[Pw_forum] problem with symmetry recognition

[giuseppe.mattioli at mlib.ism.cnr.it]

Dear Stefano, dear all.
Ok, but suppose that I would like to put a (quite useful) constraint  
like: "I say that the lattice is cubic because I know that it MUST be  
cubic (or tetragonal or monoclinic or so...), than KEEP IT CUBIC while  
performing a vc-relax, irrespective of the stress tensor! Change the  
volume, change the a/c ratio (if the lattice is, e.g., tetragonal) but  
retain the space group. I often have to perform a (boring) multi  
dimensional fit by changing the lattice parameters by hand on a grid  
near to the "expected" minimum...
I do not know exactly if Pascal's question was similar to mine;  
nevertheless I hope this can help...

Yours

Giuseppe


Quoting Stefano de Gironcoli <degironc at sissa.it>:

> this is not correct.
> if the code finds some symmetry, forces and stress components will be
> symmetrized accordingly  and vc-relax will keep it.
>
> There may be two reasons why the code does not find a symmetry:
>
> - the symmetry is not there... check your coordinates. Beware that you
> need to specify ALL the atomic positions of a given structure not just
> the inequivalent ones, hoping the code to fill the others.
>
> - the symmetry is there but involves a fractionary translations that
> cannot be written as a multiple of the FFT grid spacing.  the code
> writes a message saying that the symmetry is discarded due to a problem
> with fractionary translation and says what the translation would be in
> crystal coordinates. one can manually set the FFT dimensions (nr1,nr2,
> and/or nr3) so that the translation is in register with the FFT grid
> and the code can accept it
>
> HTH
>
> stefano
> -
> Stefano de Gironcoli - SISSA and DEMOCRITOS
>
>
>
> On 07/09/2011 07:58 PM, Vic Bermudez wrote:
>> Hi Pascal,
>>
>> 	If I understand correctly, you're trying to optimize the unit cell while
>> constraining 2 of the lattice constants to remain equal, as required by
>> symmetry. I don't think that QE allows this. If you do a "vc-relax", what
>> seems to happen is that symmetry is ignored, and all lattice constants are
>> varied independently. I've attached a script written by a colleague that
>> does repeated single-point runs for a range of lattice constants. You can
>> then fit points near the total-energy minimum to find the exact minimum. Of
>> course, this has to be done iteratively if more than one lattice constant is
>> being optimized.
>> 	Hope this helps.
>>
>> 	Best Wishes,
>> 	Vic Bermudez
>>
>> Victor M. Bermudez
>> Code 6876
>> U.S. Naval Research Laboratory
>> 4555 Overlook Ave., S.W.
>> Washington, DC 20375-5347
>>
>>
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum