[Pw_forum] PH error: k-points are odd

Konstantin Kudin konstantin_kudin at yahoo.com
Thu Aug 30 17:00:59 CEST 2007


--- Paolo Giannozzi <giannozz at nest.sns.it> wrote:

> 
> On Aug 29, 2007, at 23:42 , Konstantin Kudin wrote:
> 
> > Indeed, I was not aware of the need to run PW with 'phonon' before
> PH
> > for a general k point.
> 
> it is clearly stated in the documentation

> >  I patched a little the PH code in the CVS such that if the desired
> k
> > point can be found on a regular k space grid, one can set the input
>  
> > for
> > the dispersion calculation with some mesh and then specify that
> just
> > one k point from the mesh is of interest. The significant advantage
> of
> > this option is that the need for the prior PW run with 'phonon'
> > disappears - thus the job script can be shorter :-)
> 
> the significant disadvantage is that option 'lnscf' does exacty what
> you wanted in a much more straightforward way

 Then this is quite confusing. If 'lnscf' *is* working, why advise
people to run PW with 'phonon' first where 'lnscf' in PH by itself
would work just fine?

 I also did try 'lnscf' with a CVS version from a couple of months ago,
and it did not work there, since the failing case I posted previously
did have lnscf=.true. . The recent version works, so it must have been
fixed in the meantime. I could not find which specific commit addressed
the issue recently, that is why I thought the older version was just as
good.

 Finally, in the "clear" documentation there are 2 different options -
'nscf' and 'phonon' for PW. The manual says to run 'phonon' before PH,
at the same time lnscf=.true. in PH promises to run 'nscf' first. This
does not seem to be that particularly clear ...

 Kostya


       
____________________________________________________________________________________
Sick sense of humor? Visit Yahoo! TV's 
Comedy with an Edge to see what's on, when. 
http://tv.yahoo.com/collections/222



More information about the users mailing list