-225 was pre- response factors differ substantially for these ana-
ferred because RDX was resolved from 2-Am- lytes, being significantly higher for the di- and
DNT. These analytes co-eluted on the RTX-200. trinitroaromatics. Each matrix had interferences
Another problem with the RTX-200 was an inter- that inflated the MDLs for some of the analytes
mittent interfering peak eluting just prior to 2,4- (TNT in the Ottawa sand and 2-Am-DNT and
DNT. (This peak seemed to be associated with tetryl in the AEC soil). Nevertheless, the MDLs
plastics, but was not consistently present. Plastics (Table 3) were generally about 1 mg kg − 1 for the
contain phthalate esters, which are detected by the di- and trinitroaromatics and ten times higher for
ECD). Neither column was suitable for confirming the mono-nitroaromatics, which is consistent with
low concentrations of HMX, although a thinner the variable response factors of the ECD for these
film may have allowed this analyte to elute intact. compounds. MDLs were between 1 and 3 mg kg − 1
for the amino-nitrotoluenes, and were quite vari-
3.3. Spike reco6eries and method detection limits able for the more thermally labile nitramines and
nitrate esters.
Two blank matrices were spiked to determine Recoveries were calculated from the higher
detection limits for the analytes of interest for spikes for those analytes that had method detec-
 M.E. Walsh / Talanta 54 (2001) 427–438 435

tion limits less then 5 mg kg − 1. In all cases, (Table 3) for extraction of 25.0 g of soil with 50.0
recoveries were near 100%. Precision was best ml acetonitrile, and the MDLs were very similar
(B 4% RSD) for 1,3-DNB, TNT, and the DNT to those obtained using just 2.00 g of soil and 10.0
isomers. ml acetonitrile. The expectation of improvement
NG and PETN were not degraded during the in detection capability by using more soil in pro-
extended (18 h) sonication in a cooled sonic bath. portion to the volume of extraction solvent was
Previously in our lab, a decrease was observed in negated by the small interfering peaks introduced
recovery of NG in spiked soils (2500 mg kg − 1) by the matrix.
beyond 2 h of sonication. However, the bath at The matrix spikes/matrix spike duplicate of
that time was not cooled. Further kinetic studies field-contaminated soils showed variable recover-
with field-contaminated soils are needed to verify ies. Lacking extensive characterization of each of
that 18 h of sonication in a cooled sonic bath is these soils, the cause of the variable recoveries is
not too long for acceptable recoveries of NG. open to speculation. One difference that was visu-
Using spiked matrices to determine precision ally obvious was grain size. The finest-grained soil
and accuracy of an analytical method can pro- was that from Chickasaw, and this soil did show
duce confounding results due to the differing ana- poor recovery (41 –74%) of the spiked analytes.
lyte stability in newly spiked versus However, grain size alone did not account for low
field-contaminated aged soils [22]. This difference recovery in one of the Fort Ord samples, each of
became apparent when 25.0-g soil aliquots were which was sandy. The reason for the difference in
spiked with the analytes of interest for mine detec- stability of spiked versus aged, field-introduced
analytes is beyond the scope of this study, but this
tion (Table 1). The first soil spiked was a silt
difference greatly reduces our ability to judge
obtained locally. As stated above, 1.00 ml spike
method accuracy.
solution was added to 25.0 g of soil to yield a
bulk analyte concentration of 2.00 mg kg − 1. How-
3.4. Mine field samples
ever, this small volume of spike solution wetted
only a small portion of the soil, unlike when
3.4.1. Fort Leonard Wood soils
2.00-g soil aliquots were spiked. When the silt was
An experimental mine field was created at Fort
extracted with acetonitrile containing 25.0-mg l − 1
Leonard Wood by burying various defused anti-
3,4-DNT, recovery of all analytes was nil. The tank and anti-personnel mines. The mines were
peak height for 3,4-DNT, the internal standard manufactured in the former Yugoslavia and each
present throughout the extraction process, was contained TNT (0.10 –0.20 kg in anti-personnel
normal, thus the loss of the spiked analytes must mines and 5.50 –5.60 kg in anti-tank mines). The
have occurred during the 1 h of aging between the first set of soil samples was collected 2 months
addition of the spike solution and the extraction after the mines were buried and revealed which
solvent. To see if the loss was matrix specific, analytes are actually present in soils surrounding
duplicate 25.0-g samples of the following were emplaced land mines. Of the 143 samples re-
spiked: glass beads (25 mm), Ottawa sand, AEC ceived, only 28 had detectable analytes as deter-
soil, and the same silt where loss was observed mined by GC-ECD with confirmation by
initially. Each matrix had interferences that co- HPLC-UV. The most common analytes detected
eluted with at least one analyte, but recoveries were 2,4-DNT, 4-Am-DNT, and 2-Am-DNT,
bracketed 100% for analytes spiked onto the glass which were each found in 27 out of the 28 positive
beads and the Ottawa sand, bracketed 90% for samples and each at median concentrations
one replicate of the AEC soil and 50% for the greater than 60 mg kg − 1. 2,4,6-TNT was found in
other replicate, and was nil again for the silt. Next 24 samples and tended to be lower in concentra-
ten replicates of Ottawa sand were spiked to tion than the Am-DNTs, indicating that it is
obtain an estimate of method detection limits undergoing microbiological transformation. Me-
436 M.E. Walsh / Talanta 54 (2001) 427–438

dian 2,4,6-TNT concentration was 6.5 mg kg − 1, these analytes might be sufficiently high to make
and the maximum was over 3000 mg kg − 1 in a them valuable markers for detection of emplaced
soil collected next to a mine. The other two mines. If other research shows that these analytes
commonly found analytes were 1,3-DNB and 2,6- are important for mine detection, analytical meth-
DNT, which had median concentrations of 8.8 ods would need to be optimized for determining
and 3.1 mg kg − 1, respectively. these compounds in soils. Potentially, derivatiza-
More soils were collected 4 months after the tion via bromination or iodination could be used
mines were buried; out of 199 samples, 73 had to enhance ECD response as described for the
detectable explosives. The Am-DNT isomers were determination of aromatic amines in ammunition
the most frequently detected analytes with 71 wastewater [23,24].
detections of 2-Am-DNT and 61 detections of
4-Am-DNT. Correlation between the two meth-
ods for 2,4-DNT was similar to what has been 4. Conclusions
previously observed; the correlation coefficient
was greater then 0.99 and the slope slightly GC-ECD concentration estimates of nitroaro-
matic and nitramines in field-contaminated soils
greater than 1.00. In previous comparisons made
were compared with estimates obtained by the
between GC-ECD and HPLC-UV, there were not
standard HPLC-UV method, and there was good
a sufficient number of samples with Am-DNT
correlation between the two methods of analysis.
isomers at concentrations high enough for deter-
The GC-ECD provided improved chromato-
mination by HPLC. For this set of samples, corre-
graphic resolution and detection. Two extraction
lation coefficients were 0.951 and 0.956 for
procedures were used, both of which involved 18
2-Am-DNT and 4-Am-DNT, respectively. The
h of sonication in a cooled bath. In one method,
slopes of the least squares regression lines were
2.00 g of soil were extracted with 10.00 ml, of
not significantly different from 1.00 for either
acetonitrile, and in the second 25.0 g of soil were
analyte. The data were more scattered for these
extracted with 50.0 ml acetonitrile. Method detec-
two analytes, showing that accurate determina- tion limits were similar for these two methods
tions are more difficult to obtain than for 2,4- because matrix interferences became more pro-
DNT. Like the nitramines, the amino-DNTs are nounced when the ratio of soil to solvent was
susceptible to degradation as the GC injection increased from 1:5 to 1:2. Method detection limits
port liner becomes less and less inert with re- were around 1 mg kg − 1 for the di- and trini-
peated injections of soil extracts. On the HPLC troaromatics, about 10 mg kg − 1 for the mono-ni-
separation, these analytes elute late where the
peaks are quite broad.
Similar to the first set of samples collected from
this minefield, 2,4,6-TNT was detected generally
at lower concentrations than the Am-DNTs, ex-
cept for samples collected in contact with a mine.
The presence of the TNT biotransformation prod-
ucts implies that 2,4-DNT and 1,3-DNB biotrans-
formation products should be present as well.
These products are 2-amino-4-NT, 4-amino-2-NT,
and 3-nitroaniline, which were detected in soil
collected under a TMA-5 anti-tank mine. Unfor-
tunately, the ECD response is not strong for these
compounds because they each have only one nitro Fig. 2. GC-ECD (HP-5) chromatogram of a soil collected near
group (Fig. 2). However, the vapor pressure of a land mine.
 M.E. Walsh / Talanta 54 (2001) 427–438 437

troaromatics, 3 mg kg − 1 for RDX, 25 mg kg − 1 [4] J. Yinon, S. Zitrin, Modern Methods and Applications in
Analysis of Explosives, John Wiley and Sons Ltd., West
for HMX, and between 10 and 40 mg kg − 1 for Sussex, England, 1993, pp. 42 – 66 and 213– 224.
the nitrate esters (NG and PETN). [5] Chin.-Kai Meng, Response of Nitrogen-Phosphorus De-
Spike recovery studies revealed artifacts intro- tectors to Different Structural Nitrogen Compounds,
duced when the mass of the soil spiked was Hewlett Packard Application Brief, Hewlett-Packard
Company, Wilmington, DE, USA, February 1998.
large (25.0 g) in proportion to the volume of [6] J. Feltes, K. Levsen, D. Volmer, M. Spiekermann, J.
spike solution added (1.00 ml). Recoveries were Chromatogr. 518 (1990) 21.
excellent (around 100%) when 2.00-g soil sam- [7] K. Levsen, P. Mußmann, E. Berger-Preiß, A. Preiß, D.
ples were spiked with 1.00 ml of solution. How- Volmer, G. Wünsch, Acta Hydrochim. Hydrobiol. 21
(1993) 153.
ever, when 25.0-g soil samples were spiked with [8] A.D. Hewitt, T.F. Jenkins, On-Site Method for Measur-
5.00 ml of solution, recoveries varied from nil in ing Nitroaromatic and Nitramine Explosives in Soil and
a silt to around 80% in a sand. Matrix spikes/ Groundwater using GC-NPD, CRREL Special Report
matrix spike duplicates of field-contaminated 99-9, US Army Cold Regions Research and Engineering
Laboratory, Hanover, NH, 1999.
soils also showed inconsistency in recovery of [9] T.F. Jenkins, C.L. Grant, Anal. Chem. 59 (1987) 1326.
the spiked analytes. [10] American Society for Testing and Materials, Standard
Soils collected near emplaced mines contained Practice for Use of Electron-Capture Detectors in Gas
residues of TNT, 2,4-DNT, and 1,3-DNB and Chromatography, E 697-96, ASTM, West Conshohocken,
PA, USA, 1996.
their various microbial transformation products. [11] F. David, P. Sandra, M.S. Klee, Analysis of Nitroaromat-
The importance of these transformation prod- ics and Nitro-Polycyclic Aromatic Hydrocarbons by Cap-
ucts for land mine detection is uncertain at illary Gas Chromatography with the HP 6890
Micro-ECD, Hewlett Packard Application Note 228-380,
present. The ECD is sufficiently sensitive to de-
Hewlett-Packard Company, Wilmington, DE, USA,
tect part-per-billion concentrations of the mono- March 1997.
amino transformation products of TNT, but is [12] T.F. Jenkins, M.E. Walsh, P.W. Schumacher, P.H. Mi-
insensitive to the amino-transformation products yares, C.F. Bauer, C.L. Grant, J. AOAC 72 (1989) 890.
[13] USEPA, Test Methods for Evaluating Solid Waste, Phys-
of 2,4-DNT and DNB.
ical/Chemical Methods, SW-846 Update III, Office of
Solid Waste, Washington, DC, 1997.
[14] R.P. Murrmann, T.F. Jenkins, D.C. Leggett, Composi-
Acknowledgements tion and Mass Spectra of Impurities in Military Grade
TNT Vapor. CRREL Special Report 158, US Army Cold
Regions Research and Engineering Laboratory, Hanover,
The author gratefully acknowledges funding NH, 1971.
for this work provided by the US Army Engi- [15] T.F. Jenkins, D.C. Leggett, R.P. Murrmann, Preliminary
neer Waterways Experiment Station (Ann Investigation of the Permeability of Moist Soils to Explo-
sive Vapor. CRREL Technical Note, US Army Cold
Strong), the Defense Advanced Research
Regions Research and Engineering Laboratory, Hanover,
Projects Agency (Regina Dugan), and the US NH, 1974.
Army Environmental Center (Martin Stutz). [16] D.C. Leggett, T.F. Jenkins, R.P. Murrmann, Composi-
Technical assistance and reviews were provided tion of Vapors Evolved from Military TNT as Influenced
by Temperature, Solid Composition, Age, and Source.
by Thomas A. Ranney, Thomas F. Jenkins,
CRREL Special Report 77-16, US Army Cold Regions
Paul H. Miyares, and Vivian George. Research and Engineering Laboratory, Hanover, NH,
1977.
[17] M.E. Walsh, T.F. Jenkins, F.G. Thorne, J. Energy Mater.
13 (1995) 357.
References
[18] T.F. Jenkins, M.E. Walsh, P.G. Thorne, S. Thiboutot, G.
Ampleman, T.A. Ranney, C.L. Grant, Assessment of
[1] M.A. Rouhi, Chem. Eng. News 75 (1997) 14. Sampling Error Associated with Collection and Analysis
[2] M. Hable, C. Stern, C. Asowata, K. Williams, J. Chro- of Soil Samples at a Firing Range Contaminated with
matogr. Sci. 29 (1991) 131. HMX. CRREL Special Report 97-22, US Army Cold
[3] M.E. Walsh, T.A. Ranney, J. Chromatogr. Sci. 36 (1998) Regions Research and Engineering Laboratory, Hanover,
406. NH, 1997.
438 M.E. Walsh / Talanta 54 (2001) 427–438

[19] T.F. Jenkins, M.E. Walsh, P.G. Thorne, P.H. Miyares, [21] M.E. Walsh, T.A. Ranney, Determination of Nitroaro-
T.A. Ranney, C.L. Grant, Site Characterization at the matic, Nitramine, and Nitrate Ester Explosives in Soils
Inland Firing Range Impact Area at Fort Ord. CRREL Using GC-ECD, CRREL Special Report 99-12, US
Special Report 98-9, US Army Cold Regions Research Army Cold Regions Research and Engineering Labora-
and Engineering Laboratory, Hanover, NH, 1998. tory, Hanover, NH, 1999.
[20] T.F. Jenkins, D.C. Leggett, and T.A. Ranney, Vapor [22] C.L. Grant, T.F. Jenkins, K.F. Myers, E.F. Mc-
Signatures from Military Explosives. Part 1. Vapor Cormick, Environ. Toxicol. Chem. 14 (1995) 1865.
Transport from Buried Military-Grade TNT. CRREL [23] T.C. Schmidt, R. Haas, K. Steinbach, E. v Löw, Frese-
Special Report 99-21. U.S. Army Cold Regions Re- nius J. Anal. Chem. 357 (1997) 909.
search and Engineering Laboratory, Hanover, NH, [24] R. Haas, T.C. Schmidt, K. Steinbach, E. v Löw, Frese-
1999. nius J. Anal. Chem. 359 (1997) 497.