The ABINIT Users Mailing List ( CLOSED )

[Eduardo Cruz Silva <ecruz@ipicyt.edu.mx>]

Sorry, I forgot to post the logs of the run, these are at
http://www.abinit.org/wws/d_read/forum/logs.tgz

and the output files are at
http://www.abinit.org/wws/d_read/forum/out.tgz

GS was very well converged in both iscf 3 and 5 cases (tolvrs = 1d-8) 
and 2DEt had a very slow convergence with CG minimization, but vres2 was 
very high (6.4d07). It diverged with Anderson mixing
In all cases, nband was increased to 14.

The requested pertubations are correct, only those are needed.

I will try to restart the calculation, will let youknow what happens.
Thanks

verstraete@pcpm.ucl.ac.be wrote:
> Hello,
>
> Excellent work! You looked at the FAQ! However you didn't send us your 
> output (eventually edited to be manageable). Netiquette... Anyway:
>
> You are simulating a linear string of 4 C atoms in vacuum. Your cell 
> looks huge: having a cell that is large (VdW distances between fragments 
> in the supercell) but not well insulated (>8 or 10 A or even more in bad 
> cases) will give rise to instabilities. RF calculations are even more 
> delicate and unstable. Note that your coordinates are in A but your cell 
> is in bohr...
>
> You probably don't need all those kpoints, but that shouldn't be the 
> problem. The diemac value is ok and important when you have vacuum. Are 
> you sure you only want perturbations wrt atoms 2 to 3 (rfatpol) along 
> only z? - I presume so
>
> For the convergence, if the system is almost "metallic" (which is 
> probably the case given the symmetry) you can have convergence problems 
> and divergencies. You can try increasing the number of bands, but again 
> it would be easier to look at your output. Is it converging really 
> slowly, or diverging? Are you sure the GS was correctly converged 
> beforehand?
>
> You could start the dtset 2 with iscf 5, and set nstep to stop it before 
> it stalls due to 13 restarts, then read in the 1WF (ird1wf2 1) and re-do 
> the dtset 2.
>
> Matthieu