Processing of large-scale laser spectroscopy data

abstract

English

State-of-the-art laser-ablation based spectroscopic instruments provide data sets with increasing size, number of objects (spectra) and variables (wavelengths). In this work we concentrate namely on the Optical Emission Spectroscopic techniques, such as Laser-Induced Breakdown Spectroscopy and Laser Ablation Inductively Plasma Optical Emission Spectroscopy. However, presented algorithms are applicable also to the Mass Spec techniques. Typically, the data set is overloaded with information, analytically relevant as well as redundant, e.g. originating from noise and background. Processing such multivariate data is, thus, a challenging task demanding sophisticated approaches. A use of advanced statistical algorithms, referred to as multivariate data analysis algorithms or chemometrics, is of great interest in the contemporary data processing. Most often Principal Component Analysis and Self-Organizing Maps are implemented. Our efforts tackled mainly the dimensionality reduction in both, objects and variables. Their implementation and necessary apriori data processing is dissected. The novelty of our work leads to an algorithm providing an increased turn-around time of the multivariate data processing and to a reduction of demands on the computing power. We demonstrated our endeavors on the processing of multi-elemental maps on selected samples. In this work, the sample preparation and methodology in analysis is at the edge of our interest, exploited only to provide the data sets for further analysis.

LIBS, data, processing

POŘÍZKA, P.; KLUS, J.; PROCHAZKA, D.; VRÁBEL, J.; KÉPEŠ, E.; MODLITBOVÁ, P.; KAISER, J.

21. 10. 2018