Espectroscopia no infravermelho: predição de elementos potencialmente tóxicos em solos e sedimentos

Abstract

The assessment of potentially toxic elements (PTEs) concentrations, such as heavy metals and rare earth elements, is conventionally achieved by analytic methods which are mostly destructive, expensive and require using hazardous reagents. Thus, there is a need to develop a more practical and safer analytic method. In this context, infrared spectroscopy is presented as a promising analytical method for the prediction of PTEs contents in soil and sediments samples. This technique consists in measuring reflectance patterns originated from a beam of electromagnetic radiation at the range of infrared wavelengths which falls in the sample. The reflectance measurements coupled with the wavelengths of the emission generate a curve known as spectral signature. Given the complexity of the soil matrix, each sample has a unique and distinct spectral signature. This spectral signature is used in calibration models to predict various chemical, physical, and mineralogical soil attributes. Hence, the present work aims to build and test prediction models for potentially toxic elements concentrations by means of infrared spectroscopy in soil samples from the Itataia uranium-phosphate deposit in Santa Quitéria, Ceará State, Brazil, and in soil and bedload sediments samples from the Ipojuca river watershed, Pernambuco State, Brazil. The soil and sediments samples were collected at a 0-20 cm depth, air dried, macerated and sieved through 2 mm, 75 μm and 38 μm sieves. The concentrations of various metals were determined by inductive coupled plasma optical emission spectroscopy (ICP-OES). From the determined values, chemometric calibration models were built for the prediction of element contents using the infrared spectral measurements as entries. Spectral pretreatments were performed in order to remove from the spectra irrelevant information for the prediction and to improve the detection of important spectral features, aiming to strengthen the model. Statistical evaluation procedures were performed for confirming the technique validation and to analyze the pretreatments influence, such as the determination coefficient (R²), root mean square error (RMSE) and ellipse joint confidence region (EJCR). It was possible to generate models with reasonable performance in both study areas, for different elements, such as: La, Nd, Ce and ΣLREE in Ceará State uranium-phosphate deposit; and Al, Fe, Ti, Mo and V in Ipojuca river watershed. The results obtained in this work contribute to the establishment of infrared spectroscopy as an alternative method for the prediction of PTEs concentrations in soils, once a calibration model is developed using soils from the study area.