Electron Probe Microanalysis of Variable Oxidation State Oxides: Protocol and Pitfalls

Electron probe microanalysis of geological oxide materials relies on stoichiometric considerations to estimate the content of undetermined oxygen and thus calculate ZAF (atomic number, absorption, fluorescence) matrix correction factors, requiring the valences of cations in the corresponding software to be unambiguously defined. However, stoichiometric ZAF corrections may be problematic in the presence of other undetermined elements or variable valence state cations. Herein, we analyse several oxides containing such cations, that is magnetite (Fe3O4), haematite (Fe2O3), hausmannite (Mn3O4) and cuprite (Cu2O). We compare data re‐calculated for incorrect valence states (Method 1) with reference values, revealing incorrect results, due to an incorrect amount of oxygen used in the matrix correction. Some solid‐solution series of haematite and magnetite were also modelled in CalcZAF program to prove the relative errors when the incorrect oxygen is used. To resolve these issues, we describe two accurate methods. Method 2 uses the true valence states of analysed elements. In Method 3, all cations are analysed as metals, with the content of undetermined oxygen determined by difference. As EPMA software does not allow the use of non‐integer valences, Method 3 is applicable to cations with non‐integer or dubious valences in cases where these non‐integer valences cannot be defined. Key Points Three EPMA matrix correction methods were compared to determine the protocol and pitfalls of variable oxidation‐state oxides. Defining true valence states for variable oxidation‐state cations is critical to obtain correct ZAF factors during EPMA. Oxygen by difference is applicable for cations with non‐integer or dubious valences.