Speaker
Description
Understanding segregation phenomena is a critical aspect of integrated computational materials design. In this connection, the segregation energies plays a central role and large databases are being created to get a comprehensive overview over materials. With the availability of such databases, machine learning approaches can be used to learn the trends in the periodic table and get segregation energies even for alloys for which no data exist at present.
We present an investigation on machine learning segregation energies obtained from atomistic calculations. We will discuss the critical role of feature engineering. We analyze how different approaches based on e.g. Steinhardt parameters or bond order potentials perform in this respect. Furthermore, we show results for a variety of metallic alloys focusing on the class of transition metals. With end by discussing the challenges of machine learning approaches for segregation energies and grain boundary engineering in general.
| Speaker Country | Austria |
|---|