Machine learning-assisted LIBS for quantitative determination of heavy metals in surfactant-enhanced adsorption on PMMA nanoplastics

Abstract

Surfactant molecules can alter the hydrophobicity of nanoplastics (NPs), thereby influencing the adsorption of heavy metals onto them. In this study, a systematic comparison was conducted on the enhancement effects of three surfactants, namely cetyltrimethylammonium bromide (CTAB), sodium dodecylbenzene sulfonate (SDBS), and sodium dodecyl sulfate (SDS), on the adsorption of Cr, Pb, and Cu by poly(methyl methacrylate) (PMMA). Using laser-induced breakdown spectroscopy (LIBS), surfactant type, concentration and adsorption time were optimized, and the impact of coexisting ions was examined. Following the identification of the optimal surfactant and experimental conditions, quantitative analysis of the heavy metal adsorption capacity of PMMA was conducted using machine learning-assisted LIBS. LIBS spectra of 21 contaminated PMMA samples were collected, and wavelet transform (WT) and Savitzky-Golay (SG) were applied to remove noise and matrix effects from the spectra. Five variable selection methods were further employed to eliminate redundant variables and enhance the predictive performance of the model. SDS exhibited the most prominent enhancement among the three surfactants. The signal-to-noise ratios (SNRs) of Cr, Pb, and Cu were enhanced by 4.41, 10.58, and 3.26 times, respectively. Based on optimized input variables and model parameters, the SG-competitive adaptive reweighted sampling (CARS)-PLS calibration model demonstrated superior performance in quantitative analysis of three metal elements in PMMA: Cr (R²_p = 0.9153, MRE_p = 0.1201), Pb (R²_p = 0.9419, MRE_p = 0.0756), and Cu (R²_p = 0.9317, MRE_p = 0.2496). This study improves understanding of heavy metal adsorption by NPs in the presence of surfactants, facilitating effective control of NPs-induced risks.