Next: Cluster Functions and the
Up: Fundamental Concepts
Previous: Fundamental Concepts
Consider a model system of four electrons moving in an arbitrary
electrostatic field generated by the nuclei in a molecule. For our
purposes, it is not necessary to specify the number of these nuclei,
their types, or positions; only the general form of the electronic
wavefunction is of interest. It is convenient to describe the motions
of each electron separately by assigning them to one-electron
functions,
,
where
x1 is a
vector of the coordinates (including spin) of electron 1. In
addition, electrons are fermions, so the electronic wavefunction must
be antisymmetric with respect to interchange of the coordinates of any
pair of electrons. A traditional and very useful starting point for
such a four-electron wavefunction is the so-called Slater
determinant
 |
(2.1) |
where the
is a normalization constant. Expansion of
this determinant reveals a linear combination of products of the four
functions,
,
,
,
and
,
with the electronic
coordinates
xn distributed among them in all possible
ways. Since permutation of any two rows in the determinant -- which
is equivalent to interchanging the coordinates of any two electrons
-- changes the sign of
,
the antisymmetry principle is
maintained.
The component functions
may be chosen in a variety of ways.
For example, if the nuclear field were only a single beryllium
nucleus, the one-electron spatial functions could be constructed to
mimic the atomic 1s and 2s orbitals. For a molecular system, the
functions can be constructed as a linear combination of atomic
orbitals (AOs) in which each one-electron function represents a
molecular orbital (MO) whose AO coefficients are optimized via the
Hartree-Fock self-consistent-field (SCF) procedure.[82] A
convenient shorthand notation for this wavefunction consists of a
Dirac-notation ket containing only the diagonal elements of the above
matrix,
 |
(2.2) |
where the normalization factor is included implicitly. As discussed
in detail elsewhere in Reviews in Computational
Chemistry[77], the single-determinant wavefunction fails
to account for the instantaneous Coulombic interactions which keep the
electrons of opposite spin apart.[82]
How can we improve this so-called independent-particle approximation
such that the motions of the electrons are correlated? Often the set
of occupied orbitals (i.e., those functions which compose the
Slater determinant above) is chosen from a larger set of one-electron
functions. These ``extra'' functions are frequently referred to as
virtual orbitals and may, for example, arise as a byproduct of
the SCF procedure.2.1 Within the space described by the full set of orbitals,
any function of N variables may be written in terms of N-tuple
products of the
.
For example, a function of two variables
may be constructed by using all possible binary products of the set of
one-electron functions, e.g.,
 |
(2.3) |
where the double-summation runs over the entire set of one-electron
functions and the notation p > q indicates that only unique pairs of
functions are included. Instead of correlating the motions of a
specific pair of electrons, however, we may use a modified form of
this expansion to correlate the motions of any two electrons within a
selected pair of occupied orbitals -- say functions i and j -- using
a two-particle cluster function,
 |
(2.4) |
where the
tijab are the cluster coefficients whose specific
values are determined via the electronic Schrödinger equation (see
the next section on formal coupled cluster theory). Inserting this into
leads to the
somewhat-improved electronic wavefunction,
![\begin{displaymath}\Psi = \vert \left[ \phi_i({\mathbf x}_1) \phi_j({\mathbf x}_...
...) \right] \phi_k({\mathbf x}_3)
\phi_l({\mathbf x}_4) \rangle,
\end{displaymath}](img15.gif) |
(2.5) |
where the Dirac shorthand implies a correctly antisymmetrized
wavefunction including normalization factors as in
Eq. [2.2]. Inclusion of the cluster function, fij,
in the wavefunction produces a linear combination of Slater
determinants involving replacement of occupied orbitals
and
by virtual orbitals
and
,
such that
 |
(2.6) |
In addition, the determinantal form of the individual terms in this
expansion implies antisymmetrization of the cluster coefficients, such
that
tijab = -tjiab = -tijba = tjiba.
It should be carefully noted here that the cluster function,
fij(x1, x2), is intended to correlate the
motions of any pair of electrons placed in orbitals i and j,
and not just the motions of electrons 1 and 2. Since the Slater
determinant produces a linear combination of orbital products,
including terms such as
![\begin{displaymath}\left[\phi_i({\mathbf x}_1) \phi_j({\mathbf x}_2) +
f_{ij} ({...
...thbf x}_2) \right] \phi_k({\mathbf x}_3)
\phi_l({\mathbf x}_4)
\end{displaymath}](img19.gif) |
(2.7) |
and
![\begin{displaymath}\left[\phi_i({\mathbf x}_3) \phi_j({\mathbf x}_4) +
f_{ij} ({...
...hbf x}_4) \right] \phi_k({\mathbf x}_1)
\phi_l({\mathbf x}_2),
\end{displaymath}](img20.gif) |
(2.8) |
which differ only in their distribution of electronic coordinates, the
cluster function correlates the motion of every pair of
electrons found in orbitals
and
.
Depending on the chemical system of interest, however, it might be
more prudent to correlate the motions of electrons in orbitals k and
l rather than orbitals i and j. For example,
and
might correspond to molecular core orbitals, while
and
might correspond to the atomic or molecular valence
orbitals. Electron correlation can be particularly important in the
latter set of functions because the valence orbitals are often
directly involved in the formation of chemical bonds. In this case,
the wavefunction would be written as
![\begin{displaymath}\Psi = \vert \phi_i({\mathbf x}_1) \phi_j({\mathbf x}_2) \lef...
... x}_4) + f_{kl}({\mathbf x}_3,
{\mathbf x}_4) \right] \rangle.
\end{displaymath}](img21.gif) |
(2.9) |
On the other hand, a more intelligent approach might be to correlate
all possible pairwise combinations of orbitals in this four-electron
system, i.e.,
where the electronic coordinates are now implicit in the notation, and
the signs on individual terms arise from the permutations in the
orbital ordering needed to define the appropriate cluster functions.
However, there is no need to limit this approach to only orbital
pairs; following Harris et al.,[80] we could introduce
three-orbital cluster functions and include these in our new
wavefunction to give
If one continues this process to include all cluster functions for up
to N orbitals (four in the case discussed here), as well as
single-orbital ``cluster'' functions which account for adjustment of
the one-electron basis as other cluster functions are added, we could
obtain the exact wavefunction within the space spanned by the
.
On the other hand, we might assume that clusters larger
than pairs are less important to an adequate description of the system
-- an assumption supported by the fact that the electronic
Hamiltonian contains operators describing pairwise electronic
interactions at most[75]. We could therefore write a
four-electron wavefunction which includes all clusters of only one and
two orbitals as[83,80]
Next: Cluster Functions and the
Up: Fundamental Concepts
Previous: Fundamental Concepts
T. Daniel Crawford / crawdad@ccqc.uga.edu
23 November 1998