High-precision mapping of fluorine and lithium in energy materials by means of laser-induced XUV spectroscopy (LIXS) Spectrochimica Acta Part B: Atomic Spectroscopy

Laser-induced breakdown spectroscopy (LIBS) is a well-established elemental analysis method, thanks to negligible sample preparation, rapid analysis, and a spatially resolved sensitivity down to trace level, in any kind of matrix. State-of-the-art LIBS is operated in the optical spectral range (UV-VIS). Unfortunately, the measure- ment precision is limited by the moderate stability and repeatability of the plasma emission. The detection and sensitivity to crucial elements such as light elements and halogens are also limited. This is particularly critical for inhomogeneous materials where signal fluctuation is related to the spatial elemental distribution. To overcome these disadvantages specific, LIBS techniques arrangement are often required. Laser-induced XUV Spectroscopy (LIXS) has some intrinsic advantages for overcoming some of the above mentioned limitations and it can support the spectroscopic information collected in the UV-VIS range.


Introduction
Laser-induced breakdown spectroscopy (LIBS) is a well-established elemental analysis method. The conventional LIBS is operated in the optical spectral range (UV-VIS) and allows a rapid simultaneously detection of most of elements in a solid sample [1][2][3][4]. LIBS has also the ability for 2D spatially resolved mapping as well as depth profiling at a given location showing a local 3D mapping [1,[5][6][7][8]. Nevertheless, LIBS has also some drawbacks related to the measurement uncertainty and to the repeatability of the signal [2,3,9,10], and low of sensitivity for crucial elements such as halogens. The limited precision is due to various sources of noise, such as (i) noise due to the inhomogeneity in the laser-plasma interaction and plasma evolution; (ii) shot noise generated by the number of photons collected at the detector; (iii) the detector noise; (iv) the instrumental (thermal) drift [9]. In the last decades several efforts have been made to improve the precision of the technique and generally RSD ranges in a wide interval from 5% to 30% depending on several parameters such as experimental conditions (background environment, pressure etc.), chemical and physical characteristics of the sample (morphology, composition, homogeneity etc.), duration of the laser pulse, focusing conditions, spectral features of the analyzed elements, detection parameters (gate width and delay time, number of accumulations). Also peculiar experimental set-up have been proposed for improving the LIBS performances. As examples it was shown previously that double-pulse irradiation has improved the signalto-noise ratio in LIBS [11,12]. Moreover, an interesting hyphenated approach to enhance the signal and LIBS's sensitivity and stability, was also introduced, where the LIBS was coupled with Raman spectroscopy [13,14].
In addition, there are still some intrinsic challenges on a few key elements with poor sensitivity. For instance, it remains challenging to detect fluorine (F) and other halogens through LIBS due to the high excitation energy of F, absorption of its emission in air, as well as the moderate limit of detection (LOD) [13,15,16]. It is possible to detect fluorine with atomic transition lines in the NIR region, but it suffers from low intensity [16], corresponding to a low precision. These limitations of fluorine detection were overcome by using the F emission lines in the VUV range under vacuum. In the VUV region, expensive optics based on MgF 2 and CaF 2 materials are required due to the strong absorption of silica based light collection and imaging optics [16]. For this reason, the most common method for detecting the halogens remains the molecular detection at late stage of laser plasma emission temporal evolution, that in any case shows moderate sensitivity.
As another example, lithium (Li) is widely deployed in many industrial fields such as Li ion batteries, catalyst, photovoltaics, and other energy materials. The synthesis of Li-Manganese Oxide energy materials is rapidly performed by means of pulsed laser deposition (PLD) [17]. PLD is well appreciated for its stoichiometry, although the mass contrast between Li and Mn is often an issue for a congruent deposition and must be carefully determined in situ. Li cannot be easily quantified accurately and precisely by the conventional solid state technologies, such as X-ray fluorescence (XRF) and energy dispersive X-ray spectroscopy (EDX) [18]. For EDX, there is still a challenge to detect soft X-rays at 55 eV [18]. In contrast, inductively coupled plasma mass spectrometry (ICP-MS) has high accuracy and low LOD, which is used for Li quantification [19]. However, ICP-MS usually requires a complex sample preparation and is limited to bulk analysis [20][21][22][23]. The in-situ detection with ICP-MS is based on laser-assisted sample introduction by transporting a dry aerosol. Laser ablation ICP-MS offers excellent detection limits, but mixing effect of the dry aerosol transport limits the capabilities for low scale depth profiling at high precision [20,22].
The significant interest on a rapid and high-precision in-situ analysis of F and Li, e.g. in energy materials, requires more development. In this study, Laser-Induced XUV Spectroscopy (LIXS) was carried out for the first time. Laser plasma emission under low pressure in the XUV range (5-20 nm) was observed, from irradiation of energy materials. The emission of hydrogen-like Li 2+ (Li III) ionic line at λ =13.5 nm is strong. The element F also exhibits few ionic lines FVII and FVI in the XUV range. In the present study, lithium fluoride (LiF) was used for referencing, as well as several mixture samples (Li 2 O/Mn x O y ). The aim of this study was to investigate LIXS as an advanced method for detection of halogens and Li in energy materials. The LIXS setup's performance is tested on LiF and Li-Mn-O matrix samples. The paper is organized as follows: section 2 introduces the experimental setup, including the laser setup and the XUV spectrometer. Section 3 deals with the experimental results LIXS in XUV and UV-VIS, as well as the precision analysis of XUV and UV-Vis signals. Moreover, the quantitative analysis was carried out from the XUV and UV-VIS results for comparing the different peculiarities of the laser induced emission spectrum in the XUV and UV-VIS spectral ranges. Section 4 contains the conclusion with an outlook on possible further studies.

Laser ablation set-up
The experimental setup consisted of a Q-switched Nd:YAG (λ =532 nm) laser source (Q-smart, Lumibird) with a pulse duration of 5 ns, repetition rate 10 Hz and a pulse energy in the range of 100 mJ for plasma ignition, a sample holder, an UV-VIS spectrometer (190.0-603.9 nm) and a self-built flat-field XUV spectrometer (5.0-20.0 nm), described elsewhere [24]. The laser beam was focused about 1 mm below the sample surface resulting in a laser spot size of 200 μm. The pulse-to-pulse fluctuation in laser energy is <5%. The UV-VIS signals were acquired with a delay time of 0.2 μs, and a gate width (camera integration time) of 10 μs. The laser ablation and XUV spectrometer must be operated within the low pressure system because of the absorption of XUV in air, at an operating pressure of 10 mPa.
As shown in Fig. 1, the laser beam is focused on the sample with lens L1 (LA1253, Thorlabs, f = 200 mm) in the low pressure chamber. The sample is facing to the incoming laser beam in order to get a homogeneous ablation. The LIBS-XUV is acquired from an XUV spectrometer in the radial direction of the laser-induced plasma. Similarly, the UV-VIS is collected by a collimator and an optical fiber in a radial direction at a 5 • angle.
For the alignment of the set-up, a pilot laser was used, which has the same beam path as the laser for the ablation. The position of the laser focus on the target has to be carefully adjusted to the correct position for a high-quality XUV signal.

Echelle spectrometer (λ = 190.0-603.9 nm)
The emission from the plasma plume can be collected by using a collimator (Avantes, UV/VIS) and an optical fiber (Ocean Optics), then guided into an echelle UV-VIS spectrometer (   The XUV spectrometer consists of a source pinhole (diameter 2.5 mm) to delimit the source size, a collimation slit, a gold-coated concave variable-line-spacing (VLS) grating (Hitachi model 001-0437), and a back-illuminated X-ray CCD detector (Greateyes GE2048 512 BI UV1). The 1200 mm − 1 VLS grating compensates for the curvature of the diffracted wavefront to obtain a flat-field on the camera. The schematic of the spectrometer is shown on the left side in Fig. 1. In order to get a sharp spectrum on the CCD camera, the slit was adjusted to 50 μm. The incident angle needs to be carefully adjusted to a specific standard setting value (87 • ) as explained in a previous publication [24]. The backilluminated CCD is placed vertically at the focal plane of the spectrometer after the grating to collect the XUV spectrum. The CCD has a 2048 × 512 array with a pixel size of 13.5 × 13.5 μm 2 . The CCD camera is triggered with a 5 V pulse and 100-ms integration time.
A raw spectrum is a 2D image (X,Z) that contains the information on the line position as pixel number coordinate along the grating dispersion plane (X), and the emission divergence given by the slit width (Z). The 16-bit greyscale gives the amplitude of the observed spectral lines. Henceforth, the pixel coordinate and the grey value must be calibrated into wavelength (or photon energy) and number of photons, respectively. The calibration of the XUV spectrum with the help of NIST atomic database can be found in ref. [24].

Sample materials 2.4.1. LiF
LiF was chosen as a reference material due to the detection challenge put forward by the elements Li and fluorine (F). The 2-in. LiF plate had a thickness of 4 mm, which was supplied by Golem IMS GmbH. LiF is known as an important optical material due to its extreme low refraction index of infrared and extreme high transmission for UV light [25]. Besides, it is also a promising salt for a stable Li ion battery electrolyte [26].
LiF has a high bandgap of 13.6 eV, which makes singe photon ablation difficult. Furthermore, since there are only three electrons in the Li atomic shells, the number of transitions of neutral and ionic Li is much lower than that of heavier elements. In particular, it is known that Li III has a strong emission line at λ =13.5 nm [27]. Thus, LiF was supposed to be an appropriate sample for wavelength calibration and start of this study.

Stoichiometric mixtures Li 2 O/Mn x O y
In  Table 1.

Oxide samples
In order to reveal the "fingerprint' of oxide within the 5-20 nm wavelength range by LIBS-XUV, beside to the previous calibration samples (Li 2 O/Mn x O y ) and Li-Mn battery material, NIST glass 612 was also measured with the LIBS-XUV. NIST glass 612 is a glass support matrix with 61 trace elements, which has a nominal composition of 72% SiO 2 , 14% Na 2 O, 12% CaO, and 2% Al 2 O 3 (mass fractions) [28].

LIXS spectra
The spectra in the XUV (5-20 nm) and UV-VIS (193-603 nm) ranges from laser-induced plasma emission were acquired simultaneously without temporal evolution under low pressure. Although these experimental conditions are not optimized for elemental analysis, the obtained spectra can be used for a direct comparison of the main characteristics of the laser induced plasma spectrum in the two investigated spectral ranges. In Fig. 2, the average of 20 shots is shown for a LiF laser plasma in both spectral ranges. In the XUV (Fig. 2.a), the strongest spectral line at λ =13.5 nm is generated by the Li III ion, corresponding to the transition of 2p-1 s. Moreover, the Li III line at λ =11.4 nm and the Li II line at λ =16.7 nm were also observed in the XUV range. In addition to the Li lines, there are also several clear emission lines visible from the F VI and F VII distributed in the wavelength range of 9-15 nm. In the UV-VIS range (Fig. 2.b), the strong transitions are generated from the neutral Li and singly ionized F atom, such as the strongest Li I line at 460.3 nm and the strongest fluorine line F II at 424.6 nm. The F lines in the VIS has lower intensity than the Li lines due to its high excitation energy, while no lines were observed in the UV because the absorption of the UV emission in the air. In short, it is clear that LIXS is able to observe the transition of the higher ionization stages in the plasma while UV-VIS is able to observe the atomic and lower ionization stages in the same plasma. Moreover, the atomic and ionic Li lines show the highest line intensity in both spectral range, indicating the lower excitation energy of Li compared to F.
The electron temperature of the plasma was estimated for the XUV and UV-VIS spectra. This was accomplished with the Boltzmann plot in a LTE assumption [24]. The electron temperatures in XUV and UV-VIS were calculated to 15 eV and 0.9 eV, respectively. The difference in electron temperature from XUV and UV-VIS is based on the rapid early plasma expansion (ns time scale). The XUV signal is in fact generated from a very hot pristine plasma, with negligible background and no Bremsstrahlung. The UV-VIS radiation instead emerges from a "cooler" aged plasma (10 μs time scale in this experiment). The dynamics of the plasma expansion affects the emission characteristics of the late stages of the temporal evolution [10], while at the "zero instants" (<10 ns), the plasma is significantly more stable and in turn the corresponding spectra can be more reproducible. Moreover, the suppression of the background has a drastic effect on the improvement of the precision of the LIXS. Henceforth for the F lines in Fig. 2, the one-and-a-half order of magnitude higher plasma temperature and higher electron density led to the collisional excitation/ionization in the XUV, which is impossible to be accomplished in UV-VIS.
To obtain more insights on spectral lines, the Li III and F VI lines are shown in Fig. 2.c and Fig. 2.d. The FWHM of both Li III and F VI lines is approximately 0.05 nm, corresponding to a relative linewidth of (λ/∆λ) ~300, possibly limited by the spectrometer resolving power. On the other hand, The FWHM of Li I and F II lines are 0.08 nm and 0.06 nm, respectively. The corresponding spectral resolutions (λ/∆λ) are 5800 and 7100, respectively. The UV-VIS spectrometer has one order magnitude higher spectral resolution than the XUV spectrometer.

Repeatability and precision
Within the 20 laser shots, replicated in several measurement campaigns over months, LIXS showed higher line repeatability and precision as compared to those in the UV-VIS at the experimental condition used in the present experiment (Fig. 3). To gain insight on the repeatability and precision of the LIXS and UV-VIS measurements, n = 20 single shots were delivered to the LiF sample with the same laser energy (100 mJ). The line of Li III at λ = 13.5 nm and F VI at λ = 14.0 nm were selected for where I is the line intensity from the spectra and n is the number of repetitions of the experiments. In the present experiment, the RSD of Li III and F VI lines from the LIXS were found to be 7.1% and 10.7%, respectively. On the other hand, the RSD of Li I and F II lines were found to be as high as 23.2% and 52.6%, respectively. The emission lines in XUV delivers better precision thanks to the low background noise. In fact, even for a Planck-like distribution of the plasma emission, the largest amount of photons emitted in an XUV spectrometer is negligible. Moreover, the plasma inhomogeneity and plasma evolution plays an enormous role in the repeatability of UV-VIS especially when long detection time is applied [10] and under a unconfined expansion as found under the low pressure condition. The background noise level is attributed to the shot noise, detector noise and source noise, which is discussed further in supporting information in details. Although, as mentioned in the introduction, the analytical performance of LIBS depends on specific experimental arrangement and optimization of the analytical procedures, it has been reported that atmospheric LIBS in the UV-VIS can show high signal uncertainty and poor repeatability [9]. As previously reported the noise in LIBS consists of source noise, shot noise, detector noise and drift. The precision of the line intensity due to fluctuations in the laser plasma is called "source noise". Fu et al. [10] have reported that the significant LIBS spectral signal fluctuation (20-30% RSD) is attributed to the plasma morphology or plasma evolution, total number of density, as well as the delay time. The plasma morphology shows a poor repeatability at a later stage of the plasma, which required an appropriate delay time of the acquiring of LIBS signal. These experimental results, where temporal resolution has not been applied, clearly demonstrate that the improvement of the RSD observed in XUV with respect to UV-VIS, is mainly due to the fact that the signal of UV-VIS mainly comes from the aged stage of laser-induced plasma, which results in the poor stability and precision of UV-VIS. On the contrary, the XUV is generated during the very early stage of the laser-induced plasma, which shows a much higher stability and precision. The source noise can affect all the spectral features to some extent, such as the occurrence of transitions and continuum emission [9]. Plasma inhomogeneity may also contribute to erratic signals in the LIBS data.
Besides the source noise, the shot noise can also impact the signal precision. In this experiment, as mentioned in the experimental section, an ICCD detector was applied in UV-VIS and a CCD detector was applied in LIBS-XUV. The light collection efficiencies are different during the measurement, according to the different slit width as well as the gain of a MCP (Micro-Channel Plate) in ICCD. The intrinsic noise analysis of CCD and ICCD detector is plotted in Fig. S1. In order to reduce the contribution of the shot noise to the total RSD of the measurements, the highest possible signal intensity must be reached by optimizing the experimental conditions. Therefore, in these experiments, with reference to the highest intensity Li III and Li I lines, the precision of XUV is three times better than that of UV-VIS.

Quantitative analysis on battery materials
The calibration curve for quantifying the Li concentration from the measured LIXS intensities of Li III (13.5 nm), was derived with a set of calibration samples, as well as the Li concentration with UV-VIS intensity of Li I (460.3 nm). Fig. 4 shows the LIXS spectra measured with a laser pulse energy of 200 mJ. The laser energy has been doubled to increase the short-wavelength spectral intensity, because the LIXS intensity of the Li 2 O/Mn x O y calibration samples produced by the 100 mJ laser energy resulted in a lower SNR. In the XUV spectra, clear Li, Mn, and O emission lines were observed in the spectral range of 5 to 20 nm, which originated from four Li 2 O/Mn x O y calibration samples, respectively. The line of Li III at λ = 13.5 nm with the highest intensity (160 counts) was selected to provide clean signal and best precision among all spectral lines. On the other hand, the UV-VIS spectrum of the calibration sample is plotted in Fig. 4b, where a large number of spectral lines of Li, Mn and O can be observed. The line Li I at 460.3 nm with the highest intensity (5670 counts) is also suitable for the Li concentration calibration.
As a result, the intensities of these two reference lines at each calibration sample are displayed in Fig. 5. A linear relationship (R 2 = 0.97) was obtained between the Li concentration and XUV intensity ratio of Li III lines derived from the XUV spectra, as shown in Fig. 5a. The slope of the linear function is 12.4 counts/ wt%, showing the Li detection sensitivity in XUV. In UV-VIS, it showed higher sensitivity of 716.9 counts/ wt% as compared to XUV. This is closely related to the higher line intensities obtained in UV-VIS spectra. However, both XUV and UV-VIS show not only the capability of qualitative analysis, but also the ability to quantify the concentration of the functional elements of the battery materials. The LOD in the calibration samples was calculated using the following equation [6,17,29]: where s B is the standard deviation of the sampled background (counts) that is found to be 0.5 for XUV and 8.0 for UV-VIS, and m is the slope sensitivity of the linear calibration curve. The standard deviation of the background is sampled from three background signals within the spectral range (13.4 nm -13.6 nm and 460.2 nm -460.4 nm) of the peak of interest. The obtained LOD was 0.12% and 360 ppm for Li in XUV and UV-VIS, respectively. This is in agreement with published data [6], where the LOD for Li was determined in pegmatite minerals. In Table 2, the precision, sensitivity and LOD of XUV and UV-VIS are listed. If the LOD of XUV is poorer, the point-to-point precision is better. This is important for spatially resolved analysis of elements well above the detection limit. The precision of XUV signals was found to be a factor of 3 higher than that of UV-VIS signals.

"Fingerprint" of oxidation
Li is known to be prone to rapid oxidation effects during sample storage and/or handling. In order to avoid artefacts, such as that the analytical information is evaluated in respect to the condition in the original environment, one needs to pay attention to the occurrence of oxidative film. In the spectra of two oxide samples (LM25 and NIST glass 612), three transitions (A, B and C lines) of the O VI at 12.99 nm, 15.01 nm and 17.31 nm were observed in the laser-produced plasma, corresponding to 1s 2 2p-1 s 2 4d, 1s 2 2s-1 s 2 3p, 1s 2 2p-1 s 2 3d transitions, as shown in Fig. 6. These three lines showed relative high SNR among the spectral lines and occurred in each spectrum of these materials. On the other hand, these three lines were not observed in the LiF sample, indicating that the LiF sample was not oxidized. These three O VI lines can be considered as a rapid fingerprint of the oxide matrix to automate a rapid recognition with high reliability. The results may also allow the identification of transient oxidation layers of the measured material by using LIXS. Moreover, this fingerprint of oxidation could be potentially applied as validation for the XUV spectrometer calibration.

Conclusions
LIXS, i.e. collection of the XUV spectral emission of a laser-produced plasma, is as a novel method for element mapping in solid samples of intractable analytes such as Li and/or F. The LIXS signals (up to 7% of RSD) were observed to be 3 times more precise than the concomitant UV-VIS signals at the experimental condition employed in this work (without temporal resolution, laser ablation under low pressure and 20 accumulations). Li III has a stable and intensive emission line at 13.5 nm which can be used as the reference line for LIXS quantitative analysis of Li in solid material. Meanwhile, the F VI line at 14.0 nm can be also  were carried out with XUV and UV-VIS to implement the quantitative analysis. The calibration curve is obtained by linear fitting of the Li line intensities to Li concentration. In addition, the limit of detections (LOD) of XUV and UV-VIS for Li -Mn containing material were also calculated. The large difference in LOD of XUV and UV-VIS makes to conclude that LIXS is currently suitable for non-trace element analysis, where the spatial distribution is the crucial information.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.