© ESO, 2012

1. Introduction

The region around Sgr A^∗, the dynamical centre of our Galaxy, has been observed at various spatial scales and in different energy bands by many past and present X-ray missions. At early times dedicated monitoring campaigns using Einstein (Watson et al. 1981), Granat (Churazov et al. 1994; Pavlinsky et al. 1994), ROSAT (Sidoli et al. 2001), ASCA (Sakano et al. 2002) and BeppoSAX (Sidoli et al. 1999; in ’t Zand et al. 2004) have led to the discovery of several X-ray point sources located within the central degrees of our Galaxy.

More recently, an intensive monitoring campaign carried out with Chandra between 1999 and 2006 has resolved thousands of distinct X-ray sources in a field of 2° × 0.8° around the Galactic centre (GC; Wang et al. 2002; Baganoff et al. 2003; Muno et al. 2003b, 2004, 2006, 2009). Furthermore, starting in 2006 the inner ~25′ × 25′ around Sgr A^∗ has been monitored on an almost daily basis with Swift (Kennea & The Swift/XRT team 2006; Degenaar & Wijnands 2009, 2010), whereas a region subtending many degrees has been regularly scanned by RXTE since 1999 (Swank & Markwardt 2001) and by INTEGRAL since 2005 (Kuulkers et al. 2007b).

The plethora of X-ray sources found in the direction of the innermost parts of our Galaxy encompasses a variety of objects (e.g. Muno et al. 2004). The population of X-ray sources with luminosities of L_X ≲ 10³³ erg s^-1 (2–10 keV) is thought to be dominated by accreting white dwarfs and active stars (e.g. Verbunt et al. 1984, 1997; Muno et al. 2006; Revnivtsev et al. 2009)¹. The brightest Galactic X-ray point sources, however, have peak luminosities of L_X ~ 10³⁶⁻³⁹ erg s^-1 and can be identified with accreting neutron stars or black holes. Based on the type of companion star, these are classified as either high-mass X-ray binaries (HMXBs; donor mass ≳10 M_⊙) or low-mass X-ray binaries (LMXBs; donor mass ≲1 M_⊙).

Both LMXBs and HMXBs can be transient: such systems spend the majority of their time (years/decades) in a quiescent state during which the X-ray luminosity is generally L_X ≲ 10³³ erg s^-1, while their intensity typically reaches up to L_X ~ 10³⁶⁻³⁹ erg s^-1 during short (weeks/months) outburst episodes. This large X-ray variability is ascribed to changes in the mass-accretion rate onto the compact primary. In LMXBs this is thought to be caused by instabilities in the accretion disk or the Roche-lobe overflowing companion star. In HMXBs the compact primary accretes matter that is expelled by the companion star via a wind or a circumstellar disk, and transient behaviour may be the result of variations in the mass-loss rate of the companion or the binary geometry.

1.1. Low-luminosity X-ray transients

Repeated observations of the region around Sgr A^∗ with Chandra, XMM-Newton and Swift have revealed a population of transient X-ray sources that have 2–10 keV peak luminosities of L_X ~ 10³⁴⁻³⁶ erg s^-1 (e.g. Muno et al. 2005b; Sakano et al. 2005; Porquet et al. 2005b; Degenaar & Wijnands 2009). Their outburst amplitudes and spectral properties suggest that these objects are X-ray binaries in which the compact object accretes at a very low rate from its companion star, thereby causing a relatively low X-ray luminosity (e.g. Pfahl et al. 2002; Belczynski & Taam 2004; Muno et al. 2005b; Wijnands et al. 2006a). Earlier X-ray missions already provided a glimpse of such low-luminosity X-ray transients (Sunyaev 1990; in ’t Zand et al. 1991; Maeda et al. 1996), but the current generation of instruments exploiting sensitive and high spatial resolution X-ray imaging have considerably improved our understanding of the number and behaviour of such objects.

Observations of X-ray binaries accreting at low X-ray luminosities can address several questions related to stellar and binary evolution, as well as accretion flows at low rates. For instance, constraining the number and nature of low-luminosity X-ray transients allows us to gain more insight into the statistics of different source classes, and can serve as an important calibration point for population synthesis models (e.g. Pfahl et al. 2002; Belczynski & Taam 2004). Furthermore, the mass-accretion rate averaged over thousands of years, ⟨Ṁ⟩ _long, plays an important role in the evolution of LMXBs. First studies have shown that values of ⟨Ṁ⟩ _long < 10^-13 M_⊙ yr^-1 put tight constraints on the possible evolutionary paths, and might require an unusual type of binary such as neutron stars accreting from a hydrogen-depleted or planetary companion (King & Wijnands 2006). Monitoring observations are an important tool for estimating the time-averaged mass-accretion rates of low-luminosity transients (Degenaar & Wijnands 2009, 2010). Repeated non-detections allow us to place more stringent upper limits on their long-term averaged accretion rates and can therefore be as interesting as actual detections.

Furthermore, thermonuclear X-ray bursts observed from slowly accreting neutron stars have provided important new insight into the physics of nuclear burning on the surface of these compact objects (e.g. Cornelisse et al. 2002; in ’t Zand et al. 2005; Cooper & Narayan 2007; Peng et al. 2007; Degenaar et al. 2010a). Moreover, studying the low-luminosity transients gives insight into the accretion process at low mass-accretion rates. Some bright transient X-ray binaries that exhibit accretion outbursts with intensities of L_X ≳ 10³⁶ erg s^-1 have also been observed to undergo episodes of low-level accretion with X-ray luminosities of L_X ~ 10³³⁻³⁵ erg s^-1 (e.g. Wijnands et al. 2001, 2002; Linares et al. 2008; Sidoli et al. 2008; Degenaar & Wijnands 2009; Degenaar et al. 2011a; Fridriksson et al. 2011; Romano et al. 2011). This provides an interesting comparison with the outbursts of the low-luminosity X-ray transients.

1.2. This work

Dedicated surveys aiming to search for and to identify low-luminosity X-ray transients have the potential to unveil rare types of accreting compact objects, and can provide valuable input for theoretical models. Here, we report on the results from a joint Chandra and XMM-Newton monitoring campaign that covered a region of 1.2 square degrees around Sgr A^∗ and was carried out between 2005–2008. Wijnands et al. (2006a) reported on the initial results of our monitoring observations, discussing the first series of data obtained in 2005 June–July.

The main goal of this programme is to investigate the X-ray properties (below 10 keV) of transient objects that have low 2–10 keV peak luminosities of L_X ~ 10³⁴⁻³⁶ erg s^-1. Owing to the high concentration of X-ray point sources in the inner square degree around Sgr A^∗, as well as sensitivity limitations, such systems are often inaccessible to the current monitoring instruments. Our Chandra/XMM-Newton campaign provides the means to refine our understanding of currently known low-luminosity systems (e.g. characterise their outburst behaviour, improve estimates of their duty cycles and time-averaged mass-accretion rates), to search for new X-ray transients and to capture low-level accretion activity in bright X-ray binaries.

We present our work as follows. We describe the setup of the programme in Sect. 2, and proceed by detailing the reduction and analysis procedures in Sect. 3. The results of our temporal and spectral analysis of ten different X-ray sources are presented on a case-by-case basis in Appendix A, while in Sect. 4 we summarise the main results and highlight a few individual sources. We end in Sect. 5 where we give an overview of all transient X-ray sources located in the region covered by our campaign. We discuss the implications of our findings for understanding the nature of the GC X-ray transients and low-level accretion activity.

2. Description of the monitoring programme

2.1. Observations

Our choice to monitor the central square degree of our Galaxy was motivated by the fact that this region is populated by nearly 20 known X-ray transients, several of which undergo sub-luminous accretion episodes (Muno et al. 2005a; Wijnands et al. 2006a; Kennea et al. 2006; Degenaar & Wijnands 2009; Degenaar et al. 2011b). The relatively wide field of view (FOV; ~30′ × 30′) and large collecting area (~1100 cm² at 1 keV) of XMM-Newton make it an excellent facility for surveying sky regions down to relatively faint flux levels.

We used the data obtained with the European Photon Imaging Camera (EPIC), which consists of one PN (Strüder et al. 2001) and two MOS (Turner et al. 2001) detectors that are sensitive in the 0.1–15 keV range and have spectral imaging capabilities. The PN is an array of 12 CCDs (64    ×    200 pixels each), while the MOS units are composed of an array of seven CCDs, each consisting of 600 × 600 pixels. A micrometeorite strike damaged one of the CCDs of the MOS1, which is operated with only six detectors since then (Abbey et al. 2006).

The XMM-Newton observations are complemented by Chandra pointings that provide high spatial (sub-arcsecond) resolution and a very low X-ray background within an energy band of 0.1–10 keV. We chose the High Resolution Camera (HRC; Kenter et al. 2000) as our prime Chandra instrument, because it provides the largest FOV (~30′ × 30′), comparable in size to XMM-Newton. The HRC-I is a square micro-channel plate detector (made up of 32   768 × 32   768 pixels) that has an effective area of 225 cm² at 1 keV and is designed for imaging observations. Because the energy resolution of the HRC is poor, we obtained a few additional pointings with the Advanced CCD Imaging Spectrometer (ACIS; Garmire et al. 2003) to follow up active transients, aiming to obtain spectral information. The ACIS-I consists of a four-chip imaging array (each having 1024 × 1024 pixels), providing an effective area of 340 cm² at 1 keV and a FOV of ~16′ × 16′.

Our Chandra/XMM-Newton campaign covers 1.2 square degrees around Sgr A^∗, sub-divided into seven different pointing directions (named GC-1, GC-2, GC-3, GC-4, GC-5, GC-6 and GC-7; Wijnands et al. 2006a)². Adjacent pointings partially overlapped by a few arcminutes (see Fig. 1). The programme comprises 34 Chandra/HRC-I and 35 XMM-Newton/EPIC pointings, carried out in ten different epochs between 2005 June and 2008 September. An overview of the monitoring observations is given in Table 1. Follow-up Chandra/ACIS-I observations were performed in 2005 July (one pointing) and 2007 March–May (a series of five pointings) as listed in Table 2.

2.2. Sensitivity and X-ray images

The exposure time of individual observations was typically 5–10 ks. Depending on the spectral properties, a 5 ks Chandra/HRC-I pointing can detect sources (near aimpoint) down to a 2–10 keV luminosity of L_X ~ (3−6) × 10³³ (D/8 kpc)² erg s^-1 for photon indices of Γ = 1.0−3.0 and hydrogen column densities of N_H = (5−10) × 10²² cm^-2. The XMM-Newton observations are a factor of a few more sensitive. With our programme we reached X-ray luminosities that are a factor of ~100−1000 deeper than the sensitivity of wide-field monitoring instruments (e.g. RXTE/ASM, RXTE/PCA, Swift/BAT, MAXI, INTEGRAL), which are typically limited to L_X ≳ 10³⁵ (D/8 kpc)² erg s^-1 in their instrument passbands. In addition, XMM-Newton and Chandra provide (sub-) arcsecond spatial resolution, while these other instruments typically yield positional uncertainties of tens of arcseconds to arcminutes that can cause source confusion.

The entire programme spanned a period of 39 months (3.25 yr), for a cumulative exposure time of 412.7 ks (168.1 ks with Chandra, 244.6 ks with XMM-Newton). Subsequent pointings were separated by 2–10 months (see Table 1). The total exposure time reached in the different pointing directions is ~46–68 ks. Mosaic images of the Chandra/HRC-I and XMM-Newton/PN data are shown in Fig. 1. Apart from diffuse X-ray structures (e.g. around Sgr A^∗), these images reveal several X-ray point sources (see also Sect. 4). The locations of active transients and two persistent X-ray binaries (1E 1743.1–2843 and 1A 1742–294) are indicated by circles and the cross-hair in the centre of the images shows the position of the Sgr A^∗ complex. Figure 1 also includes a zoomed Chandra/HRC image of the inner ~1.5′ around Sgr A^∗, where three active X-ray transients were detected.

3. Data analysis

For the present study we were only interested in (candidate) transient X-ray binaries. We therefore focused on transient X-ray sources that have a 2–10 keV peak luminosity L_X ≳ 1 × 10³⁴ erg s^-1 for an assumed distance of 8 kpc, since there will be a high fraction of cataclysmic variables among the fainter objects (Verbunt et al. 1997; Muno et al. 2003c, 2009). We searched for transient sources in our Chandra and XMM-Newton data by comparing images of different epochs with one another.

The objects detected in our observations were correlated with the SIMBAD database to identify the known X-ray sources in our sample based on positional coincidence. Furthermore, we overlaid the positions of sources found in our campaign on an optical image from the Digital Sky Survey (DSS) and an infra-red image from the Two Micron All Sky Survey (2MASS), to filter out likely foreground objects (e.g. active stars). Spectral information obtained from our XMM-Newton and Chandra/ACIS-I observations also aids to identify transients that are located near or beyond the GC (i.e., at a distance of D ≳ 8 kpc). These sources will appear relatively hard in X-rays, since the softer photons (below ~2 keV) will be strongly absorbed by the interstellar medium in the direction of the GC (hydrogen column densities of several times 10²² atoms cm^-2 are typical in this region). X-ray sources with detectable emission below 2 keV are likely to be foreground X-ray active stars or cataclysmic variables that are located within a few kiloparsecs from the Sun.

To characterise the X-ray spectra and to calculate source fluxes, we fitted the obtained spectral data between 0.5–10 keV using XSpec (v. 12.6; Arnaud 1996). We used a simple powerlaw model (POWERLAW), modified by interstellar absorption (PHABS). For this we used the default abundances and cross-sections available in XSpec. Using the tool grppha, we grouped the spectra of the brightest sources to contain a minimum number of 20 photons per bin, whereas the spectra of fainter objects were binned into groups of at least 10 or 5 photons. For a small number of counts we also fitted the unbinned spectra without background subtraction using Cash-statistic (CSTAT in XSpec). Since this yielded spectral parameters and fluxes that were consistent with those obtained using the minimum χ²-method, we only report on the results obtained using the latter.

Whenever a source was detected during multiple observations, we fitted the spectra simultaneously with the hydrogen column density tied between the individual observations. We converted the deduced unabsorbed 2–10 keV fluxes into luminosities by adopting a distance of 8 kpc, unless better distance estimates were available for sources, e.g. as inferred from type-I X-ray burst analysis. Finally, we created long-term lightcurves for each transient source detected during our campaign.

The detailed data reduction and analysis procedures for both satellites are discussed in the next sections. Our analysis includes the Chandra observations performed in 2005 June (HRC) and July (ACIS), which were previously reported by Wijnands et al. (2006a).

3.1. Chandra

3.1.1. Data reduction and source detection

The Chandra data were reduced and analysed using the ciao tools (v. 4.2). The ACIS-I observations were carried out in the faint data mode with the nominal frame time of 3.2 s. As an initial step, we reprocessed the HRC and ACIS level-1 data files following the standard data preparation procedures³. Each individual pointing was inspected for periods of unusually high background. No significant background flares were found during our Chandra observations, therefore we used all data for the subsequent analysis.

We searched the data for X-ray sources by employing the WAVDETECT tool with the default “Mexican Hat” wavelet (Freeman et al. 2002). To take into account sensitivity variations across the chips, we generated an exposure map for each observation, evaluated at an energy of 4 keV (the approximate energy at which we expect to detect the largest number of photons for X-ray binaries). For each HRC-I observation, we generated images with a binning of 4, 16 and 32 pixels and ran the detection algorithm on each of the separate images with the default input parameters. This approach allowed us to cover a range of source sizes, accommodating the variation of the point spread function (PSF) as a function of off-axis angle. We adopted a recommended significance threshold that is approximately the inverse of the total number of pixels in the image (1 × 10^-8, 1 × 10^-7 and 1 × 10^-6 for HRC images binned by a factor of 4, 16 and 32 respectively), which should correspond to about one expected spurious source per image (Freeman et al. 2002). We compiled a master source list for each observation by combining the objects detected at each image resolution.

The ACIS-I images were binned by a factor of 2 and we employed WAVDETECT using a detection significance threshold of 1.5 × 10^-5. We ran the detection routine separately for the 0.5–2 and 2–10 keV bands, to be able to distinguish between soft and hard X-ray sources.

3.1.2. Count rates, lightcurves and spectra

We extracted net count rates and lightcurves for each source employing the tool DMEXTRACT (0.5–10 keV). We employed extraction regions centred on the positions found by the WAVDETECT routine and containing ~95% of the source counts. Depending on the brightness of the sources and their offset from the aimpoint of the observations, this corresponded to extraction radii of ~2–25″. Background events were collected from a source-free region that had a radius of three times that of the source region. We visually inspected lightcurves with bin times of 5, 10 and 100 s to search for features such as thermonuclear X-ray bursts.

For the ACIS-I data, we extracted source and background spectra using PSEXTRACT. Redistribution matrices (rmf) and ancillary response files (arf) were subsequently generated using the tasks MKACISRMF and MKARF, respectively. Since the HRC provides poor energy resolution, we converted the HRC-I count rates to 2–10 keV unabsorbed fluxes employing pimms (v. 4.1) and using either the spectral information deduced from our Chandra/ACIS and XMM-Newton observations, or values reported in literature (see Table 3). If a transient source was not detected, we obtained a 2σ Bayesian statistical upper limit on the source count rate using the ciao tool ASPRATES.

Both AX J1745.6–2901 and GRS 1741–2853 caused pile-up in the ACIS data. When left uncorrected, pile-up typically causes the broad-band count rate to be underestimated and the spectrum to become harder. In an attempt to circumvent these effects we used an iterative approach in which we extracted source photons from annular extraction regions with increasingly large parts of the core PSF excluded. Once the spectral parameters remained unchanged after increasing the annular radius, we assumed that the piled-up inner regions (which distort the spectral shape) were sufficiently excluded. The necessary aperture correction to the arf file was administered using ARFCORR.

3.2. XMM-Newton

3.2.1. Data reduction and source detection

In all XMM-Newton observations the EPIC cameras were operated in full window mode. Data reduction and analysis was carried using the Science Analysis Software (sas; v. 10.0.0) and following standard analysis threads⁴. Starting with the original data files, we reprocessed the MOS and PN data using the tools EMPROC and EPPROC, respectively. In order to asses the background conditions in each of the XMM-Newton observations, we extracted the full-field lightcurve for pattern-0 events with energies of ≳10 keV for the MOS, and between 10–12 keV for the PN. This revealed that some of our observations contained background flares. We excluded these episodes by selecting only data with high-energy count rates below 0.5 counts s^-1 for the MOS and below 1.0 counts s^-1 for the PN.

Source detection was carried out with the task EDETECTCHAIN, adopting the default detection likelihood of 10. We searched in two different energy bands of 0.5–2 and 2–10 keV for the PN and the MOS2. We did not include the MOS1 for source detection, since one of the CCD units was damaged by a micrometeoroid strike (Abbey et al. 2006)⁵.

3.2.2. Count rates, lightcurves and spectra

We extracted count rates for all objects in our source list using the task EREGIONANALYSE (0.5–10 keV). We also employed this tool to determine the optimum extraction region (achieving the highest signal to noise ratio) for source lightcurves and spectra. This yielded source regions with radii of ~10–100″ and a typical enclosed energy fraction of ~85–95%. For the extraction of background events we used regions with a radius three times larger than that of the source, positioned on a source-free portion of the CCD. For the observations in which our transient sources were not detected, we obtained a 2σ upper limit on the count rate using EREGIONANALYSE.

We created background-corrected lightcurves at a resolution of 5, 10 and 100 s for the PN and both MOS cameras using the tools EVSELECT and EPICLCCORR. These were visually inspected for features such as thermonuclear X-ray bursts. Source and background spectra, as well as the associated rmf and arf files, were generated using the meta task ESPECGET. The spectral data were fitted within XSpec with all model parameters tied between the three EPIC detectors.

During our observations, both AX J1745.6–2901 and SAX J1747.0–2853 became bright enough to cause pile-up in the EPIC instruments (see Sect. 4). We used the sas task EPATPLOT to evaluate the level of pile-up in the MOS and PN data, using annular regions of increasing size. Once the observed pattern distribution matched the expected one, we chose that annular size to extract source photons.

4. Results summary and highlights

Each of our Chandra and XMM-Newton observations reveals tens of distinct X-ray sources. Amongst the detected objects are the Arches cluster and Sgr A^∗, which are both complexes of X-ray point sources combined with diffuse emission structures. Several of the X-ray point sources found in our observations could be identified with known stars or had clear DSS/2MASS counterparts, which renders them likely foreground sources.

We detected two persistent X-ray binaries during our campaign: 1E 1743.1–2843 and 1A 1742–294. The former is an LMXB black hole candidate (e.g. Porquet et al. 2003), whereas 1A 1742–294 is a neutron star LMXB that displays type-I X-ray bursts (e.g. Pavlinsky et al. 1994). Our monitoring observations caught a total of six type-I X-ray bursts from this source⁶. Since the focus of our present work lies on transient X-ray sources, we do not discuss these two persistent X-ray binaries in more detail.

During our programme we detected activity from eight previously known X-ray transients, several of which exhibited multiple outbursts during our campaign. The results of our spectral and temporal analysis are presented in detail for each of the individual sources in Appendix A, where we also touch upon their long-term X-ray behaviour. In addition to the eight previously known transients, we found two X-ray sources that were detected only in a subset of our observations. Although this possibly indicates a transient nature, we argue that both are likely weak persistent X-ray sources with intensities close to the detection limit of our observations (see Sect. 4.4).

The results of our spectral analysis of these ten X-ray sources are summarised in Table 3. This table gives an overview of all observations in which a particular source was detected. For clarity we have assigned source numbers that correspond to the individual subsections of Appendix A. All fluxes listed in Table 3 and reported elsewhere in this work are given for the 2–10 keV energy band. Unless stated otherwise, luminosities were calculated from the unabsorbed flux by assuming a source distance of 8 kpc. All quoted errors refer to 90% confidence levels. Some of the detected transients were within the FOV of two different pointing directions. In these cases we report only the information extracted from the observations in which the source is nearest to the aimpoint. In the following sections we highlight some of the results of our campaign.

4.1. Lightcurve features of AX J1745.6–2901

The transient neutron star LMXB AX J1745.6–2901 was active during several of our observations and exhibited two distinct accretion outbursts during our campaign (2006 and 2007–2008; see Appendix A.2). Investigation of the X-ray lightcurves revealed two prominent features that we discuss below.

4.1.1. X-ray eclipse

During the Chandra/ACIS-I observation performed on 2007 April 6, we detected a strong reduction in the X-ray flux of AX J1745.6–2901 (left panel of Fig. 2). This event has every characteristic of an X-ray eclipse, which is caused by temporal obscuration of the central X-ray emitting region by the companion star. X-ray eclipses allow for a direct determination of the orbital period and can be used to study any possible evolution of the binary orbit (e.g. Wolff et al. 2009). Only a handful of neutron star LMXBs are known to display X-ray eclipses.

We barycentred the lightcurve using the tool BARYCEN to determine the time at which the eclipse occurred. The eclipse started on 2007 April 6 around 16:03 UTC and had a total duration of ~1600 s. The ingress of the eclipse lasted for ~400 s, while the egress time was ~200 s. The eclipse was not total, with a residual count rate of ~20% of the out-of-eclipse emission (0.5–10 keV). This profile matches the description of the eclipses that were seen by ASCA and the mid-eclipse time is consistent within the uncertainties of the ephemeris derived by Maeda et al. (1996). This leaves no doubt that this X-ray transient seen active during our campaign is AX J1745.6–2901. Other authors also reported on X-ray eclipses with similar characteristics seen during XMM-Newton (2007 March–April) and Chandra (2008 May) observations (Porquet et al. 2007; Heinke et al. 2008). A detailed study of the ASCA-detected eclipses suggested that the binary is seen at an inclination angle of ~70° and harbours a G-dwarf companion star in an ~8.4-h orbit (Maeda et al. 1996).

We extracted lightcurves in two different energy bands to investigate whether the eclipse properties are energy-dependent. We compare the 0.5–5.0 and 5.0–10 keV energy bands, because this yields similar intensities for the non-eclipsed emission. Figure 2 (left) compares the eclipse profiles. The duration, ingress and egress time is similar for both energy ranges, but the eclipse is deeper in the harder band. At the base of the eclipse the intensity in the 0.5–5.0 keV band is ~30% of the out-of-eclipse emission, whereas this is only ~10% for the 5.0–10 keV band. This may indicate that the softer X-ray photons come from a region that is more extended than the emission site of the harder photons. Similar eclipse behaviour has been observed for other LMXBs and explained in terms of a dust-scattering halo (e.g. Homan et al. 2003; Ferrigno et al. 2011).

4.1.2. X-ray burst

The Chandra/HRC data obtained on 2008 May 10 revealed an X-ray flare from AX J1745.6–2901 that started around 2008 May 11 at 00:04 UT and had a duration of ~50 s (right plot in Fig. 2). The fast rise and exponential decay shape, combined with the fact that this source is a known X-ray burster, strongly suggest that this event was a type-I X-ray burst. These are bright flashes of thermal X-ray emission that are caused by unstable thermonuclear burning of the accreted matter on the surface of the neutron star. Unfortunately, the HRC data does not allow for a spectral confirmation.

Using N_H = 21.8 × 10²² cm^-2 (see Table 3), and assuming a blackbody temperature typically seen for type-I X-ray bursts (kT_BB = 2–3 keV), the observed HRC peak count rate translates into a 0.01–100 keV luminosity of L_X ~ 5 × 10³⁷ erg s^-1 for an assumed distance of D = 8 kpc. The duration and peak intensity of the flare match other thermonuclear bursts detected from AX J1745.6–2901 (Maeda et al. 1996; Degenaar & Wijnands 2009). The duration suggests that helium is ignited in a hydrogen-rich environment (cf. Galloway et al. 2008). The persistent accretion luminosity at the time of the X-ray burst was L_X ~ 1 × 10³⁶ (D/8 kpc)² erg s^-1 (see Table 3), corresponding to ~0.5% of the Eddington rate of a neutron star.

4.2. Transient nature and X-ray bursts of SAX J1747.0–2853

4.2.1. Quiescent luminosity

The neutron star LMXB SAX J1747.0–2853 has been detected on numerous occasions and by different satellites ever since its discovery in 1998 (Appendix A.3). Wijnands et al. (2002) observed the source with Chandra in between two bright outburst and detected it at a 0.5–10 keV luminosity of L_X ~ 2 × 10³⁵ (D/6.7 kpc)² erg s^-1. Since this is 2–3 orders of magnitude higher than the typical quiescent X-ray luminosity of neutron star LMXBs, the transient nature of SAX J1747.0–2853 was cast in doubt (Wijnands et al. 2002). However, an apparent quenching of type-I X-ray bursts suggested that the accretion was suppressed at least during some intervals (Werner et al. 2004).

SAX J1747.0–2853 was active during several of our monitoring observations, but there were also epochs during which the source went undetected (Appendix A.3). We infer upper limits on the 2–10 keV luminosity of L_X ≲ (4−40) × 10³³ (D/6.7 kpc)² erg s^-1 for individual observations. This favours a classification as transient X-ray source. We further improve these constraints by using archival Chandra/ACIS-I observations performed on 2006 October 22 (obs ID 7157, ~15 ks exposure) and 2007 February 14 (obs ID 7048, ~39 ks). The combined ACIS-I image shows an excess of photons at the location of SAX J1747.0–2853 with a significance of 2.2σ. For an absorbed powerlaw model with N_H = 9.5 × 10²² cm^-2 and Γ = 2.6 (see Table 3), the net source count rate of ~6 × 10^-5 counts s^-1 translates into a 2–10 keV luminosity of L_X ~2 × 10³¹ (D/6.7 kpc)² erg s^-1. This clearly establishes the transient nature of SAX J1747.0–2853. During the Chandra observations presented by Wijnands et al. (2002) the source was thus likely detected in a low-level accretion state.

4.2.2. Short recurrence time X-ray bursts

SAX J1747.0–2853 is a known X-ray burster and we detected two such events during our campaign. The lightcurve extracted from the 2006 February XMM-Newton observation revealed a pair of type-I X-ray bursts occurring around 07:57 and 08:01 UT (Fig. 3). The time elapsed between the end of the first and the start of the second burst is 230 s (3.8 min). Both show the typical fast rise, exponential decay shape and are of similar duration (~40 s) and peak intensity (Fig. 3). We analysed the spectra of the X-ray bursts using only the MOS cameras, because the PN switched off during the bursts (possibly due to the high count rate). To avoid pile-up we extracted events from an annulus with inner radius of 20″ and an outer radius of 40″.

Limited by statistics, we cannot perform time-resolved spectroscopy; therefore we extracted the full-burst spectra (spanning ~40 s of data). A spectrum obtained from an interval of 100 s preceding the first burst was used as a background to account for the underlying persistent emission. We fitted both burst spectra to an absorbed blackbody model with the hydrogen column density fixed to the value obtained from fitting the persistent emission (N_H = 9.5 × 10²² cm^-2; Table 3). For the first X-ray burst this yields a blackbody temperature of kT_bb = 1.8 ± 0.4 keV and an emitting radius of R_bb = 5.2 ± 0.9 km. For the second burst we obtain comparable values of kT_bb = 1.4 ± 0.3 keV and R_bb = 7.4 ± 1.0 km.

Extrapolation of the blackbody fits to the energy range of 0.01–100 keV yields an estimate of the bolometric luminosity of L_bol ~ 3.6 × 10³⁷ and 3.1 × 10³⁷ (D/6.7 kpc)² erg s^-1 for the first and second burst, respectively. Using the count rate to flux conversion factor inferred from fitting the average burst spectra, we estimate that the bolometric peak luminosities of both bursts reached L_bol ~ 1.0 × 10³⁸ (D/6.7 kpc)² erg s^-1. This is close to the Eddington limit for neutron stars. Fitting the spectral data of the persistent emission shows that the source was accreting at ~1% of the Eddington rate when the burst doublet occurred. The duration of the bursts (~40 s) is typical of helium ignited in a hydrogen-rich environment (cf. Galloway et al. 2008).

The short time-interval of 3.8 min between the two consecutive type-I X-ray bursts is amongst the shortest measured for burst doublets⁷. Other bursts with similarly short recurrence times were reported from the neutron star LMXBs IGR J17480–2446 (3.3 min; Motta et al. 2011), 4U 1705–44 (3.8 min; Keek et al. 2010), 4U 1636–536 (5.4 min; Linares et al. 2009), 4U 1608–52 (4.3–6.4 min; Galloway et al. 2008) and EXO 0748-676 (6.5 min; Boirin et al. 2007; Galloway et al. 2008). Type-I X-ray bursts repeating within minutes present a challenge to our understanding of burst physics.

Current theoretical models predict that over 90% of the accreted hydrogen/helium is burned during a type-I X-ray burst, which implies that it would take at least a few hours to accumulate enough matter to power a new burst (Woosley et al. 2004). This is at odds with the detection of X-ray bursts that have short recurrence times and thus suggests that some of the initial fuel is preserved after ignition of the first burst (Galloway et al. 2008; Keek et al. 2010). One possible explanation for the occurrence of burst doublets could be that the matter is confined to certain parts of the neutron star surface, e.g. the magnetic poles, and that the bursts of a pair ignite at different locations. Alternatively, the bursts might ignite in separate layers that lie on top of each other. In this scenario the first burst causes unburned fuel to be mixed down to larger depth, which can cause the ignition of a new burst (e.g. Keek et al. 2009).

A systematic study of short recurrence time bursts shows that, on average, the second burst of a pair is shorter, less bright, cooler, and less energetic than the first (Boirin et al. 2007; Keek et al. 2010). This is suggestive of a reduced hydrogen content after the first burst has ignited and would favour the explanation that the bursts are resulting from different envelope layers rather than different areas of the neutron star (Boirin et al. 2007; Keek et al. 2010). In contrast, the two type-I X-ray bursts observed from SAX J1747.0–2853 were of similar duration (~40 s) and similar intensity, which points towards a similar fuel content.

4.3. Low-level accretion activity

Two of the X-ray transients covered by our monitoring observations were detected at low X-ray luminosities of L_X ~ 10³³⁻³⁴ erg s^-1. This is a factor of ~10–100 above their quiescent levels, yet well below their maximum X-ray intensities of L_X ≳ 10³⁶ erg s^-1. Below we briefly discuss our findings, which indicate that these sources likely exhibit low-level accretion activity outside their regular (i.e., bright) accretion outbursts. We discuss this in more detail in Sect. 5.3.

4.3.1. XMM J174457–2850.3

XMM J174457–2850.3 is a faint, unclassified transient source that displays a quiescent luminosity of L_X ~ 10³² erg s^-1 and exhibits X-ray outbursts with a peak luminosity of L_X ~ 10³⁶ erg s^-1 (Appendix A.6). We detected one such outburst during our campaign (in 2005). Several of our observations, however, detect the source at an intensity of L_X ~ (1−9) × 10³³ erg s^-1 (see Table 3 and Fig. A.6). This is a factor 10–100 higher than its quiescent emission level, yet considerably weaker than the full outbursts it displays.

As detailed in Appendix A.6, our Chandra/XMM-Newton monitoring observations support previous findings that the bright episodes of this source last only shortly and that the source spends most of its time at luminosities that are a factor ~10–100 above its quiescent level. This peculiar behaviour has led to the speculation that this unclassified X-ray source is possibly a wind-accreting X-ray binary (Degenaar & Wijnands 2010), although some Roche-lobe overflowing LMXBs also appear to display low-level accretion activity (Sect. 5.3). One such example is GRO J1744–28 (see below).

4.3.2. GRO J1744–28

The neutron star LMXB GRO J1744–28, also known as “the bursting pulsar” (Appendix A.5), undergoes outbursts that reach up to L_X ~ 10³⁷⁻³⁸ erg s^-1 (Woods et al. 1999) and has a quiescent luminosity of L_X ~ (1−3) × 10³³ erg s^-1 (Wijnands & Wang 2002; Daigne et al. 2002). GRO J1744–28 is detected during all five XMM-Newton observations of GC-4. During the first four data sets, the source intensity was found to be L_X ~ (1.5−6.5) × 10³³ erg s^-1. Within the errors this is consistent with the source being in quiescence. During the final observation (2008 September), however, the source intensity was enhanced to L_X ~ 1.9 × 10³⁴ erg s^-1 (Appendix A.5). Indications of enhanced activity above the quiescent level have been reported from this source on other occasions as well (Muno et al. 2007).

4.4. Two weak persistent objects

4.4.1. XMMU J174654.1–291542: a new X-ray source

In all XMM-Newton observations covering the field GC-6, we detect an X-ray point source at α = 17^h46^m54.15^s, δ = −29°15′42.6″ (J2000) with an uncertainty of 4″. This object is only detected above 2 keV and has no counterpart in the SIMBAD astronomical database or in DSS/2MASS images⁸. We designate this new X-ray source XMMU J174654.1–291542. Although the source is not detected in the individual HRC-I exposures, it is clearly visible by eye when all data are merged. This allows us to refine its position to α = 17^h46^m54.47^s, δ = −29°15′44.0″ (J2000) with an uncertainty of 1.5″.

Fitting all XMM-Newton spectra simultaneously to an absorbed powerlaw yields N_H = (5.2 ± 1.2) × 10²² cm^-2 and photon indices varying between Γ = 0.3−1.8 (see Table 3). This suggests that there is considerable spectral variation between the different data sets. This is illustrated by Fig. 4 (left), which compares the PN spectra obtained in 2006 September and 2008 May. If the different spectra are fitted individually, we obtain a considerable spread in hydrogen column densities of N_H = (1.6−9.0) × 10²² cm^-2 and powerlaw indices of Γ = 0.1−1.8, although the errors on both parameters are large.

We also fitted the spectra simultaneously with the powerlaw index tied between the different observations, whereas the hydrogen column density was left to vary. This resulted in Γ = 1.0 ± 0.3 and N_H = (2.5−10.2) × 10²² cm^-2, with typical errors of ~2.0 × 10²² cm^-2 (for χ_ν = 1.04 for 159 d.o.f.). Regardless of the chosen approach, we obtain 2–10 keV luminosities lying in a range of L_X ~ (0.7−3.0) × 10³⁴ erg s^-1 (assuming D = 8 kpc).

Additional evidence in support of spectral variations is provided by Fig. 4 (right), where we plot the ratio of PN source counts in the 4–10 keV and 0.5–4 keV energy bands versus the count rate over the full 0.5–10 keV range, using all five XMM-Newton observations. The biggest outlier in this graph concerns the 2006 September data set, during which the count rate in the full energy band was a factor ~2 higher than for the other observations. The count rates in the soft 0.5–4 keV band are similar for all five observations, so this difference can be completely attributed to the intensity in the harder 4–10 keV band.

XMMU J174654.1–291542 is not detected in the individual HRC-I observations, which constrains the source luminosity during those epochs to L_X ≲ (4−7) × 10³³ erg s^-1. In the merged HRC image, the source is weakly detected at a count rate of (3 ± 1) × 10^-3 counts s^-1. For N_H = 5.2 × 10²² cm^-2 and Γ = 1.0, this implies a luminosity of L_X = 6.3 × 10³³ erg s^-1. Given the small difference between the XMM-Newton detections and the Chandra upper limits, we cannot asses whether this object is truly transient or rather a persistent source that displays intensity variations by a factor of a few (see left panel of Fig. 5).

We searched through archival data to shed more light on the long-term variability of this newly identified X-ray source. We found two Chandra/ACIS pointings that cover the source region, carried out on 2006 November 2 and 2008 May 10 (obs IDs 7163 and 9559 with exposure times of ~14.3 and ~14.8 ks, respectively). During both observations XMMU J174654.1–291542 is one of the brightest objects in the field, yielding ~350–400 net photons per observation. We extracted source and background spectra for these two data sets.

There are no prominent spectral differences between the two archival Chandra observations. We obtain similar photon indices of Γ = 0.5 ± 0.5 and hydrogen column densities of N_H = (2.5 ± 1.8) × 10²² and (2.8 ± 1.7) × 10²² cm^-2 for the 2006 and 2008 data, respectively. The inferred luminosity is L_X ~ 1 × 10³⁴ erg s^-1 for both observations. The archival Chandra observations thus detect XMMU J174654.1–291542 at similar intensity levels as seen during our monitoring programme. This likely points towards a persistent nature. However, all detections occurred within a time frame of ~2.5 yr and we cannot exclude that the source is a weak transient that underwent a long outburst.

The strong spectral variation that we observe for XMMU J174654.1–291542 is reminiscent of the behaviour seen for systems that harbour a compact primary accreting from the wind of a companion star. Wind accretion occurs e.g. in neutron star or black hole HMXBs, symbiotic X-ray binaries (a subclass of LMXBs that contain a neutron star and an M-type giant) or symbiotics in which the accreting object is a white dwarf (e.g. Luna & Sokoloski 2007; Masetti et al. 2007; Heinke et al. 2009b; Romano et al. 2011). In such systems the changing spectral properties are attributed to intrinsic variations in the absorption column density due to the wind of the companion star.

Our spectral analysis of XMMU J174654.1–291542 indicates that there might be a large spread in hydrogen column densities between the different observations. The obtained spectral index of Γ = 1.0 ± 0.3 is also quite typical of wind-accreting binaries, whereas Roche-lobe overflowing LMXBs usually have softer X-ray spectra. Its X-ray spectral properties thus suggest that XMMU J174654.1–291542 may harbour a compact object accreting from the wind of its companion star.

We found a possible infrared counterpart in the Spitzer/IRAC catalogue that is consistent with our position of XMMU J174654.1–291542 (Ramírez et al. 2008). This object is relatively bright with an apparent magnitude of 10.7 mag at 3.6 micron. A quick comparison with WISE data (Wright 2008) of the recurrent nova T Coronae Borealis (T CrB), which has an absolute 3.6-micron magnitude of approximately −4 mag, suggests that the Spitzer source might be a white dwarf binary located at D ~ 6 kpc. If this object is associated with XMMU J174654.1–291542, our X-ray source might thus indeed be a symbiotic. At a distance of 6 kpc, the X-ray fluxes inferred from our campaign translate into luminosities of L_X ~ (4−10) × 10³³ erg s^-1.

4.4.2. The unclassified X-ray source AX J1742.6–2901

During the final series of Chandra/HRC observations, performed in 2008 July, we detected an X-ray point source located at α = 17^h42^m42.58^s, δ = −29°02′04.8″ (J2000). The source is weak (a total number of ~35 net counts) and detected at a large offset angle (~18′) from the aimpoint of the observation, which strongly decreases the accuracy of the source localisation. Using Eq. (5) of Hong et al. (2005), we tentatively estimate a positional uncertainty of ~9″.

Within our estimated error, the Chandra coordinates are consistent with the Swift localisation of AX J1742.6–2901. This unclassified X-ray source (not to be confused with the neutron star LMXB AX J1745.6–2901 discussed in Sect. 4.1) was discovered during ASCA observations of the GC in 1998 September (Sakano et al. 2002). The ASCA source was tentatively associated with the ROSAT object 2RXP J174241.8–290215, which was detected at a count rate of (1.9 ± 0.3) × 10^-2 counts s^-1 during a 2-ks PSPC observation in 1992 March⁹.

AX J1742.6–2901 was detected only once during our campaign, with a HRC-I count rate of (6.8 ± 2.5) × 10^-2 counts s^-1. The source is located just outside the FOV of our XMM-Newton pointings (see Fig. 1) and was not covered by the Chandra survey of Muno et al. (2009). However, the source region was observed with Swift/XRT as part of a Swift follow-up programme of unclassified ASCA sources (Degenaar et al. 2012). During the XRT observations, performed on 2008 March 7 (i.e., four months prior to our Chandra/HRC detection), a single X-ray source was detected within the FOV. The coordinates inferred from the Swift data coincide with our Chandra/HRC position. Analysis of the XRT spectral data yielded N_H = (1.6 ± 0.8) × 10²² cm^-2 and Γ = 2.9 ± 1.1, resulting in a luminosity of L_X ~ 2.8 × 10³³ erg s^-1 (Degenaar et al. 2012).

We used the reported Swift/XRT spectral parameters to convert our observed Chandra/HRC count rate into 2–10 keV fluxes (see Table 3). Assuming a distance of 8 kpc, the unabsorbed flux translates into a luminosity of L_X ~ 2.1 × 10³³ erg s^-1. For these spectral parameters the ROSAT/PSPC and ASCA/GIS count rates correspond to L_X ~ 9.2 × 10³³ and 5.1 × 10³³ erg s^-1, respectively. Given that AX J1742.6–2901 has been detected at comparable intensity levels with ROSAT (1992), ASCA (1998), Swift and Chandra (both 2008), and that the source is close to the detection limit of our HRC observations (see Fig. 5), we consider it likely that this is not a transient object but rather a weak persistent X-ray source that displays a factor of a few variability and peaks near L_X ~ 10³⁴ erg s^-1. We cannot classify the source further.

4.5. Possible transient reported by Wijnands et al. (2006a)

Wijnands et al. (2006a) reported on the possible discovery of a new sub-luminous X-ray transient, which was detected during our HRC-I observation of 2005 June 5 at a count rate of ~8 × 10^-3 counts s^-1. This object is also detected in the XMM-Newton observations carried out in 2008 September, displaying a count rate of ~(1.9 ± 0.1) × 10^-2 counts s^-1, whereas the other XMM-Newton observations yield 2σ intensity upper limits of ≲7 × 10^-3 counts s^-1. This suggests a possible transient nature. However, the XMM-Newton observations reveal that the source is detected only below 2 keV. It is therefore not a candidate transient X-ray binary, but more likely a flaring star.

5. Discussion

We have presented the results of a four-year monitoring campaign of the GC carried out with the Chandra and XMM-Newton observatories. We have covered a field of 1.2 square degrees around Sgr A^∗, which was targeted on ten different epochs between 2005 June and 2008 September. Our study focused on the behaviour of transient X-ray sources that reach 2–10 keV peak luminosities ≳1 × 10³⁴ erg s^-1 when in outburst. We detected activity of eight previously known X-ray transients during our campaign. On average, six of these were seen active each year (see Table 3). We studied their X-ray spectra and long-term lightcurves. All are highly absorbed (N_H ≳ 1 × 10²² cm^-2), indicating source distances near or beyond the GC.

We detected type-I X-ray bursts from the neutron star LMXBs AX J1745.6–2901 and SAX J1747.0–2853 (see Sects. 4.1 and 4.2). For the former we also observed a ~1600-s long eclipse in one of the X-ray lightcurves, which had similar properties as the eclipses that had previously been seen by ASCA (Maeda et al. 1996). For SAX J1747.0–2853 we found a pair of a type-I X-ray bursts that have a waiting (recurrence) time of only 3.8 min. Their similar peak intensity and duration suggests that the fuel that powered both bursts was likely of similar composition. This is an important pointer to understand the mechanism that is responsible for the unusually short recurrence time. We determined the quiescent luminosity of SAX J1747.0–2853 of L_X ~ 2 × 10³¹ (D/6.7 kpc)² erg s^-1, which firmly establishes its classification as X-ray transient (cf. Werner et al. 2004).

We uncovered episodes of low-level accretion in a luminosity range of L_X ~ 10³³⁻³⁴ erg s^-1 from two X-ray transients that are known to reach considerably higher intensities of L_X ≳ 10³⁶ erg s^-1 in full outburst (GRO J1744–28 and XMM J174457–2850.3; see Sect. 5.3). Two other transient sources that were detected during our campaign (CXOGC J174535.5–290124 and CXOGC J174538.0–290022) exhibit outbursts with similar intensities of L_X ~ 10³⁴ erg s^-1, but have never been observed in a brighter state (see Appendices A.7 and A.8).

In addition to the eight known transient systems, we detected two weak unclassified X-ray sources that, despite being undetected at some epochs during our campaign, are likely persistent and may be variable by a factor of a few (see Sect. 4.4). The first is a previously unknown X-ray source that we designate XMMU J174654.1–291542. Inspection of archival data suggests that the source may be persistent at a luminosity of L_X ~ 10³⁴ erg s^-1 (assuming D = 8 kpc). Despite its relatively steady X-ray intensity, we observed significant changes in the source spectrum on a time scale of months. Based on its X-ray spectral properties and the possible association with a Spitzer/IRAC infrared object, we tentatively classify this new X-ray source as a cataclysmic variable. The unclassified X-ray source AX J1742.6–2901 was detected with ROSAT, ASCA, Swift, and Chandra at similar intensity levels of L_X ~ 10³⁴ erg s^-1 between 1992 and 2008, which suggests that the source is likely persistent.

5.1. The population of Galactic centre X-ray transients

In Table 4 we list all known X-ray transients that are located in the region covered by our campaign. This table is an update from Wijnands et al. (2006a), in which we have included three new transients that were discovered by Swift in 2006 (Swift J174553.7–290347 and Swift J174622.1–290634; Degenaar & Wijnands 2009) and in 2011 (Swift J174535.5–285921; Degenaar et al. 2011b). For each source we list the angular distance from Sgr A^∗, the minimum and maximum X-ray luminositiy, and any other relevant information that characterises the source. We indicate which sources were active during our campaign and whether the transients displayed multiple outbursts in the past decade.

There are five confirmed and two candidate LMXBs amongst the 17 transients covered by our campaign (Table 4). The confirmed LMXBs contain a neutron star primary, as demonstrated by the detection of X-ray pulsations (GRO J1744–28) or type-I X-ray bursts (GRS 1741–2853, AX J1745.6–2901, KS 1741–293 and SAX J1747.0–2853). The candidate LMXB CXOGC J174540.0–290031 displays eclipses in its X-ray lightcurve that indicate a high inclination and an orbital period of 7.9 h. This source has been suggested to harbour a black hole based on the detection of strong radio emission (Muno et al. 2005a; Porquet et al. 2005a). Similarly, the transient X-ray source 1A 1742–289 that erupted in 1975 has been tentatively classified as a black hole LMXB based on its X-ray properties and the apparent association with a strong radio source (Davies et al. 1976; Branduardi et al. 1976).

The remaining ten X-ray transients are unclassified, although their large outburst amplitudes (a factor >100) and spectral properties render it likely that these are accreting neutron stars or black holes. A lack of infrared counterparts with K < 15 mag, in turn suggests that these transients are either LMXBs or HMXBs with companions fainter than a B2 V star (Muno et al. 2005b; Mauerhan et al. 2009). Tentative X-ray pulsations were reported for two of the unclassified transients: XMM J174457–2850.3 (~5 s; Sakano et al. 2005) and XMMU J174554.4–285456 (~172 s; Porquet et al. 2005b). These spin periods fall within the range of HMXBs. However, both results require confirmation (Sakano et al. 2005; Porquet et al. 2005b).

It is remarkable that the ten unclassified transients all have maximum 2–10 keV luminosities of L_X ≲ 10³⁶ erg s^-1 and are thus fainter than the seven confirmed LMXBs (disregarding the LMXB CXOGC J174540.0–290031, which has been argued to be intrinsically brighter and possibly strongly obscured; Muno et al. 2005a). Apart from the low outburst intensity, the X-ray spectra, outburst profiles and recurrence times are not notably different from the brighter transients (see Table 4, see also Muno et al. 2004; Degenaar & Wijnands 2009). If they are X-ray binaries, their sub-luminous character requires an explanation.

Based on statistical arguments, Wijnands et al. (2006a) argued that it is unlikely that all these objects are viewed at high inclination (thereby reducing their observed X-ray emission) and that most of them must be intrinsically faint. It has been proposed that these transients might consist of compact objects accreting from the wind of a main sequence companion (Pfahl et al. 2002). This would be a relatively inefficient process that may account for the observed low X-ray luminosities. An alternative explanation is that these are LMXBs with tight orbits and unusual donor stars (e.g. a white dwarf, brown dwarf or planet) that can only accommodate a small accretion disk (King & Wijnands 2006; in ’t Zand et al. 2009). Comparing the low-luminosity transients with weak outbursts observed from the brighter transients suggests that there might be examples of both these types of objects amongst the Galactic centre transients (see Sect. 5.3).

Out of the 17 transients listed in Table 4, we have detected eight in an active state during our campaign. Moreover, extensive monitoring of the GC region during the past decade has shown that 11 of these transients (i.e., 65%) recurred between 1999 and 2012 (see the column labelled “Rec?”), six of which even experienced three or more distinct outbursts during this epoch (the neutron star LMXBs GRS 1741–2853, AX J1745.6–2901, KS 1741–293 and SAX J1747.0–2853, and the unclassified transients CXOGC J174535.5–290124 and XMM J174457–2850.3).

No new X-ray transients with a 2–10 keV peak luminosity of L_X ≳ 1 × 10³⁴ erg s^-1 were found during our campaign. This confirms suggestions from previous authors that the majority of sources that recur on time scales less than a decade, and undergo outbursts of at least a few days, have now been identified in this region (in ’t Zand et al. 2004; Muno et al. 2009; Degenaar & Wijnands 2010). Although a new transient was discovered recently in July 2011 (Degenaar et al. 2011b,c; Chakrabarty et al. 2011; Servillat et al. 2011), previous discoveries date back to 2006 in spite of extensive monitoring with various X-ray satellites (Degenaar & Wijnands 2009).

5.2. Quiescent luminosities

In addition to the maximum 2–10 keV intensity reached during outburst, Table 4 also includes information on the quiescent luminosity of the 17 transients covered by our campaign. The listed values correspond to the minimum X-ray luminosity observed for each source and were taken from the literature. Only for SAX J1747.0–2853, KS 1741–293 and Swift J174622.1–290634 we did not find reported quiescent properties. We therefore searched the Chandra and XMM-Newton data archives for observations that covered the quiescent states of these three transients. We presented the results for SAX J1747.0–2853 in Sect. 4.2, whereas the other two are discussed below.

We note that the new transient that was discovered in 2011 (Swift J174535.5–285921) is likely associated with CXOGC J174535.6–285928 (Chakrabarty et al. 2011). This faint X-ray source was discovered in the Chandra survey of the GC, during which it was detected at a luminosity of L_X ~ 3 × 10³¹ erg s^-1 and did not display any variability on long or short time scales (Muno et al. 2009). CXOGC J174535.6–285928 likely represents the quiescent counterpart of the new 2011 transient and therefore we quote its luminosity as the quiescent level of Swift J174535.5–285921.

KS 1741–293: To determine the quiescent luminosity of KS 1741–293, we used archival Chandra observations performed in 2001, 2006 and 2007 (obs IDs 2267, 2272, 7038, and 8459) that amount to a total exposure time of ~56 ks. The composite spectrum was obtained using the tool COMBINESPECTRA and yields a 2–10 keV quiescent luminosity of L_X ~ 5 × 10³² erg s^-1 for an assumed distance of 8 kpc. The quiescent spectral properties of KS 1741–293 will be discussed in more detail in a separate work.

Swift J174622.1–290634: We found a total of seven archival Chandra/ACIS-I observations that covered the position of Swift J174622.1–290634. This source exhibited an outburst in 2006 that was first detected with Swift in mid-May. In late June, the source went undetected by Swift/XRT, indicating an upper limit on the 2–10 keV quiescent luminosity of a few times 10³³ erg s^-1 (Degenaar & Wijnands 2009). We found that the source is weakly detected in Chandra observations carried out on 2006 July 4 (obs ID 6642, ~5 ks) and 2006 August 24 (obs ID 7037, ~40 ks).

By fitting the spectra of these two Chandra observations simultaneously with the Swift outburst data, we found a joint value of N_H = (10.5 ± 3.7) × 10²² cm^-2 and photon indices of Γ = 3.0 ± 1.0 (Swift), 4.0 ± 2.8 and 6.6 ± 2.6 (Chandra; 2006 July and August, respectively). It appears that the spectrum was softening, although the source continued to be detected above 2 keV (at lower energies the source photons are strongly absorbed). The corresponding 2–10 keV luminosities are L_X ~ 1 × 10³⁴, ~4 × 10³³ and ~1 × 10³³ erg s^-1, respectively.

We can obtain additional constraints on the quiescent luminosity of Swift J174622.1–290634 by using archival Chandra data obtained in 2001 and 2004 (obs IDs 2282, 2291, 4683, and 4684; total exposure time of ~120 ks). This allows us to estimate a pre-outburst upper limit on the 2–10 keV luminosity of L_X ~ (0.7−1) × 10³² erg s^-1 (for an absorbed powerlaw model with N_H = 10.5 × 10²² cm^-2 and Γ = 3−6, assuming D = 8 kpc).

There are caveats to consider when interpreting the quiescent luminosities listed in Table 4. Although it is generally assumed that accretion is strongly suppressed during quiescence, it may not have fully come to a halt. Indeed, several transient neutron star and black hole X-ray binaries have been found to display considerable variation in their quiescent X-ray luminosity, which is attributed to residual accretion (for recent examples, see Kong et al. 2002; Hynes et al. 2004; Cackett et al. 2010, 2011; Fridriksson et al. 2011; Degenaar & Wijnands 2012). This implies that some of the minimum X-ray luminosities listed in Table 4 might correspond to low-level accretion. Literature values often only reflect a snapshot of the quiescent state and might therefore not necessarily represent the absolute minimum X-ray luminosity of these transient sources.

It is worth noting that several transients have quiescent luminosities in the range of L_X ~ 10³²⁻³³ erg s^-1 (see Table 4). Examples include the neutron star LMXBs GRS 1741–2853, AX J1745.6–2901, KS 1741–293 and GRO J1744–28. Transient neutron star LMXBs often have quiescent X-ray spectra that are composed of a soft, thermal component that dominates the spectrum below ~2 keV, accompanied by a hard emission tail that can be described by a simple powerlaw. Since any soft thermal emission is strongly absorbed by the high hydrogen column densities inferred for our sources, this implies that they must have considerable hard powerlaw emission in quiescence (cf. Jonker et al. 2003). We note that detections and upper limits of about L_X ~ 10³¹ erg s^-1 (such as obtained for SAX J1747.0–2853) do not provide strong constraints on the temperature of the neutron star, since the soft thermal emission is so heavily absorbed.

5.3. Low-level accretion activity

Although several transients covered by our campaign are known to exhibit outbursts with 2–10 keV peak luminosities of L_X > 10³⁶ erg s^-1 (see Table 4), these sources also display activity well below that level. We already discussed in Sect. 4.3 that the unclassified transient XMM J174457–2850.3 and the pulsating neutron star LMXB GRO J1744–28 were both detected during our campaign at a luminosity of L_X ~ 10³³⁻³⁴ erg s^-1, which is intermediate between their quiescent state and their full X-ray outbursts. For the former, Swift monitoring observations of the GC have shown that this low-luminosity state can persist for several weeks (Degenaar & Wijnands 2010). Indeed, this X-ray transient is more often detected at a low X-ray luminosity than in full outburst.

The two neutron star LMXBs GRS 1741–2853 and KS 1741–293 have both shown short, weak outbursts (t_ob ≲ 1 week, L_X ≲ 10³⁴⁻³⁵ erg s^-1) that preceded longer and brighter accretion episodes with a duration of several weeks/months and 2–10 keV luminosities peaking at L_X ~ 10³⁶⁻³⁷ erg s^-1 (see Appendices A.1 and A.4). Albeit on a different scale, one can argue that the neutron star LMXB AX J1745.6–2901 displayed something similar; a major outburst with a 2–10 keV peak luminosity of L_X ~ 6 × 10³⁶ erg s^-1 was detected from this system in 2007–2008, whereas a few months earlier (in 2006) it underwent an accretion episode that had a duration and intensity that were both a factor ~5 lower (see also Degenaar & Wijnands 2010). Finally, the neutron star LMXB SAX J1747.0–2853 was found to display activity at L_X ~ 10³⁵ erg s^-1 on several occasions, which is also well below its maximum outburst intensity (Wijnands et al. 2002; Campana et al. 2009). Another firm indication of low-level accretion activity is the detection of type-I X-ray bursts from SAX J1747.0–2853 and KS 1741–293 at accretion luminosities of L_X ~ 10³⁵ erg s^-1 (Chenevez et al. 2009, Linares et al. 2011; see also Kuulkers et al. 2009, for similar examples).

Taken together, it appears that all five neutron star LMXBs covered by our campaign exhibited multiple luminous outbursts with L_X ≳ 10³⁶ erg s^-1 in the past decade, but also frequently display low-level accretion activity at L_X ~ 10³³⁻³⁵ erg s^-1. For GRS 1741–2853, KS 1741–293 and AX J1745.6–2901 it was found that the weaker outbursts are shorter. The disk instability model, which is thought to provide the framework to explain the transient behaviour of LMXBs, provides a possible explanation (e.g. King & Ritter 1998; Lasota 2001). Sub-luminous outbursts may occur when only a small fraction of the accretion disk becomes ionized and accreted onto the compact object, whereas a larger part of the disk is consumed during the more luminous outbursts (see also Degenaar & Wijnands 2010).

Our campaign covered two unclassified X-ray transients that appear active at levels of L_X  ~  10³³⁻³⁴ erg s^-1 without becoming brighter. Despite being frequently active, both CXOGC J174535.5–290124 (Appendix A.7) and CXOGC J174541.0–290014 (Appendix A.8) have never been observed at a luminosity L_X ≳ 5 × 10³⁴ erg s^-1 (Muno et al. 2005b). Strikingly, as many as ten out of 17 transients covered by our campaign have 2–10 keV peak luminosities that are well below L_X ~ 10³⁶ erg s^-1 and have never been observed in a brighter state (see Table 4).

These very-faint X-ray transients are possibly connected to the “burst-only” sources: a group of neutron star LMXBs that became detectable only when exhibiting a type-I X-ray burst, while no persistent emission could be detected down to a limit of L_X ~ 10³⁶ erg s^-1 (Cornelisse et al. 2002; Wijnands et al. 2006a; Campana 2009). In analogy with the weak outbursts seen for the brighter neutron star LMXBs, we may interpret the low X-ray luminosities in terms of a small amount of matter accreted from the disk. The fact that these sources have never been observed in a brighter state could in turn imply that the entire disk (and hence the binary orbit) is relatively small. It is of note that it takes a long time to accumulate enough fuel to power an X-ray burst when the mass-accretion rate is very low. These are therefore expected to occur much less frequently than for higher accretion rates (e.g., in ’t Zand et al. 2009; Degenaar et al. 2010a).

Long-lasting activity at L_X ~ 10³³ erg s^-1, such as seen for the unclassified transient XMM J174457–2850.3, is difficult to accomodate within the disk instability model. This behaviour might find a more natural explanation in terms of wind-accretion (see the discussion in Degenaar & Wijnands 2010). Similarly, some of the other unclassified transients (e.g. CXOGC J174535.5–290124 and CXOGC J174538.0–290022) appear to spend long times at relatively low intensity levels (Muno et al. 2005b; Degenaar & Wijnands 2010). This might be an indication that in these systems the companion star does not fill its Roche lobe, but supplies matter to the compact object via a wind. On the other hand, there are a few examples of LMXBs in which the donor star does overflow its Roche lobe and that appear to persist at low X-ray luminosities of L_X ~ 10³⁴⁻³⁵ erg s^-1 (e.g. Del Santo et al. 2007; in ’t Zand et al. 2009; Campana 2009; Degenaar et al. 2010a). At least one of these harbours a hydrogen-rich companion and must thus have a relatively large orbit (Degenaar et al. 2010a). The sub-luminous character therefore remains a puzzle.

5.4. Ultra-faint X-ray transients

While our present work focused on transient X-ray sources with luminosities L_X ≳ 1 × 10³⁴ erg s^-1, we detected a number of objects that appear variable by a factor of ≳ 5–10 but remain below L_X ~ 5 × 10³³ erg s^-1 (assuming D = 8 kpc). We found several such sources in our XMM-Newton observations. They are hard (most photons emitted above 2 keV) and have no DSS/2MASS counterparts, which effectively rules out that these are foreground stars.

Two such examples are CXOGC J174451.7–285308 and CXOGC J174423.4–291741 from the Chandra catalogue of Muno et al. (2009). Both objects were detected once during our campaign at 2–10 keV luminosities of L_X ~ 3 × 10³³ erg s^-1. CXOGC J174451.7–285308 is indicated by Muno et al. (2009) as exhibiting long-term variability by a factor of ~70. CXOGC J174423.4–291741, on the other hand, is listed in this catalogue as a weak persistent source displaying a luminosity of L_X ~ 10³² erg s^-1 (using the conversion factor from photon to energy flux quoted by these authors).

As discussed in Sect. 5.3, several confirmed and candidate X-ray binaries display activity at similar intensity levels. Furthermore, Heinke et al. (2009a) found a transient object with a peak luminosity of L_X ~ 6 × 10³³ erg s^-1 in the globular cluster M15, which is likely a neutron star LMXB. Although a significant fraction of these “ultra-faint X-ray transients” found in our XMM-Newton data might be accreting white dwarfs (Verbunt et al. 1997), this suggests that there could also be X-ray binaries amongst them.

Online material

Appendix A: Results on individual sources

In this appendix we describe the properties of the eight X-ray transients that were detected during our campaign in more detail. For each source we describe the X-ray spectra and lightcurves of individual observations, as well as the flux evolution during our campaign. In particular, we attempt to constrain the luminosity, duration and recurrence time of the observed outbursts, sometimes using literature reports and archival X-ray observations. We also briefly review the historic outburst behaviour of each source, as well as any activity reported elsewhere during and after our campaign.

As mentioned in Sect. 1, the inner ~25′ × 25′ around Sgr A^∗ (corresponding to field GC-2 in our campaign) has been covered with Swift’s XRT starting in 2006 (Kennea & The Swift/XRT team 2006; Degenaar & Wijnands 2009, 2010). Those quasi-daily observations provide a partial overlap with our campaign. For sources located in this region we will therefore compare reports on the Swift data with results from our Chandra/XMM-Newton observations. Sources located at larger angular distances (≳ 20′) are covered by the INTEGRAL Galactic bulge monitoring programme, which provides observations every few days during two ~4-month windows per year (Kuulkers et al. 2007b).

We summarised the results of our spectral analysis in Table 3. Below, the individual sources are discussed in the same order as they appear in Table 3, so that the numbering of the following sections corresponds to the numbered labels in that table. To directly compare the different sources all spectra are plotted in the same energy range (2–10 keV) and all long-term lightcurves on the same intensity scale (L_X = 10³¹⁻³⁷ erg s^-1). In all plots, triangles represent XMM-Newton observations, while bullets and squares are used for Chandra/ACIS and HRC data, respectively. The last two objects listed in Table 3 are likely weak persistent X-ray sources and were already discussed in detail in Sect. 4.4.

A.1. GRS 1741–2853

Brief historic overview: this transient neutron star LMXB was discovered by the Granat observatory in 1990 (Sunyaev 1990) and has been detected in an active state many times since then (for a detailed overview, see Trap et al. 2009). The source is known to display type-I X-ray bursts (e.g. Cocchi et al. 1999), from which a distance of 7.2 kpc can be inferred (Trap et al. 2009). It is frequently active and typically exhibits outbursts that reach a 2–10 keV peak luminosity of L_X ~ 10³⁶⁻³⁷ (D/7.2 kpc)² erg s^-1, and last for a few weeks (e.g. Degenaar & Wijnands 2010). In quiescence, the source is detected at L_X ~ 1 × 10³² (D/7.2 kpc)² erg s^-1 (Muno et al. 2003a).

Activity during our campaign: we detected activity from GRS 1741–2853 in two different epochs (in 2005 and 2007). The source was first detected in outburst during the Chandra/HRC observations performed on 2005 June 5 and was also seen in the follow-up ACIS pointing of 2005 July 1 (Wijnands et al. 2006a). Over the one month time span separating these two observations, the source intensity decreased by a factor ~4 (see Table 3) from L_X ~ 6 × 10³⁵ to 1.5 × 10³⁵ (D/7.2 kpc)² erg s^-1. This may indicate that the activity was ceasing. The rise of this outburst was caught by INTEGRAL in 2005 in early April (Kuulkers et al. 2007b). If the decrease in flux signalled by the Chandra data was due to a transition towards quiescence, the duration of the 2005 outburst was therefore ~13 weeks.

We found the source again in outburst during our Chandra observations of 2007 March 12, when it was bright enough to cause pile-up of the ACIS instrument. We therefore extracted the source spectrum using a 10″−40″ annulus, avoiding the inner piled-up part of the PSF (see Sect. 3.2). The inferred luminosity was L_X ~ 2 × 10³⁵ (D/7.2 kpc)² erg s^-1. During the subsequent observation performed on 2007 April 6, GRS 1741–2853 had nearly faded by one order of magnitude (see Table 3), whereas the source was not detected on April 18 with an upper limit of L_X ≲ 3 × 10³² erg s^-1. This indicates that the source had returned to the quiescent state at that time (see Fig. A.1). The 2007 outburst of GRS 1741–2853 was covered by different satellites (Kuulkers et al. 2007a; Wijnands et al. 2007; Muno et al. 2007; Porquet et al. 2007). Swift/XRT observations detected the source with a peak luminosity of L_X ~ 1.5 × 10³⁶ (D/7.2 kpc)² erg s^-1 and constrain the outburst duration to be ≳13 weeks (Degenaar & Wijnands 2010).

X-ray spectra: a joint fit to the ACIS spectral data obtained during the two outbursts of 2005 and 2007 (Fig. A.1) yields N_H = (11.4 ± 1.1) × 10²² cm^-2 and Γ ~ 1.5−2.0 (see Table 3). These values are comparable to those found for other outbursts of GRS 1741–2853 (Muno et al. 2003a; Trap et al. 2009; Degenaar & Wijnands 2009, 2010). The count rate detected during the 2005 June Chandra/HRC pointing and upper limits inferred from observations in which the source was not detected, were converted to 2–10 keV unabsorbed fluxes using Γ = 2.0 and N_H = 11.4 × 10²² cm^-2.

Constraints on the weak 2006 outburst: apart from outbursts reaching L_X ≳ 10³⁶ (D/7.2 kpc)² erg s^-1, GRS 1741–2853 also undergoes low-level accretion activity. Swift/XRT monitoring observations of the GC exposed a weak, short outburst from GRS 1741–2853 between 2006 September 14–20, during which the source did not become brighter than L_X ~ 7 × 10³⁴ (D/7.2 kpc)² erg s^-1 (2–10 keV; Degenaar & Wijnands 2009). This is about three orders of magnitude lower than the maximum outburst luminosity exhibited by this source, yet still a factor ≳100 above its quiescent level.

The source region is covered by one of our XMM-Newton observations on 2006 September 8, which is just one week before the sub-luminous outburst detected by Swift/XRT. During these observations, GRS 1741–2853 was not detected and we can infer a 2σ upper limit on the PN count rate of ≲3 × 10^-3 counts s^-1. Using pimms with N_H = 11.4 × 10²² cm^-2 and Γ = 2.0, we can estimate that the 2–10 keV luminosity of GRS 1741–2853 was L_X ≲ 3 × 10³² (D/7.2 kpc)² erg s^-1 at that time. Therefore, one week prior to the peculiar short 2006 outburst there were no indications of enhanced activity above the quiescent level.

Activity after our campaign: Swift monitoring observations detected a new outburst from GRS 1741–2853 in 2009 September–November, which had a duration of ~4–5 weeks and reached up to L_X ~ 1 × 10³⁷ (D/7.2 kpc)² erg s^-1 (Degenaar & Wijnands 2009). The source was again active for ~12 weeks in 2010 July–October at an average luminosity of L_X ~ 4 × 10³⁵ (D/7.2 kpc)² erg s^-1 (Degenaar et al. 2010b). This further underlines the frequent activity of this neutron star LMXB, as is illustrated by the results from our Chandra/XMM-Newton monitoring campaign. It was noted by Degenaar & Wijnands (2010) that despite differences in duration and maximum intensity, the fluency of the bright outbursts of GRS 1741–2853 are very similar (with the exception of the unusual faint, short 2006 outburst).

A.2. AX J1745.6–2901

Brief historic overview: another transient that was frequently detected during our monitoring campaign is located at an angular distance of ~1.5′ from Sgr A^∗, at a position consistent with the Chandra localisation of AX J1745.6–2901 (Heinke et al. 2008). This neutron star LMXB was discovered in 1993 by the ASCA observatory, exhibits type-I X-ray bursts and its X-ray lightcurve displays eclipses that recur every ~8.4 h, corresponding to the orbital period of the binary (Maeda et al. 1996; Kennea & Skinner 1996). The source is detected in quiescence at a luminosity of L_X ~ 10³² erg s^-1 (Degenaar & Wijnands 2009). Following detections by ASCA in 1993 and 1994, AX J1745.6–2901 was never reported in outburst again until 2006, despite extensive monitoring of the source region.

Activity during our campaign and outburst constraints: we detected two distinct outbursts from AX J1745.6–2901 (in 2006 and 2007–2008). The source was first active during our XMM-Newton observations of 2006 February 27, when it displayed a luminosity of L_X ~ 4.6 × 10³⁵ erg s^-1 (assuming a distance of D = 8 kpc). In the subsequent observation performed on 2006 September 8, the source was not detected with an upper limit of L_X ≲ 3 × 10³³ erg s^-1 (assuming N_H = 21.8 × 10²² cm^-2 and Γ = 2.0). This is consistent with results obtained with Swift/XRT, which indicated that the source was active for three months between 2006 February and June, but resided in quiescence thereafter (Degenaar & Wijnands 2009).

It is unclear when the 2006 outburst of AX J1745.6–2901 started, since the position of the Sun with respect to the GC rendered this region unobservable between 2005 November and 2006 February. Since the source was not detected during our 2005 monitoring observations, the outburst must have started after 2005 October 20 (see Table 1). The time span between the Chandra observations and the first detection of AX J1745.6–2901 on 2006 February 24 (with Swift; Kennea et al. 2006) is 4 months. This constrains the outburst duration to 3–7 months.

Renewed activity of the source was reported in 2007 February, as seen by various instruments (Kuulkers et al. 2007a; Wijnands et al. 2007; Porquet et al. 2007; Degenaar & Wijnands 2009). AX J1745.6–2901 is detected at similar luminosities of L_X ~ (1−4) × 10³⁶ erg s^-1 in all our observations carried out between 2007 February and 2008 May (see Table 3). We picked up a likely type-I X-ray burst and an X-ray eclipse during these observations (see Sect. 4.1). The source intensity had decreased by nearly a factor 10 in 2008 July, and it went undetected in 2008 September. This indicates that the source had returned to the quiescent state (see Fig. A.2).

The Swift/XRT observations of the GC also suggest that AX J1745.6–2901 was continuously active since 2007 February, until it returned to quiescence in 2008 in early September (Degenaar & Wijnands 2010). The Swift observations suggest that this long outburst must have commenced between 2006 November 3 and 2007 March 6 (when the GC was unobservable due to Sun-angle constraints). This constrains the total duration of the outburst to 18–21 months (1.5–1.75 yr). The 2007–2008 outburst was a factor of ~5 longer and a factor of ~5 more luminous than the 2006 outburst (Table 3 and Fig. A.2, see also Degenaar & Wijnands 2010).

X-ray spectra: spectral analysis of the Chandra/ACIS and XMM-Newton data obtained during the 2007–2008 outburst of AX J1745.6–2901 is complicated by pile-up. We attempted to circumvent the expected effect on spectral shape and source flux by extracting source event from an annular region with a radius of 10″−40″. Figure A.2 displays the XMM-Newton/PN spectra obtained in 2006 February and 2008 March, which represent the two different outbursts of the source. We converted upper limits and HRC-I count rates into unabsorbed 2–10 keV fluxes using Γ = 2.0 and N_H = 21.8 × 10²² cm^-2.

Activity after our campaign: AX J1745.6–2901 was again reported to be active in 2010 June (Degenaar et al. 2010b,c). The outburst persisted until 2010 late-October, but the activity had ceased when the Swift monitoring observations resumed in 2011 early-February. The 2010 outburst thus had a duration of ~4–7 months. Both the length and the average 2–10 keV luminosity of this outburst are very similar to the 2006 activity.

A.3. SAX J1747.0–2853

Brief historic overview: the neutron star LMXB SAX J1747.0–2853 was initially discovered with BeppoSAX in 1998, although Granat may have detected activity from the source already in 1991 (Grebenev et al. 2002). Its neutron star nature has been established by detection of type-I X-ray bursts (in ’t Zand et al. 1998; Sidoli et al. 1999), which results in a distance estimate of D = 6.7 kpc (Galloway et al. 2008). Following its discovery, several accretion outbursts with 2–10 keV luminosities of L_X ~ 10³⁶⁻³⁷ (D/6.7 kpc)² erg s^-1 have been detected with different X-ray instruments (Wijnands et al. 2002; Natalucci et al. 2004; Werner et al. 2004; Markwardt & Swank 2004; Deluit et al. 2004).

Activity during our campaign and outburst constraints: we detected SAX J1747.0–2853 in outburst during three consecutive observations performed on 2005 October 20 (Chandra/HRC Wijnands et al. 2005b,c), 2006 February (Wijnands et al. 2006b) and 2006 September (the latter are both XMM-Newton observations). INTEGRAL detected hard X-ray activity from the source at similar epochs (2005 October and 2006 February; Kuulkers et al. 2005; Chenevez et al. 2006). Our 2006-February observation revealed a pair of type-I X-ray bursts with a recurrence time of only 3.8 min (see Sect. 4.2).

The source was not detected during subsequent observations (2007 July), from which we can estimate an upper limit of L_X ≲ 3.7 × 10³⁴ erg s^-1. This indicates that the source activity had ceased by that time (see also Fig. A.3)¹⁰. If the source was continuously active between our monitoring observations of 2005 October and 2006 September, the 2005–2006 outburst had a duration of ≳11 months. Non-detections in our data obtained in 2005 June and 2007 July constrain the duration of the active period to be ≲2.1 yr (see Fig. A.3).

Although we did not detect SAX J1747.0–2853 in 2007 July, it was seen active with INTEGRAL in 2007 October (Brandt et al. 2007). Swift/XRT follow-up observations carried out a few days later detected the source at L_X ~ 10³⁶ (D/6.7 kpc)² erg s^-1, confirming the renewed activity (Cackett & Miller 2007). The source region was not covered by our monitoring observations of 2007 September, but SAX J1747.0–2853 was not detected in XMM-Newton data obtained in 2008 March. This suggests that the new 2007-outburst had a duration of ≲7 months.

X-ray spectra: we fitted the XMM-Newton spectra obtained in 2006 February (shown in Fig. A.3) and September to constrain the spectral shape. The February observation contained two X-ray bursts (see Sect. 4.2), which were removed from the data for the purpose of fitting the persistent emission. A simultaneous fit to the two spectra yields N_H = (9.5 ± 0.2) × 10²² cm^-2 and Γ = 2.0 ± 0.1, 2.6 ± 0.1 (see Table 3).

The inferred 2–10 keV luminosities for the 2006 February and September observations are L_X ~ 2.1 × 10³⁶ and ~6.0 × 10³⁵ (D/6.7 kpc)² erg s^-1, respectively. Using N_H = 9.5 × 10²² cm^-2 and Γ = 2.6, we can estimate a 2–10 keV luminosity of L_X ~ 6.7 × 10³⁵ (D/6.7 kpc)² erg s^-1 for the HRC-I observation of 2005 October and upper limits for other epochs of L_X ≲ (4−37) × 10³³ (D/6.7 kpc)² erg s^-1 (see Fig. A.3). In Sect. 4.2 we further constrained the quiescent level of SAX J1747.0–2853 to L_X ~ 2 × 10³¹ (D/6.7 kpc)² erg s^-1. This confirms its classification as a transient X-ray source, as proposed by Werner et al. (2004).

Activity after our campaign: in 2009 February, INTEGRAL picked up an X-ray burst from SAX J1747.0–2853 without detectable persistent emission, suggesting an outburst luminosity of L_X ≲ 4 × 10³⁵ (D/6.7 kpc)² erg s^-1 (Chenevez et al. 2009). Follow-up Swift/XRT observations performed a few days later confirmed that the source was undergoing a faint outburst with L_X ~ 2 × 10³⁵ (D/6.7 kpc)² erg s^-1 (Campana et al. 2009).

In 2011 January–February, renewed activity was seen by MAXI (Suwa et al. 2011), Swift/XRT (Kennea et al. 2011) and INTEGRAL (Kuulkers et al. 2011). During this time INTEGRAL detected the first superburst ever observed from this source (Chenevez et al. 2011a). The reported fluxes translate into luminosities of L_X ~ 3 × 10³⁷ (D/6.7 kpc)² erg s^-1 (Kennea et al. 2011; Kuulkers et al. 2011; Chenevez et al. 2011a).

Summarising, in recent years SAX J1747.0–2853 has exhibited outbursts in 2006–2007, 2007, 2009 and 2011. The source is therefore very frequently active.

A.4. KS 1741–293

Brief historic overview: KS 1741–293 is a transient neutron star LMXB that was discovered in 1989 August by the TTM onboard the KVANT module of the Mir space station (in ’t Zand et al. 1991). At that time, the source displayed two type-I X-ray bursts, which revealed its binary nature and testified to the presence of a neutron star. KS 1741–293 was again detected with BeppoSAX in 1998 March (Sidoli et al. 2001) and with ASCA in 1998 September (Sakano et al. 2002), exhibiting a luminosity of L_X ~ 10³⁵⁻³⁶ erg s^-1 on both occasions (assuming D = 8 kpc). It is likely that these two observations caught the same outburst, which implies a duration of several months.

INTEGRAL detected KS 1741–293 in outburst in 2003 March and in 2004 March, during which two type-I X-ray bursts were observed (de Cesare et al. 2007). Furthermore, in 2005 July the source was serendipitously observed with Chandra and detected at a luminosity of L_X ~ 1 × 10³⁶ erg s^-1 (Degenaar et al. 2008). INTEGRAL revealed hard X-ray activity from the source in 2005 August–September (Kuulkers et al. 2007b), and Suzaku detected the source in outburst in 2005 September (Yuasa et al. 2008). It therefore seems that the source was active for several months in 2005, with an intensity similar to the 1989 and 1998 outburst levels.

Activity during our campaign: KS 1741–293 is located in the field GC-7 of our campaign and was found active during two different epochs (2007 and 2008). We first detected the source in outburst with XMM-Newton on 2007 September 6. It was not detected in subsequent XMM-Newton observations carried out on 2008 March 4, with an upper limit on the PN count rate of ≲3 × 10^-3 counts s^-1. Using the spectral parameters given in Table 3, we can infer an upper limit on the 2–10 keV luminosity of L_X ≲ 1 × 10³³ erg s^-1. This demonstrates that the outburst had ended and the source had returned to quiescence.

X-ray spectrum: we obtained a source spectrum from the 2007 XMM-Newton observation, which is shown in Fig. A.4 (only the PN data is displayed). The source was faint enough not to cause pile-up in the EPIC instruments. The spectral data can be fitted with N_H = (16.6 ± 1.8) × 10²² cm^-2 and Γ = 1.8 ± 0.3, resulting in a luminosity of L_X ~ 1.4 × 10³⁵ erg s^-1 (Table 3).

Outburst constraints: we searched through the Swift/XRT data archive in an attempt to constrain the duration of the 2007 outburst of KS 1741–293. The source field was covered during a number of pointings performed between 2007 May 23 and August 9. A total of 55 observations were carried out during this episode with an accumulated exposure time of ~57 ks.

Examination of the Swift data reveals that KS 1741–293 resided in quiescence for the larger part of this epoch, but that it exhibited enhanced activity during a short four-day interval between 2007 June 11–15, when its intensity rose to L_X ~ 3 × 10³⁴ erg s^-1. The source is not detected in composite images both before (total exposure of ~10 ks) and after (~43 ks) these dates. These considerations suggest that the 2007 outburst of KS 1741–293 was confined within an interval between 2007 August 9 and 2008 March 4, implying a duration of ≲6 months.

The short, sub-luminous outburst observed from KS 1741–293 is reminiscent of a one-week period of enhanced X-ray activity that was detected from the neutron star LMXB GRS 1741–2853 approximately four months before it started a major outburst (Degenaar & Wijnands 2009, see also Sect. A.1). The mini-outburst of KS 1741–293 occurred 1–2 months before the source erupted in its brighter and longer 2007 outburst.

The source was again active during our Chandra/HRC observations of 2008 May 10 (Degenaar et al. 2008) and July 16 (see Fig. A.4). Using the spectral parameters reported in Table 3, the observed HRC count rates can be translated into luminosities of L_X ~ 1 × 10³⁵ and ~1 × 10³⁶ erg s^-1, respectively. We obtained several Swift/XRT ToO follow-up observations spread ~2 weeks apart to constrain the properties of this outburst (Degenaar & Wijnands 2008). These suggest that the source remained active for approximately four more months following the HRC detection of 2008 May, and returned to quiescence between 2008 August 21 and September 4. We can therefore constrain the duration of the 2008 outburst to be ~3–6 months. The average luminosity inferred from the Swift observations is L_X ~ 10³⁵ erg s^-1, peaking at L_X ~ 2 × 10³⁶ erg s^-1.

Activity after our campaign: the results from our monitoring observations demonstrate that KS 1741–293 is frequently active and undergoes relatively long outbursts that last several months, which is also reflected by the historic detections of the source. KS 1741–293 was again detected in outburst with INTEGRAL in 2010 March (Chenevez et al. 2010). Reneweed activity was observed in 2011 September with Swift (2–10 keV luminosity of L_X ~ 2 × 10³⁵ erg s^-1; Linares et al. 2011) and INTEGRAL (Chenevez et al. 2011b). In the past decade KS 1741–293 was thus detected in an active state in 2003, 2004, 2005, 2007, 2008, 2010, and 2011. It appears that the outbursts of the source are typically several months in duration and reach 2–10 keV luminosities of L_X ~ 10³⁵⁻³⁶ erg s^-1.

A.5. GRO J1744–28

Brief historic overview: the “bursting pulsar” GRO J1744–28 is a transient neutron star LMXB that was discovered in 1995 with the BATSE on-board the Compton Gamma Ray Observatory (Fishman et al. 1995). Its neutron star nature was established by the detection of coherent X-ray pulsations with a frequency of 2.1 Hz, which allowed determining the orbital period of the binary (11.8 d; Finger et al. 1996). Major outbursts reaching L_X ~ 10³⁷⁻³⁸ erg s^-1 were observed in 1995 and 1996 (for D = 8 kpc, Woods et al. 1999). The source has been observed in quiescence at a 0.5–10 keV luminosity of L_X ~ (1−3) × 10³³ erg s^-1 (Wijnands & Wang 2002; Daigne et al. 2002), but enhanced activity above the quiescent level has also been reported (Muno et al. 2007, see also Sect. 4.3).

Activity during our campaign: GRO J1744–28 was detected during all our XMM-Newton observations, which went sufficiently deep to detect this source at its quiescent level. The source intensity was, however, near the detection limit of our HRC data (see Fig. A.5).

X-ray spectra: for the spectral fits we omitted the XMM-Newton data obtained in 2006 September, because they collected only few photons from GRO J1744–28. Using an absorbed powerlaw model to fit the other four XMM-Newton observations simultaneously, we obtain N_H = (9.4 ± 3.0) × 10²² cm^-2 and Γ = 2.5–3.2 (see Table 3). The inferred 2–10 keV luminosities of the first observations lie between L_X ~ (1.5−6.5) × 10³³ erg s^-1, whereas the final observation (2008 September) suggests enhanced activity at L_X ~ 1.9 × 10³⁴ erg s^-1 (see Fig. A.5). Figure A.5 compares the PN spectra obtained in 2008 March and September and illustrates the enhanced activity seen in the September observation.

With the exception of the last observation, the luminosities inferred for the XMM-Newton data are similar to what has been found for GRO J1744–28 in quiescence (Wijnands & Wang 2002). Since (non-pulsating) neutron star LMXBs often display a thermal spectrum in quiescence that is thought to be heat radiated from the surface of the neutron star, we also fitted a blackbody model to the spectral data. We excluded the last observation (2008 September) that hinted at enhanced activity.

A blackbody fit yields N_H = (5.9 ± 2.7) × 10²² cm^-2, a temperature kT_bb = 1.0 ± 0.3 keV and an emitting radius of about R_bb ~ 0.1 km (χ_ν = 0.86 for 40 d.o.f.). The inferred temperature is unusually high for a quiescent neutron star, because these are typically found at kT_bb ~ 0.2−0.3 keV (e.g., Rutledge et al. 1999). Moreover, the detection of this soft thermal component from GRO J1744–28 suffers considerably from the high Galactic extinction in the direction of the source. The mechanism responsible for the quiescent emission of GRO J1744–28 is therefore a subject of debate (Wijnands & Wang 2002; Daigne et al. 2002).

The spectral index for the powerlaw fit is very similar to that of the hard emission tails that are often detected for neutron star LMXBs above ~2–3 keV (e.g., Rutledge et al. 1999). Its origin is not well understood, but possible explanations include accretion onto the surface or the magnetic field of the neutron star, or a shock generated between the pulsar wind and the material flowing out of the donor star (e.g., Campana et al. 1998).

We used the results of the powerlaw fits to our XMM-Newton data to convert the HRC-I count rates and upper limits to 2–10 keV fluxes and luminosities (Table 3 and Fig. A.5).

A.6. XMM J174457–2850.3

Brief historic overview: XMM J174457–2850.3 is an unclassified X-ray transient that was first detected in outburst in 2001 (Sakano et al. 2005) and has been seen active on numerous occasions since (Wijnands et al. 2006a; Muno et al. 2007; Degenaar & Wijnands 2009, 2010; Degenaar et al. 2010b). Its quiescent luminosity is L_X ~ 10³² erg s^-1 (Sakano et al. 2005), and it exhibits outbursts with a peak luminosity of L_X ~ 10³⁶ erg s^-1. However, the source is often detected at intermediate intensities of L_X ~ 10³³⁻³⁴ erg s^-1 (Degenaar & Wijnands 2010, see also Sect. 4.3).

Activity during our campaign: we detected activity from XMM J174457–2850.3 several times during our campaign. As reported by Wijnands et al. (2006a), XMM J174457–2850.3 was detected both during the HRC observations performed on 2005 June 5 (see also Wijnands et al. 2005a) and the ACIS follow-up pointing carried out on July 1. The luminosities inferred from these observations are L_X ~ 8 × 10³⁵ and ~3 × 10³³ erg s^-1, respectively (assuming D = 8 kpc). This indicates that the intensity decayed by two orders of magnitude within four weeks.

Although we detected the source in several other observations during the campaign, we never found it to be as bright as in 2005. In 2006 February and September (XMM-Newton), 2007 March–May (Chandra/ACIS) and 2007 September (XMM-Newton) we detected it at L_X ~ (1−9) × 10³³ erg s^-1 (see Table 3). As argued in Sects. 4.3 and 5.3, these low-intensity states likely represent low-level accretion of the source. XMM J174457–2850.3 is not detected during the other epochs of our campaign, but the inferred upper limits are comparable to the above mentioned detections (see Fig. A.6).

X-ray spectra: XMM J174457–2850.3 is so faint that most of our observations collect only ~20–30 source photons, prohibiting accurate spectral modelling. We chose to only fit the spectra from the Chandra/ACIS observations 6603 and 6605, which collected the largest number of source photons (~50–60 each). Since the hydrogen column density remains unconstrained when left to vary freely, we fixed this parameter to the value inferred from Swift/XRT data (N_H = 7.5 × 10²² cm^-2; Degenaar & Wijnands 2010). This resulted in Γ = 1.3 ± 1.0 and 1.7 ± 1.1 for 2007 March 12 and April 30, respectively.

Within the errors this spectral shape is consistent with results obtained in other works (Sakano et al. 2002; Wijnands et al. 2006a; Degenaar & Wijnands 2010). For the remaining observations we converted the detected count rates into 2–10 keV unabsorbed fluxes by adopting the above mentioned hydrogen absorption column density and a photon index of Γ = 1.5. The spectrum extracted from the Chandra/ACIS observation performed on 2007 April 30 is plotted in Fig. A.6.

Constraints on the bright 2008 outburst: Swift/XRT uncovered a relatively bright outburst from XMM J174457–2850.3 in 2008 in late June. During the first observation in which the source was within FOV, it was detected at a luminosity of L_X ~ 1 × 10³⁶ erg s^-1. Subsequent observations revealed a decline in source intensity down to L_X ~ 5 × 10³³ erg s^-1 within one week (Degenaar & Wijnands 2010). The Swift/GC monitoring observations did not cover the source region before the outburst peak, leaving the duration of this relatively bright (L_X ≳ 10³⁴ erg s^-1) episode unconstrained (Degenaar & Wijnands 2010).

Since XMM J174457–2850.3 is not detected in our Chandra/HRC data obtained on 2008 May 10, yielding an upper limit on the 2–10 keV luminosity of L_X ~ 8 × 10³³ erg s^-1, we can infer that the luminous phase seen by Swift/XRT must have had a duration of <49 days.

Activity after our campaign: in 2010 in late July Swift/XRT again detected a relatively bright episode from XMM J174457–2850.3 with an observed peak luminosity of L_X ~ 2 × 10³⁵ erg s^-1 and a duration of 5–16 days (Degenaar et al. 2010b). In 2011, two weak flares were observed with Swift that had a maximum duration of a few days and an intensity of L_X ~ 1 × 10³⁴ erg s^-1.

In summary, three relatively bright (L_X ~ 10³⁵⁻³⁶ erg s^-1) outbursts have been observed from XMM J174457–2850.3 in 2005 (Chandra), 2008 and 2010 (Swift), whereas it is frequently detected at L_X ~ 10³³⁻³⁴ erg s^-1. Our monitoring observations support previous findings that the bright episodes of this source last only short times and that the source spends most of its time at luminosities that are a factor ~10–100 above its quiescent level (see also Sect. 4.3). This peculiar behaviour has led to the speculation that this unclassified X-ray source is possibly a wind-accreting X-ray binary (Degenaar & Wijnands 2010).

A.7. CXOGC J174535.5–290124

Brief historic overview: several of our observations reveal activity of the unclassified transient X-ray source CXOGC J174535.5–290124. This object was first discovered during Chandra observations in 2001 (Muno et al. 2003b) and has frequently been observed at L_X ~ 10³³⁻³⁴ erg s^-1 ever since (for D = 8 kpc; Muno et al. 2005b; Wijnands et al. 2005c, 2006b; Degenaar et al. 2008; Degenaar & Wijnands 2009). The source is not detected in quiescence down to a limiting luminosity of L_X ≲ 9 × 10³⁰ erg s^-1 (2–8 keV; Muno et al. 2005b).

Studies of this X-ray transient are complicated because of its close proximity to the neutron star LMXB AX J1745.6–2901, which is a factor ~100 brighter during outburst than CXOGC J174535.5–290124. The two sources are separated by only ~14″, (see Fig. 1). Therefore, sub-arcsec spatial resolution is required to resolve the two sources whenever AX J1745.6–2901 is active (see Degenaar & Wijnands 2010). On all occasions that AX J1745.6–2901 is active (see Sect. A.2), we were therefore only able to obtain information on CXOGC J174535.5–290124 from Chandra data.

Activity during our campaign: this transient was active during our Chandra observations performed in 2005 October (HRC; Wijnands et al. 2005c), 2007 March–May (ACIS), and 2008 May and July (both HRC; Degenaar et al. 2008). AX J1745.6–2901 resided in quiescence during our XMM-Newton observations performed in 2006 and 2008 September (see Sect. A.2): on both occasions we detected activity from CXOGC J174535.5–290124 (see also Wijnands et al. 2006b).

X-ray spectra: for the spectral fitting, we chose to use only the XMM-Newton data since that excludes contamination from AX J1745.6–2901, which resided in quiescence during those observations. The simultaneous fit results in N_H = (30.4 ± 8.6) × 10²² cm^-2 and Γ = 1.7−2.5 (see Table 3), yielding L_X ~ (1.9−2.4) × 10³⁴ erg s^-1. The spectral properties are therefore very similar at the two different epochs. We used these spectral parameters to convert the count rates and upper limits from the other observations into 2–10 keV unabsorbed fluxes (see Table 3). Figure A.7 shows the XMM-Newton/PN spectrum obtained on 2006 September 8.

Outburst constraints: the long-term lightcurve of CXOGC J174535.5–290124 obtained during our campaign is displayed Fig. A.7. This plot is suggestive of the source exhibiting two different outbursts in 2005–2007 and 2008, although the upper limits for the non-detections are close to the displayed level of activity. This makes it difficult to constrain the outburst and quiescent time scales from our observations.

Swift/XRT observations are not conclusive either, because the spatial resolution of Swift does not allow us to separate CXOGC J174535.5–290124 from the frequently active AX J1745.6–2901. The Swift data set indicates that CXOGC J174535.5–290124 was continuously active between 2006 July–November and from 2008 September till October (when AX J1745.6–2901 resided in quiescence).

In summary, the source was first detected during our campaign on 2005 October 20 and seen active during several observations performed between 2006–2008 with Chandra, XMM-Newton and Swift. The last detection during our campaign was obtained with XMM-Newton on 2008 September 9 (see Table 3). Swift resumed its monitoring observations of the source region on 2009 June 4 and did not detect activity from the source. This gives an upper limit on its intensity of

L_X ~ 10³³ erg s^-1 and suggests that the source was quiescent at that time (Degenaar & Wijnands 2010). If the activity seen between 2005 and 2008 was part of the same outburst, it had a duration of ≳35 months (≳3 yr; the time between the first and last detection during our campaign). If the source instead exhibited two different outbursts in 2005–2007 and 2008 (see Fig. A.7), the duration of these outbursts would be ≳18 and ~6–25 months, respectively.

Activity after our campaign: CXOGC J174535.5–290124 was not detected during the entire sample of 2009 Swift monitoring observations (Degenaar & Wijnands 2010), nor in 2010–2011. However, AX J1745.6–2901 was active from 2010 June–October, implying that during that epoch no information on the activity of CXOGC J174535.5–290124 could be obtained from the Swift observations (Degenaar et al. 2010c).

A.8. CXOGC J174541.0–290014

Brief historic overview: CXOGC J174541.0–290014 is a weak unclassified X-ray transient discovered during Chandra observations of the GC in 1999 (Muno et al. 2003b). This source was detected multiple times between 1999 and 2005, displaying luminosities of a few times 10³³ erg s^-1, whereas an upper limit on its quiescent level of L_X ≲ 8 × 10³¹ erg s^-1 was inferred (2–8 keV and assuming D = 8 kpc; Muno et al. 2005b).

Activity during and after our campaign: we detected activity from CXOGC J174541.0–290014 during our Chandra/HRC observations performed in 2005 October (Wijnands et al. 2005c). Using the spectral parameters reported by Muno et al. (2004), N_H = 18.8 × 10²² cm^-2 and Γ = 2.0, we can convert the HRC count rate into a luminosity of L_X ~ 1.2 × 10³⁴ erg s^-1. This is a factor ~2 higher than the maximum intensity previously seen with Chandra (Muno et al. 2004).

Although CXOGC J174541.0–290014 was frequently active between 1999 and 2005, we detected the source only once during our campaign. Nevertheless, our obtained upper limits of L_X ≲ (0.5−5) × 10³³ erg s^-1 (see Fig. A.8) are similar to the level of activity seen previously (Muno et al. 2005b). No activity has been reported from this source after our campaign ended.

---

Acknowledgments

This work was supported by the Netherlands organisation for scientific research (NWO) and The Netherlands Research School for Astronomy (NOVA). N.D. is supported by NASA through Hubble Postdoctoral Fellowship grant number HST-HF-51287.01-A from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS5-26555. R.W. acknowledges support from a European Research Council (ERC) starting grant. This research has made use of data obtained from the Chandra Data Archive and the Chandra Source Catalog, and software provided by the Chandra X-ray Center (CXC) in the application package CIAO. This work was in part based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA. We also made use of the Swift public data archive.

References

All Tables

All Figures