From:	Marcus G Martin <marti108 <-at-> gold.tc.umn.edu>
Date:	Tue, 4 Feb 1997 13:41:25 -0600 (CST)
Subject:	Modeling Electrostatics - summary

Here is the summary of useful responses to my question about fitting an
electrostatic potential.  The general feeling appears to be that
population analysis is not the way to do this.

my original question
   What should I do if I want to fit the electrostatic potential of a
molecule?  I have been fitting lennard-jones 12-6 force fields to alkanes
and would like to move into the (frightening) world of slightly charged
molecules.  Nothing extremely painful, such as water, but more like
alkanols or carboxyllic acids.
	The running discussion on charges appears to me to say that the
charges computed from quantum mechanics are of little use for fitting a
force field.  I would like to know if the computational community has a
feel for what the 'best' and most efficient way of fitting point charges
to these molecules would be.  thanks, I will summarize the responses if
there is interest.

[1] reply From: Konrad Hinsen T
> The running discussion on charges appears to me to say that the
> charges computed from quantum mechanics are of little use for fitting a
> force field.  I would like to know if the computational community has a
 
Right.

> feel for what the 'best' and most efficient way of fitting point charges
> to these molecules would be.
   There has been a long discussion about this topic a while ago; you should
find it in the archives. Basically, you fit the charges to reproduce the
electrostatic potential at a certain distance from the molecule (more or
less where you would find the atoms of the neighbouring molecules).
Personally, I consider it essential to use an SVD-based fitting
techniques, since otherwise you will suffer from instabilities (see any
textbook on least-squares problems for an explanation). I haven't found
anyone openly disagreeing with this statement, but on the other hand
almost no one uses SVD-based methods, because the standard programs don't
offer them.

Konrad Hinsen
 |
E-Mail:hinsen %! at !%
ibs.ibs.fr----------------------------------------------------
Laboratoire de Dynamique Moleculaire | Tel.: +33-4.76.88.99.28 Institut de
Biologie Structurale | Fax:  +33-4.76.88.54.94 41, av. des Martyrs |
Deutsch/Esperanto/English/ 38027 Grenoble Cedex 1, France |
Nederlands/Francais

[2] reply
From: Shawn Huston 
  Fit your point charges to an an obervable for the environment you
expect the model to mimic. For example,if you need to represent the
(polarized) charge distribution a moleculewill exhibit in a polar medium
 (such as water) then you want a set of point charges that yields at
least the molecular dipole moment in water. Details can be found in many
papers by W. Jorgensen (look at the early ones from the 80's where you
will find more information).
Shawn

Shawn Huston Sr. Marketing Scientist (don't forget the trailing 'h' in my
email
alias: shawnh ( ( at ) ) mdli.com) MDL Information Systems Phone:  510-895-1313 X
1145 14600 Catalina Street Fax:  510-483-4738 San Leandro, CA 94577
http://www.mdli.com
 
[3] reply
From:dew01.,at,.xray5.chem.louisville.edu(Donald E. Williams)

Hi,
	You may be interested in my paper which appeared in J. Comp. Chem.
1994, 15, 719.  The MEP of n-alkanes is not well represented by net atomic
charges.  Additional charge sites at methylene bisector locations allow a
good fit.  Also, I recently announced the software pdm97 on ccl, which
fits site multipole models to the MEP.  I assume that you got that, but if
you didn't, let me know and I will send it directly to you.

-Donald Williams dew01-0at0-xray5.chem.louisville.edu

[4] reply From: "Dr. Heinz Schiffer" 
Hi Marcus,
I think there is a misunderstanding. There are no such things like
charges of atoms in a molecule in quantum mechanics. They are not
observables. Every partitioning of the charge density of a molecule into
atomic ( or other ) subunits are completely arbitrary ! O.k.  this is a
little bit overdone. The population analysis of a molecular wavefunction
is only useful for interpretation purposes, like : which atom is more
negatively charged and is likely to be attacked by an electrophile an so
on...  The electrostatic potential of a molecule on the other hand is an
observable and rigorously defined in quantum mechanics. This potential
can be calculated point wise in the space around a molecule with most ab
intio and also with some semi-empirical programs. This potential can than
be modelled by a point charge model of a molecule, that is you adjust the
charges of the model so, that the potential  calculated with these
charges is (nearly) the same as the ab initio calculated one.  There
exist many schemes around for fitting charges to the electrostatic
potential, but they all try to model the electrostatic potential of the
molecule. Population anlysis charges are never meant to modell the
electrostatic potential of a molecule, they serve "only" for
interpretation purposes of a molecular wave-function.
The best way to get charges for your purpose is to fit them to the
electrostatic potential.
Hope this helps,
Ciao
Heinz
Dr. Heinz Schiffer
Phone ++49-69-305-2330-- Hoechst CR&T Fax ++49-69-305-81162 Scientific
Computing, G864 Email schiffer &$at$& h1tw0036.hoechst.com 65926 Frankfurt am
Main schiffer -x- at -x- msmwia.hoechst.com



Similar Messages
01/30/1997:  Modeling Electrostatics
07/07/1995:  Summary: Opinions on the quantum charges
10/25/1996:  charges and potentials
09/27/1994:  Charges
01/23/1998:  SUMMARY: QEq charges & applications
10/02/1996:  SUMARY:PD atomic charges
09/27/1994:  charges
12/10/1993:  Population analysis and dipoles
10/24/1996:   Charges
06/30/1994:  CCL:SUMMARY:elec.pot.:basisset depend.


Raw Message Text