Libs como detección de explosivos

Spectrochimica Acta Part B 64 (2009) 1028–1039

Contents lists available at ScienceDirect

Spectrochimica Acta Part B
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s a b

Analysis of explosive and other organic residues by laser induced
breakdown spectroscopy☆
V. Lazic a,⁎, A. Palucci a, S. Jovicevic b, C. Poggi a, E. Buono a
a
ENEA, FIS-LAS, Via. E. Fermi 45, 00044 Frascati (RM), Italy
b
Institute of Physics, 11080 Belgrade, Pregrevica 118, Serbia

a r t i c l e i n f o a b s t r a c t

Article history: With the aim of realizing a compact instrument for detection of energetic materials at trace levels, laser
Received 30 November 2008 induced breakdown spectroscopy was applied on residues from nine explosives in air surroundings.
Accepted 31 July 2009 Different potentially interfering organic materials were also analyzed. The residues were not uniformly
Available online 12 August 2009
distributed on an aluminum support and single-shot discrimination was attempted. For a single residue type,
large shot-to-shot fluctuations of the line intensity ratios characteristic for organic samples were observed,
Keywords:
LIBS
which made material classification difficult. It was found that both atomic and molecular emission
Explosives intensities, as well as their ratios, are strongly affected by an amount of the ablated support material, which
Organic mainly determines the plasma temperature. With respect to the spectra from the clean support, emission
Residues intensities of atomic oxygen and nitrogen are always reduced in the presence of an organic material, even if
Laser its molecules contain these elements. This was attributed to chemical reactions in a plasma containing
carbon or its fragments. Hydrogen atomic emission depends strongly on the local humidity above the
sampled point and its line intensity shows shot to shot variations up to 50%, also on a homogeneous sample.
It is argued that shock waves generated by previous spatially and/or temporally close laser pulses blow away
a relatively heavy water aerosol, which later diffuses slowly back towards the sampled point. C2 and CN
exhibit a peak emission behavior with atomic Al emission, and their variable ratio indicates an existence of
different formation or removal mechanisms from the plasma, depending on the plasma parameters and on
the composition of the organic residue. On the basis of these observations, an attempt is made to establish a
suitable procedure for data analysis and to determine the optimal experimental conditions, which would
allow for discrimination of explosives from other, potentially interfering, residues.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction [3]. An intense laser pulse is also responsible for atomization and
ionization of the material. The plasma growth and decay leads to
Explosives are notoriously difficult to detect and methods for their different processes such as: expansion, shock waves formation,
detection can address vapors or particles, dissolved or suspended continuum (bremsstrahlung) emission and light absorption by free
solids in solutions, or direct probing of solid materials [1,2]. electrons (inverse bremsstrahlung), collisions in the gas with excitation
One of the techniques recently proposed for identification of solid and relaxation of atoms/ions, chemical recombination and, as important
explosives and their residues is LIBS, which allows for multi- for LIBS, de-excitation of the species (atoms, ions and molecules)
elemental, in-situ measurements in different surroundings [2–5]. through optical emission. In LIBS, this radiation is usually detected in the
The measurements are without contact, making stand-off detection spectral range covering near UV, visible and near IR. The capability to
possible up to 130 m [6–9]. perform non-contact, long-distance, real time multi-element detection
LIBS is based on plasma generation by an intense laser pulse, which makes LIBS very interesting for different applications [3–5], including
duration is in nanosecond range or shorter. In the case of solid samples, detection of hazardous materials [8–13] and classification of
plasma is produced through laser-induced ablation of the surface layer biological agents [13–15]. However, LIBS sensitivity and accuracy
are generally lower than for some other techniques, such as ICP-OES,
mass spectrometry and gas chromatography. In addition, the laser
☆ "This paper was presented at the 5th International Conference on Laser-Induced characteristics and the acquisition parameters, the surrounding
Breakdown Spectroscopy (LIBS 2008), held in Berlin, Adlershof, Germany, 22–26 atmosphere (gas composition and pressure), interaction geometry,
September 2008, and is published in the Special Issue of Spectrochimica Acta Part B,
dedicated to that conference".
surface conditions and collecting optics geometry strongly affect the
⁎ Corresponding author. Tel.: +39 06 94005885; fax: +39 06 94005400. analytical capabilities of the method [16–21]. Under fixed experi-
E-mail address: lazic@frascati.enea.it (V. Lazic). mental conditions, the signal intensities and analytical accuracy are

0584-8547/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.sab.2009.07.035

V. Lazic et al. / Spectrochimica Acta Part B 64 (2009) 1028–1039 1029

further compromised by the so-called matrix effect [21–29], The including into model (PCA or PLS-DA) also impurities (Na, Ca and K)
matrix effect can be partially overcome by applying calibration with beside C, C2, CN, H, N, and O line intensities. However, we intentionally
matrix-matched reference samples. However, the latter is difficult to did not consider impurities here as their concentrations change with
implement in multi-elemental in-situ analyses of complex samples the origin of the organic material.
with unknown or variable matrices. In such applications, it is We analyzed nine types of residues from pure explosives and
preferable to include into calibration different types of the sample residues of six other organic materials as possible interferents. In all
matrices [28] and to take into account also the plasma parameters the cases, residues were placed on a clean aluminum support and LIBS
[28,29]. spectra were saved after each laser shot. We identified and explained
Explosives are organic compounds, containing carbon, hydrogen and some sources of changes of line intensity ratios, then proposed and
oxygen, while nitrogen is present in almost all the high explosives. discussed some procedures for recognition of explosives, based on the
Commonly, explosives are rich in N and O, and poor in H and C with line intensities and PCA analysis.
respect to other organic substances. LIBS spectra from energetic
materials normally contain atomic lines from these four elements [8–
12], and molecular bands of CN and C2. On the organic compounds, ionic 2. Experimental
lines from oxygen and nitrogen are also sometimes observed [30].
Intensity of C2 emission can be correlated with number of double CfC 2.1. Laboratory set-up
bonds in the sample [31,32]. Differently, CN spectra produced from
ablation of organic compounds in air surrounding are mainly result of The system described here is intended only for laboratory
recombination through reaction C2 + N2 = 2CN [31–33]. LIBS detection measurements and does not represent a prototype system for
of native CN bonds in organic materials was obtained only at low detection of explosive residues, which is presently in the testing
fluences of a UV exciting laser [33]. phase. The laser source is an Nd:YAG (Quantel — Ultra 200) operated
Rapid LIBS detection of energetic materials with compact instru- at 1064 nm, characterized by a super Gaussian beam profile. The laser
ments or at distance is normally performed in air surroundings, so pulse width is about 6 ns, the maximum energy is 300 mJ and the
interference from air components on the LIBS spectra must be taken maximum repetition rate is 10 Hz. The laser was triggered manually
into account. The influence of air components can be reduced, but not after repositioning of the sample. Timing between two successive
eliminated, by applying a double-pulse laser excitation [9]. In such pulses was not fixed and always longer than 10 s. The laser beam was
case, expanding shock-waves following breakdown produced by the deflected towards a horizontally placed sample and focused by a
first laser pulse, reduces locally the gas (air) pressure, so the quartz lens with 150 mm focal length. In the optimal position for the
secondary, analytical plasma expands into more rarefied atmosphere LIBS signal intensities, the measured laser spot diameter was of about
[34]. 0.94 mm at an energy of 250 mJ. This energy was used for all the
Recognition of bulk organic compounds by LIBS can be performed measurements on organic residues.
by comparison between the sample spectra and previously estab- Plasma emission was collected at an angle by a wide-angle quartz
lished library, for example through linear correlation [8,11,12,30]. objective, and directed to a spectrometer by 0.1 mm diameter quartz
Discrimination of organic materials was obtained by comparing line fiber. The fiber end was mounted on an entrance slit of a Mechelle
intensity ratios from H, C, N and O, eventually including some other spectrograph (Andor, Mod. 5000), which allows for simultaneous
elements present as impurities [8]. Molecular emission was some- spectral detection in the range 200–950 nm with resolving power
times also considered in order to improve the discrimination [8]. A (λ/Δλ) = 5000. The spectra were recorded by a gated ICCD (iStar
procedure for recognition of energetic materials is then based on DH734), whose gate aperture was synchronized with the laser burst
properly constructed algorithms [7,8], Principal Component Analysis through an optical trigger. After preliminary optimization of the
(PCA) [8–10,14,15], Soft Independent Modeling of Class Analogy gating parameters, all the LIBS measurements on residues or on
(SIMCA) [8,13], Neural Networks [10,35] or Partial Least Square clean support were performed with gate delay of 800 ns and gate
Discriminant Analysis (PLS-DA) [8]. width of 30 µs. After each laser shot, a spectrum was saved for
Analysis of residue materials placed on a support, even if the latter further analysis. This approach also allows monitoring of the shot-
has a well known composition, is difficult due to the variable amount to-shot variations of the spectral intensities.
of sample from one point to another. In such case, the previously The sample holder was placed on an X–Y micrometric table. After
discussed matrix effect must be taken into account and different each laser shot, the sample was displaced by 1.5 mm, which was
spectral libraries/models should be constructed for different support sufficient to avoid sampling in areas affected by material redeposited
materials. When probing residues by LIBS, even if placed on a fixed from previous shots.
type of support, variations in the line intensities and their ratios are
expected to be large, due to other causes such as variable amount of
ablated sample, variable quantity of the ablated support material, 2.2. Samples
possible differences in the plasma shape, temperature and electron
density, and finally possible differences in the chemical reactions As a support material for residues, we used aluminum (Antic-
occurring in the plasma. Chemical reactions depend on the plasma orodal) discs, whose surface was intentionally machined roughly
composition and its parameters, where higher reached temperatures (Fig. 1) in order to increase the sticking of the residues and to avoid
also lead to a more complete sample fragmentation and atomization. high surface reflectivity. Before sampling, the discs were cleaned in
Chemistry of LIBS plasma from organic samples is very complex, and ultrasound bath, first with pure acetone and then with distilled water.
its simulation for pure RDX explosive includes 137 species and 577 All the explosives considered, except DNT (see Table 1), are
reactions [36]. All the mentioned reasons might contribute to a failure commercial solutions with concentration of 1.0 mg/ml or 0.1 mg/ml
of residue explosive recognition from the LIBS spectra. (TATP) in methyl or ethyl alcohol, or acetonitrile. DNT was purchased
Given this background, we studied variations of the characteristic in form of crystal grains and dissolved in pure acetone before placing
atomic (C, H, N and O) and molecular (C2 and CN) lines in the LIBS it onto support. Small droplets of solution containing explosives
spectra obtained on different organic residues, not uniformly spread over aluminum surface were allowed to evaporate thermally,
distributed over the substrate. It has been reported [8] that 100% leaving unevenly distributed residues (Fig. 1). Other analyzed, not
correct residue classification (considering 3 explosives and 3 inter- energetic materials were directly distributed over the support in thin
ferents) and with an improved class separation, can be obtained when layers of uncontrolled thickness.

1030 V. Lazic et al. / Spectrochimica Acta Part B 64 (2009) 1028–1039

Table 2
Wavelengths of the emission lines used for peak analyses.

Species Peak (nm)

CI 247.8
CN 388.3
C2 516.3
HI 656.2
OI 777.4 (triplet)
NI 746.8
Al I 309.3

Molecular line intensities reach the maximum Signal-to-Noise ratio
(SNR) for gate width of about 30 µs, while for the atomic lines this
occurs for gate widths below 10 µs. In particular, C emission reaches
the maximum SNR already for gate width of 1 µs. Increasing the signal
integration time up to 3 ms does not deteriorate SNR of the lines,
which is important in a view of employing much cheaper and compact
spectrometers without gating option.
Fig. 1. Photo of the HMX residue (white) on the aluminum support; the width of the
In the following, the gate width was fixed to 30 µs in order to
area is 1.5 mm. capture the whole molecular emission. The optimal acquisition delay
was then searched for the spectra produced on clean support and for
the same in presence of diesel residues. In both cases, the optimal
3. Results delay was between 800 and 1000 ns. In all the successive measure-
ments, the gate delay was fixed at 800 ns.
LIBS spectra were initially acquired on the coal sample pressed into On all the residues considered, in addition to the emission from H,
a pellet, by accumulating over 20 laser shots with gate delay and N and O, it was possible to obtain single-shot spectra containing
width of 800 ns and 50 µs, respectively. The aim of these sufficiently intense C and CN lines. In many cases, C2 emission was also
measurements was to optimize the spectral intervals for calculation detectable. When analyzing the lines, we initially considered their
of the line peaks/integrals and corresponding background levels. The peak values (Table 2) after subtraction of an average, nearby
coal pellet was chosen due to its good homogeneity, presence of all background. Before processing, the spectra were corrected for the
the four elements of interest (C, H, N and O) and because it contained wavelength dependent instrumental response. All the files inside a
also other elements (Al, Fe, K, Mg, Ca, Na), which are often present as single directory, corresponding to one measurement run per sample,
impurities in other organic materials. Some emission lines from these were processed automatically by custom routines written under
impurities are close to the to atomic/molecular features characteristic LabView. Intensities of the lines with calculated SNR below 3.0 were
for the LIBS spectra of explosives and might lead to an erroneous automatically settled to zero.
evaluation of the characteristic line intensities and/or their back- The measured correlations between different line intensities or
ground level. their ratios were rather poor and one of the reasons is a frequently low
Temporal behaviors of the atomic and molecular emissions were SNR for C, CN and C2 peaks. Some slight improvements were obtained
also studied on the same coal sample, by changing the gate delay and by calculating the line integrals instead of their peak values, but the
using short gate widths (250 ns, 500 ns or 1000 ns). Hydrogen scattering of the data points remained large. As a consequence, by
emission disappears after about 7 µs, a time about twice longer than applying an algorithm or PCA analyses on the single shot spectra, in
that observed for nitrogen lines. Emission from CN and C2 was many cases it was not possible to discriminate the explosives from
detectable up to 30 µs, when enlarging the gate width to 5 µs. All the interfering materials. In the following, the plots relative to the line
considered lines (Table 2) have smooth, approximately exponential emission intensities and their ratios were examined in order to
decay with time constants of 1–1.5 µs for atomic lines (the longest is determine the reasons for such large fluctuations.
for H) and 12–13 µs for CN and C2. The duration of the emission lines
in the plasma was further confirmed by measurements at a fixed gate 3.1. Clean support
delay (800 ns), while increasing progressively the gate width.
First, the signal behavior on clean support was investigated. In this
case, emission from H, N and O in plasma can be attributed to the air
Table 1
Residues analyzed.
surroundings. A potential small contribution to O emission coming
from surface oxidation can be considered as a characteristic of the
Name Composition Notes support. However, this contribution should not be significant, as the
EGDN C2H4N2O6 Explosive material is Anticorodal, chosen for its low oxidation. In order to
NG C3H5N3O9 Explosive extend a comparison between plasma on clean support and the same
RDX C3H6N6O6 Explosive
in presence of residue, the laser energy was varied from 160 to 250 mJ.
TNT C7H5N3O6 Explosive, contains aromatic ring
DNT C7H6N2O4 Explosive, contains aromatic ring We observed that O and N emission intensities slightly fluctuate
PETN C6H8N4O12 Explosive from one laser shot to another, around a mean value which was lower
HMX C4H8N8O8 Explosive at the minimum energy used. The O/N ratio, measured over 30 shots
TETRYL C7H5N5O8 Explosive, contains aromatic ring at 250 mJ, remains stable within RSD b 0.05. The corresponding RSD
TATP C9H18O6 Explosive, not containing N
Diesel oil C10H22–C15H32 Might contain aromatic rings up to 25%
for O/H and H/N ratios was as high as 0.26 (Table 3). The last two
Paraffin wax C20H42–C40H82 High purity, for dental use ratios might significantly change from one day to another, depending
Grease lubricant Unknown Contains hydrocarbons on the environmental humidity. However, this fact does not explain
Coal NIST 1632b C 78%, H 5.1%, N 1.6% Volatile matter 35.4 %wt the high RSD of the line ratios that include H, particularly for the
Glue LOCTITE C5H5NO2 Main component
spectra acquired in the same hour of the same day. Successively, we
Hand cream – Containing glycerol
observed that the highest H/N and the lowest O/H ratio always


Table 3 by the shock waves as the distance between the laser spots (measured
Average peak ratios O/H, H/N and O/N and corresponding RSD over 30 laser shots, from center-to-center) was small (1.5 mm).
measured on the clean aluminum support in three different days.

Energy (mJ) O/H RSD H/N RSD O/N RSD 3.2. Effects of chemical reactions in plasma
Day 1 190 4.74 0.26 0.59 0.26 2.642 0.047
Day 2 190 6.68 0.15 0.40 0.12 2.633 0.044 We compared emission intensities of H, N and O obtained on the
Day 3 190–250 4.35 0.22 0.65 0.23 2.689 0.037 support with and without organic residues. In Fig. 3 the results are
shown for diesel, TNT and HMX. The measured H emission is higher in
presence of the residues, as expected from their molecular formula.
appeared after the first laser shot of the measurement run. Later, H/N Differently, the lines from O and N are always less intense than in the
and O/H fluctuate strongly, whereas the H line intensity seems to case of clean support. These two elements are not present in diesel,
depend on the timing interval between the successive shots. Such containing only carbon, hydrogen and some impurities, and the
effect could be explained by a reduction of local humidity above the observed O and N lines in the spectra are also coming exclusively from
sampled point, caused by the shock waves generated after ablation by dissociation of air molecules. TNT and HMX contain both O and N;
the previous laser shot. Considering that air humidity at room however, O and N line intensities are again much lower than in the
temperature is not related to single H2O molecules but rather to case of clean support. This is a clear indication that atomic O and N are
much heavier water aerosol droplets, a relatively slow aerosol partially lost due to the occurrence of some chemical reactions during
diffusion back to the proximity of the sampling spot seems realistic. the detection window. Considering the molecular composition of
A confirmation of this supposition was obtained by using another LIBS diesel (hydrocarbon), it could be deduced that a presence of carbon
set up, not well comparable with the one described in this paper. (atoms, molecules or fragments) or an excess of hydrogen in the
Although the laser energy at 1064 nm was the same (250 mJ), the plasma, is responsible for the depletion of N and O atoms (Fig. 3). If
pulse duration, beam profile, spot size and collecting optics were this depletion was caused by hydrogen, we would observe it from the
different in the two cases. These LIBS measurements were performed spectra obtained on clean support in the conditions corresponding to
at a fixed laser repetition rate (1 Hz), while constantly moving the very different atmospheric humidity. From one day to another, during
sample at different linear velocities. The spectra registered after the ablation of the support we detected changes in the average H
first laser shot within the measurement run, always exhibit the emission intensity (line integrals) up to 40%. However, such large
strongest H emission for spot-to-spot distances up to about 10 mm changes did not induce any visible variations in O and N line
(Fig. 2). The closer are the successive spots, the more laser shots are intensities. Consequently, we might deduce that a presence of atomic
needed to reach a stable H emission intensity, which might be as low carbon or carbon containing molecules/fragments in the plasma is
as 50% of the value measured by the first laser shot. If the distance responsible for a partial removal of atomic O and N through some
between the laser spots exceeds a certain value, in our case estimated chemical reactions. Placing an organic residue on the support,
to about 10 mm, the H line intensity does not vary evidently from one reduction of O and N emission intensities are evident with respect
shot to another but fluctuates around some mean value. On a to the clean support. This offers an alternative possibility to detect
homogeneous sample (as the clean Al support) these fluctuations carbon in residues and with a higher sensitivity than through its much
could be minimized by keeping constant the laser repetition rate and weaker atomic and/or molecular emissions. Emission from atomic
scanning speed, and by discarding the spectra acquired after the first carbon, after discarding the spectra not containing it, has a growing
laser shots. However, in the case of non-uniformly distributed tendency with the support ablation, although corresponding to a
residues, the ablation rate and shock wave intensity vary from shot smaller amount of residue inside the laser spot. An increase of C
to shot. Consequently, changes in H emission intensity caused by the emission with reduced quantity of residue shall be discussed later.
laser-induced shock waves must be considered during LIBS measure- The considered atomic lines have similar trends with the Al
ments in air surroundings. This fact makes more difficult to identify intensity (Fig. 3) once the latter appears in the spectra (thinner
residues through the emission intensity of this element. residue), but their ratios are not constant. In Fig. 4 some of the atomic
In the experiment described in this paper, the timing between the ratios proposed for identification of explosives by LIBS [8] are shown
laser pulses was not constant (manual triggering) and always as a function of Al emission intensity. In the case of clean support,
exceeding 10 s. Although this timing is very large, we assume that increase of the material ablation, obtained by changing the laser
the observed significant variations of H intensity were indeed caused energy from 160 mJ to 250 mJ, leads to O/H and H/N changes of about
24%, with decreasing and growing trends, respectively. Concurrently,
O/N ratio is slightly reduced by about 7%. On different residues, H/N
and O/N ratios are constant or decreasing with Al emission intensity,
except in absence of support ablation, as well observable on diesel
residue. The measured ratios of C/N and O/H have different trends
(increasing, decreasing or almost constant) with support ablation,
depending on the residue type. A special case is represented by
nitroglycerin, in which the atomic line ratios are almost independent
on Al emission in the plasma (not shown). Comparing the behavior of
these four ratios with respect to Al emission, it was not possible to
distinguish explosives from other organic residues. Increase of the
atomic emission from residues with an amount of support ablation
(Figs. 3 and 5) is caused by a much more efficient ionization of the Al
compared to that of the organic molecules. This leads to higher plasma
temperatures, as shown in Fig. 5 (diesel residue). Here, the plasma
temperatures were calculated as average values obtained from the
from the Boltzmann plot and relative to the spectra with similar Al
Fig. 2. Hydrogen peak intensity measured on clean aluminum support as a function of
intensities (cases A–C). The Boltzmann plot was applied on Al atomic
the shot number: d is the center-to-center distance between the successive laser spots. lines with the upper level energies above 4.5 eV (cases B, C) or above
The laser repetition rate is 1 Hz, and the pulse energy is 250 mJ. 4 eV (case A). This limitation was set in order to reduce self-


Fig. 3. Comparison of the integrated line intensities from O, N, H and C observed for the clean support and in the presence of organic residues: diesel, TNT and HMX.

Fig. 4. Integrated line intensity ratios H/N, O/CN, C/N and O/H as a function of Al intensity, measured on the support and on diesel, RDX and TETRYL residues; the data were smoothed
over 3–5 adjacent points.


Fig. 5. Line integrated intensity of C, C2, N, and CN, smoothed over 3 adjacent points, as a function of Al emission intensity; the sample is diesel residue. Plasma temperatures averaged
over different spectra i.e. points (encircled) are shown.

absorption effects, expected at high Al concentrations. The plasma appear, the emission intensity of C2 rises abruptly, thus indicating the
becomes hotter for increased support ablation, leading not only to a transition to a more efﬁcient fragmentation of the organic material. C
more efﬁcient line excitation, but also to higher fragmentation and and N emissions intensities show a similar initial steep increase. In the
atomization rates of residue material, as evident from Fig. 5. Changing case of diesel, the growth of N intensity is also related to an increased
from the spectra not containing Al lines to those where they start to dissociation rate of N2 present in air. Unlike the case of the atomic

Fig. 6. Integrated intensity ratios of CN/N, C2/CN, C2/N and C2/C as a function of Al intensity, measured on diesel residues; the data were smoothed over 3 adjacent points.


lines, C2 emission shows a peak behavior with Al intensity. Once the for diesel residue when considering the spectra with intermediate Al
maximum fragmentation is reached, bond breaking of C2 molecules in intensities. For diesel, where we sampled a number of points with
hotter plasma supplies further C atoms, also contributing to its thick residue coverage, I(C2) / I(CN) as a function of parameter B
intensity growth. This dissociation leaves fewer molecules available shows better linearity (Fig. 7) with respect to the previous plot
for formation of CN fragments through the reaction: (Fig. 6). However, after plotting the atomic emission lines as a
function of B (Fig. 7, bottom), which to a certain extent includes the
C2 þ N2 →2CN ð1Þ plasma parameters, we observe again an abrupt change in their
tendencies when support ablation occurs. In this figure, the arrows
Emission from CN molecules (Fig. 5) also reaches the maximum at
indicate the direction of increasing Al emission in the plasma. After
certain Al intensity, i.e., at a certain plasma temperature. For most of
reaching a certain Al threshold emission, corresponding here to about
the residues analyzed, the maximum in the CN emission appears for
5 · 105 counts, the atomic line intensities start to behave linearly with
approximately the same intensity of aluminum atomic lines. However,
parameter B. This was also confirmed for other residue types after
C2/CN ratio also exhibits a peak behavior with the support ablation, i.e.
elimination of the points with low Al intensity, and the examples are
plasma temperature (Fig. 6). This peak occurs in correspondence of the
reported in Fig. 8. Here, we can observe that the trends of I(N) = f(B)
appearance of Al emission. The concentration of N2 molecules coming
can be opposite from one explosive to another. In particular, I(N) is
from air is higher at low plasma temperatures and decreases in the
decreasing with parameter B in the cases of TNT and EGDN (not
hotter plasma due to dissociation, as observable from the trend of N
shown). H and O emissions decay with parameter B, but, contrary to
atomic emission. Given the availability of N2 molecules, if reaction (1)
their behavior as a function of Al intensity (Fig. 3), it is now possible to
was the only pathway for CN formation, its peak emission should occur
discern TNT and HMX from diesel residue. Carbon emission as a
concurrently with the C2 emission maximum (Fig. 5) and the C2/CN ratio
function of B shows large fluctuations and is not adequate for the
(Fig. 6) should be a smooth function of the Al line intensity. On the
single shot sample classification. From these results, it seems that H, N
contrary, the peak behavior of the C2/CN ratio indicates that also other
and O intensities as a function of parameter B are other important
chemical reactions leading to CN formation [36] are present in the
characteristics to consider in residue classification by LIBS.
plasma during the chosen observation window. Due to the different
behavior of molecular and atomic emission intensities in the plasma as a
3.3. Recognition of explosives residues
function of the support ablation, the ratios of their lines are not constant
(Fig. 6), their variations being particularly pronounced for very low or
As discussed in the previous section, the line intensity ratios of the
absent support ablation. Even discarding these low intensity spectra, it is
elements forming organic compounds depend strongly on the amount
clear that the considered ratios CN/N, C2/CN, C2/N and C2/C have a non-
of the ablated support material. In [8], PCA analysis was performed on
linear dependence on Al intensity.
residues of three explosives, namely RDX, TNT and composition-B
The changes of the line intensity ratios with the amount of residue
(36% TNT, 63% RDX and 1% wax), and on three different interferents
inside the sampling point could be as large as one order of magnitude.
(dust, fingerprint and lubricant oil). In that work, the residues were
This fact excludes the possibility to identify residues by a simple
also placed on aluminum support, whose ablation was observed as
comparison of the line intensity ratios relative to the sample
constituents. Also, it is evident that performing signal averaging
over the spectra corresponding to very different support ablation
rates is not adequate. When applying chemometric tools, which rely
on the trends within the data set, an abrupt change of the line
intensities and their ratios when passing from the bulk similar
analysis (thick residue) to a concurrent support ablation (thin
residue) might compromise the correct sample classification. Fur-
thermore, the line intensity ratios do not exhibit linear trends with the
support ablation (Fig. 4), even after discarding the spectra
corresponding to the absence or to a low emission from the support
material. As the molecular emissions themselves or rationed to other
lines (atomic or molecular) exhibit a peak dependence on Al ablation
(see Figs. 5 and 6), their simplistic inclusion into the model might lead
to a reduced data correlation, as we checked on our data set by
applying PCA. In order to improve classification of the organic
residues, it would be important to introduce some other parameters,
related to the plasma temperature and chemical reactions, and to
obtain a more linear behavior of the line intensities and/or their ratios
inside the data set.
Considering again chemical reaction (1), we compared the CN
emission intensity with a parameter B, defined as:

B ¼ IðC2 Þ−2⁎K1 IðNÞ ð2Þ

Here, the first term represents the C2 intensity measured from the
spectra, while the second term takes into account that the N emission
intensity detected from the spectra corresponds to a reduced N2
availability in the plasma. As explained before, both line intensities, I
(C2) and I(N), are temperature dependent. In the case of the
compounds containing nitrogen, I(N) increases with the plasma
temperature more rapidly than in the case of pure dissociation of N2 Fig. 7. Behavior of the C2/CN ratio and of the N and O emission intensities as a function
molecules from air. Numerical value of the constant K1 was chosen to of parameter B for diesel residue; the arrows at lower figure indicate a growth of Al
be 0.6, which gives similar weights to both right hand terms in Eq. (2) emission in plasma.


Fig. 8. Line intensities of N, C, H and N as a function of parameter B, for diesel, DNT, HMX and TNT residues.

well. In PCA analysis, six line intensity ratios, based on their peak In order to improve the separation between the explosives and
values, were considered: O/N, O/C, H/C, N/C and O/H. In the resulting interferents, we tested different line ratios and their sums, but
three-dimensional PCA plot, there was a good separation between the without significant improvement in PCA analysis. In particular,
considered explosives, interferents and clean support. We performed including molecular to atomic line ratios into the model deteriorates
the same type of analysis on our LIBS data set, which includes residues the results of analysis, while including the Al intensity improves them
of nine explosives, 6 interferents (Table 1) and aluminum support. but only in the case of spectra where the emission is high. One of the
Except for the latter, the spectra not containing lines of C, CN or C2 variables that seems to improve point clustering is the previously
(SNR b 3) were excluded from the analysis. Instead of the peak defined parameter B, used in addition to the six atomic ratios. In
intensities, we used here the line integrals and obtained the following alternative, we used a new variable A, defined as:
scores in PCA analysis: the first PCA components describe 68.9% (PCA1),
19.8% (PCA2) and 8.1% (PCA3) of the total variance within the data set. A ¼ B−IðCNÞ þ K2 IðCÞ ð3Þ
The results for PCA2 and PCA3 as a function of PCA1 are shown in Fig. 9.
We do not show here PCA3 with respect of PCA2 as it does not improve When defining the parameter B, we considered a loss of N2
the classification for the given data set. On the both plots from Fig. 9, the molecules due to dissociation. Now, in the new variable A, we also take
data points belonging to the clean support are well clustered on the left, into account the emission from atomic carbon in plasma, which
residues from the interferents are on the extreme right side, while in a increases with plasma temperature because of the more efficient line
middle majority of the points from the explosive residues are scattered. excitation and the dissociation of C2 molecules. To the constant K2 we
For a single residue type, increasing value of PCA1 corresponds to lower attributed value of 2.0, chosen to have similar weights of both terms for
Al intensity in spectra. Knowing this, it becomes clear that better diesel residue and considering its spectra with intermediate Al
separation of explosive residues is obtained for high support ablation, intensities. For all the non explosives (except RDX), the calculated
where the plasma is hotter and atomization is more efficient. Also, from parameter A has a smaller range of values, grouped close to zero
the graphs reported in the previous section (see Figs. 3 and 4), one can (Fig. 10). It could be observed that the H/N ratio is always above 3.7 for
see that with an increased support ablation, different atomic ratios start the tested non explosives. Inside the data set for a single residue, the H/
to have more linear dependence on Al intensity. The worst cases are N ratio is lower for higher support ablation. In Fig. 10, the straight lines
represented by the spectra with very low or absent Al intensity, encompass the area containing all the interferents, and the points
corresponding to an abrupt change in the line intensities and their outside and not overlapping with well clustered points from the
ratios. On the plots in Fig. 9, all the points belonging to NG and TATP fall support, were considered to belong to explosives. In this case, some
inside the area that encloses the interferents. Unlike the other points of TATP and NG were also correctly identified (Table 4).
explosives considered here, TATP does not contain nitrogen, so it is Repeating the PCA analysis (Fig. 9) with the six atomic line ratios
more difficult to distinguish. Some other explosives, such as EGDN, but including the parameter A into model, we obtained the results
DNT, RDX and HMX are separated from the interferents only for spectra shown in Fig. 11. Now, the points relative to a single residue are more
with high Al emission, while the spectra with low Al emission linearly distributed for high Al intensities in the spectra. Here, we
contribute to false negatives (Table 4). considered that the points belonging to the explosives are between


Fig. 9. PCA analysis for residues and clean support: the straight line of the left delimits the interferents, while the regions on the left of the lines delimit points belonging to the clean
support.

two straight lines. For some explosives, the percentage of correct 4. Conclusions
classification was lower than in the previous PCA analysis, an
erroneous classification being always related to the spectra charac- In LIBS analysis of organic residues, here performed on clean
terized by low Al lines intensity. However, some points belonging to aluminum support in air surroundings, there are different sources of
TATP and NG, previously resulting as false negatives, now are variations in the characteristic line intensities and their ratios. For the
correctly classified (Table 4). measurements in air, large changes of H emission intensity might
occur also on homogeneous samples (aluminum), as here reported for
the first time. We demonstrated that this effect is due to local changes
of air humidity, induced by the shock waves following laser ablation
by previous laser pulses. Such changes, leading to H intensity
Table 4 reduction up to 50%, were detected up to a large distance between
Percentage of correct sample classification (explosive or not explosive) according
the laser spots — in our case 10 mm at a laser repetition rate of 1 Hz.
Figs. 9–11.
The same effect was also observable in a large time scale — in order of
Name Fig. 9 Fig. 10 Fig. 11 10 s, for close sampling spots (1.5 mm). A slow diffusion of air
EGDN 65 65 68 humidity back to the focal point was explained by the presence of
NG 0 39 39 water droplet aerosols at room temperature, much heavier and less
RDX 95 94 59 diffusive than air molecules. At the beginning of LIBS sampling in air,
TNT 100 100 100
DNT 79 56 64
the much higher presence of H in the plasma after the first laser shot is
PETN 100 100 100 also reflected in the plasma parameters. This is one of the reasons
HMX 95 90 95 explaining the sometimes reported differences in the LIBS signal
TETRYL 100 100 100 between the first and successive shots, even on homogeneous
TATP 4 31 54
samples.
Diesel oil 100 100 100
Paraffin wax 100 100 100 With the successive increase of the support ablation due to a
Grease lubricant 100 100 100 smaller quantity of residue (thickness) inside the sampled spot, the
Glue LOCTITE 100 100 100 atomic line intensities (C, H, N and O) initially increase abruptly, later
Hand cream 100 100 100 following an almost linear trend. Their observed rate of growth with
Support 100 100 100
support ablation, corresponding to a hotter plasma, is not the same: as


Fig. 10. Behavior of the H/N ratio as a function of parameter A for different residues and for the substrate.

a consequence, their ratios are subject to changes which could be even spectral intensity, C2/CN trend can strongly differ from one type of
as large as one order of magnitude. Furthermore, the line intensity residue to another, thus indicating an occurrence of various chemical
ratios show a non-linear behavior with support ablation. The latter reactions in the plasma. Due to the very different behavior of atomic
fact makes more difficult to classify the samples through observation and molecular line intensities as a function of the amount of residue, a
of their trends inside the data set (chemometrics). simplistic inclusion of molecular emission features into a procedure
In the presence of any organic residue, a part of atomic O and N for residue identification might lead to a larger percentage of incorrect
formed by dissociation of molecules from air is lost from the plasma classifications.
through some chemical reactions. We identify carbon or some carbon Considering that all the line intensities and their ratios are strongly
containing molecule/fragment as being responsible for this evident O dependent on the plasma temperature, here correlated with the
and N signal depletion. This finding suggests an alternative way of support ablation, and that the plasma parameters affect both the
detecting carbon in residues through reduction of O and N emission fragmentation/atomization and the chemical reactions, the averaging/
from air, instead of using much weaker carbon atomic or molecular accumulating procedure of processing LIBS spectra is meaningful only
lines. if these spectra correspond to similar support ablation rates.
Emissions from C2 and CN show a peak behavior with the Al line We examined nine types of explosive and six types of other organic
intensity, with different respective peak positions. Their line intensity residues, which were not uniformly distributed on the support surface.
ratio shows an abrupt change from the spectra not containing Al By the previously proposed PCA analysis, which considers 6 atomic
emission to the spectra where it appears. With a further increase of Al ratios [8], enclosing properly all the interferents (identification 100%),

Fig. 11. PCA analysis using six atomic ratios and parameter A.


we classified correctly 100% only three types of the explosives. For TATP [3] D.A. Cremers, L.J. Radziemsky, Handbook of Laser-Induced Breakdown Spectros-
copy, John Wiley&Sons Ltd, 2006.
and nitroglycerin, the identification completely failed. After including an [4] J.D. Winefordner, I.B. Gornushkin, T. Correll, E. Gibb, B.W. Smith, N. Omenetto,
additional parameter into PCA analysis, which partially takes into Comparing several atomic spectrometric methods to the super stars: special
account differences in chemical reactions in plasma, these two emphasis on laser induced breakdown spectrometry, LIBS, a future super star,
J. Anal. At. Spectrom. 19 (2004) 1061–1083.
explosives were also correctly identified for the spectra with strong Al [5] L.J. Radziemski, From LASER to LIBS, the path of technology development,
emission. The percentage of incorrect identifications increased for DNT Spectrochim. Acta Part B 57 (2002) 1109–1113.
and RDX with respect to the previous PCA analysis, but occurred always [6] S. Palanco, C. López-Moreno, J.J. Laserna, Design, construction and assessment of a
field-deployable laser-induced breakdown spectrometer for remote elemental
in the case of spectra corresponding to low support ablation. We point sensing, Spectrochim. Acta Part B 61 (2006) 88–95.
out that the emission lines from impurities were not taken into account [7] C. López-Moreno, S. Palanco, F. DeLucia Jr., A.W. Miziolek, J. Rose, R.A. Walters, A.
here, although this might improve the correct classification for the given Whitehouse, J.J. Laserna, Test of a stand-off laser-induced breakdown spectros-
copy sensor for the detection of explosive residues on solid surfaces, J. Anal. At.
set of samples. This choice is justified by the variations in impurity
Spectrom. 21 (2006) 55–60.
concentrations of the organic samples (also explosives) according to [8] J.L. Gottfried, F.C. De Lucia Jr, C.A. Munson, A.W. Miziolek, Strategies for residue
their origin. explosives detection using laser-induced breakdown spectroscopy, J. Anal. At.
From the result presented in this work, we conclude that an Spectrom. 23 (2008) 205–216.
[9] J.L. Gottfried, F.C. De Lucia Jr., C.A. Munson, A.W. Miziolek, Double-pulse standoff
affordable recognition by LIBS of an unknown amount of organic laser-induced breakdown spectroscopy for versatile hazardous materials detec-
residues inside the probed spot, must consider both the plasma tion, Spectrochim. Acta Part B 62 (2007) 1405–1411.
parameters and the chemical reactions in the plasma. We also suggest [10] C. Bohling, K. Hohmann, D. Scheel, C. Bauer, W. Schippers, J. Burgmeier, U. Willer,
G. Holl, W. Schade, All-fiber-coupled laser-induced breakdown spectroscopy
that some other parameters, in addition to the line intensity ratios, sensor for hazardous materials analysis, Spectrochim. Acta Part B 62 (2007)
should be introduced into chemometric models in order to improve 1519–1527.
the linearity of the trends inside the data set. In the present paper, we [11] F.C. DeLucia Jr., A.C. Samuels, R.S. Harmon, R.A. Walters, K.L. McNesby, A. LaPointe,
R.J. Winkel Jr., A.W. Miziolek, Laser-induced breakdown spectroscopy (LIBS): a
do not explain or model really a complex chemistry occurring in the promising versatile chemical sensor technology for hazardous material detection,
plasma produced on organic residues, but rather we only show that it IEEE Sens. J. (2005) 1–9.
affects the sample classification and that it should be taken somehow [12] F.C. De Lucia Jr., R.S. Harmon, K.L. McNesby, R.J. Winkel Jr., A.W. Miziolek, Laser-
induced breakdown spectroscopy analysis of energetic materials, Appl. Optics 42
into account. We believe that further studies are necessary to (2003) 6148–6152.
understand better the processes in the plasma, and that this requires [13] C.A. Munson, F.C. De Lucia Jr., T. Piehler, K.L. McNesby, A.W. Miziolek, et al.,
more extended and complex studies to be added to those performed Investigation of statistics strategies for improving the discrimination power of
laser-induced breakdown spectroscopy for chemical and biological warfare agent
only by LIBS. Also, it requires a much more extensive data set
simulants, Spectrochim. Acta Part B 60 (2005) 1217–1224.
modeling and testing than that considered up to now. [14] A.C. Samuels, F.C. DeLucia Jr., K.L. McNesby, A.W. Miziolek, Laser-induced
The best conditions for LIBS analysis of organic residues corre- breakdown spectroscopy of bacterial spores, molds, pollens, and protein: initial
spond to an efficient atomization of the organic molecules in the high studies of discrimination potential, Appl. Opt. 42 (2003) 6205–6209.
[15] J. Hybl, G.A. Lithgow, S.G. Buckley, Laser-induced breakdown spectroscopy
temperature plasma, obtained for higher ablation rates of the support detection and classification of biological aerosols, Appl. Spectrosc. 57 (2003)
material, equivalent to a smaller amount of residue inside the sampled 1207–1215.
point. In absence of the support ablation (thick residue layer), LIBS [16] B. Sallé, D.A. Cremers, S. Maurice, R.C. Wiens, P. Fichet, Evaluation of a compact
spectrograph for in-situ and stand-off laser-induced breakdown spectroscopy
probing is equivalent to analysis of bulk explosives, which was not the analyses of geological samples on Mars missions, Spectrochim. Acta Part B 60
subject of our presented studies. During rapid, in-situ analysis of (2005) 805–815.
residues, the sample preparation must be minimized, so it is necessary [17] F. Colao, V. Lazic, R. Fantoni, A. Paolini, LIBS application for analyses of Martian
crust analogues: search for the optimal experimental parameters in air and CO2
to consider that the sample material shall not be uniformly distributed atmosphere, Appl. Phys. A 79 (2004) 143–152.
on the support. This is true also for initially uniformly distributed [18] R. Wisbrun, I. Schechter, R. Niessner, H. Schröder, K.L. Kompa, Detector for trace
powder residues, because the laser induced shock waves blow up the elemental analysis of solid environmental samples by laser plasma spectroscopy,
Anal. Chem. 66 (1994) 2964–2975.
powder, leaving a non controllable quantity of residue on the support. [19] R.A. Multari, L.E. Forster, D.A. Cremers, M.J. Ferris, The effects of sampling
One possible way to speed the correct classification, which was found geometry on elemental emissions in laser-induced breakdown spectroscopy,
to fail for some explosives when present in a large amount inside the Appl. Spectrosc. 50 (1996) 1483–1499.
[20] S. Yalçin, D.R. Crosley, G.P. Smith, G.W. Faris, Influence of ambient conditions on
focal spot, is to insert a warning into the on-line implemented, data
the laser air spark, Appl. Phys. B 68 (1999) 121–130.
analysis software — alerting that the residue is too thick! A simple, [21] B.C. Castle, K. Talbardon, B.W. Smith, J.D. Winefordner, Variables influencing the
partial removal of the residue could lead to the correct identification precision of laser-induced breakdown spectroscopy measurements, Appl.Spec-
already by the successive laser shot. trosc. 52 (1998) 649–657.
[22] J.M. Anzano, M.A. Villoria, A. Ruız-Medina, R.J. Lasheras, Laser-induced breakdown
The role of the acquisition gate delay and width on the spectroscopy for quantitative spectrochemical analysis of geological materials:
classification of organic residues should be also examined, as gate effects of the matrix and simultaneous determination, Anal. Chim. Acta 575
widths shorter than those used in this work (30 µs), might reduce the (2006) 230–235.
[23] B. Sallé, D.A. Cremers, S. Maurice, R.C. Wiens, Laser-induced breakdown
influence of chemical reactions and recombination on the acquired spectroscopy for space exploration applications: influence of the ambient
LIBS spectra. pressure on the calibration curves prepared from soil and clay samples,
Spectrochim. Acta Part B 60 (2005) 479–490.
[24] C. Chaleard, P. Mauchien, N. Andre, J. Uebbing, J.L. Lacour, C. Geertsen, Correction
Acknowledgments of matrix effects in quantitative elemental analysis with laser ablation optical
emission spectrometry, J. Anal. At. Spectrom. 12 (1997) 183–188.
[25] L. Fornarini, F. Colao, R. Fantoni, V. Lazic, V. Spizzichino, Calibration for LIBS
This project was funded by EU commission, research project PASR- analyses of bronze materials by nanosecond laser excitation: a model and an
2006-6thFP-SEC6-PR-203600. experimental approach, Spectrochim. Acta Part B 60 (2005) 1186–1201.
Authors are grateful to colleagues from ENEA: Roberta Fantoni for [26] R.E. Russo, X. Mao, H. Liu, J. Gonzales, S.S. Mao, Laser ablation in analytical
chemistry — a review, Talanta 57 (2002) 425–451.
suggesting some of the organic interferents to examine, Roberto [27] A.S. Eppler, D.A. Cremers, D.D. Hickmot, M.J. Ferris, A.C. Koskelo, Matrix effects in
Ciardi and Dino Del Bulgaro for preparing the support materials. the detection of Pb and Ba in soils using laser-induced breakdown spectroscopy,
Appl. Spectrosc. 50 (1996) 1175–1181.
[28] V. Lazic, R. Barbini, F. Colao, R. Fantoni, A. Palucci, Self-absorption model in
References quantitative Laser Induced Breakdown Spectroscopy measurements on soils and
sediments, Spectrochim. Acta Part B 56 (2001) 807–820.
[1] J.I. Steinfeld, J. Wormhoudt, Explosives detection: a challenge for physical [29] V. Lazic, R. Fantoni, F. Colao, A. Santagata, A. Morone, V. Spizzichino, Quantitative
chemistry, Annu. Rev. Phys. Chem. 49 (1998) 203–232. Laser Induced Breakdown Spectroscopy analysis of ancient marbles and correc-
[2] D.S. Moore, Instrumentation for trace detection of high explosives, Rev. Sci. tions for the variability of plasma parameters and of ablation rate, J. Anal. At.
Instrum. 75 (2004) 2499–2512. Spectrom. 19 (2004) 429–436.


[30] J.M. Anzano, I.B. Gornushkin, B.W. Smith, J.D. Winefordner, Laser-induced plasma [34] V.I. Babushok, F.C. DeLucia Jr., J.L. Gottfried, C.A. Munson, A.W. Miziolek, Double
spectroscopy for plastic identiﬁcation, Polym. Eng. Sci. 40 (2000) 2423–2429. pulse laser ablation and plasma: laser induced breakdown spectroscopy signal
[31] L. St-Onge, R. Sing, S. Bechard, M. Sabsabi, Carbon emissions following 1.064 µm enhancement, Spectrochim. Acta Part B 61 (2006) 999–1014.
laser ablation of graphite and organic samples in ambient air, Appl. Phys. A 69 [35] R. Sattmann, I. Mönch, H. Krause, R. Noll, S. Couris, A. Hatzipostolou, A.
(1999) S913–S916. Mavromanolakis, C. Fotakis, E. Larrauri, R. Miguel, Laser-induced breakdown
[32] A. Portnov, S. Rosenwaks, I. Bar, Emission following laser-induced breakdown spectroscopy for polymer identiﬁcation, Appl. Spectrosc. 52 (1998) 456–461.
spectroscopy of organic compounds in ambient air, Appl. Opt. 42 (2003) [36] V.I. Babushok, F.C. DeLucia Jr., P.J. Dagdigian, J.L. Gottfried, C.A. Munson, M.J. Nusca,
2835–2842. A.W. Miziolek, Kinetic modeling study of the laser-induced plasma plume of
[33] M. Baudelet, M. Boueri, J. Yu, S.S. Mao, V. Piscitelli, X. Mao, R.E. Russo, Time- cyclotrimethylenetrinitramine (RDX), Spectrochim. Acta Part B 62 (2007)
resolved ultraviolet laser-induced breakdown spectroscopy for organic material 1321–1328.
analysis, Spectrochim. Acta Part B 62 (2007) 1329–1334.

Libs como detección de explosivos

More Related Content

What's hot (19)

Viewers also liked (19)

Similar to Libs como detección de explosivos (20)

Libs como detección de explosivos