Assessment of atomistic data for predicting the phase diagram and defect thermodynamics. The example of non-stoichiometric uranium dioxide

Defect thermodynamic models are increasingly used to describe the evolution of materials microstructure. Unfortunately, experimental data for these models are generally scarce and difficult to obtain, which has given rise to a growing trend toward using atomic scale calculations to complement experimental data. This paper is a review concerning the way atomistic techniques can help building defect models in crystals and how efficient they are in reproducing important experimental data, such as the O M ratio and phase diagrams. The issue is addressed through the example of non-stoichiometric uranium dioxide. In this article, eleven point-defect models from the literature (defect formation energies-entropies) are presented and compared on the same footing; four of them are fitted to experimental data, while seven are obtained through atomistic calculations. This allows to compare all the models on the same basis both among themselves and with a large set of experimental data of various physical quantities, including the phase diagram near stoichiometric uranium dioxide. The defect formation energies and entropies are very different from one model to another. While the fitted models usually correctly reproduce the data sets according to which they were fitted, only two atomistic based models correctly reproduce the O M ratio diagrams, provided the oxygen molecule energy is correspondingly fitted. No model simultaneously reproduces the measured conductivity and O M ratio as functions of the oxygen potential. The difficulties of the atomistic-based models in predicting this ratio and the oxygen potential probably arise, among others, from an erroneous calculation of energy of the oxygen molecule and of the oxygen incorporation in U O 2 and also from a poor evaluation of the electron-hole formation Gibbs energy. The difficulty of obtaining reliable experimental data close to the stoichiometry might also contribute to the limited agreement between calculations and measurements, which is reason enough to reassess the behavior of the material in this stoichiometry region comprehensively, with a particular focus on the influence of uranium vacancies.