Substitution-dilution method to correct the matrix effect in multi-elementquantitative analysis by X-ray fluorescence

A mathematical model based on the dilution-addition method (DAM) for multi- elemental analysis using an X-ray fluorescence technique is proposed. The c onditions for sample preparation do not require both the unknown and standa rd samples to be similar in composition and mineralogy, and the unknown sam ple is replaced quantitatively by the standard sample, hence the denominati on substitution-dilution method (SDM). This method makes it possible to cor rect the matrix effect in multi-elemental quantitative analysis by X-ray fl uorescence for each analyte. The proposed model presents hyperbolic behavio ur of the experimental data when the X-ray fluorescence intensities are rep resented versus the substitution factor (h) for each analyte. After calcula ting the A/B parameter relations, which depend on the X-ray fluorescence in tensity of each analyte (I-i(ns)) and the substitution factor (h) and deter mining the analyte concentration in the multi-element standard sample (C-i( p)), it is possible to calculate the analyte concentration in the multi-ele ment unknown using an algorithm suggested for this purpose. This work studi es the substitution-dilution phase proposed in the method, and the factors arising from incorporation of the standard and diluent are established acco rding to the nature of the samples and the modifications. These factors mak e it possible to establish the experimental interval of analyte concentrati on, generally narrow, which corresponds to a section of the hyperbolic func tion which is so short that it can be accepted as linear. This linear model can be accepted for a wide variety of samples with a diluent/sample ratio greater than 10. The proposed linear method provides satisfactory results w hich are comparable to those calculated by applying the hyperbolic method. The proposed method (SDM)has been applied to two different types of matrice s, a binary alloy (without diluent, using the hyperbolic model) and a geolo gical sample (with diluent, using both hyperbolic and linear models). In al l cases the results were satisfactory. (C) 2001 Elsevier Science B.V. All r ights reserved.