HEURISTIC EVOLVING LATENT PROJECTIONS - RESOLVING HYPHENATED CHROMATOGRAPHIC PROFILES BY COMPONENT STRIPPING

A 'component stripping' technique is presented for unique resolution o f multicomponent systems (with more than three overlapping components) by means of the heuristic evolving latent projections method. Partial chromatographic selectivity is utilized to resolve clusters of overla pping chromatographic peaks. Analytes with partial chromatographic sel ectivity are detected and their contributions in terms of estimated co ncentration profiles and spectra subsequently subtracted to obtain a d ata matrix of reduced rank. The procedure is repeated on the rank-redu ced data until the variation corresponding to chemical components has been exhausted and the data matrix has zero 'chemical' rank in all reg ions. At this stage, the concentration profiles and spectra of all the pure chemical components should be obtained. The simplicity and effic iency of the proposed approach are proved on a simulated four-componen t mixture characterized by an unresolved profile of hyphenated chromat ographic data. The new method compares favourably with the so-called r ank annihilation evolving factor analysis both with respect to precond itions and results.