DFT+U+SOC protocol?
Posted: Wed Nov 30, 2011 11:52 pm
Dear VASPians,
I am calculating some magnetic semiconductors (mostly 3/4d)
with DFT+U+SOC calculations according to couple of different recipes,
and the results are not so consistent...
My recipes are:
1. DFT+U+SOC all in one time (both LSORBIT and LDAU on)
2. DFT+U+SOC w/ Gamma only -> DFT+U+SOC w/ full K-points with CHGCAR from previous run (ICHARG = 1)
3-1. DFT+U -> DFT+U+SOC with CHGCAR from DFT+U (ICHARG = 1)
3-2. The same with 3-1 but ICHARG = 11
4-1. DFT+SOC -> DFT+U+SOC with CHGCAR from DFT+SOC (ICHARG = 1)
4-2 The same with 4-1 but ICHARG = 11
Which one would be most reliable?
Or is there any sort of protocol in VASP community?
I usually take the advantage of 4-1, because breaking
the symmetry before applying on-site repulsion seems more reasonable.
But sometimes the total energy is lower than that of 4-2,
and sometimes higher.
Another question:
During the calculations, I often encounter
"WARNING: grid for Broyden might be to small"
With adjusting mixing parameters, I resolve(?) this WARNINGS.
The results are, however, quite often different
with different mixing parameters combinations
(AMIX, AMIX_MAG, BMIX, BMIX_MAG, and sometimes AMIN).
Why do I have this WARNING, and how can I resolve this
(other than playing with mixing parameters)?
Or, what is the most reliable protocol for choosing mixing parameters
for DFT+U+SOC calculations of magnetic semiconductors?
Many thanks in advance :-)
<span class='smallblacktext'>[ Edited ]</span>
I am calculating some magnetic semiconductors (mostly 3/4d)
with DFT+U+SOC calculations according to couple of different recipes,
and the results are not so consistent...
My recipes are:
1. DFT+U+SOC all in one time (both LSORBIT and LDAU on)
2. DFT+U+SOC w/ Gamma only -> DFT+U+SOC w/ full K-points with CHGCAR from previous run (ICHARG = 1)
3-1. DFT+U -> DFT+U+SOC with CHGCAR from DFT+U (ICHARG = 1)
3-2. The same with 3-1 but ICHARG = 11
4-1. DFT+SOC -> DFT+U+SOC with CHGCAR from DFT+SOC (ICHARG = 1)
4-2 The same with 4-1 but ICHARG = 11
Which one would be most reliable?
Or is there any sort of protocol in VASP community?
I usually take the advantage of 4-1, because breaking
the symmetry before applying on-site repulsion seems more reasonable.
But sometimes the total energy is lower than that of 4-2,
and sometimes higher.
Another question:
During the calculations, I often encounter
"WARNING: grid for Broyden might be to small"
With adjusting mixing parameters, I resolve(?) this WARNINGS.
The results are, however, quite often different
with different mixing parameters combinations
(AMIX, AMIX_MAG, BMIX, BMIX_MAG, and sometimes AMIN).
Why do I have this WARNING, and how can I resolve this
(other than playing with mixing parameters)?
Or, what is the most reliable protocol for choosing mixing parameters
for DFT+U+SOC calculations of magnetic semiconductors?
Many thanks in advance :-)
<span class='smallblacktext'>[ Edited ]</span>