[gpaw-users] Convergence

[Marcin Dulak]

Hi,

i was able to get through first few optimization steps with the default 
rmm-diis eigensolver (gpaw-0.7.2.6974) and BFGSLineSearch optimizer 
(ase-3.5.0.2113)
and the following settings, where the most important seem to be (see the 
converging script attached):
- larger width (increasing width with default Mixer caused SCF 
oscillations), if acceptable for your system!
- MixerSum
- latest 0.8.7929 setups (well, you can still converge luckily with 
setups 0.6.6300 with these special mixer settings,
but if you are at the beginning of a project please use the new ones;
note a strange guess energy and no magnetic moment with 0.6.6300 setups 
at the initial geometry step:
iter:   3  10:30:43  -1.4            -1159.97113  3               -0.0003
iter:   4  10:32:10  -1.1   -0.8     -995.11837   5      50       -0.0006,
...
iter:  18  10:51:28  -1.9   -1.4     -399.98696   3      8        -0.8680
).

To make calculations faster use explicitly the number of grid points 
divisible by 8 (along each axis), instead of giving just grid spacing h.
With 0.8.7929 setups use the default mixer settings to converge SCF faster:
mixer=Mixer(beta=0.1, nmaxold=3, weight=50.0)

Best regards,

Marcin

Samira Siahrostami wrote:
> To whom it may concern,
>
> I am doing some calculations on TiO2 surface with just one OH adsorbed 
> on the surface. The problem is it stops in the second cycle of 
> iteration and after 50 hours. It does not converge at all. For your 
> information, I am using the new set up for Ti too. I have attached the 
> script and the output file.
>
> I would be appreciated if you could give me a suggestion.
>
> Best regards,
>
> Samira
> ------------------------------------------------------------------------
>
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users

-- 
***********************************
 
Marcin Dulak
Technical University of Denmark
Department of Physics
Building 307, Room 229
DK-2800 Kongens Lyngby
Denmark
Tel.: (+45) 4525 3157
Fax.: (+45) 4593 2399
email: Marcin.Dulak at fysik.dtu.dk

***********************************

-------------- next part --------------
A non-text attachment was scrubbed...
Name: TiO2_1OH_atop_2_final.py
Type: text/x-python
Size: 1785 bytes
Desc: not available
Url : http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20110509/c6975c48/attachment.py 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TiO2_1OH_Rutile.traj
Type: application/octet-stream
Size: 4973 bytes
Desc: not available
Url : http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20110509/c6975c48/attachment.obj