The second summation in the CC expansion (Eq. 1)
represents short range correlation functions
that need to be optimized in order to obtain accurate (e + ion)
wavefunctions and to avoid pseudo-resonances that result when the two
summations
are not consistent (Berrington et al. 1987). In the present work,
the set of 
functions was optimized for each set of
symmetries of Fe I with spin multiplicity 
, and 7, in a
careful manner so as to obtain accurate energies of the
bound states of Fe I, as compared to experimental energies.
In order to keep the calculations
computationally tractable one tries to limit the bound channel set
to be as small as possible. However, for Fe I the sets of
(N+1)-electron functions needed were very large
for symmetries with
multiplicity 1 and 3, and had to be truncated in view of our present
computational capabilities on the Cray Y-MP with 64 MW memory and
2 Gb disk space limit for the
diagonalization of the dipole matrix in the STGH program.
This computational constraint affected the accuracy of the results for
the singlet and the triplet symmetries, as discussed in the next section.
The present calculation for quintets and septets symmetries included a total of
56 (N+1)-electron functions which produced nearly 1650 configurations, 150
channels, and a Hamiltonian matrix with a maximum dimension of nearly 2000.
For singlets and
triplets symmetries 65 (N+1)-electron functions were included
which gave rise to almost 2900
configurations and 152 channels; the dimension of the Hamiltonian
matrix was nearly 3000.
Table 3 presents the entire list of the (N+1)-electron correlation
configurations included (these need to be known in order to reproduce
or improve the present calculations and results).