Speakers
Description
Shape memory alloys represent a material with unique properties like shape memory effect and superelasticity that are based on diffusionless transformation between high temperature austenite (cubic) and low temperature martensite (monoclinic). Recently introduced Kwinking mechanism shed light into plastic behaviour of martensite, nevertheless, to get complete overview of this effect it is necessary to employ atomistic simulations. However, the usage of this approach is still very limited as first principles calculations are limited by the simulations cell size and 0K temperature and molecular dynamics by the precision of the interatomic potentials. To overcome these limitations we are developing the structural NiTi training data set for AI based MD potentials capable of high precision molecular dynamics simulations. These training data are produced with the help of quasi static first principles simulations. The MD potentials are then constructed via additional software packages (RuNNer or Pacemaker codes) that are based on neural networks or atomic cluster expansions. With the first version of potentials we are capable to reproduce basic behaviour of martensite. If possible then these predictions are compared with first principles data to check MD potential predictions.
