Deep Learning Models for Data-Driven Laser Induced Breakdown Spectroscopy (LIBS) Analysis of Interstitial Oxygen Impurities in Czochralski-Si Crystals

Abstract

Analytical advantages of facile and expeditious spectral data collections from laser-induced breakdown spectroscopy (LIBS) are often offset by the low-accuracy quantitative analyses offered by the technique due to non-equilibrium plasma-matrix interactions. Herein, we develop 1D convolutional neural network (CNN) and least absolute shrinkage and selection operator (LASSO) models for LIBS data analyses to predict trace amounts of interstitial oxygen (Oi) impurities in commercial Czochralski-silicon (Cz-Si) crystals with known Oi concentrations (0 - 16 ppm). While traditional spectral analyses from O I (777.2 nm) atomic lines offer poor accuracy, CNN and LASSO analyses generate excellent predictions for the Oi concentrations. Specifically, CNN-based spectral analyses uniquely identified systematic alterations in LIBS fingerprints manifested by laser-matter interactions. Our results pave the path for combining facile and voluminous LIBS data collection with deep learning driven high-fidelity data analytics.

Atomic spectrometry update: review of advances in the analysis of metals, chemicals and materials

“…Applications for nuclear samples and monitoring neutron irradiation exposure, as well as geological samples were envisaged. The 240 Pu/ 239 Pu ratio was measured by Quemet using TIMS in dissolution residues of a spent fuel solution. 184 Repeatability of better than 83 ppm was achieved in each aliquot, with interpretation of the kinetic evolution that could not be reached by conventional direct concentration measurement.…”

“…Accelerator mass spectrometry (AMS) was applied by Johansen et al to Pu isotopic ratio measurements to distinguish between fallout from the Fukushima reactor incident and the nuclear weapon detonated at Nagasaki through measurement of soils and biota. 187 The 241 Pu/ 239 Pu ratios were preferred to 240 Pu/ 239 Pu as the latter could not distinguish between most 2016 environmental samples and Fukushima contamination. The contribution of new Pu from Fukushima was assessed in samples including soil, earthworms and wild boar.…”

“…Aliquots of each were then bottled ready for use. The 235 U and 239 Pu values were assigned from gravimetric preparation and veried by IDMS, with a wider inter-calibration and characterisation campaign including prociency testing exercises.…”

“…Davari and Mukherjee have added to this list by using a one dimensional convolutional neural network and a least absolute shrinkage and selection operator (LASSO) models for LIBS data analysis . 239 The paper gives a decent description of how the models work and then applied them to LIBS data obtained while determining O impurities in the interstitial layers of Czochralski crystals. The analysis chamber was filled with 6 L min −1 He and 2 L min −1 Ar (these optimal conditions had been established in a previous study) and the O was measured at 777.19 nm.…”

Near infrared spectroscopy discriminates glutinous and non-glutinous sorghum using an approach based on typical samples and direct calibration

“…Direct discrimination of glutinous and non-glutinous sorghum can be more efficient than quantitative methods. The commonly used methods for NIR spectra discrimination, such as discriminant partial least squares (DPLS), 22 soft independent modeling of class analogy (SIMCA), 23 artificial neural networks (ANN), 24 support vector machines (SVM), 25 and convolutional neural networks (CNN), 26 require many samples for modeling. Although some studies have attempted to explain the rationale of the NIR spectra pattern recognition model, the model has yet to be fully interpreted.…”