Thanks to new supercomputer architectures, classical molecular dynamics (MD) simulations will soon reach the scale of a trillion atoms. These unprecedentedly large simulation systems will thus be capable of representing metal plasticity at the micron scale. Such simulations generate an enormous amount of data, and the challenge now lies in processing them to extract statistically relevant features for the mesoscale plasticity models (continuous-scale models).

The evolution of a material is complex as it depends on extended crystal defect lines (dislocations), whose dynamics are governed by numerous mechanisms. To feed higher-scale models, the key quantities to extract are the velocities and lengths of dislocations, as well as their evolution over time. These data can be extracted using specific post-processing techniques based on local environment characterization ('distortion score' [Goryaeva_2020], 'local deformation' [Lafourcade_2018], ‘DXA’ [Stukowski_2012]). However, these methods remain computationally expensive and do not allow for in situ processing.

We have recently developed a robust method for real-time identification of crystalline structures [Lafourcade_2023], which will soon be extended to dislocation classification. The objective of this postdoctoral project is to develop a complete analysis pipeline leading to the in situ identification of dislocations in atomic-scale simulations and their extraction in a nodal representation.