Separation of optical properties for multicomponent samples and determination of spectral similarity indices based on FEDS0 algorithm

Abstract

The spectral analysis of materials is a suitable, fast and low-cost method used for compounds identification and prediction of properties in many laboratory and industrial applications. However, in multicomponent samples, due to the huge data quantity contained in a single spectrum and overlapping of the individual spectral signatures of each component, it is necessary the use of statistical tools in order to locate and separate key information. This study presents an algorithm of operational transformations based on functionally-enhance derivative spectroscopy (FEDS), which allows to rewrite any signal in terms of its critical points (maxima, minima, and inflection points), representing in a simple and accurate way the information that define the spectral response. Here, reduction data method is added to FEDS in order to extract characteristic-signal patterns related to the spectrum (FEDS 0 ). This method was used for optical analysis of mixtures (acetone and triethylamine) and multicomponent samples (e.g., agricultural soils) in the infrared region, with the purpose to identify the adequate bands for the development of spectral comparisons and so, establish similarity indices for prediction of chemical characteristics. By comparing the FEDS 0 signals associated to samples of comparable properties, spectral coincidences higher than 74% and Pearson linear correlations up to 0.98 have been found, while for samples of different composition the spectral similarity decreases up to 20% with linear correlations lower to 0.20. This indicates that FEDS method provides high sensitivity to differences in optical spectra, which are directly related to sample composition. In general, since FEDS 0 spectra are composed only by few tens of data, that includes all relevant optical information of a sample, the spectral comparison perform through them are much more accurate than those developed by conventional FTIR signals.