[CPMD-list] problems with wannier center

[Axel Kohlmeyer]

On Tue, 28 Aug 2007, Hui-Hsu Gavin Tsai wrote:

dear gavin,


GT> dear axel,
GT> 
GT> Thanks for your quick response.

well, this time you had to wait a little longer, 
since i needed some sleep in between. ;-)

[...]

GT> The input files were taken directly from the tutorials
GT> (http://www.pci.uzh.ch/index_e.php?page=http://www.pci.uzh.ch/gruppe.hutter/e/information.html).

i can only assume, that those were using a too small basis set
to demonstrate the use of the keywords fast on a small computer
at the expense of not doing a meaningful simulation. similar 
to, e.g., doing a coupled cluster calculation with an STO-3G 
basis set, just to demonstrate how a this calculation would look like.

GT> > AK> the input reads coordinates from the restart (if present),
GT> > AK> so with every new test you start from a worse initial set
GT> > AK> of coordinates. please just visualize those coordinates
GT> > AK> from GEOMETRY.xyz...
GT> 
GT> In the input file, the keyword "restart" is shown in low case.
GT> If we understand correctly, it is commented out and will not be
GT> recognized by CPMD.

my mistake, you are correct. as you may have seen from many comments
on this mailing list by me, i consider this not a good practice.

GT> In addition, the wavefunction is first converged using the keyword
GT> "QUENCH BO" in the input file.

well, your coordinates were still far off from the initial ones, 
in the input. i can only guess that this is due to the memory
corruption induced from the (known) localization bug.

GT> The related files tested with 4 nodes and 2 nodes are attached FYI.
GT> The input files, inp1-2nodes.inp and inp1-4nodes.inp are identical.
GT> The job using 2 nodes is crashed (output files are labeled with 4
GT> nodes) and the job using 4 nodes works fine (output files are labeled
GT> with 4 nodes).

so those inputs work fine for me, but i am using a different version 
of cpmd. nevertheless, i don't quite get what your problem is?
so you already know that setting WANNIER SERIAL or changing the code
is the recommended workaround. so if you changed the code, it was
most likely incomplete, and none of your inputs disables the 
localization or uses WANNIER SERIAL to demonstrate that there is 
another bug beyond the one that is already known and corrected 
in the developers version.

GT> If you need more tests and information regarding this problem, please
GT> let us know.

all inputs work fine with the developer's version of the code,
so i don't see the need to investigate any further from my side.

cheers,
   axel.

GT> 
GT> Thanks again.
GT> 
GT> Gavin Tsai
GT> Department of Chemistry
GT> National Central University
GT> Taiwan
GT> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.