Dear Prof. Marks, Prof. Blaha, dear All,
Thank you for the rapid answer! I appreciate!
I think this resolves the issue. For practical reasons I will just use
+U only for 4f levels.
I agree with Prof. Marks who has reservations regarding the GGA+U for
4f. However, for practical reasons of checking the trends of band
dispersions in ARPES it seems the only quick solution. It's great that
it is included despite all the deficiencies!
Best,
Lukasz
On 7/24/2020 6:34 PM, Peter Blaha wrote:
I do NOT officially support 2 l values / atom for GGA+U. It may/may not
work in various programs.
The reason is simply, that in WIEN2k the U is applied only inside the
sphere and the V_orb potential is calculated from the occupations of an
orbital, varying from 0 to 1.
I.e. if n=1 (an orbital is fully occupied), it gets a potential which
shifts this state by U/2 down, on the other hand, if a particular
orbital is empty, it will be shifted up by U/2.
The problems come with relatively "delocalized" orbitals like the 5d
states of a 4f element.
Check out the "fractional charge" within the sphere as printed in
case.outputst.
4F* -0.590624 -0.590624 2.50 2.50 0.9844 F
5D* -0.208649 -0.208649 0.50 0.50 0.4864 F
You can find that Sm-4f is nearly completely confined in the sphere
(0.9844) and applicationsof U is "well defined", but one 5d electron
has less than 0.5 e inside the sphere !!!
Depending on the rare earth and on RMT it means that even a fully
occupied 5d state does NOT get shifted down due to GGA+U, but basically
all d-orbitals will be shifted up.
Am 24.07.2020 um 18:03 schrieb pluto:
Dear All,
I am trying to calculate a simple 4f compound with GGA+U,
ferromagnetic exchange (FM) and SOC. FM calculation without SOC using
the literature U settings looks reasonable, I am getting a gapped band
structure similar to anyone else.
But running SCF with GGA+U and SOC removes the gap, it also seems to
change the exchange splitting of 4f, which basically tells me that U
on 4f is not in effect anymore. Actually the problem happens with this
inorb file, when natorb=1 and nlorb=2:
1 1 0 nmod, natorb, ipr
PRATT 1.0 BROYD/PRATT, mixing
1 2 2 3 iatom nlorb, lorb
1 nsic 0..AMF, 1..SIC, 2..HFM
0.25 0.00 U J U on d levels
0.42 0.00 U J U on f levels
When nlorb is reduced to 1, and U is used only for 4f levels things
start to look "normal" also with SOC. So, it seems SOC cannot handle
nlorb>1 (perhaps just a bug).
I am using couple of years old version of WIEN2k, perhaps this has
been fixed in the meantime. Otherwise, any comment would be
appreciated!
Best,
Lukasz
_______________________________________________
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
