EUROPEAN CONGRESS AND EXHIBITION ON ADVANCED MATERIALS AND PROCESSES - EUROMAT 2021

Ab initio databases such as the NOMAD repository or the Materials Project have become a useful and quickly available tool in materials design. However, all these databases consist primarily of T=0K energies, whereas typical experimental conditions for materials design correspond to finite temperatures for which free energies are the relevant quantities. An accurate description of free energies necessitates the inclusion of various excitation mechanisms, related to fully anharmonic atomic vibrations, electrons, magnetic moments, and coupling effects. For example, it has been shown that anharmonicity, i.e., the explicit interaction of phonons with each other, can have a strong impact on thermodynamic properties.
In this presentation, our group's efforts in developing a highly accurate ab initio free energy database will be introduced. We have a long-standing expertise in computing free energies including the various excitation mechanisms mentioned above. We have developed several methods to expedite the otherwise computationally very demanding simulations. Most recently, the Two-Stage Upsampled Thermodynamic Integration using Langevin Dynamics method coupled with machine learning potentials (Moment Tensor Potentials) has enabled highly accurate free energy calculations for chemically complex multicomponent alloys. Other developments allow us, for example, to calculate ab initio free energies of liquids and to extract melting properties. An overview of the various methods and applications under the umbrella of the database will be given.