Heavy metal pollution poses a threat to the ecological environment, economic development, and human health. As a heavy metal pollutant, cadmium (Cd) is difficult to be decomposed by microorganisms, and its excessive intake can lead to cancerous lesions in the kidneys and other organs. Presently, the routine detection methods of Cd element, such as ICP-MS, ICP-OES, and AAS, require the use of chemical reagents to digest the samples, and the detection process is environmentally unfriendly and tedious. Laser-induced breakdown spectroscopy (LIBS) is a typical elemental analysis method with the advantages of a high analysis speed, multi-element measurement, and real-time detection. This has been widely studied in many fields. However, owing to the existence of the matrix effects in the samples, the collected spectral data have high bases and large background noise, and the self-absorption phenomenon is highly severe. Therefore, the stability, accuracy, and sensitivity of the analysis results must be improved. To reduce or eliminate these problems during the LIBS detection process, we constructed a polarization-resolved LIBS (PRLIBS) system to explore methods to improve the analytical ability of LIBS. Stability is promoted, and the baseline drift and background radiation are reduced using PRLIBS. We hope that our study provides a reference for the improvement of the analytical ability of LIBS.

A Cd target was selected as the object in the present study. Combined with the Fresnel equation of light wave propagation in multilayer media, the mechanism of the effect of the incident light wavelength on the light intensity transmittance is analyzed. A light-splitting transmittance model of the plasma radiation intensity in the normal incident direction was established. MATLAB software was used to simulate the model and analyze the changes in the transmitted light intensity. A Cd target was used for the actual measurement verification. The spectral data of the LIBS and PRLIBS with different laser energies and delay time were obtained, and their effects on the spectral signals were analyzed. The relative standard deviation (RSD) of the characteristic spectral line intensity of Cd was calculated to compare the stabilities of LIBS and PRLIBS.

PRLIBS has obvious measurement advantages in terms of the low energy density of the laser, and more characteristic peak signals can be collected by PRLIBS (Fig. 8). The plasma temperature increase with an increase in the laser energy, and the overall spectral characteristic signal tends to increase (Fig. 9). The phenomenon of the baseline drift in the LIBS spectrum is evident, whereas the PRLBS spectrum has almost no baseline drift problem, and the background signal is significantly reduced. This phenomenon shows that the continuous radiation and background signal are filtered out after the plasma spectral signal passes through the polarizing system, while the discrete signal is retained. The polarization characteristics of the continuous spectrum are much stronger than those of the discrete spectrum. Additionally, the RSD values of the characteristic spectral line intensity were calculated (Table 1), which in PRLIBS are less than those obtained by the LIBS method at the same detection parameters. This indicates that the light-splitting transmittance model effectively improves the stability of the plasma spectrum. With the increase in pulse energy, the light-splitting transmittance model can effectively reduce the baseline drift and background radiation and enhance the spectral resolution. This model not only retains effective information in the continuous spectrum but also improves the stability of spectral line identification. Moreover, the characteristic Cd peak signals of LIBS and PRLIBS with seven different delay time were collected to analyze their effect on the spectral signal (Fig. 10). The variation trends of the characteristic peak intensity with the delay time under the two methods are consistent.

Based on the existing LIBS detection model, the theoretical formula of the transmission ratio for different wavelengths of light directly incident onto the Glan-Thompson prism is deduced, and the PRLIBS detection model of the plasma radiation intensity is established. By collecting and comparing the LIBS and PRLIBS spectra of Cd element under different laser energy densities, it was deduced that PRLIBS can measure more characteristic signals, whereas the spectral transmittance model can reduce the baseline drift and background radiation under low-energy conditions. Moreover, the difference between the maximum and minimum values of the spectral characteristic signal intensity of PRLIBS is smaller than that of LIBS, reflecting the advantages of the PRLIBS method for characteristic spectral line recognition. The RSD values of the LIBS and PRLIBS characteristic spectral lines at laser pulse energies of 23.68 mJ and 42.52 mJ were compared. The average RSD value of the PRLIBS method was less than 10%, and the overall RSD value was smaller than that of the LIBS method, indicating that PRLIBS can effectively improve the stability of the identification of the characteristic spectral lines. The experimental results show that the model can improve the ability to identify the characteristic spectral lines of Cd plasma spectra. Additionally, the characteristic Cd peak signals of LIBS and PRLIBS with seven different delay time were collected. The variation trends of the characteristic peak intensity with the delay time under the two methods were consistent, indicating that PRLIBS does not change the delay time of LIBS.