[Pw_forum] Problem with nbnd in clusters
Stefano de Gironcoli
degironc at sissa.it
Wed Nov 21 10:39:30 CET 2007
methfessel-paxton smearing is dangerous (because it is not
positive-defnite) when k-point sampling is not sufficient.
When dealing with a "fake metal" as your molecule where the Fermi energy
should fall in gap of the spectrum I think it is more advisable to use
simple gaussian or marzari-vanderbilt smearing. The spurious negative
weights of m-p smearing the occupation of levels just above the Fermi
energy can lock some spourios state that shouldn't be there ... If you
have many kpoints this (in my experience... Nicola has a different
opinion) is never a problem.
In any case the smearing contribution for a molecule should be zero
... which is not the case in your output ... particularly so for the
problematic cases with 12 or 25 bands..
TEST> grep -e TS -e ! Fe*.out
Fe_dimer_gamma_nbnd12.scf.out:! total energy =
-121.76376536 Ry
Fe_dimer_gamma_nbnd12.scf.out: smearing contrib. (-TS) =
-0.00157785 Ry
Fe_dimer_gamma_nbnd20.scf.out:! total energy =
-121.76907297 Ry
Fe_dimer_gamma_nbnd20.scf.out: smearing contrib. (-TS) =
0.00048313 Ry
Fe_dimer_gamma_nbnd25.scf.out:! total energy =
-121.76376546 Ry
Fe_dimer_gamma_nbnd25.scf.out: smearing contrib. (-TS) =
-0.00157856 Ry
Does the problem (at least the one with 25 bands) disappears if you use
"m-v" of "gaussian" smearing ?
I would expect so... otherwise there is still something else.
hope it helps,
stefano
Cyrille Barreteau wrote:
> Dear Paolo,
>
> Maybe I found the second exception due to another very special
> phase of the moon:-) What do I win in this case:-)
>
> You will find at the following links:
> the input file, the pseudo file and the three output
> corresponding to nbnd=12,20,25 respectively.
>
> http://www-drecam.cea.fr/spcsi/cbarreteau/tempo/Fe_dimer.scf.in
> http://www-drecam.cea.fr/spcsi/cbarreteau/tempo/Fe.pw91.vdb.UPF
> http://www-drecam.cea.fr/spcsi/cbarreteau/tempo/Fe_dimer_gamma_nbnd12.scf.out
> http://www-drecam.cea.fr/spcsi/cbarreteau/tempo/Fe_dimer_gamma_nbnd20.scf.out
> http://www-drecam.cea.fr/spcsi/cbarreteau/tempo/Fe_dimer_gamma_nbnd25.scf.out
>
> thanks for having a look at this case.
>
> cyrille
>
> -
> ==================================================================
> Cyrille Barreteau | phone : +33 (0)1 69 08 29 51
> CEA Saclay | fax : +33 (0)1 69 08 84 46
> DSM/DRECAM/SPCSI | email cyrille.barreteau at cea.fr
> Batiment 462 |
> 91191 Gif sur Yvette Cedex FRANCE
> ~~~~~~~~~~~~~~~~~~~~~~~~
> http://www-drecam.cea.fr/spcsi/index.php
> http://www-drecam.cea.fr/Images/Pisp/cbarreteau/cbarreteau_fr.html
> http://www-drecam.cea.fr/Phocea/Membres/Cours/index.php
> ==================================================================
>
>
>
>
> Paolo Giannozzi wrote:
>
>
>> On Nov 20, 2007, at 15:55 , Cyrille Barreteau wrote:
>>
>>
>>
>>
>>> Does anyone has experience about the behaviour
>>> of the various diagonalization schemes?
>>> Which one is better, in which case?
>>>
>>>
>>>
>> the so-called conjugate-gradient diagonalization basically
>> never fails; it uses less memory but it is much slower than
>> the block Davidson algorithm. The latter is also quite
>> robust but there is a tiny probability to end up in the wrong
>> ground state. The only such case I know happens in some
>> high-symmetry perovskites, if you start from superposition
>> of atomic states, with occupied states only, and if the phase
>> of the moon is close to 42. The problem disappears if you
>> add a few more states, or if you start from random wavefunctions.
>> I am quite surprised to hear that there is a case in which
>> you get good results with 12 states and bad with 20.
>> I wouldn't be surprised by the opposite. Could you please
>> post one jobs showing the problem?
>>
>> Paolo
>> ---
>> Paolo Giannozzi, Dept of Physics, University of Udine
>> via delle Scienze 208, 33100 Udine, Italy
>> Phone +39-0432-558216, fax +39-0432-558222
>>
>>
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum
>>
>>
>>
>
>
> -
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
More information about the Pw_forum
mailing list