Direct standardisation in multi wavelength fluorescence spectroscopy

Abstract

The application of direct standardisation methods by chemometrics has been successful in combination with multi wavelength near-infrared spectroscopy (NIR) for prediction of chemical composition. In this paper the methods of direct standardisation are applied to the analogue case of fluorescence spectroscopy introducing problems due to slit widths, stability of lamp, scanning speed, and pH; measuring at a sensitivity and signal-to-noise ratio of several orders of magnitude better than in NIR spectroscopy. Three different methods of spectrum standardisation have been tested: full spectrum standardisation (FSS), single wavelength standardisation (SWS) and piecewise direct standardisation (PDS). The fluorescence emission spectra stem from transmission measurements of 35 sugar solutions, containing small concentrations of a large number of fluorescent substances such as phenols and amino acids, at four different instrumental configurations. The performances of the algorithms are evaluated by the partial least squares (PLS) prediction errors obtained for two sugar quality parameters, namely colour and amino-nitrogen. It is shown that full spectrum standardisation performs well in the case of amino-nitrogen, while it is necessary to use the SWS or PDS methods to get acceptable colour predictions. The root mean square errors of prediction (RMSEP) obtained in the standardisation of four different instrumental configurations are 0.7–5.6% and 1.1–13.8% higher than the RMSEP values obtained by full recalibration for amino-nitrogen and colour, respectively. As transfer standards sugar samples that are extremely stable are preferred to single chemical standards for standardisation of complex spectra from biological samples. It was not possible to transfer emission spectra of sugar samples measured at pH 7.0 (phosphate buffer) to compare with emission spectra measured in pure water, indicating fundamental chemical changes due to difference in pH.