The ABINIT Users Mailing List ( CLOSED )

[Marc Torrent <marc.torrent@cea.fr>]

Could you send me your .paw file ?
This bug is probably already corrected for the future version but I want 
to check it...

Marc


Ian D. Gay a écrit :
> On Tue, Nov 07, 2006 at 01:40:12PM +0100, Marc Torrent wrote:
>
> Sorry, I mixed up my windows abinis files, so previous information is 
> incorrect. The one which works on windows is version 5.1.3.
> Version 5.2.3 fails to read a numeric shape function on both windows 
> and linux.
>
> I tested with the same .paw file in both cases, so it's not a bug in 
> the generation of the .paw file. It appears that a bug has been 
> introduced in abinit 5.2.3 which leads to failure to read .paw files 
> containing numeric shape functions.
>
> Ian
>
>
>
> > What is the version number of your Abinit release ?
> > Is it the same as Nuno's ?
> >
> > uspp2abinit.dat is not documented (but it will be soon).
> > Marc
> >
> >
> >
> > Ian D. Gay a ?crit :
> > > Thanks for the advice. Is usp2abinit.dat documented anywhere?
> > >
> > > Actually it looks like the problem is somewhere in abinit. The '-1' type 
> > > paw files are read successfully by Nuno's recently-released windows 
> > > builds. The problem occurs for me only on Linux, both with the 
> > > distributed binary, any with my own builds. Not a compiler bug, since 
> > > it occurs with all of g95, gfortran and ifort. Maybe an uninitialized 
> > > variable somewhere? I've dug a bit, but haven't found anything yet.
> > >
> > > Ian
> > >
> > >
> > >
> > > On Mon, Nov 06, 2006 at 12:17:12PM +0100, Marc Torrent wrote:
> > > > It looks like you're getting a "numeric" shape function...
> > > > It's an hidden feature that you should'nt have to use.
> > > > Perhaps a bug...
> > > > You can impose the use of an "analytical" shape function by creating a 
> > > > file named "uspp2abinit.dat" with the following line in the directory 
> > > > used to generate the paw file:
> > > >
> > > > SHAPENUM=0
> > > >
> > > > It should create a paw file with the correct "2 0." line for the shape 
> > > > function.
> > > >
> > > > Regards,
> > > >
> > > > Marc Torrent
> > > > CEA-Bruyeres-le-Chatel - France
> > > >
> > > >
> > > > gay@sfu.ca a ?crit :
> > > > > When I attempt to use a paw data file produced by usp2abinit, version
> > > > > 1.8.2, abinit gives the message
> > > > >
> > > > > psp7in: ERROR -
> > > > > Shape function is not zero at PAW radius !
> > > > > Action: check your pseudopotential file.
> > > > >
> > > > > This happens both for the Ti file generated in paw tutorial 2, and also
> > > > > for others which I have attempted.
> > > > >
> > > > > Looking at the .paw file, I see the line 
> > > > > -1 0.                                   : shape_type(SIN),rshape
> > > > >
> > > > > and looking at the end of the file, I see that several ===== SHAPEF (l= 
> > > > > ) =====
> > > > > sections are present. The last number in each of these sections is about
> > > > > 1e-32. Is that what's supposed to be zero? Is this a rounding error
> > > > > problem?
> > > > >
> > > > > If I manually edit the .paw file to give it a shape_type of 2, then
> > > > > abinit runs successfully, and appears to give plausible results. This,
> > > > > however, seems like hacking about in the dark.
> > > > >
> > > > > Does anyone know what's going on?
> > > > >
> > > > > Ian