From the perspective of the uncertainties in chemical measurements all uncertainty sources should be part of the uncertainty of the reference material. When the primary methods are not available, interlaboratorial comparisons are used as a means of certification. The material to be distributed to the laboratories should have its homogeneity confirmed. The uncertainty due to this factor will be added to the characterization uncertainty. This work presents a homogeneity study of a lot of silicon metal of chemical degree where the uncertainty due to inhomogeneity is obtained using analysis of variance.

Interlaboratorial programs are conducted for a number of purposes: to identify problems related to the calibration of instruments, to assess the degree of equivalence of analytical results among several laboratories, to attribute quantity values and its uncertainties in the development of a certified reference material and to verify the performance of laboratories as in proficiency testing, a key quality assurance technique, which is sometimes used in conjunction with accreditation. Several statistics tools are employed to assess the analytical results of laboratories participating in an intercomparison program. Among them are the z-score technique, the elypse of confidence and the Grubbs and Cochran test. This work presents the experience in coordinating an intercomparison exercise in order to determine Ca, Al, Fe, Ti and Mn, as impurities in samples of silicon metal of chemical grade prepared as a candidate for reference material.