Hello
a few suggestions:
1. The convergence issues you are experiencing may be due to your starting
configuration; check that it is a sensible one.
2.
Use a larger smearing, and if you are using an old version of qe, try to switch
to gauss smearing,
3.
Diagonalization='cg' is very slow and inefficient; unless Davidson fails, it
is better to stick to that; otherwise
diagonalization='rmm-davidson' or 'rmm-paro'.
5) If you are not using a recent version of qe ( 7.2 or older ), avoid
using local-TF because, in some cases, it could lead to practically infinite
loops. This behaviour has been fixed in qe-7.3.
6) Interpolating potential and wave functions is not helpful for
relaxations because steps are generally longer than in dynamics.
7) conv_thr=10^-6 is a too high threshold to obtain any reliable
relaxation, it may be good for the very first steps of the relaxation but as
soon as the forces become small you will need to use a lower threshold at least
> 10^-8 to make the relaxation converge.
best regards
Pietro
________________________________
From: users <[email protected]> on behalf of Janiny
<[email protected]>
Sent: Wednesday, July 3, 2024 18:26
To: [email protected] <[email protected]>
Subject: [QE-users] [SPAM] Convergence problems
Dear all,
1) I am conducting an adsorption study, and convergence is taking too long (in
some cases, over 20 days). Additionally, there are situations where more than
100 iterations pass without generating coordinates. Is it possible to
periodically fix the coordinate generation?
2) I am using conv_thr = 10^-6. Do you suggest any other parameters that I
could adjust to improve convergence?
&control
calculation='relax'
wf_collect = .TRUE.
! restart_mode = 'restart'
max_seconds = 1.D+10
disk_io='low'
/
&system
ibrav = 0, celldm(1) =xxx,nat=xxx, ntyp= xxx,
ecutwfc = 55.0, ecutrho = 600,
occupations='smearing', smearing='marzari-vanderbilt', degauss=0.01
nspin= xxx
tot_magnetization= xxx
/
&electrons
conv_thr = 1.0e-6
mixing_beta = 0.1
mixing_mode= 'local-TF'
electron_maxstep=500
diago_thr_init = 1e-4
diagonalization = 'cg'
/
&ions
ion_dynamics = 'bfgs'
pot_extrapolation = 'second_order'
wfc_extrapolation = 'second_order'
Best regards,
_______________________________________________
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 [email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users
