From rwoods@biop.ox.ac.uk  Tue May  4 13:00:45 1993
Date: Tue, 04 May 1993 12:00:45 +0100
From: rwoods@biop.ox.ac.uk
To: chemistry@ccl.net
Message-Id: <0096BFFC.3563DFC0.21568@biop.ox.ac.uk>
Subject: MNDO ESP-Charges


Dear Netters,

In reply to Joe Leonard's question regarding ESP-charges derived from an
MNDO wavefunction, we have recently examined the use of such charges in a 
force field for carbohydrates.  We looked at the ability of these charges
to reproduce hydrogen bond energies and dipole moments.  Scaled MNDO ESP-
charges reproduced the HF/6-31G* hydrogen bond energies within about 0.5
kcal/mol.  (Work submitted to JACS)  Previously we found that HF/6-31G*
ESP-charges could lead to an excellent reproduction of the SCF-dipole
moment (J.Comput.Chem 11 (1990)297-310) so we had a look at the MNDO case.
We and others (Breneman and Wiberg, J. Comput. Chem. 11 (1990) 361) have 
noted that charge convergence requires that the ESP should be sampled at
a surprisingly high density of points.  As far as where the points need to
be located, we have found that as they should be between the VDW surface
and about 3 angstroms of it.  In our recent work with MNDO we modified the
WILLIAMS surface in MOPAC_6.0 such that the default GRID spacing was 
decreased from 0.8 to 0.5 angstroms and extended the sampling shell from
1.2 to 2.8 angstroms.  Such an arrangement produces about 750 points per
atom.  We also compared the results with a random point location method
filling the same volume around the molecules with about 1500 points per
atom (PDQC method).  The extended high density grid seemed to be 
pretty good and was a bit faster.

As far as dipole moments are concerned, with ab initio ESP-charges, 
excellent agreement between the SCF and ESP-charge-derived moments can
be expected (differences of less than 0.01 Debye).  We could not get anywhere
near as good agreement from the MNDO wavefunction.  The unscaled ESP-charges
underestimated the MNDO SCF-dipole moment by about 0.3-0.4 Debye.  The scaled
ESP-charges underestimated the HF/6-31G* SCF-dipole by about 0.1 Debye.
This feature was reported quite early (J. Comput. Chem. (1990), 11, 431) but
has not been investigated fully.

Hope this helps,

Rob Woods

Robert J. Woods
Glycobiology Institute
Department of Biochemistry
University of Oxford
South Parks Road
Oxford, OX1 3QU