Spectral screening-assisted LIBS for quantitative analysis of heavy metal elements in liquid aerosols

Abstract

Heavy metals like Cu and Zn in liquid aerosols can accumulate in the environment, posing risks to ecosystems and human health. To enhance the monitoring and prevention of heavy metal pollution, a new method was developed for quantitative analysis of heavy metals in liquid aerosols by combining LIBS with effective spectral selection. A custom chamber was designed. The hit efficiency and reproducibility of the spectra were compared with those obtained without the custom chamber. Then, the spectral data were screened using the Light Gradient Boosting Machine (LGBM) and compared with the Standard Deviation (SD) method. LGBM performed better than SD in all evaluation metrics. Finally, univariate and multivariate analysis methods were used to quantify the Cu and Zn elements in aerosols. In univariate analysis, the R_P² values of the calibration curves for Cu and Zn were 0.8390 and 0.6608, respectively. In multivariate analysis, a partial least squares regression (PLSR) model was established. Then the Recursive Feature Elimination (RFE) algorithm was combined with PLSR to optimize the features and form the RFE–PLSR model. The R_P², RMSEP, MAE and MRE of the RFE–PLSR model for Cu and Zn were 0.9876 and 0.9820, 178.8264 and 215.1126, 99.9872 and 199.9349, and 0.0499 and 0.1926, respectively. The results of this study show that LIBS technology combined with LGBM algorithm screening in the custom chamber can realize the rapid detection and quantitative analysis of heavy metal elements in liquid aerosols.