Abstract
We studied changes in laser-induced breakdown spectroscopy (LIBS) signal intensity with the thickness of a liquid layer placed on a solid substrate, where an easily evaporating methanol sample was used. For a certain optimal liquid film thickness we obtained a manifold increase of the LIBS signal from methanol. Progressive liquid film thinning leads to a reduction and a successive disappearance of laser-induced splashes; the latter condition drastically reduces the sample consumption and allows measurements to be repeated many times on a single liquid droplet. In following, we developed two methods for actively controlled deformation, i.e., thinning of a liquid droplet (volume ∼10 µl) prior to its sampling by LIBS. Control of the droplet’s height was achieved on a Si–SiO2 wafer substrate by electro-wetting in the case of water solutions or by target rotation in the case of viscous liquids. The chosen substrate also has the advantages of low cost, easy manipulation, and very high purity, thus minimizing interference with analytes. Through the droplet deformation, in a single-pulse excitation at moderate laser energy (70 mJ), we clearly detected Fe and Mn in peanut oil, which represent trace elements in edible oils (∼ 1 part per billion), according to results published in the literature.
Introduction
Laser-induced breakdown spectroscopy (LIBS) enables detection of elements present in a sample by forming and exciting a plasma with short intense laser pulses.1–4 This analytical technique rapidly evolved over the past two decades 3 and its advantages 4 include measurement capability in any environment and on any material type (solid, liquid, gas, aerosols); contactless measurements, which are feasible in situ and in real time; and moderate costs of the instruments, which might have a portable format. 5 Another advantage of LIBS with respect to conventional analytical measurements is the direct determination of some elements with minimal or no sample preparation; in the case of liquid analysis, complex sample digestion is not necessary.
Elemental analysis of liquids by LIBS is of great interest for different applications, including: monitoring of water quality,6–10 pharmaceuticals formulation, 11 medical diagnostics,12–14 applications in the food industry,15,16 and monitoring of nuclear plants17–19 and different industrial processes or mechanical equipment.8,20–22 Unfortunately, laser-driven plasma formation and excitation on or inside liquids are highly inefficient processes because a large portion of the input laser energy is lost into liquid vaporization and mechanical effects (shockwaves, bubbling, expulsion of droplets). 23 Underwater plasma usually requires dual-pulse 24 or multi-pulse 25 laser excitation to obtain useful LIBS spectra.
In laser ablation of a free liquid surface, achieved by a single-pulse laser excitation, the limits of detection are of the order of 0.1–10 parts per million (ppm); 23 here, the sample volume (>10 mL) is rapidly consumed by splashes, which also limit the laser repetition rate to 2 Hz maximum and require long focusing distances. The splashes can be reduced by sampling a frozen water solution,26,27 where the LIBS signal is strongly dependent on the ice temperature, 27 by ablating a slowly flowing liquid7,11 or a liquid jet.12,20,22 The last two approaches significantly improve the LIBS signal intensity and repeatability, but they require large sample volumes (of the order of 100 mL) and washing of the pumping circuit before analyzing another liquid.
For small available sample volumes (<1 mL), LIBS measurements could be performed on falling droplets28,29 or aerosols. 29 Breakdown formation on a falling droplet causes its disintegration, so repetition of the measurement requires a supply of new droplets. A complex triggering system becomes necessary to hit the droplet precisely, as the droplet often deviates from the focal volume; the obtained limit of detection is of the order of 1–10 ppm. Laser-induced breakdown spectroscopy measurements on liquid aerosols suffer from the light scattering from small droplets and require an exhausting system; the measured limits of detection are of the order of 1 ppm. In both sampling methods, the circuit must be washed before a new sample can be introduced.
Another experimental approach consists of placing an evaporating liquid (e.g., water) on a solid substrate and probing its residues after drying the sample close to room temperature,9,16–18,30,31 the latter condition being necessary to prevent losses of volatile components. This process is time consuming and produces a non-uniform residue distribution, which affects element quantification. The uniformity of residue distribution obtained from water droplets could be improved by increasing the substrate’s wettability, 32 as here demonstrated by employing a laser-patterned silicon wafer. Alternatively, the LIBS sensitivity on residues from water can be enhanced a few times by applying dispersive liquid–liquid microextraction followed by sample drying in an oven, 31 by metal precipitation and membrane separation, 33 or by chemical replacement of heavy metal ions on magnesium alloy. 34 These procedures involve manipulation of toxic reagents and the time-consuming sample preparation should be carried out in well-controlled laboratory conditions, which drastically reduce the advantages of LIBS (simplicity, rapidity, in-field measurements) with respect to other analytical laboratory techniques.
Unlike water-based solutions, oily liquids could not be evaporated on a solid substrate before analyzing them by LIBS. For small volumes of oil samples, a good LIBS analysis performance was obtained after smearing the liquid manually on an aluminum support; 21 however, the liquid thickness cannot be controlled in this case and the signal is affected by impurities present in the substrate.
Owing to the remarkable difficulty in obtaining highly sensitive LIBS analysis of liquids in small volumes, in this work we explore the possibility of sampling water and oil droplets on a solid substrate by changing their heights. The breakdown initiation on a solid substrate reduces the strong matrix effect characteristic of the plasma formation on liquids, 23 while active control of the droplet thickness enables optimization of the signal’s intensity and reproducibility. The final scope of this study is to achieve rapid and sensitive LIBS analysis of liquids in micro-volumes by employing a compact instrument with low operative costs.
Experimental
Instrumentation
The plasma was generated by a Nd:YAG laser (Quantel, CFR Ultra) emitting 6.5 ns long pulses at 1064 nm; applied laser energies were in the range 20–70 mJ. The beam was focused by a quartz lens, f = 100 mm, normal to a horizontally placed target, positioned slightly above the focal plane. The corresponding spot size, which was measured on a silica wafer using dimensionally calibrated images, was 0.42 mm, so the maximum applied laser fluency (at an energy of 70 mJ) was 54 ± 5 J/cm2. The plasma emission was collected at an angle of about 60° from the target plane by two quartz lenses (focal lengths, 100 mm and 70 mm) and brought to a spectrometer (Jobin-Yvon 550) by a fiber bundle arranged at the exit in a 0.1 × 12 mm2 array. At the spectrometer output an intensified charge-coupled device (ICCD) (Andor iStar DH734), was used for LIBS signal detection. The acquisition gate and delay from the laser pulse were adjusted by a delay generator (Quantum Composer 9600+) to values of 5 µs and 1 µs, respectively. Measurements were made at a laser repetition rate limited to 1 Hz in order to allow stabilization of the liquid surface before arrival of the successive probing pulse. Images of the plasma and of a liquid droplet or film were taken by a commercial photographic camera. The droplet height and area on the substrate were determined from dimensionally calibrated photographs with the help of ImageJ software (free source). The average liquid thickness on the substrate was calculated from the measured area occupied by the droplet and the known sample volume.
Samples
The liquids analyzed were methanol (>99.9%), then a solution prepared from bi-distilled water, and containing Cr at a concentration of 100 ppm, and a commercial peanut oil. The droplets placed at the substrate had precisely controlled volume (±1%), delivered by an autoclavable pipette (Labgene Scientific).
The substrates used were aluminum for methanol, and a 0.5 mm thick P-type silica wafer covered with a 285 nm-thick SiO2 layer (Graphene Supermarket, W-5 P-300) for water and oils. A water droplet was deformed into a film by electro-wetting on dielectric, 31 where one electrode was represented by an electrolytic liquid in contact with the wafer’s back plane and the other was a bunch of thin copper wires placed in contact with the liquid. The oil droplet was simply spread on the wafer by rotating the substrate on a spin-coater (Laurell Technologies, KL-SCI-20), whose speed was regulated between 18 and 150 rotations per minute (rpm). After turning on the spin-coater, we waited at least 30 s to stabilize first the rotation speed and then the liquid thickness, before starting the LIBS measurements.
Results and discussion
Effects of the Liquid Thickness on the Signal
Initially, to study the LIBS signal variation as a function of droplet height, we selected a methanol sample because it rapidly evaporates at room temperature, thus reducing the liquid thickness on the substrate. Methanol has a relatively low surface tension (22.5 mN/m at 20 ℃)
35
and a droplet of 20 µL methanol, just placed on aluminum, has an initial height of 0.72 ± 0.05 mm. During the LIBS measurements, the droplet volume and height are reduced, owing to laser-induced vaporization and splashes, and because of natural evaporation. The first laser shot, with a pulse energy of 40 mJ, did not produce any detectable line emission; the corresponding photograph (Figure 1a) shows only the expulsion of a large mass of liquid. The next few pulses generated the emission lines related to the liquid; photographs reveal the ejection of small liquid droplets (Figure 1b). From the seventh shot, ablation of the support occurs, as recognized from the presence of Al(I) lines, whose signal-to-noise ratio (SNR) jumps well above the detection limit (Figure 2a). For both C(I) and Al(I) lines, the maximum intensity occurs after the eighth pulse, and the corresponding photograph (Figure 1c) shows a very intense and uniform plasma. The successive laser pulses keep ablating the solid target, while the emission from the liquid progressively disappears until it is fully evaporated.
LIBS spectra from methanol on aluminum after (a) one, (b) four, and (c) eight sequential laser shots with energy of 40 mJ. The corresponding plasma photographs are shown on the right. Signal-to-noise ratio of C(I) (247.7 nm) and Al(I) (255 nm) lines as a function of the shot number on methanol at aluminum substrate; laser energy is (a) 40 mJ and (b) 20 mJ.
Analog measurements were performed for half the laser energy (20 mJ); the resulting emission lines are shown in Figure 2b. In this case, the emission from the support occurs earlier (after the fifth shot) and the lines reach maximum intensity after the 14th pulse, where the SNR of C(I) line is about three times less than the maximum achieved with 40 mJ pulses.
In Figure 2, it can be seen that the line intensity fluctuations are more severe at the lower laser energy, owing to the statistical nature of plasma formation in liquids. 23 Here, target ablation also starts well before the maximum C(I) emission is reached (Figure 2b) and the consequently formed nanoparticles, trapped inside the liquid, might affect the plasma formation threshold. At the higher laser energy (Figure 2a), apparently free of the target ablation during the first laser shots, the emission from C(I) changes smoothly. In this case, the LIBS signal from the support material becomes very stable after about 20 laser pulses, attributed to a disappearance of the liquid film or of small residual droplets from the target surface.
From the obtained results, we might conclude that thinning of a liquid droplet or film on a solid support increases the LIBS signal from the liquid itself, reaching a maximum for a certain thickness depending on the laser energy, where support ablation is also present. An excessive thinning of the liquid film reduces the LIBS signal because less sample material is available in the focal volume. At the higher laser energy (40 mJ), starting from the second laser shot, the C(I) line was well above the detection threshold and there were no emission lines from the target (SNR < 3), as shown in Figure 2a. Conversely, the intensity of the lower laser energy (Figure 2b) at the Al(I) line becomes stronger than that of the C(I) line after a few laser shots and, until the liquid is evaporated, it has absolute values higher than that of the excitation obtained by the 40 mJ pulses. These findings indicate that at higher laser energy the leading part of the pulse ionizes the liquid layer more efficiently so the initial plasma screens the substrate from ablation by the remnant part of the laser pulse more effectively. Besides a large LIBS signal intensity from an optimally thinned liquid layer, also when applying moderate laser pulse energies, there is another important advantage of the proposed experimental approach: liquid splashes are eliminated. In such conditions, sample consumption during the LIBS measurements is drastically reduced, thus enabling the acquisition of a large number of spectra from a small liquid volume; simultaneously, it is possible to reduce the lens focal length, i.e., to create an analytical plasma using low energy laser pulses and without cleaning the optics after each laser shot. The drawback of the method lies in the presence of emission lines from the support material, which might interfere with the analytical lines from the liquid or prevent the measurement of some trace elements also present in the substrate. 30 For this reason, in the following we have chosen a silica wafer as substrate because it is available at high guaranteed purity and relatively low cost, considering that the substrate area necessary per sample is less than 1 cm2.
Sampling of Water Droplets
Placing 20 µL of the water solution on the wafer, the initial droplet is very thick: its central height is 1.7 ± 0.1 mm because of a very large surface tension of water (73.9 mN/m at 20 ℃).
36
Applying a laser shot with energy of 40 mJ or 70 mJ, the water droplet fully disintegrates; the splashes reach nearby optical elements so they must be cleaned. The corresponding LIBS signal is very weak and the most intense Cr(I) line (425.4 nm) obtained by a pulse of 70 mJ on the solution containing 100 ppm of Cr, has a SNR of only 8 (Figure 3a). Under 40 mJ pulse this Cr(I) line was at the detection limit. In the following, we applied a pulse energy of 70 mJ.
Single-shot spectra from water droplet containing 100 ppm of Cr, placed on the wafer: (a) in the absence of an external voltage and with voltages of (b) 23.5 VDC, (c) 26 VDC, and (d) 30 VDC. Cr(I) lines from water are indicated on the bottom graph.
Switching on the voltage to produce the electro-wetting, a visible modification of the droplet shape occurs, starting from 23.5 VDC. For this voltage, the droplet’s contact area with the substrate enlarges from 17.1 mm2 to 23.4 mm2, equivalent to a change in the average droplet thickness from 1.17 mm to 0.85 mm; in the last case, the central droplet height was 1.3 ± 0.1 mm. Now, the detected LIBS spectrum also shows a broad Hγ emission (Figure 3b) while the SNR of the considered Cr(I) line is slightly improved (SNR = 12); simultaneously, the splashes are reduced but they are still present. Increasing the voltage to 26 VDC and then to 30 VDC, the splashes almost disappear and the SNR of the Cr(I) line reaches values of 36 and 68 (Figure 3c and d), respectively. In the last case, the droplet was deformed into a film occupying approximately 54 mm2, with an average thickness of about 0.37 mm. We did not observe any improvement of the LIBS signal if increasing the voltage further. We also checked the Cr(II) emission in the range 270–290 nm for the optimal voltage of 30 VDC, observing many well detectable lines (Figure 4) up to an excitation level of 10.3 eV. Here, the most prominent line, at 283.6 nm, had a SNR as high as 296. Based on the detected Cr(II) lines, the temperature inferred from the Boltzmann’s plot is very high, about 11 000 K.
Comparison between the experimentally obtained spectrum at a voltage of 30 VDC on a water droplet containing chromium, and theoretical Cr(II) lines from the NIST database.
These results demonstrate that application of electro-wetting prevents the laser-induced explosion of a water droplet and so allows repeated LIBS measurements on a single droplet with a manifold signal enhancement. It is interesting to note that once the droplet was exposed to the voltage (electro-wetted) and consequently flattened on the substrate, its shape remained unaltered after the contact electrode was removed and the voltage was turned off. The shape of the previously electro-wetted droplet remains constant over a long period (>10 min), sufficient to perform LIBS measurements without risk of signal contamination by ablation of the electrode.
Owing to a relatively large free surface of the evaporating water film, a slow progressive increase of the analyte concentrations might be expected. For this reason, timing between the sample preparation and the measurements should be reasonably controlled. Moreover, each laser shot ablates a small amount of the sample, after which the liquid surface is equilibrated by cohesive forces. Both evaporation processes lead to a reduction of the liquid film thickness during the LIBS measurements, which is slower for larger sample volumes at the equivalent initial thickness. For these reasons, in quantitative LIBS analysis of water films on a substrate, it would be opportune not to accumulate the LIBS signal but to register a series of single-shot spectra in order to control their intensity changes with time. Only the spectra with similar characteristics should be averaged to improve the SNR of the analytical lines. Here, the intensities of hydrogen lines (coming from water) or Si(I) lines (coming from the support ablation), both dependent on the liquid film thickness, could be used to monitor the evaporation of the liquid. Simultaneously, a limit for the progressive change of the analyte signal due to pre-concentration should be set, so defining a number of the laser shots (spectra) to consider.
Sampling of Oil Droplets
Another class of liquids considered here are oils, which are characterized by higher viscosity and lower surface tension than water. For example, the surface tension of peanut oil at 20 ℃ is 10.0 mN/m,
37
seven times less than that of water. Placing a 20 µL droplet of peanut oil on the wafer, the maximum liquid thickness is 0.51 ± 0.05 mm. When sending a laser pulse of 70 mJ, the splashes reach the nearby optic elements and the LIBS signal is very weak. To transform the oil droplet into a thin liquid film, we placed the substrate on the spin-coater. At the minimum rotation speed (18 rpm), the oil droplet spreads into a film with an average thickness of about 0.1 mm. Applying the laser pulse to a liquid film obtained in this way, we did not observe splashes but only the formation of a liquid aerosol (spraying). The intensity of the C(I) line at 247.7 nm increases with the rotation speeds up to 40 rpm (data not shown) and later remains constant; simultaneously, ablation of the support material still increases, as observed from the Si(I) lines at 256.8 nm. At the optimal rotation speed (40 rpm) relative to the LIBS signal from the oil, the average film thickness on the substrate is about 60 µm. In this condition, we acquired 30 consecutive spectra during sample rotation, accompanied by a slow shift (≈0.5 mm/s) of the whole spin-coater. Owing to a high plasma temperature generated on a thin liquid film, we might expect a large presence of ionic species. Here, even in the single-shot spectra, we detected ionic lines from Mn and Fe while the corresponding atomic lines, checked in the spectral range 400–425 nm, were very weak. Figure 5 shows a comparison of the spectra obtained for the clean wafer and in the presence of the oil rotated at 40 r/min. It is evident that Si(I) emission is much lower in the second case; this means that the interference of the substrate on LIBS signal from analytes is reduced.
(a) Single-shot spectrum from the wafer support, and (b) peanut oil on the wafer rotated at the speed of 40 rpm; laser energy is 70 mJ.
With the proposed simple experimental approach, it was possible to acquire at least 30 intense LIBS spectra from a single oil droplet without disintegrating it by the laser pulses and without having to clean nearby optical elements, owing to the absence of splashes. The well-detected lines from Mn and Fe in the peanut oil indicate that a high analytical sensitivity is reached, since these elements are usually present in edible oils at a concentration of 1–22 parts per billion (ppb) and 7–35 ppb, respectively. 38 However, the real detection limits of the method should be evaluated after calibration on certified oils and after corresponding error analysis.
Conclusion
Laser-induced breakdown spectroscopy sampling of a thin liquid film spread on a solid support leads to an intense high-temperature plasma, also when applying a single-pulse excitation at moderate laser energies. Simultaneously, laser-induced droplet explosion or splashes are eliminated, enabling repeated measurements to be made on a small sample volume. In the absence of splashes that might reach the optical elements, it is also possible to shorten the focal length of the laser focusing optics and so reduce the laser energy requirements.
The optimal film thickness for the LIBS measurements depends on the pulse energy and represents a compromise between the plasma intensity and the sample volume inside the focal spot. The choice of substrate, intended to be wasted, is fundamental in avoiding LIBS signal contamination by elements from the target, and we propose to use a silica wafer because of its high purity, optical finishing, and low cost. In the presence of a liquid layer, support ablation is significantly weaker than in dry conditions, or even absent, so the problem of eventual impurities interfering with the analytes is further reduced.
Here, we show that the active control of liquid film thickness is feasible for both water and oils, by electro-wetting with variable voltage or by target rotation at variable speed, respectively. The described sampling approach opens a possibility for sensitive, rapid, cost-effective LIBS analysis of liquids in small volumes and by using a compact instrument. The sample preparation does not require the use of reagents, drying, or an evacuation system, so the measurements could also be performed out of the laboratory.
Footnotes
Acknowledgments
The authors are grateful to their colleague Ivano Menicucci for assistance in realizing the system for electro-wetting.
Conflict of Interest
The authors report there are no conflicts of interest.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.