Charge defects in bulk

Message

abu · #1 Post by **abu** » Fri Jun 22, 2007 8:10 am

Hi there
I am trying to calculate defect in TiO2. I have a reasonable big supercell (192 atoms). I want to calculate point defect in bulk (putting one more Ti atom in interstitial site). With putting one more Ti atom, the system is converged nicely.
Now, I want to put Ti4+ ion instead of Ti atom. I add following extra tag in the INCAR file.
NELECT = 1024 (1028 for neutral system)
IDIPOL = 4 (I tried with 3 and did not work)
LDIPOL = .TRUE.
I am using vasp 4.6.20 version.
I did not find any converge structure, in 2nd/3rd structural iteration, I found a weird structure. i.e., some atoms come too close together and some move far away from others.
Could you please give me an idea what wrong I am doing and what should I try.
Thankx

#2 Post by **admin** » Fri Jun 22, 2007 12:35 pm

charge cannot be assigned to a specific atom in all PW and mixed-basis set calculations, by setting NELECT, the cell is
charged.

abu · #3 Post by **abu** » Fri Jun 22, 2007 3:56 pm

Yes, if we consider the cell is charged, is there any suitable way to structure minimization. As I said, setting the above criteria gives me some weird structure. Without multi-pole correction (deleting IDIPOL and LDIPOL tag from INCAR file), the structure is converged. However, the energy is around 20 eV higher than neutral cell.
<span class='smallblacktext'>[ Edited Sat Jun 23 2007, 02:31PM ]</span>

Alenushka · #4 Post by **Alenushka** » Tue Jun 26, 2007 12:57 pm

From my experience, if you keep IDIPOL, then it will be converged, I have problems only if I switch the LDIPOL on.
Here: http://cms.mpi.univie.ac.at/vasp/vasp/n ... car-dipol2

LDIPOL is needed only if you like to switch the potential coorection mode, IDIPOL is enough to have corrections:
"Both corrections (quadrupole only for charged systems) will be calculated and added to the total energy if the IDIPOL flag is set."