The ABINIT Users Mailing List ( CLOSED )

[Anglade Pierre-Matthieu <anglade@gmail.com>]

Hi Matthieu

Is it possible that, for the run that doesn't converge tolvrs, at one
or another step, the system gets in a metallic state that makes the
convergence of potential more difficult ?


On Tue, Mar 3, 2009 at 5:20 PM, matthieu verstraete
<matthieu.jean.verstraete@gmail.com> wrote:
> Ok, it's better (thanks to PMA), but still not entirely satisfactory: with
> nline 9, the tolwfr goes down, albeit more slowly, to 1.e-17 (so far).
> However the tolvrs is still stuck at 1.e-4 after 70 steps.
>
> With all the symops I get
>
>  ETOT 32  -11.444990018007     1.243E-14 9.972E-21 3.357E-20 0.000E+00
> 0.000E+00
>
> beautiful... So why is it so dependent on the non-symmorphic operations? The
> number of kpoints is the same.
>
> For the symmorphic operations in the normal cell, you no longer have
> inversion symmetry. Changing to a cell with inversion symmetry (origin in
> the middle of the C-C bond), many more of the operations are non-symmorphic;
> neglecting these you do not even have a null force on the atoms (only 8
> symops left, 343 kpt instead of 127). However, it looks like all the other
> tolerances are converging:
>
>  ETOT 21  -11.444989920147    -8.717E-12 1.188E-13 2.693E-05 8.456E-04
> 4.227E-04
>
> So, it looks like the wfr and potential convergence depends on having
> inversion symmetry, whereas getting correct forces needs some of the other
> symops (probably the 3-axes - I am now trying with a cell preserving both
> inversion and 3-axes).
>
> Matthieu
>
>
> On Tue, Mar 3, 2009 at 8:44 AM, Anglade Pierre-Matthieu <anglade@gmail.com>
> wrote:
>>
>> Hi Matthieu,
>>
>> Could it be related to the default value of nline which is sometime a
>> bit too small for large systems ? Have you tried a larger value for
>> nline in the case with removed symmetry operations ? If so, is the
>> behavior similar ?
>>
>> regards
>>
>> PMA
>>
>> On Tue, Mar 3, 2009 at 12:29 AM, Matteo Giantomassi
>> <Matteo.Giantomassi@uclouvain.be> wrote:
>> >  Hello abinists,
>> >>
>> >> I have come upon a strange behavior in abinit (not the first, of
>> >> course)
>> >> and
>> >> was wondering if anyone had heard of it or an explanation (v5.6 on
>> >> opteron
>> >> 64 bit RHEL, g95 compiled, sequential).
>> >>
>> >> In converging a ground state for graphene (what else), I need to use
>> >> only
>> >> symmorphic symmetry operations. The strange behavior is that, without
>> >> the
>> >> non-symmorphic symmetry operations, abinit does not converge the
>> >> wavefunctions below 1.e-13 or so (even after 200 steps), and the
>> >> eigenvalues
>> >> are not converged below mHa accuracy. With all the symops the
>> >> convergence
>> >> down to tolwfr 1.e-20 is effortless. Why the difference? Numerical
>> >> noise
>> >> is
>> >> one possibility, but plenty of symops are left in the calculation...
>> >>
>> >  Hi, Matthieu
>> >
>> >  I've never encountered such a kind of problem but
>> >  it might be related to a partial fulfillment of the symmetry properties
>> >  of the density and of the generated potentials.
>> >  Do you have the same number of k-points in the IBZ when non-symmorphic
>> >  operations are removed?
>> >
>> >  Just for curiosity: are you pruning non-symmorphic ops because you want
>> > to
>> >  generate a KSS file for GW calculations with abinit?
>> >  In this case please note that symmorphi=0 is not mandatory anymore
>> >  as the GW part is now able to handle non-symmorphic space groups.
>> >
>> >  Cheers,
>> >  Matteo
>> >
>> >
>> >
>> >
>>
>>
>>
>> --
>> Pierre-Matthieu Anglade
>>
>
>
>
> --
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Dr. Matthieu Verstraete
>
> European Theoretical Spectroscopy Facility (ETSF)
> Dpto. Fisica de Materiales,
> U. del Pais Vasco,
> Centro Joxe Mari Korta, Av. de Tolosa, 72,   Phone: +34-943018393
> E-20018 Donostia-San Sebastian, Spain        Fax  : +34-943018390
>
> Mail : matthieu.jean.verstraete@gmail.com
> http://www-users.york.ac.uk/~mjv500
>



-- 
Pierre-Matthieu Anglade