Multiscale simulation of high-temperature thermodynamic and mechanical properties of materials
||Multiscale simulation of high-temperature thermodynamic and mechanical properties of materials|
||SNIC Medium Compute|
||Ferenc Tasnádi <firstname.lastname@example.org>|
||2021-06-01 – 2022-06-01|
||10304 21001 |
The goal of the project is to provide computational resources to the FunMat-II competence center. The scientific goal is to establish a platform of high accuracy ab-initio molecular dynamics simulations combined with active machine learning of force fields to investigate high temperature thermodynamic and mechanical properties of materials. We primarily focus on refractory nitride materials (TiN, HfN, NbN,... and their alloys) and include high-energy density materials (materials with polymeric nitrogens). Our approach is to use ab-initio molecular dynamics simulations and static density functonal theory calculations (VASP, QE) and map the results to a particular force field (interatomic potential, moment tensor potential). The created force field is utilized in large scale classical simulations (thousands of atoms), such as dislocation motion, etc. The challenge is to develop a computational framework that extends the transferability of the chemical accurate interatomic potential to mesoscopic scales. Our approach is the recently developed technique of machine learning the interatomic potentials (MLIP, https://mlip.skoltech.ru/).
We will calculate i) high-temperature phase diagrams and elasticity of hard materials, ii) capture rare diffusion effects, iii) predict thermal conductivity via phonons or iv) develop a simulation framework to model plasticity, dislocations. A neural network based approach will be utilized to explore properties of alloys. Through the FunMat-II competence center we are in close collaboration with Seco Tools and Sandik Coromant. This industrial segment has major interest in high temperature elastic and mechanical properties of nitride alloys. We also collaborate world-leading experimentalists in the filed of high-pressure physics. Our research is supported by the Swedish strategic FunMat-II consortium and the Interdisciplinary Laboratory for Advanced Functional Materials (AFM) at Linköping University. The ab-initio calculations will be performed using VASP and Quantum Espresso (QE). The classical molecular dynamics simulations will be done using LAMMPS. The teams has experience in using VASP on GPUs and wants to port the large part of the simulations to GPUs or use a hybrid scheme.