IGR J17503-2636: a candidate supergiant fast X-ray transient

TL;DR

IGR J17503-2636 is a newly identified supergiant fast X-ray transient exhibiting fast variability, a broad spectral feature possibly indicating a magnetized neutron star, and is associated with a highly reddened OB supergiant at about 10 kpc.

Contribution

This study presents the first detailed analysis of IGR J17503-2636, identifying it as a candidate supergiant fast X-ray transient with potential magnetic neutron star characteristics.

Findings

Fast X-ray variability observed

Spectral feature suggests a magnetized neutron star

Likely associated with a reddened OB supergiant at ~10 kpc

Abstract

IGR J17503-2636 is a hard X-ray transient discovered by INTEGRAL on 2018 August 11. This was the first ever reported X-ray emission from this source. Following the discovery, follow-up observations were carried out with Swift, Chandra, NICER, and NuSTAR. We report in this paper the analysis and results obtained from all these X-ray data. Based on the fast variability in the X-ray domain, the spectral energy distribution in the 0.5-80 keV energy range, and the reported association with a highly reddened OB supergiant at ~10 kpc, we conclude that IGR J17503-2636 is most likely a relatively faint new member of the supergiant fast X-ray transients. Spectral analysis of the NuSTAR data revealed a broad feature in addition to the typical power-law with exponential roll-over at high energy. This can be modeled either in emission or as a cyclotron scattering feature in absorption. If confirmed by future observations, this feature would indicate that IGR J17503-2636 hosts a strongly magnetized neutron star with B~2e12 G.

Tables

Table 1. Table 1: Log of X-ray observations with the best-fit spectral parameters.

TELESCOPE	START	STOP	EXP.	OBSID	$n_H$		$\Gamma$		PL Flux (2–10 keV)a𝑎aa𝑎a$a$Intrinsic flux of the pegpowerlaw component in the 2–10 keV energy range.	Cstat-${\chi }_{\mathrm{red}}^2$	/d.o.f.b𝑏bb𝑏b$b$For INTEGRAL and NuSTAR, we provide the reduced ${\chi }^2$, for NICER and Swift/XRT, the C-stat.	Flux (2–10 keV)c𝑐cc𝑐c$c$Absorbed model flux in the 2–10 keV energy range.
	UTC	UTC	ks		${10}^{22}{\mathrm{cm}}^{-2}$				${10}^{-11}\mathrm{erg}{\mathrm{cm}}^{-2}{s}^{-1}$			${10}^{-11}\mathrm{erg}{\mathrm{cm}}^{-2}{s}^{-1}$
INTEGRAL	2018-08-10 15:52	2018-08-11 19:20	54.5d𝑑dd𝑑d$d$This is the IBIS/ISGRI effective exposure corrected for dead time, while each JEM-X unit has an exposure of 6 ks.	N/A	73		2.8	${}_{-0.4}{}^{}{}_{}{}^{+0.7}$	63${}_{-35}{}^{}{}_{}{}^{+103}$	0.6	/10	70
Swift/XRT	2018-08-13 19:43	2018-08-13 20:13	0.99	00010807001	13.0	$\pm$5.0	0.5	$\pm$0.6	14${}_{-2}{}^{}{}_{}{}^{+3}$	46.0	/53	8.9
NICER	2018-08-14 21:23	2018-08-14 22:43	0.76	1200320101	26.0	$\pm$5.5	2.4	$\pm$0.6	8.5${}_{-2.4}{}^{}{}_{}{}^{+4.8}$	175.0	/131	2.1
NICER	2018-08-16 02:14	2018-08-16 13:19	2.89	1200320103	5.0	$\pm$1.3	0.4	$\pm$0.3	1.6${}_{-0.1}{}^{}{}_{}{}^{+0.1}$	117.0	/141	1.3
NuSTARe𝑒ee𝑒e$e$The reported parameters are for the absorbed power-law component only and model Gabs in Table 2 (we refer to Sect. 2.2 and Table 2 for the complete NuSTAR analysis).	2018-08-22 11:23	2018-08-23 10:29	39.1	90401333002	21.0	$\pm$2.5	0.49	$\pm$0.12	4.1${}_{-0.2}{}^{}{}_{}{}^{+0.4}$	1.0258	/324	2.3
Swift/XRT	2018-08-23 04:39	2018-08-23 06:38	1.85	00088805001	56	${}_{-22}{}^{}{}_{}{}^{+27}$	2	$\pm$2	9.6${}_{-5.3}{}^{}{}_{}{}^{+31.0}$	42.0	/48	1.5

Table 2. Table 2: Spectral results obtained by fitting together the NuSTAR  FPMA and FPMB data of IGR J17503-2636 (taken from 2018 August 22 at 11:01 to August 23 at 10:26 UT)). Models and parameters are described in the text.

	Gabs		Gaussian
$N_H$ [${10}^{22}$cm-2]	21	${}_{-3}{}^{}{}_{}{}^{+2}$	24	$\pm$2
$E_C$ [keV]	12.4	$\pm$0.5	1	–
$E_F$ [keV]	7.8	$\pm$0.4	11	${}_{-1}{}^{}{}_{}{}^{+2}$
$\Gamma$	0.49	$\pm$0.12	0.51	$\pm$0.19
Fpl [${10}^{-11}$cgs]	12.0	${}_{-0.3}{}^{}{}_{}{}^{+0.4}$	19	$\pm$3
$E_{\mathrm{Fe}}$ [keV]	6.32	$\pm$0.02	6.33	$\pm$0.02
${\sigma }_{\mathrm{Fe}}$ [keV]	0.13	$\pm$0.07	0.12	$\pm$0.07
$N_{\mathrm{Fe}}$ [${10}^{-4}$ph/s/cm-2]	1.5	$\pm 0.2$	1.5	$\pm 0.2$
$E_{\mathrm{Cyc}}$ [keV]	20.1	$\pm$0.7	10.9	$\pm$0.2
${\sigma }_{\mathrm{Cyc}}$ [keV]	4.1	$\pm$0.8	3.1	$\pm$0.2
$\tau$/$N_{\mathrm{Cyc}}$	4	${}_{-1}{}^{}{}_{}{}^{+2}$	(1.2	$\pm 0.2)\times 10{}^{-3}$
Flux(0.5–100 keV)a𝑎aa𝑎a$a$Absorbed flux in the 0.5–100 keV energy range.	9.6 erg cm-2 s-1
${\chi }_{\mathrm{red}}^2/d.o.f.$	1.026	/324	0.932	/324

Equations

$$E_{cyc}=\frac{1}{(1+z)}\frac{\hslash eB}{m_{e}c}\approx \frac{1}{(1+z)}11.6\times B_{12}\mathrm{keV},$$

Full text

1 Introduction

IGR J17503-2636 is an X-ray transient discovered by the JEM-X instruments (Lund et al., 2003) on-board INTEGRAL (Winkler et al., 2003) during the Galactic Bulge observations performed on 2018 August 11 (satellite revolution 1986; Chenevez et al., 2018). At discovery, the source flux in the JEM-X data covering from 07:44 to 12:30 UT was estimated at 3.9$\times$10*-10* erg cm*-2* s*-1*(3–19 keV). The source was not initially detected in the quick-look analysis of the higher energy INTEGRAL instrument IBIS/ISGRI (Ubertini et al., 2003; Lebrun et al., 2003).

A 1 ks-long follow-up observation with Swift/XRT (Burrows et al., 2005) was carried out on 2018 August 13 at 19:41 UT in order to provide the first characterization of the source emission in the soft X-rays and improve the X-ray position down to arcsec accuracy. The preliminary analysis of XRT data revealed a variable absorption in the range (6.2–13.7)$\times$1022 cm*-2* (assuming a Galactic absorption column density in the direction of the source of 1.3$\times$1022 cm*-2*) and provided a measurement of the variable 0.3–10 keV flux during the observation in the range (2.0–8.3)$\times$10*-11* erg cm*-2* s*-1*(Chenevez et al., 2018). The most accurate localisation of the source was obtained from a 1 ks-long Chandra observation at RA(J$2000)$ = $17^{\rm h}50^{\rm m}17\aas@@fstack{s}99$, Dec(J$2000)=-26^{\circ}36^{\prime}16\aas@@fstack{\prime\prime}7$ with an associated uncertainty of 0$\aas@@fstack{\prime\prime}$7 at 90 % confidence level (the observation was carried out on 2018 August 23 at 23:31 UT; Chakrabarty et al., 2018a, b). Only eleven events from the source were recorded by Chandra, and no spectral or timing analysis could be performed (the estimated X-ray flux in the 0.5-10 keV energy band was 2.9$\times$10*-12* erg cm*-2* s*-1*). The accurate Chandra position allowed Masetti et al. (2018) to identify the IR counterpart of IGR J17503-2636 as an heavily reddened OB (super)giant star located beyond the Galactic Center at about 10 kpc, thus classifying the INTEGRAL transient as an high mass X-ray binary (HMXB; Walter et al., 2015a). The fast flaring behavior and the rapid decay in the X-ray flux suggests an association of IGR J17503-2636 to the HMXB sub-class of the supergiant fast X-ray transients (SFXTs; Sguera et al., 2005; Negueruela et al., 2006; Sguera et al., 2006; Martínez-Núñez et al., 2017), as we discuss in Sect. 3.

In this paper, we report on all available X-ray data that were collected during the first reported X-ray emission episode from IGR J17503-2636 with the instruments on-board INTEGRAL, NuSTAR, Swift, and NICER, together with our interpretations.

2 X-ray data

For all instruments, we performed spectral analysis with Xspec 12.10.0c (Arnaud, 1996). All uncertainties in the paper are given at 90% confidence level, unless stated otherwise. A summary of all observations with common spectral results is reported in Table 1; we refer to the following paragraphs for a detailed description of the spectral analysis for each X-ray facility.

2.1 INTEGRAL data

INTEGRAL observations are divided into science windows (ScWs), i.e., pointings with typical duration of $\sim$2–3 ks. To minimize calibration uncertainties and maximize the exposure, we have selected all publicly available pointings with a limited off axis angle from the source: 10 deg from IBIS/ISGRI and 4 deg for JEM-X222http://www.isdc.unige.ch/integral/analysis. All data were processed with the version 11.0 of the Off-line Scientific Analysis software (OSA) distributed by the ISDC (Courvoisier et al., 2003).

We extracted the IBIS/ISGRI and JEM-X mosaics by stacking all available data from 2018-08-10 at 15:50 to 2018-08-11 at 18:52 together. IGR J17503-2636 was detected in the IBIS/ISGRI 25–80 keV mosaic at a significance of $7\sigma$ (single trial, effective exposure time 54.5 ks) and in the JEM-X 3–25 keV mosaic at a significance of $6\sigma$ (single trial, effective exposure time 21.7 ks). We show a zoom of the IBIS/ISGRI and JEM-X mosaics around the position of IGR J17503-2636 in Fig. 1. We extracted the IBIS/ISGRI light curve in the 25–80 keV energy range and with one science window time granularity, but we could not detect any significant variability (at 3$\sigma$ confidence level). The JEM-X coverage (considering a maximum off-axis angle of four degrees) is limited to a few SCWs, not allowing for any variability study. Thus, we extracted a single spectrum integrating over the entire exposure time available for ISGRI, JEM-X1, and JEM-X2 data. The JEM-X (IBIS/ISGRI) spectra were computed in eight (five) logarithmic equally spaced bins between 3 and 35 keV (25 and 100 keV). Due to instrumental systematic uncertainties and inconsistencies between the two JEM-X units, we have limited the use of data in the range $\sim$5–25 keV for JEM-X2 and $\sim$7–25 keV for JEM-X1. These spectra (Fig. 2) could be well fit ($\chi_{\rm red}^{2}$/d.o.f.=0.6/10) with a simple power-law model (TBabs*pegpwrlaw in Xspec) with best-fit parameters reported in Table 1.

We have finally checked that during the following INTEGRAL visibility period (2018-08-17 14:03-2018-08-19 18:08), the source was not detected. We derived an upper limit on its X-ray flux of $4\times 10^{-10}$erg cm*-2* s*-1* in the 3–80 keV energy range at 3$\sigma$ confidence level (assuming a power-law photon index of 2.8).

2.2 NuSTAR data

IGR J17503-2636 was observed by NuSTAR (Harrison et al., 2013) from 2018 August 22 at 11:01 to August 23 at 10:26 (UT; ID 90401333002). After having applied to the NuSTAR data all the good time intervals (GTI) accounting for the Earth occultation and the South Atlantic Anomaly passages, we obtained an effective exposure time of 39.1 ks for both the focal plane modules A and B (FPMA and FPMB). All data were processed via nupipeline v0.46 and the latest calibration files available at the time of writing (v.20181022). The source spectra and light curves were extracted from a 80 arcsec circle centered on the source, while the background products were extracted from a region with a similar extension but centered on a region free from the contamination of both straylight and source emission. Various extraction regions were also used for the source and background products to verify that none of the timing and spectral features could be affected by some specific choices.

The FPMA and FPMB light curves of the source display a remarkable variability. We show in Fig. 3 the FPMA lightcurves of the source in two energy bands (3–10 keV and 10–60 keV) and the corresponding hardness ratio (HR) calculated with an adaptive rebinning of the lightcurves in order to achieve a signal-to-noise ratio (S/N) of at least 10 in each time bin (based on the soft lightcurve; see Bozzo et al., 2013, for more details). We verified that compatible results could be obtained from the FPMB lightcurves.

In order to investigate the origin of the pronounced variability characterizing the NuSTAR lightcurves, we carried out a timing analysis of the data by using event files where the arrival time of all recorded photons was framed at the Solar system barycenter using the barycorr tool. We accumulated light curves in the 3–60 keV energy range with bins of 0.005 s and summed FPMA and FPMB to increase the statistics. We built a power spectrum averaging on segments of 524 288 bins so that most satellite orbits contains one or more intervals. Segments resulting in an exposure of less than 50% were discarded, the others padded with zeros. Eventually, we averaged the power spectrum over 15 segments. In Fig. 4, we show the resulting power spectrum with the corresponding empirical model function for the combined white and red noise: a constant ($1.991\pm 0.002$) plus a power-law with index $-1.67\pm 0.09$.

From the power spectrum of Fig. 4, we can infer that there is no obvious periodicity. We used the best-fit model of the power spectral density (PSD) as the expected value of the power at each frequency and computed an upper limit on the pulsed fraction between 0.7 and 7% at 99% confidence level, using Eq. (A4) from Israel & Stella (1996).

Given the relatively low count-rate of the source recorded by NuSTAR and the lack of prominent HR variations in Fig. 3, we extracted and fit together the FPMA and FPMB spectra of the source accumulated by using the entire exposure time available, after applying an optimal binning (Kaastra & Bleeker, 2016). The source emission is mainly characterized by an absorbed power-law with a high-energy exponential cutoff and an iron line emission centered at $\sim$6.4 keV with equivalent width of about 0.3 keV. A fit with this simple model left, however, evident residuals especially around 10–20 keV: an additional component is required to obtain an acceptable result. We have verified that spectral fits with the most common phenomenological single-component models leave equivalent residuals (e.g., Fermi-Dirac cutoff, negative and positive power law-NPEX; see Coburn et al. 2002). We could not unequivocally determine whether this required additional component is in absorption or in emission, as both the models TBabs*(highecutpegpwrlawgabs + Gaussian) (referred as Gabs in Table 2) and TBabs*(highecut*pegpwrlaw + Gaussian + Gaussian) (Gaussian in Table 2) could successfully describe the data. Therefore, we report the best-fit parameters obtained by using both models in Table 2. In this table, $N_{\mathrm{H}}$ is the absorption column density of the Tbabs component, $E_{\mathrm{C}}$ ($E_{\mathrm{F}}$) is the cut-off (fold) energy of the highecut component, $\Gamma$ is the power-law photon index, and Fpl is the intrinsic flux of the power-law component. It should be noted that the cutoff energy could not be constrained for the model with the additional Gaussian emission feature and was fixed at 1 keV, outside of the energy range of NuSTAR data. We indicated with $E_{\mathrm{Fe}}$, $\sigma_{\mathrm{Fe}}$, and $N_{\mathrm{Fe}}$ the centroid energy, width, and normalization of the iron line, respectively. $E_{\mathrm{Cyc}}$, $\sigma_{\mathrm{Cyc}}$, and $\tau$/$N_{\mathrm{Cyc}}$ are the centroid energy, width, and depth/normalization of the absorption or emission component, possibly associated with cyclotron scattering.

2.3 Swift data

IGR J17503-2636 was observed by the Neil Gehrels Swift Observatory (Burrows et al., 2005) twice on 2018 August 13 at 19:43 (UT) for a total exposure time of 984 s and on 2018 August 23 at 04:39 (UT) for a total exposure time of 1848 s. In both cases, data were collected in photon counting (PC) mode.

The XRT data were analyzed via the standard software (Heasoft v6.25) and the latest calibration files available (CALDB 20180710). All data were processed and filtered with xrtpipeline (v0.13.4). We verified that no data were significantly affected by pileup. The source events were extracted using the photo counting observing mode from a circular region with a radius of 20 pixels (where 1 pix corresponds to $\sim 2\aas@@fstack{\prime\prime}36$), while background events were extracted from a source-free region with a similar radius. We show in Fig. 6 the background subtracted XRT lightcurve of the two observations in the 0.5–10 keV energy band corrected for point spread function losses and vignetting. The XRT spectra extracted from each observation were binned using the Kaastra & Bleeker (2016) algorithm and could be well fit with an absorbed power-law model (TBabs*pegpwrlaw in Xspec) by minimizing the Cash statistics (C-stat in Xspec). A log of XRT observations, together with the corresponding results from the spectral fits, is reported in Table 1. In Fig. 7, we show the count-rate spectra. It can be noted that during the second observation, the source faded significantly and the spectral parameters are less constrained.

2.4 NICER data

Following the discovery of IGR J17503-2636, Neutron star Interior Composition Explorer (NICER; Gendreau et al., 2016) took the opportunity to monitor this source between 2018-08-14 at 21:24 and 2018-08-16 13:20 UT with an effective exposure of $\sim$3.7 ks. We have analyzed data from the observations ID 1200320101 and 1200320103 using NICERDAS version 2018-04-24 and HEASoft 6.24 package. We reprocessed the data using the “nicerl2” pipeline. GTIs were created using the standard filtering criteria, e.g. the angle of bright Earth $>$40*∘, elevation $>$30∘*, pointing offset of $\leq$54 arcsec and excluding data collected during the passages through the South Atlantic Anomaly region. We applied the GTI for the spectral extraction using XSELECT. NICER observations of blank sky region (RXTE-6) were chosen for background measurements. We have used version 1.02 of the NICER response and effective area files for the spectral fitting.

We optimally rebinned the spectra using the algorithm by Kaastra & Bleeker (2016) and limited our energy range from 1 to 11 keV. Both NICER spectra can be well described with an absorbed power-law model (we used the model TBabs*pegpowerlaw). We report the results in Table 1 noting that we minimized the Cash statistic with background subtraction (C-stat in Xspec). The two spectra, together with the residuals from the best fit, are shown in Fig. 8.

The power density spectrum of the NICER observation is consistent with white noise from $\sim 10^{-3}$ to 500 Hz. The upper limits we could obtain on the pulsed fraction are less containing than those reported for NuSTAR data owing to the shorter exposure and high background of the NICER observations.

3 Discussion

We reported on the transient source IGR J17503-2636, whose first detectable X-ray emission was discovered with INTEGRAL on 2018 August 11. The source was followed up by Chandra, NuSTAR, Swift, and NICER.

IGR J17503-2636 was relatively faint for the instruments on-board INTEGRAL, and thus only a preliminary description of the broad-band spectrum could be obtained from the available JEM-X and IBIS/ISGRI data. INTEGRAL data showed that the X-ray emission from the source faded very rapidly, decreasing below the detection limit of JEM-X and IBIS/ISGRI within one revolution (2.7 days), supporting a fast flaring behavior. The more sensitive instruments on-board NuSTAR, Swift, and NICER revealed that significant X-ray emission from the source could be detected at least until early August 23, albeit at an intensity about a factor of $\sim$50 lower than that recorded by INTEGRAL. The Chandra observation carried out later during the same day showed a further decrease in intensity by at least another factor of $\sim$10 and down to 2.9$\times$10*-12* erg cm*-2* s*-1*(Chakrabarty et al., 2018a, b).

In the soft to hard X-ray domains (0.5–80 keV), the emission from IGR J17503-2636 is characterized by a remarkable variability on time scales from a few seconds to a few thousands of seconds, typical of what is usually observed in wind-fed high mass X-ray binaries (HMXBs; see Walter et al., 2015a, for a recent review). This variability is best appreciated by looking at the NuSTAR lightcurves which are endowed with a much higher S/N and longer exposure, compared to all other available instruments (see Fig. 3). Although within the NuSTAR observation we did not record prominent variations in the HR, a comparison between all spectra results444Note that we did not attempt to perform a combined fit between the different instruments because data from INTEGRAL, NuSTAR, NICER, and Swift were obtained at largely different epochs and the significant spectral changes recorded in these data would make it very difficult to provide a consistent interpretation of the results. Furthermore, given the large absorption column density revealed in the direction of the source, the limited S/N of the NICER and Swift data would not add significant information to the fit of the NuSTAR data that can satisfactorily cover alone a broad-band emission range extending from 3 up to 40 keV. reported in Sect. 2.1-2.4 highlights significant changes in the absorption column density (up to a factor of $\sim$10) and in the power-law slope (up to a factor of $\sim$3). These changes, as well as the much higher value of $N_{\rm H}$ compared to the expected Galactic extinction in the direction of the source (Chenevez et al., 2018), supports the idea of IGR J17503-2636 being a wind-fed HMXB. It is well known that in these systems the fast wind of a massive companion star (typically an OB supergiant) can lead to the formation of a cocoon of dense material around the accreting compact object and the local density/velocity variations of the wind can give rise to an X-ray variability on compatible time scales compared to those observed from IGR J17503-2636 (see, e.g., Martínez-Núñez et al., 2017, for a recent review).

Although the present observations do not allow to firmly establish the nature of the accreting compact object in this source, the broad-band spectral analysis revealed properties that are strongly reminiscent of what is usually observed from neutron star HMXBs (see, e.g., Klochkov et al., 2007; Walter et al., 2015b). The cut-off power-law spectrum is commonly observed in these systems, and the iron line at 6.4 keV is often observed as a consequence of the fluorescence of X-rays from the neutron star (NS) onto the surrounding stellar wind material or accretion disk. The peculiar feature around 10–20 keV that we modeled with either a Gaussian emission component or a multiplicative absorbing Gaussian profile is a known signature of a strongly magnetized NS. The interpretation of an emission feature at these energies is associated, as suggested for other HMXBs, to thermal and bulk Comptonization of magnetized bremsstrahlung seed photons along the accretion column (Ferrigno et al., 2009; Farinelli et al., 2016). The presence of an absorption feature would be explained by assuming this is a broad absorption line produced by scattering on electrons bounded around the field lines of a strong magnetic field (also known as “cyclotron line’). These features are observed in many NS HMXBs and their centroid energy provides a direct estimate of the NS magnetic field strength according to the equation:

[TABLE]

where $B_{\mathrm{12}}$ is the magnetic field in units of $10^{12}$ Gauss and $z$ is the gravitational redshift at the emission site (the above equation is valid for the fundamental CRSF; see, e.g. Staubert et al., 2018, for a recent review). If we accept this interpretation, we can conclude that the accreting object in IGR J17503-2636 is a NS endowed with a magnetic field of $\sim$2$\times$1012 G, compatible to other measured NS magnetic field strengths in HMXBs. The 4 keV width of the scattering feature is also typical for these objects and might indicate a plasma temperature of 10–20 keV, in agreement with the cutoff energy. However, this should be taken with caution, since the line width depends on the line-of-sight angle to the magnetic field. Future observations of additional episodes of enhanced X-ray emission from IGR J17503-2636 with a high sensitivity broad-band X-ray instrument as the FPMs on-board NuSTAR will hopefully be able to provide higher S/N spectra and rule out any other alternative interpretation of the feature around 20 keV from the spectral fitting.

A final convincing indication of the HMXB nature of IGR J17503-2636 is provided by the identification of the near-infrared counterpart reported by Masetti et al. (2018). This led to the specific association of the source with the class of the supergiant HMXBs555Note that other authors have questioned this conclusion, even though a firm identification of this object might require additional observations (McCollum et al., 2018).. These systems are generally divided into two sub-classes, the so-called classical systems and the supergiant fast X-ray transients (SFXTs; see, e.g., Martínez-Núñez et al., 2017, for a recent review). The former are variable but persistent systems, showing on average a luminosity that is well explained by using a wind accretion scenario onto a compact object (usually a strongly magnetized neutron stars). The SFXTs display a much prominent variability in the X-rays, alternating between hours-long outbursts reaching the typical luminosity of classical systems and extended periods of quiescence where the X-ray luminosity can decrease by up to a factor of 105–106. This behavior is far from being understood and it is still actively debated (see, e.g., the discussion in Bozzo et al., 2017). Data collected so far from IGR J17503-2636 would favor the connection with the SFXT sub-class, as the source underwent a relatively bright episode of X-ray emission at the time of the discovery with JEM-X and then progressively fainted down until it got close to the detection threshold also for Chandra (the X-ray bright phase displayed by IGR J17503-2636 was about $\sim$12 days long in total, a duration that is not uncommon in other SFXTs; see, e.g., Sguera et al., 2015). Although the total variation in the X-ray luminosity recorded so far is of $\sim$300 (see Sect. 1) and thus significantly lower than that usually measured from the SFXTs, the fact that no X-ray emission was ever recorded before from this object suggests that the true quiescent luminosity could be lower than the value measured during the Chandra observation. The overall dynamic range in the X-ray domain could, therefore, be even larger. Masetti et al. (2018) suggested from the reddening of the OB supergiant that this system is located beyond the Galactic Center at $\sim$10 kpc, and thus the outburst luminosity derived from the JEM-X flux would be of $\sim$2$\times$1036 erg s*-1*, while the NuSTAR 0.5–100 keV flux would correspond to a luminosity of $\sim$1036 erg s*-1*. This value is similar to what is observed from the faintest SFXT outbursts, that can achieve a luminosity up to $\gtrsim$1038 erg cm*-2* s*-1*(Romano et al., 2015). As a consequence, it is possible that this is a peculiarly faint system also during outbursts, explaining why no previous detection with INTEGRAL or Swift/BAT was ever reported whilst usually up to few outbursts per year are detected from the known SFXTs (see, e.g., Paizis & Sidoli, 2014; Romano, 2015; Sidoli & Paizis, 2018).

Acknowledgements

We are grateful to the NuSTAR, Swift, and NICER teams for the prompt scheduling of the ToO observations on IGR J17503-2636. We acknowledge financial contribution from the agreement ASI-INAF n. 2017-14-H.O. We acknowledges support from the HERMES Project, financed by the Italian Space Agency (ASI) Agreement n. 2016/13 U.O. GKJ acknowledges support from the Marie Skłodowska-Curie Actions grant no. 713683 (H2020; COFUNDPostdocDTU)