[Pw_forum] Convergene of Magnetic moment is very slow
Giovani Faccin
faccin.giovani at gmail.com
Fri Mar 2 18:53:20 CET 2012
Kamal,
The scf will try to optimize the total energy. Once the energy is
optimized, the magnetic moment should stop oscillating.
Each step the program prints this line: "estimated scf accuracy: XXXXX". If
XXXXX is becoming smaller over time, the calculation is converging. If not,
then you can try other beta values, as discussed here:
http://www.quantum-espresso.org/user_guide/node64.html#SECTION000137020000000000000
Just looking at the magnetization before the system gets self-consistent is
pointless, since all kind of values will show up.
A little idea that might help: be sure that you have enough kpoints. 5x5x5
can be too few for many purposes. I'd give it a try with, say, 11x11x11 to
see if the final magnetization (after the scf is concluded) changes.
Best regards,
Giovani M. Faccin
CPPP / UFMS - Brazil
2012/3/2 <ksaha at physics.udel.edu>
> Dear Users,
>
> I am running scf-iterations of a ferrimagnetic yttrium iron garnet (bcc
> primitive cell with 80 atoms) system and find that the convergence of
> magnetic moment is very slow. Can you any suggest me how to accelerate the
> overall convergence? Please see the last a few steps of magmom of the
> 100-scf steps:
>
> total magnetization = 78.61 Bohr mag/cell
> total magnetization = 78.79 Bohr mag/cell
> total magnetization = 78.78 Bohr mag/cell
> total magnetization = 78.91 Bohr mag/cell
> total magnetization = 79.25 Bohr mag/cell
> total magnetization = 79.17 Bohr mag/cell
> total magnetization = 79.22 Bohr mag/cell
> total magnetization = 79.06 Bohr mag/cell
> total magnetization = 79.08 Bohr mag/cell
> total magnetization = 79.10 Bohr mag/cell
> total magnetization = 79.12 Bohr mag/cell
> total magnetization = 79.09 Bohr mag/cell
> total magnetization = 79.15 Bohr mag/cell
> total magnetization = 79.23 Bohr mag/cell
> total magnetization = 79.17 Bohr mag/cell
> total magnetization = 79.47 Bohr mag/cell
> total magnetization = 79.45 Bohr mag/cell
> total magnetization = 79.15 Bohr mag/cell
> total magnetization = 81.33 Bohr mag/cell
> total magnetization = 82.16 Bohr mag/cell
> total magnetization = 82.24 Bohr mag/cell
> total magnetization = 82.32 Bohr mag/cell
> total magnetization = 82.53 Bohr mag/cell
> total magnetization = 84.89 Bohr mag/cell
> total magnetization = 82.93 Bohr mag/cell
> total magnetization = 78.05 Bohr mag/cell
> total magnetization = 84.82 Bohr mag/cell
> total magnetization = 84.32 Bohr mag/cell
> total magnetization = 84.69 Bohr mag/cell
> total magnetization = 85.38 Bohr mag/cell
> total magnetization = 87.34 Bohr mag/cell
>
> Also, here is the part of the script which you may want to look at:
>
> &system
> ibrav = 3,
> celldm(1) = 23.24362966, ! = a, in Bohr
> nat= 80,
> ntyp= 3,
> nspin=2,
> starting_magnetization(1)=0.0,
> starting_magnetization(2)=2.60,
> starting_magnetization(3)=0.0,
> occupations='smearing',
> smearing='mp',
> degauss=0.02,
> ecutwfc =30,
> ecutrho =240
> /
> &electrons
> electron_maxstep = 400,
> diagonalization='david',
> conv_thr = 1.0e-6,
> mixing_beta = 0.05,
> !mixing_ndim = 6
> mixing_ndim = 10
> /
> ATOMIC_SPECIES
> Y 88.90585 Y.pbe-nsp-van.UPF
> Fe 55.84500 Fe.pbe-sp-van_mit.UPF
> O 15.99940 O.pbe-van_ak.UPF
> .
> .
> .
> K_POINTS automatic
> 5 5 5 1 1 1
>
> Many thanks for your comments and suggestions,
>
> Regards,
> Kamal
>
> ==========================================================
> Kamal K. Saha
> Department of Physics and Astronomy
> University of Delaware
> Newark, DE 19716-2570
> Phone: (302) 831-8519, Fax: (302) 831-1637
>
> http://web.physics.udel.edu/about/directory/postdoc/kamal-k-saha
> ==========================================================
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120302/9e98cd7b/attachment.html>
More information about the users
mailing list