Quantitative analysis of heavy metal pollution using FTIR-ATR coupled with chemometric models

Abstract

Heavy metal contamination poses a severe threat to sustainable agriculture and public health, especially in regions where mining and industry are common. This work explores the detection of heavy metals in water using attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy in combination with chemometric modelling. Standard solutions of copper (Cu), mercury (Hg) and lead (Pb) were prepared at specific concentrations of 0.05–1.00 mg/L in order to develop calibration models. The obtained spectra were pre-processed using Savitzky-Golay smoothing, standard normal variate and deconvolution to enhance absorption characteristics and minimise spectral noise. Two regression models, partial least square regression (PLSR) and support vector machines (SVR) were developed to quantify patterns and evaluate their prediction performance. The results showed that for copper PLSR-deconvolution had the best performance (R 2 = 0.811, RMSE = 0.096, RPIQ = 1.085), for mercury SVR-SVN (R 2 = 0.920, RMSE = 0.097, RPIQ = 4.129), and SVR-deconvolution (R 2 = 0.866, RMSE = 0.106, RPIQ = 4.465) for lead. The respective pre-processing of the spectra significantly enhanced the prediction models. Thus, in conjunction with chemometric models and spectra pre-treatment, FTIR-ATR may be able to monitor heavy metal levels quickly and efficiently in environmental samples.