Reduction of interferences in graphite furnace atomic absorption spectrometry by multiple linear regression modelling

The multivariate effects of Na, K, Mg and Ca as nitrates on the electrother mal atomisation of manganese, cadmium and iron were studied by multiple lin ear regression modelling. Since the models proved to efficiently predict th e effects of the considered matrix elements in a wide range of concentratio ns, they were applied to correct the interferences occurring in the determi nation of trace elements in seawater after pre-concentration of the analyte s. In order to obtain a statistically significant number of samples, a larg e Volume of the certified seawater reference materials CASS-3 and NASS-3 wa s treated with Chelex-100 resin; then, the chelating resin was separated fr om the solution, divided into several sub-samples, each of them was eluted with nitric acid and analysed by electrothermal atomic absorption spectrome try (for trace element determinations) and inductively coupled plasma optic al emission spectrometry (for matrix element determinations). To minimise a ny other systematic error besides that due to matrix effects, accuracy of t he pre-concentration step and contamination levels of the procedure were ch ecked by inductively coupled plasma mass spectrometric measurements. Analyt ical results obtained by applying the multiple linear regression models wer e compared with those obtained with other calibration methods, such as exte rnal calibration using acid-based standards, external calibration using mat rix-matched standards and the analyte addition technique. Empirical models proved to efficiently reduce interferences occurring in the analysis of rea l samples, allowing an improvement of accuracy better than for other calibr ation methods. (C) 2000 Elsevier Science B.V. All rights reserved.