Estudo sobre o emprego de quimiometria em espectroscopia de impedância eletroquímica

Abstract

The application of a sinusoidal potential to an electrochemical system causes the rise of a current that is also sinusoidal, whose characteristics depend on the impedimetric processes that occur at the electrode-solution interface and of the applied signal frequency. This is the principle of Electrochemical Impedance Spectroscopy. Quantitative models based on this technique are generally constructed by relating the change in analyte concentration to the values of a single electrical element of an equivalent circuit used to model the behavior of spectra. The first step of the work proposed a strategy to optimize the construction of these quantitative models by correlating more than one electrical element of the circuit equivalent to the variation of analyte concentration. Multivariate models for ferrocyanide, catechol and hydroquinone quantification using multiple linear regression were built. The most significant result was observed with the multivariate hydroquinone model which presented an RMSEP of 0.79 mol L-1 against 2.51 mol L-1 achieved with the univariate model, both in the quantification of hydroquinone in spiked tap water samples. An alternative methodology for the construction of quantitative models, which does not need to use equivalent circuits, was proposed in the second stage of this work. The strategy was to construct partial least square models (PLS) using the impedance expressed as complex number (Z'+ jZ''). Quantitative models for ferrocyanide, catechol and hydroquinone were constructed and the results obtained by these models were compared to those provided by the PLS models constructed with the real impedance part (Z'), imaginary part (Z''), absolute impedance (|Z|) and the phase angle (ϕ). The third stage of the work was the construction of models to classify samples of pure and adulterated coffees with shells and sticks, using impedance data expressed as complex numbers. The techniques employed were PLS-DA and SIMCA. It was found in all studied cases that the construction of quantitative and qualitative models with the impedance expressed as a complex number led to similar or better results than the models constructed with only one of the impedance related information (Z’, Z’’, |Z| and ϕ). This is an advantage as it is not necessary to evaluate all arrays as the use of complex data will lead to better or similar results.