The Hartree-Fock
method has important limitations. Even
the Hartree-Fock limiting energy is above the true energy of a molecule. This is because correlation energy is not
included. The correlation energy can be
defined as:
Ecorrelation = Etrue – EHartree-Fock
·
This error arises because the interaction
of one electron with another is treated as the interaction with a smoothed out,
averaged electron density.
·
In fact, the position of one electron
affects the position of the other electron, because they repel one
another.
·
If one is on one side of the molecule,
the other electron is likely to be on the other side.
·
Their positions are correlated, an effect
not included in a Hartree-Fock calculation.
·
While the absolute energies calculated by the Hartree-Fock method
are too high, relative energies may still be useful.
·
If the correlation energy does not change during a process, the
energy calculated for the process with the Hartree-Fock method will be close to
the real value.
·
The correlation energy does not change very much if the bonding in
the molecule does not change much.
·
So, for example, calculations of conformational properties can be
done accurately at the Hartree-Fock level (e.g. calculating the rotational
barrier in ethane).
When the bonding in a molecule changes, though, the correlation
energy changes significantly. This
means that the Hartree-Fock method performs very poorly for calculations on
breaking chemical bonds. The
dissociation energy predicted is too high, and it may predict the wrong
products to be formed (e.g. predicting the formation of ions when in fact
neutral atoms should be formed by breaking a bond).
For treating chemical reactions, we need to use methods
which include electron correlation.
Many methods of this type use a Hartree-Fock calculation as a starting
point. An example is the use of perturbation
theory, developed by Møller and Plesset and available in many ab initio
molecular orbital programs. The term MP2 is used for
second-order perturbation theory. A
more sophisticated (and computationally demanding) approach is called configuration
interaction (CI). In this
technique, different electronic configurations are allowed to interact, to give
a better overall wavefunction. The
improved, configuration interaction, molecular wavefunction is written as a
linear combination:
Yimproved, CI = aY0 + bY1 + cY2 + …. (31)
· Y0 is the wavefunction found first (e.g. from a Hartree-Fock
calculation).
· Y1, Y2 etc. are
wavefunctions based on the original Y0 in which some
electrons have been moved into higher energy molecular orbitals which are
unoccupied in Y0.
· This gives a
better overall wavefunction, Yimproved, CI. CI calculations tend to be extremely
demanding of computer resources, because many excited electronic configurations
(Y1, Y2…) are possible. Usually only a limited number can be
included.
Recently, techniques based on density-functional theory
have become popular. The energy of the
molecule is calculated as a function (strictly a functional) of the electron
density.
· Density-functional
theory calculations on molecules usually include a Hartree-Fock
calculation.
· They are
popular because they include electron correlation, but are not as demanding of
computer resources as methods such as MP2 and CI.