Quantification of non-detectable analytes in samples of unknown stoichiometry with EPMA

JORGE TRINCAVELLI; GUSTAVO CASTELLANO

Barcelona, España

Workshop; 3rd EMAS Regional Workshop; 1998

European Microbeam Analysis Society

Conventional electron-probe microanalysis is based on the detection of characteristicradiation. The measured intensities I for each element are compared with those corre-sponding to standards of known composition, Io, and each concentration C is assessed bymeans of the so-called ZAF correction models [1]:\frac{I}{I_o} \;=\; \frac{C}{C_o}\, \cdot \, \cal ZAFwhere Co is the concentration of the corresponding element in the standard. Since theZAF correction models involve the sample concentrations, this equation implies an iter-ation.It is usual to have samples with elements undetectable with conventional energy-dispersive systems; this is the case of atomic numbers lower than 11, for which character-istic K photons have energies below 700eV. Often these elements form compounds: thisallows to obtain concentrations by means of stoichiometry [2]. When this is not the case,the presence of these elements must not be ignored by normalizing the detectable analytesto 100%, since non-detectable elements may play an important role in the absorption ofx-rays as well as in the scattering of electrons: disregarding them may lead to erroneousrelative concentrations of the heavy elements.With the aim of giving a solution to this problem, an algorithm has been developed.Besides the characteristic intensities, the value of the bremsstrahlung in the sample andone standard, as well as the backscattering coefficient are required in order to completethe set of equations, since two unknowns have been added to the concentrations of thedetectable analytes: the mean atomic number of these `hidden' elements as well as theirconcentration. Kramers expression [3] is used for modelling the continuum radiation, aswell as an adequate expression for the backscattering coefficient [4]. The resulting set ofequations has been implemented in the program MULTI [5], and several samples havebeen evaluated by means of this method.