Until now three types of benchmarks have served as indicators of the quality of the atomic models. ASCA data themselves have helped in identifying problems, but are not useful for solving them. High signal-to-noise AXAF spectra will provide extremely useful benchmarks for atomic data of general interest in high energy astrophysics.
Laboratory atomic physics experiments. It is impractical to test every atomic rate that goes into the prediction of a spectrum, but with a judicious choice of experiments, codes can be benchmarked and specific known problems may be tested. For example, the ASCA Performance Verification phase observation of the Centaurus cluster of galaxies (Fabian et al. 1994) could not be fit with existing spectral models, suggesting that the problem was with the atomic physics. Liedahl, Osterheld, & Goldstein (1995) improved the calculations for Fe XXIII and Fe XXIV, but still found discrepancies with respect to solar observations. The Fe XXIV n = 4 -> 2 to n = 3 -> 2 total rate ratio has now been measured (Savin et al. 1996) at the EBIT Electron Beam Ion Trap (EBIT) at Lawrence Livermore National Laboratory, confirming the new theoretical calculations.
As a second example, the observed line intensity of Fe XVII lambda15.01, one of the strongest lines in the solar X-ray spectrum, is discrepant with all theoretical calculations. After more than two decades of efforts to resolve the discrepancy for this line with theoretical models, line intensities from Fe XVII are only now being measured at the Livermore EBIT (Beiersdorfer et al. 1997).
Other useful laboratory measurements include wavelengths for a given ion. The inaccuracies of some of the wavelengths we are using in the plasma codes are larger than the grating resolution and will render highly suspect blind spectral fitting of the grating data using plasma codes.
Laboratory plasma physics measurements. Plasmas generated for fusion energy studies can be useful for benchmarking atomic collisional data. In particular, tokamak plasmas seeded with interesting elements have produced spectra that are not unlike astrophysical spectra. Although not all the atomic processes in tokamak plasmas are well understood (e.g. non-equilibrium ionization effects related to diffusion), these experiments have one advantage over astrophysical plasmas as benchmarks -- namely, that densities and temperatures are independently measured. Line intensity ratios and wavelength scans are among the types of measurements (Stratton et al 1985; Wargelin et al. 1997). Unfortunately, few critical assessments have been published, and fusion energy budgets continue to shrink.
Solar measurements. Solar observations are extremely useful; however, their limitations are not well known outside the solar community. Perhaps the most surprising aspect to non-solar X-ray astrophysicists is that the solar community has never conducted broad band, well calibrated measurements. Furthermore, as instrumentation has improved, the goal of high spatial and spectral resolution has narrowed the observed wavebands even more.
The Bragg crystal spectrometers on Yohkoh can provide highly specific diagnostic information, but they are limited to a few narrow spectral ranges (near Fe XXVI, Fe XXV, Ca XIX, and S XV) and are predominantly used to study highly transient plasmas (see Khan et al. 1995). They do not serve as particularly useful benchmarks.
The SOHO mission does not contain any high resolution X-ray spectrometers (the highest energy of 83 eV barely overlaps with the lowest energy of the LETG). The preference of the solar community for EUV rather than X-ray spectroscopy is not surprising given that many ions of interest in the Sun emit their strongest lines in the EUV.
Published solar X-ray spectral catalogs are also less useful than one might expect. The spectral surveys are photographic, so the flux calibration is about 25% at best (McKenzie et al. 1980). The surveys with the best spectral resolution often had even worse flux calibration (Phillips et al. 1982). The flux calibration in the soft X-rays (corresponding to the LETG band) is particularly poor and the surveys are littered with missing or questionable line identifications (Acton et al. 1985).
The diagnostically rich Fe L shell region around 1 keV is basically emitted only in solar flares. Therefore, the survey line lists conducted in this part of the spectrum have additional uncertainties involving time-dependent effects, since the spectrometers are scanned over time. Furthermore, the spectral resolution tends to be rather modest ( E/Delta E ~ 250). Examination of the catalogs derived from these observations shows a large fraction (~ 40%) of blended lines (see McKenzie et al. 1980). Thus, while the strong line intensities are useful, the uncertainties are still rather large. For weaker lines, blending becomes problematic.
Higher spectral resolution ( E/Delta E ~ 500) was obtained with the flat crystal spectrometers on the Solar Maximum Mission (SMM). Again these spectra were scanned (.01Å s-1) and narrow band (13 to 19 Å). Furthermore, the background is dominated by a non-solar contribution, such that there is no possibility of measuring the continuum flux, or of assessing the contribution of weak lines to the apparent continuum (see Schmelz et al. 1996).