1 Introduction

The ROSAT All-Sky Survey (RASS) provides unique opportunities for large-area multi-frequency studies of extragalactic objects. The catalog of $\sim$80000 sources which resulted from the recent RASS-II processing comprises the deepest soft X-ray (0.1-2.4 keV) sample yet obtained (Voges et al.1996), reaching a sensitivity of a few $\times 10^{-13}$ erg cm^-2 s^-1. Previous studies based on X-ray surveys such as the HEAO-1 Large Area Sky Survey (Wood et al. 1984) and the Einstein Extended Medium Sensitivity Survey (Gioia et al.1990; Stocke et al.1991) showed that the majority of X-ray-selected objects are extragalactic. In particular, active galactic nuclei (AGN) comprised 55 per cent of the EMSS, including 4 per cent BL Lacertae objects; normal or cooling flow galaxies another 3 per cent, and clusters of galaxies 13 per cent (Maccacaro et al.1994).

While the task of optical identification of the entire RASS catalog will take years, special attention has been paid to the regions of the ecliptic poles which have unusually long exposures. As the ROSAT satellite carried out scans on great circles along ecliptic latitudes, the two ecliptic poles were observed for half a year once per orbit for up to 32 secs each time. Therefore the total exposure of the sky region around the North Ecliptic Pole (NEP), located at $\alpha = \rm 18^h00^m$ and $\delta = 66^\circ 30$$^\prime$, reached values of $\lower.5ex\hbox{$\; \buildrel < \over \sim \;$}40000$ s (falling off with ecliptic latitude as $\sim$ arcsin (1 / cos ($\lambda$)), compared to the typical exposure for the RASS of $\lower.5ex\hbox{$\; \buildrel < \over \sim \;$}\!$400 s near the ecliptic equator. The deep ROSAT pointings with the Position Sensitive Proportional Counter (PSPC) on the NEP (Bower et al.1996) reached a total exposure of $\sim$ 79 ksec, but only data from the innermost 15.5 arcmin radius were analyzed (to minimize spurious detections due to the rib structure of the PSPC window). Thus the Survey observations of the NEP region still represent one of the deepest and spatially most extensive X-ray exposures of the sky ever performed.

Deep ROSAT pointings indicate the X-ray population expected around the NEP. For example, in eight PSPC pointings of the second Lockman hole (limiting flux of $1.7 \ 10^{-14}$ erg cm^-2 s^-1), Carballo et al.(1995) identified 51 of 89 sources of which $\sim$50 per cent are AGN and another $\sim$30 per cent are galaxies or clusters of galaxies. In the similarly deep ($\lower.5ex\hbox{$\; \buildrel \gt \over \sim \;$}1 \ 10^{-14}$ erg cm^-2 s^-1) but small-area pointings centered on the North Ecliptic Pole, Bower et al. (1996) detected 20 sources and found a higher fraction of AGN, including $\sim$60 per cent QSOs, a Seyfert 2 and a BL Lac object, as well as a cluster and two unidentified sources. Georgantopoulos et al. (1996) reported on a series of five deep ($\sim$$3 \ 10^{-15}$erg cm^-2 s^-1) PSPC pointings of fields chosen from the ultraviolet excess (UVX) survey of Boyle et al.(1990). Of the identified sources (145 of 194), 74 per cent are AGN (including QSOs and Seyfert 1 galaxies), another 13 per cent are galaxies, plus one cluster. As some AGN lie in clusters of galaxies, some confusion between the classes is probably present (Pierre et al.1994).

Another approach to analyzing the extragalactic content of the X-ray source population is the cross-correlation of deep X-ray surveys with radio surveys. A primary advantage is the substantial improvement in precision of source positions, which in turn permits cross-correlations with faint optical surveys. Broad-band properties can then be determined prior to a comprehensive optical spectroscopic program, and the effort required for the latter is substantially reduced. Simultaneous X-ray- and radio-selection produces exclusively extragalactic samples which can give insight into nonthermal processes and unified theories of AGN. Deep radio surveys (Windhorst et al. 1985; Thuan et al.1992; Benn et al.1993) and Hamilton & Helfand's (1993) cross-correlation of deep Einstein IPC pointings and VLA observations suggest that a population of faint radio-emitting starburst galaxies could provide the necessary flux and spectral hardness to make up the faint end of the XRB. However, this suggestion is not confirmed by the ROSAT-ATCA study of the QSF3 field by Boyle et al.(1993).

In previous papers (Brinkmann et al.1994; Brinkmann et al. 1995 (Paper I) and Brinkmann et al.1997 (Paper II)), we reported cross-correlations of the RASS with large scale radio surveys. Brinkmann et al.(1994) compared the RASS source list with the Molonglo Reference Catalog of 408 MHz radio sources (Large et al. 1991). Of the 546 coincidences found, 433 objects were previously spectroscopically identified. The radio - to - X-ray properties of this sample are discussed with an emphasis on implications for AGN unification schemes. Papers I and II compared the RASS sample with a source list generated from the 5 GHz Green Bank survey of the northern sky (Condon et al.1989). This radio survey covers most of the northern sky, from $0\deg \le \delta_{1950} \le +75\deg$ with a limiting sensitivity of $f\rm _r$ $\; \buildrel \gt \over \sim \;$20 mJy. The fraction of previously identified sources is much lower in this cross-correlation; of the 2127 coincidences, only 617 were previously known and their properties are discussed in Paper I. All objects were studied in follow-up observations with the VLA to obtain core fluxes and arc sec positions (Laurent-Muehleisen et al.1997) and the radio - to - X-ray properties of the large sample of unidentified extragalactic sources are discussed in Paper II.

The X-ray spectra, optical colors and morphologies, and spectral energy distributions of the ROSAT - Green Bank (RGB) sources indicate that most of the unidentified sources are quasars, although some BL Lacs, radio galaxies, Seyferts and clusters are undoubtably present. However, their properties differ subtly from those of brighter radio-loud AGN such as the 1 Jy catalog (Kühr et al.1979). These differences reflect the fact that the majority of the new RGB sources have broad-band properties between those of traditional radio-selected and X-ray-selected AGN. This is clearly shown among the RGB BL Lac objects (Laurent-Muehleisen et al.1998). Deeper X-ray and radio surveys also offer the possibility of revealing an AGN population with radio properties intermediate between the traditional radio-loud AGN (PKS and 1 Jy samples) and radio-quiet AGN.

The plan of the paper is as follows. Section 2 discusses the multifrequency observations. Sections 2.1 summarizes the radio observations, more fully presented in Kollgaard et al.(1994) and Hertz et al. (1994). Section 2.2 describes the cross-correlations with COSMOS- and APM-generated catalogs to provide optical characteristics. In Sect. 2.3 the ROSAT observations and details of the data analysis (including source detection and X-ray flux computation) are presented. Section 3 discusses the selection of sources for the multifrequency catalog and presents the catalog as three tables divided according to reliability of the X-ray-radio match and source morphology. Section 4 provides additional information on two subsets of sources. Section 4.1 presents X-ray spectral fits for the bright X-ray sources in the catalog. Section 4.2 describes the selection of candidate groups and clusters of galaxies, including individual source notes and radio-optical maps. Section 5 presents the broad-band properties of the sample, including the X-ray and radio flux distributions (Sect. 5.1), radio-to-optical and optical-to-X-ray spectral indices (Sect. 5.2), distribution of radio-to-optical spectral index (Sect. 5.3), and X-ray-to-radio and optical-to-radio flux ratios (Sect. 5.3). Comparisons to the results of Papers I and II are made throughout Sect. 5. The conclusions are summarized in Sect. 6.