[Wien] Supercell

[Stefaan Cottenier]

georg at chem.au.dk wrote:

>I'm no expert on these supercell calculations, but I guess the density of
>k-points should be the same.
>This means that if you you use 2000 k-points for generation (the number you type
>in to kgen, not the number in the IBZ) then you'll need approx. 500 generating
>k-points for a 2x2x1 supercell. However the actuall number in the IBZ may not
>be lower than the original because your symmetry is also lower
>  
>
In addition to what Georg wrote, you can find an equally dense mesh also 
in the following way:

For your original cell and original k-mesh, look at the top of 
case.klist. It looks for instance like

         1    7   11    7  154  8.0 -7.0  1.5      1000 k, div: ( 11  7 11)
         2    7   11   21  154  8.0
         3    7   11   35  154  8.0

Here you see (after the 'div') that your k-mesh is a 11x7x11 grid. If 
you now double the unit cell in the x-direction (2x1x1), you Brillouin 
zone will get half as large in that direction. This means you need half 
as many gridpoints for k in that direction. Vary the input for kgen 
until you see approximately 5x7x11 or 6x7x11. For a 2x1x2 supercell, you 
go for 5x7x5 or 6x7x6, for a 2x2x2 supercell you need 5x3x5 or 6x4x6, etc.

Stefaan