Abstract
In this work we introduce a new method for testing spectral line shape models and optimizing the parameters in any parametric model. Given some general parametric line shape and a piece of a spectrum, the method finds the optimal values for the parameters and gives a number which tells how well the spectrum under consideration is explained by that model. The number of spectral lines under analysis may be more than one, and their exact locations need not be known. This characteristic follows from the property that the method does not need the information about line positions and amplitudes at all. Thus, in the absence of a singlet line, a set of overlapping lines can also be analyzed. The analysis is carried out in the signal domain by utilizing linear prediction. Application examples of the method to a molecular spectrum measured in gas phase are given. The results suggest that the Voigt line shape, despite its common use, is not a correct model in molecular spectroscopy. Its limitations become evident when one is trying to enhance the resolution by linear prediction, which requires detailed knowledge of the line shape. Instead a stochastic model, which is also tested, turns out to be rather promising.
Get full access to this article
View all access options for this article.