[Wannier] A question based on No.23 example in wannier90 tutorial, GW calculation

[Antimo Marrazzo]

On 1/16/20 9:12 AM, 张侃 wrote:
>
> Previously, I run the silicon example, and my system with 4x4x10 
> k-points in all scf and nscf steps and including QP corrections, in 
> which no errors occurred and results can be output successfully. In 
> the current set, I only change k-pints in scf file to 10x10x28, 8x8x22 
> in gw.nscf file, and keep 4x4x10 in nscf file. QP corrections were 
> calculated on 4x4x10 grids. Here is my system in QE input form.
>
Dear Zhang,

the GW-QP grid needs to be commensurate with the NSCF grid, i.e. every 
point in the GW-QP grid must be present in the set of k-points used for 
the NSCF. You just need to factorize the grid n1xn2xn3 to know which 
subgrids are allowed.
If you use 22 k-points along k_z in the NSCF, then you can only use 
22,11 or 2 (22=11*2) k-points in the GW calculations.
Could you try if this fixes the error you get?
Thanks.

Best,
Antimo

-- 
Antimo Marrazzo
EPFL STI IMX THEOS
ME-D2 1326 (Bâtiment ME)
Station 9
CH-1015 Lausanne (Switzerland)