Error propagation as a factor in selection of measurement intervals for the determination of polycyclic aromatic hydrocarbons by second-derivative spectrofluorimetry

The most suitable wavelength intervals were selected for the determination of 4 polycyclic aromatic hydrocarbons (PAHs; benzo[g,h,i]perylene, dibenzo[ a,h]anthracene, pyrene, and triphenylene) in very complex mixtures of 11 PA Hs: anthracene, benz[a]anthracene, benzo[a]pyrene, benzo[b]fluoranthene, be nzo[g,h,i]perylene, benzo[k]fluoranthene, chrysene, dibenz[a,h]anthracene, phenanthrene, pyrene, and triphenylene. The multiple linear regression algo rithm was applied to measurements made in several wavelength intervals prev iously selected on the basis of sensitivity and minimum number of interferi ng compounds, Of the different models obtained, those displaying minimum er ror propagation in the analytical result were selected. By applying the mod els proposed in this study, we precisely and accurately determined benzo[g, h,i]perylene, dibenz[a,h]anthracene, pyrene, and triphenylene in complex mi xtures-a feat that could not be achieved by the use of constant-wavelength spectrofluorimetry in combination with second-derivative techniques.