Dear Peter (or Yun-Peng??),
Two wild guesses: 1) Did you check that the cut-off energies employed
are sufficient? 2) The 'mixing_beta' is very large; the smaller this
value, the tighter the convergence. Well, up to some limit; I would try a
value closer to 0.3 or so. (If your system has a band gap, I do not know
how the 'mv' broadening affects the occupation numbers.)
Greetings from Paris,
apsi
-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
Ari Paavo Seitsonen / [email protected] / 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 Fri, 11 Sep 2015, Peter Cybertron wrote:
Hi,
I am working in University of Florida as a postdoc.
I am exploring the spin-orbital calculations using pwscf, and I
encounter a very slow convergence like this:
> grep estimated scf.out
estimated scf accuracy < 0.00000004 Ry
... (100 similar lines)
estimated scf accuracy < 0.00000003 Ry
> grep "total energy" scf.out
...
total energy = -3667.91077719 Ry
total energy = -3667.91078176 Ry
total energy = -3667.91077642 Ry
total energy = -3667.91078300 Ry
Although the convergence of total energy seems OK (down to 0.00001 Ry ~
0.0001 eV = 0.1 meV), I still feel uncomfortable since I never encounter
such a problem without spin-orbit interactions.
I downloaded the ultrasoft and PAW pseudopotentials from PSLib and I am
using the PBE flavor of XC functional. The system I am studying is a two
dimensional magnetic insulator.
The parameters I am using are:
constrained_magnetization = 'total direction'
fixed_magnetization(3) = 0.0
lambda = 0.2
ecutwfc = 30.0 ,
ecutrho = 300.0 ,
occupations = 'smearing' ,
degauss = 0.005,
smearing = 'mv' ,
nosym=.true.
mixing_mode = 'local-TF' ,
mixing_beta = 0.9 ,
diagonalization = 'david' ,
Which parameter on earth keep my calculations non-converged?
Best,
Yun-Peng
_______________________________________________
Pw_forum mailing list
[email protected]
http://pwscf.org/mailman/listinfo/pw_forum
_______________________________________________
Pw_forum mailing list
[email protected]
http://pwscf.org/mailman/listinfo/pw_forum
