Problem DFT+U vector magnetization

[alexzai]

Hallo

In new ABINIT 7.10.4 version remain the same problems with calculation of U-correction with nspden=4.

Best regards
Alex

[amadon]

Dear Alex,

Sorry for the delay.
What is the problem with the DFT+U vector magnetization with nspden=4 ?

Best regards
Bernard Amadon

PS: Could you give your name and affiliation, thanks!

[alexzai]

Thank you for the answer.

The problem is definition of U-parameter in antiferromagnetic system with spin-orbit coupling. At calculation of U-parameter with spin-orbit-coupling the result consist of zeros and arises errors in .out file: "Matrix response chi_0 is singular Probably too many symmetries kept Matrix response chi is singular Probably too many symmetries kept" and warning in .log file:
"-- !WARNING
message: |
LDA+U+SpinOrbit is still on test
(not yet in production)
src_file: chkinp.F90
src_line: 2591"
My system is very asymmetrical and the further reduction of symmetry is not possible. I see, that this function unfortunately today is not realized and hope to realisation in future. Know you other methods of U parameter in the systems definition? Now i can run only the collinear calculation.

Best regards
Alexander Zaichenko
Göttingen

[amadon]

Dear Alexander,

Sorry, I was missing your point. The calculation of U within cRPA is implemented in ABINIT only if nspinor=1 (collinear case). It
was not written in the doc. So I have just updated it now (input variable: ucrpa) in the doc.

Best regards
Bernard

[alexzai]

Thank you for the fast answer

But i have used linear response method, not RPA and i did not know, that in GW calculation is the same problem.

Best regards
Alexander

[amadon]

In fact, my point is that the calculation of the effective Coulomb Interaction U with the constrained Random Phase Approximation (cRPA)
is not implemented in the case nspinor=2. (I am not discussing GW calculations in general).

Concerning the linear response method to compute U, to my knowledge, it was also not implemented in non collinear case.

Best regards
Bernard