Many extremely important inferences can and have been made from RASS data based simply on the X-ray observations, with minimal need for new correlative observations at other wavelengths, and these have been reported in a lengthy series of papers by Prof J. Trümper (the ROSAT P.I. at the Max-Planck-Institut für Extraterrestrische Physik, Garching) and his collaborators over the past few years. For example, the RASS point source catalog immediately yields an exceptionally accurate determination of the log N-log S diagram (number counts vs. flux) of faint X-ray sources, and thus direct and elegant information on the contribution of point sources to the diffuse X-ray background radiation. The surface brightness distribution of the more intense extended X-ray sources, both galactic and extragalactic (e.g., supernova remnants and nearby clusters of galaxies), has been easily and accurately probed with RASS data. Yet another area of immediate and crucial RASS contribution is in the coronae of nearby dwarf stars; previously cataloged, well-studied bright stars with a variety of spectral types within 10 pc of the Sun are detected in great numbers. These are but a few examples of the large variety of projects that have been probed by the ROSAT group, their collaborators, and the international community using RASS data.
Sky distribution of sources in the ROSAT All Sky Survey
The fainter RASS sources, which appear in the RASS data base in very large numbers, pose a different analysis problem, however. We know from ROSAT and also previous generations of X-ray observatories, especially the extensive observations of the Einstein Observatory, that at the the limiting flux level of RASS, the X-ray sources detected are a fascinating but exceptionally heterogeneous mix of astrophysical objects, ranging from the closest M dwarfs to high redshift QSOs. In many cases, the X-ray data taken alone cannot determine whether a RASS source is galactic or extragalactic, much less finer distinctions (and, unlike radio astronomy, there is most definitely a mixture of both at these flux levels). Thus the intrinsic X-ray luminosity of a random RASS source can easily be uncertain by a factor of 1014, lacking further information. For this reason, it has long been recognized that identification, and study of, optical counterparts is an essential companion study to large X-ray surveys.
The most ambitious optical identification effort yet attempted for a large X-ray survey is probably that devoted to the Einstein Extended Medium Sensitivity Survey (EMSS). The comparison with RASS is not inappropriate; although Einstein surveyed <10% of the celestial sphere, the flux levels and positional accuracy of source determinations from the Einstein Imaging Proportional Counter detector were not greatly dissimilar to RASS. Optical identification of sources in the EMSS have been made in an exceptionally ambitious and patient program (Stocke et al. 1991 and references therein) which has consumed almost twenty years of work, and yielded ~800 optical counterparts, including normal stellar corona, pre-main-sequence stars, interacting binaries, supernova remnants, nearby galaxies, clusters of galaxies, and, especially, active galactic nuclei of a broad range of luminosities. These identifications were largely accomplished using the only technology practical at the time, namely whatever heterogeneous collection of telescopes, instruments, and observers were available for the very large effort. Most of the work was also done before the availability of multi-object spectrographs with wide field coverage.
Although 800 identifications may seem a respectable number, there are numerous problems where the size of the EMSS identification sample rapidly dwindles and suddenly becomes dominated by the statistics of small numbers. For example, the EMSS identifications include 35 BL Lac objects. If BL Lacs evolve in their properties with redshift, and/or exhibit complex dependencies of Lx/Lopt with luminosity, as QSOs are known to do, the samples must be subdivided into further bins for analysis, and one is soon left with literally a handful of objects per bin. Similarly, QSOs must be subdivided for analysis into many bins of luminosity, redshift, and radio properties. Finally, certain objects, e.g. pre-main-sequence stars, are simply rare at these X-ray flux levels, and thus their total numbers are very small even in a seemingly large overall X-ray sample.
It seems clear that to effectively build upon this pioneering set of EMSS X-ray identifications, any new effort must exceed the 800 identifications achieved there by a substantial margin, not just a factor of two or three. If many thousands or tens of thousands of sources are to be newly identified, it is also evident that far more efficient techniques than previously available must be used, if these identifications are to be achieved on a reasonable timescale (cf., the discussion in Section 3.2). The larger X-ray sample is now available via the RASS data; the more efficient identification and analysis technique is, of course, the SDSS. RASS and SDSS are beautifully well-matched to each other via a variety of fortunate coincidences. For example, the SDSS survey region is also the region of greatest RASS X-ray sensitivity, because it lies primarily in regions of high ecliptic latitude (and thus long RASS exposure times) and also high galactic latitude (and thus regions of low interstellar photoelectric opacity, an extremely important effect at the soft ROSAT energy band). Perhaps most important, the limiting X-ray flux of RASS is such that, given the known range of Lx/Lopt for all common classes of sources, both galactic and extragalactic, even the optically faintest counterparts of RASS sources are accessible to both the SDSS photometric and spectroscopic surveys. For example, although there is considerable dispersion, the most X-ray luminous (relative to optical) normal stars, normal galaxies, BL Lacs, and AGNs are known to have log(fx/fopt) of about -1, -0.25, 1.5, and 1.0, respectively (e.g., Stocke et al. 1991). At the limiting RASS X-ray flux quoted above, this implies that the faintest optical counterparts in each of these classes will have g' magnitudes of 15, 17, 21, and 20, respectively. Thus we expect that the SDSS imaging survey will obtain highly accurate colors and magnitudes for essentially 100% of RASS counterparts, and that the spectroscopic portion of SDSS, where exposure times are set primarily to yield acceptable signal for the galaxies in the large scale structure survey, will yield excellent quality spectra for the large majority of counterparts, permitting confident identifications and in-depth subsequent analyses.
The final part of the good match between RASS and SDSS is a consequence of the modest surface density of RASS sources on the sky, about 1.5 deg-2 at 4.5 sigma detection above background. A typical SDSS spectroscopic plate thus has ~10 RASS sources, and therefore 100% of the ~15,000 RASS sources in the SDSS volume can be targeted for at least one spectroscopic observation, while only diverting 1% of the fibers from other SDSS scientific programs.
Because of the compelling nature of the above arguments, the ROSAT team, headed by Professor J. Trümper, has formed a formal collaboration with SDSS. The goal is to use the unique capabilities of the combined RASS/SDSS data to obtain a sample of ~104 X-ray sources which are not only optically identified, but all have very high quality photometry and spectroscopy from identical instrumentation. Thus not only the size, but the homogeneity of the resulting optical sample will be extraordinary and unprecedented. Further, since software tools for photometric and spectroscopic analysis are already required for the other SDSS scientific programs, this homogeneous data bank can be exploited with little further software effort. Finally, this sample will reach its final size, with automated, totally reduced photometry and spectroscopy, in 5 years, the duration of the SDSS observational effort.
The ROSAT project has had significant American scientific and financial participation from its inception, and thus large fractions of ROSAT data lie in the American and international public domain. However, in accordance with the formal agreement negotiated by NASA prior to ROSAT launch, RASS data are not included in this public data release. Since the scientific goal of the SDSS/RASS collaboration is the publication of this unique data base of X-ray/optical identifications and their properties at both wavelengths, it is clear that the raw data that backs up this work must also be made freely available. Therefore, as part of the collaboration discussed here, the MPE group has agreed to a general public release of extensive sections of the RASS data. The details of this agreement are discussed in a letter from Professor Trümper; it suffices to note here that the level of completeness, content, and detail of this release will very significantly exceed that of the ROSAT All-Sky Survey Bright Source Catalogue (1RXS), which was announced at the 1995 Würzburg meeting, currently appears on the World Wide Web, and is in press to Astronomy and Astrophysics. Thus these two data releases will be nicely complementary; one covers the entire sky, but provides only brighter sources, already processed from the raw data; the other covers the SDSS volume (25% of the sky), but goes significantly fainter, and provides raw photon data if desired, enabling detailed, ab initio source existence, spectral, and timing analyses, using any algorithm of the user's choice. In addition, the "SDSS data release" will also provide the full ROSAT All Sky Survey Catalogue (as opposed to the current "Bright" catalogue), containing 8x104sources with >5 sigma significance.
The details of the expected scientific impact of various investigations which will emerge from this RASS/SDSS effort will be discussed later in this proposal; the programs are so varied that they are more appropriately addressed in separate, discipline-oriented sections. The actual SDSS spectroscopic targeting situation is also more complex than described here, of course. Many or most RASS sources corresponding to stellar corona are too bright for SDSS spectra, nor are such spectra necessary for either identification (the chance coincidence probability of a bright star in the small X-ray positional error box is nil) or study (these are typically cataloged and well-studied objects). The number of fibers needed only for RASS sources is further decreased due to the large fraction of identifications which will prove to be QSOs, already independently targeted for spectroscopy via 5-color selection from SDSS photometric data by the SDSS QSO Working Group, who are using ~102 fibers per plate for their programs. However complications are introduced when multiple optical candidates for the RASS identification are present within the 55" minimum SDSS fiber spacing, which is often the case. These issues can be easily handled by our automated target selection software, and will be discussed below in Section 2.2.
Stocke, J. T., Morris, S. L., Gioia, I. M., Maccacaro, T., Schild, R., Wolter, A., Fleming, T.A., & Henry, J.P. 1991, ApJSuppl 76, 813.
Voges, W. 1994a, in "Basic Space Science", ed. H. J. Haubold and L. I. Onuora (New York: AIP), p. 212.
Voges, W. 1994b, in "New Horizon of X-ray Astronomy", ed. F. Makino and T. Ohashi (Tokyo: Universal Academy Press), p. 197.