2015 Volume 55 Issue 11 Pages 2391-2396
Laser-induced breakdown spectroscopy (LIBS) is a promising method for rapid determination of compositions of stainless steels in steel scrap. LIBS is widely known as a method for elemental analysis that enables a rapid determination. It has several advantages such that it can work under ambient pressure, and specimens can be tested without any pre-treatment such as acid digestion, cleaning, or polishing of surface of specimens. We applied a laboratory-build LIBS system for sorting of six types of stainless steels. The standard reference materials of JISF FXS 324–334, 335–343, and 344–349, which are respectively Fe–Ni, Fe–Cr, and Fe–Mo binary alloys, were employed for making calibration lines of them. Considering spectral interferences from emission lines of the iron matrix in these alloys, seven emission lines could be chosen. Longer gate width, shorter delay time, high stability of pulse laser energy, and more number of laser shots can decrease the fluctuation of emission intensity. Utilizing these parameters mentioned above, the sorting of stainless specimens by detecting chromium, nickel, and molybdenum could be achieved.
In recent years, recycling of steel scrap has been energetically encouraged in order to comply with the “Basic Act for the Promotion of the Recycling-Oriented Society”1) which has been enforced since 2000 in Japan. The amount of steel production in Japan has been more than 100 million tons, and produced steels have been stocked and accumulated as infrastructures, cars, constructions, etc. Around one-third of steel production in amount has been conducted by electric arc furnace (EAF), which is the major recycle route for steel scrap.2)
Steel products are recently known as the “carrier” of non-ferrous elements.3) Especially, seven elements of cobalt, chromium, nickel, manganese, molybdenum, tungsten, and vanadium has been designated as stockpiling resources by Japan Oil, Gas and Metals National Corporation (JOGMEC). The reports of material flow of mineral resources by JOGMEC have revealed that most of the seven elements listed above have been inputted as the alloying element for steel production. Above all, chromium, nickel, and molybdenum have been in production of stainless steels. Recently, the production of stainless steels in Japan has been around three million tons/year,4) which has been widely utilized as vehicles, electronic appliances, machines, etc.
The steel scrap of end-of-life vehicles, i.e., “ELV scrap”, has been one of the major targets of recycling of stainless steels. Stainless steels have been applied 20–30 kg per each vehicle, especially in an exhaust system in order to obtain higher performance of heat and corrosion resistance for better fuel consumption.5) Although the scrap of stainless steels has been classified as in-house, industrial, and obsolete, the scrap of stainless steels in ELV scrap has been unfortunately treated as obsolete scrap, which resulted just as the steel resources, not as the chromium or nickel resources.
For better recycling of stainless steel scrap, we can apply laser-induced breakdown spectroscopy (LIBS) as an encouraging and promising analytical tool6,7,8,9) to sort them according to chemical compositions. It can determine chemical compositions by remote sensing without any pre-treatment of targeted steel scrap by the combination of laser-induced plasma (LIP) generation and the detection of atomic/ionic emissions from the plasma. In the determination of chemical compositions by LIBS, the precision and repeatability is very important for reliable sorting of steel scrap. In our previous work, it has been revealed that the calibration lines of chromium and nickel obtained by LIBS using low-pressure argon gas are enough to be employed for elemental analysis of chromium and nickel in Fe-based alloys.10,11) In these studies, they made a criteria that the relative standard deviations (RSDs) of below 10% is acceptable, and linear calibration curves of chromium and nickel in Fe–Cr or Fe–Ni binary standard reference materials (hereinafter, SRMs) under a reduced argon atmosphere could be obtained from 5–20 Cr at% or 0–10 Ni at%.
The background emission from LIP generally worsens the precision of calibration lines due to the recombination emission and the bremsstrahlung.12) Therefore, we investigate precise calibration lines of chromium, nickel, and molybdenum, and furthermore, condition of mutual identification of stainless steels was investigated.
The LIBS system utilized in this study comprised an Nd: YAG pulse laser (LS-2137, LOTIS TII), a three-dimensional sample stage, planoconvex lenses for collection of the laser and the emitted radiation from the generated LIP, an optical fiber, an Echelle-type spectrometer (ME5000, Andor) and an ICCD detector (DH734-18F, Andor), as shown in our previous researches.10,11,12) Among the various types of spectrometer, an echelle-type one shows relatively high wavelength resolution because it utilizes high-order diffraction ray coming from echelle grating. The wavelength resolution Δλ can be described as Δλ/λ = kN, where k is a diffraction order, λ is an averaged wavelength, and N is the total number of diffraction grating. Although echelle-type spectrometer is a polychromenter and can achieve wide wavelength measureable and high wavelength resolution at the same time, it has relatively low sensitivity due to the utilization of high-order diffraction ray. These characteristics of echelle-type spectrometer described above is suitable for mutual elemental-based sorting of stainless steels focused in this study because it can detect multiple emission peaks at the same time with relatively high wavelength resolution. Though the sensitivity is relatively low, the sensitivity of this LIBS system is enough for detecting nickel, chromium, and molybdenum in stainless steels. The wavelength, duration, and frequency of the pulse laser are 532 nm (SHG mode), 16–18 ns, and 10 Hz, respectively. The signals from the Echelle-type spectrometer were integrated on the ICCD detector; we set the exposure time at the maximum of 1.0 s, which enables to integrate the emission signals from LIP. In order to optimize the emission signals from chromium, nickel, and molybdenum for achieving minimum fluctuation, we set delay times and gate widths of the ICCD for 1–20 μs, 1–100 μs, respectively. The amplification of a photomultiplier was also adjusted by the software interface in order to prevent saturation by the emission signals arriving at the micro-channel-plate-type photomultiplier in the ICCD. In this condition, measurement time was around 1.7 second for 10 laser shots, which was estimated in the interface. The number of laser shots and measurement time varied from 1 to 100, and around 0.17 s to 17 s, respectively. All the measurement was conducted for 10 replicates in each specimen to evaluate degree of fluctuation by calculating RSDs.
For obtaining calibration lines, that is the relationship between emission intensities and the contents of chromium, nickel, and molybdenum in the specimen, we adopted SRMs of FXS 324–334, 335–343, and 344–349 certified by The Japan Iron and Steel Federation. They were the standard reference materials of binary alloys of Fe–Ni, Fe–Cr, and Fe–Mo, respectively, and their compositions were listed in Table 1.
Six types of stainless steels, SUS301CSP, SUS304, SUS316, SUS420J2, SUS430, and SUS440C were employed for confirming the availability of LIBS measurement for mutual identification. The chemical compositions of them were listed in Table 2, focusing the components of chromium, nickel, and molybdenum. The contents of chromium, nickel, and molybdenum were determined by X-ray fluorescent analysis using ZSX Primus II (Rigaku) by the fundamental parameter method to compare the measured concentrations of chromium, nickel, and molybdenum.
Since a number of atomic/ionic emission lines from chromium, nickel, molybdenum, and iron from SRMs or stainless steels can be obtained by the combination of Echelle-type spectrometer and ICCD detector, the emission lines for calibration lines must be carefully picked up, having enough intensities and less spectral interferences. Taking them into consideration, Fig. 1 shows LIP emission spectra from the FXS 327 (Fe–Ni), 338 (Fe–Cr), 349 (Fe–Mo) stacked with those from the 4N-purity iron plate. In general, a choice of non-resonance line is favorable for making calibration lines because of avoiding self-absorption effect;13,14) an emission relating to resonance transition will be predominantly absorbed by the same kind of atoms in a ground state, resulting in decrease in the observed emission intensity. Therefore, to a maximum extent, we selected both resonance and non-resonance emission lines for evaluating a degree of this effect. In order to cancel out the fluctuation of the observed intensities, we additionally selected the Fe I emission lines as the internal standard, which are given in Table 3. Shimada et al. suggested that both the emission lines of the targeted elements had better similar upper level with internal standard lines to coordinate the excitation/de-excitation behavior.10) When the LIP is in the local-thermal equilibrium condition, the ratio R between emission lines from the elements A and B are expressed as follows:15)
Emission peak of nickel from FXS327, chromium from FXS338, and molybdenum from FXS349.
The generation of LIP and the decay of emission from it in an open-air is usually unstable in each laser shot; we should evaluate the degree of fluctuation along with the experimental parameter, which are gate width, delay time, pulse energy, and the number of laser shots. Figure 2 represents the relationship between gate widths and RSDs, which were adopted as the benchmark of fluctuation. It is clearly found that the RSDs decreased with the elevation of gate widths when the delay time, pulse energy, and the number of laser shots are fixed to be 1 μs, 130 mJ/pulse, and 100. As described in our previous study, longer gate widths for amplification of emission signals by ICCD can detect the longer part of transient emissions from the generated LIP and averaged them; therefore, we set the value of gate width to 100 μs as an optimum value.
Relationship between the RSDs from the peak area ratios and the gate widths.
As the explanation of the transient generation process of LIP, the shockwave model is widely known.7,16) At the very initial stage of LIP generation (~50 ns), atoms of the surface or in air ablated and spread at the supersonic velocity, which results the generation of shockwave. The light emission related to this spread is called “initial plasma”, mainly composed of excited/de-excited air atoms and less contains emissions from the targeted element in the surface of specimen. Inside the shockwave, the temperature of atmosphere drastically increase because of adiabatic compressive heating. In this atmosphere, targeted atoms are excited by the collision by fast electrons/atoms (collision of a first kind) or exchanging internal energies (collision of a second kind); therefore, it is important to set the timing of starting for obtaining transient emission from LIP. Figure 3 shows the RSDs and signal-to-background ratio towards the increase in delay time. There seem two types of behavior; one is that the RSD values simply increased with an increase in delay time (Type A), and the other is that they once decreased, and then increased with the increase in delay time (Type B). Type-B is found in the line pairs of Ni I 352.454 nm/Fe I 349.057 nm and Ni I 440.154 nm/Fe I 441.512 nm, and the rest of line pairs show Type-A. Changes in signal-to-background ratio are also listed in Fig. 3, showing maximum values at delay time around 2–5 μs. The values of signal-to-background are around 0.5–4 μs, and found unchanged during the delay time of 1–20 μs. The magnitude of signal-to-background depends on the original emission intensity of pairs of each peak. RSDs seems worsen with an increase in delay time, because of similar reason with described above that shorter delay time and longer gate widths can take most part of transient emission in each LIP generated from each laser shot; therefore, 1 μs of delay time was set as the optimum one.
Relationship between the RSDs from the peak area ratios and the delay times.
Figure 4 expresses the degree of RSDs vs. pulse energy in each laser shot. It is obvious that the RSDs became better with high power density of pulse laser energy. The stability of the laser energy was related to the laser energy, and thus higher stability of the laser pulse energy made the RSD values low.
Relationship between the RSDs from the peak area ratios and the pulse energies.
In LIBS measurement, the number of laser shot is very important parameters because it affects the measurement time. The relationship between RSDs and the number of pulsed laser shots are in Fig. 4, expressing better RSDs could be obtained from more laser shots, and in this study, we set the number of laser shots to 100. The experiments described above were conducted to optimize the experimental parameter, as is, gate width, delay time, pulse laser-energy, and the number of laser shot, and they are optimized to 100 μs, 1 μs, 115 mJ, and 100.
According to the optimized experimental conditions, the calibration line of nickel, chromium, and molybdenum were shown in Fig. 5. The range of concentrations of targeted elements are 0.52–15.20 wt% for Ni, 0.50–20.07 wt% for Cr, and 0.20–7.07 wt% for Mo, which configure enough concentration range for detecting nickel, chromium, and molybdenum component shown in Table 1. The presence of outliers of intensity ratio within 10 duplicates of measurements was judged by smirnov-grubbs test, with two-sided 98% confidence interval.
Relationship between the RSDs from the peak area ratios and the number of laser shots.
In Fig. 6, we can find typical characteristics of calibration lines in those of nickel. Self-absorption effect is clearly observed in the resonance line pairs of Ni I 341.476 nm/Fe I 349.057 nm, Ni I 352.454 nm/Fe I 349.057 nm, which show the bending in Fig. 5. On the other hand, the non-resonance line pair of Ni I 440.154 nm/Fe I 441.512 nm is a straight line. These tendencies are also found at the calibration lines of chromium. The line pair of Cr I 425.435 nm/Fe I 426.047 nm and Cr I 427.480 nm/Fe I 426.047 nm also show the behavior of self-absorption effect. On the other hand, the line pair of Cr I 396.369 nm/Fe I 395.667 nm does not show the self-absorption effect because of relatively high upper energy levels. The line pair of Mo I 386.411/Fe I 385.637 nm also shows the self-absorption effect, however, we can only select this line pair because there are no alternatives of Mo lines for its detection.
Calibration curves of chromium, nickel, and molybdenum by FXS SRMs, when the delay time gate width, pulse energy, and the number of laser shots were fixed at 1 μs 500 μs, 115 mJ, and 100, respectively.
Again, in general, the pair of non-resonance line tend to configure linear calibration line, and favorable for utilization of quantitative determination. In conformity to this principle, the pair of Cr I 396.369 nm/Fe I 395.667 nm was picked up for the determination of chromium. On the other hand, it does not seem suitable for choosing the pair of Ni I 440.154 nm/Fe I 441.512 nm because intensity ratios of this line are too low; therefore, we selected the line pair of Ni I 347.476 nm/Fe I 349.057 nm as an alternative. For molybdenum, there are no choice but Mo I 386.411 nm/Fe I 385.637 nm. By utilizing these calibration lines, the results of quantitative determination of nickel, chromium, and molybdenum in the samples are listed in Table 4. In the tables, the first column represents the atomic ratio between targeted elements and iron, calculating by the chemical compositions in Table 1. The atomic ratios resulting from LIBS are also listed in the second column, and the comparison of them is in the third column. By comparing the ratio between XRF and LIBS, the accuracy of LIBS ranked with those by XRF with a fundamental parameter method.
In this study, focused on the detection of nickel, around 0.1 wt% of nickel couldn’t be detected by LIBS measurement. However, a relatively lower sensitivity mentioned above is not an envelope of detection limit by LIBS,7) which has been already achieved ppm-level detection. When the LIBS measurement is conducted for stainless steels by Echelle-type spectrometer and ICCD camera for the purpose of wide-range wavelength detection, CCD sensor on the camera receives all atomic/ionic emissions from the target sample through an Echelle-type spectrometer. Therefore, the sensitivity of microchannel plate must be suppressed in order not to make a ghosting by a strong atomic/ionic emission from iron, which composes a parent phase of stainless steels. Though the set-ups in this study could not exhibit maximum performance in the sensitive detection of minor elements in the samples, they are enough to sort stainless specimens according to their composition.
In order to achieve a rapid separation of stainless steels according to their chemical compositions, a measurement by laser-induced breakdown spectroscopy (LIBS) was conducted with optimizing measurement parameters, gate width, delay time, laser energy, and the number of laser shots. Avoiding the spectral interferences with those of iron, three nickel emission lines, three chromium lines, and one molybdenum line are found to be candidates for quantitative determination of them. The parameters of gate width, delay time, laser energy, and the number of laser shots were optimized to be 100 μs, 1 μs, 115 mJ, and 100. Though the range of determined concentrations was around several thousand ppms, it was enough to separate six types of stainless steels according to the chemical compositions of nickel, chromium, and molybdenum in stainless steels.
This research is supported by a Grand-in-Aid from the Ministry of Education, Science, Sports and Culture of Japan (No. 23310049). Several parts of the set-ups were installed under a support of the A-STEP program in Japan Science and Technology Agency (No.241FT0286) and ISIJ Research Promotion Grant (Incl. Ishihara/Asada Grant) by the Iron and Steel Institute of Japan.