Convergence problems with EFIELD and NELECT
Posted: Tue Jul 31, 2012 12:34 pm
I'm trying to change the charge distribution of the system I'm working with by means of an external field (EFIELD) or with some extra charge (NELECT).
With a test cell 3x3 everything was working fine with the charge, I managed to add 5 extra electrons (more than I needed, but I was testing) and no convergence problems appeared. In the case of the external field even with the 3x3 cell, for a field bigger than 0.75 eV/Ang (and what I need is bigger ones) I was having problems with the SFC convergence (the values of the difference were going down until they started to fluctuate between E+00 and E-03 but never reaching the convergence criteria).
My actual system cell is a 6x6 (very similar to the 3x3 case), and as the case of putting extra charge (NELECT) was working fine for the 3x3, I started adding 1 extra electron to the 6x6. The system was not converging and I got a warning because it seemed that the default number of bands considered (240 for my system) was low, so I repeated the calculation with NBANDS = 300 hoping that it would help. The warning kept appearing so I increase NBANDS to 400. That time I still wasn't reaching the SFC convergence but the message disappeared. Then I noticed in the OUTCAR that the highest ocuppied band in all the kpoints was the number 198... so whay is it not enough with the default? In this case the OSZICAR shows a much more irregular behaviour (fluctuating all the time between huge values: E+04 to E+07). I don't understand why in the 3x3 I didn't have to make this corrections and if the convergence problems are due to this.
In both cases (with EFIELD and NELECT) I'm doing a preconvergence of the charge as pointed in the manual:
It is recommended to use this mode only after pre-converging the orbitals without the LDIPOL flag, and the center of charge should be set in the INCAR file (DIPOL= center of mass). The user must also ensure that the cell is sufficiently large to determine the dipole moment with sufficient accuracy.
And also using some convergence parameters for hard convergence systems:
I will be glad to provide more details of the calculation if needed. Any suggestion to fix the convergence problems in one case or the other are very welcome!!
Thanks a lot!!
<span class='smallblacktext'>[ Edited ]</span>
With a test cell 3x3 everything was working fine with the charge, I managed to add 5 extra electrons (more than I needed, but I was testing) and no convergence problems appeared. In the case of the external field even with the 3x3 cell, for a field bigger than 0.75 eV/Ang (and what I need is bigger ones) I was having problems with the SFC convergence (the values of the difference were going down until they started to fluctuate between E+00 and E-03 but never reaching the convergence criteria).
My actual system cell is a 6x6 (very similar to the 3x3 case), and as the case of putting extra charge (NELECT) was working fine for the 3x3, I started adding 1 extra electron to the 6x6. The system was not converging and I got a warning because it seemed that the default number of bands considered (240 for my system) was low, so I repeated the calculation with NBANDS = 300 hoping that it would help. The warning kept appearing so I increase NBANDS to 400. That time I still wasn't reaching the SFC convergence but the message disappeared. Then I noticed in the OUTCAR that the highest ocuppied band in all the kpoints was the number 198... so whay is it not enough with the default? In this case the OSZICAR shows a much more irregular behaviour (fluctuating all the time between huge values: E+04 to E+07). I don't understand why in the 3x3 I didn't have to make this corrections and if the convergence problems are due to this.
In both cases (with EFIELD and NELECT) I'm doing a preconvergence of the charge as pointed in the manual:
It is recommended to use this mode only after pre-converging the orbitals without the LDIPOL flag, and the center of charge should be set in the INCAR file (DIPOL= center of mass). The user must also ensure that the cell is sufficiently large to determine the dipole moment with sufficient accuracy.
And also using some convergence parameters for hard convergence systems:
Code: Select all
IMIX=1
AMIX=0.05
BMIX=0.00001
NELM=350Thanks a lot!!
<span class='smallblacktext'>[ Edited ]</span>