RX J0045.4+4154: A Recurrent Supersoft X-ray Transient in M31

N.E. White, Laboratory for High Energy Astrophysics,NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA
P. Giommi, European Space Information System, Information Systems Division of ESA, ESRIN, via G. Galilei, 00044 Frascati, Italy
J. Heise, SRON Space Research Utrecht,Astrophysics Working Group, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands.
L. Angelini and S. Fantasia, Laboratory for High Energy Astrophysics, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA

Accepted for Publication in the Astrophysical Journal Letters, Vol 445, L125

Abstract
Using data extracted from the ROSAT archive we have discovered a recurrent supersoft X-ray transient RX J0045.4+4154 in M31. The first outburst began on 1992 Feb 2 and continued for at least four days, until the end of the observation sequence. A second outburst that lasted >6 days was seen to begin on 1993 Jan 7. The X-ray spectrum on both occasions yields a characteristic blackbody temperature of ~90 eV. For a range of plausible continuum models, the hydrogen column density is 0.8-1.5*10^21 persqcm and is consistent with the source being located in M31. This implies an unabsorbed 0.1-2.0 keV peak luminosity of ~10^38 erg/s. This is the first recurrent X-ray transient to be found in M31, and is particularly notable because it is much softer than the bright X-ray transients seen in our galaxy. The spectrum is characteristic of the supersoft class of X-ray sources, which are thought to be accreting white dwarfs that have a hydrogen burning surface layer. A fit to a white dwarf model atmosphere gives a temperature of 10^6 K, the hottest found so far. This high temperature is consistent with a white dwarf mass of 1.3-1.4 Msun, approaching the Chandrasekhar limit, and burning close to the nuclear stability limit.

Introduction
ROSAT observations have established a new class of X-ray binary called supersoft sources which have a characteristic blackbody temperature of a few tens of eV, a bolometric luminosity of ~10^38 erg/s, and in some cases transient outbursts (Trumper et al 1991; Greiner, Hasinger and Kahabka 1991; Hasinger 1994). The supersoft sources are in most cases only detected below 0.5 keV, so that within our own galaxy they are mostly hidden by interstellar absorption in the galactic disk. They are, however, readily evident in external galaxies, with ~10 found in the Magellanic Clouds and at least 15 seen in M31 (Hasinger 1994; Supper 1994). Supersoft sources have now been discovered in our galaxy (e.g. Motch, Hasinger and Pietsch 1994) and in the globular cluster M3 (Hertz, Grindlay and Bailyn 1993). Blackbody fits imply Eddington limited and in some cases super-Eddington bolometric luminosities for a neutron star or white dwarf (Kahabka et al 1994). Several supersoft source show transient outbursts e.g. RX J0513.9-6951 (Schaeidt, Hasinger and Trumper 1993), some of which are recurrent (Kahabka 1995) while others appear to be stable on a timescale of years (Hasinger 1994).
Models for the supersoft sources include: the hot inner region of an accretion disk surrounding a white dwarf (Kahabka, Pietsch and Hasinger 1994); super-Eddington accretion onto a neutron star (Kylafis and Xilouris 1993); and nuclear burning of hydrogen on a white dwarf accreting material at a high rate of ~10^-7 Msun/yr (van den Heuvel et al 1992; Pacynski and Zytkow 1978). The identification of a supersoft source with the classical nova GQ Mus (Ogelmann et al 1993) and white dwarfs in several other systems (van den Heuvel et al 1992) favours the nuclear burning model. The use of white dwarf model atmospheres instead of simple blackbody fits means that super-Eddington luminosities are not required (Heise, van Teeseling and Kahabka 1995). Instabilities in the nuclear-burning layer can naturally account for the variability (recurrent transient behaviour) seen in some supersoft sources. Optical and IUE observations imply the presense of a bright accretion disk in such systems indicating high mass transfer rates >10^-8 Msun/yr (Pakull et al 1993).
Using the ROSAT archive we have found a new recurrent transient supersoft source in M31. The discovery was first announced in White et al (1994a). In this letter we report the details of the discovery, the overall source properties and how this source fits into the supersoft class.

Results
The one degree radius field of view of the ROSAT position sensitive proportional counter, PSPC, combined with its large collecting area and low background (Trumper 1983) means that any given observation makes a X-ray survey of the region of sky surrounding the target. Over the 3 yr pointed phase of the mission about 10% of the sky was covered, with varying degrees of sensitivity. Most of these data are now available in the public archive and we have undertaken a project to catalog the sources in each field (White, Giommi and Angelini 1994b). An optimized sliding-cell detect is being used to generate a list of X-ray sources for each observation. This software was originally developed to catalog sources detected in the EXOSAT low energy imaging telescopes (Giommi et al 1991). To identify highly variable objects a Kolmogorov-Smirnov (KS) test was made where the arrival time distribution of the photons for each pixel coordinate is compared with the corresponding distribution for the entire field. The result of the KS test is a chi2 value (with 2 degrees of freedom) that is used to assign an intensity value to each pixel in an image. In this way pixels where the distribution of photon arrival times is not consistent with that of the entire image will be assigned high intensity values (high chi2 values). The resulting time variability image visualize sources with strong time variability (Giommi, White and Angelini 1994). Test runs were made on a few fields of differing complexity and a time variability image produced in one of these runs revealed a transient X-ray source in an observation of M31 made in Janauary 1992.
The left panel in Figure 1 shows the image from a 45,000s ROSAT exposure of the NE region of M31 made in January/February 1992. Many sources are detected, and these are most likely either bright X-ray binaries in M31 or background AGN (Supper 1994). In the time variability image, shown in the right panel of Figure 1, one source clearly stands out. A lightcurve of that source shows an abrupt outburst began on 1992, February 2 (Figure 2). Prior to that time the source was not detected. We examined the ROSAT archive for all instances where this source position was observed, and Table 1 gives a list of the ROSAT observation request (ROR) number, the offset angle of the source in the detector, the exposure time and whether it was detected. The first outburst lasted at least until the end of that ROR, four days later. The source was not detected during the next ROSAT observation of this region in August 1992, nor in observations made in July 1991 (Supper 1994). A second outburst was seen to begin on 1993 Jan 7, and lasted for at least 5.5 days. The last observation in this sequence shows the source to be quite faint, which may indicate the end of the outburst. The first outburst is from a single ROR, but the second spans 12 different pointing positions (Table 1) which makes the analysis more complex. For each pointing position the count rate was corrected for the appropriate telescope off-axis response and local point spread function, PSF. The corrected background subtracted lightcurve of each outburst is shown in Figure 2.
The best position was estimated using eight observations made where the source was within 17 arc minutes of the pointing position. This minimizes uncertainties due to the distortion of the PSF at larger off axis angles. This gives in J2000 equinox RA = 00^h45^m29.0^s and Dec = 41\degree 54'7.6". The root-mean-square deviation of the measured positions is 13 arc second, which is consistent with the 90% confidence error radius for ROSAT PSPC detections. The source is designated RX J0045.4+4154 (= 1WGA J0045.4+4154; White et al 1994b).
The first outburst was contained in one observation, with the source at the same off-set angle. A spectrum was accumulated for the entire outburst, and a corresponding background spectrum taken from a nearby region. The detector response matrix was corrected for the effects of vignetting and the energy dependent point spread function. The spectrum was fit to a variety of simple models. A blackbody model gave a reduced chi2 of 1.01 with a temperature of 0.087 +/- 0.007 keV and an equivalent hydrogen absorption of (1.2 +/- 0.2) *10^21 cm^-2. The unabsorbed 0.1-2.0 keV flux is 6*10^-12 erg/cm2/s, which corresponds to a luminosity at 690 kpc of 3*10^38 erg/s. The best fitting model is shown in Figure 3 and the residuals given in the lower panel indicate the fit to be reasonable, although there is a single high channel around 0.8-0.9 keV. A thermal bremsstralung model resulted in a reduced chi2 of 1.32 with a temperature of 0.13 +/- 0.02 keV and absorption of (1.7 +/- 0.4)*10^21 cm^-2. A Mewe and Kaastra, meka model for emission from a solar abundance plasma in thermal equilibrium (Kaastra 1992, Mewe, Gronenschild, and van den Oord 1985) gives a reduced chi2 of 1.01 with a temperature of 0.16 +/- 0.02 keV and an absorption of 0.9 +/- 0.3)*10^21 H cm^-2. The emission measure for a source located in M31 is ~5*10^60 cm^-3, which is extremely large.
There is variability through the outburst and spectra were accumulated according to low, medium and high count rate. A blackbody model with absorption was simultaneously fit to all three spectra. The count rates were 0.017 ct/s, 0.051 ct/s and 0.073 ct/s for the low, medium and high intervals, respectively. The absorption parameter was simultaneously fitted to all three spectra, whereas the temperature and normalisation were allowed to vary. The temperature increases with intensity giving 0.061 +/- 0.014, 0.080 +/- 0.007 and 0.091 +/- 0.008 keV for the low, medium and high intervals, respectively. The corresponding 0.1-2.0 keV flux for each interval was 3.4*10^-13, 1.0*10^-12 and 1.2*10^-12 erg/cm2/s. The equivalent hydrogen column density was 1.3*10^21 persqcm and the reduced chi2 ~0.8.
In the off-state the best upper limit to any detection is during the long 30,000s exposure made in July 1991 (ROR=600066) when the source was close to on-axis (Table 1). This gives a three sigma upper limit of 9.1*10^-4 ct/s, a factor of 100 less than the peak outburst count rate. We also examined Einstein (HEAO-2) archival data from this region (see Trinchieri and Fabbiano 1991). An IPC observation including the position of RX J0045.4+4154 was made on 1979, January 11-12, but the source was not detected. The upper limit to the count rate was 4*10^-3 ct/s. The source position was close to the edge of the field of view in an Einstein HRI observations made in 1979 August and 1980 January, but it was again not detected.

Discussion
The equivalent hydrogen column density that we measure to RX J0045.4+4154 of 0.8-1.7*10^21 cm^-2 is comparable to the value expected from the galactic disk, based on HII measurements, of 0.8*10^21 cm^-2 (Stark et al 1992). This is consistent with RX J0045.4+4154 being associated with M31, rather than a foreground object. Bright transient X-ray sources are a feature of our galactic X-ray binary population (see e.g. White, Kaluzienski and Swank 1984) and it is not surprising to find a similar event in M31. Variability of the M31 X-ray sources is already well established (Primini, Forman, and Jones 1993; Trinchieri and Fabbiano 1991). A comparison of EXOSAT and Einstein images of the M31 bulge taken 4 yr apart revealed two bright transient sources (White and Peacock 1988). RX J0045.4+4154 is the first recurrent X-ray transient system to be seen in M31.
The RX J0045.4+4154 blackbody temperature of ~90 eV and bolometric luminosity of ~10^38 erg/s are unusual compared to the X-ray transients typically seen in our galaxy. These have characteristic temperatures that are much higher ranging from 1 keV in the ultra-soft black hole transients, to 5-10 keV in low mass X-ray binaries with neutron stars, to flat power-law spectra for the X-ray pulsars in high mass X-ray binaries (White et al 1984). The low temperature of RX J0045.4+4154 would have made it difficult to detect with traditional X-ray all-sky monitors, which typically cut-off at 1-2 keV (see e.g. Priedhorsky and Holt 1987). The location of RX J0045.4+4154 in M31 is one degree away from the center. If a similar source had appeared in our galaxy away from the galactic center, then even at a distance of 7 kpc it would be extremely bright, with e.g. in the ROSAT all-sky survey a count rate between 80 and 1000 ct/s, for a hydrogen column density ranging from 5*10^20 to 5*10^21 cm^-2.
RX J0045.4+4154 is most probably a member of the supersoft class of X-ray source. The 90 eV temperature is extreme for this population, where 10-50 eV is typical and most are not detected above 500 eV (Hasinger 1994). But exceptions are CAL87 in the LMC (Kahabka, Pietsch and Hasinger 1994) and RX J0925.7-475 in our galaxy (Motch et al 1994). The bolometric luminosity of RX J0045.4+4154 derived from the blackbody model is super-Eddington for an accreting white dwarf or neutron star, similar to CAL87 (Kahabka et al 1994). The transient nature of RX J0045.4+4154 is reminiscent of RX J0527.8-6954, a transient supersoft source in the LMC (Schaeidt et al 1993).
The ROSAT PSPC supersoft spectra have been characterized by Heise, van Teeseling and Kahabka (1995) as a step-function consisting of a constant flux (in erg cm ^-2s^-1) up to an edge, plus interstellar absorption. This is based on model atmospheres in the soft X-ray range which can be characterized by a slowly varying flux up to an absorption edge and a negligible flux beyond this. Moreover, supersoft source spectra are not consistent with purely scattering dominated atmospheres (Heise et al 1994). Most of the supersoft sources in the Magellanic Clouds are consistent with an edge of the order of 0.4 keV and therefore most likely represent spectra which exhibit a continuum up to the CV edge at 0.39 keV. CAL87, however, requires a step function fit at 0.71 keV, consistent with the OVII edge at 0.74 keV. A step function fit to RX J0045.4+4154 gives a narrow range of parameters: an absorption edge at 0.84 +/- 0.03 KeV with an absorption column of 6.7 (+0.9,-1.2)*10^20 cm^-2. In the soft X-ray range RX J0045.4+4154 is slightly harder than CAL87, but exhibits a lower interstellar absorption. It is compatible with an ionization edge of OVIII at 0.87 keV.
We have fit to the RX J0045.4+4154 spectra models for optically thick emission of a plasma in radiative equilibrium confined by a constant gravitational field (Heise, van Teeseling and Kahabka 1995). A grid of model LTE atmospheres at cosmic abundances (Heise 1995) in steps of 10,000 K were constructed for log(g)=8-10 (g in cm s^-2) in steps of 0.25 ranging from 2*10^5 to the Eddington limit indicated in Figure 4. The high densities in these atmospheres guarantee the validity of LTE and are appropriate close to the nuclear burning stability limit for white dwarfs. These are representative of steady nuclear burning on accreting white dwarfs. The results verify the general impression obtained from fits with a characteristic step function. Models with log(g) < 8.75 do not fit the data, because static LTE-atmospheres could not be made hot enough to show the OVIII ionization edge before blowing apart. To obtain an OVIII ionization edge at 0.87 keV requires a minimum gravity of log(g)=8.75, for which the resulting luminosity in the fits is of order 1*10^38 erg/s and the source radius is 4*10^8 cm at an assumed distance of 670 kpc. The radius, independantly found from the normalization to the fit, is consistent with the gravity used for a high mass white dwarf. We note, however, that at lower gravity (log g ~ 7) and with temperatures close to the Eddington limit non-LTE conditions are appropriate. Such models, when eventually available, might produce the OVIII ionization edge at lower gravity and somewhat lower temperatures.
Figure 4 shows the T-log(g) plane with a contour marked 0045.4+4154 where all models are expected that show OVIII as the dominant absorption edge, and in which other strong edges (CV, CVI, NVI, NVII) are ionized away. The top line is the Eddington limit (static atmosphere stability limit). Below (and almost parallel) is the line which represents the nuclear burning stability limit from the H-R diagram of Iben (1982). All stable models for on-state emission must fall between these two lines. Tracks for white dwarfs of different mass are given. Figure 4 illustrates that the white dwarf in RX J0045.4+4154 is amongst the hottest systems and could fall on either side of the nuclear stability line. If close to the stability line, the temperature and luminosity indicates a mass between 1.3 and 1.4 Msun. It corresponds to an extreme model in the high-T top-left hand area in the HR-diagram (cf Figure 2 in Iben 1982). The on-off cycle time calculated by Iben (1982) for a 1.4 Msun white dwarf accreting at a rate of 10^-7 Msun/yr is approximately consistent with the observed 340 d between outbursts, although the implied on-state time at this accretion rate (Dt_on ~ 30 d) may be too large if the 2nd outburst ended after 14 days (Figure 2). The high observed temperature suggests that RX J0045.4+4154 is an extreme supersoft system, where the white dwarf is close to its Chandrasekhar limit.

Acknowledgments
This work has made use of archival data extracted from the High Energy Astrophysics Science Archive Research Center, HEASARC operated by the NASA Goddard Space Flight Center. We thank Peter Kahabka for useful discussions, and Walter Lewin and David Helfand for advice and encouragement.

References

Giommi, P., et al. 1991, Ap. J. 378, 77.
Giommi, P., White, N.E., and Angelini, L., 1994 In "Astronomical Data Analysis Software and Systems" IV H. Payne and J. Hayes ed. ASP Conference series, in Press
Greiner, J., Hasinger, G., and Kahabka, P., 1991 A&A 246, L17
Hasinger, G., 1994, The Evolution of X-ray Binaries, AIP conference proceedings 308, p611.
Heise, J. 1995, in Cape Workshop on Magnetic Cataclysmic Variables, to be published in the Astron. Soc. of the Pacific Conference Series, Ed. D.A.H.Buckley.
Heise, J., van Teeseling, A., and Kahabka, P., 1994, A&A 288, L45
Heise, J., van Teeseling, A., and Kahabka, P., 1995, A&A submitted
Hertz, P., Grindlay, J.E. and Bailyn, C.D., 1993, Ap.J. 410, L87.
Iben, I., 1982, Ap.J. 259, 244.
Kaastra J.S. 1992, An X-Ray Spectral Code for Optically Thin Plasmas (SRON-Leiden Report)
Kahabka, P., Pietsch, W., and Hasinger, G., 1994, A&A 288, 538.
Kahabka, P., 1995, A&A submitted
Kylafis, N.D., and Xilouris, E.M., 1993, A&A 278, L43.
Mewe, R., Gronenschild, E.H.B.M., and van den Oord, G.H.J. 1985. A&A 62, 197.
Motch, C., Hasinger, G., and Pietsch, W., 1994, A&A 284, 827.
Ogelman, H., Orio, M., Krautter, J., and Starrfield, S., 1993, Nature 361, 331.
Paczynski, B., and Zytkow, A.N., 1978, Ap.J. 222, 604.
Pakull, M.W. et al, 1993, A&A 278, L39
Priedhorsky, W.C., and Holt, S.S., 1987, Sp. Sci. rev. 45, 291.
Primini, F.A., Forman, W., and Jones, C., 1993, Ap.J. 410, 615.
Schaeidt, S., Hasinger, G. and Trumper, J., A&A 270, L9.
Stark, A.A., Gammie, C.F., Wilson, R.W., Bally, J., Linke, R.A., Heiles, C., Hurwitz, M. 1992 Ap.J.Suppl. 79, 77
Supper, R., 1994 The Evolution of X-ray Binaries , AIP conference proceedings 308, p631.
Trinchieri, G., and Fabbiano, G., 1991, Ap.J. 382, 82.
Trumper, J. 1982, Adv. Space Res. 2, 241.
Trumper,J. et al 1991 Nature 349, 579.
Van den Heuvel, E.P.J., Bhattacharya, D., Nomoto, K. and Rappaport, S.A., 1992, A&A 262, 97.
White, N.E., Kaluzienski, J.L., and Swank, J.H., 1984, High Energy Transients in Astrophysics, AIP conference proceedings 115, p31.
White, N.E., Giommi, P.G., Angelini, L., and Fantasia, S., 1994a, IAUC 6064.
White, N.E., Giommi, P.G., and Angelini, L. 1994b, IAUC 6100.
White, N.E., and Peacock, A., 1988, Memorie S.A.It., 59, 7.

Table 1: ROSAT PSPC Observations used that include RX J0045.4+4154 -------------------------------------- ROR Exposure Offset Detect (s) (') -------------------------------------- 600065 28323 0.425 N 600066 30812 0.032 N 600067 27970 0.489 N 600121 45038 0.253 Y 600309 2576 0.459 N 600310 2721 0.351 N 600311 1352 0.282 N 600312 2415 0.260 N 600313 2672 0.303 Y 600314 2702 0.378 Y 600315 3033 0.488 Y 600329 1762 0.392 N 600330 2564 0.258 Y 600331 2875 0.144 N 600332 2954 0.091 Y 600333 3042 0.178 Y 600334 2888 0.294 N 600335 2363 0.425 Y 600349 2822 0.386 N 600350 2683 0.257 N 600351 2836 0.133 N 600352 2632 0.080 Y 600353 3007 0.174 Y 600354 2372 0.291 N 600355 1914 0.423 N 600369 2520 0.454 Y 600370 2866 0.349 N 600371 2082 0.270 N 600373 3136 0.292 Y 600374 2692 0.373 N 600375 3196 0.483 Y -------------------------------------

Figure 1: The image on the left is the M31 field containing RX J0045.4+4154. On the right is the time variability image. The single bright spot indicates that this source is highly variable.

Figure 2: The upper and lower panels show the lightcurves of the two outbursts from RX J0045.4+4154. The accumulation time is 1482 s.

Figure 3: The upper panel shows the ROSAT PSPC spectrum of RX J0045.4+4154 along with the best fitting blackbody model. The lower panel shows the residuals from the model.

Figure 4: The temperature-log(g) plane for the model atmosphere fits to the overall spectrum from RX J0045.4+4154. The models are bounded by the Eddington limit (top curve). The range of models with fits within the 1-sigma chi2-levels are indicated in the area marked 0045.4+4154. The thick line represents the expected relation for steady nuclear burning in accreting white dwarfs (c.f. Iben, 1982). Tracks for nuclear burning white dwarfs of different mass are indicated.

This image (which does not appear in the ApJ letter) illustrates the position of RX J0045.4+4154 relative to the nucleus of M31. On the left is a mosaic, made by Steve Snowden, that uses the inner region of the PSPC detector where the PSF is optimal. In these pointings RX J0045.4+4154 was not in outburst. Binning and smoothing the data for display purposes in the image has reduced the apparent sensitivity to point sources. On the right is an optical image obtained from the digitized sky survey of the Palomar plates using skyview.

This image (which does not appear in the ApJ letter) illustrates the position of RX J0045.4+4154 relative to the nucleus of M31. On the left is a mosaic, made from all of the long (> 10,000s) PSPC exposures and using the full field of view. RX J0045.4+4154 was in outburst during some of these PSPC observations. On the right is an optical image obtained from the digitized sky survey of the Palomar plates (generated using skyview). The position of the transient is indicated.

This image (which does not appear in the ApJ letter) illustrates the 1.8*1.8 arc min optical region around RX J0045.4+4154 taken from the Palomar survey. The error circle radius is 13 arc sec, and there is no obvious candidate in the center. This image was generated using skyview.
Back to the HEASARC