The ABINIT Users Mailing List ( CLOSED )

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

Dear Ravi,

Larger kptrlen means more k-point samplings,
so it should produce more converged result.
So, energy convergence can be one check-point.
(Relative energies (i.e. energetics) will have faster convergence
 with respect to k-points than absolute total energies.)

K-point sampling may also be relevant to geometry optimization.
When I calculated a Urea crystal (No.113 , P-421M),
Gamma-point sampling was bad for geometry optimization.
K-point sampling was needed to have a plausible geometry
comparable with an experimental geometry.
(I am afraid that I do not have more detailed specific data, but
 it was performed for testing purpose a long long ago,
 so I cannot find the log immediately.
 So the above description is based on my memory only ... )

Best wishes,
Masayoshi

On 2004/04/07, at 5:51, rabrol@us.ibm.com wrote:

> Hi AbInit Users,
> I am working with molecular crystals and
> am trying to setup a k-grid.
> I used prtkpt=1 option, so that abinit
> can suggest a kptrlatt and shiftk for me.
> Here's the input file i used:
>
> -------------------input file begins---------
> ndtset 1
>
> ecut 10.0         # Maximal kinetic energy cut-off, in Hartree
>
> kptopt 1          # Option for the automatic generation of k points,
>                   # taking into account the symmetry
>
> prtkpt 1
>
> toldfe  1.0d-6
>
>
> #Definition of the unit cell
> acell 8.098  5.953  8.652  angstrom
> angdeg 90  124.4  90
>
> #Definition of the atom types
> ntypat 2          # There are two types of atoms
> znucl 6 1         # 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 types are carbon 
> and hydrogen
>
>
> #Definition of the atoms
> natom 36          # There are two atoms
> typat 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 1  1 2 1 2 1 2 1 2 1 2 1 2 1 2 
> 1 2 1 1
>                   # They are of type 1 or 2, that is, carbon or 
> hydrogen respectively
> xangst
> .............................coordinates for 36 atoms go here
> ------------------------end of input file-----------------------------
>
> here are the three choices suggested by abinit on output:
>
> -----------------------------output file------------------------------
>  Grid#    kptrlatt         shiftk         kptrlen       nkpt  iset
>
>     2     1   0   0      5.0000E-01      1.4792E+01       1     1
>           0   2   0      5.0000E-01
>           1   0   1      5.0000E-01
>    20     1   1   1      0.0000E+00      1.6350E+01       2     5
>           2   0   1      0.0000E+00
>           1   1   0      0.0000E+00
>
>    21     1   2   1      0.0000E+00      2.5271E+01       3     5
>           2   0   1      0.0000E+00
>           1   2   0      0.0000E+00
>
>  List of best grids, ordered by nkpt.
>   (stop at a value of kptrlen 20% larger than the target value).
>   (the merit factor will tend to one or two in 3 dimensions)
>   (and to one, two or four in 2 dimensions)
>
>     nkpt   kptrlen    grid#  merit_factor
>      1    1.4792E+01     2      0.9853
>      2    1.6350E+01    20      0.6654
>      3    2.5271E+01    21      1.6379
>
>  For target kptrlen=    2.0000E+01, the selected grid is number    21,
>      giving kptrlen=    2.5271E+01 with nkpt=     3
> ------------------------end of output file--------------------------
>
> abinit uses default value of kptrlen = 20 to select grid # 21 out of 
> the
> three best grids. how should i determine what value of kptrlen to use
> as my k-grid selection will depend on that.
> the description of kptrlen input variable is not clear on this.
>
> thanks,
> ravi