3 Comparison with existing catalogues

3.1 Correlation with the Einstein survey

The first comprehensive X-ray survey of the LMC was performed with the Einstein Observatory in the energy range 0.15-4 keV. Catalogues of discrete sources derived from the Imaging Proportional Counter (IPC) data were published by LHG81 and after a re-analysis by WHHW91. Both lists comprise together 140 distinct X-ray sources from which 12 were not covered by PSPC pointings (within 52$^\prime$). From the remaining 128 sources 78 (a fraction of 61%) were detected in the PSPC observations (within a correlation radius of 90$^{\prime\prime}$). In Table 2 the 50 IPC sources with no detection in the PSPC are summarized. The table gives the source number from LHG81 (Col. 1), the WHHW91 name (Col. 2), the number of PSPC observations which covered the IPC position (within 52$^\prime$, Col. 3), the minimum and maximum off-axis angle (Cols. 4 and 5) and the minimum and maximum exposure (Cols. 6 and 7). 67 IPC sources in the PSPC FOV were both detected by LHG81 and WHHW91 from which 55 (82%) were detected by the PSPC. The lower PSPC detection rates of 38% (15 out of 39) IPC sources found by WHHW91 only and 36% (8/22) IPC sources found by LHG81 only are compatible with the expected higher percentage of spurious detections in these subsamples. Extrapolating from the more secure sources (those detected by LHG81 and WHHW91) that $\sim$18% of the IPC sources were not recovered by the PSPC, most likely due to time variability, results in about 50% spurious detections in the number of Einstein sources which are either only detected by LHG81 or only by WHHW91. A similar conclusion was drawn by Schmidtke et al. 1998, from ROSAT HRI observations of Einstein sources. Sources 29, 55, 58, 63, 64 and 96 from LHG81 were not detected by the HRI although the observations were sensitive enough. Neither of these sources were detected by the PSPC (96 was not in the FOV), casting doubt on their reality. Table 2 shows several additional IPC detections (e.g. sources LHG 37, 65, 75, 76, 81, 84 and WHHW 521.5-6921, 523.7-6923, 524.2-6937...) which were observed many times by the PSPC and never detected.

  

Table 2: Einstein IPC sources not detected by the PSPC

In Table 4 77 Einstein IPC sources detected by the PSPC are given. In addition to the columns described for Table 1 Col. 2 gives the source number from LHG81, Col. 3 the source name from WHHW91, Col. 4 the distance between Einstein IPC and ROSAT PSPC position and Col. 7 the IPC count rate. For source 603 the IPC sources 522.5-6759 and 522.6-6801 (WHHW91) are within the correlation radius of 90$^{\prime\prime}$, both are part of the HII complex N44. To estimate the confidence of the identification, the 90% statistical error of the PSPC position is given. A systematic error of 7$^{\prime\prime}$ should be added. The error on the IPC position is estimated to 40$^{\prime\prime}$ by LHG81 (1$\sigma$).

The IPC and PSPC count rates of 76 sources from Table 4 ($\rm LMC\,X-3$ is not included because no IPC count rate is given in LHG81 and WHHW91) are compared in Fig. 1 where the sources identified with an SNR (see Sect. 3.3.1) are marked with a circle. A linear fit to the SNR count rates gives a mean conversion factor of 3.2 for IPC to PSPC count rates. The scatter shows however that the conversion rate is strongly dependent on the detailed X-ray spectrum and in particular on the absorption. Clear variability on long-term time scales is only seen for at most three sources: AB Dor, CAL83 and $\rm LMC\,X-4$ which are known to be highly variable. The different energy bands and comparison of only two epochs makes it difficult to look for time variable sources.

  

Figure 1: Einstein IPC vs. ROSAT PSPC count rates. Sources identified with SNRs are marked with circles. The line is a linear fit to the SNR data points

3.2 Comparison with the ROSAT all-sky survey

The Bright Source Catalogue (BSC) of the RASS was published by Voges et al. (1996). A correlation of the PSPC catalogue from the pointed observations with the BSC catalogue yields 55 sources within a distance of 60$^{\prime\prime}$. They are summarized in Table 5. The count rates are shown in Fig. 2. The SNR count rates scatter little around the line of ratio 1.0, as expected for constant sources. The only exception is N49 with a too low BSC count rate. Sources variable by more than a factor of 3.0 are the high mass X-ray binary (HMXB) $\rm LMC\,X-4$ and the Be X-ray transient $\rm A0538-66$ which was in outburst during the RASS (see also HP99).

  

Figure 2: ROSAT all-sky survey vs. PSPC pointing count rates. Circles mark SNRs which follow the line of equal count rates

3.3 Literature search with SIMBAD

To identify the PSPC sources with known objects we cross-correlated our catalogue with the SIMBAD data base operated at CDS. SIMBAD contains predominantly stars but also includes catalogues of sources from other wavelength bands together with references. This allows to define samples of sources with the same type (SNRs, SSSs, X-ray binaries, foreground stars and background objects) which can be used to classify the newly discovered sources using the global X-ray properties characteristic to the source class.

3.3.1 Supernova remnants

Table 6 lists 46 SNRs and candidates (indicated with SNR?) found in the literature and detected in ROSAT PSPC pointings. Many of them were already seen by the Einstein IPC. The majority appears as extended X-ray sources with typical extent of $10\hbox{$^{\prime\prime}$}-40\hbox{$^{\prime\prime}$}$.However the value can only be regarded as indicator for the extent because it is determined by a Gaussian approximation of the intensity profile. For sources with deviating profile the parameters determined by the ML algorithm become more and more unreliable. For this reason the count rates were re-determined for the sources already mentioned in Sect. 2.

From preliminary investigations of the PSPC catalogue Haberl et al. (1998) have shown that extent and likelihood for the extent (ML$_{\rm ext}$) in combination with the hardness ratios can be used to characterize the class of SNRs. In Fig. 3 extent and ML$_{\rm ext}$ are plotted for all PSPC sources detected in the inner 18$^\prime$ of the detector where the PSF of the instrument is best. The sources from Table 6 are marked with a square (filled for known SNRs and open for candidates; crossed squares for new candidates classified in this work - see below). The known X-ray binaries, SSS, foreground stars and AGN should be point sources and are marked with hexagons. On the left part of the diagram (left to the majority of unknown sources marked with a cross) many point sources appear with small extent. These are all detections with very good counting statistics (more than 400 counts) in which deviations of the PSF from the assumed Gaussian shape become significant, resulting in an artificial extent. This includes also most of the known SNRs which shows that the two parameters can only be used as indicator for source extent. Nevertheless the known SNRs are clearly distinguished showing the highest ML$_{\rm ext}$ values. For some of the SNR candidates found in the literature (mainly new candidates suggested by radio observations, Filipovic et al. 1998) the source parameters derived from the PSPC observations support their classification. This is particularly true for PSPC source 712 which was detected by Einstein and suggested as SNR by WHHW91 (see Table 6). For source 687 source extent and hardness favour an SNR identification and the nearby foreground star proposed by Cowley et al. (1997) as optical counterpart for $\rm RX\,J0528.6-6836$ is probably unrelated to the X-ray source. There is one unknown source (93) which is located in the area of the diagram where only SNRs are found. The hardness ratios are also compatible with such an interpretation. Other new sources with the highest ML$_{\rm ext}$ values, but below the majority of known SNRs, need to be investigated in detail as promising SNR candidates.

  

Figure 3: Source extent and extent likelihood for PSPC sources with off-axis angle less than 18$^\prime$ (crosses). SNRs are marked with squares (filled: secure, open: candidates, crossed: classified in this work) and known point sources (X-ray binaries, SSSs, stars and AGN) with hexagons

3.3.2 X-ray binaries and supersoft sources

X-ray binaries belong to the objects with hardest X-ray spectrum in the ROSAT energy band. From an investigation of the variability on time-scales of days to years of the PSPC sources presented in this work, HP99 proposed seven new candidates for X-ray binaries. Most of them are probably HMXBs and one is a candidate for a low mass X-ray binary (LMXB). The HMXB nature of one of them (source 1225, $\rm RX\,J0544.1-7100 = 1SAX\,J0544.1-710$) was confirmed by the independent detection of X-ray pulses by the BeppoSAX satellite (Cusumano et al. 1998). Together with the previously known X-ray binaries their properties as determined from the PSPC observations are summarized in Table 7. It should be noted that the count rates (Col. 5) are only representative for a single observation. For more information on variability see HP99.

Supersoft sources were established as a new class of X-ray sources after the ROSAT discoveries of five new such objects. Before ROSAT only two (CAL83 and CAL87) were known from Einstein observations. SSSs are characterized by very soft X-ray spectra resulting in PSPC hardness ratios HR1 and HR2 close to -1.0. Foreground absorption may however increase HR1. HP99 proposed two new SSS candidates from their hardness ratios and X-ray variability. The properties of the known SSSs in the LMC which are all detected in PSPC observations are summarized in Table 7.

3.3.3 Background AGN and galaxies

Relatively few of the PSPC sources were identified with background objects like AGN and galaxies (Table 8), although many are expected from the $\log N - \log S$ distribution of extragalactic objects. Using a sensitivity of $6\ 10^{-3}$ cts s^-1 for a typical 2000 s observation and the $\log N - \log S$ distribution from Hasinger et al. (1998), one would expect of the order of 5 background objects per square degree in LMC areas where the absorbing column density is 3 10²¹ cm^-2 (e.g. Dickey et al. 1994). The PSPC observations cover in total an area of 58.6 square degrees, including areas where the absorption is only by the galactic contribution of about 6 10²⁰ cm^-2. There the number of background objects can be up to 10 per square degree. Also large areas are covered by observations deeper than 2000 s, resulting in a lower limit of about 300 background objects expected in the catalogue.

Most of the known AGN were optically identified due to ROSAT follow-up observations of Einstein sources (e.g. Schmidtke et al. 1998). Three galaxies and a group of galaxies are proposed as optical counterparts from their positional coincidence. However, in these cases the error on the X-ray position is large and the identifications need to be confirmed as is also the case for source 1367.

3.3.4 Foreground stars

The source classes described so far contain bright X-ray sources which were in many cases already detected by X-ray instruments launched before ROSAT. Also ROSAT sources identified optically are already published. The high sensitivity of the ROSAT PSPC allowed for the first time to study the X-ray emission from normal stars. A large number of stars (mainly active corona late type stars) are therefore expected in the PSPC source catalogue in the direction of the LMC. Since many of these stars are detected in X-rays the first time, PSPC sources were "identified" with stars by correlation with the SIMBAD catalogue, i.e. by positional coincidence only. Few stars were identified by follow-up optical work published in the literature and six sources were proposed by HP99 as foreground stars from their X-ray spectral and temporal properties. In total up to 57 of the PSPC sources investigated in our fields correlate with foreground stars (Table 9). In 10 of these cases the identification is uncertain due to a large distance to the optical position and/or uncertain X-ray position.

A flux ratio $f_{\rm x}/f_{\rm opt}$ in the X-ray and optical band was calculated from the 0.1-2.4 keV count rate and optical magnitude $m_{\rm GSC}$ from the GSC from log($f_{\rm x}/f_{\rm opt}$) = log(PSPC counts/s $10^{-11})+0.4 m_{\rm GSC} + 5.37$(Maccacaro et al. 1988; Voges et al. 1999). We use here the more complete GSC instead of V or B magnitudes available for a smaller number of sources from the correlations with SIMBAD. A clear correlation of $f_{\rm x}/f_{\rm opt}$ with spectral type of the stars is seen with A and F stars as weakest X-ray emitters while dMe stars show highest log($f_{\rm x}/f_{\rm opt}$) values around -1.0 (Table 9).

3.4 Source classification

In Tables 6-9 in total 144 source "identifications" are summarized. In many cases these are based on optical identifications of previously known X-ray sources. Identifications from positional coincidence only (mainly foreground stars) are supported by their X-ray properties like hardness ratios and X-ray to optical flux ratios. In the following we investigate this sample of identified sources to find properties unique to the different types of X-ray emitters with the aim of a classification of the unidentified sources. Kahabka et al. (1999) have classified ROSAT PSPC sources in the Small Magellanic Cloud using hardness ratio criteria. Figure 4 (top) shows the hardness ratios of identified sources which have errors on both hardness ratios of less than 0.25. The different source classes occupy partially overlapping areas of the diagram. There are however parts of the parameter space where only sources from one class are found. SSSs exhibit HR2 below -0.70 and only foreground stars are found with $\rm HR2\gt-0.70$ and $\rm HR1<0.25$. In the range $\rm HR1\gt.25$ and $\rm -0.70<HR2<0.25$ only SNRs are located. In Fig. 4 (bottom) the hardness ratios of unidentified PSPC sources are drawn. The areas covered by different source classes are indicated by thick lines in both diagrams. Since most of the 758 PSPC sources have large error bars on the hardness ratios a restrictive selection was used to classify new sources. This yields a smaller number of sources, but maximizes the probability for giving the correct classification.

A more detailed classification requires the knowledge of foreground absorption to the sources. Because this varies across the LMC, sources with the same intrinsic X-ray spectrum are distributed over a range of hardness ratios, in particular HR1 which is most sensitive to absorption. The selection criteria used here are relatively insensitive to absorption. Only stars may be misclassified when higher absorption increases HR1 above 0.25.

  

Figure 4: Hardness ratios of known PSPC sources (top). X-ray binaries are marked with a hexagon, SSSs with crossed square, SNRs with square, stars with x, and AGN with triangle. The thick lines separate areas where only members of a single class are found. Hardness ratios of new PSPC sources with unknown nature (bottom). Only sources with error on HR1 and HR2 less than 0.25 are shown

Table 3 summarizes the selection criteria used to find new promising candidates for SNRs and SSSs in the LMC, foreground stars and background objects. The numbers of sources which obey the selection criteria (and not already included in a previous selection) are given in Col. 3. The number of finally classified sources are found in Col. 4. Sources on the hardness ratio dividing lines (error bar crossing the line) were included when the errors on HR1 (EHR1) and HR2 (EHR2) are both less than 0.25. As expected three of the HR-selected SNR candidates (153, 887 and 1293) are very likely identified with stars in the GSC, visible as bright point-like objects on the DSS images. These three were re-classified as foreground stars which is also consistent with their low $f_{\rm x}/f_{\rm opt}$ ratio (see below). One further HR-selected SNR candidate (327) was removed from the classification because it is located near the south ecliptic pole and far from the LMC. The classified sources are summarized in Table 10 and indicated by [SNR], [SSS] or [fg Star] in the remark column.

Supernova remnants are characterized by constant X-ray flux. The SNR candidates (from literature and from this work) were therefore investigated for temporal variability. In case of more than two PSPC detections a chi-square test against a constant count rate was performed as in HP99. The more appropriate hard band (0.5-2.0 keV) was used. One case with clear variability was found for source 440 (probability more than 5 $\sigma$), which rules out the identification with an SNR. The detections of sources 93, 540, 712 and 1063 are consistent with constant flux, in agreement with their proposed SNR nature.

In the upper right part of Fig. 4 one finds all, stars, SNRs, AGN and X-ray binaries and it is not possible to classify the hard sources uniquely according to their hardness ratios. The very hard sources ($\rm HR1\gt.75$ and $\rm HR2\gt-0.10$) are candidates for either SNRs, AGN or X-ray binaries. Thirty-seven sources are classified as [hard] in Table 10. An inspection of the DSS images in the error circle of these hard sources reveals galaxy-like extended objects in three cases (101, 653 and 1184). All are located far from the LMC supplying further evidence that they are background objects unrelated to the LMC, in particular source 1184 is classified as AGN below. Also sources 418 and 1189 are located in areas where no LMC objects are expected. From the remaining hard sources many show empty error circles on the DSS images - at least those without indication of X-ray extent and position errors less than 20$^{\prime\prime}$. In three cases (482, 747 and 1181) even optical follow up observations failed to identify the optical counterpart (see Table 10). This suggests either background AGN or LMC objects optically too faint to be identified. If located in the LMC their PSPC count rates indicate typical X-ray luminosities of $\rm 10^{34}-10^{35}~erg~s^{-1}$, too high for cataclysmic variables, but consistent with LMXBs. The latter consist of a neutron star and a late type dwarf star, too faint for current optical instrumentation to be detected at the distance of the LMC, consistent with the empty error box. It is remarkable that most of these LMXB candidates are located around the optical bar of the LMC where also the SSSs are found. SSSs and LMXBs evolve from low-mass stars and belong to an older population compared to the HMXBs, many of which are found in areas of more recent star formation (e.g. in LMC4, HP99).

  

Table 3: Classification criteria

AGN and stars can be distinguished from their flux ratio $f_{\rm x}/f_{\rm opt}$ in the X-ray and optical band (e.g. Fig. 14 in Voges et al. 1999). Calculating $f_{\rm x}/f_{\rm opt}$ for a subsample of the PSPC catalogue with clear optical identifications and available visual or blue magnitudes (most from Schmidtke et al. 1994) yields similar results. $f_{\rm x}/f_{\rm opt}$ for some identified AGN, SSSs, X-ray binaries and foreground stars is shown in Fig. 5. The six identified AGN, gather near log($f_{\rm x}/f_{\rm opt}$) = 0 with HR1 above 0.8. However also SSSs and X-ray binaries can show high $f_{\rm x}/f_{\rm opt}$ ratio, even exceeding that of AGN. Stars have the lowest $f_{\rm x}/f_{\rm opt}$ and usually softer spectra (smaller HR1). For unidentified PSPC sources with a nearby GSC entry $f_{\rm x}/f_{\rm opt}$ was calculated using again the optical magnitude given in the GSC. The cases with error on HR1 of less than 0.25 are included in Fig. 5. Five sources with $\rm HR1<0.25$ were not classified by criteria described above and show log($f_{\rm x}/f_{\rm opt}$) <-0.5, compatible with foreground stars. Two more candidates are found near identified foreground stars with $\rm HR1<0.75$ and log($f_{\rm x}/f_{\rm opt}$) <-2. These seven sources are classified as [fg Star] in Table 10. Source 1163 with $\rm HR1=-1$ and low $f_{\rm x}/f_{\rm opt}$ (unusual for SSSs and also white dwarfs using examples from Thomas et al. 1998) is probably unrelated with the GSC entry as the GSC position is also outside the X-ray error circle.

Three unidentified PSPC sources exhibit $\rm HR1\gt.25$ and high log($f_{\rm x}/f_{\rm opt}$) above -1.0 (1, 37 and 1184). All three are located in the outermost regions of the 10$^\circ$ by 10$^\circ$ field covered by PSPC observations and are probably unrelated to the LMC. Source 1184, mentioned before to have a possible galaxy as likely optical counterpart, correlates as well as source 37 with a radio source. Based on the high $f_{\rm x}/f_{\rm opt}$ they are classified as [AGN] in Table 10. The three AGN candidates are relatively bright PSPC sources with two of them also detected in the RASS (1 and 37). Their optical brightness given in the GSC, range from 15.0 to 15.3 mag. The majority of expected AGN in the PSPC catalogue is fainter and to find them by utilizing their $f_{\rm x}/f_{\rm opt}$ ratio more complete optical catalogues of the LMC area are required.

  

Figure 5: Flux ratio $f_{\rm x}/f_{\rm opt}$ as function of hardness ratio 1. The optically identified sources are marked with different symbols (X-ray binaries: hexagon, SSSs: crossed square, foreground stars: x and AGN: triangle). Unidentified sources with error on HR1 less than 0.25 and nearby GSC entry are shown with HR1 error bars

Additional PSPC sources likely related to foreground stars were obtained from the correlation with the GSC. A conservative selection of sources with position uncertainty r₉₀ of less than 20$^{\prime\prime}$ and a distance to GSC entries of less than r₉₀ yields 29 objects. Removing unclear cases after visual investigation of the DSS images (several objects with similar brightness exist in the X-ray error circle) and objects with optical extent indicated in the GSC yields 16 stellar-like sources. Two of them are already classified as foreground stars from their low $f_{\rm x}/f_{\rm opt}$ ratio. The remaining 14 are classified as [stellar] in Table 10. In general this sample consists of X-ray fainter objects compared to the hardness ratio classified foreground stars (which have better photon statistics and therefore smaller errors on the hardness ratios). None of the 16 sources shows significant X-ray extent. However from their optical brightness - GSC magnitudes range from 10.1 to 14.8 mag - early type LMC stars not can be excluded (e.g. HMXBs in the LMC identified with Be stars have magnitudes at the lower end of the brightness distribution, cf. Haberl et al. 1997).

3.5 Source distribution

The spatial distribution of the 758 PSPC sources in the catalogue is shown in Fig. 6. The 144 identified sources from Tables 6-9 (including literature candidates) are marked according to their source class. In Fig. 7 the sources belonging to the LMC (SNRs, SSSs and X-ray binaries) are shown. Candidates and sources classified in this work are included. As was noted already by HP99 there is a significant concentration of X-ray binaries in and around the LMC4 supergiant shell. In contrast SSSs are only detected along the rim of the optical bar of the LMC. This includes all new SSS candidates from HP99 and this work, supporting their proposed nature. Many SSSs may be hidden inside the bar where the soft X-rays are strongly attenuated by photo-electric absorption.

  

Figure 6: Distribution of X-ray sources in the LMC region detected by the ROSAT PSPC (small circles). The source positions are plotted on a weak grey scale image (0.1-2.4 keV) for orientation. Identified sources are marked with different large symbols for different source classes; cross: SSS, square: SNR, circle: X-ray binary, cross + circle: foreground star, cross + square: background object

  

Figure 7: Distribution of PSPC detected LMC sources: SNRs (square), SSSs (cross) and X-ray binaries (circle) including new candidates from this work. The background image is the same as in Fig. 6