Premium
Information‐theoretic chemometric analyses of Raman data for chemical reaction studies
Author(s) -
Chew Wee
Publication year - 2011
Publication title -
journal of raman spectroscopy
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.748
H-Index - 110
eISSN - 1097-4555
pISSN - 0377-0486
DOI - 10.1002/jrs.2660
Subject(s) - raman spectroscopy , chemometrics , principal component analysis , biological system , entropy (arrow of time) , chemistry , chemical species , acetic anhydride , chemical reaction , analytical chemistry (journal) , logarithm , computer science , mathematics , artificial intelligence , optics , thermodynamics , physics , organic chemistry , chromatography , mathematical analysis , catalysis , biology
A methodology of multivariate chemometric techniques based on the information‐theoretic approach was applied for elucidating chemical reaction information from a Raman data array R m ×ν that arises from in situ reaction monitoring. This reaction‐induced dynamic dataset R m ×ν can be contaminated by random cosmic ray spikes found in the midst of characteristic spectral variations associated with the disappearance or emergence of Raman active reactants, intermediates and products. Such spurious cosmic spikes were identified and removed using a novel and fast numerical approach based on maximum and minimum spectral entropy principles while preserving the genuine reaction‐induced spectral variations. Subsequently, the band‐target entropy minimization (BTEM) algorithm, a minimum spectral entropy based self‐modeling curve resolution technique, was applied to recover the pure component spectra of Raman active chemical species. Information gain through the chemometric analyses was calculated using information entropies with base 2 logarithm. This sequence of information‐theoretic chemometric analyses (or transinformations) was successfully tested on the reaction spectral data obtained from alcoholysis of acetic anhydride, which contains four Raman active chemical species. It is envisioned that this series of multivariate statistical analyses will be useful in chemical reaction studies and process analytical technology (PAT) applications that utilize in situ Raman spectroscopy to monitor transient dynamic changes in chemical concentrations, and also in Raman microscopy/imaging data containing spatial variations. Copyright © 2010 John Wiley & Sons, Ltd.