orbital polarization calculation

[Long Zhang]

Dear ELK users and developers,

Is there a possible way to do the "orbital polarization" calculation of PRB 63, 235114 (2001) using ELK?
Wien2k has it listed as an option of the orbitally dependent potential V_orb as function of (l,m).
I notice we actually have the save variables (vmatmt, vmatmti, vmatmto from moddftu.f90), so I am wondering if one can do the same calculation in the dft+u mode with ELK.

Thanks,
Long

[Lars Nordström]

Dear Long!

I am not 100% sure what you are aiming at, but the paper you refer to discuss an implementation in Wien2k that is corresponding to the DFT+U implemented in ELK, with spherical invariance etc …
(but if I recall correct the Wien2k implementation still uses a spherically non-invariant basis, in contrast to ELK.)

The orbital polarization mentioned in Wien2k is probably another, pre-DFT+U, approach which is simpler but less correct, by Brooks et al.
I compare these two approaches in a paper using ELK, PRB 78, 100404(R) (2008).

Best wishes/ Lars

[Long Zhang]

Dear Prof. Nordström,

Thank you for pointing me to the right place!
At the moment my goal is to calculate magnetic properties of 4f compounds.
I was suggested to exploit the fixed orbital polarization calculation,
like, for example, the gga+soc+op calculation in
https://iopscience.iop.org/article/10.1088/0953-8984/26/41/416001
where the authors used LMTO but I want to do the same with FP-LAPW.

I had a quick read of the 2008 PRB. It’s very insightful. I think what I plan to do is completely contained in the tensor moment decomposition analysis.
I am still in reading the follow up paper PRB 80, 035121 (2009) that's exactly corresponding to the current implementation, and at the same time trying to understand the provided example.

Best Regards,
Long

[Long Zhang]

Dear all,

I have a quick question about the fixed tensor moment (FTM) calculation.
Does it always need to be carried out with dft+u with a finite value of U? or can I use it in a magnetic state calculation with SOC but without U?
I am asking because the provided examples are all with a finite U. I am not sure if that's a must for FTM calculations.

Thanks,
Long

[Lars Nordström]

Yes Long, you are right in principle that makes sense. As it is implemented now you need DFT+U in order to calculate tensor moments, but you can set U to zero ...

Depending on your choice of input (inptftu) it would correspond to:
setting both U and J to zero (1)
all l+1 number of parameters to zero (2 or 3)
or U to zero (5).
The first version (1) would still give a small DFT+U effect from higher Slater parameters ...
I would recommend to always use version 5, but since no other code has it implemented some referees might be confused ... but in this case it is definitely simplest.
best
/Lars

[Long Zhang]

Dear Prof. Nordström,
Thank you very much to clarify this.
Best,
Long