The ABINIT Users Mailing List ( CLOSED )

["Andrew M. Rappe" <rappe@sas.upenn.edu>]

Dear Group,

I do not think that d-local will help.  In fact, I guess that it could
make things a bit worse, possibly causing a ghost in the 4p state.

In my view, the problem is the highly localized d orbitals.  When one
uses a very large rc for the 3d of copper, the resulting d orbital is
not localized enough, and it increases the neighboring interactions,
raising B.

The optimized pseudopotential method does a better job of permitting
smaller rc's and low planewave cutoffs than T-M for problem elements
like Cu (see http://opium.sourceforge.net).

Of course, the ultrasoft or PAW methods would do a yet better job of
describing the localized 3d than any norm-conserving psp approach,
since they have extra localized functions to help treat states like
the Cu 3d.  But abinit doesn't have these at the current time.

Additional, LDA overbinds relative to GGA, as is well known.
Therefore if you are using LDA, that could cause a somewhat higher B.

Good luck with Cu!
Andrew

On Tue, Feb 25, 2003 at 04:41:54PM +0100 or thereabouts, Thomas Maxisch wrote:
> Dear Gustavo,
> 
> >I have always used the Troullier-Martins pseudopotentials available from 
> >the
> >web. Could the problem be related to this?
> 
> May be. Try to use a pseudopotential with the d of even f orbital as local 
> part. Actually the pseudopotential from the webpage is the only one which I 
> know that has the s orbital as local part. But I am not totally sure.
> 
> Hope this helps,
> Thomas
> 
> 
> -- 
> __________________________________________________________________________
> 
> Thomas Maxisch
> EPFL SB-ITP/CPNMC
> PHB-Ecublens
> CH-1015 Lausanne
> Switzerland
> Tel : +41 (0) 21 693 45 16
> Fax : +41 (0) 21 693 44 44
> Email : Thomas.Maxisch@epfl.ch
> __________________________________________________________________________