[Wannier] [Fwd: Wannier90 kpoints issue]

[Jonathan Yates]

On 2 Oct 2008, at 18:02, Lex Kemper wrote:

>
> Hello,
>
> I get an error from wannier90 that I don't know how to fix. I am
> attempting to do a calculation for a relatively small system, 5 atoms,
> in a body-centered tetragonal lattice. When I try to create the pwscf
> inputfile, wannier90 gives me this error after the first 2 bvectors.
>
> kmesh_get_bvector: Not enough bvectors found
>
> I'm not sure what this means or how to fix it. Any suggestions?
>
> The (relevant) portions of the input/output are below.
>
>  |                    Distance to Nearest-Neighbour Shells        |
>  |                    ------------------------------------        |
>  |          Shell             Distance (Ang^-1)          Multiplicity
>  |          -----             -----------------           
> ------------        |
>  |             1                   0.265096                       
> 2        |
>  |             2                   0.423354                       
> 6        |
>  |             3                   0.423356                       
> 2        |
>  |             4                   0.530192                       
> 2        |

Lex,

  The short answer to your question is that you either need to specify  
the lattice vectors to greater precision, or set
  kmesh_tol=0.00001 (10x lower than the default)
    The former is aesthetically 'nicer' but no harm will come from the  
latter.


  The error comes from the part of w90 which sets up the finite  
difference formula (B1 in MV). It has to analyse the lattice of  
kpoints to find shells of nearest neighbours. From the section of  
output file above you can see that shells 2 and 3 are very nearly  
degenerate: in fact if you think about the symmetry of your lattice  
they should be degenerate (I'm guessing that the finite precision in  
your lattice vectors has broken this).
   In this circumstance w90 gets a bit confused, hence the error. We  
could code it to be more robust to this, although this is messy (and  
hides the real problem which is the lack of precision) so I prefer not  
to. But I will improve the error message to make it more helpful!

  I assume you are using pwscf. What would be useful is a program/ 
script to convert from pwscf input to w90 input keeping 'enough'  
precision. Any volunteers?

  Jonathan

General note: I'm in the middle of transferring institutions. If  
you've sent me mail in the last couple weeks and I haven't replied -  
sorry, I am (slowly) catching up.


-- 
Department of Materials
University of Oxford
Parks Road, Oxford, OX1 3PH, UK
http://www.tcm.phy.cam.ac.uk/~jry20/