[Pw_forum] Charge density overwrite?

[Paolo Giannozzi]

On Wednesday 17 May 2006 18:22, Conor Hogan wrote:

> I've seen some notes in the ChangeLog stating "do not overwrite the
> charge density after a non-scf calculation". However, this is not my
> experience with 3.0 (*.rho file) and even now with 3.1
> (*.save/charge-density.xml file). An nscf run, as far as I can see,
> overwrites the charge density 

too bad: this shouldn't happen. Not that it is harmful in any way:
the charge density that is written to file should be the scf one 
that has just been read. If you end up with a strange charge
density, please comment line 56 in PW/punch.f90 :
  IF ( .NOT. lscf .AND. .NOT. lbands ) CALL sum_band()
which recalculates the charge density. I think this is completely
useless now.

> which could of course give problems later in life, if one is not 
> careful... (at least a "diff" says the file has changed)

a "diff" on a binary file is not especially significant

> In particular, I use the pw_export.x facility to calculate optical
> spectra later. I guess I should be using the scf .rho, .pot. and 
> the nscf .save files?

you should use the nscf .save files, which should contain all you need.
The potential is no longer written, since it can be recalculated from the
charge density

Paolo
-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy