THE QUIESCENT COUNTERPART OF THE PECULIAR X-RAY BURSTER SAX J2224.9+5421

Perhaps the most tantalizing burst-only source is SAX J2224.9+5421 (J2224 hereafter). It was detected with the BeppoSAX Wide-Field Camera (WFC) as a ≃10 s long X-ray flash on 1999 November 6 (Gandolfi 1999). Initially dubbed GRB 991106, multi-wavelength follow-up observations failed to detect a fading afterglow and cast doubt on a gamma-ray burst (GRB) nature (e.g., Antonelli et al. 1999; Gandolfi et al. 1999; Frail 1999; Jensen et al. 1999; Gorosabel et al. 1999). Cornelisse et al. (2002b) demonstrated that the properties of the event were consistent with a thermonuclear X-ray burst, which would identify J2224 as a new neutron star LMXB. A source distance of D ≲ 7.1 kpc was inferred by assuming that the X-ray burst peak did not exceed the Eddington limit (L_edd = 2 × 10³⁸ erg s⁻¹).

No accretion emission could be detected with the WFC around the time of the X-ray burst, implying a 2–28 keV luminosity of L_X ≲ 2 × 10³⁶(D/7.1 kpc)² erg s⁻¹ (Cornelisse et al. 2002b). Rapid follow-up observations with the BeppoSAX Medium-Energy Concentrator Spectrometer (MECS), performed ≃8 hr after the X-ray burst detection, revealed only one source within the WFC error circle that was detected at a 0.5–10 keV luminosity of L_X  ≃  8 × 10³²(D/7.1 kpc)² erg s⁻¹ (Antonelli et al. 1999). This raises the question whether the X-ray burst was ignited when the neutron star was accreting at a very low level, or whether it exhibited a faint, undetected accretion outburst that had ceased within 8 hr of the X-ray burst detection.

In this work, we present a deep XMM-Newton observation to search for an X-ray counterpart of J2224, and to find clues to its nature and unusual behavior.

The field around J2224 was observed with XMM-Newton for ≃51 ks on 2013 May 29 UT 06:04–20:24 (ID 0720880101). X-ray data was obtained with the European Photon Imaging Camera (EPIC), which consists of two Metal Oxide Semiconductor (MOS) detectors (Turner et al. 2001) and one PN camera (Strüder et al. 2001). All instruments were operated in the full frame imaging mode with the medium optical blocking filter applied. We reduced and analyzed the data using the Science Analysis software (version 11.0). The original data files were reprocessed using the tasks emproc and epproc. The observation was free from background flaring.

Figure 1 (left) displays the composite X-ray image of all three detectors. There are eight faint X-ray sources within the BeppoSAX/WFC error circle that were detected with all three EPIC instruments. The source positions (determined using the task edetect$\_$chain) are listed in Table 1. Sources 1 and 3 were previously detected at similar intensities in Swift/XRT data of the field (named SAX J2225-1 and SAX J2225-2, respectively; Campana 2009).

Zoom In Zoom Out Reset image size

Table 1. Positions and Basic Properties of Detected X-Ray Sources

Source	R.A.	Decl.	Error	MOS Count Rate	PN Count Rate	PN Hardness Ratio
	(J2000)	(J2000)	('')	(10−3 counts s−1)	(10−2 counts s−1)
1	22 24 52.94	+54 22 38.2	1.6	3.9 ± 0.3	1.3 ± 0.1	0.94 ± 0.17
2	22 25 07.87	+54 21 30.2	1.7	3.2 ± 0.3	0.8 ± 0.1	0.73 ± 0.15
3	22 24 40.99	+54 24 54.4	1.6	3.8 ± 0.3	1.5 ± 0.1	$0.06^{+0.37}_{-0.06}$
4	22 24 49.65	+54 23 10.1	2.3	0.7 ± 0.2	0.17 ± 0.03	$0.22^{+0.58}_{-0.22}$
5	22 24 50.64	+54 22 05.1	2.4	0.6 ± 0.2	0.13 ± 0.03	1.22 ± 0.46
6	22 24 43.99	+54 20 59.9	2.2	0.6 ± 0.2	0.23 ± 0.05	1.54 ± 0.52
7	22 24 33.82	+54 21 48.9	2.2	0.7 ± 0.2	0.24 ± 0.04	1.99 ± 0.34
8	22 24 53.74	+54 24 54.9	2.1	0.6 ± 0.2	0.23 ± 0.04	2.29 ± 0.32

Notes. The quoted positional uncertainties are at 90% confidence level, and are a combination of the statistical error from the detection algorithm and an estimated systematic uncertainty of 15 (Watson et al. 2009). Count rates errors are at the 1σ level of confidence. We consider sources 1 and 4 as possible counterparts for J2224.

Download table as:  ASCIITypeset image

We determined the count rate for each source by employing the task eregionanalyse, using circular regions with radius of 20'' (400 pixels) centered on the source positions. For the background we used a circular region with a radius of 60'' placed on a source-free part of the CCD. Source and background spectra were extracted using especget, which also generated the response files. We co-added the two MOS spectra and their response files with epicspeccombine. Using grppha, the spectral data was grouped to contain a minimum of 15 photons per bin. We used XSpec (version 12.7; Arnaud 1996) to perform spectral fits in an energy range of 0.5–10 keV.

To obtain a rough characterization of the spectral shape, we determined the hardness ratio for each source using the PN data. For the present purpose, we defined this quantity as the ratio of counts in the 2–10 and 0.5–2 keV bands. This information is included in Table 1.

We explored their X-ray spectra using two models that are widely used for neutron star LMXBs at low luminosities: a simple power law (pegpwrlw in XSpec) and a neutron star atmosphere model (for which we chose nsatmos; Heinke et al. 2006). For the latter we always fixed the neutron star mass (M = 1.4 M_☉) and radius (R = 10 km) as well as the distance (D = 7.1 kpc) and the normalization (unity). The neutron star temperature was then the only free fit parameter. Thermal bolometric fluxes were estimated by extrapolating the nsatmos model fits to the 0.01–100 keV energy range. In all spectral fits, we accounted for interstellar absorption by including the tbabs model in XSpec, using the wilm abundances and vern cross-sections (Verner et al. 1996; Wilms et al. 2000).

Quasi-simultaneous optical and ultraviolet (UV) coverage was provided by the Optical Monitor (OM; Mason et al. 1996). The position of J2224 was observed with the OM operated in image mode. Exposures of 2000, 2460, and 2200 s were obtained using the V (λ_c  ≃  5407 Å), UVW1 (λ_c  ≃  2905 Å), and UVW2 (λ_c  ≃  2070 Å) filters, respectively. Reduction, source detection, and photometry were performed using the meta-task omichain.

None of the X-ray sources within the BeppoSAX uncertainty of J2224 are detected in the UVW2 image. However, sources 2 and 3 both have a potential counterpart in the V and UVW1 wavebands. This is illustrated in Figure 1 (right), which shows the OM V-band image.

Sources 5–8 have the hardest spectra among the detected sources, as indicated by their high hardness ratios (Table 1). Although a limited number of counts prohibits a detailed analysis, their spectra can be described by a power law with photon indices of Γ  ≃  1–3 and hydrogen column densities that are well in excess of the Galactic value in the direction of J2224 (N_H  ≃  4.6 × 10²¹ cm⁻²; Kalberla et al. 2005). This renders them unlikely counterparts to the X-ray burster. Their unabsorbed fluxes lie in the range of F_{X, unabs}  ≃  (1–10) × 10⁻¹⁴ erg cm⁻² s⁻¹ (0.5–10 keV).

It is more likely that these hard X-ray sources are (obscured) background active galactic nuclei (AGNs) or Galactic accreting white dwarfs (i.e., cataclysmic variables (CVs); see also Section 3.2). Both types of objects are abundant in the sky. For example, assuming a number density of ≃ 3 × 10⁻⁵ pc⁻³ (e.g., Schwope et al. 2002), we expect ≃ 3 CVs within the WFC error circle. Furthermore, down to a flux of ≃ 10⁻¹⁴ erg cm⁻² s⁻¹, about 3 × 10² AGNs are expected per square degree of sky (Brandt & Hasinger 2005). That implies that within the WFC uncertainty, roughly three AGNs with fluxes of ≳ 10⁻¹⁴ erg cm⁻² s⁻¹ are expected.

Sources 2 and 3 are the only ones with possible optical/UV associations. Based on the total number of V-band objects detected within the 17' × 17' OM field of view, we estimate the probability of a chance alignment within the 16 and 17 positional uncertainties of sources 2 and 3 to be ≃ 5 × 10⁻³ and ≃ 6 × 10⁻³, respectively.

For source 2 we measure AB magnitudes of V = 16.41 ± 0.01 mag and UVW1 = 17.45 ± 0.05 mag. The corresponding flux densities are f_V = (1.1 ± 0.1) × 10⁻¹⁵ erg cm⁻² s⁻¹ Å⁻¹ and f_UVW1 = (3.8 ± 0.2) × 10⁻¹⁶ erg cm⁻² s⁻¹ Å⁻¹ (uncorrected for reddening). The X-ray spectrum of source 2 cannot be described by a neutron star atmosphere model ($\chi _{\nu }^2 > 3$), but fits a power law with Γ = 2.0 ± 0.3 and N_H  ≃  (6.9 ± 2.3) × 10²¹ cm⁻². The resulting 0.5–10 keV unabsorbed flux is F_{X, unabs}  ≃  (4.9 ± 0.7) × 10⁻¹⁴ erg cm⁻² s⁻¹.

By multiplying the obtained V flux density with the width of the filter (≃684 Å), we obtain a ratio of the X-ray to optical flux of F_X/F_V ≃ 0.3 for source 2. Quiescent neutron star LMXBs typically have a much higher X-ray flux compared to that in the optical/UV band (a ratio of ≃10 or higher; e.g., Homer et al. 2001; Russell et al. 2006; Hynes & Robinson 2012). We therefore consider it unlikely that source 2 is the counterpart to J2224. It is plausible that it is a background AGN (see also Section 3.1). The X-ray spectra of these objects can typically be described by a Γ  ≃  2 power-law model, and their X-ray/optical flux ratios fall in a wide range of F_X/F_V  ≃  0.1–10.

Source 3, for which we measure V = 18.08 ± 0.07 mag (f_V = (2.3 ± 0.2) × 10⁻¹⁶ erg cm⁻² s⁻¹ Å⁻¹) and UVW1 = 19.10 ± 0.19 mag (f_UVW1 = (8.4 ± 1.4) × 10⁻¹⁷ erg cm⁻² s⁻¹ Å⁻¹), also seems too optically/UV bright for a quiescent neutron star LMXB. This is supported by the fact that neither a power law nor a neutron star atmosphere model, or a combination of the two, can satisfactory describe the X-ray spectrum of source 3 (all trials resulted in $\chi _{\nu }^2 > 1.5$ for 53–55 degrees of freedom (dof)). The data is instead better fitted with, e.g., a combination of a blackbody and an optically thin thermal plasma model (mekal; yielding $\chi _{\nu }^2 = 1.07$ for 52 dof), or a double mekal model ($\chi _{\nu }^2 = 1.14$ for 52 dof). Such models are often used to describe the spectra of (magnetic) CVs (e.g., Ramsay et al. 2004; Baskill et al. 2005). Indeed, we expect several CVs are present within the WFC error circle of J2224 (see Section 3.1).

Source 4 stands out by showing a relatively soft X-ray spectrum that can be described by an absorbed power-law model with $\Gamma =2.9^{+2.2}_{-1.0}$ and $N_{\rm H} = 2.8^{+4.5}_{-2.2}\times 10^{21} \,\mathrm{cm}^{-2}$ (Table 2). The obtained hydrogen column density is consistent with the Galactic value in the direction of J2224. A neutron star atmosphere model provides a good fit, yielding a neutron star temperature (as seen by an observer at infinity) of kT^∞ = 50  ±  4 eV and N_H = (4.8 ± 0.2) × 10²¹ cm⁻². The obtained 0.5–10 keV luminosity is L_X  ≃  5 × 10³¹(D/7.1 kpc)² erg s⁻¹. The X-ray spectrum of source 4 is shown in Figure 2 (right), and the fit results are summarized in Table 2.

Zoom In Zoom Out Reset image size

Table 2. Spectral Results of the Two Candidate X-Ray Counterparts

Parameter (Unit)/Model	Source 1		Source 4
	pow	pow+nsatmos	pow	nsatmos
NH (× 1021 cm−2)	6.7 ± 1.9	7.0 ± 2.7	$2.8^{+4.5}_{-2.2}$	4.8 ± 0.2
Γ	1.7 ± 0.2	1.7 ± 0.2	$2.9^{+2.2}_{-1.0}$
kT∞ (eV)		<61		50 ± 4
χ2/dof	52.4/58	52.3/57	10.9/16	16.0/17
FX, abs (× 10−14 erg cm−2 s−1)	6.0 ± 0.5	6.0 ± 0.6	0.4 ± 0.2	0.2 ± 0.1
FX, unabs (× 10−14 erg cm−2 s−1)	7.8 ± 0.8	8.0 ± 1.5	$0.6^{+1.8}_{-0.1}$	0.9 ± 0.4
LX (× 1032(D/7.1 kpc)2 erg s−1)	4.7 ± 0.5	4.8 ± 0.7	$0.4^{+1.0}_{-0.1}$	0.5 ± 0.2
Lth (× 1032(D/7.1 kpc)2 erg s−1)		<3.1		1.4 ± 0.5
Thermal contribution		<24%

Notes. Quoted errors represent 90% confidence levels. F_{X, abs} and F_{X, unabs} represent the 0.5–10 keV absorbed and unabsorbed flux, respectively. L_X denotes the 0.5–10 keV luminosity and L_th the 0.01–100 keV thermal luminosity. The bottom row gives the contribution of the thermal nsatmos component to the total unabsorbed 0.5–10 keV model flux.

Download table as:  ASCIITypeset image

The spectral shape of source 4 is typical for a quiescent neutron star LMXB. Combined with its location near the center of the BeppoSAX/WFC error circle, and lack of an optical/UV association, source 4 is a strong candidate counterpart to J2224. The intensity inferred from our XMM-Newton observation is well below the sensitivity of previous X-ray observations that covered the source region (e.g., with BeppoSAX and Swift; Antonelli et al. 1999; Campana 2009), which explains why this object was not found before.

A neutron star atmosphere model fails to describe the spectral data of source 1 ($\chi _{\nu }^2 > 3$). However, a power law provides an adequate fit, yielding Γ = 1.7 ± 0.2 and N_H  ≃  (6.7 ± 1.9) × 10²¹ cm⁻² (comparable to the Galactic value toward J2224). The inferred 0.5–10 keV luminosity is L_X  ≃  5 × 10³²(D/7.1 kpc)² erg s⁻¹. The X-ray spectrum of source 1 is shown in Figure 2 (left).

Although the spectral shape of source 1 is harder than that typically observed for quiescent neutron star LMXBs, there are a few sources with similar properties (see Section 4). Combined with its location near the center of the BeppoSAX/WFC error circle and lack of a bright optical/UV association, we cannot discard source 1 as a potential counterpart. This object was detected during Swift/XRT observations obtained in 2006–2008 at L_X  ≃  8 × 10³²(D/7.1 kpc)² erg s⁻¹ (Campana 2009), i.e., similar as during our XMM-Newton observation.

We probed the upper limits on any thermal emission by adding an nsatmos component to the best power-law fit. This yielded kT^∞ ≲ 61 eV and a fractional contribution of the thermal component to the total unabsorbed 0.5–10 keV model flux of ≲24%. The spectral results of source 1 are listed in Table 2.

J2224 is one of several sources for which BeppoSAX detected an X-ray flash without detectable accretion emission (e.g., Cornelisse et al. 2002b). Follow-up observations were performed with Chandra for some of these objects (several years later) and revealed weak candidate X-ray counterparts with L_X ≲ 5 × 10³² erg s⁻¹. This suggested that they were likely transient neutron star LMXBs that exhibited faint (L_X ≲ 10³⁶ erg s⁻¹) accretion outbursts when the X-ray bursts were ignited (Cornelisse et al. 2002a). Indeed, several of these sources were later detected during faint accretion outbursts (e.g., SAXJ1828.5–1037, SAXJ1753.5–2349, and SAXJ1806.5–2215; Hands et al. 2004; Degenaar & Wijnands 2008; Del Santo et al. 2010; Altamirano et al. 2011b).

The quality of the WFC data of J2224 was very low, imposing considerable uncertainty on the interpretation of this event (Cornelisse et al. 2002b). However, the similarities between the X-ray flash from J2224 and that of the other sources (i.e., duration, peak flux, overall spectral properties, indication of softening during the decay, and location in the Galactic plane) renders it likely that J2224 too is a bursting neutron star (Cornelisse et al. 2002b). Here we identify two possible quiescent neutron star LMXBs within the WFC error circle that provide support for this interpretation.

During our long XMM-Newton observation, we detected eight weak X-ray sources within the 32 BeppoSAX/WFC uncertainty of J2224. They had 0.5–10 keV luminosities in the range of L_X  ≃  (0.5–5) × 10³²(D/7.1 kpc)² erg s⁻¹, which is typical for transient neutron star LMXBs in quiescence (see, e.g., Jonker et al. 2004). However, based on the X-ray spectral properties and X-ray/optical luminosity ratios, we can discard six of these as potential counterparts to the X-ray burster, leaving only sources 1 and 4 as candidates.

Source 1 was detected at L_X  ≃  5 × 10³²(D/7.1 kpc)² erg s⁻¹, with a spectrum best described by a power-law model with an index of Γ = 1.7. Emission from a neutron star atmosphere with kT^∞ ≲ 61 eV could contribute up to ≃24% to the total unabsorbed 0.5–10 keV flux. Only a handful of quiescent neutron star LMXBs have similarly hard X-ray spectra that lack detectable thermal emission. The small LMXB subclass of AMXPs (which display coherent X-ray pulsations during accretion outbursts) exhibit Γ  ≃  1.5 power-law spectra with ≲40% attributed to thermal emission (e.g., SAX J1808.4–3658, Swift J1749.4–2807, and IGR J18245–2452; Campana et al. 2008; Heinke et al. 2009; Degenaar et al. 2012a; Linares et al. 2014). Their hard spectra are ascribed to their relatively strong magnetic field. The only non-pulsating neutron star with a very hard (Γ  ≃  1.7) quiescent spectrum is the LMXB and X-ray burster EXO 1745–248 (e.g., Wijnands et al. 2005a; Degenaar & Wijnands 2012). Its irregular quiescent properties have been interpreted in terms of ongoing low-level accretion. Thus, source 1 could be the counterpart of J2224, but it would imply that it is an unusual neutron star (perhaps with a relatively strong magnetic field or exhibiting low-level accretion).

Source 4, on the other hand, has a soft X-ray spectrum like the majority of quiescent neutron star LMXBs. This source was detected at L_X  ≃  5 × 10³¹(D/7.1 kpc)² erg s⁻¹, and its spectrum can be described by a neutron star atmosphere model with kT^∞  ≃  50 eV. This relatively low temperature is consistent with the expectations for a neutron star that exhibits faint accretion outbursts (Wijnands et al. 2013). We therefore tentatively identify source 4 as the counterpart of the X-ray burst detected with BeppoSAX in 1999.

The duration, repetition rate, and energetics of X-ray bursts depend on the conditions of the ignition layer, such as its thickness, temperature profile, and chemical abundances. These conditions are sensitive to the (local) accretion rate onto the neutron star so that different accretion regimes give rise to X-ray bursts with distinct properties (e.g., Fujimoto et al. 1981; Fushiki & Lamb 1987). X-ray bursts that ignite in a pure He layer are typically short (seconds), whereas the presence of H in the ignition layer prolongs the duration (minutes).

Upper limits obtained for the accretion emission of the burst-only sources suggests they occupy the lowest accretion regime (≲ 0.01 L_Edd; Cornelisse et al. 2002b). Classical burning theory prescribes that in this regime unstable burning of H should trigger a mixed He/H X-ray burst with a duration of ≃ 100 s (e.g., Fujimoto et al. 1981). This is much longer than the short (≃ 10 s) X-ray bursts detected from J2224 and other BeppoSAX burst-only sources. This apparent discrepancy was addressed by Peng et al. (2007), who showed that for the low inferred mass-accretion rates, sedimentation of heavy elements significantly reduces the amount of H in the ignition layer. This would alter the X-ray burst properties, possibly accounting for the observed short duration of the X-ray bursts (Peng et al. 2007).

The fact that no persistent X-ray emission was detected above L_X  ≃  8 × 10³²(D/7.1 kpc)² erg s⁻¹ within ≃8 hr after the X-ray burst detection raises the question whether J2224 may have been accreting at ≲ 1 × 10⁻⁵ L_Edd when the burst ignited. This is right at the boundary below which theoretical models predict that no X-ray bursts can occur (e.g., Fushiki & Lamb 1987). We consider this scenario unlikely, because the chance probability of detecting this ≃10 s event would be very low; for a typical ignition column depth of y  ≃  10⁸ g cm⁻² and a mass-accretion rate of $\dot{M}\,{\simeq}\, 10^{-13} \,M_{\odot }\, {\rm yr}^{-1}$, the time to accumulate enough material to produce this X-ray burst would be ≃8 yr.

Regardless whether source 1 or source 4 is the true counterpart, we consider it more likely that J2224 is a transient neutron star LMXB that exhibited an X-ray burst during the decay of a (short) faint accretion outburst. In particular, the behavior of J2224 is strikingly similar to that of the AMXP Swift J1749.4–2807. This source rapidly decayed to a level of L_X  ≃  1 × 10³³  erg s⁻¹ within a day after it was discovered through the detection of an X-ray burst (Wijnands et al. 2009). Its behavior remained a puzzle until it exhibited a ≃2 week long accretion outburst with an average 0.5–10 keV luminosity of L_X  ≃  10³⁶ erg s⁻¹ several years later (e.g., Altamirano et al. 2011a). This suggests that the peculiar X-ray burst had likely occurred during the decay of its previous accretion outburst. A similar scenario can be envisioned for SAX J2224.9+5421.