ON THE EXTRACTION OF SPECTRA OF COMPONENTS FROM SPECTRA OF MIXTURES - A DEVELOPMENT IN FACTOR THEORY .2.

The theory of factor analysis developed in Part I [A. Muller and D. Steele, Spectrochim. Acta 46A, 817 (1990) [1] is extended to cases where m > 3, m being the number of factors. The effect of increasing the number of independent parameters by m(m-1) for T′ and the subsequent changes produced in the transformed eigenspectra, are investigated. Applications to actual spectra of mixtures of: (A) butan-2-one, ethyl ethanoate, cyclohexane and oct-1-ene (B) butan-2-one, ethyl ethanoate, cyclohexane and 2-methylpentane and (C) oct-1-ene, butan-2-one, ethyl ethanoate, cyclohexane and 2-methyl-pentane are presented. Using standard target techniques, an alternative approach is developed for the m> 3 case. This is based on transformation of the leading three eigenspectra, followed by identification and elimination of a component in a repetitive manner, until completion. The problems associated with the implementation of the Beer-Lambert Law are explored the need for accurate determination of a suitable wavenumber block to ensure the removal of the correct amount of each component, is examined. The ability to recognize at least one of the m component spectra per programme cycle is demonstrated. © 1990.