[Pw_forum] Two Fermi energies
stefano de gironcoli
degironc at sissa.it
Sat Sep 19 00:35:27 CEST 2015
the starting magnetization option is used to break the up/dw symmetry
in the electronic states at the first scf iteration and select which
magnetic configuration is considered (you can suggest a ferromagnetic
configuration or various type of antiferromagnetic ones depending on how
you define the starting_magnetization variables).
The code then goes on seeking a self consistent solution without
forcing the value of the total magnetization and occupying each level
according to its position w.r.t. a single Fermi energy.
Usually (there is no guarantee, but it usually happens) the final
magnetic configuration is the same (ferromagnetic/antiferromagnetic) as
the one initially suggested and, for a given magnetic configuration, the
particular values of the starting_magnetization variables do not affect
the final self consistent result (but may affect how many iterations are
necessary to reach self-consistency).
How many electrons end up to be up or dw is determined by the self
If the system does not want to be magnetic, no matter which
starting_magnetization you chose, the final result would be non magnetic
The total_magnetization option corresponds to a different situation in
which you force the system to have a given total magnetization
(therefore a given number of Nup and Ndw electrons)
This is imposed by defining two distinct Fermi energies (one for up and
one for dw electrons) and if they end up to be different it can be
interpreted as having put the system in a uniform magnetic field.
Another option is to force the atomic magnetizations (defined by
integrals in spheres surrounding the atoms) to have certain values.
Check the constrained_magnetization details.
This does not impose a different value of Nup/Ndw. It produces only one
On 18/09/2015 23:07, Ary Junior wrote:
> I'm using the variable starting_magnetization(#) instead of
> tot_magnetization in a spin-polarized calculation (nspin = 2), with
> smearing and degauss > 0.0 (lgauss is .true.)... The result is that
> for some starting magnetization values, there are more bands occupied
> with spin up than spin down, with partial occupations at the top of
> the valence band... In that case, I was expecting to see the values of
> the variables ef_up and ef_dw in the output, instead of only the value
> of the variable ef...
> My question is: why in the file PW/src/input.f90 the variable
> two_fermi_energies is .true. only if tot_magnetization /= -1._DP? In
> the calculation described above, the subroutine weights
> (PW/src/weights.f90) calls the subroutine gweights
> (PW/src/gweights.f90) only once, and the variable ef is set with the
> bisection method implemented in the file PW/src/efermig.f90....
> Shouldn't it call gweights twice for ef_up and ef_dw?
> Thank you very much!
> Ary Ferreira
> CNPq postdoctoral fellow
> TU München
> Pw_forum mailing list
> Pw_forum at pwscf.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the users