Novel application and comparison of multivariate calibration for the simultaneous determination of Cu, Zn and Mn at trace levels using flow injectiondiode array spectrophotometry

Three different calibration approaches: Partial Least Squares, Unfold Parti al Least Squares and n-way Partial Least Squares are compared in terms of e xplained variance, root mean square error of calibration and root mean squa re error of cross-validation. Attention is also focused on the application of genetic algorithms to spectral data as a way to obtain an improvement in calibration accuracy. Influence of initial starting conditions for the gen etic algorithms (population size, mutation probability, % initial terms etc .) was investigated by means of factorial experimental designs. A simple fl ow injection manifold coupled to a diode array spectrophotometer and multiv ariate calibration were employed in order to determine Ni, Cd, Pb, Cu, Zn a nd Mn. Calibration was not successful for the three first elements. Rapid d etermination of the microamounts of Cu, Zn and Mn was performed in the pres ence of concomitant ions with a sample throughput of 180 samples h(-1). Sin ce three metals were simultaneously determined, this corresponds to 540 det erminations h(-1). The effect of different matrix interferences (Fe, Al, Ca , Mg, Cr, Sr, Co) was studied in natural and synthetic water samples. Chemi cal masking with NaF was assayed for the interferent ions Fe and Al. The ty pical recoveries ranged from 91 to 94% with typical relative standard devia tion between 5 and 10 %. (C) 1999 Elsevier Science B.V. All rights reserved .