[Solved] gwcalctyp 09: Renormalization factor

GW, Bethe-Salpeter …

Moderators: maryam.azizi, bruneval

Locked
gabriel.antonius
Posts: 58
Joined: Mon May 03, 2010 10:34 pm

[Solved] gwcalctyp 09: Renormalization factor

Post by gabriel.antonius » Sat May 18, 2013 4:45 pm

Hi every one,
When making a one-shot GW calculation with contour deformation, I find that with gwcalctyp 09,
the renormalization factor used is 1.0000
Is that a bug? Is there any purpose in this?
And what is the difference anyway between gwcalctyp 02 and 09? It is not documented so well, and there is no automatic test with gwcalctyp 09 !

Here is an abipy input with silicium

Code: Select all

"""One-shot GW calculations on silicon."""
from abipy.htc import Launcher

calc = Launcher('02-si/calc')

calc.set_bindir('/Users/Antonius/Work/Software/abinit/7.2.2-private/build2/src/98_main/')
calc.set_pseudodir('/Users/Antonius/Work/Software/abinit/7.2.2-private/tests/Psps_for_tests/')
calc.set_pseudos('14si.pspnc')

# Bands
nband_occ = 4
nband_unocc = 12
nband_tot = nband_occ + nband_unocc
nband_buff = nband_tot / 10
nband_extra = nband_tot + nband_buff

groundstate = {
    'tolvrs' : 1e-2,
    'iscf' : 7,
    'ixc' : 7,
    'optforces' : 0,
    'nband' : 3 * nband_occ,
    'nbdbuf' : nband_occ,
    'kptopt' : 1,
    'ngkpt' : 3 * [2],
    'nshiftk' : 4,
    'shiftk' : [[0.5,0.5,0.5],
                [0.5,0.0,0.0],
                [0.0,0.5,0.0],
                [0.0,0.0,0.5]],
    }

wavefunctions = {
    'iscf' : -2,
    'getden' : -1,
    'tolwfr' : 1e-2,
    'nband' : nband_extra,
    'nbdbuf' : nband_buff,
    }

screening = {
    'optdriver' : 3,
    'irdwfk' : 1,
    }

sigma = {
    'optdriver' : 4,
    'nkptgw' : 1,
    'kptgw' : [0,0,0],
    'bdgw' : [nband_occ-2, nband_occ+7],
    'irdwfk' : 1,
    'irdscr' : 1,
    }

gw_options = {
    'gwcalctyp' : '09',
    'awtr' : 1,
    'symchi' : 1,
    'symsigma' : 0,
    'gwpara' : 2,
    'nfreqim' : 2,
    'nfreqre' : 4,
    'freqremax' : '40 eV',
    }

ecut = 5.
planewaves = {
    'ecut' : ecut,
    'ecutwfn' : ecut,
    'ecutsigx' : 2 * ecut,
    'ecuteps' : 2.5,
    'nband' : nband_tot,
    }

kpoint_grid_unshifted = {
    'kptopt' : 1,
    'ngkpt' : 3 * [2],
    'nshiftk' : 1,
    'shiftk' : [0,0,0],
    }

silicon_cell = {
    'acell' : 3*[10.217],
    'rprim' : [[0.0,0.5,0.5],
               [0.5,0.0,0.5],
               [0.5,0.5,0.0]],
    'ntypat' : 1,
    'znucl' : 14,
    'natom' : 2,
    'typat' : [1,1],
    'xred' : [3*[0.0], 3*[0.25]]
    }

options = {
    'istwfk' : '*1',
    'enunit' : 1,
    'prtvol' : 10,
    'npfft' : 1,
    }


calc.ndtset = 4
calc.set_variables(groundstate, 1)
calc.set_variables(wavefunctions, 2)
calc.set_variables(screening, 3)
calc.set_variables(sigma, 4)

calc.set_variables(gw_options)
calc.set_variables(kpoint_grid_unshifted)
calc.set_variables(planewaves)
calc.set_variables(silicon_cell)
calc.set_variables(options)

calc.link_io(3, 2, 'wfk')
calc.link_io(4, 2, 'wfk')
calc.link_io(4, 3, 'scr')

calc.make()
calc.run()
print calc.report()
Last edited by gabriel.antonius on Sun May 19, 2013 4:20 pm, edited 1 time in total.
Gabriel Antonius
Université du Québec à Trois-Rivières

User avatar
gmatteo
Posts: 291
Joined: Sun Aug 16, 2009 5:40 pm

Re: gwcalctyp 09: Renormalization factor

Post by gmatteo » Sun May 19, 2013 12:35 pm

Hi Gabriel,

I find that with gwcalctyp 09,
the renormalization factor used is 1.0000
Is that a bug? Is there any purpose in this?


I would say that it is done on purpose since the factor Z is not used
in this particular case.
gwcalcty x9, indeed, corresponds to the model self-energy of Faleev that is static (see PRL 93, 126406 (2004)).
The QP correction are obtained by diagonalizing the model self-energy instead of using the standard perturbative
approach that requires the computation of Z.

And what is the difference anyway between gwcalctyp 02 and 09? I


x2 corresponds to the "true" GW approximations (non-hermitian, frequency dependent), whereas x9
corresponds to the model self-energy of Faleev (static, Hermitian)

and there is no automatic test with gwcalctyp 09 !


There are tests for gwcalctyp 28 and 29. I would say that using gwcalctyp 09 does not make much sense
since, in Faleev's method, the model self-energy is always solved self-consistently.

Locked