The ABINIT Users Mailing List ( CLOSED )

[Matthieu Verstraete <mjv500@york.ac.uk>]

> ---------- Forwarded message ----------
> On Jan 5 2006, Zhigang Wu wrote:
>
> > Dear Matthieu,
> >
> > When I use anaddb to calculate Tc, I found that it I set
> > telphint  0
> > kptrlatt
> >           32 0  0
> >           0  32 0
> >           0  0  32
> >
> > the calculated lambda is decreasing with the kptrlatt dimension, i.e.,
> > when I set kptrlatt
> >           64 0  0
> >           0  64 0
> >           0  0  64
> >
> > lambda is 1/8 of previous calculated value. It should be very easy to
> > fix it.
> >
> > Zhigang Wu

> Hmmm, strange. Could you be more specific about the rest of the in/outputs
> and what you are doing? If you use the same GKK file for 32x32x32 and just
> change the kptrlatt, it should be complaining loudly. We have some results
> for lambda which agree well with litterature, so I would be surprised if
> there was an additional factor of 1/nkpt too many (which is what is
> suggested by your mail)
>
> Which version of abinit are you using?
>
> Matthieu

> Date: Fri, 6 Jan 2006 10:46:16 -0500 (EST)
> From: "Zhigang (Michael) Wu" <z.wu@gl.ciw.edu>
> To: mjv500@york.ac.uk
> Subject: Re: el-phon
>
> So you mean the kptrlatt must be the SAME as the kmesh in GKK file? I
> raised kptrlatt to check convergence, and I found (1) no complaining; (2)
> lambda and a2F(w) values decreses with kptrlatt^3. The version I used is
> 4.6.5.
>
> Thanks,
>
> Zhigang

Ok, that should be an error: anaddb does not have the eigenvalues on the
64^3 grid, so it can't be using them. The only change must be in the
number of kpoints or the number of tetrahedra (hence the /8).

You really need to redo everything with 64^3 kpoints if you want more
precision.

Matthieu

--
================================================================
Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
Dept. of Physics, University of York,     tel: +44 1904 43 22 08
Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14