Homogeneity testing of microanalytical reference materials by electron probe microanalysis (EPMA)

Homogeneity testing of candidate reference materials requires distinguishing the effects of measurement uncertainty of the analytical method from true compositional variations within the material. Many in situ microanalytical techniques do not allow classical ANOVA homogeneity testing due to the infeasibility of truly replicated analyses on the same analysis volume. This also applies to the analysis of beam-sensitive and light element-bearing materials by electron probe microanalysis (EPMA). This reality has led me to reconsider the homogeneity index approach used in the testing of microanalytical reference materials by EPMA. Based on statistical considerations, I show that the homogeneity index is suitable for statistical significance testing using F and chi-squared statistics and allows estimating the contribution of compositional heterogeneity to the total uncertainty budget of the referenced values. However, there are problems of bias and masking of small compositional variations by measurement uncertainty. This contribution shows the strong impact of the total number of measurements on the resolution of a microanalytical homogeneity study and discusses how to quantify the relative contribution of heterogeneity to the total uncertainty budget. I present an example of EPMA to illustrate this approach and show some pitfalls and limitations in its application.