Optimisation using a surface response methodology of a system with drift: Application to a flow analysis system with electrochemical detection Original Research Article

The aim of this work is to optimise the multivariate signal obtained by means of a continuous flow system with electrochemical detection when three phenolic compounds (2,4-dimethylphenol, phenol and 4-nitrophenol) are analysed. Firstly, an experiment is carried out according to a central composite design for the factors percentage of methanol and pH of the carrier solution, taking as response of the design the net analyte signal (NAS) for each analyte. The system of analysis has drift that has to be taken into account when optimising the analytical signal. To do so, between each two individual experiments, a reference experiment is inserted in the experimental conditions corresponding to the centre of the design. This allows one to directly correct the response of the design taking as linear correction factor the variation of NAS between a previous reference sample and the subsequent one. However, a detailed analysis of the data suggests the interaction between the experimental factors and the system of analysis, which is confirmed with a second experiment. As a consequence, correcting the voltammograms by subtracting the contribution of the principal component related to drift is proposed, which is carried out separately for analytes and blank solutions, and next estimating the NAS. Using this corrected NAS as response of the design, significant quadratic models with high explained variance for each studied compound are obtained. In addition, an optimum is found when common conditions are desired for all the analysed compounds using the desirability function, being found at 1.8% of methanol and pH 5.9.