[Wien] Okay to combine mBJ, spinorbit, and LDA+U?

Fri Aug 17 14:12:17 CEST 2012

What Fabien says is correct, but I am not sure that I completely agree with
everything. As he says, mBJ has a potential which gives fairly good band
structures for many oxides as it uses a highly level effective potential,
but has the disadvatage that this is not a functional derivative so the
energy is incorrect. The forces are consistent with the potential, so one
can in principle minimize the internal atomic positions.

Returning to the original question, is it better to use LDA+U, GGA+U or
PBE, WC or PBEsol to first optimize the atomic positions before using mBJ
or -so? That is a tricky question. I will argue that if the relaxed
positions in mBJ are a better match to those of LDA+U, then it is
reasonable to use that. Of course they might not be similar and the relaxed
mBJ positions might be nonsense, in which case you need to try and decide
from other evidence (xray refinements) what to use.

While not automated, with -so you can always do the forces by hand as a
numerical derivative to check that the positions are not stupid.

---------------------------
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
"Research is to see what everybody else has seen, and to think what nobody
else has thought"
Albert Szent-Gyorgi
 On Aug 17, 2012 4:12 AM, "tran at theochem.tuwien.ac.at" <
tran at theochem.tuwien.ac.at> wrote:

> Lattice relaxation and atomic positions relaxation are two different
> things. In WIEN2k there is no automatic lattice relaxation
> (the stress tensor would need to be implemented), i.e., the
> lattice constants a, b and c can only be optimized manually (with
> x optimize).
>
> The automatic (using forces) relaxation of atomic positions
> is possible only for functionals for which the forces are calculated
> correctly (GGA, GGA+U, onsite hybrids). This is not the case with
> spin-orbit coupling or the full hybrid functionals recently implemented.
>
> For mBJ, the problem is a little bit different. The forces are calculated
> correctly, but (in my opinion) are nonsense and should not be used
> for the atomic position relaxation. The problem is that mBJ is a potential
> for which there is no associated energy functional, and therefore the
> forces will be zero for a geometry which does not correspond to a
> minimum on a total-energy surface which does not even exist.
>
> My recommendation is to never use mBJ for structure optimization
> (lattice constant and atomic positions).
>
> F. Tran
>
> On Fri, 17 Aug 2012, Zhiyong Zhu wrote:
>
> > Dear Prof. Laurence Marks,
> >
> > I have a question about the force minimization in mBJ calculations.
> > According to the Userguide, the forces in mBJ calculations are not
> > correct, which means that the lattice relaxation using forces on atoms
> > in mBJ calculations is impossible. According to your description below,
> > however, it seems that the lattice relaxation is possible even in mBJ
> > calculations, if we use MSR1a in *.inm file. Is my understanding correct?
> >
> > Another question is about the force minimization in calculations with
> > spin-orbit coupling. According to the wien2k website
> > (http://www.wien2k.at/reg_user/limitations/), lattice relaxation in
> calculations
> > with spin-orbit coupling is also not possible, because the Pulay
> corrections
> > to the forces are not correct in the presence of spin-orbit coupling.
> > So my concern is that whether the MSR1a option in *.inm file can overcome
> > this problem?
> >
> > Thank you very much.
> >
> > Regards,
> > Zhiyong Zhu
> >
> > On 8/16/12, Laurence Marks <L-marks at northwestern.edu> wrote:
> > > If it is a decent insulator I would do LDA+U directly; often it
> > > converges better and since the lattice parameter and forces change you
> > > do not gain much by first doing LDA/GGA. Normally LDA+U is stable,
> > > often more stable that LDA/GGA. Volume optimization should be done
> > > first, then min_lapw or MSR1a. Better is to do MSR1a or min_lapw at
> > > each volume.
> > >
> > > Then add -so, mBJ as appropriate with the optimized positions.
> > > However, I not sure if mBJ+U is appropriate (I doubt that it is). It
> > > might be that LDA+U positions are a better approximation for mBJ, not
> > > sure. One way is to minimize the forces with mBJ using MSR1a (not
> > > min_lapw/PORT) and compare them to LDA+U. If they are the same then
> > > you are in good shape, needs testing. Maybe someone has....
> > >
> > > N.B., it is completely fine to minimize positions in mBJ using MRS1a
> > > -- do not use min_lapw/PORT, it will not be correct. MSR1a does not
> > > care that the energy is incorrect whereas min_lapw/PORT does.
> > >
> > > On Thu, Aug 16, 2012 at 2:44 PM, Jeff Spirko <spirko at lehigh.edu>
> wrote:
> > >> Is it okay to use spinorbit and LDA+U with mBJ?
> > >>
> > >> I would guess it is done like this:
> > >>  * Check that forces <10 mRy/au with plain LDA or GGA.  Reduce via
> > >> min_lapw.
> > >>  * Volume optimization (if desired) with plain LDA or GGA to reduce
> > >> absolute pressure.
> > >>  * Set up LDA+U (Sec 4.5.6) and use -orb flag from now on.
> > >>  * Need to converge LDA+U???
> > >>  * Follow mBJ instructions (Sec 4.5.9).
> > >>  * After mBJ+LDA+U is converged, follow spinorbit instructions (Sec
> > >> 4.5.5).
> > >>  * For spinpolarized, check whether atoms became nonequivalent
> > >> (affects case.inso, case.inorb,
> > >>     case.indmc, case.in1c, basically any input file with atom lists or
> > >> indices)
> > >>  * touch .fulldiag (necessary because klist can change???)
> > >>  * Do final run with -so -orb
> > >>
> > >> Best regards,
> > >> --
> > >> Jeff Spirko   spirko at lehigh.edu   WD3V   |=>
> > >>
> > >> The study of non-linear physics is like the study of non-elephant
> > >> biology.
> > >> _______________________________________________
> > >> Wien mailing list
> > >> Wien at zeus.theochem.tuwien.ac.at
> > >> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> > >
> > >
> > >
> > > --
> > > Professor Laurence Marks
> > > Department of Materials Science and Engineering
> > > Northwestern University
> > > www.numis.northwestern.edu 1-847-491-3996
> > > "Research is to see what everybody else has seen, and to think what
> > > nobody else has thought"
> > > Albert Szent-Gyorgi
> > > _______________________________________________
> > > Wien mailing list
> > > Wien at zeus.theochem.tuwien.ac.at
> > > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> > >
> >
> > --
> >
> > ------------------------------
> > This message and its contents, including attachments are intended solely
> > for the original recipient. If you are not the intended recipient or have
> > received this message in error, please notify me immediately and delete
> > this message from your computer system. Any unauthorized use or
> > distribution is prohibited. Please consider the environment before
> printing
> > this email.
> > _______________________________________________
> > Wien mailing list
> > Wien at zeus.theochem.tuwien.ac.at
> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> >
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20120817/0621958a/attachment.htm>