Comparison of Algorithms for Baseline Correction of LIBS Spectra for Quantifying Total Carbon in Brazilian Soils

Abstract

LIBS (Laser-Induced Breakdown Spectroscopy) is a versatile technique for multi-element analysis, offering rapid analysis with minimal sample preparation. However, challenges persist in baseline correction due to background radiation, causing non-linear interference with emission lines. This necessitates proper baseline correction for accurate quantification of elements in LIBS spectra. Our study focuses on the quantification of total carbon in soil samples, conducting a comparative analysis of filters and methods for noise removal and baseline correction in LIBS spectra. We exhaustively tested all combinations of filters and methods, optimizing their parameters to establish the strongest correlation between corrected spectra and carbon content in a training sample set. We then evaluated these combinations using optimized parameters on a separate test sample set. Through rigorous evaluation, we found that the combination of the Savitzky-Golay filter and the 4S Peak Filling method achieved the most effective baseline correction. This combination demonstrated a Pearson's correlation coefficient of 0.93 and a root mean square error of 0.21. Notably, it outperformed a linear regression model based solely on the carbon emission line at 193.04 nm, which yielded a correlation of 0.91 with a root mean square error of 0.26. Our proposed procedure presents a novel approach for baseline correction in LIBS spectra, opening new possibilities for the development of multivariate methods utilizing specific spectral ranges.