[Pw_forum] Error in vc-relax

Konstantin Kudin konstantin_kudin at yahoo.com
Mon Sep 17 20:16:12 CEST 2007


 Wow, this was like a very entertaining post! I nominate Axel for the
appropriate prize !!! Kids, don't play with the fire that is QE ! :-)

 By setting the entertainment aspect aside, I suspect that most users
would rather much prefer to get the job done quickly while remaining
for the most part ignorant about the code intrinsics, and move on with
their life, as opposed to diving deep into the manuals and PW forum
archives in order to figure out how to force the code to do what is
needed.

 So, from this angle, some sort of *automatic* solution would be
desirable. I have used vc-relax quite a lot for 2D cell optimizations.
The code crashes each time there is a symmetry decrease. I am certain
that changes in the cell geometry cause such effects, for example, when
the input cell is hexagonal, and the first step makes it non-hexagonal.
So, then, for example, the system goes from 8 symmetry operations to 4,
and the code crashes. A fix would involve either making the code
respect the input symmetry such that the cell changes do not alter it,
or, the code could re-initialize with the new symmetry when a problem
is detected, sort of like doing an internal restart.

 My examples are all huge, so I won't post them yet.

 Kostya

--- Axel Kohlmeyer <akohlmey at cmm.chem.upenn.edu> wrote:

> On Mon, 17 Sep 2007, Paolo Giannozzi wrote:
> 
> hi paolo,
> 
> PG> Since complaints about the variable-cell optimization
> PG> show up more often than not: can anybody provide a
> PG> complete test job (i.e. one that can be rerun), with
> 
> my impression is more often, that this is a manifestation
> of vc-relax being a less straightforward to use method and
> that people don't always understand all the involved subtleties.
> as a consequence of that, one is tempted to use vc-relax
> in a way that is not working well, or for calculations 
> (lattice constant determination) where there may be other
> more efficient and reliable ways to obtain this information.
> 
> i can sympathize very much with these ideas, since i had
> them all myself and it was a (sometimes painful) learning
> process to get a better understanding of where i was 
> going wrong.
> 
> PG> complete output, showing a problem in a reproducible
> PG> way, in a reasonable amount of computer time, in a
> PG> recent versions of Q-E?
> 
> to some level, we'll just have to get used to people trying
> things with vc-relax that cannot work (you have to get
> burned to stay away from the fire!) or do not work in the
> way they hoped for. other than that a few more extensive 
> explanations, pointers to (more) commented tutorial examples, 
> also in the INPUT_PW file and perhaps an adjusted default for 
> the cell mass (as cesar has pointed out a couple of times), 
> could handle the majority of the cases.
> 
> cheers,
>    axel.
> 
> p.s.: i agree with your assessment, that when somebody posts
> an incomplete input, that the chance of somebody completing it
> and using it for testings is (almost) zero. posting only an
> input fragement, should be ok (and saves a lot of bandwidth and
> storage), when people are not certain about their input flags
> altogether, but if a problem persists, there is not alternative
> to making the whole input available, so that somebody can (try to) 
> re-run the calculation elsewhere.
> 
> PG> 
> PG> Paolo
> PG> ---
> PG> Paolo Giannozzi, Dept of Physics, University of Udine
> PG> via delle Scienze 208, 33100 Udine, Italy
> PG> Phone +39-0432-558216, fax +39-0432-558222
> PG> 
> PG> 
> PG> 
> PG> _______________________________________________
> PG> Pw_forum mailing list
> PG> Pw_forum at pwscf.org
> PG> http://www.democritos.it/mailman/listinfo/pw_forum
> PG> 
> 
> -- 
>
=======================================================================
> 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.
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
> 



      ____________________________________________________________________________________
Fussy? Opinionated? Impossible to please? Perfect.  Join Yahoo!'s user panel and lay it on us. http://surveylink.yahoo.com/gmrs/yahoo_panel_invite.asp?a=7 



More information about the Pw_forum mailing list