1. Introduction

Measurement of elements in single aerosol particles, especially in small droplets, could provide new insight into several processes in the fields of research and industry. Applications can be found, for example, in cloud formation research, in analysis of very low sample volumes in bioanalytics, in monitoring of emissions [1], and in the characterization of microbes [2] and nanoparticles [3]. Laser-induced breakdown spectroscopy (LIBS) is a technique for a fast elemental analysis of the solids, liquids and gases and mixtures of these such as aerosols and can thus be utilized in single particle measurements. One of the major possibilities of single particle LIBS measurements is to perform a sensitive online analysis of liquids with nanoliter-range sample consumption as several processes suppress the signal in the direct LIBS analysis of liquid samples [4, 5]. The technique could be utilized in process and waste water control in metal production and power industry facilities, and in water treatment plants. LIBS analysis of water solutions by single droplet sampling has been previously presented in the papers by Janzen et al. [6], Groh et al. [7], Cahoon and Almirall [8] and Järvinen et al. [9, 10]. In [9] and [10], we presented a single particle LIBS methodology where the individual droplet is dried after generation and the residual particle containing the concentrated trace elements and additive salt is introduced to the LIBS plasma.

Many referred applications require a measurement of a femtogram-level trace element mass per each particle. In the analysis of water by single droplet sampling, 1 ppb concentration limit of detection corresponds detection of only 0.5 fg in a 100 μm droplet. Such sensitivity is believed to be attainable for a real-time LIBS system if the measurement conditions are optimized and the pulse-to-pulse signal fluctuations minimized. Based on previous studies [3, 6, 11], the LIBS signal from a single droplet or a single dry particle is affected by the particle location and size at the moment of plasma initiation, and the applied laser fluence. However, the exact dependency and importance of these factors are not studied for precisely controlled particles and for low laser pulse energies of 1–20 mJ. The term ’particle’ is used throughout this article to describe either a liquid droplet, a completely dry aerosol particle formed from the droplet by drying or a droplet which is at the stage of fast drying and between these two extremes.

The location of the aerosol particle with respect to the focal point of the laser affects the time the analyte atoms diffuse and equilibrate to the plasma and the location of the mass transfer origin [12–14]. Thus, analyte emission has temporal and spatial variation, yet plasma emission is typically collected only from a certain slice of the plasma volume to a spectrograph using a fixed detection time window. Using similar particles, a lower detection limit was found when the particle was precisely trapped [9] instead of moving along a 100 μm wide air flow [10]. The water content in the particle alters the size and composition of the single particle and affects the diffusion rate and plasma temperature via the energy that is required for the vaporizing and molecular dissociation of the particle [15]. Insufficient fluence with relation to particle size causes an incomplete vaporization of the particle and the elemental composition of the plasma may differ from the original particle composition. It also reduces the linearity of the analyte calibration curve [16]. On the other hand, unlike in the analysis of nebulized droplet clouds or any bulk sample, increasing the fluence significantly above the complete vaporization threshold does not add analyte mass in the plasma but only increases the temperature of different species and speeds up the expansion of the plasma.

In this study, LIBS signal from single particles was investigated when the water content and the exact position of the particles were varied. Single water droplets containing the known concentrations of Pb, Al, and Mn were trapped and levitated in the LIBS focal volume using electrodynamic balance (EDB) [17] trap and the vertical position of the particles was adjusted by an electrostatic force. The combination of LIBS and EDB has been previously presented in [18] and [9]. Moreover, Dutouquet et al. [19] demonstrated the LIBS analysis of a particle cloud levitating in a RF discharge cell. To our knowledge, this is the first time the LIBS analysis was done for single particles that were scanned across the focal spot of the laser beam with a resolution of a few micrometers. The amount of water in the aerosol particle was controlled by varying the drying time between the droplet generation and plasma initiation for monitoring the LIBS signal dependency on the level of analyte preconcentration. Scattering from the particle was used for determining the location and size of the particle and for the automatic operation of the setup. Also, LIBS signal-to-noise ratio versus excitation laser fluence was measured at different ICCD gate delay times to estimate the optimal pulse energy and gating. The results can be applied to LIBS analysis of any precisely controlled micrometer sized aerosol particles regardless of trapping technique or aerosol origin. They provide means to improve the single particle LIBS analysis with regard to sensitivity and reproducibility.

2. Experimental

Figure 1 represents the experimental EDB-LIBS setup schematically. The EDB-LIBS principle and the used EDB electrode structure and the electric potentials are described in more detail in [9]. A piezoelectric droplet generator (MJ-AB-01-40-6MX, MicroFab Technologies Inc.) followed by a washer having −500 V potential are used to inject a charged single droplet between pairs of EDB electrodes. The droplet generator produces monodisperse droplets of about 74 μm diameter. The cylindrical electrodes are aligned along the center axis of a hexagonal measurement chamber. By controlling the electrode potentials the droplet can be trapped in the electric field and levitated with 1.5 μm precision at the focus of LIBS excitation laser beam. While levitating, the droplet dries into a dry residual particle in 3–10 s depending on the relative humidity (RH) of the surrounding air. The analyte elements in the droplet are preconcentrated due to the evaporation of the liquid matrix. The size of the residual particle depends on the total impurity concentration of the sample solution, which was conveyed to the droplet generator. Thus, each excitation laser pulse will sample almost equal trace element mass which is located in a well defined and repeated position.

 

Fig. 1 The EDB-LIBS configuration. L1=pulsed Nd:YAG, L2=532 nm CW laser, DM=dichroic mirror, W=window, EM=energy meter, HC=hexagonal chamber, EL=cylindrical electrodes, DG=droplet generator, CH=droplet charger, IF1&2=interference filters, P=pinhole, FL=flashlamp, H=liquid sample hose.

Download Full Size | PDF

After ejection from the generator orifice, the droplet oscillates vertically between the electrodes due to gravitational restoring force. The mass of the droplet is reduced by evaporation and the particle automatically settles at the focal point after complete drying. However, to measure LIBS signal versus the preconcentration level, the particle had to be driven to the LIBS focal point and the oscillation amplitude had to be suppressed before significant drying takes place. The gravitation affecting the droplet can be compensated by upward pointing electric force which is realized by applying a DC potential between upper and lower EDB electrodes. Thus, during the drying process, the DC potential was reduced continuously according to experimentally determined exponential function to settle the movement of the drying droplet already after 1 s levitation. Scattering from the droplet is used to determine if the droplet is sufficiently settled to the trapping point as discussed further in the text.

The droplet trapping requires control over the AC field frequency during the droplet drying. The levitation can be performed for different sample solutions with fixed device parameters if the total impurity concentration in the droplets stays constant, and if the RH inside the chamber doesn’t change significantly. The former condition was previously realized by adding soluble salt, such as NaCl, to the sample solution an amount that is much greater than the analyte concentration [9]. However, the added salt interferes with the detection of certain analyte emission lines and have been shown to cause matrix effects [20]. In this study, the EDB-LIBS analysis was done for the first time without using additive chemicals in the droplet. As illustrated in Fig. 1, the particle is illuminated using a 20 mW 532 nm CW laser module, and two aspheric lenses and a pinhole collect scattered green light from a small volume around the center point of the chamber. A PMT (H11706-01, Hamamatsu K.K.) detects the scattered light which provides feedback for the AC frequency adjustment algorithm. The PMT signal is also used for determining when the particle has settled and ready to be analyzed. As the particle slows down at the LIBS focal point, a 100 ms running average signal from the PMT first rises steeply, then decreases due to reduction in the droplet size, and then reaches a plateau level. After attaining the plateau, that is above certain threshold level, particle’s movement is considered stopped and the system is ready to trig the LIBS excitation pulse. The trigger signal is sent to the pulse laser right after the predetermined drying time from the droplet ejection has elapsed. The trigger signal activates the function of the laser Q-switch and a pulse is emitted at the next operating cycle of the 10 Hz laser. The scattering based trigger system also enables the measurements of high purity liquid droplets that are analyzed before the complete drying.

1–15 mJ laser pulse having a temporal FWHM of 7.8 ns and FWHM diameter of 6 mm from an actively Q-switched Nd:YAG (NT 342/1/UVE, Ekspla Ltd.) emitting at 355 nm is focused on the trapped particle by a 30 mm focal length aspheric lens (84339, Edmund Optics Ltd.). The diameter of the focal spot was measured to be 19 μm. The plasma emitted light is collected at 120° angle from the direction of the incoming laser pulse using a 50 mm focal length and 25 mm diameter achromatic doublet lens (65976, Edmund Optics Ltd.). The light is focused into a spectrometer with another doublet lens having a focal length of 100 mm (65979, Edmund Optics Ltd.). An ICCD camera (DH340T-18U-E3, Andor Technology Plc.) is coupled to the Czerny-Turner spectrometer (250is, Bruker Corp.) having a 1200 grooves/mm grating. The slit of the spectrometer was set at 27 μm which is two times the ICCD pixel width and the theoretical maximum resolution of the system is 0.16 nm. The pixels in the 2048×512 CCD cell are vertically binned to speed up the signal processing program. Unless separately emphasized in the Results and discussion section, the ICCD gate delay and width times were 1 μs and 20 μs, respectively.

After the laser pulse and the following data acquisition, the whole procedure is repeated starting with a generation of a new droplet. Depending on the desired degree of drying and the RH, the period of single-shot measurements is 2 s at the shortest. Hence, the maximum sampling rate is 30 particles/min or 6.4 nl/min. The RH can be lowered by flushing the chamber with compressed dry air or nitrogen. The droplet generation, trapping and laser pulse triggering are automated using a DAQ card (USB-6363, NI Corp.) and LabVIEW software. The same software controls also the xenon flashlamp (L7684, Hamamatsu K.K.) and the CMOS camera (DCC1645C, Thorlabs Inc.) seen in Fig. 1. They are used for taking 5 μs exposure still images and videos of the levitating particle. The water samples measured in this study, were prepared by dissolving PbCl₂ and the hydrous compounds of Al(NO₃)₃ and MnCl₂ in deionized water and diluting the solutions in volumetric flasks. The water samples were found to contain trace amounts of calcium due to impurities in the used chemicals and deionized water.

3. Results and discussion

3.1. Effect of droplet evaporation stage

Figure 2 presents an example spectrum showing the response to Mn, Al and Pb. The LIBS signal dependency on water content in the particle was measured by varying the time between the droplet generation and the LIBS excitation laser pulse as described in the Experimental section. The excitation pulse energy was maintained at 4 mJ and the signal was collected from Al 396.2 nm and Mn 403.1 nm emission lines. The concentration of both trace metals was 1 ppm corresponding 0.2 pg of Al and Mn in each droplet. The RH of the air inside the chamber was 25%. The results presented in Fig. 3 show a rise in the signal until drying time of about 6 s. After the rise, the signal settles at a saturation level which is assumed to correspond a situation where all the water has evaporated from the droplet. The time of complete drying, 6 s, agrees well with the theoretical drying time for 74 μm water droplet in RH=25% air [21]. The signal remains constant at the saturation level even after significantly longer trapping times than presented in Fig. 3. The main reasons for the signal reduction at short drying times are the incomplete vaporization of the large droplet and the consumption of energy for phase transition. They result in lower temperature plasma having a lower analyte species density compared with the analysis of a completely dried particle. At longer drying times, the decreasing moisture content in the particle increases the signal until complete drying. The relation between sample moisture and LIBS signal was reported in a recent study by Chen et al. [22]. Large droplets are more difficult to settle accurately at the focal point due to their mass and initial velocity. It was not possible to settle the droplet faster in the focal volume than about 1 s which determined the shortest investigated drying time.

 

Fig. 2 EDB-LIBS spectrum from water sample containing 1 ppm of Mn and Al and 0.6 ppm of Pb using 4 mJ excitation laser pulse energy. The spectrum is an average of 20 single-shot spectra. The CN and ${\text{N}}_2^{+}$ bands originate from the ambient air.

Download Full Size | PDF

 

Fig. 3 LIBS signal from Al and Mn as function of drying time. The insets show 30μm×30μm CMOS camera images of the particle at 1 s and 6 s after the droplet launch.

Download Full Size | PDF

3.2. Effect of micrometer displacements in the single particle position

The effect of micrometer-scale displacements between the analyzed particle and the focal point of the excitation laser beam on the LIBS signal from Pb 405.8 nm emission line was investigated using the ability to control the particle location in the EDB chamber. The dependency is expected to be independent of the selected transition. The water sample contained 20 ppm of Pb which corresponds 4 pg per droplet, and all the analyzed droplets were dried completely into a dry residual particles. The LIBS excitation pulse energy was kept at 5 mJ which produced plasma in the chamber even without a particle in the trap. The vertical displacement of the particle was determined from the still images using the flashlamp and the CMOS camera. The inset in Fig. 4 presents the vertical displacement of the centroid of a 5.5 μm particle from the excitation beam focus for each DC voltage values between the upper and lower electrodes. The ordinate in the upper graph of Fig. 4 is an average LIBS signal of 20 single-shot spectra. At each measurement point, the plasma emission was guided to the spectrometer with constant light collecting efficiency. The signal remains stable while the particle lies in the interval which has a width of approximately 1.5 times the focal spot diameter. After the offset of ±15 μm, there is a sharp signal decrease setting the collected analyte emission at a level of less than one seventh of the emission collected at the focus. These high-signal and low-signal levels are assumed to correspond the region where the laser pulse can directly interact with the particle and the region where plasma forms first in air and then surrounds the particle by expansion. In the latter, the analyte emission may be localized to a certain part of the plasma cloud [15]. The high signal was observed also when the particle was only partly located in the focal area. When the particle is well located at the focal spot, the fluctuation of the single-shot signals is about 15%. The lower graph shows the relative standard deviation (RSD) of the averaged signal. It is obtained by dividing the standard deviation of the mean by the average signal of 20 single-shots. The RSD is less than 4% when the particle is located at the focus. Applying a DC voltage different from the value that exactly cancels out the effect of gravitational force causes the particle to vibrate slightly in the vertical direction. Thus, moving away from the focus increases the deviation in the successive particle positions which also contributes to the rise in signal RSD.

 

Fig. 4 LIBS signal from Pb and signal RSD as function of particle position. The error bars in the upper graph show the standard deviation of the mean signal whereas the signal fluctuation between successive points is due to a systematic error. Particle displacement versus bias voltage between upper and lower electrodes is shown in the inset.

Download Full Size | PDF

3.3. Pulse energy in single particle measurements

Effect of LIBS excitation laser fluence was studied by varying the pulse energy between 1.5 mJ and 15 mJ. Atomic lines are visible with pulse energies greater than 1.5 mJ which is labeled as the threshold for the plasma generation E_th in Fig. 5(a). At this fluence, more than 50% of the excitation pulses do not vaporize the 3 μm residual particle but merely break it into small fractions that are still visible in the CMOS camera screen. The fractions levitate between the electrodes until they are removed by flushing air through the chamber. When the pulse energy is adequate to vaporize the particle in the trap, no fractions are observed and flushing is not required. Figure 5(a) shows the LIBS signal-to-noise ratio (SNR) of Pb 405.8 nm emission line as a function of excitation pulse energy using 700 ns, 1 μs and 5 μs ICCD gate delay times. In the measurement of SNR, the signal is the background subtracted peak height and noise is defined as three times the standard deviation of the background. The emergence of the background is considered the primary reason for decreasing SNR at higher pulse energy values. Less important factors are the larger diameter of the emitting vapor, the higher degree of ionization, and the temperature dependency of the electronic partition function Z(T) of Pb atom. The emitting vapor is more confined using a combination of short delay and low pulse energy as shown in images 1–4 in Fig. 5(b). Thus, more analyte atoms are within the narrow field-of-view of the spectrometer which is considered the reason for higher SNR maximum at short gate delays. Using a delay time less than 1 μs can nevertheless be impractical due to the high sensitiveness of the SNR to the changes of pulse energy. Short delay also requires pulse energy to be near E_th and may cause signal fluctuation through the incomplete vaporization of the particle. The error bars in Fig. 5(a) present the standard deviation of the single-shot SNR values. Similar dependence as seen in Fig. 5(a) was found for other atomic transitions having the upper state energy less or comparable to 4.4 eV which is the upper state energy of the probed Pb 6p7s ${}^3{\text{P}}_1^o\to 6{\text{p}}^2\hspace{0.17em}{}^3{\text{P}}_2$ transition [23]. The measured transitions include Al 396.2 nm, Mn 403.1 nm, and Ca 422.7 nm lines. The requirement of higher plasma temperature and thus higher pulse energy is expected when probing transitions involving significantly higher upper energy states. For example, the SNR of the Zn 481.1 nm emission line having the upper state energy of 6.7 eV [24] was found to increase 145% when the pulse energy was raised from 6 mJ to 10 mJ.

 

Fig. 5 a) LIBS SNR of 20 ppm Pb as function of excitation laser pulse energy using 700 ns, 1 μs and 5 μs gate delays. The spline fit curves are guides for the eye and have no physical interpretation. b) Pb 405.8 nm emission from plasma using different pulse energies and 1 μs (1 and 2) and 5 μs (3 and 4) gate delays. Imaged through a bandpass interference filter using 100 ns ICCD gate width.

Download Full Size | PDF

The LIBS signal response to Pb mass was found to be linear in the concentration interval of 4–40 ppm Pb in the sample solution or equivalently 0.8–8 pg of Pb in each particle. The applied pulse energy was 6 mJ which was found to be optimal in the detection of Pb in 3 μm particles. If the trapped particle is notably larger, for example, due to a high water impurity concentration, the signal saturates due to the incomplete vaporization of the particle mass. Higher pulse energy is then required for the linear dependency. The change in the pulse energy will also affect the value of the optimal gate delay according to Fig. 5(a). 40 ppm was the highest Pb concentration measured in this study due to the poor solubility of the PbCl₂ salt.

4. Conclusions

The LIBS signal dependency on the water content and the exact position of single particles was quantified. The measurements were carried out using an additive-free EDB-LIBS technique which enables the control of the particle position and the liquid aerosol particle size while holding other conditions fixed. The key benefit in the single particle analysis of liquids is the preconcentration of the trace elements as even a very small amount of water in the particle was found to depress the LIBS signal. The signal from both investigated elements saturated after 6 s period of droplet evaporation. When the particle was scanned across the laser beam path, the region of the high signal was observed where the particle could directly interact with the LIBS excitation pulse. Increasing the excitation pulse energy significantly above the plasma formation threshold energy is not advisable in a single particle LIBS scheme where the mass to be vaporized is limited. In the detection of several trace metals in micrometer sized particles, pulse energy around 6 mJ with gate delay of 1 μs was considered optimal in terms of SNR and pulse-to-pulse repeatability. However, the optimal pulse energy value also depends on the plasma imaging area and laser focusing optics. Based on the obtained results, simultaneous variation in the particle location and size and excitation pulse energy can fluctuate the single particle LIBS signal of successive pulses even orders of magnitude. On the other hand, for many analytical purposes, the information from a single-shot spectrum is sufficiently accurate if the three factors are optimized and stabilized. The current findings pave the way for enhanced LIBS analysis for monitoring liquid and fine particle emissions, for industrial process and quality control and for atmospheric and biochemical research.