[Pw_forum] c_bands: eigenvalues not converged (GIPAW)
Jarkko.Vahakangas at oulu.fi
Tue Jan 29 10:15:35 CET 2013
Hi Emine and all.
One thing i do to understand if it is something about gipaw.x or pw.x, is to increase the threshold to the same of the gipaw calculation, (1d-14 in your case) in the pw run and see if one is really in the "noise range" of the calculation due to cutoffs/pseudos/XC.
Works well for me, especially for small systems.
Thanks for advice E. I tried to increase PW's conv_thr to 1d-14 and luckily i was able to reach that particular convergence after adding of extra bands and change of mixing_mode for my small gap system (17 atoms). Then I launched the gipaw.x module and g-tensor...
Possible reasons of this warning has been discussed many, many times in this forum so i will skip.
But if it is a GIPAW specific problem, for example if the pw calculation is done with a setting that GIPAW doesnt understand therefore cannot solve the correct hamiltonian the wavefunctions are read for, (I think this was happening for the isolated system corrections like martyna-tuckermann )
then please write again and we can discuss more in detail.
and during the magnetic susceptibility calc. the only attentions were following:
k-point # 3 of 20 pool # 1
ATTENTION: ik= 3 ibnd= 63 eigenvalues differ too much!
ATTENTION: ik= 3 ibnd= 64 eigenvalues differ too much!
k-point # 4 of 20 pool # 1
I do not believe that this could be a pw calculation error because similar input for system which contains wider gap works smoothly. What is a good thing is that now i do not get any error of c_band convergence. But should I be worry about ibnd attentions?
Gipaw's ethr is 1d-14 as a default. Can that be reduced and does it make any sense?
Thanks again & BR,
Jarkko Vähäkangas (Ph.D. cand.)
Department of Physics, NMR-group, University of Oulu,
FI-90014 UNIVERSITY OF OULU, FINLAND,
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the users