1. Introduction

The threat of homemade explosives (HMEs) using materials such as urea nitrate (UN), ammonium nitrate (AN) and the accelerant nitromethane (NM) is a matter of growing concern [1,2]. A technique capable of rapid, wide area-scanning for trace explosives is highly desirable, as it facilitates early detection. Despite much research, however, such a system currently does not exist [3]. To be useful beyond the laboratory, a technique must display high sensitivity, immunity to false positives, ease of use, and detection range greater than ~1 meter (to facilitate rapid area search). A promising laser-based method for remote detection that meets many of these requirements uses photodissociation followed by laser-induced fluorescence (PD-LIF) [4,5]. We have successfully used this technique on military explosives [6], and have shown its potential to detect trace quantities utilizing a single laser to achieve detection with a single laser pulse.

Since most explosives have exceedingly low vapor pressures [7], it is desirable for real-world applications that PD-LIF is used to detect traces of the condensed-phase materials themselves and not their ambient vapors. The photodissociation of NM has been well studied; however, the focus has been primarily on its gas phase [8]. Much less is known about the photodissociation of AN and UN. In this work, we use PD-LIF to probe condensed-phase HMEs in an ambient environment. We vaporize the condensed-phase samples with the same tunable ultraviolet (UV) laser used for PD-LIF. Vaporization/PD-LIF, requires the near-simultaneous occurrence of the following steps: (1) vaporization of the explosives molecules; (2) photo-induced dissociation of the explosives molecules, generating NO photofragments; (3) resonant excitation of the NO fragments; (4) fluorescence from the excited state of the NO fragments. All four steps occur within or – in the case of step (4) – also immediately following a single laser pulse.

The most promising version of PD-LIF requires the LIF (step 4 above) to be at shorter wavelengths than the laser, existing in a spectral regime where few other signals occur. Thus, a very low background signal and immunity to false positives can be achieved via spectral filtering of light from unwanted sources including scattered laser light, red-shifted fluorescence from non-explosive materials, and signals from atmospheric NO, which exists predominantly in its vibrational ground state. The blue-shifted signal relies on the photogeneration of vibrationally-excited NO fragments (step 2 above), preferably via the same laser wavelength (λ) used for LIF. Vibrationally-excited NO fragments are critical to this detection technique, as they allow the resonant photoexcitation to occur at a λ longer than the fluorescence emission.

Limiting ourselves to the excitation of the first electronically excited state A ²Σ⁺, (v’ = 0 or 1) and wavelengths in the 200-250-nm region, several combinations of excitations (exc) and fluorescence (fluor) are possible. Most of the relevant Frank-Condon factors are similar, within a factor of 4, with the exception of the (v’, v”) = (1, 2) transition, which is ~100x smaller [9,10], thus eliminating the possibility of either exciting or observing fluorescence due to this transition. The remaining set of combinations is as follows [9,10] (see Fig. 1 ):

 

Fig. 1 Photodissociated NO energy levels for the vaporization/PD-LIF detection process. NO fragments are optically excited from their vibrationally-excited electronic ground state to an electronically excited state. The blue-shifted fluorescence is used for detection. Several excitation-emission processes are shown.

Download Full Size | PDF

Process (a) above has been the basis for the PD-LIF detection method described in Reference [6] for military grade explosives. Process (b) probes the same v” = 1 level as process (a), but has certain practical drawbacks and has not been pursued. Process (c) detects NO fragments in their v” = 2 level, and has been reported in Reference [4,5] for TNT vapor.

In this paper we seek to expand the previous results and explore the feasibility of PD-LIF to the HME materials AN, UN, and NM. To this end, we performed high-resolution excitation-emission studies of these materials and of the reference material 2,6-dintrotroloune (DNT). The blue-shifted PD-LIF relies on a high vibrational temperature of the NO fragments, i.e., on a significant population of v” = 1, 2, or 3.

2. Experimental

Solid DNT crystals (from Aldrich) were melted into films. AN powder and NM liquid were used as received from Kinepak. UN (from TCI America, Inc.) was dehydrated and pressed into solid pellets. All PD-LIF measurements used bulk quantities of the above-mentioned materials to obtain detailed spectra. Prior work [6] with trace TNT has demonstrated the utility of the technique for trace detection using dropcast samples of ~1 μg/cm². Additionally, the amount of material dissociated per pulse and factors, such as laser fluence, that influence that dissociation were studied. The HMEs examined in this paper exhibited dependencies generally similar to TNT.

The UV absorption of these compounds is important to the efficiency of PD-LIF. We measured the absorption of UN by forming a thin film of UN (4.5 mg mass formed via dropcasting from acetone solution) on a fused silica substrate and placing it in a UV-visible spectrophotometer (Hitachi U-4001). The measured absorption cross section, σ, at 236 nm was 4.5 x 10⁻²⁰ cm²/molec. The absorption cross sections of the other compounds were obtained from the literature: σ(236 nm) of DNT in solution [6], NM [11], and AN [12] are 5 x 10⁻¹⁷, 6 x 10⁻²⁰ and 1 x 10⁻¹⁸ cm²/molec respectively. Therefore, the newly studied compounds AN, UN, and NM are orders of magnitude more transparent than DNT.

Two types of PD-LIF spectral measurements were performed: excitation and emission. In both, the photodissociation and the excitation of NO were performed with the same laser, but the experimental configuration and the photodetection were somewhat different. The laser was a pulsed solid state laser equipped with an optical parametric oscillator and a frequency doubling module (Continuum 9030 Powerlite II with Panther EX). The pulse duration was ~7 ns full width at half maximum, the pulse repetition rate was 30 pps, and the energy was ~2 mJ/pulse with a beam size of ~1 cm². The spectral linewidth in the deep UV was 0.03 nm. In what follows, all λ‘s reported are in air, and all measurements were performed under ambient temperature and pressure conditions.

When obtaining excitation spectra, the laser beam was focused into a ~0.1-cm² spot at the sample, at ~60° from normal, with the fluence being 10 mJ/cm²/pulse. The photodetector was positioned ~10 cm directly above the sample, with no collecting optics. It consisted of a Cs-Te solar blind photomultiplier tube (PMT), Perkin-Elmer MD 1922, with negligible dark counts, single-photon sensitivity, and quantum efficiency of 10%, and custom optical filters that were placed in front of the photomultiplier tube. The filters were designed to transmit radiation near ~225 nm (including the 226 nm and 224 nm fluorescence) with a ~0.3% efficiency (optical density, OD 2.5), while rejecting the excitation wavelengths at 236 nm or longer (OD > 11). The fluorescence data consisted of 90-pulse averages. This setup was similar to that previously used to demonstrate PD-LIF detection of military grade explosives [6].

In obtaining emission spectra, the laser was focused to a 2 mm x 2 mm (0.04 cm²) spot on the sample, with fluence of 25-30 mJ/cm²/pulse (1.0-1.2 mJ pulses). The fluorescence was collected ~15 cm from the sample via a curved f/4 mirror focused onto the entrance slit of a dual-grating monochromator (Horiba Jobin-Yvon Gemini 180) with 2400 grooves/mm gratings. An interference filter strongly rejecting laser light near 236 nm while passing light near 226 nm (40% transmission) was placed at the entrance slit of the monochromator for some materials. In these cases, the reported signal has been adjusted by 2.5x for comparison with non-filter data. Photodetection occurred via a Cs-Te PMT (Hamamatsu R7154), connected to a 5 MHz pre-amplifier and boxcar averager (Stanford SRS 250), which used a 250 ns gate window centered about the incident laser pulse. The fluorescence data are 900-pulse averages.

3. Results and discussion

The excitation spectra in Fig. 2 show clear evidence of vibrationally and rotationally hot NO photodissociated from all four compounds NM, UN, AN and DNT. For all our compounds, the (v” = 1)→(v’ = 0) excitation (236 nm) has the strongest signal. The (v” = 2)→(v’ = 0) excitation (247 nm) is also evident though considerably weaker, especially in UN. NM has the strongest signal overall. NM also displays a clean signal when the excitation is at ~243 nm. We attribute this to excitation of the (v” = 3)→(v’ = 1) transition, followed by fluorescence at 224 nm; the (v’ = 1)→(v” = 1) emission should be slightly offset (224 nm; see process d3 Fig. 1) from the other emission at 226 nm, from v’ = 0, though still within the filter passband. Emission data in Fig. 3 below confirm this shift. The (v” = 3) → (v’ = 0) excitation at 258.5 nm is also evident in NM, though not displayed in Fig. 2.

 

Fig. 2 PD-LIF excitation spectra: blue-shifted fluorescence as a function of excitation λ. Insets are close-ups of (v” = 1) excitation. Symbols are experimental data; lines are simulated NO fluorescence with the following parameters (and uncertainties): (a) NM: (v” = 1)→(v’ = 0), T_Rot = 1000 K ± 500 K (near 236 nm) ; (v” = 2)→(v’ = 0), T_Rot = 500 K ± 200 K (near 247 nm) ; (v” = 3)→(v’ = 1), T_Rot = 1000 K ± 500 K (near 243 nm); (b) UN: (v” = 1)→(v’ = 0), T_Rot = 300 K ± 100 K; (c) AN: (v” = 1)→(v’ = 0), T_Rot = 1000 K ± 500 K; (v” = 2)→(v’ = 0), T_Rot = 1000 K ± 500 K; (d) DNT: (v” = 1)→(v’ = 0), T_Rot = 1000 K ± 500 K; (v” = 2)→(v’ = 0), T_Rot = 2000 K ± 700 K.

Download Full Size | PDF

 

Fig. 3 PD-LIF emission spectra for various HMEs and DNT. Symbols are experimental data. Spectral resolution was 0.53 nm, except for 236 nm excitation of NM (0.25 nm) and 243 nm excitation of UN (0.25 nm) and DNT (0.74 nm). NM data in a) have been offset for clarity. Lines are simulated NO emission with T_Rot = 350 K. In a) lines correspond to (v’ = 0)→(v” = 0) transition; in b) lines correspond to (v’ = 1)→(v” = 0,1) transitions.

Download Full Size | PDF

For any given excitation spectrum, the relative magnitudes of the peaks at 236, 247, and 243 nm are indicative of the relative populations of the v” = 1, 2, and 3, respectively, of the NO photofragments. It is evident that the four energetic compounds are photodissociated in different ways. The NO fragments of AN and DNT have large populations in v” = 2, 30%-50% of the v” = 1 population. The NO fragment from UN is vibrationally the coldest; while NM dissociates into vibrationally very hot NO with a remarkably high population in v” = 3. The high vibrational temperatures of NO fragments are qualitatively consistent with similar observations in the photodissociation of TNT vapor [4,5] and RDX vapor [13,14], where populations have been observed up to at least v” = 2 and v” = 3, respectively.

A comparison of the signal magnitudes across the spectra in Fig. 2 reveals large variations, of almost two orders of magnitude, in the yield of vibrationally excited NO, with AN having the lowest value and NM the highest. This ranking does not match the ordering of the molecular absorption cross sections listed above. In particular, UN and NM are nearly equally transmissive at 236 nm, and DNT is by far the most absorptive, while the signal size in the excitation spectra of UN and DNT is nearly the same. We conclude that the magnitude of the photo-absorption is not a good predictor of the PD-LIF signal strength under our experimental conditions. One source for this apparent discrepancy may be the relatively high thickness of the samples in our experiments, which is greater than the penetration depth of even the more transparent UN, ~30 μm. When the sample thickness is appreciably higher than the photon penetration depth, most of the incident radiation is absorbed irrespective of the molecular absorption cross section. Still, the differences in the PD-LIF signal strength and spectrum across all the materials examined here is remarkable. These differences may be attributed to at least two factors related to the photochemistry of the compounds: differences in the photo-dissociation pathways leading to different yields of NO fragments; and, keeping in mind that our PD-LIF experiments do not sense NO in v” = 0, even if the NO yield is the same, the dissociation pathways leading to vastly different vibrational temperatures.

Overlaid on the experimental data in Fig. 2 are simulated excitation spectra of NO, obtained using the LIFBASE software package [9,10] assuming a linewidth of 0.03 nm (laser bandwidth). Simulated spectra assume (v” = 1, 2)→(v’ = 0) excitation (236 and 247 nm respectively), and also (v” = 3)→(v’ = 1) excitation (243 nm). All data match the NO simulations very well, showing significant NO rotational energy. The effective rotational temperatures, T_Rot, (with uncertainty estimates) of the various vibrational states of X ²Π are listed in Fig. 2 and are all well above room temperature. A recent study [8] of the gas-phase dissociation of NM followed by (v” = 0)→(v’ = 0) excitation of the resultant NO observed T_Rot = 600 K, similar to our condensed phase results of 1000 K, 500 K, and 1000 K for v” = 1, 2, 3 excitations respectively. Another recent gas-phase study [13,14] of nitramine military explosives observed NO with T_Rot = 20 K. This is in contrast to our results for the military explosive DNT and the HMEs. The differences may be due to differences in the structure of the parent compounds or differences caused by the condensed-phase dissociation process. A more detailed analysis of the possible causes leading to differences in T_rot can be found in Reference [6].

Spectral characterization of the NO fluorescence from A ²Σ is shown in the PD-LIF emission spectra of Fig. 3. Excitation at 236.2 nm and 247 nm (Fig. 3(a)) populates the v’ = 0 state, yielding (v’ = 0)→(v” = 0) emission at 226 nm for all compounds except AN at 247 nm excitation. Excitation at 243 nm, (v” = 3)→(v’ = 1), yields two distinct peaks corresponding to the (v’ = 1)→(v” = 0,1) transitions at 215 and 224 nm for all compounds except AN. The transition probabilities [9,10] for these two transitions predict a 215 nm:224 nm signal ratio of ~3:1. We observe a very different ratio in which the 215 nm signal appears to be suppressed. This may be due to self absorption by nearby ground-state (v” = 0) NO fragments. A similar suppression of emission to the ground state is apparent in other ratios, including the emission following (v” = 3)→(v’ = 0) excitation of NM at 259 nm (not shown). Given [15] σ_ΝΟ = 1x10⁻¹⁸ cm²/molec, ~1 μm of material must be photolyzed and form NO in order to achieve an absorption in excess of 50%, which is feasible at these λ’s. In all cases, the simulated NO spectra matched the data very well for T_Rot = 350 K, except NM at 224 nm for which the apparent spectral broadening, implying higher T_rot, may have been due to unfiltered laser leakage. Unlike in the excitation spectra, which probe the population distribution of the nascent v” states (modulated by collisional effects before photoexcitation), the emission spectra of Fig. 3 reflect the population distribution of the v’ states being pumped by the laser. Noting our laser bandwidth is 0.03 nm, one would expect a similarly narrow range of rotational states to be populated upon LIF. The observed T_Rot of the populated A ²Σ state is a convolution of the narrow range of rotational states excited by the laser (0.03 nm bandwidth corresponds to T_Rot ~1 K) and any collisional effects that broaden this range on a time scale shorter than the duration of the fluorescence. The collisional thermalization of rotational states of NO (X, v” = 2) is quite efficient, with a rate constant of ~4 x 10⁻¹⁰ cm³molec⁻¹s⁻¹ [16]. Assuming similar rate constants for the NO (A, v’) states, the collisional thermalization of the rotationally selected excited molecules takes places within less than 1 ns at atmospheric pressures, in qualitative agreement with our observation.

The photochemical phenomenology presented in Figs. 2 and 3 strongly indicates the potential of noncontact detection of the homemade explosive constituents AN, UN, and NM using PD-LIF at several wavelengths in the UV. Figure 4 summarizes these results, comparing the relative strengths of the PD-LIF signal for the homemade explosives and for the military grade DNT and TNT (the latter derived from Reference 4 and related unpublished data). This Figure indicates the relative ease of the eventual detection of the respective compounds, as well as the preferred excitation and emission wavelengths, or their combinations. While a claim of full detection will require extensive statistical analysis of the PD-LIF signal obtained under different conditions and in the presence of potential interferents, we note that numerous non-explosive substrates were also examined including silicon, silica, and aluminum and they all showed no blue-shifted photons. Thus it appears that the probability of false alarm when searching for these explosives on the listed substrates is negligible. Taken in aggregate, the data presented here can serve as the basis for the design of systems engineered to remotely detect trace explosives of interest. Given that the UN and NM signal levels are similar to that of DNT, we expect (based on prior work [17] with DNT) that detection ranges can exceed 10 m.

 

Fig. 4 Comparison of PD-LIF signal strengths for various explosives. Data are derived from excitation spectra at 10 mJ/cm² fluence. HMEs are noted in red.

Download Full Size | PDF

4. Conclusions

In summary, laser-based detection of several condensed-phase HME constituents was demonstrated under ambient conditions using PD-LIF. Since detection can be achieved within a single pulse, a detection system with a suitable high-repetition laser could rapidly scan a large area for dangerous explosives. Furthermore, since the relative populations of the v” = 1, 2, 3 states differ amongst the compounds, the PD-LIF emission spectrum may be useful in distinguishing among them.