Quantitative spectroscopic methods have given birth to a number of widely used analytical tools, yet being a subject of active research and development. In spite of numerous advantages, some difficulties still exist. These include, but are not restricted to broad and highly overlapping bands, typical for multicomponent mixtures. These issues present particular difficulties for exploratory studies and the analysis of unknown systems.

Chemometrics and closely related approaches such as independent component analysis (ICA) can offer a solution in this situation. One branch of the research in this field focuses on the development and practical application of algorithms aimed at the decomposition of experimental spectra X of a multicomponent mixture onto a matrix of concentration profiles A and a spectral profile matrix S. Methods of formal modeling that does not require any prior knowledge about the system under study (its composition, concentration range, etc.) are employed in such an analysis.

We have proposed a simple, fast and reliable method of spectroscopic determination of metals in complex multicomponent mixtures. The analysis is based on decomposition of UV-VIS spectra of EDTA-metal complexes by means of chemometrics. We have compared the performance of Alternating Least Squares (ALS) with that of ICA methods (MILCA, SNICA, SIMPLISMA, RADICAL, FASTICA and JADE) in the identification and quantitative analysis of metals (Ca, Zn, Mg, Co, Cu and Mn) in complex mixtures. The results are presented as a series of experimental case studies, including spectroscopic analysis of metal model mixtures as well as real objects (multivitamin drugs, juices, etc.) in the UV-VIS region.

The results have shown that MILCA and MCR-ALS algorithms outperform other methods under study and are capable of providing robust results in well-defined accuracy ranges. By our estimate, the analytical errors in recovered concentrations is at the level of several percent, while the localization error of peak positions is comparable to instrumental uncertainties (below 1 nm for the techniques used).