THE ChaMPlane BRIGHT X-RAY SOURCES—GALACTIC LONGITUDES l = 2°–358°

CBS 7 has a hard spectrum that is best fit with a power law with Γ = 1.0 ± 0.1; fits with thermal models only place a lower limit to the plasma temperature of kT ≳ 50 keV. Photometry of the optical source inside the error circle gives log (F_X/F_R)_u = 1.3 ± 0.1. The VRI colors of this source are blue compared to the bulk of the surrounding stars, and there is marginal evidence for excess Hα emission with Hα − R = −0.5 ± 0.2. We present optical follow-up spectra that confirm the Hα emission lines in CBS 7 elsewhere (Servillat et al. 2012). These characteristics are typical for accreting binaries with compact objects. Using the derived N_{H, X}, we find a distance d = 1.2 ± 0.5 kpc, and an absolute magnitude M_V ≈ 10.0. We conclude that CBS 7 has a low-mass, unevolved donor and is likely a CV rather than a quiescent wind-accreting Be X-ray binary. Sources belonging to the latter category can have similarly hard X-ray spectra but their massive donors are much brighter in the optical.

The X-ray light curve of CBS 7 is clearly variable, with recurring deep and shallow dips (Figure 3). We ran a Lomb–Scargle period-search algorithm (Scargle 1982) on the barycenter-corrected photon arrival times. Hong et al. (2012) give details of the timing analysis. The power-density spectrum between 20 s and 20,000 s shows peaks with >99% significance at P₁ = 4392 ± 290 s and P₂ = 8772 ± 957 s. The longer period corresponds to the spacing of the deep dips, while the shorter one is the spacing between the deep and shallow dips. Figures 4(a) and (b) show the Chandra light curves folded on P₁ and P₂. We examined the time-resolved spectral parameters using quantile analysis, but see no indication for variations correlated with count rate. This could be partly due to poor statistics.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

CBS 7 is also serendipitously included in an XMM-Newton observation taken on 2003 August 14 (ObsID 0152780201; 81 ks). We retrieved the data from the archive and performed further processing with SAS 10.0 following instructions in the XMM-Newton ABC Guide¹¹ and the SAS Threads.¹² After filtering out time intervals with background flares, ∼37 ks (EPIC-PN), 59.2 ks (EPIC-MOS1), and 62.1 ks (EPIC-MOS2) of exposure time remains. Periods consistent with the values of P₁ and P₂ show up as significant peaks in the periodograms of the PN light curves. For the MOS1 and MOS2 data only the shorter period is significantly detected, but the folded light curves do not show convincing variability. On the other hand, folding the PN and MOS count rates on P₂ and P₂/2 produces light curves that are qualitatively similar to the Chandra light curves and give a smoother result than the periods derived from the XMM-Newton data (Figure 4(c)).

To investigate the spectrum, we focused on the data from the PN camera given its larger effective area compared to the MOS cameras and Chandra/ACIS at higher energies. We extracted counts from a 20'' source aperture and corrected for background using a nearby source-free region on the same chip. Fitting a thermal-bremsstrahlung model to the PN spectrum, grouped to have at least 20 counts bin⁻¹, sets a lower limit to the temperature of kT  >  30 keV. A power-law fit gives parameters that are consistent with those in Table 2, viz., Γ = 1.30 ± 0.06 and N_{H, X} = (2.4 ± 0.2) × 10²¹ cm⁻² (χ²_ν = 1.07, 91 dof). The XMM-derived value for N_{H, X} is thus better constrained than the value obtained by fitting the Chandra spectrum, and gives a distance of 1.4 ± 0.1 kpc; this is the value we include in Table 4. Note, however, the caveat in Section 4.1.1 regarding our distance estimate for this source. The residuals show a systematic excess between 6 and 7 keV compared to this model, which prompted us to add a Gaussian-shaped emission line at a fixed position of 6.4 keV or 6.7 keV to mimic an Fe Kα fluorescent or He-like emission line. While the former does not give sensible line parameters, including a 6.7 keV line removes the systematic residuals. In the latter case, we find a marginal (95%) significance for the presence of the line using the method of Bayesian posterior predictive probability values (e.g., Protassov et al. 2002). The spectrum can be described by Γ = 1.36 ± 0.07, N_{H, X} = (2.6 ± 0.3) × 10²¹ cm⁻², and an FWHM of 0.6 ± 0.2 keV (χ²_ν = 0.98, 89 dof). The unabsorbed flux of F_{X, u} = 6.3 × 10⁻¹³ erg s⁻¹ cm⁻² (0.3–8 keV) is about 25% higher than found with Chandra.

CBS 17 is detected ≲10 pixels away from the chip edge, so we use caution when considering its properties derived from the Chandra data. The hard power-law spectrum (Γ = 0.8 ± 0.2; or kT > 43 keV for a thermal model) suggests that this source is accretion powered. TiO absorption bands (5847–6058 Å, 6080–6390 Å, 6651–6852 Å) are clearly visible in our Hydra spectrum of the candidate counterpart, and their strength points at a late-K or early-M spectral type. The absence of the gravity-sensitive CaH lines around 6382 and 6389 Å  indicates that this star is a giant (Figure 5). Hα is seen in emission, which supports the true association with the X-ray source. On these grounds we conclude that CBS 17 is likely a symbiotic binary where an evolved late-type star transfers mass to a hotter companion, in many cases a white dwarf. Many "canonical" symbiotics are thought of as wind-accreting systems as the strong wind from the evolved companion manifests itself through nebular emission lines excited by the high-energy photons created by the accretion process. We do not see such prominent emission lines in CBS 17. Mass transfer would therefore have to take place through Roche lobe overflow.

Zoom In Zoom Out Reset image size

The observed NIR colors of CBS 17 are only consistent with the spectral classification if the extinction is lower (N_H ≈ 3 × 10²¹ cm⁻²) than the one derived from X-ray spectral fitting (Figure 2(b)). A lower value for N_H of about 3.5 × 10²¹ cm⁻² is also suggested by comparing the continuum slope of the Hydra spectrum to artificially reddened template spectra. Formally, the error bar on N_{H, X} allows such low values (the difference is ∼2.3σ), but the discrepancy can also indicate that there is material inside the system obscuring the X-ray emitting region but not the late-type giant. Therefore, we do not use N_{H, X} to estimate a distance, but adopt N_H = (3.25 ± 0.25) × 10²¹ cm⁻² to calculate a spectroscopic distance based on the allowed range of spectral type. A K7 III giant has M_V = +0.4, while an M1 III giant has M_V = −0.2, resulting in d = 4.1–5.6 kpc, and log L_X = 33.2 ± 0.2.

CBS 17 is included in three XMM-Newton observations. It is commented on by Angelini & White (2003, AW03; their source 1) as a serendipitous detection and possible AGN that stands out in a 6.2–6.8 keV Fe-band image of the first of these observations. Kaaret et al. (2006) also list CBS 17 among the serendipitous detections in the first and second XMM-Newton observation, as well as in the Chandra observation analyzed here. Their reported fluxes indicate long-term variability but are based on the assumption (for all their sources) of a power-law spectrum with Γ = 1.5 and N_H = 3.1 × 10²¹ cm⁻². This motivated us to do a more detailed analysis. Table 5 lists all X-ray observations for CBS 17.

We extracted spectra following the same procedures as for CBS 7 and restricted the analysis to the EPIC-PN data. First, we fitted each spectrum individually. The results, summarized in the upper part of Table 6, indicate spectral and/or flux variability but the errors are large. Therefore, we also tried fitting all spectra simultaneously, forcing various combinations of parameters to be the same for each epoch. Keeping N_{H, X} and the power-law slope and normalization the same gives an unacceptable fit with χ²_ν = 1.81 and 278 dof for the overall fit, which confirms the variability. Allowing only N_{H, X} to vary for each epoch does not result in a good fit either (overall χ²_ν = 1.69, 275 dof). The same is true if we fix N_{H, X} and Γ but allow the normalization to vary; in this case an unsatisfactory fit is found for epochs 1 and 2. If we let Γ and the normalization vary and keep N_{H, X} the same, a good fit is found for each epoch (Table 6, bottom). In this case, we see variations of Γ, and the intrinsic flux varies up to ∼60%. The largest flux change is seen between two XMM-Newton observations (epochs 1 and 3), so this finding is unaffected by the fact that CBS 17 lies close to the chip edge in the Chandra observation.

Figure 6 shows the data and best model fits for fixed N_H. The spectrum from epoch 1 shows small positive residuals between 6 and 7 keV that could correspond to the Fe K line reported by AW03. The good fits achieved with models without an Fe K line do not warrant adding such an emission component. When we do include a Gaussian-shaped emission line, we find parameter values that are consistent with the results of AW03, viz., 6.24 ± 0.15 keV and 0.66 ± 0.24 keV for the line center and width, Γ = 1.7 ± 0.2, N_{H, X} = (0.9 ± 0.2) × 10²² cm⁻², and F_{X, u} = 4.8 × 10⁻¹³ erg s⁻¹ cm⁻² (χ²_ν = 0.91 for 39 dof). Here we re-grouped the spectrum to ⩾15 (instead of 20) counts bin⁻¹ for a slightly better resolution. Only the spectrum from epoch 1, during which the source was at its softest, shows a hint of a line. The absence of this line in the Chandra data could be due to the poorer sensitivity at higher energies.

Zoom In Zoom Out Reset image size

Follow-up studies are needed to uncover the nature of CBS 17. With high-resolution optical spectroscopy the binary status can be established, and the orbital period and companion masses can be constrained. To test if the companion is (close to) Roche lobe filling, one can look for the signature of a distorted companion in the optical/NIR light curves.

Our optical spectra show that CBS 3, 4, 11, 12, 20, and 21 are G or K stars (Table 3). We have not obtained spectra for the candidate counterparts to CBS 6 and 10, yet. Considering their dereddened 2MASS colors, they are probably late-type stars as well, with spectral types around early G and early M, respectively (Figure 2(b)). The X-rays of late-type stars are emitted by a hot coronal plasma, which is confined by magnetic fields that are generated by a solar-like dynamo operating in the convective outer layers. See Güdel (2004) for a review.

The X-ray-derived extinction values for these eight stars are all relatively low, and consequently the Drimmel extinction-versus-distance curves place them nearby. Our estimates or upper limits on L_X are consistent with X-ray luminosities of nearby F–K stars (log L_{0.1-2.4 keV} ≈ 26.5–29.5; Schmitt & Liefke 2004) and active binaries (log L_{0.1-2.4 keV} ≈ 28–32; Dempsey et al. 1993, 1997) as measured by ROSAT, and of active binaries in open clusters (log L_0.3-7 keV ≈ 28–31; van den Berg et al. 2004) as measured by Chandra. If we adopt the N_{H, X}-derived distances to calculate absolute magnitudes M_V or M_R, we find that CBS 4, 10, 11, and 21 are likely main-sequence stars or BY Dra-type active binaries, and CBS 3 is likely a subgiant or an RS CVn-type active binary. CBS 12 and 20 are dwarfs or subgiants, either single or in an active binary. While one should be wary when using our N_{H, X}-derived distance estimates, we note that these sources are relatively nearby, and that non-flaring active stars or binaries are often not internally absorbed. Low-resolution stellar X-ray spectra are often found to be better fitted with a sub-solar coronal metal abundance Z than with a solar abundance (Güdel 2004). We also see this in the results of our fits to the spectra of CBS 3, 6, 10, 11, 12, 20, and 21.

With kT ≈ 4 keV and log (F_X/F_R)_u ≈ −0.9, CBS 4 appears to be more active than most of the sources in this category, which have kT ≲ 1 keV. Here we could be seeing the effect of a coronal flare during the first ∼20 ks of the observation. To study temporal variations in the energy spectrum of CBS 4, we resort to quantile analysis as there are too few counts to fit, and compare, spectra extracted from different sub-intervals of the exposure. In quantile analysis, the energy values that divide the energy distribution of the photons in certain fixed fractions are used as spectral diagnostics. This approach is more powerful than measuring the number of counts in fixed energy bands (as is done when using hardness ratios), as the errors on the diagnostic are less sensitive to the underlying spectral shape. Following Hong et al. (2004) we choose the median energy (E₅₀), and the 25% and 75% quartiles (E₂₅, E₇₅) to characterize the spectra. By comparing the observed quantiles with those expected for a spectral model of choice, the X-ray spectral parameters and N_{H, X} can be constrained. As expected for a coronal flare, the spectrum of CBS 4 is harder when it is bright compared to quiescence, where the spectrum can be described by emission from a ∼1 keV thermal plasma. For intermediate count rates (∼10 counts ks⁻¹) the source moves down in the quantile diagram which is the signature of two different temperatures contributing more or less equally. Flaring has been frequently observed in M dwarfs, but also in earlier-type dwarfs (see Güdel 2004 and references therein). What is puzzling is the trend of spectral hardening toward the end of the observation, when the count rate is at its lowest.

CBS 3 is also an X-ray variable, showing a flare-like event near the end of the exposure (Figure 3). We see no indication for a change in its X-ray spectrum during the flare.

CBS 12 is potentially a long-term variable. Besides in the observation we analyze here, it is detected in ObsID 757 (2000 August 14, 13.4 ks GTI) with a significantly higher count rate: 9.8 ± 1.2 versus 4.3 ± 0.2 counts ks⁻¹. The energy quantiles in the two ObsIDs are consistent.

The bright star that is matched to CBS 5 is HD 97434. The O7.5III classification by Walborn (1973) is in better agreement with our Hydra spectrum than other classifications found in the literature. Using its colors and spectral type, Vazquez & Feinstein (1990) derive A_V = 1.5 ± 0.2 and M_V = −5.7 ± 0.15, where the errors were estimated by us (since none were given by Vazquez & Feinstein) and set to reasonable values based on the information provided in their study, and the compilation of absolute magnitudes of O stars in Vacca et al. (1996). This gives a distance to CBS 5 of d = 2.8 ± 0.4 kpc. The X-rays from single early-type stars are believed to arise from shock-heated plasma formed by instabilities in the strong winds emanating from such massive stars. The spectra can be described by the combination of two thermal components with kT ≈ 0.3 and 0.7–1 keV (Güdel & Nazé 2009). This agrees with our results. We find log (F_X/F_R)_u ≈ −4.0, and log (L_X/L_bol) ≈ −6.3 (for d = 2.8 kpc), also typical for O stars.

Bhatt et al. (2010) analyzed XMM-Newton data of HD 97434 taken two months before the Chandra observation. To facilitate comparison, we fit our spectrum with their adopted model, i.e., a 2T APEC thermal plasma model (xsapec). Our data are not good enough to constrain Z. Setting Z/Z_☉ ≡ 0.21, i.e., the value found by Bhatt et al., we find N_{H, X} =(3 ± 1) × 10²¹ cm⁻², kT₁ = 0.24 ± 0.05 keV, and kT₂ = 0.6 ± 0.1 keV, suggesting no change in these parameters between the epochs. Bhatt et al. assume that HD 97434 is a member of the open cluster Trumpler 18 located at 1.55 ± 0.15 kpc (Vazquez & Feinstein 1990). However, the spectroscopic distance (see above) places it behind the cluster by ∼3σ. When adopting a common distance, the Chandra luminosity is two to three times lower than the XMM-Newton value, indicating long-term variability.

The temperature derived for CBS 19 (kT = 4⁺²_{− 1} keV) indicates a high level of activity. Our Hydra spectrum of the candidate counterpart shows a clear Hα emission line with equivalent width (EW) = −6.0 ± 0.8 Å. At the poor S/N of our spectrum, the continuum looks featureless except for the Na I D doublet around 5890 Å, which could be interstellar in nature. This makes spectral classification challenging. At first sight, the optical and NIR colors of the candidate counterpart seem to disagree (Figure 2). The former suggest a spectral type of late-F to early-M, whereas the latter point at a late-M star. Such cool M stars show prominent TiO absorption bands in their optical spectra, but these are notably absent in CBS 19.

One explanation is that CBS 19 is a young PMS star surrounded by a warm, circumstellar disk that causes excess NIR emission. In fact, after accounting for the extinction derived from N_{H, X}, the location of CBS 19 in the NIR color–color diagram (Figure 2(b)) agrees with the locus of the classical T Tauri stars (CTTS), a subclass of PMS stars that are still actively accreting from their circumstellar disk (Meyer et al. 1997). The prominent but relatively weak (compared to CVs) Hα emission is also in line with this classification. The X-ray properties of CBS 19 are typical for PMS stars, where X-rays are generated by a high level of magnetic activity, with a possible contribution of (relatively soft) X-rays from the accretion process (e.g., Feigelson et al. 2007). We note that the disk can be a source of local extinction, and therefore our estimates of the distance (d = 3.0 ± 0.8 kpc) and luminosity (log L_X = 31.8 ± 0.3 erg s⁻¹) based on the Drimmel model may be too high.

The Li i 6708 Å line is a well-known indicator of youth. Detection of this line in high-resolution optical spectra of CBS 19 would confirm its T Tauri nature.

Our Hydra spectrum of the candidate counterpart to CBS 8 classifies it as an early/mid-K star. The X-ray spectrum of CBS 8 is rather hard for steady, stellar coronal emission. For a power-law model we find Γ = 1.9 ± 0.2; for a bremsstrahlung model we find kT = 5.5^+2.3_{− 1.3} keV. The count rate is stable throughout the 48 ks observation, so it is unlikely that the hard spectrum is the result of a typical coronal flare which can have a decay time of up to ∼15 ks (Güdel 2004). CBS 8 appears in our database four more times, always with a lower count rate (at a level of >3σ). The largest difference is found with respect to ObsID 3807 (2002 September 24, 23 ks GTI) where the source is detected with 2.7 ± 0.4 counts ks⁻¹ compared to 10.7 ± 0.5 counts ks⁻¹ in the observation analyzed here, taken a year later. In ObsID 3807 the source appears softer, with E₅₀ = 1.2 ± 0.1 keV compared to E₅₀ = 1.64 ± 0.04 keV. The hardening and brightening of the source could point toward an increased level of coronal activity occurring on a ∼one year timescale.

The distance derived from N_{H, X} gives an absolute magnitude M_V = 0.4^+1.2_{− 0.9}, pointing to the star being a luminosity-class III giant (for a K3 III star, M_V = +0.8). Also from Figure 2 it can be seen that the spectral type better agrees with the optical/NIR colors if we assume that the star is a giant instead of a dwarf. The Hα absorption line is very weak, a sign that the line may be filled in by excess Hα emission. We tentatively classify CBS 8 as an active RS CVn-type binary.

The X-ray spectrum of CBS 14 is well fitted with a 2T MeKaL model with kT₁ ≈ 0.9 keV and kT₂ ≈ 2.5 keV, and minimal absorption (N_{H, X} < 2 × 10²⁰ cm⁻²). Apart from small-amplitude variations, the count rate steadily declines during the ∼9.5 ks observation. The spectrum becomes softer when the source gets fainter (Figure 7). This is typical for a coronal flare, but since the source was already bright at the start of the observation it is impossible to say if the light curve shows the characteristic "fast rise, slow decay" profile of a flare.

Zoom In Zoom Out Reset image size

The error circle of CBS 14 contains an extremely red object with V − K_s = 9.4. Its optical/NIR colors place this object among the dwarfs with a spectral type of M7 or later; these are the so-called ultracool dwarfs. This is illustrated in the color–color diagrams of Figures 2 and 8, which also include nearby ultracool dwarfs. By comparing its J magnitude, and I − J and J − K_s colors with those of ultracool dwarfs at known distances (Phan-Bao et al. 2008), we estimate an approximate spectral type of M9 and d ≈ 24 pc.

Zoom In Zoom Out Reset image size

The area around CBS 14 is included in the field of view of the EPIC-MOS cameras in two XMM-Newton observations: 0093670501 (2001 March 2; 14 ks) and 0207300201 (2004 February 22; 34 ks); the source lies outside the field of the EPIC-PN camera. After filtering out background flares, ∼13 ks and ∼15 ks of exposure time remain. No source is seen at the location of CBS 14. Upper limits on the flux are estimated by first extracting all counts (0.3–8 keV) from within 20'' of the Chandra source position. Using Gehrels (1986), we compute the 3σ upper limit on the number of counts. The background contribution is estimated from an annulus around the source position. The resulting limit on the net source counts is converted to a count rate limit using the value of the exposure map at the source position. With the spectral model of Table 2, we estimate that the most restrictive 3σ flux limit comes from the 2001 observation and is ∼1.6 × 10⁻¹⁴ erg s⁻¹ cm⁻². It is possible that the spectrum is softer when the source is not flaring, but for a kT = 0.5 keV MeKaL spectrum the upper limit on the flux changes only 10%. The flux from the Chandra observation is uncertain due to the pileup. A conservative lower limit is obtained by assuming that the detected count rate is the incident count rate. This implies an unabsorbed flux of >1.6 × 10⁻¹² erg s⁻¹ cm⁻² (0.3–8 keV), meaning that CBS 14 was ≳100 brighter during the Chandra observation compared to the time of the XMM-Newton observations. Therefore, our estimate of (F_X/F_O)_u is very uncertain as our X-ray and optical observations were not simultaneous.

CBS 14 was not detected in the ROSAT All Sky Survey. The detection limit of ∼0.015 counts s⁻¹ (0.1–2.4 keV; Huensch et al. 1998) gives an upper limit to the intrinsic flux of ∼1.7 × 10⁻¹³ erg s⁻¹ cm⁻² for a kT = 0.5 keV MeKaL model and N_{H, X} = 2 × 10²⁰ cm⁻². The lower limit to the Chandra flux is 1.4 × 10⁻¹² erg s⁻¹ cm⁻² when extrapolated to the ROSAT band, implying a flux increase of at least a factor of ∼8.

Spectroscopic follow-up is needed to test our tentative classification of CBS 14 as an ultracool dwarf. It is not obvious what an alternative explanation could be. The X-ray spectrum points at negligible extinction and excludes the option of a background AGN or a heavily obscured object to explain the red optical/NIR colors. The number of random interlopers inside the area searched for counterparts is N_ran = 0.5, so there is a reasonable chance that the red object is a chance coincidence. This implies that the true counterpart has R > 24.

The hard spectrum of CBS 7 suggests it is a magnetic CV (e.g., Heinke et al. 2008), although some dwarf novae in quiescence have hard spectra as well (e.g., Balman et al. 2011). Magnetic CVs can be divided in two subclasses. In polars, the magnetic field is strong enough (B ≳ 10 MG) to lock the spin to the orbital motion. Intermediate polars (IPs) have fields of moderate strength (B ≈ 1–10 MG), which are too weak to force synchronism. It is not clear which of these classes CBS 7 belongs to. Our follow-up optical and NIR data indicate that the longer of the two X-ray periods is likely the orbital period (P_b) and we refer to Servillat et al. (2012) for a detailed discussion. The origin of the shorter period is less clear. If the system is synchronized, the variability could be caused by the two poles or accretion spots on the white dwarf being partially (self)eclipsed. If the system is asynchronous, the shorter period could reflect the white dwarf spin period (P_s), which would then be half the orbital period by chance. The ratio of the periods (P_s ≈ 0.5 P_b) is relatively high compared to the typical ratio for IPs; most have P_s ≲ 0.3 P_b. It would add CBS 7 to a small but growing number of "near-synchronous" IPs that may be evolving toward synchronism (Norton et al. 2008; Hong et al. 2012). As the X-ray spectral properties of CBS 7 do not vary with count rate, the dips are likely not caused by a variation of the local amount of absorbing material. It is not uncommon for magnetic CVs to be internally absorbed, which can cause our distance and L_X estimates to be overestimated.

Traditionally, symbiotic binaries were thought of as soft X-ray emitters. This picture was mostly based on ROSAT observations that were limited by its soft response (Mürset et al. 1997). Recently, an increasing number of symbiotics have been detected at harder energies (Kennea et al. 2009; Luna et al. 2010; van den Berg et al. 2006). Kennea et al. (2009) suggest that for RT Cru, T CrB, CD−57 3057, and CH Cyg the hard X-rays are thermal and come from an accretion-disk boundary layer around a massive, non-magnetic white dwarf. The potentially high white dwarf masses make them interesting as candidate Type Ia supernova progenitors. CBS 17 could be a member of this class of "hard" symbiotics. Besides the hard spectrum, similarities with this class include the weak optical emission lines and possibly an internal absorption component. On the other hand, our analysis suggests that the long-term variability is not caused by variations in the column density, whereas in the systems discussed by Kennea et al. (2009) the variations were found to result from variations in the intrinsic absorption.

Unlike earlier-type M stars that have a radiative core and convective envelope, ultracool dwarfs are fully convective with cool and effectively neutral atmospheres. Multi-wavelength diagnostics indeed mark a change in magnetic activity around spectral type M7, but a larger sample is needed to get a clearer picture: only 12 ultracool dwarfs have been detected in X-rays (Berger et al. 2010; Robrade et al. 2010). In this context, it is important to establish the nature of CBS 14. Its colors place it between the M9 and L0 dwarfs. So far, there is no firm X-ray detection of a dwarf later than M9; the L2-dwarf Kelu-1 with a marginal detection of four photons is the only possible exception (Audard et al. 2007). With log L_0.3-8 keV/L_bol ≳ −0.7 (averaged over the observation, for a spectral type M 9) and a duration of at least 8 ks, CBS 14 could have been caught during one of the strongest flares seen in ultracool dwarfs. The upper limit on the quiescent luminosity from the XMM-Newton data is 1 × 10²⁷ erg s⁻¹ (0.3–8 keV). This is consistent with the limits for other M9 stars but not very constraining as typical quiescent luminosities are a few times 10²⁶ erg s⁻¹ (Berger et al. 2010; Robrade et al. 2010).

We estimate the expected number of AGNs in our sample using the cumulative X-ray point-source density as a function of flux (log N–log S distribution) derived for high Galactic latitudes. To this end, we determine the flux limit of an observation by calculating the flux of a source observed at the aimpoint with 250 net counts (0.3–8 keV). Here we assume a Γ = 1.7 power-law spectrum and a column density equal to the integrated Galactic N_H along the line of sight. We do this for each of the 63 observations that we searched for bright sources; the values for the observations analyzed here are included in Column 10 of Table 1. The average flux limit of an observation is higher than the value thus computed as the sensitivity decreases with offset angle from the aimpoint. As a first-order correction for the vignetting, we multiply the flux limits for the aimpoint by the ratio of the value of the exposure map at the aimpoint and the mode of the exposure map. This gives a correction factor of 1.09 for ACIS-S and of 1.04¹³ for ACIS-I. We find that the number of AGNs predicted is rather uncertain. The log N–log S curves (0.3–8 keV) from Kim et al. (2007) predict a total of 19–57 AGNs, where the range is defined by the 2σ error margins in the parameters of the log N–log S equation. If we use the extinction maps from Schlegel et al. (1998), which on average predict a higher integrated Galactic column density, the number of expected AGNs is 13–42. In fact, we have one confirmed AGN and six candidate AGNs (Sections 3.3 and 3.4). However, if we take into account statistical errors (which are not included in the ranges given above), and the possibility that the Galactic extinction is still underestimated (neither the Drimmel nor the Schlegel extinction models have been verified externally in an extensive way), the predicted and observed number of AGNs are not so discrepant as they might seem at first sight. In any case, it is likely that most of the sources in Section 3.4 are indeed AGNs.

Using our classifications to separate the Galactic sources in our sample from the extragalactic sources (including the six AGN candidates), we have computed the cumulative source density versus flux for the Galactic sources only. We do this by adding up the contribution of each source and normalizing it by the area in which it could have been detected based on the flux limits of each of the 63 observations searched. As in this case we only consider Galactic sources, we have used a different spectral model to calculate the flux limits than in Section 4.2.1. To account for the typically softer spectra of the Galactic sources (see Table 2), we use a power law with photon index Γ = 3 and correct for only 25% of the integrated N_H along the line of sight. This value of Γ is only appropriate for coronal sources, but we find that Γ = 1.7 and correcting for the total integrated N_H give the same results. The log N–log S curve for the 0.3–8 keV band is shown in Figure 9. At the high end of the flux range (≳ 5 × 10⁻¹³ erg s⁻¹ cm⁻²) lie the accreting sources CBS 7 and CBS 17, the early-type star CBS 5, the highly variable source CBS 14, and the stellar coronal source CBS 10. The nine sources below this limit are all likely late-type active stars.

Zoom In Zoom Out Reset image size

We do not have enough statistics to do a detailed analysis, but can make a qualitative statement. The slope α of the curve log N(> S)∝S^−α appears flatter than the slope for an isotropic distribution (α = −1.5) that is expected for a truly homogenous angular distribution like that of AGNs, or for a population of local sources that is sufficiently close as to seem isotropic. The slope is closer to the value expected for a disk distribution (α = −1). The derived distances indeed place our Galactic sources up to 3–5 kpc away (Table 4). Other Galactic plane surveys find a similar flattening of the Galactic log N–log S distributions (e.g., Hertz & Grindlay 1984), which for the XGPS is mostly accounted for by soft stellar coronal sources (Motch et al. 2010). The resulting Galactic source density is roughly similar to the value from the XGPS: we find about 6–15 sources deg⁻² above a limiting flux of 5 × 10⁻¹⁴ erg s⁻¹ cm⁻² (0.3–8 keV) compared to ∼3 sources deg⁻² (0.4–2 keV) and ∼15 sources deg⁻² (2–10 keV) in the region studied by the XGPS (Motch et al. 2010). Galactic sources are outnumbered by AGNs above this flux limit (by a factor of three to seven according to Kim et al. 2007).

In this study, we have avoided the central, most obscured part of the plane, which enabled us to use the properties of the optical and NIR counterparts for source classification. Would Chandra have detected our sources if they were located in the central bulge, and if so, can we use them as templates to constrain the nature of more distant obscured sources? We answer this question for the specific case of the 10' × 10' region around Sgr A*—the Galactic center region (GCR)—where Chandra found a large number (thousands) of X-ray point sources, most of which remain unidentified at other wavelengths. We apply the canonical extinction for the GCR of N_{H, X} = 6 × 10²² cm⁻² to our intrinsic source fluxes and put them at a distance of 8.5 kpc. The deep pointings of the Sgr A* field reach a sensitivity of at least 5 × 10⁻¹⁵ erg s⁻¹ cm⁻² (S/N ⩾ 3 between 2 and 8 keV for 700 ks of stacked exposures; Hong et al. 2009). All (candidate) AGNs except CBS 2, the two (likely) Galactic accretion-powered sources CBS 7 and CBS 17, and the RS CVn binary CBS 8 would fall above this X-ray limit. Only two sources would be bright enough in the NIR to be detectable in our GCR K_s-band observations (Laycock et al. 2005), viz., CBS 8 and CBS 17. With K_s ≈ 15 and 14.1, respectively, they would lie above or close to the NIR confusion limit, which is K_s = 15.4 in most of the region, and K_s = 14.5 in the most crowded region within 1' from Sgr A*. With higher-resolution images, obtainable with adaptive optics imaging, they would be easily detectable. However, as their spectra do not show strong signatures of activity or accretion (not in the optical, at least), it is difficult to distinguish them from the random coincidences with late-type evolved stars, which abound in the old bulge. While Laycock et al. (2005) showed that at most 10% of the unidentified sources around Sgr A* have such bright counterparts, it is important to recognize that it is not only young wind-fed quiescent Be-HMXBs, which have received much of the focus so far, that contribute to this NIR "bright" population of X-ray sources, but also active binaries of the RS CVn-type and symbiotic binaries.