[Pw_forum] Re: Relativistic GGA ?

[Axel Kohlmeyer]

On Tue, 24 Apr 2007, Lazaro Calderin wrote:

LC> Thanks Andrea. 
LC> 
LC> I have been trying ... but I haven't found a way to run relativistic GGA
LC> jobs with pwscf.
LC> 
LC> Below are the input options and the corresponding errors that I get for
LC> some quick and dirty tests for fully relativistic GGA jobs. 
LC> 
LC> The following tests were done for Hg using TM and US fully relativistic
LC> GGA(PBE) pseudopotentials and PWSCF version 3.2. 
LC> 
LC> ------------------------------------------------------------------------
LC> Using: GGA(PBE) fully relativistic Troullier-Martins pseudopotential for
LC> Hg
LC> 
LC> 1)
LC>     noncolin=.true.,
LC>     lspinorb=.true.,
LC>     starting_magnetization(1)=0.0,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from stres_gradcorr : error #         1
LC>      noncollinear stress + GGA not implemented
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

note, this message comes obviously from the subroutine calculating
the stress tensor. so if you can go without, it might still work,

cheers,
  axel.

LC> 
LC> NOTE: The same test (same input file) but using relativistic LDA psp
LC> runs without errors. 
LC> 
LC> 2)
LC>     noncolin=.false.,
LC>     lspinorb=.true.,
LC>     starting_magnetization(1)=0.0,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from setup : error #         1
LC>      wrong beta functions
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC> 
LC> 3)
LC>     noncolin=.false.,
LC>     lspinorb=.false.,
LC>     starting_magnetization(1)=0.0,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from setup : error #         1
LC>      wrong beta functions
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC> 
LC> 4)
LC>     nspin=1,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from setup : error #         1
LC>      wrong beta functions
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC> 
LC> 
LC> -----------------------------------------------------------------------
LC> Using: GGA(PBE) fully relativistic ultra soft pseudopotential for Hg
LC> (the same happens with BLYP)
LC> 
LC> 
LC> 1)
LC>     noncolin=.true.,
LC>     lspinorb=.true.,
LC>     starting_magnetization(1)=0.0,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from stres_gradcorr : error #         1
LC>      noncollinear stress + GGA not implemented
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC> 
LC> NOTE: The same test (same input file) but using relativistic LDA psp
LC> runs without errors. 
LC> 
LC> 2)
LC>     ncolin=.false.,
LC>     lspinorb=.true.,
LC>     starting_magnetization(1)=0.0,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from setup : error #         1
LC>      US j-average not yet implemented
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC> 
LC> 
LC> 3)
LC>     ncolin=.false.,
LC>     lspinorb=.false.,
LC>     starting_magnetization(1)=0.0,
LC> 
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC>      from setup : error #         1
LC>      US j-average not yet implemented
LC> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LC> 
LC> 
LC> So far it seems that: 
LC> 
LC> I)   Fully Relativistic GGA is not implemented yet for spin polarized
LC> systems. 
LC> 
LC> II)  It is not implemented yet neither for non-spin polarized systems
LC> using US pseudopotentials. 
LC> 
LC> III) For Troullier-Martins kind of psps I get an error related to the KB
LC> projectors ( wrong beta functions ) for the non-spin polarized
LC> calculations. So it could be that it is possible to do the fully
LC> relativistic GGA in this case, but I must figure out first why the
LC> "wrong beta functions" error in this case.
LC> 
LC> Is that the actual situation or I am missing something?
LC> 
LC> I will do the same kind of tests for the scalar relativistic options
LC> soon.
LC> 
LC> Thanks, L.
LC>  
LC> 
LC> >It is possible but not well tested. It should work well 
LC> >for nonmagnetic systems, while it could have 
LC> >problems for magnetic systems. In any case if you find
LC> >strange behavior please report it to the forum.
LC> >
LC> >Andrea
LC> >
LC> >
LC> >On Wed, 2007-04-18 at 16:00 +0200, Lazaro Calderin wrote:
LC> >> Hi guys, 
LC> >> 
LC> >> Is it possible to do relativistic GGA calculations with pwscf?
LC> >>         
LC> >> Thanks, Lazaro
LC> 
LC> 
LC> _______________________________________________
LC> Pw_forum mailing list
LC> Pw_forum at pwscf.org
LC> http://www.democritos.it/mailman/listinfo/pw_forum
LC> 

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