SPECTROPOLARIMETRIC EVIDENCE FOR RADIATIVELY INEFFICIENT ACCRETION IN AN OPTICALLY DULL ACTIVE GALAXY^*

Deep X-ray surveys have revealed many bright X-ray point sources with weak or no emission lines in their optical spectra (e.g., Elvis et al. 1981; Comastri et al. 2002). Called X-ray bright optically normal galaxies or "optically dull" active galactic nuclei (AGNs, the name adopted here), these objects require an AGN to produce their high X-ray luminosities but lack the broad or narrow emission line signatures of AGN accretion. Optically dull AGNs make up ∼15% of luminous (L_X > 10⁴² erg s^–1) point sources in deep X-ray surveys and ∼25% of those at z < 1 (Trump et al. 2009a; Trouille et al. 2009; Yan et al. 2011). Three major paradigms exist to explain their X-ray brightness and optical dullness: (1) obscuration of both narrow and broad emission lines, (2) dilution by host galaxy starlight, and (3) a physically distinct accretion flow. We explore the evidence for and against each paradigm in turn.

The standard AGN unified model has had great success in using obscuration to explain the differences between optically selected Type 1 (broad-line) and Type 2 (narrow-line) AGNs (Krolik & Begelman 1988; Antonucci 1993). In the simplest interpretation of this model, all AGNs have the broad optical emission lines and strong UV/optical continua of Type 1 AGNs, but along certain lines of sight these features are obscured within a dusty "torus" a few parsecs from the black hole. In an obscured object the narrow emission lines remain visible because they are excited beyond the obscuring material, and so the lack of a broad-line region (BLR) and weaker optical continuum of Type 2 AGNs are attributed to obscuration. Similarly, Comastri et al. (2002) and Civano et al. (2007) suggested that optically dull AGNs have the same physical engine as a Type 1 or Type 2 AGN, but with additional obscuration within a few parsecs completely blocking the ionizing continuum radiation from exciting even the narrow emission lines. Rigby et al. (2006) instead suggest that optically dull AGNs are obscured by extranuclear (>100 pc) gas and dust in the host galaxy, blocking the narrow lines along the observer's line of sight. No matter the physical location of the obscuring material, it must preferentially absorb the optical emission, since optically dull AGNs remain X-ray bright. Indeed, more than half of optically dull AGNs are relatively unobscured (N_H < 10²² cm⁻²) in the X-rays (Severgnini et al. 2003; Page et al. 2003). Preferentially obscuring the optical emission while remaining X-ray unabsorbed would require extreme gas-to-dust ratios not observed in other AGNs (Maiolino et al. 2001).

Optically dull AGNs could also be Type 2 AGNs with narrow emission lines diluted by a bright host galaxy. Indeed, Moran et al. (2002) showed that many z ∼ 0 AGNs would appear optically dull if observed at z ∼ 1, since the host galaxy would occupy more of the spectroscopic slit or fiber and consequently overwhelm the AGN emission lines. Hubble Space Telescope (HST)/Advanced Camera for Surveys (ACS) images additionally show that many optically dull AGN hosts at z < 1 are edge-on (Rigby et al. 2006) or have a second galaxy falling within the spectroscopic aperture (Trump et al. 2009b). However, 10%–20% of local (undiluted) AGNs are optically dull (La Franca et al. 2002; Hornschemeier et al. 2005). And even after removing the host galaxy light by decomposing the spectral energy distribution, ∼1/3 of optically dull AGNs have anomalously high X-ray to optical flux ratios (Trump et al. 2009b).

Another possibility is that optically dull AGNs have different accretion physics due to low accretion rates (Yuan & Narayan 2004). Models have long predicted that an AGN with a low accretion rate ($\dot{m} \equiv L_{\rm bol}/L_{\rm Edd} \lesssim 0.01$) will have a radiatively inefficient accretion flow (RIAF) within some truncation radius R_t, with R_t defined as where the collisional cooling time is comparable to the accretion time (Begelman et al. 1984; Narayan et al. 1995). Beyond R_t, accretion will remain in a standard geometrically thin and optically thick disk with a thermal blackbody spectrum (e.g., Shakura & Sunyaev 1973). However, within R_t there are too few collisions to couple the ions and electrons, and the gas becomes a two-temperature plasma. The electrons are cooled by bremsstrahlung, synchrotron, and Compton up-scattering, while the ions remain at the virial temperature. This means the flow is geometrically thick and optically thin. A RIAF then lacks much of the optical/UV blackbody emission of an optically thick accretion disk, and consequently cannot ionize and/or excite the broad or narrow optical emission lines seen in other AGNs (Yuan & Narayan 2004; Trump et al. 2011). Because the ions in the RIAF are only marginally bound, such AGNs should also have strong outflows and be consequently radio luminous (Meier 2001). Optically dull AGNs, especially those with high X-ray to optical flux ratios, are indeed observed to be have higher ratios of radio to optical/UV luminosity than Type 1 AGNs (Trump et al. 2009b, 2011). Much of the optical and infrared light in RIAF AGNs may be synchrotron radiation associated with the radio jet (Ho 2009).

Each of these three paradigms has a different signature in spectropolarimetry. Anisotropic obscuration of the narrow emission lines would leave telltale reflected emission lines in the polarized spectrum (Nagao et al. 2004). On the other hand, if an optically dull AGN is a Type 2 AGN diluted by a host galaxy, its polarized flux would be equally diluted. (A diluted optically dull AGN might exhibit polarized broad emission lines like those seen in Type 2 AGNs, but the host galaxy would probably overwhelm them to a non-detectable level.) And if optically dull AGNs have RIAFs we might observe a featureless polarized continuum, either from the synchrotron emission associated with the stronger radio jet (e.g., Jannuzi et al. 1994; Cohen et al. 1999) or because we are viewing the naked, lineless continuum (from the disk beyond the RIAF) reflected by a scattering surface in the host galaxy (Antonucci & Miller 1985; Ogle et al. 1999; Kishimoto et al. 2001). We summarize the polarization signatures of each paradigm in Table 1.

We used Subaru/Faint Object Camera and Spectrograph (FOCAS) to observe the two brightest optically dull AGNs of Trump et al. (2009b), 095849+013220 and 100036+024929. The objects have multiwavelength observations from the Cosmic Evolution Survey (COSMOS; Scoville et al. 2007), a survey based on a 1.7 deg² HST/ACS treasury program (Koekemoer et al. 2007). Both targets have identifications and redshifts from the XMM-COSMOS AGN spectroscopic campaign with Magellan/IMACS (Trump et al. 2009a). Despite their absorption line optical spectra, they are confirmed as bona fide AGNs using their XMM-Newton data (Cappelluti et al. 2009; Brusa et al. 2010): each has an X-ray luminosity of L_{0.5 − 10keV} > 3 × 10⁴² erg s⁻¹, requiring an AGN (Hornschemeier et al. 2001). We show properties of 095849+013220 and 100036+024929 in Table 2.

Trump et al. (2009b) used AGN plus galaxy fits to the optical photometry to suggest that both of these targets are roughly 45% AGN and 55% host galaxy in the i band. Table 2 also gives the X-ray to optical flux ratios, given by X/O = log f_X/f_O = log (f_{0.5 − 2keV}) + i_AB/2.5 + 5.352. The second target, 100036+024929, is a good candidate to be a normal AGN diluted by a host galaxy because of its X/O(AGN) = −0.6, fitting nicely in the traditional "X-ray AGN locus" of −1 < X/O < 1 (Maccacaro et al. 1988). However, the first target, 095849+013220, is anomalously X-ray bright and optically dull, with X/O = 1.1 and X/O(AGN) = 1.4. Since starlight dilution would increase the optical emission and decrease the X/O ratio, the high X/O ratio for the first target means it cannot be explained by starlight dilution. Instead 095849+013220 is a good candidate to host a RIAF (Trump et al. 2009b).

Spectropolarimetric observations of both targets were undertaken using the Faint Object Camera and Spectrograph (Kashikawa et al. 2002) on the 8.2 m Subaru telescope. Each target was observed for 6 hr of total integration in a 08 wide center slit. We used the R58 filter and the 150 l mm^–1 grism, resulting in an observed spectral range of 5800–10000 Å with ∼2.8 Å pixel^–1 resolution. During spectropolarimetry observations on FOCAS, a crystal quartz Wallaston prism splits the incident light into ordinary and extraordinary components, which then pass through a rotating achromatic half-wave plate before being recorded simultaneously on the CCD. Instrumental polarization of Subaru/FOCAS using the center slit is negligible (<0.05%), as confirmed by our unpolarized standard star exposures. The 6 hr observations were divided into nine sets of 10 minute exposures at each of four half-wave plate position angles 0°, 45°, 225, and 675.

We used the unpolarized standard star G191B2B to flux calibrate and remove telluric absorption lines, and used the polarized standard star HD 251204 (Turnshek et al. 1990) to determine the absolute position angle. We reduced individual frames using the FOCAS IRAF cookbook and then calculated the normalized flux difference for each individual frame: F_θ = (f^o_θ − f^e_θ)/(f^o_θ + f^e_θ). We then combined the normalized flux difference for each angle θ and calculated the Stokes parameters: Q = 0.5(F₀ − F₄₅) and U = 0.5(F_22.5 − F_67.5). We additionally computed the Stokes parameters using the method of Miller et al. (1988), which accounts for calibration differences between the "ordinary" and "extraordinary" light positions, and found no significant differences in the resultant Q or U.

The degree of polarization is given by $P \equiv \sqrt{Q^2+U^2}$ and the polarization angle by $\theta \equiv 0.5 \arctan (U/Q)$. Errors in U and Q bias the total polarization P in the positive direction, and we correct for this bias using the simulation-derived correction of Patat & Romaniello (2006). Galactic interstellar polarization is <0.2% in the direction of both targets (Heiles 2000) and is therefore disregarded.

The total flux, degree of polarization, and polarization angle for the RIAF candidate and dilution candidate AGNs are shown in Figures 1 and 2, respectively. We bin the degree of polarization and polarization angle by 20 pixels (56 Å in the observed frame) to improve their signal-to-noise ratio (S/N).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is first evident that several narrow emission lines are revealed in the unpolarized Subaru/FOCAS spectra, as compared to our previous Magellan/IMACS spectra with much lower S/N. The RIAF candidate, 095849+013220, may have weak Hα emission, although it is strangely blueshifted from the absorption line redshift. The diluted candidate, 100036+024929, shows strong Hα and [Nii] emission (which was missed in the previous Magellan/IMACS spectrum because of a chip gap) and weak Hβ and [Oiii] emission.

Neither of the polarized fluxes in Figures 1 and 2 show evidence for reflected AGN emission lines such as Hα, Hβ, or [Oiii]. The Nai Dλ5892 absorption feature might have increased percent polarization, although it is significant at only ∼2σ in both spectra. If the increased percent polarization of Na i D is real, it is probably caused by the lower starlight dilution at that wavelength, and so does not represent an actual increase in polarized flux.

The RIAF candidate shows significant continuum polarization, while the diluted candidate does not. The mean continuum polarization of 095849+013220 is 0.78% ± 0.07% at an angle of 104% ± 4% degrees. The mean continuum polarization in the blue, measured at 4400 Å < λ < 5050 Å, is even stronger: 1.37% ± 0.16%. Subtracting the host galaxy component, which contributes ∼55% of the total emission at λ_rest ∼ 5000 Å, suggests that the mean polarization of the AGN component is ∼1.7% overall and ∼3.0% in the blue. The polarized continuum roughly doubles from ∼6500 Å to ∼4500 Å, consistent with a typical unobscured quasar continuum of f_λ ∼ λ^−1.6 (Vanden Berk et al. 2001).

In addition to the two optically dull AGNs from COSMOS, we observed the cluster BL Lac candidate MS 1455-X2 from Hart et al. (2009) and discuss its spectropolarimetry results in the Appendix.

Neither of our targets show evidence for polarized broad or narrow emission lines, which suggests that the lines are not missing due to anisotropic absorption as in a standard AGN "torus" model (Antonucci 1993). Instead, our RIAF candidate has a blue polarized continuum. Here we place limits on polarized emission line flux and investigate the two possible causes of the continuum polarization: synchrotron emission or scattering.

4.1. Limits on Polarized Emission Lines

4.2. Continuum Polarization from Synchrotron Emission

4.3. Continuum Polarization from Scattering

We present Subaru/FOCAS spectropolarimetry for two optically dull AGNs: the RIAF candidate 095849+013220 and the galaxy-diluted Type 2 AGN candidate 100036+024929. Neither source exhibits reflected broad emission lines, with 3σ upper limits of ≲2% polarized broad lines. This rules out simple torus obscuration as the reason behind the lack of broad emission lines. Despite the lack of polarized broad-line emission, in the RIAF candidate we additionally observe a polarized continuum. The blue spectral slope of the polarized flux suggests that it represents the intrinsic AGN continuum, either observed directly as synchrotron emission or as reflected by clumpy dust. Because we observe the direct emission from the central AGN engine without broad emission lines, we suggest that this source has an intrinsically weaker or nonexistent BLR. The spectropolarimetry is another piece of evidence that the optically dull AGN 095849+013220 has a RIAF.

We thank Takashi Hattori and the staff of Subaru for excellent support during and after observations. We thank Kenta Matsuoka for help in reducing the FOCAS data. Masaomi Tanaka was invaluable in interpreting the spectropolarimetry results. We also thank Hien Tran and Makoto Kishimoto for help in interpreting a polarized continuum in AGNs. We thank the referee for helpful comments which improved the quality of this work. J.R.T. acknowledges support from NSF/DDEP grant 0943995 and NSF grant AST-0808133, and with C.D.I. acknowledges support from AST-0908044. F.C. acknowledges support from the Blancheflor Boncompagni Ludovisi foundation and the Smithsonian Scholarly Studies. B.C.K. acknowledges support from NASA through Hubble Fellowship grant HF-51243.01 awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS 5-26555. Y.T. acknowledges support from JSPS grants 17253001 and 19340046.

Hart et al. (2009) reported that the cluster object MS 1455-X2 is luminous in X-ray and radio emission and yet lacks emission lines in its optical spectrum. Its basic properties are given in Table 2. Its optical colors indicate emission blueward of 4000 Å in excess of that expected for a lineless red galaxy, and so MS 1455-X2 is similar to the RIAF candidate we observe in COSMOS. Indeed, Hart et al. (2009) suggest that it is a low-luminosity BL Lac object and predict that such objects may be common in clusters and can provide the feedback necessary to heat cluster cores.

We observed MS 1455-X2 for a total of 160 minutes, divided into four sets of 10 minute exposures at each of the four half-wave plate position angles. We used the same instrumental parameters and reduction process as described in Section 2. The shorter exposure time and faintness of the target (r = 20.05) mean that we cannot put meaningful limits on the amount of emission line polarization for MS 1455-X2. However, we detected a significant total polarization of P = 2.57% ± 0.47% at 4800 < λ_rest < 7700 Å. This high degree of polarization strongly supports the hypothesis that this source is a low-luminosity BL Lac AGN, especially considering some amount of starlight dilution from the host galaxy.