A Simple Potential Model Criterion for the Quality of Atomic Charges
Anselmo E. de Oliveira, Paulo H. Guadagnini, Roberto L. A. Haiduke, and Roy E. Bruns*
Instituto de Quı ´mica, UniVersidade Estadual de Campinas, CP 6154, 13083-970 Campinas, SP, Brazil
ReceiVed: December 17, 1998
The simple potential model has been shown to be useful in relating core electron binding energies measured
in the X-ray region with mean dipole moment derivatives obtained from experimental infrared vibrational
intensities. The importance of including relaxation corrections to the experimental 1s ionization energies of
sp, sp
2
, and sp
3
hybridized carbon atoms are investigated here. Although relaxation energies obtained from
6-31G(d,p) and 6-311++G(3df,3p) basis sets using ΔSCF calculations show differences of about 1 eV for
most molecules studied, relative differences are of the order of 0.1 eV. Exceptions are the CO, CO
2
, COS,
and CS
2
molecules where discrepancies are larger. Relaxation energy corrections improve simple potential
model fits with mean dipole moment derivatives for all carbon atom models but is most pronounced for the
sp hybridized atoms. The simple potential model corrected for relaxation energies is investigated as a criterion
for testing the quality of Mulliken, CHELPG, Bader and GAPT carbon atomic charges calculated from MP2/
6-311++G(3d,3p) wave functions. The GAPT charges are in excellent agreement with the experimental
mean dipole moment derivatives (within 0.067e) and provide superior statistical fits to the simple potential
model when compared with those obtained for the other charges.
Introduction
Recently, the simple potential model
1
proposed 30 years ago
by Siegbahn and collaborators has been shown to be useful in
relating core electron binding energies measured in the X-ray
region with polar tensor invariant quantities obtained from
experimental infrared vibrational intensities.
2
Separate models
relating the carbon 1s core electron binding energies to their
mean dipole moment derivatives were found for sp, sp
2
, and
sp
3
hybridized atoms. Furthermore, the models’ parameter values
were shown to be inversely dependent on the carbon atom
covalent radii and identified with the Coulomb repulsion
integrals involving core and valence electrons. This was
confirmed in our later study relating 2p and 3p electron
ionization energies of Si and Ge with their atomic polar tensor
matrix traces.
3
The potential model is expected to accurately relate core
electron binding energies and mean dipole moment derivatives
if two conditions are fulfilled: (1) the relaxation energies of
the ionization process being investigated are negligible or
constant and (2) the mean dipole moment derivatives can be
identified with atomic charges. These two assumptions are
investigated in this work.
To maintain its simplicity and usefulness for the interpretation
of experimental results, the potential model was not designed
to contemplate the reorganization of electron densities in
molecules during the ionization process. Relaxation energies
can be used to adjust the experimental ionization energies to
compensate for this reorganization so that the modified energies
are appropriate for use in simple potential model applications.
These adjustments, whether using values calculated from either
the empirical equivalent cores method
4-6
or the ΔSCF method
7-9
from 6-31G(d,p) wave functions, improve agreement of potential
model fits of experimental carbon mean dipole moment deriva-
tives to experimental 1s carbon atom ionization energies of the
fluorochloromethanes.
2,10
However, the relaxation energies
calculated by each of these methods are not constant. For
example, 6-31G(d,p) ΔSCF relaxation energies for the fluoro-
chloromethanes have variations, more than an order of magni-
tude larger than the estimated experimental errors in their
measured ionization energies. Further improvement of the
potential model fits might be expected, especially for molecules
with sp hybridized carbon atoms since their 1s ionization
energies for CO, CO
2
, OCS, and CS
2
were found to result in
large deviations when mean experimental dipole moment
derivatives are used as atomic charges. Here more extensive
wave functions, calculated with a 6-311++G(3d,3p) basis set
are applied to the 1s ionization energies investigated in our
previous work in order to further test the importance of
relaxation effects on simple potential model applications.
The second assumption is not so easily tested. There is no
universally accepted method of calculating atomic charges, and
no experimental technique is available to measure them directly.
The mean dipole moment derivative, on the other hand, can be
determined using only experimental fundamental vibrational
frequency and infrared intensity data and common molecular
parameters obtained from experimental sources, such as atomic
masses, molecular dipole moments, bond distances, and angles
between bonds. Futhermore, it can be calculated from the same
molecular orbital wave functions used to calculate other kinds
of charge estimates. In fact, population analysis using mean
dipole moment derivatives, also called GAPT (generalized
atomic polar tensor) charges, has been proposed by Cioslowski
11
and does not require any direct reference to the basis set used
to calculated the molecular wave function. However, dipole
moment derivatives, just as dipole moments, have not been
considered reliable sources of atomic charge values since
molecules do not appear to be describable by spherical nonde-
formable charge distributions centered on their nuclei. Besides
the static contribution, results of molecular orbital calculations
* Corresponding author. Fax: 55-19-788-3023. E-mail: bruns@
iqm.unicamp.br.
4918 J. Phys. Chem. A 1999, 103, 4918-4924
10.1021/jp984777e CCC: $18.00 © 1999 American Chemical Society
Published on Web 06/06/1999