Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Contents lists available at ScienceDirect Spectrochimica Acta Part B: Atomic Spectroscopy journal homepage: www.elsevier.com/locate/sab Comparison of light capturing approaches in Laser-Induced Breakdown Spectroscopy (LIBS) for multichannel spectrometers María Gabriela Fernández-Manteca a, Marina Martínez-Minchero a, b, Asier García-Escárzaga d, e, Alain A. Ocampo-Sosa a, f, Jesús Mirapeix a, b, c, José J. Valdiande b, José Miguel López-Higuera a, b, c, Luis Rodríguez-Cobo a, b, c, Adolfo Cobo a, b, c, * a Instituto de Investigación Sanitaria Valdecilla (IDIVAL), Santander, Spain b Photonics Engineering Group, Universidad de Cantabria, Santander, Spain c CIBER de Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Instituto de Salud Carlos III, Madrid, Spain d Department of Prehistory, Institute of Environmental Science and Technology (ICTA-UAB), Universitat Autònoma de Barcelona, Bellaterra, Spain e Department of Archaeology, Max Planck Institute for Geoanthropology, Jena, Germany f Servicio de Microbiología, Hospital Universitario Marqués de Valdecilla, Santander, Spain A R T I C L E I N F O A B S T R A C T Keywords: LIBS technique requires the spectroscopic analysis of the light emitted by a laser-induced plasma plume. One LIBS spectroscopy challenge of the different approaches to capture the plasma light emission is the significant shot-to-shot varia­ Optical plasma emission tions of the plume inhomogeneities, position, and morphology. This is even more challenging when multichannel Multichannel spectrometer CCD spectrometers are used, because the light should be homogeneously divided among multiple capturing Köhler optics optical fibers (typically up to 8 fibers) with stable spectral efficiency for all channels. Otherwise, any further analysis of the atomic emission spectra involving multiple channels, such as line intensity ratios, Boltzmann plots, or calibration-free LIBS, could be compromised by the morphology-dependent spectral artifacts induced by the collection optics. In this work, we assess the performance of several collection optics in terms of overall capturing efficiency and channel-to-channel variations due to changes in plasma morphology. Results clearly show that this could be an issue even with the approaches with the best spatial homogenization, including optical fibers and Köhler optics. 1. Introduction Many optical solutions have been proposed in the literature [4]. A common approach involves a single or multiple lens arrangement that Laser-Induced Breakdown Spectroscopy (LIBS) offers information on projects the plasma image over the slit of the spectrometer. With this the composition of a sample at the atomic level by the spectral analysis solution, shape variations are translated and sampled at the slit plane. of the optical emission of a plasma induced by a pulsed laser generated Besides, atomic emission lines for some elements are mainly located in during the material ablation [1]. For typical samples, plasma has the the ultraviolet spectral range (<400 nm), so finding suitable achromatic shape of a plume, up to several millimeters in size, but with significant lenses is difficult, and chromatic aberration could produce shot-to-shot shot-to-shot variations of the plume position, morphology, and spatial spectral variations. An alternative is the use of a pierced parabolic inhomogeneities of electronic temperature and density [2,3]. This rep­ mirror [8] that focuses the plasma emission onto the spectrometer resents a challenge for the capturing optics that carries the optical without chromatic aberration and high efficiency due to the large area of emission to the spectrometer. The broad spectral range of interest, from the mirror. However, any imaging optical arrangement could suffer from the far ultraviolet (<200 nm) to the near-infrared (>900 nm), also af­ variable collected light intensity due to plasma shape variations in each fects the collection optics design. Common applications (such as quan­ laser shot. titative analysis) try to perform a spatial integration of the light emitted This problem could be mitigated by using an optical fiber at the by the plume while reducing chromatic aberrations so that shot-to-shot image plane that integrates the captured light over their core area and variations can be minimized. numerical aperture [5] and, after guiding, produces a stable light * Corresponding author at: Instituto de Investigación Sanitaria Valdecilla (IDIVAL), Santander, Spain. E-mail address: adolfo.cobo@unican.es (A. Cobo). https://doi.org/10.1016/j.sab.2023.106617 Received 16 November 2022; Received in revised form 9 January 2023; Accepted 10 January 2023 Available online 14 January 2023 0584-8547/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Fig. 1. LIBS experimental setup. “spectral artifacts” the unwanted shot to shot amplitude variations in different spectral ranges of the captured spectrum that cannot be attributed to naturally occurring changes in plasma emission but to changes in the light splitting performed by the collection optics that affect the capturing efficiencies of each channel. One proposed solution uses a fiber homogenizer, a short (typically about 5 cm in length) silica rod placed in front of the fiber bundle [10]. More complex arrangements have been proposed, particularly the use of the Köhler illumination principle. This arrangement of lenses is typically used in microscopes to obtain a diffuse illumination at the sample plane irrespective of the shape and light spatial distribution of the light source [11]. In the case of LIBS capturing optics, a Köhler configuration can project the plasma emission on a homogenous spot at the object plane, thus significantly reducing shot-to-shot fluctuations. A LIBS-specific design was proposed recently by Lucchi et al. [12]. It is based on two plano-convex lenses, and achieves an excellent homogenization perfor­ Fig. 2. Photograph of the experimental model of the plasma plume. mance. However, it has the drawback of a large spot at the object plane (~4 mm size) that results in poor collection efficiency at the optical fiber pattern over the spectrometer slit. Choosing different focal lengths of plane for typical fiber core diameters. An improved design with high lenses, plasma images smaller or larger than the core size can reduce the homogenization and a smaller spot (i.e., <1 mm diameter) would need intensity fluctuations. A more common arrangement in published ex­ more lenses and typically larger working distances for the optics. Be­ periments is the use of a bare multimode, high core diameter (up to 1 sides, attaining good performance over the required wide spectral range mm) optical fiber connected to the spectrometer, placed at several is always challenging. millimeters from the plasma. This produces a stable capturing efficiency, In this work, we assess the performance of several collection optics in almost free from chromatic aberrations, and with sufficient capturing terms of overall capturing efficiency and channel-to-channel variations efficiency due to the proximity to the plasma plume [6]. due to changes in the shot-to-shot plasma morphology for a widely used However, the problem of shot-to-shot fluctuations is even more LIBS setup involving a multichannel (8 channels) spectrometer. challenging when a multichannel CCD spectrometer is used, i.e., a set of multiple independent but synchronized CCD spectrometers in different 2. Materials and methods spectral ranges, that require a fiber bundle to split the collected plasma emission among all of them. A non-comprehensive random survey of 50 2.1. Experimental setup LIBS experiments published in 2022 shows that 50% of them use this kind of instrument. Incidentally, all LIBS systems deployed or to be Experiments were performed with the LIBS setup shown in Fig. 1. deployed in Mars rovers are based on multichannel spectrometers [7]. We used a Q-switched Nd:YAG laser (Lotis LS-2134D-MH) emitting This widespread use can be attributed to their combination of broad 12 ns pulses at 1064 nm with a 10 Hz repetition rate. The laser beam was spectral range, high resolution, high optical throughput, and low cost focused on the sample with a fused silica lens of 75 mm focal length. The [9]. sample could be moved in three axes with a motorized stage, and the For multichannel spectrometers, plasma emission should be captured sample's surface was observed with a coaxial RGB camera. The optical by a bundle of optical fibers (typically up to 8 fibers), and different emission of the plasma was captured by the different optics used in this channels analyze different spectral regions. Shot-to-shot spectral varia­ study and sent to an 8-channel CCD-based spectrometer [10] ULS2048- tions of the captured spectra are unavoidable in LIBS due to changes in USB2-RM) with a total spectral range from 178 to 889 nm, and variable plasma properties, but a homogenous, repeatable splitting of light in­ resolution between 0.015 and 0.06 nm. An 8-fiber bundle with fused tensity and spectral content among the channels is expected. Otherwise, silica multimode optical fibers of 200 μm core diameter and 220 μm any further analysis of the atomic emission spectra involving different cladding diameter was used to split the incoming light to the different channels, such as line intensity ratios, Boltzmann plots, or calibration- channels. The fibers of the bundle are arranged in a circular pattern with free LIBS, could be compromised by the morphology-dependent spec­ a total diameter of 800 μm. tral artifacts induced by the collection optics. We understand by To verify the influence of plasma morphology variations, we 2 M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Fig. 3. Example of Iron LIBS spectrum. The inset shows the spectral range overlapped in two adjacent channels. Amplitude difference can be attributed to the different gratings and blaze wavelength sin each channel. designed an experimental model of the plasma plume. It is based on a undesirable shot-to-shot channel splitting variations. small (4 mm height by 3 mm wide) incandescent light bulb covered by a Polytetrafluoroethylene (PTFE) diffuser. This lamp has a wide spectral 2.2. Collection optics range of emission and size slightly larger than a typical plasma plume. It was placed on an XYZ automatic positioner, and its position was scanned Several designs of collection optics have been tested. All of them within a 2x2x2mm cube to simulate shot-to-shot changes in the plasma have been added to the optical fiber bundle that splits the light to the plume. Fig. 2 shows a photograph of the experimental model. eight channels of the spectrometer (the fiber bundle comprises eight For each position, the overall capturing efficiency was derived from fibers of 200 μm core diameter each one), and placed at an angle of the wavelength integration of channels 4–7 (360–680 nm), while the about 45◦ with respect to the horizontal. The tested optical arrange­ homogenization performance was estimated from the mean intensity ments are the following (See Fig. 5): ratio at three overlapped wavelength ranges: 399–401 nm between channels 4&5, 489–492 nm (channels 5&6), and 561–564 nm (channels 1. Nothing added: the 8-fiber bundle is placed close to the plasma 6&7). Best arrangements correspond to those with higher collection plume (10 mm distance) and aligned for maximum captured light efficiency and more stable intensity ratios at channel edges, that is, less intensity. dependence of the intensity ratio on the shift in position. 2. A multimode, 1-m length, 1 mm core diameter homogenizing Additionally, LIBS experiments were performed on an ingot of pure fiber is connected to the fiber bundle, and one extreme is placed Iron obtained in a levitation induction furnace to assure >99.99% pu­ close to the plasma plume (10 mm distance). In this case, the rity. Iron was chosen due to the high number of emission lines and single overall diameter of the 1 mm fiber is slightly larger than the fiber element composition. The ingot was cut flat and polished to further bundle diameter (800 μm), so most of the captured light is uni­ assure sample homogeneity. To verify if the light reaching different formly projected on the eight cores of the fiber bundle. The nu­ channels changes on a shot-to-shot basis, we rely on the fact that a merical aperture is the same for the optical fiber and the fiber spectral range of a few nanometers is duplicated in adjacent channels of bundle (NA = 0.22). the spectrometer. This overlap allows us to monitor changes in the 3. An aspheric UV-enhanced collimator (model 74-UV, Ocean collected intensity for the same wavelengths in different channels, a Insight) is placed before the 1 mm diameter homogenizing fiber situation that should not occur in the case of an ideal collection optics. and aligned to the plasma plume. To perform these experiments, the plasma emission from a burst of 100 4. The same aspheric UV-enhanced collimator is directly connected laser shots (30 mJ pulse energy) at five spatial points of the sample's to the fiber bundle without the homogenizing fiber. surface were collected with the 8-channel spectrometer, using 1 μs gate 5. A commonly used arrangement of two plano-convex lenses with delay and 1 ms acquisition time. The crater size was 200 μm in diameter, 50 and 40 mm focal lengths, respectively, is aligned with the resulting in an irradiance of 8 GW/cm2. First 5 spectra in each spatial plasma plume at the focal distance of the first lens (50 mm), and point were discarded as cleaning shots. Those conditions are represen­ the plasma plume is imaged onto the 1 mm homogenizing fiber tative of a typical LIBS experiment. with magnification M = 0.85. Due to the expected size of the Fig. 3 shows an example of the captured spectrum from the Iron plasma, its image on the surface of the optical fiber is of the order sample. The inset depicts the Fe I emission lines at 561.56 and 562.45 of the fiber core. nm, captured in both channels 6 and 7. Other Fe I emission lines in the 6. The same optical arrangement of two lenses (50 mm + 40 mm) is overlapped range were identified and measured: 399.74 nm and 400.52 used directly with the fiber bundle, without the 1 mm homoge­ nm in channels 4&5, and 490.33 nm in channels 5&6. Tracking the in­ nizing fiber. tensities ratio of the same emission line in adjacent channels indicates 3 M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Fig. 4. Four Köhler designs optimized for collection efficiency, homogenization, and spectral range of operation. Fig. 5. Summary of the tested optical arrangements for capturing the plasma emission. Table 1 Table 2 UV fused silica plano-convex lenses used in the optical arrangements #5-#12 to Comparison of the overall (spectrally integrated) capturing efficiency of the capture the plasma emission [13]. *Design wavelength is 588 nm for all lenses. different optical setups (relative to the one with the highest performance). Reference Focal Diameter Radius of Center Edge Collection optics Capturing efficiency Length (inches) Curvature Thickness Thickness Only 8-fiber bundle 80.9% (mm)* (mm) (mm) (mm) 1 mm fiber 76.1% LA4148 50.2 1′′ 23.0 5.8 2.0 Aspheric collimator +1 mm fiber 67.1% LA4306 40.1 1′′ 18.4 7.1 2.0 Aspheric collimator 78.8% LA4725 75.3 1” 34.5 4.4 2.0 Imaging 2 lenses 50 + 40 mm + 1 mm fiber 75.7% LA4102 200.7 1” 92.0 2.9 2.0 Imaging 2 lenses 50 + 40 mm 82.5% LA4647 20.1 1/2” 9.2 4.3 1.8 Imaging 2 lenses 50 + 75 mm + 1 mm fiber 88.0% LA4936 30.1 1/2” 13.8 3.4 1.8 Imaging 2 lenses 50 + 75 mm 100.0% Köhler 4 lenses +1 mm fiber 46.4% Köhler 4 lenses 51.7% 7. An arrangement of two plano-convex lenses with 50 and 75 mm Köhler 2 lenses +1 mm fiber 73.5% Köhler 2 lenses 77.2% focal lengths is aligned with the plasma plume at the focal dis­ tance of the first lens (50 mm), and the plasma plume is imaged onto the 1 mm homogenizing fiber with magnification M = 1.5. In this case, the size of the plasma plume image larger than the 4 M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Fig. 6. Measurement of the light intensity ratio in overlapped spectral windows for different optical arrangements. The vertical axis is the normalized ratio. Average is performed for the three spectra channel edges. Only a representative selection of measurements is shown: a) 1 mm optical fiber, b) Collimator added, c) Two-lenses Köhler design; d) Four-lenses Köhler design, e) Bare 8-fiber bundle, f) Two-lenses (50 + 40 mm focal lengths). 5 M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Table 3 efficiency and, therefore, affect the signal-to-noise ratio of emission lines Comparison of the homogenization performance of the different optical ap­ and any subsequent qualitative or quantitative analysis. We used the proaches with a single figure of merit calculated as the maximum variation of the experimental model of the plasma plume (small light bulb) to compare light intensity ratio averaged at the three overlapped spectral windows. The the capturing efficiency of the different optical configurations. Collec­ lower the ratio variation, the better. tion efficiency was estimated from the spectral integration of channels 4 Collection optics Avg. ratio variation to 7 (360 to 680 nm), as most of the spectral content of the incandescent Only 8-fiber bundle 59.4% bulb is concentrated in that range. Table 2 shows a summary of the re­ 1 mm fiber 5.3% sults. The spectral efficiency is shown relative to the best-performing Aspheric collimator +1 mm fiber 10.8% configuration, which is the two-lenses optics with 50 and 75 mm focal Aspheric collimator 34.4% lengths. This result was expected due to the large size of the first lens (1′′ Imaging 2 lenses 50 + 40 mm + 1 mm fiber 41.1% Imaging 2 lenses 50 + 40 mm 94.2% diameter) and the low numerical aperture (NA = 0.19). The configu­ Imaging 2 lenses 50 + 75 mm + 1 mm fiber 26.5% ration with the lowest performance is the 2-lenses Köhler design, with a Imaging 2 lenses 50 + 75 mm 94.7% relative efficiency of approximately 50%. Interestingly, the bare optical Köhler 4 lenses +1 mm fiber 23.7% fiber of 1 mm core diameter placed at 10 mm from the plasma plume Köhler 4 lenses 37.0% reached a 76% relative efficiency. Nevertheless, there is only a two-fold Köhler 2 lenses +1 mm fiber 19.0% Köhler 2 lenses 50.4% difference between the worst and best performing arrangement, which should have a minor impact in a typical LIBS experiments. core of the capturing fiber, so only a portion of the plasma area is 3.2. Shot-to-shot variations in the captured light sampled. 8. The same optical arrangement of two lenses (50 mm + 75 mm) is To estimate the ability of each optical configuration to reduce the placed directly before the fiber bundle, without the 1 mm ho­ shot-to-shot variations, we have measured the light intensity detected in mogenizing fiber. the overlapped spectral ranges of different channels and its variation 9. A Köhler design with two lenses proposed by Lucchi et al. [12], due to plasma movements. The experimental model of the plasma plume placed before the 1 mm fiber. It is a simpler design but with lower (small light bulb) was placed at different points within a cube of capturing efficiency. 2x2x2mm. At each spatial point, the light intensity was calculated by 10. The same Köhler design of two lenses, but directly connected to integrating the signal at the following overlapped spectral windows: the fiber bundle. 399–401 nm for channels 4 and 5; 489–492 nm for channels 5 and 6; and 11. A Köhler design based on four lenses is placed before the 1 mm 561–564 nm for channels 6 and 7. For ideal capturing optics, the ratio of homogenizing fiber. The design is shown in Fig. 4, and details are light intensity of two channels at the same spectral window should given in the following paragraphs. remain constant irrespective of the plasma position if a perfect ho­ 12. The same Köhler design of 4 lenses, but directly connected to the mogenization of the light over the 8-fiber bundle is expected. Any fiber bundle. change in the light intensity ratio in those measurements will translate to spectral artifacts in the captured spectrum in LIBS experiments with Collection optics #11 in the previous list is a new Köhler design real plasma plumes. proposed in this work that tries to optimize the collection efficiency, Fig. 6 shows a selection of the light intensity ratio measurements for homogenization, and spectral range. Several designs were analyzed representative cases, including the best and worst-performing optics. using Zemax's OpticStudio® to assess their performance for this appli­ The vertical axis shows the normalized light intensity ratio at the indi­ cation; they are shown in Fig. 4. cated spectral window and channels. Only the horizontal spatial scan­ The first two designs - (a) and (b) in Fig. 4 - use two lenses and ning of ±2 mm of the model of the plasma plume was considered. achieve a good homogenization and a small spot size (slightly over the It can be seen from Fig. 6 that the best-performing optics is the 1 mm 800 μm diameter target to illuminate the 8-fiber bundle fully), for a core-diameter optical fiber, with a maximum variation of the light in­ spectral range from 300 to 900 nm. However, smaller wavelengths are tensity ratio of about 5% (averaged in the three overlapped spectral spread over a larger area, as seen in the simulated spot diagrams for the windows at the channel edges). In contrast, other configurations, such as 200 nm wavelength in Fig. 4 (a)’ and (b)’. This behavior can be the two-lenses arrangement, show much more variations, close to 100%. explained by the sharp refractive index increment of fused silica in the Adding a collimator to the optical fiber results in higher variations UV region [13]. An improved performance was found with the four (lower homogenization), while the Köhler designs do not provide a lenses design shown in Fig. 4 (c), achieving a maximum spot size of 800 better homogenization as expected. μm for the entire wavelength range from 200 to 900 nm. This is the To get an overall ranking of the performance of the different ap­ Köhler design #11 implemented for the experiments. The other Köhler proaches, for each one, the maximum variation of the light intensity design - Fig. 4 (d) - corresponds to the one proposed by Lucchi et al. It ratio was calculated at the three spectral windows, and the results were has excellent spectral and spatial homogenization, but with a spot size of averaged to get a single figure of merit. The results are shown in Table 3. 4 mm, thus resulting in poor collection efficiency with typical optical From Table 3, it can be seen that the best approach is the use of the fiber sizes. Table 1 shows the details of the lenses chosen for the designs. bare 1 mm optical fiber connected to the 8-fiber bundle. As expected, For each optical arrangement, an experiment with the model of the adding this homogenizing fiber improves the metrics for all configura­ plasma plume and a LIBS experiment on an Iron sample were performed. tions. However, the two lenses arrangement, widely used in reported In the next section, we show the results concerning the overall light- LIBS experiments, results in very poor performance if no homogenizing capturing efficiency of each one and, most importantly, their perfor­ fiber is used. Additionally, Köhler designs do not improve homogeni­ mance in terms of the shot-to-shot variations of the channel splitting of zation over using an optical fiber as capturing element. Most interest­ light and its impact on the LIBS spectrum analysis. ingly, the best-performing approach still has a 5.3% mean variation of the light intensity ratio at the same spectral window in adjacent chan­ 3. Results and discussion nels, meaning that the effect of variable light splitting is not negligible and should be considered in the analysis of the LIBS spectral data. 3.1. Overall light capturing efficiency Different approaches for light capturing result in different optical 6 M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Fig. 7. Correlation between the intensities of the same emission line of Fe I when measured by two adjacent channels in the overlapped spectral ranges. Only the results for four of the configurations are included (including the one with the best and worst performance). In each row, three charts corresponding to one emission line in each spectral range are shown, from top to bottom: a) 1 mm optical fiber, b) bare 8-fiber bundle, c) two-lenses arrangement (50 + 75 mm focal lengths) and d) two-lenses Köhler design. 3.3. Impact on LIBS experiments nm, 561.56, and 562.45 nm collected in the overlapped spectral ranges. Fig. 7 shows a representative set of the results from all the tested To assess the impact of shot-to-shot plasma plume variations in the collection optics. In each chart, the intensity of the same line, measured inter-channel distribution of light in LIBS experiments, we have per­ in two different channels, is plotted for all the laser shots, and the cor­ formed ablation experiments with the above-described setup on a relation coefficient R2 is calculated. For ideal capturing optics with no sample of pure Iron. For each design of collection optics, we calculated spectral artifacts, a value of 1 is expected. It can be seen that, again, the the intensity of the same emission line of Fe I at 399.74, 400.52, 490.33 use of a bare fiber optic before the 8-fiber bundle gives the best 7 M.G. Fernández-Manteca et al. Spectrochimica Acta Part B: Atomic Spectroscopy 201 (2023) 106617 Table 4 taken into account in any further spectral processing, for example, for Performance of all the tested collection optics in terms of inter-channel varia­ any quantitative LIBS analysis involving emission lines in different tions of the emission lines in LIBS experiments. channels. Collection optics Avg. R2 coeff. Avg. RSD Only 8-fiber bundle 0.74 7.2% Declaration of Competing Interest 1 mm fiber 0.92 4.1% Aspheric collimator +1 mm fiber 0.74 4.9% Aspheric collimator 0.83 4.0% The authors declare that they have no known competing financial Imaging 2 lenses 50 + 40 mm + 1 mm fiber 0.77 5.7% interests or personal relationships that could have appeared to influence Imaging 2 lenses 50 + 40 mm 0.74 7.2% the work reported in this paper. Imaging 2 lenses 50 + 75 mm + 1 mm fiber 0.91 3.9% Imaging 2 lenses 50 + 75 mm 0.76 7.1% Köhler 4 lenses +1 mm fiber 0.89 4.3% Data availability Köhler 4 lenses 0.59 7.2% Köhler 2 lenses +1 mm fiber 0.81 6.6% Data will be made available on request. Köhler 2 lenses 0.52 10.3% Acknowledgments homogenization, with R2 values close to 1 in all the three overlapped spectral ranges (first row of Fig. 7). In contrast, the two-lenses Köhler This work was supported by the R + D project PID2019-107270RB- design has the lowest R2 values (last row of Fig. 7). It is clear from the C21 (funded by MCIN/ AEI /10.13039/501100011033) and by Plan figures that shot-to-shot ratios of the selected emission lines suffer a Nacional de I + D + and Instituto de Salud Carlos III (ISCIII), Sub­ significant variability that only can be attributed to an unwanted vari­ dirección General de Redes y Centros de Investigación Cooperativa, able light splitting between the channels. Ministerio de Ciencia, Innovación y Universidades, Spanish Network for A summary of the performance for all the tested collection optics is Research in Infectious Diseases (REIPI RD16/0016/0007), CIBERINFEC shown in Table 4. In this table, the correlation coefficients (as shown in (CB21/13/00068), CIBER-BBN (BBNGC1601), cofinanced by European Fig. 7) have been averaged for the five emission lines at the three Development Regional Fund “A way to achieve Europe”. A. O.-S was different spectral ranges. Additionally, the relative standard deviation financially supported by the Miguel Servet II program (ISCIII- (RSD) of the ratio of the emission lines has been calculated for the 100 CPII17–00011). AGE was supported by the Catalonia Government laser shots and averaged for the five emission lines. It can be seen that throughout a Beatriu de Pinós fellowship (grant number 2020 BP two designs give the best results: single 1 mm optical fiber (#2) and the 00240). We thanks Dr. A. Picón for providing the pure iron ingot and same fiber with added imaging lenses (#7), in both cases with an helpful suggestions to improve the experimental setup. averaged R2 coefficient of 0.9 and averaged RSD of 4%. The additional lenses also improve the collection efficiency. Removing the homoge­ References nizing fiber greatly increases variability in all cases. These results show that, even for one of the best homogenization [1] L. Radziemski, D. Cremers, A brief history of laser-induced breakdown spectroscopy: from the concept of atoms to LIBS 2012, Spectrochim. Acta B At. approaches to prevent shot-to-shot inter-channel variations, some small Spectrosc. 87 (2013) 3–10. and non-negligible spectral artifacts are still present, and this should be [2] C. Aragón, J.A. Aguilera, Characterization of laser induced plasmas by optical taken into account in any further spectral processing, for example, for emission spectroscopy: a review of experiments and methods, Spectrochim. Acta B At. Spectrosc. 63 (9) (2008) 893–916. any quantitive LIBS analysis involving emission lines in different [3] Y. Fu, Z. Hou, T. Li, Z. Li, Z. Wang, Investigation of intrinsic origins of the signal channels. uncertainty for laser-induced breakdown spectroscopy, Spectrochim. Acta B At. Spectrosc. 155 (2019) 67–78. 4. Conclusions [4] T. Li, S. Sheta, Z. Hou, J. Dong, Z. Wang, Impacts of a collection system on laser- induced breakdown spectroscopy signal detection, Appl. Opt. 57 (21) (2018) 6120–6127. We have analyzed different designs of collection optics typically used [5] S.V. Shabanov, I.B. Gornushkin, J.B. Winefordner, Radiation from asymmetric by the LIBS research community to study their performance when used laser-induced plasmas collected by a lens or optical fiber, Appl. Opt. 47 (11) (2008) 1745–1756. with multichannel spectrometers routinely employed to increase LIBS [6] R. Noll, Laser-Induced Breakdown Spectroscopy: Fundamentals and Applications, experiments' spectral range and resolution. Springer, 2012. The tested configuration of collection optics involves several optical [7] R.C. Wiens, X. Wan, J. Lasue, S. Maurice, Laser-induced breakdown spectroscopy in planetary science, in: Laser-Induced Breakdown Spectroscopy, Elsevier, 2020, arrangements placed before the fiber bundle used by this kind of spec­ pp. 441–471. trometer: homogenizing single-core optical fiber, two lens systems with [8] D. Redoglio, U. Perini, S. Musazzi, LIBS collection optics: Comparative analysis of different focal lengths, and two Köhler designs with improved homog­ different mirror-based configurations, in: In 2014 Fotonica AEIT Italian Conference on Photonics Technologies, IEEE, 2014, May, pp. 1–4. enization capabilities. [9] J.E. Carranza, E. Gibb, B.W. Smith, D.W. Hahn, J.D. Winefordner, Comparison of To assess the homogenization performance of each setup, we nonintensified and intensified CCD detectors for laser-induced breakdown leverage the overlapped spectral range in several adjacent channels of spectroscopy, Appl. Opt. 42 (30) (2003) 6016–6021. [10] https://www.avantes.com/products/fiber-optics/fiber-optic-homogenizer/. the spectrometer. As they measured the same wavelengths, any shot-to- Accessed 14 November 2022. shot variation could be attributed to homogenization issues. [11] R. Fischer, B. Tadic-Galeb, P. Yoder, Optical System Design, McGraw-Hill Our results show that even for one of the best homogenization ap­ Education, 2008. proaches to prevent shot-to-shot inter-channel variations (namely, a [12] J. Lucchi, M. Martinez, M. Baudelet, Homogenization of plasma emission collection for multichannel spectrometers, Appl. Spectrosc. 73 (10) (2019) 1228–1236. long single-core optical fiber without any additional lenses), some small [13] “UV Fused Silica Plano-Convex Lenses, Uncoated.” Thorlabs, https://www.thorlab but non-negligible spectral artifacts are still present, and this should be s.com/newgrouppage9.cfm?objectgroup_id=123. Accessed 14 November 2022. 8