Optical spectra of five new Be/X-ray binaries in the Small Magellanic Cloud and the link of the supergiant B[e] star LHA 115-S 18 with an X-ray source

Abstract

INTRODUCTION

High-mass X-ray binaries (HMXBs) are stellar systems consisting of a massive, early-type star (of O or B spectral type) and a compact object (neutron star or black hole). The material lost by the companion star (the donor) either through strong stellar winds or a circumstellar disc is accreted on to the compact object resulting in the formation of supergiant X-ray Binaries (sgXRBs) and Be/X-ray Binaries (BeXRBs), respectively. Predominantly the compact objects in these systems are pulsars with spin periods in the 1–1000 s range (e.g. Knigge, Coe & Podsiadlowski 2011). Depending on the available material and the geometry of the orbit of the systems, their X-ray emission can be either persistent or variable in time-scales of days up to several months. Their typical luminosity ranges between ∼10³⁴ (for low-activity systems) up to 10³⁸ erg s⁻¹ (for outbursting systems).

In the BeXRBs, the most numerous subclass of HMXBs (Liu, van Paradijs & van den Heuvel 2005, 2006), the donor is a non-supergiant B star (luminosity class III-V) whose spectrum shows or, has at some time in the past shown, Balmer lines in emission (the so-called ‘Be phenomenon’; e.g. Porter & Rivinius 2003). This emission is produced by an equatorial disc of ionized material that has been expelled from the star due to its high (close to the critical limit) rotational velocity. Subsequently, part of this material is accreted on the compact object. Most BeXRBs are transient systems (e.g. Reig & Roche 1999) which can produce outbursts with luminosities in the range of 10³⁶–10³⁷ erg s⁻¹ (type I outbursts, which occur at periastron and last a few days) or even stronger with luminosities ≳10³⁷ erg s⁻¹ (type II outbursts, which are more rare and occur at irregular intervals). However, there are also persistent sources which display lower luminosity levels (∼10³⁴–10³⁵ erg s⁻¹; for a review see Reig 2011).

In the case of sgXRBs (Charles & Coe 2006; Liu et al. 2006, and references therein), the donor is a supergiant O or B-type star (luminosity class I-II). Depending on the mass-transfer mechanism these systems are divided in Roche lobe overflow (RLOF; e.g. Lamers, van den Heuvel & Petterson 1976) and wind-fed (e.g. Lutovinov et al. 2013). For systems in the first subclass, the steady mass-transfer rate through the Roche lobe is high enough to lead to the formation of an accretion disc around the compact object, resulting in persistent systems with luminosities up to ∼10³⁸ erg s⁻¹. In wind-fed systems, the donor star loses mass through a strong radial stellar wind (with mass-loss rates between 10⁻⁸–10⁻⁶ M_⊙ yr⁻¹). As the compact object lies in a close orbit around the donor, it becomes a persistent X-ray source with much lower luminosity (in the range ∼10³⁵–10³⁶ erg s⁻¹). These systems are referred as ‘classical’. The advent of INTEGRAL has unveiled new populations of wind-fed systems: the supergiant Fast X-ray Transients (SFXTs; Negueruela et al. 2006) and the heavily obscured sgXRBs (Walter et al. 2006). The SFXTs display flaring activity which lasts from few minutes to several hours with a luminosity increase from ∼10³³–10³⁴ erg s⁻¹ to ∼10³⁶–10³⁷ erg s⁻¹. On the other hand, the heavily obscured sgXRBs are actually similar to the ‘classical’ wind-fed sgXRBs but the compact object is deeply embedded in a dense absorbing environment. Their H i column density can be as high as N_H ∼ 10²⁴ cm⁻² thus suppressing significantly their observed X-ray luminosities. For comparison, the measured absorbing column density for the ‘classical’ systems is of the order of ∼10²¹–10²² cm⁻² (for a review see Kaper, van der Meer & Tijani 2004; Chaty 2011).

The Small Magellanic Cloud (SMC) is an excellent laboratory to study the HMXBs, since it is nearby (D = 60 kpc; Hilditch, Howarth & Harries 2005) and well covered by the Chandra and XMM–Newton X-ray observatories that can detect sources down to L_X ∼ 10³³ erg s⁻¹ (i.e. reaching luminosities of non-outbursting sources). Moreover, it does not suffer from large extinction and distance uncertainties that often hamper studies of HMXBs in the Milky Way (MW). It also has a relatively uniform metallicity among the young populations, and a well-determined star formation history (Harris & Zaritsky 2004). Most importantly it is host to a large number (∼90) of HMXBs (Haberl & Pietsch 2004; Coe et al. 2005; Antoniou et al. 2009a,b, 2010). Out of these systems only one is a sgXRB, source SMC X-1 (Webster et al. 1972), which is the only persistent accreting X-ray pulsar (P_spin ∼ 0.71 s; Lucke et al. 1976) in the SMC fed through RLOF. This system has a B0 supergiant companion (Webster et al. 1972) with an orbital period of 3.89 d (Tuohy & Rapley 1975) and an X-ray luminosity of ∼9.5 × 10³⁷ erg s⁻¹ in the 0.2–12.0 keV energy band (e.g. XMMSL1; Saxton et al. 2008). In contrast, in the MW the number of confirmed or suspected supergiant systems is much higher (∼32 per cent of the total number of HMXBs; Liu et al. 2006; Chaty 2011).

Although the number of known HMXBs in the SMC has increased dramatically in the last decade, only recently we started having a picture of the spectral classification of their donor stars (e.g. McBride et al. 2008; Antoniou et al. 2009b). This is important since it can yield valuable information on the evolution of massive binary stellar systems. Following our previous work (Antoniou et al. 2009b), we used the multiple-object mode of the AAOmega spectrograph, a fibre-fed optical spectrograph on the 3.9 m Anglo-Australian Telescope (AAT), to obtain optical spectra of confirmed and candidate HMXBs, in order to identify BeXRBs and determine their spectral types.

In this paper, we present the results of this spectroscopic campaign. Its structure is the following. In Section 2, we describe the sample of sources and in Section 3 we discuss the observations and the data reduction. In Section 4, we present the selection criteria of candidate BeXRBs. In Section 5, the spectral classification of the BeXRBs and the comparison of their spectral types with previous results are discussed. In Section 6, we discuss the properties of the overall population of BeXRBs in the SMC and we compare them with the BeXRB population in the MW. We also discuss the nature of the X-ray source CXOU J005409.57−724143.5 that is associated with the supergiant B[e] star LHA 115-S 18 (hereafter S 18, Henize 1956; also known as AzV 154; Azzopardi, Vigneau & Macquet 1975). A summary of the main results of this study is given in Section 7.

SAMPLE

The sample used in this work is derived from studies of X-ray sources detected with the Chandra and XMM–Newton X-ray observatories. As our basic sample, we use the catalogue of HMXB candidates detected in the Chandra shallow survey of the SMC (Antoniou et al. 2009a), which were identified based on the location of their optical counterparts in the (V, B − V) colour–magnitude diagram (CMD). The chance-coincidence probability for a Chandra X-ray source to be associated with an OB star is estimated to be ∼20 per cent (Antoniou et al. 2009a). This approach allowed us to identify candidate HMXBs even when we could not detect X-ray pulsations in the X-ray data, a tell-tale signature of BeXRB pulsars. It also allowed us to identify objects of lower X-ray luminosities than it would be impossible based on the detection of X-ray pulsations.

In this work, we use the sample of Antoniou et al. (2009a), which includes the most likely optical counterpart of 158 Chandra sources with X-ray luminosities as low as L_X ∼ 4 × 10³³ erg s⁻¹, of uncertain or unknown spectral types. Moreover, this sample is supplemented by 211 additional sources detected in various XMM–Newton observations of the SMC reaching L_X ∼ 3.5 × 10³³ erg s⁻¹ (Haberl & Pietsch 2004; Antoniou et al. 2010), which also have uncertain or unpublished spectral types. We were able to obtain spectra for 133 Chandra and 151 XMM–Newton sources in total.

OBSERVATIONS AND DATA ANALYSIS

AAOmega spectroscopy

Although optical photometry for these sources has identified them as candidate HMXBs, and it is a powerful tool for identifying large samples of such objects, only optical spectroscopy can unambiguously confirm this classification, and provide additional information on the nature of these systems.

The optical spectra for this study were acquired during two nights of service time (on 2008 July 26 and September 19), using the multi-object mode of the AAOmega spectrograph (Sharp et al. 2006), a double-arm fibre-fed optical spectrograph (up to 400 fibres) on the 3.9 m AAT fed by the 2 Degree Field (2dF) robotic fibre positioner. A summary of the observing runs is presented in Table 1. In addition, flat-field and arc (FeAr+CuAr+CuHe+CuNe) calibration exposures were taken each night for each setup.

The data reduction was performed with the 2dfdr¹ v4 package with default values. However, we did not perform the sky subtraction built in 2dfdr as we followed a different approach than the standard procedure. The steps taken during the 2dfdr process included: (i) bias subtraction and flat-fielding; (ii) wavelength calibration; (iii) combination of individual exposures for the same field. The extraction of the individual spectra was performed with the extract command of the figaro v5.6-6 package of starlink (Shortridge et al. 2004).

After the extraction of all spectra from the AAOmega data, we performed the sky subtraction and initial characterization of the spectra. Flux calibration was not attempted since the wavelength-dependent throughput of each fibre is different and, ideally, a flux standard should be observed through each fibre. Nevertheless, our analysis is not affected because our classification criteria are based mainly on the presence or absence of spectral lines and not their absolute intensity. In addition, we can use the relative intensity of nearby lines, as the majority of the lines used in the classification are between 3900 Å and 4700 Å (with the exception of the Hα line), where the fibre response is fairly flat.

We considered for further analysis objects with S/N ratio above 20 (i.e. 400 counts) in the blue (4100 Å–4300 Å) as well as the red (6410 Å–6450 Å) parts of the spectrum. For the sky spectra, since the final sky spectrum resulted from the combination of several spectra, we set a limiting S/N ratio of 15 (i.e. 225 counts) in each of the two bands. After this selection, a total of 25 sky and 130 source spectra were kept from the September 19 observing run, while from the July 26 run we kept 5 sky and 53 spectra for further analysis.

The field of each sky fibre was examined visually (using the images from the OGLE-II project;² Udalski et al. 1998) in order to ensure that the spectra were not contaminated by any nearby source. After this process, 10 spectra were selected for further analysis from the south field (nine observed on September 19 and one on July 26) and two from the north field (observed on July 26). These sky spectra were combined into a median sky spectrum for each observation date and field.

All object and sky spectra were corrected for small residual wavelength offsets by measuring the positions of strong sky emission lines at 5577.3 Å and 6300.3 Å for the blue and the red band, respectively. In order to account for throughput variations between the object and sky fibres, the fluxes for these sky emission lines were measured for each stellar spectrum and the corresponding sky spectrum was scaled in order to match the measured intensity of the lines. Then the rescaled sky spectrum was subtracted from the corresponding object spectrum.

Although this method is sufficient for the subtraction of sky emission from each spectrum, it is not sufficient for the removal of the contaminating interstellar emission in the stellar spectra. The selected fields in the SMC show strong and spatially variable diffuse emission from H ii regions and supernova remnants. Among the interstellar emission lines, Hα is the strongest one but, at the same time, it is also a critical feature for the classification of BeXRBs. Ideally, sky subtraction would be performed with sky fibres placed within few arcsec from each source, in order to correct for the local interstellar contamination. However, due to hardware limitations we cannot place two fibres closer than 30 arcsec,³ leaving us with the only option of measuring an average diffuse emission spectrum for each field.

By measuring a mean sky spectrum, we can remove a large part of the interstellar emission background but there may still be some residual contamination. This is indicated by the presence of typical interstellar emission lines (such as [O iii] λ5007 and [S ii] λλ6716,6731) in the sky subtracted spectra.

Optical and infrared data for star S 18

Optical photometry for star S 18 (Henize 1956, or AzV 154 after Azzopardi et al. 1975) has been derived from the Optical Gravitational Lensing Experiment (OGLE) online database (see footnote 2; Udalski, Kubiak & Szymanski 1997; Szymanski 2005). The retrieved data were obtained in the Bessell I band. There are 327 observations between 1997 June and 2000 November, with typical photometric errors of 0.003 mag. Although photometry for this star also exists in the MAssice Compact Halo Object (MACHO) data base (Alcock et al. 1997, 1999), these data show unphysically large-scale scatter of up to ∼1 mag compared to the OGLE data. We attribute this to confusion with a star of similar brightness located ∼ 4 arcsec from S 18 (for comparison the median seeing of the MACHO survey is ∼ 3 arcsec; Alcock et al. 1997).

We used the catalogue compiled by Bonanos et al. (2010) to retrieve the infrared (IR) photometric properties of star S 18 (discussed in Section 6.3.1): J = 12.349 ± 0.033 mag, H = 11.931 ± 0.038 mag, K_s = 11.109 ± 0.026 mag, [3.6 μm] = 9.177 ± 0.042 mag, [8 μm] = 6.966 ± 0.022 mag, [24 μm] = 4.786 ± 0.007 mag.

X-ray data for source CXOU J005409.57−724143.5

In the ∼9.4 ks long Chandra observation obtained in 2002 July 04, we detected this X-ray source with an absorption-corrected X-ray luminosity of ∼3.5 × 10³³ erg s⁻¹, (0.5–7.0 keV; assuming a power-law spectrum with photon index Γ = 1.7 and an absorbing column density of N_H = 6.23 × 10²⁰ cm⁻², based on the average Galactic H i column density along the line of sight of this field; Dickey & Lockman 1990) at an off-axis angle of ∼5 arcmin (Zezas, in preparation). The small number of net counts (⁠|$9^{+3}_{-4}$|⁠) did not allow us to derive and model the X-ray spectrum for this source.

Despite the low significance of its intensity (1.9σ above the background) this is a solid detection (see Kashyap et al. 2010 for a discussion of the detection and intensity significance). This source was also detected by XMM–Newton on 2003 Dec. 18 (ObsID 0157960201) and reported in the XMM Serendipitous Source Catalogue (3xmm-dr4 Version⁴) as 3XMM J005408.9−724144. We reanalyzed these data with the XMM–Newton Science Analysis System (sas v12.0.1). After processing the raw data with the epchain and emchain tasks, we filtered any bad columns/pixels and high background flares (excluding times when the total count rate deviated more than 3σ from the mean), resulting in 14.8, 18.7 and 17.2 ks net exposures for the European Photon Imaging Camera (EPIC) Metal Oxide Semi-conductor (MOS1), MOS2, PN cameras, respectively. We only kept events of patterns 0–4 for the PN and 0–12 for the MOS detectors. Source detection was performed simultaneously in five energy bands (0.2–0.5, 0.5–1.0, 1.0–2.0, 2.0–4.5 and 4.5–12.0 keV) for each of the three EPIC detectors with the maximum likelihood method (threshold set to 7) of the edetect_chain task. At the position of CXOU J005409.57–724143.5 in the EPIC PN camera, there is source XMMU J005409.2−724143 with coordinates RA = 00:54:09.16 (J2000.0), Dec. = −72:41:43.46 (J2000.0) and a positional error of 1.6 arcsec (at the 1σ level, including the relative as well as the absolute astrometric uncertainty). The edetect_chain task lists this source with 39 ± 9 counts (source-detection likelihood |${\rm DET\_ML = 24.6}$|⁠) at an off-axis angle of 4.89 arcmin in the 0.2–12.0 keV energy band.⁵ The absorption-corrected X-ray flux, assuming a spectral model of an absorbed power law with a column density N_H = 6.23 × 10²⁰ cm⁻² and spectral slope Γ = 1.7, is then (1.6 ± 0.4) × 10⁻¹⁴ erg cm⁻² s⁻¹. At the distance of the SMC, this corresponds to absorption-corrected luminosity of |$L_{\rm X}^{\rm abs.} = (7.1 \pm 1.7) \times 10^{33}\, {\rm erg\, s^{-1}}$| (0.2–12.0 keV). We note that source XMMU J005409.2−724143 was not detected with either of the EPIC MOS detectors, while the XMM Serendipitous Source Catalogue lists it with slightly different coordinates and characteristics (⁠|${\rm DET\_ML = 34.6}$| and mean F_X ∼ (1.0 ± 0.4) × 10⁻¹⁴ erg cm⁻² s⁻¹), which are consistent with our measurements within the errors. Novara et al. (2011) who also analysed these data do not report this source since they only focus on bright sources.

We then extracted the PN source spectrum from an 8.5 arcsec radius aperture and a background spectrum from a 75 arcsec source-free region at the same distance from the readout as the source region. Unfortunately, the high background in combination with the small number of detected counts did not allow us to perform any spectral analysis.

The area around source CXOU J005409.57−724143.5 was also observed with XMM–Newton on 2006 November 01 (ObsID 0404680201). Following the same analysis procedure as above, we obtain a 30.5 ks net exposure for the EPIC PN camera. Despite the almost double exposure time compared to the 2003 observation, this time the source was undetected. We used the BEHR tool⁶ (Park et al. 2006) and measured an intensity upper bound of 10.3 counts (at the 99 per cent confidence level) at the location of the source. This corresponds to an observed X-ray luminosity of |$L_{\rm X}^{\rm abs.}\sim 4.7 \times 10^{32}$| erg s⁻¹ (0.2–12 keV) assuming the same spectral parameters and distance as for the other XMM–Newton observation. This is a factor of 10 lower than the previously derived source intensity.

SELECTION OF CANDIDATE BeXRBs

Here, we use the spectra extracted as described in the previous section in order to identify the spectral types of the studied sources. However, since contamination by the interstellar emission can be significant, for our analysis we selected objects with minimum contamination based on the width of their Hα emission line and their [S ii]/Hα ratio (c.f. Antoniou et al. 2009b). A minimum full width at half-maximum of the Hα emission line (FWHM|$_{\rm {H}_\alpha }$|⁠) is used to eliminate objects with too narrow emission, since BeXRBs have broader emission lines than the interstellar component (Coe et al. 2005). The selection of objects with [S ii]/Hα ratio smaller than in the average ‘sky’ spectrum also helps to eliminate objects with contamination from diffuse interstellar emission (mainly supernova remnants).

The minimum FWHM|$_{\rm {H}_\alpha }$| was based on the width of the Hα line measured in the sky spectra separately for each observation as the use of different gratings resulted in different spectral resolutions. For the July 26 data, there are only three sky spectra of suitable S/N (two from the north and one from the south field) which result in an average interstellar Hα width of |$\langle \mathrm{FWHM}_{\rm {H}_\alpha }\rangle = 5.70\pm 0.07$| Å. For the September 19 data, 9 good sky spectra were used resulting in an interstellar Hα width of |$\langle \mathrm{FWHM}_{\rm {H}_\alpha }\rangle = 1.94\pm 0.03$| Å. We selected for further analysis objects with an FWHM|$_{\rm {H}_\alpha }$| at least 3σ above the average width of the interstellar Hα emission. This means that only objects with FWHM|$_{\rm {H}_\alpha }$| > 5.91 Å for July 26, and FWHM|$_{\rm {H}_\alpha }$| > 2.03 Å for September 19, were considered for further examination.

In order to determine the average [S ii]/Hα ratio for the interstellar emission, we used again the sky spectra. We found [S ii]/Hα = 0.28 ± 0.11 for the south field (September 19) and [S ii]/Hα = 0.38 ± 0.02 for the north field (July 26). As no Hα emission was present in the only sky spectrum of the south field observed on July 26, we used the [S ii]/Hα ratio measured in the September 19 run (since the flux of the lines is independent of the spectral resolution). These values are in agreement with the [S ii]/Hα ratio (≥0.4) expected in environments with supernova remnants, indicating that the interstellar medium (ISM) in these regions is to a large degree shock excited. Thus, all sources considered as BeXRB candidates should have a [S ii]/Hα ratio smaller than the maximum values found for each field, in addition to the low limit on the FWHM|$_{\rm {H}_\alpha }$|⁠.

After applying these criteria, we are left with 21 sources out of the 272 initial targets that are BeXRB candidates and which are considered for spectral classification. These include 18 Chandra and 3 XMM–Newton sources.

The final spectra were normalized by subtracting the stellar continuum (after a spline fit), using the dipso v3.6-3 package of starlink (Howarth et al. 2004).

SPECTRAL CLASSIFICATION

Spectral classification criteria

As seen in the previous section, the selection of sources for further classification was based on their broad Hα lines, a key characteristic of BeXRBs. In Figs 1, 2 and A1, we present the spectra of the 21 sources selected for further analysis. As expected, by selection, they exhibit strong Hα emission. The left-hand panels in these figures show the blue part of the spectrum with the diagnostic lines for spectral-type classification marked, while the right-hand panels focus on the Hα line. Shaded areas indicate bad columns and/or sky subtraction residuals. We see that two sources (CH4-8 and CH4-2 in Fig. A1) show asymmetric Hα profiles, while two more sources (CH3-7 and CH5-6 in Fig. 1), show double-peaked Hα emission. Although Herbig Ae/Be stars present most of the time a double-peaked Hα profile (Vieira et al. 2003), we can safely rule out this possibility since the position of our objects in the V, B − V CMD (see fig. 3 of Antoniou et al. 2009a) ensures that they are not pre-main sequence objects. One more source (CH6-20) shows evidence for P-Cygni profiles (Fig. 2).

Figure 1.

The spectra of the five new BeXRBs identified in this work (including the tentative BeXRB CH7-19). Shaded areas indicate wavelength ranges of bad columns and/or sky subtraction residuals.

Open in new tabDownload slide

Figure 1

– continued

Open in new tabDownload slide

Figure 2.

Spectrum of the optical counterpart to the Chandra source CXOU J005409.57−724143.5 (CH6-20) identified as a sgB0[e] star (Zickgraf et al. 1989), which is the known sgB[e] star LHA 115-S 18 (Henize 1956). There is clear presence of emission lines of He ii, permitted and forbidden Fe lines, and Balmer lines with P-Cygni profiles.

Open in new tabDownload slide

Figure 3.

Comparison of spectral-type distributions of the different samples considered in this work: ‘Antoniou09’ (blue, right diagonal line filled) and ‘McBride08’ (green, left diagonal line filled) corresponds to the spectral-type distributions obtained from Antoniou et al. (2009b) and McBride et al. (2008), respectively; ‘sample’ (red solid line, ‘o’ filled) and ‘new’ (black solid line) correspond to the BeXRB sample studied in this work (excluding sources CH7-19 and CH6-20) and only the new sources (excluding source CH6-20) from this work, respectively, as defined in Section 6.1. (For sources extending over more than one class their spectral type is split equally between the encompassed class bins, e.g. a B0–B2 object will split into 1/3 in B0, B1 and B2 spectral type, respectively.)

Open in new tabDownload slide

The spectral-type classification is based on the scheme of Evans et al. (2004), which was later applied to BeXRBs by McBride et al. (2008) and Antoniou et al. (2009b), supplemented by the atlas of Gray & Corbally (2009). In order to classify OB stars, the use of metal lines is normally preferred, but because of the low metallicity of the SMC they are much weaker and difficult to detect. Thus, we classify the spectra based on a combination of He i, He ii and metal lines. The spectral lines used for our classification are presented in Table 2.

In Table 3, we present the list of the HMXBs identified in our study, along with the classification derived from this and previous studies. In this table each source is identified with an ID of the type CH/XMM F-NN, where CH stands for Chandra and XMM for XMM–Newton sources, F is the field number, and NN is the source ID from the studies of Antoniou et al. (2009a), and Antoniou et al. (2010), respectively. We find 20 sources of B spectral type which in combination with their strong Hα lines and X-ray emission make them BeXRBs, and one X-ray emitting supergiant system with also strong emission in the Hα line and evidence for a strong wind, which makes it a possible HMXB. Next, we discuss the new BeXRBs and the supergiant system identified in this work; sources that have been already classified as BeXRBs are discussed in detail in the Appendix.

For each source, we measured the equivalent width (EW) of the Hα line as the ratio of the flux in the region of Hα (6553–6576 Å) over the continuum at the position of Hα as calculated by a linear fit to the flux of line-free regions adjacent to Hα (i.e. at 6530–6540.5 Å and 6630–6650 Å). These measurements are presented in Table 4.

Discussion of individual sources

• CXOU J005409.57−724143.5 (source CH6-20) – classified as sgB0[e]. This is probably the most interesting object in our sample (see Fig. 2). Not only all the Balmer lines are in emission but we also observe He ii, as well as permitted and forbidden Fe lines in emission. Moreover, all the Balmer lines and some He ii lines present P-Cygni profiles, typical of supergiant B[e] stars (Shore, Sanduleak & Allen 1987; Zickgraf et al. 1989; Lamers et al. 1998). However, due to the lack of characteristic lines (which are absent even in high-resolution spectra, cf. Graus, Lamb & Oey 2012), we cannot identify the spectral type of this source. Through a model atmosphere fit Zickgraf et al. (1989) concluded that the stellar temperature corresponds to a B0 spectral type. According to the classification criteria of Table 2, the spectrum should also show He ii λ4686 for this spectral type, which is however absent in our spectrum. Nevertheless, the star is known to show variability in this line (Shore et al. 1987; Morris et al. 1996). To the best of our knowledge, there are no other classifications in the literature, thus we classify source CH6-20 as a sgB0[e]. This makes this source the second X-ray source associated with a supergiant in the SMC. We further discuss the nature of source CH6-20 in Section 6.3.

• CXOU J005504.40-722230.4 (source CH5-6) – classified as B1–B5. The He ii λλ4200, 4686 lines are absent so this spectrum is of B1 spectral type or later. As the He i λ4471 line is present without any sign of the Mg ii λ4481 line, we can deduce that the source must be earlier than B5. No other line that would help us to further constrain the classification is visible, thus we only propose a B1–B5 spectral-type range for source CH5-6.

• CXOU J005723.77−722357.0 (source CH3-7) – classified as B2. The He i λλ4144, 4387, 4471 absorption lines are clearly detected. Since no He ii λλ4200, 4686 lines are visible the source has a spectral type of B1 or later. Moreover, the absence of the Mg ii λ4481 line and the presence of Si iii indicates that the star is earlier than B2.5, which is also supported by the presence of the O ii λ4415–4417 lines and the O ii+C iii λ4640–4650 blend. The Si iv λλ4116, 4088 and Si ii λ4128–4132 lines are not present, which means that all criteria for a B2 star are completely fulfilled for this previously unclassified source.

• XMMU J010519.9−724943 (source XMM1-2) – classified as B3–B5. The absence of the He ii λλ4200, 4686 lines supports a spectral type of B1 or later. We see no evidence for the O ii+C iii λ4640-4650 blend, so the source must be later than B3, as this blend disappears after this class. Moreover, this source cannot be later than B5, since the He i λ4471 line is present but without any sign of the Mg ii λ4481 line. We thus assign to the previously unclassified source XMM1-2 a spectral type of B3–B5. Noteworthy are the weak Hγ emission and the almost filled-in (by emission) Hδ line.

• XMMU J010620.0−724049 (source XMM1-3) – classified as B9. The strong presence of the Mg ii λ4481 line is characteristic of a late B-type star. Moreover, the Mg ii λ4481 line is clearly stronger than the He i λ4471 line, which immediately places this source in the spectral range later than B8. The Ca ii K λ3933 line is also present in later B types and it becomes a dominant metal line in A-type stars (being much stronger than Mg ii). In our case, the Ca ii K λ3933 line is much stronger than the usual interstellar emission line seen in all other spectra, but it is weaker than the Mg ii λ4481 line, so we conclude that this, previously unclassified, star has a B9 spectral type.

• CXOU J004941.43−724843.8 (source CH7-19) – classified as B1–B5. This source displays an Hα emission line with an FWHM|$_{\rm {H}_\alpha } = 5.21$| Å which is marginally lower than our limiting Hα FWHM (FWHM|$_{\rm {H}_\alpha } = 5.70\pm 0.07$| Å for July 26; c.f. Section 4). Nevertheless, we decided to further examine this source. The most striking features of its spectrum are the emission in the Balmer lines series and the presence of the [O iii] λλ4959, 5007 lines. As the sky subtraction is not perfect, the [O iii] lines are residuals of the contribution from the surrounding environment. A previous classification for this source has been given by Murphy & Bessell (2000) as a potential Planetary Nebula (source 61 in their list). Their analysis showed that only 33 per cent of the good candidates are real Planetary Nebulae, while source 61 is not characterized as a good candidate as its properties are closer to emission stars. Taking into account that their position accuracy was not better than 12 arcsec and that there is a star cluster in the same region (Bica & Schmitt 1995), it is possible that this source was misclassified as a potential Planetary Nebula. In addition, the FWHM|$_{\rm {H}_\alpha } = 5.21$| Å translates to a rotational velocity of v sin i ∼ 240 km s⁻¹ which is within the normal range of rotational velocities for Be stars (Steele, Negueruela & Clark 1999). Thus, the major contributor of the spectral lines is considered to be of stellar nature than interstellar. The absence of both He ii λλ4200, 4686 lines suggests that the source is of spectral type B1 or later, but not later than B5 due to the absence of the Mg ii λ4481 line. Thus, we tentatively classify the previously unclassified source CH7-19 as B1–B5, although the Be nature is uncertain due to the marginal width of its Hα emission line.

DISCUSSION

New Be/X-ray binaries

In Table 3, we present the X-ray and optical properties of the BeXRBs studied in this work. From the 21 classified sources, 12 are in full agreement (within 0.5 spectral type) with the previous studies of McBride et al. (2008), and Antoniou et al. (2009b), while 3 have later spectral types (by 1 to 1.5 type) than previous results. Most importantly, we identify four new BeXRBs and one candidate sgXRB system (discussed in detail in Section 6.3). Although source CH7-19 is included in the table we exclude this source from further analysis, as its identification as an emission line star is tentative due to the possibly significant contamination of its spectrum by interstellar emission (see Sections 4 and 5.2). In the following discussion, we exclude from any comparisons with other samples source CH6-20 (the sgB[e]) since it does not belong to the same population as the luminosity class III-V BeXRBs. The remaining BeXRBs is hereafter referred to as the ‘sample’.

The four new BeXRBs sources (hereafter referred to as ‘new’) are: CXOU J005504.40−722230.4 (CH5-6), CXOU J005723.77−722357.0 (CH3-7), XMMU J010519.9−724943 (XMM1-2) and XMMU J010620.0−724049 (XMM1-3). The spectral types for sources CH3-7 and XMM1-3 (B2 and B9, respectively) are accurate to ±0.5 subclass. XMM1-3 is the BeXRB with the latest spectral type (B9) known in the SMC. For the remaining new BeXRBs, we can provide only a range of spectral types: B1–B5 for the source CH5-6, and B3–B5 for the source XMM1-2.

All but source CH3-7 (V = 14.71 mag), are faint objects with V-magnitudes in the range 16.4–17.9 mag (see Table 3) and have rather noisy spectra (see Fig. 1) which hampers their spectral classification. They are also faint X-ray sources with typical X-ray luminosities ∼10³⁴ erg s⁻¹, outside the typical range of luminosities of outbursting BeXRBs. Overall our sample of BeXRBs spans three orders of magnitude in X-ray luminosity and it includes faint BeXRBs for which it is not possible to detect X-ray pulsations.

Spectral-type distributions

By combining our results with these from previous studies of the BeXRBs in the SMC (e.g. McBride et al. 2008; Antoniou et al. 2009b), we can obtain a more complete sample of the properties of the BeXRB population in the SMC. In Fig. 3, we plot a histogram of the spectral-type distribution of our sample and new sources separately (as defined in Section 6.1), along with the samples of the previous studies by McBride et al. (2008) and Antoniou et al. (2009b). In order to account for the uncertainty in the spectral-type classification, we split sources extending over more than one class equally between the encompassed class bins (e.g. a B0–B2 object will split into 1/3 in B0, B1 and B2 spectral class, respectively.) By inspecting this figure, we see that there is a trend for our sample to extend to later spectral types than previous works. In order to assess the significance of this trend, we compared the spectral-type distributions of these samples using a two-sample Kolmogorov–Smirnov test (KS test; Conover 1999). We find that this trend is significant at more than 99 per cent confidence level, which further indicates that the sources in our sample are skewed towards later types (see Fig. 4 for the cumulative distributions of the spectral types in these three samples).

Figure 4.

Cumulative distributions of the spectral types of BeXRBs populations in the SMC for our sample (solid blue, excluding source CH6-20) compared with the samples (dashed green) of Antoniou et al. (2009b) and McBride et al. (2008). By applying the Kolmogorov–Smirnov test, we find that our sample is different from the previous ones, at more than 99 per cent confidence level (the probability to reject the null hypothesis, that the two distributions come from the same parent distribution, is given above each plot). This indicates that our sample is skewed to later spectral types.

Open in new tabDownload slide

In Fig. 5, we also present the unabsorbed X-ray luminosities of the sources identified in this study (new identifications are presented again separately) along with the maximum X-ray luminosity of known BeXRBs in the SMC, as given in table 1 of Rajoelimanana, Charles & Udalski (2011). As this list is not homogeneous, we transformed these values to the 0.5–7 keV energy band that we used in our study, assuming a power law with photon index Γ = 1.7 and N_H = 6 × 10²⁰ cm⁻². We obtained the maximum luminosities for our sources from table 1 of Rajoelimanana et al. (2011). For the sources that are not in this catalogue, we used the unabsorbed luminosities presented in Zezas (in preparation) and Antoniou et al. (2009a, 2010). We have also assumed that the X-ray luminosities given in Rajoelimanana et al. (2011) are unabsorbed. In any case, the difference between the unabsorbed and absorbed luminosities for the assumed model and energy band is not important for the purpose of the comparison presented here.

Figure 5.

Unabsorbed X-ray luminosities (L_X) of SMC BeXRBs plotted against the spectral type of their optical counterparts. The blue circles represent the maximum observed L_X for sources obtained from table 1 of Rajoelimanana et al. (2011) and after transforming these values to the 0.5–7 keV energy band that we used in our study (assuming a power law with photon index Γ = 1.7 and N_H ∼ 6 × 10²⁰ cm⁻²). The red asterisks (‘Known BeXRBs’) represent sources from our sample (see Table 3) with a previous classification and their maximum luminosities are derived either from table 1 of Rajoelimanana et al. (2011) or from Zezas (in preparation, Chandra sources) and Antoniou et al. (2010, XMM–Newton sources). The cyan asterisks (‘New BeXRBs’) represent new sources identified in this work (see Section 6.1) with their luminosities taken from Zezas (in preparation) and Antoniou et al. (2010).

Open in new tabDownload slide

In Fig. 5, we see a weak trend for lower luminosity sources to be associated with sources of wider spectral-type range, and extending to later types. A possible physical explanation for this trend may lie in the nature of the BeXRBs, especially if there is a correlation between the spectral type and the size of the equatorial disc. Although this is a very intriguing prospect, we should note that the observed trend could be the result of a selection bias: the list of Rajoelimanana et al. (2011) gives by construction the historically maximum detected X-ray luminosity for these sources, while our sample, which mainly contributes to the lower luminosity, later spectral-type sources, includes sources for which the X-ray luminosity was measured from a single snapshot.

In Fig. 6, we compare the average orbital period (P_orb) and eccentricity (e) as a function of spectral type of the BeXRB populations in the SMC and MW. Data for the orbital periods of BeXRBs in the SMC are taken from Rajoelimanana et al. (2011), and for the MW from Townsend et al. (2011). For the eccentricities, we take all data (for both SMC and MW) from Townsend et al. (2011). In Fig. 7, we present the cumulative distributions of the spectral types, orbital periods and eccentricities for MW and SMC BeXRBs. For the SMC, we used the spectral types of sources as derived from this work and the previous studies of McBride et al. (2008) and Antoniou et al. (2009b), while for the MW, we take the data from the review paper of Reig (2011). We find marginal evidence for difference between the spectral type distributions of SMC and MW BeXRBs (∼99.9 per cent confidence based on the KS test), in contrast to previous studies which found no evidence for difference between the BeXRB populations in these two galaxies (e.g. McBride et al. 2008; Antoniou et al. 2009b). We attribute this difference to the use of larger samples for both the SMC and the MW than previous studies, and the extension of the former to later spectral types. However, given the important implications of this difference for the evolution of BeXRBs in the SMC, we consider it as only tentative at this point. On the other hand, we do not find any evidence for difference in the distribution of the orbital parameters (P_orb, e) in SMC and MW BeXRBs, based on the KS test. This result provides an indication for the distribution of the kick velocities imparted on the pulsars during the supernova explosion. The vector of the kick and the mass of the secondary object (in the initial binary) will affect the orbit of the neutron star around it. Since the spectral types (which reflects the donor mass of the BeXRBs) are not dramatically different between the SMC and the MW, and the orbital-element distributions (period and eccentricity) are not different at a statistically significant level, we can deduce that there is an indication that the kicks imparted to the pulsars during the supernova explosions have similar strengths in the SMC and the MW. Furthermore, support comes from previous studies, where the measured kicks in the MW BeXRBs (v ∼ 15 ± 6 km s⁻¹; van den Heuvel et al. 2000) are found to be comparable with these of the SMC BeXRBs (v ∼ 30 km s⁻¹; Coe 2005, v < 15–20 km s⁻¹; Antoniou et al. 2010).

Figure 6.

The orbital periods (P_orb) and eccentricities (e) as a function of the spectral types of the optical counterparts of BeXRBs populations in the SMC and the Milky Way (binned to 1 spectral type). For the orbital periods, the solid blue and the dashed green lines correspond to the samples of BeXRBs in the SMC (after Rajoelimanana et al. 2011) and in the Milky Way, respectively (after Townsend et al. 2011). For the eccentricities, we use data from Townsend et al. (2011) for BeXRBs in the SMC (dotted red) and Milky Way (dashed green). The error bars indicate the 1σ standard deviation for each of the two parameters within each spectral-type bin.

Open in new tabDownload slide

Figure 7.

Cumulative distributions of the spectral types, orbital periods (P_orb) and eccentricities (e) for BeXRBs in the MW (solid blue) and the SMC (dashed green). For the spectral-type distributions, the data come from this work, McBride et al. (2008) and Antoniou et al. (2009b), while for the MW we take the data from the review paper of Reig (2011). Data for the P_orb of BeXRBs in the SMC are taken from Rajoelimanana et al. (2011), and for the MW from Townsend et al. (2011). For the e, we take all data (for both SMC and MW) from Townsend et al. (2011). The Kolmogorov–Smirnov test probabilities are given over each plot. We find marginal evidence for difference between the spectral type distributions of SMC and MW BeXRBs (at ∼99.9 per cent confidence), while we do not find any evidence for difference in the distribution of the orbital parameters (P_orb, e) in SMC and MW BeXRBs (see Section 6.2).

Open in new tabDownload slide

The case of the supergiant B[e] source CXOU J005409.57−724143.5

Optical and infrared properties

The optical counterpart of source CXOU J005409.57−724143.5 (CH6-20) is classified in this work as a supergiant star. If the X-ray source is an accreting binary, it would be only the second to be found in the SMC after SMC X-1 (Webster et al. 1972). Its optical counterpart is star LHA 115-S 18 (Henize 1956) also known as AzV 154 (Azzopardi et al. 1975), a well known bright emission-line star that has been studied systematically since 1956 (Shore et al. 1987; Zickgraf et al. 1989; Morris et al. 1996; Massey & Duffy 2001; van Genderen & Sterken 2002, and references therein), and more recently in Clark et al. (2013). In Fig. 8, we present a finding chart of the X-ray source and its corresponding optical counterpart.

Figure 8.

Finding chart of source CXOU J005409.57−724143.5 (or CH6-20 in Antoniou et al. 2009a) from an OGLE-III I-band image (Udalski et al. 2008). The dimensions of the field are 24.7 arcsec × 24.7 arcsec. The cyan circle indicates the positional error-circle of the Chandra source (1.5 arcsec radius, see Antoniou et al. 2009a for details) and the magenta x-symbol indicates the location of star number 7 from the OGLE-III SMC105.6 map, which is spatially coincident with star LHA 115-S 18 (Henize 1956).

Open in new tabDownload slide

According to the classification criteria for supergiant B[e] stars (Lamers et al. 1998), these objects exhibit strong Balmer emission lines (usually with P-Cygni profiles indicating mass-loss), low excitation permitted and forbidden lines (e.g. Fe ii), and strong near- or mid-IR excess (due to hot circumstellar dust). Moreover, they are rather stable photometrically (variation of the order of ∼0.1–0.2 mag) and spectroscopically, unlike the S Doradus/Luminous Blue Variable (LBV) stars which exhibit similar spectra (Zickgraf et al. 1986).

As discussed in Section 5.2, star S 18 presents the typical spectral characteristics of B[e] stars. In addition, it is very bright in the near-IR showing a very large colour excess of J-[3.6 μm] = 3.17 ± 0.05 mag, compared to the main sequence and supergiant B stars (most Be stars show a colour excess of J-[3.6 μm] ∼ 0.8 mag; Bonanos et al. 2010).

In Fig. 9, we present the light curve of star S 18, using the OGLE-II data in the I filter (as discussed in Section 3.2), clearly indicating a highly variable source. In addition, S 18 exhibits spectroscopic variations (e.g. in the lines of He ii λ4686, C iv λ1550, N iv λ1487; Shore et al. 1987). This is in contrast with the photometric stability of typical sgB[e] stars, and it is more similar to the strong variability of LBV stars (e.g. Zickgraf et al. 1986). However, there is growing evidence that the sgB[e] and the LBV sources are not distinct classes (Morris et al. 1996; van Genderen & Sterken 2002; Clark et al. 2013). In order to investigate if it is an accreting binary source, we searched for modulation of its optical emission resulting from an orbital period. Due to the significant variability, we detrended the light curve following Schurch et al. (2011), by fitting a second-order polynomial to each year-long segment of the data. Then, we obtained the Lomb–Scargle periodogram (Lomb 1976; Scargle 1982) for the entire detrended light curve. We did not find any periodicity at the 90 per cent confidence level in agreement with the results of Clark et al. (2013). The significance level was estimated by simulating light curves based on the noise characteristics of the data and repeating the analysis for each simulated light curve.

Figure 9.

OGLE-II I-band light curve of the optical counterpart of source CH6-20 (ROSAT observations were obtained before the OGLE-II survey, while Chandra and XMM–Newton observations were obtained later). The horizontal line indicates the median value of ∼12.6 mag, and the errors on the photometric points are 0.003 mag. The observed aperiodic variability is larger than what expected for a normal supergiant B[e] star and it is more similar to the variability that Luminous Blue Variables display (see Section 6.3.1).

Open in new tabDownload slide

X-ray properties

In order to explain the spectral variations of S 18, Shore et al. (1987) suggested the presence of a hot companion, possibly a helium star or a neutron star. But no further evidence of such an object existed as the X-ray observations at the time were not sensitive enough to observe any emission from accretion on to a putative compact object. Nowadays, Chandra and XMM–Newton are able to routinely detect sources with X-ray luminosities down to ∼10³³ erg s⁻¹ at the distance of the SMC (with 10–30 ks exposures) and have succeeded in detecting source CXOU J005409.57−724143.5 (CH6-20) which is spatially coincident with star S 18. In the study of Antoniou et al. (2009a), its proposed optical counterpart is star O_6_311169 (00:54:09.53, -72:41:42.9) from the OGLE-II catalogue (Udalski et al. 1998) or Z_2611188 (00:54:09.57, −72:41:42.9) from the Magellanic Clouds Photometric Survey (MCPS) catalogue (Zaritsky et al. 2002), which is coincident with star S 18. With this work, we confirm that the optical counterpart of the source CXOU J005409.57−724143.5 (CH6-20) is the supergiant B[e] star S 18.

This source has only been detected in the X-ray band in only two epochs (see Section 3.3). The lack of previous detections with ROSAT or Einstein indicate that it is persistently at a low-luminosity state (L_x < 10³⁵ erg s⁻¹) since at higher luminosities it would have been detected already in one of the several pointings with these observatories. Although the intensity of CH6-20 is only ∼2σ above the background in the Chandra survey, the independent XMM–Newton detection with a source-detection likelihood of DET_ML = 24.6 during the first XMM–Newton observation ensures that the source is not spurious.

The estimated peak L_X at periastron from the values obtained from Zickgraf et al. (1989) is too high for this source to remain undetected for so long. On the other hand, when the values from Mokiem et al. (2007) are used, the estimated peak L_X is consistent with the measured luminosity and comparable to the detection limits of the Chandra and XMM–Newton observations. However, the above luminosities correspond to the peak L_X for this source when the neutron star is located at periastron, assuming a stable mass-loss rate. If the X-ray observations were obtained when the neutron star is away from the periastron and/or a lower mass-loss rate phase (as for example it would be expected from its significant optical variability) then the X-ray luminosity would be even lower than that estimated above.

Furthermore, a clumpy wind with a high mass-loss rate may provide an alternative interpretation for the nature of this source. In this case, the high column density towards the neutron star due to the wind may obscure its X-ray emission. A column density of N_H = 3 × 10²⁴ cm⁻² (assuming a power-law spectrum with a photon index Γ = 1.7 in the 0.5–7 kev energy band) would be enough to attenuate a source with intrinsic L_X = 10³⁷ erg s⁻¹ (which is the upper range in its X-ray luminosity based on the parameters from Zickgraf et al. 1989) down to L_X = 5.7 × 10³³ erg s⁻¹, while an N_H = 1 × 10²⁴ cm⁻² would be enough to attenuate a source with intrinsic L_X = 7 × 10³⁴ erg s⁻¹ (the luminosity expected from the parameters of Mokiem et al. 2007) down to L_X = 1.6 × 10³³ erg s⁻¹. Both of these estimates are half the value we found with Chandra. However, we do have significant detection below 2 keV in both the Chandra and XMM–Newton observations (⁠|$6_{-2}^{+3}$| and 39 ± 11 counts, respectively), and only upper limits or marginal detections above 4 keV, which do not allow us to set any useful constraints on the spectral parameters of this source.

There is an emerging subclass of highly obscured wind-fed sgXRBs, discovered in recent INTEGRAL observations (Walter et al. 2006). To date there are three known X-ray sources with B[e] companions: IGR J16318−4848 (Walter et al. 2003; Filliatre & Chaty 2004), which is considered the prototype of highly obscured wind-fed sgXRBs, CI Cam/XTE J0421+560 (Clark et al. 1999; Boirin et al. 2002), the first HMXB with a sgB[e] companion, and Wd1-9 (Clark et al. 2008), a probable colliding-wind system. Recently, Bartlett et al. (2012) presented a detailed study of CI Cam observed in 2003 with XMM–Newton. This system has a B0-2[e] supergiant companion (Clark et al. 1999) and shows a heavily absorbed (N_H ∼ 4.4 × 10²³ cm⁻²) power-law (Γ ∼ 1) spectrum reaching L_X ∼ 4.1 × 10³³ erg s⁻¹ (3–10 keV) in quiescence. In 1998, it showed an outburst reaching a flux of ∼4.8 × 10⁻⁸ erg cm⁻² s⁻¹ (2 Crab) in the 2–10 keV energy band (Smith et al. 1998). Assuming a distance of 5 kpc, the outburst luminosity is ∼5.7 × 10³⁶ erg s⁻¹. The nature of the compact object in this system as well as in IGR J16318−4848 is unclear (Bartlett et al. 2012), although there is some indication that at least in the case of IGR J16318−4848 it might be a neutron star (Filliatre & Chaty 2004).

The main difference of these systems from the ‘classical’ sgXRBs is that they are much more absorbed in the X-ray band. The compact object is deeply embedded in the dense material of the wind and the absorbing column density changes during the orbit (it may increase up to 10 times close to the eclipse; Manousakis & Walter 2011). Moreover, material away from the compact object (like the material around the sgB[e] star) may contribute to the absorption (e.g. as in IGR J16318−4848; Walter et al. 2006). In this case, the source exhibits strong IR excess, attributed to free–free and bound-free emission from hydrogen in a circumstellar envelope (or disc; Wisniewski et al. 2007). At longer wavelengths such as those observed by Spitzer, the IR excess indicates the presence of warm dust in the circumstellar envelope or the disc. In Fig. 10, we present the location of source CH6-20 in a near-infrared (NIR) colour–colour diagram. This plot is an updated version of fig. 4 by Graus et al. (2012), where in addition to the BeXRBs we include all highly obscured sgXRBs identified so far by INTEGRAL. These sources show luminosities in the 10³³–10³⁴ erg s⁻¹ range, they are highly variable and they have absorbing column densities in the 10²³–10²⁴ cm⁻² range. We find that source CH6-20 shows redder NIR colours than the other BeXRBs which place it closer to the colours of the obscured supergiant HMXBs.

Figure 10.

J − H versus H − K_s colour–colour diagram of BeXRBs with available 2MASS data. Black dots show sources from our sample and grey dots show sources from the samples of McBride et al. (2008) and Antoniou et al. (2009b). We present also X-ray sources with a known sgB[e] companion (triangles) and the highly obscured sgXRBs identified so far by INTEGRAL (cyan polygons; Manousakis 2011). Our source CH6-20 is presented as a purple diamond.

Open in new tabDownload slide

An alternative scenario for the nature of the CH6-20 is that it is a colliding-wind binary (e.g. Pollock 1987). These systems consist of two orbiting massive stars in orbit, and their low-luminosity X-ray emission (L_X ∼ 10³³–10³⁴ erg s⁻¹) is produced by shocks at the collision front of the winds. Clark et al. (2013) point out that although S 18 (CH6-20) and the wind-fed sgXRB CI Cam exhibit many similarities, their long-term photometric and spectroscopic behaviours are not quite the same, and argue for a colliding-wind nature for the former system instead of an accreting binary. This is supported by the location of source CH6-20 in the NIR colour–colour diagram (Fig. 10) which is also consistent with that of the colliding-wind system Wd1-9 (Clark et al. 2008) and their similar X-ray luminosity (unabsorbed L_X ∼ 4 × 10³³ erg s⁻¹ for Wd1-9, in the 0.5–8 keV band; Clark et al. 2008). The luminosities of both CH6-20 and Wd1-9 are well within the range of colliding-winds systems (∼10³²–10³⁵erg s⁻¹, Stevens, Blondin & Pollock 1992; Gagne et al. 2012). Although colliding-wind systems also have hard X-ray spectra, a possible discriminating feature from accreting pulsar binaries is that the spectra of the former have a cut-off at ∼10 keV, and they show effectively no photons with energies above 10 keV (Pollock, private communication). Unfortunately the weak X-ray emission of this source does not allow us to detect any X-ray photons above 2 keV. This in combination with its highly variable optical light curve that does not allow us to measure its orbital parameters, hampers the distinction between these two scenarios. However, if the heavily obscured sgB[e] scenario proves to be correct then CXOU J005409.57−724143.5 (CH6-20)/S 18 will be the first extragalactic heavily obscured sgXRB, and the second sgXRB in the SMC.

CONCLUSIONS

In this paper, we presented our results from optical spectroscopic observations of the optical counterparts of X-ray sources detected in the Chandra and XMM–Newton surveys of the SMC. We used the AAOmega spectrograph at the AAT to observe sources identified in Zezas (in preparation) and Antoniou et al. (2009a, 2010), with the aim to identify new HMXBs and determine their spectral types. We identified five new BeXRBs and one known supergiant system which we associate with an X-ray source. We confirmed the previous classifications (within 0.5 spectral type) of 12 sources, while for three sources our revised classification, with higher resolution and S/N data, result in later (by 1–1.5 subclass) spectral types.

We were able to classify BeXRBs with X-ray luminosities over three orders of magnitude. The selection of the parent sample from Chandra and XMM–Newton observations of the SMC, allows us to extend our census of BeXRBs to almost quiescent luminosities. A comparison of the populations of BeXRBs in the SMC and the MW with respect to their spectral types reveals a marginal evidence for difference. However, we find no statistically significant differences for their orbital parameters (periods and eccentricities). This result further supports other lines of evidence for similar supernova kick velocities between the low-metallicity SMC and the MW.

Finally, we discuss the X-ray, optical and IR properties of source CXOU J005409.57−724143.5. This intriguing source is associated with the well-known supergiant star LHA 115-S 18. Its optical and X-ray properties do not allow us to distinguish between a colliding-wind system or an sgXRB. If the second scenario proves to be correct, then this source would be the first extragalactic sgXRB with a B[e] companion.

We would like to thank the referee, Phil Charles, whose comments and suggestions helped to improve the paper. We would like to thank the staff of the Anglo-Australian Observatory for obtaining the data used in this work. We thank A. Manousakis and R. Walter for useful discussions regarding the class of heavily obscured sgXRBs, and A. Pollock for advice regarding the class of colliding-wind binaries. GM acknowledges support by State Scholarships Foundation of Greece in the form of a scholarship. AZ acknowledges support by the Chandra grant GO3-14051X, the NASA grant NNX12AL39G and the EU IRG grant 224878. Space Astrophysics at the University of Crete is supported by EU FP7-REGPOT grant 206469 (ASTROSPACE). VA acknowledges support by NASA grant NNX10AH47G issued through the Astrophysics Data Analysis Program, and Chandra grant GO3-14051X.

This research has made use of NASA's Astrophysics Data System, SIMBAD data base (operated at CDS, Strasbourg, France), and data products from the Optical Gravitational Lensing Experiment and 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. Figures were generated using the matplotlib⁷ library in the python⁸ programming language.

REFERENCES

APPENDIX A: SUPPLEMENTARY SPECTRAL CLASSIFICATION RESULTS

In this section, we review the spectral classification for the sources not discussed in Section 5.2 (15 in total). For sources that we confirm previous classifications we only give the spectral classification, while for sources with updated classifications we give a detailed account of the basis for these new classifications. We also present all spectra in Fig. A1.

• CXOU J004814.15−731004.1 (source CH4-8) – classified as B1.5. This source is classified as B1.5, in full agreement with Antoniou et al. (2009b).

• CXOU J004903.37−725052.5 (source CH7-1) – classified as B1-B5. The absence of both He ii λλ4200, 4686 lines and of the Mg ii λ4481 line combined with the presence of the He i λ4471 line, constrain the spectral type in the B1–B5 range, in agreement with McBride et al. (2008) who classified this source as ∼B3.

• CXOU J004913.57−731137.8 (source CH4-2) – classified as B3–B5. For this source, we obtained spectra on both nights (2008 July 26 and September 19). The spectral range B3–B5 is determined by the absence of the He ii λλ4200, 4541 and 4686 lines, and the O ii+C iii λ4640-4650 blend, as well as the stronger He i λ4471 line compared to the Mg ii λ4481 line. The resulting spectral type is later than the previous classification of B1.5 in Antoniou et al. (2009b), where the O ii+C iii λ4640-4650 blend and Si iv λλ4088, 4116 lines were identified.

• CXOU J004929.74-731058.5 (source CH4-5) – classified as B1–B5. The absence of the He ii λλ4200, 4686 lines, and the combination with the clear presence of the He i λ4471 line and the absence of Mg ii λ4481, indicates a spectral type in the B1–B5 range. Antoniou et al. (2009b) provided a spectral type of B1, based on the presence of the O ii+C iii λ4640-4650 blend and Si iv λλ 4088, 4116 lines, which are not detected in our deeper, higher resolution spectra.

• CXOU J005057.16−731007.9 (source CH4-3) – classified as B1–B3. Source CH4-3 was also observed both nights (2008 July 26 and September 19). The absence of the He ii λλ4200, 4686 lines and the presence of the O ii λ4415-4417 and O ii+C iii λ4640-4650 blend limit the spectral type in the B1–B3 range. This source has been classified previously as B0.5 by Antoniou et al. (2009b) due to the weak presence of the He ii λ4686 line, which is not detected in our spectra.

• CXOU J005153.16−723148.8 (source CH5-3) – classified as B0.5. Given the weak He ii λ4686 line, and the absence of the He ii λλ4200, 4541 lines, all the criteria for a B0.5 star are fulfilled, resulting in a little later type than the previous classification of O9.5-B0 by McBride et al. (2008), although in the latter work there are no details about the specific lines that led to this classification.

• CXOU J005205.61−722604.4 (source CH5-1) – classified as B3-B5. The combination of the presence and relative strength of the He i λλ4009, 4026 and 4144 lines, and the absence of the Mg ii λ4481 line limits the spectral type to the B3–B5 range. This is later than the previous classification of B1–B1.5 by McBride et al. (2008), for which no details about the specific lines are presented.

• CXOU J005208.95−723803.5 (source CH6-1) – classified as B1-B3. This source is classified as B1–B3, in full agreement with Antoniou et al. (2009b).

• CXOU J005245.04−722843.6 (source CH5-12) – classified as B0. The clear presence of the He ii λλ4541, 4686 lines combined with the fact that the He ii λ4200 is absent constrains the spectral type to B0, since for earlier types this line is of comparable strength to the He ii λ4541 line. Thus, all criteria for a B0 star are fulfilled, which improves the previous wider classification of O9–B0 by McBride et al. (2008).

• XMMU J005255.1−715809 (source XMM2-1) – classified as B1-B3. The absence of the He ii λλ4200, 4541, 4686 lines and the presence of the O ii+C iii λ4640-4650 blend, allow us to classify source XMM2-1 as B1-B3, which is in marginal agreement with the previous classification of B0-B1 by McBride et al. (2008).

• CXOU J005355.25−722645.8 (source CH5-16) – classified as B0. The He ii λλ4541, 4686 lines are present, but not the He ii λ4200 line, indicating that this is a B0 star (earlier types display the He ii λ4200 line in at least comparable strength to the He ii λ4541 line). Antoniou et al. (2009b) have classified this source as B0.5 based on the absence of the He ii λ4541 line, in contrast to our deeper spectrum, where it is clearly seen.

• CXOU J005455.78−724510.7 (source CH6-2) – classified as B1.5-B3. The presence of the He ii λλ4200, 4686 lines and the absence of Si iv λ4116 limit the spectral range to an as early-type as B1.5, while the presence of the O ii+C iii λ4640-4650 blend limits the spectral type to no later than B3. This is marginally consistent with the previous classification of B1–B1.5 by Antoniou et al. (2009b), based on the presence of the Si iv λλ 4088, 4116 lines, which however are not detected in our deeper spectra.

• CXOU J005456.34−722648.4 (source CH5-7) – classified as B0.5. The absence of the He ii λ4200 line and the presence of the He ii λ4686 line suggest a spectral type of B0.5. The slightly earlier classification of B0 by Antoniou et al. (2009b) was based on the presence of the He ii λ4200 line, which is not seen in our deeper spectrum.

• CXOU J005605.42−722159.3 (source CH3-18) – classified as B2. The Si iii λ4553 line is clearly present but without any sign of the Si iv λλ4088, 4116 lines. This combination is in agreement with the criteria set for B2 stars. This classification is later than the previous classification of B1 by McBride et al. (2008).

• CXOU J005736.00−721933.9 (source CH3-3) – classified as B1–B5. The absence of the He ii λλ4200, 4686 lines, in combination with a clearly present He i λ4471 line and the absence of the Mg ii λ4481 line, suggest a spectral range of B1–B5. This classification result is in agreement with the also broad classification of B0–B4 by Antoniou et al. (2009b).

Figure A1.

The spectra of BeXRBs studied in this work with previously known classifications. Shaded areas indicate wavelength ranges for bad columns and/or sky subtraction residuals.

Open in new tabDownload slide

Figure A1

– continued

Open in new tabDownload slide

Figure A1

– continued

Open in new tabDownload slide

Figure A1

– continued

Open in new tabDownload slide

Figure A1

– continued

Open in new tabDownload slide