Procedure to build a signal transfer set, independent of the target analytes, between a portable fluorimeter based on light-emitting diodes and a master fluorimeter

Abstract

The need of performing “in situ” analytical determinations together with the availability of high-power deep UV-LEDs have led to the use of fluorescence spectroscopy. However, it is necessary to register excitation-emission matrices (EEM) to obtain three-way data which can be decomposed using parallel factor analysis for enabling the unequivocal identification of the analytes. In this context, the feasibility of transferring EEM between a portable fluorimeter based on LEDs and a master fluorimeter based on a xenon source has been recently reported without losing analytical quality. To build the transfer function, the signals of the same N samples must be recorded in the portable and in the master fluorimeter. In literature, these samples always contained the target analytes so the EEM signal transfer methodology is very limited in practice. Therefore, the challenge is to search for a set of samples whose EEM enable to perform the signal transfer without previously knowing the target analytes.

The aim of this work is the design of a procedure to build N mixtures of P fluorophores so the N EEM would be optimal for the signal transfer. Five criteria have been defined a priori to identify the quality of a transfer set made up of N EEM. Then, a procedure has been designed to obtain the n mixtures of the P fluorophores “in silico” using the Pareto front of the optimal solutions and a desirability function to choose the desired N EEM.

The procedure has been used to find five mixtures of the three chosen fluorophores for the signal transfer (coumarin 120, DL-Tyrosine and DL-Tryptophan) which are chemically different from the analytes of interest (enrofloxacin and flumequine) and are contained in a different matrix. These two analytes are antibiotics which have maximum residue limits set in the EU legislation in force.

The correlation coefficients between the experimental reference spectra and the PARAFAC spectral loadings of the data registered with the master fluorimeter were greater than or equal to 0.999 in all cases. On the other hand, the correlation coefficients obtained with the portable fluorimeter ranged from 0.900 to 0.950 once the procedure was applied to the two antibiotics. Therefore, the unequivocal identification of the analytes was ensured.