2 Cross-identification of the ROSAT all-sky survey bright source catalogue with SIMBAD

2.1 BRASS-SIMBAD correlation

Studying the clustering properties of identified X-ray sources in a multi-wavelength parameter space can help finding domains in which the probability of association with a given class of object is higher than on the average. For instance, at low galactic latitude the overwhelming domination of stellar coronae makes the discovery of new CVs a difficult task as they only account for $\approx$1% of the total number of sources (Motch et al. 1996) whereas their frequency is much higher among either hard, or soft and bright selected sources.

If completely identified flux limited samples are obtained using catalogue cross-correlation or work at the telescope, then the probability that X-ray emission from a given source originates from a particular class of emitters (star, AGN, etc.) may be eventually estimated in a quantitative manner. This allows the computation of a statistical identification using as input the restricted set of multi-wavelength parameters available for all sources.

RGPS sources identified in SIMBAD are far from making a complete sample. SIMBAD is roughly optical flux limited, although at very different levels depending on the nature, stellar or extragalactic, of the object. As most of the recent additions are coming from the literature, the whole collection is highly heterogeneous. Clearly, the set of SIMBAD identifications cannot be used to provide real probabilities of identification. However, it may be used to construct selected samples of presumably enhanced probabilities.

A first analysis was made in 1991, using X-ray and optical characteristics of a limited number of stars, AGN, CVs and accreting binaries identified in preliminary and partial releases of the RASS (Motch 1991a). The selection criteria used to build the source samples studied in this paper are derived from this early analysis.

The recently published ROSAT all-sky survey bright source catalogue (BRASS) by Voges et al. (1996) constitutes a sample of larger size and of much better quality than the one originally used in 1991. Therefore, for the sole aim of illustrating in an extensive manner the X-ray selection criteria used in this work, we decided to cross-correlate BRASS with SIMBAD. We underline that the associations between bright RASS sources and SIMBAD entries used in this paper have not been human screened and cannot be considered individually as real identifications. Nevertheless their overall statistical properties may still be used for our purpose.

The BRASS covers the whole sky whereas the RGPS is by definition limited to the galactic plane area. However, selection rules arising from the BRASS catalogue can usefully be applied to the restricted RGPS area. The variability of parameters with galactic latitude or $N_{\rm H}$ can also be better studied using the whole sky catalogue.

The first step of the cross-correlation consisted in extracting from SIMBAD all objects located within 3$^\prime$ from the ROSAT source, retrieving the optical position and associated error, the first identifier, possible ROSAT name, optical magnitudes, spectral types, full object type, number of references and Einstein measurements when available. In a second step, we retained from the SIMBAD cross correlation log entries having a non null intersection between the ROSAT 90% confidence circle and the SIMBAD error circle. As the error radius of Einstein sources is underestimated in SIMBAD, we updated it manually to 50$^{\prime\prime}$. When several SIMBAD objects remained, we retained the one closest to the X-ray position. With a mean 90% error radius of 27$^{\prime\prime}$ for BRASS sources, and a total number of $\sim$1.46 10⁶ SIMBAD objects having in most cases small positional errors ($\leq$ 1$^{\prime\prime}$), we expect about 120 spurious matches between a SIMBAD and a ROSAT source. This estimate ignores effects resulting from inhomogeneities in the surface density of SIMBAD and BRASS entries. We list in Table 1 some statistics of the cross-correlation.

Clearly, the large majority of SIMBAD matches are real since only $\sim$1.5% of the sources with one or more SIMBAD entries are expected to be spurious. Considering the restrictive aim of this work, we did not investigate in detail the reasons explaining the relatively large number of sources with multiple SIMBAD matches. Probably, the high surface density of SIMBAD entries in some specific small regions of the sky, multiple stellar systems and overall the fact that a number of astrophysical objects still appear as distinct SIMBAD objects could account for this effect.

  

Table 1: BRASS-SIMBAD correlation

Most of the proposed SIMBAD identifications of BRASS sources (61%) are with stars and only 20% with galaxies and AGN. The high identified stellar fraction obviously reflects the emphasis put on stars in SIMBAD.

The quality of the cross-correlation may be judged from Fig. 1 which shows the relation between Einstein IPC and ROSAT PSPC count rates for the 1244 Einstein sources recovered in the BRASS. IPC rates were extracted from the corresponding measurement headers in SIMBAD. The fractions of ROSAT/Einstein sources with one, two or more than 2 matches in SIMBAD are 77%, 16% and 7% respectively. The ratio of the count rates of the two instruments strongly depends on source spectrum with an additional scatter due to variability. We show for comparison two relations, PSPC = 1 $\times$IPC and PSPC = 6 $\times$ IPC corresponding to hard (power law photon index = 0) and soft (thermal spectra with logT = 5.8) sources respectively. As expected, most sources lie between these two extreme cases. The outstanding source at IPC = 0.02cts s^-1 and PSPC = 20cts s^-1 is the low-mass X-ray binary transient EXO 0748-676 which was detected by Einstein in the low state before its discovery by EXOSAT in outburst.

  

Figure 1: The relation between Einstein IPC and ROSAT PSPC count rates for 1244 Einstein sources identified with BRASS entries. The two lines represent relations expected for soft and hard sources

2.2 X-ray and optical properties of identified BRASS sources

For source classification purposes, the most interesting parameters are flux ratios in various energy bands, including the conventional X-ray hardness ratios, but also $F_{\rm X}$/$F_{\rm opt}$ ratios as well as optical colours. Another important information for galactic studies is the line of sight absorption which can be estimated from H I and CO data.

Arguments involving the $F_{\rm X}$/$F_{\rm opt}$ ratio have been used by many authors in the context of the Einstein Observatory surveys at high and low galactic latitudes (e.g., Maccacaro et al. 1982; Hertz & Grindlay 1984). ROSAT offers a slightly improved spectral response compared to Einstein and allows PSPC hardness ratios to be used as additional information, at least for relatively bright sources.

2.2.1 Hardness ratios

We plot in Fig. 2 the positions in the hardness ratio diagram of BRASS sources cross-identified in SIMBAD with different classes of objects.

The exact energy range used for the computation of hardness ratios changes with the version of the Scientific Analysis System Software (SASS; Voges et al. 1992). Whereas the 93 sources discussed later in this paper were the output of SASS-I (see Sect. 3.1), the BRASS catalogue (Voges et al. 1996) uses SASS-II processing in which the hardness ratios are defined as:

$${HR1}\ = \frac{(0.5-2.0)-(0.1-0.4)}{(0.1-0.4)+(0.5-2.0)} \ {\rm (SASS-II)}$$

$${HR2}\ = \frac{(1.0-2.0)-(0.5-1.0)}{(1.0-2.0)} \ {\rm (SASS-II)}$$

where (A-B) is the raw background corrected source count rate in the A-B energy range expressed in keV.

Active coronae populate the central part of the HR1/HR2 diagram with some tail extending towards hard spectra. Stellar coronae are known to exhibit a range of temperatures between 3 10⁶K and 10⁷K, with the most active and luminous stars also exhibiting the highest kT and hardest spectra. In this diagram, isolated white dwarfs are all found at $HR1 \leq -0.9$. At the BRASS flux level, the vast majority of active coronae are located within 100 pc from the Sun (e.g. Guillout et al. 1996b) and except for the early type stars and some of the most luminous late type binaries, the effects of interstellar absorption are negligible in both X-ray colours.

In general, the AGN found in SIMBAD do not exhibit HR2 smaller than $\approx$ -0.2 but have a large scatter in HR1 which is closely related to galactic foreground absorption (see below). However, ROSAT has discovered some AGN with very steep spectra which are so far barely represented in SIMBAD (e.g. Greiner et al. 1996).

The bulk of the cataclysmic variables populate the upper right quadrant, avoiding very hard and very soft HR2 values. Among outstanding sources we find at HR1 = 0.95 $\pm$ 0.05 and HR2 = -0.63 $\pm$ 0.08 the peculiar soft IP candidate RX J1914.4+2456 which exhibits strong interstellar absorption (Haberl & Motch 1995; Motch et al. 1996). The supersoft X-ray emission from GQ Mus = Nova Muscae 1983 at HR1 = -0.05 $\pm$ 0.27 and HR2 = -0.87 $\pm$ 0.47 was discovered by Ögelman et al. (1993). Cataclysmic variables with HR1 $\leq$ 0.1 are all polar or soft intermediate polar systems.

Most X-ray binaries exhibit very hard X-ray hardness ratios resulting both from the usually high interstellar absorption towards these remote sources and from an intrinsically hard spectrum (kT $\geq$ 2 keV with sometimes local photoelectric absorption such as in Be/X-ray binaries for instance). The two objects at the lower left corner are supersoft sources in the Large Magellanic Cloud. The galactic supersoft source RX J0925.7-4758 (HR1 = 1.00 $\pm$ 0.01 and HR2 = -0.36 $\pm$ 0.08) exhibits an intrinsically soft spectrum affected by strong interstellar absorption. The source at HR1 = 0.64 $\pm$ 0.02 and HR2 = 0.09 $\pm$ 0.02 is the LMC transient 1A 0538-66 and that at HR1 = 0.20 $\pm$ 0.03 and HR2 = 0.31 $\pm$ 0.04 is the peculiar M giant binary HD 154791/A 1704+241.

  

Figure 2: Position of various classes of X-ray emitters in the HR1/HR2 diagram. Asterisks in the Stars diagram represent white dwarfs. The scatter in HR2 of white dwarfs is due to the large error on this ratio for such soft sources. For non-degenerate stars and AGN, we only show sources with errors less than 0.2 on both ratios. The percentages of stars and AGN remaining amount to 40% and 51% of the total sample respectively

2.2.2 Hardness ratios and $F_{\rm X}$/$F_{\rm opt}$

The most informative diagrams are without doubt those involving optical information, basically in the form of the $F_{\rm X}$/$F_{\rm opt}$ ratio. The X-ray to optical flux ratio can be defined as log($F_{\rm X}$/$F_{\rm opt}$) = log(PSPC count rate) + V/2.5 - 5.63, following the expression used by Maccacaro et al. (1982) for the Einstein medium sensitivity survey and assuming an average energy conversion factor of 1 PSPCcts s^-1 for a 10^-11 erg cm^-2 s^-1 flux in the range of 0.1 to 2.4 keV. We show in Fig. 3 the position of stars, AGN, X-ray binaries and cataclysmic variables in the X-ray colour versus $F_{\rm X}$/$F_{\rm opt}$ ratio diagram.

Although stars and AGN have similar X-ray colours, their mean X-ray to optical ratios are obviously quite different and the two populations are well separated in the HR1/2 $F_{\rm X}$/$F_{\rm opt}$ diagram. Stars with HR1 close to -1 are not recorded as white dwarfs in SIMBAD. Their soft X-ray emission could either be due to a particularly low temperature corona or to an unrecognized degenerate companion or nature. They do not appear in Fig. 2 because of their large errors on HR2. X-ray binaries are essentially recognizable from their hard X-ray spectra and usually large $F_{\rm X}$/$F_{\rm opt}$. The low $F_{\rm X}$/$F_{\rm opt}$ X-ray binary tail consists of high mass X-ray binaries. Cataclysmic variables exhibit a large range of X-ray colours and $F_{\rm X}$/$F_{\rm opt}$ ratios and can be somewhat confused with both the AGN and the most active part of the stellar population. However, the addition of a B-V or U-B optical index would allow to distinguish further between these overlapping populations.

  

Figure 3: Position of various classes of identified BRASS sources in the HR1 (left) or HR2 (right) / Log($F_{\rm X}$/$F_{\rm opt}$) diagram. Small dots represent stars, large dots AGN, squares are X-ray binaries and crosses are cataclysmic variables. For stars and AGN only sources having errors on HR1 or HR2 smaller than 0.2 are plotted

2.2.3 Interstellar absorption

An efficient way to discriminate between local and remote populations of X-ray sources is to use the anisotropy produced by the Galaxy, essentially in terms of scale height and interstellar absorption. This effect is illustrated in Figs. 4 and 5 which display the position of AGN, CVs and X-ray binaries in the integrated H I/hardness ratio diagram. For this study we used $N_{\rm H}$ data from Dickey & Lockman (1990).

Most active stars detected in the RASS are little affected by interstellar absorption and their X-ray colours do not vary with galactic latitude nor with integrated column density.

Apparently, the shape of the low energy spectrum of AGN is rather constant with the consequence that hardness ratio 1 is well correlated with the integrated H I column density as shown in Fig. 4. Actually, this relation is well defined and we used it for preselecting AGN candidates with a high rate of success. However, the HR2 distribution is relatively peaked around a value of 0.1 and does not vary with $N_{\rm H}$. This is probably due to a selection effect against highly absorbed AGN which are likely to be missing in the mostly optically selected SIMBAD sample.

With X-ray luminosities up to 10^32-33 erg s^-1, cataclysmic binaries can be detected to distances as large as 1 kpc or more at the BRASS sensitivity. There is indeed a slight tendency that cataclysmic variables exhibiting the hardest HR1 are preferentially found at low galactic latitudes but the large variety of spectra emitted in this class (from very soft polars to intrinsically absorbed intermediate polars) somewhat blurs the picture.

In contrast, luminous X-ray binaries are seen by ROSAT at very large distances in deeply absorbed regions of the galactic plane and apart from a few cases (e.g. supersoft sources), their hardness ratio HR1 is close to +1, indicating that all X-ray photons below 0.4 keV are blocked by interstellar absorption. Effects of interstellar absorption are also seen on hardness ratio HR2 which involves higher energy bands. Not surprisingly, Fig. 5 shows that sources lying in the most absorbed direction of the Galaxy (and therefore lower galactic latitudes) are also the hardest in HR2.

  

Figure 4: Variation of HR1 with integrated galactic H I column density for various classes of X-ray emitters. For AGN identifications, only sources with error on HR1 smaller than 0.2 are plotted

  

Figure 5: Variation of HR2 with integrated galactic H I column density for various classes of X-ray emitters. For AGN identifications, only sources with error on HR2 smaller than 0.2 are plotted

2.3 Conclusions

Stars and AGN numerically dominate the ROSAT X-ray sky. These two populations can be easily distinguished using optical flux information. Searching for rarer X-ray emitters can be made much more efficient by selecting particular domains of the X-ray and optical source parameter space. Some classes of X-ray sources preferentially populate particular regions of the HR1/HR2 diagram. For instance, isolated hot white dwarfs, high magnetic field polars and unabsorbed supersoft sources all show very soft spectra. These three classes may be distinguished further from their X-ray to optical flux ratio. X-ray binaries and hard intermediate polar CVs may also be detected from their particularly hard X-ray emission in HR1 and HR2. However, since the hard region also contains some very active coronae and absorbed AGNs, the selection criterion is less efficient than for other classes of sources. The soft HR2 / hard HR1 region has special interest since it is essentially void of stars, AGN, CVs and classical XRBs. This is the place where one expects to find luminous galactic supersoft sources. These general guidelines were used to select the bulk of the sources discussed in the next section.