Solving chemical classification problems using polarized Raman data
Solving chemical classification problems using polarized Raman data
复制标题
使用偏振拉曼数据解决化学分类问题
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
M. Hedegaard
中科院分区：
文献类型：
--
作者：
S. Hassing;K. D. Jernshøj;M. Hedegaard
在解决化学分类问题时，多元分析已被证明是一种重要的数学工具。非偏振光谱数据、红外、近红外和紫外可见吸收数据以及非偏振振动拉曼数据通常通过双向化学计量学方法进行分析，例如主成分分析(PCA)。当不同分子的非极化光谱几乎相同时，主成分分析的识别率很低，甚至失败。与吸收过程相反，拉曼散射可以提供偏振数据。通过数学模拟表明，用极化的振动拉曼数据代替非极化的拉曼数据可以显著改善主成分分析的结果。这一改进源于分子信息量的增加，这现在可用于振动数据的主成分分析，因为通过去极化比(DPR)表示的散射的偏振性质对不同分子性质的微小变化非常敏感，而对样品和实验变化不敏感。对于具有一定对称性的分子，DPR的变化可能是由对称性的降低引起的，而对于高度对称的分子，非色散喇曼模通常变成色散的。对于色散模式，DPR与波数有关的变化也可能是由于允许的电子跃迁的小的能量移动造成的。我们表明，极化数据中固有的信息量的增加，为将分类问题的解决与揭示由于生物分子结构的各种扰动和变化而导致的生物物理中不同性质和过程的细节相结合开辟了新的可能性。研究还表明，增加对分子信息的获取使在分析生物功能时经常遇到的分子变化能够在体外检测到，这些变化反映在激发电子态的变化中。版权所有©2010 John Wiley&Sons，Ltd.
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.