I did a calculation based on VASP. The goal is to exact band structure from wannier90 package.
The initial projection for wannier90 is essential, so I used Random initial projections to obtain the wannier functions. But the following convergence of the “Wannierisation” procedure is rather harder. After 7000 iterations, the convergence window is still around 1.0E0.
I think the issue may comes from the initial projection. Then, I assigned the atoms the orbital projection for the initial projection. As a results the num_wann is not equal to num_bands, which is from NBANDS in VASP. However, the convergence of the “Wannierisation” procedure is difficult, too.
So I wander if anyone can offer me some tips? Thanks a lot!!
PS: the system contains 48 atoms, which presents NBANDS = 184 by default.
Dear Henk,
Thank you for your comment. I have modified title of the question and been waiting for answers.
Best regard,
Jiaduo Zhu
Dear Jiaduo,
The initial projections should be chosen according to the chemical environment/electronic structure of your system of interest. Have a look at: Phys. Rev. B 92, 165134 – Published 30 October 2015 (link attached)
Best regards,
Alberto
https://journals.aps.org/prb/abstract/10.1103/PhysRevB.92.165134
Your WF centres are supposed to be atomic p and s-orbitals, but their corresponding spreads are way out of range, meaning that you haven't chosen the right energy window and energy bands for wannierisation. You need to fix this first!
As for the site symmetrization, as far as I know, it doesn't work with non-collinear calculations. It also needs the input file "seedname.dmn", which I don't think can be taken from VASP. So, you may need to make it manually.