Building of mid-infrared spectral signature of pesticides using functionally-enhanced derivative spectroscopy (FEDS)

Pesticides are synthetic substances widely used in several processes related to agriculture-based food production. Their monitoring is important to evaluate their impact on environment, ensure the food quality, and for the developing of new methods and strategies that help to understand their dynamics from the moment they are released until their end disposal. In this way, the use of mid-infrared spectroscopy (mid-IR spectroscopy) and chemometric methods have been widely explored. However, the strong overlap of spectral signals in complex analytical matrices is recognized as the main limitation of reflectance-based methods for monitoring of this kind of pollutant. In this work, the use of MID-IR spectroscopy by attenuated total reflectance (ATR), in conjunction with functionally-enhanced derivative spectroscopy (FEDS), is proposed as non-destructive technique for fast and simple monitoring of pesticides. Thus, high-resolution dynamic spectral signature (HDSS) for six pesticides commonly used in tropical crops were used as model substances; as analytical matrix was used the peel of banana (e.g., Musa paradisiaca). In addition, concentration effect was evaluated to mimic real conditions. Data recording was obtained by mid-IR spectroscopy (from 4000 to 600 cm⁻¹) and HDSS were evaluated by Pearson's correlation coefficient and spectral similarity FEDS index between the spectra and the reference signals. Based on these correlation parameters, some selection criteria are proposed to estimate the spectral zones where the pesticide signature persists even at relative low concentrations. For the six pesticides, the spectral region considered coincide with those bands of high polarity bonds in the pesticide molecules. Additionally, a good linear prediction models that correlate the relative absorbance intensity and the peak area of the mid signal with the pesticide concentration level are obtained in the spectral ranges of interest. The model performance is defined by Pearson's coefficients r_f greater than 0.97, residual prediction deviation RPDs (root mean square errors - RMSEs) greater (lower) than 5 (3), and limits of detection (LODs) lower than 1 g/kg. Furthermore, this methodology and results are successfully employed to detect Chlorpyrifos residues in banana peel.