Multivariate Approach to Detection of Uranium Using LIBS

Investor logo

Warning

This publication doesn't include Faculty of Economics and Administration. It includes Central European Institute of Technology. Official publication website can be found on muni.cz.
Authors

KLUS Jakub MIKYSEK Petr PROCHAZKA David POŘÍZKA Pavel NOVOTNÝ Jan HRDLIČKA Aleš NOVOTNÝ Karel KAISER Jozef

Year of publication 2016
Type Conference abstract
MU Faculty or unit

Central European Institute of Technology

Citation
Description This work aims to study the possibilities of LIBS concerning the detection of uranium precipitated in sandstones. The main challenge in uranium detection is its complex emission spectrum with very high density of spectral lines. As is it stated by Chinni et al. the number of uranium lines measured using a hollow cathode discharge lamp exceeds 5000 in the range 384.8 to 908.4 nm. Considering the spectral line broadening such line density significantly exceeds resolution of usually utilized spectrometers. In this paper two new analytical approaches are suggested to determine the relative content of uranium in the studied specimen. Prior to the LIBS technique the X-ray fluorescence analysis (XRF) was utilized for uranium presence verification. To employ this task an orthogonal double-pulse (DP) LIBS setup was utilized in order to reach high sensitivity, maintaining low crater diameter, subsequently enabling high spatial resolution chemical mapping. Sample of sandstone was mapped with spatial resolution 0.1 mm and crater diameter 50 um. The mapped area was 15x15 mm, which resulted in a chemical map of 150x150 pixels. Resulting dataset contains over 586 million data points; hence the utilization of multivariate method, principal component analysis (PCA), for mapping of uranium in selected ores is proposed. PCA is one of the fundamental projection methods in multivariate data analysis (MVDA). The input for PCA is matrix of sizes N×K, where N stands for number of objects (in our case laser pulses/ measurements) and K stands for variables (spectra recorded on detector). Statistically, PCA finds projections called principal components (PC) that approximate the data in the least squares sense. There are as many PCs as is the rank of input matrix. To each PC, a loadings vector is assigned, which defines the linear combination of variables in the original K-space that PC represents. For the reason that PCA computation is computationally extensive, representative set of 50x50 measurement points was chosen to compute the PCs. The resulting chemical map was compared to three other approaches, first of them being the PCA analysis of the whole dataset resulting in 0.99997 correlations. The second one was rather classical approach of fitting the U II line detected at 409.02 nm with pseudo-Voigt profile and the integral intensity under the fitted line taken as analytical signal. And the third one was based on assumption of seeming background increase in the short region of spectra (590 – 595 nm). The last two correlated with factor 0.947 and 0.955, respectively.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.