A&A 484, 783-800 (2008)
DOI: 10.1051/0004-6361:20078768

Multi-wavelength observations of Galactic hard X-ray sources discovered by INTEGRAL[*],[*]

I. The nature of the companion star

S. Chaty1 - F. Rahoui2,1 - C. Foellmi3 - J. A. Tomsick4 - J. Rodriguez1 - R. Walter5

1 - Laboratoire AIM, CEA/DSM - CNRS - Université Paris Diderot, Irfu/Service d'Astrophysique, Bât. 709, CEA-Saclay, 91191 Gif-sur-Yvette Cedex, France
2 - ESO, Alonso de Cordova 3107, Vitacura, Casilla 19001, Santiago 19, Chile
3 - Laboratoire d'Astrophysique, Observatoire de Grenoble, BP 53, 38041 Grenoble Cedex 9, France
4 - Space Sciences Laboratory, 7 Gauss Way, University of California, Berkeley, CA 94720-7450, USA
5 - INTEGRAL Science Data Centre, Chemin d'Écogia 16, 1290 Versoix, Switzerland

Received 30 September 2007 / Accepted 15 December 2007

Context. The INTEGRAL hard X-ray observatory has revealed an emerging population of highly obscured X-ray binary systems through multi-wavelength observations. Previous studies have shown that many of these sources are high-mass X-ray binaries hosting neutron stars orbiting around luminous and evolved companion stars.
Aims. To better understand this newly-discovered population, we have selected a sample of sources for which an accurate localisation is available to identify the stellar counterpart and reveal the nature of the companion star and of the binary system.
Methods. We performed an intensive study of a sample of thirteen INTEGRAL sources, through multi-wavelength optical to NIR photometric and spectroscopic observations, using EMMI and SofI instruments at the ESO NTT telescope. We performed accurate astrometry and identified candidate counterparts for which we give the optical and NIR magnitudes. We detected many spectral lines allowing us to determine the spectral type of the companion star. We fitted with stellar black bodies the mid-infrared to optical spectral energy distributions of these sources. From the spectral analysis and SED fitting we identified the nature of the companion stars and of the binary systems.
Results. Through spectroscopic analysis of the most likely candidates we found the spectral types of IGR J16320-4751, IGR J16358-4726, IGR J16479-4514, IGR J17252-3616, IGR J18027-2016: They all host OB type supergiant companion stars, with IGR J16358-4726 likely hosting an sgB[e]. Our spectra also confirm the supergiant O and B nature of IGR J17391-3021 and IGR J19140+0951. From SED fitting we found that IGR J16418-4532 is a (likely OB supergiant) HMXB, IGR J16393-4643 a (likely BIV-V star) HMXB, and IGR J18483-0311 a likely HMXB system. Through accurate astrometry, we rejected the proposed counterparts of IGR J17091-3624 and IGR J17597-2201, and we discovered two new candidate counterparts for each source, both suggesting an LMXB from SED fitting. We confirm the AGN nature of IGR J16558-5203. Finally, we report that NIR fields of four sources of our sample exhibit large-scale regions of absorption.
Conclusions. The majority of these systems are high-mass X-ray binaries hosting supergiant companion stars. We therefore confirm that INTEGRAL reveals a dominant class of obscured and short-lived high-energy binary systems, and we suggest an association of these systems with regions of the Galaxy exhibiting large-scale absorption. Stellar population models should take these systems into account for realistic estimates of high-energy binary systems in our Galaxy.

Key words: infrared: stars - X-rays: binaries - X-rays: general - stars: supergiants

1 Introduction

The hard X-ray INTEGRAL observatory was launched on the 17th of October 2002, and since then, it has performed a detailed survey of the Galactic plane. The ISGRI detector on the IBIS imager (Lebrun et al. 2003) has discovered many new hard X-ray sources[*], including binary systems, pulsars and AGNs, all these so-called IGR sources being reported in Bird et al. (2007) and Bodaghee et al. (2007). One of the most important achievements of the INTEGRAL observatory to date is that it is revealing hard X-ray sources which were not easily detected in earlier soft X-ray (typically $\leq$10 keV) observations, bringing to light a previously hidden part of a population of highly obscured high-energy binary systems in our Galaxy. These objects share characteristics that have rarely been seen (see Dean et al. 2005). They are high-mass X-ray binaries (HMXBs) hosting a neutron star orbiting an OB companion star, in some cases a supergiant star (see e.g. Filliatre & Chaty 2004; and Pellizza et al. 2006), some of them possibly being long-period X-ray pulsars. Many of these new sources are highly absorbed, exhibiting column densities higher than about $10^{23} \mbox{ cm}^{-2}$, and are concentrated in directions tangential to Galactic arms, for instance the Norma arm (see Chaty & Filliatre 2005; Tomsick et al. 2004a; and Walter et al. 2004b), the richest arm of our Galaxy in high-mass star forming regions. This short-lived population hosts the likely progenitors of extremely compact binary objects, which are good candidates of gravitational wave emitters, and might constitute a key sample in the understanding of the evolution of high-energy binary systems.

Among these HMXBs hosting an OB supergiant companion star, two classes, which might overlap, appear (Chaty & Rahoui 2006). The first class is constituted of intrinsically highly obscured hard X-ray sources, exhibiting a huge local extinction. The most extreme example of these sources is the highly absorbed source IGR J16318-4848 (Filliatre & Chaty 2004). The second class exhibits fast and transient outbursts, with peak fluxes of the order of $10^{-9} \mbox{ erg~s}^{-1}~\mbox{cm}^{-2}$ in the 20-40 keV band, and lasting only a few hours. This last characteristic is very unusual among HMXBs. For this reason they are called Supergiant Fast X-ray Transients (SFXTs, Negueruela et al. 2006b). These SFXTs have faint quiescent emission, and their hard X-ray spectra require a black hole or neutron star accretor. Among these sources, IGR J17544-2619 (Pellizza et al. 2006) seems to be the archetype of this new class of HMXBs, with long quiescent periods (Zurita Heras & Chaty, in prep.).

Even if the INTEGRAL observatory can provide a localisation that is accurate above 10 keV ($\sim$$2^\prime$), it is not accurate enough to pinpoint the source at other wavelengths, which is necessary to reveal the nature of these sources. So the first step in the study of these sources is to look for an accurate localisation of the hard X-ray sources by X-ray satellites such as XMM-Newton, Swift or Chandra. While XMM-Newton and Swift provide positions good enough to restrict the list of possible counterparts to a small number, only Chandra gives unique identifications in most cases (see for instance the multi-wavelength study of four INTEGRAL sources in the direction of the Norma arm by Tomsick et al. 2006, via Chandra localisation). Once the source position is known to better than several arcseconds, the hunt for the optical counterpart of the source can begin. However, a difficulty persists, due to the high level of interstellar absorption in this region of the Galaxy, close to the Galactic plane. Near-infrared (NIR) observations, thanks to accurate astrometry, and photometric and spectroscopic analysis, allow for the nature of these sources to be revealed, constraining the spectral type of the companion star and the type of the binary system.

Finally, mid-infrared (MIR) observations are required to understand why these sources exhibit a strong local absorption, by characterizing the nature of the absorbing matter, and determining whether it is made of cold gas or dust, or anything else. Based on the optical/NIR observations we report here, we have performed MIR observations with the VISIR instrument on ESO VLT/UT3. These observations, focusing on the characterization of the presence, temperature, extension and composition of the absorbing material constituting the circumstellar medium enshrouding the obscured sources, are described in a companion paper by Rahoui et al. (2008).

Here we report on an intensive multi-wavelength study of a sample of 13 INTEGRAL sources belonging to both obscured and SFXT classes, for which accurate X-ray localisations are available, aimed at identifying their counterparts and constraining the nature of the companion star and of the binary system. We first describe the ESO optical/NIR photometric and spectroscopic observations, and build the spectral energy distributions (SEDs) in Sect. 2. We then review hard X-ray properties of each source, and report the results of our optical/NIR observations in Sect. 3. We give general results and discuss them in Sect. 4, and provide our conclusions in Sect. 5.

2 Observations

The multi-wavelength observations that we describe here are based on astrometry, photometry and spectroscopy on 13  INTEGRAL sources indicated in Table 1. They were performed at the European Southern Observatory (ESO, Chile), in 2 domains: in optical ( $0.4{-}0.8 \mbox{ } \mu \mbox{m}$) with the EMMI instrument and in NIR ( $1{-}2.5 \mbox{ } \mu \mbox{m}$) with the SofI instrument, both with the 3.5 m New Technology Telescope (NTT) at La Silla Observatory. Our optical and NIR observations were carried out as part of the programme ESO # 073.D-0339, in the visitor mode[*].

Table 1: Sample of sources. Name and coordinates of the sources: position (RA, Dec, J2000.0), galactic longitude and latitude (l,b), uncertainty (Unc. in arcmin) and reference (Ref1) of the discovery of the source by INTEGRAL; and position (RA, Dec, J2000.0), uncertainty (Unc.) and reference (Ref2) of the most accurate position by the satellite indicated in parenthesis.

Table 2: Results in Optical. We indicate the name of the source, the date and UT time of the observations, the airmass (AM), the exposure time in seconds (ET) and the B, V, R, I and Z magnitudes. Z-band magnitudes are instrumental.

Table 3: Log of optical and NIR spectra. We indicate the name of the source, the telescope used, the date and UT time of the observations, the airmass (AM), the exposure time in seconds in optical and NIR (blue and red grisms respectively). All spectra were obtained at ESO/NTT with EMMI and SofI instruments, except the optical spectra of IGR J17252-3616 and IGR J18027-2016 obtained at ESO/3.6 m with EFOSC2.

Table 4: Results in NIR. We indicate the name of the source (C1, C2, etc. indicate the different candidates as labeled in the finding charts of Fig. 2), the date and UT time of the observations, the airmass, the exposure time (Exptime) in seconds, and the J, H and $K_{\rm S}$ magnitudes.

2.1 Optical observations

On 2004 July 10 between UT 0.0 and 11.0 we obtained optical photometry in B, V R, I and Z bands of the sources given in Table 2 with the spectro-imager EMMI, installed on the NTT. We used the large field of EMMI's detector, with the images binned by a factor of 2, giving an image scale of $0{\rlap.}^{\prime \prime}332$ /pixel and a field of view of $9{\rlap.}^{\prime}0$ $\times $ $10{\rlap.}^{\prime}0$. The photometric observations were performed with an integration time between 1 and 30 s for each exposure, as reported in Table 2. We observed five photometric standard stars of the optical standard star catalogue of Landolt (1992): PG 1633+099, PG 1633+099A; PG 1633+099B; PG 1633+099C and PG 1633+099D.

We also carried out optical spectroscopy on 2004 July 9, taking 30 spectra of the source IGR J17391-3021 with the low-resolution grism #1 of EMMI, with a slit of $1{\rlap.}^{\prime \prime}0$ providing a resolution of about 350, and a spectral range between 4370 and $10~270~\AA$ with a dispersion of about $7.3~\AA$. Each individual spectrum had an exposure time between 60 and 120 s, giving a total integration time between 720 and 1440 s, as reported in Table 3. We observed the spectro-photometric standard star LTT 7379 taken with a similar airmass to flux calibrate the optical spectra.

Finally, through internal Director's Discretionary Time (DDT) we obtained long-slit low-resolution spectra of 2 sources - IGR J17252-3616 and IGR J18027-2016 - with the spectro-imager EFOSC2 installed on the 3.6 m telescope of La Silla Observatory.

2.2 NIR observations

We performed NIR photometry in J, H and $K_{\rm s}$ bands of the sources given in Table 1 on 2004 July 08-11 with the spectro-imager SofI, installed on the NTT. We used the large field of SofI's detector, giving an image scale of $0{\rlap.}^{\prime \prime}288$ /pixel and a field of view of $4{\rlap.}^{\prime}92$ $\times $ $4{\rlap.}^{\prime}92$. The photometric observations were obtained by repeating a set of images for each filter with 9 different 30 $^{\prime \prime }$ offset positions including the targets, following the standard jitter procedure allowing us to cleanly subtract the sky emission in NIR. The integration time varied between 10 and 50 s for each individual exposure, giving a total exposure time between 108 and 450 s. The NIR photometry results are given in Table 4. We observed three photometric standard stars of the faint NIR standard star catalogue of Persson et al. (1998): sj9157, sj9172 and sj9181.

We also carried out NIR spectroscopy with SofI between 0.9 and $2.5 \mbox{ } \mu \mbox{m}$, taking 12 spectra using alternatively the low-resolution Blue and Red grisms respectively, half of them with the $1{\rlap.}^{\prime \prime}0$ slit on the source and the other half with an offset of 30 $^{\prime \prime }$, in order to subtract the NIR sky emission. Each individual spectrum has an exposure time from 60 to 180 s, giving a total integration time between 720 and 2160 s in each grism, as reported in Table 3.

2.3 Data reduction

We used the IRAF (Image Reduction and Analysis Facility) suite to perform data reduction, carrying out standard procedures of optical and NIR image reduction, including flat-fielding and NIR sky subtraction.

We performed accurate astrometry on each entire SofI $4{\rlap.}^{\prime}92$ $\times $ $4{\rlap.}^{\prime}92$ field, using all stars from the 2MASS catalogue present in this field[*] (amounting to $\sim$1000 2MASS objects). The rms of astrometry fit is always less than $0{\rlap.}^{\prime \prime}5$, and we obtained a pixel scale in xy axis of $-0{\rlap.}^{\prime \prime}28783$ and $0{\rlap.}^{\prime \prime}28801$ /pixel respectively. The finding charts including the results of our astrometry are shown in Figs. 2 and 3 for all sources of our sample.

We carried out aperture photometry, and we then transformed instrumental magnitudes into apparent magnitudes using the standard relation: $mag_{\rm app} = mag_{\rm inst} - Zp - ext \times AM$where $mag_{\rm app}$ and $mag_{\rm inst}$ are the apparent and instrumental magnitudes, Zp is the zero-point, ext the extinction and AM the airmass. We used for the extinction extJ = 0.06, extH = 0.04 and $ext_{K{\rm s}} = 0.10$ typical of the La Silla observatory. The log of the observations and the results of photometry are given in Tables 2 and 4 for the optical and NIR respectively.

Table 5: GLIMPSE fluxes (in mJy) for the sources IGR J16393-4643, IGR J16418-4532, IGR J18027-2016 and IGR J18483-0535. The fluxes for the other sources are given in Rahoui et al. (2008).

Table 6: Summary of parameters we used to fit the SEDs of the sources. We give their name, the interstellar extinction in magnitudes $A_{\rm i}$, the X-ray extinction of the source in magnitudes $A_{\rm x}$ and the parameters of the fit: the extinction in the optical $A_{\rm v}$, the temperature $T_\ast $ and the $\frac{{R}_{\ast}}{{D}_{\ast}}$ ratio of the companion (more details on these parameters are given in the text). The 90%-confidence ranges of these parameters are given in parenthesis. We also give the reduced $\chi ^2$. The parameters for the other sources are given in Rahoui et al. (2008).

We analyzed the optical spectra using standard IRAF tasks, subtracting bias and correcting for flat field, and we used the IRAF noao.twodspec package in order to extract spectra and perform wavelength and flux calibrations. The optical spectra were reduced and flux-calibrated in ``F-lambda'' units -  $\mbox{ erg~cm}^{-2}~\mbox{s}^{-1}~\AA$.

NIR spectra were reduced using IRAF by flat-fielding, correcting the geometrical distortion using the arc frame, shifting the individual images using the jitter offsets, combining these images and finally extracting the spectra. The analysis of SofI spectroscopic data, and more precisely the sky subtraction, was difficult due to a variable sky, mainly in the red part of the blue grism, causing some wave patterns. The target spectra were then corrected for the telluric lines using a median of various standard stars observed with the same configuration during the corresponding nights. All spectra, optical and NIR, are finally shifted to the heliocentric rest frame.

2.4 SEDs

Once the most likely counterpart was identified through astrometry, photometry and spectroscopy when available, we were able to build the optical/NIR SEDs of all sources of our sample, shown in Fig. 1. While these SEDs are mainly based on optical and NIR photometry from this paper, we also add MIR observations ( $5{-}20
\mbox{ } \mu \mbox{m}$) obtained with the VISIR instrument on Melipal, the 8 m third Unit Telescope (UT3) of the ESO Very Large Telescope (VLT) at Paranal Observatory, reported in the companion paper by Rahoui et al. (2008). We also put MIR data taken from the Spitzer GLIMPSE survey, reported in Rahoui et al. (2008), and in Table 5 for the sources not included in this companion paper. We include in these SEDs data from X-ray observations taken with INTEGRAL/IBIS for all the sources, XMM for IGR J16418-4532, IGR J17252-3616 and IGR J19140+0951, RXTE for IGR J16358-4726, IGR J17091-3624, IGR J17597-2201 and IGR J17391-3021, ASCA for IGR J16320-4751, IGR J16393-4643 and IGR J16479-4514, BeppoSAX for IGR J18027-2016, and finally INTEGRAL/JEM-X for IGR J18483-0311.

We fitted the optical to MIR observing data with a black body emission reproducing the stellar emission. The broad band SEDs of these sources were modelled using an absorbed black body emission component, representing the companion star emission ($D_{\ast}$ and $R_{\ast}$ are respectively the distance and radius of the star):

...ac{hc}{{\lambda}k{T}_\ast}}-1}\right]~~~~~~~{\rm in~W~m}^{-2}.
\end{displaymath} (1)

The free parameters of the fits were the absorption in the V band $A_{\rm v}$, the companion star black body temperature $T_\ast $ and its $\frac{{R}_{\ast}}{{D}_{\ast}}$ ratio. The fits were performed using a $\chi ^2$ minimization.

Best-fitting parameters for individual sources, as well as corresponding $\chi ^2$ and 90%-confidence ranges of parameters, are reported in Table 6 for the sources IGR J16393-4643, IGR J16418-4532, IGR J17091-3624, IGR J17597-2201, IGR J18027-2016 and IGR J18483-0311, and in Rahoui et al. (2008) for the remaining sources. The IGR J16558-5203 SED has not been fitted, because of its AGN nature. In this table we also give the interstellar extinction in magnitude, $A_{\rm i}$, obtained from the neutral hydrogen column density  $N_{\rm HI}$, and the X-ray extinction in magnitude, $A_{\rm X}$, obtained from X-ray observations. Both extinctions were converted to magnitudes using the conversion between $N_{\rm H}$ and $A_{\rm v}$ given by Predehl & Schmitt (1995). $N_{\rm HI}$ has been computed using the $N_{\rm H}$ tool from HEASARC[*] (Dickey & Lockman 1990). Since $N_{\rm HI}$ is the total galactic column density along the line of sight, it is likely overestimated compared to the real value at the distance of the sources.

We overplot on the SEDs of Fig. 1 the best-fitted model to our observations. The dip seen at $\sim$$\times $ 1013 Hz is due to silicate absorption present in our extinction model (more details on this model are given in Rahoui et al. 2008).

...\includegraphics[angle=-90,width=5.7cm,clip]{igr19140_he+fit3.ps} }
\end{figure} Figure 1: SEDs of all these INTEGRAL sources, showing the observations from hard X-rays to MIR wavelengths. In each case, we overplot the black body emission representing the stellar spectral type of the companion star, which is given in Table 7 for the sources IGR J16393-4643, IGR J16418-4532, IGR J17091-3624, IGR J17597-2201, IGR J18027-2016 and IGR J18483-0535, and in Rahoui et al. (2008) for the remaining sources. See Sect. 2.4 for more details on the X-ray data. From top to bottom, and left to right: IGR J16320-4751, IGR J16358-4726, IGR J16393-4643, IGR J16418-4532, IGR J16479-4514, IGR J17091-3624, IGR J17252-3616, IGR J17391-3021, IGR J17597-2201, IGR J18027-2016, IGR J18483-0311 and IGR J19140+0951.
Open with DEXTER

\par\mbox{\subfigure[IGR~J16320-4751 (3$^{\prime \prime}$\space {...
...it Chandra})]{ \includegraphics[width=5.3cm,clip]{I17391-Ks.eps} }}
\end{figure} Figure 2: Finding charts of the studied INTEGRAL sources, observed at ESO NTT telescope in the infrared $K_{\rm S}$ band (2.2  $\mbox{ } \mu \mbox{m}$). Size: 1$^\prime $ $\times $ 1$^\prime $; North is to the top and East to the left. They are all centered apart from IGR J16558-5203, because the source was caught at the edge of the CCD. We overplot the most accurate localisation available to date.
Open with DEXTER

\par\mbox{\subfigure[IGR~J17597-2201 (4$^{\prime \prime}$\space {...
...it Chandra})]{ \includegraphics[width=5.3cm,clip]{I19140-Ks.eps} }}
\end{figure} Figure 3: Fig. 2 continued.
Open with DEXTER

3 Results on individual sources

All the sources studied in this paper were discovered with the IBIS/ISGRI detector onboard the INTEGRAL observatory. The sample of 13 sources, along with their position, uncertainty, and references about their discovery and position are given in Table 1. We present in the following our results on each source, for which we followed the same strategy. We first observed the field in optical and NIR, we performed accurate astrometry, and we derived the photometry of the candidates inside the X-ray satellite error circle. We then analyzed the optical/NIR spectrum of the most likely candidate, when available. We give the results of the optical-MIR SED fitting.

3.1 IGR J16320-4751

IGR J16320-4751 was discovered on 2003 February by Tomsick et al. (2003) at the position ${\rm RA} = 16^{\rm h}32{\rlap.}^{\rm m}0$, Dec = $-47^{\circ}51^\prime$ (equinox J2000.0; uncertainty $2^\prime$). Follow-up XMM-Newton observations localized the source at $16^{\rm h}32^{\rm m}01{\rlap.}^{\rm s}9$ $-47^{\circ}52^\prime27^{\prime \prime}$ with 3 $^{\prime \prime }$accuracy (Rodriguez et al. 2003; Rodriguez et al. 2006). It is a heavily absorbed variable source with $N_{\rm H}\sim2.1$ $\times $ $10^{23} \mbox{ cm}^{-2}$, and a hard X-ray spectrum fitted by an absorbed power-law, with $\Gamma\sim1.6$(Rodriguez et al. 2006). Soft X-ray pulsations have been detected from this source at a period of $P\sim1309$ $\pm$ 40 s with XMM-Newton and $P\sim1295$ $\pm$ 50 s with ASCA observations, these pulsations being the signature of an X-ray pulsar (Lutovinov et al. 2005b). An orbital period of $8.96\pm0.01~{\rm days}$ was found from a Swift/BAT lightcurve extending from 2004 December 21 to 2005 September 17 (Corbet et al. 2005), and of 8.99 $\pm$ $0.05~{\rm days}$ with INTEGRAL (Walter et al. 2006). Putting the spin and orbital periods of this source on a Corbet diagram (Corbet 1986) suggests a supergiant HMXB nature. IGR J16320-4751 might have been persistent for at least 8 years, since this source is the rediscovery of a previously known ASCA source AX J1631.9-4752.

We performed accurate astrometry of the field (rms of fit = $0{\rlap.}^{\prime \prime}49$), and overplot the 3 $^{\prime \prime }$ XMM-Newton error circle in the finding chart of Fig. 2. We give the infrared magnitudes of all the candidate counterparts in Table 4. Two candidate counterparts had been proposed for this source (Rodriguez et al. 2003), however the XMM-Newton 3 $^{\prime \prime }$ error circle made the ambiguity disappear, accurately localizing the candidate labeled 1 in Fig. 2, and making it the most likely counterpart (2MASS J16320215-4752289; Rodriguez et al. 2006). This result is in agreement with Negueruela & Schurch (2007) rejecting candidate 2 on the basis of 2MASS photometry. In addition, Candidate 1 is much more absorbed than Candidate 2, as can be seen from the optical and NIR magnitudes given for both candidate counterparts in Tables 2 and 4, since Candidate 1 is invisible in the optical, but becomes as bright as Candidate 2 in the $K_{\rm S}$ band. There are at least two faint candidate counterparts (labeled 3 and 4) inside the error circle, therefore it would be useful to obtain a more accurate position to unambiguously pinpoint the correct counterpart. However, the faintness of Candidates 3 and 4 tends to rule them out as counterparts, and in the following we will consider Candidate 1 as the most likely counterpart of this source.

NIR spectra of candidate 1 of IGR J16320-4751 are shown in Fig. 4. We report the detected lines in Table 8. There are only a few lines visible in the blue NIR spectrum, probably because it is very faint and absorbed. The red NIR spectrum exhibits a very red continuum, and the presence of absorption and emission lines: the Pa(7-3) emission line, the Brackett series with P-Cygni profiles between 1.5 and $2.17 \mbox{ } \mu \mbox{m}$, and $He {\sc i}$ at $2.166 \mbox{ } \mu \mbox{m}$ (perhaps with P-Cygni profile). The presence of these narrow and deep Paschen and $He {\sc i}$ lines, associated with P-Cygni profiles, are typical of early-type stars, and more precisely of luminous supergiant OB stars (Caron et al. 2003; Munari & Tomasella 1999), which is therefore the likely spectral type of the companion star. Furthermore, the presence of a wide Br $\gamma$ emission line constrains the spectral type to a O supergiant or even O hypergiant (Hanson et al. 2005). We also took optical and NIR spectra of Candidate 2: they do not exhibit any emission lines, and seem typical of a late-type star. Such a spectral type would therefore be hard to reconcile with the wind accretion hard X-ray spectra and the localisation of this source in the Corbet diagram. Therefore both astrometry and spectroscopy allow us to exclude Candidate 2 as a candidate counterpart. These results strengthen Candidate 1 as the real counterpart of IGR J16320-4751. From the nature of the companion star and of the compact object, we derive that this source belongs to the very obscured supergiant HMXB class, hosting a neutron star. This result is also in agreement with the fit of its SED, computed in Rahoui et al. (2008) and shown in Fig. 1.

3.2 IGR J16358-4726

IGR J16358-4726 was discovered on 2003 March 19 by Revnivtsev et al. (2003b) at the position (RA Dec J2000.0) = ( $16^{\rm h}35{\rlap.}^{\rm m}8$, $-47^{\circ}26^\prime$, $1{\rlap.}^{\prime}5$ uncertainty). This hard X-ray source was observed for 25 700 s serendipitously with Chandra during a scheduled observation of SGR 1627-41 on 2003 March 24. Chandra localized the source at ( $16^{\rm h}35^{\rm m}53{\rlap.}^{\rm s}8$, $-47^{\circ}25^\prime41{\rlap.}^{\prime \prime}1$) with $0{\rlap.}^{\prime \prime}6$accuracy (Kouveliotou et al. 2003). We point out however that the uncertainty may be somewhat underestimated since the source was $9{\rlap.}^{\prime}7$ from the Chandra aimpoint so that the point-spread function is significantly broadened. This source is a transient source, its hard X-ray spectrum being well fitted with a heavily absorbed power-law: $\Gamma\sim0.5$ and $N_{\rm H}\sim3.3$ $\times $ $10^{23} \mbox{ cm}^{-2}$, with the presence of Fe K$\alpha$ emission line (Patel et al. 2007). By performing detailed spectral and timing analysis of this source using multi-satellite archival observations, Patel et al. (2007) have detected 5880 $\pm$ 50 s periodic variations, which could be due either to the spin of a neutron star, or to an orbital period, and they identified a 94 s spin-up in 8 days, corresponding to a mean spin period derivative of 1.9 $\times $ 10-4 s/s, pointing to a neutron star origin. Assuming that this spin up is due to accretion, they estimate the source magnetic field to be between 1013 and 1015 G, suggesting that the compact object might be a magnetar. These observations suggest that this source is an X-ray pulsar at a distance of $\sim$6-8 kpc. No radio emission has been detected.

...id.eps}\includegraphics[width=8.8cm,clip]{IGR_J19140+0951_id.eps} }
\end{figure} Figure 4: From top to bottom, and left to right: combined blue and red grism NIR spectra of IGR J16320-4751 (Candidate 1), IGR J16358-4726, IGR J16479-4514 (Candidate 1) and IGR J19140+0951 (the y axis is in arbitrary units).
Open with DEXTER

A 2MASS counterpart has been suggested by Kouveliotou et al. (2003) based on the accurate Chandra position (2MASS J16355369-4725398, with J=15.41, H=13.44, K=12.59). We performed accurate astrometry of the field (rms of fit = $0{\rlap.}^{\prime \prime}45$), and overplot the $0{\rlap.}^{\prime \prime}6$ Chandra error circle, as shown in Fig. 2. The 2MASS counterpart is at 1.2 $^{\prime \prime }$ from the centre, therefore outside the error circle. On the other hand, this 2MASS counterpart might be a blended object, since it shows an extension towards the east, right at the position of the error circle. A better spatial resolution would allow us to confirm or not whether the 2MASS candidate is the real counterpart of this source. However, its brightness favors this candidate, and in the following we consider it as the candidate counterpart. Optical and NIR magnitudes of this 2MASS candidate are given in Tables 2 and 4 respectively.

The NIR spectra of IGR J16358-4726 are shown in Fig. 4. We report the detected lines in Table 8. The NIR spectrum is very faint and extremely absorbed, however we detect some lines even in the blue part of its spectrum, mainly $He {\sc ii}$ emission lines. The red part of the NIR spectrum exhibits a red continuum, and the presence of absorption and emission lines, with tentative P-Cygni profiles: the H Brackett series with P-Cygni profiles between 1.5 and $2.2 \mbox{ } \mu \mbox{m}$, and $He {\sc i}$ and $He {\sc ii}$ absorption lines. The presence of these lines, associated with P-Cygni profiles, are typical of an OB supergiant star, which is therefore probably the spectral type of the companion star. In this case it would be a supergiant HMXB. In addtion, we clearly detect the forbidden [ $Fe {\sc ii}$] line at $2.22\mbox{ } \mu \mbox{m}$ (and tentatively the allowed $Fe {\sc ii}$ line at $1.98\mbox{ } \mu \mbox{m}$), suggesting that the companion star is a sgB[e] star. Furthermore, it is interesting to note that Rahoui et al. (2008) propose, using an independent method of SED fitting, that the companion might be a sgB[e] star. Our result would therefore be in agreement with the fit of its SED, shown in Fig. 1.

3.3 IGR J16393-4643

IGR J16393-4643 was discovered by Malizia et al. (2004) at the position (RA, Dec, J2000.0) = ( $16^{\rm h}39{\rlap.}^{\rm m}3, -46^{\circ}43^\prime$) ($2^\prime$uncertainty). The improved position from XMM-Newton/EPIC is (RA Dec, J2000.0) = ( $16^{\rm h}39^{\rm m}05{\rlap.}^{\rm s}4, -46^{\circ}42^\prime12^{\prime \prime}$) ( $4^{\prime \prime}$ uncertainty) which is compatible with that of 2MASS J16390535-4642137 (Bodaghee et al. 2006). It is a persistent, heavily-absorbed ( $N_{\rm H}= 2.5$ $\times $ $10^{23} \mbox{ cm}^{-2}$), and hard ( $\Gamma=1.3$ $\pm$ 1.0) wind-accreting pulsar, and a pulse period of 912.0 $\pm$ $0.1~{\rm s}$ was discovered in the ISGRI and EPIC light curves, characteristic of a spin period of an X-ray pulsar (Bodaghee et al. 2006). The high column density and hard spectral index suggest that IGR J16393-4643 is an HMXB. This source exhibits large variations in intensity, and shows X-ray lines. An orbital period of 3.6875 $\pm$ 0.0006 days has been detected in Rossi-XTE data, implying a mass function of 6.5 $\pm$ $1.1\mbox{ }M_{\odot}$ (or up to $14\mbox{ }M_{\odot}$ if the orbit is eccentric; Thompson et al. 2006), this lower limit on the mass confirming that the system is an HMXB.

We performed accurate astrometry of the field (rms of fit = $0{\rlap.}^{\prime \prime}54$). Inside the error circle, there is the 2MASS candidate counterpart proposed by Bodaghee et al. (2006) (2MASS J16390535-4642137), and in addition there are 3 more candidates, labeled 1 to 4 in Fig. 2. They are located at 1.7, 3.1, 3.4 and $3{\rlap.}^{\prime \prime}3$ respectively from the centre of the error circle. Optical and NIR magnitudes of these candidates are given in Tables 2 and 4 respectively. A more precise localisation of the hard X-ray source is therefore necessary to know which is the real counterpart of the source, however Candidate 1 seems to be the most likely counterpart based on the astrometry and on its NIR brightness.

We fitted its SED with the model described in Sect. 2.4, taking the optical and NIR magnitudes of Candidate 1. We obtain a stellar temperature $T_{\ast} = 24~400~{\rm K}$, typical of a B spectral type companion star. The other parameters are given in Table 6, and the SED is shown in Fig. 1. By taking the $R_\ast/D_\ast$(=2.21 $\times $ 10-11) minimizing $\chi ^2$, and assuming a radius $R_{\ast} = 10 \mbox{ }R_{\odot}$ typical of a BIV-V spectral type companion star, we derive a distance of $D_\ast = 10.6 \mbox{ kpc}$. On the other hand, by taking the minimum radius of a B supergiant star, i.e. at least $20 \mbox{ }R_{\odot}$, we derive a distance of nearly $20.4\mbox{ kpc}$, probably too large to be plausible. The fit therefore is entirely consistent with its HMXB nature and favours a BIV-V companion star.

3.4 IGR J16418-4532

IGR J16418-4532 was discovered on 2003 February 1-5 at the position (RA Dec J2000.0) = ( $16^{\rm h}41{\rlap.}^{\rm m}8$, $-45^{\circ}32^\prime$, $2^\prime$ uncertainty), towards the Norma region (Tomsick et al. 2004b). XMM-Newton localized the source at ( $16^{\rm h}41^{\rm m}51{\rlap.}^{\rm s}0, -45^{\circ}32^\prime25^{\prime \prime}$) with 4 $^{\prime \prime }$ accuracy (Walter et al. 2006). XMM-Newton observations have shown that it is a heavily absorbed X-ray pulsar exhibiting a column density of $N_{\rm H}\sim 1.0$ $\times $ $10^{23} \mbox{ cm}^{-2}$, a peak-flux of $\sim$80 mCrab (20-30 keV), and a pulse period of 1246 $\pm$ 100 s (Walter et al. 2006). This source is an SFXT candidate, as proposed by Sguera et al. (2006) using INTEGRAL observations. A $3.75~{\rm day}$ modulation was found in Rossi-XTE/ASM and Swift/BAT lightcurves, with a possible total eclipse, which would suggest either a high binary inclination, or the presence of a supergiant companion star (Corbet et al. 2006). The latter case would be consistent with the position of this object in the Corbet diagram.

We performed the astrometry of this source with 1139 2MASS stars (rms of fit = $0{\rlap.}^{\prime \prime}38$). The field is shown in Fig. 2. The centre of the XMM-Newton error circle is $2{\rlap.}^{\prime \prime}2$ away from a bright 2MASS source (2MASS J16415078-4532253, labelled Candidate 1), as pointed out by Walter et al. (2006), however there are at least three more possible counterparts in the $4^{\prime \prime}$ XMM-Newton error circle, not present in the 2MASS catalogue, at 1.9, 1.3 and $2{\rlap.}^{\prime \prime}1$ respectively from the centre of the error circle. These additional candidate counterparts are labeled 2 to 4 respectively. We give their NIR magnitudes in Table 4. From the brightness in NIR we favor Candidate 1 as the candidate counterpart.

The fit of its SED, shown in Fig. 1, and taking optical and NIR magnitudes as described in Sect. 2.4, gives a stellar temperature $T_{\ast} = 32~800~{\rm K}$, suggesting an OB spectral type companion star. The other parameters of the fit are given in Table 6. The $R_\ast/D_\ast$ ratio which minimizes $\chi ^2$ (=3.77 $\times $ 10-11) allows us to derive a minimal distance of 13 kpc for a supergiant, which is plausible. This source is therefore an HMXB, and could be a supergiant HMXB, consistent with its position in the Corbet diagram. We point out however its membership of the SFXT class is uncertain, based on its X-ray behaviour (Zurita Heras & Chaty, in prep.).

3.5 IGR J16479-4514

IGR J16479-4514 was discovered on 2003 August 8-9 at the position (RA Dec J2000.0) = ( $16^{\rm h}47{\rlap.}^{\rm m}9, -45^{\circ}14^\prime$), uncertainty $\sim$$3^\prime$, by Molkov et al. (2003). XMM-Newton observations localized the source at ( $16^{\rm h}48^{\rm m}06{\rlap.}^{\rm s}6, -45^{\circ}12^\prime08^{\prime \prime}$) with a $4^{\prime \prime}$accuracy (Walter et al. 2006). These XMM-Newton observations have shown a column density of $N_{\rm H}= 0.77{-}1.2$ $\times $ $10^{23} \mbox{ cm}^{-2}$. This source has recurrent outbursts, making it a fast transient with a peak-flux of $\sim$120 mCrab (20-60 keV) (Sguera et al. 2005, 2006). There is an IR source IRAS 16441-4506 in the error circle.

We performed accurate astrometry of the field (rms of fit = $0{\rlap.}^{\prime \prime}41$), shown in Fig. 2. There is a bright J = 12.9, H = 10.8 and K = 9.8 2MASS source (2MASS J16480656-4512068) $1{\rlap.}^{\prime \prime}1$ away from the centre of the XMM-Newton error circle, which is the closest 2MASS source, suggested by Walter et al. (2006) as being a candidate counterpart. In addition, we find another candidate counterpart inside the error circle, at $2{\rlap.}^{\prime \prime}6$ from the centre of the error circle, labeled 2 in Fig. 2. The NIR magnitudes of both candidate counterparts are given in Table 4. However, on the basis of the astrometry (close to the error circle centre) and the photometry (NIR brightness) we favour Candidate 1 as the candidate counterpart.

NIR spectra of Candidate 1 are shown in Fig. 4. We report the detected lines in Table 8. We find emission of H (Brackett series) between $1.5{-}2.17\mbox{ } \mu \mbox{m}$, and also $He {\sc i}$, $He {\sc ii}$ and $Fe {\sc ii}$ emission lines. These NIR spectra are typical of a supergiant OB star, therefore strengthening Candidate 1 as the likely counterpart of this source, the companion star of IGR J16479-4514 having an OB spectral type. Furthermore, the presence of a Br $\gamma$ emission line and $He {\sc i}$ at $2.11\mbox{ } \mu \mbox{m}$ absorption line constrains the spectral type to a late O or peculiar early B supergiant (Hanson et al. 2005). This source would therefore be a supergiant HMXB system, probably belonging to the SFXT class. Its SED, consistent with the HMXB nature of the system, is shown in Fig. 1.

3.6 IGR J16558-5203

IGR J16558-5203 was discovered at the position (RA Dec J2000.0) = ( $16^{\rm h}55{\rlap.}^{\rm m}8, -52^{\circ}03$) with a 2$^\prime $ uncertainty (Walter et al. 2004a). This source has a ROSAT counterpart (1RXS J165605.6-520345) at the position ( $16^{\rm h}56^{\rm m}05{\rlap.}^{\rm s}60, -52^{\circ}03^\prime45{\rlap.}^{\prime \prime}5$), which allowed them to reduce the uncertainty to 8 $^{\prime \prime }$ (Stephen et al. 2005). Recently, the position has been refined by Swift observations at ( $16^{\rm h}56^{\rm m}05{\rlap.}^{\rm s}73,
-52^{\circ}03^\prime41{\rlap.}^{\prime \prime}18$) with a $3{\rlap.}^{\prime \prime}52$ error circle radius (Malizia et al. 2007).

We performed accurate astrometry of the field, shown in Fig. 2, and find that there is a bright 2MASS object inside the $3{\rlap.}^{\prime \prime}52$ Swift error circle. The NIR magnitudes of this object are given in Table 4. This object is clearly extended on the NIR images, suggesting an extragalactic source. This is in agreement with the result from Masetti et al. (2006), who showed that this object is a Seyfert 1.2 AGN at a redshift of 0.054, exhibiting H$\beta$ and $O {\sc iii}$ emission lines.

3.7 IGR J17091-3624

IGR J17091-3624 was discovered at the position (RA Dec J2000.0) = ( $17^{\rm h}09{\rlap.}^{\rm m}1, -36^{\circ}24^\prime38^{\prime \prime}$), uncertainty $\sim$$3^\prime$(Kuulkers et al. 2003). Follow-up analysis of archival data from the TTM telescope aboard the KVANT module of the MIR orbital station revealed that this source had been detected in several observations performed on Oct. 1-10, 1994, at the position (RA Dec J2000.0) = ( $17^{\rm h}09^{\rm m}06^{\rm s}, -36^{\circ}24^\prime07^{\prime \prime}$), error radius about 0.8$^\prime $. This TTM position is within 0.6$^\prime $ of the INTEGRAL position (Revnivtsev et al. 2003a). It is also coincident with the SAX source 1SAX J1709-36. A variable radio counterpart has been detected in follow-up radio observations at the position ( $17^{\rm h}09^{\rm m}02{\rlap.}^{\rm s}3\pm0.4, -36^{\circ}23^\prime33^{\prime \prime}$), giving an error circle of 5 $^{\prime \prime }$ radius (Rupen et al. 2003). Since this source exhibited some radio emission it has been classified as a galactic X-ray binary, and a candidate microquasar, probably hosting a black hole. It exhibits a large variability in X-rays, and it is uncertain if this source is a Be/X-ray binary or an LMXB. Capitanio et al. (2006) confirm the low absorption ( $N_{\rm H}\leq 1$ $\times $ $10^{22} \mbox{ cm}^{-2}$) and Comptonised spectrum of this source. Negueruela & Schurch (2007) proposed the 2MASS J17090199-3623260 source as a candidate counterpart, based on photometric catalogues. They report that this object is a late F8 V companion star. However, a more accurate position has been obtained by Kennea & Capitanio (2007): (RA Dec J2000.0) = ( $17^{\rm h}09^{\rm m}07{\rlap.}^{\rm s}6,
-36^{\circ}24^\prime24{\rlap.}^{\prime \prime}9$), error radius about $3{\rlap.}^{\prime \prime}6$, which excludes the association of the high energy source with the radio source.

We have performed accurate astrometry of the field, shown in Fig. 2, and we overplot the $3{\rlap.}^{\prime \prime}6$ Swift error circle. We find that the claimed counterpart 2MASS J17090199-3623260 of Negueruela & Schurch (2007) is well outside this error circle; we can therefore reject it. We report the discovery of two blended candidate counterparts inside the error circle, labeled 1 and 2 in Fig. 2 respectively, located at $0{\rlap.}^{\prime \prime}5$ and $0{\rlap.}^{\prime \prime}4$ from the centre of the error circle. We give the NIR magnitudes of these two candidate counterparts in Table 4. We need a more precise position in order to pinpoint the real counterpart of this source, but we favour Candidate 1 as the likely counterpart, on the basis of its NIR brightness.

With only three infrared magnitudes for the two candidate counterparts labeled C1 and C2, the fit of their SEDs is not very accurate, with the model described in Sect. 2.4. In addition, there is no counterpart of this source in the Spitzer GLIMPSE survey (3.6 and $4.5 \mbox{ } \mu \mbox{m}$). We can therefore not firmly conclude the nature of this binary system. However, it is interesting to note that the $R_\ast/D_\ast$ 90% best fit values are very low (=1.15 $\times $ 10-11). From these values we can compute the range of maximal radius that the massive star would have if it was at the maximal distance inside our Galaxy, at $D_\ast=30\mbox{ kpc}$: we find $R_{\rm\ast max}=[6.7{-}12.8]\mbox{ }R_{\odot}$. Since the maximal radius of the companion star would be $\sim$ $13\mbox{ }R_{\odot}$, this star cannot be either a supergiant, or a giant. Alternatively, it could be a main sequence early-type star, but then located very far away in our Galaxy. We therefore conclude that this object is more probably an LMXB system in the Galactic bulge.

3.8 IGR J17252-3616

IGR J17252-3616 was discovered on 2004 February 9 in the Galactic bulge region, at the position ( $17^{\rm h}25{\rlap.}^{\rm m}2,
-36^{\circ}16^\prime$) by Walter et al. (2004a). Observations performed by XMM-Newton on 2004 March 21 by Zurita Heras et al. (2006) localized the source at ( $17^{\rm h}25^{\rm m}11{\rlap.}^{\rm s}4,
-36^{\circ}16^\prime58{\rlap.}^{\prime \prime}6$) with 4 $^{\prime \prime }$ accuracy. This source is a rediscovery of the EXOSAT source EXO 1722-363, based on similar timing and spectral properties. It has been shown to be a heavily absorbed ( $N_{\rm H}\sim1.5$ $\times $ $10^{23} \mbox{ cm}^{-2}$) and persistent source, exhibiting apparent total eclipses, and a hard X-ray spectrum with either an absorbed Compton (kT $\sim$ 5.5 keV and $\tau\sim7.8$) or a flat power law ( $\Gamma\sim0.02$), a very large column density, and Fe K$\alpha$ line at 6.4 keV. The detection of a spin period of $\sim$413.7 s and an orbital period of $\sim$ $9.72~{\rm days}$ securely classified this source as a binary X-ray Pulsar (Zurita Heras et al. 2006). Thompson et al. (2007) refined the orbital period to $P = 9.7403~{\rm days}$ with data collected over more than 7 years. They also derived a mass function of the system of $11.7 \mbox{ }M_{\odot}$, suggesting an HMXB nature. In addition, we point out that putting the spin and orbital periods of this source on a Corbet diagram (Corbet 1986) allows us to suggest a supergiant HMXB nature.

We performed accurate astrometry of the source, and overplot the XMM-Newton error circle, as shown in Fig. 2. There is a bright 2MASS counterpart, 2MASS J17251139-3616575 ($K_{\rm S}$ = 10.7, labelled Candidate 1), inside the error circle, as suggested by Walter et al. (2006). However, in addition to the 2MASS source, there are three fainter candidate counterparts inside the 4 $^{\prime \prime }$ XMM-Newton error circle, labelled 2 to 4. All these candidates are located at 0.7, 1.4, 3.7 and $3{\rlap.}^{\prime \prime}9$ from the centre of the error circle. One of these is blended with the 2MASS counterpart. IGR J17252-3616 would therefore benefit from a more accurate localisation, however based on astrometry and NIR brightness, we favour Candidate 1 as the most likely candidate counterpart. NIR magnitudes of these candidate counterparts are given in Table 4.

....eps}\includegraphics[width=9cm,clip]{IGR_J17391-3021_NIR_id.eps} }
\end{figure} Figure 5: Top panel: combined flux calibrated spectra of the two sources IGR J17252-3616 Candidate 1 ( left) and IGR J18027-2016 ( right) respectively. The optical spectra were obtained with 3.6 m/EFOSC, and the blue and red grism NIR spectra with NTT/SofI. We combined EFOSC2 and SofI spectra by applying a scaling factor of 18.4035 for IGR J17252-3616 and 2.417 for IGR J18027-2016, the difference being due to a different calibration between EFOSC2 and SofI, and to different exposure times. These two factors were obtained by dividing the flux in the overlap region between EFOSC2 and SofI spectra. The spectra are extremely red because of the galactic reddening. Besides the low S/N of the spectra, a number of narrow absorption lines could be detected. The y axis of the spectra are given in `` $\lambda F_\lambda $'' units: $\mbox{ erg~s}^{-1}~\mbox{cm}^{-2}$. Bottom left panel: IGR J17391-3021 flux calibrated optical spectrum, with the y axis given in `` $\lambda F_\lambda $'' units: $\mbox{ erg~s}^{-1}~\mbox{cm}^{-2}$. Bottom right panel: combined blue and red grism NIR spectra of IGR J17391-3021, with the y axis in arbitrary units.
Open with DEXTER

Optical and NIR spectra of the 2MASS counterpart (Candidate 1) are shown in Fig. 5. We report the detected lines in Table 8. We find emission lines of H (Paschen and Brackett), $He {\sc i}$ and $He {\sc ii}$ lines with likely P-Cygni profiles at $2.06\mbox{ } \mu \mbox{m}$, and perhaps $Fe {\sc ii}$. From these NIR spectra, typical of an OB star, we conclude that the companion star of this source has an OB spectral type. Furthermore, the presence of $He {\sc i}$ in absorption and with P-Cygni profile constrains the spectral type to a B supergiant (Hanson et al. 2005). This result is in agreement with the suggestion by Zurita Heras et al. (2006), based on the presence of this bright 2MASS source in the error circle. This source is therefore an HMXB system, probably hosting a supergiant companion star, in agreement with the results by Thompson et al. (2007), and in agreement with its position in the Corbet diagram. The SED, consistent with the HMXB nature of the system, is shown in Fig. 1.

3.9 IGR J17391-3021

IGR J17391-3021 was discovered on 2003 August 26 at the position (RA Dec J2000.0) = ( $17^{\rm h}39{\rlap.}^{\rm m}1, -30^{\circ}21{\rlap.}^{\prime}5$) with a 3$^\prime $ uncertainty by Sunyaev et al. (2003). Chandra observations performed on 2003 October 15 localized the source at ( $17^{\rm h}39^{\rm m}11{\rlap.}^{\rm s}58, -30^{\circ}20^\prime37{\rlap.}^{\prime \prime}6$) with $\sim$ $1^{\prime \prime}$ accuracy (Smith et al. 2006). It is a rediscovery of a previously known ASCA and Rossi-XTE source AX J1739.1-3020 = XTE J1739-302 (Smith 2004). Negueruela et al. (2006a) have suggested that the optical counterpart of IGR J17391-3021 is a USNO A2.0 source (USNO B1.0 0596-058586) with B=17 and R=12.9, and that the NIR counterpart is a bright 2MASS source (2MASS J17391155-3020380 with J=8.600 $\pm$ 0.021, H=7.823 $\pm$ 0.027and $K_{\rm S}$ = 7.428 $\pm$ 0.023; Smith et al. 2006). They also showed that the source is highly reddened, with a variable absorption column density between outbursts. During both low and high level states it exhibits hard X-ray spectra. The fast X-ray transient behaviour (typical of an SFXT) and neutron star spectrum have been confirmed by analyzing archival INTEGRAL data, with flares lasting between 30 min and 3 h (Sguera et al. 2005), the bright ones reaching 330 mCrab (Türler et al. 2007). Negueruela et al. (2006a) state that this source is most likely an HMXB, however the outbursts are shorter than expected for HMXBs or Be/NS binaries, consistent with the SFXTs.

Table 7: Summary of results and characteristics of our studied sample of INTEGRAL sources. We give the name of the sources, the region of the Galaxy in the direction in which they are located, their spin and orbital period, the interstellar column density ($N_{\rm H}$I), the absorption derived from optical to infrared observations ($N_{\rm H}$IR), the absorption derived from X-ray observations ($N_{\rm H}$X), the spectral type (SpT) of their most likely candidate obtained or confirmed by spectroscopy (spec) or by fitting the SED (sed), the nature and type of the binary system, and the reference (Ref) to the spectral type. More details on each source are given in Sect. 3. Details on how we obtained $N_{\rm H}$I, $N_{\rm H}$IR and $N_{\rm H}$X are given in Sect. 2.4. Type abbreviations: AGN = Active Galactic Nucleus, B = Burster (neutron star), BHC = Black Hole Candidate, D = Dipping source, LMXB = Low-Mass X-ray Binary, HMXB = High-Mass X-ray Binary System, OBS = obscured source, SFXT = Supergiant Fast X-ray Transient, T: Transient source, XP: X-ray Pulsar. The classification as SFXT is still subjective, since we lack some accurate observations on a long-term scale for most of the sources. The spectral types come from optical/NIR spectroscopy, reported in the references given in the column Ref. c: this paper, h: Hannikainen et al. (2007), m: Masetti et al. (2006), neg: Negueruela et al. (2006a), nes: Nespoli et al. (2007).

We performed accurate astrometry of the field (rms of fit = $0{\rlap.}^{\prime \prime}47$), and overplot the Chandra error circle, as shown in Fig. 2. Optical and NIR magnitudes of the candidate counterpart are given in Tables 2 and 4 respectively. However the NIR magnitudes are out of the domain of linearity of SofI, which explains the discrepancy with the 2MASS magnitudes. We find that this 2MASS source is a blended source, however from the astrometry, it is clear that the 1 $^{\prime \prime }$ Chandra error circle confirms that the 2MASS source is the counterpart of this source.

We examine the optical and NIR spectra shown in Fig. 5. We report the detected lines in Table 8. In the optical spectrum, we first detect interstellar lines at 5800, 5890-6 (Na$_{\rm I}$ doublet), H$\alpha$, H lines (Paschen), $He {\sc i}$ and $He {\sc ii}$ emission lines, N$_{\rm I}$ lines. The NIR spectrum is very rich in H lines from the Paschen and Brackett series, and also in $He {\sc i}$ and $He {\sc ii}$ emission lines, some exhibiting P-Cygni profiles, and $O {\sc i}$lines. All these lines are characteristic of early-type stars, more precisely of a supergiant O star, in agreement with the O8Iab(f) spectral type derived by Negueruela et al. (2006a). IGR J17391-3021 is therefore a supergiant HMXB, located at a distance of $\sim$ $2.3~{\rm kpc}$ (Negueruela et al. 2006a), and not a Be as stated in Bird et al. (2006). The SED, consistent with the HMXB nature of the system, is shown in Fig. 1.

We can determine the column density along the line of sight, using the equivalent width of the Na$_{\rm I}$ doublet = $9~\AA$; E(B-V) = 0.25 $\times $ $W(\AA) = 2.25 \mbox{ mag}$ (using Munari & Zwitter 1997); $N(H {\sc i}+\rm H2) = 5.8$ $\times $ 1021 $\times $ E(B-V) = 1.3 $\times $ $10^{22}~{\rm atoms/cm}^2$ (using Bohlin et al. 1978), the column density is consistent with the one indicated in Table 7.

3.10 IGR J17597-2201

IGR J17597-2201 was discovered at the position (RA Dec J2000.0) = ( $17^{\rm h}59{\rlap.}^{\rm m}7, -22^{\circ}01^\prime$) with $\sim$$2^\prime$ uncertainty (Lutovinov et al. 2003). XMM-Newton observations localized the source at ( $17^{\rm h}59^{\rm m}45{\rlap.}^{\rm s}7, -22^{\circ}01^\prime39^{\prime \prime}$) with a 4 $^{\prime \prime }$ accuracy (Walter et al. 2006). It exhibits an absorption of $N_{\rm H}= 4.5$ $\pm$ 0.7 $\times $ $10^{22} \mbox{ cm}^{-2}$, or 2.70 $\pm$ 0.15 $\times $ $10^{22} \mbox{ cm}^{-2}$ when a partial-covering absorber is included. This source, associated with the Rossi-XTE source XTE J1759-220, is a Low-Mass X-ray Binary system, a late-type transient type I X-ray burster, therefore hosting a neutron star, with a dipping behavior of $\sim$$30\%$ with $\sim$5 min dip duration (Markwardt & Swank 2003).

We performed accurate astrometry of the field, shown in Fig. 2. Walter et al. (2006) proposed a 2MASS counterpart for this source (2MASS J17594556-2201435), however it is well outside the error circle, towards the south-west, at 5 ${\rlap.}^{\prime \prime}$1 from the centre of the error circle. There are one bright and two faint candidate counterparts which are well inside the XMM-Newton error circle (labeled 1, 3 and 5 and located respectively at 2.25, 0.95 and $2{\rlap.}^{\prime \prime}05$ from the centre of the XMM-Newton error circle). In addition, there are one faint and one bright candidate counterparts, lying on the error circle (labeled 2 and 4, and located respectively at 3.8 and $3{\rlap.}^{\prime \prime}9$ from the centre of XMM-Newton error circle). We give the NIR magnitudes of these candidate counterparts in Table 4. Since the previously proposed counterpart is outside the error circle, and since we propose new candidate counterparts, this source deserves further observations to find which one is the right counterpart.

By fitting the SED of the brightest candidate labeled 1, which is also the closest to the error circle centre, with the model described in Sect. 2.4, we obtain a stellar temperature $T_\ast=31~700$ K, consistent with an O type star. The other parameters are given in Table 6, and the SED is shown in Fig. 1. However our spectral fit is not accurate enough to unambiguously constrain the spectral type of the companion star, because we have only the three NIR magnitudes (we did not detect any MIR counterpart in the Spitzer GLIMPSE survey). Therefore, taking into account the errors, this SED fitting is consistent at 90% both with an LMXB and HMXB system. Assuming the source is a supergiant, with a minimum radius of $20 \mbox{ }R_{\odot}$, and taking a mean $R_\ast/D_\ast$ of 2 $\times $ 10-11, we obtain a distance of $35.2\mbox{ kpc}$, which is not plausible. Alternatively, taking a radius $R_{\ast} = 10 \mbox{ }R_{\odot}$ typical of a B V companion star, we derive a distance of $D_{\ast}=15.2$ kpc, a more reasonable estimate although still high. We therefore conclude that the source is most likely an LMXB, consistent with its X-ray properties, given the detection of type I X-ray bursts.

3.11 IGR J18027-2016

IGR J18027-2016 was discovered at the position (RA Dec J2000.0) = ( $18^{\rm h}02^{\rm m}46^{\rm s}, -20^{\circ}16{\rlap.}^{\prime}3$) (Revnivtsev et al. 2004). XMM-Newton observations allowed to localize the source at ( $18^{\rm h}02^{\rm m}42{\rlap.}^{\rm s}0,
-20^{\circ}17^\prime18^{\prime \prime}$) with 4 $^{\prime \prime }$ accuracy (Walter et al. 2006). The source was originally called IGR J18029-2016, and it is associated with the SAX source SAX J18027-2017. The absorption is $N_{\rm H}= 9.1$ $\pm$ 0.5 $\times $ $10^{22} \mbox{ cm}^{-2}$(Walter et al. 2006). It is an X-ray pulsar, and a transient source, an eclipsing HMXB, likely at a distance of 10 kpc. An orbital period of 4.57 days and a pulse period of 139.47 s were reported by Hill et al. (2005). These authors have derived the system parameters, with an excentricity of $e \leq 0.2$, a mass function of $f(M) \sim17$ $\pm$ $5\mbox{ }M_{\odot}$, implying a mass of the companion of $18.8{-}29.3 \mbox{ }M_{\odot}$.

We performed accurate astrometry of this source (rms of fit = $0{\rlap.}^{\prime \prime}51$), showed in Fig. 2. There is a 2MASS counterpart (2MASS J18024194-2017172) at $\sim$ $1^{\prime \prime}$ from the centre of the XMM-Newton error circle, which is the candidate counterpart proposed by Walter et al. (2006). However there is another bright source well inside the error circle, labeled Candidate 2. There is also at least one faint counterpart inside the error circle. We give the NIR magnitudes of Candidate 1 and Candidate 2 in Table 4. Based on the proximity to the error circle centre, and its NIR brightness, we favour Candidate 1 as the likely candidate.

The optical and NIR spectra of Candidate 1 are shown in Fig. 5. We report the detected lines in Table 8. In the optical spectrum we detect hydrogen ($\alpha$ to $\zeta$) and $He {\sc ii}$ emission lines, and in the NIR spectrum we detect H (Paschen and Brackett series), $He {\sc i}$ and $He {\sc ii}$ emission lines, some tentatively exhibiting P-Cygni profiles. These NIR spectra are typical of a supergiant OB star, which is therefore likely the spectral type of the companion star: IGR J18027-2016 is therefore a supergiant HMXB system. Fitting the data with the model described in Sect. 2.4, we obtain a stellar temperature $T_\ast = 20~800~{\rm K}$, which is typical of a B supergiant star. The other parameters are given in Table 6, and the SED is shown in Fig. 1. By taking the $R_\ast/D_\ast$minimizing $\chi ^2$ (3.7 $\times $ 10-11), and assuming a typical radius of a B supergiant star, i.e. $R_\ast=20\mbox{ }R_{\odot}$, we derive a distance of $D_\ast = 11.9~{\rm kpc}$. We therefore favour the supergiant nature of the companion star, with a B spectral type, in agreement with the results derived from the spectra.

3.12 IGR J18483-0311

IGR J18483-0311 was discovered at the position (RA Dec J2000.0) = ( $18^{\rm h}48{\rlap.}^{\rm m}3, -03^{\circ}11^\prime$), with uncertainty $\sim$$2^\prime$ (Chernyakova et al. 2003). The Swift observations allowed Sguera et al. (2007) to refine the position of the source to ( $18^{\rm h}48^{\rm m}17{\rlap.}^{\rm s}17, -03^{\circ}10^\prime15{\rlap.}^{\prime \prime}54)$, uncertainty $3{\rlap.}^{\prime \prime}3$. The Swift refined position allowed them to identify an optical counterpart from the USNO-B1.0 and 2MASS catalogue located at ( $18^{\rm h}48^{\rm m}17{\rlap.}^{\rm s}2, -03^{\circ}10^\prime16{\rlap.}^{\prime \prime}5$) with magnitudes R=19.26, I=15.32, J=10.74, H=9.29 and K=8.46. The source exhibits a column density of $N_{\rm H}= 9$ $\times $ $10^{22} \mbox{ cm}^{-2}$, a $\Gamma=1.4$ and an energy cutoff at $E_{\rm cut}=22$ keV (Sguera et al. 2007). Timing analysis of Rossi-XTE/ASM light curve have allowed investigators to derive an orbital period of 18.55 $\pm$ 0.05 days (Levine & Corbet 2006). Sguera et al. (2007) also report that the source contains a pulsar with a spin period of 21.0526 $\pm$ 0.0005 s with a pulse fraction of 65 $\pm$ 10%, while the highly reddened optical counterpart suggests by analogy with other such systems that the source is an HMXB. They further argue that it is a likely Be system due to its localisation in a Corbet diagram (Corbet 1986), but they cannot rule out an SFXT, although the typical ratio between maximum and minimum luminosity they observed is at least 10 times lower than that of typical SFXT.

We performed accurate astrometry of this source (rms of fit = $0{\rlap.}^{\prime \prime}67$), shown Fig. 2. There is a bright 2MASS counterpart 2MASSJ 18481720-0310168 inside the Swift error circle, which is the candidate counterpart proposed by Sguera et al. (2007). However this source seems to be blended with another fainter source, at 2.5 $^{\prime \prime }$ from the centre of the error circle. We give the NIR magnitudes of the bright candidate counterpart in Table 4.

Fitting the SED of this counterpart with the model described in Sect. 2.4 allows us to derive a stellar temperature of $T_\ast = 22~500~{\rm K}$ at the limit of 90% of $\chi ^2$, which is typical of a B star. The other parameters are given in Table 6, and the SED is shown in Fig. 1. The stellar temperature is consistent with a B spectral type companion star. Since the $R_\ast/D_\ast$ ratio minimizing $\chi ^2$ is 2.15 $\times $ 10-10, the distance of this source would then be 0.9 kpc if the companion star is a main sequence star (with a typical stellar radius of $R_\ast=3\mbox{ }R_{\odot}$), 1.5 kpc for a sub-giant and 2.7 kpc for a supergiant star (with a typical stellar radius of $R_\ast=20\mbox{ }R_{\odot}$). This source exhibits a strong NIR excess, which might indicate the presence of a disk/wind such as in massive stars. Furthermore, its position in the Corbet diagram is between Be and wind accretor-supergiant X-ray binary systems. Although we cannot firmly conclude on the spectral type and class, this SED shows that this source is an HMXB system.

3.13 IGR J19140+0951

IGR J19140+0951 was discovered on 2003 March 6-7 at the position (RA Dec J2000.0) = ( $19^{\rm h}13^{\rm m}55^{\rm s}, +9^{\circ}51{\rlap.}^{\prime}6$), uncertainty 1$^\prime $) by Hannikainen et al. (2003). Chandra observations performed on 2004 May 11 localized the source at ( $19^{\rm h}14^{\rm m}4{\rlap.}^{\rm s}232, +09^{\circ}52^\prime58{\rlap.}^{\prime \prime}29$) with $0{\rlap.}^{\prime \prime}6$ accuracy (in 't Zand et al. 2006). Fitting the hard X-ray spectrum from Rossi-XTE observations allowed investigators to derive a $\Gamma\sim1.6$ and $N_{\rm H}\sim6$ $\times $ $10^{22} \mbox{ cm}^{-2}$(Swank & Markwardt 2003) with variations of $N_{\rm H}$ up to $\sim$ $10^{23} \mbox{ cm}^{-2}$ (Rodriguez et al. 2005). An orbital period of $13.55~{\rm days}$ was found from timing analysis of Rossi-XTE data, with an X-ray activity detected with Rossi-XTE/ASM as early as 1996 (Corbet et al. 2004), confirming the binary nature of the source. Rodriguez et al. (2005), after a comprehensive analysis of INTEGRAL and Rossi-XTE data, showed that the source was spending most of its time in a faint state but reported high variations of luminosity and absorption column density. It is a persistent HMXB with evidence for the compact object being a neutron star rather than a black hole, exhibiting a variable absorption column density, and a bright iron line (Rodriguez et al. 2005). This source has other names: IGR J19140+098 (Hannikainen et al. 2003), and EXO 1912+097. in 't Zand et al. (2006) proposed the 2MASS source 2MASS J19140422+0952577 as the NIR counterpart of this source (J= 8.55, H= 7.67 and $K_{\rm S}$ =  $7.06\mbox{ mag}$).

We performed accurate astrometry of the field (rms of fit = $0{\rlap.}^{\prime \prime}35$), shown in Fig. 2. The accurate error circle allows us to confirm the 2MASS source 2MASS 19140422+0952577 as the candidate counterpart, located at $4{\rlap.}^{\prime \prime}6$ North of a bright 2MASS source (2MASS J19140417+0952538, $K_{\rm S}$ = $6.27\mbox{ mag}$). We give the NIR magnitudes of the deblended candidate counterpart in Table 4. These deblended magnitudes we derive are different from the magnitudes given in the 2MASS catalogue, because both 2MASS sources are spatially very close in the 2MASS data.

We show the NIR spectra of this source in Fig. 4. We report the detected lines in Table 8. The spectrum is dominated by H (Paschen and Brackett series), and $He {\sc i}$ and $He {\sc ii}$ in emission. The NIR spectra are typical of an OB spectral type companion star, and the narrowness of the lines suggests a supergiant type, which is consistent with the results by Nespoli et al. (2007) for a B1I stellar type companion, derived from spectra obtained at ESO/NTT/SofI. This classification has been refined to B0.5I based on spectra obtained at UKIRT (Hannikainen et al. 2007), making IGR J19140+0951 a supergiant X-ray binary. Both classifications confirm the supergiant HMXB nature of IGR J19140+0951, hosting a neutron star. The SED, consistent with the HMXB nature of the system, is shown in Fig. 1.

in 't Zand et al. (2006) also suggest a MIR counterpart at 8.3 $\mu$m found in the Midcourse Space Experiment (MSX, Mill 1994), however Rahoui et al. (2008) show that in 't Zand et al. (2006) are in fact reporting the summed flux of a blended source, composed of the MIR counterparts of both IGR J19140+0951 and the bright 2MASS source at the South, which is unrelated to IGR J19140+0951. Rahoui et al. (2008) give the MIR fluxes of the counterpart of IGR J19140+0951.

4 Discussion

We begin this Section by giving a summary of the results of all individual sources. We continue by showing the large scale environment of these sources, and report the presence of absorption in their environment. We then recall the general characteristics of HMXBs, before discussing the results obtained, in the context of the INTEGRAL era.

4.1 Summary of results of the sample of studied sources

We found by spectroscopy of the most likely candidate counterparts the spectral types of IGR J16320-4751, IGR J16358-4726, IGR J16479-4514, IGR J17252-3616 and IGR J18027-2016: they are all of supergiant OB types, IGR J16358-4726 likely hosting a sgB[e] companion star. We also confirm the supergiant O nature of IGR J17391-3021, and the supergiant B nature of IGR J19140+0951. By fitting the SED of the most likely candidates we found that IGR J16418-4532 is an HMXB, probably hosting a supergiant OB spectral type companion star, that IGR J16393-4643 is an HMXB system probably hosting a BIV-V companion star, and that IGR J18483-0311 is very likely an HMXB system. By accurate astrometry we rejected the counterpart proposed for IGR J17091-3624 and IGR J17597-2201, we propose two new candidate counterparts for each of them, which by SED fitting we found consistent with an LMXB nature. We confirm the AGN nature of IGR J16558-5203.

We summarize in Table 7 the results obtained both from spectroscopy and SED fitting to the sample of studied INTEGRAL sources. We give the results of the column density in optical/IR, spectral type of the companion stars, and type of sources for all individual sources of our sample. We also give all the parameters known about these sources, such as the spin and orbital period, and column density derived from X-ray observations, in order to facilitate the following discussion. The interstellar column density and absorption in the optical and NIR domain have been derived from our observations, given in Table 6 for the sources IGR J16393-4643, IGR J16418-4532, IGR J17091-3624, IGR J17597-2201, IGR J18027-2016 and IGR J18483-0535, and in Rahoui et al. (2008) for the remaining sources. We then converted the column density into absorption in magnitudes, using the relation given in Cardelli et al. (1989): $N_{\rm H}/A_{\rm v}= 1.87$ $\times $  $10^{21} \mbox{ cm}^{-2}/\mbox{ mag}$.

The results, concerning the spectral type of the companion star, given by these SED fittings are in agreement with those directly derived from spectroscopy, when available. On the other hand, they allow us to derive the likely spectral type of the companion star, when there is no spectroscopic information. We can therefore conclude that most of these obscured sources host luminous, massive, hot and early-type companion stars, i.e. of OB spectral type, most of them being evolved stars of the supergiant spectral class.

From these SEDs, all reported in Fig. 1, we can directly compare the optical to MIR with the hard X-ray domain integrated flux (corresponding to the energy output). All sources for which both fluxes are comparable are supergiant HMXBs: IGR J16418-4532, IGR J16479-4514, IGR J17252-3616, IGR J17391-3021, IGR J18027-2016, IGR J19140+0951. These comparable optical-MIR and hard X-ray fluxes are expected from the HMXB nature of the systems. Most of the sources for which the hard X-ray flux is much higher than the optical-MIR flux are HMXBs, hosting either a supergiant or B type companion star: IGR J16320-4751, IGR J16358-4726, IGR J16393-4643. IGR J17091-3624 and IGR J17597-2201 are the only cases with a hard X-ray flux higher than the optical-MIR flux, consistent with their LMXB nature. Finally, there is only one case where the hard X-ray flux is much less than the optical-MIR flux: the HMXB IGR J18483-0311. However we have to be cautious because the multi-wavelength observations of these SEDs were not taken simultaneously, and the hard X-ray fluxes are variable.

\par\mbox{\subfigure[IGR~J16358-4726]{ \includegraphics[width=6.8...
...17391-3021]{\includegraphics[width=6.8cm,clip]{I17391-J-rbc.eps} }}
\end{figure} Figure 6: Large fields of 4 studied sources exhibiting high absorption along their line of sight. Size: $5.5^\prime $ $\times $ $5.5^\prime $; North is to the top and East to the left. We overplot a white circle to indicate the position of the INTEGRAL sources.
Open with DEXTER

4.2 Environment

Four sources of the studied sample exhibit large-scale regions of absorption in the NIR images, very close to the line of sight of the hard X-ray source: IGR J16358-4726, IGR J16418-4532, IGR J16479-4514 and IGR J17391-3021. We show in Fig. 6 the large field J band image of these sources (except for IGR J16479-4514 for which we show the large field $K_{\rm S}$ band image). This absorption region might be due to the presence of extended molecular clouds and/or $\mbox{H~{\sc ii}}$ regions. The presence of such highly absorbed regions is not surprising close to these peculiar hard X-ray sources, since they must be strongly linked to their formation. For instance, the Norma arm region is one of the richest star-forming regions of our Galaxy, where many high-mass stars form and evolve (Bronfman et al. 1996). There exists therefore a high probability for binary systems made up of high-mass stars to form, and evolve into HMXBs (see for instance Tauris & van den Heuvel 2006). This might explain why so many binary systems with high-mass companion stars and progenitors of compact objects such as black holes or neutron stars have formed in this region of our Galaxy.

Bodaghee et al. (2007) have compared the spatial distribution of HMXBs discovered by INTEGRAL - for which the distance is known - and of star-forming complexes, mainly OB regions reported by Russeil (2003). They have shown that their spatial distribution is similar, suggesting that HMXBs are associated with these complexes. The discovery of large-scale absorption regions in the direction of these sources is therefore not surprising, since the formation sites of HMXBs are closely linked to rich star forming regions. Indeed, the short life of HMXBs prevents these systems from migrating far away from their birthplace. Characterisation of these large-scale absorption regions, and measurement of the metallicity of stars hosted by these regions are required to clarify this situation.

4.3 HMXBs in the INTEGRAL era

HMXBs are separated in two distinct groups. The first group contains the majority of the HMXB systems, constituted of known or suspected Be/X-ray Binary systems (BeXBs), called Be/X-ray transients. In Be systems, the donor is a Be star and the compact object is a neutron star typically in a wide, moderately eccentric orbit, spending little time in close proximity to the dense circumstellar disk surrounding the Be companion (Coe 2000; Negueruela 2004). X-ray outbursts occur when the compact object passes through the Be-star disk, accreting from the low-velocity and high-density wind around Be stars, and exhibiting hard X-ray spectra.

The second group of HMXB systems contains the Supergiant/X-ray Binaries (SXBs), where the compact object orbits deep inside the highly supersonic wind of a supergiant early-type star, which plays the role of the donor star (Kaper et al. 2004). The X-ray luminosity is powered either by accretion from the strong stellar wind of the optical companion, or by Roche-lobe overflow. In a wind-fed system, accretion from the stellar wind results in a persistent X-ray luminosity of 1035-36 erg/s, while in a Roche-lobe overflow system, matter flows via the inner Lagrangian point to form an accretion disc. In this case, a much higher X-ray luminosity ($\sim$1038 erg/s) is then produced during the outbursts.

In the pre-INTEGRAL era, known HMXBs were mostly BeXBs systems. For instance, in the catalogue of HMXBs of Liu et al. (2000), there were 54 BeXBs and 7 SXBs identified, out of 130 HMXBs, representing a proportion of 42% and 5% respectively. Then, between the two last editions of HMXB catalogues (Liu et al. 2000, 2006), the proportion of SXBs compared to BeXBs has increased, with the first HMXBs identified in the INTEGRAL data. The third IBIS/ISGRI soft $\gamma$-ray survey catalogue (Bird et al. 2007), spanning nearly 3.5 years of operations, contains 421 sources detected with the INTEGRAL observatory, of which 214 ($\sim$50%) were discovered by this satellite. This catalogue, extending up to 100 keV, includes 118 AGNs, 147 X-ray binaries (79 LMXBs and 68 HMXBs), 23 Cataclysmic Variables, 23 other objects, and 115 still unidentified objects. Among the 68 HMXBs, 24 have been identified as BeXBs and 19 as SXBs, representing a proportion of 35% and 28% respectively. The proportion of BeXBs, relative to the total of HMXBs, has decreased by a factor 1.2 while the proportion of SXBs has increased by a factor 5.6 between the catalogue of Liu et al. (2000) and the one of Bird et al. (2007). Related to this increasing proportion of SXBs, the other highlight of the INTEGRAL catalogue is the emergence of the SFXT class, with 12 INTEGRAL sources being firm or possible candidates.

Our studied sample of INTEGRAL sources allows us to add four newly identified SXBs which were not classified as such in Bird et al. (2007): IGR J16320-4751, IGR J16358-4726, IGR J17252-3616 and IGR J18027-2016. The other sources that we have identified in this paper as supergiants were already considered SXBs, based either on spectral classification - IGR J17391-3021 and IGR J19140+0951 - or on X-ray properties - the SFXT candidates IGR J16418-4532 and IGR J16479-4514. With these new SXBs, the proportion of SXBs has reached that of BeXBs: 35% of all HMXBs for each population. This clearly shows that the launch of INTEGRAL has drastically changed the statistical situation concerning the nature of HMXBs, by revealing a new dominant population of supergiant X-ray binaries, which are purely wind accretor systems.

Although this came as a surprise, it is a posteriori consistent with the fact that these hard X-ray emitters are sources ideally detected by INTEGRAL, as discussed in Lutovinov et al. (2005a), Dean et al. (2005) and Bodaghee et al. (2007). INTEGRAL, observing at energies higher than the threshold above which photoelectric absorption becomes negligible in most matter, can easily detect bright sources above a few tens of keV, while they are not detectable below, and therefore had remained hidden up to now. HMXBs accreting by stellar wind create a naturally dense and highly absorbing circumstellar wind compared to Roche lobe overflow in LMXBs, hiding the X-ray emission in a similar way to Seyfert 2 AGNs (Dean et al. 2005; Malizia et al. 2003).

4.4 The nature of INTEGRAL HMXBs

Let us now consider our results, summarized and put together with other characteristics of these sources in Table 7. In this sample of 13 INTEGRAL sources, concentrated both towards the direction of the Norma arm and the Galactic centre, we classified 10 HMXBs (8 sources hosting sgOB and 2 sources hosting BIV-V companion stars), 2 LMXBs and 1 AGN. We clearly confirm the predominance of HMXBs hosting supergiants, as opposed to those hosting Be companion stars: in our sample, 80% of HMXBs host compact objects (probably neutron stars) orbiting around OB supergiant secondaries. As already discussed, this result is in agreement with the increase of SXBs in the HMXBs population. However, quantitatively, the proportion of SXBs in our studied sample is higher than in the Bird et al. (2007) catalogue. There are two reasons: first, our sample is much more limited; second, most of the sources of our sample are in the direction of the Norma arm, which is associated with rich star-forming regions, the natural birthplace of massive stars, as discussed in Sect. 4.2. Most of these new INTEGRAL sources are wind accretors, consistent with their location in the Corbet diagram (Corbet 1986): nearly all of the HMXBs discovered by INTEGRAL, for which both spin and orbital periods have been measured, are located in the upper part of this diagram, among other wind accretors, typical of supergiant HMXBs. These systems exhibit extra absorption by a factor of $\sim$4 compared to the average of all HMXBs, and they are X-ray pulsars, with longer pulsation periods than previously known HMXBs (Bodaghee et al. 2007).

In the companion paper by Rahoui et al. (2008) it is shown that only 3 sources - IGR J16318-4848, IGR J16358-4726 and IGR J16195-4945 - of a sample of 12 INTEGRAL sources exhibit a MIR excess, due to an absorbing component enshrouding the whole binary system. The MIR emission of the remaining sources comes from the supergiant companion star, showing that absorbing material is enshrouding the compact object for most of the sources (preliminary results of this paper were described in Chaty & Rahoui 2006). We are therefore observing two classes of INTEGRAL sources, i.e. highly absorbed sources such as IGR J16358-4726, for which the extreme representant is IGR J16318-4848, and SFXTs such as IGR J17391-3021, for which the archetype is IGR J17544-2619.

As shown in Table 7, these classes share similar properties - for instance they both host supergiant companion stars - however they do not seem to have exactly the same configuration, and one way to explain their different characteristics can be found in their excess in absorption, which does not seem to have the same origin. It is caused by two different phenomena in the case of the highly absorbed sources: the observations from hard X-ray to MIR domains suggest the presence of absorbing material concentrated around the compact object, and also some dust and/or cold gas, perhaps forming a cocoon, enshrouding the whole binary system (Filliatre & Chaty 2004; Rahoui et al. 2008). Their characteristics might be explained by the presence of a compact object (neutron star or black hole) orbiting within the dense wind surrounding the companion star. On the other hand, in the case of SFXTs, characterised by fast X-ray outbursts, the presence of the absorbing material seems concentrated around the compact object only, and MIR observations show that there is no need for any other absorbing material around the whole system (Rahoui et al. 2008). Their characteristics might be explained by the presence of a compact object (neutron star or black hole) located on an eccentric orbit around the companion star, and it would then be when the compact object penetrates the clumpy circumstellar envelope that outbursts are caused. However the situation is more complicated, since some SFXT sources are also highly obscured, and the intrinsic absorption derived from X-ray observations vary for some INTEGRAL sources (see e.g. the case of IGR J19140+0951; Prat et al. 2008).

The X-ray transient or persistent nature of these sources might therefore be related to the geometry of these systems, and it is possible to distinguish the nature of both classes by speculating on the geometry of these systems (see also Chaty & Rahoui 2006). However only the knowledge of their orbital periods, which is currently unknown for most of these sources, will allow us to confirm or not this view. Many questions are still open, most of them related to the interaction between the compact object and the supergiant star, and to the nature of the absorbing material, which seems to be present in the environment either of the compact object, or of the whole binary system, and even both. To better understand the general properties of these systems, we will need to better characterise the dust, its temperature, composition, geometry, extension around the system, and to investigate where this dust or cold gas comes from. Finally, probably the most important question to solve is whether this unusual circumstellar environment is due to stellar evolution OR to the binary system itself. To answer this question, this dominant population of supergiant HMXBs now needs to be taken into account in population synthesis models.

5 Conclusions

We have performed an extensive study of a sample of 13 INTEGRAL sources, through optical and NIR photometry and spectroscopy, performing for each source accurate astrometry, identifying candidate counterparts, deriving their optical and NIR magnitudes, and obtaining spectra for most of them. In addition, we built and fitted the SED of these sources, from MIR to hard X-rays. We thus identified the nature of the companion stars and of the binary systems by spectroscopy for 7 of these sources, and by fitting their SED for 5 of them. We finally reported NIR fields of four of these sources, which exhibit large-scale regions of absorption, probably linked to their formation process. We then discussed the existence of this dominant population of supergiant HMXBs in our Galaxy, born with two very massive components: a population which only recently has been revealed by the INTEGRAL observatory.

Thus, it clearly appears that a study of this new population of supergiant HMXBs, constraining the nature and orbital parameters of these systems, and linking them to population synthesis models, will provide a better understanding of the evolution of HMXBs. These systems are probably the primary progenitors of neutron star/neutron star or neutron star/black hole mergers. There is therefore the possibility that they are related to short/hard $\gamma$-ray bursts, and also that they might be good candidates for gravitational wave emitters. Because most of these sources are obscured, the ``Norma arm''-like sources can only be studied in the hard X-ray and infrared domains. A joint study of these sources with multi-wavelength hard X-ray, optical, NIR, MIR (and radio) observations is therefore necessary.

Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere, Chile (ESO Programme 073.D-0339). We acknowledge Jorge Melnick for special DDT time at 3.6 telescope on EFOSC2. S.C. thanks the ESO staff and especially Valenti Ivanov and Emanuela Pompei for their invaluable assistance during the run when we performed these optical and NIR observations. J.A.T. acknowledges partial support from Chandra award number GO5-6037X issued by the Chandra X-Ray Observatory Center, which is operated by the Smithsonian Astrophysical Observatory for and on behalf of the National Aeronautics and Space Administration (NASA), under contract NAS8-03060. This research has made use of NASA's Astrophysics Data System, of the SIMBAD database, operated at CDS, Strasbourg, France, and of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.



Copyright ESO 2008