[Pw_forum] convergence of sigma 5 GB

[Nicola Marzari]

Provided your cell looks good in xcrysden (i.e. no errors in using an
odd unphysical structure) I would start reducing the beta mixing to 0.2,
or 0.1, or 0.05 - see if any is sufficient to converge, and if so, which
is faster. Try using some of the more sophisticated mixing schemes.

Before doing that, note that your pseudopotential is ultrasoft, so
you need to have an ecutrho between 6 and 12 times larger than
ecutwfc (ultrasoft pseudopotentials allow for smoother wavefunctions,
but the charge density is what it is - it can't be made smoother.
so the usual relation ecutrho=4*ecutwfc, that is correct for a
rho=|wfc|^2, breaks if the wfc is smoother, and the real rho is
reproduced by |wfv|^2 + augmentation charges, as it happens in ultrasoft
pseudos).

Last, given that your unit cell is much longer around z, you should
sample with a 4 4 2 mesh, not a 4 4 4. Maybe a 2 2 1 mesh would
be good enough to get things started, and then you increase it at the
end for the last iterations around the converged geometry.

Note also that typically a 1 1 1 shift (i.e. 2 2 1 1 1 1, or
4 4 2 1 1 1) gives better sampling earlier - but something you could
test in a unit cell of 4 atoms only, with one atom displaced by 5-10%:
check how the forces converges to the true value as you increase the
sampling using n n n 0 0 0 or n n n 1 1 1 meshes.

				nicola


On 06/10/2014 19:52, Ravi Kiran wrote:
> Dear Quantum espresso users,
>
> I am looking to study the properties of sigma 5 grain boundary with 80
> atoms.I am using scf calculations for initial convergence. I am using
> the following script. The problem is that the values dont converge even
> after 100 iterations .Can anyone please tell me how can I converge the
> total energy in order to obtain relaxed atomic coordinates.
> &control
>      calculation='relax',
>      restart_mode='restart',
>      prefix='Ni',
>      tprnfor =.t.,
>      pseudo_dir = '/psuedo/',
>      outdir='/psuedo/tmp',
> /
>   &SYSTEM
>       nat =80,
>      ntyp = 1,
>      degauss = 0.02,
>      nspin = 2,
>      starting_magnetization = 0.62,
>      ecutwfc = 24,
>      ibrav = 0,
>      occupations = 'smearing',
>      smearing = 'mv',
>   /
>   &ELECTRONS
>         conv_thr = 1e-6,
>      mixing_beta = 0.3,
>      /
> &IONS
>   /
> &CELL
> /
>
> ATOMIC_SPECIES
>   Ni 58.69d0 Ni.pz-nd-rrkjus.UPF
> K_POINTS automatic
>     4 4 4   0 0 0
> ATOMIC_POSITIONS
> Ni  0.000000  0.000000  0.500000
> Ni  0.000000  0.000000  0.250000
> Ni  0.400000  0.200000  0.500000
> Ni  0.400000  0.200000  0.250000
> Ni  0.200000  0.600000  0.500000
> Ni  0.200000  0.600000  0.250000
> Ni  0.800000  0.400000  0.500000
> Ni  0.800000  0.400000  0.250000
> Ni  0.600000  0.800000  0.500000
> Ni  0.600000  0.800000  0.250000
> Ni  0.100000  0.300000  0.500000
> Ni  0.100000  0.300000  0.250000
> Ni  0.700000  0.100000  0.500000
> Ni  0.700000  0.100000  0.250000
> Ni  0.500000  0.500000  0.500000
> Ni  0.500000  0.500000  0.250000
> Ni  0.300000  0.900000  0.500000
> Ni  0.300000  0.900000  0.250000
> Ni  0.900000  0.700000  0.500000
> Ni  0.900000  0.700000  0.250000
> Ni  0.500000  0.000000  0.375000
> Ni  0.500000  0.000000  0.125000
> Ni  0.300000  0.400000  0.375000
> Ni  0.300000  0.400000  0.125000
> Ni  0.100000  0.800000  0.375000
> Ni  0.100000  0.800000  0.125000
> Ni  0.900000  0.200000  0.375000
> Ni  0.900000  0.200000  0.125000
> Ni  0.700000  0.600000  0.375000
> Ni  0.700000  0.600000  0.125000
> Ni  0.200000  0.100000  0.375000
> Ni  0.200000  0.100000  0.125000
> Ni  0.000000  0.500000  0.375000
> Ni  0.000000  0.500000  0.125000
> Ni  0.600000  0.300000  0.375000
> Ni  0.600000  0.300000  0.125000
> Ni  0.400000  0.700000  0.375000
> Ni  0.400000  0.700000  0.125000
> Ni  0.800000  0.900000  0.375000
> Ni  0.800000  0.900000  0.125000
> Ni  0.000000  0.000000  0.750000
> Ni  0.000000  0.000000  1.000000
> Ni  0.600000  0.200000  0.750000
> Ni  0.600000  0.200000  1.000000
> Ni  0.200000  0.400000  0.750000
> Ni  0.200000  0.400000  1.000000
> Ni  0.800000  0.600000  0.750000
> Ni  0.800000  0.600000  1.000000
> Ni  0.400000  0.800000  0.750000
> Ni  0.400000  0.800000  1.000000
> Ni  0.300000  0.100000  0.750000
> Ni  0.300000  0.100000  1.000000
> Ni  0.900000  0.300000  0.750000
> Ni  0.900000  0.300000  1.000000
> Ni  0.500000  0.500000  0.750000
> Ni  0.500000  0.500000  1.000000
> Ni  0.100000  0.700000  0.750000
> Ni  0.100000  0.700000  1.000000
> Ni  0.700000  0.900000  0.750000
> Ni  0.700000  0.900000  1.000000
> Ni  0.800000  0.100000  0.625000
> Ni  0.800000  0.100000  0.875000
> Ni  0.400000  0.300000  0.625000
> Ni  0.400000  0.300000  0.875000
> Ni  0.000000  0.500000  0.625000
> Ni  0.000000  0.500000  0.875000
> Ni  0.600000  0.700000  0.625000
> Ni  0.600000  0.700000  0.875000
> Ni  0.200000  0.900000  0.625000
> Ni  0.200000  0.900000  0.875000
> Ni  0.500000  0.000000  0.625000
> Ni  0.500000  0.000000  0.875000
> Ni  0.100000  0.200000  0.625000
> Ni  0.100000  0.200000  0.875000
> Ni  0.700000  0.400000  0.625000
> Ni  0.700000  0.400000  0.875000
> Ni  0.300000  0.600000  0.625000
> Ni  0.300000  0.600000  0.875000
> Ni  0.900000  0.800000  0.625000
> Ni  0.900000  0.800000  0.875000
> CELL_PARAMETERS
> 14.87  0     0
> 0    14.87   0
> 0     0    26.60
>
>
>
> --
> Thanks
> Ravi
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>

-- 

----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL