Origin-dependent finite electric field results
Posted: Thu Jul 21, 2016 8:49 am
I am trying to do finite electric field calculations on a system with multi-atom defects. I am using PAW_PBE and LDAU=.TRUE., with LASPH = .TRUE. For the finite field calculations I apply a field in the Z direction (tetragonal cell) with e.g. EFIELD_PEAD = 0.0 0.0 0.0005
My problem is that the results depend on the geometric origin. If the defect is located in the centre of the cell I get a particular total energy, which is independent of the polarity of the applied field. However, if I do a displacement of all of the ions along Z the energy changes by about 10 meV, and the sign of the change depends on the polarity of the field even though the defect is symmetric in that direction. Applying the opposite polarity field results in a change of the energy by ~20 meV.
The change appears at the start of the third phase of the PEAD calculation, when the actual field is switched on, e.g.:
DMP: 4 -0.675991128405E+03 -0.26193E-08 -0.34084E-08 9632 0.772E-08 0.210E-08
DMP: 5 -0.675991128407E+03 -0.15571E-08 -0.98430E-09 9632 0.249E-08-0.801E-10
DMP: 6 -0.675991128407E+03 -0.56752E-09 -0.58434E-09 9632 0.130E-08 0.419E-09
N E dE d eps ncg rms rms(c)
SDA: 1 -0.675997231284E+03 -0.61029E-02 -0.23834E-05 9632 0.596E-05 0.000E+00
DMP: 2 -0.675997233165E+03 -0.18802E-05 -0.17977E-05 9632 0.271E-05 0.344E-05
DMP: 3 -0.675997234521E+03 -0.13568E-05 -0.10493E-05 9632 0.143E-05 0.229E-05
DMP: 4 -0.675997235358E+03 -0.83621E-06 -0.77002E-06 9632 0.112E-05 0.156E-05
This behaviour is independent of the KPOINT sampling and the symmetrisation (ISYM=0 does not change it).
If I do similar calculations with fields aligned in the Y or X+Y directions, no polarity dependence is observed.
What is going on here?
(One point that may be relevant is that the total energy of this particular defect structure, relative to other structures of the same defect composition, is strongly dependent on the Hubbard parameters. That is, with four structures for the defect, the energies of three shift by a constant on a change in LDAUU, but this more symmetric defect changes in energy by a substantially different amount.)
My problem is that the results depend on the geometric origin. If the defect is located in the centre of the cell I get a particular total energy, which is independent of the polarity of the applied field. However, if I do a displacement of all of the ions along Z the energy changes by about 10 meV, and the sign of the change depends on the polarity of the field even though the defect is symmetric in that direction. Applying the opposite polarity field results in a change of the energy by ~20 meV.
The change appears at the start of the third phase of the PEAD calculation, when the actual field is switched on, e.g.:
DMP: 4 -0.675991128405E+03 -0.26193E-08 -0.34084E-08 9632 0.772E-08 0.210E-08
DMP: 5 -0.675991128407E+03 -0.15571E-08 -0.98430E-09 9632 0.249E-08-0.801E-10
DMP: 6 -0.675991128407E+03 -0.56752E-09 -0.58434E-09 9632 0.130E-08 0.419E-09
N E dE d eps ncg rms rms(c)
SDA: 1 -0.675997231284E+03 -0.61029E-02 -0.23834E-05 9632 0.596E-05 0.000E+00
DMP: 2 -0.675997233165E+03 -0.18802E-05 -0.17977E-05 9632 0.271E-05 0.344E-05
DMP: 3 -0.675997234521E+03 -0.13568E-05 -0.10493E-05 9632 0.143E-05 0.229E-05
DMP: 4 -0.675997235358E+03 -0.83621E-06 -0.77002E-06 9632 0.112E-05 0.156E-05
This behaviour is independent of the KPOINT sampling and the symmetrisation (ISYM=0 does not change it).
If I do similar calculations with fields aligned in the Y or X+Y directions, no polarity dependence is observed.
What is going on here?
(One point that may be relevant is that the total energy of this particular defect structure, relative to other structures of the same defect composition, is strongly dependent on the Hubbard parameters. That is, with four structures for the defect, the energies of three shift by a constant on a change in LDAUU, but this more symmetric defect changes in energy by a substantially different amount.)