1. Introduction

Lithium (Li) is a light element important to a variety of fields including geochemistry, energy storage, nuclear energy, forensics, and tritium production [1,2]. The detection and quantification of Li and its isotopes is critical to each of these fields. For example, in petrological processes, Li partitions to the melt and can therefore be used as a geochemical tracer [1]. In the field of energy storage, Li fractionation is studied to track Li-ion transfer and loss in Li-ion batteries crucial to understanding aging mechanisms and the design of improved battery materials [3,4]. In addition, Li is a key component of several nuclear applications, such as some fluoride salt-cooled high temperature nuclear reactors that employ solid nuclear fuel and a molten salt coolant. Li$_2$BeF$_4$ (a mixture of 2LiF-BeF$_2$) is a salt composition of interest to many molten salt reactor designs. For the use of this coolant, Li must be highly enriched in $^7$Li to prevent overproduction of tritium ($^3$H) during reactor operation, hence an understanding of Li isotopic composition is needed. Li/Be elemental ratios also must be measured, as this ratio is critical to the chemical and thermophysical properties of the coolant [5]. In contrast to molten salts with high $^7$Li concentrations, the application of tritium ($^3$H) production relies on LiAlO$_2$ ceramics that are enriched in $^6$Li for use in tritium-producing burnable absorber rods (TPBARs) [6]. When irradiated with neutrons, $^6$Li produces $^3$H which is absorbed by a Zircaloy-4 getter component of the TPBAR, and is later extracted for use in defense and energy applications.

Numerous techniques have been demonstrated for Li elemental and isotopic analysis including several mass spectrometry methods such as inductively coupled plasma mass spectrometry (ICP-MS) [5], and atom probe tomography (APT) [7,8]. Optical methods have also been explored, including LIBS [9–11], laser induced fluorescence (LIF), [12] and absorption spectroscopy (AS) [9,13,14]. The advantage of optical methods in comparison to mass spectrometry is the rapid data collection time, the ability to analyze gas, liquid, and solid phases, and no need for sample preparation, enabling high throughput analyses. In addition, LIBS and LIF allow for non-contact analysis that can be performed at standoff distances [15,16] which can be a significant advantage if analyzing hazardous substances, such as molten salts or neutron irradiated materials. LIBS offers the additional benefit of a relatively simple experimental set-up, which can make the technique portable and flexible for field deployment [2,17]. However, challenges remain in the accurate, quantitative analysis of Li and its isotopes via LIBS, in particular how to overcome line broadening and line shape distortions that impact Li spectral features.

As a passive sensing method, LIBS uses laser ablation (LA) to generate a plasma that emits light at wavelengths characteristic of species in the target. Laser-produced plasmas (LPP) are inherently transient media and therefore inhomogeneous in both space and time. These inhomogeneities are strongly influenced by plasma generation conditions, in addition to the nature and pressure of the ambient gas, which can in turn impact Li isotopic analysis. Isotopic shifts present in atomic and molecular spectra allow for the detection and analysis of isotopes in solids, liquids, or gases via optical spectroscopy tools [15]. The ability to detect isotopes depends strongly on linewidth and line shape. The linewidth to isotopic shift is critical for isotopic analysis, and is dictated by the resolution of the detection system and phenomena contributing to line broadening (e.g., Stark, Doppler, van der Waals, instrumental) [15]. Line broadening becomes increasingly important when attempting to resolve finely-spaced isotopic shifts, such as those for Li isotopes, the greatest of which is $\approx$15.6 $\pm$ 0.3 pm for a Li I resonance line (2s$^2$ - 2p$^2$) at $\approx$ 670.7 nm [18,19]. This transition also has fine structures, with D₁ and D₂ components at 670.776/670.791 nm for $^7$Li and 670.791/670.806 nm for $^6$Li [20]. Each of these fine structures also has several hyperfine structures [21,22].

In addition to line-broadening mechanisms, factors such as self-absorption and self-reversal can impact analytical measurements and isotopic detection. Self-absorption occurs when some of the radiation emitted by a source is re-absorbed before exiting the source and eventually reaching the detector [23]. Self-absorption distorts spectral line profiles because absorption is strongest at the line center and weakest at the wings, however, the presence of self-absorption is challenging to evaluate based on line shape alone. Self-absorption may manifest in spectral profiles as saturation at the line center. Additionally, self-absorption may be more clearly identified in the shape of a calibration curve, where emission intensity is plotted versus known analyte concentration. In the absence of self-absorption, we expect emission intensity to grow proportionally with analyte concentration. However, if self-absorption is present, at high analyte concentrations the emission intensity saturates. Self-reversal is a distinctly different phenomenon that is attributed to spatial gradients in plasma temperature and electron number density [23]; when a photon is emitted by an atom or ion in the hot, dense central region of the plasma and is transported through a cooler, less dense periphery it can be re-absorbed. Self-reversal can be observed in spectral profiles with the presence of a dip in emission intensity at the line center [24]. Several experimental parameters, such as the nature and pressure of the ambient gas, can impact self-absorption and self-reversal phenomena observed in LPPs [25], leading to line distortions that make Li isotopic analysis via LIBS challenging.

In this work, we characterize the influence of the nature and pressure of the ambient gas on Li I emission spectral features produced via nanosecond (ns) laser ablation (LA) of a LiAlO$_2$ target material. We report spatially and temporally resolved Li I spectral features in Ar, N$_2$, and He gases. The impact of ambient gas composition and pressure on LPP spectral features, such as linewidth and line shape (i.e., the presence of a central dip) are investigated over the pressure range of 0.05 - 100 Torr. Lastly, experimental parameters leading to minimal line broadening and line shape distortions are used for collecting Li I emission spectra from LPPs generated via LA of LiAlO$_2$ targets with varying $^6$Li/$^7$Li ratios. Chemometrics, specifically principal component analysis and regression (PCA and PCR, respectively), are used to classify LiAlO$_2$ target materials based on $^6$Li/$^7$Li ratios, and produce a calibration curve.

2. Methods

2.1 Experimental methods

The experimental set-up used in this work is schematically described in Fig. 1. A Nd:YAG laser (Continuum, Surelite) with 1064 nm wavelength, $\approx$6 ns full width half maximum (FWHM), operated at 10 Hz, with a laser fluence of $\approx$12 J/cm$^{2}$ was used for producing plasmas from a LiAlO$_2$ target with natural abundance of $^6$Li (7.59 %) and $^7$Li (92.41 %) isotopes positioned in a vacuum chamber. The vacuum chamber was placed on a x-y translator to easily move between targets and prevent drilling. Multiple optical windows and ports for laser entrance, light collection, and electronics were included in the chamber design. A pressure gauge, vacuum pump, and gas lines were also attached to the chamber to control ambient gas pressure. The measurements were carried out in Ar, N$_2$, and He gas environments. The environment was varied from 0.05 - 100 Torr for investigating ambient gas pressure effects.

 

Fig. 1. Schematic of the experimental set-up used for collecting optical emission spectra. Acronyms used in the schematic are defined as follows: C- cube polarizer; WP- wave plate; L- lens; ICCD- intensified charged coupled device.

Download Full Size | PDF

An optical system consisting of two plano-convex lenses was used for imaging the plasma plume onto the slit of a 0.75 m spectrograph (SpectraPro HRS-750, Princeton Instruments) positioned orthogonal to the plasma expansion direction for spectral measurements. The light dispersing system is equipped with three gratings (300, 1200, and 2400 grooves/mm) and an intensified CCD (ICCD, PiMAX4, Princeton Instruments). Broad spectral measurements were evaluated using the spectrograph-ICCD combination with 300 grooves/mm grating, while the 2400 grooves/mm grating was used for high-resolution spectral measurements of Li I transitions in the range of 670.7749 - 670.807 nm. The maximum resolution of the detection system is $\approx$ 12 pm (at FWHM) at 632 nm measured using a HeNe laser. For isotopic analysis, six LiAlO$_2$ target materials with the following $^6$Li/$^7$Li were analyzed: 0.0821 (natural abundance), 0.3812, 0.3540, 0.3100, 0.569, and 0.7130. These values of $^6$Li/$^7$Li were measured via ICP-MS.

2.2 Chemometrics for isotopic analysis

The Li I transition centered at $\approx$670.8 nm was fit with a Voigt function, and the peak central position (in nm), FWHM, and peak area were determined from the Voigt fit. In addition, the area under the raw data curve was determined. These parameters were used for calculating principal components (PCs) used for generating calibration curves via principal component regression (PCR). All analysis was performed in Python using scipy, pandas, and sklearn libraries. To generate a calibration curve (i.e., predicted $^6$Li/$^7$Li versus known $^6$Li/$^7$Li measured via ICP-MS), PCR was performed.

3. Results and discussion

A broad survey spectrum of LiAlO$_2$ LPPs is given in Fig. 2. Spectral features were recorded in 100 Torr Ar, 5 mm from the target at a gate delay/width of 20 $\mu$s/10 $\mu$s. At this later time in the LPP lifecycle, the plasma is dominated by Al and Li atomic emission features. Prominent Al I and Li I transitions are labelled in Fig. 2 [26] and include: Al I at 308.22 nm (32435.5 - 0 cm$^{-1}$), Al I at 309.27 nm (32436.8 - 112.1 cm$^{-1}$), Li I 323.26 nm (30925.6 - 0 cm$^{-1}$), Al I at 394.40 nm (25347.8 - 0 cm$^{-1}$), Al I 396.15 nm (25347.8 - 112.1 cm$^{-1}$), Li I at 610.37 nm (31283.1 - 14904.0 cm$^{-1}$), and $^7$Li D$_1$ at 670.665 nm (14904.00 - 0 cm$^{-1}$), $^7$Li D$_2$ at 670.791 nm (14903.66 - 0 cm$^{-1}$), $^6$Li D$_1$ at 670.791 nm, and $^6$Li D$_2$ at 670.806 nm. It is noted that the D lines for $^6$Li and $^6$Li appear as a single broad peak centered at $\approx$670.8 nm in the Fig. 2 spectrum. Li I 670.8 nm is isolated from other transitions and has the greatest isotopic shift ($\approx$15.8 pm) and is therefore well suited for isotopic analysis using many commercially available spectrographs. Other Li I transitions in the visible spectral range have isotopic shifts of only $\le$3.5 pm [22].

 

Fig. 2. Broad spectrum of LiAlO$_2$ LPP (with natural Li isotopic abundance) generated in 100 Torr Ar, at 5 mm from the target, at a gate delay/width of 20 $\mu$s/10 $\mu$s. The spectrum is averaged over 10 laser shots. The spectrum was collected using the 300 grooves/mm grating.

Download Full Size | PDF

In order to accurately distinguish isotopic splitting with varying $^6$Li/$^7$Li ratios, the line broadening and distortions must be minimized. In this work, we select the Li I 670.8 nm transition with the largest isotopic shift of $\approx$15.8 pm for all analyses. Several experimental conditions must also be considered for minimizing line broadening and distortions, including ambient gas and pressure, position in the plasma, time after plasma onset when measurements are made (i.e., gate delay), and integration time (i.e., gate width). Each of these parameters is important for obtaining a strong emission signal (and high signal-to-background ratio) and minimal line broadening required for isotopic analysis employing LIBS.

3.1 Ambient gas and pressure effects

Here, Ar, N$_2$, and He were considered as background gases in order to avoid any gas-phase chemistry (e.g., oxidation) that readily occurs for Al in oxygen-containing environments [27]. Ar, N$_2$, and He background gases are often used in analytical applications since they allow for collisional interactions leading to plume confinement, increased persistence and emission signal, and efficient plasma cooling [28–31]. Ar in particular has been shown to lead to more uniform plasma properties, such as temperature and density, and hence a reduction in self-reversal effects [28,32,33].

Spectral features of Li I 670.8 nm were collected in Ar, N$_2$, and He gases from 0.1 - 100 Torr, with select spectra reported in Fig. 3(a)-(c). All measurements were made at a distance of 5 mm from the target surface and at 20 $\mu$s after the onset of plasma generation. The gate width used was 10 $\mu$s. This position and time were selected because Li I 670.8 nm is a ground state transition, and hence is prone to self-absorption when/where Li number density is high (i.e., close to the target, at early times). Hence, we delayed measurements to later times in the plasma lifecycle and selected a distance relatively far from the target surface where we expect self-absorption to be minimal, but we can still collect a strong emission signal.

 

Fig. 3. Emission spectra of Li I ($\approx$670.8 nm) taken in (a) Ar, (b) N$_2$, and (c) He gases at varying pressures from 100 to 0.1 Torr as noted in each sub-figure. The gate delay/width used were 20 $\mu$s/10 $\mu$s. Each spectrum is averaged over 10 laser shots.

Download Full Size | PDF

Spectral features in Fig. 3 show that for all gases at 100 Torr, Li I is significantly broadened and a central dip is observed. The observation of central dip or self-reversal is most severe in 100 Torr N$_2$ in comparison to Ar and He. As pressure is reduced from 100 Torr, we observe different effects in each of the gases. The central dip in Ar is observed at 100 Torr, but not at 75 and 50 Torr, then is apparent again in 20, 10, and 1 Torr. At the lowest pressure of 0.1 Torr Ar, no central dip is present. The central dip in N$_2$ is significant (reaching a minimum close to 0 counts), and is present in all spectra from 100 Torr to 1 Torr, and is absent at 0.1 Torr. In 100 Torr He, there is a central dip with a peak still present at the center, which disappears by 50 Torr. A central dip is still present in spectra from 20 - 10 Torr He. Minor peak distortion is observable at 1 Torr He, and none is seen at 0.1 Torr. In all gases, linewidth decreases as pressure is reduced, and at lower pressure levels the central dip is either absent or less severe than at the higher pressures.

To better understand changes in Li I lineshape with varying pressure and ambient gas, the FWHM of experimentally collected Li I spectra were determined and compared to the FWHM of a Voigt fit with no central dip. This Voigt fit can be referred to as a ’reconstructed’ line profile and determined based on the wings of the experimentally measured Li I spectral features. Peak fitting was performed by removing points corresponding to the central dip, and then fitting the remaining data points. The Voigt fit therefore has a different line shape and FWHM than the experimentally collected data. An example of an experimental Li I 670.8 nm spectrum and a reconstructed line profile/Voigt fit is shown in Fig. 4 for 100 Torr N$_2$.

 

Fig. 4. Example experimental emission spectra of the Li I transition at $\approx$670.8 nm with a central dip and the corresponding Voigt fit. This spectra was collected in 100 Torr N$_2$ at 5 mm from a LiAlO$_2$ target (natural Li isotopic abundance) at a gate delay/width of 20 $\mu$s/10 $\mu$s.

Download Full Size | PDF

FWHMs of measured spectra and corresponding Voigt fits at various pressure levels are reported in Fig. 5(a), (b), and (c) for Ar, N$_2$, and He, respectively. All FWHM values were determined from experimentally measured spectra (and corresponding Voigt fits) collected at 20 $\mu$s after plasma onset, 5 mm from the LiAlO$_2$ target surface, consistent with spectra reported in Fig. 3. Error bars reported in Fig. 5 are calculated from the standard deviation of three duplicate spectra, where each spectrum is averaged for 10 laser shots. For all gases, Li I FWHM is maximum at 100 Torr, and decreases as pressure is reduced. Li I FWHM reaches a minimum and plateaus in all gases by 0.1 Torr, although the minimum FWHM value varies between gases. The lowest FWHM value of $\approx$33 pm is achieved in 0.1 Torr He. In contrast, FWHM in 0.1 Torr Ar and N$_2$ is $\approx$40 pm.

 

Fig. 5. FWHM as a function of pressure for (a) Ar, (b) N$_2$, and (c) He gases. FWHM is reported for recorded spectral data (i.e., Experimental), and for the Voigt fit that excludes a central dip (i.e., Voigt).

Download Full Size | PDF

Differences in spectral features (Fig. 3) and FWHM versus pressure trends (Fig. 5) may be attributed to several environmental and experimental factors that impact plasma cooling, and thus self-reversal. Such factors include: (1) pressure of the ambient gas, (2) ambient gas medium (molecular versus atomic gas), (3) thermal conductivity of the ambient gas, and (4) position and timing of spectral measurements (here, 20 $\mu$s gate delay, 5 mm from the target). For all of these factors, a critical consideration is how these impact plasma inhomogeneities in electron temperature and electron number density. Electron temperature gradients in particular have been found to have a more significant impact on self-reversal than electron number density [34].

First, the pressure of the ambient gas, studied here between 0.05–100 Torr, significantly impacts LPP confinement and hence cooling. At higher pressures, plume confinement is more significant, leading to increased collisions between plume and ambient gas species. These collisions lead to cooling of the plasma periphery and greater temperature gradients than at reduced pressure levels. A comprehensive study on the expansion dynamics of Li atoms in different background gases can provide additional insight into the effects of background gas pressure.

In addition, whether the ambient medium is a molecular or atomic gas impacts its ability to cool the LPP. The average energy of an atom or molecule defined in statistical mechanics is:

The thermal conductivity of the ambient gas will also impact temperature gradients in the plume. Of the three gases studied here, He has the greatest thermal conductivity, and thus is most effective at cooling the plasma. This effective cooling, particularly at pressures between 10-100 Torr contributes to the presence of a central dip and significant line broadening. In contrast, Ar thermal conductivity is lower, and therefore LPP temperature in Ar is more uniform. Ar as a background gas has been shown to lead to more uniform spatial distributions of temperature and density in LPPs [33]. The spectral features recorded in higher He pressures showed a central peak in the self-reversed line. A similar observation was reported previously for Ba$^+$ resonance lines and was attributed to resonance scattering in the plasma [35].

Lastly, it is essential to note that all FWHM values were determined from Li I spectra taken at a fixed position (5 mm) and time (20 $\mu$s) in LPP evolution for all gases. Expansion dynamics of plume species in varying background gases and pressures will change due to differences in confinement effects [29]. We expect Ar and N$_2$ confinement effects to be similar (comparable masses), although they will result in different temperature and density gradients, as mentioned previously. However, in He the LPP expansion is expected to be greater due to its light mass. Therefore at a fixed position of 5 mm from the target surface in gases and pressures that confine the plume differently, spectral measurements are taken where temperature gradients are expected to be significantly different between environments. For example, in 100 Torr Ar and N$_2$, 5 mm from the target surface is close to the plume edge, whereas in 100 Torr He, we collect spectral from closer to the center of the plume.

From Fig. 5, we find the minimum FWHM ($\approx$33 pm) is observed in 0.1 Torr He. For all gases and pressure levels, FWHM plateaus at a value between $\approx$33 - 39 pm, and never reaches the instrumental linewidth ($\approx$12 pm). This plateau in FWHM higher than the instrumental linewidth could be attributed to the fact that several phenomena contribute to line broadening in Li LPPs that cannot be overcome with the use of emission-based spectroscopic tools. First, there are several fine and hyperfine structures, in addition to the $\approx$ 15.8 pm isotopic shift, that contribute to the width of the Li I ($\approx$670.8 nm) transition. The $^7$Li fine structures of D1 and D2 (670.7760/670.7911 nm) have a difference of 15.1 pm, and the $^6$Li D1 and D2 (670.7918/670.8068 nm) have a difference of 15.0 pm. An example spectra from the LiAlO$_2$ target with natural Li isotopic abundance is shown in Fig. 6 with fine structures labelled. Each of these fine structures also have hyperfine structures. Prior work employing an optical comb synthesizer measured 19 hyperfine components for 12 resolvable features for $^{6,7}$Li D lines spread over a span of 20 GHz (0.03 nm) [21]. For the FWHM values for Li I reported in Fig. 5 contribution from $^6$Li is minimal, given the target analyzed has natural Li isotopic abundance with only $\approx$7.59 % $^6$Li.

 

Fig. 6. Li I spectrum with D$_1$ and D$_2$ of $^6$Li and $^7$Li labelled. Relative intensities of D$_1$ and D$_2$ are for Li with natural isotopic abundance. The spectrum was collected in 0.1 Torr He gas at 5 mm from the target at a gate delay/width of 20 $\mu$s/50 $\mu$s. The spectrum was averaged over 10 laser shots.

Download Full Size | PDF

In addition to the presence of an isotopic shift, fine, and hyperfine structures, line broadening mechanisms inherent in LPPs contribute to the measured linewidths of Li I. Line broadening mechanisms that must be considered when measuring emission from LPPs include Natural, Stark, Doppler, resonance, and van der Waals all contribute to the measured linewidth of a selected transition. The estimated natural and resonance linewidth contributions of $^7$Li D2 transition are 0.35 pm and 0.65 pm respectively [9]. Typically, Stark and Doppler are the prominent line-broadening contributors in LPPs [15]. The impact and significance of each of these parameters vary with plasma properties, species in the plasma, timing, and the nature and pressure of the ambient gas. The FWHM of a transition $\Delta \lambda _{s}$ will be Stark broadened due to electron impact and is given by [36]:

 

Fig. 7. The estimated Li I (670.8 nm) linewidth at various temperatures and electron densities.

Download Full Size | PDF

Since the selected Li I transition contains fine and hyperfine components, any broadening mechanism will affect all of these structures individually. The measured linewidth represents the sum of linewidths of various fine and hyperfine components and instrumental broadening. Thus, the minimum line broadening we can achieve while also having a measurable emission signal is $\approx$30 pm. This finding suggests that even with the improved spectral resolution, we are limited by line-broadening mechanisms inherent to emission-based optical techniques. Active sensing methods such as LAS would be required to obtain clearly separable D$_2$ ($^7$Li), D$_1$ ($^7$Li) + D$_2$ ($^6$Li), D$_1$ ($^6$Li) peaks [9,13].

To quantify the extent of distortions in the Li I transitions (observed in Fig. 3), the values of FWHM of experimental and Voigt fits were used to calculate a coefficient. This coefficient is defined as follows: $(SA)^{\alpha }$ = $\Delta$ $\lambda$ / $\Delta$ $\lambda _0$, where $\alpha$ is an empirically derived constant equal to -0.54, $\lambda$ is the FWHM of the original, experimentally measured spectral line, and $\lambda _0$ is the FWHM of the corrected/reconstructed line that is representative of the original spectral line had there not been any central dip [38]. Figure 4 provides an example of experimental versus reconstructed lines. This SA coefficient is referred to in prior work as the self-absorption coefficient [38], however here is used to quantify the extent of self-reversal in Li I spectral profiles, so we refer to this coefficient either simply as a coefficient or self-reversal coefficient, given the presence of the central dip is indicative of the self-reversal phenomenon. When the coefficient is equal to 1, the FWHM of experimental and Voigt fits agree and line distortions are negligible. All errors given in Fig. 8 are standard deviations calculated from three duplicate spectra.

 

Fig. 8. Self-reversal coefficient as a function of pressure for Ar, N$_2$, and He gases. Coefficients reported here were calculated from data provided in Fig. 5.

Download Full Size | PDF

From results in Fig. 8, we find that the coefficient increases from 0.4 - 0.6 for all gases to 1 as pressure decreases. The coefficient varies between gases. We find that for pressures of 5 Torr and lower, the coefficient in He is greater than in Ar and N$_2$, suggesting self-reversal is reduced in He over this pressure range. For example at 1 Torr He, the coefficient for He is $\sim$ 0.93, versus $\sim$ 0.65 in Ar and $\sim$ 0.60 in N$_2$. At 0.1 Torr coefficients for all gases have approached values of $\sim$ 0.89, 0.93, and 1 for N$_2$, Ar, and He, respectively. Between 10-100 Torr, we observe an interesting trend: the self-absorption/reversal coefficient increases when the background pressure $\ge$ 10 Torr and it indicates that LPPs become more homogeneous at high background pressures. As the background pressure increases $\ge$ 10 Torr, temperature and electron number density gradients are reduced due to enhanced collisions caused by plume confinement [39].

3.2 Space and time evolution

Based on results reported in Figs. 5 and 8, we find low pressure (i.e., 0.1 Torr) He gas provides minimal line broadening and distortions for a given spatial and temporal window (i.e., 5 mm from the target surface, 20 $\mu$s after plasma onset). At 0.1 Torr He, we then collected spatially and temporally resolved Li I spectra to understand how the intensity and linewidth change with position and time (Fig. 9). For accurate isotopic analysis via LIBS, it is important to achieve a strong signal (or signal-to-background ratio) and minimal line broadening [31]. Hence intensity and linewidth (i.e., FWHM) were determined from spectral features given these parameters are critical to analytical measurements. In order to obtain strong emission signals, higher temperature conditions are needed. These high-temperature conditions are present at early times in the LPP lifecycle. In contrast, the Stark broadening becomes negligible at later times in LPP evolution, leading to reduced linewidths. Thus, the timing of spectral measurements must be selected appropriately to achieve a balance between strong emission signal and reduced linewidth. From spectral features reported in Figs. 9(a) (d), we observe no evidence of line shape distortions.

 

Fig. 9. Spatially and temporally-resolved Li I spectral features, and corresponding intensity and FWHM. LPPs were generated from a LiAlO$_2$ target with natural $^{6,7}$Li abundance. (a) Spatially-resolved Li I spectra (collected at a gate delay/width of 20 $\mu$s/50 $\mu$s), (b) intensity of Li I (670.8 nm), and (c) FWHM as a function of distance from the target. (d) Time-resolved Li I spectra (taken at 5 mm from the target surface), (e) intensity and (f) FWHM as a function of time. Progressive gate widths (noted as x-error bars) were used to account for reduced emission intensity at later times after plasma onset, as follows: 10 $\mu$s (from 1-10 $\mu$s), 20 $\mu$s (from 10-45 $\mu$s) and 30 $\mu$s (from 50-100 $\mu$s). All y-error bars represent the standard deviation calculated from three duplicate spectra. Each spectrum is averaged over 10 laser shots.

Download Full Size | PDF

For spatially-resolved Li I spectra shown in Fig. 9(a), we find that intensity decreases approximately linearly with increasing distance from the target (Fig. 9(b)), and FWHM is approximately constant with distance (Fig. 9(c)).

From time-resolved Li I spectra (Fig. 9(d)), we find emission intensity is strongest at early times and decays rapidly with increasing time after plasma onset. By $\approx$ 20 $\mu$s, Li I intensity reaches a plateau (Fig. 9(e)). Li I emission intensity remains relatively strong from $\approx$20 - 100 $\mu$s because this is a ground state coupled transition that persists even at relatively low plasma temperatures that are present at later times. FWHM as a function of time shows a similar trend to intensity, and reaches a minimum by $\approx$ 20 $\mu$s (Fig. 9(f)). The spatio-temporal analysis of Li I 670.8 nm transition showed that the FWHM approaches a minimum value at all distances when the gate delay is $\sim$ 20 $\mu$s.

3.3 Isotopic analysis

We find that a 0.1 Torr He ambient gas environment, with an observation position of 5 mm from the target surface, and a at a gate delay of 20 $\mu$s are parameters well-suited for the analysis of Li isotopes in LPPs. These parameters lead to minimal line broadening (FWHM of $\approx$33-35 pm) and distortions, while also achieving reasonable emission intensity. Hence, we perform a case study using six LiAlO$_2$ targets with varying $^6$Li/$^7$Li ratios (0.082 - 0.713) to determine if we can develop a calibration curve with high accuracy, and ultimately predict Li isotopic composition from emission spectral features. All recorded Li I spectra appear as a single broad peak with some differences in peak central position and width that may be attributed to varying amounts of $^6$Li and $^7$Li. Example Li I spectra for select LiAlO$_2$ targets are shown in Fig. 10. However, using typical peak fitting and analysis procedures alone, it is difficult to distinguish between samples with varying isotopic ratios. Hence, we performed principal component regression (PCR) in order to plot predicted $^6$Li/$^7$Li from emission spectra versus known $^6$Li/$^7$Li values.

 

Fig. 10. Example Li I spectra from LiAlO$_2$ targets with varying Li isotopic ratios. LPPs were generated in 0.1 Torr He, 5 mm from the target surfaces, at a gate delay/width of 20 $\mu$s/50 $\mu$s. Spectra were averaged over 100 laser shots.

Download Full Size | PDF

To perform PCR, 50 duplicate Li I spectra for each LiAlO$_2$ target were collected, where each spectrum was an average of 100 laser shots. A train/test split of 75/25 was used, where the testing group was randomly selected from all spectral data collected and the remainder was used for training. Hence, 300 total data instances (i.e., spectra) were collected, 225 for training/75 for testing. Results from PCR for these data collection parameters are provided in Fig. 11. It should be noted that this figure includes 75 data points, however several are overlapping. The regression line has an R$^2$ value of 0.90, a slope of 0.9223 $\pm$ 0.0355 (close to the ideal value of 1), and a y-intercept of 0.00704 $\pm$ 0.01615. These results indicate that predicted and known values of $^6$Li/$^7$Li show good agreement. Hence, distinguishing between target materials with varying Li isotopic ratios is possible with high accuracy for LIBS measurements made in 0.1 Torr He, 20 $\mu$s after plasma onset, 5 mm from the target surface using a spectrograph with $\lambda$/$\Delta \lambda$ $\approx$50000. In addition to PCR, principal component analysis (PCA) was performed and mean absolute percent error (MAPE) was calculated. A MAPE of 4.11 % was achieved for the 75/25 train/test split and data collection parameters used to generate the calibration curve shown in Fig. 11.

 

Fig. 11. Calibration curve generated from PCR of Li I spectra.

Download Full Size | PDF

Prior work by Wood and Shattan noted that self-absorption can impact calibration curves at high $^6$Li fractions of 0.855 - 0.950 for spatially-integrated measurements performed in 40 mTorr Ar [11]. A slight saturation effect was observed at higher $^6$Li fractions which they attributed to self-absorption. Here, no such flattening is observed, which can be due to several factors, including lower $^6$Li fractions were analyzed in this work, and different experimental parameters (e.g., ambient gas pressure, spatially-resolved analysis, timing) were used for data collection. While the results shown in Fig. 11 are promising, there are several additional parameters that can be explored to improve the $^6$Li/$^7$Li predictions from emission spectra. Factors that can be considered include methods for improving signal-to-noise ratio, such as performing spatially-integrated measurements. In addition, factors such as number of spectra to save, and how many laser shots are used to generate each spectrum can be explored while trying to reduce data collection time, and avoiding target drilling.

4. Conclusion

This work reports the impact of ambient gas and pressure on line broadening and line distortions of Li I 670.8 nm emission spectral features in Ar, N$_2$, and He at various pressure levels. At high background pressure levels, self-reversal was evident for all background gases used although self-reversal effects were more severe in N$_2$ in comparison to He or Ar. Self-reversal was found to be significantly reduced at low-pressure levels of $\approx$0.1 Torr for all gases. At low-pressure levels, FWHM of the Li I 670.8 nm transition reaches a minimum value of $\approx$ 33-40 pm in all gases even though the spectrograph provides an instrumental resolution of $\approx$12 pm. The presence of fine and hyperfine structures, a $\approx$15.8 pm isotopic shift, and line broadening mechanisms inherent in LPPs (particularly Doppler) lead to partially resolved spectral profiles for samples containing varying Li isotopic ratios. These results also indicate that it is very challenging to obtain well-resolved Li isotopic shifts even with the use of higher spectral resolution emission spectroscopy instrumentation considering spontaneous emission from the plasma requires higher temperature conditions. Hence, active sensing methods such as LAS or LIF of laser ablation plumes, which can monitor the Li populations at lower temperatures (thereby reducing the Doppler effect) with negligible instrumental width, are required to observe clearly-resolved isotopic splitting of $^6$Li and $^7$Li. Finally, we report the Li isotopic spectral measurements with optimized plasma conditions for varying $^6$Li/$^7$Li ratios. We demonstrate that with the use of chemometrics we can predict $^6$Li/$^7$Li ratios.