[CPMD-list] Extra channels in Vanderbilt PP
Ari.P.Seitsonen at iki.fi
Ari.P.Seitsonen at iki.fi
Sun Sep 15 21:27:12 CEST 2002
Dear Fred,
In case that you haven't got any reply to your answer yet...
(Everyone would answer to the list, wouldn't one?!?!)
> In small molecule calculations, it is common to use basis sets that are at
> least 'double zeta', i.e. two functions per l channel, and often augmented
> by a polarization function (higher l value).
>
> So, when generating PPs, do you follow the same policy as you would with
> gaussian or slater type orbitals, or do you only put in channels for the
> occupied space, and then increase the plane-wave cutoff?
There's a small misunderstanding here: The double zeta etc is related
to the _basis_ set, whereas the projectors in the Vandertbilt pseudo
potentials are for describing the valence-core interaction, and has
thus directly nothing to do with the basis set. Thus the number
projectors per l channel is in principle not related to the size of
the basis set, unless some of the projectors require a higher cutoff.
The advantage of more projectors is to increase the transferability,
or the accuracy, of the pseudo potential, i.e. the valence electrons
feel a potential which is close to the true one caused by the nuclear
charge and the core electrons if they'd be included.
> I've been having problems with convergence properties of the USPPs within
> cpmd3.5.3, and was therefore curious if a smaller potential would help.
Well, you can always try e.g. Troullier-Martins pseudo potentials,
however there's no guarantee that the system would converge even then.
Do you have a metallic system, degenerate/radical one, ...?
Greetings,
apsi
--
-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
Tel +41 1 635 44 97 / Fax +41 1 635 68 38 / GSM +41 79 719 09 35
Anschrift: Physikalisch Chemisches Institut (PCI), Universität Zürich (UniZh)
Indirizzo: Winterthurerstraße 190, CH-8057 Zürich
Address: Schweiz / Svizzera / Suisse / Svizra / Switzerland
More information about the CPMD-list
mailing list