3 Individual X-ray sources

The EPIC images of the central 30$^\prime$ of M 31 (Fig. 1 represents part of EPIC PN image) contain more than 100 discrete X-ray sources as well as unresolved emission near the centre (see Paper I for more details). The detections include sources seen with Einstein , ROSAT , or Chandra  (TF,PFJ,Su97,Su01,G2000) as well as new sources. The transient source discovered by Chandra  (G2000) is clearly seen in both XMM-Newton  observations (Sect. 3.2). In the June observation, there is a bright new transient which was not seen previously (Sect. 3.1). Several other bright sources demonstrated significant variability on half of a year time scale, so that uncertainties of count rate measurements do not overlap for the two observations. We have analyzed 66 sources with limiting count rate 4 counts/ks and found 10 variables. These sources are listed in Table 2 along with their XMM  positions, June and December count rates in MOS1, and any relevant identifications (though only count rates in MOS1 are cited, we selected the sources based on the data from all three XMM-Newton  instruments). The source positions were derived from comparison with Chandra  calibrated images of the central region of M 31. The estimated positional error of 1 $^{\prime\prime}$-3 $^{\prime\prime}$ (in diameter) depends on the distance of the source from the boresight axis, its intensity and nearby sources. We estimate 3 $^{\prime\prime}$ as a conservative error limit for the XMM  positions presented in this paper.

Taking into account the total number of sufficiently bright detected sources, we conclude that at least $\sim$15% of all sources in the central field of M 31 are variable on a time scale of several months. We believe, however, that this value should be considered a lower limit, because the sensitivity of our analysis was a function of the source flux, and the variability of many of the fainter sources would not be detected. Below we discuss in more detail several individual variable sources in M 31.

3.1 X-ray Nova in M 31

We found a new bright X-ray source in EPIC PN and EPIC MOS images from the June observation. The coordinates of the new source are presented in Table 2. The source was neither detected in previous Einstein  and ROSAT  observations nor has been reported from Chandra  observations. During the June observations the flux from the source was $(1.57\pm 0.09)\times 10^{-13}$ erg s^-1 cm^-2 (0.3-10 keV), which corresponds to a luminosity of $\sim$ $1.1\times 10^{37}$ erg s^-1 in the cited energy band assuming a power-law spectrum (see Paper III). The source flux faded to below the background level before the next XMM-Newton  observation of the same field on Dec. 28, 2000. The upper limit for the count rate for this second observation was less than 6.4 counts/ks (2$\sigma$ upper limit for MOS1). The detected luminosity of the source during the June 25 observation, as well as its several-month time scale to fade to quiescence, is typical for bright X-ray transients in our Galaxy (see reviews by Tanaka & Shibazaki 1996; Chen et al. 1997).

3.2 ${\mathsfsl {Chandra}}$ transient

A bright transient source designated CXO J004242.0+411608 was discovered with Chandra  during a 17.5 ks observation of the core of M 31 on Oct. 13, 1999 (G2000). We detected an X-ray source at the same location during both our XMM-Newton  PV observations. The flux of the source measured on June 25, 2000 was equal to $(5.5\pm 0.5) \times 10^{-13}$ erg s^-1 cm^-2 ( $L_x \sim 3.8\times 10^{37}$ erg s^-1). The source was detected again on Dec. 28, 2000 with flux of $(4.2 \pm 1.0)\times 10^{-13}$ erg s^-1 cm^-2. Including both the Chandra  and XMM  detections, the source has remained bright for more than 14 months, with only slight possible fading ($\sim$25%) between the two XMM-Newton  observations separated by half of a year. The X-ray luminosity and spectral shape detected by XMM-Newton  (Paper III) correspond to those of the hard/low state of Galactic black hole transients (e.g., Tanaka & Lewin 1995). Typically Galactic X-ray transients fade significantly in less than $\sim$100 days (see Chen et al. 1997 for a review); however, a long plateau in a hard spectral state was observed from the X-ray transient GRS 1716-249 (Sunyaev et al. 1994; Revnivtsev et al. 1998). We should mention that our sampling does not allow us to distinguish reliably between single and repeated outbursts. It will be interesting to follow the evolution of CXO J004242.0+411608 during future planned observations with XMM-Newton .

3.3 Supersoft 865-s pulsator

Figure 2: Power density spectra for a supersoft source in M 31 obtained from each of the three EPIC instruments during the observation on June 25, 2000. The peak power from each of the three instruments occurs at a frequency corresponding to to a period of $\sim$865 s. The vertical dotted line denotes the best fit value of the period found by the epoch-folding method. Detection thresholds are indicated by horizontal dashed lines labelled with the approximate probability that any of the 1500 or 1300 frequency bins in each PDS (for MOSs and PN respectively; no averaging applied) would have a noise value exceeding the indicated power level (Vaughan et al. 1994).

We detected significant oscillations in the X-ray flux from one of the bright variable sources in our field (source #9 in Table 2). The Fourier power density spectra (PDS) from each of the three independent EPIC instruments (Fig. 2) show a highly significant periodic signal from this source during the June observation, with a best-fit period of $865.5 \pm 0.5$ s. The folded light curve for this source (Fig. 3) is quasi-sinusoidal with an amplitude of about 40%.

In June 2000 this 865-s pulsator had a spectrum similar to the three sources in the XMM  field previously identified by Kahabka (1999) as supersoft source (SSS) candidates based on their ROSAT /PSPC hardness ratios and luminosities (one was identified with a supernova remnant). The temperature of the blackbody spectral fit was ${\it kT}_{\rm bb}=61\pm 2$ eV with reduced $\chi^2=1.5$ (for EPIC-MOS data). All four SSSs exhibit blackbody-like spectra with effective temperatures in the range ${\it kT}\sim 50$-150 eV and no detectable X-ray emission above $\sim$1.5 keV (see Paper III for more detailed discussion of spectral parameters). Such sources are typically interpreted as accreting white dwarfs in binary systems, powered by nuclear burning of the accreted matter on their surfaces (see Kahabka & van den Heuvel 1997 for a review).

The source count rate during the June 25 observation was equal to 44 counts/ks in the 0.3-1.5 keV energy range (MOS1, see Table 2). By Dec. 28 the source count rate had faded down to a level below 5 counts/ks (MOS1); thus, it was not possible to detect pulsations during the second observation. All three supersoft sources in the NE portion of our field (sources ## 8, 9 and 10) faded between the June and December observations (see Table 2). We are aware of a possible sensitivity degradation in this region of MOS1; however, the result is confirmed by MOS2 and PN data. In particular, the count rate of the supersoft pulsator as detected by MOS2 dropped down from $41 \pm 1.5$ counts/ks in June to $3 \pm 1.5$ counts/ks in December, while many other sources in the field showed no significant variability or became brighter in December.

We note that there is a nearby source ( ${00}^{\rm h}{43}^{\rm m}{21}\hbox{$.\!\!^{\rm s}$ }{2}, \delta={+41}\hbox{$^\circ$ }{17}\hbox{$^\prime$ }{52}\hbox{$^{\prime\prime}$ }$, J2000) with a harder spectrum which is much fainter than the supersoft pulsator in the June observation but dominates this region of the sky for the December observation. The distance between two sources is about 10 $\hbox{$^{\prime\prime}$ }$, so that they are spatially resolved with XMM-Newton , but their counts are not completely spatially separated on the detector. Because the relative flux of the harder source was much lower during the June observation, it did not significantly affect our timing and spectral analysis of the SSS. The ROSAT /PSPC would barely be able to resolve the two sources, so ROSAT 's RX J0043.3+4117 (Su01#236) may include contributions from both of them, although its position coincides with that of the harder source rather than the SSS pulsator.

The 865.5-s period is the shortest among all known SSSs (Greiner 2000). Interpreted as the binary orbital period, it would be too short to accommodate a main-sequence companion and would suggest a degenerate secondary. It may be more plausible to assume that the pulsations indicate that the white dwarf possesses a magnetic field large enough to modulate the X-ray emission yet not so large that the spin and orbital periods are locked, e.g., as in intermediate polars or DQ Her stars (see review by Cordova 1995). However, the absorbed luminosity of the object $\sim$ $1.7 \times 10^{37}$ erg s^-1 (0.3-1.5 keV) is several orders of magnitude higher then typical for intermediate polars luminosity range $L_{\rm X}=10^{31}$-10³⁴ (see e.g. Patterson 1994). The high luminosity and transient nature of the pulsator may indicate steady burning in a post-nova stage, as has been observed in a few classical and symbiotic novae (Kahabka & van den Heuvel 1997). An alternative explanation for the nature of the SSS could be a double degenerate polar similar to RX J1914.4+2456 (Haberl & Motch 1995; Ramsay et al. 2000). An interpretation of the source as a foreground object is unlikely due to the lack of an optical counterpart in Dec. 2000 XMM  OM images up to the limiting magnitude of $\sim$19 in the B filter.

3.4 Variability of other sources within a single observation

We have searched for coherent periodic modulation on time scales from $\sim$10 to $\sim$1000 s for about 60 of the brightest sources in the XMM-Newton  field of view; however, only in one case (see previous subsection) was significant variability detected. The 90% upper limits to periodic modulation fractions, calculated as the ratio of the sine amplitude to the constant flux level for periods of 10 000, 300, and 10 s were obtained in all other cases. These upper limits vary from 3.6% for the brightest source to $\sim$30% for the faintest sources of the sample.

The lack of detectable variability for many of the individual sources in M 31 may seem surprising compared with rich variability observed from Galactic sources. Our observations so far have included mainly the bulge of M 31, where low-mass X-ray binaries (LMXBs) are most prevalent. Such systems commonly show dips, bursts, and quasi-periodic variability rather than coherent pulsations. We expect to detect more X-ray pulsars during planned XMM  observations of M 31 in fields along the disk of M 31, where population I stars and high-mass X-ray binaries (HMXBs) dominate. We must note also that the sensitivity to variability depends strongly on the brightness of the source, and hence our data are not very sensitive to the variability of the many faint sources.

We have also looked for non-periodic variations on all accessible timescales for the same set of objects. No X-ray burst has been detected, however our sensitivity to bursts is restricted to a relatively small luminosity interval by the low count rate of most sources and by the Eddington limit.

Figure 3: Light curve of the pulsating supersoft source in M 31, folded using a period of 865.5 s. The data is from the June 25, 2000 observation, and counts from all three EPIC instruments have been used. Contribution of the background is within the error bars.