Dear Charlie,
I join Martin's recommendation to try LDA+U. By looking at the
non-integer total magnetic moment per unit cell 14.88 muB, I guess you
obtained (semi)metallic state, which might or might not be what you
want. Pure GGA sometimes leads to metallic solution for iron oxides
(even for those they are good insulators). I don't know how pronounced
the semi-metallic character should be for greigite, probably much more
than for magnetite, but even then the value of 14.88 muB looks quite far
from 16.00 muB. (For magnetite you should reach this integer value even
without LDA+U.)
Anyway, the LDA+U approach will also help you stabilizing the desired
spin structure quite efficiently. I would be surprised if the moments
flipped when you have some U (a few eV) applied to 3d states of Fe. In
fact, you can use LDA+U just to get close to desired spin structure and
then try to remove it ....
And additional note: when reducing the symmetry of magnetite be aware of
the fact that it is a mixed-valence compound (and I guess greigite might
be similar case) and has a tendency to localize the minority electron
within the octahedral sublattice. When you split the eight octahedral Fe
into four sorts they can readily differentiate to Fe2+ and Fe3+. (Even
in the cubic phase.) Which again may or may not what you want to have in
your model. (With +U you will definitely get the Fe2+ / Fe3+ scenario,
unless you keep all octahedral Fe equivalent.)
Regards,
Vojtech
On 17-Apr-19 12:00, Penny, Charles wrote:
Dear all,
Thanks for your prompt and helpful replies.
I have run runfsm initially before switching to runsp, but no luck. It
immediately converges back to the ferrimagnetic solution.
This raises two further thoughts/questions in my mind.
1) I am assuming that runfsm is NOT an acceptable replacement for
runsp with regards to projecting the total energy onto the Heisenberg
model. In any case I obtain a large moment in the interstitial region
which likely invalidates such a move.
2) I find it interesting that this process works without problem for
Fe3O4, but seems presently rather futile with Fe3S4. Simply swapping O
for S in the lattice (not quite, of course) appears to create such
instability in alternative spin configurations. Is anyone aware of
anything I could look into which might be able to explain this, or is
it just a case of 'bad luck' for this material and move on if I can't
get any improvement?
Re: Magnetite
An interesting material without a doubt. Luckily I'm not investigating
the low temperature phase, otherwise I might have more of a headache!
Regards,
Charlie
------------------------------------------------------------------------
*From:* Wien <[email protected]> on behalf of
Stefaan Cottenier <[email protected]>
*Sent:* 17 April 2019 08:50
*To:* A Mailing list for WIEN2k users
*Subject:* Re: [Wien] Magnetic moments converging in a different
direction to the one they are defined
Dear Martin,
You trigger long-forgotten memories... ;-)
Amazing that you remember that talk. That was the only way in which
these results were ever communicated, there was no paper. The reason
for this was that we did observe very clear and smooth energy
dependences upon continuous rotation of the Fe-moments in different
ways and for different configurations, yet the overall picture was a
messy paradox (= looking at one individual result would lead to one
conclusion, looking at another individual result would lead to an
opposite conclusion).
There has always been the feeling that once we should revisit this,
but it never got realized so far. Perhaps, if you feel like...
Best regards,
Stefaan
> -----Original Message-----
> From: Wien <[email protected]> On Behalf Of
> pieper
> Sent: Tuesday, April 16, 2019 12:51 PM
> To: A Mailing list for WIEN2k users <[email protected]>
> Subject: Re: [Wien] Magnetic moments converging in a different
direction to
> the one they are defined
>
> Fe3O4 being an old but unsatisfied love of mine a few additional
> comments:
>
> Determining exchange constants by spin reversal only makes sense if the
> changes in electronic structure are small (see e.g. P. Novak et. al
PHYSICAL
> REVIEW B 71, 184433, 2005).
>
> This (usually) works best in insulators, it is a delicate problem in
metals, and
> much more so in Fe3O4 with its Vervey transition. Here a very intricate
> coupling between electronic and structural degrees of freedom is at
work. I
> seem to recall that the low temperature phase is a comlicated mess
(see e.g.
> Novak et al, PRB 61, 1256, 2000 and references therein). So
complicated that,
> as far as I remember, in the early 2000nds S.Cottenier, R.
Laskowski, J. Rusz,
> M. Rots and P. Novak gave a talk on a Wien2k Workshop calculating
> exchange interactions in magnetite using the non-collinear magnetism
> version NCM-Wien2k. Unfortunately I don't have time to search for
> literature on that one, but you probably don't want to get into NCM
anyway.
>
> However, I don't think you can avoid DFT+U or +EECE - at least not
for Fe3O4.
> I am sure you can find a lot of literature on DFT+U and +EECE of Fe3O4,
> among others by Novak, Madsen, ... This may introduce an additional
> parameter in your comparisons of your structures. If you are using
an older
> version of Wien2k, upgrade! Wien26_16 had a bug with
> DFT+U (see
> https://www.mail-
> archive.com/[email protected]/msg15590.html).
>
> Best regards,
>
> Martin Pieper
>
>
> ---
> Dr. Martin Pieper
> Karl-Franzens University
> Institute of Physics
> Universitätsplatz 5
> A-8010 Graz
> Austria
> Tel.: +43-(0)316-380-8564
>
>
> Am 2019-04-16 10:29, schrieb Penny, Charles:
> > Dear all,
> >
> > I am running spin-polarised calculations on a range of iron-spinel
> > structures (namely, magnetite (Fe3O4), maghemite (gamma-Fe2O3) and
> > greigite (Fe3S4)) with the objective of calculating magnetic exchange
> > energies in these minerals. This requires calculating total energies
> > of lot of different spin configurations. This process has worked well
> > for magnetite and maghemite, but I have encountered a problem with
> > greigite.
> >
> > When I run a calculation on a spin configuration of greigite that
> > isn’t the ferrimagnetic ground state (e.g. a ferromagnetic
> > configuration) the calculation converges to the ferrimagnetic
> > solution, with the sublattice moments pointing in opposing directions.
> >
> >
> > In the examples below, I have used a low-symmetry unit cell with
> > eight unique iron atoms which allows me to calculate the required
> > number of spin configurations for estimating J_ij. Atoms 1-4
> > correspond to A site iron atoms in the spinel structure, atoms 5-8
> > correspond to B site iron atoms and atoms 9-16 are sulphur atoms. In a
> > ferrimagnetic system the A and B sites have opposing moments and
> > sulphur atoms are non-magneitc.
> >
> > When I define a ferrimagnetic spin configuration, the calculation
> > proceeds as expected, with the final moments looking like;
> >
> > rkmax_8_k_500.scf::MMINT: MAGNETIC MOMENT IN INTERSTITIAL =
> > -0.05116
> >
> > rkmax_8_k_500.scf::MMI001: MAGNETIC MOMENT IN SPHERE 1 =
> > 2.47349
> >
> > rkmax_8_k_500.scf::MMI002: MAGNETIC MOMENT IN SPHERE 2 =
> > 2.47348
> >
> > rkmax_8_k_500.scf::MMI003: MAGNETIC MOMENT IN SPHERE 3 =
> > 2.47348
> >
> > rkmax_8_k_500.scf::MMI004: MAGNETIC MOMENT IN SPHERE 4 =
> > 2.47348
> >
> > rkmax_8_k_500.scf::MMI005: MAGNETIC MOMENT IN SPHERE 5 =
> > -3.01699
> >
> > rkmax_8_k_500.scf::MMI006: MAGNETIC MOMENT IN SPHERE 6 =
> > -3.01699
> >
> > rkmax_8_k_500.scf::MMI007: MAGNETIC MOMENT IN SPHERE 7 =
> > -3.01699
> >
> > rkmax_8_k_500.scf::MMI008: MAGNETIC MOMENT IN SPHERE 8 =
> > -3.01699
> >
> > rkmax_8_k_500.scf::MMI009: MAGNETIC MOMENT IN SPHERE 9 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI010: MAGNETIC MOMENT IN SPHERE 10 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI011: MAGNETIC MOMENT IN SPHERE 11 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI012: MAGNETIC MOMENT IN SPHERE 12 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI013: MAGNETIC MOMENT IN SPHERE 13 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI014: MAGNETIC MOMENT IN SPHERE 14 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI015: MAGNETIC MOMENT IN SPHERE 15 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMI016: MAGNETIC MOMENT IN SPHERE 16 =
> > -0.03675
> >
> > rkmax_8_k_500.scf::MMTOT: SPIN MAGNETIC MOMENT IN CELL =
> > -14.88108
> >
> > Final energy;
> >
> > rkmax_8_k_500.scf::ENE : ********** TOTAL ENERGY IN Ry =
> > -43322.30312592
> >
> > However, when I define a ferromagnetic spin configuration the system
> > converges to a ferrimagnetic solution with final moments;
> >
> > k_500_rkmax_8.scf::MMINT: MAGNETIC MOMENT IN INTERSTITIAL =
> > 0.05118
> >
> > k_500_rkmax_8.scf::MMI001: MAGNETIC MOMENT IN SPHERE 1 =
> > -2.47348
> >
> > k_500_rkmax_8.scf::MMI002: MAGNETIC MOMENT IN SPHERE 2 =
> > -2.47347
> >
> > k_500_rkmax_8.scf::MMI003: MAGNETIC MOMENT IN SPHERE 3 =
> > -2.47346
> >
> > k_500_rkmax_8.scf::MMI004: MAGNETIC MOMENT IN SPHERE 4 =
> > -2.47346
> >
> > k_500_rkmax_8.scf::MMI005: MAGNETIC MOMENT IN SPHERE 5 =
> > 3.01697
> >
> > k_500_rkmax_8.scf::MMI006: MAGNETIC MOMENT IN SPHERE 6 =
> > 3.01697
> >
> > k_500_rkmax_8.scf::MMI007: MAGNETIC MOMENT IN SPHERE 7 =
> > 3.01697
> >
> > k_500_rkmax_8.scf::MMI008: MAGNETIC MOMENT IN SPHERE 8 =
> > 3.01697
> >
> > k_500_rkmax_8.scf::MMI009: MAGNETIC MOMENT IN SPHERE 9 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI010: MAGNETIC MOMENT IN SPHERE 10 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI011: MAGNETIC MOMENT IN SPHERE 11 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI012: MAGNETIC MOMENT IN SPHERE 12 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI013: MAGNETIC MOMENT IN SPHERE 13 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI014: MAGNETIC MOMENT IN SPHERE 14 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI015: MAGNETIC MOMENT IN SPHERE 15 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMI016: MAGNETIC MOMENT IN SPHERE 16 =
> > 0.03675
> >
> > k_500_rkmax_8.scf::MMTOT: SPIN MAGNETIC MOMENT IN CELL =
> > 14.88103
> >
> > Final energy is the same as in the ferrimagnetic case;
> >
> > k_500_rkmax_8.scf::ENE : ********** TOTAL ENERGY IN Ry =
> > -43322.30312578
> >
> > Charge distance looks like it converges in both cases. Note that
> > whilst the two calculations have the same saved name, they are in
> > completely different CASE files. To outline my procedure, I initially
> > call,
> >
> > instgen -ask
> >
> > And define the moments as ‘u u u u u u u u n n n n n n n n’ for a
> > ferromagnetic calculation and ‘u u u u d d d d n n n n n n n n’
> > for a ferrimagnetic calculation.
> >
> > Then,
> >
> > init -b -sp -numk 500 -rkmax 8.00
> >
> > runsp -ec 0.0001 -cc 0.0001 -fc 1.0 -p -i 200
> >
> > I am at a loss as to what is going on and can’t find anything in
> > the mailing list to explain this. Any help on this matter would be
> > greatly appreciated.
> >
> > Regards,
> >
> > Charlie
> > _______________________________________________
> > Wien mailing list
> > [email protected]
> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> > SEARCH the MAILING-LIST at:
> > http://www.mail-
> archive.com/[email protected]/index.html
> _______________________________________________
> Wien mailing list
> [email protected]
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at: http://www.mail-
> archive.com/[email protected]/index.html
_______________________________________________
Wien mailing list
[email protected]
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/[email protected]/index.html
_______________________________________________
Wien mailing list
[email protected]
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/[email protected]/index.html
_______________________________________________
Wien mailing list
[email protected]
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/[email protected]/index.html
