The ABINIT Users Mailing List ( CLOSED )

["Matteo Giantomassi" <Matteo.Giantomassi@uclouvain.be>]

> hi
>
> i would appreciate if someone could explain how to set up the
> self-consistency
> parameters in GW:
>

 Dear Manuel,

> 1. the mixing parameter for the density

 rhoqpmix should be changed only if you are experiencing problems
 in converging the self-consistent GW calculation, in
 particular when the matrix elements of the Hartree potential
 start oscillating during the self-consistent run.
 Generally, the default value (1.0) works pretty well.

> 2. the final criterium for convergence (quasiparticle energies
> invariance?)

 At present, there's not convergence criterion that can be used to stop
 the self-consistent (SC) procedure.
 One should manually inspect the QP results obtained at each iteration
 in order the understand whether the calculation reached convergence or not.

 There are two quantities defining the degree of convergence of the SC GW
run.
 The most important one is DeltaE that gives the change of the QP energy
 with respect to the previous step.
 DeltaE should be as small as possible, possibly for all bands specified
by bdgw.

 Another quantity worth checking is the difference between E(N)_pert and
E(N)_diago
 E_diago is obtained by diagonalizing the self-energy operator
 in the QP representation while E_pert is obtained by using
 the standard perturbative approach that only involves the
 diagonal matrix elements of Sigma.
 At convergence, the values reported in the two columns should be
 very close as the self-energy in the (converged) QP basis set is almost
diagonal

 Best Regards
 Matteo Giantomassi