[Pw_forum] symmetry breaking in QE and VASP
himm0013 at umn.edu
Tue Jan 29 08:38:11 CET 2013
The orbital ordered states can be found (if they exist) by using the
"starting_ns_eigenvalue" variable in the &system namelist. This
suggests the desired orbital occupations when the default choice takes
another path. If such an orbitally ordered phase exists, then you can
compare its energy to the phase where orbital ordering does not occur.
Usually, the orbital ordered state will have lower energy.
This is related to the fact that DFT+U leads to several local energy minima
and sometimes one needs to use physical intuition to find them.
An example of this situation is given for FeO in the tutorials, for example
A similar situation probably occurs also for LTO.
University of California at Santa Barbara
On Mon, Jan 28, 2013 at 6:43 PM, Hanghui Chen <chenhanghuipwscf at gmail.com>wrote:
> Dear QE developers,
> This is a comment rather than a question.
> In QE, if the system has some symmetry (say cubic symmetry), and even
> if you intentionally turn off the symmetry (set nosym = .TRUE.), the
> resulting self-consistent charge density in fact still has the original
> cubic symmetry (i.e. three 3 t2g orbitals have the occupancy and so do the
> two eg orbitals). Interestingly, if you do the same procedure in VASP, even
> the atomic structure has the cubic symmetry, the symmetry could be
> spontaneously broken in the self-consistent charge density. This is
> particular the case when the Hubbard U is turned on. Because of this, it is
> extremely difficult to stabilize an insulating LaTiO3, since the three t2g
> orbitals always have the same occupancy in QE. But in VASP, the symmetry is
> spontaneously broken and one t2g orbital is largely occupied and an
> insulating ground state (d^1 configuration) can be stabilized with a
> reasonable U (~7 eV).
> I am wondering whether there is some automatic symmetrization in
> charge density after each consistent iteration? And sometimes, when the
> physical system does have an orbital ordering, how can we correctly do the
> calculations (i.e. avoid missing the orbital-ordered state) with the
> rotationally invariant LDA+U in QE?
> Thank you very much.
> Dr. Hanghui Chen
> Department of Physics
> Columbia University
> Pw_forum mailing list
> Pw_forum at pwscf.org
Department of Materials
University of California at Santa Barbara, CA
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the users