[QE-users] Effect of nbnd on vc-relax

[Giovanni Cantele]

Dear Jing,

generally speaking, if you treat the system as a semiconductor (no
occupations keywork, no nbnd specified), the default number of bands will
be the number of electrons divided by two.
For metallic systems or systems that are treated as they were metallic
(occupations = 'smearing') the systems increases this number of bands by a
factor (if I remember well by 20%), because
this is needed in the convergence of the scf cycles when a subset ob bands
may cross the Fermi level.

So, if QE automatically sets the number of bands to 960, it means that you
have 960/1.2 = 800 KS states. It is clear that if your system is metallic
or if it has a very small gap, the scf cycle will never
(or quite hardly) converge if you set nbnd = 800 (nor the code will output
an error, that is instead issued if you set nbnd < nelec/2, for scf/relax).

Concerning the time needed to complete the calculation, this depends on a
number of factors: number of electrons (high in your case), number of
k-points (doubled for spin-polarized calculations
as yours),  starting geometry (far or close to the equilibrium one?),
number of needed plane waves (your cutoff is not that small). Moreover, it
can significantly depend on the number of processes
that you allocate on your supercomputer and on the way you use the parallel
features of QE.

So, the answers to your questions:
- And how should I determine an appropriate value of nbnd before starting
any calculation?
The easiest way (for relax/scf) is to let QE choose the default, bearing in
mind that if your system has small gaps, is metallic or is spin polarized
you need occupations = 'smearing'.

- Is this a normal timescale for a system with >100 atoms?
It can be normal, but there is not sufficient information to establish
whether or not you're efficiently exploring the parallelization features of
QE.

Giovanni

-- 

Giovanni Cantele, PhD
CNR-SPIN
c/o Dipartimento di Fisica
Universita' di Napoli "Federico II"
Complesso Universitario M. S. Angelo - Ed. 6
Via Cintia, I-80126, Napoli, Italy
e-mail: giovanni.cantele at spin.cnr.it <giovanni.cantele at spin.cnr.it>
Phone: +39 081 676910
Skype contact: giocan74

ResearcherID: http://www.researcherid.com/rid/A-1951-2009
Web page: https://sites.google.com/view/giovanni-cantele/home


Il giorno sab 3 giu 2023 alle ore 22:18 Jing Lian Ng <jinglian at utexas.edu>
ha scritto:

> Hello all,
>
> I am trying to do a vc-relax calculation on y-Fe2O3 with 160 atoms.
> Currently I am facing convergence issue with the input script below. With
> nbnd = 800, the calculation usually finishes in ~ 1 hr on a supercomputer,
> but fails to reach convergence even on the first BFGS step. When I removed
> the nbnd parameter, QE automatically set the number of Kohn-Sham states to
> 960 and the calculation was able to proceed. However, the relaxation
> calculation is taking more than 72 hours and hasn't reached completion (It
> is still running now, on 15th BFGS cycle). Is this a normal timescale for a
> system with >100 atoms? And how should I determine an appropriate value of
> nbnd before starting any calculation?
>
> ############ Input Script ####################
> &SYSTEM
>    ibrav           = 14
>    A               = 8.48998
>    B               = 8.48998
>    C               = 25.46994
>    cosAB           = 0
>    cosAC           = 0
>    cosBC           = 0
>    nat             = 160
>    ntyp            = 3
>    ecutwfc         = 65
>    ecutrho         = 780
>    nbnd            = 800
>    occupations     = 'smearing'
>    degauss         = 0.001
>    smearing        = 'mv'
>    nspin           = 2
>    starting_magnetization(1) = 0.5
>    starting_magnetization(2) = -0.5
>    starting_magnetization(3) = 0
>  /
>  &ELECTRONS
>    conv_thr        = 0.00001
>    mixing_mode     = 'plain'
>    mixing_beta     = 0.3
>    electron_maxstep = 200
>  /
>  &IONS
>  /
>  &CELL
>    !cell_dofree     = 'volume'
>  /
> ATOMIC_SPECIES
>   FeO   55.84500  Fe.pbesol-spn-kjpaw_psl.0.2.1.UPF
>   FeT   55.84500  Fe.pbesol-spn-kjpaw_psl.0.2.1.UPF
>     O   15.99900  O.pbesol-n-kjpaw_psl.0.1.UPF
> ##########################################################
>
> Thanks,
> Jing Lian Ng
> Graduate student at CHE, University of Texas at Austin
>
> _______________________________________________
> The Quantum ESPRESSO community stands by the Ukrainian
> people and expresses its concerns about the devastating
> effects that the Russian military offensive has on their
> country and on the free and peaceful scientific, cultural,
> and economic cooperation amongst peoples
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20230605/601d7a78/attachment.html>