>>>> The polarization orbital (only one) is applied atop of other orbitals in
>>>> the basis, i.e. you cannot have P for both s and d. That's why DZP-DZP
>>>> sounds weird.
Hmmm... I did not know that. Thanks!
To be frank, I need to be less lazy and go deeper into the Siesta paper (it is
here, in front of me, 34 pages of fun) to understand better those basis sets
and other things about it.
What is the meaning of TZTP-TZTP-TZTP in the paper of Suarez et al (Journal of
Physics: Condensed Matter 16, 5453)?
>>>> Well, this has nothing to do with BSSE.
This is something I am not so sure. Maybe there is nothing to do directly, but
there is something wrong somewhere. Let me try to explain the problem from the
beginning (I don't remember if I already did that).
In PRB 67, 214103, the authors calculate the enthalpy of solvation of a carbon
atom in alpha-iron as follows (using VASP):
H_n=E(Fe_n+C_1)-E(Fe_n)-E(C)
where E(C) is the total energy per atom of graphite. To try to reproduce those
results with Siesta taking BSSE into consideration, I did that:
H_n=E(Fe_n+C_1)-E(Fe_n+C_1(ghost))-E(Fe_n(ghost)+C_1)
Bad results. I guess it is because, for both E(Fe_n+C_1(ghost)) and
E(Fe_n(ghost)+C_1), the iron unit cell is the same as E(Fe_n+C_1), which is
distorted by the carbon atom. So in order to calculate H_n properly I should
take this deformation energy into consideration, shouldn't I? If so, how to do
that?
I tried to consider a new term in the difference, E(Fe_n)-E(Fe_n*), where
E(Fe_n*) is the iron atom unit cell in the distorted geometry but without the
carbon atom. With that, my intention was to compute the energy required to
deform the cell. Is there anything wrong, mathematical or physically speaking,
in doing so? Anyway, once again, bad results, compared to those given by VASP.
Thanks and sorry if I am bothering everyone with this subject.
Roberto
________________________________
From: Oleksandr Voznyy <o.voz...@usherbrooke.ca>
To: SIESTA-L@listserv.uam.es
Sent: Saturday, December 6, 2008 7:01:54 PM
Subject: Re: [SIESTA-L] Still energy differences and BSSE corrections
> >>>> What is DZP-DZP for Fe (...) ????
> Double zeta plus polarization for both 3d and 4s.
The polarization orbital (only one) is applied atop of other orbitals in the
basis, i.e. you cannot have P for both s and d. That's why DZP-DZP sounds weird.
> other complicated things. How to take into account the deformation caused by
> the presence of carbon? Not only local deformation, but also the change in
> the cell shape?
Well, this has nothing to do with BSSE. Normally you should compare relaxed
systems, not the same geometries.
With a non-complete basis you may have an error in final geometry, e.g. when
you study the relaxation of the molecule on the surface, the overlap of basis
functions changes, and ghosta wouldn't help here. But this error should be
small.
In case of ghosts, you can always relax the system in the presence of ghost
orbital (usually fixing it). But again the effect of the ghost on the geometry
is likely to be small, although the total energy would change significantly.
> Of course, I know you cannot apply the correction only when it interests
> you...
That's the point. You can always try to use a more complete basis from the very
beginning (TZP) hoping that BSSE would be smaller. But in my experience, TZP
reduces significantly the total energy, but the BSSE value remains comparable
to DZP.
