Identification of unknown pure component spectra by indirect hard modeling

Indirect hard modeling (IHM) is a physically motivated spectral analysis principle. It utilizes nonlinear spectral hard models generated by peak fitting of the pure spectra. This approach allows the consideration of various nonlinear effects such as peak variations or spectral shifts. Compared to established methods, less calibration samples are required and basic calibration transfer is performed inherently. To extend the applicability of IHM, which currently requires knowledge of the pure component spectra, two methods for the identification of pure spectra are presented in this work. These methods work automatically on a mathematically objective basis and do thus not depend on the expertise of the user. As IHM relies on an underlying physical picture of the spectra, the relevant information in the input data is exploited very efficiently especially for selective spectra, and nonideal spectral behavior is captured throughout the identification process. Compared to established SMCR methods the number of required spectra is reduced. The first method, complemental hard modeling (CHM), is introduced for the case that a single pure spectrum is unknown. The method is based on a deconvolution approach and only requires a single mixture spectrum as input data. The second method, hard modeling factor analysis (HMFA), is conceptually related to SMCR methods. It allows the identification of all pure spectra in a completely unknown mixture from a limited set of mixture spectra. As shown in this work, even highly collinear data can be employed.