6.3.2 Sample dilution to facilitate linear regression

Some sample preparation procedures can be used to obtain a homogeneous distribution of the sample in a given matrix. Such procedures include thorough homogenization by mixing the sample with a binding agent and pressing into pellets or by mixing with a fluxing agent and further fusion in beads. Pressing is simpler and cheaper, but the final homogeneity can be worse than that obtained with fusion.

Both procedures lead to the dilution of the sample, and therefore minimize the influence of the original sample in the final matrix, which for large dilution rates can be considered as constant (e.g. that of the binder or fluxing agent). Calibrations can be made by measuring synthetic standards and validated using certified reference materials. As a rule, the final concentration of the element must be no greater than 2-3 %, to avoid departures from linearity arising from attenuation effects.

The typical relative accuracy of such procedures is better than 5 %. The major disadvantages are the more laborious sample preparation and the deterioration of the detection limits. As a rule, sample dilution techniques are applied for the determination of major constituents in complex samples. Trace elements can be prepared with a much lower dilution, and their concentrations determined from a second measurement, this time incorporating a calculation of the attenuation corrections using the already known concentrations of major elements.

IAEA QXAS include the program "Direct comparison of count rates", which can be accessed from the initial Menu option "Quantitative analysis" to implement simple linear regression calibrations.