Comparison of various peak evaluation techniques for limited and extended dynamic ranges in flow injection systems

摘要

Various peak evaluation and calibration techniques have been compared for two-line and gas-diffusion FIA systems, designed for the determination of iron(II) and ammonium, respectively. The univariate peak evaluation techniques comprised of peak height, peak area, peak width, slope value at the first inflection point and triangulated peak height and area. All these techniques were utilised to condense each FIA peak to one single value prior to fitting a second order polynomial to the data by regression analysis. When multivariate data analysis was performed each peak was represented by a large number of values evenly distributed over the entire FIA peak and partial least squares (PLS) regression was applied. The injected standard solutions covered a large concentration range, three orders of magnitude for both system types, and a relative viscosity (eta/eta(0)) range of 1.000-1.128 for the two-line system and 1.000-1.104 for the gas-diffusion system. Triangulated peak area data can replace peak height and peak area for univariate evaluation in the two-line FIA system applied for standards within the limited, linear concentration range, and become the measure of choice or the extended, non-linear concentration range. For the gas-diffusion method the peak height, the peak area, the slope value at the first inflection point and the triangulated peak height or area work equally well in the limited concentration range. For the extended concentration range the only univariate peak evaluation technique that was found to work satisfactorily nias peak area by triangulation. PLS regression provides an alternative, universal solution to the problems with matrix variations and large concentration ranges but in a given situation it is not always superior to the best of the various univariate approaches. The PLS calibration technique requires a much larger number of standards in comparison with the univariate approaches. (C) 1998 Elsevier Science B.V. [References: 13]