SPECTRAL RESOLUTION AND PREDICTION OF SLIT WIDTHS IN FLUORESCENCE SPECTROSCOPY BY TWO- AND THREE-WAY METHODS

摘要

Modern spectrofluorometers have several instrumental settings to be adjusted and decided on before sample measurement, e.g. excitation and emission slit widths, emission scan velocity and spectral ranges to be recorded. The influences of these settings on the recorded spectra are crucial, particularly when applying full fluorescence spectra in the analysis of a given problem. The effect on the fluorescence emission spectra when changing the slit widths is studied in detail by recording the emission spectra of an ovalene standard block at all possible excitation (3-15 nm) and emission (3-20 nm) slit width combinations. By the two-way curve resolution method alternating regression (AR) it is possible to resolve the emission spectra into three hidden spectra describing the coarse, medium coarse/fine and fine structure of the recorded spectra. By the three-way methods PARAFAC and Tucker it is possible to separate the effects of both the excitation and emission slit widths on the recorded spectra. An analogous analysis of a natural sugar sample results in a one factor PARAFAC solution, probably because of the large amount of different substances found in a table sugar sample resulting in rather featureless emission spectra not so susceptible to influence by the instrumental settings. Finally, it is demonstrated that two-way unfold PLS, PARAFAC and Tucker regression models are able to predict the excitation and emission slit widths from the recorded emission spectra. The root mean square errors of prediction (RMSEP) are about 0.5 nm (R ≈ 0.990) for the excitation slit and 0.3 nm (R ≈ 0.999) for the emission slit.