Screening of new Entropy Stabilized Oxides by DFT calculations and machine learning
Résumé
The entropy stabilized oxides (ESO) are the MOx oxides where the metallic site M is a mixing of at least 5 elements with a composition between 5 and 35%. At high temperatures, these materials are stabilized by configuration entropy. The ESO may have some particular properties, like a high dielectric constant and can be used as catalysts, supra-conductors and electrolytes in Li-batteries. Considering a base of 16 cationic elements with a 1% step variation in composition, this leads to 2.4 billion of possible ESO. Our work aims to predict through DFT calculation and machine learning which ESO could be meta-stable and at what temperature they stabilize. We use the Special Quasi-random Structure (SQS) to model the mixing. Then, some DFT calculations are used to calculate the formation enthalpy of oxides. Thanks to a convex hull code that we have developed, we are able to determine which phases are the most stable. The entropy, considered only as configuration entropy is added with the equation : -R∑_i▒〖x_i ln(x_i)〗. With the 2.4 billion of possible ESO, we use machine learning to predict the formation enthalpy instead of calculating it by DFT. Based on 16 simple oxides, 120 binary mixings et 219 quinary mixings, we can predict the formation enthalpy with an RMSE of 5,12 kJ/mol.