Speaker
Description
Local chemical composition of grain boundaries (GB) usually differs from that of bulk crystal. Region, where two or more grains are connected often comes with an extra space that usually serves as a location where various impurities tend to segregate. Effect of segregated impurities has been extensively studied using ab initio calculations and molecular dynamics simulations. While the former approach imposes limitation on size of the computational cell, the latter approach brings limited accuracy due to less reliable interatomic potentials. The most promising way, how to overcome these limitations, is the molecular dynamics with interatomic potentials generated during the on-the-fly machine learning (ML) within ab initio simulations based on the density functional theory (DFT). Reliability of this approach is tested on the example of Tin segregation at selected grain boundaries in bcc iron. Supercells of different size can be used to estimate the effect of concentration on the segregation energy.
To generate reliable description of atomic interactions, we used on-the-fly machine learning as has been implemented in the recent version of the Vienna Ab initio Simulation Package. Using these interactions, one can calculate segregation energy for number of possible segregation sites to determine the preferred one.
