The ABINIT Users Mailing List ( CLOSED )

["D. R. Hamann" <drh@bell-labs.com>]

Hi Eduardo,

Let me add one more piece to the very good advice you've already 
gotten.  With metallic occupation, gamma-point phonon calculations 
involve a first-order fermi energy, which will be printed in the log 
file if you set printvol 10 (grep for fermie1).  I recall that your 
tsmear was very small, and this could have contributed to an instability 
related to this (oscillation or divergence).  Since you are aiming to 
simulate an isolated molecule, I would guess that a considerably larger 
tsmear could still be small compared to the HOMO-LUMO gap, which is all 
you need.  (I assume the HOMO is partly occupied or degenerate, or you 
wouldn't need metallic occupation.)

Maybe I meant two more pieces of advice.  I'd use a smaller tolvrs in 
your first dataset, say 1.d-12.  I'd also insert an additional 
calculation before the RF calculation to make sure your unoccupied GS 
wave functions are converged, since the self-consistent potential isn't 
sensitive to these and tolvrs doesn't guarantee their convergence.  With 
metallic occupation, these are used in a special way in calculating the 
first-order wave functions, and poor convergence could lead to 
inaccuracy.  To do this, put prtden1 1 in the first dataset, in the 
second data set put getden2  -1, getwfk2  -1, iscf2  -2, and tolwfr2  
1.d-12.  getwfk3  -1 in the RF calculation will then pick up these 
WFK's.  These steps probably don't influence the convergence problem, 
but will increase the accuracy.

Good luck.

Don Hamann

Eduardo Cruz Silva wrote:

> Hi Mathieu
>
> Thanks for your advice about restarting the calculation, it did work 
> pretty well!
>
> The idea is to simulate isolated molecules, and about the box size, I 
> checked the charge density at the faces of the box, and it was less 
> than 1e-4, compared to the density peak around 0.4, I thougth it was 
> low enough to avoid interactions.
>
> Thanks!!
> Best regards
>
> verstraete@pcpm.ucl.ac.be wrote:
>
> > Better! Ok, from the occupations you have for the different kpoints, 
> > your images are clearly interacting. You still haven't told us if you 
> > want to simulate isolated molecules (in which case you need a bigger 
> > box) or a monoatomic wire, in which case your z direction should be 
> > smaller for the wire to be continuous, and the x and y should be 
> > larger to avoid interwire interactions.
> >
> > try to have fun anyway, this is the most painful part of abinit (sc 
> > convergence with vacuum).
> >
> > Matthieu


--
D. R. Hamann                            Phone: 908-582-4454
Director, Theoretical Materials         Fax:   908-582-4868
 Physics Research (retired)             email: drh@physics.bell-labs.com
Bell Laboratories 
Lucent Technologies
700 Mountain Ave, Room 1D-371
Murray Hill, NJ 07974-0636 USA