Analytica Chimica Acta 689 (2011) 198–205 Contents lists available at ScienceDirect Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca Parametric signal fitting by gaussian peak adjustment: A new multivariate curve resolution method for non-bilinear voltammetric measurements Santiago Cavanillas, José Manuel Díaz-Cruz ∗ , Cristina Ari ˜ no, Miquel Esteban Departament de Química Analítica, Universitat de Barcelona, Martí i Franquès, 1-11, E-08028, Barcelona, Spain article info Article history: Received 5 November 2010 Received in revised form 26 December 2010 Accepted 10 January 2011 Available online 18 January 2011 Keywords: Voltammetry Multivariate curve resolution (MCR) Non-linearity Gaussian peak adjustment (GPA) Potential shift Peak broadening abstract A new methodology based on the fitting of signals to parametric functions is proposed for the multivariate curve resolution (MCR) analysis of overlapping and peak-shaped voltammetric signals which progres- sively get broader or narrower and move along the potential axis, thus causing a dramatic loss of linearity. The method is based on the least squares fitting of gaussian functions at both sides of the peaks by using adjustable parameters for the peak height, position and symmetry. It consists of several home-made programs written in Matlab environment, which are freely available as supplementary material of the present work. The application to the systems Zn(II)–oxalate, and to the phytochelatin PC 5 in a wide pH range provides excellent results as compared to these of more conventional linear methods, which raises good expectations about future application to electrochemical and even non-electrochemical data. © 2011 Elsevier B.V. All rights reserved. 1. Introduction The application of Chemometrics to electroanalytical measure- ments has been quite scarce up to now as compared to the widespread use in the analysis of spectroscopic data [1–3]. Among the reasons for that, it can be mentioned the electrochemical tradi- tion of hard modelling through fundamental equations (Fick’s Laws, Nernst Equation) and the lack of linearity of many electrochemical processes, which prevents an accurate use of many chemometrical methods designed for linear data [3]. Multivariate curve resolution (MCR) methods constitute a good example for that. The especially versatile alternating least squares algorithm (MCR-ALS) was developed by Tauler et al. for spectro- scopic data [4] and in a few years it has become quite popular [5], but most of their applications have remained in the field of spectro- scopic measurements. As for the electroanalytical uses of MCR-ALS, they started quite early [6], but the bilinearity requirements of the method have restricted them to voltammetric measurements on a few particular systems which ensure such a linear behaviour. This is the case of electrochemically inert metal complexes, where the signals of the free metal, the complexes and, eventually, the free ligand remain at fixed potential along the whole experiment [7]. Unfortunately, this is not the usual situation in metal complexa- ∗ Corresponding author. Tel.: +34 93 403 91 16; fax: +34 93 402 12 33. E-mail address: josemanuel.diaz@ub.edu (J.M. Díaz-Cruz). tion studies by voltammetry. In most cases, the complexes are not totally inert from an electrochemical point of view, i.e., they sig- nificantly dissociate during the time taken by the measurement, which causes a progressive shift of the signals along the potential axis and, hence, a loss of data linearity. This can be qualitatively detected by a too large number of com- ponents or latent variables required to reproduce the matrix in principal component analysis (PCA) or partial least squares (PLS) with an acceptable error. However, from a quantitative point of view, it is difficult to measure the extent of non-linearity. In sys- tems where the number of expected components is known, the percentage of linearity could be measured as the percentage of PCA data reproduction with this prefixed number of components, but this is quite an exceptional situation in voltammetry. In the most favourable situations, the linearity decrease is not important and MCR-ALS can be applied with reasonable accuracy, but quite often the potential shift of the signals is so huge that MCR- ALS analysis becomes impossible or provides unrealistic results. In order to solve this problem, the shiftfit [8,9] and pHfit [10] algorithms were recently proposed. The first one corrects the data matrix from signal movements prior to the application of MCR- ALS. For this purpose, it optimises by least squares the potential shift of every pure voltammogram with respect to a reference posi- tion. This program works reasonably well when only a few signals are moving without overlapping with each other. The second algo- rithm can solve more intricate systems like those encountered in voltammetric pH titrations by imposing a shape restriction to the 0003-2670/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2011.01.017