[Pw_forum] problem of convergence

[Cyrille Barreteau]

Grazie Mille Stefano. Quesaco local-TF?
What is exactly this local-TF mixing scheme?
Do you have any reference? It seems very efficient
for slabs (and probably clusters).

  cyrille



On Lun 5 décembre 2005 9:25, Stefano de Gironcoli a écrit :
> try changing mixing_mode to local-TF.
> It should damp charge sloshing.
>
>  stefano
>
> Cyrille Barreteau wrote:
>
>>Dear pwscf users and developpers,
>>
>>I am trying to do a pwscf calculation of a 100 iron surface.
>>I first performed a scf calculation on a 11 layers thick
>>slab without any difficulty. But to get a better description
>>of the surface I have tried to perform a calculation with
>>a thicker slab: 20 atomic layers. Then comes the difficulty:
>>the scf cycle never converges. After 46 iterations the
>>estimated scf accuracy is huge (see below).
>>Any hint to improve convergency?
>>  -decreasing mixing?
>>  -changing occupation
>>  -changing degauss...
>>
>>  thank you in advance
>>     Cyrille
>>
>>====================================================
>>     iteration # 46     ecut=    35.00 ryd     beta=0.30
>>     Davidson diagonalization with overlap
>>     ethr =  2.46E-03,  avg # of iterations =  6.7
>>
>>     total cpu time spent up to now is 228848.87 secs
>>
>>     total energy              = -1200.62214988 ryd
>>     estimated scf accuracy    <  4756.34124938 ryd
>>
>>     total magnetization       =    10.95 Bohr mag/cell
>>     absolute magnetization    =    11.56 Bohr mag/cell
>>=====================================================
>>
>>Here is the input file:
>>
>>==============================================
>> &control
>>    calculation='scf'
>>    title='Fe 001 20 layers +   vacuum'
>>    restart_mode='from_scratch',
>>    prefix='fe_surf_001_gga'
>>    pseudo_dir = '/home/barreto/SOFTWARE/pseudo/',
>>    outdir='tmp'
>> /
>> &system
>>    ibrav = 6, celldm(1) = 5.42351511431617127800 , celldm(3)=13,
>>     nat= 20, ntyp= 1,nbnd=150,nspin=2,
>>    ecutwfc = 35.0, ecutrho=250.,starting_magnetization(1)=0.3,
>>    occupations='smearing', smearing='methfessel-paxton', degauss=0.025
>> /
>> &electrons
>>    conv_thr =  1.0d-8
>>    mixing_beta = 0.3
>> /
>>ATOMIC_SPECIES
>> Fe 55.847 Fe_us_gga_d2.1.8.pseudo.UPF
>>ATOMIC_POSITIONS
>>Fe   0.000000000000000  0.000000000000000   -5.25000000000000
>>Fe   0.500000000000000  0.500000000000000   -4.75000000000000
>>Fe   0.000000000000000  0.000000000000000   -4.25000000000000
>>Fe   0.500000000000000  0.500000000000000   -3.75000000000000
>>Fe   0.000000000000000  0.000000000000000   -3.25000000000000
>>Fe   0.500000000000000  0.500000000000000   -2.75000000000000
>>Fe   0.000000000000000  0.000000000000000   -2.25000000000000
>>Fe   0.500000000000000  0.500000000000000   -1.75000000000000
>>Fe   0.000000000000000  0.000000000000000   -1.25000000000000
>>Fe   0.500000000000000  0.500000000000000   -0.75000000000000
>>Fe   0.000000000000000  0.000000000000000   -0.25000000000000
>>Fe   0.000000000000000  0.000000000000000    0.25000000000000
>>Fe   0.500000000000000  0.500000000000000    0.75000000000000
>>Fe   0.000000000000000  0.000000000000000    1.25000000000000
>>Fe   0.500000000000000  0.500000000000000    1.75000000000000
>>Fe   0.000000000000000  0.000000000000000    2.25000000000000
>>Fe   0.500000000000000  0.500000000000000    2.75000000000000
>>Fe   0.000000000000000  0.000000000000000    3.25000000000000
>>Fe   0.500000000000000  0.500000000000000    3.75000000000000
>>Fe   0.000000000000000  0.000000000000000    4.25000000000000
>>Fe   0.500000000000000  0.500000000000000    4.75000000000000
>>Fe   0.000000000000000  0.000000000000000    5.25000000000000
>>K_POINTS automatic
>> 14 14 1 0 0 0
>>==========================================================
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>