Dear Guanping

You are right in both cases. In SIESTA, the MeshCutoff does not
control the basis, only the fineness of its proyection in the
operations that are done in a cartesian mesh in real space. Usually,
you get a fast convergence with the mesh, (I know no case
non-pathological case with a mesh bigger than 1000 Ry, and usually
even 400-500 is enough). Magnetism is usually sensitive, so be careful
to check total convergence of both the energy and the totalspin/
mulliken charges(spins)...

Also, the finer the mesh, the bigger the DM file. There is another
parameter, MeshSubDivisions, which may help you there.

About the radii, it's a bit more complicate. Look at
http://www.iop.org/EJ/abstract/0953-8984/16/30/008 .

Regards

On Wed, Mar 10, 2010 at 07:14, Guangping Zhang <[email protected]> wrote:
> DEAR Lucas,
>
> First ,thank you for your attention to my question again.Now after several
> readings ,I get some concept of the PAO.EnergyShift.
>
> In ideal,we use a basis set with extending to infinite, but  for reality
> calculation and in fact,the orbitals are zero or near zero,so for
> efficiency,we set the orbitals to be zero beyond some place named rc
> (orbital-confining cutoff radii).We provide the PAO.EnergyShift to determine
> the value of the rc for the orbitals with first-zeta used in the basis
> set,Thus ,the rc's are different for different atom species! For
> multi-zeta,the second-zeta ,the third-Dear zeta orbits and so on are 
> determined
> using PAO.SplitNorm.I.E. the second zeta of the second-zeta orbit carries
> the half of the PAO.SplitNorm norm of the whole orbit,and the third carries
> the one fourth of the PAO.SplitNorm of the whole orbit,the split radius is
> defined accordingly.The larger PAO.EnergyShift the smaller the rc ,and the
> more efficient but the less accurate!
>
> Am I right ?Do I understand correctly?
>
> But for MeshCutoff,I only know it defines the plane wave cutoff for the
> grid.Then I turned to the plane wave cutoff, some book said the larger the
> plane wave cutoff,the more plane waves used in the calculation and the lee
> efficient but the more accurate the result!But for siesta calcilation ,the
> basis set is fix at the beginning independent of the MeshCutoff.
>
> In your reply ,you said :'MeshCutoff controls the number of points in 3D
> grid to represent, say, the charge density of your system.
> The bigger the MeshCutoff the more points you have and calculation is closer
> to "converged".'
>
> So the bigger the MeshCutoff ,the denser points we get.But it does not
> affect the profile of the whole point grid,only the density!
>
> Am I right ?
>
> Respect to your kind reply!
>
>
>
> ------------------
> BEST REGARDS!
>
> Guangping Zhang
> -----------------------------------
> Atom and Melecular Physics
> Physics and Electronics College
> Shandong Normal University
> Shandong,Jinan,China
> ------------------------------------



-- 
Lucas Fernández Seivane
Ph. D. Candidate
Universidad de Oviedo - CINN

Responder a