DEM L241, A SUPERNOVA REMNANT CONTAINING A HIGH-MASS X-RAY BINARY

The compact X-ray source appears point-like and has a luminosity ∼2 × 10³⁵ erg s⁻¹ at the Large Magellanic Cloud (LMC) distance of 50 kpc. No extended PWN is visible in the immediate vicinity and the diffuse emission close to the point source is thermal. As shown in Figure 4, the Chandra telescope point-spread function (PSF) accounts very well for the appearance of the source. The source X-ray spectrum is a power law and it is appreciably harder than that from the diffuse parts of the remnant.

Zoom In Zoom Out Reset image size

It is very unlikely that the object is a background active galactic nucleus (AGN). There are an average of 0.2 AGN deg⁻² at this flux level (Kim et al. 2007) so the chance of finding one even within the 2' × 5' remnant is only ≈6 × 10⁻⁴. Furthermore, the X-ray spectral index is −1.3 which is harder than the mean index of −1.9 (with dispersion of 0.5) for radio-quiet QSOs.

The X-ray spectrum is also too hard to be that of a foreground normal star or even that of the optical counterpart O5III(f) star. This star itself is expected to be an X-ray source but with a soft spectrum and luminosity only 1% of the flux we observe here (Chlebowski 1989). Colliding winds in a binary system could be more energetic but the luminosity of observed systems (e.g., W140; Pollock et al. 2005) is still an order of magnitude less than that of this object. We conclude that the source is neither a foreground nor a background object and is probably associated with the remnant.

The Chandra X-ray spectrum is shown in Figure 5. Data are from a circular region with radius 4'' centered on the source and background was taken from a surrounding 5'' wide annulus. The count rate was 0.057 s⁻¹ which, in the ACIS instrument which integrates for 3.2 s, leads to a 3% chance of two events being recorded as one. This pileup is not severe but enough to distort the spectrum. We therefore incorporated a pileup correction, jdpileup, in the Chandra CIAO/SHERPA spectral analysis. Results are listed in Table 1. The best fit was

a power law with index −1.28 ± 0.08 and is shown in the figure. Without the pileup correction the best-fit index was −1.10 ± 0.06. (The uncertainties given are 1σ.) The index measured with XMM, free of pileup but with larger background subtraction, was −1.57± 0.05 (Bamba et al. 2006). The lack of agreement could be due to an inadequate pileup correction, larger uncertainties than quoted, or a variable source. The Chandra- and XMM-measured luminosities are the same. The luminosity range in Table 1 is due to the observed variability. We also fit an xsdiskbb model to the data since this is appropriate for accretion-powered sources and there is reason to believe this object may be part of a binary system. The fit is reasonable but the absorption is too low. If this is an accretion-powered system, a more elaborate model is needed.

Zoom In Zoom Out Reset image size

The bright optical counterpart, a V = 13.5 O5III(f) star, is easily visible within the Head of the remnant in Figure 1. This had been noted as a possible counterpart to the Einstein source CAL 60 (from the catalog of Long et al. 1981) by Crampton et al. (1985) who published a finding chart and spectrum. Since the source CAL 60 includes the diffuse emission as well as the point source, we will refer to the point source as CXOU 053600.0-673507. Our X-ray position is 22 S of the XMM position and <06 from this O star. X-ray positions are listed in Table 2 with uncertainty the larger of that from counting statistics or the difference between positions from the two halves of our observation. The systematic error associated with Chandra positions is 06 at 90% confidence. The star HD 269810 is also in the field and our X-ray position for this star is 05 from the optical and Two Micron All Sky Survey position, so the registration of the Chandra field is good and we adopt 06 as the uncertainty of the X-ray position.

The source is variable. The count rate in the second part of the observation increased 25% over that obtained in the first part. No variability was seen on timescales shorter than ∼10⁴ s although the observation was not sensitive to periods shorter than 6 s or to pulsed fractions <20%. This object cannot be an unresolved PWN as was reasonably inferred from the XMM observation by Bamba et al. (2006).

Since in ≈12 hr the Chandra flux varied 25% we searched for variability in past observations. These were all of limited sensitivity but could show variability extremes. Table 3 lists imaging observations which have detected this remnant. We inspected the archival Einstein and ROSAT fields and, in all cases could distinguish the Head and Tail of the remnant but could not separate the point source in the Head from the diffuse emission. The diffuse and compact parts of the remnant are only well separated in XMM and Chandra observations. Indeed, the spatial resolution and low count rate in Einstein and ROSAT observations make it impossible to do this if the point source is no stronger than in our observation. In these past observations only 10%–20% of the counts in the Head should be from the point source and total counts from the Head range from ∼30 to ∼200 with the higher rates from detectors with the lower spatial resolution. In all cases the counting rate was about the same from the Head and Tail regions as it is in the Chandra observation, so this is reassuring.

In order to compare observations of the diffuse source, we have converted count rates to flux levels with PIMMS, assuming a thermal spectrum with kT = 0.5 keV and N_H = 2 × 10²¹ cm⁻². To avoid large absorption corrections at energies below 0.5 keV, we calculated the flux in the range 0.5–5.0 keV and these results are summarized in Table 3. Uncertainties are counting statistics only. The luminosity, L_X, of the compact source is calculated for the Chandra range 0.3–10 keV. The agreement of past observations is good, as indeed it should be for the diffuse emission. Any variation of the point source less than a factor of ∼3 is undetectable in the Einstein and ROSAT data. The past observations of this remnant therefore show that there have been no major outbursts at these times.

We searched for regular pulsations in the Chandra data and found nothing significant. Since the ACIS instrument integrates for 3.2 s, the search was only valid for periods between 6.4 s and 6 hr and the signal of 2500 counts limited the search to pulsed fraction above ≈20%. Bamba et al. (2006) searched XMM and ASCA data in the period range 0.15–500 s at about this same sensitivity with null result. A short period and/or a ≈10% pulse fraction is quite possible.

We have measured the optical spectrum of this star with the SAAO 1.9 m telescope at Sutherland, South Africa, in 2011 October and 2012 April. The spectrum from 2011 October 24 is shown in Figure 6. This spectrum is almost identical to that of Crampton et al. (1985), indicating no long-term spectral variability. The velocity we measure from the Balmer (mean of Hβ and Hγ) absorption lines is 324 ± 7 km s⁻¹ so it is clearly in the LMC and an association with the remnant is possible. Further spectra on 2012 April 12 and 15 yielded velocities of 301 and 296 km s⁻¹, respectively (and with similar uncertainties). These values are consistent with the Crampton et al. (1985) mean value of 304 km s⁻¹, as well as the variability by 34 km s⁻¹ that was noted then. The small velocity change noted between our April spectra are indicative of a long-period system, likely tens of days or more.

Zoom In Zoom Out Reset image size

IR photometry in the J, H, and K_S wavebands of the optical counterpart was obtained during the month of 2011 November using the 1.4 m IRSF telescope at SAAO's Sutherland observing station. These showed no significant variability to within 0.1 mag.

Moderate-resolution radio observations of this area were made during surveys at 8.6 and 4.8 GHz (Dickel et al. 2005), at 1.4 GHz (Hughes et al. 2007), and at 843 MHz (Mauch et al. 2003). The best resolution is at 8.6 GHz with a half-power beam width of 20''. An overlay of the X-ray image on the 8.6 GHz radio contours is shown in Figure 7. The X-ray emission has been smoothed with a 5 arcsec Gaussian. The H ii region is clearly delineated by the contours with its two brightest parts N59A on the west and N59B on the east as designated by Henize (1956). Because of the overlap with the bright H ii region it is difficult to separate a possible contribution of the SNR to the radio emission. The rms noise should give a 5σ limit of about 1 mJy beam⁻¹ but the contamination could make the limit a few mJy beam⁻¹. The fact that no radio emission appears to match with the X-ray emission is unusual. Most SNRs with significant X-ray emission are readily visible at radio wavelengths as well. Apparently the surrounding H ii region complex masks the non-thermal emission from the SNR.

Zoom In Zoom Out Reset image size

With the current resolution, any point radio source would be difficult to distinguish from the somewhat irregular background, but the absence of any feature at the position of the point source suggests that it would be less than a few mJy. For comparison, the X-ray only pulsar 0537-6910 in N157B is less than .06 mJy at 22 cm (Crawford et al. 1998) and PSR 0540-693 in the composite SNR of the same name is 0.4 mJy at 640 MHz. N206 is a shell-like radio SNR with a radio jet-like feature with an X-ray point source at the tip, but no radio point source or pulsar at <.02 mJy (Williams et al. 2005). Thus, the radio limit on the point X-ray source is not very significant or unexpected.

Another way to distinguish between the thermal emission from H ii regions and the non-thermal synchrotron emission from SNRs is the radio spectrum. H ii regions have a nearly flat spectrum with a slope, or spectral index, near 0 while PWNe have a modest decrease in brightness as frequency increases with an index of about −0.2 to −0.3. Shell SNRs have average spectral index slopes of −0.5 with young remnants being steeper and old ones flatter. Figure 8 shows a four-wavelength radio spectral index map of the entire region. Production of this image required convolution of all the data to the 48.5 × 45.0 arcsec resolution of the 36 cm image. Obviously, part of the X-ray Tail is missing in the radio but other areas could contain some synchrotron emission from a PWN or very old SNR in addition to the thermal emission from the H ii regions. From the spectral index map, we cannot distinguish the separation of various components.

Zoom In Zoom Out Reset image size

For comparison with optical wavelengths, we used images from the Magellanic Cloud Emission Line Survey (MCELS; Smith & the MCELS Team 1999). These images were taken at the UM/CTIO Curtis Schmidt telescope at Cerro Tololo Inter-American Observatory (CTIO). The detector, a Tek 2048 × 2048 CCD with 24 μm pixels, gave a scale of 2.3 arcsec pixel⁻¹ and a resulting angular resolution of approximately 2.6 arcsec. The narrow-band images were taken with filters centered on the [O iii] (5007 Å, FWHM = 40 Å), Hα (6563 Å, FWHM = 30 Å), and [S ii] (6724 Å, FWHM = 50 Å) emission lines. The image shown in Figure 1 has been reduced using the IRAF software package for bias subtraction and flat-field correction, and the astrometry was derived based on the HST Guide Star and USNO-A catalogs. This image, however, has not been flux-calibrated or continuum-subtracted.

Because of the presence of the bright H ii region, a good way to depict the SNR is to show the [S ii]/Hα ratio. The [S ii] doublet at 671.7 and 673.1 nm is enhanced by shocked material so any region with an [S ii]/Hα ratio greater than about 0.5 is considered an SNR. In Figure 9, we show the ratio superimposed on contours of the X-ray emission. There is clearly a good correlation except on the eastern edge where the bright H ii-region component N59B overwhelms emission from the remnant.

Zoom In Zoom Out Reset image size

No X-rays are observed from the outer shock traced by the [S ii] filaments. This implies that the temperature of hot gas in the shock is <8 × 10⁵ K. The Sedov model then can be used to estimate an age limit. Assuming E₀ = 10⁵¹ erg and using the densities given in Table 5, we derive an age of >(5–7) × 10⁴ yr.

There are two major parts to this object. The X-ray spectra of the Head and the Tail are sufficiently similar to indicate that they are parts of the same remnant rather than two separate objects and we will so treat them. Data were extracted from the three regions outlined in Figure 3: the Head or S region, the bright NW strip which accounts for 80% of the emission from the Tail, and an NE strip where there is fainter emission which might be indicative of a shock. After background subtraction the number of events in these regions was: S, 7960; NW, 4640; and NE, 1020. The first two regions contain enough events to support an interesting spectral analysis but the faint NE region yields less information. Background was taken from large source-free regions northeast and southwest of the remnant and was 35% of the remnant signal in the S region, 85% of the NW signal, and 200% of the NE signal. The spectra of these regions are shown in Figure 10. The visually distinct spectra of the S and NW regions can be explained almost entirely by differences in absorption of the emitted X-rays. This can be seen by inspection of the softest emission and is greatest in the region between the Head and Tail. The visual gap in the X-ray image which separates the Head from the Tail is most prominent in the softest energy band and is caused by absorbing material. This is not surprising considering the complex surrounding environment. The spectra of the bright structure just south of the O star and the surrounding outer part of the Head were examined separately and were, within uncertainties, the same. The bright structure is therefore plausibly explained as a density enhancement within the SNR or an extension of the remnant structure along the line of sight.

Zoom In Zoom Out Reset image size

We used the xsvpshock model, which accounts for non-equilibrium ionization in a one-dimensional shock, to fit the NW region which shows least absorption. Metal abundances were fixed at 0.3 solar and then allowed to vary to fit the data. It was necessary to vary only O, Ne, Mg, and Fe. We then fit the S and NE regions with the same spectrum allowing only the interstellar medium (ISM) absorption and timescale parameters to vary. Best-fit parameters are listed in Table 4. The NW fit is excellent, the S fit is good and the NE fit is as good as any that can be achieved with the limited number of events. These fits are shown in Figure 10. The difference between fits of the S and NW regions is entirely due to different absorption of the O, Ne, and Fe emission lines. The fit to the NE region requires a lesser value of the ionization timescale (τ_μ). Since the NE region comprises the outer part of the expanding remnant, this is expected.

The formal 1σ uncertainties in the S and NW spectral parameters due to counting statistics are ≈15% for the absorption (N_H), <10% for the temperature (kT), ∼40% for the abundances, and ∼50% for the ionization timescale (τ_μ). The statistical uncertainties of the data points, however, allow acceptable fits over a large range of parameters, e.g., the temperature can be varied by ±50% and reasonable fits can be obtained with small changes in abundances (although the Fe abundance is more sensitive than the lighter elements). Table 4 lists CIAO/SHERPA best-fit spectral parameters for the xsvpshock and the equilibrium xsvmekal models.

The variability excludes an unresolved PWN such as that surrounding the Vela Pulsar. Furthermore, the Vela PWN has L_x = 2 × 10³³ erg s⁻¹, considerably less than observed here. The compact object is too bright to be a Cas A-type central compact object (CCO, with L_x = 10³³–10³⁴ erg s⁻¹). The luminosity is compatible with an anomalous X-ray pulsar (AXP, with L_x = 10³⁵–10³⁶ erg s⁻¹) but the spectrum is harder than that of most AXP (McGill Pulsar Group 2012). The hard spectrum suggests an accretion-powered binary which, comprising an O star and a neutron star, would be a high-mass X-ray binary (HMXB). This is unexpected. Only two other X-ray binaries have been discovered within observable SNRs—SS 433, with a B or A giant primary star, within the galactic remnant W50 and recently, a Be X-ray pulsar system, SXP 1062, within a faint Small Magellanic Cloud (SMC) remnant (Hénault-Brunet et al. 2011).

Although we have no evidence other than location that the O star under discussion and the X-ray source form a binary system, we suspect that the object within DEM L241 is the third HMXB to be found within an SNR. The compact object, in this HMXB, if a neutron star, would accrete material from the O-star wind and changes in luminosity would be caused by changes in the density of the wind. Scaling from the HMXB Cen X-3, the moderate luminosity and lack of observed outbursts imply (very roughly) a separation of ∼1 AU and an orbital period of weeks. At the moment this is speculation but the velocity of the O star has indeed been observed to vary. No regular pulsations in the X-rays have been found but the observations were not sensitive to pulsed fractions below ∼20%.

The fact that the supernova (SN) precursor must have been more massive than the O-star companion implies a precursor mass of 25, perhaps 40 M_☉, thought to be the mass necessary for collapse to a black hole. It is therefore possible that the compact object, which we have assumed to be a neutron star, could actually be a black hole. The X-ray luminosity, however, is unusual for a black hole. Known black hole binaries are either bright with L_x ⩾ 10³⁷ erg s⁻¹ or are transient systems in quiescence with L_X < 10³³ erg s⁻¹ (Remillard & McClintock 2006).

However, the HMXB in DEM L241 has a luminosity (2 × 10³⁵ erg s⁻¹), intermediate between these extremes. Interestingly, this is similar to the luminosity of SS433. While there is as yet no direct dynamical measurement of the compact object mass in SS433, all the indirect evidence strongly supports it being a ∼15 M_☉ black hole accreting at extremely high mass transfer rates from a highly evolved supergiant primary (see, e.g., Blundell et al. 2008). The reason that its X-ray luminosity is so low (∼10³⁵ erg s⁻¹) is that it is observed at high inclination (it is eclipsing), and hence we only receive X-rays scattered from the surrounding material into our line of sight. That is, it is an "accretion disk corona" system.

As discussed earlier, there is marginal evidence that L_x of the HMXB in DEM L241 does not vary much, but we consider it unlikely that it too would be a high-inclination system. Consequently, a black hole is certainly possible but we have no strong indication that this might be the case. A measure of the velocity variation of the O5III(f) companion would provide a test of this hypothesis. This remnant could become the first example of a remnant known to be created by stellar collapse to a black hole.

Throughout this discussion it is assumed that the precursor star was originally bound to the O5III(f) star. After the SN, the compact remnant of this remained bound to the O star to form the present HMXB. The explosion itself probably determined the size and temperature of the remnant. The binary environment of the precursor may have shaped the remnant morphology.

The X-ray morphology of the SNR looks like that of a PWN with an extended elliptical structure showing no obvious shell and a point X-ray source near one end. However, all diffuse emission, including that close to the point source, is well fit with a thermal spectrum. There is no evidence for a power-law component although a 10% contribution cannot be excluded. The X-ray extent along the major axis of about 80 pc makes the object one of the larger of the known SNR. The SN may have exploded between the two major H ii regions and the expansion was impeded in the east–west direction. However, although the SNR and the H ii region complex do appear coincident in the sky there is no indication in their structures that they have interacted so they could be at different distances along the line of sight.

The vpshock fits indicate an age of τ_μ/n_e ∼ 7 × 10¹¹/0.14–0.27 s = (0.8–1.6) × 10⁵ yr, about twice the lower limit estimated using the Sedov model. This age is old for a radio SNR as the expanding shock will have slowed sufficiently to reduce the shock acceleration of relativistic particles that create the radio synchrotron emission. Thus, the age can explain both the relatively flat radio spectral index and the weak radio emission in general.

Since the plasma is close to equilibrium we also fit the X-ray spectra with the xsvmekal equilibrium model. Table 4 gives the best-fit result of this model to the S and NW spectra. In this case, the two spectra have been fit separately. The S region, the Head surrounding the compact object requires somewhat higher temperature and abundances than the NW Tail region. Absorption is still higher in the S than in the NW region.

Although there is considerable uncertainty in the spectral fit parameters and in the volumes used for the three regions, enhanced abundances and derived densities indicate that we are seeing radiation from SN debris above that from the normal ISM in the Magellanic Clouds. Table 5 gives the plasma densities and masses calculated from the vpshock results of Table 4. The excess of O, Ne, and Mg above that expected is 2.5, 0.8, and 0.15 M_☉, indicating the explosion of a massive star. Woosley & Weaver (1995) have calculated the masses of elements ejected from explosions of massive stars. This amount of O can come from a precursor with mass >25 M_☉ and the Ne from a precursor with mass >40 M_☉. Since there is variation in the different Woosley & Weaver models and uncertainty in our measured abundances, we can only conclude that the precursor was massive. Since it evolved first, it should have been more massive than the present O5III(f) star, so this is expected. Martins et al. (2005) give the mass of an isolated O5III(f) star as about 40 M_☉. However, Rappaport & Joss (1983) show evidence that the companion stars in HMXBs are "undermassive for their luminosity," so a precursor mass estimate of >25 M_☉ is more conservative.

The remnant parameters derived here differ somewhat from those obtained from the XMM observation by Bamba et al. (2006). Our X-ray flux and luminosity is 30% higher because we use a different energy band. Our total mass is 30% lower because we assume different volumes for the emitting regions. Bamba et al. (2006) assume ellipsoidal volumes for the Head and Tail whereas we use cylindrical volumes with different dimensions. The Chandra best-fit spectra have higher temperature and greater enrichment. Remarkably, the total thermal energy content is the same for the two analyses.

The elongated structure may be due to a "blowout" to a lower density region in the north. If so, this was a major event. The Tail now has twice the volume of the Head and contains 40% of the thermal energy of the hot gas. On the other hand, perhaps the unusual elongation of the remnant is associated with the unusual central source which is quite likely an HMXB. The massive binary might have produced a disk of material which channeled the ejecta into two lobes. Or, the precursor star itself could have been a Wolf–Rayet star with strong wind which shaped a cavity now filled by the remnant. Or, rotation of the precursor could have produced an asymmetrical explosion.

There is an interesting parallel with the morphology of the Galactic remnant W50 which is also large and elongated with dimension 95 × 200 pc (at a distance of 5.5 kpc; Lockman et al. 2007). The binary SS 433 lies at the center of W50 and produces opposing relativistic jets which have inflated and elongated the remnant in the EW direction. The optical manifestation of these jets is spectacular (Margon et al. 1980). SS 433 is an unresolved X-ray source with luminosity 3 × 10³⁵ erg s⁻¹, the same as that of the DEM L241 source.

DEM L241, with dimension 32 × 77 pc, is also elongated but only about one-third the size of W50. From the O star to the northern end of DEM L241 is 57 pc, about half the length of 120 pc for the longer arm of W50. Perhaps the Tail of DEM L241 was inflated by a single northern jet and the structure south of the O star is due to energy deposited by the oppositely directed jet. Since we see no other evidence for jets, this probably happened in the past and is no longer operating. The lesser size of DEM L241 also indicates a weaker jet and/or a younger remnant.

Despite intensive searches, no other examples of SS 433/W50 systems have been found. Here in the LMC is a system with similarities that suggest further investigation might be worthwhile.

As a matter of interest, we list here the compact objects now known to be associated with Magellanic Cloud SNR. The LMC and SMC contain ∼60 remnants. About 45 of these are known to emit X-rays (Smith 1999; Williams et al. 2000; Inoue et al. 1983; Filipovic et al. 1998). So far, 34 have been observed by Chandra and the results may be viewed in the online catalog Seward et al. (2012b). Structure of the brighter remnants has been well resolved with moderate observing times and much detail is visible. One can see shock waves, interior structure, clumpy material, central objects, and PWNe. Chandra data have been used to compile Table 6, an inventory of X-ray-emitting compact objects and PWNe within Magellanic Cloud SNRs. Diffuse thermal radiation from the shells or interiors has not been included. Faint unresolved sources, for which no pulsations have been detected, could contain a contribution from a small PWN so the luminosities are listed under the header "PSR and PWN." All luminosities have been corrected to the 0.5–10 keV range. Ten of the 35 Magellanic Cloud SNRs observed by Chandra have PWNe or compact objects. The luminosity range is a factor of 10⁴ and there is great variety of type: one soft gamma-ray repeater (SGR), two rotation-powered PSR/PWNe, one CCO, two PWNe with no associated compact object, two X-ray binaries (XRB) (including this observation), and two unclassified faint objects.