PLS-Computation and simultaneous spectrophotometric determination of traces of TI(IV),Mo(VI), and Sn(IV)

The partial least square method (PLS) has been applied to the simultaneous spectrophotometric determination of Ti(IV), Mo(VI), W(VI), and Sn(IV) which are similar in chemical properties and often coexist in samples. Highly sensitive chromogenic reactions are known between them and salicyl fluorone (SAF) in the presence of cetyl trimylammonium bromide (CTMAB), a surfactant. Their absorption spectra overlap seriously. The statistical results for 16 sets of calibration samples indicate that the additives of all samples at wavelength over 520 nm (|ΔA| < 0.1) are better than that at the wavelength range of 500 to 520 nm except set 2 and set 14 in which the W content was at maximum (25 μg per mL) and the optimum point is at 530 nm. The computation error is the smallest when the interval of wavelength taken uniformly from 530 nm to two sizes is 2 to 6 nm and corresponding number of wavelength is 2 to 7. In order to obtain both smaller computation error and shorter computation time, the suitable number of calibration samples should be 3 to 4 times that of components after considering fully concentration ratios of all components. Twenty sets of simulation samples were used to determine the number of meaningful factors (abstract components). The results of 16 sets of calibration samples with different contents of Ti, Mo, W, and Sn indicate that the PLS method is superior to the conjugate gradient (CG) method and linear programming (LP) method due to application of the calibration samples. The features of PLS method are discussed