4. Optical identifications
4.1. Extragalactic sources
Active galactic nuclei account for almost 70% of all extragalactic identifications in the Einstein Extended Medium Sensitivity Survey (EMSS; Stocke et al. 1991) which has a median sensitivity close to that of the ROSAT all-sky survey. With this X-ray sensitivity AGNs have reddening free V magnitudes in the range of 14 to 20 (Gioia et al. 1984). Redshifted broad and narrow emission lines are easily recognizable optical signatures even at low signal to noise ratio. From HI and CO data we find that the clearest part of the 'full' region has only 2-3 mag whereas the most absorbed part has 15 mag. Therefore we do not expect to identify many AGNs with our instrumental spectroscopic limit of V 18 and may consider our only identified AGN at V = 16.3 as outstanding. The situation is even more unfavourable for clusters of galaxies which constitute the second most numerous class of extragalactic sources. Their intrinsic optical faintness and the high galactic stellar foreground make their recognition on a CCD image almost impossible in such a low latitude region. Finally the lack of marked optical signatures renders identification of BL Lac objects probably hopeless.
4.2. White dwarfs
Because of the very pronounced effect of interstellar absorption on their intrinsically very soft X-ray spectra, most of the white dwarfs detected by ROSAT undergo very little interstellar absorption and thus appear basically unreddened at optical wavelength. The hottest DA and DO members which are the most likely to be serendipitously discovered by ROSAT exhibit only weak and shallow H or He absorption lines (Wesemael et al. 1993) which are sometimes difficult to detect for the optically faintest objects. Therefore, we rather considered the very blue Rayleigh-Jeans like continuum as the definite signature of a white dwarf. Of the three white dwarfs detected in our sample area one was already catalogued as such (GD 394). The optically brightest of the two new white dwarfs was found recently in a list of UV excess objects in the galactic plane (LAN 121; Lanning & Meakes 1994).
4.3. Early type stars
Einstein X-ray observations have shown that early type stars may have soft X-ray luminosities as large as a few 1033 erg s-1. Pallavicini et al. (1981) demonstrated that the soft X-ray and bolometric luminosities of O-B stars were tightly correlated with 10-7 . Among the 300 stars earlier than B5 listed in SIMBAD for our Cygnus survey area, we detect two of the three optically brightest (B 6.0) (HR 8154, B = 4.99, O8e and HD200310, B = 5.16, B1Ve) but we miss the overall brightest one (HR 8047, B = 4.69, B1.5nne). These stars have photometric distances of the order of 600 pc and are probably members of the CYG OB7 association. In these two cases, the X-ray luminosities are consistent with normal coronal emission from the OB star. Our third detection is HR 8106 which is a B9III/Ap star rather than a hot early type star and the inferred X-ray luminosity ( 1 1030 erg s-1) is consistent with that of a late type active companion star.
4.4. Active coronae
The fact that many late type stars were bright soft X-ray sources was one of the important discoveries of the Einstein observatory (Vaiana et al. 1981). X-ray coronal activity is now known to be a common feature of many stars of spectral type later than about A5 (e.g. Rosner et al. 1985 and references therein). Observed stellar X-ray luminosities are in the range of 1026 -1031 erg s-1. They increase with rotation rate (Pallavicini et al. 1981) and decrease with stellar age (Micela et al. 1988), pre main sequence stars being about 1000 times more luminous than old main sequence objects. Although a satisfactory theory accounting for all aspects of stellar X-ray activity is not yet available, the currently accepted picture is that X-ray emission originates from a hot stellar corona. By analogy with what is known from the Sun the bulk of X-ray emission is thought to come from magnetically confined loop-like structures emerging at the stellar surface and which are likely to be generated in the subphotospheric convective layers by a dynamo mechanism.
Unfortunately, there is no bright spectral signature of coronal activity at optical wavelength. However, chromospheric and coronal activities are known to be well correlated in the Sun and late type stars in general (see e.g. Reimers 1989 and references therein). In particular, the Ca II H&K chromospheric emission lines are very sensitive measurements of stellar activity which behaves like X-ray emission in being more intense in fast rotating and young stars (Wilson 1963, 1966). Studies based on Einstein X-ray data have shown that the strength of these emissions is indeed well correlated with X-ray luminosity (e.g. Schrijver 1983, Maggio et al. 1987, Fleming et al. 1988).
4.4.1. Optical spectroscopy
Using medium resolution blue optical spectra of 100 stars associated with RGPS sources Guillout (1996) has established and calibrated the relation between Ca II H&K and PSPC flux. Basically, Ca II H&K and soft X-ray emissions are found to be correlated both in flux and luminosity with and . These relations hold over 2 decades in flux and over 3 decades in luminosity. The existence of a flux/flux correlation leaves no doubt that the luminosity relation is real and not an artifact of the distribution in distance. A rms scatter of about a factor 2 in flux / luminosity exists around the mean trend with a maximum range of a factor 10. This scatter is probably caused by the long term (solar cycles) and short term (flares) variability of the coronal/chromospheric activity since our spectroscopic measurements were obtained on the average 2 years after the X-ray survey observations. In spite of this dispersion the above relations may still be extremely valuable tools for identifying active coronae.
The distances in the log () / log () diagram between the position of each active corona and the mean relation have an apparently Gaussian distribution with = 0.30 (Guillout 1996). We show in Fig. 5 the distribution of these distances for our sample in Cygnus. In order to quantify the Ca II identification criterion we computed for each star , the formal probability that if the star is responsible for the X-ray emission its chromospheric emission appears fainter or equal to the observed value.
Several M stars which were optically too faint for medium resolution spectroscopy were observed at low resolution ( 3500- 7500 Å ; FWHM resolution 14 Å). In these cases, we considered that the detection of Balmer emission unambiguously identifies the X-ray source with the late type star. It is known that active M dwarfs tend to have stronger Balmer over Ca II H&K emission ratios than earlier types (Rutten et al. 1989) and that H and X-ray emissions tightly correlate (Fleming et al. 1988). In order to check the validity of our Me star identifications we computed H fluxes using our spectra and the V magnitudes extracted from the literature or from our own measurements. X-ray luminosities were estimated assuming 10-11 S erg cm-2 s-1 where S is the PSPC count rate in cts s-1. For RX J2104.1+4912 (index 7) we used the quiescent X-ray count rate (see Sect. 6). In the / diagram (Fig. 6) our Me stars occupy a region similar to that occupied by the Me stars identified in the EMSS (Fleming et al. 1988) or by Skumanich et al. (1984). Considering the remaining photometric errors of 0.3 mag and keeping in mind the slightly different X-ray energy ranges between ROSAT and Einstein we consider that the agreement is good and leaves no doubt that we have correctly identified these X-ray sources.
4.4.2. Positional coincidence with GSC
Visual examination of our Guide Star Catalogue finding charts readily showed that in many cases the X-ray position was quite close to a rather bright (typically V 12) star usually unreferenced in SIMBAD. In order to obtain an additional identification criterion we computed the a priori probability of positional coincidence by systematically scanning the GSC catalogue within from the ROSAT position. Care was taken to remove duplicate stars in the regions of plate overlaps. We show on Fig. 7 the density of GSC entries held in a ring-shaped area centered on the X-ray position as a function of the radius. The high density excess of stars close to the X-ray position shows that a sizeable fraction of our X-ray sources have indeed counterparts in the GSC catalogue. For instance, 59 GSC entries are found within from the X-ray position whereas the expected number of random matches is only 4. This example shows that in the galactic plane at least, systematic cross-correlation of the ROSAT all-sky survey sources with GSC entries should allow efficient source selection if not identification.
For each X-ray position with a GSC entry of magnitude V within r90 we defined the a priori probability of a spurious match within r90 as = where is the number of GSC entries of magnitude brighter or equal to V found within from the ROSAT source. In a similar fashion we define the a priori probability of identification of the X-ray source with the GSC star as . When no GSC entry was found within r90 we set = 0.0 and used the Ca II H&K emission alone as identification criterion.
We show in Fig. 8 the histograms of for the samples of candidate stars having a Ca II measurement compatible with the X-ray emission (i.e. with a probability of more than 2% to belong to the mean Ca II / X luminosity relation) and for those proposed on the only basis of their proximity to the X-ray position. The two distributions look alike and by integrating the probabilities one expects in total 0.6 among 35 and 1.7 among 53 false cases in the spectroscopic and positional selected samples respectively. This gives us confidence in the possibility to identify X-ray sources with rather bright GSC entries, all presumably active coronae, without obtaining time consuming medium resolution spectroscopy.
Because of the late interactive re-analysis of X-ray survey data in this field, we could spend significantly more observing time per source in the 'inner' area than in the 'full' area. Consequently, the majority of our spectroscopic sample is in the 'inner' area. For the sources located outside the 'inner' area, we concentrated on X-ray locations without bright GSC candidates relying on the probability of positional coincidence to identify the un-investigated X-ray sources.
Note that for convenience we arbitrarily assigned a position probability of 1.0 to our optical identifications with AGN, white dwarfs and Me and late Ke stars. A couple of stars with GSC positions inside r90 (sources index 13 and 80) were eventually found slightly outside when using SIMBAD coordinates.
4.4.3. Combined identification criteria
The positional and Ca II probabilities have different natures since we are dealing on one hand with the probability that a bright unrelated star falls into the ROSAT error circle and on the other hand with the probability that a given star is actually related to the X-ray source once it is found close to the ROSAT position. As a rule of thumb we decided to set the probability boundaries such as the number of spurious or missed identifications among the considered samples was of the order of 1. Since in the 'full' area, we have 44 candidate active coronae with a GSC match and no spectroscopy available and the same number of candidate stars with spectroscopic observations, we decided to identify with active coronae all X-ray sources having either 98% or 2%. Stars without measured Ca II flux or with 2% and 95% 98% were considered as potential optical counterparts and marked with a '?' in columns 'Class' and 'Identification' of Table 11 and 12. We count 13 such 95% confidence level identifications with ML 8 whereas only 4 such random associations are expected for a set of 128 sources. This illustrates further the usefulness of the Guide Star Catalogue for identifying RASS sources at low galactic latitudes.
We list in Table 2 the statistics of the origin of identifications with active coronae for the original 'inner' and 'full' areas. The distribution in count rates of the two identified samples is shown in Fig. 9.
Table 2. Origin of active coronae identifications for sources with maximum likelihood larger than 8
Among the 44 candidate stars located in the 'full' area and having medium resolution spectra, we find 4 cases (all upper limits) where the probability that the observed Ca II emission is compatible with X-ray emission is smaller than 2% whereas the normal distribution would predict only one such case. For two of these sources (RX J2055.3+5025, index 15 and RX J2054.1+4942, index 44) the optical counterpart is bright (V 11) and the a priori chance probability of positional random coincidence derived from the GSC surroundings is correspondingly very small, typically less than 1%. These two bright candidates of spectral types F8 and G0 have strong photospheric Ca II flux and any emission will have lower contrast than in later type stars. Velocity differences between emission and absorption components produced by fast rotation or binarity will further decrease the Ca II contrast. In the two remaining cases (RX J2132.5+4849, index 51 and RX J2054.6+5120, index 72), the probability of random coincidence is much larger ( 5%) and we shall not consider these two identifications. All 8 stars for which we only had Ca II upper limits compatible with the observed ROSAT X-ray flux also had 98% and on this basis were considered as reliable optical identifications.
© European Southern Observatory (ESO) 1997
Online publication: July 8, 1998