Application of partial least-squares regression to the resolution of highly correlated spectra. Simultaneous spectrofluorimetric determination of Al(3+), Ga(3+) and In(3+)

The simultaneous spectrofluorimetric determination of mixtures of aluminium, gallium and indium as their 8-hydroxyquinoline complexes following extraction into chloroform is studied. The high collinearity of the spectra hindered their resolution by multiple linear regression (MLR) methodology; therefore, experimental data were processed by partial least-squares regression (PLSR) methodology. A previous step in the study of three-dimensional fluorescence spectrum is possible to select the best information to quantify this system with high collinearity. Finally the optimal conditions for quantitation, the best data preprocessing procedure and the most suitable spectral mode for calibration were established. Using an external set allowed the three analytes to be determined simultaineously at concentrations below 1 mug ml(-1) with errors less than 10% for aluminium and indium, and 15% for gallium.