1. Introduction

Laser-induced breakdown spectroscopy (LIBS) uses a focused laser pulse to breakdown a small target area to generate a radiating mixture of ions, electrons, and neutral fragments of the target material for spectroscopic studies. This fully or partially ionized plasma can in general emit detectable radiation that could reveal the identities of its components. It has attracted much attention because it is a fast and robust technique of emission spectroscopy [1] and might be suitable for standoff detection of materials, such as chemical explosives [2,3]. Conventional LIBS commonly uses nanosecond laser pulses. Ionization typically occurs in femtoseconds and the initial temperature of the plasma is relatively low as it is governed by the oscillatory energy of the electrons in the laser field. However, prolonged illumination by intense nanosecond laser pulses can continue to heat the plasma to high temperatures through collisional absorption processes, such as inverse bremsstrahlung [4]. Molecular fragments usually do not survive in such high-temperature plasma environment and would decompose into their atomic constituents and be further ionized. The spectral signature of such fully decomposed and ionized plasma would only contain the atomic features and the target molecular signatures would be lost. Even in the event that some residual molecular fragments were still present, their spectral signatures would be crowded by the typically strong atomic spectral lines or masked by the continuum background radiation of the hot plasma. This is particularly true with nitrocompounds, such as most organic energetic materials. Recent studies [3] showed the nanosecond LIBS spectra of multiple organic explosives to be primarily atomic lines of C, H, O, N, and some molecular lines, such as CN and $C_2$. One could only use measures, such as the relative ratios of these lines which carry large statistical uncertainties as a crude method to identify the explosives. One needs more specific molecular signatures, such as those of the ${\mathrm{NO}}_x$, to identify first of all that nitrocompounds are being detected. It is therefore important to study different LIBS environments that would favor better survivability of key molecular signatures, such as ${\mathrm{NO}}_x$.

In recent years, with the development of high-power picosecond and femtosecond lasers, ultrashort LIBS (ULIBS) has gathered considerable research attention [5–17] for more than the simple reason of higher laser power and laser intensity. First of all, even though the higher laser fields of an ultrashort pulse would have generated a higher initial plasma temperature, the much shorter pulse structure prevents further collisional heating of the plasma to the much higher temperatures of nanosecond LIBS plasma [15]. This lower-temperature breakdown-plasma allows higher survivability of the molecular fragments. It also generates a lower continuum background [15]. Second, lower-temperature plasma has shorter electron-ion recombination time, allowing rapid decay of the atomic line emissions. These decay times could be $<200\mathrm{ns}$ for ULIBS [5] as compared to $>5\mathrm{ms}$ for nanosecond LIBS [17]. On the other hand, molecular emissions could have longer time decays, such as the excited B state of the nitric oxide (NO), which has a lifetime of over a microsecond [18]. Time-gated detection techniques could be employed to separate the atomic and molecular signatures, using detectors, such as a gated intensified optical imager. Atomic line emissions could be eliminated as early as $\sim 200\mathrm{ns}$ after the femtosecond laser pulse, when the molecular signal is still quite early in its decay and thus still has relatively large amplitudes for better detection. Most of the recent studies in ULIBS were conducted under controlled pressures and millijoule level pulse energies in a laboratory setting, using targets, such as ethanol vapor and methane [8,9] and biological materials varying from albumin, barley, corn, wheat, to bacteria [10–16]. In these studies, ULIBS showed fluorescence signatures of molecular fragments, such as CH, $C_2$, CN, $N_2$, and $N_2^{+}$ [8–16]. These are very promising results for ULIBS to be a useful technique for molecular identification of target materials.

Another advantage of using ultrashort laser pulses in LIBS is that they can be manipulated to focus longitudinally and transversely in air at standoff distances due to linear group velocity dispersion (GVD) and nonlinear self-focusing [19,20]. Remote focusing of ultrashort pulse for standoff detection has been demonstrated previously [21,22]. The combination of molecular spectral signatures and remote generation of high-intensity laser pulse is the unique approach in applying ULIBS to standoff detection of chemicals.

The goal of this research work was to focus on the fluorescence emitted from the plasma generated in an ultrashort laser-pulse-induced surface breakdown. A standoff breakdown was performed in the laboratory by using a telescope to focus the beam tens of meters away onto a target to investigate the aspects of using ULIBS to detect surrogate explosive chemicals. Since large explosive molecules could be broken into small, excited fragments and radicals using femtosecond laser pulses, the experiment concentrated on detecting signatures emitted from partially broken molecular fragments. Most explosive molecules emit in the mid-infrared region of 3–10 μm [23,24]. At standoff distances of many tens of meters, it would be very difficult to detect these radiations because of the low sensitivities of mid-infrared detections. However, for fragments, such as ${\mathrm{NO}}_2$, NO, ${\mathrm{CH}}_3$, and CN, they could be excited to higher electronic states with higher orders of vibrational transitions. The fluorescence radiation emitted from these electronic states could lie in the ultraviolet to visible and near-infrared region with vibrational modifications to their emission spectra, as reported for the NO, ${\mathrm{NO}}^{+}$, and ${\mathrm{NO}}_2$ molecular spectra [25,26]. Visible to near-infrared ${\mathrm{NO}}_x$ signatures are especially useful in atmospheric remote sensing because they propagate with little attenuation in the atmosphere and common silicon detectors could be used. These spectral features could then provide the necessary signatures to identify the presence of different nitrocompounds.

This report focuses on the LIBS study using a femtosecond laser. Nanosecond LIBS on the nitrate compounds had been carried out at the initial phase of our experiment to test run the experimental setup. The results collected in the nanosecond LIBS showed familiar spectral characteristics, such as atomic line spectra and high continuum backgrounds. They served only to validate the experimental configuration and are not presented here. Instead, we are reporting on the observation of an ${\mathrm{NO}}_x$ signature in the visible wavelength when femtosecond LIBS was applied with appropriate laser pulse configurations. Suppression and reduction of the atomic line emissions and continuum background so often seen in nanosecond LIBS spectra were also observed at these laser pulse formats. The resulting improvement of the SNR might have contributed to the revelation of the visible ${\mathrm{NO}}_x$ signature in our experiment.

2. Experimental Setup

The lasers used in this experiment were a 10 Hz Ti:sapphire femtosecond laser (TFL) that can generate a 50 fs pulse with a central wavelength of 800 nm and a single shot Nd:YAG laser at 1064 nm that can generate a 10 ns, 30 mJ pulse. The nanosecond laser was used to generate nanosecond LIBS spectra as controls. The TFL has both a regenerative amplifier (Regen) and a five-pass amplifier that can deliver a pulse energy of over 1 J before compression (up to 600 mJ after compression). The pulse energy used in the experiment was controlled by attenuating the stretched pulse with high-quality reflective filters of optical density 1.0 and higher. The pulse energies obtained after compression were $(20\pm 1)\mathrm{mJ}$, $(50\pm 1)\mathrm{mJ}$, and $(70\pm 1)\mathrm{mJ}$. The final laser pulse duration could also be changed by varying the distance between the pair of compressor gratings to overcompensate or undercompensate the compression by the gratings. The final laser pulse will necessarily retain certain amount of positive or negative frequency chirps. However, for the pulse lengths of 100 fs to 1 ps used in this experiment, the residual chirps were relatively mild and the physics of ULIBS has a very mild dependence on the frequency chirp in the laser pulse.

The compressed pulse was down-collimated through a $2:1$ all-reflective telescope to a beam diameter of $\sim 2.5\mathrm{cm}$. A reflective collimator was used here to prevent nonlinear focusing effects that could occur inside transmissive optics. For this reason, reflective optical elements were used extensively in this experiment except at a few inconvenient locations described below. The laser beam was then reflected off several dielectric and gold mirrors before arriving at a telescope composed of a plano-concave negative lens [1 in. (2.5 cm) diameter] and a large [10 in. (25 cm) diameter] concave spherical mirror. Figure 1 shows the schematic of the experimental setup.

In order to adjust the laser focus, the negative lens was mounted on a translation stage. The last mirror before the negative lens was an elliptical gold mirror with a 1 in. (2.5 cm) minor axis. The combined footprint of the gold mirror, the negative lens, and the mounting hardware was minimized to reduce the loss in laser energy and the spatial distortion from the clipping of the ongoing laser beam. On its way to the target, the beam was reflected off a few more mirrors to increase the total propagation distance to more than 20 m to simulate the standoff breakdown distances anticipated. The energy deposited on the target was approximately 50% of the output compressor. This energy loss was due to the clipping in the telescope and the low reflection efficiencies of some of the mirrors.

 

Fig. 1. Schematic of experimental setup that shows the laser beam path and the timing of camera trigger.

Download Full Size | PDF

The spectrometer used in this experiment was an imaging spectrometer (Acton SpectraPro 0.3 Meter Triple Grating Monochromator/Spectrograph). In order to improve the resolution and UV transmission of the collection lens, two fused silica plano-convex lenses placed back to back (spherical side facing each other to reduce spherical aberration) were used. The combined focal length of the collection lens set was chosen such that the emitted radiation from the laser breakdown of the chemical target was imaged onto the input slit of the spectrometer with an $f$-number of 4, matching the acceptance of the spectrometer. At the exit port, a CCD camera (Foculus FO432TB, pixel size: $4.65\mathrm{\mu m}\times 4.65\mathrm{\mu m}$) was positioned at the image plane of the entrance slit of the spectrometer. To measure the resolution of the detection setup, a target resolution card was placed on the object plane. The target resolution card was imaged at the imaging plane of the spectrometer output with the grating positioned at the zero-order (specular reflection). The smallest image size that can be resolved was measured at 35 μm. However, the width of the slit was set at a convenient value of 50 μm. In this experiment, two different gratings with groove densities of $150$ and $600\mathrm{gr}/\mathrm{mm}$ were used. Taking into consideration of the slit width, the spectral resolution of the spectrometer for these two gratings were 1 and 0.3 nm, respectively, and verified with narrow atomic emission lines from metallic targets.

The spectrometer wavelength calibration was carefully performed with Newport-Oriel spectral calibration lamp 6034 (mercury-neon) for wavelengths between 365 and 546 nm. The calibration lamp was placed at the target location to eliminate any alignment error of the optical axis of the collection optics with the optical axis of the spectrometer that may cause an apparent wavelength shift. In addition, a pure copper target was also used for the known copper emission lines between 510 and 578 nm as further verification of the accuracy of the wavelength calibration of the spectrometer. The accuracy of the wavelengths measured in this experiment is better than 1 nm.

The laser alignment was routinely performed at 10 Hz while data collection was done in single shots. The laser pulses from the 10 Hz TFL can be down-selected as single shots by placing a shutter at the output of the Regen, which ran at 10 Hz. The shutter opened on demand to allow only one amplified pulse to leave. As shown in Fig. 1, the electronic timing control of the TFL then sent out a signal earlier than the laser pulse to the SRS DG535 digital delay pulse generator, which in turn triggered the CCD camera after a time delay of 48.75 ms. The CCD camera’s exposure time was $\sim 0.1\mathrm{ms}$, more than enough to capture the plasma emission. To monitor the energy of the laser pulse used for the breakdown, a photodiode was placed behind the last dielectric mirror to capture the leakage radiation. The photodiode signal was monitored with an oscilloscope, and the photodiode signal was calibrated against a laser energy meter directly measuring the laser energy on target.

Two samples were used as target chemicals, sodium nitrate (${\mathrm{NaNO}}_3$) and potassium nitrate (${\mathrm{KNO}}_3$), both for their NO-containing chemical structure as surrogates to actual explosives. These chemical samples were prepared in a way to obtain fresh, uniform, and known quantity at each breakdown. To form uniform and relatively smooth chemical distribution, a 10% concentration solution of the chemical was poured inside a small pool ($30\mathrm{mm}\times 40\mathrm{mm}\times 0.6\mathrm{mm}$) made in the middle of a plastic plate covered by aluminum tape. The filled pool containing the chemical was placed under an incandescent light bulb to dry. After complete water evaporation, it is estimated that the surface density of the target chemical is $\sim 0.06\mathrm{gm}/{\mathrm{cm}}^2$. The target plate was mounted on a two-dimensional translation stage where it could be translated up, down, left, and right. These two degrees of freedom allowed us to get a fresh spot after each laser breakdown. The one crucial parameter that was varied in this experiment was the laser intensity.

In this experiment, the change of the laser intensity at breakdown was achieved by changing pulse power and laser spot size on the target. Three sets of data were taken with each set having a different value for the laser power (100, 50, and 33 GW), which was in turn achieved through adjusting the laser pulse energy and laser pulse length. Data were collected for various footprints of the breakdown while the pulse power remained unchanged. The change of the laser spot size on target was accomplished by adjusting the position of the 1 in. (2.5 cm) negative lens in the telescope.

The laser spot diameter on the target was changed systematically as shown in Table 1. The laser spot diameter on target was measured by imaging through the spectrometer at the zero-order dispersion setting (specular reflection) of the grating. The input slit was set wide open at 3 mm to accept the full image of the laser spot on target. Color filters and neutral density filters were used in front of the spectrometer input slit to block plasma light and to allow only attenuated laser light scattered from the target laser spot to enter the spectrometer. The spot diameters of the laser spot on target were obtained as the FWHM values of the image line outs, where the demagnification factor of 1.5 for the collection lens was taken in account. Figure 2 shows a plot of the spot diameter as a function of the focused position of the laser with respect to the target plate. Negative values correspond to focused positions in front of the target plate. For these positions, the profiles of the laser spot on target were more uniform where the central clipping of the laser beam was partially filled in as a consequence of the spherical aberration of the telescope. These focal positions were therefore used in obtaining data in the ULIBS experiment.

Table 1. Matrix of Laser Parameters Used in This Experiment

View Table

 

Fig. 2. Measured spot diameter (FWHM) as function of laser focal positions at target. Positive values indicate focal positions behind the target plane.

Download Full Size | PDF

As mentioned earlier, the laser pulse length was modified by adjusting the separation of the compression gratings, and measured at the output of the laser using a second-order autocorrelator. The distance of the sample is 20 m from the laser. The bandwidth of the laser will incur a GVD spread of the laser pulse of about 20 fs in that distance [21]. Therefore, there will be a 20% discrepancy in the laser power and intensity on target for the shortest 100 fs laser pulse used in this experiment. This discrepancy will progressively decrease for the longer pulses used. Since the range of laser pulse lengths encountered in the experiment extended over 1 order of magnitude, significance of this discrepancy is suppressed in the next section where we will not distinguish the nominal laser pulse lengths measured at the output of the laser from those at the target location. In addition, we have deliberately imparted a positive chirp on the laser power so that the pulse length on target would always be about 20 fs longer. This is to help suppress the formation of filaments before the geometric focus of the telescope even for the shortest pulse length of 100 fs. Filament formation was checked during the experiment and was found not occurring for the entire 20 m of propagation distance. It has been estimated that the effective $f$-number associated with the focusing telescope is small enough to restrict possible filamentation until the last 15 cm for the highest laser power of 100 GW. Effects such as clipping of the beam on the Newtonian telescope also might have severely distorted the wavefront and prevented filament formation.

3. Data Analysis and Discussion

The experiments began with generating nanosecond LIBS spectra using the Nd:YAG laser. For pulse energies in the range of 10–30 mJ and laser spot size in the similar range of the femtosecond laser, all spectra were dominated by large background radiation from the nanosecond plasma produced by the Nd:YAG laser. The experiment then proceeded to use the femtosecond laser. As discussed earlier, experimental parameters were systematically varied and thus the results collected in this experiment were also categorized accordingly, as shown in Table 1. When categorized with respect to laser power, there are three sets of data under laser powers of 100, 50, and 33 GW. Each set can further be subdivided into three subsets using three different values of laser pulse energies with the appropriately adjusted laser pulse lengths. The size of the laser footprints was also changed from the smallest size allowed by the optics in the telescope to bigger sizes until the signal disappeared. Therefore, laser intensity on target was varied with all three parameters: footprint area of the focused beam, energy, and pulse duration. In this experiment, there were three sets of power and 11 sets of spot size for each power value. However, since each laser power value was obtained by three sets of energy and pulse duration, there were altogether 99 sets of data taken. Three shots were taken for each set of energy, pulse duration, and spot diameter values. Since the spectrometer used in the experiment was an imaging spectrometer, the spectrum obtained could be analyzed as a function of space within the laser spot on target. However, in this experiment, we have only attempted to study the spectra within the “hot spot” of the laser on target. Data collected were averaged by integrating spatially over the FWHM of the laser footprint.

A prominent feature in the fluorescence spectrum of the target chemicals was observed between 400 and 420 nm for both ${\mathrm{NaNO}}_3$ and ${\mathrm{KNO}}_3$, at various laser parameters used in the experiment. Figure 3 shows a set of spectra for ${\mathrm{NaNO}}_3$ at various laser spot sizes on target for a laser power of 100 GW. Case A where the spot size was only slightly smaller than case B, appeared to have substantial amount of continuum background and several atomic lines showing. This is probably a consequence of the nonuniformity of the target surface leading to a variation in the laser coupling efficiency. Case A may have more laser energy coupled and thus generating a higher-temperature plasma with higher backgrounds. In comparison, Fig. 4 shows typical spectra of control chemicals, such as sodium chloride (NaCl), potassium chloride (KCl), and a bare surface of aluminum. Other than knowing atomic lines from aluminum, potassium, and chlorine in the neighborhood of this spectral region, the prominent feature seen in Fig. 3 is absent in Fig. 4. It indicates that the nitrate component in the target chemicals may be responsible for this feature. Two approaches were used to investigate this spectral feature, temporal property and spectral identification.

 

Fig. 3. Typical spectral signals obtained from ${\mathrm{NaNO}}_3$ in ULIBS using 10 mJ, 100 fs laser pulses at different laser diameters on target: A, 0.608 mm; B, 0.628 mm; C, 0.740 mm; and D, 0.896 mm. The prominent broadband feature between 400 and 420 nm appeared for both ${\mathrm{NaNO}}_3$ and ${\mathrm{KNO}}_3$ for various laser energies, pulse lengths, and spot diameters.

Download Full Size | PDF

 

Fig. 4. Spectral signals from breakdown of a focused laser pulse on (a) NaCl (blue, short dashed line), (b) KCl (red, dashed–dotted line), and (c) Al (green, long dashed line). The spectral lines near 400 nm for Al are ionic aluminum lines, the line near 404 nm for KCL is a potassium line, and the line near 422 nm for KCl and NaCl is a chlorine line.

Download Full Size | PDF

The central wavelength of this feature near 410 nm raises the suspicion that it could simply be the second harmonic of the drive laser at 400 nm, though the large shift in the second-harmonic wavelength of 10 nm makes it not very plausible. However, there is still a finite possibility for this to occur because the shift could be the result of a preferentially phase-matched frequency within the broad bandwidth of the driving femtosecond laser. Since second-harmonic generation is a nonlinear process, it could only happen near the peak of the femtosecond laser for a time interval close to the pulse length of the laser pulse. For the laser pulse parameters studied in this experiment, the compressed pulse length is no more than a picosecond and the stretched pulse length is approximately 200 ps. Even for a poorly compressed laser pulse with substantial long time pedestals, the second harmonic that could be generated by the laser pulse could not have a temporal structure of more than 200 ps. Therefore, a temporal diagnostic would be useful in studying the spectral feature from the two nitrates. A fast photodiode, Thorlabs high-speed photodetector SV2 with a rated rise time of 175 ps and a fall time of 150 ps, connected to a LeCroy LC584AL 1 GHz bandwidth, $8\text{Giga-samples}/s$, single-shot oscilloscope with an air-core solid-shield heliax coaxial cable was used to study the ${\mathrm{NaNO}}_3$ spectral feature shown in Fig. 3. The spectrometer was put in the monochromator position with the output slit window set at a central wavelength of 410 nm and a slit width of 20 nm. The photodiode-oscilloscope setup was estimated to have a temporal response of about 500 ps. The intrinsic temporal response of the photodiode-oscilloscope setup was measured using an aluminum target with spectrometer set at both the fundamental wavelength of 800 nm and the wavelength of interest of 410 nm. Adequate neutral density filters were used to keep the photodiode response below saturation for the two wavelengths, respectively. The signal at 410 nm from the aluminum was simply the plasma continuum light from the aluminum surface breakdown. It is shown as the blue (dotted) curve in Fig. 5. It has a FWHM of about 425 ps, which was the same for the pulse width measured at the fundamental 800 nm (not shown), and thus agreed with the estimated response time. This indicates that the photodiode is suitable for detecting temporal events that are substantially longer than 425 ps and the plasma recombination light of laser-generated aluminum breakdown plasma in our experiment was much less than the intrinsic response of the photodiode used. With this photodiode, the spectral feature at 410 nm from ${\mathrm{NaNO}}_3$ was measured and is shown as the red (solid) curve in Fig. 5. Other than some oscillatory behavior present in both the blue and red curves (residual impedance mismatch in the cable), the red curve shows a temporal decay when compared to the photodiode response (blue dotted) curve. A simple fitting of an exponential decay revealed a decay time constant of about 1.3 ns. This temporal measurement of the spectral feature at 410 nm for the nitrate compound ruled out its origin as the second harmonic generated by the pump laser in the chemical.

 

Fig. 5. Oscilloscope traces obtained from a fast photodiode placed at the output slit of the monochromator at 410 nm with a 20 nm bandwidth: (a) signal from Al target (blue, dotted line) and (b) signal from NaNO3 (red, solid line). Oscillations are from reflections in the coaxial cable. Curve in (a) shows a FWHM of 425 ps for the intrinsic time response of the photodiode and oscilloscope. Curve in (b) has a $\sim 1.3\mathrm{ns}$ exponential decaying tail.

Download Full Size | PDF

To further analyze this spectral feature, higher groove density grating ($600\mathrm{gr}/\mathrm{mm}$) was used with a measured resolution of 0.3 nm. The high-resolution spectrum shown in Fig. 6 has relatively broad structures (linewidths of 2–3 nm) that could be of molecular origins. These structures were found to be compatible with the $\beta$-system NO and the visible system ${\mathrm{NO}}_2$ signatures [25,26]. In the spectral region from 400 to 430 nm, NO radicals can emit at regions of 402.78–404.18 nm (0–12), 411.36–412.79 nm (1–13), and 420.07–421.52 nm (2–14). These emission signatures represented the transition $B{}^2\mathrm{\Pi }\leftrightarrow X{}^2\mathrm{\Pi }$, where numbers in parentheses are the vibrational states. Also, ${\mathrm{NO}}_2$ radicals emit at 408.1, 410.2, 413.3, and 427.0 nm. In addition to the signatures mentioned, other signatures of NO [26,27] such as the 380 nm (0–11) signature have been observed in our experiment, but the signal strength was rather weak from the combined low spectral response of our grating and silicon-based detection system at wavelengths less than 400 nm. Similarly, a possible signature at 358 nm (0–10) for ${\mathrm{NO}}_x$ could not be observed because it lies on the edge of our detector’s spectral sensitivity range. In addition, we have observed other NO signatures around 520 nm (6–20) [26,27], but they require higher laser fluences and the laser parameters were quite different from those chosen for this study. That signature will be studied in more detail in future experiments. The signature that appeared in the 400–430 nm region was chosen to be analyzed and studied extensively because of their prominent intensities and also their constancy in appearance for our laser parameters. It had been shown that such signatures could occur when a trace of oxygen is introduced into active nitrogen [25,26]. Since air contains both nitrogen and oxygen, there was a possibility that the surface breakdown might have formed excited ${\mathrm{NO}}_x$ radicals directly from air. However, when other similar targets, such as plain aluminum, plain copper, and other nonnitrate salts, such as NaCl and KCl were used, as shown in Fig. 4, these ${\mathrm{NO}}_x$ related signatures were not observed indicating such formation of nitro-radicals from air did not occur in our experiment. It has been brought to our attention since the initial writing of this report that the ${\mathrm{NO}}_x$ signature at 400–430 nm has been present in recent nanosecond LIBS studies of organic explosives (Fig. 3 in [3]), although there is no explicit recognition or discussion of this signature in [3].

 

Fig. 6. Spectral signature of ${\mathrm{NaNO}}_3$ obtained using $600\mathrm{groove}/\mathrm{mm}$ grating.

Download Full Size | PDF

Data were generally taken with a lower-resolution grating of $150\mathrm{gr}/\mathrm{mm}$ for better SNR. The NO signature showed more variations at this spectral resolution but the general broadband structure did not change much. Figures 7 and 8 represent the signatures obtained from ${\mathrm{KNO}}_3$ and ${\mathrm{NaNO}}_3$ using $150\mathrm{grooves}/\mathrm{mm}$ grating for the same laser intensity. However, the laser parameters used were different. In Fig. 7, the pulse duration was 100 fs and laser pulse energy was 10 mJ. In Fig. 8, the pulse duration was 350 fs and laser pulse energy was 35 mJ. The signatures appeared to be quite different for these two figures. At shorter pulse durations (100 fs) the ${\mathrm{NO}}_x$ signature appeared quite clearly for both chemicals with the absence of atomic emissions. However, for longer pulse durations (350 fs), in addition to atomic emissions of aluminum and calcium (from impurities) appearing for both chemicals, the ${\mathrm{NO}}_x$ signature diminished substantially for ${\mathrm{NaNO}}_3$ and disappeared entirely for ${\mathrm{KNO}}_3$. In general, the ${\mathrm{NO}}_x$ signature appeared for short duration pulses and sufficiently high laser power. It disappeared or got weaker when the pulse duration was longer and the power was lower. These observations were true for both chemicals, and became more apparent through further data reduction. It is interesting to note in Fig. 8, especially for ${\mathrm{KNO}}_3$ on the left, that substantial continuum background radiation and atomic lines are present at these longer pulse lengths. They are reminiscent of the nanosecond LIBS spectra we observed during the initial phase of the experiment when a nanosecond laser was used for test runs.

 

Fig. 7. ${\mathrm{KNO}}_3$ (left) and ${\mathrm{NaNO}}_3$ (right) spectra around 400 nm, for laser parameters of 10 mJ, 100 fs, and 0.608 mm spot diameter.

Download Full Size | PDF

 

Fig. 8. ${\mathrm{KNO}}_3$ (left) and ${\mathrm{NaNO}}_3$ (right) spectra around 400 nm, for laser parameters of 35 mJ, 350 fs, and 0.608 mm spot diameter.

Download Full Size | PDF

In order to show more clearly the behavior of these signatures as functions of the laser parameters, the spectral region of 400–430 nm was processed. The continuum background, if any, was estimated with a linear fit. Any obvious atomic or ionic lines in this region were removed together with the estimated continuum background. The signal was then integrated between 400 and 430 nm and averaged over the number shots for each data subset. These results were then presented as bubble plots in Figs. 9–11, where the size of the bubble represented the strength (integrated intensity) of the signature normalized to unity. At laser parameters where there is no measurable signal a miniature bubble symbol is used to indicate experiments attempted at these parameters.

 

Fig. 9. Bubble plot of signal strength of NaNO3 (left) and KNO3 (right) against laser pulse energy and pulse duration, with the bubble sizes scaled to the relative amplitude of the signal strength. When no measurable signal was observed, a marker of 0.0 bubble size was introduced to mark the attempt for that laser parameter set. The full data of 81 laser parameter sets collapsed into only nine sets for these plots. The red lines show the loci of constant laser powers of 33, 50, and 100 GW.

Download Full Size | PDF

 

Fig. 10. Bubble plot of signal strength of KNO3 (top) and NaNO3 (bottom) against laser pulse duration and spot diameter, with the bubble sizes scaled to the relative amplitude of the signal strength. When no measurable signal was observed, a marker of 0.0 bubble size was introduced to mark the attempt for that laser parameter set. Low signal amplitudes at large spot diameters and long pulse durations could be attributed to overall low laser intensities.

Download Full Size | PDF

 

Fig. 11. Bubble plot of signal strength of KNO3 (left) and NaNO3 (right) against laser intensity and pulse duration, with the bubble sizes scaled to the relative amplitude of the signal strength. When no measurable signal was observed, a marker of 0.0 bubble size was introduced to mark the attempt for that laser parameter set. For ${\mathrm{KNO}}_3$, an upper bound on the laser pulse duration (indicated by the red line) exists where signal was observed for all laser intensities attempted in the experiment. For ${\mathrm{NaNO}}_3$, the pulse duration limitation has yet to be reached and signal was still observed up to the maximum pulse duration of 1 ps.

Download Full Size | PDF

The first two laser parameters to be examined were the laser pulse energy and the laser pulse lengths. Since the ratio of the two parameters is the laser power, the bubbles in Fig. 9 formed three lines representing the three different laser powers (100, 50, and 33 GW) examined in this set of experiments. For the ${\mathrm{NaNO}}_3$ plot, the signal diminished when the power was decreased, but was relatively strong at the 100 and 50 GW laser power levels. For the KNO3 plots, the signal was the strongest when the laser power equals to 50 GW, but it disappeared when the power was reduced to 33 GW. Note that although there were 99 sets of data, they collapsed into only nine for these plots because of the choice of the laser parameters. The reduction in signal for 100 GW in ${\mathrm{KNO}}_3$ may seem at odd with the ${\mathrm{NaNO}}_3$ where the signal remained large at 100 GW. In general, the ${\mathrm{NO}}_x$ signal strength diminished when the laser-generated plasma became too hot and ${\mathrm{NO}}_x$ radicals further dissociate. Higher laser power usually translates into high laser intensity and thus hotter plasma. So it is actually not surprising to see the ${\mathrm{NO}}_x$ signal being smaller at 100 GW for ${\mathrm{KNO}}_3$. The observation that for ${\mathrm{NaNO}}_3$ the ${\mathrm{NO}}_x$ signal remained strong at the highest power may indicate ${\mathrm{NaNO}}_3$ has a higher intensity threshold for generating the ${\mathrm{NO}}_x$ signal.

However, when the integrated signal was plotted against other laser parameters, such as pulse durations and spot diameters as in Fig. 10, all 99 datasets were represented. It should be noted the independent laser variables in the experiment were the pulse energy, pulse width, and spot size. Laser power and laser intensities were derived quantities. Figures 9 and 10 were plotted with independent laser parameters as x and y axes, while Fig. 11 below was plotted with the dependent parameter laser intensity on one axis and the independent parameter laser pulse width on the other.

In both of the two plots in Fig. 10, most of the strong signals appeared at shorter pulse lengths and smaller spot diameters. In the case of ${\mathrm{KNO}}_3$, no signal was observed for laser pulse length of 700 and 1050 fs. As for NaNO3, the dependence on laser pulse widths was not as apparent but quite diminished signal strengths can be seen for 750 and 1000 fs data points. The enhancement of the signal for shorter pulses could be the effect of high laser power associated with shorter pulses. Similarly, the dependence on spot diameters may also be an indication that higher laser intensities enhanced the signal strength. Thus, it would be interesting to plot the integrated signal strength against another set of laser parameters, the derived quantity of laser intensity and the laser pulse length, as in Fig. 11. This may help to reveal more clearly the dependences of the signal on the laser parameters and possibly the underlying mechanisms for the signal generation.

In Fig. 11, it can be seen that higher laser intensities led to larger magnitudes for the observed signature, such as the column of bubbles at 100 fs for ${\mathrm{KNO}}_3$ on the left and most of the columns for ${\mathrm{NaNO}}_3$ on the right. At the same time, laser pulse length was also very important. Substantial signal strengths could be generated at lower laser intensities for relatively long pulses (500–700 fs) as shown near the bottom of the ${\mathrm{KNO}}_3$ plot on the left, although that is not as apparent for the ${\mathrm{NaNO}}_3$ plot on the right. Instead, for ${\mathrm{NaNO}}_3$, a trend of diminishing signal strength develops as pulse lengths were increased for laser intensities $<25\mathrm{TW}/{\mathrm{cm}}^2$. This negative effect of long pulse lengths also showed up for the ${\mathrm{KNO}}_3$ dataset. In fact, for ${\mathrm{KNO}}_3$, laser pulse lengths of 750 and 1050 fs did not produce any observable signals. As for higher laser intensities in the ${\mathrm{KNO}}_3$ case, laser pulses shorter than 750 fs turned off the generation of the ${\mathrm{NO}}_x$ spectral signature. Also, as pointed out above, there seems to be different laser intensity thresholds for the two nitrates. ${\mathrm{KNO}}_3$ plot on the left in Fig. 11 showed larger signal strengths near the bottom where laser intensities were quite low, while the ${\mathrm{NaNO}}_3$ plot on the right has large signal data points near the top for the highest laser intensities. It has been noticed during the course of the experiment that NaNO3 produced weaker surface breakdown compared to KNO3 for the same laser parameters. This could be again caused by a difference in the overall laser coupling efficiency of the two compounds. It was observed that the two salts precipitated out of the water solution differently during preparation of the target surfaces, since they have quite different solubilities in water. The resultant surface morphologies of the two nitrates are noticeably different.

It appears that there is a correlation between the laser intensity and the laser pulse length in generating the molecular signature. For the laser parameters used in this experiment, there seems to be a range of laser pulse lengths and laser intensities that would produce substantial signal strengths. Effects such as the variable laser coupling efficiency due to target surface nonuniformity may have led to some of the fluctuations in the signal strengths shown in Figs. 10 and 11, even with the three-shot averaging performed in the data analysis. It appears to be difficult to define some clear and sharp boundaries or thresholds in the laser parameter for the onset of the ${\mathrm{NO}}_x$ signature. An attempt to mark the region of this laser parameter range is loosely marked for the ${\mathrm{KNO}}_3$ plot on the left in Fig. 11 with a red line. As for NaNO3, the pulse length limit for extinguishing the signal seems to be a bit beyond the pulse length parameter that was attempted in this experiment. However, based on the similarity of the two compounds and their behaviors shown in Fig. 11, that pulse length limit is expected to be around 1 ps also. Although short laser pulse length is one of the prerequisites to obtain high laser intensity, this laser parameter requires more considerations with respect to generating molecular spectral signatures. Upon closer examination of the spectra obtained at these high laser intensities, as shown in Fig. 8, there was often a much elevated background in the spectra together with atomic spectral features showing up at the same time. It is very likely that at the high laser intensities produced by ultrashort pulse lasers, picosecond or shorter pulses could already be too long to prevent complete breakdown of the target chemicals into their component elements and the formation of a fully ionized plasma region. The substantially higher background radiation from a fully developed plasma region could easily mask the spectral signature of any residual ${\mathrm{NO}}_x$ radicals.

4. Conclusions

We have presented the results of an experiment where a short pulsed laser (100 fs–1 ps) was used instead of a nanosecond pulsed laser to generate the breakdown spectra in LIBS of ${\mathrm{KNO}}_3$ and ${\mathrm{NaNO}}_3$ covered surfaces. This ultrashort pulse LIBS was conjectured to have advantages over the traditional nanosecond LIBS in being capable of partially preserving the molecular structure of the target sample and in creating a relatively low background plasma radiation. The data collected and analyzed in this experiment indeed verified that these advantages exist for the ULIBS. A molecular signature in the region of 400–430 nm was observed and it was identified to be the band progression of the fluorescence radiation of the nitro-radical fragments from the chemicals. However, it was also discovered that the notion of subnanosecond (picosecond) laser pulses being short enough to avoid excessive ionization in regular LIBS may be overly optimistic. For laser intensities around ${10}^{13}\mathrm{W}/{\mathrm{cm}}^2$ that were used in this experiment, laser pulse lengths much shorter than 1 ps were found to be necessary to preserve the molecular signatures. The identification of the 420 nm ${\mathrm{NO}}_x$ signature and that it occurs only for femtosecond to picosecond pulse formats showed the benefit of operating LIBS using ultrashort lasers as compared to nanosecond LIBS. Combining this with the additional benefits of high laser intensities at extended distances from femtosecond laser propagation and the accessibility of inexpensive detectors in the visible spectrum for the ${\mathrm{NO}}_X$ signature, it appears that there are certain advantages to the use of ultrafast LIBS for standoff detection applications.