Hi Everyone,
Thank you for your quick responses.
@ Alaxander:
>Does this work use SIESTA??? I thought that non-local XC functionals are
not implemented in >SIESTA and the van der Waals interactions are added a
posteriori as a force-field correction. >(i.e. the basis set has no
relevance to vdW at all (in SIESTA))
Yes this work is using siesta, it is the trunk version. The stable versions
of siesta do not have this feature yet and need to rely on force field
corrections.
@ Herbert:
You mentioned for BSSE:
>You can use this in siesta for energies, but not forces or geometry
optimisation.
I'm actually interested in forces since I intend to do phonon calculations,
does this mean that BBSE corrections for the force constant calculation
will not play much of a role?
I have another question I tried doing calculations for my system with TZDP
basis with 8 nodes and the memory requirement per thread was < 1.75 GB, but
upon using a basis set > TZDP (given below) the memory requirement suddenly
jumps upto > 6 GB per thread. I was wondering if using TZDP+BSSE or diffuse
orbitals would reduce the memory requirement?
Basis for C:
%block PAO.Basis
C.vdw 2 0.0
n=2 0 3 P 0
0.0 0.0 0.0
n=2 1 3 P 3
0.0 0.0 0.0
%endblock PAO.Basis
Regards,
Juzar Thingna
Department of Physics,
Center for Computation Science and Engineering,
National University of Singapore.
On Mon, Jan 17, 2011 at 4:45 PM, Herbert Fruchtl <
[email protected]> wrote:
> These are different, but related. issues:
>
> BSSE is an error that comes from the incompleteness of atom-centred basis
> sets.
> Basically, if you add atoms, you add basis functions. If you move atoms
> closer
> together, they can use each other's basis functions, which "improves" the
> basis. The result is an artificial lowering of energies for compact
> systems,
> and therefore an attractive force. You can fix this by using counterpoise
> correction, i.e. placing "ghost atoms" (basis functions without electrons
> or
> nuclei) where the atoms would be. You can use this in siesta for energies,
> but
> not forces or geometry optimisation. Just use the negative atomic number
> (i.e.
> -6 for a "ghost C").
>
> Van-der-Waals forces are mainly based on dispersion, which is neglected or
> badly
> described by most DFT functionals. You can add an empirical force field
> term,
> based on work by Grimme. This is implemented in Siesta 3.0 (see section
> 6.20 of
> the manual).
>
> And finally, long-range interactions need diffuse basis functions. This
> also
> reduces BSSE (but doesn't remove it completely, and won't give you
> dispersion).
> You find an application example in Phys Rev B 79, 075441 (2009). In
> practice,
> you have to optimise basis functions. There is an example script in the
> siesta
> distribution, but it's tricky.
>
> HTH,
>
> Herbert
>
> Quoting Juzar Thingna <[email protected]>:
>
> > Hello Everyone,
> >
> > I'm working with vdw interractions for C in a double wall carbon
> nano-tube.
> > What I have realized is that using the vdw requires very good basis set
> and
> > the regular DZP doesn't work anymore. Even TZDP is not sufficient,
> according
> > to Guillermo Roma慨-Pe愉ez et al. PRL, 103, 096102 (2009), they needed a
> > TZDP optimized basis + BSSE corrections.
> > Please could some explain of how to actually apply BSSE corrections?
> > Also I read some discussions on the previous threads about using diffuse
> > orbitals instead of BSSE, how do i use diffuse orbitals in siesta?
> >
> > If both the above options can be used I would like to know which one is
> > computationally more efficient.
> >
> > P.S. If possible please could you provide examples.
> >
> > Regards,
> >
> > Juzar Thingna
> > Department of Physics,
> > Center for Computation Science and Engineering,
> > National University of Singapore.
> >
>
>
> --
> Herbert Fruchtl
> Senior Scientific Computing Officer
> School of Chemistry, School of Mathematics and Statistics
> University of St Andrews
> --
> The University of St Andrews is a charity registered in Scotland:
> No SC013532
>
>
> ------------------------------------------------------------------
> University of St Andrews Webmail: https://webmail.st-andrews.ac.uk
>
>
