Free Access
Issue
A&A
Volume 555, July 2013
Article Number A65
Number of page(s) 48
Section Catalogs and data
DOI https://doi.org/10.1051/0004-6361/201321165
Published online 01 July 2013

© ESO, 2013

1. Introduction

thumbnail Fig. 1

Optical image of the Andromeda Galaxy (DSS2 blue) with X-ray source regions from the Chandra HRC-I catalogue ~50% encircled count fraction radius) and a circle marking the outer rim of the detector FOV (image obtained from the NASA Sky View website).

Open with DEXTER

M 31 is an ideal target for X-ray observations since it is the closest spiral galaxy and similar to our own galaxy. The outside view and the distance of M 31 enable us to monitor several hundred X-ray sources in just one telescope pointing. The Einstein X-ray observatory (e.g. Trinchieri & Fabbiano 1991) and the ROSAT satellite (Primini et al. 1993; Supper et al. 1997, 2001) gave a first insight into the X-ray population of M 31, followed by the instruments of Chandra (e.g. Kong et al. 2002b; Kaaret 2002; Williams et al. 2004; Voss & Gilfanov 2007a) and XMM-Newton (e.g. Pietsch et al. 2005b; Stiele et al. 2008, 2011) with better spatial resolution and deeper observations.

Several individual X-ray sources in M 31 have been discussed in previous publications (e.g. sources showing X-ray dips, pulsations, bursts, or ultra-luminous X-ray outbursts, for references see Sect. 4 and Table 1). In addition, many source classes have been studied. The main class of M 31 X-ray sources are X-ray binaries (XRB), in which a compact object like a white dwarf, a neutron star, or a black hole accretes matter from a companion star. Such sources can show a highly variable or transient behaviour (e.g. Osborne et al. 2001; Williams et al. 2006). The subclass of globular cluster X-ray sources (GlC), which are mostly low-mass XRBs (LMXBs), has been covered in several papers (e.g. Di Stefano et al. 2002; Trudolyubov & Priedhorsky 2004; Peacock et al. 2009; Barnard et al. 2012b). Another subclass of XRBs are defined by their super-soft X-ray spectrum (SSS; e.g. Di Stefano et al. 2004). Most of these SSSs have been found to be associated with optical novae (Pietsch et al. 2005a, 2007; Orio 2006). Another class of X-ray sources are supernova remnants (SNR), which can be spatially resolved in M 31 using X-ray, optical, or radio observations (Kong et al. 2002a; Sasaki et al. 2012, and references therein).

Beginning in 1999, the Local Group galaxy M 31 (Andromeda Galaxy) has been regularly observed with the Chandra X-ray Observatory. The satellite High Resolution Mirror Assembly (HRMA, see Weisskopf 1988) was frequently pointed at the centre of M 31, which allows us to examine the long-term evolution of the source intensities within the central field of the galaxy. We used 64 observations of the Chandra High Resolution Camera in imaging mode (HRC-I, see Murray et al. 1997) from 1999 to 2012. The single Chandra HRC-I observation with the longest exposure time has been discussed in Kaaret (2002). Williams et al. (2004) have discussed snapshots covering the whole galaxy until 2002.

The observations from 2006 to 2012 are part of an extensive monitoring program (41 observations with about 20 ks exposure), to investigate X-ray sources associated with optical novae. These observations have been searched for novae in Henze et al. (2010) and Henze et al. (2011). We now use these monitoring observations, along with all previous archival HRC-I observations, to systematically analyse the time variability of all sources detected in the M 31 central field with Chandra HRC-I.

We used the HRC-I detector, which has a wide field of view (32′ × 32′) and a very high angular resolution (below 1′′) for sources close to the pointing of the telescope1. The dense HRC-I monitoring of M 31 allows us to create the deepest (about 1 Ms of total exposure time) catalogue of X-ray sources at the centre of M 31 and provide a detailed long-term variability study for each source. Figure 1 shows an image of the Digitized Sky Survey 2 (DSS2) taken with a blue filter and the approximate HRC-I field of view (FOV).

2. Observations and data reduction

As mentioned above, there are 64 Chandra HRC-I observations to date, pointed to within 3′ from the centre of M 31 (RA: 00h42m44.33s, Dec: +41° 16′ 07.50′′, J2000, Skrutskie et al. 2006). We excluded two ~1 ks observations with high background count rates (ObsID 00270 and 01569), which would not have contributed new sources to the catalogue and would have deteriorated the signal-to-noise ratio when merging the individual observations. Table A.1 contains information on all HRC-I observations used. The observations from 1999 to 2002 mostly have an exposure of ~1.2 ks (one with 2.5 ks and one with 5.2 ks), except for ObsID 1912 (46.7 ks), which was used by Kaaret (2002) for his investigation. From 2004-12-06 till 2005-02-21 there were six HRC-I observations of the M 31 centre, with exposures between ~20 and ~50 ks to monitor the variability of M 31*, the source at the M 31 nucleus. The period from 2006 to 2012 is covered by the ~20 ks HRC-I observations of the XMM-Newton/Chandra nova monitoring programme and additional Chandra guaranteed-time observations of the Max-Planck-Institut für extraterrestrische Physik (MPE). In this period we have very regularly spaced data points with about the same exposure. This monitoring delivers the best input for the time variability studies.

2.1. Source detection

The data for each observation were reprocessed using the Chandra Interactive Analysis of Observations software package (CIAO, Fruscione et al. 2006) version 4.4 and the Chandra Calibration Database (CalDB, Graessle et al. 2007) version 4.4.7, to receive a level-2 event list according to the Chandra Standard Data Processing (SDP). For each observation we created a light curve of the whole detector and removed events at times of high background from the event file (good time interval (GTI) filtering) using the Munich Image Data Analysis System (MIDAS) software package provided by the European Southern Observatory (ESO). We only used events in the pulse-invariant (PI) channel range from 48−293, which according to the CIAO science threads increases the contrast between sources and background2.

As the point spread function (PSF) of the Chandra HRMA causes higher source-blurring with increasing off-axis angle, we used five images with different resolution (similar to Kaaret 2002). This reduces computation time, since the detection does not have to run on the entire detector FOV at highest resolution. The highest resolution image covers 200′′ × 200′′ with 0.13′′ pixel size (HRC-I detector resolution). The second, third, and fourth image cover 400′′ × 400′′, 800′′ × 800′′ and 1400′′ × 1400′′, with a binning of 2 × 2, 4 × 4 and 8 × 8 detector pixels. The fifth image covers the entire detector and has a 16 × 16 binning.

The size of the PSF increases by a factor of ~30 from the centre to an off-axis angle of ~15′. where the effective exposure decreases to about 70%. This means that the spatial resolution and the sensitivity of the catalogue depend on the off-axis angle of a source.

Using CIAO tools, we created an exposure map for each image, correcting for vignetting effects of the HRMA and for the efficiency of the detector. After testing various background models (by subtracting background maps from images and inspecting residuals), we found that the XMM-Newton Science Analysis System (SAS, Gabriel et al. 2004) task esplinemap delivered the best results for modelling background contributions and the diffuse emission of M 31. To separate point-source contributions, we performed a detection run with the CIAO tool wavdetect with the sensitivity parameter sigthresh = 10-4. We chose a slightly higher sensitivity than recommended in the CIAO science threads3 to avoid missing any detections. Later we removed detections below a significance of 3σ. To avoid spurious detections at the detector edges, we used a detection mask to exclude areas where the effective exposure is below 50% of the maximum exposure, due to vignetting effects of the mirrors and detector efficiency. The scales of the wavelets were optimised to detect point sources in the relevant region of each image. The esplinemap tool used the wavdetect output to remove sources above a likelihood of 5σ from the image and then created a background map, fitted to the re-binned image with 16 spline nodes.

With these background maps as input we ran wavdetect again on every image, with the same parameters as in the first run, to create a final source list for each image.

We merged the output source lists of the five images. The lists cover regions from radius 0−95′′, 90−200′′, 190−400′′, 380−700′′, and 665−1315′′. Sources that were found in two images in an overlap-region of two lists were taken from the higher resolution image.

Based on the wavdetect count rates, we calculated the 0.2−10 keV flux, corrected for Galactic foreground absorption (NH = 6.6 × 1020  cm-2, Stark et al. 1992) towards M 31, and used a power-law spectrum with a photon index of Γ = 1.7. For the luminosity calculation, we assumed a distance of 780 kpc to M 31 (Stanek & Garnavich 1998) and isotropic radiation, which has been used in many previous X-ray studies, although there are more recent papers on the distance to M 31 (e.g. Riess et al. 2012, 752 ± 27  kpc).

To improve the astrometry of individual observations we compared the source-list positions with the Kaaret (2002) catalogue, which is based on the longest individual observation, and then corrected systematic offsets. The Kaaret (2002) catalogue has been astrometrically corrected after cross-correlation with the Two Micron All-Sky Survey catalogue (2MASS, see Skrutskie et al. 2006). After this correction, the mean angular distance to the Kaaret (2002) sources was 0.23′′. Since Kaaret (2002) states that systematic uncertainties are about 0.15′′, we estimated a conservative systematic 1σ uncertainty of 0.38′′ for the catalogue.

2.2. Merged image

We merged the individual observations to find possible faint sources that have not been detected in the individual observations.

We corrected for the systematic offset against the Kaaret (2002) catalogue by re-projecting the events for each observation. We again produced images at the five resolutions (see Sect. 2.1), but aligned to the coordinates of the centre of M 31 instead of the centre of the detector. Thereby, we ensured that a given pixel coordinate in one image corresponds to the same RA/Dec and pixel coordinate in the images of each observation. Thus, we were able to add up the images pixel by pixel, instead of merging the event lists, which would have become too large to handle for some of the programs and would have used up too much computational resources.

The images and exposure maps of the individual observations were merged. We created new background images, accepting only merged areas or which the effective exposure is above 70% of the maximum. This was necessary because the corners of the detector fields do not overlap farther off-axis in the merged image as the roll angle of the satellite changed during the course of the observations, leading to a fast drop in exposure (see Fig. 2).

For these images, we ran the detection pipeline described in Sect. 2.1.

The longest exposure time of the merged image of all 62 observations adds up to 1.02 Ms in the central area (see Figs. 1416).

thumbnail Fig. 2

Colour image of the merged HRC-I exposure maps, showing the entire detector FOV. The dashed black circle marks the outer limit of the catalogue (see Fig. 1). The dashed white circle includes the approximate area with >70% effective exposure. The colour bar shows the effective exposure time value (in seconds). The innermost black region marks the approximate 1 Ms effective exposure region.

Open with DEXTER

2.3. Spectral information

Table 3

Photon count distribution observed with Chandra ACIS in the 0.2 keV to 10 keV photon energy range.

To distinguish between XRBs with a harder spectrum or SSSs with a super-soft X-ray spectrum, we checked the hardness ratios (HR) of correlating X-ray sources from previous catalogues (see Sect. 3.2). In some cases we also checked the hardness information from previous studies on individual sources, in which data from the XMM-Newton European photon-imaging camera (EPIC, Strüder et al. 2001; Turner et al. 2003) detectors, the Swift X-ray telescope (XRT, Gehrels et al. 2004), or the Chandra Advanced CCD Imaging Spectrometer (ACIS, Garmire et al. 2003) were analysed. Where we could not obtain spectral information from the identification with known X-ray sources, we extracted hardness information from one adjacent detection with Chandra ACIS, obtained from the Chandra data archive4, although there were many more observations available in most cases. We used observations with the ACIS-I or ACIS-S CCD array.

We extracted ACIS source counts in the 0.2−1 keV and 1−10 keV band (see Table 3). The counts were extracted from regions with ~95% encircled count fraction PSF size and were not corrected for background contributions. For the ACIS-I observations the background was negligible. In the ACIS-S observation (ObsID 14197, ~37.5 ks exposure) we extracted nine counts in the background. The obtained count ratio given in Table 3 can be used to clearly distinguish hard X-ray spectra from super-soft spectra. In all cases the source counts in the bands were of similar order, indicating a hard spectrum. For an SSS spectrum most of the counts would be expected in the 0.2−1 keV band. The SSS selection criteria have been discussed by Kong et al. (2002b) for Chandra ACIS and by Pietsch et al. (2005b) for XMM-Newton EPIC.

For two outburst sources (No. 116 and No. 300, see Sect. 4.1), which were only observed with Chandra HRC-I, we give a rough hardness estimate based on HRC-I HRs, as has been successfully demonstrated by Pietsch et al. (2007).

The Chandra observatory proposers guide5 suggests this procedure.

3. Analysis and results

3.1. Catalogue compilation

To produce the source catalogue from the detection lists of the individual and merged observations (for sources with a significance above 3σ), we matched positions of sources in each observation using a script implemented in Python6. We consecutively merged the detection lists of the individual and merged observations by adding sources that were not correlating with any source position in the preliminary merged list within a radius of twice the 50% encircled count fraction PSF size. This conservatively exceeds the 3σ positional errors given by wavdetect and in addition accounts for systematic effects, which are caused mainly by the strong degradation of the PSF size with increasing off-axis angle. While building this list, we excluded sources that were only detected in one observation and had a source significance below 5σ. The parameters adopted for each source were derived from the most significant detection with wavdetect. We then removed sources with a maximum source significance below 4σ and screened the list. During the screening, we removed one duplicate of a source, spurious detections around very bright sources, faint and far off-axis sources with bad positions and false sources found at the detector edges with unrealistic wavelet scales. Overall, there were 11 369 individual detections with wavdetect. After the merge and after applying the >3σ cut, there were 384 sources in the catalogue. The subsequent significance cuts and the screening reduced the number of sources from 384 to the final 318.

Because the detection was optimised to find point sources, we also removed known supernova remnants (SNRs included in Sasaki et al. 2012) from the catalogue when the source was split into multiple spurious detections, because of the extent of the SNR. We discuss these separately in Sect. 4.2. The wavdetect tool did not find the correct position for the far off-axis neighbouring sources Nos. 20 and 28 in the catalogue (with the merging parameters), therefore we determined the detection properties from an observation (HRC-I ObsID 5925) where one of the sources was too faint to be detected. The properties of Nos. 18 and 21 were also determined from single detections (HRC-I ObsID 5926 and 5927) since they were too close together to be resolved during the automatic compilation of the catalogue. We also added the positions for the double nucleus of the galaxy (No. 164 and 165 see Fig. 3) from ObsID 10684, where the nucleus is resolved.

The final catalogue includes 318 sources, 20 of which are only detected in the merged images. It is sorted ascending in right ascension (J2000). Throughout we refer to specific sources in the catalogue with their source number (see SRC_ID column in Sect. 3.5) preceded by “No.”.

The systematic positional errors increase the farther off-axis a source is located because the PSF of the Chandra HRMA degrades with increasing off-axis angle. Figures 1416 show the source regions of the HRC-I catalogue plotted over the merged images at different resolutions. The sources are marked by circles with ~90% encircled count fraction PSF radius for 1 keV photons at the source position.

thumbnail Fig. 3

Logarithmically scaled contour image resolving the M 31 nucleus (ObsID 10684). Coordinate system in RA and Dec. Image smoothed with Gaussian function of 2 pixel FWHM and contours are chosen at 0.5, 1.1, and 1.5 counts per pixel to illustrate the separation of the two sources at the core of M 31 (double-core east No. 165 and west No. 164). Nos. 162, 169, 170, and 172 were not active in this observation. The region size is the 90% encircled count fraction PSF.

Open with DEXTER

thumbnail Fig. 4

Example of light-curve plots showing X-ray luminosity (left axis, 0.2−10 keV) and photon count rate (right axis) over modified Julian date. Fluxes with 1σ errors are given as black dots and 3σ upper limits as red triangles.

Open with DEXTER

3.2. Correlations

The catalogue was correlated with the X-ray catalogues, listed below (abbreviation, reference, and assumed 68% confidence errors) to identify previous detections and classifications of sources:

We considered two source positions to be correlated if their angular distance (d) from each other meets the criterion (1)where σ1,2 are the 68% confidence errors of the positions, of detection 1 and 2, respectively. Assuming that the probability distribution of a source position follows a Rayleigh distribution, the above formula (1) results in 99.73% completeness of the correlation (i.e. the probability, that we will detect all real counterparts, see e.g. Watson et al. 2009).

The catalogue was also compared to source catalogues in other wavelengths. In the near-infrared we used the 2MASS catalogue (Skrutskie et al. 2006), while in the optical we used the LGGS catalogue (Massey et al. 2006), for which we visually inspected the images by creating finding charts for all X-ray sources as described in Sect. 3.6. For the comparison with radio sources, we used the Braun (1990) catalogue (hereafter B90), which is based on Very Large Array (VLA) radio observations. The classification criteria are discussed in Sect. 4 and summarized in Table 5. To find out which sources are associated with globular clusters in M 31 we compared the catalogue positions with globular clusters from Galleti et al. (2007) (and checked classifications in the latest online version of the Revised Bologna Catalog RBCv5, August 2012). To identify X-ray sources with novae in M 31, we correlated the catalogue with optical nova positions obtained from the M 31 nova monitoring project7.

3.3. Variability

To investigate the X-ray variability of the point sources, we developed a Linux shell script, calling various CIAO tools, to obtain the source counts and upper limits for each source in each observation. We extracted source counts from the event files (with dmlist), using circular extraction regions. As radii, we used the 90% encircled count fraction PSF radius at 1 keV, calculated with mkpsfmap at the source position. Then we subtracted background counts calculated from the background maps accordingly (with dmstat). Using the CIAO tool aprates, we then calculated the source statistics. We executed two aprates runs with a 68% and 99.7% confidence level for the errors and upper limits (corresponding to 1σ and 3σ confidence for a Gaussian distribution). Luminosities were calculated from the aprates count rate output, as described in Sect. 2.1.

For each source, we calculated the fraction (ndet) of the number of >1σ data points in the light curves over the number of observations, in which the source is in the detector’s FOV. Additionally, we introduced two measures of variability to the source catalogue (see Stiele et al. 2008; Kong et al. 2002b).

One is the variability significance (svar) (2)where F is the flux and the σ are the errors associated with the flux. For the extreme values Fmax and Fmin (fmax and fmin in the primary source catalogue table) we used the flux values, where F − σ is the highest and F + σ is the lowest flux respectively (adopted from Sturm et al. 2013). If a 3σ upper limit lower than Fmin was available, we used it as the minimum flux instead, and in place of its uncertainty σmin in Eq. (2) we used the 1σ upper limit.

The second variability parameter is the variability ratio (fvar), which stands for the ratio of maximum over minimum flux Fmax/Fmin of the source. To calculate these properties we used data points for which the significance of the respective source given by aprates was higher than 1σ.

In addition to the overall variability significance svar between the highest and lowest flux over 13 years, we calculated the maximum of the variability significance (ssvar) between neighbouring observations. The observations have different spacings in time, but the highest variability between them is still a good measure for the variability on time scales from about ten days to one year.

Using the data in Table 2, we plotted the HRC-I light curves for all sources over the last thirteen years (visualised using matplotlib by Hunter 2007). The variability plots show brightness as count rates (label on right ordinate) and luminosity in M 31 (label on left ordinate, assuming the power-law spectrum from Sect. 2.1 and a distance of 780 kpc) with a significance higher than 1σ, as a function of time in modified Julian date (MJD). The data points are filled circles (black) with 1σ error-bars, showing the mean brightness during one observation. Whenever the source was outside the detector FOV of the relevant observation, there was no corresponding data point. When the source significance was below one, the 3σ upper limit of the source flux was marked by a red triangle. The observations span about 12.5 years from 1999-11-30 to 2012-06-01. Sources Nos. 2, 4, 5, 8, 16, 22, 24, 44, 52, 90, 303, 315, and 318 have no detection >1σ in the individual observations therefore we cannot show light curves for them.

For 18 sources in the catalogue the 90% encircled count fraction PSF overlaps with an adjacent source, thus causing some cross-talk (see “XT“ flag in Sect. 3.5). As can be seen in Figs. 316, this overlap is mainly critical for the sources in the crowded core region of M 31.

All light-curve plots are available in the online version of this publication in Fig. B.1.

3.4. Light-curve classes

We derived two light-curve classes using the following selection criteria from the variability parameters defined in Sect. 3.3: highly variable (HV):  ndet >0.3  and  (ssvar >4  or  svar >5), and outburst (OB):  ndet <0.3  and  fvar >15  and  svar >5.

HV sources are detected in >70% of the observations and show a strong variability between neighbouring observations and/or high variability at longer time scales in the HRC-I monitoring (e.g. No. 104 in Fig. 4).

OB sources show a significant intensity increase by at least a factor of 15 and are not detected in at least 70% of the observations when in the FOV. We furthermore divided OBs into three subclasses:

  • Short outbursts (SOB, e.g. No. 12 in Fig. 4), which decay to a quiescent level below the sensitivity limit in less than about 500 days.

  • Long outbursts (LOB, e.g. No. 184 in Fig. 4), which need more than 500 days to decay.

  • Activity periods, which show a strong variability during an outburst in X-rays (AP, e.g. No. 185 in Fig. 4).

Some OBs may have been classified as transient X-ray sources in the literature, because they show one outburst and then disappear again. However, sources in M 31 are only detected if they are brighter than about 1   ×    1036 erg s-1. Therefore we cannot be sure whether a source is not still reasonably bright before and after OBs, since Garcia et al. (1998) found quiescent states of Galactic XRBs to be much lower (between ~1033  erg  s-1 and ~1034  erg  s-1) than the sensitivity limit. Good examples for such behaviour are No. 181 (Fig. 4) or Nos. 152 and 198 in the catalogue, which show outbursts, but are still detected in between. Sources with very long APs can also be classified as HV instead of OB.

thumbnail Fig. 5

Population plot showing the source variability ratio fvar versus the highest photon count rate of a source. Blue diamonds are OBs, green squares are HVs, red circles are GlCs, black dots are novae, and black crosses are the remaining sources.

Open with DEXTER

Figure 5 shows a scatter plot of the variability factor (fvar) versus the highest photon count rate for all sources with variability information in the catalogue (created using TOPCAT8). The plot shows a branching of variable sources at higher luminosities. The OBs by definition have a higher fvar, the brighter they become, the HVs can have a low fvar at high luminosities (see also Fig. 7). Owing to the sensitivity limit of the observations, sources at lower luminosities generally have a lower variability factor.

3.5. Table description

The catalogue consists of two tables, containing the 318 sources, the primary table with general source information (Table 1) and a secondary table with detailed source variability information (Table 2).

The primary table (Table 1) contains the source number (SRC_ID, Col. 1), the RA and Dec coordinates (J2000) in degrees (RA, Dec, Cols. 2, 3), three times the statistical 1σ positional error (from wavdetect, not including systematic errors due to centroid offsets far off-axis9) and the systematic catalogue error from offset correction and reference catalogue, see Sect. 2.1 (error, Col. 4), the net source counts and corresponding error (NET_COUNTS, NET_COUNTS_ERR, Cols. 5, 6), background counts in the source region (BKG_COUNTS, Col. 7), the net count rate and corresponding error (NET_RATE, NET_RATE_ERR, Cols. 8, 9), the background count rate and corresponding error (BKG_RATE, BKG_RATE_ERR, Cols. 10, 11), the exposure time of the most significant detection of the source (EXPTIME, Col. 12) and the highest source significance found in any observation (SRC_SIGNIFICANCE, Col. 13). These parameters are part of the wavdetect output.

In addition, we give the number of the highest resolution image the source is detected in (res, Col. 14, see Sect. 2.1), the energy flux and its error (flux, flux_err, Cols. 15, 16), luminosity, assuming the source lies in M 31 (see Sect. 2.1) and error (lum, lum_err, Cols. 17, 18), and the distance to the nearest neighbouring source (dist_NN, Col. 19).

From the light curves of each source we give the parameters svar, fvar, ndet, ssvar, fmax, fmin, and errors (Cols. 20 to 27) and the derived light curve classification (lc_class, Col. 28), described in Sect. 3.3. When a source is detected only in the merged images or only once in the detector FOV, we cannot give any light curve information and the variability values are set to 0.0.

We list the X-ray correlations of the sources with other X-ray catalogues, described in 3.2 (XID, Col. 29). Due to the high X-ray source density in the central area of M 31, there can be confusion of X-ray sources, especially in the case of catalogues obtained with lower-resolution instruments, since they do not resolve all sources. The column contains a string of correlations that provide the source No. in the catalogue, followed by the name of the reference catalogue, the source number or name, and if available, the type of the source, correlating with the catalogue source (e.g. [132, “PFH2005”, “310 XRB;”]).

Table 4

Prominent optical identifications up to about 1′′ separation from one of the X-ray sources.

Table 5

Source classes and criteria in the catalogue.

Names of optical novae and globular clusters associated with an X-ray source are listed separately (nova, glc, Cols. 30, 31). For the globular clusters we also list the quality flag of the RBCv5 (1: confirmed, 3: controversial object, 6: confirmed star, see Galleti et al. 2007), stating whether the globular cluster has been confirmed. Source No. 85 in the catalogue correlates with a former GlC candidate and has since been confirmed to be a foreground star. Source No. 49 correlates with a controversial object from the RBCv5, and we classified it as a GlC candidate.

We then give the source type (class, Col. 32), based on the identifications in the X-ray, radio, and optical wavelengths, as described in Sect. 3.2 and the criteria listed in Table 5. Candidates are designated by ⟨  ⟩.

thumbnail Fig. 6

Light curve plot of the GlC X-ray source No. 72, showing X-ray luminosity (left axis, 0.2−10 keV) and photon count rate (right axis) over modified Julian date.

Open with DEXTER

Where necessary, we comment (comment, Col. 33) on special properties of a source. Comments refer to special light curve properties of a source (see Sect. 3.4) or note the lack of a light curve (no lc) for some sources that are only found in the merged images or are too far off the optical axis (we did not include data points where the extraction region overlaps with the detector edge, see Sect. 3.3). For some sources we added a bad position flag (bad position) in the comments, since we found that the CIAO tool wavdetect did not find the centroid position for some faint and far off-axis sources. Where the identification with known X-ray sources or counterparts in radio or optical data was unclear because of the bad position, we provided no classification for the sources based on these identifications (in Col. 32). If the 90% encircled count fraction extraction region of a source overlaps with another extraction region, this is highlighted by “XT“ followed by the overlapping source No. (e.g. XT[164]). For each statement in the “comments” column that is not derived from our analysis, we give the abbreviated relevant reference.

The time variability table (Table 2) contains the observation numbers (ObsID, Col. 1), start date of the observation (date, Col. 2), the modified Julian date at the start of the observation (MJD, Col. 3), and three columns for each source, giving its flux and upper and lower 68% confidence limits (Cols. 4 to 957). When a source is not detected >1σ, we give a 3σ upper limit in the upper confidence limit column and zero for flux and lower confidence limit. When a source is out of the FOV, the three flux columns are zero.

Sources that are not included in previous dedicated M 31 catalogues are listed in Table 10. For the new sources as for all other sources in the catalogue, we give a light curve classification (see Sect. 3.4) and hardness information where possible. For some faint sources we can only give the positions since we do not have any significant light curve information.

3.6. Optical counterparts

The search for optical counterparts is important to further constrain the nature of an X-ray source. We checked finding charts derived from broad- and narrow-band images provided by the Local Group Galaxy Survey project (LGGS, see Massey et al. 2006). Where we found a prominent and not yet clearly identified optical source within the 3σ error circle and up to a separation of about 1′′ of the catalogue source, we list the source name from the Massey et al. (2006) catalogue, with visual magnitude and optical colours in Table 4. We regard an optical counterpart as prominent if we were able to visually identify it as the brightest and at the same time closest counterpart in the LGGS images. Table 4 lists the closest LGGS counterpart to the corresponding X-ray position. Expected counterparts in this list could be galactic foreground stars, background galaxies or high-mass stars (as companion stars in high-mass X-ray binaries in M 31, see 4.1.5). We did not include in this list confirmed objects from earlier studies such as galaxies, foreground stars, globular clusters, or supernova remnants.

4. Discussion

In the central field of M 31 there are many X-ray sources that have previously been classified or identified into different classes according to their X-ray spectra or at least HRs, their time variability, and optical and radio correlations (see X-ray catalogues in 3.1). Since HRC-I has only limited spectral resolution, we had to rely on previous catalogues or neighbouring Chandra ACIS, Swift XRT, or XMM-Newton EPIC observations to estimate the hardness of a source (see Sect. 2.3).

New in this catalogue, however, are the good source positions in the inner area of the field made possible through the high resolution of the HRC-I detector. The long integrated exposure time improves the detection of faint sources. Table 5 gives an overview of the criteria for different classifications in the catalogue, which are discussed in more detail throughout this section. Table 5 also lists the statistics for the number of sources clearly identified in the catalogue and the number classified using the selection criteria.

For most sources we were able to analyse the time variability in detail. This enabled us to give a classification based on light curves, which has not been possible to this extent in previous catalogues. However, there are still sources without classification in the catalogue, because we have no significant light curve information on them and thus cannot obtain a light curve class (see Sect. 3.3).

In general, the variability we infer from the light curves can be caused by intensity or spectral changes, resulting from anisotropic radiation of a rotating object, full or partial occultation of the X-ray source, or a change in energy output of the emitting source.

In the catalogue we classify 195 sources as located in M 31 (61.3%) and 13 foreground and background sources (4.1%). One hundred and ten sources remain unclassified (34.6%, see Table 5). Details of the classifications are given below. We regard a source as identified when the source has been clearly identified in previous studies and this identification is compatible with the HRC-I analysis. Otherwise we regard a source as classified when it has only been a candidate in previous studies or we obtain the classification only from the variability analysis (candidates indicated by ⟨  ⟩ in the catalogue).

4.1. X-ray binaries

According to theoretical models, X-ray binaries (XRB) are thought to consist of a compact object (black hole, neutron star, or white dwarf) and a donor star of various types from which matter is accreted onto the compact object.

From the phenomenological point of view, we divide X-ray binaries into globular cluster (GlC) sources (optically identified), XRBs in the field with a hard X-ray spectrum (probably a black hole or a neutron star as the compact object), and XRBs with a super-soft (SSS, nova) X-ray spectrum (probably a white dwarf as the compact object).

While for hard sources a significant part of the photons are detected at energies >1 keV, for SSS sources the majority of photons are detected below 1 keV.

4.1.1. Globular cluster sources

We found 32 of the X-ray sources in the catalogue to be associated with confirmed M 31 GlCs (excluding novae). From the variability criteria (see Sect. 3.4), we classified 16 of them as highly variable (HV, e.g. No. 72, shown in Fig. 6) and 2 as OB. Fourteen GlC sources have no variability class, because they are too faint in X-rays to detect significant variability. In addition, there are two GlC sources associated with optical novae (Nos. 81 and 150 in Sect. 4.1.6). The errors of both the X-ray positions and the globular cluster catalogue (Sect. 3.1) are small and the density of globular clusters in the central region of M 31 is too low to lead to a significant contribution of chance coincidences. This allows us to state with high confidence that the X-ray sources we found to be correlated with GlCs are actually inside an M 31 globular cluster.

Clark (1975) and Fabian et al. (1975) found that bright X-ray sources in GlCs in the Milky Way are LMXBs and that the density of LMXBs per stellar mass in GlCs is higher than in the field. They explained this fact by two different dynamical mechanisms of binary formation (three-body scattering and tidal capture). Based on these findings, we assume that all the GlC X-ray sources are LMXB candidates. Since all GlC sources with significant detections in their light curves show high variability between neighbouring observations and are thus classified as HV or OB sources, we classified all sources in the field showing the variability class HV or OB as XRB candidates. As shown in Table 5 and in the following sections, other source types can either be excluded due to time variability behaviour, spectral information, or identifications at other wavelengths.

The classification scheme is additionally supported by the fact that similar time variability in the 0.2 to 10 keV X-ray energies has been observed in Galactic X-ray sources that have been confirmed to be XRBs (e.g. Homan et al. 2010). The variability in the HRC-I light curves might be enhanced by spectral changes of a source.

As mentioned above, the GlC sources that have not been classified by the light curve variability are most probably also LMXBs. Therefore, several LMXBs will not be classified by the light curve criterion in the field either.

There are no other source types besides the SSS emission of optical novae in M 31 that correlate with the OB and HV sources and their distribution becomes more dense towards the centre of M 31, following the density of stellar populations in the galaxy. Therefore we are very confident, that the OB and HV sources are indeed XRB candidates located in M 31, because we would expect background or foreground objects to be randomly distributed across the FOV.

4.1.2. XRB variability

Figures 7 and 8 show the spread of X-ray luminosity LX we observe for HV and OB XRB (HVXB, OBXB) sources in the field of M 31. The OBXBs all show a high variability factor (as seen in Fig. 5). Several sources are detected in a very deep observation at a flux below the most stringent upper limits. Most HVXBs do not fall below the sensitivity limit during the observed period of thirteen years, but show a more confined LX range. Some of them have a particularly low LX spread, which could hint at an underlying source class within the HVXBs.

thumbnail Fig. 7

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all HV XRB sources (HVXB) in the field of M 31. The x-axis is labelled with the catalogue source number (No.). When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow.

Open with DEXTER

thumbnail Fig. 8

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all OB XRB sources (OBXB) in the field of M 31. The x-axis is labelled with the catalogue source number (No.). When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow.

Open with DEXTER

Phenomenological discussion of some XRB candidates, with hard X-ray spectrum: No. 3 shows an AP, starting in 2004 (see Trudolyubov et al. 2006). It was first detected by Galache et al. (2007a).

No. 47 shows an SOB starting on 2009-11-07 followed by an AP lasting until the end of the HRC-I monitoring in 2012. In an ACIS-I observation (ObsID 11275 from 2009-11-11), the source shows a hard spectrum (see Table 3).

No. 49 is an X-ray burster (see Pietsch & Haberl 2005) showing a highly variable light curve. The source is associated with a GlC candidate (see RBCv5 in Sect. 3.2).

No. 54 shows two high-luminosity states, each lasting for several years. During the low state (luminosity around 5   ×    1037 erg s-1), the variability is low and during the high state (luminosity around 2   ×    1038 erg s-1), the variability is high. From 2008 until 2010-02-15 (ObsID 9825) the source steadily brightened until it reached the high state. During this high state, the source shows variability similar to AP sources. This system has been discussed as being a black-hole X-ray binary, showing spectral state changes in Barnard et al. (2011).

Table 6

Additional Chandra detections for recurrent outburst source No. 246.

Table 7

Additional detections for recurrent outburst source No. 290.

Table 8

Additional observations for recurrent outburst source No. 313.

No. 57 shows two APs in 2004 (Stiele et al. 2008) and 2007 (Henze et al. 2008a). The source is not bright enough to classify it as OB (see criteria in Sect. 3.4). We still classify the source as an XRB candidate, based on the literature.

No. 58 is not confirmed as an fgStar candidate as in Stiele et al. (2011), because there is no star visible in the Massey et al. (2006) images and we find a highly variable light curve. Therefore we classify the source as an XRB candidate.

No. 65 shows an SOB, lasting a few months and starting on 2008-09-01 (in ACIS-I ObsID 9523) with a hard X-ray spectrum (see Table 3). In an observation on 2008-07-15 (ObsID 9522) it was not detected. It is also detected with ACIS-I on 2008-10-13 (ObsID 9524) and not detected after 2008-11-27 (ObsID 9521), which constrains the duration of the outburst.

No. 67 shows an AP starting 2009-05-29 (see Galache et al. 2009; Kong et al. 2009). The source was active untill 2011-03-10 in HRC (ObsID 13180).

No. 82 shows an SOB in the globular cluster B112, starting on 2012-06-01 (Hofmann et al. 2012). In following monitoring observations with the Swift XRT, the source showed a rapid decay within a few weeks. It is one out of only two OB sources we found in globular clusters (see Sect. 4.1.2).

No. 84 is a very faint source in the catalogue. Williams et al. (2005c) have detected the source in an outburst in their ACIS-I monitoring programme and discussed it as a recurrent transient, since it was already detected by the ROSAT HRI instrument (see Primini et al. 1993).

No. 104 is the most luminous HV source in the catalogue. It has been discussed as a Z-track XRB by Barnard et al. (2003).

No. 116 shows an SOB that is only detected with HRC-I, but not discovered with ACIS. We extracted 34 counts in the soft HRC-I band (PI channel 50 to 130), 69 counts in the medium band (131 to 200), and 25 counts in the hard band (201 to 280). This count distribution hints at a hard X-ray spectrum (see Pietsch et al. 2007).

No. 127 shows an SOB in an HRC-I observation starting on 2006-11-13. The ACIS-I ObsID 07064 (2006-12-04) indicates a hard X-ray spectrum (see Table 3).

No. 128 shows an AP lasting from 2006-12-04 (ACIS-I ObsID 7064) until 2010-02-04 (ACIS-I ObsID 11278) with a hard X-ray spectrum (see Table 3).

No. 132 shows an LOB that we cannot clearly constrain, since the source is already bright in the first HRC-I observation. It was discovered by Murray et al. (1999) in October 1999. Trudolyubov et al. (2001) argued that the source shows an outburst of at least 400 days, with a plateau phase of about one year, which would indicate a black-hole binary.

No. 149 is the second ultra-luminous X-ray (ULX) source detected in M 31 (see Henze et al. 2012b,a). It was also observed with Swift XRT (see Esposito et al. 2013) and in radio observations by Middleton et al. (2013).

No. 153 and 158 are clearly resolved as two sources in the catalogue (see also Kong et al. 2002b). In Barnard et al. (2012a) the source is treated as one source. We give two light curves that show some cross talk.

No. 161 shows an SOB and is first detected in ACIS-I ObsID 12973, starting on 2011-08-25. In the following ACIS-I observation (ObsID 12974) from 2011-09-28 the source can be classified as hard (see Table 3).

No. 181 has been reported as an X-ray transient by Shirey (2001). In the HRC-I monitoring light curve, with lower sensitivity limit, we detect the source as HV.

No. 185 shows an AP from 2008-05-31 (ACIS-I ObsID 9529) till 2010-03-05 (ACIS-I ObsID 11279). In HRC-I observations we detect the source as HV. The source has a hard spectrum, derived from XMM-Newton observations (Pietsch et al. 2008b).

No. 201 shows a long-term variability, with two LOBs at the beginning and end of HRC-I observations. After the second outburst, the source stays in a high-luminosity state, which slowly declines over the course of three years and is still at 3   ×    1037 erg s-1 at the end of the HRC-I monitoring.

No. 222 shows an SOB in just one of the HRC-I observations (ObsID 7284 on 2006-09-30). The ACIS-I ObsID 7064 on 2006-12-04 indicates a hard spectrum (see Table 3).

No. 234 is the first ULX source found in M 31 (see Henze et al. 2009a; Kaur et al. 2012; Middleton et al. 2012; Barnard et al. 2012a; Nooraee et al. 2012).

No. 237 is an HV source. In Nov. 2010 it dropped in luminosity from ~1038  erg  s-1 and after that showed high variability. This indicates a spectral-state change.

No. 240 shows a very slow variability. It is not classified as an HV because the count rate has large errors when the source is brightest.

No. 247 has an SOB starting 2010-10-19 (Barnard et al. 2012a) in ACIS-I ObsID 12160, where we find a hard spectrum (see Table 3).

No. 259 shows a faint AP from 2006-06-05 until 2010-12-12, which is not significant enough for the outburst criteria. The ACIS-I observation ObsID 7064 on 2006-12-04 indicates a hard spectrum (see Table 3).

No. 275 is an X-ray burster in a globular cluster system, which is discussed in Pietsch & Haberl (2005).

No. 276 shows dips with a 107 min X-ray period (Mangano et al. 2004) and was thus classified as an X-ray binary. In the HRC-I light curve the source does not show high enough variability to be classified as HV.

No. 278 is at the position of a confirmed globular cluster at the edge of an SNR in the optical images (see Massey et al. 2006), which might cause some contamination of the light curve of the GlC source. However, the SNR has to be very faint in X-rays, since we do not see an extended source at the position in the merged X-ray images.

No. 279 shows an AP starting 2004-05-23, discussed in Williams et al. (2005b). The last detection in the HRC-I light curve is on 2005-02-21.

No. 285 shows a very peculiar light curve with two very well separated luminosity states and has previously been found to have a 2.78 h period in Trudolyubov et al. (2002). Barnard et al. (2006) have argued, that there is another component to the variability, caused by precession of the accretion disc of the compact object.

No. 296 shows an SOB starting on 2008-08-21, discovered with Swift XRT (Pietsch et al. 2008a; Galache et al. 2008c; Henze et al. 2008b; Galache et al. 2008a).

No. 299 is an HV source. One of its brightness peaks has been reported in Nooraee et al. (2009).

No. 300 shows one SOB and is only detected with HRC-I. To estimate the hardness of the spectrum, we extracted 629 counts in the soft HRC-I band (PI channel 50 to 130), 1440 counts in the medium band (131 to 200), and 479 counts in the hard band (201 to 280). This distribution indicates a hard X-ray spectrum (Pietsch et al. 2007). Recurrent outburst hard X-ray sources: No. 15 shows three APs in the light curve, the first of which is discussed in Osborne et al. (2001) and Williams et al. (2006). The source is discussed as a recurrent transient in Stiele et al. (2011), since the source was already observed with the ROSAT observatory (see Primini et al. 1993). We detected the latest AP, with the source still active at the end of the HRC-I monitoring, on 2012-06-01.

No. 40 shows many recurrent SOBs about every 1.2 years (see Henze et al. 2007; Pietsch et al. 2010a; Galache et al. 2007b).

No. 111 shows recurrent SOBs. One is discussed in Garcia et al. (2001). The source was not active in ACIS-I 9521 (2008-11-27) and shows one outburst in HRC-I ObsID 10684 (2008-12-18). We detected a second SOB starting in ACIS-I ObsID 12164 (2011-02-16), which we also observed one day later in HRC-I ObsID 13178. If there is a quasi-periodic recurrence time, the HRC-I data would suggest an outburst period of ~2 years.

No. 140 shows recurrent SOBs and a hard spectrum in ACIS-I (Kong et al. 2002b) and XMM-Newton EPIC observations, separated by ~6 years.

No. 147 shows either short recurrent OBs about every year, which are missed in some of the HRC-I monitoring observations, or it shows two APs.

No. 162 shows an SOB in HRC-I ObsID 7283 (2006-06-05). ACIS-I observation (ObsID 7138 from 2006-06-09) closely after the HRC-I detection (ObsID 7283 from 2006-06-05) indicates a hard spectrum (see Table 3). The earliest detection of the outburst was visible in an ACIS-I observation from 2006-05-26 with a total of nine counts. In previous observations there was an outburst only seen with ACIS-I in an observation starting on 1999-10-13 (ObsID 303), from which we extracted 26 counts. Because of this outburst, the source is part of the Kong et al. (2002b) catalogue and thus is classified as a source with recurrent outbursts.

No. 189 shows recurrent SOBs in the HRC-I light curve. The source brightens from 2009-11-07 till 2010-02-26. The ACIS-I observation ObsID 11279 on 2010-03-05 indicates a hard spectrum (see Table 3). The source is active till the ACIS-I ObsID 11839 from 2010-06-23.

No. 246 shows multiple SOBs of which we caught three in a very bright state in the HRC-I observations. There are many additional observations with the Chandra ACIS detector (see Table 6), filling in some of the HRC-I observation gaps. The source seems to have bright outbursts about every year, lasting up to about three months.

No. 290 shows OBs lasting up to several months about every year during the time span covered by HRC-I observations. Table 7 lists additional observations from other instruments, where the source is visible in the X-ray images, which fill in some of the time covered by HRC-I light curves. The XMM-Newton EPIC HR of the source indicates a hard X-ray spectrum (see 2XMM catalogue by Watson et al. 2009).

No. 313 shows recurrent OBs first discovered with the Einstein and ROSAT X-ray observatories. In XMM-Newton EPIC observations it showed a harder X-ray spectrum (see 2XMM catalogue by Watson et al. 2009). In the HRC-I light curves we detected four OBs. It shows both LOBs and SOBs, with the SOBs about nine months apart. Table 8 gives additional observations in which the source is detected.

4.1.3. GlC population

We found that after applying the variability criteria for XRB selection, the ratio of hard XRBs classified as HV (HVXB) to hard XRBs classified as OB (OBXB) is 2.3/1 in the field of M 31. For the hard XRBs found in GlCs we obtained a ratio of 7.5/1 (HVXB/OBXB). For this statistical analysis we only used sources inside the ~70% effective exposure radius (see Fig. 2) to avoid effects by the inhomogeneous coverage outside this circle. Fig. 9 visualises that most GlC sources are HVXB sources with a very low variability factor. This can also be seen in Fig. 5, where the GlC sources concentrate in the HV source branch. Because we have fewer sources (and thus higher statistical uncertainty) in the GlCs, we estimated the probability that this difference could occur by chance. We used binomial statistics and assumed that the true ratio is 2.3/1 (from the field sources, where we have a sample of 97 sources, 29 OBXB, and 68 HVXB). The probability to obtain a ratio of 7.5/1 after 17 trials (2 OB and 15 HV sources in GlCs) is only 8%. This may indicate that the abundance fraction of outburst sources in GlCs might be lower than in the field. The reason for this possible difference could be differing formation processes of LMXBs in the GlC and field environments. However, there is also the possibility of additional OBs in GlCs, masked by interfering HV sources, which could not be resolved if they were located in the same GlC. A higher OB fraction in the field might also be caused by high-mass X-ray binaries (HMXBs, e.g. Be-XRBs with Type II outburst, see Reig 2011) in the sample, which were not optically identified. It has been discussed by Clark (1975), Fabian et al. (1975), and recently by Voss & Gilfanov (2007b) that XRBs in GlCs are likely to form through tidal capture in the dense star population of a globular cluster, which they also used to explain an over-abundance of XRBs in GlCs.

Overall, the ~70% effective exposure radius of the catalogue contains 154 confirmed GlCs from the RBCv5 (see Sect. 3.2). These are associated with 27 X-ray sources, suggesting that at least 17% of GlCs in M 31 host an XRB. This is a very conservative lower limit, since the observations of M 31 are not deep enough to detect all XRB sources.

4.1.4. Low-mass X-ray binaries (LMXB)

Following the variability analysis, the majority of X-ray sources in M 31 seem to be LMXBs (see Sect. 4.1.2). When one of those variable sources showed a hard X-ray spectrum, we classified it as an XRB candidate (⟨ XRB ⟩). The HRC-I observations mainly cover the bulge of M 31, which hosts a predominantly old stellar population. Therefore, we would only expect very few HMXB candidates in the catalogue (for an estimate see Voss & Gilfanov 2007a).

thumbnail Fig. 9

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all OB or HV GlC sources. The x-axis shows the source No. in the catalogue. When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow. The blue diamonds mark the sources that are classified as OB sources in the catalogue.

Open with DEXTER

Whether the compact object is a neutron star or a black hole can to some extent be decided from X-ray spectra or some variability properties (soft X-ray sources with transient outbursts are most likely black hole systems, see Tanaka 2000). The variability of XRBs in the light curves could in part be due to the transition from one spectral state into another, where hardness ratios in the 0.2-10 keV X-ray band would change. Such state changes in X-ray binaries have been discussed in Hasinger & van der Klis (1989). For similar results from the RXTE satellite see Homan et al. (2010).

The Eddington luminosity (proportional to the mass of the compact object in XRBs) describes the limit at which the outward photon pressure becomes so high that no additional material can be gravitationally accreted. Hence we can derive a rough estimate for the mass of the compact object from the luminosities at which different spectral states of a source exist (see discussion of No. 54 in Barnard et al. 2011). If the mass is between about 1.4 solar masses (M) and 3.0 M, we would expect a neutron star. If the mass exceeds 3.0 M, the compact object is probably a black hole (see Tanaka 2000).

Our sample contains four sources exceeding the Eddington limit for a 3.0 M object. Two of these are the ULX sources that have already been discussed as BH XRBs in previous studies (e.g. Kaur et al. 2012; Middleton et al. 2013). From HRC-I data alone a clear distinction between BH and NS systems was not possible since there was not sufficient simultaneous spectral information.

Short bursts (at time scales of ~one minute) or pulsations in the short-term light curves (during an observation) of a source are strong evidence for a neutron star binary. Bursts (as have been found for Nos. 49 and 275 by Pietsch & Haberl 2005) are thought to arise from short nuclear burning flashes on the surface of a neutron star. For a recent review on neutron star binaries see Bhattacharyya (2010).

Garcia et al. (1998) have discussed the separation of some galactic neutron star and black hole binaries using the variability range of OBs. The sensitivity limit of the HRC-I observations for sources in M 31 does not allow one to measure high enough ratios between the highest and lowest X-ray luminosity to apply the criteria proposed by Garcia et al. (1998) to the sample of sources.

4.1.5. High-mass X-ray binaries (HMXB)

HMXBs are expected in star-forming regions, which are mainly located in the disk of M 31. So far, there are no confirmed HMXBs in M 31. We would classify a source as an HMXB candidate when it showed XRB-like time variability in the light curves (see criteria in Sect. 4.1.2), had a hard X-ray spectrum and showed an optical counterpart with colours consistent with a high-mass star in the Massey et al. (2006) images.

thumbnail Fig. 10

Light-curve plots of LOB X-ray sources associated with optical novae, showing X-ray luminosity (left axis, 0.2−10 keV) and photon count rate (right axis) over modified Julian date.

Open with DEXTER

thumbnail Fig. 11

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all OB or HV SSS or nova sources. The x-axis is labelled with the catalogue source number (No.). When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow. The blue diamonds mark the sources that are classified as OB sources in the catalogue. The black dots mark the sources that are associated to optical novae.

Open with DEXTER

Source No. 301 is bright in X-rays, has a prominent optical counterpart (LGGS and Hubble Space Telescope (HST) images), and shows no radio emission in the reference catalogues. The colours and magnitude of the optical counterpart (see Table 4) and the hardness ratios of the X-ray emission would be consistent with an HMXB in M 31 (see Sturm et al. 2013, for a comparison with HMXBs in the Small Magellanic Cloud). However, the errors of the X-ray hardness and the optical colours are too large to exclude an AGN as the source of the X-rays. We detected no Hα emission at the source position and have no spectrum of the optical counterpart. The HRC-I X-ray light curve is too faint to show significant variability.

thumbnail Fig. 12

Contour plots of the logarithmically scaled merged HRC-I images with dashed circular regions at the radio positions with approximate radio radii (Fesen et al. 1999; Sjouwerman & Dickel 2001) of the known supernova remnants within 90′′ of the center of M 31. The images are smoothed with a Gaussian function of 7-pixel FWHM and contour levels are at 0.53, 0.56, 0.6, and 0.7 counts per pixel. North is up and east to the left. The zoom is slightly different (for region size and SNR names see Table 9). The X-ray catalogue source regions are circles with ~90% encircled count fraction radius.

Open with DEXTER

4.1.6. Super-soft sources and novae

The class of SSS in M 31 is thought to be associated with cataclysmic variables (CV), which are binary systems with a white dwarf as the accreting object. These systems are also responsible for nova explosions (see Pietsch et al. 2007). Most of the sources with super-soft hardness ratios (photon counts mostly below 1 keV) are associated with nova explosions, which are discussed in detail in Pietsch et al. (2005a), Pietsch et al. (2007), Henze et al. (2010, 2011). A nova explosion is thought to be caused by a thermonuclear runaway in the hydrogen layer, which is accumulated onto the surface of the white dwarf via accretion from the outer envelope of a companion star. Mainly depending on the mass of the white dwarf and the accretion rate of the system, there can also be a state during which the accumulated matter on the white dwarf surface undergoes steady hydrogen burning without a nova explosion (see Nomoto et al. 2007). We detected six SSSs in the catalogue that are not associated with known optical novae (Nos. 64, 184, 229, 294, 295, and 314).

The catalogue contains 45 X-ray sources associated with optical novae, 12 of which have not previously been published and will be discussed in detail by Henze et al. (in prep.). One type of novae shows very short outbursts in the light curves (12 classified as OB sources), while others show a very steady X-ray outburst over several years. Because of these long outbursts, sources Nos. 79, 100, and 123 are classified as HV sources because they show the longest outbursts we observe for novae (e.g. Figs. 10 and 11). No. 123 even shows a re-brightening in the 2011/2012 monitoring observations (see Fig. 10).

Figure 11 gives an overview of SSSs and novae classified as significantly time variable in the catalogue. It can be seen that most novae are classified as OB sources. There are three OB SSS without optical nova association (Nos. 184, 295, and 64), which might be X-rays from a nova missed in optical observations. Nos. 294 and 229 are HV SSSs and show a particularly low variability factor, which, as described above, could be caused by steady hydrogen burning on the surface of a white dwarf. Discussion of individual SSS: No. 64 is an SSS showing an SOB that is not associated with a reported optical nova (see Henze et al. 2010). However, the nova outburst might have been missed in optical observations.

Table 9

Source statistics for SNRs within 90′′ of the centre of M 31, extracted from the 1.02 Ms merged high resolution image.

Nos. 81 and 150 are classical novae in an M 31 globular cluster (see Henze et al. 2009b, 2013).

No. 184 is not detected in a 29.07 ks ACIS-I observation (ObsID 7064 on 2006-12-04), where it should be bright according to the HRC-I light curve (see No. 184 in Fig. 4), which shows a plateau during this time. This indicates a super-soft X-ray spectrum, since the Chandra ACIS detector has a lower sensitivity than HRC-I at lower energies (see Pietsch et al. 2007). For a 50 eV black-body spectrum, the conversion factor from HRC-I to ACIS-I count rates would be 0.081 according to WebPIMMS. Accordingly converting the HRC-I (ObsID 7285) count rate of 1.95 × 10-3   counts  s-1, we would still expect 4.5 photon counts during the ACIS-I observation. This means that either the spectrum is even softer than we estimated above, or the source was unexpectedly fainter during the time of the ACIS-I observation and then re-brightened. The HRC-I light curve is typical for the SSS phase of an optical nova several months after outburst, but no known optical nova correlates with its position.

No. 205 is detected at the position of nova M31N 2002-08b (for the optical light curve see Lee et al. 2012). In two Hα images from the Massey et al. (2006) survey on 2002-09-12, the bright Hα emission is visible at the source position, which is most likely associated with the nova. This is the first detection in X-rays from M31N 2002-08b (see Table 10).

No. 229 is a super-soft source candidate, showing a 217 s pulse period (see Trudolyubov & Priedhorsky 2008).

No. 294 is an HV and classified as an SSS in Pietsch et al. (2005b), but does not show a nova outburst and is detected throughout the thirteen years of monitoring. It also has an optical counterpart in an Hα image of the LGGS (see Table 4).

thumbnail Fig. 13

Light curve of No. 99 during HRC-I observation 13178, showing an X-ray flare at the end of the observation. The plot displays counts over time in seconds from the start of the observation, with a binning of 100 s.

Open with DEXTER

No. 295 is a super-soft source that could be the counterpart of a nova explosion missed in the optical. It has a 865 s period that is discussed in Osborne et al. (2001). Its spectrum is similar to galactic super-soft sources (see Trudolyubov et al. 2001).

No. 314 shows a faint SOB (see Haberl et al. 2007). The source re-brightened in a Swift follow-up observation and has a super-soft spectrum according to Kong & Di Stefano (2007). This may be emission from a missed optical nova.

Table 10

New X-ray sources not previously published in dedicated M 31 catalogues (Sect. 3.2).

thumbnail Fig. 14

HRC-I merged high-resolution image (0.13′′ pixels) with source positions and source No. from the catalogue. The region size is the 90% encircled count fraction PSF.

Open with DEXTER

thumbnail Fig. 15

HRC-I merged medium-resolution image (0.52′′ pixels) with source positions and source No. from the catalogue. The region size is the 90% encircled count fraction PSF.

Open with DEXTER

thumbnail Fig. 16

HRC-I merged low-resolution image (2.08′′ pixels) with source positions and source No. from the catalogue. The image shows the overlap of the whole detector at different roll angles. The region size is the 90% encircled count fraction PSF.

Open with DEXTER

4.2. Supernova remnants

After a supernova explosion, a shock front propagates through the interstellar medium (ISM), heating it to temperatures at which thermal X-rays are emitted. There can also be strong radio emission from synchrotron radiation, and optical emission lines are seen where the ISM is ionized. Depending on the age of the SNR, the extent can be large enough to be resolved as an extended source of X-rays, radio, or optical light even at the distance of M 31. The catalogue contains six sources thought to be associated with supernova remnants from the Sasaki et al. (2012) catalogue (some of their sources, overlapping with the FOV, are not contained in the catalogue due to the lower sensitivity of the HRC-I detector (compared with XMM-Newton EPIC) towards the edges of the FOV). Aside from those, we classified one new SNR candidate (No. 272) that shows X-ray extent, radio emission, and hardness ratios (obtained from Stiele et al. 2011) consistent with thermal X-ray emission. No. 304 is a confirmed SNR and is discussed as being associated with a low-mass X-ray binary in Williams et al. (2005a).

By visually inspecting the merged high-resolution image and smoothing with a 7-pixel FWHM Gaussian function, we found extended X-ray sources at the positions of the three radio SNRs discussed in Sjouwerman & Dickel (2001) (see Fig. 12). Of these remnants only SNR 101 has a counterpart in the source catalogue (No. 215). SNRs 101 and 95 have previously been detected in X-rays and are discussed in Kong et al. (2003). However, our deeper image of SNR 95 is much more convincing. For SNR 80 ([B90]80 in Braun 1990) we report the first observation in X-rays (Fig. 12 leftmost image), close to another bright X-ray source. In the 1.02 Ms merged high-resolution image we found 446 source counts in a visually determined region (see Table 9), which corresponds to a significance of about 3.2σ.

In the right-most image of Fig. 12 we see a very dim X-ray emission at the position of SN 1885A (Fesen et al. 1999), which is not included in the catalogue. Source properties were extracted from a circular region with a radius of 0.7′′ (see estimated extent in Sjouwerman & Dickel 2001) from the merged high-resolution image at the position of SN 1885A. We obtained 28 counts and a significance of 2.6σ, which is not sufficient to state a clear detection. To the south-east of SN 1885A lies a transient X-ray source, which we found not to be associated with the supernova in accordance to Kaaret (2002), since the positional errors plus the maximum extent of the source are clearly separated. Table 9 shows the circular regions we estimated for each SNR and the counts and significance we obtained for each of these regions using the CIAO tool aprates to compute source statistics.

4.3. Foreground stars and background galaxies (AGN, Gal, GCl)

There are sources in the catalogue that are not located in M 31. There are X-ray-bright foreground stars in our own galaxy, which we identified by correlating the source positions with known foreground stars (fgStar) in other X-ray catalogues (see Sect. 3.1) and visually inspecting images from the LGGS for clearly visible foreground stars. The advantage of the catalogue for these identifications is that they are more precise for most sources because of the good positional accuracy of Chandra HRC-I and the long exposure available to us. The catalogue contains four confirmed fgStars and seven fgStar candidates, which are statistically distributed across the FOV, as expected. For No. 85, 262, 289, and 291 we give the colours of the optical counterparts in Table 4.

The luminosities of fgStars are erroneously overestimated in the catalogue because they are much closer to us than M 31, and we used the distance to M 31 to calculate all luminosities.

AGN in the background should show no X-ray extent in the images, have a hard spectrum, and show moderate variability (see Barnard et al. 2012b). They are super-massive black holes at the centre of a galaxy, accreting matter. The real X-ray luminosity is underestimated in the catalogue because these sources are much farther away from us than M 31. AGN can also show strong radio emission. The radio emission in AGN is expected to arise from synchrotron radiation connected to acceleration processes in the jet of the super-massive black hole (see Cavagnolo et al. 2010). The catalogue only contains one confirmed AGN and one candidate that we adopted from previous X-ray catalogues (Sect. 3.1). They show very faint light curves and therefore we cannot draw any conclusions about variability.

We found no identifications of X-ray sources with normal galaxies (Gal) or galaxy clusters (GCl) in the background or with CVs or other known sources in the foreground of the M 31 central field. Discussion of individual foreground and background objects: Nos. 44 and 45 show radio emission (see B90: Braun 1990) and a harder spectrum (Stiele et al. 2011), which suggests they are AGN candidates.

No. 85 correlates with the position of a confirmed foreground star from the RBCv5 (see Galleti et al. 2007) and the light curve shows one significant outburst, which is caused by a short flare in the light curve of HRC-I ObsID 10882.

No. 99 is an fgStar candidate with time variability due to flaring activity in HRC-I ObsID 13178 (see observation light curve in Fig. 13).

No. 207 shows an optical counterpart (Table 4) and radio emission (source [B90]96). From the two extended optical sources, the radio emission and the soft X-ray spectrum at the position, the source could be a galaxy merger in the background whose X-ray emission is caused by gas of the ISM, heated during the collision.

4.4. M 31 core

For the largely overlapping central sources of M 31 we propose a new distribution of discrete X-ray sources in the catalogue (see Fig. 3).

Nos. 164 and 165 are at the position of the double core of M 31 (see Garcia et al. 2010). They show variability very similar to the XRB light curves and are classified as HV sources in the catalogue. In Hubble Space Telescope images of the M 31 core (M 31*), a double structure, corresponding to the two X-ray source positions of Nos. 164 and 165 is visible as the most luminous region around the dynamical centre of the galaxy. This circum-nuclear environment is discussed as a region of very high star density around a central black hole in Bender et al. (2005) and Li et al. (2009). Li et al. (2011) recently discussed No. 164 as the X-ray source of the central super-massive black hole (M 31*) and provided a detailed Chandra light curve with an outburst in X-rays. The maximum of this outburst has only been observed with Chandra ACIS and is thus not visible in the HRC-I light curve of No. 164.

No. 167 is a very soft X-ray source to the south of the core of M 31 (see Garcia et al. 2000).

Nos. 170 and 172 are close to No. 167 and show cross-talk from this source, so that we cannot classify them separately.

5. Summary and conclusions

We presented the deepest X-ray catalogue of the M 31 central field available to date, which contains 318 discrete X-ray sources with detailed variability information (spanning 13 years). From the HRC-I data analysis, we were able to provide a crude classification of X-ray binary (XRB) candidates based on their long-term time variability. The catalogue contains light curves for 97 XRB candidates in the field of M 31 and 18 XRB candidates in globular clusters (GlC).

This classification of XRBs according to their time variability is the main step forward of the HRC-I analysis compared with previous studies. Many of the XRB candidates have only been classified as ⟨ hard ⟩ X-ray sources in one of the most recent X-ray catalogues of M 31 in Stiele et al. (2011).

Thirty-six percent of all sources in the catalogue are classified as XRB candidates with a hard X-ray spectrum (X-rays from optical novae make up another 14% of the catalogue sources). Nine XRB candidate sources show recurrent outbursts. The analysis may suggest that outburst sources are less frequent in GlCs than in the field of M 31. We detected one SNR in the central region of M 31 that is seen in X-rays for the first time and discovered one previously unknown SNR candidate. We presented some promising optical counterparts to the X-ray sources that may be spectrally identified with future optical follow-up observations.

We observed 61 new sources that had not yet been published in a dedicated M 31 catalogue. Seventeen of them have never been detected before, 12 sources are new X-ray sources associated with optical novae, 19 sources have only been detected in the Chandra source catalogue (see Evans et al. 2010), and 13 sources have only been discussed in separate papers or Astronomer’s Telegrams, but are not included in previous catalogues.

The catalogue contains a large sample of XRBs with variability behaviour and can thus provide important input for a better understanding of XRBs and the X-ray source population of the M 31 central field in general.


Acknowledgments

We thank the anonymous referee for constructive comments that helped to improve the clarity of the paper. This research has made use of data obtained from the Chandra Data Archive and the Chandra Source Catalog, and software provided by the Chandra X-ray Center (CXC) in the application packages CIAO, ChIPS, and Sherpa. Based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA. The XMM-Newton project is supported by the Bundesministerium für Wirtschaft und Technologie/Deutsches Zentrum für Luft- und Raumfahrt (BMWI/DLR FKZ 50 OX 0001) and the Max Planck Society. This research has made use of NASA’s Astrophysics Data System. This research has made use of the VizieR catalogue access tool, CDS, Strasbourg, France. This research has made use of SAOImage DS9, developed by Smithsonian Astrophysical Observatory. This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. Based on observations made with the NASA/ESA Hubble Space Telescope, obtained from the data archive at the Space Telescope Science Institute. STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. The Second Palomar Observatory Sky Survey (POSS-II) was made by the California Institute of Technology with funds from the National Science Foundation, the National Geographic Society, the Sloan Foundation, the Samuel Oschin Foundation, and the Eastman Kodak Corporation. The Oschin Schmidt Telescope is operated by the California Institute of Technology and Palomar Observatory. We acknowledge the use of NASA’s SkyView facility (http://skyview.gsfc.nasa.gov) located at NASA Goddard Space Flight Center. We acknowledge the use of public data from the Swift data archive. This research has made use of data and/or software provided by the High Energy Astrophysics Science Archive Research Center (HEASARC), which is a service of the Astrophysics Science Division at NASA/GSFC and the High Energy Astrophysics Division of the Smithsonian Astrophysical Observatory. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France. M. Henze acknowledges support from the BMWI/DLR grant FKZ 50 OR 1010 and from an ESA fellowship. R. Sturm acknowledges support from the BMWI/DLR grant FKZ 50 OR 0907.

References

Appendix A: Observations

Table A.1

Chandra HRC-I observations used in this study.

Appendix B: HRC-I light curves

thumbnail Fig. B.1

Light curve plots showing X-ray luminosity (left axis, 0.2–10 keV) and photon count rate (right axis) over modified Julian date for all catalogue sources. Fluxes with 1σ errors are given as black dots and 3σ upper limits as red triangles.

Open with DEXTER

All Tables

Table 3

Photon count distribution observed with Chandra ACIS in the 0.2 keV to 10 keV photon energy range.

Table 4

Prominent optical identifications up to about 1′′ separation from one of the X-ray sources.

Table 5

Source classes and criteria in the catalogue.

Table 6

Additional Chandra detections for recurrent outburst source No. 246.

Table 7

Additional detections for recurrent outburst source No. 290.

Table 8

Additional observations for recurrent outburst source No. 313.

Table 9

Source statistics for SNRs within 90′′ of the centre of M 31, extracted from the 1.02 Ms merged high resolution image.

Table 10

New X-ray sources not previously published in dedicated M 31 catalogues (Sect. 3.2).

Table A.1

Chandra HRC-I observations used in this study.

All Figures

thumbnail Fig. 1

Optical image of the Andromeda Galaxy (DSS2 blue) with X-ray source regions from the Chandra HRC-I catalogue ~50% encircled count fraction radius) and a circle marking the outer rim of the detector FOV (image obtained from the NASA Sky View website).

Open with DEXTER
In the text
thumbnail Fig. 2

Colour image of the merged HRC-I exposure maps, showing the entire detector FOV. The dashed black circle marks the outer limit of the catalogue (see Fig. 1). The dashed white circle includes the approximate area with >70% effective exposure. The colour bar shows the effective exposure time value (in seconds). The innermost black region marks the approximate 1 Ms effective exposure region.

Open with DEXTER
In the text
thumbnail Fig. 3

Logarithmically scaled contour image resolving the M 31 nucleus (ObsID 10684). Coordinate system in RA and Dec. Image smoothed with Gaussian function of 2 pixel FWHM and contours are chosen at 0.5, 1.1, and 1.5 counts per pixel to illustrate the separation of the two sources at the core of M 31 (double-core east No. 165 and west No. 164). Nos. 162, 169, 170, and 172 were not active in this observation. The region size is the 90% encircled count fraction PSF.

Open with DEXTER
In the text
thumbnail Fig. 4

Example of light-curve plots showing X-ray luminosity (left axis, 0.2−10 keV) and photon count rate (right axis) over modified Julian date. Fluxes with 1σ errors are given as black dots and 3σ upper limits as red triangles.

Open with DEXTER
In the text
thumbnail Fig. 5

Population plot showing the source variability ratio fvar versus the highest photon count rate of a source. Blue diamonds are OBs, green squares are HVs, red circles are GlCs, black dots are novae, and black crosses are the remaining sources.

Open with DEXTER
In the text
thumbnail Fig. 6

Light curve plot of the GlC X-ray source No. 72, showing X-ray luminosity (left axis, 0.2−10 keV) and photon count rate (right axis) over modified Julian date.

Open with DEXTER
In the text
thumbnail Fig. 7

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all HV XRB sources (HVXB) in the field of M 31. The x-axis is labelled with the catalogue source number (No.). When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow.

Open with DEXTER
In the text
thumbnail Fig. 8

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all OB XRB sources (OBXB) in the field of M 31. The x-axis is labelled with the catalogue source number (No.). When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow.

Open with DEXTER
In the text
thumbnail Fig. 9

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all OB or HV GlC sources. The x-axis shows the source No. in the catalogue. When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow. The blue diamonds mark the sources that are classified as OB sources in the catalogue.

Open with DEXTER
In the text
thumbnail Fig. 10

Light-curve plots of LOB X-ray sources associated with optical novae, showing X-ray luminosity (left axis, 0.2−10 keV) and photon count rate (right axis) over modified Julian date.

Open with DEXTER
In the text
thumbnail Fig. 11

Variability plot showing the spread of luminosity (left y-axis) and count rate (right y-axis) for all OB or HV SSS or nova sources. The x-axis is labelled with the catalogue source number (No.). When the lowest count rate was derived from an upper limit of the light curve, the lower count rate is marked by a downward arrow. The blue diamonds mark the sources that are classified as OB sources in the catalogue. The black dots mark the sources that are associated to optical novae.

Open with DEXTER
In the text
thumbnail Fig. 12

Contour plots of the logarithmically scaled merged HRC-I images with dashed circular regions at the radio positions with approximate radio radii (Fesen et al. 1999; Sjouwerman & Dickel 2001) of the known supernova remnants within 90′′ of the center of M 31. The images are smoothed with a Gaussian function of 7-pixel FWHM and contour levels are at 0.53, 0.56, 0.6, and 0.7 counts per pixel. North is up and east to the left. The zoom is slightly different (for region size and SNR names see Table 9). The X-ray catalogue source regions are circles with ~90% encircled count fraction radius.

Open with DEXTER
In the text
thumbnail Fig. 13

Light curve of No. 99 during HRC-I observation 13178, showing an X-ray flare at the end of the observation. The plot displays counts over time in seconds from the start of the observation, with a binning of 100 s.

Open with DEXTER
In the text
thumbnail Fig. 14

HRC-I merged high-resolution image (0.13′′ pixels) with source positions and source No. from the catalogue. The region size is the 90% encircled count fraction PSF.

Open with DEXTER
In the text
thumbnail Fig. 15

HRC-I merged medium-resolution image (0.52′′ pixels) with source positions and source No. from the catalogue. The region size is the 90% encircled count fraction PSF.

Open with DEXTER
In the text
thumbnail Fig. 16

HRC-I merged low-resolution image (2.08′′ pixels) with source positions and source No. from the catalogue. The image shows the overlap of the whole detector at different roll angles. The region size is the 90% encircled count fraction PSF.

Open with DEXTER
In the text
thumbnail Fig. B.1

Light curve plots showing X-ray luminosity (left axis, 0.2–10 keV) and photon count rate (right axis) over modified Julian date for all catalogue sources. Fluxes with 1σ errors are given as black dots and 3σ upper limits as red triangles.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.