5 Source classes

The X-ray spectral characteristics of the catalogued sources are given as two hardness ratios (the soft hardness ratio - HR1 and the hard hardness ratio - HR2), the source extent (assuming a Gaussian model) and the count rates in a spectrally soft and hard band. This information is used to characterize a source and to allow a source classification. We previously applied some selections to this catalogue in order to derive the sample of super-soft X-ray sources of the SMC (Kahabka et al. 1994). In Kahabka & Pietsch (1996) we applied another set of selections in order to derive the sample of spectrally hard X-ray binary candidates with luminosities in excess of $\rm 3\ 10^{35}\ {\rm erg}\ {\rm s}^{-1}$. These subsets comprise only a small number of sources (four supersoft and 7-13 hard). The majority of sources (209) have not been classified previously.

In a recent paper Filipovic et al. (1998) presented results of a comparison between the ROSAT PSPC catalogue presented here and the Parkes catalogue of sources towards the SMC at radio frequencies (1.42, 2.45, 4.75, 4.85 and 8.55 GHz). They found 27 sources in common to both surveys (cf. Table 3 in their paper). These include 14 SNRs (with two SNR candidates), eight background sources and three HII regions.

We now define other classes, set up the selection criteria, identify the members, derive the relevant distributions of the source properties and discuss the implications in terms of population studies. The positional distribution of the classified sources using criteria established here (Table 3) is shown in Fig. 2.

  

Table 3: Definition of the source classes and the selection criteria

Remark: additional classes from refined classification: AB = AGN or hard X-ray binary source and H = HII region.

  

Figure 2: Distribution of classified X-ray detections (this work) in the field of the SMC plotted over high resolution HI image of the SMC (from Stanimirovic et al. 1999). Selected classes are: super-soft sources (red), hard X-ray binaries (blue), AGNs (black), SNRs (green) and foreground stars (yellow). Big/small circles are drawn for the high/low luminosity subclass respectively. For the AGNs and hard X-ray binaries the refined classification is used (cf. Table 1)

5.1 The high luminosity super-soft sources

This class was extensively discussed in Kahabka et al. (1994) and in Paper I. Applying the strict selection criteria outlined in these papers ($HR1+\delta HR1<-0.8$ and count rate $\rm \gt.015~s^{-1}$) we obtain candidates classified as class = Sl in Table 1. Sources Nos. 20, 62, 133, 135 and 181 were discussed in Paper I. Sources Nos. 6, 112 and 173 coincide with struts of the PSPC detector system and are classified as possible detector artifacts.

5.2 The low luminosity super-soft sources

Low luminosity super-soft sources are not yet established as a new class of objects, but they are expected to exist e.g. as hot central stars of planetary nebulae (PNe) which, if they are on the cooling track, can attain luminosities in the range $\rm 10^{34} - 10^{36}~ {\rm erg}\ {\rm s}^{-1}$. This would translate into PSPC count rates for sources in the SMC of $\rm \sim7\ 10^{-5}$ - $\rm 7\ 10^{-3}\ s^{-1}$ i.e. below our threshold of $\rm 0.015\ s^{-1}$ for high luminosity super-soft sources.

There are three objects which are fulfilling these criteria and they are classified as class = Sw in Table 1. The source No. 38 is found at a large off-axis angle of 43' and is extended (Gaussian $\sigma=214''$). This may indicate either diffuse structure or an artifact. The source No. 141 coincides with a Simbad star of 17.5 mag and with the planetary nebula L357 in the catalogue of Meyssonnier & Azzopardi (1993). It is therefore a real candidate. Source No. 243 is found at an off-axis angle of 20' and coincides with the inner PSPC support structure and may be an artifact. This leaves us with one firm low-luminosity super-soft source No. 141 (RX J0059.6-7138).

5.3 The "stronger'' hard X-ray binary candidates

The sample of X-ray binary candidates with luminosities in excess of $\rm 3\ 10^{35}\ {\rm erg}\ {\rm s}^{-1}$ has been given in Paper I. The selection criteria were $HR1-\delta HR1\gt.5$, $HR2-\delta HR2 \gt.3$ with extent likelihood $LH_{\rm extent}<50$ and count rate $\rm \gt.015~s^{-1}$. We classify these sources as class = Bl in Table 1.

The sources Nos. 3, 69, 83, 84, 100, 157, 158, 160 and 242 were selected as candidate X-ray binaries. In Paper I sources with Nos. 26, 105 and 153 were rejected mainly due to the absence of time variability. For these sources radio emission was found in the radio survey of Filipovic et al. (1998) and they are candidates for AGNs.

The source No. 103 (RX J0054.9-7226) is identified with the XTE J0055-724 = 1SAX J0054.9-7226 source (Marshall et al. 1998; Israel 1998). The detection of pulsations with a period of 59 s with BeppoSAX confirm the X-ray binary nature of this source.

Stevens et al. (1998) identified early type emission-line stars through colour indices and H$\alpha$ emission for the sources with catalogue indices 3, 69, 103, and 158.

The source No. 153 (RX J0100.7-7211) was considered to be consistent with a background AGN shining through the SMC bulge (Paper I). Sources Nos. 157 and 160 were found to coincide with detector struts and were rejected accordingly (Paper I).

5.4 The "weaker'' hard X-ray binary candidates

Weak hard X-ray binaries are an interesting class of X-ray objects as they have been predicted to exist and their number is expected to be large especially in galaxies of low metallicity like the SMC. In previous work (Bruhweiler et al. 1987; Wang & Wu 1992) candidates for such sources have been found and either classified as low luminosity Be systems or as background objects.

Here, we are searching for candidates of this class by applying the same selection criteria as for the strong (or higher luminosity) hard X-ray binaries as outlined in Paper I: $HR1-\delta HR1\gt.5$, $HR2-\delta HR2 \gt.3$ and extent likelihood $LH_{\rm extent}<50$. The only difference is to select objects with count rates of $\rm <0.015\ s^{-1}$, i.e. with luminosities below $\rm \sim 3\ 10^{35}\ {\rm erg}\ {\rm s}^{-1}$ assuming a standard spectral model for the source flux (Paper I). There are 60 such objects and we tentatively classify these sources as class = Bw. This class is a substantial fraction (25%) of the total catalogue entries and turns out to be the class with most members.

We find 15 of these objects which coincide with Einstein detections. This may reflect that we are considerably deeper in sensitivity than the Einstein survey. Some 17 objects were found to coincide with a Simbad source. However, such a correlation could be misleading as most of the distances to the Simbad sources are too large (>60'') to be considered reliable. Only three sources have a distance to a Simbad source of <50'' and may be considered to be identified. This is a very small fraction of all catalogued sources here.

Four sources are found close to the inner ring of the PSPC window support system and may be artifacts. A more thorough investigation of these sources appears to be required as these sources can still be real. Another 46 sources have been found inside the inner support ring of the PSPC detector and are considered as firm candidates. This comprises 22% of all reliable entries in the catalogue. Sources found outside the detector ring suffer due to less accurate positions.

"Screening'' of the catalog using hardness ratios derived from simulated power-law slope -0.8 spectra for hard X-ray binaries, and power-law slope -2.0, and slope -2.6 spectra for "radio loud'' and "radio quiet'' AGNs respectively gives 43 firm weak hard X-ray binary candidates. Six previous hard X-ray binary candidates are consistent with AGNs and 11 candidates are consistent with either class (cf. Table 1).

Recent observations towards the SMC with ASCA, BeppoSAX and Rossi-XTE established 10 new X-ray pulsars in this galaxy (Table 4). Most (if not all) of them appear to be connected with a Be-type donor star. Pulsation periods in the range 3-345 s have been determined. This range in pulsation periods is covered by the range of pulsation periods found in the galactic Be-star X-ray binaries of $\sim 4-1500$ s (cf. van den Heuvel & Rappaport 1987). An orbital period has been estimated only for the two systems AX J0051-722 and XTE J0053-724 with 110-120 and 139 days. The orbital period deduced for AX J0051-722 is in agreement with the relation between pulsation period and orbital period found by Corbet (1986), while the orbital period estimated for XTE J0053-724 is twice the predicted period. Five of these new X-ray pulsars may have a counterpart in our SMC X-ray catalogue. Source 98 (RX J0053.9-7226) has been discovered with Rossi-XTE = RXTE (cf. Levine et al. 1996) in an outburst and 46.6 s pulsations have been found (Corbet et al. 1998). This confirms the correct classification of this source. In addition, source 79 (RX J0051.3-7216) may be identical with AX J0051-722 (Corbet et al. 1998) also confirming the correct classification. Source 89 coincides with the transient source RX J0052.9-7158 of Cowley et al. (1997) and AX J0051-73.3 (Yokogawa et al. 1998b) coincides with RX J0050.7-7316 (cf. Cook 1998; Cowley et al. 1997; Kahabka 1998). This source with catalog index 72 apparently both fits to the AGN and the hard X-ray binary class but it is a strong candidate for a hard X-ray binary.

The derived number of 51 hard X-ray binary candidates may be compared with the number of X-ray binaries predicted from the population synthesis calculations of Dalton & Sarazin (1995) for the SMC. These calculations predict 46 X-ray binaries with luminosities in excess of $10^{34}\ {\rm erg}\ {\rm s}^{-1}$, the lower sensitivity limit of our SMC X-ray survey.

  

Table 4: New X-ray pulsars discovered in recent ASCA, BeppoSAX and Rossi-XTE observations of the SMC

Refs. Corbet et al. 1998 (1), Lochner 1998b (2), Buckley et al. 1998a (3), Lochner 1998a (4), Israel et al. 1998 (5), Marshall et al. 1998 (6), Israel 1998 (7), Yokogawa & Koyoma 1998a (8), Kahabka & Pietsch 1998 (9), Yokogawa & Koyoma 1998b (10), Kahabka 1998 (11), Cook 1998 (12), Schmidtke & Cowley 1998 (13), Yokogawa & Koyama 1998b (14), Yokogawa & Koyoma 1998c (15), Wojdowski et al. 1998 (16), Hughes 1994 (17), Clark et al. 1997 (18), Israel et al. 1997 (19), Buckley et al. 1998b (20), Yokogawa & Koyama 1998d (21), Imanishi et al. 1998 (22), Chakrabarty et al. (1998a) (23), Wilson & Finger (1998) (24), Chakrabarty et al. (1998b) (25).
^*) MACHO has seen a 0.708 day period. This is consistent with a rotation period of a Be star.

  

Table 5: Upper panel: Previously detected SNRs identified in the ROSAT sample, classified as SNR candidates and extended structure. The Einstein No. refers to Wang & Wu (1992). Source radio name is given following Filipovic et al. (1998)

5.5 Supernova remnants and other extended structures

SNRs have been identified in the SMC by work done in the radio, optical and X-ray regime (Mills et al. 1982, 1984). In the work of Ye & Turtle (1993), some 15 SNRs and SNR candidates are detected in a 843 MHz survey.

We applied the selection criteria likelihood of extent $LH_{\rm extent} \gt 50$ and $HR1-\delta HR1 \gt-0.8$ in order to derive a candidate sample of SNRs and other extended structures. We find 19 objects fulfilling these criteria (including four SNRs detected by Einstein, although they have a $LH_{\rm ext}$ below our classification threshold: sources 86, 128, 177 and 182). These sources are classified as class = R in Table 1. The sources with catalogue number 95, 132, and 136 may be detector artifacts. They have the class = R/D. For 13 sources an Einstein identification has been found. We give the Einstein number and the identification for these sources in Table 5. 12 of these sources actually are known SNRs. The bright young oxygen-rich SNR 0102-723 (Amy & Ball 1993) has two entries (182 and 183) in our catalog due to the merging of two pointed observations at different off-axis-angles. Entry 183 is the more accurate one. An additional classified source (34) correlates with a SNR proposed by Filipovic et al. (1998). We find new candidate SNRs in our X-ray survey: RX J0101.8-7249 (source 165) and RX J0112.7-7207 (223) were not reported before, while RX J019.4-7301 (245) was already detected with Einstein (BKGS 30, Bruhweiler et al. 1987), however not classified as SNR.

We miss a few well known SNRs in the SMC with this selection. They have an extent likelihood ratio >10 and fulfill the criteria of being extended, suggesting we have chosen too strict an extent criterion in order to be on the secure side. It also suggests that there are still unrecognized SNRs in our sample (see discussion in Filipovic et al. 1998). We list these SNRs in the lower part of Table 5.

5.6 Foreground stars

It is not trivial to select this sample just from the X-ray characteristics. Stars are coronal emitters with temperatures in the range of a few $\rm10^6$ to $\rm10^7$ K. The HR1 would then fall into the regime $HR1 +\delta HR1<0.5$, $HR2-\delta HR2\gt-0.8$ and the likelihood of extent $LH_{\rm extent}<50$. Actually, all galactic foreground stars detected in a $8^{\circ}\times8^{\circ}$ field centered on the SMC and observed during the RASS have values of HR1 <0.1 (Kahabka & Pietsch 1993). This means that we may be too conservative in selecting stars in our sample with the criteria mentioned above.

We find 19 candidates and we classify these sources as class = F in Table 1. For seven objects, a Simbad match exists. Six of these identifications appear to be reliable as the distance to the Simbad source is <60''. Five of them correlate with stars of spectral type O, A or F and they are given in Table 6. Source 138 coincides with an O star. Assuming a conversion factor of $0.76\ 10^{37}\ {\rm erg}\ {\rm cts}^{-1}$ we derive for the O-star in the SMC an X-ray luminosity of $10^{34}\ {\rm erg}\ {\rm s}^{-1}$ from the measured count rate. The nature of the other 14 sources remains unclear.

  

Table 6: Optically identified X-ray selected candidates identified with Simbad stars

5.7 Background objects and AGNs

Background objects are selected as $HR1-\delta HR1\gt.5$, $HR2+\delta HR2<0.3$ and extent likelihood $LH_{\rm extent}<50$. The same criterion has already been applied in Paper I in order to identify a possible sample of AGNs in the high luminosity X-ray binary candidate sample. We find 20 sources to fulfil these criteria. Source 148 correlates with a SNR and has to be removed from this class. Three sources correlate within a radius of <60'' with a Simbad star (either foreground or SMC). We refine and extend the classification of candidate AGNs by comparing the measured hardness ratios with the predictions from simulated power-law spectra of slope -2.0 and -2.6. We find 53 candidates for class = A and 62 candidates if we also consider class = AB in Table 1. Class = AB means a hard X-ray binary nature is also possible due to the hardness ratio criteria.

Tinney et al. (1997) present 10 quasars behind the SMC. Seven of them are covered by the fields X1 and C (cf. Table 2) and are listed in Table 7. None of these quasars was detected in the radio survey of Filipovic et al. (1998). QJ0102-7546, one of the three quasars not covered by our survey was detected in the RASS field (Kahabka & Pietsch 1993).

  

Table 7: "Tinney'' quasars found in pointings X1 and C of the SMC X-ray survey

We independently calculate the number of background sources in the analyzed field by taking the distribution of the neutral hydrogen into account. By making use of the standard log$(N)-\log(S)$ of the soft extragalactic X-ray background (Hasinger et al. 1993) and by taking into account the absorption due to the SMC (from the HI image of Stanimirovic et al. 1999) the expected number of background sources with absorbed fluxes in excess of $10^{-13}\ \rm erg\ {\rm cm}^{-2}\ {\rm s}^{-1}$ has been determined. We derive a number of 10 background sources for our covered field. As we expect to be complete for this flux limit we consider 10 background sources as the lower limit. The lower end of the flux distribution extends to $10^{-14}\ {\rm erg}\ {\rm cm}^{-2}\ {\rm s}^{-1}$. In case of completeness we expect to detect 519 background sources. The fact that we classify 53-62 sources as background sources is consistent with these numbers.