Atomistic simulations

We perform simulations with the general utility lattice program (GULP) (Gale 1997), where Coulomb interactions are computed using the Ewald sum method; and with the large-scale atomic/molecular massively parallel simulator (LAMMPS) (Plimpton 1995), where Coulomb interactions are computed with the particle–particle particle–mesh (pppm) method (Eastwood et al. 1980). For calculations based on the THB1 potential, we use the custom version of LAMMPS modified by Mahendran (Mahendran et al. 2017), where the Coulomb interaction is computed with the Wolf summation method (Wolf 1995). Preliminary tests allowed us to verify that the different codes and methods yield the same lattice constants and energies.

Pressure is imposed by rescaling the simulation box and monitoring the stress tensor as computed by LAMMPS. The cell angles are constrained to 90∘, and the geometry and atom positions are optimized several times until the target pressure is reached and all forces are smaller than 10-9 eV.Å-1. Supercells containing lattice defects are constructed with Atomsk (Hirel 2015), and atomic structures are visualized with VESTA (Momma and Izumi 2011).