TWO NEW BURSTING NEUTRON STAR LOW-MASS X-RAY BINARIES: SWIFT J185003.2–005627 AND SWIFT J1922.7–1716

On 2011 June 24, Swift/BAT triggered and located an unknown source (trigger 456014; Beardmore et al. 2011a). Based on the soft nature of the BAT detection and the proximity to the Galactic plane, the trigger was tentatively identified as a previously unknown Galactic transient and dubbed Swift J185003.2–005627 (J1850 hereafter). The detection of a fading X-ray counterpart in follow-up XRT observations provided an arcsecond localization of the source (Beardmore et al. 2011a). Based on the properties of the XRT data and the refined analysis of the BAT trigger data,⁵ it was suggested that this event was a thermonuclear X-ray burst (Markwardt et al. 2011; Beardmore et al. 2011b).

J1850 was not detected in the Swift/BAT hard X-ray transient monitor on or after the time of the BAT trigger, with a 1σ upper limit on the daily averaged 15–50 keV count rate of 7 × 10⁻⁴ counts s⁻¹ cm⁻² (Krimm et al. 2011). However, the source was detected at an average 15–50 keV intensity of (3.5 ± 1.0) × 10⁻³ counts s⁻¹ cm⁻² (corresponding to a flux of ≃ 2 × 10⁻¹⁰ erg cm⁻² s⁻¹ for a Crab-like spectrum) between 2011 May 18 and May 26. Archival searches back to 2005 February did not reveal any similar detections of J1850 (Krimm et al. 2011).

Swift J1922.7–1716 (hereafter J1922) is a transient X-ray source that was discovered during the Swift/BAT hard X-ray survey of 2004 December–2005 March (Tueller et al. 2005a, 2005b). Swift/XRT follow-up observations allowed for the identification of a soft X-ray counterpart (Tueller et al. 2005b). Non-detections with INTEGRAL (2003–2004; Falanga et al. 2006) and Swift (2006 October–November; Kennea et al. 2011) testified to its transient nature.

Falanga et al. (2006) discussed Swift, RXTE, and INTEGRAL data taken between 2005 July and November. The broadband spectrum and timing properties were found to be typical of an LMXB in a faint X-ray state, but the nature of the compact accretor (i.e., a neutron star or a black hole) could not be established. A combined blackbody and power-law model provided an adequate description of the broadband spectrum, with a hydrogen column density N_H ≃ (1.5–2.0) × 10²¹ cm⁻², spectral index Γ ≃ 1.4–1.8, and temperature kT_bb ≃ 0.4–0.6 keV. The resulting 0.1–100 keV flux was ≃ (3.3–3.9) × 10⁻¹⁰ erg cm⁻² s⁻¹ (Falanga et al. 2006).

Renewed activity of J1922 was seen with MAXI and Swift/XRT in 2011 August (Nakahira et al. 2011; Kennea et al. 2011). Inspection of the Swift/BAT transient monitor data revealed that the source was detected in hard X-rays starting in 2011 July. It was seen at a mean 15–50 keV count rate of (1.5 ± 0.3) × 10⁻³ counts s⁻¹ cm⁻² (≃ 1 × 10⁻¹⁰ erg cm⁻² s⁻¹ for a Crab-like spectrum) until early-August, after which the hard X-ray flux decreased (Kennea et al. 2011). The 2011 intensity was similar to the average BAT count rate of the source in 2005.

Swift/BAT triggered on J1922 on 2011 November 3 (Barthelmy et al. 2011).⁶ Preliminary analysis of the BAT and XRT data revealed that the trigger was very likely caused by an X-ray burst (Degenaar et al. 2011d). Optical spectroscopy revealed clear He- and H-emission lines, supporting an LMXB nature (Wiersema et al. 2011; Halpern & Skinner 2011). The presence of H-lines implies that the donor star is H-rich and therefore the binary orbital period must be P_orb ≳ 1.5 hr (Nelson et al. 1986). Upper limits on the quiescent optical counterpart (>23.4 mag in the r and g bands) are consistent with an M-dwarf or evolved star and suggest P_orb ≲ 5 hr (Halpern & Skinner 2011).

We generated standard BAT data products (15–150 keV) for the trigger observations using the batgrbproduct tool. In all cases, the spacecraft started slewing toward the trigger location when the source had already faded into the background. We extracted single BAT spectra of the pre-slew data employing the tool batbinevt. The standard geometrical corrections and the BAT-recommended systematical error were applied with batupdatephakw and batphasyserr, respectively. Since we use only pre-slew data, we generated a single response matrix for each trigger by running the task batdrmgen. The spectra were fitted between 15 and 35 keV with XSpec (ver. 12.7; Arnaud 1996).

The XRT data cover an energy range of 0.5–10 keV and consist of a combination of photon counting (PC) and windowed timing (WT) mode. In the PC mode, a two-dimensional image is acquired, whereas in the WT mode, the CCD columns are collapsed into a one-dimensional image to reduce the frame time. The WT is typically used when the count rate exceeds ≃ 1–2 counts s⁻¹, because higher rates cause considerable pile-up in the PC mode.

All Swift data were reduced using the Swift tools (ver. 3.8) within the heasoft package (ver. 6.12), and employing the latest calibration data (ver. 3.8). We made use of the online tools to obtain XRT data products and obtain a global characterization of the persistent emission (Evans et al. 2009).⁷ For the detailed analysis of the X-ray bursts, we manually extracted XRT light curves and spectra using XSelect (ver. 3.8). Exposure maps were generated using the tool xrtexpomap, and subsequently used to create ancillary response files (arfs) with xrtmakearf. The latest redistribution matrix files (rmfs) were taken from the calibration database (caldb).

The XRT spectra were grouped to contain a minimum of 20 photons per bin and fitted between 0.5 and 10 keV in XSpec.

The UVOT data were obtained using a variety of optical and ultraviolet (UV) filters in a wavelength range of ≃ 1500–8500 Å (see Poole et al. 2008). We used a standard aperture of 5'' to extract source photons and a source-free region with a radius of 10'' as a background reference. Magnitudes and light curves were extracted using the tools uvotsource and uvotmaghist.

For our spectral analysis with XSpec, we use a blackbody model (BBODYRAD), a simple power law (POWERLAW), or a combination of both. In all fits, we included the PHABS model to account for neutral hydrogen absorption along the line of sight, for which we employed the default XSpec abundances (Anders & Grevesse 1989) and cross-sections (Balucinska-Church & McCammon 1992).

The bolometric accretion luminosity is typically a factor ≃ 2–3 higher than observed in the 0.5–10 keV energy band (in 't Zand et al. 2007). We therefore apply a correction factor of 2.5 to the 0.5–10 keV luminosity (L_X, determined from the Swift/XRT data) to estimate the bolometric accretion luminosity (L_bol). We subsequently use this to estimate the mass-accretion rate onto the neutron star (during outburst) via the relation $\dot{M}_{\mathrm{ob}}=RL_{\mathrm{bol}}/GM$, where G = 6.67 × 10⁻⁸ cm³ g⁻¹ s⁻² is the gravitational constant, and M and R are the mass and radius of the neutron star, respectively. We adopt canonical values of M = 1.4 M_☉ and R = 10 km.

It is often useful to express the accretion luminosity and mass-accretion rate in terms of the Eddington limit. We will adopt L_EDD = 3.8 × 10³⁸ erg s⁻¹, which is the empirical limit determined from X-ray bursts that display photospheric radius expansion (PRE; Kuulkers et al. 2003).

3.1.1. XRT

Figure 1 shows the outburst evolution of J1850 as seen with Swift/XRT. Following the BAT trigger on 2011 June 24, the source was observed almost daily for 27 days until 2011 July 21. In the days following the X-ray burst, the intensity remained steady at ≃ 5 counts s⁻¹ (WT), but the source started fading around 2011 July 5 (11 days after the BAT trigger). It eventually went undetected with the XRT starting on 2011 July 16 (22 days after the BAT trigger). Considering the sensitivity of the XRT, a non-detection in ≃ 3.3 ks of data between 2011 July 16 and 21 (Table 1) suggests that the source luminosity had dropped below a few times 10³³ erg s⁻¹ (depending on the spectral shape, 0.5–10 keV). This indicates that the outburst had ceased and that the source had returned to quiescence. We further constrain the quiescent emission in Section 3.3.

Zoom In Zoom Out Reset image size

The outburst observed with the XRT had a duration of ≃ 3 weeks, but the BAT already detected activity from the source on 2011 May 18–26 (see Section 1). If this was part of the same outburst, then it had a duration of ≳ 8 weeks. Lack of X-ray data before this epoch does not allow us to constrain the onset of the activity any further.

To obtain a global characterization of the outburst spectrum and flux, we fitted the PC and WT data simultaneously. The hydrogen column density was fixed between the two data sets, whereas the other fit parameters were left to vary freely. The results of our spectral analysis are summarized in Table 2.

Table 2. Results from Spectral Fitting of the Averaged Outburst Data

Year	NH	Γ	kTbb	Rbb	χ2ν (dof)	FX	LX	Lbol
Data Mode	(1022 cm−2)		(keV)	(km)		(10−10 erg cm−2 s−1)	(1035 erg s−1)
Swift J185003.2–005627
2011	1.1 ± 0.1	...	...	...	1.1 (669)	...	...	≃ 7
WT	...	1.6 ± 0.2	0.7 ± 0.1	1.8 ± 0.2	...	2.3 ± 0.2	3.8 ± 0.3	...
PC	...	0.9 ± 0.5	0.7 ± 0.1	1.7 ± 0.3	...	1.1 ± 0.1	1.8 ± 0.2	...
Swift J1922.7–1716
2005–2006	0.17 ± 0.08	...	...	...	0.9 (209)	...	...	≃ 17
WT	...	1.5 ± 0.4	0.4 ± 0.1	6.6 ± 0.7	...	2.8 ± 0.3	7.7 ± 0.8	...
PC	...	1.6 ± 0.4	0.4 ± 0.1	5.8 ± 0.7	...	2.2 ± 0.3	6.0 ± 0.9	...
2011–2012	0.15 ± 0.03	...	...	...	1.0 (660)	...	...	≃ 27
WT	...	1.7 ± 0.1	0.7 ± 0.1	3.8 ± 0.4	...	6.1 ± 0.2	16.8 ± 0.5	...
PC	...	1.6 ± 0.2	0.5 ± 0.1	4.7 ± 0.4	...	1.7 ± 0.1	4.7 ± 0.3	...

Notes. All quoted errors refer to 90% confidence levels. The outburst data were fitted to a combined power-law and blackbody model (POWERLAW+BBODYRAD), modified by absorption (PHABS). F_X gives the total unabsorbed model flux in the 0.5–10 keV band and L_X represents the corresponding luminosity assuming distances of 3.7 and 4.8 kpc for J1850 and J1922, respectively. L_bol represents the estimated bolometric accretion luminosity averaged over the outburst (see the text).

Download table as:  ASCIITypeset image

Fitting the spectra with either a single power-law or blackbody model yields reduced chi-square values of χ²_ν ≳ 1.3 for 673 degrees of freedom (dof). A combined power-law and blackbody model provides a better description of the data and results in N_H ≃ (1.1  ±  0.1) × 10²² cm⁻², Γ ≃ 0.9–1.6, and kT_bb ≃ 0.7 keV (χ²_ν = 1.1 for 669 dof; Table 2). The blackbody component contributes ≃ 45%–65% to the total unabsorbed 0.5–10 keV model flux. These spectral parameters are typical for low-luminosity neutron star LMXBs in the hard state (e.g., Lin et al. 2009; Linares 2009; Armas Padilla et al. 2011). The average PC and WT outburst spectra are shown in Figure 2.

Zoom In Zoom Out Reset image size

The corresponding 0.5–10 keV model flux is F_X ≃ 1.7 × 10⁻¹⁰ erg cm⁻² s⁻¹, when averaged over the PC and WT data. For a distance of D = 3.7 kpc (see Section 3.2), this translates into a mean outburst luminosity of L_X ≃ 2.8 × 10³⁵ erg s⁻¹ (0.5–10 keV). This allows for an estimate of the average bolometric accretion luminosity of L_bol ≃ 7 × 10³⁵(D/3.7 kpc)² erg s⁻¹, which corresponds to ≃ 0.2% of the Eddington limit (Section 2.4).

To investigate whether any spectral evolution occurred along the X-ray outburst, we compared the ratio of counts in the 2–10 keV and 0.5–2 keV energy bands (Figure 3). This suggests that the spectrum softens (i.e., the hardness ratio becomes smaller) as the intensity decreases. This behavior has been seen for several black hole and neutron star LMXBs transitioning from the hard state to quiescence (Armas Padilla et al. 2011 and references therein).

Zoom In Zoom Out Reset image size

3.1.2. UVOT

3.2.1. BAT Burst Peak

Figure 4 displays the 15–25 keV BAT light curve of trigger 456014 at 2 s resolution. The light curve looks like a single peak centered at t ≃ 2 s after the trigger. The source is visible for ≃ 25 s from ≃ 10 s prior to the BAT trigger to ≃ 15 s thereafter. Slewing of the spacecraft started at t ≃ 35 s, when the source had already faded into the background. We used an interval of 20 s centered around the peak (covering t = −5 to 15 s) to extract the average BAT spectrum.

Zoom In Zoom Out Reset image size

The BAT spectrum is soft, with no source photons detected above ≃ 35 keV. It fits to a blackbody with kT_bb ≃ 2.3 keV and R_bb ≃ 5.6 km, resulting in χ²_ν = 1.0 for 7 dof (assuming D = 3.7 kpc and fixing N_H = 1.1 × 10²² cm⁻²). Extrapolating the model fit to the 0.01–100 keV energy range yields an estimate of the bolometric flux of F_bol ≃ 7.5 × 10⁻⁸ erg cm⁻² s⁻¹. For a duration of ≃ 20 s, we estimate a (bolometric) fluence of f_BAT ≃ 1.5 × 10⁻⁶ erg cm⁻² (Table 3).

Table 3. Spectral Parameters of the X-Ray Bursts

Year	Δt	NH	kTbb	Rbb	χ2ν (dof)	Fbol
Data Mode	(s)	(1022 cm−2)	(keV)	(km)		(erg cm−2 s−1)
Swift J185003.2–005627
2011 June 24
BAT	0–20	1.1 fix	2.3 ± 0.6	5.6+6.5−5.6	1.0 (7)	(7.5 ± 4.2) × 10−8
XRT/WT	101–201	1.1 fix	0.81 ± 0.03	5.4 ± 0.1	1.1 (66)	(1.0 ± 0.1) × 10−9
XRT/WT	202–430	1.1 fix	0.73 ± 0.03	4.4 ± 0.1	1.2 (68)	(4.5 ± 0.1) × 10−10
Swift J1922.7–1716
2011 November 3
BAT	0–20	0.16 fix	2.4 ± 0.5	6.7+4.0−6.7	1.7 (8)	(6.8 ± 3.2) × 10−8
XRT/WT	136–210	0.16 fix	0.80 ± 0.07	5.3 ± 0.3	0.9 (84)	(5.0 ± 0.3) × 10−10
XRT/WT	211–440	0.16 fix	0.59 ± 0.09	5.0 ± 0.6	1.2 (151)	(1.4 ± 0.1) × 10−10
2011 December 2
BAT	0–20	0.16 fix	2.0 ± 0.5	14.9+10.7−14.9	1.0 (9)	(10.2 ± 6.3) × 10−8

Notes. All quoted errors refer to 90% confidence levels. The parameter Δt indicates the time since the BAT trigger. The burst data were fitted to a blackbody model (BBODYRAD), modified by absorption (PHABS), with the hydrogen column density (N_H) fixed to the value obtained for the outburst fits. When calculating the emitting radii (R_bb), we assumed distances of 3.7 and 4.8 kpc for J1850 and J1922, respectively. F_bol gives an estimate of the bolometric flux, which was obtained by extrapolating the blackbody fits to the 0.01–100 keV energy range. We note that the emitting radii inferred from simple blackbody fits are expected to be underestimated due to the fact that electron scatterings in the neutron star atmosphere harden the spectrum resulting in a color temperature that is larger than the effective temperature by a factor ≃ 1.5 (e.g., Suleimanov et al. 2011). Consequently, the inferred radius is underestimated by that same factor.

Download table as:  ASCIITypeset image

While the average count rate over the 20 s interval is 2.0 × 10⁻² counts s⁻¹, the peak of the BAT data is a factor ≃ 3 higher. We therefore estimate a bolometric peak flux of F_{bol, peak} ≃ 2.3 × 10⁻⁷ erg cm⁻² s⁻¹. Assuming that the peak was equal to the empirical Eddington limit inferred from PRE bursts (Section 2.4) places the source at a distance of D = 3.7 kpc. The data do not reveal signatures of PRE, which implies that the burst may have been sub-Eddington. This distance estimate should therefore be regarded as an upper limit.

3.2.2. XRT Burst Tail

Observations with the narrow-field instruments commenced ≃ 100 s after the BAT trigger. The inset of Figure 1 displays the XRT data, which shows a clear decay in count rate from ≃ 20 counts s⁻¹ at the start of the observation, leveling off to ≃ 5 counts s⁻¹ about 300 s later. This suggests that the total duration of the X-ray burst was ≃ 400 s (≃ 7 minutes).

The statistics of the XRT data do not allow for a detailed time-resolved spectroscopic analysis. To investigate whether the data exhibit the characteristic spectral softening seen in the tails of X-ray bursts, we extracted two separate spectra for t = 101–201 and t = 202–430 s after the burst trigger (all WT data).

We use an interval of ≃ 500 s of PC data, obtained between t = 525 and 1025 s, to subtract the underlying accretion emission. Fitting this persistent spectrum to an absorbed power-law model yields an unabsorbed 0.5–10 keV flux of F_X ≃ 4.8 × 10⁻¹⁰ erg cm⁻² s⁻¹. This suggests that J1850 was accreting at ≃ 0.5% of the Eddington rate when the X-ray burst occurred.

The XRT spectra along the burst tail are best described by an absorbed blackbody model. We fix N_H = 1.1 × 10²² cm⁻² (Section 3.1), and find that there is evidence of cooling along the ≃ 300 s decay tail from kT_bb ≃ 0.81 ± 0.03 keV to kT_bb ≃ 0.73 ± 0.03 keV, with corresponding blackbody emitting radii of R_bb ≃ 4–5 km (Table 3). The obtained spectral parameters are typical of X-ray burst tails and provide further support that the BAT was triggered by a thermonuclear event.

To investigate the shape of the decay, we fitted the XRT count rate light curve (binned by 10 s) to a power law and an exponential function (Figure 5). For both decay functions, we include a constant offset to represent the underlying persistent X-ray emission. The power-law fit yields a decay index of α = −2.1 ± 0.1 and a normalization of ≃ 3.1 × 10⁵ counts s⁻¹ (χ²_ν = 0.7 for 30 dof). For the exponential function, we obtain a decay time of τ = 84.7 ± 4.9 s and a normalization of ≃ 54 counts s⁻¹ (χ²_ν = 0.5 for 30 dof).

Zoom In Zoom Out Reset image size

We apply a count rate to flux conversion factor inferred from fitting the average XRT burst spectrum and integrate both decay fits between t = 20 s (the time when the BAT intensity had returned to the background level) until t = 400 s (when the XRT intensity had leveled off to its persistent value). This yields an estimate of the fluence along the XRT burst tail of f_XRT ≃ (3.8–8.5) × 10⁻⁷ erg cm⁻². Adding this to the value obtained for the BAT data suggests a total fluence of f_tot ≃ (1.9–2.4) × 10⁻⁶ erg cm⁻² for this event (which is corrected for the underlying persistent emission). For a distance of D = 3.7 kpc, the corresponding radiated energy is E_b ≃ (3.1–3.9) × 10³⁹ erg.

We searched the light curves of all individual XRT observations for occurrences of X-ray bursts. Apart from the 2011 June 24 event, no other bursts were found.

J1850 is within the field of view (FOV) of three archival XMM-Newton observations obtained in 2003 (Table 1). The source is not detected in any of these observations. We derive a 95% confidence upper limit on the count rate of ≲ 1 × 10⁻² counts s⁻¹ for the PN data, and ≲ 1 × 10⁻³ counts s⁻¹ for the MOS cameras, after applying the prescription for small numbers of counts given by Gehrels (1986). Using pimms (ver. 4.5), we estimate the corresponding upper limit on the quiescent luminosity.

Since the quiescent spectral shape is unknown, we explored both a power-law spectral model with Γ = 1–3 and a blackbody of kT_bb = 0.2–0.3 keV (with N_H = 1.1 × 10²² cm⁻²), which are typical values found for the quiescent spectra of neutron star LMXBs (e.g., Asai et al. 1998; Rutledge et al. 1999; Wijnands et al. 2005a; Degenaar et al. 2012a). This results in an estimated upper limit on the 0.5–10 keV quiescent luminosity of L_q ≲ (0.5–2.8) × 10³²(D/3.7 kpc)² erg s⁻¹.

4.1.1. XRT

Swift covered two different outbursts of J1922 in 2005–2006 and 2011–2012. The source was found active in all XRT observations that were carried out between 2005 July 8 and 2006 June 18. However, it is not detected with the XRT in a set of pointings taken between 2006 October 30 and November 14. This demonstrates that the source had returned to quiescence (Section 4.3). The XRT data suggest that J1922 was active for at least 345 days in 2005–2006.

Hard X-ray monitoring data also indicate that the source experienced a long outburst. The BAT light curve (15–150 keV) shows that it appeared active right at the start of the survey (2004 December; Tueller et al. 2005a, 2005b), until it faded ≃ 20 months later in 2006 July. Examination of the pre-processed INTEGRAL/IBIS-ISGR light curve of J1922 (17–80 keV), which is publicly available via the High-Energy Astrophysics Virtually ENlightened Sky (HEAVENS; Lubiński 2009; Walter et al. 2010),⁸ suggests a similar outburst length as inferred from the BAT data. This is in agreement with the XRT results.

The average XRT spectrum of the 2005–2006 outburst can be described by a single absorbed power-law model, yielding N_H ≃ (2.9 ± 0.3) × 10²¹ cm⁻² and Γ = 2.1 ± 0.1 (χ²_ν = 0.9 for 213 dof). Analysis of the broadband 2005 spectrum required the addition of a soft emission component (Falanga et al. 2006). We obtain similar results for the 2011–2012 data (see below). Therefore, we included a blackbody component to the power-law fit for comparison.

The combined power-law and blackbody fit results in N_H ≃ (1.7 ± 0.1) × 10²¹ cm⁻², Γ ≃ 1.5–1.6, kT_bb ≃ 0.4 ± 0.1 keV, R_bb ≃ 6–7 km (for D = 4.8 kpc), and χ²_ν = 0.9 for 209 dof (Table 2). The blackbody component contributes ≃ 20% to the total 0.5–10 keV unabsorbed flux. These results are similar to those obtained for the broadband 2005 spectrum (Falanga et al. 2006).

The 0.5–10 keV unabsorbed flux (averaged over the PC and WT data) for the combined model fit is F_X ≃ 2.5 × 10⁻¹⁰ erg cm⁻² s⁻¹. This translates into a mean outburst luminosity of L_X ≃ 6.9 × 10³⁵ erg s⁻¹ for a distance of D = 4.8 kpc (Section 4.2). Assuming that the bolometric luminosity is a factor ≃ 2.5 higher than the intensity in the 0.5–10 keV band, we estimate an average accretion luminosity of L_bol ≃ 1.7 × 10³⁶(D/4.8 kpc)² erg s⁻¹, or ≃ 0.5% of the Eddington limit (Section 2.4).

Swift/BAT and MAXI monitoring observations revealed that J1922 entered a new outburst around 2011 July 17 (Kennea et al. 2011). Between 2011 August 13 and 2012 March 15, J1922 was observed during a number of XRT pointings that all detected it in outburst. However, the source is no longer seen with the XRT on 2012 May 25 and follow-up observations performed between 2012 June 9 and 21 (Table 1). This suggests that the source had returned to quiescence (Figure 6 and Section 4.3). The 2011–2012 outburst had a duration of ≳ 240 days (0.7 yr). If the Swift/BAT hard transient monitor caught the start of it, then the duration is constrained to ≲ 310 days (0.9 yr). The pre-processed publicly available IBIS-ISGR data do not cover this second outburst from J1922.

Zoom In Zoom Out Reset image size

Fitting the average XRT spectrum of the 2011–2012 data with a simple absorbed power-law model yields N_H ≃ (2.7 ± 0.1) × 10²¹ cm⁻² and Γ ≃ 2.0 ± 0.1 (χ²_ν = 1.2 for 664 dof). The fit can be improved by adding a blackbody component, which results in N_H ≃ (1.5 ± 0.1) × 10²¹ cm⁻², Γ ≃ 1.6–1.7, kT_bb ≃ 0.5–0.7 keV, R_bb ≃ 4–5 km (for D = 4.8 kpc), and χ²_ν = 1.0 for 660 dof (Table 2). The 2011–2012 spectral data and model fit are shown in Figure 7.

Zoom In Zoom Out Reset image size

The best fit yields an average 0.5–10 keV luminosity for the 2011–2012 outburst of L_X ≃ 1.1 × 10³⁶(D/4.8 kpc)² erg s⁻¹, and an estimated bolometric accretion luminosity of L_bol ≃ 2.7 × 10³⁶(D/4.8 kpc)² erg s⁻¹ (≃ 1% of the Eddington limit). The blackbody component contributes ≃ 20% to the total unabsorbed 0.5–10 keV model flux. The spectral properties and intensity of the 2011–2012 outburst are comparable to that observed in 2005–2006 (Table 2).

Investigation of the ratio of counts in different energy bands (0.5–2 and 2–10 keV) does not reveal any evident spectral evolution along the outburst. It is of note that the dynamical range spanned by the observations of J1922 is narrower than that of J1850, for which the spectrum appeared to be softening with decreasing intensity (Section 3.1).

4.1.2. UVOT

Inspection of UVOT data reveals a possible counterpart at the position of J1922 in all filters (Table 4). This object is not present in Digitized Sky Survey (DSS) images (Barthelmy et al. 2011), and is not detected during the observations in which J1922 had faded into X-ray quiescence (2006 October–November and 2012 May–June; see Figures 8 and 9). This strongly suggests that the source detected in the UVOT images represents the optical/UV counterpart of J1922. As such, the UVOT data provide a subarcsecond localization of the LMXB (Barthelmy et al. 2011).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 4. UVOT Observations and Results for J1922

Obs ID	Date	Band	Exp.	Magnitude
			(ks)
35174001	2005 Jul 8	um2	1.7	16.13 ± 0.03
		u	0.3	16.12 ± 0.03
		v	0.5	17.11 ± 0.06
		uw1	1.1	16.05 ± 0.03
		uw2	2.3	16.05 ± 0.03
35174003	2005 Oct 1	b		17.31 ± 0.04
		um2	0.4	16.08 ± 0.03
		u	2.7	16.17 ± 0.03
		v	0.9	16.98 ± 0.05
		uw1	1.8	16.03 ± 0.03
		uw2	3.6	16.04 ± 0.02
35471001	2006 Mar 14	uw2	1.5	17.62 ± 0.03
35471002	2006 Jun 4	v	0.3	17.45 ± 0.10
35471003	2006 Jun 18	v	1.0	17.42 ± 0.05
35471004	2006 Oct 30	v	5.1	>20.69
35471005	2006 Nov 3	v	1.0	>19.74
35471006	2006 Nov 10	v	4.2	>20.42
35471007	2006 Nov 14	v	3.8	>20.24
35471008	2011 Aug 13	um2	1.0	16.52 ± 0.04
35471009	2011 Aug 30	uw1	1.2	16.34 ± 0.03
35471010	2011 Aug 31	u	4.2	16.45 ± 0.02
506913000	2011 Nov 3	b	0.2	17.13 ± 0.06
		um2	0.2	17.80 ± 0.06
		u	0.2	16.42 ± 0.05
		v	0.2	17.40 ± 0.14
		uw1	0.2	17.64 ± 0.05
		uw2	0.2	16.03 ± 0.05
		wh	0.4	16.64 ± 0.04
35471011	2011 Nov 4	b	0.1	17.08 ± 0.09
		um2	0.2	15.98 ± 0.06
		u	0.1	17.02 ± 0.06
		v	0.1	17.54 ± 0.28
		uw1	0.1	15.97 ± 0.06
		uw2	0.3	15.85 ± 0.04
35471012	2011 Nov 5	b	0.1	17.31 ± 0.11
		um2	0.2	18.15 ± 0.07
		u	0.1	16.44 ± 0.08
		v	0.1	17.71 ± 0.30
		uw1	0.2	16.26 ± 0.07
		uw2	0.3	16.25 ± 0.05
35471013	2012 Mar 9	u	0.4	17.12 ± 0.07
		uw1	0.3	17.11 ± 0.09
35471014	2012 Mar 14	u	1.0	17.08 ± 0.04
35471015	2012 May 25	u	1.2	>20.65
35471016	2012 Jun 9	uw1	1.9	>20.94
35471017	2012 Jun 11	uw2	2.0	>21.21
35471018	2012 Jun 13	uw1	0.9	>20.64
35471019	2012 Jun 15	uw2	0.7	>20.62
35471020	2012 Jun 17	uw1	0.3	>19.84
35471021	2012 Jun 21	uw1	2.3	>21.14

Notes. Quoted errors on the observed magnitudes correspond to a 1σ confidence level. In the case of a non-detection, a 3σ upper limit is given.

Download table as:  ASCIITypeset image

We have listed the magnitudes and upper limits of all UVOT observations in Table 4. These magnitudes are not corrected for interstellar extinction. Fits to the 2005–2006 and 2011–2012 XRT spectra yield N_H ≃ 1.6 × 10²¹ cm⁻², which suggests a visual extinction of A_V ≃ 0.7 mag (Güver & Özel 2009). The upper limits obtained with the UVOT for the quiescent state (v ≳ 20 mag, Table 4) are consistent with the constraints from ground-based telescopes (>23.4 mag in the r and g bands; Halpern & Skinner 2011).

4.2.1. BAT Burst Peak

4.2.2. XRT Burst Tail

J1922 went undetected during four consecutive XRT observations carried out between 2006 October 30 and November 14 (Obs IDs 35471004–7). We obtain count rate upper limits for these individual exposures of ≲ (1–3) × 10⁻³ counts s⁻¹ (0.5–10 keV). Summing the four observations (≃ 13.8 ks in total) reveals a small excess of 11 photons at the source position (see Figure 8), whereas five photons are expected from a set of three randomly placed background regions. The inferred count rate is 4.3 × 10⁻⁴ counts s⁻¹. For N_H = 1.6 × 10²¹ cm⁻², a power-law model with Γ = 1–3, or a blackbody with kT_bb = 0.2–0.3 keV, we estimate a 0.5–10 keV quiescent luminosity of L_q ≃ (0.4–0.9) × 10³²(D/4.8 kpc)² erg s⁻¹.

We note that this quiescent detection occurred within ≃ 4 months after the end of a long (≃ 2 yr) accretion outburst of J1922. It is possible that the neutron star became considerably heated during this prolonged outburst and did not yet cool down at the time of the Swift/XRT observations (Section 6.2). In this case, the true quiescent level could be lower than inferred here.

Inspection of an archival 78.6 ks long Suzaku observation, performed on 2007 April 10, reveals a possible weak detection at the position of J1922 in the combined XIS image. In the source region, a total of ≃ 450 photons are detected, whereas ≃ 350 are expected from the background. We estimate a source count rate of ≃ 1.3 × 10⁻³ counts s⁻¹. Using the range of spectral parameters stated above, we obtain a 0.5–10 keV luminosity of L_q ≃ (0.6–1.0) × 10³²(D/4.8 kpc)² erg s⁻¹. This is comparable to our estimate obtained from the Swift/XRT observations.

On 2012 May 25, the source was not detected during a single Swift/XRT observation with an upper limit on the count rate of ≲ 2.5 × 10⁻³ counts s⁻¹ at 95% confidence level (Gehrels 1986). This translates into a 0.5–10 keV quiescent luminosity upper limit L_q ≲ (2.2–2.5) × 10³²(D/4.8 kpc)² erg s⁻¹, and strongly suggests that the 2011–2012 outburst had ceased.

We obtained a series of Swift follow-up observations between 2012 June 9 and 21 (Table 1). The source is not detected in the combined XRT image (total exposure time of ≃ 9.2 ks when including the observation of 2012 May 25). We infer an upper limit on the count rate of ≲ 6.8 × 10⁻⁴ counts s⁻¹ (95% confidence; Gehrels 1986). Using the range of spectral parameters given above, we obtain a 0.5–10 keV upper limit of L_q ≲ (0.4–1.4) × 10³²(D/4.8 kpc)² erg s⁻¹.

The observable properties of X-ray bursts (e.g., the duration, recurrence time, and radiated energy) depend on the conditions of the ignition layer, such as the thickness, temperature, and H-abundance. These can drastically change as the mass-accretion rate onto the neutron star varies, such that there exist distinct accretion regimes that give rise to X-ray bursts with different characteristics (Fujimoto et al. 1981; Bildsten 1998). The overall similarities between the properties of the X-ray bursts of J1850 and J1922 (Table 5) suggest that these events were ignited under similar conditions.

In the accretion regime of J1850 and J1922 (≃ 0.5%–1% of Eddington), the temperature of the burning layer is expected to be low enough for H to burn unstably. This may in turn trigger He ignition in an H-rich environment, which typically results in ≃ 10–100 s long bursts (e.g., Fujimoto et al. 1981; Bildsten 1998; Galloway et al. 2008). The detection of H-emission lines in the optical spectrum of J1922 suggests that the neutron star is accreting H-rich material (Wiersema et al. 2011; Halpern & Skinner 2011). This implies that the fuel triggering the X-ray burst could have indeed contained H. Given the similarities in burst properties, J1850 may therefore be expected to host an H-rich companion star as well.

The observed burst duration of t_b ≃ 400 s (τ ≃ 85–100 s) is considerably longer than that typically observed for H-rich X-ray bursts (Chenevez et al. 2008; Linares et al. 2012). The burst profiles may look very different depending on the energy band in which they are observed (Chelovekov et al. 2006; Linares et al. 2012), and most bursts known to date have been detected with instruments that cover higher energies than Swift/XRT (typically >2 keV, e.g., RXTE, INTEGRAL, BeppoSAX, Fermi; Cornelisse et al. 2003; Chenevez et al. 2008; Galloway et al. 2008; Linares et al. 2012). However, there are several X-ray bursts observed with Swift/XRT, Chandra, or XMM-Newton (i.e., in the same energy band as the bursts observed from J1850 and J1922) from neutron star LMXBs accreting at ≃ 1% of Eddington that do show the expected duration of ≃ 10–100 s (e.g., Boirin et al. 2007; Trap et al. 2009; Degenaar & Wijnands 2009; Degenaar et al. 2012b).

The uncommon properties of the bursts from J1850 and J1922 are further illustrated by their energetics. The estimated radiated energies of E_b ≲ (3–7) × 10³⁹ erg are higher than those typically found for normal X-ray bursts, yet lower than those classified as intermediately long (Chenevez et al. 2008; Linares et al. 2012). There are several possible explanations that may account for the unusual burst properties.

First, for some LMXBs, the first X-ray burst that occurred during a new outburst was found to be significantly longer than later bursts (e.g., Aql X-1 and Cen X-4; Fushiki et al. 1992; Kuulkers et al. 2009). It is thought that at the start of the outburst the neutron star is relatively cold and therefore a thicker layer of fuel can build up before the ignition conditions are met. Since the duration of an X-ray burst depends on the cooling time of the ignition layer, a thicker layer would result in a longer (and more energetic) X-ray burst. By considering the expected burst recurrence time, we can assess whether this scenario might be applicable to J1850 and J1922.

With the burst energetics at hand, the ignition depth can be estimated as y = E_b(1 + z)/4πR²Q_{nuc, burst}, where z is gravitational redshift, R is the neutron star radius, and Q_{nuc, burst} = 1.6 + 4X MeV nucleon⁻¹, the nuclear energy release during an X-ray burst given an H-fraction X at ignition (e.g., Galloway et al. 2008). For a neutron star with M = 1.4 M_☉ and R = 10 km (i.e., z = 0.31), we estimate y ≃ 10⁻⁸ g cm⁻² for the bursts of J1850 and J1922. The recurrence time that corresponds to a given ignition depth is $t_{\mathrm{rec}} \simeq y(1+z)/\dot{m}$, where $\dot{m}$ is the accretion rate onto the neutron star surface per unit area. For J1850 and J1922, we infer an average accretion rate during outburst of $\dot{m}=\dot{M}/4\pi R^2 \simeq 300$ and ≃ 2300 g cm⁻² s⁻¹, respectively (assuming that the emission is isotropic). This would imply an expected recurrence time on the order of a few days to two weeks.

The BAT transient monitor revealed activity from J1850 ≃ 5 weeks before the BAT trigger of 2011 June 24. J1922 had been accreting for ≃ 3 months prior to its BAT trigger on 2011 November 3, as evidenced by Swift and MAXI (Sections 1.1 and 1.2). Comparing this to the expected burst recurrence time of ≲ 2 weeks suggests that the BAT triggers were likely not the first X-ray bursts that occurred during the outbursts of J1850 and J1922. The BAT coverage starts at 15 keV, which implies that it is not optimally sensitive to detect events as soft as X-ray bursts. These are therefore easily missed.

A second scenario that is worth considering is that some X-ray bursts display prolonged tails that last up to an hour (e.g., EXO 0748–676 and the "clocked burster" GS 1826–24; in 't Zand et al. 2009). These are explained as the cooling of layers below the ignition layer that were heated by inward conduction of energy generated during the X-ray burst. These long cooling tails are characterized by fluxes and fluences that are two orders of magnitude lower than the prompt burst emission. In the case of J1850 and J1922, however, the fluence in the BAT peak and XRT tail are of similar magnitude. This indicates that the long duration of these X-ray bursts is likely caused by a different mechanism.

Alternatively, the long burst duration may be explained in terms of a relatively large H-content in the ignition layer. The presence of H lengthens the duration of nuclear energy generation via the rp-process (Schatz et al. 2001). There are two additional effects that likely contribute in making the bursts observable for a longer time. At low accretion rates, the temperature in the neutron star envelope is expected to be low compared to brighter bursters. This implies that a thicker fuel layer can accumulate before reaching the critical ignition conditions, resulting in a longer burst. Furthermore, J1850 and J1922 accreted at a relatively low level of ≃ 0.5%–1% of Eddington. For such low persistent emission levels, the tails of the X-ray bursts are visible for a longer time (if observed with a sensitive instrument), which can add to a longer burst duration. The burst recurrence time at the accretion rates inferred for J1850 and J1922 is considerably lower than for higher rates. This can account for the fact that most observed X-ray bursts are shorter than those seen for these two sources.

Intermediately, long X-ray bursts are typically observed from sources accreting at ≃ 0.1%–1% of Eddington. These events are both longer (≳ 10 minutes) and more energetic (≃ 10^40–41 erg) than we have observed for J1850 and J1922 (e.g., in 't Zand et al. 2008; Falanga et al. 2009; Linares et al. 2009; Degenaar et al. 2011b). Some of these sources are strong candidate ultra-compact X-ray binaries. These contain an H-depleted companion star so that the neutron star accretes (nearly) pure He (e.g., in 't Zand et al. 2008). In absence of H-burning, the temperature in the accreted envelope is relatively low, so that a thick layer of He can build up before it eventually ignites in a long and energetic X-ray burst. A similarly long X-ray burst has also been observed, however, from a source that accretes at ≃ 0.1 of Eddington and shows a strong H-emission line in its optical spectrum. This testifies to the presence of an H-rich companion (Degenaar et al. 2010). In this case, it may be that unstable H-burning could not immediately trigger He ignition, allowing the development of a thick layer of fuel (Cooper & Narayan 2007; Peng et al. 2007).

To summarize, the burst duration and energetics of the X-ray bursts observed from J1850 and J1922 appear to fall in between that of normal and intermediately long X-ray bursts (cf. Chenevez et al. 2008; Linares et al. 2012). Similar bursts have also been observed from the persistent neutron star LMXB 4U 0614+09 (Kuulkers et al. 2010; Linares et al. 2012). We considered several possible scenarios that may account for the relatively long burst duration, and find it most likely that it is due to the ignition of a relatively thick layer of (H-rich) fuel. Our findings suggests that there may not exist two distinct groups of X-rays bursts, but rather a continuous range of burst durations and energies. This supports the idea proposed by Linares et al. (2012) that the apparent bimodal distribution is likely due to an observational bias toward detecting the longest and most energetic X-ray bursts from slowly accreting neutron stars.

J1850 could not be detected in archival XMM-Newton observations with an estimated upper limit of L_q ≲ (0.5–3.0) × 10³²(D/3.7 kpc)² erg s⁻¹. J1922 is detected in quiescence with Swift/XRT and Suzaku at a 0.5–10 keV luminosity of L_q ≃ (0.4–1.0) × 10³²(D/4.8 kpc)² erg s⁻¹. These quiescent levels are common for neutron star LMXBs (e.g., Menou et al. 1999; Garcia et al. 2001; Jonker et al. 2004).

It is thought that the accretion of matter onto the surface of a neutron star compresses the stellar crust and induces a chain of nuclear reactions that deposit heat (Haensel & Zdunik 1990). This heat spreads over the entire stellar body via thermal conduction and maintains the neutron star at a temperature that is set by the long-term averaged accretion rate of the binary (Brown et al. 1998; Colpi et al. 2001).

During quiescent episodes, the neutron star is expected to thermally emit X-rays providing a candescent luminosity of $L_{\mathrm{q,bol}} = \langle \dot{M} \rangle Q_{\mathrm{nuc}} / m_u$, where Q_nuc ≃ 2 MeV is the nuclear energy deposited in the crust per accreted baryon (Haensel & Zdunik 2008; Gupta et al. 2007, but see Degenaar et al. 2011a), m_u = 1.66 × 10⁻²⁴ g is the atomic mass unit (i.e., Q_nuc/m_u ≃ 9.6 × 10¹⁷ erg g⁻¹), and $\langle \dot{M} \rangle$ is the long-term accretion rate of the binary averaged over ≃ 10⁴ yr (i.e., including both outburst and quiescent episodes). The latter can be estimated by multiplying the average accretion rate observed during outburst ($\dot{M}_{{\rm ob}}$) with the duty cycle of the binary (i.e., the ratio of the outburst duration and recurrence time).

The outburst of J1850 observed with Swift in 2011 had a duration of ≃ 8 weeks (0.17 yr) and an estimated bolometric accretion luminosity of L_bol ≃ 7 × 10³⁵(D/3.7 kpc)² erg s⁻¹ (corresponding to $\dot{M}_{\mathrm{ob}}\simeq 6\times 10^{-11} \, {M_{\odot }\, {\rm yr}}^{-1}$). The outburst duration and recurrence time are not known for J1850, but we can make an order of magnitude estimate based on the constraints of the quiescent luminosity.

Within the deep crustal heating model, a quiescent luminosity of ≲ 3 × 10³²(D/3.7 kpc)² erg s⁻¹ would suggest a time-averaged mass-accretion rate of $\langle \dot{M} \rangle \lesssim 5 \times 10^{-12} \, {M_{\odot }\, {\rm yr}}^{-1}$. For $\dot{M}_{\mathrm{ob}}\simeq 6\times 10^{-11} \, {M_{\odot }\, {\rm yr}}^{-1}$, the corresponding duty cycle is ≲ 10%. If J1850 typically exhibits outbursts with a duration of 8 weeks, then the expected recurrence time would be ≳ 2 yr. We regard this as a lower limit, since the expected recurrence time increases for a longer outburst duration, or a quiescent thermal luminosity that is lower than the assumed upper limit of 3 × 10³² erg s⁻¹. This estimate is consistent with the fact that the Swift/BAT hard transient monitor did not detect any other outbursts from J1850 back to 2005 February (Krimm et al. 2011).

Since two outbursts with a relatively well-constrained duration have been observed for J1922, we can reverse the above reasoning and estimate the quiescent luminosity that is expected based on the observed duty cycle. The source was discovered when it exhibited an outburst of ≃ 20 months in 2005–2006. Renewed activity was observed from the source 5 yr later in 2011–2012, when it accreted for ≃ 8–10 months. For both outbursts, we estimate a similar bolometric accretion luminosity of L_bol ≃ (2–3) × 10³⁶ erg s⁻¹, which suggests a mean mass-accretion rate of $\dot{M}_{\mathrm{ob}}\simeq 3\times 10^{-10} \, {M_{\odot }\, {\rm yr}}^{-1}$ (Table 2). If we assume a typical outburst duration of ≃ 1.3 yr and a recurrence time of ≃ 5 yr (i.e., a duty cycle of ≃26%), then we can estimate a long-term averaged mass-accretion rate of $\langle \dot{M} \rangle \simeq 1\times 10^{-10} \, {M_{\odot }\, {\rm yr}}^{-1}$ (equivalent to ≃ 6 × 10¹⁵ g s⁻¹).

If the outburst behavior observed over the past decade is typical for the long-term accretion history of J1922, then the estimated $\langle \dot{M} \rangle$ should give rise to L_{q, bol} ≃ 6 × 10³³(D/4.8 kpc)² erg s⁻¹. This is considerably higher than the observed L_q ≃ 1 × 10³²(D/4.8 kpc)² erg s⁻¹ (0.5–10 keV). Although the bolometric luminosity may be a factor of a few higher than that measured in the 0.5–10 keV band, the luminosity remains lower than expected based on the crustal heating model. Limited by a low number of counts, we cannot constrain the shape of the quiescent spectrum of J1922 with the current available data. It is possible that the quiescent emission contains a significant non-thermal component (e.g., Campana et al. 2005; Wijnands et al. 2005b; Heinke et al. 2009; Cackett et al. 2010b; Degenaar & Wijnands 2012; Degenaar et al. 2012a), which would increase the discrepancy between the expected and observed quiescent thermal emission.

Provided that the heating models are correct, a plausible explanation for this apparent mismatch could be that the time-averaged mass-accretion rate is lower than estimated (e.g., because the recent outburst behavior is not representative for the long-term accretion history). Alternatively, the neutron star could be cooling faster than assumed in the standard paradigm (Page et al. 2004). A comparison with theoretical cooling models of Yakovlev & Pethick (2004) would suggest enhanced cooling due to the presence of kaons in the neutron star core.

Four neutron star LMXBs have been closely monitored after the cessation of their very long (≳ 1 yr) accretion outbursts, which has revealed that the neutron star temperature was gradually decreasing over the course of several years (Wijnands et al. 2001, 2003; Cackett et al. 2008, 2010a; Degenaar et al. 2009, 2011c; Fridriksson et al. 2010, 2011; Díaz Trigo et al. 2011). Recently, similar behavior has been observed for a transient neutron star LMXB in Terzan 5 that exhibited a much shorter accretion outburst of ≃ 10 weeks (Degenaar & Wijnands 2011; Degenaar et al. 2011a). These observations can be explained as cooling of the neutron star crust, which became considerably heated during the accretion outburst and needs time to cool down in quiescence. Monitoring and modeling this crustal cooling provides the unique opportunity to gain insight into the properties of the neutron star crust and core (Rutledge et al. 2002; Wijnands 2005; Shternin et al. 2007; Brown & Cumming 2009; Degenaar et al. 2011a; Page & Reddy 2012).

With its relatively low quiescent luminosity, long outburst duration, and low extinction, J1922 would be a promising target to search for crustal cooling now that its recent outburst has ceased.