AN HST/COS OBSERVATION OF BROAD Lyα EMISSION AND ASSOCIATED ABSORPTION LINES OF THE BL LACERTAE OBJECT H 2356–309

The most accurate measurement of the redshift of H 2356–309 (R.A.: $\rm 23^d59^m07\buildrel{\mathrm{s}}\over{.}9$; decl.:$\rm -30^h37^m40\buildrel{\mathrm{s}}\over{.}9$) was given by the 6dF Galaxy Redshift Survey (Jones et al. 2004, 2009). The redshift, z = 0.16539 ± 0.00018 or cz = 49582 km s⁻¹, was determined by cross-correlating the host galaxy spectrum with a set of predefined template spectra following the procedure used for the 2dF Galaxy Redshift Survey (Colless et al. 2001). The rms uncertainty is 47 km s⁻¹.

H 2356–309 is classified as an HBL because of its high X-ray flux (Giommi & Colafrancesco 2005). Multiband observations of H 2356–309 from radio to very high energy (VHE, E > 100 GeV) suggested that the broadband spectral energy distribution (SED) can be simply fitted with a synchrotron self-Compton (SSC) model with a double-peak structure (H.E.S.S. Collaboration et al. 2010), in which the peaks in X-ray and VHE are produced via synchrotron radiation and inverse Compton scattering, respectively.

The host of H 2356–309 is a normal giant elliptical galaxy with a profile that can be well-fitted with a de Vaucouleurs law (Falomo 1991). Interestingly, with HST/WFPC2, a blue companion was discovered about 12 away from the host galaxy, or at a separation of 4.6 kpc assuming the same redshift (Falomo & Ulrich 2000). We discuss the possible association of the detected absorption lines with this blue companion in Section 3.

H 2356–309 was observed with the HST/COS on 2013, June 15, for 17.0 ks with the medium-resolution (Δv ∼ 18 km s⁻¹), far-UV grating G130M (1135 Å <λ < 1450 Å) as HST program 12864. The main science goal of this program was to search for broad H i Lyα and highly ionized metal absorbers corresponding to foreground superstructures; we will present findings in this regard in a separate paper. The unexpected scientific result of the observations we present here is the broad Lyα emission feature intrinsic to the BL Lac object as well as a trio of narrow absorption features at approximately the same redshift.

The flux-calibrated, one-dimensional spectra for each exposure were obtained from the Mikulski Archive for Space Telescopes and combined using the standard IDL procedures described in detail by Danforth et al. (2010). The combined data show a continuum flux level of ∼1.7 × 10⁻¹⁵ erg cm⁻² s⁻¹ Å⁻¹ and a median signal-to-noise ratio (S/N) of ∼12 per seven pixel resolution element.

The absolute wavelength calibration is crucial to the measurement of emission and absorption features. We quantified the uncertainty in the wavelength calibration by comparing the low-ionization Galactic absorption features detected in our COS spectrum with the H i 21 cm emission data. In the direction of H 2356–309, Kalberla et al. (2005) shows a nice H i profile with a velocity relative to the local standard of rest (LSR) of v_LSR = −5 km s⁻¹. We then measured the centroids of as many low-ionized Galactic absorption lines (most are Si ii, Si iii, and C ii lines) as we can find in the COS data. We found that the observed centroids are consistent with the LSR frame with an uncertainty of ∼10 km s⁻¹.

To accurately determine the far-UV continuum shape, Galactic extinction must be taken into account. We normalized the data in the range 1180 Å < λ < 1250 Å with a linear continuum, masked out regions of narrow emission (geocoronal airglow) and absorption (Galactic and extragalactic absorption features), and fitted a Voigt profile to the Galactic damped Lyα line (see Figure 1, left panel). The resulting column density $\log N_{\rm H\,\scriptsize{I}}\ (\rm cm^{-2})=19.94\pm 0.03$⁶ implies an extinction of $E(B-V)=N_{\rm H\,\scriptsize{I}}/5.8\times 10^{21}\ {\rm cm^{-2}}=0.015\pm 0.01$ via Shull & van Steenberg (1985). We assume no reddening in the AGN host galaxy. We use this extinction to deredden the observed spectrum by applying the IDL routine fm_unred.

Zoom In Zoom Out Reset image size

We then redshifted the spectrum into the rest-frame of the blazar and fitted the broadband, dereddened continuum with a power law, $F_{\lambda } = F_{912} (\lambda /912{\rm \ \mathring{\rm{A}}})^{-\alpha _{\lambda }}$. To accurately characterize the continuum, we manually identified all the regions with strong absorption/emission features and excluded those regions in our fit. We also smoothed the spectrum with a bin size of seven pixels. We find a power-law index of α_λ = 1.58 ± 0.30 and a normalization of F₉₁₂ = (2.63 ± 0.16) × 10⁻¹⁵ erg cm⁻² s⁻¹. Using HST/COS, Shull et al. (2012) studied 22 AGNs and found a photon index of <α_λ > =1.32 ± 0.14 for the far UV composite spectra. This is consistent with what we found in H 2356–309. Figure 1 shows the spectrum between 1160 Å and 1450 Å in the observer's frame. The red dashed line is the continuum model. The two shadowed areas indicate the Galactic H i Lyα absorption (left) and the blazar H i Lyα emission which will be discussed later. While the overall fit is reasonably good, we clearly see deviations at some wavelength regions, particularly around 1450 Å. Such deviations are mainly caused by COS calibration uncertainties. The presence of fixed pattern noise can lead to 5%–10% uncertainties in the absolute flux calibration.⁷ Also, the detector edges are less sensitive and less well-calibrated than the middle of the detector since the spectrum is smeared out over a larger region of the detector. Such deviation will not affect our estimation of the ionizing photons at λ < 912 Å.

Due to the deviation of a single power-law fit to the broadband continuum at long wavelengths, we limited the fit to the range of 1210–1222 Å (in the rest frame of the blazar) when measuring the Lyα emission line flux. We also ignored the two H i absorption line regions at ∼1215.5 Å and 1217 Å (the spectral regions labeled with horizontal red lines in the right panel of Figure 1). We fitted the line in the rest-frame of the BL Lac and found that a single Gaussian profile fits the emission line very well (see the right panel of Figure 1). We found the line central wavelength is λ_obs = 1215.53 ± 0.14 Å, corresponding to cz = 49541 ± 40 km s⁻¹. This is consistent with the velocity of the host galaxy within 1σ uncertainty. We also found that the FWHM is 5.43 ± 1.30 Å, or 1340 ± 321 km s⁻¹, and the equivalent width is EW = −1.03 ± 0.22 Å. These values are slightly higher than those measured in Stocke et al. (2011), but at a substantially greater distance. We also derived an Lyα line intensity of (1.68 ± 0.36) × 10⁻¹⁵ erg cm⁻² s⁻¹ and a luminosity of $(9.53\, \pm \, 2.02) \times \rm 10^{40}\rm \ erg\ s^{-1}$.

Following Stocke et al. (2011), we investigate the implication of the blazar Lyα emission line on our understanding of the BLR around the nucleus. We begin by calculating the expected Lyα flux under the "nebular hypothesis." We assume the ionizing flux is isotropic and the BLR clouds are optically thick to the ionizing photons. Using case-B recombination, the emission lines such as Hβ can be estimated from the total number of the ionizing photons; Lyα luminosity can be calculated by the ratio of the specific intensity between Lyα and Hβ (see Stocke et al. 2011 for detailed equations). We extrapolate the fitted continuum beyond the Lyman limit and estimate a Lyα emission line flux of 2.03 × 10⁻¹¹ erg cm⁻² s⁻¹, four orders of magnitude higher (over-prediction factor, or OPF = 1.2 × 10⁴) than what we detect in our COS observation.

One way to explain the overprediction of the Lyα flux is simply that the ionizing flux seen by the BLR clouds is much weaker than what we see due to the beaming effect of the blazar. The BLR clouds are likely exposed to the emission from either the accretion disk or the jet, or a combination of both (Yuan & Narayan 2014). If the emission from the jet dominates, then we can estimate the required Lorentz factor (Γ = 1/(1 − β²)^−1/2) and Doppler factor (δ = [Γ(1 − βcos θ)]⁻¹). Here, β = v_j/c and θ is the viewing angle. The OPF can be related to δ as ${\rm OPF}=\delta ^{5-\alpha _{\lambda }}$ (Urry & Padovani 1995). Using the power-law index from our continuum fitting, we find δ = 16.1. Since Γ ⩾ (δ + 1/δ)/2 and sin θ ⩽ 1/δ, we obtain a minimum Γ of 8.1 and a maximum viewing angle of 36 (Urry & Padovani 1995). As observed in Stocke et al. (2011), if the true Γ is typically as suggested Γ = 3–5, our somewhat large value would imply a covering factor of ∼3%.

On the other hand, if the BLR clouds are mainly ionized by the flux from the accretion disk, then the derived Lorentz factor is likely to be a lower limit since the jet emission seen by the BLR must be much weaker (Stocke et al. 2011). Detailed modeling may help separate the relative contributions from the disk and jet emission (see, e.g., Yuan & Cui 2005; Allen et al. 2006).

A second, more likely interpretation is that instead of a weak ionizing flux, the weakness of the H i emission line rather reflects a lack of line-emitting gas, either because of a small covering fraction of the BLR clouds or because the BLR clouds are optically thin. Similar to the BL Lac objects studied in Stocke et al. (2011), the warm gas in the BLR of H 2356–309 would have a mass around 10⁻⁴–10⁻⁵ M_☉. A weak BLR implies a weak accretion disk and a weak accretion process with low radiative efficiency, such as those models suggested by the advection-dominated accretion flow (ADAF, see Yuan & Narayan 2014 for a review).

Finally, the broad line width suggests that the host galaxy of H 2356–309 is unlikely to be responsible for the observed Lyα emission line. A similarly broad Lyα emission line was detected in an HST/Faint Object Spectrograph observation of M 87 (Sankrit et al. 1999). By comparing with a previous off-nucleus observation, Sankrit et al. (1999) concluded that the majority of the Lyα emission comes from the central region of M 87. While several recent observations have suggested that giant ellipticals may not be as "dead" as one expects and may host a large amount of cold gas (see, e.g., Thom et al. 2012; Werner et al. 2014), the detected Hα line widths are significantly lower (Werner et al. 2014) than the Lyα line discussed in this paper.

A number of Galactic and intergalactic absorption lines were identified in the HST/COS spectrum of H 2356–309 and analyzed by Fang et al. (in preparation). In this paper, we focus on the three absorption lines that were identified as absorbers intrinsic to the blazar. Two of them are H i Lyα absorbers (see the right panel of Figure 1). We also marginally detected the stronger line λ1031.93 of the O vi doublet at the blazar redshift.

The absorption lines were fitted with custom IDL routines using a Voigt-based line profile (see Danforth et al. 2011; B. A. Keeney 2012, private communication). The stronger H i Lyα (EW = 166 ± 17 mÅ) is located at cz = 49498 ± 4 km s⁻¹, well within the rms uncertainty of the blazar redshift (see the top panel of Figure 2). Its profile is well fit with b = 29 ± 5 km s⁻¹ and $\log N_{{\rm H\,\scriptsize{I}}} (\rm cm^{-2})= 13.64\, \pm\, 0.09$. To estimate the line significance level (expressed in Gaussian σ) we make use of the empirical relation derived in Keeney et al. (2012) and Danforth et al. (2011, their Equation (4)): SL = 0.095 W_λ(S/N/pixel)/b^0.62 = 9.0. Here, W_λ is the line EW and (S/N/pixel) is the S/N per COS pixel (see Table 1).

Zoom In Zoom Out Reset image size

A weak absorption feature is detected at 1202.3 Å. No commonly known Galactic interstellar medium (ISM) lines are located around this wavelength; the nearest N i lines are around 1200 Å. Such feature also cannot be redshifted H i Lyα or Lyβ. This feature, however, if it belongs to an absorber intrinsic to the BL Lac object, fits well with the stronger transition of the O vi λλ1031.93, 1037.67 doublet. We tentatively identify this feature as an O vi absorption line at the redshift of H 2356–309. Our Voigt profile shows a line EW of 58  ±  14 mÅ  a column density of $\log N_{{\rm O\,\scriptsize{VI}}} \,(\rm cm^{-2}) = 13.73 \,{\pm}\, 0.12$, and cz = 49497  ±  8 km s⁻¹.

Giving the proximity of the z = 0.16509 O vi and H i absorption lines (|Δv| = 3 km s⁻¹), it is interesting to consider the possibility that a single-phase absorber is responsible for both absorption lines. In this case, we can estimate the temperature of the absorber by assuming the broadening of both lines is a combination of thermal and non-thermal (such as turbulence) effects. We can write $b^2 = 0.129^2 (T/A) + b^2_{{\rm nt}}$, where b and b_nt are the Doppler-b parameter and the non-thermal velocity in units of km s⁻¹, respectively. By solving this equation for the H i and O vi lines, we can then find the temperature and the non-thermal velocity of the absorber. It is unlikely that the Doppler-b parameters for both lines are at their best-fit values, since in this case $b({\rm H\,\scriptsize{I}}) < b({\rm O\,\scriptsize{VI}})$ and the temperature would then be negative. However, we can estimate the upper limit of the temperature by adopting the 1σ upper limit for the Doppler-b parameter of the H i line and the 1σ lower limit for the O vi line. We found a maximum temperature of T = 7.7 × 10⁴ K and a minimum non-thermal velocity of b_nt = 11.4 km s⁻¹ at this temperature.

Such an absorber is unlikely to be in a pure collisional ionization equilibrium. The ratio between the measured column densities of the two species implied a temperature range of 10^5.2–10^6.5 K for metal abundances between 0.1 and 1 solar abundance. This temperature range is inconsistent with the upper limit we inferred from the Doppler-b parameters of the H i and O vi lines. As suggested by Tripp et al. (2008), the line ratio can also rule out the non-equilibrium, radiatively cooling process modeled by Gnat & Sternberg (2007)

We also investigate the physical condition of the absorber under the assumption of photoionization. Numerous H i absorbers with associated O vi absorption were discovered in the past decade (see, e.g., Danforth & Shull 2005; Danforth et al. 2006; Tripp et al. 2008). These absorbers are distributed in the intergalactic medium (IGM) and are likely to be under the influence of the cosmic UV background. Adopting the Haardt & Madau (1996, 2012) UV background, we estimate the physical properties of the absorber using the photoionization code CLOUDY (ver. 13.02; last described by Ferland et al. 2013). The calculation stops when the H i column density reaches $\log N_{{\rm H\,\scriptsize{I}}} (\rm cm^{-2}) = 13.53$. We find that an ionization parameter of log ξ = −0.9 for a metal abundance of log Z = −0.5 is necessary to produce the observed O vi column density. Here, Z is metallicity in units of solar abundance and the ionization parameter is defined as ξ = ϕ/n_Hc, where ϕ is the surface flux of ionizing photons, n_H is the hydrogen density, and c is the speed of light. However, the required n_H is 10^−5.5 cm⁻³, suggesting a path length of 1.5 Mpc for the absorber. The Hubble broadening at this scale is b_H ≈ 50 km s⁻¹, assuming a Hubble constant of 70 km s⁻¹ Mpc⁻¹ (Danforth & Shull 2008), and is larger than the observed line width. Therefore, we also rule out the photoionization model of the O vi–H i absorber in the IGM, unless the metallicity is much higher than the typical value for a low-density IGM environment.

This absorber also cannot be associated with the ISM/halo gas in foreground galaxies. Using HST/WFPC2, Falomo & Ulrich (2000) found a companion galaxy about 1'' away from the sightline toward H 2356–309. A recent Magellan/IMACS survey also found several galaxies within 150'' (or an impact distance of 400 kpc at the blazar redshift) of this sightline (Williams et al. 2013; R. J. Williams 2014, private communication). Using the same UV background we find even if we apply the solar metal abundance, the required pathlength would still be around 1 Mpc which is far too large for an absorber to be within a galaxy.

The O vi–H i absorber is ∼100 km s⁻¹ away from the blazar and can be classified as the so-called "proximity absorber" (Weymann et al. 1979; Foltz et al. 1986). It is likely that the absorber is ionized by the UV radiation from the blazar. We consider two scenarios: the absorber is either located in the ISM of the host galaxy of H 2356–309 or in the BLR. We ran the CLOUDY calculation with the input spectrum derived in the previous section and find the observed column densities can be achieved in both scenarios with reasonable physical parameters. If the absorber is located in the ISM, then assuming solar abundance and a hydrogen density of 1 cm⁻³, we find the absorber has an ionization parameter of log ξ = −1.2, a size of 2.6 kpc, and a distance to the source of 2.7 kpc. For the BLR case, we also assume a solar abundance but with a hydrogen density of 10¹⁰ cm⁻³. We find an ionization parameter of log ξ = −1.4, a size of 0.026 pc, and a distance to the source of 0.034 pc.

The second H i Lyα absorption line is located at the red side of the broad Lyα emission line with a velocity of cz = 49738 ± 7 km s⁻¹. This is about 150 km⁻¹ away from the emission line redshift. Such a phenomenon is frequently observed in the Lyα forest spectra of quasars (see, e.g., Scott et al. 2002); however, this is the first time it was detected in the BL Lac spectrum. It normally suggests the influence of inflow or a large-scale peculiar velocity (see, e.g., Loeb & Eisenstein 1995). The derived properties of this H i absorber are similar to those of the H i Lyα at z = 0.1651 at a weaker significance level (see Table 1).