Abstract: Objective　Laser-induced breakdown spectroscopy (LIBS) offers many advantages, such as simultaneous multi-element measurement, rapid analysis, real-time online detection, and no need for sample pretreatment. Laser-induced plasma is generated by the interaction of pulsed laser with matter, and its spectrum includes continuous background radiation and characteristic discrete spectral lines of elements. Strong continuous background radiation can severely interfere with the measurement and analysis of discrete spectral lines of elements. Researchers try to eliminate the influence of continuous background on characteristic spectral line measurement by setting CCD acquisition delay. However, the setting of CCD acquisition delay may vary with experimental conditions and samples. For detectors that cannot set delay and gate width, it is more meaningful to conside using some algorithms to suppress the influence of continuous background on characteristic spectral lines. On the other hand, researchers have found that the polarization degree of continuous background spectra in LIBS is greater than that of characteristic spectral lines. Considering the different polarization characteristics of characteristic spectral lines and continuous background, polarization resolved LIBS technology has therefore been used to suppress continuous background. In order to improve the quantitative analysis performance of LIBS for soil samples, the polarization resolved LIBS combined with continuum background subtraction algorithm (CBSA) is used in this work to suppress continuous background radiation. 318 Methods　We focus on the LIBS data of the soil samples, and propose a method combining polarization resolved LIBS with CBSA to suppress continous background radiation. An extreme value algorithm is used to process spectral signals, the original LIBS spectral points are selected by searching for local maxima in the spectrum and labeled as SO, while the continuous background spectral points are identified by searching for local minima and labeled as SB. Using cubic spline interpolation, the original LIBS spectrum and the continuous background spectrum are plotted with their respective nodes, and the envelope lines of the original LIBS spectrum and the continuous background spectrum are obtained. The spectrum SC is obtained by subtracing the continuous background spectrum SB from the original LIBS spectrum SO at the corresponding position. Results and Discussion　We conduct a comparative analysis of the LIBS original spectrum and the polarization resolved LIBS using CBSA, and evaluated the signal-to-background ratio (SBR) and the relative standard deviation (RSD) of the characteristic spectral lines, and the root mean square error of calibration (RMSEC) of the quantitative model for heavy metal element Cr. Compared with the original spectrum of LIBS, the adoption of CBSA increases the SBRs of the four characteristic spectral lines Fe I:404.581 nm, Pb I:405.78 nm, Ca I: 422.67 nm and Mg I:518.361 nm by 4.0, 8.6, 14.4 and 5.5 times respectively, with an average increase of 8.1 times for SBR. As for polarization resolved LIBS, the adoption of CBSA increases the SBRs of the four characteristic spectral lines by 12.5, 18.6, 22.5 and 8.6 times respectively, with an average increase of 15.6 times. By using CBSA, the RSDs of the four characteristic spectral lines are decreased by 0.8%, 0.3%, 2.1% and 0.6% respectively, with an average increase of 1.0%. And for polarization resolved LIBS, by using CBSA, the RSDs of the four characteristic spectral lines are decreased by 3.5%, 2.8%, 3.7% and 1.4% respectively, with an average increase of 2.9%. The RMSEC of Cr is reduced from 0.088 to 0.053 for polarization resolved LIBS combined with CBSA, representing a 40% reduction. Conclusions　The polarization resolved LIBS combined with CBSA significantly suppresses continuous background radiation and effectively enhances the quantitative detection capability of LIBS for soil samples. Compared with the original LIBS spectrum, the polarization resolved LIBS combined with CBSA results in an average SBR improvement of 15.6 times for the characteristic spectral lines Fe I: 404.581 nm, Pb I: 405.78 nm, Ca I: 422.67 nm, and Mg I: 518.361 nm, an average RSD decreases of 2.9%, and the RMSEC of the heavy metal element Cr decreasing by 40%. It is shown that the polarization resolved LIBS combined with CBSA can provide a high-precision analytical approach for quantitative analysis of soil using LIBS.

Key words: spectroscopy, laser-induced breakdown spectroscopy, polarization, continuum background subtraction, soil