An accurate determination of finite temperature free energies is the basis for addressing a wide range of topics in materials science. Examples are the determination of phase diagrams, reaction paths and kinetic barriers by sampling free energy surfaces. For such calculations the determination of free energy differences or of derivatives on the free energy surface requires an accuracy in the statistical averages of well below 1 meV. While such accuracy can be easily obtained using empirical potentials, it becomes numerically prohibitive when using fully parameter-free ab initio techniques. In the talk various coarse-graining schemes will be presented that allow a numerically efficient yet highly accurate sampling of the high-dimensional configuration space. Using this approach we demonstrate its performance in describing all relevant free energy excitation mechanisms up to the melting temperature and its power in the design of novel high-strength steels. |
![]() |