Max-Planck-Institut für Eisenforschung, Düsseldorf, Germany
Towards high-throughput atomistic microstructure–mechanics simulations
High-throughput atomistic simulations are well established to explore the chemical composition space of materials as well as the structural space of individual defects like grain boundaries (GBs). Similarly, the response of individual defects like cracks or dislocations to mechanical loads can be studied by systematically varying temperature and applied stress [1,2]. However, the mechanical response of engineering materials is dominated by the interaction of a multitude of different defects and the competition of multiple deformation mechanisms. For example, to understand nanocrystal plasticity, one needs to take into account the adaptation of the entire GB network, GB glide and migration processes, and dislocation nucleation, interaction, pinning, and absorption processes [3,4]. This requires on the one hand a systematic variation of the GB character distribution, the GB network, and also of the GB interfaces, and on the other hand analysis methods that allow for statistical analysis and identification of spatiotemporal correlation of deformation events . Similarly, realistically modeling fracture requires the inclusion of existing dislocations, GBs, and precipitates. While the steady increase in the number of CPUs in HPC systems now enables farming large-scale simulations with different microstructures, the toolsets to generate and analyze such structures are still not widely accessible nor parallelized.
Here we present examples of how microstructures can be generated, tested under complex boundary conditions, and analyzed in a quantitative way [4,5,6]. By analyzing the pain points of current approaches, we want to pave the way toward a systematic high-throughput sampling of the microstructure space in atomistic simulations of mechanical properties.