From chemistry-request@ccl.net Sun Jun  1 21:14:48 2003
Received: from sungoddess.ccs.yorku.ca (sungoddess.ccs.yorku.ca [130.63.236.144])
	by server.ccl.net (8.12.8/8.12.8) with ESMTP id h521EmgC007681
	for <chemistry:at:ccl.net>; Sun, 1 Jun 2003 21:14:48 -0400
Received: from curl.gkcl.yorku.ca (curl.gkcl.yorku.ca [130.63.232.224])
	by sungoddess.ccs.yorku.ca (8.12.9/8.12.8) with ESMTP id h521El0O011583
	for <chemistry:at:ccl.net>; Sun, 1 Jun 2003 21:14:47 -0400 (EDT)
Received: (from chan@localhost)
	by curl.gkcl.yorku.ca (8.10.0/8.10.0) id h521Ei102277
	for chemistry:at:ccl.net; Sun, 1 Jun 2003 21:14:44 -0400 (EDT)
From: Wai-To Chan <chan:at:curl.gkcl.yorku.ca>
Message-Id: <200306020114.h521Ei102277:at:curl.gkcl.yorku.ca>
Subject: Re: calculating atomic charges with AIMPAC
To: chemistry:at:ccl.net
Date: Sun, 1 Jun 2003 21:14:44 -0400 (EDT)
X-Mailer: ELM [version 2.5 PL3]
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


<<<<<<<<<<<<<<<,
What "chemical intuition"? The one that comes from looking for many years
to Mulliken populations in gaussian outputs? That is not science, it is
prejudgment. Whatever interpretation we want to give to them, Bader charges
are unique and well defined properties of the electron density of the
molecule or solid under study. Both experimental and good theoretical
electron densities agree on the calculated Bader charges of hundreths of
compounds well characterized so far. On the other hand, Bader charges
>>>>>>>>>>>>>>>>>>>>>>>>>>>>

  Mulliken charges could indeed give counterintuitive results. I remember
it has a serious problem related to basis-set sensitivity. 
  I am not sure what experimental measurements you referred to. Bader
chages, bond paths, lone pairs and so forth are defined in terms of the topology
of electron density a measurable qauntity. So it is only natural 
that theoretical AIM properties should match experimental measurements. 
If you were referring to the transferability of the properties of atoms
or molecular fragments as defined by Bader then AIM may be said to be a 
predictor of experimental measurement in certain respect. For the purpose of 
calculating partial atomic charges AIM is not always a good choice. For 
large size molecules containing elements of ring structures 
the code in Gaussian is very vulnerable to error. PROAIMV is more reliable
but it could be very time consuming. For organometallic systems I suspect
that the use of ECP would cause problems in the analysis of Laplacian 
and perhaps bond critical points as well. 
Use AIM in cases where it provides better
prediction or correlation of whatever you are calculationg. Off the top of my
head I remember AIM does give good correlation with proton affinity for
some carbonyl system. I have done calculations on systems for which
the Natural population charges analysis methods also gave reliable results
for interpretation of site-specific reactivies on small metal clusters. 

Wai-To Chan


<<<<<<<<<<<<<<<<<<<<<<<<<
are not enough to analyze the electron densities. The whole multipolar
series (charges are nothing but the zero order poles) should be considered
instead. Several recent works by Popelier, for instance, would made a
good reading in this respect.

In any case, the charges quoted in the original question look like if
the wavefunction is wrong by some reason. Perhaps a calculation
not converged.

   Regards,
             Victor Lua~na
>>>>>>>