The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@rc.m-kagaku.co.jp>]

Hi,

I am not familiar with metal cluster cases, but
my  first guess based on my experience with metal slabs  is like this:

- try "iprcel 45" (or 55,  65 ...).
        http://www.abinit.org/Infos_v4.6/vargs.html#iprcel
  This option has been discussed sometimes about metallic slabs.
  You might want to find some e-mails in the forum archive.

- increase nbands ?
   Ag has 11 electrons (4d^10 + 5s^1), so your cluster has 11*8=88
   electrons, so at least 44 bands are occupied, and I am not sure that
   only 6 empty bands are enough. You might want to plot DOS to see
   the situation.

- "trial and error" with diemac ?

- anything else ? (to forum readers)

In any case, let us enjoy the process of "trial and error" and
share our experience. We may say, following Euclid,
        "There is no royal road to convergence". ;-)

Good continuation and a good weekend,
Masayoshi

On 2005/11/10, at 4:06, tlahuicef@yahoo.com wrote:

> Hello guys.
> I got a big problem here.
> I want to do a metallic cluster optimization, but it seems is not ok  
> at the moment.
> So any advice good be great.
>
> #Ag Cluster optimization using BFGS.
>
> ionmov  3              # Use the modified Broyden algorithm
> ntime   100             # Maximum number of Broyden "timesteps"
> tolmxf  1.0d-5         # Stopping criterion for the geometry  
> optimization : when
>                        # the residual forces are less than tolmxf, the  
> Broyden
>                        # algorithm can stop
> occopt 4
> tsmear 0.04
> xcart 5.000000000 4.301809057 7.177053398
> 7.177053398 5.698190943 5.000000000
> 5.000000000 7.643092678 6.677619295
> 2.822946602 5.698190943 5.000000000
> 5.000000000 7.643092679 3.322380705
> 3.322380705 2.356907321 5.000000000
> 5.000000000 4.301809057 2.822946602
> 6.677619295 2.356907322 5.000000000
>
> toldff  1.0d-6         # Will stop the SCF cycle when, twice in a row,
>                        # the difference between two consecutive  
> evaluations of
>                        # forces differ by less than toldff (in  
> Hartree/Bohr)
> #toldff debe ser 10 veces menor que tolmxf!
>
> nband 50            # Only one band is occupied. In order to get the  
> energy,
>                      # forces ..., there is no need to compute more  
> than one band.
>
> #---------------------------------------------------------------------- 
> ---------
> #The rest of this file is similar to the t11.in file, except
> #that xcart has been moved above, and that toldfe has been replaced
> #by toldff, see above.
>
> nsym 1
> #Definition of the unit cell
> acell 10 10 10    # The keyword "acell" refers to the
>                   # lengths of the primitive vectors (in Bohr)
>
> #Definition of the atom types
> ntypat 1          # There is only one type of atom
> znucl 47          # The keyword "znucl" refers to the atomic number of  
> the
>                   # possible type(s) of atom. The pseudopotential(s)
>                   # mentioned in the "files" file must correspond
>                   # to the type(s) of atom. Here, the only type is  
> Hydrogen.
>
>
> #Definition of the atoms
> natom 8           # There are two atoms
> typat 8*1         # They both are of type 1, that is, Hydrogen
>
> #Definition of the planewave basis set
> ecut 20.0        # Maximal kinetic energy cut-off, in Hartree
>
> #Definition of the k-point grid
> kpt 0 0 0
> nkpt 1            # Only one k point is needed for isolated system,
>                   # taken by default to be 0.0 0.0 0.0
>
> #Definition of the SCF procedure
> #iscf 5 # Algorithm for SCF
> nstep 250        # Maximal number of SCF cycles
> #toldfe is no more defined, as toldff is used above...
> diemac 2      # Although this is not mandatory, it is worth to
>                   # precondition the SCF cycle. The model dielectric
> diemix 0.5        # function used as the standard preconditioner
>                   # is described in the "dielng" input variable  
> section.
>                   # Here, we follow the prescriptions for molecules
>                   # in a big box