Thermal and mechanical properties of high-entropy transition-metal oxides

In this work, we used the neuro-evolution potential (NEP), which is an interatomic potential constructed using machine learning, and which runs on GPUs, to investigate thermophysical properties of the 5-component MgCoNiCuZnO5 compound. Data set were constructed for the 5 binary oxides as well as ternary, quaternary and quinary oxides in a supercell in which atoms were randomly displaced and forces and total energies recorded. Using this potential, we first investigated atomic distortions which are due to different atomic sizes of the transition metal (TM) elements. It was found that static distortions were on the order of 0.11 Ang for TMs and 0.18 Ang for O, while dynamic ones, due to thermal fluctuations in the atomic displacements, increased with temperature and reached 0.21 Ang for TMs and 0.27 Ang for O at a temperature of 800K. Elastic properties such as bulk modulus and yield strength showed a decrease as temperature T was increased. Finally, we will report on the thermal conductivity of this material as a function of T, and the possibility of tuning the distortions and thus its thermal conductivity by changing the stoichiometry of the compound.