A Novel Method for

Remotely Detecting
Trace Explosives
Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

The development of a technique with the ability

to detect trace quantities of explosives at a
» Events in recent years have led to an
increased need for improvements in our
abilities to detect explosives. In numerous
distance is of critical importance. In numerous situations, the capability to detect explosives
situations when explosive devices are prepared, sensitively, accurately, and rapidly could have great benefit
to national security both at home and abroad. Some of the
transported, or otherwise handled, quantifiable
homeland situations that would benefit from improve-
amounts of the explosive material end up on ments in explosives detection include screening passen-
surfaces. Rapid detection of these chemical gers and luggage at airports and other sensitive locations,
residues in a noninvasive standoff manner would and screening vehicles and people along the perimeters of
high-value installations such as federal buildings. Abroad,
serve as an indicator for attempts at concealed
the improvised explosive device (IED) problem is clearly
assembly or transport of explosive materials in need of creative solutions to mitigate the devices’ very
and devices. We are investigating the use of a damaging effects. Much research has been conducted to
help solve these problems; however, it is unlikely that any
fluorescence-based technique to achieve the
one solution will suffice.
necessary detection sensitivity. If we view the problem of explosive devices as a
timeline, from the initial planning to construction of the
devices and ultimately to their detonation, we can see
that it is preferable to detect the activities as early in the
process as possible. The term “left of boom” derives from
displaying the explosives creation-to-detonation timeline
from left to right and is sometimes used to refer to the
concept of early explosives detection. Figure 1 shows a
graphical representation of this process.
To address these concerns, we have begun several
efforts to both understand and help solve the difficult prob-
lem of explosives detection. In particular, we are focus-
ing on a solution for rapid and remote detection of trace
amounts of explosives, with the idea that such a capability
would significantly improve our ability to respond left of
boom. As we discuss later, bomb making is a messy process
A NOVEL METHOD FOR REMOTELY DETECTING TRACE EXPLOSIVES

Improvise
Gather and Evaluate
Obtain Develop CONOPS, Plan Perform
provide BOOM after-action
funds organization tactics, attack attacks
materials activities
devices

FIGURE 1. The interest in detection of explosives is focused on the early timeline activities, including the planning
stages, although after-action post-detonation evaluation could determine the detonation products for forensic pur-
poses. After-action activities on the part of the perpetrators would include evaluating the effects of the explosion.
CONOPS stands for concept of operations.

that leaves behind quantifiable residues on a variety of sur- in the photodissociation process is initially produced with
faces. Trace-detection techniques could thus be used foren- excess vibrational energy. This excess energy is extremely
sically, identifying devices and bomb-related activities, significant in that it not only allows us to distinguish
such as assembly, earlier in the process and thus prevent- explosives from smog, but also, as we show later, allows
ing the need for more difficult countermeasures later on. us to devise a detection scheme that should be relatively
To be practically useful, the technique must be rapid and immune to many common types of optical clutter (see the
sensitive. Additionally, long-range capabilities that could sidebar “Optical Clutter” on page 34).
scan from distances of at least ten meters open up numer- We have performed extensive studies [1] of the
ous possibilities in which large areas of interest could be bomb-making process to determine the sensitivity levels
rapidly assessed for indications of explosives activity. For necessary for detecting microscopic explosives residues.
a brief explanation of the issues we face in trace detection When explosive devices are prepared, transported, or
of various explosives, see the sidebar “Explosives and Their otherwise handled, quantifiable amounts of the explosive
Characteristics” on page 30. material end up on both people and related surfaces. The
We have been investigating a multistep technique in primary vectors of transport are the hands and feet of the
which the first step is a dissociation of polyatomic materi- bomb makers and bomb handlers. Even when people do
als into diatomic molecules. The technique is known as not directly contact explosive material, they may trans-
photodissociation followed by laser-induced fluorescence fer quantifiable amounts of residue that they pick up via
(PD-LIF). The laser powers required for this technique are secondary transfer. Such a transfer occurs when some-
significantly lower than those of some other methods, and one directly handling explosives touches a surface such
there is a good possibility of successful detection by using as a countertop or door handle, and then someone else
eye-safe lasers. When nitro-bearing explosives are illumi- touches that surface. The second person will pick up a
nated with ultraviolet (UV) light within their absorption quantifiable amount of explosives upon contact with the
band (see Figure 2) and of a sufficient intensity, the mol- surface and may then transfer it to additional surfaces.
ecules dissociate (the PD step of PD-LIF), creating, among We carefully quantified the amounts that were trans-
other things, fragments of nitric oxide (NO). Identification ferred to surfaces under a variety of conditions. These
of this NO photofragment through laser-induced fluores- surfaces, such as vehicles, luggage, clothing, walkways,
cence (LIF) forms the basis of our detection technique. NO and doors, could potentially be used as early indicators
(not to be confused with nitrous oxide, N2O) can be found of activity. Although we identified large variability in the
in smog. There is, however, a very important difference amount of explosives residues transferred to surfaces,
between the NO produced in the dissociation of explosives in most instances, a majority of surfaces of interest had
and that which can occur as a pollutant. The atmospheric quantities of at least 1 μg/cm2. We use this quantity as
pollutant exists in its ground state, while the NO produced a rough estimate of the required sensitivity for a foren-
 Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

100

Cross section (10–18 cm2/molecule)

sic technique. Note that this areal coverage amounts, on
average, to only a few monolayers of explosives.
80 TNT

Trace-Detection Techniques 60
Given the characteristics detailed in the sidebar on
explosives, what sorts of techniques are possible, if we 40
keep in mind the constraint that we require something
rapid, sensitive, and remote? A number of techniques 20

are currently in use for point detection (at a location),

0
including ion-mobility spectrometry (in use in many 200 300 400 500 600
airports), gas-chromatography mass spectrometry (gen- Wavelength (nm)
erally a laboratory technique), and a somewhat newer
NO
technique based on the fluorescence quenching of poly-

Absorption (arbitrary units)

mers. The detector employing the latter is known as Ground
Fido. The Fido explosives detector is built by Nomadics, state
Vibrationally
Inc. (a business unit of ICx Technologies, Inc.) and is excited
based on a proprietary technology developed by Timothy 2700 K
Swager at MIT. The product is so named because its level
of detection is comparable to that of a highly trained
explosives-detection dog. All these techniques can detect
minute quantities of explosives (generally in the pg-to- 200 210 220 230 240 250 260
ng range). However, they all share one common property Wavelength (nm)
that makes them ill suited for standoff detection. They =N =O =C
all require ingestion of the explosives into their detec-
FIGURE 2. The analysis of explosives is improved with the
tors. In other words, they must come into direct physical detection and distinction of NO units. These optical absorp-
contact with the explosive molecules themselves. Given tion spectra of molecular species show the improvement.
the vanishingly low vapor pressures of explosives (Fig- The top graph is of TNT, a polyatomic molecule in the solid
phase. Note that the optical absorption spectra of other
ure 2), it does not appear feasible for us to operate these
organic nitro-bearing explosives are generally similar to that
techniques remotely.
of TNT because they have a broad peak in the deep ultravio-
We found that, in general, a useful detection tech- let (UV). The bottom graph is of nitric oxide (NO), a diatomic
nique must fulfill two potentially conflicting require- molecule in its gas phase. Both the absorption spectra of
ments. First, it must have high sensitivity: it must be able the ground (solid lines) and the first vibrationally excited
(dashed lines) states are shown. Photodissociation, the first
to detect a signal generated from as little as a single mono-
step of the photodissociation laser-induced fluorescence
layer of material. For the technique to operate remotely, (PD-LIF) process, creates the distinct NO spectrum from
this signal must be accessible from several meters’ dis- the generally featureless spectrum of TNT. Also note that
tance. None of the above techniques appears to us to be PD-LIF creates NO in an excited state, which allows for a
able to meet this distance requirement. Second, the tech- low-clutter detection technique to be applied.

nique must have high specificity. The signal generated

by explosives should be readily distinguishable from the generated by a source other than explosives and that do
signal generated by other common materials. We use the not share the same optical characteristics as our signal.
term chemical clutter to refer to nonexplosive materials While optical clutter can be addressed via optical filtering
with optical characteristics similar to those of explosives. techniques, chemical clutter could pose a more serious
In our case, these materials might include other NO-con- difficulty. Additional considerations for a successful tech-
taining compounds. False alarms may also occur as the nique include its practicality—the system should have a
result of optical clutter (see the sidebar “Optical Clutter”). minimum of components and complexity to facilitate ease
We use the term optical clutter to refer to photons that are of use in a real-world situation.
A NOVEL METHOD FOR REMOTELY DETECTING TRACE EXPLOSIVES

Explosives and Their Characteristics

Some compounds are easier to detect than others.

Explosives may be broken in the plastic explosive C4. These per quadrillion, in the case of octa-
down into two general classes: materials make up a large percent- hydro-1,3,5,7-tetranitro-1,3,5,7-
nitro/nitrate-based and non-nitro/ age of the explosives used in impro- tetrazocane (HMX). Low vapor
nitrate-based. Non-nitro/nitrate- vised explosive devices. pressure places severe limitation on
based explosives are derived from To develop and/or assess the a technique that relies on the ambi-
materials such as peroxides, e.g., utility of a trace-detection tech- ent vapor above the material for
triacetone triperoxide (TATP), per- nique, we have to understand some detection. The solid residue itself
chlorates, and azides. While these of the physical characteristics of holds much more potential signal
explosives clearly pose a threat, these explosives. Figure A shows than the vapor phase.
our studies focused upon the more the chemical structure and vapor • Nitro-based explosives all have at
common nitro-based explosives. pressure of some of the more com- least one NO2 group. When nitro-
The military primarily uses these mon nitro-based explosives. Other based explosives detonate, one
types of explosives; they include key characteristics of explosives are of the ultimate end products is N2
compounds such as nitroglycerin, the following: gas, which is very stable due to its
2,4,6-trinitrotoluene (TNT), and • Explosives have extremely low N-N triple bond. The formation of
hexahydro-1,3,5-trinitro-1,3,5-triaz- vapor pressures, ranging from this low-energy product from the
ine (RDX), the active component parts per million to less than parts higher-energy starting material is

We concluded that optical techniques appear to have A more promising technique utilizes Raman spec-
the greatest potential for remote-detection capability. troscopy. It has been known for some time that there is a
A variety of approaches to optical detection have been weak but distinct interaction between photons illuminat-
attempted [2], and we found it instructive to examine them ing a material and the vibrational modes (phonons) of the
in the context of the above-mentioned constraints. UV illu- material. The result of this interaction is a scattered pho-
mination of explosives induces a relatively strong fluores- ton with either slightly more or less energy than that of
cence signal. However, it is not very specific to explosives, the incident photon, depending upon whether a phonon
since many materials fluoresce under UV illumination and has been absorbed or created. This inelastic scattering
their fluorescence spectrum is often nonspecific. Thus this process creates a spectral fingerprint of the material spe-
technique does not fulfill the specificity constraint; optical cific to its vibrational structure; thus it is expected to have
clutter would be a significant problem. Other techniques reasonably low clutter. While these Raman energy shifts
generate a reflection signal based on the absorption peaks are successfully used to identify materials at very close
in explosives. Again, optical clutter appears to be a prob- range in the laboratory, the scattered signal is so weak that
lem because the spectra are broad and not necessarily detection at long range would require the combination of
unique to explosives. Furthermore, only a very thin layer large collection optics and long integration times.
of material (monolayer) is being interrogated; thus it will An alternate approach, which removes many of the
not absorb much light beyond the ultraviolet. Additionally, difficulties associated with the indistinct spectra of the
and perhaps more importantly, the explosives distribution materials themselves, is to dissociate the materials and
is nonuniform, such that much of the reflected signal may use instead signals related to their constituents. Laser-
not be dominated by the explosives themselves but by the induced breakdown spectroscopy (LIBS) is one such tech-
surface upon which they reside. nique, which in its most common incarnation analyzes
 Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

accompanied by the release of a 10–6 broad and nondistinct. There is a

Nitroglycerin
sizable amount of energy. Note broad absorption peak in the ultra-
that inorganic nitrate-based explo- violet (UV), and the explosives

Relative concentration in air

sives have NO3 groups instead of TNT will fluoresce in the visible as a
10–9
NO2 groups (ammonium-nitrate result of UV illumination.
fuel oil is one such example). Our
PETN references
initial work focused on organic RDX (C4)
a. National Research Council, Existing
materials that contain NO2 10–12 and Potential Standoff Explosives
groups. Detection Techniques (The National
• The explosives residues relating Common feature: Academies Press, Washington,
NO2
to bomb activities can be, on aver- D.C., 2004).
age, only a few monolayers thick. 10–15 b. J. Steinfeld and J. Wormhoudt,
“Explosives Detection: A Challenge
However, they are nonuniform,
HMX for Physical Chemistry,” Ann. Rev.
taking the form of mounds or Phys. Chem., vol. 49, 1998, pp.
clumps at the microscopic level. =N =O =C 203–232.
This nonuniformity means that
some percentage of a surface of FIGURE A. The vapor pressures of some
nitro-bearing explosives are quite low and
interest may be bare, containing
span orders of magnitude, indicating the dif-
no explosive material at all. ficulties of trace detection of vapors [a, b].
• The optical absorption spec- TNT is 2,4,6-trinitrotoluene, RDX is hexa-
trum (Figure 2 in the main text) hydro-1,3,5-trinitro-1,3,5-triazine, PETN is
of nitro-bearing explosives is pentaerythritol tetranitrate, and HMX is octa-
hydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocane.

the atomic constituents. LIBS uses tightly focused, high- possibility (as we discuss later) that successful detection
energy laser pulses to dissociate the materials into their can occur if eye-safe lasers are used (again in contrast to
constituent atoms via the formation of a microplasma. LIBS). When nitro-bearing explosives are illuminated
The atomic emission spectra allow identification of the with UV light within their absorption band (see Figure
atoms and estimation of their relative abundances. A 2) and of a sufficient intensity, the molecules dissociate
nitrogen/oxygen ratio consistent with explosives is used and create, among other things, fragments of NO. Identi-
as an indicator of the presence of a specific explosive. fication of this NO photofragment forms the basis of our
While the LIBS signal is strong, the relative ratios can be detection technique. As mentioned earlier, the NO prod-
easily obscured by the existence of other nitrogen and/or uct of PD has excess vibrational energy when compared
oxygen-bearing materials present on surfaces. In other to the potential chemical-clutter components of NO in
words, LIBS is highly susceptible to clutter [3, 4]. In smog, for example.
addition, there is the disadvantage of using high-intensity An additional advantage of having produced the NO
lasers that aren’t eye safe. molecule via photodissociation is that we are now deal-
ing with a gas-phase diatomic molecule (as compared to
Photodissociation Followed by a solid-phase polyatomic molecule), which has a much
Laser-Induced Fluorescence more distinct spectrum that can be used as an identify-
PD-LIF is a dissociation-based technique in which the ing fingerprint. The lower portion of Figure 2 displays
polyatomic materials are dissociated into diatomic mol- the absorption spectrum of NO. Note the distinct lines, as
ecules, as opposed to the atoms produced via LIBS. The compared to the broad spectrum of 2,4,6-trinitrotoluene
laser powers required for the PD-LIF process are sig- (TNT). Also note the different spectra for the ground and
nificantly lower than those of LIBS, and there is a good excited states of the molecule. As can be seen from the
A NOVEL METHOD FOR REMOTELY DETECTING TRACE EXPLOSIVES

S 2Σ

Vibrational
Electronic

levels
states
2. Optically pump NO

Energy
(same laser pulse as step 1) 2 X 2Π
3
Pulsed UV laser 3
2 Photo-
detector
Vibrationally 226 nm
236.2 nm excited
Internuclear distance
nitric oxide
1
3. Detect higher-energy
1. Vaporize and photons
photodissociate Eliminates red-shifted
solid explosive �� h� fluorescence clutter
�� excited NO vapor
Explosives residue
(nitro-bearing)
=N =O =C

FIGURE 3. The PD-LIF explosives-detection technique can be understood as a three-step process. In the first step, a mate-
rial with a distinct signature is created by vaporizing and dissociating solid explosives and forming gaseous NO fragments. In
the second step, these fragments are optically pumped to a higher-energy state. Finally, the higher-energy fluorescence emis-
sion is detected. The wavelength of the fluorescence and the wavelength required to induce fluorescence are both very pre-
cise, which provides the specificity of the technique.

figure, the energy difference between the two is ~2700 K: state in a standard atmosphere.
thus the excited state is not occupied under standard 3. The resulting NO fluorescence, which occurs
ambient temperatures. In PD-LIF, detection of the vibra- at a shorter wavelength than that of the laser, is
tionally excited NO is achieved via laser-induced fluores- used for detection.
cence. Here, a photon further pumps the vibrationally Note that all these steps occur within a single laser
excited NO fragment into an electronically excited state. pulse. The exquisite specificity of this technique derives
The electronically excited NO molecules rapidly fluoresce from the precise wavelengths involved in the LIF process.
as they return to their ground states. Because of the initial It is highly unlikely that many other materials will pro-
excess of vibrational energy, this fluorescence occurs at a duce photons at precisely 226 nm in response to illumi-
shorter wavelength (226 nm)—i.e., higher energy—than nation at precisely 236 nm, since few will share the exact
the exciting laser photons (236.2 nm). Signal photons that spectral structure of the excited NO. Thus we can utilize
are of shorter wavelengths, or shifted toward the blue end a narrow-wavelength source and a narrowband detec-
of the spectrum relative to the laser photons, are critical tor to specifically detect these NO fragments with a high
to minimizing optical clutter. A schematic of this multi- degree of sensitivity and specificity. We note also that it is
step (photodissociation/vaporization—photoexcitation— fortuitous that the nitro-bearing explosives absorb quite
fluorescence) detection process is shown in Figure 3. In well at the UV wavelength needed to pump NO. Because
brief the steps are the following: of this absorption, we can accomplish the multiphoton
1. The first photon is absorbed by the explosives process with a single laser pulse, significantly simplifying
and very rapidly (<<1 ns) vaporizes and dissoci- the necessary equipment.
ates the explosives’ components into fragments Advantages of this detection method include a rela-
including the vibrationally excited NO. tively strong fluorescence signal—which can be gener-
2. A second photon pumps the vibrationally ated by using eye-safe excitation-laser intensities—and a
excited NO to an electronically excited state. low false-alarm rate. The low false-alarm rate is expected
This pumping must happen within a few ns, because relatively few processes produce light at shorter
which is the lifetime of the vibrationally excited wavelengths than the source laser. In contrast, most other
 Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

processes, such as scattering or traditional fluorescence, performed primarily in a close-range geometry, so that
are single-photon processes in which each incident pho- the PMT and filters were positioned 6 cm directly above
ton yields another of equal or lesser energy. The PD-LIF the sample. No collection or collimation optics were used
process involves multiple photons, the first of which in the close-range configuration. In this geometry, we
creates a molecule with excess energy and the second estimate a collection efficiency of 4 × 10–7 of all emitted
of which further excites that molecule. As the molecule signal photons.
relaxes to its ground state (see Figure 3), photons are A variety of explosives in a variety of morphologies
emitted with greater energy than that of the incident were studied. The explosives included 2,6-dinitrotoluene
photons. As such, this technique isn’t susceptible to false (DNT), TNT, pentaerythritol tetranitrate (PETN), and
alarms from traditional fluorescence processes, for which hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX). DNT, as
the resultant photons typically are at longer wavelengths. received from Sigma-Aldrich, was in the form of small
The sidebar on optical clutter shows how these higher- granules roughly the size of salt crystals. Studies were
energy emitted photons are distinguishable from typi- performed on macroscopic mounds (several milligrams)
cal fluorescence coming from the surfaces on which the
explosives reside.

PD-LIF Detection Measurements

Previous studies by others [5] have examined the PD-LIF
Photodetector
signature from explosives vapors. We, however, probed the
solid and liquid explosives themselves [6, 7] as they offer
the potential for much greater signal, since the vapor pres-
sure of explosives is very low. The four main components
in single-laser PD-LIF laboratory experiments shown in
Pulsed UV laser
Figure 4 are the excitation laser, the photodetector, the
bandpass optical filters, and the samples being tested. In
our studies, the laser was a tunable pulsed Continuum
9030 system, which provides variable laser output from
215 to 310 nm. The laser output was an ~1 cm2 beam of
7 ns pulses, emitted at a rate of 30 pulses per second, with
energies of 2 to 3 mJ per pulse. The spectral linewidth was
0.03 to 0.04 nm around the wavelengths of interest. All Optical filter
wavelengths reported are the values in air (approximately
0.03% less than the wavelengths in a vacuum).
The fluorescence detection subsystem consisted of
a solar-blind (tuned to eliminate the background solar
radiation) photomultiplier tube (PMT) with narrowband
filters. Narrowband filters are essential in these experi- TNT residue
on sand
ments because they are needed to suppress the scattered
laser light. The Perkin Elmer cesium-tellurium channel
photomultiplier had single-photon sensitivity, negligible
dark counts, and a quantum efficiency of 10% near the FIGURE 4. The four primary components of the experi-
signals of interest (220 to 230 nm). The PMT was used mental detection setup are shown: the pulsed UV laser, the
high-rejection optical filters (which are normally attached
in both a photon-counting and linear mode (dependent
directly to the photodetector and allow only signal photons
upon photon flux) and was calibrated for short laser to reach the detector while blocking the scattered laser pho-
pulses at various cathode voltages in order to ensure a tons), the photomultiplier tube photodetector, and the sam-
linear response at a given flux level. Experiments were ple being tested (here it is sand covered with traces of TNT).
A NOVEL METHOD FOR REMOTELY DETECTING TRACE EXPLOSIVES

Optical Clutter
Reduce the effects of clutter on false-alarm rates by taking measurements where spurious
fluorescence isn’t.

Detection of spurious radia- fact opens up the possibility of and shorter wavelengths than the
tion, or optical clutter, causes operation of a detection system laser). Generally, the scattered
an increase in false-alarm rates. that relies upon a very small num- photons (at the laser wavelength)
Potential sources of optical clut- ber of signal photons (perhaps as are the strongest source of clutter.
ter can be either ambient or laser- few as one) for detection. The Fluorescence processes (the domi-
induced radiation, as shown in PD-LIF detection system we are nant form of red-shifted clutter) can
Figure A. pursuing operates in the deep UV yield signals as strong as ~10% of
Ambient Clutter. While solar and thus aims to utilize just such the laser scatter, depending on the
radiation is a dominant source of single photons. material and wavelengths chosen.
clutter in the visible and near infra- Laser-Induced Clutter. Note that many materials fluoresce
red, and thermal radiation can be Laser-induced clutter can be sub- when irradiated with UV light (just
problematical in the infrared, the divided into three broad spectral visit your local roller rink). Photons
deep ultraviolet (UV) is quite clutter categories: laser scatter, i.e., pho- that are created with more energy
free. Below approximately 300 nm tons at the laser wavelength; red- than the energy of incident photons
(the solar-blind region), the absorp- shifted clutter (photons with less are quite rare because they require
tion due to atmospheric ozone pre- energy and longer wavelengths the addition of excess energy. Thus
vents almost any light from reaching than the laser); and blue-shifted blue-shifted clutter is much weaker
the surface of the earth [a]. This clutter (photons with more energy than in either of the other two

of these solid granules, in addition to a liquid form, which explosives samples display the same multipeak struc-
was obtained by heating the granules to 80°C. With the ture with a maximum signal at the excitation wavelength
exception of the liquid DNT measurements, all other of 236.2 nm and a shoulder near 236.3 nm. They also
measurements were performed at room temperature display a secondary peak at 236.9 nm, again with an
under ambient atmospheric conditions. Military-grade accompanying shoulder.
TNT was studied in a solid pellet form, as a trace coat- The similarity of these signatures implies that all
ing on sand, and dropcast from a dilute acetone solution. nitro-bearing explosives can be detected via this tech-
Military-grade PETN was studied in the form of a white nique without the need to fine-tune the laser parameters
powder. RDX was studied both as a trace coating on sand to each individual explosive type. The vapor pressures
(8% by weight) and as the dominant component (~90%) for these compounds differ significantly—by over four
in the putty-like C4 plastic explosive. orders of magnitude at room temperature—while their
Figure 5 displays the results of fluorescence detec- signal strengths are within one order of magnitude. These
tion measurements of DNT, TNT, C4 (RDX is the active measurements indicate that the observed signal is not
component), and PETN. Data were taken at a fluence related to the ambient vapors of the materials, but rather
of 10 mJ/cm2/pulse (1-mJ pulses over 0.1-cm 2 area). the condensed phase itself. This is an important point in
Data points represent 6-pulse averages, with the excep- our consideration of solid-phase analysis: a technique
tion of the background measurements on the bare silica that relies upon detection of the ambient vapors of the
substrate (shown as the open squares in the top graph) explosives will be severely limited due to their very low
for which the data points are 60-pulse averages. All vapor pressures [8, 9].
 Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

Solar-blind regions. Anti-Stokes Raman scat-

region Solar spectrum tering involves excess energy that
(photons/s/cm2/nm) 1012 comes from the vibrational energy
10 8
Solar illumination (daytime) (phonons) of the material scattering
Solar flux

104 is 1015–1017 photons/s/cm2 the light. Processes that involve the

simultaneous absorption of more
100 Upper-bound flux
than one photon (multiphoton pro-
10–4 (MIT LL measurements)
cesses such as our PD-LIF) can
200 400 600 800 1000 1200
Wavelength (nm) also produce a blue-shifted signal.
The cross sections for all these pro-
cesses are exceedingly weak, and
Optical clutter
thus this regime offers a useful por-
Laser Red-shifted clutter tion of the spectrum in which to do
100 Traditional fluorescence low-clutter detection.
Relative signal

Laser scatter

10–3
10–6 Blue-shifted clutter reference
10–9 Anti-Stokes, multiphoton
a. A. Thompson, E.A. Early, J. DeLu-
10–12
10–15 isi, et al., “The 1994 North Ameri-
More energy than laser λ Less energy than laser can Interagency Intercomparison of
Ultraviolet Monitoring Spectroradi-
FIGURE A. Ambient clutter is present whenever the system is in an environ- ometers,” J. Res. Natl. Inst. Stand.
ment where sources other than the laser are present. Such clutter is shown In Technol., vol. 102, no. 3, 1997, pp.
the upper image, specifically for solar-induced scatter. The laser-induced clut- 279–322.
ter exists in three wavelength regions. The laser scatter (at the laser’s wave-
length) and the lower-energy red-shifted scatter do not affect the PD-LIF
measurement, which occurs at a higher-energy blue-shifted wavelength.

The bottom graph of Figure 5 displays the predicted 236.2 nm, complete photolysis of TNT takes about 10
fluorescence spectrum of NO, assuming the PD-LIF exci- laser pulses, as shown on the right in the figure. On the
tation (excitation from the first vibrationally excited state basis of these results, we can estimate that in our specific
of its electronic ground state to the vibrational ground experimental configuration we were detecting 20 ng TNT
state of its first electronically excited state). We obtained with each laser pulse (2 μg/cm2 times 0.1 cm2/10 pulses).
these results by using the LIFBASE software package Also displayed in Figure 6 is the photo response from a
[10,  11]. Comparison of the experimental data to the control sample (the bare silicon wafer), which is signifi-
computed NO spectrum provides clear evidence that the cantly weaker than that of the TNT, indicating a relatively
measured signal is generated by excited NO species. high signal-to-noise ratio for detection.
We used calibrated quantities of explosives dissolved
in acetone to demonstrate trace-level detection. These Remote Detection Projections
explosives were dropcast on silicon wafers to yield areal Our trace-detection results along with the spectra of the
concentrations of 2 μg/cm2 (shown in Figure 6), similar different explosives indicate a technique with the required
to concentrations expected from fingerprint residues. We sensitivity to detect the trace amounts of interest. Keeping
investigated both RDX and TNT in this manner and mea- in mind the eventual application of PD-LIF to the stand-
sured signals roughly half of those measured for the bulk off detection of these residues, we need to estimate the
materials. Laser-induced photolysis of the trace material detection range of a notional system. For a first step, we
is evident in the top portion of the left-hand figure. For calculate the relative efficiency of this process, compared
laser fluences of 10 mJ/cm2/pulse and a wavelength of to the alternative of Raman spectroscopy. (While compar-
 A NOVEL METHOD FOR REMOTELY DETECTING TRACE EXPLOSIVES

25
ison to LIBS is also possible, PD-LIF and Raman-based
DNT (liquid)
20 detection have in principle similar low false-alarm rates.)
Signal at 226 nm
(photons/pulse)

15
The Raman cross section for TNT has been reported as
2 × 10–31 cm2/sr/molecule for visible excitation and 10–28
10
cm2/sr/molecule for ultraviolet resonant excitation, and
5 therefore the angle-integrated values are 2.5 × 10–30 cm2/
molecule and 10–27 cm2/molecule, respectively [12, 13].
0
12 The spontaneous Raman scattering process is linear in
TNT (solid) fluence, and therefore the cross section is a constant. In
Signal at 226 nm
(photons/pulse)

8 contrast, the PD-LIF is nonlinear in fluence because it is

a multiphoton process. A cross section will therefore be
4
fluence dependent. Furthermore, it will be an effective
parameter rather than a true molecular property.
0
By analogy to a linear process such as Raman scatter-
12 ing, the number of fluorescence photons collected in our
C4 (solid) experimental setup, Nphoton, can be expressed as follows:
Signal at 226 nm
(photons/pulse)

8
N photon = F n σ eff η ,

4
where F is the laser fluence (in photons per unit area), n is
0
the number of molecules being irradiated, σeff is the effec-
tive cross section of the overall PD-LIF process, and η is
PETN (solid)
2 the experimental collection efficiency. We have estimated
Signal at 226 nm
(photons/pulse)

the PD-LIF cross section, σeff, on the basis of our dropcast

results at a fluence F = 10 mJ/cm2/pulse (1.25 × 1016 pho-
1
tons/cm2). At this fluence, we collect Nphoton ~10 photons/
pulse. We can estimate the number of irradiated mole-
0 cules, n, by using the areal density of the dropcast TNT
1.0
(2 μg/cm2), the laser spot size (0.1 cm2), and the molecu-
NO lar weight of TNT (227 g/mole).
Signal (arbitrary units)

fluorescence
Given that ~10 laser pulses were required to fully pho-
tolyze the film, and that the total number of molecules in
0.5
the film was 5 × 1014, we estimate that n = 5 × 1013 mol-
ecules/pulse were irradiated. Note that the average thick-
ness of the dropcast TNT is ~12 nm (the density of TNT is
0.0
235 236 237 238 1.65 g/cm3), similar to our measured photolysis rate per
Pump laser wavelength (nm) pulse. However, we observed experimentally that 10 pulses
were required to photolyze the film, rather than one pulse,
FIGURE 5. The response of NO-based materials to pump as may be indicated by the average film thickness. These
laser wavelength is selective. The shifted peaks and shoul- results are reconciled by noting, as is evident from optical
ders show a marked response to the NO first excited state. images of the films, that the films are not uniform. Rather,
The top graph also includes a background measurement of
they consist of islands of different sizes and, presumably,
the response of the silica substrate without explosives (open
squares). The bottom graph is the predicted fluorescence of
thicknesses. Using our collection efficiency of η = 4 × 10–7,
NO, assuming excitation from its first vibrationally excited we calculate σeff = 4 × 10–23 cm2/molecule for TNT at
state. Note that the emission from these materials is actu- 10 mJ/cm2/pulse fluence. This is a factor of 4 × 104 higher
ally at the higher-energy blue-shifted 226 nm wavelength. than that achievable with Raman scattering in the opti-
 Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

20

Laser ablation

Signal at 226 nm (arbitrary units)

TNT
15

10

Bare silicon wafer

0
1 cm 0 10 20 30 40 50 60
Pulse number

FIGURE 6. Dropcast TNT was analyzed to determine the trace-detection levels available with PD-LIF. On the left is a photo
image of dropcast TNT on a silicon wafer with the laser spot (causing photolysis) labeled. On the right are time-series data
displaying signal photons as a function of laser pulse (the top curve for TNT is offset for clarity from the bottom bare silicon).

mal condition of UV excitation, and ~107 higher than that mate the fluence dependence of σeff . This dependence is
of the more conventional visible/near-IR Raman scatter- compared to the Raman cross section, which remains fixed
ing. Making the simplifying assumption that the number with fluence, in Figure 7. As can be seen in the figure, PD-
of molecules probed via PD-LIF is fixed at n = 5 × 1013 LIF compares favorably with Raman scattering as a process
molecules/pulse, we use our data at other fluences to esti- enabling standoff detection of TNT and other nitro-bearing
explosives. Using these estimates for the PD-LIF cross sec-
10–22 tion, we estimate that it will be feasible to achieve detection
10–23 of trace levels of explosives with a single pulse of eye-safe
Cross section (cm2/molecule)

PD-LIF
10–24
UV illumination at distances of tens of meters.

10–25
Future Work
10–26 To achieve these detection ranges (tens of meters) would
UV
10–27
require ~30 cm optics and a laser capable of producing
at least 5 mJ pulses of light at 236.2 nm. While the laser
10–28
Raman used in our laboratory measurements produces nearly
10–29 IR these pulse energies, it has several drawbacks. Its pulse
10–30
repetition rate is low (30 Hz), which makes rapid area
0 5 10 15 20 25 30 scanning impossible. It is also large and requires signifi-
Fluence (mJ/cm ) 2
cant input power. Moreover, it requires precise alignment
and as such is not robust enough for low-maintenance
FIGURE 7. The TNT effective cross sections as a function field measurements. Finally, its bandwidth (0.03 nm) is
of fluence for PD-LIF (squares) clearly outperform Raman- not well matched to the explosives signal peak (~0.5 nm;
detection cross sections [14, 15], whose UV upper and infra-
see Figure 5).
red lower bounds are denoted by dashed lines. Note that
PD-LIF cross-section estimates assume that the number of
John Zayhowski of Lincoln Laboratory’s Laser Tech-
particles probed per pulse was constant with fluence, and nology and Applications Group has developed a design
equal to the measured value at 10 mJ/cm2. for a solid-state laser that would meet all these specifi-
A NOVEL METHOD FOR REMOTELY DETECTING TRACE EXPLOSIVES

cations. It would not only be more robust, smaller, and cal clutter, extensive studies are required to verify a low
lower power than the commercial system, but it would false-alarm rate for a wide variety of substrates. Thus
also facilitate rapid area scanning and improve the sys- our continued efforts are twofold: testing the new equip-
tem’s signal-to-noise ratio (via a 0.5 nm laser bandwidth ment for standoff distance capabilities and evaluating
matched to the explosives’ signal bandwidth). We have the specificity of the technique with other materials and
identified the necessary crystal family and completed pre- substrates to eliminate or at least significantly reduce the
liminary fluorescence studies. These efforts should pro- false-alarm rate.
vide us with a stable high-power 236.2 nm laser source
capable of the desired high standoff distances. On the Acknowledgments
basis of our examination of some nitrate-bearing materi- We thank John Zayhowski for his efforts in design-
als (fertilizer, soil, and manure), which showed no evi- ing the next-generation laser necessary to continue
dence of a PD-LIF signal, we expect that there will be few this work in increasing the detection range. His
false alarms generated by other materials—i.e., we do not efforts stand out as a vital component for our future
expect nitrates to be a source of chemical clutter. research. This work was sponsored under the auspices
Although our initial work does not indicate the pres- of the Lincoln Laboratory National Technology and
ence of significant sources of false alarms due to chemi- Industrial Base Program. n
 Charles M. Wynn, Stephen Palmacci, Roderick R. Kunz, and Mordechai Rothschild

References About the Authors

1. R. Kunz, J.D. Hybl, J.D. Pitts, M. Switkes, R. Herzig-Marx, Charles M. Wynn is a technical staff
F.L. Leibowitz, L. Williams, and D.R. Worsnop, private com- member in the Submicrometer Technology
munication. Group, where he is working on laser-based
2. D.S. Moore, “Instrumentation for Trace Detection of High trace detection techniques. He joined
Explosives,” Rev. Sci. Instr., vol. 75, no. 8, 2004, pp. 2499– Lincoln Laboratory in 2000. He earned
2512. bachelor’s, master’s, and doctoral degrees
3. Kunz, private communication. in physics from the University of Connecti-
4. Moore, “Instrumentation for Trace Detection.” cut, Carnegie Mellon University, and Clark
5. T. Arusi-Parpar, D. Heflinger, and R. Lavi, “Photodissocia- University, respectively.
tion Followed by Laser-Induced Fluorescence at Atmo-
spheric Pressure and 20° C: A Unique Scheme for Remote Stephen Palmacci is an assistant staff
Detection of Explosives,” Appl. Opt., vol. 40, no. 36, 2001, pp. member in the Submicrometer Technology
6677–6681. Group. He has worked for more than 35
6. C.M. Wynn, S. Palmacci, R.R. Kunz, et al., “Experimental years on laser-based research at Spectra
Demonstration of Remote Optical Detection of Trace Explo- Physics Laser Analytics division and Lin-
sives,” Proc. SPIE, vol. 6954, 2008, 695407. coln Laboratory.
7. C.M. Wynn, S. Palmacci, R.R. Kunz, K. Clow, and M. Roth-
schild, “Detection of Condensed-Phase Explosives via Laser-
Induced Vaporization, Photodissociation and Resonant
Excitation,” Appl. Opt., vol. 37, no. 31, 2008, pp. 5767-5776. Roderick R. Kunz is a senior staff mem-
8. Wynn, “Experimental Demonstration.” ber in the Submicrometer Technology
9. Wynn, “Detection of Condensed-Phase Explosives.” Group. His current focus is on the analysis
10. J. Luque and D.R. Crosley, “LIFBASE: Database and Spectral and implementation of sensors for chemi-
Simulation Program,” SRI International Report MP 99-009 cal and explosives defense and threat
(1999). signature phenomenology. He received
11. J. Luque and D.R. Crosley, “Transition Probabilities and his bachelor’s degree in chemistry from
Electronic Transition Moments of the A 2Σ+–X 2Π and Rensselaer Polytechnic Institute and his
D 2Σ+–X 2Π Systems of Nitric Oxide,” J. Chem. Phys., vol. 111, doctoral degree in chemistry from the Uni-
no. 16, 1999, pp. 7405–7415. versity of North Carolina at Chapel Hill.
12. L. Nagli and M. Gaft, “Raman Scattering Spectroscopy
for Explosives Identification,” Proc. SPIE, vol. 6552, 2007, Mordechai Rothschild is the leader of
65520Z. the Submicrometer Technology Group. His
13. L. Nagli, M. Gaft, Y. Fleger, and M. Rosenbluh, “Absolute interests are in standoff chemical sensing,
Raman Cross-Section of Some Explosives: Trend to UV,” Opt. microelectronics, optical lithography, and
Mater., vol. 30, no. 11, 2008, pp. 1747–1754. nanophotonics. He received a bachelor’s
14. Nagli, “Raman Scattering Spectroscopy.” degree in physics from Bar-Ilan University
15. Nagli, “Absolute Raman Cross-Section.” in Israel and a doctoral degree in optics
from the University of Rochester.