Energy calculation instability for hcp Co(0001) surface

[tak]

I had optimized 5-layer 2x2 hcp Co cell where top 3 layers were relaxed. After the successful optimization and energy calculation, I used the optimized geometry to re-optimize the geometry and calculate the energy again. Interestingly there was a difference between two as much as 0.49 eV. I repeated the calculation 2 more times, and found that the energy was nearly evenly distributed from the average +- 0.245eV. I also found that if I add adsorbate on the surface this instability did not occur. Attached below are INCAR and KPOINT. I used PW91 with USPP, but PBE-PAW also showd similar instability. Could anyone explain what might cause the energy instability on the clean surface? I greatly appreciate your help.

Thank you.



INCAR:
SYSTEM = Co x 20
ISTART = 0
ISMEAR = 1
SIGMA = 0.2
ISPIN = 2
ENCUT = 400
NBANDS = 290
IBRION = 2
NSW = 100
POTIM = 0.2
LREAL = Auto

KPOINTS:
Co 20
0
G
4 4 1
0 0 0

[forsdan]

Just to eliminate the obvious difference first. Is the magnetic structure for the slab the same at the end of the runs? If you restart with a new geometry the electronic relaxation might have converged to another magnetization, i.e. a local minimum, which would explain the energy difference. Adjust the MAGMOM tag to aid convergence to the lowest energy state in that case.

Cheers
Dan


<span class='smallblacktext'>[ Edited Fri Jun 03 2011, 12:58AM ]</span>

[tak]

Hi Dan,

Thank you very much for your advice, and it worked!! Our system was down for a while, and I could not run the calculation until yesterday, but now it's back. I increased MOGMON, and the energy is stable, and it was lowered more than 2eV comparing to the previous results.

Thank you.

Tak