Pattern recognition applied to vapour-phase infrared spectra: characteristics of νOH bands

Abstract

With the commercial availability of infrared spectrum libraries in computer accessible form, a new application of these libraries, the investigation of structure-spectrum correlations, has become possible in addition to the conventional application, the identification of unknown substances on the basis of spectral search in the library. This paper deals with the νOH bands of alcohols, also including diols and diol monoether type compounds, investigating correlations between chemical structure and the features of spectra. For monoalcohols, the frequencies measured in dilute solution and vapour phase, respectively, are correlated; the former were taken from literature, the latter from the vapour phase spectrum library of EPA. For the other types of compound, vapour phase νOH frequencies of free and hydrogen bonded hydroxy groups were studied, and the differences correlated with the hydrogen bonded structures. Peak maxima were determined by a peak picking program; in the case of overlapping bands Fourier self-deconvolution was applied prior to peak picking. For more seriously overlapping bands least-squares curve fitting was applied. In studying correlations, a cluster analysis method (KMEANS) was also used.