[Pw_forum] Convergence of relax of slab calculation
Ari P Seitsonen
Ari.P.Seitsonen at iki.fi
Thu Aug 6 16:47:22 CEST 2015
Dear Ben Liew,
If one is lucky, the system might converge to the right magnetic ground
state with a "random guess". Yet if the system has several local minima,
or the initial guess is very far from the true magnetic ordering, one
might get very bad initial densities (total and spin density, or spin-up
and spin-down densities), leading to difficulties in convergence and maybe
even divergence. Yet-another-issue might be bad description of the
electronic structure with the GGA's, I seem to remember that in the case
of MnO2 the GGA gives reasonable results, even if the electronic band gap
is naturally much too small (and using the smearing should help to get a
convergence anyway).
I would myself start by trying a weakly ferromagnetic guess (for
example 'starting_magnetization(1) = 0.2' and seeing if something happens.
If it converges, or doesn't, I would then see if an anti-ferromagnetic
solution might also make sense (two types of Mn atoms in cell, etc).
Greetings,
apsi
-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
Ecole Normale Supérieure (ENS), Département de Chimie, Paris
Mobile (F) : +33 789 37 24 25 (CH) : +41 79 71 90 935
On Wed, 5 Aug 2015, ben liew wrote:
> Dear Bahadir,
> Thanks for your comments. I will try using a lower mixing_beta and change
> the mixing_mode='local-TF'.
>
> Dear Ari Paavo Seitsonen,
>
> Thanks. I have no knowledge on the magnetism of my system. How could I know
> what value of starting_magnetization that I should use in my slab
> calculation? Does the starting_magnetization affect the convergence of my
> calculation?
>
> Thank you
>
>
> Best Regards,
> Ben Liew
> Ph.D Student
> Fuel Cell Institute,
> The National University of Malaysia,
> 43600 Bangi,
> Selangor, Malaysia.
> Contact no. : +6016 552 0878
> Email : liewkienben at gmail.com
>
> On Wed, Aug 5, 2015 at 5:20 PM, Bahadır salmankurt <bsalmankurt at gmail.com>
> wrote:
> Dear Ari P Seitsonen,
> I didnt know that mixing_mode='local-TF' effect initial magnetic
> moments. thanks for this. By the way, Do you know about e field
> configuration? for example, for slab + molecule systems whose atomic
> positions are in 0.5 and 0.65 along z direction , respectively, what
> must values of the emaxpos and eopreg be?
> Bests
> Bahadir
>
> 2015-08-05 10:59 GMT+03:00 Ari P Seitsonen <Ari.P.Seitsonen at iki.fi>:
>
> Dear Ben Liew,
>
> Without looking at the structure itself, adding to the
> previous comment about the algorithm for mixing, you have
> not really given any initial magnetic moments even if you
> have 'nspin = 2': The 'starting_magnetization(1) = 0.0'
> does not give any preferential spin ordering. Do you know
> something about this, is the system expected to be
> ferromagnetic, antiferromagnetic, ...?
>
> Tiny issues, I usually do the first calculation without
> the surface dipole as it makes the convergence always a
> bit trickier; and your
> 'B ~= 2 * A', but you give the same number of k points in
> both directions. (I usually tend to use smearing schemes
> where the occupations are guaranteed to be in the physical
> range of [0:1], but this is just a personal preference and
> should thus be ignored)
>
> Greetings,
>
> apsi
>
> -=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*
> =-
> Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi /
> http://www.iki.fi/~apsi/
> Ecole Normale Supérieure (ENS), Département de Chimie,
> Paris
> Mobile (F) : +33 789 37 24 25 (CH) : +41 79 71 90 935
>
>
> On Wed, 5 Aug 2015, ben liew wrote:
>
> Dear PWSCF users,
>
> Hi, I am a new user of Pwscf. I am working on
> slab calculation for pyrolusite MnO2 (110)
> surface with 3 atomic layers calculation and
> fixed bottom 2 layers and only top atomic
> layer is allowed to be relaxed. I have also
> include the dipole correction for the
> calculation to counteract with the interaction
> between slabs. However, my calculation doesn't
> seems to be converged as the scf accuracy is
> not decreasing over each iteration. Below is
> the estimated scf accuracy after hundreds of
> iterations, it is not converging
>
> estimated scf accuracy <
> 373.64728976 Ry
> estimated scf accuracy <
> 106.52177156 Ry
> estimated scf accuracy <
> 96.97567125 Ry
> estimated scf accuracy <
> 207.03807187 Ry
> estimated scf accuracy <
> 132.53045831 Ry
> estimated scf accuracy <
> 77.19101827 Ry
> estimated scf accuracy <
> 112.41885075 Ry
> estimated scf accuracy <
> 93.27430429 Ry
> estimated scf accuracy <
> 116.68049786 Ry
> estimated scf accuracy <
> 587.09711027 Ry
> estimated scf accuracy <
> 374.45312501 Ry
> estimated scf accuracy <
> 106.89811469 Ry
> estimated scf accuracy <
> 96.84898888 Ry
> estimated scf accuracy <
> 207.83006869 Ry
> estimated scf accuracy <
> 132.74588293 Ry
> estimated scf accuracy <
> 77.09445200 Ry
> estimated scf accuracy <
> 112.41688597 Ry
> estimated scf accuracy <
> 93.57796273 Ry
> estimated scf accuracy <
> 116.25663912 Ry
> estimated scf accuracy <
> 587.39968959 Ry
> estimated scf accuracy <
> 374.91438307 Ry
> estimated scf accuracy <
> 106.71641516 Ry
> estimated scf accuracy <
> 96.75392736 Ry
>
> I have include my input file as followed. Can
> anyone tell me is that any problem with my
> input file?
>
> &control
> calculation = 'relax'
> tstress=.true.
> tprnfor=.true.
>
> pseudo_dir='/home/ben/Downloads/espresso-5.1.2/pseudo'
> outdir='./vc3'
> prefix='pyrolusite'
> tefield=.true.
> dipfield=.true.
> /
> &system
> ibrav=8,
> A=2.8563, B=6.2199, C=20.0
> nat=18, ntyp=2,
> starting_magnetization(1)=0.0
> nspin=2
> ecutwfc=37.0
> ecutrho=250.0
> occupations='smearing'
> smearing='m-v'
> degauss=0.01
> edir=3
> emaxpos=0.5
> eopreg=0.1
> eamp=0.001
> /
> &electrons
> electron_maxstep=500
> conv_thr=1.0d-6
> mixing_beta=0.7
> /
> &ions
> ion_dynamics='bfgs'
> /
> ATOMIC_SPECIES
> Mn 54.9380 Mn.pbe-spn-kjpaw_psl.0.3.1.UPF
> O 15.9994 O.pbe-kjpaw.UPF
> ATOMIC_POSITIONS (angstrom)
> O 1.428129 3.109734 0.000000 0 0
> 0
> O 1.428129 0.000241 0.652526 0 0
> 0
> Mn 1.428129 3.109896 1.883398 0 0
> 0
> O 0.000000 1.885646 1.883398 0 0
> 0
> O 0.000000 4.334118 1.883398 0 0
> 0
> Mn 0.000000 0.000062 1.883398 0 0
> 0
> O 0.000000 5.014614 4.987475 0 0
> 0
> O 0.000000 1.205401 4.987475 0 0
> 0
> Mn 0.000000 3.109816 4.987475 0 0
> 0
> Mn 1.428129 0.000027 4.987475 0 0
> 0
> O 1.428129 3.109734 3.767387 0 0
> 0
> O 1.428129 0.000241 3.114269 0 0
> 0
> O 1.428129 3.109734 6.207560
> O 1.428129 0.000241 6.860698
> Mn 1.428129 3.109896 8.091569
> O 0.000000 1.885646 8.091569
> O 0.000000 4.334118 8.091569
> Mn 0.000000 0.000062 8.091569
> K_POINTS (automatic)
> 5 5 1 0 0 0
>
> Your comments and suggestion is much
> appreciated.Thank you.
>
>
> Best Regards,
> Ben Liew
> Ph.D Student
> Fuel Cell Institute,
> The National University of Malaysia,
> 43600 Bangi,
> Selangor, Malaysia.
> Contact no. : +6016 552 0878
> Email : liewkienben at gmail.com
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
>
More information about the users
mailing list