As a preliminary step of my calculations I tried to optimize the geometry of bulk MoS2 starting from the cell in the database.
With LCAO and PBE (see attached script) I obtain an interlayer spacing larger than the experimental one (lattice constant c=12.86 A), as expected, since Van der Waals interactions are not correctly reproduced. Thus, I tried to add either Grimme D2 or D3 corrections with default parameter values (the scripts are also attached). While the inclusion of D2 correction actually slightly reduces the interlayer spacing (c=12.26 A), the D3 correction seems to induce a repulsive force. The trajectory file from the optimization shows a steady increase of the interlayer spacing: I stopped the geometry optimization after several hours and 50 optimization steps with c=14.6 A. I tried to change the numerical accuracy parameters and the basis, but without luck: any attempt with D3 correction gives rise to an always-expanding optimization.
Am I doing anything wrong or the D3 correction is not supposed to work in this case?
Thank you in advance for any suggestion.
Luca Donetti
