Multivariate curve resolution of nonlinear ion mobility spectra followed by multivariate nonlinear calibration for quantitative prediction Víctor Pomareda a, b, ⁎, Ana V. Guamán a, b , Masoumeh Mohammadnejad b, c , Daniel Calvo a , Antonio Pardo a , Santiago Marco a, b a Departament d'Electrònica, Universitat de Barcelona, Martí i Franqués 1, 08028, Barcelona, Spain b Artificial Olfaction Lab, Institute for Bioengineering of Catalonia, Baldiri i Rexach, 4‐8, 08028, Barcelona, Spain c Faculty of Chemistry, Bu-Ali Sina University, Hamedan, Iran abstract article info Article history: Received 30 January 2012 Received in revised form 12 April 2012 Accepted 8 June 2012 Available online 18 June 2012 Keywords: Ion mobility spectrometry Multivariate curve resolution Gas phase ion chemistry Multivariate calibration In this work, a new methodology to analyze spectra time-series obtained from ion mobility spectrometry (IMS) has been investigated. The proposed method combines the advantages of multivariate curve resolution- alternating least squares (MCR-ALS) for an optimal physical and chemical interpretation of the system (qualita- tive information) and a multivariate calibration technique such as polynomial partial least squares (poly-PLS) for an improved quantification (quantitative information) of new samples. Ten different concentrations of 2- butanone and ethanol were generated using a volatile generator based on permeation tubes. The different con- centrations were measured with IMS. These data present a non-linear behavior as substance concentration in- creases. Although MCR-ALS is based on a bilinear decomposition, non-linear behavior can be modeled by adding new components to the model. After spectral pre-processing, MCR-ALS was applied aiming to get infor- mation about the ionic species that appear in the drift tube and their evolution with the analyte concentration. By resolving the IMS data matrix, concentration profiles and pure spectra of the different ionic species have been obtained for both analytes. Finally, poly-PLS was used in order to build a calibration model using concentration profiles obtained from MCR-ALS for ethanol and 2-butanone. The results, with more than 99% of explained var- iance for both substances, show the feasibility of using MCR-ALS to resolve IMS datasets. Furthermore, similar or better prediction accuracy is achieved when concentration profiles from MCR-ALS are used to build a calibration model (using poly-PLS) compared to other standard univariate and multivariate calibration methodologies. © 2012 Elsevier B.V. All rights reserved. 1. Introduction Ion mobility spectrometry (IMS), is a simple, portable and sensitive instrumental analytical technique that provides rapid response to trace levels of volatile organic compounds. IMS ionizes volatile compounds with different methods, namely radioactive, ultraviolet light, corona discharge, etc., and ions are separated based upon their mobilities in weak electric fields at atmospheric pressure [1]. It was developed in the early 1970s and it has been mainly used in security applications as chemical warfare agent detection [2,3], screening for illicit substances [4] and protection against explosives among others [1,5]. However, in recent times, IMS applications in other areas have been explored such as environmental monitoring [6,7] and pharmaceutical applications [8] and, gradually it is expanding its range of applications to food and beverage applications [7,9,10], clinical applications [11–14] and indus- trial applications [7,15–17], among others. Although IMS technology has demonstrated good performance in many scenarios, it has some disadvantages. In particular, in IMS instru- ments based on radioactive ionization sources the ionization of the sample occurs by ion/molecule reactions rather than by direct ionization of the an- alyte, thus hindering IMS qualitative and quantitative analyses [18]. In IMS atmospheric chemistry, the formation of protonated molecules (MH + ) is due to an effect of proton transfer from the reactant ions (pre- dominantly hydronium ions: (H 2 O) n H 3 O + ) to the analyte molecules; this chemical process could occur either in the reactant region or in the drift tube. Additionally, when the concentration of analyte molecules is ad- equately high, a proton-bound dimer is formed as a result of the clustering of protonated monomer with an additional analyte molecule [1,19]. Protonated monomer : M þðH 2 OÞ n H 3 O þ ↔MH þ þðn þ 1ÞH 2 O ð1Þ Proton À bound dimer : MH þ þ M↔MH þ M ð2Þ High concentrations also favor the formation of proton-bound tri- mers, proton-bound tetramers and so on, but normal IMS spectra do not show these ion species due to their extremely short lifetimes. Chemometrics and Intelligent Laboratory Systems 118 (2012) 219–229 ⁎ Corresponding author at: Institute for Bioengineering of Catalonia (IBEC), Baldiri i Rexach, 4‐8, 08028, Barcelona, Spain. Tel.: +34 93 403 11 18; fax: +34 934 021 148. E-mail address: vpomareda@ibecbarcelona.eu (V. Pomareda). 0169-7439/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.chemolab.2012.06.002 Contents lists available at SciVerse ScienceDirect Chemometrics and Intelligent Laboratory Systems journal homepage: www.elsevier.com/locate/chemolab