Review of Element Analysis of Industrial Materials by In-Line Laser—Induced Breakdown Spectroscopy (LIBS)

: Laser-induced breakdown spectroscopy (LIBS) is a rapidly developing technique for chemical materials analysis. LIBS is applied for fundamental investigations, e.g., the laser plasma matter interaction, for element, molecule, and isotope analysis, and for various technical applications, e.g., minimal destructive materials inspection, the monitoring of production processes, and remote analysis of materials in hostile environment. In this review, we focus on the element analysis of industrial materials and the in-line chemical sensing in industrial production. After a brief introduction we discuss the optical emission of chemical elements in laser-induced plasma and the capability of LIBS for multi-element detection. An overview of the various classes of industrial materials analyzed by LIBS is given. This includes so-called Technology materials that are essential for the functionality of modern high-tech devices (smartphones, computers, cars, etc.). The LIBS technique enables unique applications for rapid element analysis under harsh conditions where other techniques are not available. We present several examples of LIBS-based sensors that are applied in-line and at-line of industrial production processes. operation during a three-month test run has demonstrated the potential of LIBS for in-line process analysis.


Introduction
Laser-induced breakdown spectroscopy (LIBS) is a versatile technique for the analysis of the chemical composition of many classes of materials. Solids, liquids, gases, powders, biological, and organic material, aerosols, and micro and nanoparticles are investigated and the type and abundance of chemical elements in such materials is determined by LIBS [1][2][3][4][5][6][7][8]. This laser-based analytical method is fast and robust, enables multi-element detection, does not require laborious sample preparation, and can be employed for field measurements under harsh conditions. For these reasons, LIBS is becoming one of the key techniques for element analysis of complex materials besides X-ray fluorescence spectrometry (XRF), spark optical emission spectrometry (spark OES), prompt gamma neutron activation analysis (PGNAA), and others [9]. LIBS has enormous potential for various applications and LIBS-based sensors are employed in many areas: industry (e.g., for materials analysis on-site or in-line with production processes), security (e.g., for remote detection of hazardous materials such as CBRNE threats [10]), mineralogy and geological materials [11,12], cultural heritage [13,14], biomedicine (e.g., identification of bacteria [15][16][17][18]), deep-sea inspection [19,20], space exploration (e.g., for analysis of Martian rocks using the LIBS sensors in ChemCam and SuperCam on board of the NASA Mars rovers Curiosity and Perseverance [21][22][23][24]), environmental measurements [25][26][27][28], and many others. The number of scientific papers on LIBS that are published per year is increasing rapidly ( Figures A1 and A2 in Appendix A).
In LIBS, material is pulsed-laser ablated from the surface of a sample and the optical radiation of the laser-induced plasma (LIP) is analyzed by optical emission spectroscopy [29,30]. Short nanosecond laser pulses are employed for the sampling of material, typically, as stable and robust nanosecond laser sources operating at various wavelengths, pulse energies and repetition rates are available. However, ultrashort femtosecond lasers offer several advantages for LIBS compared to conventional nanosecond lasers and femtosecond-LIBS experiences a growing number of applications (e.g., [31][32][33][34][35][36][37][38]). In most studies, the optical emission of atomic and molecular species in the plasma is analyzed spectroscopically. Furthermore, recent studies have successfully demonstrated the remote isotope analysis by LIBS-based molecular isotope spectroscopy [39,40].
For the quantitative analysis of the chemical composition of plasma and sample calibration-based methods are frequently employed. Certified reference materials (CRM) are measured with the LIBS system and calibration curves for the analyte species are derived by univariate or multivariate regression procedures. With chemometric methods and machine learning techniques, large spectral data sets can be evaluated and the extraction of information from multidimensional data cubes can be automated [41,42]. The calibrationbased approach enables quantifying trace concentrations (ppm range). However, matrixmatched reference materials have to be measured under the same conditions as the sample materials. With laboratory-based analytical methods such as laser-ablation inductively coupled plasma mass spectrometry/optical emission spectroscopy (LA-ICP-MS/OES) much smaller element concentrations (ppb range) can be resolved [43][44][45][46][47]. Calibration-free (CF) LIBS is complementary to calibration-based LIBS and can be employed to quantify the major elements in a sample material without measuring CRMs. In CF-LIBS, the concentration of major components is derived directly from measured LIBS spectra by modeling the laser-induced plasma and the optical plasma emission [48,49]. This approach is of interest as matrix-matched calibration samples are not required and constraints regarding the control of experimental parameters are less stringent.

Plasma Emission and Multi-Element Detection
For a laser-induced plasma that is optically thin and in local thermodynamic equilibrium (LTE), the number of photons ∆n ki emitted within the time interval ∆t due to transitions between two atomic quantum states k and i is given by [Equation (1)]: ∆n ki /∆t = N A ki g k exp[−E k /k B T e ]/Z(T e ), (1) where N is the number of neutral atoms of the respective chemical element in the plasma, A ki the transition probability, g k the degeneracy factor of state k, E k the energy of the upper level of transition, k B the Boltzmann constant, Z(T e ) the partition function, and T e the electron temperature (= plasma temperature). The number of detected photons n det depends strongly on the measurement setup and the instrumentation used [Equation (2)]: n det = ∆n ki /∆t × (∆Ω/2 π) τ g γ det.
The solid angle covered by the collection optics is ∆Ω, the measurement time (gate width) is τ g , and the total detection efficiency is γ det (which includes efficiencies of the light guiding system with mirrors, lenses, and fibers, of the optical imaging, and of spectrometer and electronic detector). As an example, we discuss the ablation of an iron sample containing 10 ppm of carbon impurity and the optical emission of carbon in the laserinduced plasma. With nanosecond lasers, the ablated mass per pulse is typically around 300 ng, corresponding to 3.2 × 10 15 Fe atoms and N = 3.2 × 10 10 C atoms (ignoring ions in the following). The most intense emission line of C is in the UV range at wavelength λ ki = 193.09 nm and the corresponding spectroscopic parameters are A ki = 3.7×10 8 s −1 , g k = 3, and E k = 7.7 eV. For a typical plasma temperature of T e = 9000 K the partition function of neutral C is Z = 9.8. The C atoms are emitting ∆n ki /∆t ≈ 1.8 × 10 14 photons per second. For a typical LIBS setup equipped with an Echelle spectrometer and ICCD detector and with ∆Ω/2π ≈ 10 −2 , τ g ≈ 10 µs, and γ det ≈ 1.25 × 10 −5 , the total detected emission signal for carbon is n det ≈ 430 photons. As a consequence of such rather low signals, the limits of detection (LOD) and of quantification (LOQ) in LIBS are typically in the ppm range (or higher, depending on the element considered). With double-pulse LIBS the emission of the laser-induced plasma can be enhanced and lower LOD values are achieved [1,6]. The measured signals depend also on the type of sample material (matrix effect) and very different limits for the same analyte may be obtained for different matrices. The measured LOD and LOQ values for many analyte elements in different matrices can be found in the literature [50]. The LOD values for solid materials are summarized in the table of elements in Figure A3 (Appendix A).
The compositional analysis of industrial materials requires multi-element detection capability. Modern electronic components used in mobile phones, computers, and cars are comprised of up to 60 chemical elements, for example. In LIBS, many different elements can be measured at the same time, making this technique especially suited for the analysis of materials with complex composition. The number of photons emitted from a laser-induced plasma per atom in the plasma and per time can be calculated considering the spectroscopic parameters of the involved atomic transition and the plasma temperature [51]. From Equation (1) this element-specific emission rate (EMRA) is calculated by [Equation (3)]: EMRA = (∆n ki /∆t)/N = A ki g k exp[−E k /k B T e ]/Z(T e ).
(3) Figure 1 shows the calculated emission rate of most chemical elements for a homogeneous and optically thin plasma in (local) thermodynamic equilibrium at T e = 10,000 K. Figure 1. Number of photons emitted per atom and second for optically thin and homogeneous LTE plasma at temperature T e = 10,000 K. This emission rate of atoms EMRA is calculated for most chemical elements using the most intense emission lines of neutral species (upper part). The ionization energy of neutral atoms versus the atomic number of elements (lower part). The sign "*" stands for multiplication.
For each element, the most abundant isotope is selected and at least the two most intense emission lines of neutral atoms are evaluated [52,53]. Ions are not considered here as most species are neutrals in LIBS plasma measured at longer delay times with respect to the laser pulse. The emission rate varies by almost 12 orders of magnitude for the different elements. The temperature T e = 10,000 K is typical in LIBS measurements employing nanosecond laser pulses with energy density (fluence) Φ = 1−100 J/cm 2 for ablation. At very high nanosecond laser fluence of Φ~10 4 −10 5 J/cm 2 the plasma species can have high ionization stages (e.g., Ti 12+ and Sn 11+ ). The corresponding plasma temperatures are around 10 5 −10 6 K and plasma emission is obtained in the extreme UV and soft X-ray spectral ranges. This regime has so far not been explored for technical applications in analytics. The lower part of Figure 1 displays the ionization energy of neutral atoms. The plasma emission and the ionization energies of elements are correlated. High energy levels are sparsely populated and atoms with high excited state and ionization energies have relatively weak emission.
In Figure 2 the emission rate of atoms in ideal plasma is presented according to the group of chemical elements in the periodic table. The numbering of groups follows the scheme of the International Union of Pure and Applied Chemistry [54]. In group 1, the alkali metals (lithium to francium) show intense emission (EMRA ≈ 10 6 − 10 7 photons/atom × sec), whereas the hydrogen emission lines are weak in comparison (EMRA ≤ 10 4 photons/atom × sec). The alkaline earth metals (beryllium to radium, group 2), the transition metals (group 3 including lanthanoids and groups 4−12), and the elements in group 13 (boron to thallium) have rather intense emissions. From group 14 to 18 the emission rates show a trend to lower values and much larger variations for different elements within the same group. For halogens (group 17) and noble gases (group 18) the lightest elements have the faintest emissions with EMRA ≈ 100 and 10 −3 photons/atom × sec for fluorine and helium, respectively. Elements with high plasma emission intensities such as alkali metals, alkaline earth metals, and lanthanoids are measured with low LOD values, for example~10 ppm for Mg and Na and~2 ppm for Y. For halogens, on the other hand, the emission intensity is much lower and the reported LOD values are much higher, for example~300 ppm for F,~1000 ppm for Cl, and~5000 ppm for Br. In real LIBS plasma, the emission intensities may be significantly lower than the calculated values. Radiation self-absorption, plasma non-homogeneity, plasma-chemical reactions, and ejection of non-luminous particles from the sample material are influencing the optical emission of the plasma. A secondary excitation of the laser-induced plasma can enhance the optical signals, homogenize the plasma [55] and improve the detection sensitivity. Techniques employed for secondary excitation in LIBS include double laser pulses [56][57][58][59][60][61], electric spark discharges [62][63][64][65], and radiofrequency [47, 66,67], and microwave [68][69][70] radiation.
The reported LOD values of the different elements measured in solid samples and the calculated emission rates of elements are correlated. There is a clear trend to lower LOD for higher EMRA for most elements ( Figure A4 in Appendix A).

LIBS Analysis of Industrial Materials
The rapid development of stable and robust high power laser sources, of efficient and broadband spectrometers, and of fast and sensitive detection systems has stimulated many scientific studies on LIBS in the last years. Various research groups and business enterprises are developing LIBS systems, components, and software for technical applications in the field, i.e., out-of-laboratory. Many classes of materials that are used in large scale in industrial production are investigated by LIBS. Some materials classes are listed in the following in alphabetical order. For each class of materials some references to recent scientific publications are given.

Technology Materials
Materials and energy are among the most important resources in industrial production and are the basis for the development of new high-tech materials, novel products, and advanced devices. Technology metals are especially important to produce such materials and devices and the future need for various metals is increasing. Neodymium, for example, is used mainly to produce permanent magnets and laser crystals and the need for Nd in year 2030 is expected to be 3.8× larger than the present world production of this metal [431]. Plastics are another important resource material due to their high versatility and their use in many areas. Plastics have grown enormously in importance over the last few decades and 359 million tons of plastics were produced worldwide in 2018.

Precious Metals and Minerals
The production volume of high-tech devices such as smartphones, computers, high energy density rechargeable batteries, flat panel displays, clean energy applications, and components for cars and e-cars is strongly increasing. For the industrial fabrication of such devices, large amounts of special materials are required. For example, the number of smartphones sold worldwide to end-users increased from 122 Mio devices in 2007 to more than 1520 Mio devices in 2019 [432] (for details see Figure A5 in Appendix A). Around 1300 Mio smartphones (with 1300 Mio rechargeable batteries) and 300 Mio PC's and Laptops (with 140 Mio rechargeable batteries) have been sold worldwide in 2008. The production of only these devices consumed 3, 16, and 23% of the annual global mine production of metals Ag, Pd and Co, respectively [433]. More than 50% of the global mine production of noble metals Pt and Pd and more than 80% of Rh are used only for exhaust gas catalyst systems of cars.
The global production of precious and special metals ("technology metals") by mining has drastically increased in recent years to cope with the demands. Figure 3 shows the global mine production of some technology metals in years 1990-2017 normalized to the total production since 1900 (compiled from data of the U.S. Geological Survey [434]). For many metals more, than 80% of the total material mined since 1900 has been produced in the last few decades [433,435,436]. The available reserve of metals (in units of years) is estimated from the materials reserve (tons) known in 2019 and the amount of material (tons/year) produced in 2017 [434]. For several metals the reserve is only a few decades. The supply of more than 50 chemical elements of economic value is considered to be at risk [437]. An example of the depletion of natural resources is indium. In year 2007 the global production of In metal was 510 tons, the known global reserves 11,000 tons, and the estimated remaining time until exhaustion 19 years [431]. This risk clearly conflicts with the need of In, which is expected to increase by 3.3× (until 2030) over the present world production just for the fabrication of displays and photovoltaic devices. . Global mine production of some technology metals in years 1990 to 2017 normalized to the production since 1900 (REO is rare earth oxide; PGM is platinum group metals including Pd, Pt, Ir, Os, Rh, and Ru; Si refers to silicon metal). Reserve calculated from known materials reserve and annual mine production. Data taken from [434].
Technological metals and other materials are continuously transferred from natural geogenic resources to anthropogenic resources (i.e., products, goods, buildings and other infrastructure, waste in man-made deposits) and to the environment. The global reserve for copper, for instance, is around 955 Mio tons. In 1920, 930 and 25 Mio tons were contained in geogenic and anthropogenic resources, respectively. In year 2000, the distribution has drastically changed to 495 Mio tons in geogenic resources, 360 Mio tons in anthropogenic resources, and 105 Mio tons spread in the environment [438].

Plastics Materials
The use of polymers and plastics in industry and many other areas has been strongly growing over the last decades. In 2018, the plastic production reached 359 million tons worldwide and 62 million tons in Europe [439]. China is the biggest producer of plastics worldwide (30%). Most of the plastics in Europe (EU-28 + Switzerland and Norway) are used for packaging, building/construction, automotive purposes, electronics, household, and agriculture. The share of polymers by resin type is for polyethylene (PE-LD and PE-HD) 29.7%, for polypropylene (PP) 19.3%, for polyvinyl chloride (PVC) 10%, for polyurethane (PUR) 7.9%, for polyethylene terephthalate (PET) 7.7%, for polystyrene (PS) 6.4%, and for others (ABS, PBT, PC, PMMA, PTFE, etc.) 19%, see Figure 4. The increasing demand for plastics requires efficient treatment of the post-consumer plastic waste in order to avoid uncontrolled release of the material into the ecosphere, e.g., as marine litter [440]. The collection of post-consumer polymer and plastic waste reached 29.1 million tons in Europe in 2018 (including 17.8 Mt plastics packaging waste). This waste material was recycled (32.5%), used for energy recovery (42.6%), and disposed as landfill (24.9%), see Figure 4. The evolution of plastic post-consumer waste treatment in Europe from 2006 to 2018 shows a strong increase of materials recycling (4.7 to 9.4 Mt) and energy recovery (7.0 to 12.4 Mt) and a substantial decrease of landfill (12.9 to 7.2 Mt). The European Strategy for Plastics in a Circular Economy [441] is aiming to transform the production and use of plastic material and products and to further increase the recycling rates for plastic waste. Different pathways of plastic recycling are assessed to reduce the emission of greenhouse gases [442]. The PVC fraction must be removed from the collected plastic waste prior to waste treatment. The removal of PVC protects machines that are processing recyclate material from damage and avoids the formation of reactive substances (e.g., HCl) and toxins in energy recovery processes.

Secondary Raw Materials
The exploitation of natural deposits and the production of primary raw materials most likely cannot be scaled up with the increasing demands. Depletion of (known) natural reserves, low abundance, and economic and environmental issues are obstacles in increasing the mine production. Secondary raw materials are produced by the recycling of end-of-life products and articles (e.g., waste), of by-products (e.g., from industrial production), and of materials after their initial use. The recovery of raw materials from anthropogenic resources is an important strategy to avoid severe shortage of commodities and to stabilize the materials supply chains. "Urban mining" of secondary raw materials has substantial economic and environmental impact as materials are used more efficiently and materials flows become manageable (circular economy [443], impact on materials flow cost accounting [444,445]). The mining of urban resources is also more energy-efficient than the production of primary raw materials. The mining of gold (average Au concentration in ore~5 g/ton) produces approx. 17,000 tons of CO 2 per ton of Au metal [446]. The CO 2 gas emission is in large part due to the energy required to extract the ore from deep lying natural deposits and to process it. Printed circuit boards (PCB) of computers and cell phones, on the other hand, have Au concentrations of 150 and 300 g/ton, respectively, and are easily accessible [447]. Another example is the recycling of Cu and Al, which saves more than 85% and up to 95%, respectively, of the energy required to produce the primary raw materials by mining.
The production of high-quality secondary raw materials from waste streams requires the inspection of the waste composition and the separation into different fractions and materials classes. In the recycling industry, various sensor technologies are employed in sensor-based sorting [448]. The sorting systems are using optical sensors (e.g., detecting the color or absorption/reflection/fluorescence of waste pieces) [449], X-ray transmission sensors (e.g., measuring the atomic density and element composition), magnetic and Eddy current sensors (e.g., detecting Fe and non-Fe metals), spectroscopic sensors (e.g., detecting specific absorption bands in the infrared (IR) and emission lines and bands in the ultraviolet (UV) and visible (VIS)), gamma ray-based sensors, and acoustic sensors.

In-Line Application of LIBS in Industrial Production
The major strengths of the LIBS technique are its robustness and versatility and its capability for rapid and stand-off multi-element detection. This enables LIBS to be used for in-line chemical sensing in industrial production under harsh conditions. Laboratorybased element analytical techniques (e.g., LA-ICP-OES/MS) are more accurate than in-line compatible techniques such as LIBS. However, in-line techniques allow for continuous sampling, the reduction of sampling errors, and the real-time detection of rapid changes in the materials composition. Hence, analytical data obtained by a less accurate but continuously performed in-line measurement may have advantages over more accurate data measured in the laboratory that are less representative. In the following we present some recent examples for in-line, at-line, and in-situ measurements of various materials (alphabetical order).

Coal and Coal Ash
Coal is still one of the major resources for the global production of electricity. Around 39% of the total electricity production worldwide was based on coal in 2015 [450]. Moreover, coal is an important raw material in various industries. The combustion efficiency of coal depends on various material properties such as chemical composition (major elements C, H, O, N, and S), heat value, moisture content, volatile matter, fixed carbon, and ash content [92,451]. In order to optimize the efficiency of power generation and to reduce the environmental pollution (e.g., by emission of SO 2 ) a technology for rapid in-line or at-line chemical analysis is needed.
Coal is usually analyzed by standardized ASTM laboratory methods (American Society for Testing and Materials), which requires several days for the results to be obtained. Prompt gamma neutron activation analysis (PGNAA) and X-ray fluorescence (XRF) are methods enabling in-line analysis of coal. However, radiation safety issues are of relevance for both methods. Furthermore, PGNAA requires the use of an isotope source and XRF does detect only the heavier elements (Z ≥ 11). LIBS is a promising candidate for coal quality detection due to its advantages of real-time, in-situ, and multi-element measurement capability. A schematic of a LIBS measurement system installed in a coal-fired power plant is shown in Figure 5a [452,453]. The LIBS system was installed above a conveyor belt to analyze coal material that was transported from the crushing station to the coal bunkers (PPL Generation's Montour Power Station). For calibration, an artificial neural network (ANN) model was developed and trained with calibration samples (ASTM reference analysis). Figure 5b shows results from validation tests of the in-line LIBS system using approx. 120 coal samples that were grabbed and then analyzed by LIBS and a laboratory method. The LIBS results for the Sulphur concentration (red and black symbols, Figure 5b) were in good agreement with the lab reference analyses (black lines, Figure 5b). The measurement of S was required for optimal operation of the SO 2 reduction system of the plant. Besides S, the LIBS analyzer was detecting also Al, C, Ca, K, Mg, Na, Fe, Si, and Ti in real-time. The ash content in coal transported on a conveyor belt can be monitored online by PGNAA. However, the neutron radiation in PGNAA represents a potential health hazard requiring strict regulatory demands. LIBS does not require radioactive materials for operation and LIBS systems can be designed more compact and of less weight compared to PGNAA systems. Laser Detect Systems has developed one of the first mineral analysis systems using LIBS and has pioneered the in-line analysis of coal ash [110]. The system was installed at a conveyor belt of a coal mine in South Africa for a four-month field trial ( Figure 6a). The analytical performance of the LIBS system was compared to a PGNAA system installed in the same line. LIBS analyzed the surface of the coal material on the belt while the PGNAA signal depended on the volume of the irradiated material. The field trial demonstrated successful online coal ash content monitoring by the LIBS analyzer. The key elements in coal and coal ash, C, Mg, Al, Si, Ca, Fe, and Ti, were measured and the in-line ash quantification by LIBS and PGNAA had the same accuracy with a mean absolute error of ±0.5% (Figure 6b).
The concentration of unburned carbon in fly ash is an important criterion for evaluating the combustion efficiency of coal-fired power plants and the commercial value of fly ash as a secondary raw material. A high concentration of unburned carbon can have adverse impact on the combustion efficiency and on the value of the ash. The carbon concentration is presently measured off-line by manually obtaining fly ash samples from the precipitator ash hoppers or flue gas streams and sending the samples to a laboratory. Depending on the laboratory test procedure employed, the results may not be available for up to 24 h. This procedure takes several hours and delays the combustion optimization process. For online measurements a LIBS system and a two-stage cyclone measurement system were developed to quantify the carbon content in fly ash in real-time [106]. The cyclone system in combination with a 1 ns pulse-width laser enabled to eliminate the effect of CO 2 on the unburned carbon content. A schematic diagram of a boiler control system using LIBS in a coal-fired power plant for fly ash measurement is shown in Figure 7a. The unburned carbon content in different fly ash samples as measured by LIBS and by standard chemical analysis methods (Japanese Industrial Standards JIS) are compared in Figure 7b. For improved quantitative analysis the plasma temperature correction method was used. The results of measurements by LIBS and JIS method were consistent with R 2 = 0.9052 and RMSEP = 3.9% in the measurement range of 14.0 to 53.2% of unburned carbon. This demonstrates the feasibility of LIBS for real-time measurement of fly ash contents in power plants. Some earlier studies on the on-line analysis of unburned carbon in fly ash and of combustion products in industrial boilers and furnaces are reported in [454,455], respectively.

Metal Melts
The element analysis of metal melts in the production is usually performed off-line by laboratory-based analytical techniques such as XRF and spark OES. A liquid sample is taken and measured after cool-down and solidification. This is a time-consuming process, making the real-time monitoring of the melt composition impossible. With LIBS the melt can be measured directly, i.e., without sampling, and real-time analysis becomes feasible.
The chemical analysis of aluminum in a primary aluminum smelter has been reported recently [188]. The measurement system was installed at a casting launder system (Figure 8a). The melt temperature was around 730 • C. Fourteen trace elements (Fe, Si, Cu, Ni, Ti, Cr, Mn, Sn, V, Ga, Zn, Sb, Mg, Na) were measured in the melt and the results were correlated with laboratory measurements on corresponding solid samples. The trace elements Cu, Cr, Mn, and Sn were quantified down to ppm levels and volatile elements, e.g., Na, were measured in real-time down to ppm levels ( Figure 8b). It was concluded that the in-line LIBS analysis of many technically important trace elements in the primary aluminum melt was fully competitive with the off-line laboratory analysis of solid process samples in terms of accuracy and precision. Some earlier reports on the analysis of aluminum melts are [190][191][192]. The LIBS analysis of molten steel has been reported by several groups, e.g., in [194,[197][198][199][200][201][202][203]. In a recent publication, a hollow refractory lancet was immersed into the metal melt to pass the laser light through the surface slag layer onto the liquid steel underneath and to pass the optical radiation of the laser-induced plasma backwards to the optical detection system (using a Cassegrain telescope) and spectrometer (Figure 9a) [197]. The setup was designed to protect all optical and electronic equipment against the high-temperature environment near the steel ladle.  [197].
In laboratory test runs the elements Si, Mn, Cr, Ni, and V in molten steel samples were measured. In the steel plant the elements C (0.21-0.27 wt%), Si (0.52-0.63 wt%), and Mn (1.20-1.38 wt%) were analyzed quantitatively by LIBS and by spark OES as reference. The predicted concentration of test samples was close to the reference concentration with small relative root mean square errors of prediction RMSEP (Figure 9b). The results obtained in the steel plant were not as good as the results in the laboratory. However, the accuracy achieved in the in-line measurements approached the steel plant's requirements. From these results it was concluded that the developed LIBS system is promising for the in situ analysis of melt steel in the steelmaking industry. Immersion probes for LIBS analysis of liquid metals including steel were developed prior by other groups [199][200][201].

Minerals
For the exploration and efficient excavation of mineral quarries the composition of rock has to be measured during the drilling process. Analysis of minerals inside the drill hole is not feasible; however, the drill dust can be extracted with a dust hose and analyzed. Figure 10a shows a LIBS analyzer for continuous in-line analysis of the drill dust [102]. The LIBS system is mounted on a mobile drill rig. The results from a LIBS measurement campaign in a quarry are shown in Figure 10b. Dust samples were collected at different horizontal positions of the drill rig and at various drill depths. The Aluminum concentration measured as a function of drill depth and horizontal position varied from 0 to 1.5 m%. From such elemental maps a spatial model of the mineral deposit can be determined, and the excavation process can be optimized.
In another study, the ability of LIBS to provide in-line analyses of phosphate ores under industrial conditions was demonstrated [233]. Impurities in the phosphate rock significantly affect the ability to efficiently recover phosphate from the rock in the production plant and produce on-grade products. The most significant variables are CaO, MgO, Fe 2 O 3 , and Al 2 O 3 . A rugged LIBS sensor was developed and installed above a conveyer belt in an open phosphate mine (Four Corners Mine, FL, USA). A photograph of the installed LIBS system is seen in Figure 11a

Metal Scrap
Metal scrap is an important resource material for various industries. In 2017, approx. 57 million metric tons of selected metals were recycled from scrap in the U.S. (an amount equivalent to about 47% of the apparent supply of those metals). Iron and steel accounted for about 89% of the recycled metal and about 88% of the apparent supply [456]. The use of recycled scrap metals in place of virgin iron ore is beneficial to the environment (e.g., energy savings by 75%) and for every ton of new steel made from scrap steel approx. 1.1 tons of iron ore and 0.6 tons of coal are saved [457]. Aluminum is one of the few materials that is completely recyclable. The production of recycled aluminum is 92% more efficient than the production of new aluminum. Per year, 5 Mt of aluminum are recycled in the U.S. and Canada [458].
LIBS was applied for the in-line monitoring of steel scrap on a conveyor belt transporting the scrap metal to an electrical arc furnace (EAF) in a steel plant [102]. The real-time measurement of the content of key elements in the scrap allows stabilizing the process of furnace charging and steel making. The optical unit of the developed LIBS system was installed over the conveyor line ( Figure 12a). The main components included a 3D scanner and a laser light section sensor to measure the geometry of the scrap pieces. The scanner optics directed and focused the laser beam onto the scrap in a wide field of the cross-section of the conveyor (1.2 m × 2 m, varying filling level). The charging operation of the EAF is controlled by determining the mass flow of key elements (e.g., Si) from the element concentration measured by LIBS (Figure 12b). Parameters such as the belt speed, the filling height, and the average scrap density have to be taken into account to determine the element mass flow. The availability of data in real-time enables to adjust the charging process before the charging is finished. LIBS analysis of steel scarp was reported by other groups as well [215].
The LIBS technique has been applied also to the inspection of aluminum scrap for metal recycling [6,209,210,214]. Wrought and cast Al alloy pieces have been identified using a belt conveyor sorting system by measuring the LIBS signals for Al, Ti, and Si achieving a mass throughput of up to 4 tons/hour [6]. In another study, scrap of Al alloys containing different amount of Mg and Si was sorted by LIBS. For Al scrap pieces of 40-110 mm size the sorting throughput was 3-5 metric tons per hour and the sorting purity was 98% [209].

Nuclear Material
In the nuclear industry the analysis of nuclear and other materials before, during, and after production and utilization is required for safe and economic operation. This includes different processes in the nuclear fuel cycle such as mining of ore, fabrication of fuel, power plant operation, fuel reprocessing, and spent fuel storage (Figure 13a) [259]. Laser spectroscopy techniques such as LIBS, laser-induced fluorescence (LIF), and cavityring down spectroscopy (CRDS) are employed for analysis due to their elemental and molecular selectivity and high sensitivity. The inherent advantages of LIBS make it an efficient method for the analysis of hazardous samples in harsh environments. The nuclear industry is one of the fast-growing fields of LIBS application [248]. The development of stand-off LIBS systems enables for remote and in situ inspection of samples that are at large distance from the LIBS sensor (i.e., many meters).  Figure 13b shows a stand-off LIBS system using an optical telescope for the characterization of high-level radioactive waste at the THORP nuclear reprocessing plant in UK [263]. Optical access to the material was possible via a 1 m thick lead-glass radiation shield window. The LIBS system was used for remote identification of an unknown solid material that accumulated on the basket surface. The perforated basket was used in the processing of spent fuel. Remote analysis of this surface contamination was necessary due to the difficulties in taking a sample from behind the radiation shield and the subsequent difficulties with laboratory analysis. The LIBS analysis showed that the contaminant material was rich in zirconium and molybdenum (mainly zirconium molybdate which forms during the reprocessing of spent fuel).
Remote analysis of materials in nuclear fusion reactors is another important application for stand-off LIBS [249,257]. During operation of a tokamak fusion reactor, the inner walls of fusion chambers and divertors (plasma-facing components, PFC) are severely interacting with the plasma. As a consequence, the PFCs are subject to erosion, re-deposition of eroded material, and retention of fuel. The performance of fusion tokamaks such as ITER [459] can be influenced by such processes. Figure 14a shows the schematic of a stand-off LIBS system installed at the Experimental Advanced Superconducting Tokamak (EAST) [257,460]. The Nd:YAG laser beam (1064 nm, 5 ns, 180 mJ) was focused on the wall surface on the high-magnetic-field side using a quartz lens (f = 3 m) mounted at port H of the EAST device. The emission of LIBS plasma was collected in a backward direction using an optical telescope. The LIBS spectra measured in situ showed spectral signals of multiple elements (D, H, Li, Mo, W, Ti, La, Fe, and Si). The signals of Mo, W, C, and La were from the substrate materials of tiles, the signals of D and H came from the fuel (H was used for isotope experiments). The Li signal was caused by Li wall conditioning and Ti, Fe, and Si were due to impurities in the Li co-deposited layer.

Refractory Materials
Many industrial processes at high temperature such as metal making, furnace annealing, and sintering require high-temperature stable refractory products such as bricks, etc. The reuse and recycling of spent refractory materials have high potential to reduce the production of waste and the consumption of primary raw materials. The estimated amount of spent refractories is up to 28 million tons per year [461]. For high-grade recycling, the different types of refractory materials have to be identified and sorted with respect to their chemical composition and impurities have to be removed. The refractory materials are modified on the surface due to the interaction with the high-temperature processed material (e.g., liquid steel). Therefore, the surface layer is not representative for the bulk. LIBS has been used for the analysis of spent refractory materials [353,354]. When several laser pulses are applied on the same position of the specimen the contamination layers on the surface can be removed and the composition of the bulk material retrieved [208].
A demonstrator of a LIBS-based spent refractory sorting machine is shown in Figure 15a. The LIBS sensor is installed above the conveyor belt. An end-of-life refractory brick with a modified surface layer and unmodified bulk is shown in Figure 15b. Three LIBS measurement spots are marked with a white rectangle.

Rubber
Rubber has outstanding material properties such as mechanical elasticity, viscoelasticity, dielectric strength, thermal stability, resisting power against chemicals, morphological flexibility, and durability, and is used in a wide range of applications. About 70% of the annual global production of rubber is used for tire production and retreading [462]. A key parameter for the fabrication of rubber is the concentration of the vulcanization agents Sulfur and Zinc oxide (ZnO). The properties of the material (e.g., elasticity, stiffness, wear) depend on the amount of S and ZnO, and for the production of rubber of high quality the concentrations have to be controlled precisely in the process.
XRF and PGNAA cannot be employed for in-line measurements in the rubber production due to radiation hazards and other techniques were not available. Recently, LIBS has been employed for the first time to quantify ZnO and S directly in the tire rubber production [357]. The system was optimized to measure the optical emission of S and Zn from the rubber plasma in air (Figure 16a). Plasma excitation in collinear double-pulse geometry and detection of plasma emission with time-gated detectors was employed to resolve the weak sulfur lines in the near-infrared range. The element S and ZnO were quantified in three different sample materials (natural rubber NR, styrene-butadiene rubber SBR, and butadiene rubber BR) that were prepared from the most important polymers used in production (Figure 16b). The mean error of the prediction of concentrations RMSEP is ≤0.07 wt% for S and ≤0.33 wt% for ZnO for all polymer types. The results demonstrated that the vulcanizing system of rubber can be quantified under ambient conditions with a LIBS in-line sensor. Earlier attempts on tire rubber analysis by LIBS in the production are reported in [363,364].

Steel Grade Detection in Casting
In industrial steel production the casting of liquid steel into slabs, blooms, and billets is a frequently used process. The continuous casting of steel from different heats produces slabs that may have different chemical composition. The detection of different steel grades is important for the identification of the slabs. Moreover, the detection of transition zones from one steel grade to the other can improve productivity and cost-efficiency, e.g., by the reduction of steel waste. LIBS has been proposed for analyzing the chemical composition of cast steel and the at-line monitoring in the steel casting process has been successfully demonstrated [463][464][465].
The steel slabs are covered by different layers of varying thickness, which poses a substantial challenge for surface-analytical techniques such as LIBS. A schematic of LIBS at-line measurement of hot steel slabs in the continuous steel casting process is shown in Figure 17a. The bulk steel material (a) is covered by a scale (oxide) layer (c), mold powder (d), and dirt. The thickness of this surface layer can exceed several 100 µm. For LIBS analysis of the bulk material the surface layer has to be removed locally, e.g., by a sequence of laser pulses that precede the LIBS measurement and ablate the slab surface ("laser cleaning"). The LIBS signal of the Pb (I) line at 405.78 nm measured at-line on a hot steel slab in motion shows the transition from one steel grade to another (Figure 17b) [463]. The number of laser shots corresponds to the position of LIBS measurements on the steel slab along the casting direction (5 mm distance between two subsequent laser shots at pulse repetition frequency of 5 Hz). For comparison, the Ni (I) 341.47 nm line intensity did not change at the transition zone (inset of Figure 17b). The concentration of Pb in the two steel grades was 0 and 0.17 wt% (for Ni the concentration was 1.75 and 0.88 wt%). The LIBS signals obtained from the sample surface in real-time and the statistical analysis of signals allowed to discriminate special steel grades and to predict the distribution of elements in the intermixed transition zone of the cast slabs. The combination of LIBS measurements and Artificial Neural Network (ANN) methods for signal evaluation has also been used for the quantitative elemental analysis of cast steel along the slab length [465].

Steel Slags
Steel slags are multi-component oxide materials that are produced in large quantity in industrial steel production. For the control of the steelmaking processes and for the recycling of the metallurgical slag materials suitable analytical techniques are required. The standard method is XRF. However, LIBS requires less time for the analysis of slags than XRF [466]. For some applications the quantitative analysis of the major components of metallurgical slags is sufficient. This task can be accomplished by calibration-free LIBS (CF-LIBS) where the concentration of major elements is calculated directly from LIBS spectra of the samples without the need to measure reference materials. This approach is of interest if minor and trace elements are not relevant, analysis time is important, and reference materials are not available. The CF-LIBS method has been employed to analyze metals [467][468][469][470], rocks [471][472][473], and biological materials [474,475].  The analysis of slag samples at high temperature is important when short time-toanalysis is required. In order to study the stability of CF-LIBS analysis against sample temperature, ceramic slag samples were heated in a box furnace to high temperature and measured during cool-down in air [374]. The calculated concentration values C CF showed only weak variation with sample temperature up to 275 • C. Larger deviations in concentration were observed at higher sample temperature. The ablation rate, the self-absorption of radiation, the plasma expansion dynamics, and the plasma parameters may depend on the sample temperature [477]. The evaluation of data has led to the conclusion that the CF-LIBS method enables to quantify individual constituents with concentrations ≥1 wt%. This result agrees with the conclusions from theoretical investigations [371,478] that the inhomogeneity of plasma is a major limiting factor for the quantitation of smaller concentrations by CF-LIBS.
In the crude steel production (Linz-Donawitz process) converter slag is a by-product which can be used as raw material in other industrial branches, e.g., for road construction. The chemical composition of the slag is varying, and chemical analysis is required before further use of the material. An automated LIBS measurement system has been developed to analyze the major oxide components of the liquid slag (T = 600−1400 • C) while it is transported in a ladle to slag pits [373]. Figure 19a is a camera view into the slag ladle showing a solidified crust at the slag surface and the laser-induced plasma plume. A measuring probe guides the laser beam onto the slag surface and the plasma radiation from the liquid to the detection unit. The probe is moving across the slag surface during the LIBS measurement (2 min/meas.). The mass fraction of the major oxides was determined by calibration curves for approx. 50 slag ladles and compared to the XRF reference mass fraction (Figure 19b). Similar results were obtained for the liquid slag (R 2 = 0.992, Figure 19b) and solid pressed powder samples (R 2 = 0.997, data not shown). Stable operation during a three-month test run has demonstrated the potential of LIBS for in-line process analysis.

Waste Electrical and Electronic Equipment
The amount of waste electrical and electronic equipment (WEEE) produced per year has increased exponentially in the last 20 years, reaching 50 million metric tons in 2018 [479]. WEEE is mainly composed of iron/steel, plastics, non-ferrous metals, glass, and printed circuit boards (PCBs). For efficient recycling and recovery of valuable materials from this increasing waste, new technologies for fast and accurate chemical identification of WEEE components are needed [412]. Figure 20 shows an application of LIBS for the inspection of end-of-life PCBs from disassembled mobile phones [208]. The optical sensors used for sample recognition within the measuring volume and the LIBS sensor (Figure 20a) were part of a large demonstrator system aiming at automated disassembly, separation, and recovery of valuable materials from WEEE [480]. The photograph of a PCB from a mobile phone (top) and a LIBS raster scan of the PCB with the obtained chemical image of Ta (bottom) are shown in Figure 20b. High intensities of Ta are represented by orange/red color. The LIBS scan of the complete PCB (108 × 42 mm 2 ) was repeated several times to penetrate the housings of the electronic components and to access the bulk material. The chemical information gained by LIBS can be used in the next step for selective laser unsoldering and removal of the identified components (e.g., capacitors) from the PCB [208].
In related studies, LIBS sensors have been developed to control a hydrometallurgical Cu recovery process in discarded PCBs [412] and to analyze heavy metals and brominated flame retardants in polymers and WEEE pieces on a conveyor belt [347].

Waste Polymers
For the recycling of waste polymers and the use of polymer recyclates as secondary raw material, the identification of different types of polymers and the detection of contaminations such as surface layers and heavy metals is important. The polymer polyvinylchloride (PVC, monomer formula C 2 H 3 Cl) is used in different segments such as building and construction, packaging, automotive, electrical and electronic, and textile. PVC is among the top three materials by market share accounting for 10% of the total European demand for plastics ( Figure 4). More than 42% of the collected post-consumer plastic waste in Europe is used for energy recovery. For the energy recovery, the PVC fraction has to be sorted out from the polymer waste to avoid the formation of HCl and other detrimental or toxic substances.
LIBS enables discriminating different types of polymers and detecting contaminations. PVC can be identified by measuring the Chlorine emission line, which is unique for this polymer type. However, this task is challenging because of the low emission rate (Figure 1) and rather high LOD values for Cl ( Figure A4 in Appendix). The in-line measurement of waste polymers for the identification of PVC in an industrial waste materials sorting plant is shown in Figure 21a [343]. Material from municipal waste plastic collection containing different types of plastic pieces and impurities was measured on the conveyor belt. The LIBS spectra (100 measurements/sec) were evaluated in real-time comparing the NIR range with the Cl emission line at around 837.6 nm. LIBS spectra of two different polymer waste samples measured in-line and of a pure PVC reference sample are shown in Figure 21b. PVC pieces were identified by a high correlation of spectra of the waste and reference materials (e.g., sample 1). Waste polymers of low optical reflectivity are difficult to measure by standard NIR reflectance sensors, but they are easy to measure with LIBS as this signal is largely independent of the sample color. Similar measurements were performed on polymer recyclate material to identify impurities such as PVC and surface contaminations in recycled PET flakes [481]. LIBS analysis of heavy metals and halogens in waste polymers has been reported by several groups, e.g., [345,347,482,483].
In the production of plastics, new polymers are often diluted with recycled material. The properties (mechanical, color, chemical) of the diluted polymer should be monitored to keep it within specifications. This is a challenging task as the chemical and mechanical properties are usually tuned via a huge variety of additives such as inorganic coloring pigments, flame retardants, and various thermal and photochemical stabilizers. LIBS can be used for the elemental monitoring of recycled plastics in the production process. A demonstration of LIBS monitoring in polymer production is illustrated in Figure 22 [418]. The produced polymer material is measured at the extrusion orifice of an industrial extruder in a recycling plant (Figure 22a). In addition to the elemental analysis by LIBS, other parameters of the recycled material such as color and strength can also be measured. The monitoring of chemical composition of polymers is shown in Figure 22b. The measured LIBS intensities for Ti and Sb (from additives) are changing with extrusion time as the raw polymeric material (ABS) is increasingly replaced by recycled plastic material (granulate from casings of electronic waste) [418]. After completion of transition from raw to recycled material (at approx. 20 min time) the elemental signals reach a plateau level. The elemental monitoring allows to control the polymer composition and to automatically discard undesirable fractions of the recycled material.

Welds
Welding processes are one of the most commonly used joining technologies. Defects in the weld metal reduce the safety and integrity of a weldment. For the welding of stainless steel, the chemical composition of the weld metal determines the solidification of the steel and the weld metal quality. For the inspection of weld seams, various destructive and non-destructive methods are employed after completion of the welding process. Inspection during the welding process would save time and effort. LIBS can be used for in-situ weld pool monitoring during the welding process. Tungsten inert gas welding of austenitic stainless steel was monitored by LIBS to measure in-situ changes of the chemical composition [402]. Figure 23a shows the schematic of an in-situ LIBS monitor in a welding process. The normalized intensities for elements Cr, Ni, and Mn recorded during welding at various positions on the weld metal are shown in Figure 23b. After solidification of the weld pool the intensity for Mn strongly dropped, whereas intensities for Cr and Ni were almost unchanged. The formation of Mn vapor above the weld pool and condensation of Mn on the weld metal surface was concluded from the measurements. The results proved that LIBS can be used in situ to inspect the TIG welding process.

Conclusions
The major strengths of LIBS from the application point of view are the versatility, the multi-element detection, and the field suitability of the method. LIBS enables for fast measurements without or with only little sample preparation reaching detection limits in the low ppm range, typically. The major limitation of LIBS is its rather low sensitivity ("ppm barrier"), which does not compete with laboratory-based laser analytical techniques such as LA-ICP-MS and LA-ICP-OES. These methods have better analytical performance in terms of LOD and LOQ reaching values in the ppb range, typically. However, the field suitability of LIBS enables to use LIBS-based sensors for in-line and at-line measurements in industrial production and for other applications under harsh conditions out of the laboratory. The ongoing rapid development of laser sources, efficient spectrometers, and sensitive light detection systems is a driving force for the further development of robust LIBS systems and of hand-held LIBS devices. New solid state lasers with high repetition rate and high average power, e.g., advanced fiber lasers and compact Nd:YAG lasers, are supporting this development. Fast element analysis of primary and secondary raw materials, of semi-finished workpieces, and of finished goods is an area with large growth potential. LIBS-based sensors can contribute to the efficient use of resource materials and the accurate chemical monitoring of materials in production. The progress in LIBS measurement technology and the increasing demands for efficient production processes will continue triggering the development of new in-line, at-line, and on-site applications of this laser-analytical method in the near future.  Data Availability Statement: The datasets analyzed or generated during the study are not publicly archived. All datasets are compiled from data available in the cited references.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. Figure A1 shows the annual number of scientific publications on LIBS as retrieved from the SCOPUS database in June 2021 using the search terms "laser induced breakdown spectroscopy" and "laser induced plasma spectroscopy" [484].  China and the United States account for more than 35% of the scientific publications on LIBS or LIPS. Figure A3 shows the table of elements with the LOD values obtained by LIBS measurements of solid sample materials (data taken from www.LIBS-info.com [50]). For each element, the LOD values reported in several publications are averaged. The number above the atomic symbol is the number of publications used. For some elements only one publication was available. The number below the atomic symbol is the average LOD value in ppm.  Figure A4 shows the correlation of reported LOD values [50] and calculated emission rate EMRA of elements (calculated for LTE plasma at temperature T e = 10,000 K, average over the most intense emission lines for each element). A higher EMRA favors lower LOD values for most of the elements (order of magnitude estimate). Figure A4. Correlation of reported LOD values for LIBS analysis of solid materials with calculated emission rate EMRA of LIBS plasma for various elements. The sign "*" stands for multiplication. Figure A5 shows the total number of smartphones sold to end-users in the time period 2007 to 2021 (values for 2020 and 2021 are estimates) [432].