The Quantum Defect Orbital (QDO) (Simons 1974; Martín & Simons 1976; Lavín et al. 1992) formalism and its relativistic (RQDO) (Karwowski & Martín 1991; Martín et al. 1993) version have been described in detail. We shall, thus, only mention here some aspects of the theory that are relevant to the present calculations.
The relativistic quantum defect orbitals are determined by solving analytically a quasirelativistic second-order Dirac-like equation with a model Hamiltonian that contains the quantum defect (Karwowski & Martín 1991). This model Hamiltonian allows for an effective variation of the screening effects with the radial distance and as a consequence, the radial solutions behave at least approximately correctly in the core region of space, and display a correct behaviour at large radial distances. These can be expected to be, in most cases, the most relevant regions that contribute to the transition integral. The relativistic quantum defect orbitals lead to closed-form analytical expresions for the transition integrals. This allows us to calculate transition probabilities and oscillator strengths by simple algebra and with little computational effort. The LS coupling scheme has been adopted in all the studied transitions.
One aspect of the computational procedure which has to be investigated is
how to correct the transition operator for the neglect of explicit
core-valence correlation in the quantum defect orbitals. It is well known
that core-valence correlation significantly contributes to certain
transition matrix elements. Laughlin (1989) has pointed out
that when matrix elements are calculated with wavefuntions obtained from a
valence-electron equation, core-polarization corrections introduced in the
transition operator are especially adequate. Different forms for these
corrections can be found in the literature. As in some of our previous
computations (Lavín et al. 1993) we have adopted
(Bielínska-Waz 1992) 
where 
is the core-polarizability and 
is a cutoff radius.
This expression offers the great advantage, unlike the one formerly
proposed by Caves & Dalgano (1972), of retaining total
analyticity in the RQDO transition integrals.
In the Relativistic Quantum Defect Orbital (RQDO) method energy level data are required in order to obtain the quantum defects. We have taken the available energy data from the most recent critical compilation (Kelly 1987). For CuXVI to Kr XXIII we have used the energy values by Sugar et al. (1990) and for RbXXIV to MoXXIX those reported by different authors (Sugar et al. 1990; Jupén et al. 1991; Ekberg et al. 1992).
Concerning the accuracy of these energy data, the energy levels
reported in the critical compilation (Kelly 1987) are only
the lines that have been observed (not calculated). The rest of the
experimental sources (Sugar et al. 1990; Jupén et al.
1991; Ekberg et al. 1992) are quite more explicit as
to the procedures followed in obtaining their data. Sugar et al.
(1990) employed tokamak-generated plasmas for Cu to Mo ions, and
laser-produced plasmas for Cu to As ions. Jupén et al. (1991)
used the JET tokamak for their experimental measurements. And Ekberg
et al. (1992) produced linear plasmas in a line-focused laser beam.
In the three cases the wavelengths for the different transitions observed
were combined with calculated ones (through relativistic methods). Then,
the difference between these quantities was plotted against the atomic
number for the isoelectronic sequence, smooth curves being obtained in all
cases. This fact allowed the authors to make interpolations for
non-observed ions. The wavelengths were then used to derive energy levels.
Sugar et al. (1990) claim the uncertainty of observed
wavelengths to be 
and that of their assignments to be of
. Jupén et al. (1991) conclude that their results confirm
Sugar et al. (1990)'s findings and claim the tokamaks to be
highly competitive light sources for accurate spectroscopy of highly charged
ions. Finally, Ekberg et al. (1992) remark that from their
isoelectronic comparisons smooth energies for ions not observed by them
have also been obtained, and all their results are in agreement with those
reported by Sugar et al. (1990).
The above analyses made us confident in the correctness of the aforementioned energy data and decided to employ them in our calculations. Still there we some gaps in the energy data as regards some of the presently studied ions. These gaps have been filled by our own energy values calculated with the MCDF code as developed by Dyall et al. (1989).
The ionization energies (I.P.) are also needed for our calculations. The values corresponding to the ions with Z-values up to 36 have been taken from Kelly (1987). However, for obtaining the ionization energies from RbXXIV upwards, we have used a Z-dependent expresion for extrapolating the I.P. which has been obtained by applying a standard least-squares fitting routine for the data comprised between Z = 24 (CrXI) and Z = 36 (KrXXIII). The quality of the fit can be evaluated through its correlation factor (r = 0.9996465).
Oscillator strengths have been computed both with
and without explicit introduction of core polarization effects in the
dipole transition operator. For the former, we have taken core
polarizabilities from Fraga et al. (1976). Since there is no
analytical way of obtaining the cutoff radius 
, we have chosen a
value equal to the core mean radius, calculated in accord with an
expression given by Chichkov & Shevelko (1981). In the RQDO
calculations where the standard dipole-length transition operator is
employed Q(r) = r, core-valence polarization is always accounted for
implicitly through the semiempirical parameter of the model Hamiltonian.
The MCDF method, as implemented in the code we have employed (Dyall et al. 1989), is a generalization of the Dirac-Fock formalism, using a multiconfigurational trial function in the variation of the total energy. Within this model, an atomic state function is represented as a linear combination of configuration state functions built from antisymmetrized products of standard Dirac orbitals. These orbitals are eigen-functions of the one-electron total angular momentum operator rather than of the spin and orbital momentum operators independently (as it is the case in the non-relativistic multiconfigurational Hatree-Fock method). For details of the atomic MCDF model we refer to the review by Grant (1988).
The extended average level (EAL) mode (Dyall et al. 1989) has been used in the present calculations. We have chosen the results evaluated with the Babushkin gauge, given that they seem to be the most adequate ones in this context, given that in the relativistic limit, the Babushkin gauge corresponds to the length form of the matrix elements.
In order to achieve
accurate results with this method, the most important configurations have
been included. For the ground configuration 
, several
authors (Flambaum & Sushkow 1978; Biémont & Hansen
1986) found that the highly excited 
configuration gives
the largest interaction. Similarly, the excited 
configuration interacts strongly with states of the 
configuration (Dembczynski & Rebel 1984). The important
perturbation that appears between 
and 
is particularly important for the 
term, as Biémont (1986a) noticed. Consequently, these
configurations have been included in our MCDF calculations for the ground
and excited states. Computations which add the far-off configurations in
which one electron belongs to highly excited orbitals have not been
performed, given that the effects of these interactions may be considered
to be negligible in calculating both energy levels and oscillator
strengths.