[QE-users] Convergence to an AFM ground state
Nicola Marzari
nicola.marzari at epfl.ch
Sun Feb 9 21:47:52 CET 2020
Dear Daniel, David,
do the calculations need to be done with non-collinear magnetism?
I have very little experience with those, except the suspicion that they
converge very very slowly, if they converge at all (the torque on the
spin is negligible compared to the energy change in the functional).
If you can/want to try a collinear calculation, my perennial advice
would be to use the pseudopotentials and setup provided by
https://www.materialscloud.org/work/tools/qeinputgenerator
[Of course you might want to add initial moments that are compatible
with the AFM solution.]
The pseudo suggested above are the "best" pseudopotentials among all
those tested for QE: https://www.materialscloud.org/discover/sssp/.
Mn is a particularly difficult element to pseudize - so I would be very
wary of using any pseudo that has not been tested carefully.
If the pseudopotentials are faithful (not at all a given - hence the
effort above for verification) then any self-consistent state will be a
proper minimum of your functional. Magnetic systems have many minima,
and the one you fall upon depends on how large is its basin of
attraction - how to go from one mimimum to another is a bit of an art,
but choosing intial moments can help a lot.
Can you try the protocol above and report?
nicola
On 09/02/2020 21:28, Daniel Kaplan wrote:
> Hello Paolo and David.
> I tested the exact same system with ecutrho=4,6,8,10*ecutwfc to no avail.
> Perhaps there's something else still missing.
> Yours,
> Daniel
>
>
> On Sun, Feb 9, 2020, 20:05 David Guzman <davgumo at me.com
> <mailto:davgumo at me.com>> wrote:
>
> Paolo,
> Thanks for the comment. Even with appropriately chosen cutoffs for
> the kinetic energy and charge density, the AFM solution is difficult
> to converge with PAW pseudopotentials (suggested from website).
> The system I was trying (FeSb2) is less complicated than Daniel’s
> system. I thought it was an isolated case, but now that Daniel is
> also having problems maybe there’s more to it.
>
> Regards,
> David Guzman
> Brookhaven National Laboratory
>
>> On Feb 9, 2020, at 12:38 PM, Paolo Giannozzi
>> <p.giannozzi at gmail.com <mailto:p.giannozzi at gmail.com>> wrote:
>>
>>
>> On Sun, Feb 9, 2020 at 3:36 PM David Guzman <davgumo at me.com
>> <mailto:davgumo at me.com>> wrote:
>>
>> Not sure [...] if there are extra setting that should go along
>> with those potentials.
>>
>>
>> the only extra setting with USPP and PAW is that you may (and
>> often, you should) use a cutoff for the charge density (ecutrho),
>> that differs from 4*ecutwfc (default value).
>>
>> Paolo
>>
>>
>>> On Feb 9, 2020, at 9:10 AM, Daniel Kaplan
>>> <danielkaplan137 at gmail.com
>>> <mailto:danielkaplan137 at gmail.com>> wrote:
>>>
>>>
>>> Hello All!
>>>
>>> I'm trying to calculate a system with a *known* AFM ground
>>> state. In the attached example, I provide the input data I'm
>>> using for CuMnAs -- a tetragonal anti-ferromagnet.
>>> I've started on this project by first executing the examples,
>>> and particularly FeO.
>>> I tested this example against all sorts of variations:
>>> different functionals, with/without U, and so on. Without any
>>> further tweaking, the system always converged to the AFM
>>> ground-state, provided the initial moments were also oriented
>>> in the AFM configuration.
>>>
>>> Which makes my failure in this (CuMnAs) system even more
>>> puzzling. Firstly, *without* any constraints, the system does
>>> not converge to an AFM state.
>>> 1. Using 'constrained_magnetization=total' leads to
>>> completely wrong results, with a divergent "Magnetic field".
>>> 2. A more-or-less sensible result can be obtained with
>>> 'constrained_magnetization='atomic' (as shown), however, the
>>> resultant magnetization is not altogether anti-ferromagnetic.
>>> Note that the system is in general endowed with PT-symmetry.
>>> The resultant eigenvalues *DO NOT* show this and you can also
>>> see the disparity in the magnetic moments of the Mn atoms, as
>>> well as eigenvalue difference of more than 1meV for some
>>> bands and k-points.
>>> 3. This behavior is /weakly/ dependent on lambda. I tried
>>> fiddling around with the values. A certain increase worsens
>>> the results, then seems to improve it, only to worsen again.
>>> What is a reason value for the constraint, in your
>>> estimation? I take it to be 5% of the unperturbed energy
>>> (i.e., energy without the constraint).
>>> 4. Testing the *exact* same system on different software
>>> (VASP, in this case), converged very well to the AFM state
>>> (i.e., PT symmetry was recovered to less than 1meV).
>>>
>>> What am I doing wrong, therefore?
>>> I would appreciate any advice.
>>> Yours thankful,
>>> Daniel Kaplan
>>> Dept. of Condensed Matter Physics
>>> Weizmann Institute of Science
>>> <scf.in <http://scf.in>>
>>> <scf.out>
>>> _______________________________________________
>>> Quantum ESPRESSO is supported by MaX
>>> (www.max-centre.eu/quantum-espresso
>>> <http://www.max-centre.eu/quantum-espresso>)
>>> users mailing list users at lists.quantum-espresso.org
>>> <mailto:users at lists.quantum-espresso.org>
>>> https://lists.quantum-espresso.org/mailman/listinfo/users
>> _______________________________________________
>> Quantum ESPRESSO is supported by MaX
>> (www.max-centre.eu/quantum-espresso
>> <http://www.max-centre.eu/quantum-espresso>)
>> users mailing list users at lists.quantum-espresso.org
>> <mailto:users at lists.quantum-espresso.org>
>> https://lists.quantum-espresso.org/mailman/listinfo/users
>>
>>
>>
>> --
>> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
>> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
>> Phone +39-0432-558216, fax +39-0432-558222
>>
>> _______________________________________________
>> Quantum ESPRESSO is supported by MaX
>> (www.max-centre.eu/quantum-espresso
>> <http://www.max-centre.eu/quantum-espresso>)
>> users mailing list users at lists.quantum-espresso.org
>> <mailto:users at lists.quantum-espresso.org>
>> https://lists.quantum-espresso.org/mailman/listinfo/users
> _______________________________________________
> Quantum ESPRESSO is supported by MaX
> (www.max-centre.eu/quantum-espresso
> <http://www.max-centre.eu/quantum-espresso>)
> users mailing list users at lists.quantum-espresso.org
> <mailto:users at lists.quantum-espresso.org>
> https://lists.quantum-espresso.org/mailman/listinfo/users
>
>
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
>
--
----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
Director, National Centre for Competence in Research NCCR MARVEL, EPFL
http://theossrv1.epfl.ch/Main/Contact http://nccr-marvel.ch/en/project
More information about the users
mailing list