Re: [SIESTA-L] BSSE and pi-stacking

[Gregorio García Moreno]

Hi all again
As I told you, I have a problem with pi-stacking distance, which is 
smaller than the typical experimental value .According with your 
suggestions I have solved my problem from two different ways: .
a) I have reduced PAO.EnergyShift to 0.15 eV (before 0.30 eV), 
calculated pi-staking distance 3.548

b) I have used a DZ2P for S and C (my system has C, H and S) basis sets. 
However, I don't know it is well built.  Calculated pi-staking distance 3.52

%block PAO.Basis
C           2                    # Species label, number of l-shells
 n=2   0   2                         # n, l, Nzeta
   0.0   0.0
   1.000      1.000
 n=2   1   2  P   2                   # n, l, Nzeta, Polarization, NzetaPol
   0.0     0.0
   1.000      1.000
S           2                    # Species label, number of l-shells
 n=3   0   2                         # n, l, Nzeta
   0.0      0.0
   1.000      1.000
 n=3   1   2 P   2                   # n, l, Nzeta, Polarization, NzetaPol
   0.0      0.0
   1.000      1.000
%endblock PAO.Basis

Does DZ2P basis set well built? If so, which is the most correct 
solution? I think that a)........

Thanks

Gregorio




El 03/06/2010 15:47, Pablo A. Denis escribió:
Hi Gregorio,

                       The BSSE is HUGE, and the results without BSSE 
are meaningless. It can be 0.5 eV for non bonded interactions. If you 
working with molecules you can do the BSSE optimization by hand or 
write a short script. For the benzene dimer the changes in the 
equilibrium distance when you optimize including BSSE are around 0.4 A.
In some cases you may need to consider the monomer deformation energies.

You can try to reduce the PAO.EnergyShift to 0.001 Ry, this helps a 
lot to reduce BSSE. I have tried to use a TZP basis set but as you 
said it was not useful, maybe you need a optimized basis set.


Best Regards,

                       Pablo



----- Original Message ----- From: "Herbert Fruchtl" 
<herbert.fruc...@st-andrews.ac.uk>
To: <siesta-l@uam.es>
Sent: Thursday, June 03, 2010 8:44 AM
Subject: Re: [SIESTA-L] BSSE and pi-stacking


The Van-der-Waals correction is an additional attractive force, 
correcting something that traditional GGAs don't describe. It lowers 
interaction energies and shortens bonds.

BSSE is a spurious and unphysical attractive force. It also lowers 
interaction energies and shortens bonds. If you already have this 
error and add VdW correction, you will make things worse.

We once did a quick study on the sandwiched benzene dimer, to find 
the "correct" S6 parameter (the scaling for the VdW interaction) for 
Siesta's default basis sets. We abandoned this, because the 
well-depth of the interaction was already massively overestimated 
through BSSE (I vaguely remember a factor of 8) compared to 
high-level literature results.

I don't know a solution to this problem. Arguing that BSSE somehow 
simulates VdW interaction is of course dodgy. Besides the massive 
overestimation of its magnitude, it will have the wrong distance 
dependence. I don't know a solution to this problem, except 
increasing the size of the basis set, but I suspect it would have to 
be very big to remove BSSE. In the molecular case with Gaussian basis 
sets this is well studied, but I suspect that the limited range of 
basis functions in Siesta makes things worse.

  Herbert

Ricardo Faccio wrote:
Hi Gregorio
When you talk about Pi-stacking... do you mean Van der waals 
interactions?
If this is the case remember the drawbacks of:
LDA: cancellation of errors with a near overbinding tendency
GGA: VdW interactions are totally neglected
Please, consider the use of the new VdW xc-potential introduced in 
the new
versoin of siesta.
Regards
Ricardo
p.s.: it has nothing to do with BSSE

Ok, thanks.
I tried with TZP basis set, but the results don't improve enough.
Respect to use BSSE, I agree with you about the problem to use with my
system, since the are two indentical molecules (interconvertible by
symmetry) in the unit cell.
Finally, I don't know how to include diffuse functions in SIESTA.

Thanks

Gregorio


El 03/06/2010 12:49, Herbert Fruchtl escribió:
If you mean counterpoise corrected geometry optimisation, that's
tricky, and definitely not implemented in SIESTA. Counterpoise is 
only
really straight-forward for dimers (or at least small numbers) of
fairly rigid molecules. In a crystal, you would have to make sure 
that
your supercell is big enough (you can't CP correct the interaction of
a molecule with its image), and I don't know the implications for
K-points (my first inclination would be to say if you need more than
one K-point, your system is too dense for counterpoise to be
meaningful).

Have you tried adding diffuse functions? BSSE is a consequence of the
incompleteness of localised basis sets. If you describe the
wavefunction far away from the nuclei better, it should improve, 
but I
don't know how quickly.

HTH,

  Herbert

Gregorio García Moreno wrote:
Hi all
I am studing polymers pi-stacked. I have obtained good crystal
parameters, except  for the pi-stacking distance, which is smaller
than experimental data. Here, you recomended me that I should use 
TZP
and BBSE. I have tried  with TZP, but the results do not improve
respect to DZP. However, I have searched in literature, and to apply
BSSE could solucionate my problem.
In the SIESA-list I have not found how to optimize with BSE, can
someone help me?
Thanks in advance

Gregorio



--
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 


<<attachment: gjgarcia.vcf>>