Aplicación de técnicas quimiométricas a la resolución de señales electroquímicas solapadas

Abstract

In this Doctoral Thesis, the problem of selectivity in electrochemical signals has been tackled by using chemometric techniques. The aim of the work was the resolution of hardly overlapped electrochemical signals making use of novel statistical techniques for the design and interpretation of experiments. In this way, several mathematic algorithms were developed in collaboration with an Italian research group from Universita degli Studi di Modena e Reggio Emilia (Modena-ITALY). Amongst the statistical techniques applied, Fast Wavelet Transform (FWT) and Artificial Neural Networks (ANN) can be underlined, as well as several multivariate calibration methods: Multilinear Regression (MLR), Partial-Least Squares Regression (PLS) and Principal Component Regression (PCR).The chemometric methodology developed has been applied to certain electrochemical systems of different complexity: the first one, inorganic, constituted by thallium and lead ions, and the other one, organic, constituted by four chlorophenols: 4 methyl-3-chlorophenol (MF), 4-chlorophenol (CF), 2,4-dichlorophenol (DCF) and 2,4,6-trichlorophenol (TCF). Both systems were studied with amperometric electrochemical sensors: the hanging mercury drop electrode, for the inorganic mixtures, and Sonogel-Carbon electrodes, recently developed by our research group, for the organic system. With respect to the inorganic system, low errors were obtained for the individual concentration values of the two metallic ions in every mixture. In the case of the system composed by four chlorophenols, the best predictions (the lowest errors) of concentration were for the TCF analyte; the rest of the chlorophenols registered an increasing order of errors as follows: MF < DCF < CF. Finally, it has to be noticed that the grade of overlapping studied in both situations was very high; the concentration values of the organic chemical species were always very close to the Sonogel-Carbon detection limit for this compounds and under the values reported in literature; besides, the processes of feature selection based on wavelet transform with the aim of resolve an electrochemical system implies a rather novel application.