Optical identification of XMM sources in the Canada–France–Hawaii Telescope Legacy Survey

Abstract

1 INTRODUCTION

Sky surveys play a key role in astronomy as they provide the basic data for the characterization of different populations of astronomical objects and their evolution over cosmic time. Here, we take advantage of two surveys of the same field: (i) the XMM–Large Scale Structure survey (XMM–LSS; Pierre et al. 2004) to survey the X-ray sky to a relatively low-flux limit of 1 × 10⁻¹⁵ erg s⁻¹ cm⁻² in the 0.5–2 keV band and (ii) the wide part of Canada–France–Hawaii Telescope Legacy Survey (CFHTLS; Cuillandre & Bertin 2006) to observe limited portions of the sky to faint magnitude limits in five optical bands.

The vast majority of the AGN known today were discovered during optical surveys like the bright Palomar Quasar Survey (Palomar–Green, PG; Schmidt & Green 1983), the Large Bright Quasar Survey (Hewett, Foltz & Chaffee 1995), the Hamburg–ESO (European Southern Observatory) Quasar Survey (HES; Reimers, Koehler & Wisotzki 1996) the Two-degree-Field Quasar Survey (2dF; Croom et al. 2001) and the Sloan Digital Sky Survey (SDSS; Schneider et al. 2007). Such optical colour selection based on wide photometric bands can miss the obscured AGN. The SDSS has identified many type 2 AGN based on the presence of narrow emission lines in their spectra without broad lines, and with emission-line ratios typical of AGN (Kauffmann et al. 2003; Zakamska et al. 2003). Alternatively, X-ray surveys include AGN without narrow emission lines or those at high redshift where emission-line selection is not possible. This is why the identification of X-ray sources down to low optical magnitudes is an important step towards elucidating AGN activity.

There are a number of existing surveys in various energy bands in the X-ray wavelength range such as the Einstein Extended Medium Sensitivity Survey (Maccacaro et al. 1982; Stocke et al. 1991), the ROSAT Bright Survey (Schwope et al. 2000) and the ASCA survey (Ueda et al. 2003), but they all are of relatively poor sensitivity. There are however deep surveys with ROSAT, such as the UDS (ROSAT Ultra Deep Survey; Lehmann et al. 2001) and the ROSAT Deep Survey (Hasinger et al. 1998). They reach an X-ray flux limit of ∼10⁻¹⁵ erg cm⁻² s⁻¹, however over very small regions in the sky. With the Chandra and XMM–Newton X-ray telescopes in operation, X-ray surveys have been significantly boosted. Surveys carried out using XMM–Newton and Chandra include the Subaru/XMM–Newton Deep Survey (Ueda et al. 2008), Large Area Lyman Alpha Survey (Wang et al. 2004), CLASXS (Chandra Large Area Synoptic Survey; Yang et al. 2004), Chandra Multiwavelength Project (Kim et al. 2004), Chandra Deep Field-South (Szokoly et al. 2004; Luo et al. 2008), Chandra Deep Field-North (Barger et al. 2003), Bootes Survey (Brand et al. 2006), HELLAS2XMM (XMM High Energy Large Area Survey; Cocchia et al. 2007) and the XMM–COSMOS survey (Brusa et al. 2007; Cappelluti et al. 2009; Trump et al. 2009). Most of these X-ray surveys are either deep/ultra-deep pencil beam surveys or shallow large-area surveys.

The XMM–LSS survey is a medium depth survey conducted with XMM–Newton in the CFHTLS region. It is designed to provide a well-defined statistical sample of X-ray-selected galaxy clusters out to a redshift of unity, over a large area suitable for cosmological studies (Pierre et al. 2004). Apart from finding new galaxy clusters, XMM–LSS will also provide X-ray point-like sources down to low-flux levels, of which AGN might represent ∼95 per cent (Pierre et al. 2007) as they are known to be strong X-ray emitters (Fabbiano, Kim & Trinchieri 1992) in comparison to normal galaxies. These AGN are thought to be at the origin of most of the X-ray background (Giacconi et al. 2002; Alexander et al. 2003). Recently, about 80 per cent of the background has been resolved in the 2–10 keV energy range by deep Chandra and XMM–Newton observations (e.g. Worsley et al. 2005; Hickox & Markevitch 2006; Carrera et al. 2007). Optical identification of such XMM sources is of great importance to characterize the AGN population as a function of redshift and address specific issues related to the importance of the AGN phenomenon in galaxy formation and evolution of the intergalactic medium. There is an overlap between the XMM–LSS and surveys at different wavelengths: the Visible Multiobject Spectrograph (VIMOS) Very Large Telescope (VLT) Deep Survey (VVDS; Le Fèvre et al. 2004) and the CFHTLS in the optical, the UKIRT (United Kingdom Infrared Telescope) Infrared Deep Sky Survey (Dye et al. 2006; Lawrence et al. 2007) in the near-infrared (IR), the Spitzer Wide Area InfraRed Extragalactic Legacy Survey (SWIRE; Lonsdale et al. 2003), radio observations from the VLA at 1.4 GHz (Bondi et al. 2003) and at 325 and 74 MHz (Cohen et al. 2003), 610 MHz observations from the Giant Metrewave Radio Telescope (GMRT) (Bondi et al. 2007) and ultraviolet (UV) observations with GALEX (Arnouts et al. 2005; Schiminovich et al. 2005). In this work, we use version 1 of the XMM–LSS X-ray catalogue (Pierre et al. 2007). Sources were entered in the catalogue if the likelihood of detection in either of the survey bands is greater than 15, and if the observed flux is larger than a 0.5–2 keV flux limit of 1 × 10⁻¹⁵ erg cm⁻² s⁻¹.

In this paper, we present optical spectroscopic identification of these XMM sources in ∼3 deg² of the CFHTLS down to g < 22 mag. This spectroscopic campaign is intermediate between large scale but shallower surveys like the 2dF (Croom et al. 2001) or SDSS (Schneider et al. 2007) and deeper but spatially restricted surveys like the VVDS (Gavignaud et al. 2006) which has a limiting magnitude of I_AB < 24, the AGN survey in the COSMOS field (Trump et al. 2009) reaching a magnitude limit of i < 23.5 and optical identification of XMM sources in the European Large Area ISO Survey (ELAIS-S1) field reaching down to R∼ 24.5 (Feruglio et al. 2008).

Identifications of X-ray sources in the XMM–LSS have been already obtained by Tajer et al. (2007) and Garcet et al. (2007). The former authors survey sources detected in 1 deg² with F_{2-10 keV} > 10⁻¹⁴ erg cm⁻² s⁻¹ in the 2–10 keV band at a significance ≥3σ and report 122 unambiguous identifications. The later authors present identification of 99 sources with R≤ 22 mag. Here, we increase by a factor close to 5 the number of identifications in this field. This paper is organized as follows. We present the selection of targets in Section 2; Section 3 describes the spectroscopic observations; identifications are discussed in Section 4; Section 5 gives the discussion of results and conclusions are summarized in the final section. Throughout this paper, we adopt a cosmology with H₀= 70 km s⁻¹ Mpc⁻¹, Ω_M= 0.27 and Ω_Λ= 0.73.

2 OPTICAL AND X-RAY SURVEYS

The basic data sets used to select the targets for this survey are from the XMM–LSS (Pierre et al. 2004) and the wide synoptic component (W1) of the CFHTLS.1

2.1 CFHTLS

The CFHTLS is an ambitious imaging programme that has been carried out at the 3.6-m Canada–France–Hawaii Telescope (CFHT) using the wide-field prime focus MegaPrime equipped with MegaCam, a 36 CCD mosaic camera each of them with 2048 × 4612 pixels². The pixel scale is 0.185 arcsec, thus giving a total field of view of 0.96 × 0.94 deg². The CFHTLS consists of a deep survey in four fields (D1, D2, D3 and D4) each covering about 1 × 1 deg² and a shallower survey on four wide fields (W1, W2, W3, W4) each covering 7 × 7 deg² in u*, g′, r′, i′ and z′ filters. Final limiting magnitudes in the wide fields should be of the same order in the five bands and typically i∼ 24.5. The internal accuracy of the astrometric solution (band to band) is better than 1 pixel rms over the entire MegaCam field, whereas the external astrometric solution is around 0.25 arcsec rms (Schultheis et al. 2006).

2.2 XMM–LSS

The XMM–LSS is a medium depth large-area X-ray survey designed to map LSSs in the Universe (Pierre et al. 2004) and is located around coordinates, RA = 2^h and Dec. =−4° (see Fig. 1). Pierre et al. (2007) present the source list obtained from the observations of the first 5.5 deg² (pertaining to 45 XMM pointings) in the 0.5–2 and 2–10 keV bands and above a detection likelihood of 15 in either bands. They also provide the list of optical objects extracted from the CFHTLS catalogue and located within a radius of 6 arcsec around each X-ray source. The layout of the XMM–LSS pointings (big circles) is shown in Fig. 1. We extracted from this catalogue the list of X-ray sources having optical counterparts from the CFHTLS brighter than g′= 22 mag. We thus arrived at 829 XMM sources having an optical counterpart within 6 arcsec of the X-ray source and brighter than g′= 22 mag. The positions of these sources are marked in Fig. 1 with open small circles. Also marked with crosses on these open small circles are the sources that are reliably identified in the course of this work.

Figure 1

Layout of the 45 XMM–LSS pointings (large closed circles). The positions of our initial sample of 829 sources with g′ < 22 mag and flux in the 0.5–2 keV band larger than 1 × 10⁻¹⁵ erg cm⁻² s⁻¹ are marked with open small circles. Crosses show the XMM sources for which reliable spectroscopic identifications are obtained in this work. The upper half of the XMM–LSS pointings do not have corresponding optical coverage in the CFHTLS. The larger dashed circles are the three AAT pointings.

Open in new tabDownload slide

3 OBSERVATIONS

The spectroscopic observations were performed with the AAOmega system (Sharp et al. 2006) at the 3.9-m Anglo-Australian Telescope (AAT), during the nights of 2006 September 25–27 and 2007 September11–13. AAT observations were focused on three regions and are shown as three larger circles in Fig. 1. AAOmega involves the multi-object fibre feed from the 2dF fibre positioner system (Lewis et al. 2002) linked to an efficient and stable bench-mounted dual-beam spectrograph. The fibre positioner can place 392 fibres of 2 arcsec projected diameter within a 2° field of view. Within each configuration there is a minimum target separation of 30 arcsec imposed by the physical size of the fibre buttons (Miszalski et al. 2006). The dual-beam AAOmega spectrograph, with both blue and red arms, was used in its default low-resolution configuration. In the blue arm, the 580V volume phase holographic (VPH) grating was used, providing wavelength coverage between 3700 to 5800 Å sampled at 1 Å pixel⁻¹. In the red arm, the 385R VPH grating was used, providing wavelength coverage from 5600 to 8800 Å sampled at 1.6 Å pixel⁻¹. The red and blue segments are spliced together between 5600 and 5800 Å, thus giving continuous wavelength coverage between 3700 to 8800 Å at a resolution of ∼1300. Of the 392 fibres, 25 fibres were assigned to random positions uniformly distributed over the 2 deg² to sample the background sky and eight fibre positions were assigned to guide stars within the field. For each plate configuration (each 2 degree field was observed in two to three plate configurations), the observations were broken into blocks of up to 2 hours to minimize flux losses. Each block consists of a quartz-halogen flat-field exposure, a composite arc lamp frame for wavelength calibration and three to four science exposures. Total exposure time on each target varies from 10 min to 2 h depending on their brightness. The spectra were not flux calibrated.

Reduction of the observed spectra was performed using the AAOmega's data reduction pipeline software drcontrol. The two-dimensional images were flat-fielded, and the spectrum was extracted (using a Gaussian profile extraction), wavelength-calibrated and combined within drcontrol. Redshift estimate was performed using the autoz code (this was kindly provided by Scott Croom). All spectra, identifications and redshift determinations were checked manually. The reduced optical spectra of the 487 identified XMM sources are available upon request.

4 IDENTIFICATION AND CLASSIFICATION OF X-RAY SOURCES

From the initial sample of 829 optical counterparts to XMM sources brighter than g′= 22 mag, we were able to obtain spectra for a total of 693 sources. Of these, 487 spectra (∼70 per cent) are of sufficient signal-to-noise ratio to identify their nature. Standard classification schemes (e.g. Caccianiga et al. 2008) were used to classify these objects from the spectral features detected in their spectra. The g′-band magnitude distribution of our observed sample of 693 sources (dashed histogram), the 487 spectroscopically identified XMM sources (solid histogram) and the 206 unidentified sources (hatched region) are shown in the top-left panel of Fig. 2. The bottom-left panel of Fig. 2 shows the redshift distributions of broad-line AGN (BLAGN) (cross hatched region), emission-line galaxies (ELGs) (sparsely shaded region) and absorption-line galaxies (ALGs) (darkly shaded region). In the top right panel of Fig. 2 is shown the X-ray flux (0.5–2 keV) against the optical g′-band magnitude for BLAGN (filled circles), ELGs (crosses), ALGs (filled triangles), stars (open squares) and unidentified sources (open circles). The bottom-right panel of Fig. 2 shows the 0.5–2 keV flux distribution of the observed sources (dashed histogram), the spectroscopically identified sources (solid histogram) and the unidentified sources (hatched region). The distribution of the angular separation between the X-ray and optical positions of the 487 sources is shown in Fig. 3. Also shown in this figure are the histograms of the offsets in α and δ (α_XMM−α_optical; δ_XMM−δ_optical) along with a Gaussian fit to the distributions. The fitted Gaussian function does not reproduce well the real distribution beyond 4 arcsec. 406 sources (83 per cent) are within the 4 arcsec error circle and their identification is probably secure. It is possible that some of the optical counterparts with larger offsets (>4 arcsec) are not correctly identified. The Δδ distribution is found to be symmetrically distributed around zero, whereas the Δα distribution is offset by about 0.15 arcsec. This shift in α might be due to systematics in the astrometric corrections applied in the XMM–LSS catalogue (Pierre et al. 2007). The intrinsic X-ray luminosity of these sources plotted against their redshift is shown in Fig. 4 together with the curve marking the position of a source with a 0.5–2 keV flux limit of 1 × 10⁻¹⁵ erg cm⁻² s⁻¹.

Figure 2

Top-left panel: g′ magnitude distribution of XMM sources observed in this work (dashed histogram), sources with certain identifications (solid histogram) and unidentified sources (hatched histogram). Bottom-left panel: redshift distribution of BLAGN (cross hatched region), emission-line objects (hatched region) and ALGs (shaded region). Top-right panel: plot of the X-ray flux at 0.5–2 keV (erg cm⁻² s⁻¹) band against the optical g′ magnitude for BLAGN (filled circles), ELGs (crosses), ALGs (filled triangles), stars (open squares) and unidentified sources (open circles). Bottom-right panel: X-ray flux distribution in 0.5–2 keV (erg cm⁻² s⁻¹) band for our initial sample (dashed histogram), the spectroscopically identified sources (solid histogram) and the unidentified sources (hatched region).

Open in new tabDownload slide

Figure 3

Top panels: offsets between the XMM and optical (XMM–optical) positions (left-hand panel) and the distribution of the angular separation between the XMM and the optical positions (right-hand panel) of the spectroscopically identified X-ray sources. Circles in the top-left panel have radius of 4 and 6 arcsec, respectively. Bottom panels: offset histograms for Δα (left-hand panel) and Δδ (right-hand panel). The solid line is the Gaussian fit to the distribution of offsets. Note the fit is not a good representation of the data and an offset of about 0.15 arcsec is present in the α direction.

Open in new tabDownload slide

Figure 4

The observed 0.5–2 keV X-ray luminosity of the XMM sources as a function of redshift. Filled circles are for BLAGN, crosses are for ELGs and filled triangles are for ALGs. The line shows the luminosity calculated as a function of redshift for a 0.5–2 keV flux limit of 1 × 10⁻¹⁵ erg cm⁻² s⁻¹.

Open in new tabDownload slide

Among the spectroscopically classified 487 sources, we find 265 BLAGN (∼54 per cent of the identifications), 116 ELGs (∼24 per cent of the sample), 32 ALGs (∼7 per cent of the sample) and 74 stars (∼15 per cent of the sample). Among the 116 ELGs, based on the Baldwin–Phillips–Terlevich (BPT) diagnostic diagram (Baldwin, Phillips & Terlevich 1981), emission lines in 55 ELGs are consistent with them being powered significantly by AGN and 61 by starbursts (see below). A summary of these numbers is given in Table 1. Details of these 487 objects are given in Table 2. For most of the sources with spectral lines in our spectroscopic sample, we were able to obtain a reliable estimate of the redshift. This is because for these sources, we were able to identify two or more lines in their spectra. We assign a quality flag (indicative of the reliability of the redshift) Q= 1 for those sources. For sources, with only one broad or narrow emission/absorption line in their optical spectra, the redshift estimate is not secure as it is degenerate with more than one possible redshifts. For such sources, we assign a quality flag Q= 2. For some sources, the spectral classification is unambiguous, however the spectra is of poor quality. For these sources, we have assigned a quality flag Q= 3. For 206 sources, spectral classification was not possible due to the poor quality of their spectra. Such sources have faint optical magnitudes, and this is clearly seen in the top-left panel in Fig. 2. In our 487 identified sources, 432, seven and 48 sources have quality flags Q1, Q2 and Q3, respectively.

4.1 Broad-line AGN

Objects having at least one of the emission lines (Lyα, C ivλ1549, C iii]λ1909, Mg iiλ2800, Hβ and Hα) of width >2000 km s⁻¹ are classified as BLAGN. They include type I Seyferts and quasars. These are sources, in which we are able to have an unobscured view of their central nuclear region (Antonucci 1993). We detect 265 such objects (see Table 1) in our sample corresponding to 54 per cent of the identifications. A few examples are shown in Fig. 5. Redshifts are between 0.15 < z < 3.87 with a median of z_med= 1.68 (see Fig. 2).

Figure 5

Examples of BLAGN optical spectra. In each panels, we mark the locations of prominent emission lines and the corresponding redshift of the source.

Open in new tabDownload slide

We also looked for the presence of broad absorption line (BAL) quasars in our spectroscopic sample. These BAL quasars are characterized by the presence in their spectra of strong absorption troughs blueshifted relative to the QSO emission redshift by 5000 to 50 000 km s⁻¹. Seven of the XMM sources we have identified are BALs at z > 1.5 based on the presence of strong C iv absorption. Further details on these objects will be reported in a specific paper focused only on BALs (Stalin, Srianand & Petitjean, in preparation).

4.2 Absorption-line galaxies

Objects visually identified by the 4000 Å continuum break and absorption features such as Ca ii-HK λλ3934,3968 and with no obvious presence of emission lines are classified as ALGs. We do not impose any limit on the equivalent width of emission lines. Some examples of observed ALGs are shown in Fig. 6. We have identified 32 XMM sources as ALGs. They lie in the redshift range 0.04 < z < 0.34. The possible nature of this sample of ALGs is further discussed in Section 5.2.

Figure 6

Examples of optical spectra of ALGs. Redshifts are indicated on each panels. The Na absorption line is marked and the region where Ca H&K lines are found is indicated as Ca.

Open in new tabDownload slide

4.3 Emission-line galaxies

Sources with narrow emission lines, but with no obvious AGN features in their optical spectra (e.g. high ionization and/or broad lines) are classified as ELGs. Examples of ELG spectra are shown in Fig. 7. Typical emission lines are O ii λ3727, Hβ, [O iii]λ4959,5007, Hα, etc. Other features include Ca ii–HK λλ3934,3968 absorption, the continuum break at ∼4000 Å and narrow [Ne iii]λ3869 emission. We have identified 116 ELGs in the redshift range 0.02 < z < 0.76 (median z_med= 0.26).

Figure 7

Examples of optical spectra of ELGs. The redshift of each galaxy is given in the corresponding panel. The prominent emission lines are marked.

Open in new tabDownload slide

This classification does not rule out the presence of some underlying AGN activity in ELGs. The dominant energy source (AGN or starburst) can be identified using the commonly used BPT diagnostic diagrams (Baldwin, Phillips & Terlevich 1981). The ratios of nebular emission lines ([O iii]λ5007/Hβ) and ([N ii]λ6593/Hα) are used to distinguish the ionizing source that is between thermal continuum from starbursts and non-thermal AGN continuum. The BPT diagram for our sample of ELGs is shown in Fig. 8 together with the two empirical relations commonly used to classify the emission-line objects into AGN (points above the curves) or starbursts (below the curves):

1

2

Figure 8

The BPT diagnostic diagram for the emission-line objects with [O iii], Hβ, [N ii] and Hα detected in their optical spectra. Objects above and below the lines are powered by AGN and starbursts, respectively. The solid line is that of Kewley et al. (2001) and the dotted line is that of Kauffmann et al. (2003).

Open in new tabDownload slide

Equation (1) is given by Kauffmann et al. (2003) and equation (2) by Kewley et al. (2001). However, there are sources for which we are not able to measure both the [N ii]/Hα or the [O iii]/Hβ line ratio. In such cases, an ELG is thought to be powered by an AGN if it has log([N ii]/Hα) > −0.2 or log([O iii]/Hβ) > 0.9 and by a starburst otherwise if we take Kauffmann et al. (2003) relation. On the other hand, if we consider Kewley et al. (2001) relation an ELG is thought to be powered by an AGN if it has log([N ii]/Hα) > 0.3 or log([O iii]/Hβ) > 1.0. Based on the above line ratio diagnostics and following Kauffmann et al. (2003), of the 116 ELGs, 61 are powered significantly by starbursts and the remaining 55 are significantly powered by AGN (so about 50 per cent). Alternately if we consider Kewley et al. (2001) relation, in the sample of 116 ELGs, 82 are powered by starbursts and 31 are powered by AGN. Another criterion to separate the ELGs with X-ray emission dominated by stellar processes rather than AGN activity is to look into their X-ray to optical flux ratio (f_X/f_o; Fiore et al. 2003, see Fig. 5). If f_X/f_o < 0.01 it is a normal star-forming galaxy, and if f_X/f_o > 0.1 it is an AGN (Kim et al. 2006). We define log(f_X/f_o) as

3

where f_X is the flux in the 0.5–2 keV band and g′ is the optical magnitude. The constant comes from the conversion of AB g′ magnitude into monochromatic flux and integration of the monochromatic flux over the g′ bandwidth assuming a flat spectrum. Fig. 9 shows f_X/f_o versus the intrinsic X-ray luminosity in the 0.5–2 keV band, log[L_{(0.5-2 keV)}] for the BLAGN, the two classes of ELGs found from the BPT diagram (those significantly powered by AGN and starburst) and ALGs. From the figure, it is clear that the BLAGN in our spectroscopic sample occupy a distinct location in the log(f_X/f_o) versus log(L_{(0.5-2 keV)}) plane. ELGs and ALGs of our sample mostly occupy the region with intrinsic X-ray luminosity L_{0.5-2 keV} < 10⁴³ erg s⁻¹. It thus seems that at L_{0.5-2 keV} < 10⁴³ erg s⁻¹ the ELG population in our sample is made of a mixture of normal star-forming galaxies, low-luminosity AGN and galaxies powered by starbursts with high star formation rates. In all further discussions, we however use the Kauffmann et al. (2003) relation to separate ELGs powered predominantly by AGN and starburst, respectively.

Figure 9

X-ray to optical flux ratio (f_X/f_o) plotted against the intrinsic X-ray luminosity L_{0.5-2 keV}.

Open in new tabDownload slide

4.4 Stars

As in other deep X-ray surveys, our sample also contains about 15 per cent (74 objects) of galactic stars as optical counterparts to XMM sources (at z= 0). They are typically G-, K- and M-type stars whose X-ray emission is caused by magnetic activity (Brandt & Hasinger 2005). A few spectra are shown in Fig. 10.

Figure 10

Examples of the optical spectra of stars. Their spectral types are given on each panel.

Open in new tabDownload slide

4.5 Multiple optical counterparts

In our spectroscopic sample, there are eight cases for which two XMM sources have the same optical counterpart within 6 arcsec of each of the XMM sources. The details of these eight sources are given in Table 3. In all these cases, the optical counterpart is a quasar. We also find 16 XMM sources having more than one optical counterpart within 6 arcsec of them. Spectra of all possible counterparts of the 16 XMM sources were obtained, and all are classified based on their spectral appearance. The details of these sources are shown in Table 4. Of these, the two XMM sources, XLSS J022600.1−035955 and XLSS J022630.7−050550, are associated with a quasar–galaxy pair. It is more than probable that the XMM source counterpart is the quasar. Note that XLSS J022536.4−050011 is also associated with a radio source in the National Radio Astronomical Observatory VLA Sky Survey (NVSS) with a 1.4 GHz flux density of 12 mJy (see Table 5).

4.6 Multiwavelength counterparts

4.6.1 Correlation with radio surveys

We cross-correlated the 487 newly identified XMM sources with the publicly available NVSS (Condon et al. 1998) at 1.4 GHz. Only 21 XMM sources are detected in the NVSS. We also cross-correlated our 487 identified sources with low-frequency radio observations performed in this field with the GMRT at 240 and 610 MHz (Tasse et al. 2007). There are 12 objects detected at 240 MHz, while another 22 objects are detected at 610 MHz. In total, 34 objects are detected in one of the radio bands. The details of the XMM sources with radio detections are given in Table 5.

4.6.2 Correlations with infrared surveys

There is some overlap between the XMM–LSS field and one of the fields observed by the Spitzer Wide-area Infrared Extragalactic Survey (SWIRE; Lonsdale et al. 2003). In SWIRE, observations were performed with the Infrared Array Camera (IRAC) at 3.6, 4.5, 5.8 and 80 μm and with the Multiband Imaging Photometer for Spitzer (MIPS) at 24, 70 and 160 μm to a 5σ depth of 4.3, 8.3, 58.4, 65.7 μJy and 0.24, 15 and 90 mJy, respectively (Tajer et al. 2007). Correlating our sample of optically identified XMM sources (487 in total) with SWIRE, we find that approximately 50 per cent (239 sources) are detected in SWIRE. Of these, 133 are BLAGN, 36 are stars, 50 and 20 are ELGs and ALGs, respectively. The positions of these X-ray sources in the IR colour–colour diagram (3.6–4.5 versus 5.8–8.0 μm) are shown in Fig. 11. Optical identification and description of SWIRE sources over a larger area in the CFHTLS will be reported in Stalin et al. (in preparation).

Figure 11

Positions of the various kinds of XMM sources in the IR colour–colour diagram. Here filled circles are BLAGN, crosses are ELGs, filled diamonds are ALGs and open circles are stars. The dotted line is the empirical region of Stern et al. (2005) used to separate AGN from stars and galaxies.

Open in new tabDownload slide

5 DISCUSSION

5.1 Optical-to-X-ray slope (α_ox)

The broad-band spectral index α_ox of any source characterizes the UV to X-ray spectral energy distribution by assuming that the rest-frame flux emitted at 2500 Å can be connected to the one at 2 keV with a simple power law. This is a simple measurement of the amount of X-ray radiation emitted mostly by non-thermal processes with respect to the amount of UV radiation emitted mostly by thermal processes (Kelly et al. 2007). We estimated α_ox for each of the spectroscopically identified BLAGN and ELGs (as some of the ELGs are powered by AGN) in our sample. For this, we have converted the observed i′-band magnitudes to fluxes following the definition of the AB system (Oke & Gunn 1983)

4

where and are, respectively, the flux and magnitude in the i′ band. The luminosity at the frequency corresponding to 2500 Å in the rest frame is calculated following Stern et al. (2000),

5

where ν₂ is the observed frequency corresponding to the i′ band, is the observed flux in the i′ band, ν₁ is the rest-frame frequency corresponding to 2500 Å and D_l is the luminosity distance. An optical spectral index α_o=−0.5 (Anderson et al. 2007) is assumed (S_ν∝ν^α).

The luminosity at rest frame 2 keV is obtained using a similar equation as equation (5), assuming an X-ray spectral index of α_X=−1.5 (Anderson et al. 2007).

Thus, the broad-band spectral index α_ox is obtained as

6

where L_X and L_opt are the rest-frame monochromatic luminosities at 2 keV and 2500 Å (erg s⁻¹ Hz⁻¹), respectively. The distribution of α_ox (not corrected for intrinsic and galactic absorption) for the sample of spectroscopically identified BLAGN is shown in Fig. 12. A mean value of α_ox=−1.47 ± 0.03 is found for BLAGN which is consistent with the values found in the literature (Strateva et al. 2005).

Figure 12

Distribution of α_ox for the spectroscopically identified BLAGN in our sample of XMM sources.

Open in new tabDownload slide

5.1.1 X-ray luminosity (L_X) versus optical luminosity (L_opt)

Previous studies of optically selected quasars have found a strong correlation between the optical and X-ray monochromatic luminosities (Anderson et al. 2007 and references therein). The relation between both luminosities is of the form L_X∝L^β_opt. Different values have been derived for β varying typically from 0.7 (Pickering, Impey & Foltz 1994; Wilkes et al. 1994) to 0.8 (Avni & Tananbaum 1986) and up to unity or slightly larger (La Franca et al. 1995; Green et al. 2009). The plot giving the rest-frame monochromatic luminosity at 2 keV versus the rest-frame optical luminosity at 2500 Å of all the optically identified X-ray sources is shown in Fig. 13 (top panel). It is apparent that a strong correlation exists for BLAGN. Part of the scatter in the X-ray to optical relation in Fig. 13 (top panel) might be related to the variability of the AGN as the X-ray and optical observations are far from being simultaneous, but we do not expect this effect to be very important. The fit for only the BLAGN gives

7

with a linear correlation coefficient r= 0.98. This is slightly lower than the slope of 1.12 found by Green et al. (2009), however within the range of values found by Steffen et al. (2006) through different regression methods. Note that to perform the fit we excluded radio-loud AGN that can have additional UV and X-ray flux associated with the radio jet (Wilkes & Elvis 1987; Worrall et al. 1987; Worrall & Birkinshaw 2006). Also, radio-loud AGN are found to be two to three times brighter in X-ray for the same optical magnitude (Zamorani et al. 1981; Shen et al. 2006). Therefore, we have used only the sample of BLAGN, excluding the nine AGN which are found to be detected in the radio (see Table 5).

Figure 13

Top panel: plot of the rest-frame monochromatic luminosity at 2 keV versus the rest-frame monochromatic luminosity at 2500 Å. Bottom panel: plot of α_ox against the rest-frame monochromatic luminosity at 2500 Å. Symbols have the same meaning as in Fig. 4. The solid line is the linear least-squares fit to only the BLAGN.

Open in new tabDownload slide

5.1.2 α_oxversus optical luminosity (L_opt)

The existence of a correlation between L_X and L_opt in BLAGN in turn implies that α_ox also correlates with the X-ray and UV monochromatic luminosities at 2 keV and 2500 Å, respectively. Fig. 13 (bottom panel) shows the trend of α_ox with the rest-frame monochromatic luminosity for our spectroscopic sample. For BLAGN, from a linear least-squares fit we found

8

This is similar to the value of 0.060 ± 0.007 found by Green et al. (2009), however flatter than the value of 0.137 ± 0.008 found by Steffen et al. (2006). We do not find any significant correlation between α_ox and L_opt for our sample of ELGs.

5.1.3 Evolution ofα_ox

The plot of α_ox versus redshift is shown in Fig. 14 (top panel). The fit to the data in our sample of BLAGN with α_ox=A z+B gives A=−0.021 ± 0.015 and a linear correlation coefficient of 0.07. It thus seems that for BLAGN, α_ox does not depend strongly on redshift. This is consistent with the claims of non-evolution already published in the literature (Avni & Tananbaum 1986; Vignali et al. 2003; Strateva et al. 2005; Steffen et al. 2006; Just et al. 2007; Green et al. 2009; however, see Yuan et al. 1998; Bechtold et al. 2003). We also tried to check for the evolution of α_ox for BLAGN after the removal of its luminosity dependence. For this, we subtracted the α_ox obtained through the best-fitting α_ox−L_opt regression (equation 8) from the observed α_ox. We find that this resultant Δα_ox for BLAGN shows no trend with redshift (Fig. 14, bottom panel).

Figure 14

Top panel: α_ox versus redshift for BLAGN and ELGs (as some of the ELGs are powered by AGN). The solid line is the linear least-squares fit to only the BLAGN. Bottom panel: residual in α_ox plotted against redshift for only BLAGN. Symbols are the same as in Fig. 4.

Open in new tabDownload slide

It can be seen in Fig. 14 that for z < 0.5 (i) most of the α_ox measurements are smaller than −1.6 and (ii) the scatter in the values of α_ox is much larger than for higher redshifts. The main reason for this is that the low-redshift sources are mostly ELGs (50 per cent of them being significantly powered by AGN). The optical spectrum of these objects first is significantly/predominantly contaminated by starlight from the host galaxy, and secondly the derivation of α_ox using the optical and X-ray spectral index of AGN might not be appropriate in ELGs.

An ideal sample to look for the correlations between L_X and L_opt, and α_ox with either L_X, L_opt or z is the one that fills the L–z plane. However, in practice it is difficult, and most of the samples in the literature used for such correlation searches were based on the objects that cover only a narrow range in the L–z plane. Such a sample is also likely to be contaminated with sources with unrelated physical processes, such as BALs and radio-loud AGN (Green et al. 2009). We wish to point out that the linear least-squares regression analysis presented here is a simple-minded approach compared to the detailed statistical analysis performed by Green et al. (2009) that includes upper limits. However, as discussed in earlier sections, our results are broadly in line with those available in the literature.

5.2 The absorption-line galaxies

The existence of an intriguing population of galaxies with strong X-ray emission (41 < log L_X < 44 erg s⁻¹ in the rest-frame energy range 2–8 keV; Rigby et al. 2006) was first noted from observations with the Einstein Observatory (Elvis et al. 1981). The X-ray emission reveals the presence of an AGN without other sign of activity (no emission line is seen in their optical spectrum). Further observations with the ROSAT, Chandra and XMM–Newton have supported the existence of this population (Griffiths et al. 1995; Fiore et al. 2003; Caccianiga et al. 2007; Garcet et al. 2007). Such galaxies are now referred to as X-ray bright optically normal galaxies (XBONGs).

Their exact nature is unknown, although several hypotheses have been advocated in the literature to explain their properties. It has been argued that heavy obscuration by the Compton-thick gas covering the nuclear source could prevent ionizing photons from escaping. In this hypothesis, the obscuration must occur in all directions, and should not be restricted to the torus, because the sources lack both broad and narrow lines in their spectra. This idea might be correct for at least some XBONGs (Spoon et al. 2001). Dilution of nuclear emission from the host galaxy starlight could play a role (Georgantopoulos & Georgakakis 2005). Such an effect might be important if the ground-based spectroscopic observations are performed with relatively wide slits. Severgnini, Caccianiga & Braito (2003) have shown on three objects of their sample that adequate observations (narrow slit, accurate positioning of the slit, etc.) reveal the missing emission lines. They argue that survey observations are sometimes inadequate to completely unveil the true nature of the X-ray sources. However, the intrinsic optical continuum emission of XBONGs is weak compared to that of quasars (Comastri et al. 2002) suggesting that the emission-line luminosities are also intrinsically small. XBONGs could be extreme BL Lacs in which the featureless non-thermal continuum is much weaker than the host galaxy starlight. Indeed, there is one case of an XBONG being associated with a BL Lac object (Brusa et al. 2003), and it has been argued by Fossati et al. (1998) that XBONGs could belong to the low-luminosity tail of the blazar sequence. However, large calcium break and radio quietness in several XBONGs argue against them being BL Lacs (Fiore et al. 2000). Finally, XBONGs might be powered by an inner radiatively inefficient accretion flow plus an outer radiatively efficient thin accretion disc (Yuan & Narayan 2004).

In our sample, we detect 32 X-ray sources having only absorption lines and devoid of any emission lines. This represents approximately 6 per cent of our sample of observed sources. They have redshifts ranging between 0.04 and 0.34. These galaxies are located in the peculiar regions of the IR colour–colour diagram (see Fig. 11) well outside the AGN location and also have a distinct location in the L_opt–L_X plane (see Fig. 13). We further analysed the optical spectra of these sources to see if they could be BL Lacs by looking at the shape of the continuum around the Ca H&K break at 4000 Å. The detection of a significant reduction of the Ca H&K break when compared to normal elliptical galaxies can be considered as an indication of the presence of a substantial non-thermal nuclear continuum emission. To identify the shape of the continuum around 4000 Å, we define D_n (relative flux depression across the Ca H&K break) as

9

where F⁺ and F⁻ are the average fluxes in the region 4050–4250 and 3750–3950 Å, respectively, in the source rest frame (Caccianiga et al. 2007). We adopt a limit of D_n < 40 per cent for a source with no emission line to be a BL Lac object (Caccianiga et al. 2007). Of the 32 sources in our sample, 17 have D_n < 40 per cent and hence could be BL Lac candidates. To further investigate the possibility for these 17 sources to be BL Lac candidates, we looked for any sign of radio emission from these objects (this is another indication of the presence of non-thermal emission) both in the literature and in the NVSS. Three among the 17 sources with D_n < 40 per cent are found to be radio emitters. It is thus likely that these three objects (XLSS J022509.7−050950, XLSS J022609.9−045805, XLSS J022659.2−043529) are BL Lacs. It is also probable that emission lines in ALGs are weak and could be revealed by deeper observations. It would be interesting to conduct follow-up observations of our sample of 32 ALGs to study this population in more detail.

6 CONCLUSIONS

We have presented the optical spectroscopic classification of 487 X-ray sources. They were drawn from an initial sample of 829 X-ray sources detected in the course of the XMM–LSS survey that overlap one of the wide fields of the CFHTLS. The sources have a detection threshold of 15 in either the 0.5–2 or 2–10 keV bands and are brighter than 22 mag in the optical g′ band. This sample of 829 X-ray sources is thus flux limited in both X-ray and optical bands. We observed 693 of these objects with the AAOmega system at the AAT of which spectroscopic identification was possible for 487 sources (our sample is therefore approximately 70 per cent complete). A large fraction of these X-ray sources were identified as BLAGN (∼54 per cent) based on the full width at half-maximum of the emission lines present in their spectra. In addition to these BLAGN, 74 sources are identified as galactic stars, 32 are ALGs and 116 are ELGs. Based on the BPT diagram, we find that the emission lines in 55 (61) of the 116 ELGs are powered significantly by an AGN (starburst). For the BLAGN, α_ox is found to correlate well with the rest-frame monochromatic luminosity at 2500 Å (see also Just et al. 2007; Krumpe et al. 2007; Gibson, Brandt & Schneider 2008; Green et al. 2009). The rest-frame X-ray and UV monochromatic luminosities at 2 keV and 2500 Å, respectively, are found to be closely correlated for BLAGN. No dependence of α_ox with redshift is found. This is similar to the results by Just et al. (2007) and Green et al. (2009) but in contrast to Yuan et al. (1998) and Bechtold et al. (2003) studies who found a correlation of α_ox with redshift. We detect 32 X-ray-emitting galaxies with no sign of AGN activity in their optical spectra. They are found to be located in a well-defined part of the IR colour–colour diagram. It would be very interesting to perform more focused observations to investigate the true nature of what seems to be a particular population of X-ray sources.

We thank the anonymous referee for his/her critical comments that led to a significant improvement of this paper. We also thank the present and former staff of the Anglo-Australian Observatory for their work in building and operating the AAOmega facility. This work used the CFHTLS data products, which are based on observations obtained with MegaPrime/MegaCam, a joint project of CFHT and CEA/DAPNIA, at the CFHT which is operated by the National Research Council (NRC) of Canada, the Institut National des Science de l'Univers of the Centre National de la Recherche Scientifique (CNRS) of France, and the University of Hawaii. This work is based in part on data products produced at TERAPIX and the Canadian Astronomy Data Centre as part of the CFHTLS, a collaborative project of NRC and CNRS. CSS, PP and RS gratefully acknowledge support from the Indo-French Center for the Promotion of Advanced Research (Centre Franco-Indien pour la Promotion de la Recherche Avance) under contract no. 3004-3.

REFERENCES