[Pw_forum] a question about nbnd setting in band structure calculation

[Paolo Giannozzi]

On May 11, 2007, at 24:21 , Peter Winey wrote:

>  When I run the "bands" calculation for Al, I found that the  
> results depend on the setting of "nbnd". I tried nbnd=8 (the value  
> used in the example) and nbnd=10.
> The bands below the Fermi level are the same in both cases,  
> however, some
> of the empty bands are clearly different.

if you want to be sure to get all the N bands you want, you need to  
calculate
a few more than N. Iterative diagonalization requires a good starting  
point.
If for some reason the final bands are not present in the starting  
manifold,
you can easily miss them. This never happens (with the notable exception
of some ferroelectric perovskites) in scf calculations, even if you  
use the
strict minimum of bands, because sooner or later the 'good' bands  
find their
way into the lowest-energy manifold. In non-scf calculation it may  
happen
that a few of the highest bands are not what they are supposed to be,  
though.

Paolo
---
Paolo Giannozzi, Democritos and University of Udine, Italy