X-RAY AND OPTICAL OBSERVATIONS OF THE UNIQUE BINARY SYSTEM HD 49798/RX J0648.0–4418

Interacting binaries containing accreting white dwarfs constitute a large fraction of the population of bright Galactic X-ray emitters. These systems, including both persistent and transient sources, display a rich variety of properties, determined mainly by the strength of the white dwarf's magnetic field and the nature of the mass donor star (see, e.g., Kuulkers et al. 2006). Most accreting white dwarfs are found in cataclysmic variables, where accretion proceeds through Roche-lobe overflow. The mass donor in these systems is typically a main-sequence star or another white dwarf, but sub-giant and giant companion stars have also been observed. Symbiotic systems, where a white dwarf accretes from the stellar wind of a massive supergiant, have also been detected as X-ray sources.

The system discussed here is the only known X-ray binary composed of a white dwarf (RX J0648.0–4418) and a hot sub-dwarf star (HD 49798). Hot sub-dwarfs are evolved low mass stars that lost most of their hydrogen envelopes and are believed to be in the helium-core-burning stage (for a recent review on hot sub-dwarfs see Heber 2009). They are spectroscopically classified into different types: sdB (with effective temperature T ∼ 25, 000–28,000 K), sdOB (with T ∼ 33, 000–40,000 K), and sdO (with T > 40, 000 K). One possible mechanism responsible for the loss of their hydrogen envelopes is non-conservative mass transfer in a binary (Podsiadlowski & Han 2004).

Being one of the brightest hot sub-dwarfs (apparent magnitude V = 8.3), HD 49798 has been the object of several studies in the optical/UV wavebands. The first spectroscopic observations led to its classification as a sub-dwarf of O6 spectral type (Jaschek & Jaschek 1963) and showed radial velocity variations, later found to be caused by orbital motion with a period of 1.55 days (Thackeray 1970; Stickland & Lloyd 1994). No spectroscopic evidence for a secondary could be found, leading to the early suggestion that the companion could be a white dwarf (Thackeray 1970). Soft X-ray emission from the direction of HD 49798 was first detected with the Einstein Observatory in 1979, but it could be studied in detail only with a ROSAT satellite observation performed in 1992. The ROSAT data showed a very soft spectrum and led to the discovery of a strong modulation with a period of 13.2 s (Israel et al. 1997). While the periodicity clearly indicated the presence of a compact object, the poorly constrained spectral fit resulted in a large uncertainty in the X-ray luminosity, making it impossible to distinguish between a neutron star and a white dwarf.

Thanks to a long observation carried out with the XMM-Newton satellite in 2008 May, we could establish that the subdwarf's companion is a white dwarf. Its X-ray luminosity, well constrained by the high quality XMM-Newton spectrum, is much smaller than that expected from a neutron star accreting in the stellar wind of HD 49798. The measurement of the X-ray pulse delays induced by the orbital motion, and the discovery of the X-ray eclipse, allowed us to derive the X-ray mass function and to constrain the system's inclination. This information, coupled with the accurately measured optical mass function, gives the masses of the two stars: M_sd = 1.50 ± 0.05 M_☉ for the subdwarf HD 49798 and M_WD = 1.28 ± 0.05 M_☉ for its white dwarf companion RX J0648.0–4418 (Mereghetti et al. 2009). The main parameters of this binary system are summarized in Table 1.

Table 1. Main Parameters of the HD 49798/RX J0648.0–4418 Binary System

Parameter	Symbol	Value	Referencea
Orbital period	Porb	1.5476666 ± 0.0000022 days	(1)
Eccentricity	e	0	(2)
Inclination	i	79°–84°	(1)
Distance	d	650 ± 100 pc	(3)
HD 49798
Mass	Msd	1.50 ± 0.05 M☉	(1)
Effective temperature	Teff	46,500 K	(4)
Magnitudes		U = 6.758, B = 8.017, V = 8.287	(5)
Surface gravity	log g	4.35 cgs	(4)
Radius	Rsd	1.45 ± 0.25 R☉	(3)
Luminosity	L	3 × 1037 erg s−1	(3)
RX J0648.0–4418
Mass	MWD	1.28 ± 0.05 M☉	(1)
Radius	RWD	3000 km
Luminosity	LX	1032 erg s−1
Spin period	P	13.18425 ± 0.00004 s	(1)
Period derivative	$\dot{P}$	−5 × 10−13 s s−1 $<\dot{P}< 9 \times 10^{-13}$ s s−1

Note. ^aReferences. (1) Mereghetti et al. 2009; (2) Stickland & Lloyd 1994; (3) Kudritzki & Simon 1978; (4) Hamann 2010; (5) Landolt & Uomoto 2007.

Download table as:  ASCIITypeset image

RX J0648.0–4418 is one of the most massive white dwarfs currently known and the one with the shortest spin period. This binary is also particularly interesting since systems of this kind, where a massive white dwarf accretes from a helium star, have been proposed as possible progenitors of type Ia supernovae (Iben & Tutukov 1994; Wang et al. 2009). Here, we report a comprehensive analysis of all the XMM-Newton observations and the results of a search for optical pulsations.

XMM-Newton performed four short observations of RX J0648.0–4418 in 2002 and a longer one in 2008 (see Table 2). To look for possible variability, the 2002 pointings were taken at different orbital phases.⁵ Based on the optical ephemeris of the system (Stickland & Lloyd 1994), the 2008 observation was scheduled to include the orbital phase 0.75, in order to look for the presence of an X-ray eclipse. This orbital phase was not covered by previous X-ray observations.

Here, we concentrate on data obtained with the EPIC instrument (0.1–12 keV), consisting of two MOS and one pn CCD cameras (Turner et al. 2001; Strüder et al. 2001). During all the observations, the three cameras were operated in Full Frame mode (time resolution of 73 ms and 2.6 s for pn and MOS, respectively) and with the medium optical blocking filter. The data were processed using SAS version 9.0.

The source photons for the timing analysis were extracted from a circular region of 30'' radius and their arrival time corrected to the solar system barycenter and for the orbital motion using the system parameters given in Table 1. The source spectra were extracted from a slightly smaller circle (20'' radius), in order to avoid possible contamination at high energy from a source of similar brightness and harder spectrum located 70'' away from RX J0648.0–4418. The spectra were rebinned to have at least 30 counts per energy channel and to oversample by a factor three the instrumental energy resolution. The background spectra were extracted from nearby regions, which were on the same chip as the source and where no sources were detected. Observation B was affected by high particle background, therefore we excluded it from the spectral analysis.

2.1. Variability

The background-subtracted light curves of RX J0648.0–4418 in the total (0.1–10 keV), soft (0.1–0.5 keV), and hard (0.5–10 keV) energy ranges are plotted as a function of the orbital phase in Figures 1 (pn) and 2 (sum of the two MOS). These figures clearly show that a significant flux is also detected during the X-ray eclipse, with net count rates of (2.8 ± 0.3) × 10⁻² counts s⁻¹ in the pn and (1.7 ± 0.2) × 10⁻² counts s⁻¹ in the sum of the two MOS. The presence of X-ray emission during the eclipse is confirmed by the image shown in Figure 3, which was accumulated selecting events in the time interval corresponding to the eclipse.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Besides the variation due to the eclipse, the light curves shown in Figures 1 and 2 indicate the possible presence of some variability. In particular, the count rates in 2002 September (Observation D) were slightly larger than in the 2008 observation, especially at energies above 0.5 keV. This variability is confirmed by the spectral analysis described below.

2.2. Spectral Analysis

We first analyzed the X-ray emission detected during the eclipse by extracting the pn and MOS spectra corresponding to the orbital phase interval 0.73–0.77. The three spectra were fitted together obtaining good results either with a power law with photon index Γ = 2.8 ± 0.3 or with a thermal bremsstrahlung with temperature kT_Br = 0.55^+0.3_{− 0.2} keV (χ²/dof = 17.2/16 and 24.7/16, respectively). In both cases we fixed the interstellar absorption to the value N_H = 2.7 × 10¹⁹ cm⁻² derived from the 2008 un-eclipsed pn data. The X-rays seen during the eclipse of the white dwarf are most likely due to emission from HD 49798, but even in the case of a different origin, we expect that this emission is present at all orbital phases. Therefore we included its contribution in all the subsequent fits to the non-eclipsed RX J0648.0–4418 emission.

The 2008 out-of eclipse pn and MOS spectra were extracted excluding the orbital phase interval 0.71–0.79 and fitted jointly with different models.⁶ Single component models gave unacceptable results, while good fits could be obtained with two-component models consisting of a blackbody plus either a power law or a thermal bremsstrahlung. The best-fit parameters are summarized in Table 3. The results differ slightly from those we reported earlier (Mereghetti et al. 2009) due to the inclusion in the present analysis of the eclipse component. The blackbody component, which accounts for most of the flux, is similar in both models, with temperature kT_BB ∼ 40 eV and bolometric luminosity L_BB ∼ 8 × 10³¹ erg s⁻¹ (for d = 650 pc). The hard component provides an additional luminosity of ∼8 × 10³⁰ erg s⁻¹ (0.2–10 keV, corrected for the absorption). To check whether the uncertainties in the eclipse spectrum affect the overall results, we repeated the analysis with different bremsstrahlung temperatures of the fixed component in the range kT_Br = 0.3–0.8 keV and found that the best-fit values for the non-eclipse spectra change by less than 10%.

Table 3. Results of Phase-averaged Spectroscopy of the 2008 Observation (pn+MOS)

Parameter	Blackbody + Power law	Blackbody + Bremsstrahlung
NH (cm−2)	<2.4 × 1019	<1.6 × 1019
kTBB (eV)	38.9+1.4− 1.2	40.0+1.1− 1.3
RBBa (km)	17.9+3.2− 1.6	16.3+1.7− 1.0
Γ	1.6 ± 0.1	...
FPLb (erg cm−2 s−1)	(1.7 ± 0.1) × 10−13	...
kTBrem (keV)	...	7.9+2.7− 1.7
FBremc (erg cm−2 s−1)	...	(1.53 ± 0.08) × 10−13
LBBd (erg s−1)	8.9 × 1031	8.2 × 1031
kTEcl (keV)	0.55e	0.55e
FEclc (erg cm−2 s−1)	4.3 × 10−14e	4.3 × 10−14e
χ2/dof	151.1/133	158.7/133

Notes. All the errors are at the 90% c.l. for a single interesting parameter. ^aBlackbody emission radius at infinity, for d = 650 pc. ^bObserved flux of the power law component 0.2–10 keV. ^cObserved flux of the bremsstrahlung component 0.2–10 keV. ^dBolometric luminosity of the blackbody component, for d = 650 pc. ^eFixed eclipse component.

Download table as:  ASCIITypeset image

Also the spectra of the 2002 observations rule out single component models, but, owing to their lower statistics, the best-fit parameters have larger uncertainties compared to those obtained from the 2008 data. To search for possible long term spectral variations we fitted the 2002 spectra using a power-law plus blackbody model with photon index, temperature, relative normalization, and absorption fixed at the 2008 best-fit values. The eclipse component was always included, with all the parameters fixed as described above. For simplicity we restricted the comparison to the pn data, which are those with the smallest statistical uncertainties. The results are shown in Figures 4 and 5, referring respectively to the observations of 2002 May (A+C) and 2002 September (D). The residuals obtained in this way (middle panel of each figure) suggest that in 2002 September only the power-law component was brighter than in 2008, while in 2002 May the largest residuals are in the low-energy part of the spectrum, where the blackbody component dominates. Allowing the power-law normalization to vary in the 2002 September fit, we obtained good results with a power-law flux ∼30% higher than in 2008 (see residuals in the lower panel of Figure 5). For the 2002 May spectrum, we kept all the power-law parameters fixed and allowed the blackbody component to change. This resulted in a best-fit temperature kT_BB = 31.6 ± 0.2 eV and emitting radius R_BB = 42.5 ± 1.5 km (see residuals in lower panel of Figure 4). Table 4 summarizes all the parameters obtained in this analysis of the 2002 observations.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 4. Spectroscopy Results of the 2002 Observations (pn Data)

Parameter	2002 May (A+C)	2002 September (D)
NH (cm−2)	2.7 × 1019a	2.7 × 1019a
kTBB (eV)	31.6 ± 0.2	39.6a
RBB (km)	42.5 ± 1.5	16.6a
Γ	1.57a	1.57a
FPLb (10−13 erg cm−2 s−1)	1.7a	2.2 ± 0.2
LBB (erg s−1)	2.2 × 1032	8.4 × 1031
kTEcl (keV)	0.49a	0.49a
FEcl (10−14 erg cm−2 s−1)	4.2a	4.2a
χ2/dof	41.8/31	34.7/34

Notes. All the errors are at the 90% c.l. for a single interesting parameter. ^aFixed. ^bObserved flux of the power law component 0.2–10 keV.

Download table as:  ASCIITypeset image

2.3. Timing Analysis

A best-fit pulse period P = 13.18425 ± 0.00004 s was derived for the 2008 observation by Mereghetti et al. (2009). After correcting the times of arrival to the solar system barycenter and for the effects of the orbital motion of the system, we searched for pulsations in the 2002 observations obtaining the values reported in Table 2. No long-term variations in the spin period could be inferred by comparing the 2002 and 2008 values. A linear fit gives a 90% c.l. interval −5 × 10⁻¹³ s s⁻¹ $<\dot{P}< 9 \times 10^{-13}$ s s⁻¹ for the period derivative.

The folded light curves at soft and hard X-rays are shown in Figure 6. No significant variations are seen by comparing the 2002 and 2008 light curves. On the other hand, all data sets show a clear difference between the two energy ranges. The soft band, where the blackbody component dominates, has a nearly sinusoidal profile with a pulsed fraction of 56%. Above 0.5 keV the light curve is instead characterized by two peaks, of unequal intensity and out of phase with respect to the soft pulse.

Zoom In Zoom Out Reset image size

2.4. Phase-resolved Spectroscopy

On the basis of the spectral results presented above, clearly showing the presence of two distinct components, we performed phase-resolved spectroscopy choosing different phase intervals tailored to the light curves seen in the soft and hard energy ranges. We concentrate here on the 2008 data (excluding the eclipse). The 2002 data gave fully consistent, but less constrained, results, owing to their smaller statistics. In all the phase-resolved spectral fits discussed below we included the fixed eclipse component, as was done for the phase-averaged analysis.

We first searched for spectral variations as a function of the pulse phase in the soft component, by extracting four spectra corresponding to two pairs of phase intervals selected on the basis of the 0.1–0.5 keV pulse profile: pulse minimum/maximum (phases 0.25–0.65/0.65–1.25) and pulse decline/rise (phases 0–0.4/0.4–1). All the spectra were fitted with the power-law plus blackbody model keeping N_H linked to a common value. The blackbody temperatures obtained in the four spectra were all consistent with the value kT_BB ∼ 40 eV found in the phase-averaged spectrum. Acceptable χ² values were obtained by forcing a common value of kT_BB and fitting together the minimum and maximum or the rise and decline spectra. Thus, we conclude that all the variations in the low energy pulse profiles can be interpreted as a change in the intensity of the blackbody component that remains at a constant temperature.

In order to investigate the phase dependence of the hard component, we extracted three spectra corresponding to the intervals marked in Figure 6: first peak (P1, phases 0.15–0.35), second peak (P2, phases 0.65–0.85), and inter-peak (IP, phases 0.35–0.65 and 0.85–1.15). The three spectra were fitted with a blackbody plus power-law model, keeping absorption and temperature fixed to the phase averaged values obtained with the pn (N_H = 2.7 × 10¹⁹ cm⁻² and kT_BB = 39.6 eV). The results are given in Table 5. The spectrum during the first peak is slightly softer than at other phases, as shown by the confidence level contours for the power-law photon index and normalization plotted in Figure 7.

Zoom In Zoom Out Reset image size

Table 5. Results of Phase-resolved Spectroscopy of the 2008 Observation (pn Data)

Parameter	Peak 1	Peak 2	Inter-peak
NHa (cm−2)	2.7 × 1019	2.7 × 1019	2.7 × 1019
kTBBa (eV)	39.6	39.6	39.6
RBB (km)	15.1 ± 0.8	17.0 ± 0.7	16.6 ± 0.4
Γ	1.9 ± 0.2	1.4 ± 0.15	1.39 ± 0.15
FPLb (10−13 erg cm−2 s−1)	1.55 ± 0.25	2.6 ± 0.3	1.2 ± 0.2
χ2/dof	38.4/28	43.8/32	83.7/62

Notes. All the errors are at the 90% c.l. for a single interesting parameter. ^aFixed. ^bObserved flux of the power law component 0.2–10 keV.

Download table as:  ASCIITypeset image

We observed HD 49798/RX J0648.0–4418 with the University of Cape Town CCD photometer (O'Donoghue 1995) at the 1.9 m telescope of the South African Astronomical Observatory in Sutherland on 2009 December 25 and 27. The observations were carried out in the U filter, with a time resolution of 4 and 6 s, respectively. The first observing run lasted 29.1 minutes, the second run lasted 60.2 minutes.

A timing analysis of the optical data did not reveal any significant signal at, or close to, the frequency of the X-ray pulsations down to a level of 1.5 mmag (December 25) and 0.6 mmag (December 27). These limits correspond respectively to a pulsed flux of 1.5 × 10⁻³ and 0.6 × 10⁻³ photons cm⁻² s⁻¹ Å⁻¹ at 3600 Å, several orders of magnitude above the extrapolation to the optical U band of the X-ray blackbody spectrum. Even considering the low energy extrapolation of the power law seen during Peak 1, i.e., the spectral component with the steepest slope (Γ = 1.9 ± 0.2), gives a 3600 Å flux of a few 10⁻⁶ photons cm⁻² s⁻¹ Å⁻¹, well below our upper limit. However, other mechanisms not directly related to the spectral components seen in X-rays might in principle contribute in the optical/UV band, and the above limit is the best currently available in the optical band for this pulsar.

The results of X-ray observations with XMM-Newton spanning six years indicate that the properties of RX J0648.0–4418 are rather stable. Only minor changes were seen in the spectrum, but it is not possible at this stage to attribute them to the different orbital phases of the observations or to a long term variability. The spectrum and luminosity seen with XMM-Newton are consistent with those measured with ROSAT in 1992 and Einstein in 1979, although the uncertainties in the latter data are rather large. No significant variations were seen in the spin period, pulsed fraction, and shape of the pulse profile. Given that this is the only known X-ray-emitting binary composed of a hot sub-dwarf and a white dwarf, it is worth examining different possibilities for the origin of the observed X-ray emission.

The upper limit on the white dwarf secular spin-down ($\dot{P}<9 \times 10^{-13}$ s s⁻¹) implies a rotational energy loss I_WD 4π² P⁻³ $\dot{P}$ ≲ 1.5 × 10³⁶ erg s⁻¹ (for a moment of inertia I_WD = 10⁵⁰ g cm²). This is sufficiently high to power the observed luminosity of ∼10³² (d/650 pc)² erg s⁻¹, with an efficiency similar to that seen in rotation-powered neutron stars. Even if no convincing examples have been discovered, the existence of white dwarf analogs of rotation-powered neutron stars is an intriguing possibility (see, e.g., Usov 1993). Among isolated white dwarfs, even the one with the fastest rotational velocity (RE J0317–853, P = 726 s, Vennes et al. 2003), can develop a maximum potential drop along open field lines⁷ of only ΔV = Ω² B R³/c² ∼ 7 × 10⁹ (726 s/P)² (B/660 MG) (R/5000 km)³ V, despite its high magnetic field (B ∼ 170–660 MG). This should be compared with the much higher value ∼10¹⁵–10¹⁷ V typically reached in radio pulsars. White dwarfs in binary systems can rotate much more rapidly. After RX J0648.0–4418, the next fastest rotators are found in the dwarf nova WZ Sge (P = 28.87 s; Patterson et al. 1998) and in the intermediate polar AE Aqr (P = 33 s; Patterson 1979). Their magnetospheres can, in principle, accelerate particles to high energies, but still few orders of magnitude below that reached in radio pulsars. A value ΔV ∼ 10¹⁴ V could be attained in RX J0648.0–4418, but only in the case of a strong magnetic field, which, as we discuss below, seems rather unlikely. The possible detection of pulsed hard X-rays (10–30 keV) in AE Aqr has been interpreted as evidence for non-thermal processes powered by the white dwarf rotational energy (Terada et al. 2008), implying an efficiency of ∼0.1%. AE Aqr spins down at a rate of 5.6 × 10⁻¹⁴ s s⁻¹, most likely as a result of the magnetic propeller effect. Its K-type main-sequence companion transfers mass trough Roche-lobe overflow, but the accretion stream is disrupted by the magnetic field of the white dwarf and most of the mass is ejected (Wynn et al. 1997). AE Aqr exhibits a hard X-spectrum and strong rapid variability at all wavelengths, contrary to RX J0648.0–4418 which shows a nearly constant flux, dominated by a thermal-like component. The significant differences, except for the short spin-period, between these systems disfavor the presence of a rotation-powered mechanisms in RX J0648.0–4418.

We believe that accretion provides a much more plausible and natural explanation for the X-ray emission observed from RX J0648.0–4418. The Roche lobe of HD 49798 has a radius R_L ∼ a (0.38+0.2 log (M_sd/M_WD)) = 3.2 R_☉, where a = 5.5 × 10¹¹ cm is the orbital separation. The sub-dwarf, with a radius R_sd = (1.45 ± 0.25) R_☉ (Kudritzki & Simon 1978), is significantly smaller than its Roche lobe. Therefore, accretion can only occur via stellar wind capture. Hot sub-dwarfs have relatively weak stellar winds, compared to main-sequence or giant stars of the same spectral type, but HD 49798 is one of the few sdO stars for which evidence of a significant mass loss has been reported. The P Cygni profiles of its N v and C iv lines indicate a mass loss in the range 5 × 10⁻¹⁰–10⁻⁸ M_☉ yr⁻¹ and a wind terminal velocity v = 1350 km s⁻¹, reached at only 1.7 R_sd (Hamann et al. 1981). A more precise estimate of $\dot{M}= 3\times 10^{-9}$ M_☉ yr⁻¹ has been obtained thanks to a recent analysis of the optical/UV spectra based on improved wind models (Hamann 2010). The accretion rate onto the white dwarf can be estimated as $\dot{M}_{{\rm WD}}\sim \dot{M}$ (R_a/2a)² = 8 × 10⁻¹³ M_☉ yr⁻¹, where R_a ∼ 2GM_WD/v² is the accretion radius. The corresponding accretion luminosity is 3 × 10³¹ (R_WD/3000 km)⁻¹ erg s⁻¹, in good agreement with the observed value.

The presence of two components in the X-ray spectrum of RX J0648.0–4418 is similar to what is observed in many cataclysmic variables of the polar and intermediate polar classes. In these systems, the magnetic field starts to dominate the motion of the accreting matter at large distances from the white dwarf, channeling the flow onto the star's magnetic pole(s), and even preventing the formation of an accretion disk in the systems with the strongest fields. Shocks are formed in the accretion column above the white dwarf and the hot plasma in the post-shock region emits hard X-rays. This hard component is usually fit with multi-temperature plasma models (see, e.g., Cropper et al. 1999; Yuasa et al. 2010), but, as a first approximation, it can be described by a thermal bremsstrahlung with temperature kT ∼ 30 (M/M_☉) R⁻¹₉ keV, where M is the white dwarf mass and R₉ its radius in units of 10⁹ cm. Part of the hard X-ray emission may be intercepted and reprocessed by the white dwarf surface, giving rise to UV and soft X-ray thermal emission, with blackbody temperature of tens of eV. The relative importance of the soft and hard components depends on several factors, including geometrical effects, the white dwarf mass, magnetic field, accretion rate per unit surface, with the consequence that the observed values of the ratio L_soft/L_hard span a large range (Evans & Hellier 2007).

The value L_soft/L_hard ∼ 10 observed for RX J0648.0–4418 is rather high, but not uncommon. For example, similar or larger values are found in EU UMa, DP Leo, and RX J1007–2016 (Ramsay & Cropper 2004). The blackbody temperature we derived for the soft component as well as the dimensions of the emitting surface is well within the range of values measured for the polars. On the other hand, the best-fit temperature of the bremsstrahlung component, kT_TB ∼ 8 keV, is rather small if one considers that, based on the relation mentioned above, massive white dwarfs should have the highest shock temperatures. This discrepancy might have to do with the unique properties of this white dwarf, which accretes from a low density stellar wind rather than through Roche-lobe overflow. The intensity and configuration of the magnetic field is also relevant for the properties of the X-ray emission, and might explain the differences between this system and classical cataclysmic variables. An upper limit on the dipolar field of RX J0648.0–4418 can be set by the requirement that its magnetospheric radius, R_M, be smaller than the corotation radius R_COR = (GM_WD P²/4π²)^1/3 = 9 × 10⁸ cm. Since R_M is defined by balancing the magnetic pressure and the ram pressure of the accreting matter, this translates in an upper limit on the white dwarf magnetic moment μ ∼ 2 × 10²⁹ g cm³. However, the above condition does not exclude the possibility that the source be in the sub-sonic propeller regime, in which R_M < R_COR, but the matter is too hot to accrete (Davies et al. 1979). Using the equation that defines the transition between the accretion regime and the subsonic propeller regime (Bozzo et al. 2008), we obtain an upper limit μ < 3 10²⁸ g cm³, corresponding to a field of the order of 1 kG at the white dwarf's surface.

Early type stars emit X-rays with relatively soft thermal spectra (kT ∼ 0.5 keV) and luminosity proportional to the bolometric luminosity. The average relation L_X ∼ 10⁻⁷ L_bol has been derived from the observations of main sequence, giant and supergiant stars of O spectral type. The X-ray emission is believed to originate from hot plasma heated by shocks and instabilities in the strong stellar winds of these stars (Pallavicini 1989). Despite their very high effective temperature, O type sub-dwarfs are much fainter than main-sequence stars, and, up to now, X-ray observations provided only an upper limit of ∼10³¹ erg s⁻¹ for the sdO star BD −3°2179 (D'Antona et al. 1983).

The X-rays we detected during the eclipse are well described with a bremsstrahlung of temperature 0.5 keV and luminosity 2 × 10³⁰ erg s⁻¹. Given the bolometric luminosity of HD 49798 L_bol = 10^3.9 L_☉, this corresponds to a ratio L_X/L_bol = 7 × 10⁻⁸. Thus, both from the spectral and luminosity point of view, the observed X-ray flux is consistent with emission from HD 49798. To our knowledge, this could be the first detection of X-ray emission from a sdO star. Further observations of this system, as well as searches for X-ray emission from other members of this class, can provide interesting information on the properties of X-ray production in stellar winds spanning a large range of parameters from hot sub-dwarfs to supergiants.

HD 49798/RX J0648.0–4418 is the only known X-ray binary containing a massive white dwarf accreting from a hot sub-dwarf. In principle, other systems of this kind, with a luminosity of ∼10³² erg s⁻¹, could be detected even at distances of several kpc by the current X-ray satellites. However, due to the very soft spectrum, the observed flux is very sensitive on the amount of interstellar absorption. For example, a column density of 3 × 10²¹ cm⁻² would reduce by one order of magnitude the observed EPIC count rate of RX J0648.0–4418, making it difficult to identify other sources of this kind in the Galactic plane only through X-ray observations (RX J0648.0–4418 is at Galactic coordinates l = 254°, b = −19°).

On the other hand, also from the point of view of the optical properties, this system seems quite uncommon. In fact, HD 49798 is rather atypical compared to other sdO stars (Heber 2009), which on average are less massive (M ∼ 0.5 M_☉), less luminous, and have a higher surface gravity. This is probably a result of its evolution in a close binary in which a common envelope phase occurred, as clearly indicated by several arguments. The low H abundance (X_H ∼ 0.19; Kudritzki & Simon 1978) suggests that HD 49798 is the stripped core of an initially much more massive and larger star. Also the high abundance of N and low abundance of C confirm that its present surface layers once belonged to the outer part of the hydrogen-burning core of a massive star. Slightly different evolutionary scenarios have been proposed to explain this system.

One possibility is that HD 49798 descended from a fairly massive progenitor that began to fill its Roche lobe before helium ignition (Case B evolution) and is currently in a core He-burning phase (Kudritzki & Simon 1978; Iben & Tutukov 1993). In this case, the mass of the He star is related to that of its progenitor by M_He = 0.043M^1.67 (Iben & Tutukov 1985), implying that HD 49798 was originally a star of ∼8 M_☉. Two other possibilities have been considered by Bisscheroux et al. (1997). The first one is that HD 49798 consist of a degenerate CO core, surrounded by a helium envelope with a He-burning shell at its base. In this scenario, it would be the core of a progenitor of ∼5 M_☉ which lost mass during a common envelope event when it was on the early asymptotic giant branch (AGB). Alternatively, HD 49798 could be burning hydrogen in a shell, being the remnant of a star which shed its envelope while on the thermally pulsing AGB. However, this possibility is disfavored since it predicts a mass of only ∼0.6 M_☉ for the sub-dwarf.

At the observed mass transfer rate, with only a small (<10⁻³) fraction of the total mass lost in the subdwarf's wind accreted, the white dwarf mass is increasing very slowly. At the end of the current He-burning phase, HD 49798 will expand and fill the Roche lobe, giving rise to a much higher mass transfer rate. This will produce a higher X-ray luminosity and possibly bright outbursts like the one seen in the helium nova V445 Puppis (Woudt et al. 2009), which could be the descendent of a system similar to HD 49798/RX J0648.0–4418. The amount of accreted mass, ∼0.3–0.5 M_☉ (see Figure 1 of Iben & Tutukov 1994), is sufficient to bring the, already quite heavy, white dwarf above the Chandrasekhar limit. However, the fraction of mass that is effectively retained depends on the rate and composition of the accreting matter, on the mass, composition, and temperature of the white dwarf, and on the poorly known relevance of other factors, such as, e.g., rotation, magnetic fields, and wind outflows. Thus, the fate of RX J0648.0–4418 is uncertain.

Evolutionary computations for RX J0648.0–4418 have been recently performed (Wang & Han 2010) assuming the mass accumulation efficiency that takes into account the wind mass loss triggered by the He-shell flashes (Kato & Hachisu 2004). These indicate that RX J0648.0–4418 will reach a mass of 1.4 M_☉ after only a few 10⁴ years of Roche-lobe overflow, during which ∼(5–6) × 10⁻⁶ M_☉ yr⁻¹ of He-rich matter are steadily converted to C and O, while the unburned matter is ejected by the system trough an optically thick wind. Wang & Han (2010) assumed that RX J0648.0–4418 is a CO white dwarf that will explode as a type Ia supernova when the Chandrasekhar limit is reached. The fast rotation can increase the mass stability limit above the standard value for non-rotating stars, thus systems like RX J0648.0–4418 could be the progenitor of overluminous type Ia supernovae. While an ONe composition seems more likely for RX J0648.0–4418 in view of its large mass, also in this case there are some uncertainties and other factors that might play a crucial role. For example, it has been pointed out that massive white dwarfs can have a CO composition as a result of rapid rotation (Dominguez et al. 1993, 1996).

If RX J0648.0–4418 is an ONe white dwarf, an accretion induced collapse (AIC) might occur, instead than a type Ia supernova explosion, leading to the formation of a millisecond pulsar (MSP). There is strong evidence that MSPs are old neutron stars which underwent magnetic field decay and spin-up due to accretion of mass and angular momentum in low mass X-ray binary systems (LMXRBs; Bhattacharya & van den Heuvel 1991). This evolutionary scenario is supported by the observation of short period X-ray pulsations in the persistent and/or burst emission of several LMXRBs, by the small eccentricity of MSPs with white dwarf companions in the Galactic disk,⁸ and by the recent discovery of PSR J1023+0038 (Archibald et al. 2009), the long sought missing-link between the LMXRB and MSP phases. On the other hand, the formation of an MSP by AIC of accreting white dwarfs (e.g., Bailyn & Grindlay 1990) has been considered to reconcile the apparent discrepancy in the birthrate of LMXRBs and MSPs, and it cannot be excluded that a fraction of the MSP population is formed through AIC. The discovery of PSR J1903+0327, a Galactic plane MSP with a high eccentricity orbit (e = 0.44) around a solar mass main-sequence star (Champion et al. 2008; Freire et al. 2011), gives further evidence for the possibility of directly forming young MSP. The high spin rate and low magnetic field of RX J0648.0–4418 make it an ideal candidate for the direct formation of an MSP though AIC.

HD 49798/RX J0648.0–4418 is a unique system, but its X-ray properties are similar in several respects to those seen in white dwarfs of the polar and intermediate polar classes. Its spectrum consists of a very soft, strongly pulsed, thermal-like emission plus a harder component dominating above ∼1 keV. The two components have different pulse profiles and show small uncorrelated variations on long timescales. These similarities suggest that the X-ray emission properties depend mainly on the physical conditions near the white dwarf, with little influence of the large scale accretion scenario (wind accretion in RX J0648.0–4418 wrt Roche-lobe overflow in cataclysmic variables).

With a dynamically measured mass of 1.28 ± 0.05 M_☉, RX J0648.0–4418 is one of the most massive white dwarf currently known and the one with the shortest spin period, only a factor ∼5 larger than the break-up limit. It is also the only known white dwarf accreting from the wind of a hot sub-dwarf star. This system is the outcome of a common envelope evolution, most likely of an original pair of stars with mass of ∼8–10 M_☉. Its future evolution might lead to the formation of a neutron star through AIC or to the explosion of a type Ia supernova.

Both possibilities have interesting implications. AIC of a fast-spinning and low-magnetic-field white dwarf could be a promising scenario for the direct formation of non-recycled millisecond pulsars. In the second case, being the result of the evolution of relatively massive stars (∼8–9 M_☉), this would be a type Ia supernova formation channel with a short delay time. Future X-ray and optical observations can lead to more accurate determination of the system parameters, providing crucial information to investigate in more detail the past and future evolution of this system. For example, further observations of the orbital phases including the eclipse can give a more precise measurement of its duration, a better characterization of the X-ray emission from HD 49798, as well as some information on the structure of its wind by measuring possible variations of N_H. High-resolution spectroscopy of the soft X-ray spectral component and deeper observations above a few keV are needed to better compare the properties of this peculiar X-ray binary with those of other accreting white dwarfs, and, in particular, to better constrain the magnetic field geometry and intensity.

We thank Alak Ray, Ulrich Heber, and Stephan Geier for interesting discussions. We acknowledge financial contribution from the agreements ASI-INAF I/009/10/0 and I/032/10/0. P.E. acknowledges financial support from the Autonomous Region of Sardinia through a research grant under the program PO Sardegna FSE 2007–2013, L.R. 7/2007 "Promoting scientific research and innovation technology in Sardinia."