Solving chemical classification problems using polarized Raman data
Article first published online: 6 MAY 2010
Copyright © 2010 John Wiley & Sons, Ltd.
Journal of Raman Spectroscopy
Volume 42, Issue 1, pages 21–35, January 2011
How to Cite
Hassing, S., Jernshøj, K. D. and Hedegaard, M. (2011), Solving chemical classification problems using polarized Raman data. J. Raman Spectrosc., 42: 21–35. doi: 10.1002/jrs.2666
- Issue published online: 14 JAN 2011
- Article first published online: 6 MAY 2010
- Manuscript Accepted: 28 FEB 2010
- Manuscript Received: 9 JUL 2009
- polarized Raman data;
- polarization ratio;
- chemometric methods
When solving chemical classification problems, multivariate analysis has proven to be an important mathematical tool. Unpolarized spectroscopic data, IR, NIR, and UV-Visible absorption data and unpolarized vibrational Raman data, are typically analyzed by two-way chemometric methods, e.g. principal component analysis (PCA). When the unpolarized spectra of the different molecules are almost identical, the PCA results in low recognition ratios or even fails. In contrast to absorption processes, Raman scattering can provide polarized data. It is shown, by using mathematical simulations, that the outcome of the PCA can be improved considerably by using the polarized, vibrational Raman data instead of the unpolarized data. The improvement stems from the increased amount of molecular information, which is now available for the PCA of the vibrational data, because the polarization properties of the scattering, expressed through the depolarization ratio (DPR), is very sensitive to small changes in distinct molecular properties and insensitive to sample and experimental variations. For molecules possessing some symmetry, a change of the DPR may be induced by a decrease in symmetry and for highly symmetric molecules non-dispersive Raman modes typically become dispersive. For dispersive modes, a wavenumber-dependant change of the DPR may also result from a small energy shift of an allowed electronic transition. We show that the increased information content inherently present in the polarized data, opens up new possibilities for combining the solution of classification problems with an unveiling of details of the different properties and processes in bio-physic due to various perturbations and changes of the structure of the bio-molecules. It is also demonstrated that the increased access to molecular information enables in vitro detection of molecular changes often encountered when analyzing biological functions, which are reflected in changes in the excited electronic states. Copyright © 2010 John Wiley & Sons, Ltd.