Astron. Astrophys. 323, 853-875 (1997)

6. Discussion

We summarize in Table 8 the final list of identifications for all 24 sources resulting from our follow-up optical/X-ray studies. Spectral types were updated using our own determinations. In total we have discovered 5 very likely new massive X-ray binaries with an additional 2 good candidates needing further confirmation. Table 9 lists the main optical and X-ray characteristics of the 7 new massive X-ray binary candidates. Distances, X-ray luminosity and luminosity ratios are subject to various sources of error such as count rate statistics and unknown X-ray energy distribution, inaccuracies in the interstellar absorption and absolute photometric calibrations. A typical error of 25% to 50% on all three quantities is probably realistic. The same errors apply to data listed in Table 10.

Table 8. Summary of optical identifications. A.C. stands for active coronae. Spectral types are from our own determination or from the literature. Horizontal lines divide the three groups of candidates defined in Table 4

Table 9. Optical and X-ray characteristics of the new OB/X-ray candidate systems. In this table we use the revised spectral types derived from our optical observations and the maximum X-ray count rates measured from survey or pointed PSPC observations and extracted using EXSAS. The last column lists the ratio of the maximum to minimum count rate observed between survey and pointed observations when significant variability is detected. The first five objects can be considered as very good OB/X-ray candidates whereas the last two require final confirmation of their X-ray excess

Table 10. Optical and X-ray characteristics of the known OB/X-ray systems detected by ROSAT during the all-sky survey at . About 5% of the galactic sky is not covered by this study (see Section 2.1). Distances, / and are computed using the automatic process described in Sect. 2.4 and the results of the SASS analysis. The first group consists of Be/X-ray systems and the second group gathers disk and wind fed high mass X-ray binaries

6.1. Interlopers

The cross-correlation in position of large catalogues with the unprecedented number of sources present in the ROSAT all-sky survey unavoidably leads to numerous spurious matches. Further selection of associations with extreme X-ray versus optical characteristics such as the one performed here enhances the fraction of wrong identifications and our present study clearly illustrates the need to ensure identification using X-ray and especially optical follow-up observations.

Several points of interest concerning the statistics of positional coincidence may be noted. First, all but 4 sources have a confirmed or a likely alternative optical counterpart within the 90% confidence radius. Among the 8 OB/X-ray associations outside the SASS 95% confidence radius (group 2) only one, SS 73 49 is confirmed. These evidences demonstrate the reliability of the survey ROSAT XRT positions and show that the centering statistics is well enough understood at least phenomenologically to be used as a constraining tool for systematic optical identification. Second, 3 sources listed in group 1 have a confirmed alternative optical counterpart physically unrelated to the OB star. A fourth case may be that of LS IV -12 70. This rate would be close to the number of spurious matches within the 95% confidence error radius that was estimated in Section 2.5 on the basis of the size of the cross-correlated samples. These three to four cases do appear in the first group consistent with the expectation in Section 2.5 that spurious associations will be preferentially found among the high / objects. However, since similar comprehensive optical follow-up has not been carried out for the remainder of the 108 OB/X-ray associations no definite conclusions about the validity of the estimates for spurious matches can be drawn.

6.2. Comparison with previously known OB/X-ray binaries

Among the 13 massive X-ray binaries known before the launch of ROSAT and detected by the SASS analysis in the galactic plane survey, there are 8 Be/X-ray systems and 5 disk-fed or supergiant wind-fed binaries (see Table 10). Most of the detected pre-ROSAT Be/X-ray systems are transient sources and apart from EXO 2030+375 which was probably caught in weak outburst (Mavromatakis 1994) all other systems have X-ray luminosities typical of the quiescent state ( = 5 10³³ - 5 10³⁴ erg s^-1 in the energy range 0.1-2.4 keV and corrected for photoelectric absorption).

The distribution in optical spectral types of the 7 new massive X-ray binary candidates compares well with that of the previously known systems. In particular, we do not find accreting candidates with spectral types later than B2 and having X-ray luminosities above 10³² erg s^-1 among the candidates earlier than B6, although the possibility to detect an excess of X-ray luminosity is in principle larger for the later stars. The absence of good accreting candidates later than B2 is also not an artifact of our selection criteria since our threshold of / = 3 10^-6 implies a limiting of 10³¹ erg s^-1 at B5V and 7 10³¹ erg s^-1 at B2V. In fact we do detect an excess of X-ray emission in the range of 2-7 10³¹ erg s^-1 from four B3V-B5V stars (HD 38087, HD 38023, HD 36262 and HD 53339). The youth of the OB associations in which these four extreme stars are located favours an explanation in terms of X-ray activity from a pre-main sequence low mass companion.

The distribution in X-ray luminosities of the four best Be/X-ray candidates is also comparable with that of known Be/X-ray systems in quiescence. On the other hand, the two additional candidates, HD 161103 and SAO 49725 have significantly lower X-ray luminosities.

The un-absorbed X-ray luminosities of the five first rank candidates clearly indicate that the accreting object is a neutron star or less probably a black hole. All five sources have HR2 0.5 similar to that exhibited by the majority of the known X-ray binaries, probably reflecting the hard intrinsic energy distribution. Unfortunately, in all cases, at most 100 photons were collected from the new sources preventing more detailed spectral modelling. Binary stellar evolution models predict the existence of five times more Be + white dwarf than Be + neutron star binaries (Pols et al. 1991). Waters et al. (1989) compute that white dwarfs accreting from a Be envelope should display X-ray luminosities in the range from 10²⁹ to 10³³ erg s^-1. So far there is no established Be + white dwarf systems. Using ROSAT PSPC observations Haberl (1995) argues that Cassiopeiae could well be an example of a white dwarf accreting from the dense circumstellar envelope of a Be star although the hard X-ray properties of the source are consistent with those of an accreting neutron star in a widely separated orbit (White et al. 1982; Waters 1989). In our sample, only HD 161103 and SAO 49725 have low enough X-ray luminosities to qualify as Be + white dwarf candidates if their X-ray luminosity excess is confirmed. It may be noted that these two candidates have large Balmer emission revealing the presence of a high density envelope. A white dwarf could easily accrete enough matter from such a dense circumstellar material in order to produce the 10³² erg s^-1 detected by ROSAT. The X-ray spectrum of an accreting white dwarf in a Be envelope may not be necessarily as hard as that of a neutron star (see discussion in Meurs et al. 1992).

Two quite different mechanisms may account for the huge variations in X-ray luminosities often observed in these binaries. First, the motion of the compact star along an eccentric orbit with a period of weeks or longer may produce periodic outbursts when the X-ray source crosses the densest parts of the envelope close to periastron. Second, most Be stars are optically variable and this variability has usually been attributed to dramatic changes in the size and density of the equatorially condensed circumstellar envelope responsible for the Balmer and infrared emission. This second mechanism may produce large variations of the X-ray luminosity whatever is the orbital phase.

Our sample of four first rank new Be/X-ray systems probably illustrates all these possible configurations. Although the time interval between the X-ray and optical observation may introduce some additional scatter, it may be significant that these four Be stars display H equivalent widths smaller than those of the previously known sources exhibiting X-ray outbursts (e.g. A1118-61, A0535+26, EXO2030+375). This suggests that the weakness of the envelope could to some extent explain the low persistent X-ray luminosities of LS 992 ( 1.4 10³³ erg s^-1) and the absence of strong recorded outbursts from BSD 24- 491. On the other hand the relatively weak H emitting Be star LS 1698 had a recorded bright outburst in the early seventies indicating that some of our new Be/X-ray sources could be hard X-ray transients in the quiescent state. These sources may undergo an outburst on the occasion of a future ejection of matter in the circumstellar envelope. LS 992 may be a particular case since it exhibited a factor 100 variation between survey and pointed observations. Optical spectroscopy contemporaneous to the pointed low state observations shows relatively weak H emission at the same level as 7 months before. We could have here a low peak X-ray luminosity outbursting system where the main mechanism for variability is orbital motion. During the follow-up pointed observations two among of our seven new candidate X-ray binaries (LS 992 and LS 1698) displayed a PSPC count rate below the detection threshold of the survey. This illustrates the strong variability of these sources and the well known fact that any new X-ray survey leads to the discovery of new members of this class.

6.3. Hard X-ray emission

A soft thermal bremsstrahlung component with a luminosity of typically 10³⁴ erg s^-1 is observed from massive X-ray binaries with a supergiant counterpart (Haberl et al. 1994). This emission arises probably in a bow shock around the neutron star traveling through the dense stellar wind of the star. The X-ray luminosity of LS 5039 observed by ROSAT might be fully accounted for by this thermal emission. Also the hardness ratios are consistent with the values found for Vela X-1 from a pointed ROSAT observation (HR1 = 0.99 0.01 and HR2 = 0.62 0.01). The relation between X-ray luminosity in the thermal component and the luminosity in the hard spectrum originating at the neutron star strongly depends on the wind density and the system parameters and it is therefore difficult to predict the luminosity at higher energies.

Be/X-ray binaries on the other hand rarely show a soft component. For instance, the X-ray spectrum from X Persei is well represented by a power law with exponential high-energy cutoff between 0.1 and 12 keV as the combined ROSAT and BBXRT results show (Haberl 1994). For these systems the spectra might be extrapolated to higher energies, however the cutoff energy lies in general outside the ROSAT band and is not known.

A mean colour excess of E(B-V) = 1.0 for the new OB/X-ray binaries corresponds to = 5.5 10²¹ cm^-2 (Predehl & Schmitt 1995). Assuming a power law spectrum with a photon index in the range of 0 - 2 and a cutoff energy 10 keV (White et al. 1983, Tanaka 1986) implies that 1 PSPC count s^-1 corresponds to an absorbed 2-10 keV flux in the range of 4 10^-11 - 7 10^-10 erg cm^-2 s^-1 or 2-30 Uhuru count s^-1 (Forman et al. 1978). The presence of a soft component will obviously decrease the hard X-ray flux corresponding to a given PSPC count rate. Therefore, with an estimated limiting count rate of 1 Uhuru count s^-1, highly depending on confusion effects in the galactic plane region, the new X-ray sources were most probably below the detection threshold of the all-sky surveys carried out by Uhuru and Ariel V (except for LS 1698 = 4U 1036-56 and LS I +61 235 which contributes to the Uhuru source 4U 0142+61). The absence of detection in the HEAO A-1 X-ray catalogue (Wood et al. 1984) above a 2-10 keV flux of 5 10 ^-12 erg cm^-2 s^-1 is also compatible with the recorded PSPC count rates.

6.4. Distribution of the new sources in the galaxy

The distribution of the new massive X-ray binary candidates in the Galaxy follows the spiral arm structures traced by H II regions (Georgelin & Georgelin 1976) and open clusters (Vogt & Moffat 1975). SAO 49725 and LS 992 are located on the local arm in opposite directions as seen from the Sun, BSD 24- 491 and LS I +61 235 are in the Perseus arm and LS 1698, HD 161103 and LS 5039 probably all belong to the Sagittarius-Carina arm. Comparing with optical absorption maps from Neckel & Klare (1980) shows that all new ROSAT detected massive X-ray binary candidates are in regions of relatively low interstellar absorption. This is not surprising considering the sensitivity of the PSPC count rate to interstellar extinction and the selection resulting from the correlation with optical catalogues. Stars with detailed spectral types are probably complete down to B 9-10 and the Luminous Star catalogue out of which about half of our optical input sample was extracted is complete down to B 12. Therefore, the optical identifications reported here tend to be close to the completeness level of the optical input catalogue, clearly suggesting that several other systems with similar X-ray luminosities but without optical entries, because associated with fainter optical objects, may well be present in the ROSAT survey. This is consistent with the fact that among the 13 known massive X-ray binaries listed in Table 10, 5 had no known entries in optical catalogues before their X-ray discovery. In a future paper we will report on some of these new ROSAT discoveries.

In spite of the patchiness of interstellar absorption and incompleteness of the optical catalogues it may be possible to estimate the space density of massive X-ray binaries (see e.g. Meurs & van den Heuvel 1989) and the low end of their X-ray luminosity function using the ROSAT all-sky survey. Such a study may allow to constrain the contribution of this population to the overall hard X-ray emission of the Galaxy and more precisely to the hard X-ray galactic ridge detected by EXOSAT (Warwick et al. 1985).

© European Southern Observatory (ESO) 1997

Online publication: May 26, 1998

helpdesk.link@springer.de