[CPMD-list] Adapted geometry optimization went abnormally after several thousands steps..

[Axel Kohlmeyer]

On 6/17/06, osirixu-cpmd at yahoo.com.cn <osirixu-cpmd at yahoo.com.cn> wrote:
> Dear cpmders,

dear rex,

> steps with adapted convergence criteria, the
> optimzatio went abnormally and reported many "NaN"
> strings (the output file is also appended below). It
> seems the LAMBDA was going bigger and bigger. Is it
> because of the using of PCG MINIMIZE and LBFGS?

unlikely. that is usually only an indicator that something
has gone wrone before.

>       Well, I have tried with ODIIS, but I find it is
> also too hard to reach a convergence. In addition, I

this is strange, you don't have any complicated elements
in there, so DIIS should work fine.

> also encountered the same problem when running BOMD
> after several hundreds of md-steps.
>
>       Any suggestions?

does the same happen when you keep the supercell size fixed?

please look closer at the output. basically, your supercell is
exploding (in y-direction).

>
>     greetings,
>        Rex
>     June 18, 2006
>


>  **   CELL ABSOLUTE                       **
>  **     4.73030 4.31381 5.19292    0 0 0                       **
>  **   REFERENCE CELL ABSOLUTE                       **
>  **      5 5 5.5   0 0 0                       **

this is a pretty small cell for a gamma point only run.

>  CELL PARAMETER (VOLUME=    675.4196):

>  CELL PARAMETER (VOLUME=    503.8368):

>  CELL PARAMETER (VOLUME=    794.8757):

>  CELL PARAMETER (VOLUME=   1027.5242):

>
>  WARNING| SPLINE REGION SMALLER THAN MAXIMUM G-VALUE

at this point your cell is far to large to fit into the precalculated
spline range from the reference cell in your input.
the whole wavefunction optimization _cannot_ work properly
anymore.

please take warnings in the CPMD output seriously until
you _know_ they are harmless.

regards,
     axel.



> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://cpmd.org/mailman/listinfo/cpmd-list
>
>


-- 
=======================================================================
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.