I sincerely hope you are not Italian, then you do not deserve any answer;-(
Otherwise, have you used GGA or LDA in your calculations? GGA has problems with non-collinearity, L(S)DA does not. However there are some tricks to avoid GGA problems, although not actually solving them, and there are suggestions how to solve them ... None are to my knowledge yet implemented in the standard code.
The L(S)DA implementation in Elk is truely non-collinear!
The GGA is not really, due to the imperfect standard GGA-formalism.
Best regards,
Lars
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
thanks for answering despite of my nationality and sorry for my late reply!
I also have to admit that my question was badly formulated.
Let me recap to make some points clear. When you say
The L(S)DA implementation in Elk is truly non-collinear!
You mean the Kuebler trick, i.e.
To be more precise I was wondering if the elk code included an exchange and correlation potential which has a dependency on the propagation vector of a spin spiral.
I tried to dig into the code by following the variables vqlss and vqcss (f.t.r propagation vector in lattice and Cartesian coordinates) and I did not find anything like that, but of course I could be very wrong.
I believe that the problem can be more generally stated as the accuracy of the assumption that the exchange-correlation magnetic field is parallel to the local magnetization.
As a consequence, are there xc functional in Elk for which the above assumption is relaxed?
Thanks again for your support and your kindness,
best regards,
Pietro Bonfa'
Last edit: Pietro Bonfa 2016-06-22
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
You are right about m(r) and B_xc(r) being parallel at each point in space by construction. This is true of all (semi-) local functionals in elk.
There are a couple of "truly" non-collinear functionals in this sense-- functionals which do not assume m(r) and b_xc(r) are parallel at each r. (i) Phys. Rev. Lett. 111, 156401; this functional was implemented initially in elk. But we just never could converge it. So right now it is not a part of elk. (ii) exact exchange +lsda correlations. This is implemented in elk and is truly non-collinear. But this is a very expensive calculation.
Best
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
One final point: I saw that EXX cannot be used with spin spirals (init0.f90 line 130).
However, Dr. Florian Eich's thesis (http://www.diss.fu-berlin.de/diss/receive/FUDISS_thesis_000000094776) shows how to extend the EXX approach to wavefunctions written in the form of spin spirals, correct?
Thank you very much again for all the help.
Best regards,
Pietro
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear All,
I was asked by a referee to check my non-collinear spin calculations with a true non-collinear exchange and correlation functional.
I quickly grep(ped) the elk code for some references (mainly PRB or PRL articles, for example PhysRevLett.111.156401) but I could not find anything.
Can experienced users/developers please update me about the current implementation of non-collinear spin functionals (both in Elk or in other codes)?
Thanks in advances for your time.
Best regards,
Pietro Bonfa'
Hi Pietro!
I sincerely hope you are not Italian, then you do not deserve any answer;-(
Otherwise, have you used GGA or LDA in your calculations? GGA has problems with non-collinearity, L(S)DA does not. However there are some tricks to avoid GGA problems, although not actually solving them, and there are suggestions how to solve them ... None are to my knowledge yet implemented in the standard code.
The L(S)DA implementation in Elk is truely non-collinear!
The GGA is not really, due to the imperfect standard GGA-formalism.
Best regards,
Lars
Dear Dr. Nordström,
thanks for answering despite of my nationality and sorry for my late reply!
I also have to admit that my question was badly formulated.
Let me recap to make some points clear. When you say
You mean the Kuebler trick, i.e.
To be more precise I was wondering if the elk code included an exchange and correlation potential which has a dependency on the propagation vector of a spin spiral.
I tried to dig into the code by following the variables vqlss and vqcss (f.t.r propagation vector in lattice and Cartesian coordinates) and I did not find anything like that, but of course I could be very wrong.
I believe that the problem can be more generally stated as the accuracy of the assumption that the exchange-correlation magnetic field is parallel to the local magnetization.
As a consequence, are there xc functional in Elk for which the above assumption is relaxed?
Thanks again for your support and your kindness,
best regards,
Pietro Bonfa'
Last edit: Pietro Bonfa 2016-06-22
Hi Pietro,
You are right about m(r) and B_xc(r) being parallel at each point in space by construction. This is true of all (semi-) local functionals in elk.
There are a couple of "truly" non-collinear functionals in this sense-- functionals which do not assume m(r) and b_xc(r) are parallel at each r. (i) Phys. Rev. Lett. 111, 156401; this functional was implemented initially in elk. But we just never could converge it. So right now it is not a part of elk. (ii) exact exchange +lsda correlations. This is implemented in elk and is truly non-collinear. But this is a very expensive calculation.
Best
Dear Dr. Sharma,
that's extremently clear, thank you.
One final point: I saw that EXX cannot be used with spin spirals (init0.f90 line 130).
However, Dr. Florian Eich's thesis (http://www.diss.fu-berlin.de/diss/receive/FUDISS_thesis_000000094776) shows how to extend the EXX approach to wavefunctions written in the form of spin spirals, correct?
Thank you very much again for all the help.
Best regards,
Pietro