Forensic Science International 232 (2013) 160–168

Screening for trace explosives by AccuTOFTM-DART1:An in-depth validation study

Edward Sisco a,*, Jeffrey Dake b, Candice Bridge b

a University of Maryland College Park, Department of Chemistry and Biochemistry, Chemistry Building, College Park, MD 20742, United Statesb United States Army Criminal Investigation Laboratory, Gillem Enclave, GA 30297, United States

A R T I C L E I N F O

Article history:

Received 12 December 2012

Received in revised form 18 June 2013

Accepted 5 July 2013

Available online

Keywords:

Ambient ionization mass spectrometry,

Trace detection, Gas chromatography mass

spectrometry

A B S T R A C T

Ambient ionization mass spectrometry is finding increasing utility as a rapid analysis technique in a

number of fields. In forensic science specifically, analysis of many types of samples, including drugs,

explosives, inks, bank dye, and lotions, has been shown to be possible using these techniques [1]. This

paper focuses on one type of ambient ionization mass spectrometry, Direct Analysis in Real Time Mass

Spectrometry (DART-MS or DART), and its viability as a screening tool for trace explosives analysis. In

order to assess viability, a validation study was completed which focused on the analysis of trace

amounts of nitro and peroxide based explosives. Topics which were studied, and are discussed, include

method optimization, reproducibility, sensitivity, development of a search library, discrimination of

mixtures, and blind sampling. Advantages and disadvantages of this technique over other similar

screening techniques are also discussed.

� 2013 Elsevier Ireland Ltd. All rights reserved.

Contents lists available at ScienceDirect

Forensic Science International

jou r nal h o mep age: w ww.els evier . co m/lo c ate / fo r sc i in t

1. Introduction

Screening techniques for the analysis of trace explosives areutilized in a wide range of situations including both laboratory andfield based applications. While the necessary characteristics of ascreening technique will vary based upon circumstances, there area number of benchmarks these techniques will have to be able toaccomplish in all cases. Among these characteristics, a usefulscreening technique must be able to rapidly detect a number ofpossible threats with minimal sample preparation. Furthermore,the sensitivity of the screening technique should be equal to orexceeding that of the confirmatory technique. Other qualities of agood screening technique include a high level of accuracy andreproducibility, the ability to detect threats even in complexmatrices, and the minimization of false alarms. One such techniquewhich has exhibited all of these characteristics is DART-MS.

DART-MS is an atmospheric pressure ionization technique thatwas first discussed by Cody et al. in 2005 [2] and has since become acommercially available source. In general, the DART source takesadvantage of the production of metastable gas atoms generatedfrom a glow discharge plasma in a heated gas stream. The heatedgas stream interacts with the sample and depending onoperational conditions, a number of ionization pathways can

* Corresponding author. Tel.: +1 301 975 2093, fax: +1 301 417 1312.

E-mail addresses: esisco@umd.edu, edward.sisco611@gmail.com (E. Sisco),

jeffrey.dake@us.army.mil (J. Dake), candice.bridge@us.army.mil (C. Bridge).

0379-0738/$ – see front matter � 2013 Elsevier Ireland Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.forsciint.2013.07.006

occur [3]. This technique allows for sample introduction directlyinto the gas stream between the source and the mass spectrometerinlet, providing a way to analyze a sample under atmosphericconditions using high temperature and high voltage, without ahigh potential.

Even though DART-MS is a relatively new technique, it hasalready been applied to a number of different forensic specimens,including bank dye, inks, and illicit drugs [4–7]. Similarly DART-MShas also been applied to the detection of counterfeit pharmaceu-ticals, identifying adulterants in food, and chemical warfare agents[8–10]. The use of this technique for the detection of explosives hasbeen discussed; however, little work has been done in establishinga comprehensive method optimization which measures thesensitivity in the detection of explosives, as well as studyinghow it compares against currently employed techniques. Nilleset al. [11] has shown that detection of a number of differentexplosives is possible using DART-MS. Furthermore, it was shownthat a detection of these compounds off of a number of surfaces, aswell as detection in a number of different solutions can beaccomplished [11]. Detection of trace explosives within latentfingerprints using DART-MS has also been discussed [12]. Somelimits of detection have been reported elsewhere, however theyhave included only a limited number of explosives [2,12,13].

In addition to DART-MS, there are a number of other techniqueswhich are being used to screen for trace explosives. These includeion mobility spectrometry (IMS) [14–17], chemiluminescence [15],canine detection [18,19], infrared spectroscopy [15,20], Ramanspectroscopy [21–24], chemical field test kits, and other ambient

E. Sisco et al. / Forensic Science International 232 (2013) 160–168 161

pressure ionization mass spectrometry techniques [25–33]. Whileeach technique presents its own advantages and disadvantages,DART-MS has been shown to be as formidable of a screeningtechnique as any other. DART-MS could also be combined with aconfirmatory technique, such as high performance liquid chroma-tography mass spectrometry (HPLC-MS) or gas chromatographymass spectrometry (GC-MS) [34,35] to provide a powerfulcombination of techniques for the analysis of explosives.

This paper shows the capabilities and limitations of DART-MSas a useful screening tool for the analysis of trace explosives.Because this study was completed for a screening scenario, it hasbeen optimized for qualitative (not quantitative) analysis. Adiscussion of the variety of subjects which were analyzed isdiscussed for both nitro based and peroxide based explosives.These subjects include method optimization, method reproduc-ibility, limits of detection, construction of a search library, analysisof mixtures, and the ability of the method to analyze unknowns. Inall twenty-three explosives were analyzed, two peroxide-basedexplosives and twenty-one nitro based explosives. Also discussedis the development and implementation of calibration compounds,outside of the traditional polyethylene glycol (PEG) employed inDART analysis.

2. Materials and methods

2.1. Solvents, standards, & sampling material

All of the explosives used were purchased as analytical standards from

AccuStandard (New Haven, CT, USA) at concentrations of either 100 mg/mL or

1000 mg/mL. A list of the explosives which were analyzed is presented in Table 1.

Two pre-made mixtures were also used in this study–a 14 component HPLC

mixture purchased from AccuStandard as well as a 6 component mixture which was

purchased from ThermoScientific (Waltham, MA, USA). Serial dilutions of the

individual explosives and the 14 component mixture were completed using

methanol, purchased from Sigma-Aldrich (St. Louis, MO, USA).

Several different mass calibrants and independent quality assurance quality

control (QA/QC) compounds were also required for this study. For the detection of

nitro-based explosives, the mass calibrants used were polyethylene glycol 600

(Acros Organics, Geel, Belgium) and a series of saturated fatty acids (Supelco,

Bellefonte, PA, USA). A mixture of glycol ethers was used as the mass calibrant for

the peroxide explosives, with individual glycol ethers being purchased from

AccuStandard. The independent QA/QC compounds which were used included:

Table 1List of explosives used in the study as well as the abbreviations used in the paper.

Explosive Abbreviation

Nitroaromatics

Trinitrotoluene TNT

2,4-Dinitrotoluene 2,4-DNT

2,6-Dinitrotoluene 2,6-DNT

Tetryl Tetryl

Ammonium Picrate AP

Picric Acid PA

Nitrobenzene NB

1,3-Dinitrobenzene 1,3-DNB

1,3,5-Dinitrobenzene 1,3,5-TNB

2-Nitrotoluene 2-NT

3-Nitrotoluene 3-NT

4-Nitrotoluene 4-NT

2-Amino-4,6-Dinitrotoluene 2-A-4,6-DNT

4-Amino-2,6-Dinitrotoluene 4-A-2,6-DNT

Ringed Nitros

Cyclotrimethylenetrinitramine RDX

Cyclotetramethylenetetranitramine HMX

Straight chained Nitros

Nitroglycerin NG

Ethylene Glycol Dinitrate EGDN

Diethylene Glycol Dinitrate DEGDN

Pentaerythritol Tetranitrate PETN

Erytritol Tetranitrate ETN

1-Nitroglycerin 1-NG

Peroxides

Triacetone Triperoxide TATP

Hexamethylene Triperoxide Diamine HMTD

oleic acid, purchased from Acros, and methyl decanoate, purchased from

AccuStandard.

Glass microcapillaries were used to introduce the samples into the DART gas

stream. The 90 mm closed capillaries were purchased from Corning Incorporated

(Corning, NY, USA). Before analysis, the capillaries were introduced into the gas

stream to burn off any contaminants which may be present on the rods. This was

especially necessary for the positive ionization mode, where a background signal due

to the phalates from the plastic container was readily detectable on unheated rods.

2.2. Development of mass calibration mixtures

During the validation study, it was found that the mass calibration compound,

PEG 600, was unresponsive in the sub 150 amu mass region, as well as at the lower

temperatures and voltages required to analyze peroxide explosives. Therefore, it

was necessary to develop new mass calibration mixtures. To aid in the calibration of

the low mass range for negative mode, a fatty acid mixture was developed and

implemented. The fatty acid mixture contained approximately 100 mg of hexanoic

acid, octanoic acid, decanoic, acid, dodecanoic acid, myristic acid, palmitic acid,

stearic acid, eicosanoic acid, docosanoic acid, and tetracosanoic acid, dissolved in

2 mL of hexane, to give a solution concentration of approximately 50 mg/mL. A

tunable mass range of 100–400 amu was obtained by monitoring the [M-H]� and

[M + O2-H]� ions.

For the peroxide explosives it was necessary to have a calibrant which responded

well in the positive ionization mode, at a low mass range, low temperature, and low

voltage. It was determined that a mixture of glycol ethers provided the desired

characteristics. The mixture was developed to calibrate a mass range of 75–

380 amu–which encompasses all of the peaks identified for the peroxide explosives.

Nine glycol ethers were chosen and included: 2-Methoxyethanol, Di(ethylenegly-

col)methyl ether, 2-Ethoxyethanol, 2-Hexoxyethanol, 2-(2-n-Hexoxy)-ethanol,

2(2-butoxyethoxy)ethanol, and 2(Phenoxyethanol). The mixture was made by

diluting 50 mL of each of the nine glycol ethers in 1650 mL of methanol, giving a

solution concentration of approximately 30 ng/mL. Using this mixture, the

calibration points included protonated ions [M + H]+ as well as the protonated

dimers [M + M + H]+. Both of these mass calibration mixtures provided more than

the necessary number of masses for calibration, as well as calibration correlation

coefficients lower than 1 � 10�12. They were also cross compared to PEG 600, under

conditions in which both calibrants responded well, to verify mass accuracy.

2.3. Parameters for AccuTOF-DART

The instrument which was used in the study was a JEOL (Toyko, Japan)

AccuTOFTM mass spectrometer (JMS-T100LC) coupled with an IonSense (Saugus,

MA, USA) DART1 source. Helium was used as the ionizing gas with a flow rate of

1.75 L/min. For analysis in negative mode, the needle voltage was 3000 V, the

electrode 1 voltage was �200 V, and the electrode 2 voltage was �225 V.

Parameters of the mass spectrometer inlet included a varied orifice 1 voltage and an

orifice 2 voltage of �5 V. The ring lens voltage was �3 V, and the detector voltage

was 2500 V. A peaks voltage of 600 V was also used. For analysis in the positive

mode, needle, electrode 1, and electrode 2 voltages were 3000 V, +200 V, and

+225 V, respectively. The orifice 2 voltage was +5 V with a ring voltage of +10 V. A

peaks voltage of 500 V was used with a detector voltage of 2500 V. In both instances,

the orifice 1 and gas stream temperature were varied.

Two individual methods were developed for the detection of different

explosives. One method was developed in negative ion detection mode for

explosives containing nitro groups while the second method was developed for

positive ion detection of peroxide explosives. Within each method, three different

parameters were varied in order to optimize the technique, the first of which was

the orifice 1 voltage. The orifice 1 voltage provides a mechanism to control the

amount of fragmentation which a molecule undergoes, with high orifice 1 voltages

typically increasing fragmentation. A range of voltages from �10 V to �70 V were

evaluated for the negative mode, while a range of +5 V to +20 V were chosen for the

positive mode, due to the fragility of the peroxide explosives. The second parameter,

which was altered, was the gas stream temperature. Increasing the gas temperature

can increase desorption, although thermal decomposition may occur if the

temperature is increased too far. In this study the gas temperature for negative

mode was varied from 150 to 300 8C, while the temperature was altered from 100 to

200 8C for positive mode. The final parameter which can be altered is the addition of

a dopant to the gas stream. A dopant can work by providing a softer ionization cloud

in addition to the formation of adduct ions via incorporation of dopant ions with the

sample molecules. Four dopants were tested in negative mode and one dopant was

evaluated in positive mode.

To analyze the effects of these parameters in an orderly fashion, a matrix of

different orifice 1 voltages and gas temperatures was first completed. Once an

optimal voltage and temperature was chosen, the effects of dopant addition were

studied. Optimization was based off of the response from TNT, RDX, and PETN for

negative mode and TATP and HMTD for positive mode. A mass range of 100–

600 amu was monitored for negative mode, and a mass range of 60–400 amu was

monitored for positive mode in order to remove the greatest amount of background

possible without losing necessary response peaks. All spectra were calibrated and

centroided prior to analysis.

E. Sisco et al. / Forensic Science International 232 (2013) 160–168162

2.4. GC-MS methods

In order to compare the limits of detection for the DART-MS to current

confirmatory techniques, explosives were also analyzed by GC-MS. The nitro

explosives were analyzed by CI-GC-MS, using an Agilent 6890N gas chromatograph

with a 5973 mass spectrometer (Santa Clara, CA, USA) while the peroxides were

analyzed by EI-GC-MS, using an Agilent 7890A gas chromatograph with a 5975C

mass spectrometer. The CI-GC-MS method utilized a 15 m J&W Scientific 123-5012

DBS column (Agilent Technologies) with an outer diameter of 320 mm and a film

thickness of 0.25 mm. Helium carrier gas was used at a flow rate of 3.0 L/min, with

methane employed as the ionization gas. The inlet temperature was 175 8C with a

1 mL splitless injection. The oven was ramped from an initial temperature of 50 8C,

which was held for 1 min, to 220 8C at a rate of 20 8C/min. The oven was then held at

220 8C for 4 min. A mass range of 42–350 m/z was scanned.

The EI-GC-MS method utilized a 30 m HP-5MS 5% Phenyl Methyl Siloxane

column (Agilent Technologies) with a 250 mm outer diameter and a 0.25 mm film

thickness. Hydrogen carrier gas was used at a flow rate of 0.045 L/min. A 1 mL

splitless injection was introduced into the front inlet at 150 8C. The oven was

ramped from an initial temperature of 50 8C to 200 8C at a rate of 10 8C/min, and

then held at 200 8C for 20 minutes. A mass range of 15–550 m/z was scanned.

2.5. Limit of detection determinations

The limit of detection analysis was completed by creating serial dilutions from

the stock solutions of explosives. Each dilution was run in triplicate on the DART-

MS, with one microliter of sample being deposited onto a clean glass microcapillary

via a microsyringe. A new microcapillary was used for every replicate to prevent

concentration buildup on the rod with repeated analyses. In all other portions of the

study, solutions were sampled by dipping a new capillary tube directly into the

sample vial. It was determined in previous studies that this sampling method

results in approximately 1–2 mL being deposited onto the rod. The limit of detection

was defined as the level at which the signal was at least three times greater than the

background levels for all three replicates. The peak which was required to be above

background was one that was unique to that particular explosive (or pair of

explosives, as explained in Section 3.4). The nitrate peak was not monitored for limit

of detection determinations as its signal does not provide useful discriminating

information about the identity of the explosive, even though lower limits of

detection could be accomplished through monitoring an increase in the nitrate

signal. Furthermore, the presence of a nitro group does not provide sufficient

information to consider it a presumptive test for an explosive, as many nitro

containing compounds are not an explosive.

3. Results and discussion–negative mode (nitro basedexplosives)

3.1. Effects of altering DART parameters

The first portion of the validation study investigated the effectsof altering analysis parameters on the response of differentexplosives. In order to evaluate these effects, three explosives,which were representative of each class of nitro based explosive,were chosen. TNT was chosen as a representative nitroaromatic,RDX as a representative ringed nitro, and PETN as a representativestraight chained nitro. The first parameter, which was analyzed,was gas stream temperature. It was found that as the gas streamtemperature was increased from 150 8C to 225 8C, the intensity ofall three explosives increased. When the gas temperature wasraised above 225 8C, the intensity of the peaks began to decrease,possibly due to thermal decomposition. The 75 8C increase intemperature, from 150 8C to 225 8C, provided up to a factor of fiveincrease in the signal of the base peak. Fig. 1 depicts the averagesignal of TNT obtained using a number of different parameters.

After completing several other portions of the study, it wasfound that two explosives, DEGDN and EGDN, responded poorly at225 8C, but were more readily ionized at 125 8C. Up to a sixteen-fold increase in signal was observed for the detection of DEGDN bylowering the temperature, as shown in Fig. 2. This is likely due tothe small size and fragility of these two molecules, withdegradation accelerated at the elevated temperatures. All otherexplosives showed maximum intensity at, or near, 225 8C.

Orifice voltage was the second parameter which was studied. Ofthe orifice 1 voltages which were analyzed, �10, �30, and �50 V, didnot show any differences in fragmentation patterns. Furthermore,

as the orifice voltage was increased in this range, up to a two-folddecrease in signal was observed. When the orifice voltage was raisedto �70 V, increased fragmentation was seen; however, overall signalintensities were greatly diminished. Tuning was difficult toaccomplish at �70 V as the mass calibrants themselves were alsoheavily fragmented at this voltage. Voltage switching between�10 V and �70 V was also attempted, however signal intensitydropped significantly compared to runs in which the voltage was notswitched. Since the purpose of the study was to evaluate DART as ascreening technique, the voltage which provided the greatestsensitivity, �10 V, was chosen.

The final parameter, which was analyzed, was the addition of adopant species to the gas stream. This was accomplished by using acapped GC-MS vial with a shortened melting point capillary tubepierced through the rubber septum, in order to steadily introduceonly a small amount of the dopant species at a time. Additionalmethods of dopant introduction were attempted, however thismethod was found to produce the most consistent results, inaddition to being the easiest method to implement. The effects offour dopants were studied: acetone, chloroform, methylenechloride, and trifluoroacetic acid (TFA). The TFA dimer produceda strong background ion at 226.9779 amu, which was in betweenthe two TNT peaks (at 226 and 227 nominal amu), and was shownto hinder the detection of both peaks. Also, due to the proximity inmass between TFA and TNT, it could be possible that a poorlycalibrated run using TFA as a dopant could incorrectly be mistakenfor TNT. An example of the TFA background peak is shown in Fig. 1.The chloride ion producing dopants, methylene chloride andchloroform, were shown to enhance the signal for both RDX andPETN, however it decreased the signal of TNT and several othernitroaromatic explosives. No adduct ions were produced in thepresence of either dopant. Acetone was shown to increase thesignal intensity of the all explosives, even though it too did notproduce adduct ions.

From these results, two separate operating parameters wheredeveloped to screen for nitro based explosives. A ‘‘high tempera-ture’’ method was developed using a 225 8C gas stream tempera-ture, a �10 V orifice 1 voltage, and an acetone dopant. This methodallowed for detection of all but one of the nitro-containingexplosives analyzed. In order to screen for that explosive, EGDN, aswell as to increase the sensitivity for DEGDN, a ‘‘low temperature’’method was also developed which was identical to the ‘‘hightemperature’’ method, with the exception that the gas streamtemperature was lowered to 125 8C.

3.2. Method reproducibility

A study on the reproducibility of the method for the analysis of anumber of explosives was undertaken. Six different explosives,TNT, 2,6-DNT, PETN, DEGDN, RDX, and HMX, were analyzed todetermine how reproducibly the system detected the peaks relatedto these explosives. To do so, the three most abundant peaks fromeach explosive were monitored to determine if the calculatedmasses fell within �0.005 amu of the theoretical masses. The�0.005 amu tolerance was chosen based on manufacturer specifica-tions. Forty-eight replicates were completed in eight runs, spanningseveral days. In addition to monitoring the three masses from eachexplosive, six masses from the calibration compound, and the mass ofthe independent QA/QC compound were also monitored. It was foundthat for the six explosives analyzed, there were only 11 of 864measured masses which did not fall within the �0.005 amu tolerancefrom the theoretical mass. Of those eleven masses 5 were associatedwith DEGDN, 3 were associated with HMX, 2 with PETN, and 1 wasassociated with 2,6-DNT, giving a 98.7% agreement. Oleic acid, theQA/QC compound, had an accurate mass in 100% of the measure-ments. PEG, the mass calibrant in which six masses were monitored

Fig. 1. Effects of altering DART parameters on the response of TNT. The optimized parameters (1) are compared to analysis without a dopant (2), at a low temperature (3), and

at a high voltage (4). The effects of other dopants are also shown (5,6). Note the background ion at 226.978 m/z in the run using TFA as a dopant (6).

E. Sisco et al. / Forensic Science International 232 (2013) 160–168 163

demonstrated 2 out of 96 missed masses. Overall, a percent accuracyof 98.8% was achieved.

Upon further analysis of the failed masses, it was found that therewere two situations which caused the masses of the peaks to fall outof the tolerance range. The first situation occurred when a shoulderwas present on the peak of interest. The presence of the shouldercaused the center of the mass peak to shift when the spectrum wascentroided. If the shoulder caused a large enough shift, thecentroided peak was shown to fall outside of the acceptable massrange. The other situation which caused a peak to fall outside of thetolerable range occurred when the peak had both low intensity andpoor resolution. It was found that the peaks without a shoulderwhich failed had an intensity of less than 1000 counts and a massresolution of less than 4000. These two situations explained all butone of the masses which fell outside of the range.

3.3. Limit of detection

The limit of detection was determined for twelve of the nitroexplosives on DART-MS and compared to their limit of detection on

Fig. 2. Effects of temperature on the response of DEGDN. A significant increase in s

CI-GC-MS. Table 2 represents the comparison of limits of detectionbetween the two techniques. For eight of the twelve explosivesanalyzed the limit of detection was greater on DART-MS than on CI-GC-MS. Two of the components, 2,4-DNT and 2,6-DNT, had limits ofdetection on DART-MS equal to that of CI-GC-MS. Additionally, twoexplosives had limits of detection which were more sensitive in CI-GC-MS than the ‘‘high temperature’’ method of DART-MS - DEGDNand EGDN. This is likely due to the fact that nitrate, which was notbeing monitored due to its ambiguity, is the predominant fragment.The limit of detection of both DEGDN and EDGN were determinedusing the ‘‘low temperature’’ method as well. The ‘‘low temperature’’method allowed for a lower limit of detection for these twoexplosives, giving DART-MS increased sensitivity over CI-GC-MS forthe detection of DEGDN. In the case of EGDN, DART-MS was not assensitive as CI-GC-MS even at the lower temperature run.

3.4. Developing a search library

In order to appropriately analyze mixtures and complete blindsampling, a search list for the explosives which were analyzed was

ignal is noted when the temperature is lowered from 225 8C (1) to 125 8C (2).

Table 2Comparison of LOD between DART-MS and CI-GC-MS.

Explosive DART-MS LOD CI-GC-MS LOD

TNT 0.25 ng 1 ng

2,4-DNT 0.50 ng 0.50 ng

2,6-DNT 0.50 ng 0.50 ng

Tetryl 1.0 ng 5 ng

AP 50 ng >100 ng

RDX 0.50 ng 25 ng

HMX 10 ng >1000 ng

NG 5 ng 50 ng

EGDN 100 ng* 10 ng

DEGDN 10 ng* 25 ng

PETN 5 ng 50 ng

ETN 1 ng 50 ng

The asterisk indicates LODs which were determined using the ‘‘low temperature’’

method.

E. Sisco et al. / Forensic Science International 232 (2013) 160–168164

developed. In the development of the search list, it was determinedthat there are three types of peaks which exist for the explosivecompounds–unique peaks, shared ion peaks, and mass overlappeaks. These types of peaks can be used to explain a number ofdifferent situations which arose in the analysis of mixtures andunknowns. A representative spectrum from each of the explosivesanalyzed and cataloged is presented in the Supplementary data.

Unique peaks were defined as a peak which could be assigned toonly one explosive. These peaks are extremely useful in identifyingexplosives–though not all explosives were shown to have uniquepeaks. In the case of NB and 1,3-DNB, no unique peaks were found.Furthermore, isomeric explosives such as 2,4-DNT and 2,6-DNThave nearly all peaks in common with one another–the majordifference being the relative intensity of the peaks to one another.Similar trends were also seen in 2-A-4,6-DNT and 4-A-2,6-DNT.

In many instances, a particular peak could be assigned tomultiple different explosives and was labeled as a shared ion peak.In shared ion peaks, the fragments from each of the individualexplosives have identical molecular formulas, and therefore themolecular masses are also identical. An example of a shared ionpeak is presented in Fig. 3. A number of shared ion peaks have beenidentified and were found to occur between explosives of the sameclass (i.e. two or more nitroaromatics share the sample peak). Itwas determined that in a mixture or real world sample, when asearch list was being used, it may be difficult to assign anindividual peak to one particular explosive, and therefore all

Fig. 3. An example of a shared ion peak. Tetryl, picric acid, and ammonium picrate were

possible explosives were listed as potential matches for thesepeaks. By identifying other peaks present in the spectra, it ispossible to rule out one or more of the explosives listed for a sharedion peak

The other types of peaks identified were mass overlap peaks,which can be difficult to interpret. Mass overlap peaks are definedas peaks from two or more ions, from different explosives, whichhave theoretical masses within �0.005 amu of one another. Each ofthe peaks has a different molecular formula, and therefore a slightlydifferent molecular mass. An example of a mass overlap instance isshown in Fig. 4. A list of mass overlap occurrences is presented inTable 3. Since a number of mass overlap peaks involve Tetryl andDEGDN, identification of peaks pertaining to these explosives can bedifficult, as a number of possibilities can occur depending on thecalibration of the individual run. If a run is well calibrated, a particularpeak may hit with the correct assignment solely, or as a combinationof possible peaks. If the calibration is slightly off, then the peak may bemis-assigned as a hit from the explosive which shares the massoverlap peak–as opposed to not being assigned the proper identity.Using this information, it is possible to explain the presence of certainpeaks when they are not expected. Furthermore, an isotopic peakfrom a nominal mass one less than a mass overlap peak will also fallwithin the 0.005 amu tolerance of the mass overlap peaks, which addsa third possible peak which can be detected, depending on theexplosive(s) present.

Examining the mass overlap occurrences it was noted that themass overlap occurrences exhibited nearly identical character-istics. In all instances one of the mass overlap peaks is a straightchain nitro (DEGDN and/or EGDN) while the other is a nitroaro-matic (typically Tetryl). Furthermore, in every instance, thedifference in mass between the two peaks which are not isotopesis always 2.68 mmu. The differences in chemical formulas are alsoidentical in every occurrence. This could be attributed to similarfragmentation and recombination patterns amongst explosives aswell as the similarities in the molecular formula ratios amongst thetwo classes. Using these phenomena, all anomalies which wereencountered in the analyses of mixtures and blind samples couldbe explained.

3.5. Analysis of mixtures

To evaluate the ability of the technique to detect multipleexplosives simultaneously, two mixtures were analyzed. The firstmixture was a 6 component mixture containing TNT, 2,6-DNT,

all shown to form the oxygenated trinitrobenzene with a nominal mass of 227 amu.

Fig. 4. An example of mass overlap. The M* ion of Tetryl (A) and a dimer of EGDN (B) have masses which lie within 5 mmu of one another.

E. Sisco et al. / Forensic Science International 232 (2013) 160–168 165

PETN, RDX, NG, and EGDN–with concentrations ranging between20–40 mg/mL. It was analyzed using both the ‘‘high temperature’’and ‘‘low temperature’’ methods. The ‘‘low temperature’’ methodwas used solely to screen for DEGDN and EGDN. A representativespectrum of a ‘‘high temperature’’ run is shown in Fig. 5. In thefifteen replicates which were completed using the ‘‘high tempera-ture’’ method, all explosives except EGDN were identified in everyreplicate by the searching against the search list. More specifically,peaks which were unique to TNT, PETN, RDX, and NG weredetected in all replicates, and peaks which are shared between 2,6-DNT and 2,4-DNT were detected in all replicates. Only one replicateproduced a hit for EGDN. Difficulties in detecting EGDN wereexpected since it responds poorly at the higher temperature.However, analysis of the mixture using the ‘‘low temperature’’method also only produced one hit for EGDN, indicating thatcompetitive ionization may be occurring which is hindering thedetection of EGDN in a mixture.

The second mixture was a 14 component mixture containingTNT, 2,4-DNT, 2,6-DNT, Tetryl, NB, 1,3-DNB, 1,3,5-TNB, 2-NT, 3-NT,4-NT, 2-A-4,6-DNT, 4-A-2,6-DNT, RDX, and HMX. This mixture wasfar more complex than the previous mixture; however, similarresults were obtained. A representative spectrum of this mixture isshown in Fig. 6. In all replicates, peaks for every explosive wereidentified after being searched against the library. Unique peakswere identified in all replicates for TNT, Tetryl, RDX, HMX, 4-A-2,6-DNT, and 1,3,5-TNB. Detection of unique peaks for 2,4-DNT, 2-A-4,6-DNT and 2,6-DNT as well as the nitrotoluenes, NB, and 1,3-DNBwere not achieved, due to the ion sharing which occurs betweenlike explosives. Even though all explosives were detected, ananomaly was observed as some searches produced hits for DEGDN,due to mass overlap with a nitroaromatic explosive. As Tetryl andDEGDN share a number of the same nominal masses, hits forDEGDN were produced in each replicate. However, it was possibleto rule out the presence of DEGDN by searching for the presence ofisotopes and through the absence of additional DEGDN peaks.

Table 3A list of the mass overlap occurrences which were identified.

Nominal mass Theoretical mass Assignmen

228 228.020452 TNT 13C iso

228.022983 DEGDN

228.025662 TNT -or- 2-

242 242.023305 Tetryl 13C i

242.026057 EGDN -or-

242.028736 Tetryl

258 258.018227 Tetryl 13C i

258.020972 EGDN -or-

258.023651 Tetryl

259 259.026057 Tetryl 13C i

259.028797 DEGDN

259.031476 4-A-2,6-DN

3.6. Blind sampling

A blind sampling study was also completed to evaluate theability of the technique to identify unknowns and evaluate any biason the part of the examiners. In this study, twelve mock-casesamples were prepared by an independent examiner. The sampleswere run, in triplicate, using the high temperature method inaddition to the low temperature method–which was used solely toscreen for DEGDN and EGDN. The results of the study are shown inTable 4. In every sample, peaks unique to the explosive presentwere identified, with two or more unique peaks being present in allbut two samples. Furthermore, the blank sample produced nosearch list hits, indicating that the search list was not falselyidentifying peaks which were due to background contaminants.The blind sampling results demonstrate that there was no biaspresent in either the method or the examiners, and that the use of asearch list to detect the presence of an explosive was reliable.

4. Results and discussion–positive mode (peroxide basedexplosives)

4.1. Effects of altering DART parameters

Since the peroxide based explosives readily produce positiveions, a positive ion detection method was also developed. The twoperoxide explosives which were analyzed were TATP and HMTD.As stated in Section 2.2, a glycol ether mixture was used as a masscalibrant, and methyl decanoate was the independent QA/QCcompound. When the method was optimized, it was found thatmaximum signal was achieved with a gas temperature of 125 8Cand an orifice 1 voltage of +5 V, and no dopant. Increasing thetemperature from 125 8C up to 200 8C provided a marginaldecrease in the signal of both TATP and HMTD. The largestdecrease in signal occurred when ammonium hydroxide wasintroduced as a dopant. The signal for TATP was decreased by

t Formula Mass difference

tope C7H5N3O6 2.53 mmu

C4H8N2O9

A-4,6-DNT C7H6N3O6 2.68 mmu

sotope C7H5N4O6 2.75 mmu

DEGDN C4H8N3O9

C7H6N4O6 2.68 mmu

sotope C7H5N4O7 2.75 mmu

DEGDN C4H8N3O10

C7H6N4O7 2.68 mmu

sotope C7H6N4O7 2.75 mmu

C4H9N3O10

T C7H7N4O7 2.68 mmu

Fig. 5. A representative spectrum of the six component mixture. Five of the six components are readily detectable.

E. Sisco et al. / Forensic Science International 232 (2013) 160–168166

approximately a factor of four, while the signal of HMTD wasdecreased by approximately half. No additional fragmentation oradduct production were noted with the introduction of ammoni-um hydroxide. Fig. 7 shows the average signal intensities of thebase peak of TATP and HMTD under a number of differentconditions.

4.2. Method reproducibility

A reproducibility study identical to that carried out for thenegative mode was completed in the positive mode. The threemost abundant peaks for both TATP and HMTD were monitored inaddition to six peaks for the glycol ether mixture and one peak formethyl decanoate. With a �0.005 amu tolerance, it was determinedthat all masses measured fell within the threshold limit, providing

Fig. 6. A representative mass spectra of the 14 component mixture. Select peaks are highl

noted.

100% mass accuracy for the positive mode. The reason 100% massaccuracy may have been achieved is due to the lower number massesanalyzed (392 in positive mode versus 976 in negative mode) as wellas overall higher abundance of the three most abundant peaks in TATPand HMTD, as compared to explosives like DEGDN and HMX, whichrepeatedly had low intensity responses.

4.3. Limit of detection

The limit of detection for both TATP and HMTD weredetermined using the optimized method on the DART-MS, andwere compared to the limit of detections observed on the EI-GC-MS. For both TATP and HMTD, the limit of detection of the DART-MS was determined to be 10 ng. An identical limit of detection wasfound for TATP on EI-GC-MS, while the HMTD was undetectable

ighted to indicate which explosives were detected. Not all peaks which produced are

Table 4Results of the blind sampling study.

Sample Contents # of unique peaks Additional explosives hits

Shared ion Mass overlap

1 AP 4* *PA, Tetryl None

2 Blank 0 None None

3 PETN 2 ETN None

4 HMX 4 RDX None

5 RDX 3 HMX None

6 DEGDN 2 EGDN Tetryl, TNT, & 2-A-4,6DNT

7 ETN 3 None None

8 2,6-DNT 3* *2,4-DNT, TNT None

9 2-A-4,6-DNT 3* *4-A-2,6-DNT DEGDN

10 Blank 0 None None

11 NG 1 None None

12 Tetryl 4 AP/PA & TNT DEGDN & EGDN

The column titled ‘‘# of unique peaks’’ identifies how many peaks were detected that can only be attributed to that explosive. An asterik indicates the explosive shares all

peaks with another explosive (also indicated with an asterik) and therefore the number of peaks unique to that pair of explosives which were detected is listed. The last two

columns indicate the identities of hits which were produced and are shared between explosives either due to shared ions or mass overlap.

E. Sisco et al. / Forensic Science International 232 (2013) 160–168 167

using this technique. These results, as well as those presented forthe negative mode, illustrate the ability of DART to be an extremelyuseful screening technique for a number of explosives.

4.4. Analysis of mixtures

A 1:1 mixture of TATP and HMTD was prepared and analyzed toevaluate the ability for the DART-MS to detect both explosivessimultaneously. A search list was also created for the peroxideexplosives, and it was found that two peaks were shared ion peaksbetween TATP and HMTD. TATP had one additional peak whichwas unique to that explosive while HMTD had four unique peaks.When the mixture was analyzed using the positive method, allreplicates produced results where at least one unique peak forHMTD was detected. In 75% of the replicates two or more peaksunique to HMTD were detected. Furthermore, both of the peaksshared by TATP and HMTD were detected in all replicates. Roughly25% of the replicates produced the sole unique TATP peak, with

Fig. 7. Average intensity of the base peak of TATP (A) and HMTD (B) under a number of d

the dopant used. Runs where a dopant was not introduced are indicated with ‘‘ND’’.

intensities under 6% relative to the base peak in all cases. Fromthese results it is unclear whether the lack of detection of theunique TATP in many of the sample runs is due to the competitiveionization of HMTD or if it is due to the intensity for the uniqueTATP peak being below the 1% relative intensity threshold utilizedin the searches. Lowering the detection threshold to 0.5% relativeintensity did lead to 95% detection of the unique TATP peak,indicating it is likely low abundances that was leading to the lack ofdetection.

4.5. Blind sampling

A twelve unknown blind sampling study was completedidentically to that used for the negative mode. It was determinedthat the blank runs produced no search list hits, indicating nobackground ions were being identified as peaks belonging to anexplosive. Furthermore, all samples in which HMTD was presentlead to the detection of unique HMTD peaks in addition to the two

ifferent conditions. The parameters listed are gas temperature, orifice 1 voltage, and

E. Sisco et al. / Forensic Science International 232 (2013) 160–168168

shared ion peaks. Samples in which TATP was present lead to thedetection of just the two ion shared peaks, further supporting thetheory that the unique TATP peak is not detected due to lowabundances. The results of this study indicate that TATP and HMTDcan be differentiated by the presence or absence of the uniqueHMTD peaks.

5. Conclusions

The DART-MS technique provides an extremely powerfulscreening tool for the analysis of trace explosives. It has beenshown to accurately and reproducibly detect a wide range ofexplosives, with sensitivities equal to or exceeding that of currentanalytical techniques. However, due to the lack of a seconddimension, such as a separatory technique, confirmatory identifi-cation of specific explosives can be difficult due to similarfragmentation patterns of explosives in the same class. Detectingmultiple explosives simultaneously is easily achieved with thetechnique, and the use of a search list to analyze unknowns hasbeen demonstrated to be highly effective. While false positives canbe encountered in analysis, they are easily accounted for by sharedions and mass overlap. Furthermore, the use new mass calibrationcompounds expand the possibilities of potential calibrants outsideof the standard PEG 600.

While this study looked at a number of explosives, there isadditional work which could be completed. This includes, lookingat solvent background effects for analysis by DART, determinationof possible substrate interferences, and analysis of real-world preand post blast samples. Also, additional studies could look into thepossibility of adding a second dimension to the analysis to allowfor confirmatory analysis to be completed using DART-MS alone.As a screening tool, the DART-MS shows promise as a leadingtechnique in the detection of trace explosives, as well as a slew ofother forensic samples.

6. Disclaimer

The opinions or assertions contained herein are the privateviews of the author and are not to be construed as official or asreflecting the views of the Department of the Army or theDepartment of Defense.

Acknowledgments

The authors would like to recognize the help and adviceprovided by Dr. Alice Mignerey of the University of Maryland, Dr.Jeff Salyards of the United States Army Criminal InvestigationsLaboratory, and Dr. Greg Gillen of the National Institute ofStandards and Technology. The primary author also acknowledgessupport from the Department of Defense–Science, Mathematics,and Research for Transformation fellowship.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.forsciint.2013.07.006.

References

[1] F.M. Green, T.L. Salter, P. Stokes, I.S. Gilmore, G. O’Connor, Ambient mass spec-trometry: advances and applications in forensics, Surf. Interface Anal. 42 (2010)347–357.

[2] R.B. Cody, J.A. Laeamee, D.H. Durst, Versatile new ion source for the analysis ofmaterials in open air, Anal. Chem. 77 (2005) 2297–2302.

[3] E.S. Chernetsova, G.E. Morlock, I.A. Revelsky, DART mass spectrometry and itsapplications in chemical analysis, Russ. Chem. Rev. 80 (2011) 235–255.

[4] A.M. Pfaff, R.R. Steiner, Development and validation of AccuTOF-DARTTM as ascreening method for analysis of bank security device and pepper spray compo-nents, Forensic Sci. Int. 206 (2011) 62–70.

[5] R.W. Jones, R.B. Cody, J.F. McClelland, Differentiating writing inks using directanalysis in real time mass spectrometry, J. Forensic Sci. 51 (2006) 915–918.

[6] S.E. Howlett, R.R. Steiner, Validation of thin layer chromatography with AccuTOF-DARTTM detection for forensic drug analysis, J. Forensic Sci. 56 (2011) 1261–1267.

[7] R.R. Steiner, R.L. Larson, Validation of the direct analysis in real time source for usein forensic drug screening, J. Forensic Sci. 54 (2009) 617–622.

[8] E. Chernetsova, P. Bochkov, G. Zatonskii, R. Abramovich, New approach to detect-ing counterfeit drugs in tablets by DART mass spectrometry, Pharm. Chem. J. 45(2011) 306–308.

[9] J.M. Nilles, T.R. Connell, H.D. Durst, Quantitation of chemical warfare agents usingthe direct analysis in real time (DART) technique, Anal. Chem. 81 (2009) 6744–6749.

[10] J. Hajslova, T. Cajka, L. Vaclavik, Challenging applications offered by direct analysisin real time (DART) in food-quality and safety analysis, Trends Anal. Chem. 30(2011) 204–218.

[11] J.M. Nilles, T.R. Connell, S.T. Stokes, D.H. Durst, Explosives detection using directanalysis in real time (DART) mass spectrometry, Propel. Explos. Pyrot. 35 (2010)446–451.

[12] F. Rowell, J. Seviour, A.Y. Lim, C.G. Elumbaring-Salazar, J. Loke, J. Ma, Detection ofnitro-organic and peroxide explosives in latent fingermarks by DART- and SALDI-TOF-mass spectrometry, Forensic Sci. Int. 221 (2012) 84–91.

[13] L. Song, A.B. Dykstra, H. Yao, J.E. Bartmess, Ionization mechanism of negative ion-direct analysis in real time: a comparative study with negative ion-atmosphericpressure photoionization, J. Am. Soc. Mass Spectrom. 20 (2009) 42–50.

[14] M. Nambayah, T.I. Quickenden, A quantitative assessment of chemical techniquesfor detecting traces of explosives at counter-terrorist portals, Talanta 63 (2004)461–467.

[15] D.S. Moore, Instrumentation for trace detection of high explosives, Rev. Sci.Instrum. 75 (2004) 2499–2512.

[16] R.G. Ewing, M.J. Waltman, D.A. Atkinson, Characterization of triacetone triper-oxide by ion mobility spectrometry and mass spectrometry following atmospher-ic pressure chemical ionization, Anal. Chem. 83 (2011) 4838–4844.

[17] G.A.E. Karpas, Z. Karpas, Ion Mobility Spectrometry, Second Edition, CRC Press,2005.

[18] K.G. Furton, L.J. Myers, The scientific foundation and efficacy of the use of caninesas chemical detectors for explosives, Talanta 54 (2001) 487–500.

[19] R.J. Harper, J.R. Almirall, K.G. Furton, Identification of dominant odor chemicalsemanating from explosives for use in developing optimal training aid combina-tions and mimics for canine detection, Talanta. 67 (2005) 313–327.

[20] D.S. Moore, Recent advances in trace explosives detection instrumentation, SensImaging. 8 (2007) 9–38.

[21] D. Moore, R. Scharff, Portable Raman explosives detection, Anal. Bioanal. Chem.393 (2009) 1571–1578.

[22] E.M.A. Ali, H.G.M. Edwards, I.J. Scowen, Raman spectroscopy and security appli-cations: the detection of explosives and precursors on clothing, J. Raman Spec-trosc. 40 (2009) 2009–2014.

[23] C. Eliasson, N.A. Macleod, P. Matousek, Noninvasive detection of concealed liquidexplosives using Raman spectroscopy, Anal. Chem. 79 (2007) 8185–8189.

[24] J. Almog, S. Kraus, B. Glattstein, ETK-an operational explosive testing kit, J. Energ.Mater. 4 (1986) 159–167.

[25] N. Talaty, C.C. Mulligan, D.R. Justes, A.U. Jackson, R.J. Noll, R.G. Cooks, Fabricanalysis by ambient mass spectrometry for explosives and drugs, Analyst 133(2008) 1532–1540.

[26] J.F. Garcia-Reyes, J.D. Harper, G.A. Salazar, N.A. Charipar, Z. Ouyang, R.G. Cooks,Detection of explosives and related compounds by low-temperature plasmaambient ionization mass spectrometry, Anal. Chem. 83 (2010) 1084–1092.

[27] I. Cotte-Rodrıguez, H. Chen, R.G. Cooks, Rapid trace detection of triacetonetriperoxide (TATP) by complexation reactions during desorption electrosprayionization, Chem. Commun. 9 (2006) 953.

[28] I. Cotte-Rodrıguez, Z. Takats, N. Talaty, H. Chen, R.G. Cooks, Desorption electrosprayionization of explosives on surfaces: sensitivity and selectivity enhancement byreactive desorption electrospray ionization, Anal. Chem. 77 (2005) 6755–6764.

[29] N. Na, C. Zhang, M. Zhao, S. Zhang, C. Yang, X. Fang, Direct detection of explosiveson solid surfaces by mass spectrometry with an ambient ion source based ondielectric barrier discharge, J. Mass Spectrom. 42 (2007) 1079–1085.

[30] I.A. Popov, H. Chen, O.N. Kharybin, E.N. Nikolaev, R.G. Cooks, Detection ofexplosives on solid surfaces by thermal desorption and ambient ion/moleculereactions, Chem. Commun. 13 (2005) 1953–1955.

[31] Y. Song, R.G. Cooks, Atmospheric pressure ion/molecule reactions for the selectivedetection of nitroaromatic explosives using acetonitrile and air as reagents, RapidCommun. Mass Spectrom. 20 (2006) 3130–3138.

[32] Z. Takats, J.M. Wiseman, R.G. Cooks, Ambient mass spectrometry using desorptionelectrospray ionization (DESI): instrumentation, mechanisms and applications inforensics, chemistry, and biology, J. Mass Spectrom. 40 (2005) 1261–1275.

[33] M. Zhang, Z. Shi, Y. Bai, Y. Gao, R. Hu, F. Zhao, Using molecular recognition of b-cyclodextrin to determine molecular weights of low-molecular-weight explosivesby maldi-tof mass spectrometry, J. Am. Soc. Mass Spectrom. 17 (2006) 189–193.

[34] E. Holmgren, H. Carlsson, P. Goede, C. Crescenzi, Determination and characteri-zation of organic explosives using porous graphitic carbon and liquid chroma-tography-atmospheric pressure chemical ionization mass spectrometry, J.Chromatogr. A 1099 (2005 Dec) 127–135.

[35] R. Schulte-Ladbeck, U. Karst, Determination of triacetonetriperoxide in ambientair, Anal. Chim. Acta 482 (2003) 183–188.