Orthogonal signal corrected two-dimensional (OSC 2D) correlation infrared spectroscopy Yuqing Wu a, * , Isao Noda b , Filip Meersman c , Yukihiro Ozaki d a Key Lab for Supramolecular Structure and Material of Ministry of Education, Jilin University, No. 2699 Qianjin Street, Changchun 130012, PR China b The Procter & Gamble Company, 8611 Beckett Road, West Chester, OH 45069, USA c Department of Chemistry, Katholieke Universiteit Leuven, Celestijnenlaan 200 F, B-3001 Leuven, Belgium d School of Science and Technology, Research Center for Environment Friendly Polymers, Kwansei-Gakuin University, Gakuen, Sanda, Hyogo 669-1337, Japan Received 26 November 2005, received in revised form 15 February 2006; accepted 6 March 2006 Available online 24 April 2006 Abstract Orthogonal signal correction (OSC) is a chemometrical data processing technique used for removing the information unrelated to the target variables based on the constrained principal component analysis (PCA). The combined use of OSC filtering and two-dimensional (2D) correlation analysis, which is called orthogonal signal corrected 2D (OSC 2D) correlation spectroscopy, is proposed in the present study to enable one to develop high quality of 2D correlation spectra by eliminating any information unrelated to the external variables. A set of temperature-dependent infrared spectra of poly(N-isopropylacrylamide) (PNiPa) in aqueous solutions, as well as the simulated spectra developed by adding different random noise spectra or a systematic noise spectrum of contaminating water after multiplied with a random weight factor to the experimental spectra, were used as examples. The results provided by OSC 2D were compared to those obtained by 2D without OSC filtering, and OSC 2D spectrum has demonstrated its substantial power in eliminating signals that are unre- lated to the external variable and the great improvement in the synchronous spectrum. Ó 2006 Elsevier B.V. All rights reserved. Keywords: Orthogonal signal correction; Two-dimensional correlation spectroscopy; Infrared spectroscopy; Poly(N-isopropylacrylamide) 1. Introduction Orthogonal signal correction (OSC), introduced first by Wold et al. [1], has been widely applied to the removal of unwanted variations from various kinds of experimental data, such as near-infrared (NIR) spectra [2–7], X-ray pow- der diffraction [8], fluorescence spectroscopy [9], and 13 C cross-polarization/magic angle spinning NMR spectrosco- py [10,11]. There are several approaches to the scheme gen- erally known as the OSC filtering. The basic idea is to remove a certain portion of the systematic variations of data not directly affected by the controlling variable, such as temperature and pressure. The simplest approach is that of direct orthogonalization (DO) proposed by Andersson [12], where the entire data matrix is directly pretreated to remove any part that is not affected proportionally by the external variables. Other approaches include those of Wold et al. [1] and Fearn [13]. In the former, some portion of the PLS score, which is not affected by the external variable, is removed. In the latter approach, the portion of data matrix, which is not proportional to the first PLS weight loading, is removed. The original aim of OSC pretreatments was to produce a calibration equation to predict concentrations from spectra for future samples. For example, OSC was used for partial least squares (PLS) regression of NIR spectra, since NIR spectra often contain major sources of variation that are of little or no predictive value. These approaches all work in a similar manner with the same objective of removing the systematic variations of data that are not caused by 0022-2860/$ - see front matter Ó 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.molstruc.2006.03.030 * Corresponding author. Tel.: +86 431 5168730; fax: +86 431 5193421. E-mail address: yqwu@jlu.edu.cn (Y. Wu). www.elsevier.com/locate/molstruc Journal of Molecular Structure 799 (2006) 121–127