See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6636166

High-sensitivity detection of TNT

Article in Proceedings of the National Academy of Sciences · January 2007

DOI: 10.1073/pnas.0609789104 · Source: PubMed

CITATIONS READS
215 2,481

6 authors, including:

Michael Pushkarsky Ilya Dunayevskiy

DRS Technologies Pranalytica Inc.
67 PUBLICATIONS 1,485 CITATIONS 17 PUBLICATIONS 833 CITATIONS

SEE PROFILE SEE PROFILE

Alexei Tsekoun Chandra Patel

Pranalytica Inc. University of California, Los Angeles
39 PUBLICATIONS 1,691 CITATIONS 261 PUBLICATIONS 9,353 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Chandra Patel on 05 February 2015.

The user has requested enhancement of the downloaded file.

High-sensitivity detection of TNT
Michael B. Pushkarsky*†, Ilya G. Dunayevskiy*, Manu Prasanna*, Alexei G. Tsekoun*, Rowel Go*,
and C. Kumar N. Patel*‡§
*Pranalytica, Inc., 1101 Colorado Avenue, Santa Monica, CA 90401; and ‡Department of Physics and Astronomy, University of California,
Los Angles, CA 90095

Contributed by C. Kumar N. Patel, November 6, 2006 (sent for review October 26, 2006)

We report high-sensitivity detection of 2,4,6-trinitrotoluene (TNT) by

using laser photoacoustic spectroscopy where the laser radiation is
obtained from a continuous-wave room temperature high-power
quantum cascade laser in an external grating cavity geometry. The
external grating cavity quantum cascade laser is continuously tunable
over ⬇400 nm around 7.3 ␮m and produces a maximum continuous-
wave power of ⬇200 mW. The IR spectroscopic signature of TNT is
sufficiently different from that of nitroglycerine so that unambiguous
detection of TNT without false positives from traces of nitroglycerine
is possible. We also report the results of spectroscopy of acetylene in
the 7.3-␮m region to demonstrate continuous tunability of the IR
source.

quantum cascade lasers 兩 high-power lasers 兩 continuous-wave

operation 兩 room temperature operation 兩 TNT detection

D etection of illegally transported explosives has become

important since the global rise in terrorism subsequent to
the events of September 11, 2001. Although not a choice of
Fig. 1. FTIR absorption spectrum of TNT.

suicide bombers, 2,4,6-trinitrotoluene (TNT) is a potent explo-

sive for which techniques for detection on a person’s body or in (6) using an optical parametric oscillator with a zinc-germanium-
one’s baggage is considered important for assuring safety of phosphide nonlinear crystal pumped by a flash-lamp pumped
airports and air travel. As with detection of other similar Q-switched erbium-chromium doped yttrium-scandium-gallium-
compounds, such as chemical warfare agents, any detection garnet laser at 2.8 ␮m.
scheme that claims to detect these targets must exhibit accept-
able receiver operational characteristic (ROC) that assures Results and Discussion
detection at very low levels without an unacceptable level of false IR Absorption of TNT. TNT exhibits a strong but broad absorption
alarms (1, 2). The molecular mass of TNT (C7H5N3O6) is almost feature at ⬇7.3 ␮m (Fig. 1), a region that is not accessible using
exactly identical to the molecular mass of nitroglycerine high-power lasers such as CO2 lasers used for high-sensitivity PA
(C3H5N3O9) even though the chemical compositions of the two detection of many gases including ammonia, SF6, and ethylene.
molecules are very different (TNT, 227.131 Da vs. nitroglycerine, L-PAS derives its high detection sensitivity from a combination
227.0872 Da). The nearly same molecular masses often lead to of factors that include remarkably low-noise PA detection
problems for unambiguous detection of TNT using techniques technique and high continuous wave (CW) power laser radia-
that rely on measuring the molecular mass of the species. On the tion. The ability to distinguish the target species from potential
other hand, the differences in the chemical structure between interferents is afforded by tunability of the laser because of
TNT and nitroglycerine lead to noticeably different infrared unique IR absorption fingerprints of each chemical compound
(IR) absorption signatures (3), making it possible to distinguish (1). In the region of 7.3 ␮m, we have recently developed a
between the two. However, the detection of TNT in vapor phase relatively high-power CW/RT QCL source that is tunable over
is hampered by its low vapor pressure of ⬇2 ⫻ 10⫺4 torr at 25°C. the spectral region where TNT has the absorption feature shown
In this work, we report on studies of detection of TNT by using in Fig. 1.
room-temperature (RT) quantum cascade laser (QCL)-based
photoacoustic spectroscopy (QCL-PAS). The high sensitivity Characteristics of the High-Power Continuously Tunable Laser. Fig. 2
afforded by laser-based photoacoustic spectroscopy (L-PAS) (4) shows the CW/RT laser output wavelength from the external
shows that the vapor-phase detection of TNT at an ambient grating cavity (EGC) QCL as a function of the grating tuning
temperature of ⬇25°C is possible. angle and representative Fourier transform IR (FTIR) spectra of
Previously, CO and CO2 lasers have been used for photo-
acoustic (PA) spectroscopic detection (3, 5) of vapors of explo-
Author contributions: C.K.N.P. designed research; M.B.P., I.G.D., M.P., A.G.T., and R.G.
sives. However, both of these laser sources are step tunable, and
performed research; M.B.P., I.G.D., and C.K.N.P. analyzed data; C.K.N.P. wrote the paper;
neither of the lasers is able to access the strong absorption and M.P. generated algorithms for continuous tuning of the external grating quantum
features of TNT that lie in the 6.0–7.5 ␮m region. Quantum cascade lasers.
cascade lasers (QCLs), with their continuous tunability, should The authors declare no conflict of interest.
be the right sources for the detection of TNT and other species Abbreviations: CDA, clean dry air; CW, continuous wave; EGC, external grating cavity; FTIR,
that do not absorb strongly in the 9–11.5 ␮m region. Finally, the Fourier transform IR; PA, photoacoustic; PAS, PA spectroscopy; L-PAS, laser-based PAS; QCL,
small size, simplicity, and potential long-term reliability of QCLs quantum cascade laser; RT, room temperature.
†Present
(as seen from the long-term reliability data on other semicon- address: Daylight Solutions, 13029 Danielson Street, Suite 203, Poway, CA 92064.
ductor lasers) could make QCL-PAS a more desirable technique §To whom correspondence should be sent at the
* address. E-mail: [email protected].
for the detection of TNT compared with that reported in a study © 2006 by The National Academy of Sciences of the USA

19630 –19634 兩 PNAS 兩 December 26, 2006 兩 vol. 103 兩 no. 52 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0609789104
Fig. 2. Tuning characteristics of CW/RT operation of the 7.3-␮m EGC QCL.

the output taken at seven grating angles. Continuous tunability

over ⬎350 nm has been demonstrated with highest single mode
CW/RT power of ⬎200 mW. We believe that with improved

APPLIED PHYSICAL
optics and optimized antireflection coatings on the grating side
of the QCL gain chip and partially transmitting coating on the

SCIENCES
output side of the QCL, the tunability can be extended over a
wider region with power outputs higher than those seen in Fig.
2. The single mode output as well as continuous tuning charac-
teristics of our computer-controlled CW/RT EGC QCL were
investigated at high resolution by using the FTIR spectrometer,
which confirmed mode hop free operation and single frequency Fig. 4. QCL-PAS measurements for acetylene as a function of acetylene
output with a linewidth of ⬍600 MHz, limited by the FTIR concentration.
resolution. Details are given in Materials and Methods.

Confirmation of Continuous Tuning Through PA Spectroscopy of spectral region. The very close similarity in shapes, locations, and
Acetylene. Additional confirmation of narrow single frequency intensities of the PA spectrum with the simulated spectrum con-
output and mode hop free tuning was obtained by carrying out firms that the CW/RT ECG QCL, whose wavelength is controlled
spectroscopy of acetylene (0 0 0 11 11 band) shown in Fig. 3. The by the computer algorithm that simultaneously adjusts the grating
upper trace shows a simulated acetylene spectrum using HITRAN angle, the QCL drive current, and the external cavity length, indeed
data (7) in the spectral region covering 7,250–7,450 nm. The lower produces mode hop free tunable power output over the range of
trace shows PA spectra of 10 ppm acetylene in clean dry air (CDA) wavelengths in Fig. 3.
at a total pressure of 300 torr, obtained by using the radiation from We have looked at one specific absorption peak of acetylene
the CW/RT ECG QCL that was tuned discreetly to 700 points in the from the spectrum shown in Fig. 3 to estimate the detectivity for

Fig. 3. High-resolution HITRAN simulated absorption spectrum of acetylene (Upper) and measured QCL-PAS spectrum of 10 ppm acetylene in CDA (Lower) at
a total pressure of 300 torr.

Pushkarsky et al. PNAS 兩 December 26, 2006 兩 vol. 103 兩 no. 52 兩 19631
Fig. 5. Measured QCL-PAS absorption spectra of TNT at four different Fig. 6. QCL-PAS spectrum of purified TNT sample at three different temper-
temperatures (see text for details). atures using the smart grid tuning algorithm (see text).

acetylene. Fig. 4 shows the L-PAS peak as a function of acetylene QCL-PAS spectrum consists of two distinct features, one cen-
concentration. From these data plotted in the top half of the tered at ⬇7,380 nm that matches the expected absorption feature
figure, we estimate a 1␴ detection capability of ⬇2.5 ppb for of TNT and the second centered at ⬇7,300 nm that arises from
acetylene. the yet unknown impurity in the commercial-grade TNT. The
unknown impurity was seen to be located on the surface of the
Detection of TNT. For exploring detection of TNT, whose absorp- TNT sample, and the 7,300-nm feature gradually disappeared as
tion spectrum (Fig. 1) shows a broad but strong absorption the TNT sample was kept at 100°C for 48 h while flushing the
feature in the same wavelength region where PA detection data sample with CDA.
of acetylene are obtained in Fig. 3, we provided a continuous Fig. 6 shows the measured L-PAS spectrum of ‘‘purified’’
flow of CDA over a sample of TNT,¶ and the emerging gas was sample of TNT vapor in a background of room air with relative
continuously analyzed by our PA spectrometer. The temperature humidity of ⬇40% at 25°C. The spectrum matches the expected
of the TNT sample could be controlled from RT to 60°C. The gas position and width well. The spectra were taken by using 300
transport lines from the TNT sample chamber to the PA cell discreet wavelengths determined by the computer using the
were maintained at 78°C and the PA cell was maintained at 60°C smart grid algorithm that skips the wavelengths corresponding to
to prevent condensation of TNT vapors either in the transfer the known strong absorption features of water vapor. We
lines on in the PA cell. The upper temperature limit was set by conjecture that the disappearing peak at 7,300 nm could be used
the PA cell microphone, whose sensitivity begins to degrade
significantly above 60°C but is not a limitation for future
operation of the cell at higher temperatures using appropriate
high-temperature microphones.
Fig. 5 shows an L-PAS spectrum obtained when the TNT
sample was kept at 24°C, 35°C, 45°C, and 55°C, respectively. The
PA spectrum qualitatively matches the shape of FTIR feature at
7,400 nm shown in Fig. 1. We comment on three specific aspects
of the PA spectrum. The first is that a number of sharp
absorption features arising from residual water vapor in the
system (as verified by using water vapor absorption spectra
obtained from HITRAN simulations) occur at certain wave-
lengths in the same region of wavelengths. These were avoided
by using a ‘‘smart grid’’ of laser wavelengths that skips these
specific wavelengths as the computer provides the tuning in-
structions to the EGC QCL.
The second is that the QCL-PAS spectrum is significantly
broader than would be expected by looking at Fig. 1. In fact, the

¶TNT samples (commercial grade) were obtained from Naval Air Warfare Center, Weapons Fig. 7. Calculated vapor pressure of TNT (using Clapeyron fit) vs. tempera-
Division (China Lake, CA). The pedigree of the samples is unknown. ture and the measured PA signal strength at various temperatures.

19632 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0609789104 Pushkarsky et al.

 at a level of 0.1 ppb (⬇1.01 pg cm⫺3) with a signal-to-noise ratio
(S/N) of 1 (i.e., TNT at temperatures as low as 5°C). It should be
noted, however, that the relationship of vapor pressure and
temperature is dependent on the Clapeyron fit to the measured
data.

Conclusions
The optical absorption shape information provided by QCL-PAS
should permit easy discrimination between TNT and many
interferents including nitroglycerine and other explosives as well
as a host of commonly available fertilizers because the optical
absorption reflects the chemical structure of the molecule rather
than simple parameters such as overall mass or shape. QCL-PAS
would provide high-sensitivity, low-false-positive detection of
TNT desirable for many security-related applications. The com-
pact size and relative simplicity of the operation of the QCL
source should make QCL-PAS a preferred technique for looking
for explosives in critical homeland security applications.
Fig. 8. Laser power vs. QCL drive current for the 7.3-␮m gain chip operated
in a Fabry–Perot geometry (the same chip used in the EGC geometry in these Materials and Methods
studies). High-Power CW Operating RT QCLs. The 7.3-␮m QCL epi material
was grown using molecular-beam epitaxy by the Center for
Quantum Devices (Northwestern University, Evanston, IL).

APPLIED PHYSICAL
in the future to tag the origin and age of the TNT sample for After cleaving, the 3-mm-long, 10.6-␮m-ridge-width chips were
forensic purposes. However, a confirmation of the conjecture

SCIENCES
mounted epi-side down on AlN substrates by using Au–Sn
will have to await getting samples of different age and origin eutectic solder (9). The chip carrier was maintained at 25°C by
from Naval Air Weapons Station China Lake (China Lake, CA). using a thermoelectric cooler (TEC). Heat was removed from
In either case, the shape and location information provides a the TEC by using a copper block maintained at 25°C. Operated
powerful tool for QCL-PAS to uniquely identify TNT and in a Fabry–Perot geometry, the 7.3-␮m QCLs (uncoated facets)
minimize effects of interference. generated multimode CW/RT power output of ⬎80 mW (laser
The third aspect of the measured spectra (in Figs. 5 and 6) that power vs. QCL drive current are shown in Fig. 8). The QCL
deserves mention is that the signal feature in the ⬇7,380 nm power output was analyzed by using an FTIR spectrometer and
absorption region grows rapidly as the TNT temperature is in- is shown in Fig. 9. At the highest drive current, producing
creased from 24°C to 55°C as would be expected form the tem- CW/RT power output of 80 mW, as expected we see a broad
perature dependence (ref. 8 and references therein) of the vapor spectrum of power output when the QCL is operating in a
pressure of TNT shown in Fig. 7. We have plotted the measured Fabry–Perot geometry.
QCL-PAS data for TNT on the same plot, anchoring the 50°C
QCL-PAS data on the vapor pressure vs. T plot. An acceptable Continuous Tuning of EGC QCLs. To produce a single frequency
correlation is seen between the vapor pressure data and the PA laser output that is conveniently tunable over the broad spectral
signal amplitude. region represented by the gain features of the QCL chip (and as
seen in Fig. 9), the QCL gain chips, with uncoated facets, were
Sensitivity of TNT Detection. From the lowest temperature (24°C) incorporated into an EGC setup with the total length of the
at which the PA spectrum is shown, we can estimate the detection external cavity of 4 cm with appropriate beam expanding and
sensitivity from known vapor pressure data for TNT. The vapor collimating 4-mm-diameter f/0.7 aspheric antireflection coated
pressure of TNT is seen to be (from Clapeyron fit curve) ⬇3 ppb ZnSe collimating lenses. The grating incidence angle could be
at 24°C (⬇30 pg cm⫺3). Comparing the L-PAS signal with the precisely and reproducibly adjusted by using a computer-
noise floor shown in Fig. 7, we estimate that we can detect TNT controlled linear actuator with an integrated encoder. The
grating position and therefore the length of the EGC were
accurately controlled by using another linear piezoelectric linear
actuator.
The principal differences between the QCL wavelength tuning
scheme described earlier (10) and the present work arise from
the much shorter external cavity length (25 vs. 100 cm) and the
center frequency of the gain curve of the QCL gain chip (7.3 vs.
6.3 ␮m). The shorter external cavity length and the different
refractive index of the gain chip compared with that used in ref.
10 require different numerical values of tuning constants that
were obtained through characterizing the output of the EGC-
QCL with a high-resolution FTIR spectrometer. Continuous,
mode hop free, real-time tuning was accomplished by a com-
puter-controlled algorithm that simultaneously adjusted the
grating angle, the external cavity length, and the QCL drive
current.
We obtained an overall tuning range of ⬇350 nm centered
around 7,350 nm (Fig. 2), with highest single frequency optical
power of nearly 200 mW. (Note that the CW/RT EGC QCL power
Fig. 9. FTIR analysis of the spectrum of the laser output (without the EGC output is nearly twice the CW/RT Fabry–Perot QCL power output
geometry) with the drive current of 1.35 A. shown in Fig. 8 because the Fabry–Perot output shows the output

Pushkarsky et al. PNAS 兩 December 26, 2006 兩 vol. 103 兩 no. 52 兩 19633
per each facet. The external grating reflection feedback to one facet near perfect match between the measured line positions and the
of the QCL in the external grating geometry improves the operating linewidths with those obtained from a HITRAN simulation pro-
characteristics of the QCL by reducing the total optical loss inside vides convincing proof that the computer-based algorithm produces
the laser cavity and provides all of the QCL output from one facet.) a mode hop free tuning even with uncoated facet QCL chip.
Continuous tuning was demonstrated by recording output spectra High-quality antireflection coating (11) on the external grating side
over the entire spectral tuning range by using a high-resolution of the chip would essentially eliminate many of the complexities of
FTIR. The FTIR spectra, in their highest resolution mode, provided the computer-based tuning algorithm and increase the tuning range
information about the linewidth of the EGC QCL, which was ⬍600 of this QCL gain chip.
MHz limited by the FTIR resolution. A convincing proof of
We thank the Naval Air Warfare Center, Weapons Division (China
continuous tuning over a significant range of the tuning range and Lake, CA) for providing the TNT samples used in these studies. This
maintenance of the very narrow output linewidth while tuning was work was supported in part through Defense Advanced Research
obtained by measuring a QCL-PAS spectrum of 10 ppm acetylene Projects Agency Contract HR0011-04-C-0102 (Approved for Public
in 300 torr of CDA as described earlier and shown in Fig. 3. The Release, Distribution Unlimited).

1. Webber ME, Pushkarsky MB, Patel CKN (2005) J Appl Phys 97:113101. Chance K, Clerbaux C, Dana V, Devi VM, et al. (2003) J Quant Spectrosc Radiat
2. Pushkarsky MB, Webber ME, Macdonald T, Patel CKN (2006) Appl Phys Lett Transfer 82:5–44.
88:044103. 8. Oxley JC, Smith JL, Shinde K, Moran J (2005) Propellants Explosives Pyro-
3. Claspy PC, Pao Y-H, Kwong S, Nodov E (1976) App Opt 15:1506–1509. technics 30:127–130.
9. Tsekoun A, Go R, Pushkarsky MB, Razeghi M, Patel CKN (2006) Proc Natl
4. Pushkarsky MB, Webber ME, Patel CKN (2003) Appl Phys B 77:381–385.
Acad Sci USA 103:4831–4835.
5. Crane RA (1978) Appl Opt 17:2097–2102. 10. Pushkarsky MB, Tsekoun A, Dunayevskiy IG, Go R, Patel CKN (2006) Proc
6. Todd MW, Provencal RA, Owano TG, Paldus BA, Kachanov A, Vodopyanov Natl Acad Sci USA 103:10846–10849.
KL, Hunter M, Coy SL, Steinfeld JI, Arnold JT (2002) Appl Phys B 75:367–376. 11. Maulini R, Mohan A, Giovannini M, Faist J, Gini E (2006) Appl Phys Lett
7. Rothman LS, Barbe A, Benner DC, Brown LR, Camy-Peyret C, Carleer MR, 88:201113-1–201113-3.

19634 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0609789104 Pushkarsky et al.

View publication stats