Recent advance of spectroscopic instruments has allowed us to obtain a large amount of spectral data in machine readable forms.
High resolution molecular spectra contain abundant information on structures and dynamics of molecules. However, extraction
of such useful information necessitates a procedure of
spectral assignment in which each spectral line is assigned a set of quantum numbers. This procedure has traditionally been performed by making
use of regular patterns that are obviously seen in the observed spectrum. However, we often encounter complex spectra in which
such regular patterns may not be readily discerned. The purpose of the present work is to search for new methods which can
assist in assigning such complex molecular spectra. We wish to devise computer aided techniques for
picking out regular patterns buried in a list of observed frequencies which look like randomly distributed. We hope that we may depend on great computational power of modern computers.
Previously [1], [2], we have proposed a method, which we tentatively refer to as “second difference method” and suggested that this technique
may be developed as a useful tool for analysis of complex molecular spectra. This method has been tested with success on the
observed spectrum of a linear molecule DCCCl [1]. We have also presented a further test using an artificial data corresponding to an infrared spectrum of a linear molecule
HCCBr [2]. However, we recently encountered a set of data for which the method in the original form did not work well.