1. INTRODUCTION
The development of novel online and insitu methods for the classification of materials is still an important field of research in different disciplines, such as industrial process control and environmental diagnostics. In this context, characterization of explosives and housing materials for mine detection applications, but also for anti-terror purposes is still very critical. The UNO estimated worldwide between 60 and 100 millions of mines that have to be detected and then destroyed. Each year about 25000 people are involved in accidents caused by mines (Altmann, 2001). Special danger comes from mines that are housed by plastic materials and therefore cannot be detected by conventional geo-radar or metal detection methods.
Optical methods are very promising because they operate in non-contact, they are very sensitive, give real-time results, and in most cases, there is no need for sample preparation. When using miniaturized laser sources and fiber optics laser spectroscopy offers the possibility to develop compact laser sensors for various analytical applications. In this context, laser-induced breakdown spectroscopy (LIBS) is very interesting, because this method allows analyzing solid state samples in a real-time operation mode (Sattmann et al., 1998; Wainner et al., 2001; Lancaster et al., 1999; Bublitz et al., 2001). This is highlighted in this paper, for the characterization of explosives in different environments with special emphasis on the detection of hidden explosives, for example, improving conventional mine detection prodder by using the LIBS technique.
A nonmagnetic mine prodder is designed for manually locating anti-personnel mines that are buried at depths up to 0.25 m. The device consists of three main components, a handle, an extension tube, and the probe itself. Typical operational modes are kneeling or upright position of the operator while carefully digging in the soil. The prodder itself is a lightweight and robust tool; objects in the soil are tracked down by direct contact with the probe. The objective of this investigation is to improve the mine prodder by adoption of laser and fiber technology without much change in the conventional handling of the device.
LIBS are not only a tool for diagnostics but also for material processing. Therefore, laser drilling and online diagnostics of the material processed can be achieved if the laser power is sufficiently high as demonstrated recently in many cases (Fernandez et al., 2005; Trusso et al., 2005; Gamaly et al., 2005; Gavrilov et al., 2004; Mulser & Schneider, 2004).
The output power of a microchip-laser at 1064 nm can be amplified by a factor up to 100, when using an Ytterbium-fiber amplifier system, resulting in a peak power of several MW. Such power levels allow drilling a hole through the housing of a mine or a suitcase, and simultaneously analyzing the material processed, as well as the inner material by means of LIBS. In the case of energetic materials, this can be done without ignition of explosives when applying power levels below 400 mJ/pulse.
2. FIBER OPTIC LIBS: SENSOR
2.1. LIBS Basics
In general, the atomic composition of explosives, plastics, and soil is given by the following elements: carbon (C), nitrogen (N), oxygen (O), and hydrogen (H) (Fig. 1). When focusing an intense pulsed laser (λ = 1064 nm, Δt < 1 ns, Ep > 0.1 mJ) on the surface to be investigated, plasma will be generated instantaneously. Surface molecules are fragmented into their atomic compounds and excited atoms, molecules and molecular compounds, as well as electrons generated by the plasma. When the laser pulse is off, no energy is transferred to the sample, and the ions and electrons recombine as well as the excited states, which results in the emission of light. In Figure 2, the temporal evolution of laser-induced plasma is shown schematically. Spectrally-resolved detection of the plasma emission gives information on the atomic composition of the sample (“first order fingerprint”). This can be used for characterization of different samples, if the atomic composition is different from the sample being sampled. Plastics and explosives are composed of the same elements, and therefore, cannot be classified by only measuring the LIBS intensities. However, when using the temporal evolution of the LIBS intensities for specific wavelengths, a data analysis can be developed that allows a selective sample analysis. Depending on the absorption cross-section for the materials under investigation, the laser-induced plasma temperature is different for different materials. Plasma temperatures up to 12000 K are obtained when using laser pulses of a few mJ. This effect results in characteristic LIBS intensity decay times for different materials. The decay times are typically in the order of several tens to a few hundreds of nanoseconds (“second order fingerprint”) depending on the applied laser power and the material to be analyzed.
Atomic composition of two explosives (a) TNT (trinitrotoluene) and (b) HNS (hexanitrostilbene).
Time-evolution of LIBS.
To investigate these effects in detail, the fundamental output of a diode laser pumped and passively Q-switched Cr4+:Nd3+:YAG microchip-laser (Zayhowski, 1996; Voss et al., 2001) (pulse duration Δt ≤ 1 ns, pulse repetition rate ν = 200 Hz, pulse energy Ep = 0.4 mJ, emission wavelength λ = 1064 nm (Alphalas GmbH, Göttingen, Germany)) is focused on the surface of the sample under investigation (diameter d ≈ 200 μm). Then plasma is generated. The LIBS emission is recorded by a gateable optical multichannel analyzer (OMA-4 system, EG&G, time-resolution > 5 ns). The plasma emission is measured at 388 and 248 nm, which is attributed to cyanide (CN) and atomic carbon. Temporal- and spectrally-resolved LIBS spectra are shown in Figure 3 for trinitrotoluene (TNT) (a), for the plastic polyamide (b), as well as the LIBS intensity decay for the CN-line at 388 nm for TNT (c), and polyamide (d). The strong carbon emission is because of the relatively high amount of carbon in these materials, but the CN emission is explained by a reaction of the carbon emitted from the material and the nitrogen from the ambient air, which is also ionized by the plasma. If the experiments are performed under vacuum conditions, the C-emission is still present, but the CN-emission at 388 nm vanishes. Also the polyamide spectrum (Fig. 3c) shows calcium (CA) contaminations of the plastic material around 390 nm that easily can be detected simultaneously by the LIBS technique.
Time- and spectrally resolved LIBS intensity spectra of TNT (a) and polyamide (b). The decay of the CN-intensity at 388 nm is analyzed for TNT (c) and polyamide (d). When fitting single exponentials to the data points decay times τ = 136 ns and τ = 81.5 ns are obtained for trinitrotoluene (TNT) and polyamide (PA). The laser excitation was at 1064 nm and the pulse power about 5 mJ.
In conclusion, time-resolved recording of LIBS intensities for the C- or the CN-plasma emission after excitation with short laser pulses (Δt < 1 ns), allows the classification of different explosives and plastics even though the atomic composition of the material is the same. In the case of explosives, no ignition of the material is obtained for pulse energies up to 400 mJ. In Figure 3b and Figure 3d, the estimated 1/e plasma decay times for the CN-emission (388 nm) are shown for TNT and the plastic material (PA), respectively. In a first order approximation, the plasma intensity decay can be described by a single exponential. Therefore, single exponentials are fitted to the data points resulting in decay times τ = 136 ns and τ = 81.5 ns for TNT and PA, respectively. In Table 1, LIBS decay times for laser excitation at 1064 nm and observation wavelengths 248 nm (CI) and 388 nm (CN) are summarized for different materials. When storing such data on a microcomputer, they can be used for online and insitu characterization of samples by comparing them with real time LIBS measurements (Romano et al., 2004; Schade et al., 2004).
Summary of LIBS 1/e decay times for different materials and fixed experimental conditions
2.2. Miniaturized Laser for LIBS Applications
The development of smart, portable, and rugged LIBS fiber optic sensors, for example, upgrade for a conventional mine prodder or as a sensor for the detection of hidden explosives in anti-terror purposes, requires the supply of a powerful and miniaturized laser source. This is done in this investigation by applying a passively Q-switched Cr4+:Nd3+:YAG laser as a seed laser for a fiber amplifier. The Cr-ions in the laser crystal, act as a passive Q-switch. Therefore, when pumping the Cr4+:Nd3+:YAG laser crystal (Stankov, J. 2002, private communication) with a continuous wave (cw) diode laser (λ = 808 nm, P = 2 W), the microchip-laser emits pulses below 700 ps in duration, a repetition rate of up to 20 kHz, and a pulse energy of up to 25 μJ. In Figure 4, the set-up for a microchip-laser as used in this investigation (a) together with the temporal profile of a single pulse (b), the change of the repetition rate dependence on the laser current (c) and the pulse to pulse stability (d) are shown. For amplification of the microchip-laser output, an Ytterbium fiber amplifier (Höfer et al., 2001; Limpert et al., 2001), which is also pumped by a high power cw diode laser (P = 50 W, λ = 976 nm) is used (Fig. 5a). The microchip-laser operates as a seed-laser and pulse amplification with a factor of up to 100 is possible (Geiser et al., 2004). This gives a peak power in the order of a few MW at the laser wavelength 1064 nm, which allows drilling holes into the housing of plastic mines (Fig. 5b), and synchronously monitoring the material characteristics by time-resolved recording of LIBS intensities.
(a) Schematic and photo of a passively Q-switched Cr4+:Nd3+:YAG microchip laser, (b) temporal pulse shape, (c) variation of repetition rate, and (d) pulse to pulse stability.
(a) Yb-fiber amplifier with Cr3+Nd4+:YAG microchip laser as seed-laser. (b) Drilling holes into the casing of a plastic mine.
3. LIBS CLASSIFICATION OF MATERIALS
With respect to practical applications as a fiber optic laser sensor, for example, for mine detection, the time-resolved LIBS method is simplified, such as LIBS intensities in two time windows, with respect to the laser pulse recorded. Then, the ratio of these intensities is calculated and used as a measure for the classification of the material to be investigated. The first time window opens directly after the laser pulse for a typical duration of 50 ns (“early LIBS intensity”), while the second is activated with a delay of 50 ns with respect to the laser pulse for a duration of 200 ns (“late LIBS intensity”). For fixed excitation conditions (diameter laser spot on surface, pulse power), the ratio of “late” to “early” LIBS intensities for C- or CN-emission is a characteristic measure for different materials. This is shown in Figure 6 when recording the CN-emission and different materials.
(a) CN-emission of the explosive TNT. (b) Time-integrated ratios of “early” and “late” LIBS intensities (“area ratio”) for different materials and laser excitation at 1064 nm.
In Figure 6a, the spectrum of the CN emission is shown for the explosive TNT in the spectral range of 350 to 440 nm with an intensity maximum at 388 nm. Figure 6b shows the measured ratios of “late” to “early” LIBS intensities (“area ratio”) for two explosives TNT and HNS and polyamide. The results show that different materials can be classified by time-integrated LIBS, even though the material composition consists of the same elements. The accuracy of the classification strongly depends on the experimental parameters, especially the temporal setting of the integration windows. These parameters need to be determined accurately in laboratory experiments to optimize the sensitivity of the method, and then stored in a database. A set-up as used in the first practical applications for the characterization of explosives and mine casings is shown in Figure 7a. A photo of a laser induced plasma directly in front of a laser mine prodder on a TNT sample is shown in Figure 7b. The accuracy of the fiber optic laser prodder is optimized when using self-learning neuronal networks for real-time data analysis. Then the intensity of the LIBS signal, the time evolution, and the “area ratio” are used for classification. As an example, the system acquires this data by analyzing the time-evolution of CN-plasma emission for one sample under investigation, this information is stored in a 10 × 10 matrix, for example, on position 2 × 10 (Fig. 7c). Then the next sample is analyzed by the same procedure. The system compares the result with all previous measurements, if it is identical, it is stored on the same matrix element, if not, and a new element is generated. In such a way, 100 different samples can be classified when using a 10 × 10 matrix as discussed in this paper. When calibrating the system with reference data from well known mine casing materials, for example, the matrix element 2 × 10 can be identified as mine casing material of type y.
(a) Configuration of a LIBS-based mine detection needle for online and insitu classification of explosives. (b) Photo of a laser induced plasma on TNT directly in front of the laser prodder. (c) Map of neurons for material classification when using self learning neuronal networks for real-time data analysis.
By this miniaturization of lasers and the application of fiber optics, the LIBS method can easily be integrated into a conventional mine prodder to improve the conventional mine searching process, but the same method can also be applied for the detection and verification of explosives in hidden environments, such as suitcases or any type of luggage, which will be interesting for fighting terror attacks, for example, at airports or any public location.
4. SUMMARY
The results show that samples with same atomic composition (for example, C, N, O, and H), can be classified with sufficiently high accuracy by recording LIBS intensities for CN- or C-emission in two different time windows, with respect to the exciting laser pulse followed by real-time data analysis, using the self-learning neuronal networks. The practical capability of the method is demonstrated as different explosives such as TNT, HNS, H5, or tetryl, but also plastic materials such as PVC or PA are analyzed. Beside classification of materials, contaminations in materials can be monitored by this technique. An example is the detection of CA contaminations in explosives, for example, in HNS but also in PA. Here CA is introduced during material processing. Therefore, the detection of impurities in materials may give important hints from the history of the material itself. Mine casing materials are investigated and can be classified by the same method.
The use of miniaturized and powerful lasers for the generation of plasma and fiber optics offer new and very interesting fields of applications: The integration of a LIBS sensor into a conventional mine prodder or the online and insitu detection of hidden explosives. In both cases, real-time information is achieved and the analysis is performed in non-contact modus. This will be very important in order to improve safety aspects and the speed for analysis for clearing mines.
ACKNOWLEDGMENT
Financial support by the German Ministry of Defense under contract No. 005/B-3D023/X5165 is gratefully acknowledged.
