From:	mforster \\at// nibsc.ac.uk (mforster)
Date:	Tue, 17 Mar 1998 15:29:59 GMT
Subject:	summary

Dear Net Chemists

Here is a summary of responses to my recent question concerning
calculation of intermolecular interactions by electronic structure
methods. There were no responses to the part of the query asking
how the use of numerical basis sets, as opposed to contracted gaussians,
would alter the effect of basis set superposition error.

Here is the question itself.

> Could some Quantum chemistry experts please offer some advice or
> comments on the computation of intermolecular interactions by
> quantum methods.
>
> I noticed a recent paper in which the the difference in the (SCF)
> energy of an ion pair at some separation (r) relative to the the sum
> of the energies of the individual ions was used to estimate the stabilisation
> of the ion pair. Does this approach neglect basis set superposition effects
> where the 'effective' basis set size grows as the two separate molecules
> are brought together. In addition how would such calculations be affected
> by numerical basis sets (eg as used by DMOL).

Here are the responses:

From: scheiner ^at^ chem.siu.edu (Steve Scheiner)

intermolecular interactions
    
in reply to your question on the net about interactions, your question is a
good one, and one that has generated quite a bit of work over the years.
the approach you mention may or many not have included a BSSE correction;
you would have to read the details to know for sure.  about the second part
of your question, the basis set has a great deal to do with the result
computed. so i can't answer your question except to refer you to the
original paper.
  for more details, you might consult a book which i recently edited for
wiley.  the title of the book is Molecular Interactions.  you would
probably be most interested in chapters 3, 5 and 9.
-----
Steve Scheiner, Professor
scheiner;at;chem.siu.edu
Dept. of Chemistry & Biochemistry
Mail Stop 4409
Southern Illinois University
Carbondale  Illinois 62901
ph: 618/453-6476  fax: 618/453-6408




From: noertema at.at theochem.tu-muenchen.de (Folke Noertemann)

Dear Dr Forster,
 since I am not really an expert, I may be telling you things
that you are already aware of.
Of course the effective basis, as you call it, grows if
the distance between two molecular fragments is decreased.
However, it is possible to correct the energies for the BSSE
and so I think your first question cannot be answered without
having seen the paper.
Moreover the BSSE does not depend on the distance alone but
also one the quality of the basis. It is for example possible
to get a very low BSSE with a minimal basis-set calculation
because for reasonble distances only the diffuse functions
centered on fragment a can be used for the representation
of orbitals on fragment b.

Unfortunately I do not have any experience at all with
numerical basis sets and thus I am unable to comment on
your second question.

I hope this is of some help to you,
  F. Noertemann

Folke Noertemann
Lehrstuhl f. Theoretische Chemie
TU Muenchen
Lichtenbergstr.4
85748 Garching
noertema at.at theochem.tu-muenchen.de


From: Richard Wheatley 

This is an extremely widely studied area of chemistry.  Despite the many
hundreds of postdoc-years and hundreds of postgrad-years spent on it,
there is still no way of doing accurate, cheap calculations for any system
with more than about 20 electrons or more than about 3 degrees of freedom.
By accurate I mean guaranteed within about 10%, and by cheap I mean less
than a few months of supercomputer time for the surface.  Don't let anyone
convince you otherwise!

Having said that, SCF calculations will be
(1) in error due to the incorrect polarizability of the negative ion.
This is easy to look up (both SCF and correlated results are known).
Also the (fairly small) dispersion interaction energy is neglected.  If
the ions are not spherical, SCF will also get the multipoles wrong and
this will give major errors.
(2) free of BSSE (reasonably!) provided that at least 10 - preferably more
- Gaussian functions are used per spin-orbital, so 100 Gaussians would
just be tolerable for F- and Na+, for example.  The Counterpoise method
can always be used to estimate BSSE if in doubt.

I don't know about numerical basis sets.  Note that BSSE gets MUCH WORSE,
not better, for correlated calculations with the same basis set.

For light reading, try B Jeziorski, S S Xantheas, A J Stone or possible
some of the P W Fowler work.  Or me.

Best wishes
 
Richard Wheatley, Department of Chemistry, University of Nottingham.



  Dr Mark J Forster Ph.D.
  Principal Scientist
  Informatics Laboratory
  National Institute for Biological Standards and Control
  Blanche Lane, South Mimms,
  Hertfordshire EN6 3QG, United Kingdom.

  Tel 	01707 654753
  FAX 	01707 646730
  E-mail 	mforster : at : nibsc.ac.uk



Similar Messages
03/13/1998:  intermolecular interactions
12/10/1995:  basis stes
08/01/1996:  Re: CCL:M:Heat of formation calculation using MOPAC.
03/16/1995:  Re: CCL: BSSE
08/01/1995:  Spin contamination, effect on energy and structure.
03/16/1995:  Re: CCL:BSSE (Quantum ab initio)
12/16/1994:  Spin contamination & AM1 "ROHF" versus UHF
05/01/1997:  calculation of chemical shifts (summary)
02/23/1999:  SUMMARY: colour of molecules in the solid state (fwd)
06/28/1995:  Re:POSTED RESPONSES: Quantitative assessment of novel ligands


Raw Message Text