HIERARCHICAL ANALYSIS OF MOLECULAR-SPECTRA

A method of characterizing molecular spectra is proposed here. Hierarchical trees are generated from molecular spectra by following the branching of peaks as resolution is changed continuously. The trees are analyzed using several techniques developed in the classification and multivariate analysis literature. The information obtained from these techniques includes (1) dimensionality of geometrical representations of the trees; (2) clustering statistics; and (3) number of distinct time scales inherent in the spectra. Several model spectra are studied to establish the utility of the hierarchical analysis and a portion of a vibronic spectrum of NO2 is also investigated.