Spectrochimica Acta Part B 62 (2007) 1433 – 1442 www.elsevier.com/locate/sab Efficient plasma and bubble generation underwater by an optimized laser excitation and its application for liquid analyses by laser-induced breakdown spectroscopy ☆ Violeta Lazic a,⁎, Sonja Jovicevic b , Roberta Fantoni a , Francesco Colao a a b ENEA, FIS-LAS, V. E. Fermi 45, Frascati (RM), Italy Institute of Physics, 11080 Belgrade, Pregrevica 118, Serbia Received 29 November 2006; accepted 8 October 2007 Available online 18 October 2007 Abstract Laser-induced breakdown spectroscopy (LIBS) measurements were performed on bulk water solutions by applying a double-pulse excitation from a Q-Switched (QS) Nd:YAG laser emitting at 1064 nm. In order to optimize the LIBS signal, laser pulse energies were varied through changing of the QS trigger delays with respect to the flash-lamp trigger. We had noted that reduction of the first pulse energy from 92 mJ to 72 mJ drastically improves the signal, although the second pulse energy was also lowered from 214 mJ to 144 mJ. With lower pulse energies, limit of detection (LOD) for Mg in pure water was reduced for one order of magnitude (34 ppb instead of 210 ppb). In order to explain such a phenomenon, we studied the dynamics of the gas bubble generated after the first laser pulse through measurements of the HeNe laser light scattered on the bubble. The influence of laser energy on underwater bubble and plasma formation and corresponding plasma emission intensity were also studied by photographic technique. From the results obtained, we conclude that the optimal first pulse energy should be kept close to the plasma elongation threshold, in our case about 65 mJ, where the gas bubble has its maximum lateral expansion and the secondary plasma is still well-localized. The importance of a multi-pulse sequence on the LIBS signal was also analyzed, where the pulse sequence after the first QS aperture was produced by operating the laser close to the lasing threshold, with the consequent generation of relaxation oscillations. Low-energy multi-pulses might keep the bubble expansion large prior to the probing pulse, but preventing the formation of secondary weak plasmas in multiple sites, which reduces the LIBS signal. The short interval between the pre-pulses and the probing pulse is another reason for the observed LIBS signal enhancement. © 2007 Published by Elsevier B.V. Keywords: LIBS; Water; Laser; Plasma; Bubble 1. Introduction LIBS technique is a powerful tool for in-situ elemental measurements and in different surroundings (gas, water), and a wide range of applications has been developed in the past two decades [1,2]. Spectrally resolved plasma emission generated by laser ablation of solid targets or by breakdown in medium (gaseous or liquid) allows determining qualitatively the elemental sample com☆ This paper was presented at the 4th International Conference on Laser Induced Plasma Spectroscopy and Applications (LIBS 2006) held in Montreal, Canada, 5–8 September 2006, and is published in the Special Issue of Spectrochimica Acta Part B, dedicated to that conference. ⁎ Corresponding author. Tel.: +39 06 94005885; fax: +39 06 94005400. E-mail address: lazic@frascati.enea.it (V. Lazic). 0584-8547/$ - see front matter © 2007 Published by Elsevier B.V. doi:10.1016/j.sab.2007.10.019 position. Concentration of the elements in the sample could be also retrieved by applying an appropriate calibration on reference materials, and in some cases, also by Calibration-Free approach. A critical analysis of different quantitative LIBS procedures was recently reviewed in [3]. One of the growing requests for LIBS technique regards in-situ analyses of water solutions, which is important for environmental control [4–7], for monitoring of cooling or waste waters from industry, thermo-electric and nuclear power plants [7–9], then for geological and marine researches [10], for study of chemical reactions in the liquid phase etc. A number of papers reported LIBS analysis of waters in presence of water–air interface, by applying single [5–15] or double-pulse excitation [16]. For most of the measured elements, LOD's obtained by ablation of steady-state 1434 V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 water surface [5–12] are in order of 1–100 ppm, except for calcium, magnesium, chromium and alkali elements, for which somewhat lower detection limits have been achieved. Improvement of the detection limit has been obtained by ablating a surface of liquid jets [12–16], where the splashing effects were reduced, and a detection limit of 0.1 ppb for sodium was reported [16]. However, in some cases direct measurements on bulk water are required, so in the absence of liquid–gas interface. Examples include detection of leakages in industrial plants, measurements of biological activity, determination of nutrients and pollution in deep waters, characterization of sub-glacial waters, etc. Focusing a short laser pulse with sufficient energy into a liquid, a dielectric breakdown takes place generating plasma [17], which further absorbs the incoming part of the laser beam. Rapid heating of the liquid is followed by its explosive expansion and formation of a gas bubble [18]. Intensity of the plasma emission produced in bulk water is generally lower than at water–air interface due to several factors that include: water absorption of the laser and plasma emission and their scattering on suspended particles and micro-bubbles [19], radiation shielding by the high density plasma [20] and fast quenching in the dense medium. Furthermore, the emission lines are strongly broadened by the high electron plasma density [21,22]; all mentioned effects lead to a relatively poor signal in a single pulse LIBS measurements [4,23,24]. Much better analytical performances of underwater LIBS could be obtained by applying a double-pulse laser excitation technique [14,24–28]. In such case, the first laser pulse produces a cavitation bubble in water [18,28], while the second, probing pulse excites the plasma inside the bubble. After the second laser pulse, a relatively intense and narrow spectral emission can be observed due to gaseous state inside the bubble and consequently reduced plasma quenching. As a result, the LOD's below 1 ppm were obtained for some elements directly analyzed from bulk waters [25,27]. An additional LOD improvement (up to one order of magnitude) could be obtained by a proper signal post-processing, as it was recently demonstrated [25, 29]. Another factor to take into consideration for double-pulse LIBS applied on bulk liquids or immersed samples is the timing between the pulses. In Ref. [28], it is reported that the maximum LIBS signal is obtained if the second pulse hits the sample when the gas bubble produced by the first pulse reaches its expansion maximum. For this study, the authors used two Nd:YAG lasers operated at 532 nm, and measured the maximum bubble expansion after 100 μs from the first laser pulse. Dynamic of the laser generated bubbles in liquids was also studied for laser medical applications (Ref. [30] and cited papers) with the aim to avoid an excessive tissue damage, and also in the attempt to obtain efficient tissue ablation or high efficiency of the shock-wave generation (laser lithotripsy). The time evolution and maximum radius of the laser-induced bubble in a certain liquid are strongly dependent on the experimental conditions [30], such as laser wavelength, pulse duration and numerical aperture after the focusing lens. One of the methods for studying the gas bubble dynamic is laser scattering [31,32]. The laser-induced bubbles have a diameter in the order of 0.010–1 mm [17], so the scattering of light in the visible can be described by Mie's theory [31,33] or optical ray geometry. The final scope of this research was to improve the sensitivity of LIBS technique applied on bulk liquids, which we intend to employ for sub-glacial lake exploration. To this aim, we studied the influence of the laser energy, divided in two or more nanosecond pulses, on underwater plasma emission. Dynamics of the lateral gas bubble expansion after the first laser pulse was measured by light scattering techniques and for different laser pulse energies. Shape of the laser-induced plasma was photographed both by conventional photo camera and by an ICCD, after applying one or two laser pulses. The LIBS signal intensity after the second laser pulse was also measured and compared to the results obtained by above mentioned techniques. Influence of low-energy multi-pulse laser sequence on the LIBS signal was also considered. For a few representative laser excitation conditions, calibration graphs were generated for solutions containing magnesium, manganese or chromium and their, LOD's were determined. Different experimental aspects, important for the improvement of the LIBS signal intensity are also discussed. 2. Experimental 2.1. LIBS set-up The underwater plasma was produced by a Q-Switched (QS) Nd:YAG laser operated at 1064 nm, with maximum pulse energy of about 300 mJ and with a repetition rate of 10 Hz. The QS trigger was externally controlled, thus to have a possibility to extract also two laser pulses during single lamp flashing. In such a case, the separation between two laser bursts was typically varied between 50 and 100 μs, which together with the first trigger delay, determines the energy partition between the two pulses. The triggering scheme and examples of pulse energy partitions are reported in our recent papers [25,29]. Two plano-convex lenses, whose equivalent focal length in air was 25 mm, focused the laser beam. The second lens was in contact with the liquid contained in a beaker. Laser beam was fixed at an angle with respect to the vertical position in order to avoid that the gas bubbles, formed by the breakdown and moving upwards, deposit on the focusing lens and further disturb the laser transmission (Fig. 1). The signal was collected by a wide angle optical system, at 60° with respect to the laser beam, and brought to the monochromator (Jobin Yvon 550) by means of a fiber bundle (26 fibers of 0.1 mm diameter) arranged into array 0.1 ×2.6 mm2 at the exit. The chosen monochromator grating has 1200 g/mm, and as a detector, an intensified CCD (ICCD Andor Intraspec) was used. The ICCD was electrically triggered contemporary with the second QS trigger and acquisition gate and delay were properly chosen, as discussed in Section 3.1. For generating of the calibration curves on different solutions, the ICCD gain was settled to value two. 2.2. Scattering measurements The gas bubble produced by the first laser pulse was illuminated by a 35 mW HeNe laser. The scattered signal was measured at 90° using the same collection optics as for LIBS or at 20° in forward direction where the light was collected with a plano-convex lens. In both cases, the full angle aperture for the scattered light collection was larger than 15° in order to minimize angle dependent signal ripples characteristic for Mie 1435 V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 Fig. 1. Experimental lay-out for scattering measurements. scattering [31–33]. The signal was brought to a photomultiplier (PMT) by a 6 mm diameter fiber bundle. Between the fiber bundle exit and the PMT, an interferential filter centered at the HeNe laser emission was placed. The PMT used for the present experiment was a Hamamatsu R928. The high voltage power supply feeded a divider designed for high linearity; the output anode was pre-amplified and AC coupled with a Tektronix 2430A oscilloscope to record light transient. An additional gate circuit was used, which electronically switches off the PMT gain during the time elapsing between consecutive laser shots. Gated operation allowed us to use a CW laser, minimizing problems related with the high mean photon flux incident on the PMT cathode due to light scattering by hydrosols and particles suspended in water. However, this background scattering was always present and the PMT was operated at relatively low voltage (425 V) in order to avoid nonlinearities in its response. Considering that the laser produced gas bubble might be elongated, and in order to measure only its lateral expansion, a 1 mm high slit, with precisely adjustable vertical position, was placed in proximity to the optical window of the beaker wall. Position of laser generated plasma and consequently the bubble center might depend on the laser pulse energy [17,25]. In order to keep the illumination constant across the bubble, a negative lens ( f = − 50 mm) was placed before the slit, which horizontal width was restricted to 5 mm. The power distribution through the slit, scanned vertically in 3 mm range, was checked by a power-meter and resulted uniform within 5%. This range of the slit positions enclosed both the bubble produced by low laser energy pulse (8 mJ) and the one produced at the maximum laser energy (300 mJ, single pulse) (see Fig. 1). 2.3. Samples Calibration graphs for quantitative LIBS analyses were generated on reference solutions, adding properly MgSO4, MnCl2·7H2O or Cr(CH3COH)3 to a pure (milli-Q) water. The corresponding minimum element concentrations were 5 ppm for Mn and Cr, and 1 ppm for Mg. These elements had been studied in our previous work [25] and Mg was chosen as a common component of waters, Mn as an element related to bioactivity in waters and Cr as an example of contaminant. The measurements of the bubble dynamics by the laser scattering technique were performed on tap water, whose standard impurities content was previously determined [26]. In order to avoid disturbances from eventual nanoparticle formation, which for example were clearly observed (clustered) after long measurement sessions in solutions containing MgSO4, water in the beaker was exchanged each 30 min of the measure- ments. During water exchange, the focusing lens was always cleaned to avoid accumulation of the gas bubbles produced by the breakdown. 3. Results 3.1. Calibration for quantitative LIBS analyses For three examined solutions the spectral emission from the analytes (Mg, Mn or Cr) upon double-pulse laser excitation, was acquired at different central monochromator wavelengths. It was found that the most intense lines from these elements correspond to the ionic emission in UV region (see Table 1) although the overall instrument spectral response has its maximum in visible region, where atomic emission of these elements could be found. Such high ionic emission intensities are an indicator of a high plasma temperature, here not measured. In the first run of measurements, the laser energy was settled to its maximum. The most intense plasma emission was obtained when the first QS trigger was delayed for t1 = 155 μs from the flash-lamp trigger and the second pulse delayed for Δt = 75 μs from the first one (we call Δt as an interpulse delay). In this case, the laser pulse energies were E1 = 92 and E2 = 214 mJ. The optimized acquisition gate delay, measured on 5 ppm Mg solution, corresponded to 600 ns from the second laser pulse. The gate width was settled to 600 ns, and its further extension up to 2 μs did not produce appreciable changes of the Signal-to-Noise Ratio (SNR). The same acquisition parameters were also used for the other solutions (Cr and Mn). For each concentration of one solution six replicated measurements were performed by applying 1000 laser shots. The spectra acquired at each laser shot were registered separately into the columns inside a single file. Successively, we applied the data filtering procedure described in our recent paper [25] and here briefly reviewed. The aim of the data filtering is to eliminate the contribution of the plasma continuum, which partially masks the useful signal, from the spectra where the emission from the analytical line is very weak or absent. Such spectra could be often observed in underwater Table 1 LIBS detection limits for two different laser excitations E1 = 92 J E1 = 72 J Element Wavelength (nm) E2 = 214 mJ E2 = 144 mJ Mg Mn Cr Notes 279.5 257.6 283.6 210 ppb 2450 ppb – t1 = 155 μs, Δt = 75 μs Two pre-pulses 34 ppb 390 ppb 920 ppb t1 = 145 μs, Δt = 75 μs Five pre-pulses 1436 V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 LIBS where strong shot-to-shot signal fluctuations are present. A program written under LabView finds a spectrum between 1000 registered spectra, which has the maximum analytical peak intensity Pmax (after background subtraction). Further, the program selects the spectra with the analytical peak intensity above the defined threshold, for example 0.1 Pmax, then sums this spectra and save into a new file to be used for successive analyses (here used for calibration). In this way, the SNR, so of the detection limit, could be improved up to a factor 7, as reported in [25]. After the data filtering, the calibration graphs were generated for two solutions, and the corresponding LOD's were found from 3σ criteria (Table 1). Keeping the QS delays as above and reducing the laser pulse energies to E1 = 37 and E2 = 156 mJ by lowering the current through the laser flash lamp, we measured comparatively the LOD for Mg, and it was slightly higher (0.33 mg/l) than at the maximum energy (0.22 mg/l). However, in the successive experiment we varied QS delays at lower total laser energy and we registered a very strong LIBS signal enhancement for t1 = 145 μs and interpulse Δt = 75 μs. In this case, the laser pulse energies measured by the energy-meter were E1 = 72 and E2 = 144 mJ respectively. For this laser setting, repeating the measurements on the solution containing different Mg concentration we found LOD of 0.034 mg/l, i.e. almost one order of magnitude lower than at the maximum laser energy. One possible explanation for strong plasma emission enhancement after the second laser pulse with mid first pulse energy could be searched in dynamics of the gas bubble expansion after the first laser pulse. In literature, the reported maximum gas bubble expansions in liquids after a nanosecond laser-induced breakdown are in the range 50 μs–300 μs [17], depending on the laser energy and its wavelength, as well as on the focusing conditions (numerical aperture after the focusing lens). Maximum LIBS signal was measured for the second laser pulse arrival when the gas bubble radius is at its maximum [21]. In order to verify whether the LIBS signal enhancement that we had observed is caused by the maximum bubble expansion for Δt = 75 μs at mid first laser pulse energy (65 mJ) we introduced laser scattering measurements as described in following. second laser pulse, the HeNe beam was sent through 1 mm high slit, moved with 1 mm step, and scattered signal was measured at different slit positions. At maximum laser energy here used (280 mJ) the signal could be observed for three slit positions (over 3 mm), while up to 170 mJ the signal is existent for only two slit positions. In all the cases, the same slit position gave the signal maximum and corresponding scattered light intensity as a function of time is depicted in Fig. 2b. The first, narrow peak, correspond to the laser pulse arrival, which produce the plasma continuum emission also in the spectral range transmitted by the interferential filter. In Fig. 2a, the scattering signal obtained without slit is also reported. From this figure, we could observe that the first bubble collapse occurs between 240 μs and 320 μs, followed by the first rebound. At the maximum energy, also the second and weak third rebounds were detected (Fig. 2b). The signal shape in presence of the slit is quite different from the one in the absence of the aperture. With the slit, the signal is more symmetrical, and the highest peak, as well as the longest first collapse period, is observed at a relatively low energy, thus indicating largest lateral bubble expansion (Eq. (1)). In the absence of the slit, the peak value is larger for higher laser pulse energies due to the plasma and hypothesized bubble elongation, so the scattering from the peripheral bubble volume increments the peak value. In such case, also the peak is shifted towards shorter time because the peripheral bubble expansion time is shorter then the central one, as it was confirmed from the measurements with moving the slit. Both for Mie scattering and geometrical optics, the 3.2. Laser scattering measurements For scattering measurements, we initially substituted the plastic beaker with one made of glass. However, the strong reduction in the plasma emission was always observed already after 100 laser shots, due to air bubble deposition on the focusing lens. As such an effect was not observed when using plastic beaker, it might be supposed that the electrostatic charges, induced by the shock waves, were forming in presence of large glass surface. For this reason, again the plastic beaker was placed and modified to have an optical window for the HeNe beam entrance. Dynamics of the laser-induced bubble as a function of the laser energy was monitored by settling a single QS trigger (t1 =210 μs) and by changing the current through the flash lamp. With increasing of the laser energy, the produced plasma becomes more elongated. In order to monitor only the lateral bubble expansion, which we consider more important for the LIBS signal after the arrival of the Fig. 2. Signal from the light scattered at 90° on the gas bubble produced by the laser with pulse energy 65 mJ (dots) and 280 mJ (solid): a) without slit; b) with slit; Tc is measured first collapse period at lower energy; 1–3 indicate the bubble rebounds at the higher laser energy. V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 1437 intensity of the light scattered by the spherical bubble of radius R is proportional to R2 for all scattering angles [33,34]. Consequently, the gas bubble diameter is proportional to the square root of the PMT voltage V [34]. The maximum bubble radius achieved Rmax is proportional to the first collapse period Tc through Rayleigh relation [30]: Rmax ¼ 1:83 Tc pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi q 0 =ð p0 pv Þ ð1Þ where ρ0 =998 kg/m3, p0 is hydrostatic pressure and pv is vapor pressure inside the bubble. The same relationship approximately describes also elongated bubbles, but in this case, Rmax corresponds to the radius of a sphere having the same volume as the elongated bubble [30]. The first collapse period Tc was determined from the intersection of linearly fitted decaying signal on PMT and the final voltage level (see Fig. 2b) [35]. The scattered signal peak Vmax and the first collapse period of the bubble as a function of the laser pulse energy, measured with the slit, are shown in Fig. 3a and b. In both cases, the maximum lateral bubble expansion corresponds to the laser energy of 65 mJ. Up to this energy, Vmax value increases approximately linearly. After achieving the maximum, the gas bubble expansion starts to decrease as the breakdown threshold is reached also out of the focal volume. Fig. 3. Peak of the scattered signal at 90° (a) and the first bubble collapse period (b) as function of the laser pulse energy, measured with the slit. Fig. 4. Scattered signal measured with the slit as a function of time for three sets of the laser pulse energies; QS delay is t1 = 210 μs. Peaks corresponding to the bubbles from new micro-plasma centers inside illuminated volume are indicated by arrows. Therefore, a smaller part of the available optical energy is coupled to the mechanical energy (bubble expansion) in the focal point. The measured values of Vmax and Tc are correlated with factor 0.91. The slightly different behavior between these two parameters could be observed for the laser pulse energies between 65 mJ and 125 mJ, where probably the bubble elongation or multi-bubble formation takes place inside the illuminated section, which corresponds to about 1.5 mm length along the laser axis. The measurements with higher spatial resolution were not possible as the scattered signal becomes too low. On the other hand, further increasing of PMT supply voltage leads to the signal distortion, so it was not applied here. However, there is an evidence of the longitudinal plasma expansion in the illuminated section also from the shape of the scattered signal (Fig. 4). For the energy of 45 mJ, the waveform is symmetrical, thus indicating the bubble sphericity. Increasing the energy to 65 mJ, a weak feature at shorter times appears, that could be attributed to another, much weaker micro-plasma i.e. bubble formed inside the observed volume. Contemporary, the central peak still increased. Further increasing of the laser energy leads to more intense scattered signal at shorter times due to more intense micro-plasma out of focus, and the central peak starts to decrease as discussed above. For laser energies of 170 mJ or higher, two distinct peaks could be observed before the central one, and this can be attributed to another two centers for micro-plasma i.e. bubble formation inside the observed volume. These plasmas are weaker then the plasma in focus and the produced bubbles at these centers decay earlier. The central peak is strongly reduced. From the results reported in this section, we might conclude that for the present focusing conditions, the optimal energy of the first laser pulse, used to prepare LIBS analyses by the second pulse, is about 65 mJ. Further increase of the first pulse energy leads to the LIBS signal deterioration due to the plasma elongation and reduced lateral bubble expansion. These results are consistent the lower measured values of LOD's by LIBS when using the maximum laser energy than when applying E1 = 37 and 1438 V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 3.3. Laser pulse and plasma shape Previously, the laser pulse energies in double excitation were measured by an energy-meter, applying one or two QS triggers. The energy of the pulses were varied by changing the QS trigger delays, where longer first pulse delays correspond to its higher energy and consequently to lower energy of the following pulse. Observing the laser pulse shape by a fast photo-diode, we noticed that for shorter first QS delays with respect to the flashlamp trigger, the relaxation oscillations occur resulting in a multi-pulse sequence (Fig. 5). For t1 = 145 μs or shorter, five pulses are present, whose energies are tending to equalize with reducing of the QS delay. Now, it seems clear that lowering laser pulse energies we had obtained better LIBS signal due to multi-pulse excitation. In the first measurements, we had the first pulse energy of 92 mJ obtained for t1 = 155 μs, so in the presence of a weak secondary pre-pulse. Its energy was estimated to about 13% of the total prepulse energy, and consequently the first pre-pulse had about 80 mJ. In the second experiment, 5 pre-pulses were present with total energy of 72 mJ and the corresponding first pre-pulse energy was about 50 mJ. However, it should be clarified if the observed LIBS signal improvements in the second case were due to: Fig. 5. Shape of the first laser pulse, detected by a fast photo-diode, for different delays of the QS trigger with respect to the flash-lamp trigger. E2 = 156 mJ. In the latter case, the first laser pulse produces the bubble of smaller radius (Fig. 3a), so the LIBS signal is consequently smaller [26]. However, the differences in the lateral bubble expansion between the first laser pulse energies of 72 mJ and 92 mJ are very small (3% according to Tc measurements) and from the scattered signal waveforms they could be even neglected for time interval of 75 μs, used for the second pulse arrival. Consequently, the measured differences in the lateral bubble expansions cannot explain the LIBS signal increase for almost one order of magnitude with the applied energy reduction from 92 mJ and 72 mJ. In [28] the LIBS signal improvement of about factor 10 was measured when hitting the gas bubble having a diameter two orders of magnitude larger and that is not our case. If we compare the second laser pulse energy in the two cases (144 mJ and 210 mJ), it could be hypothesized that more pronounce plasma elongation at higher energy reduces the amount of the energy transmitted to the focal point, i.e. to the center of the bubble produced by the first pulse. Unfortunately, by using a doublepulse technique from single laser source, the energies of the two pulses cannot be varied independently, as required for rigorous measurements of the second pulse influence on the LIBS signal. However, from Fig. 3 an appreciable difference in the energy coupled to the focal volume cannot be observed when the single pulses with analogue energies are applied. Therefore, we further investigated the reason for the LIBS signal improvement at lower laser energies through characterization of the laser pulse form and by photographing the plasma shape. a) larger lateral plasma expansion after low energy multi-pulse sequence b) More localized breakdown (less elongated plasma) c) Short interval between the last pre-pulse and the probing pulse. First, we compared the lateral bubble expansion after the first laser pulse sequence for the two settings from above and for single pulse (t1 =210 μs) with the same energy as the pulse sequence (Fig. 6). Both from the scattered signal intensity and the first collapse period, the lateral bubble expansion results smaller for the lower energy setting (t1 =145 μs, E1 = 72 mJ) than for the higher one (t1 = 155 μs, E1 =92 mJ). In the first case, the intermediate pulse Fig. 6. Scattered signal after the first laser pulse sequence for two settings used in LIBS measurements, and comparative signal measured in presence of single pulse (t1 = 210 μs) with equivalent energy. V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 Fig. 7. Plasma shape photographed by a digital camera: a) t1 = 145 μs, Δt = 80 μs, 5 pre-pulses, ETot 1 = 72 mJ, E2 = 140 mJ; b) t1 = 155 μs, Δt = 80 μs, 2 pre-pulses; ETot 1 = 110 mJ, E2 = 123 mJ; c) t1 = 165 μs, Δt = 80 μs, 1 pre-pulse, E1 = 181 mJ, E2 = 57 mJ. occurring about 50 μs from the first one, clearly increases the slope of the bubble expansion. For the interval used for the second laser pulse arrival (Δt= 75 μs) the scattered light intensity approximately reaches that one obtained at higher energy setting. In the latter case, a portion of the scattered signal for this timing also comes from the bubble formed out of the focus, as discussed earlier. We might conclude that the splitting of the pulse into sequence of the pulses with the same total energy does not improve, or at least — not significantly, the lateral bubble expansion in the cases here examined. Consequently, the gas bubble dynamics is not responsible for the observed LIBS signal improvement. 1439 Observing photographs of the underwater plasma, taken with simple digital camera in the presence also of the second laser pulse, we may note that increasing the first laser pulse (or pulse sequence) energy, the secondary plasma changes shape from the spherical to elliptical (Fig. 7a and b). Further increase of the first pulse energy leads to a clear formation of the secondary plasma in multiple sites (Fig. 7c). The presence of multiple plasmas after the second pulse was also photographed by the ICCD camera, triggered contemporary with the last laser pulse and using acquisition gate of 10 μs. In Fig. 8a the secondary plasma produced after low-energy multipulse sequence followed by the energetic pulse, has a spherical shape with a relatively high intensity (analogue to Fig. 7a). Increasing the pre-pulse energy closely spaced spherical plasmas are produced (Fig. 8b) after the second shot, but they loose in intensity and in volume (analogue to Fig. 7b). Further increase of the pre-pulse energy causes gas bubble elongation or multi-bubble formation along the laser path. Then, after the second (probing) pulse, the distance between new formed plasma sites might increase (Fig. 8c) and intensity of the plasma formed farther from the focusing lens is consequently reduced (similar to Fig. 7c). An excessive energy of the pre-pulses causes not only the strong plasma elongation but also its shift towards the focusing lens (Fig. 8d). The secondary plasmas are again formed in multiple sites but they are no more symmetrical and have weak intensities. If we look the LIBS signal and laser pulse energy, measured as a function of the interpulse delay (Fig. 9), the steep increase of the LIBS signal accumulated over 200 laser shots, is caused by the laser pulse instabilities present for Δtb 68 μs. Above this value of interpulse delay, initially both the line and plasma continuum emission follow the energy oscillations caused by the current switching through the flash lamp (period about 8 μs). However, increasing the interpulse delay, and in this way also the interval from the last pre-pulse, the plasma intensity decays faster than the second pulse energy although the bubble radius before the second pulse arrival still increases (Fig. 6) and this should increment the Fig. 8. Plasma images taken by the ICCD after the second laser pulse, with integration time 10 μs; the laser focusing lens is on the left side. The energy of the first laser pulse (multi-pulse) is progressively increased from (a) to (d). Data for the pulse energies are not available. 1440 V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 Fig. 9. Plasma continuum and Mg+ (279 nm) peak emission multiplied 2× (left), and second laser pulse energy (right) a function of the interpulse delay. The first QS trigger delay is t1 = 145 μs. LIBS signal. All these indicate that a relatively short timing between the last pre-pulse and the second laser pulse might be a reason for an additional LIBS signal enhancement. In a pioneering work relative to double-pulse LIBS underwater [24], two Nd:YAG laser at 1064 nm were used with similar energies (E1 = 30–76 mJ, E2 = 125 mJ) and in analogue focusing conditions as in our experiment. The measured LIBS signal as a function of interpulse delay had a maximum for Δt = 18 μs and this cannot be attributed to the maximum bubble expansion caused by the first laser pulse, which generally occurs after more than 100 μs (in our measurements — after 120–160 μs). For interpulse delay in order of 100 μs, the measured signal was more than 6 times lower with respect to the optimal delay (Δt = 18 μs). Also in a recent publication [28], the LIBS signal intensity obtained after the second laser pulse when using two laser sources at 532 nm, does not follow the gas bubble radius for small interpulse delays, but starts to increase with shortening Δt bellow 40 μs. This indicates that a relatively short timing between two pulses, in our case obtained by a sequence of the pre-pulses, may be another reason for observed plasma emission enhancement after lowering the laser pulse energies. In our experiment, using single laser source, it was not possible to vary the interpulse delay independently from the laser pulse energies. However, by reducing the current through the flash lamp for the first QS trigger delay t1 =155 μs, and helped by the laser energy oscillations due the current switching, for Δt= 73 μs we found the same set of the total pre-pulse and the probing pulse energies as in the case of t1 = 145 μs, Δt=74 μs. In the latter case, the observed much stronger LIBS signal (Fig. 10) could be attributed only to the presence of multi-pulse sequence shown in Fig. 5 as the differences in the bubble expansion could be neglected in correspondence of the probing pulse arrival (Fig. 6). Moreover, Fig. 10 shows an increased continuum emission, thus indicating LIBS plasma having a density higher than that observed in case of a simple double-pulse excitation. We hypothesize that the observed, almost tenfold increase of SNR (Fig. 10a) is due to increased absorption coefficient of the vapor bubble after its excitation by the last pre-pulse. Consequently, it could be deduced that the presence of a pre-pulse sufficiently close to the probing one (here the distance between the last pre-pulse and the probing one was about 5 μs) maintains the plasma inside the bubble. Then, the last laser pulse is coupled better with the bubble and a much stronger plasma emission (i.e. LIBS signal) is so obtained. 4. Conclusions Fig. 10. LIBS signal accumulated over 100 shots for gate delay zero from the probing pulse and gate width 10 μs. The pre-pulse and the second pulse energies are equal (E1 = 72 mJ, E2 = 128 mJ) in both cases (t1 = 145 μs and t1 = 155 μs). Mg concentration is 5 ppm. In this paper, we studied the causes for a tenfold LIBS signal improvement underwater that we had obtained by lowering the laser pulse energies. The studies were performed by analyzing LIBS signals, laser pulse energies and shapes, then by applying scattering measurements to study the gas bubble dynamics and by taking photographs of the plasma. V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 Laser produced plasma underwater exhibits elongation above a certain energy threshold, in our case at about 65 mJ, which also represents a limiting value for the maximum lateral gas bubble expansion. Above this limit, the plasma is elongated and generates the gas bubbles in multiple sites and with smaller diameters. Hitting the laser bubble with an appropriately timed second laser pulse produces a relatively strong secondary plasma, so the LIBS signal. This signal is maximal if the primary plasma (i.e. the gas bubble) was produced in a single site and with the first pulse energy corresponding to the maximum lateral bubble expansion. In such case, the secondary plasma formed on the seeding bubble, is strongly symmetric and intense. Further increase of the first laser pulse energy leads to the formation of spherical secondary plasmas in multiple sites, seeded on the previously formed bubbles. The overall secondary plasma emission is then reduced as the plasma formed closer to the focusing lens partially absorb the second laser pulse, which radiation is also scattered from there present bubbles. Both effects are responsible for decrease of the optical coupling to the largest bubble present in the focal point. Excessive first laser pulse energy leads to an intense breakdown before the focal point and to a number of very small bubbles, so the second laser pulse cannot be coupled in proximity of the focal point both due to the light absorption and to scattering by the bubbles formed closer to the focusing lens. The resulting secondary micro-plasmas are very weak and asymmetrical and the LIBS signal is very weak or even absent. Besides keeping the first laser pulse energy close to the plasma elongation limit, an appropriate splitting of the first laser pulse into multi-pulse sequence, here obtained by inducing the relaxation oscillations after the first QS aperture, could further improve the LIBS signal. The first, most intense pulse of the sequence produces well-localized gas bubble, whose radius is further expanded by the successive pulses of the sequence. In this way, still keeping the bubble expansion high, the plasma elongation is avoided and the second pulse is well-coupled in the focal point. We also demonstrated that a relatively short timing between the last pulse of the sequence and the second, probing pulse, also contributes to a manifold plasma intensity, therefore LIBS signal enhancement. For optimized laser excitation by low-energy multi-pulses, and by performing the spectra filtering procedure [25], we obtained the next detection limits for LIBS applied on bulk liquids: 0.034 ppm for Mg, 0.38 ppm for Mn and 0.92 ppm for Cr. These detection limits could be further improved by using two separate laser sources, the first one operating in a multipulse regime, in order to hit the gas bubble by the second laser when it reaches the expansion maximum. The latter occurs at 120–160 μs from the first pulse according to the scattering measurements. Strong improvement of underwater LIBS signal by limiting the laser pulse energies has also an importance in the system miniaturization in view of real in-situ analyses. Further research would address the influence of the second laser pulse energy on the LIBS signal by employing of two independent laser sources. Studies of the increased, resonantlike laser absorption by weak plasma inside the gas bubble, here 1441 excited by low-energy intermediate pulses, should be also performed. Acknowledgments The authors gratefully acknowledge Dr. Valeria Spizzichino (ENEA fellowship) for assistance in ICCD imaging and to Isabelle Rauschenbach (University of Munster, Germany) for photographs by a digital camera. Special thanks to Dr. Antonio Palucci (ENEA Frascati) for assistance with the laboratory equipment. The research has been conducted within the projects MIAO (sensor micro-systems for application in hostile environments), and supported also within the PANDORA project (sub-glacial lakes exploration) conducted within the Technology Sector of PNRA (Italian National Program of Researches in Antarctica). Both the projects are funded by Italian Ministry for Scientific Research (MUR). References [1] J.D. Winefordner, I.B. Gornushkin, T. Correll, E. Gibb, B.W. Smith, N. Omenetto, Comparing several atomic spectrometric methods to the super stars: special emphasis on laser induced breakdown spectrometry, LIBS, a future super star, J. Anal. At. Spectrom. 19 (2004) 1061–1083. [2] L.J. Radziemski, From LASER to LIBS, the path of technology development, Spectrochim. Acta Part B 57 (2002) 1109–1113. [3] V. Lazic, L. Caneve, F. Colao, R. Fantoni, L. Fornarini, V. Spizzichino, Quantitative elemental analyses of archaeological materials by laser induced breakdown spectroscopy (LIBS) — an overview, SPIE Vol. 5857 (2005), special issue on “Optical Methods for Arts and Archaeology (Eds. R. Salimbeni and L. Pezzati), p. 58570H 1–12. [4] S. Koch, W. Garen, M. Müller, W. Neu, Detection of chromium in liquids by laser induced breakdown spectroscopy (LIBS), Appl. Phys. A 79 (2004) 1071–1073. [5] L.M. Berman, P.J. Wolf, Laser-induced breakdown spectroscopy of liquids: aqueous solutions of nickel and chlorinated hydrocarbons, Appl. Spectrosc. 53 (1998) 438–442. [6] G. Arca, A. Ciucci, V. Palleschi, S. Rastelli, E. Tognoni, Trace element analysis in water by the laser-induced breakdown spectroscopy technique, Appl. Spectrosc. 51 (1997) 1102–1105. [7] O. Samek, D.C.S. Beddows, J. Kaiser, S.V. Kukhlevsky, M. Liška, H.H. Telle, J. Young, Application of laser-induced breakdown spectroscopy to in situ analysis of liquid samples, Opt. Eng. 39 (2000) 2248–2262. [8] P. Fichet, P. Mauchien, J.F. Wagner, C. Moulin, Quantitative elemental determination in water and oil by laser induced breakdown spectroscopy, Anal. Chim. Acta 429 (2001) 269–278. [9] J. Wachter, D.A. Cremers, Determination of uranium in solution using laser-induced breakdown spectroscopy, Appl. Spectrosc. 41 (1987) 1042–1048. [10] C. Fabre, M.C. Boiron, J. Dubessy, M. Cathelineau, D.A. Banks, Palaeofluid chemistry of a single event: a bulk and in-situ multi-technique analyses (LIBS, Raman Spectroscopy) of an Alpine fluid (Mont-Blanc), Chem. Geol. 182 (2002) 249–264. [11] B. Charfi, M.A. Harith, Panoramic laser-induced breakdown spectrometry of water, Spectrochim. Acta Part B 57 (2002) 1141–1153. [12] R. Knopp, F.J. Scherbaum, J.I. Kim, Laser induced breakdown spectroscopy (LIBS) as an analytical tool for the detection of metal ions in aqueous solutions, Fresenius J. Anal. Chem. 355 (1996) 16–20. [13] P. Yaroschyk, R.J.S. Morrison, D. Body, B.L. Chadwick, Quantitative determination of wear metal in engine oils using laser-induced breakdown spectroscopy: a comparison between liquid jets and static liquids, Spectrochim. Acta Part B 60 (2005) 986–992. 1442 V. Lazic et al. / Spectrochimica Acta Part B 62 (2007) 1433–1442 [14] L. St-Onge, E. Kwong, M. Sabsabi, E.B. Vadas, Rapid analysis of liquid formulations containing sodium chloride using laser-induced breakdown spectroscopy, J. Pharm. Biomed. Anal. 36 (2004) 277–284. [15] C.W. NG, F. Ho, N.H. Cheung, Spectrochemical analyses of liquids using laser-induced plasma emission: effects of laser wavelength on plasma properties, Appl. Spectrosc. 51 (1997) 976–983. [16] A. Kuwako, Y. Uchida, K. Maeda, Supersensitive detection of sodium in water with use of dual pulse laser-induced breakdown spectroscopy, Appl. Opt. 42 (2003) 6052–6056. [17] P.K. Kennedy, D.X. Hammer, B.A. Rockwell, Laser-induced breakdown in aqueous media, Prog. Quantum Electron. 21 (1997) 155–248. [18] H. Sobral, M. Villagran-Muniz, Temporal evolution of the shock wave and hot core air in laser induced plasma, Appl. Phys. Lett. 77 (2000) 3158–3160. [19] N.F. Bunkin, A.V. Lobeyev, Influence of dissolved gas on optical breakdown and small-angle scattering of light in liquids, Phys. Lett. A 229 (1997) 327–333. [20] D.X. Hammer, E.D. Jansen, M. Frenz, G.D. Noojin, R.J. Thomas, J. Noack, A. Vogel, B.A. Rockwell, A.J. Welch, Shielding properties of laser induced breakdown in water for pulse durations from 5 ns to 125 fs, Appl. Opt. 36 (1997) 5630–5640. [21] A. De Giacomo, M. dell'Aglio, F. Colao, R. Fantoni, Double pulse laser produced plasma on metallic target in seawater: basic aspects and analytical approach, Spectrochim. Acta Part B 59 (2004) 1431–1438. [22] A. Escarguel, B. Ferhat, A. Lesage, J. Richou, A single laser spark in aqueous medium, J. Quant. Spectrosc. Radiat. Transfer 64 (2000) 353–361. [23] R. Knopp, F.J. Scherbaum, J.I. Kim, Laser induced breakdown spectroscopy (LIBS) as an analytical tool for the detection of metal ions in aqueous solutions, Fresenius J Anal. Chem. 355 (1996) 16–20. [24] D.A. Cremers, L.J. Radziemski, T.R. Loree, Spectrochemical analyses of liquids using the laser spark, Appl. Spectrosc. 38 (1984) 721–729. [25] V. Lazic, F. Colao, R. Fantoni, V. Spizzichino, Laser Induced Breakdown Spectroscopy in water: improvement of the detection threshold by signal processing, Spectrochim. Acta Part B 60 (2005) 1002–1013. [26] A. De Giacomo, M. Dell'Aglio, F. Colao, R. Fantoni, V. Lazic, Double-pulse LIBS in water bulk and on submerged bronze samples, Appl. Surf. Sci. 247 (2005) 157–162. [27] W. Pearman, J. Scaffidi, S.M. Angel, Dual-pulse laser-induced breakdown spectroscopy in bulk aqueous solution with an orthogonal beam geometry, Appl. Opt. 42 (2003) 6085–6093. [28] A. Casavola, A. De Giacomo, M. Dell'Aglio, F. Taccagna, G. Colonna, O. De Pascale, S. Longo, Experimental investigation an modeling of double pulse laser induced plasma spectroscopy under water, Spectrochim. Acta Part B 60 (2005) 975–985. [29] V. Lazic, F. Colao, R. Fantoni, V. Spizzichino, Recognition of archeological materials underwater by laser induced breakdown spectroscopy, Spectrochim. Acta Part B 60 (2005) 1014–1024. [30] A. Vogel, J. Noack, K. Nahen, D. Theisen, S. Busch, U. Parlitz, D.X. Hammer, G.D. Noojin, B.A. Rockwell, R. Birngruber, Energy balance of optical breakdown in water at nanosecond to femtosecond time scales, Appl. Phys. B 68 (1999) 271–280. [31] R.G. Holt, L.A. Crum, Mie scattering used to determine spherical bubble oscillations, Appl. Opt. 29 (1990) 4182–4191. [32] W.J. Lentz, A.A. Atcheley, D.F. Gaitan, Mie scattering from a sonoluminescing air bubble in water, Appl. Opt. 34 (1995) 2648–2654. [33] M. Kerker, The scattering of light and other electromagnetic radiation, Academic, New York, 1969. [34] B.P. Barber, S.J. Putterman, Light scattering measurements of repetitive supersonic implosion of a sonoluminescing bubble, Phys. Rev. Lett. 69 (1992) 3839–3842. [35] P. Hubert, K. Jöchle, J. Debus, Influence of shock wave pressure amplitude and pulse repetition frequency on the lifespan, size and number of transient cavities in the field of an electromagnetic lithotripter, Phys. Med. Biol. 43 (1998) 3113–3128.