Speaker
Description
Understanding the atomic-scale plasticity of bcc transition metal tungsten is crucial for many industrial applications, including components of nuclear fusion reactors that face extreme levels of irradiation. The modeling of dislocations as well as irradiation induced defects, however, remains extremely challenging as it requires simulations of large atomic ensembles with DFT accuracy. In this work, we present a new interatomic potential based on the Atomic Cluster Expansion (ACE) [1] that is able to reproduce properties of extended crystal defects in tungsten with unprecedented accuracy and computational efficiency, exceeding those of the recent machine-learning potentials.
ACE is a novel approach that uses the local atomic environment as a descriptor for a complete and efficient representation of atomic properties. The effort for evaluating ACE has been shown to scale linearly with the number of neighbors, irrespective of the order of the expansion. The current work will describe the construction of the ACE potential for tungsten, from the specificities of the descriptor and database to the particular fitting procedure that is based on forces rather than energies. We will demonstrate that the new ACE model is able to simulate a broad variety of crystal defects in a remarkable agreement with DFT results. Particular emphasis will be given to the analysis of stresses that are mandatory to reproduce correctly the mobility of point defects, dislocations, and their complexes. In addition, for dislocations the relaxation volume tensor, calculated from the stress tensor variation along the minimum energy migration path, is directly linked to the Non-Schmid behaviour of the material.
[1] R. Drautz (2019) PRB 102, 024104
| Speaker Country | Germany |
|---|