1 Introduction

Despite intensive research efforts, the detection and identification of hazardous materials, like explosives and chemical or biological agents [1, 2], as well as other compounds, are still a persisting challenge. Hence, advancement of optical methods, particularly laser-based methods [3–10], which can detect explosive residues on different surfaces with high selectivity and sensitivity, is of great interest to homeland security and forensic science. Particularly, development of abilities providing specific detection of traces, or bulk quantities of explosives in short times, at point, proximal and standoff distances [11], which can be used under different scenarios, are required.

Indeed, a variety of approaches have been pursued during the last decades [1–9], including vibrational spectroscopies, which are considered as most appealing means. For instance, Raman spectroscopy [5, 11–25] and its coherent variants, coherent anti-Stokes Raman scattering (CARS) [26–32] and stimulated Raman scattering (SRS) [33] have been used for explosives detection over large or short distances. These methods, based on spontaneous or coherent Raman scattering, provide relatively sharp spectral features in a broad spectral range, leading to signatures that are strongly related to specific compounds and to their spatial distribution, and ultimately allowing explosives identification, even on strongly interfering surfaces.

While standoff detection was mainly used for identification of bulk quantities of explosives or of single large particles, positioned on distant targets, the point and proximal detection focused on uncovering the presence of single particles or of particles residing from fingerprints. Detection of these residues on a variety of substrates (glass, metals and clothing) was achieved using integrated tabletop microscope-based Raman systems [20–22] or a homebuilt compact system [23–25]. These approaches were used not only for measurement of Raman spectra, but also for Raman imaging of explosives residues. Imaging was achieved by point-mapping, involving laser spot or sample raster-scanning through the x and y spatial dimensions, and resulting in successive Raman spectra at each particular x, y position. The drawback of this approach is that it is extremely time-consuming.

Consequently, obtaining even moderate-size Raman images, while operating the system with short integration times for each measurement point, can take several hours [23]. For instance, mapping of a 1 × 1 cm sample of potassium nitrate (KNO3) particles, with 50 μm steps, by the compact Raman spectrometer took 8 h [24]. Essentially, by photographing the sample, prior to data acquisition, it was possible to selectively monitor spatially resolved particles and to map only them at reduced sampling times [24, 25]. The photo-guided sampling allowed a 15-fold reduction in acquisition time compared to raster-scanning and thus significant improvement in system performance. It is obvious that this reduction factor can change, depending on the size of the individual mapped particles and on the extent of surface coverage.

Nevertheless, in addition to spot illumination that is used to obtain Raman point-maps, other methods including line or wide-field excitation, introduced by Delhaye and Dhamelincourt [34], can be used to acquire line-maps and direct global images, respectively [35]. Line-mapping is considered to be an extension of point-mapping, where the laser beam is focused as a line and the sample can be raster-scanned along either the x- or y-axis and the spectral and spatial information can be collected by a two-dimensional (2D) charge-coupled device (CCD) detector. In fact, the spatial data of the Raman signal from the sample are consisted in a line that is projected parallel to the entrance slit of the spectrometer, and the spectral information is dispersed perpendicularly. By scanning the sample perpendicularly to the focused line, the second spatial dimension is monitored on the CCD. When wide-field sample illumination is used, the Raman signal is collected and projected on the CCD, while using filters for wavelength selection in front of it. This implies that from a single measurement of a particular field of view of a sample, illuminated by an incident laser, the spatial points of the Raman image at a specific wavenumber can be monitored simultaneously [35].

Actually, some studies discussed the advantages of these imaging methods, while emphasizing theoretically the achievable signal-to-noise ratios (SNRs) [36] and the effect of the power density [37]. In addition, this approach was also used in experiments that were performed in the past. Bowden et al. [38] showed that it is possible to use a 2D CCD detector, combined with a modified Raman microscope to obtain sequences of Raman spectra from a ca. 100 μm long line on a sample of mixed TiO2 powders and the components of a zircon mineral. Other studies used this technique for imaging of different biological samples, including fingerprints [39], bacterial spores [40] and milk powder [41], as well as 1-μm polystyrene beads [42] and geological samples [43].

Of particular importance is the comparison of different illumination methods that was performed on well-defined test samples, composed of 5 μm squares of silicon [35]. This study showed that line-excitation is the fastest approach for acquiring Raman images. This method also has the advantage that it provides higher spectral resolution and the possibility to monitor several characteristic Raman features in a sample, unlike the wide-field imaging which might be suitable only when a specific band is needed. It should be pointed out that some of these experimental measurements were carried out by the use of a tabletop Ti:Sapphire laser [40] or large commercial imaging microscope systems [35, 38], limiting the use of this approach to measurements carried out in a laboratory.

In this article, we intend to examine and demonstrate the potential of our compact system to successfully measure Raman spectra and Raman maps of explosive traces, while utilizing a configuration that uses a narrow line-shaped laser source for excitation. To the best of our knowledge, this mode of operation was not evaluated yet for detection and imaging of this kind of particles, which usually have various shapes and sizes. It is anticipated that realization of this operation mode might lead to improved performance, with minimal changes in the configuration of the compact system. Our results demonstrate that this approach shortens the acquisition time extensively by a factor of about twenty relative to point-excitation, and can therefore be considered as improved means for detection and imaging of explosives and other compounds.

2 Experiment

The compact Raman spectrometer for line-excitation measurements, schematically shown in Fig. 1, is mainly based on the previous setup [24, 25], with the addition of a cylindrical lens (CL), as well as the removal of a confocal hole, at the spectrometer slit entrance. Briefly, the system consisted of a handheld green laser at 532 nm, driven by a dc power supply for long-term operation and intensity control. For all measurements, a 35-mW laser beam excited the sample. The beam was turned by a dichroic mirror (DM) and passed though the CL, 40 mm focal length (f.l.), to narrow the beam in the x direction. This beam was then turned by a 90° tiny prism and focused on the sample through a spherical lens (L1), 65 mm f.l. The lens functioned as focusing/collecting/collimating lens that initially collimated the narrowed Gaussian laser beam in the y direction and focused it in the x direction. This led to a line beam characterized by a Gaussian profile with 13 mm × 150 µm full width half maximum, as measured by the knife-edge method [44]. In fact this beam length is larger than the size of the sampled particles (in the y direction), leading to factors of overlap between the areas of the incident laser beam and the particle, <1, depending on the size of the sampled particle. The calculated beam radius for point-excitation and the depth of focus for both beams were ~20 µm and ~0.5 mm, respectively. Accordingly, the intensity of the incident laser beam is larger by about five orders of magnitude for point-excitation than for line-excitation. The samples were placed on an x-axis motorized translation stage and raster-scanned through sampling, at fixed intervals within the defined area. This stage was combined with a manual z translational stage, for finding the optimal focus of the sample, positioned in front of L1.

Fig. 1
figure 1

Schematic of the compact Raman spectrometer for line-scan measurements, including the following components: handheld green laser, dichroic mirror (DM), cylindrical lens (CL), spherical lens (L), an x, z translational stage, long-wavepass edge filter (F) and a detector consisting of a spectrometer and an intensified charge-coupled device (ICCD)

Full size image

The backscattered spontaneous Raman signal was collected and collimated by the same L1 lens and filtered by a long-wavepass edge filter. This signal was focused using an f/4 lens (L2) on the entrance slit of a 0.14 m spectrometer with an aperture ratio of f/3.88, equipped with an air-cooled 1024 × 1024 intensified CCD (ICCD) camera. The synchronized x movement of the stage and the ICCD readout was computer-controlled by a newly developed LabVIEW software application, which acquired successive Raman signals for line- or point-excitation across the sample.

The spatial data were measured by the ICCD on a line parallel to the entrance slit of the spectrometer (y-axis), while dispersing the spectral data vertically to it. During the Raman measurements, full vertical binning (FVB) or image readout modes of the ICCD were employed, providing the total intensity of each of the peaks along its vertical pixels or at a specific pixel, respectively.

Moreover, monitoring the intensity of the dominant peaks in the spectra, along each of the vertical pixels of the ICCD, while scanning the sample in the x direction, enabled obtaining Raman line-maps of the different samples. The samples were translated by 50 μm steps, and the Raman signal was obtained by FVB or image readout, using two accumulations of 100 ms and 0.5 s integration times, respectively. The samples contained particles of potassium nitrate, 2,4-dinitrotoluene (DNT), 2,4,6-trinitrotoluene (TNT) and 1,3,5-trinitro-1,3,5-triazacyclohexane (RDX). These samples included individual particles, consisting single or several grains of the same compound, or of two different compounds, which were prepared by randomly depositing them on particular glass microscopic slides. The size and volume of the particles were obtained from the photographs of the individual measured particles. The size of individual particles in the x, y directions was in the 0.5–1.5 mm range, with roughly estimated thicknesses, (z) direction, of 0.3–0.5 mm for the KNO3 particles and 150–300 μm for DNT, RDX and TNT.

3 Results and discussion

3.1 Comparison of Raman signals, obtained by point- and line-excitation

Representative Raman spectra of KNO3, DNT, TNT and RDX particles, measured with the compact Raman spectrometer, using point- and line-excitation, are shown in Fig. 2. These displayed spectra were treated by subtracting the underlying broad background and by calibrating the wavelength, based on the procedure of Ref. [23]. Specifically, the outcome of point-excitation for KNO3, DNT, TNT and RDX is shown in Fig. 2a, d, g and j, respectively, and the corresponding results for line-excitation appear in Fig. 2b and c, e and f, h and i and k and l. For panels (c), (f), (i) and (l) of Fig. 2, image readout was performed, while for the rest FVB.

Fig. 2
figure 2

Representative Raman spectra of particles of potassium nitrate (KNO3) [(a)–(c)], 2,4-dinitrotoluene (DNT) [(d)–(f)], 2,4,6-trinitrotoluene (TNT) [(g)-(i)] and 1,3,5-trinitro-1,3,5-triazacyclohexane (RDX) [(j)–(l)]. The spectra were measured with the compact Raman spectrometer operating by point- [(a), (d), (g) and (j)] and line-excitation [(b), (c), (e), (f), (h), (i), (k) and (l)] with image readout for (c), (f) (i) and (l) and full vertical binning for the rest. The measurements were performed using integration times of 1 s for all samples, with ten accumulations for KNO3 and DNT and twenty accumulations for TNT and RDX samples, respectively

Full size image

It is seen that the spectral features related to a specific compound are similar for the two measurement modes; however, the SNRs of the spectra differ. For instance, the calculated SNRs, determined from the ratio of the dominant peak height of KNO3 (1054 cm−1) to the standard deviation of the background in Fig. 2a–c, were found to be about 120, 40 and 13, respectively. Similar behavior was also encountered in measurements of the Raman spectra of DNT (1350 cm−1) (Fig. 2d–f) and of TNT (1365 cm−1) (Fig. 2g–i) and RDX (880 cm−1) (Fig. 2j–l), implying that the SNRs obtained by point-excitation are higher than those by line-excitation with FVB readout of the ICCD. The signal intensities and the SNRs decreased even more upon line-excitation and image readout. The spectral SNRs obtained for point- and line-excitation could be attributed to the different measurement conditions. In particular, considering the factor of overlap, between the area of the incident laser beam and the particle, in case of line-excitation (see below), as well as the lower collection efficiency of the Raman scattering line-excitation spectrometer, it is reasonable to obtain reduced SNRs for this operation mode. During image readout, the SNR is reduced even more, due to the reading of specific pixels of the ICCD. Consequently, the Raman intensities, obtained by line-excitation and FVB or image readout, can be strongly affected by the size of the sampled particles, provided that the length of the exciting line is kept constant.

Roughly, threefold to fourfold ratios were observed for the Raman signal intensities of the KNO3 (Fig. 2a–b), DNT (Fig. 2d, e) and RDX (Fig. 2j, k) samples, respectively, obtained by point- and line-excitation. On the other hand, for the TNT (Fig. 2g, h) sample, the intensities were almost similar. This can be understood by considering that the spontaneous Raman signal intensity, I Raman, is proportional to \( I_{\text{Raman}} \left( {\frac{{{\text{d}}\sigma }}{{{\text{d}}\varOmega }}} \right)LN \) IA, where \( \left( {\frac{{{\text{d}}\sigma }}{{{\text{d}}\varOmega }}} \right) \) is the differential Raman scattering cross section for a specific transition, L is the length of the interaction region, N is the molecular number density, I is the exciting laser beam intensity, and A is the overlap factor between the areas of the incident laser beam and the particle in the sample. According to this expression, the Raman signal should be linearly reduced as a function of the laser beam intensity, but it should be increased due to the larger sampled area of the particle upon line-excitation and FVB readout. Indeed, by considering an overlap factor of about 0.4 for line-excitation of KNO3 and DNT, the threefold ratio between the Raman signal intensities obtained by point- and line-excitation seems reasonable. A similar behavior was encountered for the particles of TNT and RDX that had overlap factors of 0.9 and 0.25, respectively.

Actually, if the exciting laser intensities are below the damaging threshold of the sample [43] and the particles should be completely overlapped by the exciting line, similar Raman signal intensities should be expected for the two measurement approaches. Nevertheless, the collection efficiency of the Raman scattering line-excitation spectrometer is probably lower, leading to somewhat reduced signal. It is expected that the Raman signal should rise, upon line-excitation, if the number of scatterers is increased, providing that FVB readout is used.

To check this behavior, the dependence of the Raman signal for KNO3 particles of different sizes, representing a variety of overlap areas, was measured. This was achieved by measuring the average intensity of the dominant peak (1054 cm−1) above that of the background (in the vicinity of the peak) in each spectrum. The Raman signal resulting from different KNO3 particles, with point- (red squares) and line-excitation (blue squares), was plotted as a function of the estimated volumes, covered by the laser beam, and is shown in Fig. 3. Considering that the same particles with similar thicknesses were sampled by the two modes of operation and that the excitation conditions were similar for each mode, then the sampled volume was mostly affected by the size of the particles in two different directions. Actually, the spot size was smaller, while the illuminating line in the y direction was larger than the size of the measured particles. Therefore, the different thicknesses (<0.5 mm) of the particles (z direction), which were less than the depth of focus of the incident beam, and the size of the particles in the y direction, played the major role in determining the sampled volume for point- and line-excitation measurements, respectively.

Fig. 3
figure 3

Intensities of the dominant Raman peak of potassium nitrate particles on glass substrates as a function of their volumes, obtained through operation of the compact Raman spectrometer with point (red) and line (blue) sampling and full vertical binning readout

Full size image

At any rate, for both operation modes, a linear growth of the Raman signal as a function of the volume of the KNO3 particles is observed and the signal obtained by point-excitation is higher. Again, this is reasonable if it is considered that only partial overlap between the incident laser beam and the measured particles occurs. Nevertheless, the use of line-excitation and FVB readout led to Raman signal intensities of the same order of magnitude as those obtained by point-excitation.

3.2 Detection and imaging by line-excitation

Based on the above, it is anticipated that in scenarios where only detection is required, it would be possible to use the line-excitation mode of operation with FVB readout. Indeed, by scanning the sample and using line-excitation, it was possible to detect the presence of particles. This can be seen from comparison of Fig. 4a, b, which shows a photograph, obtained by a smartphone camera, presenting two KNO3 particles (round shape) and one of DNT, consisting of several needles, deposited on a glass slide and their respective detection, as indicated by the bright lines in the ICCD readout data. These lines represent the intensities of the dominant peaks in the Raman spectra, i.e., 1054 and 1350 cm−1, reflecting the presence of KNO3 and DNT particles at positions that match their appearance in the photograph, Fig. 4a. This mode of operation resulted in very short acquisition times of only 1 min, for a scan of a 5 × 5 mm area, limited by the time of movement of the motorized stage. Essentially, this time depends not only on the scanned area, but also on the scanning step size; however, the fast detection of the presence of specific particles in a relatively large area is of much importance.

Fig. 4
figure 4

a A photograph presenting the individual particles of potassium nitrate (KNO3) and 2,4-dinitrotoluene (DNT) on the glass substrate, b a view of the detector readout, for operation of the compact Raman spectrometer in line-scanning mode with full vertical binning (FVB), corresponding to 1054 and 1350 cm−1. These readouts show the intensities at the positions of the particles in (a) and consequently evidence for presence of KNO3 and DNT, c and d x, y Raman point- and line-maps obtained by FVB and image readout, showing the KNO3 (red) and DNT particles (blue), respectively. The normalized intensity scales for the Raman signal in (b), as well as c and d are shown to their right, respectively

Full size image

Besides, by considering the high sensitivity achieved by line-excitation and the feasibility of sample scanning in the x direction, it seemed important to test whether mapping of individual particles would be possible and whether it would be preferable over point-excitation. Therefore, Raman spectra of KNO3 and DNT, obtained by point-, Fig. 4c, and line-excitation, Fig. 4d, and FVB and image readout, respectively, were measured, allowing to retrieve the intensities of the dominant peaks for each of them. These intensities, obtained at each point of measurement at the 1024 lines of the ICCD, represent the available scatterers for point- or line-excitation. The obtained intensity matrices allowed creating Raman spectral maps of the samples. As can be seen from Fig. 4c, d, the Raman images, measured by point- and line-excitation, for KNO3 (red) and DNT (blue) particles, resemble very much the optical images of the samples, Fig. 4a. Under the conditions of our experiment, the measurement time required to obtain the Raman map of the 5 × 5 mm area by point-excitation (steps of 50 μm) was 135 min, while for line-excitation mapping (similar steps) the time was cut to only 7 min, resulting in a reduction factor of ~20.

This approach was also used for obtaining the Raman maps for RDX and TNT samples, as shown in Fig. 5a, b. These Raman maps, based on the intensities of the dominant peaks at 880 and 1365 cm−1, also show evidence for the presence of particles of the corresponding compounds, which match well their positions and shapes in the optical images, Fig. 5c, d. Actually, even some of the very small particles appearing in Fig. 5c were observed in the Raman map, but their observation is limited by the very short acquisition times. Yet, as mentioned above, the line-mapping of the same 5 × 5 mm area with steps of 50 μm reduced the acquisition time to only 7 min, decreasing further the time required for mapping, already achieved by the photo-guided sampling [24]. Hence, the line-spectrometer is suitable for line-mapping of explosives, although it suffers from poorer spatial resolution. This is since the scattered photons propagate from the particle to the spectrometer with a line profile, preventing the use of the confocal hole for rejection out-of-focus Raman signal.

Fig. 5
figure 5

x, y Raman line-maps showing the presence of a 1,3,5-trinitro-1,3,5-triazacyclohexane and b 2,4,6-trinitrotoluene particles, as obtained from the intensity distribution of their dominant and characteristic peaks at 880 and 1365 cm−1, respectively. The corresponding photographs of the samples appear under the Raman line-maps in panels (c) and (d). The normalized intensity scales for the Raman signal in panels (a) and (b) are shown to their right

Full size image

4 Summary

The use of the compact Raman spectrometer with line-excitation and scanning of the samples with FVB or image readout allowed fast detection and imaging. The power densities for point- and line-excitation were very different, but the acquisition times required for the Raman maps with the latter were much faster, pointing to its preference. The short acquisition times for detection and mapping can be shortened even more if line-excitation with higher intensities and photo-guided sampling will be used. This suggests the possibility of using our setup as a fast, compact and relatively low-cost system for detection and mapping of explosives and a variety of other samples.