Ab initio and classical atomistic modelling of structure and defects in crystalline orthorhombic polyethylene : Twin boundaries, slip interfaces, and nature of barriers

We study the stability of twin boundaries and slip in crystalline orthorhombic polyethylene by means of density functional theory (DFT), using a nonempirical, truly nonlocal density function, and by means of classical molecular dynamics (MD). The results show that, in accordance with experimental observations, there is a clear preference to chain slip over transverse slip for all considered slip planes. The activation energy for pure chain slip lies in the range 10–20 mJ/m2 while that for transverse slip corresponds to 40–280 mJ/m2. For the (11¯0)-slip plane the energy landscape is non-convex with multiple potential energy minima, indicating the presence of stable stacking faults. This suggests that dissociation of perfect dislocations into partials may occur. For the two low-energy twin boundaries considered in this work, {110} and {310}, we find that the former is more stable than the latter, with ground state energies corresponding to 8.9 and 28 mJ/m2, respectively. We have also evaluated how well the empirical MD simulations with the all-atom optimized potential for liquid MD simulations (OPLS-AA) and the coarse-grained united atom (UA) potential concur with the DFT results. It is found that an all-atom potential is necessary to partially capture the γ-surface energy landscapes obtained from the DFT calculations. The OPLS-AA predicts chain slip activation energies comparable with DFT data, while the transverse slip energy thresholds are low in comparison, which is attributed to weak close ranged monomer repulsion. Finally, we find that the H-H interaction dominates the slip activation. While not explicitly represented in the UA potential, its key role is revealed by correlating the DFT energy landscape with changes in the electron distributions and by MD simulations in which components of the OPLS-AA intermolecular potential are selectively silenced. © 2017 Elsevier Ltd