Chemometrics -Assisted Laser Induced Breakdown Spectroscopy of High Background Radiation Area (HBRA) Geothermal Field Matrices

In this study, the utility of LIBS with chemometrics techniques namely PCA, PLS, ANNs and SIMCA has been investigated and demonstrated in performing trace quantitative and explorative analysis of High Background Radiation Areas (HBRA) geothermal field matrices (rocks, soils), for the purpose of analyzing atomic and molecular signatures, so as to characterize and evaluate the impact of HBRA geothermal discharges on the surrounding environment. Analytical performance tests based on (multi- signal) standard addition method were done for the elements in the concentration range of 10-150 ppm for the trace elements and 0.1-1.5% for Ti. The classical calibration yielded to predicted concentrations not close (> 10 ppm) to the true/measured concentrations hence the use of chemometrics techniques (PLS and ANNs) for more accurate prediction of the elements' concentration in soils and rocks respectively. PCA and SIMCA were applied on the samples' LIBS spectral signatures. Linear calibration curves from classical univariate approach with R²> 0.84 were obtained for the lines from which the limit of detection were calculated and found to be: 2.4 ppm, 5.1 ppm, 3.1 ppm, 7.6 ppm, 0.012 % for As, Cr, Cu, Pb and Ti in soils and 8.3 ppm, 6.1 ppm, 9.0 ppm, 3.0 ppm, 0.018 % for As, Cr, Cu, Pb and Ti in rocks respectively. The concentrations of Cr, Cu and Pb in soils were within the range recommended by Environmental Protection Agency. Pearson correlation coefficients showed that HBRAs (geothermal) were uniquely characterized by positive correlation of As and Cr concentrations while NBRAs had negative correlation of Cu with Pb and Ti. PCA and SIMCA classified samples in the categories of sampling sources i.e. HBRA (geothermal), HBRA (non-geothermal) and NBRA (geothermal) based on full spectrum signatures in the 200 - 980 nm range, hence demonstrating the capability of PCA and SIMCA in identifying similarities and differences between unknown samples based on their LIBS spectra. The results obtained indicate that although univariate multi signal calibration is applicable to a limited degree in LIBS quantitative analysis, chemometrics performs better quantitative calibration and subsequent modeling of spectra in relation to analyte concentrations. (author)