[gpaw-users] convergence issue with spin-polarised system

[Chengjun Jin]

Hi All,

Just a little feedback. 

By using smaller mixing like 0.05 does improve the convergence. 

Regards,
Chengjun

________________________________________
From: Ask Hjorth Larsen [asklarsen at gmail.com]
Sent: Saturday, November 16, 2013 6:40 PM
To: Chengjun Jin
Cc: gpaw-users
Subject: Re: [gpaw-users] convergence issue with spin-polarised system

Hello Chengjun

I think GPAW's mixers do not handle spin all that well.  The total
magnetic moment, even in the first step which converges, is still
changing rather significantly.  It is slightly strange that the
calculations is considered to have converged.

Try a more careful MixerSum(0.06, 5, 100.0) or so.

If this won't work you can perhaps use a higher Fermi temperature.  As
long as you are doing geometry optimization this is probably less
critical than it would be for a binding energy calculation.

(Consider parallelizing more over k-points, e.g. parallel={'domain':
(1, 2, 2), 'band': 1}.  In future versions you can do e.g.
parallel={'kpt': 8}.)

Best regards
Ask

2013/11/16 Chengjun Jin <chej at fysik.dtu.dk>:
> Dear All,
>
> I am trying to run a spin-polarised system, a slab with NO molecule on top. I have difficulty to converge the electroinic SCF loop.
> Attached are the traj and the output files.
>
> Any suggestion is apprecitated.
>
>
>
> The input files is below.
>
>   8 import numpy as np
>   9 from ase.visualize import view
>  10 from ase.io import read, write
>  11 from ase.optimize import QuasiNewton
>  12 from gpaw import GPAW, Mixer, MixerSum, FermiDirac
>  13 from ase.constraints import FixAtoms
>  14
>  15
>  16 system=read('scat.traj')
>  17 #initial_magnetic_moments on N and O is(0.6, 0.4)
>  18 system.set_pbc(111)
>  19 #view(system)
>  20
>  21 calc = GPAW(
>  22            mode='lcao',
>  23            basis={'Au':'dzp',  'O':'dzp','N':'dzp' }, #XXX
>  24            h=0.2,
>  25            xc='PBE',
>  26            kpts=(6,6,1),
>  27            occupations=FermiDirac(width=0.1),
>  28            txt='scat.out',
>  29            mixer=MixerSum(),
>  30            maxiter=1000,
>  31            usesymm=True
>  32            )
>  33
>  34 system.set_calculator(calc)
>  35
>  36 relax = QuasiNewton(system, trajectory= 'qn_scat.traj', logfile='qn_scat.log')
>  37 relax.run(fmax=0.01)
>
>
>
>
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users