Hello,
I'm calculating different slabs with a net dipole moment, and I've encountered some problems when using the dipole correction. I had problems converging the calculations and therefore started to first do a scf calculation without the dipole correction as suggested here (https://www.vasp.at/wiki/Electrostatic_corrections) before the actual calculations, which helped with convergence. However, for some calculations, it might converge faster if I directly use a dipole correction from the start. I'm interested in changes in the DOS and the LOCPOT between different slabs, and I realized that doing the calculations the two different ways does not give the same result.
In more detail:
version 1 of my calculations:
1. scf without a dipole correction
2. scf calculation with a dipole correction starting from the output of step 1
3. non-scf calculation with a dipole correction to obtain DOS
version 2:
1. scf calculation with a dipole correction
non-scf calculation with a dipole correction to obtain DOS
I use the same structure, dipole correction settings and so on. The only difference between the two calculations are: in version 1, I'm using ALGO=All for the first two steps, and then ALGO=Normal, and in version 2, I always use ALGO=Normal.
My questions are:
1. In version 1, I see a change in the dipole correction going from step 2 to step 3 (i.e. from the scf to the non-scf calculation), but there is hardly any change in version 2 (e.g. the correction is -0.861271 eA after step 2 and -0.921371 after step 3, whereas in version 2 it is -0.916304 eA after step 1 and -0.915168 eA after step 2). Should I expect to see a change of the dipole correction going from the scf to the non-scf calculation?
2. The overall dipole correction is not the same in both cases, and also if I look at e.g. the LOCPOT, I see that the remaining electric field in my material is not the same for the two cases. Is there a way to judge which result is more accurate?
Thank you in advance for your help.
Best,
Edith
