For this project we used the code PARCAS by Kai Nordlund which was modified to meet our requirements as described previously. We are not authorized to publish all the parts of the code that we had to change. Our part can be downloaded here.
We encountered difficulties when we used the Nosé-Hoover thermostat and pressure control with systems where the atoms are far apart. If the distance between the particles is bigger than the cutoff radius of the potentials the forces will be necessarily zero. This causes problems in the Verlet integrator since the propagation of the positions - as it can be seen from eq.(4) - is determined exclusively by the accelerations, i.e. the forces, and independent of the velocities. Within the time constraints of this project this problem had to remain unsolved. The simulation of homogeneous nucleation was therefore performed entirely with the Gear predictor-corrector algorithm and the Berendsen thermostat.
We used a cubic box with side lengths of 70.0Åcontaining 343 Ar and 343 Cu atoms. Periodic boundary conditions have been applied in all directions. In the beginning the atoms were arranged on a bcc lattice, i.e. the box was subdivided into 7x7x7 unit cells each holding two atoms, as can be seen on figure fig. 4. The initial velocities were set according to the Maxwell-Boltzmann distribution at 300K.
The positions and its derivatives were integrated with a sixth order Gear predictor-corrector algorithm for approximately 10ns of real time. The Berendsen thermostat with a coupling constant of 10fs was applied to the Ar atoms only to hold the temperature about 300K.
|