Ab initio and classical atomistic thermodynamics, phase stability and plasticity of crystalline and semi-crystalline polyethylene

Dnr:

SNIC 2017/1-463

Type:

SNAC Medium

Principal Investigator:

Pär Olsson

Affiliation:

Malmö universitet

Start Date:

2017-11-01

End Date:

2018-11-01

Primary Classification:

10304: Condensed Matter Physics

Webpage:

- Kebnekaise at HPC2N: 80 x 1000 core-h/month
- Tetralith at NSC: 48 x 1000 core-h/month

The purpose of this project is to investigate (i) thermodynamic properties, (ii) phase stability and (iii) plasticity mechanisms in semi-crystalline polyethylene (PE). This project is performed in collaboration with Tetra Pak.
For (i) and (ii) we will use quantum mechanical modelling (using quantum espresso and VASP) to elucidate phase stability and extract thermodynamic properties of the crystalline part of PE (i.e. heat capacity, thermal expansion, Gibbs free energy etc.). To this end we will utilize the quasi-harmonic approximation in conjunction with the newly established vdW-cx exchange-correlation functional (developed by Per Hyldgaard et al.), which requires a substantial amount of phonon calculations of large supercells, which are very computationally demanding. Two morphoogies will be considered, orthorhombic and monoclinic phases. We intend to perform the phonon calculations on the Triolith system, whose local node setup is beneficial to reduce the file overhead, which speeds up the phonon calculations. We estimate the need of about 60 000 core-hours per month over a year for this effort at the triolith system.
For the plasticity modelling we will use both DFT and classical atomistic modelling. Typically we intend to investigate twinning and martensitic transformation by means of DFT modelling. This effort requires the modelling of transition states for large supercells (e.g. through the utilization of the Nudged elastic band method). To this end we will use the Kebnekaise system. A typical job would require about 56-84 nodes.
We will also use LAMMPS to investigate multiaxial deformation mechanisms in the crystalline and amorphous parts. The slow strain rate combined with utilization of numerically expensive all-atom force fields require large paralellization on multiple (up to 112) cores for about one week per simulation on Kebnekaise.
In total, we estimate the mothly use to 80 000 core hours per month on Kebnekaise.