This thread is locked.Only browsing is available.
Top Page > Browsing
Question about non-collinear MLWF calculation
Date: 2014/11/22 02:27
Name: Limillas   <>

Hi all,

I have a problem using non-collinear (NC) calculation Maximally-Localized Wannier Functions (MLWF) in OpenMX.

In NC case, the wave functions calculated from DFT are spinors (combination of two spin components). Therefore the Wannier functions generated from these spinors should also be spinors.

In my calculation, the inner window has 22 spin bands (11 if spin degenerate). Therefore 11 spinor Wannier functions should be enough to calculated these bands. However the program told me "Bands number with inner window is larger than Wannier function number."

If I choose to use 22 Wannier function instead, I will get good band interpolation. However from the output file, the spinsize of this case is only 1. I think it should be 2. The Wannier functions generated this way might be collinear and doesn't include any information about different spin components.

Could you give me any advice on this problem?

Thank you!

- Limillas
Page: [1]

Re: Question about non-collinear MLWF calculation ( No.1 )
Date: 2014/11/23 08:26
Name: T. Ozaki


Suppose that you calculate a hydrogen atom with a single s-orbital by the NC method.
Then, you have two spinor Kohn-Sham orbitals, one corresponds to spin-up and the other
spin-down. The same is true for your case. You have 22 spin bands in the inner window.
Thus, you have to specify 22 Wannier functions. Also, the spinsize of 1 in the output file
is also correct, since the spin-up and spin-down are not distinguish a priori, but they are
obtained as a result in the NC calculations if spin-orbit coupling is neglected.



Page: [1]