The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

PS:
Just forgot to mention the following: to get an accurate extrapolation
of the dielectric matrix you need som empty bands. So if the problem
comes just from inaccurate preconditioning you may want to add from 4
to 10 conduction bands in you calculation.
PMA

On 2/2/07, Anglade Pierre-Matthieu <anglade@gmail.com> wrote:
> Hi,
>
> 1) Have you tested non default mixing: iscf=7 or 2 (with eventually
> small diemix ?)
> 2) for some such systems we have witnessed that state degeneracy may
> prevent convergence. This usually can be solved with electronic
> temperature. Sometimes tsmear as low as 30 Kelvin is enough. Have you
> tryed this ?
> 3)Note that it may have been useful to include the output from the 10
> first iterations. because It may tell if the preonditioner works
> correctly. Did it ?
> 3') You may try to tune the computation of the dielectric matrix with
> dielam and diegap (eventually diecut too). If you do so I'm interested
> in being informed of potential accuracy improvements you will get.
> 3") Since you have no real vacuum you may also want to use iprcel 6X
> instead of 4X. If this really helps, please let me know.
>
> 4) (not related with previous remarks) Your cell looks small for just
> one kpt. Is really converged enough ?
>
> regards.
>
> PMA
>
> On 2/2/07, zmq.lyao@gmail.com <zmq.lyao@gmail.com> wrote:
> > Dear Abinit users,
> >      I encounter a problem when I do a defect calculation. When I perform a 
> structure optimization of a supercell with a vacancy, the SCF can't converge. 
> The force and energy always oscillate and don't decrease. Because of this, the 
> structure optimization can't continue. I have tried the iprcel and diemix 
> parameter, but it still don't work. Anybody has some suggestions to solve this 
> problem? Thanks in advance.
> >
> > Regard,
> > Zeming
> >
> > The input files is as following :
> > ----------------
> > optcell 0
> > ionmov  2
> > ntime  100
> > nstep  200        # Maximal number of SCF cycles
> > dilatmx 1.05
> > ecutsm  0.5
> > toldff = 1.0D-4
> > tolmxf = 1.0D-3
> > prtgeo 1
> > prtden 0
> > prtwf 0
> >
> > iprcel 49
> > diemix 0.7
> > ecut 70
> >
> > #Definition of the unit cell
> > acell 14.703085  14.703085  22.539888       # 2X2X2
> > angdeg 90 90 120
> > chkprim 0
> > brvltt 0
> > #Definition of the atom types
> > ntypat 2
> > znucl 57 8
> > natom 39
> > typat 8*1 7*2 8*1 16*2
> > xangst
> >  1.123   1.945   1.455
> >  1.123   1.945   7.418
> >  1.123   5.835   1.455
> >  1.123   5.835   7.418
> >  4.492  -0.000   1.455
> >  4.492  -0.000   7.418
> >  4.492   3.890   1.455
> >  4.492   3.890   7.418
> >  1.123   1.945   3.856
> >  1.123   1.945   9.820
> >  1.123   5.835   3.856
> >  1.123   5.835   9.820
> >  4.492  -0.000   9.820
> >  4.492   3.890   3.856
> >  4.492   3.890   9.820
> >  2.246   0.000   4.509
> >  2.246   0.000  10.473
> >  2.246   3.890   4.509
> >  2.246   3.890  10.473
> >  5.615  -1.945   4.509
> >  5.615  -1.945  10.473
> >  5.615   1.945   4.509
> >  5.615   1.945  10.473
> >  2.246   0.000   2.107
> >  2.246   0.000   8.071
> >  2.246   3.890   2.107
> >  2.246   3.890   8.071
> >  5.615  -1.945   2.107
> >  5.615  -1.945   8.071
> >  5.615   1.945   2.107
> >  5.615   1.945   8.071
> >  0.000   0.000   0.000
> >  0.000   0.000   5.964
> >  0.000   3.890   0.000
> >  0.000   3.890   5.964
> >  3.369  -1.945   0.000
> >  3.369  -1.945   5.964
> >  3.369   1.945   0.000
> >  3.369   1.945   5.964
> >
> >
> > #Definition of the k-point grid
> > kptopt 0
> > kpt 0 0 0
> >
> >
> >
>
>
> --
> Pierre-Matthieu Anglade


--
Pierre-Matthieu Anglade