[Pw_forum] Electron Phonon interaction: simple-minded questions

Paolo Giannozzi giannozz at nest.sns.it
Tue Nov 7 10:35:01 CET 2006

On Nov 3, 2006, at 18:38 , Andrea Marini wrote:

> My question is: as I am using a MP grid the shifted points (the  
> ones with zero weight) are still points of the 2x2x2 grid (like k 
> (2) and k(5)) . Why does PW calculates the states for these points  
> twice ?

because in this particular case you have k(1)+q=k(5)

> Is there any way to avoid it ?

for sure there is, but it requires a programming effort that is not
justified by a small performance gain in special cases only

> 2. Why does PW performs a band structure calculation for every q- 
> point of a subset of the 2x2x2 grid. Can I calculate the states of  
> the whole grid once and use it for each q-point phonon calculation ?

I think this is the way it is usually done, at least in recent versions

> The point is that, while I get marvelous Eliashberg functions for a  
> metal like Al when I look at  the el-ph matrix elements between the  
> fermi surface and high energy bands I'd expect them to go to zero  
> as a function of the upper band energy, while it is not. Does any  
> of you have experience of the high energy behavior of the el-ph  
> matrix elements ?

I haven't any experience. High-energy bands are likely to be less  

> 4. When I try to calculate the matrix elements for many states I  
> get this error message already in the sc run
>      task #         1
>      from cgramg1_k : error #         1
>      negative or zero norm in S
> This error disappears when I increase the energy cutoff.

the entire machinery is based on the assumption that the number
of states is much smaller than the number of plane waves, so when
this is no longer true it may break down

Paolo Giannozzi, Democritos and University of Udine, Italy

More information about the users mailing list