Introduction: Atomic data is needed in many areas of
scientific endeavor, the development of controlled thermonuclear
fusion being one example. Energy level structure, autoionization
rates, cross-sections for excitation, ionization, charge exchange, and
recombination all enter into model calculations. In geophysical
studies of the atmosphere, for example, the emission features of
atomic oxygen associated with branching ratios for decay from
state are related to the abundance of oxygen in the
thermosphere. Astrophysics has a longstanding need for large amounts
of data. A successful ``opacity'' project for generating such data has
produced vast amounts of data, but all calculations were all performed
in the LS approximation and do not include relativistic effects or
nuclear effects. Heavy elements also are gaining in importance, not
only as a test of theories: data for rare earth elements, for example,
are needed for the development of high intensity discharge lamps, and
data for other elements are relevant to environmental clean-up
problems.
Atomic structure is an area where accuracy is extremely important. In
spectroscopy, observations are related to energy differences. It is
these differences that need to be predicted to spectroscopic accuracy,
which, in many applications, may be defined as a fraction of a
cm-1. This translates to an error of less than 4.5
au. in an energy difference. Even for relatively small
systems, this requirement may present a considerable challenge. A
number of physical effects may need to be included to achieve the
desired accuracy.
Achievements: The inclusion of correlation in the motion of
electrons in many-electron systems is computationally challenging. In
light atoms, using a non-relativistic formalism together with a
parallel version ![[*]](http://www.emsl.pnl.gov/images/latex2html/foot_motif.gif)
of our atomic structure package, called MCHF
a full-correlation study is feasible, put the magnitude of the problem
increases exponentially with the number of electrons. But many atomic
processes, such as transition probabilities, are outer-electron
phenomena. Through the use of systematic methods within a series of
models, we are able to determine properties along with estimates of
uncertainties. This is illustrated in figure 21 where
results are presented for an intercombination line in C III that has
been of great interest in astrophysics. Note the final error bar for
the calculation.
 |
New Opportunities: As accuracy is increased, or heavier atomic
systems are considered, it becomes desirable to use a fully
relativistic formalism that includes the effect of the finite size of
the nucleus. In heavy atoms the latter is an extremely important
effect. In our work to date, we have modified the General
Relativistic Atomic Structure Package (GRASP) developed at Oxford
under the direction of I.P. Grant. In particular, dynamic memory
allocation has been included along with an eigenvalue solver relying
only on matrix vector multiplication so that sparse matrix methods may
be used. Though the methodology of this code is quite different, its
structure is quite similar to that of MCHF. In fact, the recently
published GRASP92 ,
has been modularized in a similar fashion. We believe this code
should be modified for massively parallel computing, as supported for
the Cray T3D.
With this implementation, our methodology could be extended to heavier systems. In such systems relativistic effects and correlation cannot be separated. Rapid progress could also me make on a collaboration with G. Malli for the study the energy levels structures of the transactinides where relativistic correlation plays an extremely important role.