The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear Abram,

After a quick examination of your input file, it seems that you should
at least change toldfe 1d-6 (a very loose value for geometry  
optimization) to
toldfe 1d-12, or even better, switch to the use of tolvrs (recommended).

If I remember correctly, Gian-Marco Rignanese tried to compute  
ferromagnetic Ni
several years ago with ABINIT (and succeeded), and he observed that  
the smearing of the
occupation function is critical. He needed a very small value of tsmear,
on the order of 0.001-0.005, no more (I do not remember whether it  
was with occopt 4, but more likely
with one of the cold smearings). For such a small value of tsmear,  
you might need an even
finer sampling of k points.

I guess more could be said with the availability of part of your  
output file ...

Xavier


On 12 Apr 2006, at 01:50, avanderg wrote:

> Hello All,
>
> I am trying to find the equlibrium lattice constant for  
> ferromagnetic FCC Ni.  Unfortunately I am running into several  
> problems:
>
> 1)the latice constant of the conventional cell becomes far too  
> large compared to experimental (abinit ->8.9367bohr ;experimental - 
> >6.6598 bohr).
>
> 2) this should be a 1 degree of freedom problem (optcell = 1) but  
> it takes about 13 hours to get the above result which seems like a  
> long time to me (and I have done convergence tests on things like  
> kpts and ecut).
>
> 3)if I use optcell = 2 (letting rprim be optimized as well as  
> acell), the system becomes tetragonal even after inputting a fcc  
> crystal. I wouldn't think anything would break the symmetry  and  
> the system would stay fcc.
>
> Thanks in advance for the help
>
> Abram Van Der Geest
>
>
> Input File:
>
> #Crystalline Nickel: looking for Eq acell
>
>
> kptopt   1
> kptrlatt -17 17 0
>              0 17 0
>              0 17 -17
> shiftk   0.0 0.0 0.0
>
> toldfe   1.0d-6
> getwfk   -1
>
> nsppol   2               #Ni is magnetic
> spinat  0.0 0.0 2.0  #breaking symmetry using Hund's rule
>
> #options for optimizing lattice parameters
> optcell 1
> ionmov  3
> ntime  25
> dilatmx 1.6  #abinit says anything less is too small
> ecutsm  0.5
>
> #Input Variables for both
> occopt    4
> tsmear  0.04
>
> nband  20
>
> #Def of unit cell
> acell     3*4.7092       #a of the primitive cell
> rprim     0.0 0.5 0.5    #FCC prim vectors
>          0.5 0.0 0.5
>          0.5 0.5 0.0
>
> #def of atom
> ntypat    1              #1 type of atom
> znucl     28             #and it is Ni
>
> natom     1              #1 atom
> typat     1              #what is the type of each atom
>
> xred      0.0 0.0 0.0    #location of each atom
> ixc       1              #LDA Teter Pade parametrization
> ecut      55             #50 = medium ecut
> diemac    1.0d6          #preconditions SCF
> nstep     20