Chemometric handling of spectral-temporal dependencies for liquid chromatography data with online registering of excitation-emission fluorescence matrices

Abstract In this work, the generation and posterior chemometric resolution of third-order data, obtained from samples processed by liquid chromatography (LC) with online registering of excitation-emission fluorescence matrices (EEM) is reported. Samples were instrumentally processed in a relatively short time, and neither an intentional reduction of the linear flow rate nor an unconventional fluorescence instrument were required. Through the inclusion of external circuitry based on open-source hardware, the occurrence time of each individual fluorescence intensity reading was recorded. For the reported instrumental setup, irregular signal sampling was verified. In order to consider samples-specific time measurements, the PARAFAC (Parallel Factor Analysis) algorithm, and the derived APARAFAC (Augmented-PARAFAC) strategy, were adapted. The functional information was employed during the computational stages, through the development and implementation of smoothing strategies. To tackle differences between the rate of spectral acquisition and the rate of change in the concentration of the mobile fluorophores, Expectation Maximization was implemented. Data from samples with one calibrated analyte (Vitamin B6-Pyridoxine), in presence of uncalibrated interferents, were modeled. In order to preserve the original data structure, unfolding data operations were minimized. The resolved profiles of all species were in agreement with the corresponding chromatographic and spectral references. Results suggest that the effects derived from the loss of trilinearity previously reported in the literature for LC-EEM data, depend on interpretation and subsequent modeling of the data. The reported strategies can be useful with other flow techniques and kinetics.