lable at ScienceDirect

Analytica Chimica Acta 892 (2015) 1e9

Contents lists avai

Analytica Chimica Acta

journal homepage: www.elsevier .com/locate/aca

In-source collision induced dissociation of inorganic explosives formass spectrometric signature detection and chemical imaging*

Thomas P. Forbes*, Edward SiscoNational Institute of Standards and Technology, Materials Measurement Science Division, Gaithersburg, MD, USA

h i g h l i g h t s

* Official contribution of the National Institute of Stsubject to copyright in the United States.* Corresponding author.

E-mail address: thomas.forbes@nist.gov (T.P. Forbe

http://dx.doi.org/10.1016/j.aca.2015.06.0080003-2670/Published by Elsevier B.V.

g r a p h i c a l a b s t r a c t

� In-source CID enhanced detection ofelemental inorganics up to 1000-fold.

� In-source CID optimization of poly-atomic oxidizers enhanced detectionup to 100-fold.

� Optimal CID identified at transitionfrom breaking ionic salt to molecularanion bonds.

� Trace detection of inorganic explo-sives at nanogram levels wasdemonstrated.

� Oxidizer particles were chemicallyimaged directly from latentfingerprints.

a r t i c l e i n f o

Article history:Received 30 March 2015Received in revised form2 June 2015Accepted 7 June 2015Available online 8 July 2015

Keywords:In-source collision induced dissociationMass spectrometryFuel-oxidizer mixturesInorganic explosivesLaser desorption/ionizationChemical imaging

a b s t r a c t

The trace detection, bulk quantification, and chemical imaging of inorganic explosives and componentswas demonstrated utilizing in-source collision induced dissociation (CID) coupled with laser desorption/ionization mass spectrometry (LDI-MS). The incorporation of in-source CID provided direct control overthe extent of adduct and cluster fragmentation as well as organic noise reduction for the enhanceddetection of both the elemental and molecular ion signatures of fuel-oxidizer mixtures and other inor-ganic components of explosive devices. Investigation of oxidizer molecular anions, specifically, nitrates,chlorates, and perchlorates, identified that the optimal in-source CID existed at the transition betweenfragmentation of the ionic salt bonds and molecular anion bonds. The chemical imaging of oxidizerparticles from latent fingerprints was demonstrated, including both cation and anion components inpositive and negative mode mass spectrometry, respectively. This investigation demonstrated LDI-MSwith in-source CID as a versatile tool for security fields, as well as environmental monitoring and nu-clear safeguards, facilitating the detection of elemental and molecular inorganic compounds at nanogramlevels.

Published by Elsevier B.V.

andards and Technology; not

s).

1. Introduction

Significant research has been conducted advancing the accu-rate and sensitive detection of explosives, most frequently com-mon military-grade nitrated organic explosives, includingnitramines, nitroaromatics, and nitrate esters such as

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e92

cyclotrimethylenetrinitramine (RDX), trinitrotoluene (TNT), andpentaerythritol tetranitrate (PETN). These explosives have beendetected using a range of analytical techniques, including ionchromatography (IC) [1,2], capillary electrophoresis (CE) [3,4],high performance liquid chromatography (HPLC) [5], and mostnotably, ion mobility spectrometry (IMS) [6e13] and mass spec-trometry (MS) [14e22]. The rapid analysis times and cost effec-tiveness of these field compatible instruments have led towidespread deployment of ion mobility spectrometry. Neverthe-less, developments in ambient and atmospheric pressure ioniza-tion mass spectrometry have provided selective and sensitivealternatives, including a range of analyte ionization techniquesrequiring no or minimal sample preparation such as desorptionelectrospray ionization (DESI) [23] and direct analysis in real time(DART) [24]. A number of reviews and books provide extensivedetails on the many ambient ionization techniques and specifics ofexplosives detection [25e30].

The threats posed by homemade and improvised explosivedevices (HMEs/IEDs) continue to increase, necessitating thegrowth of techniques for the trace detection of their widelyvarying signatures. These HMEs and IED components includeunstable primary explosives (e.g., peroxides [31e34], azides, andfulminates), fuel-oxidizer and self-initiating mixtures, nitratedsugar alcohols, and radiological dispersion devices (RDDs)[34e37]. In the present work, we focus on the detection ofinorganic explosives and components, a rising area of research.Fuel-oxidizer mixtures may contain both organic (e.g., fuel-oil,sugar, or nitromethane) and inorganic (e.g., potassium chlorate,ammonium nitrate, or aluminum) components. Evans-Nguyenet al. [35] and Forbes and Sisco [34] recently demonstratedenhanced detection of elemental inorganic components of ex-plosives/explosive devices using in-source collision induceddissociation (CID) combined with liquid based ambient ionizationsources, transmission mode desorption electrospray ionization(TM-DESI) and desorption electro-flow focusing ionization(DEFFI). The incorporation of in-source CID led to 10-folde100-fold improvements in detection of elemental inorganic ions byincreasing fragmentation of adducts and clusters formed byinorganic compounds and decreasing chemical noise fromorganic compounds [34,35,38]. In addition, tuning of in-sourceCID has enabled control over charge reduction from the doublycharged to singly charged cations of alkaline earth metal ions[38]. These previous studies utilized liquid-based ion sources andfocused on how in-source CID affected the response of organicexplosives and inorganic elemental components of explosivesand explosive devices. The detection of inorganic particles wasdemonstrated by incorporating a 5% nitric acid aqueous solution.However, to circumvent the need for nitric acid, we present aninvestigation of laser-based desorption and ionization of inor-ganic particles and residues. Here, we investigate the effects ofin-source CID on the detection of inorganic explosives, bothelemental and molecular, highlighting inorganic oxidizers.

The analytical performance and detection optimization forinorganic explosives (e.g., fuel-oxidizer mixtures) using in-sourceCID is presented. CID was investigated using a laser desorption/ionization mass spectrometry (LDI-MS) system. The integrationof elevated in-source CID potentials led to increased ion accel-eration and collisions with atmospheric gas molecules, relative totypical CID (declustering) potentials, in the mass spectrometerinlet region. These more frequent and higher energy collisionsled to fragmentation of inorganic adducts and clusters. Thecharacteristic in-source CID response for inorganic ions wasdependent on the ion composition (e.g., elemental: Kþ vs mo-lecular: ClO4

�) as well as the originating salt composition (e.g.,ammonium nitrate vs potassium nitrate). The optimized

detection of inorganic explosive components was then used todemonstrate the direct detection of oxidizer particles from acomplex matrix (i.e., latent fingerprints), bulk quantification fromextracted collection regions that included both inorganic andorganic components, and chemical imaging of particles fromlatent fingerprints.

2. Experimental methods

2.1. Materials and sample preparation

Salts (potassium chlorate: PC, potassium perchlorate: PPC,potassium nitrate: PN, ammonium nitrate: AN, and calciumammonium nitrate: CAN) and metals (zinc, aluminum, magne-sium, strontium, and cobalt) were purchased from Sigma Aldrich(St. Louis, MO) and Inorganic Ventures (Christiansburg, VA),respectively. The salts were either crushed with a metal spatulafor direct particle analysis or dissolved in water and diluted torequired concentrations for parametric analyses. Similarly, ICP-MS standards from Inorganic Ventures were purchased at1 mg mL�1 in an aqueous solution with 0.1 %e3 % nitric acid (vol/vol) and further diluted as required. Liquid chromatography-massspectrometry (LC-MS) Ultra Chromasolv® grade water was pur-chased from Sigma Aldrich (St. Louis, MO) and used for all samplepreparation. Parametric investigations and chemical imagingwere conducted on LDI commercial substrates from HudsonSurface Technology (Old Tappan, NJ). Solution aliquots of 1 mL ofeach analyte concentration (500 pg mL�1 to 1 mg mL�1, dependingon the experiment) were spotted onto the patterned LDI sub-strates (PSU0106000: mFocus MALDI Plate). Direct detection andchemical imaging of complex matrices were conducted fromlatent fingerprints deposited by an anonymous volunteer directlyonto a blank LDI substrate (PSG0103000: NonFocus HST Array forImaging). Latent fingerprints were collected per NIST InstitutionalReview Board Case 407 protocol and no volunteers handled orwere exposed to exogenous inorganic chemicals. Fingerprintdeposition and collection was followed by the dry transfer ofparticles of relevant inorganic oxidizer onto the latent finger-print. Explosive residues generated during a homemade materialsynthesis of a simulated explosive device were collected on ad-hesive sheets (samples provided by Signature Science, Charlot-tesville, VA). Three individual samples were collected from a fuel-oxidizer mixture synthesis containing potassium perchlorate andicing sugar. Each 15 mm � 20 mm sample section of the adhesivesheet, containing collected inorganic and organic compounds,was placed into a 5 mL vial and extracted into 2 mL of water,mixed for 10 min on a rotating mixer at 20 rpm, and then vor-texed for 60 s. Standards for linear least squares calibration wereprepared gravimetrically using milligram quantities of crystallinesalt analytes with manufacturer designated purities of �99.0%(AT21 comparator balance, Mettler Toledo, Columbus, OH).Nominal solution concentrations of 1 mg mL�1 were preparedand gravimetrically diluted.

Certain commercial products are identified in order toadequately specify the procedure; this does not imply endorsementor recommendation by NIST, nor does it imply that such productsare necessarily the best available for the purpose.

2.2. Instrumentation

2.2.1. Mass analyzer and imagingA 4000 QTrap MS/MS system, consisting of a hybrid triple

quadrupole/linear ion trap, (Applied Biosystems/MDS Sciex, FosterCity, CA/Toronto, Canada) and a 35 mm extended capillary heatedinterface was used in this experiment (Fig. 1), with additional

Fig. 1. Schematic representation of the laser desorption/ionization (LDI) mass spec-trometry imaging setup, including in-source collision induced dissociation (CID).

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e9 3

details of the instrument parameters and operation found in theliterature [20,39]. Briefly, experiments included the following sys-tem parameters: 138 kPa curtain gas (N2), 83 kPa ion source gas 1(N2), 0.0 kPa ion source gas 2 (N2), 100 �C interface temperature, ±10 V entrance potential, and 1.1 � 10�3 kPa to 1.3 � 10�3 kPaoperating vacuum. In-source CID of various analyte adducts/clustersand organic compounds was achieved at the instrument orifice,between the curtain plate and skimmer cone (Fig. 1). The extent ofCID was manipulated through the instrument's declustering po-tential in the differentially pumped region. Principally, the charac-terization of in-source CID of inorganic explosives was conductedutilizing the first quadrupole for mass analysis and the remainingquadrupoles as ion guides. In-source CID potentials were variedfrom ±25 V to ±375 V for positive and negative mode experiments.Samples were interrogated across a specifiedm/zwindow for 5 s ateach CID potential interval. Tandem mass spectrometry (MS/MS)product ion scans were used for the identification of various ad-ducts/clusters and the characterization of molecular ion fragmen-tation. Here, a specific mass of interest was selected by the firstquadrupole, fragmented by the second quadrupole, and scanned bythe third quadrupole. The laser desorption/ionization source (de-tails below) was synchronized with the instrument and dataacquisition software for mass spectrometry imaging (MSI) experi-ments. Details of peripheral device synchronization can be found inthe literature [39]. Chemical image datawas collected from 5mm�5 mm regions of interest divided into 100 mm � 100 mm pixels andscanned at a constant 12.5 mm/min velocity in a unidirectionalpattern. Imaging and image analysis of the mass spectrometric datawas completed withMSiReader, a MATLAB-based freeware package(v0.04, W. M. Keck FT-ICR Mass Spectrometry Laboratory, NorthCarolina State University) [40].

2.2.2. Laser desorption/ionizationAn atmospheric pressure matrix assisted laser desorption ioni-

zation source (AP-MALDI, MassTech, Inc., Columbia, MD) wascoupled to the QTrap Triple-Quadrupole mass spectrometer andused without alteration. The system provided direct analysis ofinorganic species without matrix-assisted ionization, solvent, orother sample preparation. Briefly, the LDI system included a355 nm wavelength Ng:Yag laser, pulsed at 150 Hz, with 150 mJpulse�1 and 3 nse5 ns duration. A ± 3000 V potential was appliedto the sample substrate for all positive and negative mode experi-ments, respectively (Fig. 1).

2.2.3. Support instrumentationSupplementary bulk quantification of extracted samples was

completed using ion chromatography (IC) and inductively coupledplasma (ICP)-MS. A Dionex 5000-IC system (Thermo Scientific,Bannockburn, IL) with a CS-18 cation ion exchange column wasutilized. The analysis included a 25 mL injection, 30 �C columntemperature, 0.5 mL min�1 eluent flow rate of 7 mM meth-anesulfonic acid buffer in ion chromatography grade water (Sig-maeAldrich), and a 35 �C detector compartment with an 11 mAsuppression on the total conductivity detector. A calibration curvefor potassium chlorate in the range of 0 ng mL�1 to 50 ng mL�1 wascreated and used to quantify triplicate injections of the extractions.Similarly, the extracted samples were also quantified using an iCAP-Q ICP-MS (Thermo Scientific, Bannockburn, IL). The samples wereintroduced into an argon plasma at 0.4 mL min�1 via nebulization.Again, a calibration curve ranging from 0 ng mL�1 to 50 ng mL�1 wascreated and used to quantify triplicate measurements of eachextraction sample.

2.3. Evaluation of uncertainty

The uncertainty in quantification of unknown samples fromlinear least squares calibration was evaluated in accordance withstandard guidelines [41,42]. Sources of measurement uncertaintywere considered from the gravimetric preparation of standard so-lutions and the linear least squares calibration. The uncertainty inthe calibration standard concentration included components fromthe solute purity, balance calibration, and fluctuations in roomtemperature. The purity of the compounds used here was given as� 99.0% by the manufacturer. Without additional information, apurity and uncertainty of 99.5 ± 0.5% was assumed, resulting in anexpanded uncertainty (k ¼ 2) of 1.0%. The uncertainty in the taredweighing included the reproducibility (2 mg, AT21 comparatorbalance, Mettler Toledo), readability (1 mg), and the sensitivity/linearity of the balance calibration function (±8 mg). These un-certainties from the balance technical specifications were used toestimate the total expanded uncertainty (k ¼ 2) as 0.33%. Standardpreparation was completed in an environmentally controlled lab-oratory with ambient temperature of (23 ± 1)�C. For the millilitervolumes considered here and assuming a coefficient of volumeexpansion for water of 2.1 � 10�4 �C�1, the expanded uncertainty(k ¼ 2) in the volume due to temperature was 0.04%. Theseexpanded uncertainties were combined in quadrature for a totaluncertainty in the calibration curve standard concentration of 1.1%.

The uncertainty in the quantification of samples was evaluatedfrom a linear least squares calibration curve comprised of n un-weighted data points (xi, yi) and p unknown sample measurements.Here x and y represented the standard concentration (c) and themeasurement intensity (e.g., signal intensity, I). The calibrationcurve was given by, y ¼ mxþ b, where m and b are the slope andintercept. The standard uncertainty in the unknown concentrationdue to variability in the response was determined by,

uðco; IÞ ¼ S=mffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1=pþ 1=nþ ðco � cÞ2=Sxx

q, [41] where S is the root

mean square error (determined by curve fitting with MATLAB,Version 2013a, The Mathworks Inc., Natick, MA) [43], co is theevaluated concentration of the unknown sample, c is the average

concentration from the curve values, Sxx ¼P ðci � cÞ2, and ci is the

concentration of the ith calibration point. The uncertainty in anunknown concentration due to variability in the reference cali-bration values, ci, was approximated as, uðco; xÞ ¼ uðciÞ=n [41]. Asdemonstrated below, the uncertainty in the quantification of un-known samples due to variability in the system response out-weighed the variability due to the reference standardconcentrations.

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e94

3. Results and discussion

3.1. Mass spectral characteristics of explosive oxidizers

Inorganic explosive oxidizers were analyzed by laser desorp-tion/ionization mass spectrometry (LDI-MS) with in-source CID.Sample substrates were oriented orthogonal to the mass spec-trometer inlet as displayed in Fig. 1. Themain oxidizers investigatedhere included PC, PPC, PN, AN, and CAN. Fig. 2 demonstrates therepresentative spectra for 1 mg PPC in positive and negative ionmode at high (þ300 V) and low (�100 V) in-source CID potential,respectively. The application of an elevated in-source CID potentialled to acceleration of ion clusters and adducts, resulting inincreased fragmentation. The increase in fragmentation of largerclusters and adducts enhanced the detection of elemental inorganicions by 10-folde1000-fold (depending on the compound) for LDI-MS.

In positive mode at high in-source CID (Fig. 2a), the massspectrum for PPC shows elemental cations and isotopic distribu-tions for potassium (m/z 39 Kþ), chromium (m/z 52 Crþ), iron (m/z56 Feþ), and a potassium chlorine cluster (m/z 113 K2Clþ). Theisotopic distribution and tandem mass spectrometry (MS/MS) ofthe m/z 113 peak corroborated identification (Fig. 2a(inset)). Thechromium and iron peaks were identified in all positive modespectra and attributed to components of the metal substrate usedfor parametric experiments. In negative mode at lower in-sourceCID (Fig. 2b), the mass spectrum for PPC displayed major poly-atomic anion peaks and isotopic distributions for perchlorate (m/z99 ClO4

�) and a potassium perchlorate cluster (m/z 237 KðClO4Þ2�),which was confirmed with an MS/MS scan at 54 V (Fig. 2b(inset)).

The remaining inorganic oxidizers displayed similar massspectra in positive and negative ion mode. Briefly, in high CID

Fig. 2. Representative LDI mass spectra of 1 mg potassium perchlorate sampled for 30 s inrespectively. Insets: corresponding MS/MS product ion spectra of the m/z 113, K2Clþ, and m/MS.

positivemode, PC and PN displayed cation peaks for potassium (m/z39 Kþ), chromium (m/z 52 Crþ), and iron (m/z 56 Feþ). The potas-sium chlorine cluster (m/z 113 K2Clþ) observed for PPC was alsoobserved for PC. The molecular cation for ammonium fell outsidethe detectable m/z range of the present instrument and could notbe observed. The negative ion mass spectra of PC at low in-sourceCID displayed the chlorate anion (m/z 83 ClO3

- ), and a potassiumchlorate cluster (m/z 205 KðClO4Þ2�), which displayed the correctisotopic distributions (MS/MS in supporting informationFigure S1a). Both nitrate (m/z 62 NO3

�) and nitrite (m/z 46 NO2�)

were observed for PN, AN, and CAN in negative mode. Similar to PCand PPC, a cluster of potassium and nitrate (m/z 163 KðNO3Þ2�) wasobserved for PN and corroborated by MS/MS (Figure S1b).

3.2. In-source CID optimization

Following an initial characterization of the main peaks observedfor each inorganic oxidizer, an in-depth characterization of in-source CID was completed using residues/reconstituted particlesfrom solution deposits. In-source CID was utilized to directly con-trol the extent of cluster and adduct formation and fragmentation.In addition, the extent of molecular anion fragmentation could bedirectly controlled with in-source CID. As discussed in detail else-where [34,44e46], unlike the specificm/z fragmentation of MS/MS,CID impartially induces acceleration of all ions and collisions withatmospheric gas molecules in the MS inlet region. Here, eachoxidizer was interrogated while varying the extent of in-source CIDthrough the instrument declustering potential. Potentials wereincreased from ±25 V to ±375 V in 25 V increments for positive andnegative mode experiments, respectively, collecting a mass spec-trum for 5 s at each interval.

(a) positive and (b) negative mode, at high (þ300 V) and low (�100 V) in-source CID,z 237, KðClO4Þ2� , clusters at ± 54 V collision cell voltage in positive and negative mode

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e9 5

Fig. 3 displays the MS signal (integrated peak area), normalizedby the maximum signal, for the anions of each inorganic oxidizer asa function of increasing in-source CID potential. Data points anduncertainties represent the average peak area and standard de-viations for 5 to 8 replicate samples. The signal response of thesepolyatomic inorganic anions differed from those previous found fororganic explosives and elemental cations [34]. Previous workinvestigating organic explosives demonstrated a nearly mono-tonically decreasing response as a function of in-source CID po-tential. However, as seen in Fig. 3, the molecular anions, derivedfrom salt starting materials, displayed an initial increase in signalwith in-source CID, a local maximum, followed by a decrease ofvarying degree. In addition to variability in the rate of decrease insignal beyond the maximum, the in-source CID potential foroptimal signal intensity was dependent on the analyte. Thesephenomena were determined to be directly related to the ionicbond strength of the starting ionic material (e.g., potassium chlo-rate), and the bond strength/molecular stability of the molecularanion (e.g., chlorate). In the regime of increasing signal withincreasing CID potential, the stronger the ionic bond, the lower theinitial signal strength (i.e., at � 25 V CID) and the higher the in-source CID potential at the maximum signal. Here, the MSresponse trends correlated with the energy necessary to break theionic bonds and fragment salt-anion clusters (e.g. , Fig. 2b). Asdemonstrated in Fig. 3, the overall salt bond strength, via enthalpyof formation, increased from AN (�365.6 kJ mol�1), CAN (unknowncombination of ammonium nitrate and calcium nitrate with en-thalpies of formation of �365.6 kJ mol�1 and �938.2 kJ mol�1,respectively), PC (�397.7 kJ mol�1), PPC (�432.8 kJ mol�1), to PN(�494.6 kJ mol�1), resulting in optimal in-source CID potentialsof �100 V, �125 V, �125 V, �150 V, and �275 V, respectively.

Following the optimal extent of CID for breaking up the cation-anion ionic bond and salt-anion clusters, further increase in theenergy and frequency of collisions resulted in fragmentation of theanion molecule, e.g., ClO3

�. The overall anion bond strength, againvia enthalpy of formation, increased from chlorate (�98.3 kJ mol�1),to perchlorate (�131.4 kJ mol�1), to nitrate (�206.6 kJ mol�1). Thedirect visualization of these trends in Fig. 3 was convoluted by anumber of factors, specifically variability in the energy threshold forbreaking the salt ionic bonds among compounds and the mixture ofammonium nitrate and calcium nitrate found in CAN. For example,when comparing the responses of PN and AN, at a mid-range in-

Fig. 3. Normalized signal abundance as a function of in-source CID (declustering)potential magnitude for inorganic molecular anions from 1 mg explosive oxidizers:NO3

� from AN, NO3� from CAN, -◊-ClO3

� from PC, -C- ClO4� from PPC,

NO3� from PN. Data points and uncertainty expressed as the average peak area and

standard deviations for 5 to 8 successive replicate samples, respectively.

source CID potential (e.g., �200 V), a larger fraction of the impartedenergywent into breaking of the potassiumenitrate bond relative tothe ammonium-nitrate bond. More specifically, the majority of theavailable energy went into breaking the potassium-nitrate ionicbond, while only a fraction was needed to break the ammonium-nitrate ionic bond and the remaining available energy went intofragmenting the nitrate anion (Fig. 3). The enthalpy of formation,and bond strength, of the PN salt was the highest considered hereand resulted in a wide local maximum and minimal fragmentationof the nitrate anion.

The anion fragmentation behavior was further investigated us-ing tandem mass spectrometry (MS/MS) of AN, PC, and PPC.Product ion mass spectra for each anion resulted in expectedfragments (Figure S2). Similar to scanning the declustering poten-tial, the collision cell voltage was scanned from 25 V to 100 V in 5 Vincrements collecting a product ion mass spectrum for 5 s at eachinterval. Using the optimal in-source CID potential found above, theMS/MS spectra were taken at �100 V, �125 V, and �150 V CIDpotential for AN, PC, and PPC, respectively. Fig. 4a displays theresponse for each anion as a function of collision cell voltage. Theanions were fragmented, identifying transitions from nitrate intonitrite (m/z 46 NO2

�); chlorate into chlorite (m/z 67 ClO2�) and

hypochlorite (m/z 51 ClO�); and perchlorate into chlorate, chlorite,and hypochlorite (Fig. 4 and S2).

Fig. 4. Signal intensity, i.e., integrated peak area, as a function of collision cell voltagefor inorganic molecular anions in negative mode. (a) Response of anions originatingfrom PPC: -C- m/z 99 ClO4

� , PC: -◊- m/z 83 ClO3� , and AN: NO3

� . Signal intensityof relevant anions in (b)e(d) also include fragments for nitrite: m/z 46 NO2

� ,chlorite: m/z 67 ClO2

� and hypochlorite: m/z 51 ClO�. Data points and un-certainty expressed as the average peak area and standard deviations for 5 successivereplicate samples, respectively.

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e96

Similar to the results demonstrated for the molecular anions inFig. 3, Fig. 5a displays the MS response of a few main clustersobserved as a function of in-source CID. The clusters, KðClO3Þ2�,KðClO4Þ2�, and KðNO3Þ2�, confirmed the results presented in Fig. 3,revealing higher initial signals for the weaker salts and higher in-source CID potentials for maximum signal intensity. As observedin previous investigations [34], elemental ions demonstrated theopposite response, asymptotically increasing in signal withincreasing in-source CID potential (Fig. 5bed). Fig. 5b displays therelative response of the potassium cation from PC, PPC, and PN. Theelemental cations cannot be further fragmented and therefore didnot demonstrate any reduction in signal above a local maximum forthe range of CID potentials investigated here. However, anasymptotic approach to a maximum signal was observed. Fig. 5cdisplays the response for the metal fuel zinc, the chloride anionfrom PPC, and calcium cation from CAN, all exhibiting the expectedtrends for elemental ions. Additional metal fuels such as aluminumand magnesium fell outside the m/z range optimized for thisinvestigation and were not readily detected. In-source CID providesthe critical fragmentation of clusters to enhance the detection ofthese inorganic explosives. The trace detection of explosive devicesignatures may also require detection of non-explosive compo-nents. For example, radiological dispersion devices incorporateobtainable industrial radionuclides used in medical radiotherapy,sterilization, or power generation, to be dispersed by a primaryexplosive. In-source CID was employed in Fig. 5d to demonstratethe detection of stable isotope surrogates for cobalt (59Co repre-senting 60Co) and strontium (88Sr representing 90Sr) in place of theradionuclide isotopes.

3.3. Inorganic oxidizer detection and chemical imaging

The framework developed above for optimization of in-sourceCID of both molecular and elemental inorganic components was

Fig. 5. Normalized signal abundance as a function of in-source CID (declustering)potential magnitude for (a) inorganic clusters in negative mode and (b)e(d) elementalcations in positive mode from 1 mg explosive oxidizers, metal fuels, and stable isotoperadionuclide surrogates. (a) Negative mode inorganic clusters from PC: -◊- m/z 205KðClO4Þ2� , PPC: -C- m/z 237 KðClO4Þ2�, PN: m/z 163 KðNO3Þ2� . Positive mode (b)potassium cations (m/z 39 Kþ) from PPC, PC, and PN; (c) metal fuel zinc ( m/z 65Znþ) and elemental components of PPC ( m/z 35 Cl�) and CAN ( m/z 40 Caþ);and (d) stable isotope radionuclide surrogates cobalt ( m/z 59 Coþ) and strontium( m/z 88 Srþ). Data points and uncertainty expressed as the average peak area andstandard deviations for 5 to 8 successive replicate samples, respectively.

leveraged for the detection of fuel-oxidizer mixtures using an LDI-MS instrument with in-source CID. The laser desorption ionizationsource was fitted with a CCD camera and on-screen display formanual interrogation of regions of interest (ROIs) and individualparticles. Fig. 6 displays an image and mass spectrum associatedwith the direct examination of a particle and residue area spiked ona latent fingerprint. For demonstration purposes, potassiumperchlorate crystals were crushed and dry transferred onto adeposited latent fingerprint on a nickel chromium substrate. TheROIs identified in Fig. 6a were analyzed using both automatedmeandering (serpentine) scans at 10 mm min�1 and direct manualinterrogation of the particle (Fig. 6b). Distinct peaks correspondingto chlorate (m/z 83), perchlorate (m/z 99), and the previouslyidentified potassium perchlorate cluster (m/z 237 KðClO4Þ2�) wereobserved. Given preexisting knowledge of the inorganic oxidizer,the optimal in-source CID potential (�150 V) was utilized for thedetection of potassium perchlorate. However, for unknown sam-ples, rapid scans through CID potentials within the transition fromcrystal fragmentation to anion fragmentation, demonstrated inFig. 3, would cover the optimal parameter value for a range of ox-idizers. This mode of specific user defined interrogation may beapplied to the detection of other explosive particles (organic:nitramines, nitroaromatics, nitrate esters, peroxides, etc.),

Fig. 6. (a) Optical images of potassium perchlorate particles and residues on a latentfingerprint. (b) Representative LDI-MS mass spectra for the 2 min manual interrogationof the identified particle and adjacent residue from user defined ROIs (i) and (ii). (c)LDI-MS mass spectrum for the 45 s scan of a 1 mL sample from the full print liquidextraction. The base peak absolute intensity is noted in the upper right hand corner ofeach panel.

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e9 7

radionuclide particles, and gunshot residues/particles from finger-prints and forensic tape pulls.

In addition, for bulk detection and avenues toward quantifica-tion of specific analytes dry transferred onto fingerprints, the latentfingerprint containing potassium perchlorate particles/residueswas extracted from the substrate surface in 200 mL of water. Theprint extract was then deposited in 1 mL aliquots and allowed to drybefore direct analysis by LDI-MS. Fig. 6c displays the mass spectrumfor a 45 s scan of the extracted aliquot deposit, clearly displayingthe chlorate and perchlorate peaks previously identified. Additionalpeaks likely associated with the fingerprint material and othercontaminants were also observed. The extraction process alsoprovided sufficient dissolution of the potassium perchlorate crys-tals to significantly reduce the intensity of the cluster at m/z 237(KðClO4Þ2�) and introduce a potassium-chlorate-perchlorate clus-ter at m/z 221 ðKðClO3ÞðClO4Þ�Þ, previously not observed (Fig. 6c).

Next, we investigated the bulk quantification capabilities for theanalysis of inorganic explosive containing residues using LDI-MS andin-source CID. For this analysis trace residues were collected on ad-hesive sheets during the synthesis of homemade potassium chlorateand icing sugar materials for a simulated fuel-oxidizer mixture de-vice. The three samples, labeled A, B, and C, were extracted andcalibration curveswereprepared asdescribed above. For LDI-MS,1mLaliquots were deposited onto sample plates and the solvent allowedto evaporate under ambient conditions. The calibration curve con-sisted of samples in the range of 0 ng mL�1 to 50 ng mL�1. Each samplewas interrogated by the laser for 15 s and mass spectral data wascollected and summed. In-source CID potentials of þ350 Vand�125Vwere used for the quantification and calibration curves ofthe elemental potassium cation and molecular chlorate anion. Thecalibration curves for potassiumand chloratewere linear in the rangeof 0 ng mL�1 to 25 ng mL�1 and 0 ng mL�1 to 50 ng mL�1, respectively(Figure S3). The calibration curves were utilized for quantification ofextracted samples, as well as to approximate the limits of detection(LOD) for the LDI-MS system. The LOD was defined as thedepositedmass generating an analyte-specific peak that resulted in asignal-to-noise ratio of 3:1. From potassium chlorate, the LODsmeasured for the potassium cation (Figure S3a) and chlorate anion(Figures S3b) were both at the nanogram level. Similarly, nitratebased oxidizers, AN and CAN, also demonstrated nanogram levelLODs, for detection of the nitrate anion.

The three unknown samples (A-C) were then each measured intriplicate and the results listed in Table 1. The standard uncertaintyfrom the linear least squares calibration was evaluated as describedin the Methods section. The potassium signal for sample A wasmeasured above the linear region of the calibration curve andexcluded. For confirmation, the potassium cation from the 2 mLextracted solutionswas also analyzed using ICP-MSand IC. As shownin Table 1, the LDI-MS quantification between the potassium cationand chlorate anion were comparable to each other, as well asconsistent with the potassium cation measurements across thealternative techniques. However, due to the nature of the

Table 1Mass of potassium chlorate detected from samples AeC. Calculated masses based onmeasurements of the potassium cation and chlorate anion using LDI-MS, ICP-MS,and IC. Standard uncertainties in the quantification of unknowns due to variability inthe system response from linear least squares calibration were evaluated asdescribed in the Methods section (k ¼ 1).

Technique Sample A Sample B Sample C

Kþ (mg) ClO3� (mg) Kþ (mg) ClO3

� (mg) Kþ (mg) ClO3� (mg)

LDI-MS e 34.5 ± 6.9 9.0 ± 2.9 8.1 ± 6.8 2.4 ± 2.6 1.7 ± 6.9ICP-MS 36.5 ± 0.5 8.5 ± 0.5 2.1 ± 0.5IC 36.3 ± 0.4 9.7 ± 0.4 2.0 ± 0.4

measurement, the uncertainty in the LDI-MS system response wasgreater than the other techniques. The ICP-MS and ICmeasurementsdirectly analyzed the relatively homogenous extracted solutions.Replicatemeasurements by thesemethodswere in essence taken onthe same sample. On the contrary, LDI-MS relied on the deposition ofsolution phase aliquots and analysis of the solid phase remainingafter solution evaporation. This significant difference in the mannerof sample preparation can result in non-uniform surface concen-trations and thicknesses, which in turn affect the actual amount ofmaterial interrogated. These aspects may have led to the large un-certainties in the calibration curve and bulk quantification mea-surements by LDI-MS. This method was clearly inadequate forquantification for masses near the limit of detection (e.g., sample C).In spite of this potential shortcoming, the direct analysis of particleshas the potential to provide additional information, specifically,molecular speciation of individual particles, beyond the bulkextraction and analysis of analyte distributions. Similarly, massspectrometry imaging (MSI) of the spatial distribution of particlesand chemical distributions of analytes on a surface or within acomplexmatrixmight provide complementary forensic information.

Similar to the method above, the LDI-MS systemwas utilized forthe chemical imaging of organic and molecular inorganic com-pounds from a latent fingerprint. Again, particles of potassiumperchlorate or calcium ammonium nitrate were dry transferredonto latent fingerprints on a nickel/chromium substrate andimaged without sample preparation, pretreatment, or the use ofmatrix. Prints were imaged in 100 mm � 100 mm pixels across in-dividual 5 mm � 5 mm regions of interest at constant12.5 mm min�1 velocity. Fig. 7 demonstrates the chemical imagesof an endogenous lotion component, behentrimoniumm/z 368 Mþ,perchloratem/z 99 ClO4

�, and the potassium perchlorate clusterm/z 237 KðClO4Þ2�. The lotion component, behentrimonium, wasimaged in positive mode and detectable at a relatively wide rangeof in-source CID potentials (Fig. 7a). The potassium component ofthe PPC particles was unsuccessfully resolved in positive mode athigh in-source CID due to interference with potassium from thelatent fingerprint as well as the underlying substrate surface. Theelemental potassium cation distribution was dominated by themachining marks left on the sample substrate surface (image notshown). The perchlorate (Fig. 7b) and potassium perchlorate cluster(Fig. 7c) distributions were clearly resolved in negative mode MS atlow in-source CID (�150 V) and demonstrated colocalization(Fig. 7d), further substantiating the identification of the m/z 237cluster.

Fig. 7e and f displays a surface map of the lotion component andthe corresponding intensity line scan across a series of print ridges,respectively. The technique demonstrated spatial resolution on theorder of 100's mm, sufficient for an overall characterization of theanalyte distributions, yet insufficient for differentiating muchsmaller micron size particles or tertiary ridge detail. Alternativehigh resolution laser desorption [47] and time-of-flight secondaryion mass spectrometry (ToF-SIMS) [48] systems can provide thesub-micrometer spatial resolution necessary for particles sizingand pore imaging.

The chemical distribution imaging of another oxidizer, calciumammonium nitrate, successfully identified the calcium cation (m/z40 Caþ), the nitrate anion (m/z 62 NO3

�), and a calcium nitratecluster atm/z 226 CaðNO3Þ3� (Figure S4). The ROI was interrogatedtwice, first at high in-source CID (þ300 V) in positive mode for thechemical distribution of the elemental calcium cation, and then atlow in-source CID (�100 V) in negative mode for the distributionsof nitrate and the calcium nitrate cluster. The nitrate and calciumnitrate cluster collocated demonstrating the particle/residue dis-tribution (Figure S4b). However, the calcium cation and nitratechemical images demonstrated the same general distributions, but

Fig. 7. LDI-MSI of endogenous organic and polyatomic inorganic ions at low in-source CID: (a) endogenous lotion component e behentrimonium m/z 368, (b) perchlorate m/z 99ClO4

� , (c) potassium perchlorate cluster m/z 237 KðClO4Þ2�, and (d) the colocalization map of the lotion (blue) and potassium perchlorate particles/residues (represented byperchlorate ion distribution, red). (e) and (f) display a surface map of the lotion component and the corresponding intensity line scan across a series of print ridges, respectively. (Forinterpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e98

were not as precisely collocated (Figure S4a). In some instances thecalcium cation image displayed a much smaller size distribution ofindividual particles as compared to the nitrate distributions. Theseartifacts may be due to a number of artifacts, including non-uniform distributions of calcium nitrate and ammonium nitrate,slight variability in the two consecutive scans, variability in theinitial particle shape and the post-scan particle shape (the first scaninterrogated and ablated the apex of a particle, while the secondscan experienced a wider mid-section of the particle), variability inthe particle/residue distribution at different sample depths, or ni-trate dissolution into the eccrine component of the fingerprint. TheLDI-MS system enabled direct control over the laser power andpulse frequency, providing a potential means for investigating theapplicability of depth profiling with this technique.

4. Conclusions

The analysis and detection of explosive device signatures from atransient collection of homemade and improvised explosives anddevices remains a clear area of interest. Inorganic compoundsincluding fuel-oxidizer and self-initiating mixtures, azides, andradionuclides, play a critical role in an assembly of HMEs and IEDcomponents. Here, we presented the coupling of in-source colli-sion induced dissociation (CID) with atmospheric pressure laserdesorption/ionization mass spectrometry (LDI-MS) for the detec-tion, bulk quantification, and chemical imaging of fuel-oxidizermixture signatures. In-source CID enhanced detection ofelemental inorganic ions by 10-fold to 1000-fold through thefragmentation of larger adducts and clusters, as well as through thereduction in organic chemical noise. Similarly, in-source CID wasutilized to identify the optimal CID potential, providing 10-fold to100-fold increase in signal, for the detection of oxidizer polyatomicanion components, specifically, nitrates, chlorates, and perchlo-rates. This optimization was identified at the transition between

energetically breaking the ionic salt bonds and fragmenting themolecular anion. The optimized detection of elemental and mo-lecular inorganic ions enabled the detection of inorganic explosivecomponents at nanogram levels. The developed framework wasleveraged for their detection from a complex matrix, the bulkquantification of realistic samples collected during explosive syn-thesis, and chemical imaging of oxidizer particles from latent fin-gerprints. The use of LDI-MS for quantification demonstratedinherently large uncertainty in the system response. Future workwill explore the use of inkjet printing for the controlled depositionof calibration standards and the incorporation of an internalstandard. Precision drop-on-demand inkjet printing has the po-tential to provide micron-scale uniform sample deposits for linearleast squares calibration and quantification. In addition, the LDI-MSsystem enabled direct control over the laser power and pulse fre-quency, providing a potential means for investigating high densityinorganic particles for environmental monitoring, radionuclideisotopic ratio analysis, and biological tissue imaging. Future workwill also aim to explore these opportunities.

Acknowledgments

The U.S. Department of Homeland Security Science and Tech-nology Directorate sponsored a portion of the production of thismaterial under Interagency Agreement IAA HSHQDC-12-X-00024with the National Institute of Standards and Technology.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.aca.2015.06.008.

T.P. Forbes, E. Sisco / Analytica Chimica Acta 892 (2015) 1e9 9

References

[1] G.-h.L. Lang, K.M. Boyle, The analysis of black powder substitutes containingascorbic acid by ion chromatography/mass spectrometry, J. Forensic Sci. 54(2009) 1315e1322.

[2] G.W. Dicinoski, R.A. Shellie, P.R. Haddad, Forensic identification of inorganicexplosives by ion chromatography, Anal. Lett. 39 (2006) 639e657.

[3] J.P. Hutchinson, C.J. Evenhuis, C. Johns, et al., Identification of inorganicimprovised explosive devices by analysis of postblast residues using portablecapillary electrophoresis instrumentation and indirect photometric detectionwith a light-emitting diode, Anal. Chem. 79 (2007) 7005e7013.

[4] J.J. Corr, J.F. Anacleto, Analysis of inorganic species by capillary electro-phoresis�mass spectrometry and ion exchange chromatography�massspectrometry using an ion spray source, Anal. Chem. 68 (1996) 2155e2163.

[5] M. Bottegal, L. Lang, M. Miller, et al., Analysis of ascorbic acid based blackpowder substitutes by high-performance liquid chromatography/electrosprayionization quadrupole time-of-flight mass spectrometry, Rapid Commun.Mass Spectrom. 24 (2010) 1377e1386.

[6] R.G. Ewing, D.A. Atkinson, G.A. Eiceman, et al., A critical review of ion mobilityspectrometry for the detection of explosives and explosive related com-pounds, Talanta 54 (2001) 515e529.

[7] B.M. Kolakowski, Z. Mester, Review of applications of high-field asymmetricwaveform ion mobility spectrometry (FAIMS) and differential mobility spec-trometry (DMS), Analyst 132 (2007) 842e864.

[8] M. Najarro, M.E. Davila Morris, M.E. Staymates, et al., Optimized thermaldesorption for improved sensitivity in trace explosives detection by ionmobility spectrometry, Analyst 137 (2012) 2614e2622.

[9] J.R. Verkouteren, J. Lawrence, G.A. Klouda, et al., Performance metrics based onsignal intensity for ion mobility spectrometry - based explosive trace de-tectors using inkjet printed materials, Analyst 139 (2014) 5488e5498.

[10] S. Ehlert, A. Walte, R. Zimmermann, Ambient pressure laser desorption andlaser-induced acoustic desorption ion mobility spectrometry detection ofexplosives, Anal. Chem. 85 (2013) 11047e11053.

[11] X. Liang, Q. Zhou, W. Wang, et al., Sensitive detection of black powder by astand-alone ion mobility spectrometer with an embedded titration region,Anal. Chem. 85 (2013) 4849e4852.

[12] T. Khayamian, M. Tabrizchi, M.T. Jafari, Analysis of 2,4,6-trinitrotoluene,pentaerythritol tetranitrate and cyclo-1,3,5-trimethylene-2,4,6-trinitramineusing negative corona discharge ion mobility spectrometry, Talanta 59 (2003)327e333.

[13] M. Tabrizchi, V. Ilbeigi, Detection of explosives by positive corona dischargeion mobility spectrometry, J. Hazard. Mater. 176 (2010) 692e696.

[14] D.R. Justes, N. Talaty, I. Cotte-Rodriguez, et al., Detection of explosives on skinusing ambient ionization mass spectrometry, Chem. Commun. (2007)2142e2144.

[15] C. Szakal, T.M. Brewer, Analysis and mechanisms of Cyclo-trimethylenetrinitramine ion formation in desorption electrospray ionization,Anal. Chem. 81 (2009) 5257e5266.

[16] Z. Takats, I. Cotte-Rodriguez, N. Talaty, et al., Direct, trace level detection ofexplosives on ambient surfaces by desorption electrospray ionization massspectrometry, Chem. Commun. (2005) 1950e1952.

[17] I. Cotte-Rodríguez, Z. Tak�ats, N. Talaty, et al., Desorption electrospray ioniza-tion of explosives on surfaces: sensitivity and selectivity enhancement byreactive desorption electrospray ionization, Anal. Chem. 77 (2005)6755e6764.

[18] J.M. Nilles, T.R. Connell, S.T. Stokes, et al., Explosives detection using directanalysis in real time (DART) mass spectrometry, Propell., Explos. Pyrot 35(2010) 446e451.

[19] E. Sisco, J. Dake, C. Bridge, Screening for trace explosives by AccuTOF™-DART®: An in-depth validation study, Forensic Sci. Int. 232 (2013) 160e168.

[20] T.P. Forbes, T.M. Brewer, G. Gillen, Desorption electro-flow focusing ionizationof explosives and narcotics for ambient pressure mass spectrometry, Analyst138 (2013) 5665e5673.

[21] Y. Zhang, X. Ma, S. Zhang, et al., Direct detection of explosives on solid surfacesby low temperature plasma desorption mass spectrometry, Analyst 134(2009) 176e181.

[22] J.F. Garcia-Reyes, J.D. Harper, G.A. Salazar, et al., Detection of explosives andrelated compounds by low-temperature plasma ambient ionization massspectrometry, Anal. Chem. 83 (2010) 1084e1092.

[23] Z. Tak�ats, J.M. Wiseman, B. Gologan, et al., Mass spectrometry sampling underambient conditions with desorption electrospray ionization, Science 306(2004) 471e473.

[24] R.B. Cody, J.A. Laram�ee, H.D. Durst, Versatile new ion source for the analysis ofmaterials in open air under ambient conditions, Anal. Chem. 77 (2005)2297e2302.

[25] G.A. Harris, A.S. Galhena, F.M. Ferna

ndez, Ambient sampling/ionization massspectrometry: applications and current trends, Anal. Chem. 83 (2011)4508e4538.

[26] D.R. Ifa, C. Wu, Z. Ouyang, et al., Desorption electrospray ionization and otherambient ionization methods: current progress and preview, Analyst 135(2010) 669e681.

[27] J. Yinon (Ed.), Counterterrorist Detection Techniques of Explosives, Elsevier,B.V., Amsterdam, 2007.

[28] M. Marshall, J.C. Oxley (Eds.), Aspects of Explosive Detection, Elsevier, B. V.,Amsterdam, 2009.

[29] O. Dalby, D. Butler, J.W. Birkett, Analysis of gunshot residue and associatedmaterialsdA review, J. Forensic Sci. 55 (2010) 924e943.

[30] M. M€akinen, M. Nousiainen, M. Sillanp€a€a, Ion spectrometric detection tech-nologies for ultra-traces of explosives: a review, Mass Spectrom. Rev. 30(2011) 940e973.

[31] I. Cotte-Rodríguez, H. Hern�andez-Soto, H. Chen, et al., Situ trace detection ofperoxide explosives by desorption electrospray ionization and desorptionatmospheric pressure chemical ionization, Anal. Chem. 80 (2008) 1512e1519.

[32] I. Cotte-Rodriguez, H. Chen, R.G. Cooks, Rapid trace detection of triacetonetriperoxide (TATP) by complexation reactions during desorption electrosprayionization, Chem. Commun. (2006) 953e955.

[33] S. Parajuli, W. Miao, Sensitive determination of Triacetone Triperoxide ex-plosives using electrogenerated chemiluminescence, Anal. Chem. 85 (2013)8008e8015.

[34] T.P. Forbes, E. Sisco, Mass spectrometry detection and imaging of inorganicand organic explosive device signatures using desorption electro-flowfocusing ionization, Anal. Chem. 86 (2014) 7788e7797.

[35] K.M. Evans-Nguyen, A. Quinto, T. Hargraves, et al., Transmission modedesorption electrospray ionization (tm-desi) for simultaneous analysis ofpotential inorganic and organic components of radiological dispersion devices(RDDs), Anal. Chem. 85 (2013) 11826e11834.

[36] P.M. Flanigan, J.J. Brady, E.J. Judge, et al., Determination of inorganicimprovised explosive device signatures using laser electrospray massspectrometry detection with offline classification, Anal. Chem. 83 (2011)7115e7122.

[37] J. Pavlov, A.B. Attygalle, Direct detection of inorganic nitrate salts by ambientpressure helium-plasma ionization mass spectrometry, Anal. Chem. 85 (2012)278e282.

[38] T.P. Forbes, Rapid detection and isotopic measurement of discrete inorganicsamples using acoustically actuated droplet ejection and extractive electro-spray ionization mass spectrometry, Rapid Commun. Mass Spectrom. 29(2015) 19e28.

[39] T.P. Forbes, E. Sisco, Chemical imaging of artificial fingerprints by desorptionelectro-flow focusing ionization mass spectrometry, Analyst 139 (2014)2982e2985.

[40] G. Robichaud, K. Garrard, J. Barry, et al., MSiReader: an open-source interfaceto view and analyze high resolving power MS imaging files on matlab plat-form, J. Am. Soc. Mass Spectrom. 24 (2013) 718e721.

[41] S.L.R. Ellison, M. Rosslein, A. Williams, Eurachem/CITAC Guide: QuantifyingUncertainty in Analytical Measurement, Second ed., 2000.

[42] B.N. Taylor, C.E. Kuyatt, NIST Technical Note 1297, Guidelines for evaluatingand expressing the uncertainty of NIST measurement results, 1994. http://www.nist.gov/pml/pubs/tn1297/.

[43] MATLAB, The MathWorks Inc, Natick, MA, 2013.[44] A.T. Blades, P. Jayaweera, M.G. Ikonomou, et al., Studies of alkaline earth and

transition metal Mþþ gas phase ion chemistry, J. Chem. Phys. 92 (1990)5900e5906.

[45] P. Jayaweera, A.T. Blades, M.G. Ikonomou, et al., Production and study in thegas phase of multiply charged solvated or coordinated metal ions, J. Am.Chem. Soc. 112 (1990) 2452e2454.

[46] J.T. Watson, O.D. Sparkman, Introduction to Mass Spectrometry, John Wiley &Sons, Ltd, West Sussex, 2007.

[47] C. Elsner, B. Abel, Ultrafast high-resolution mass spectrometric finger poreimaging in latent finger prints, Sci. Rep. 4 (2014) 1e6.

[48] M.J. Bailey, M. Ismail, S. Bleay, et al., Enhanced imaging of developed fin-gerprints using mass spectrometry imaging, Analyst 138 (2013)6246e6250.