X-ray bolometric corrections for Compton-thick active galactic nuclei

TL;DR

This study derives X-ray bolometric correction factors for Compton-thick AGN using IR and X-ray models, finding consistency with existing relations and suggesting the corona's emission is isotropic, aiding understanding of obscured AGN energetics.

Contribution

It provides the first systematic analysis of bolometric corrections for local Compton-thick AGN, comparing different models and examining orientation effects.

Findings

Mean log10 κ_Bol ≈ 1.44 with 0.2 dex scatter.

No significant dependence of κ_Bol on N_H.

X-ray corona emission appears isotropic.

Abstract

We present X-ray bolometric correction factors, $\kappa_{Bol}$ ($\equiv L_{Bol}/L_X$), for Compton-thick (CT) active galactic nuclei (AGN) with the aim of testing AGN torus models, probing orientation effects, and estimating the bolometric output of the most obscured AGN. We adopt bolometric luminosities, $L_{Bol}$, from literature infrared (IR) torus modeling and compile published intrinsic 2--10 keV X-ray luminosities, $L_{X}$, from X-ray torus modeling of NuSTAR data. Our sample consists of 10 local CT AGN where both of these estimates are available. We test for systematic differences in $\kappa_{Bol}$ values produced when using two widely used IR torus models and two widely used X-ray torus models, finding consistency within the uncertainties. We find that the mean $\kappa_{Bol}$ of our sample in the range $L_{Bol}\approx10^{42}-10^{45}$ erg/s is log$_{10}\kappa_{Bol}=1.44\pm0.12$ with an intrinsic scatter of $\sim0.2$ dex, and that our derived $\kappa_{Bol}$ values are consistent with previously established relationships between $\kappa_{Bol}$ and $L_{Bol}$ and $\kappa_{Bol}$ and Eddington ratio. We investigate if $\kappa_{Bol}$ is dependent on $N_H$ by comparing our results on CT AGN to published results on less-obscured AGN, finding no significant dependence. Since many of our sample are megamaser AGN, known to be viewed edge-on, and furthermore under the assumptions of AGN unification whereby unobscured AGN are viewed face-on, our result implies that the X-ray emitting corona is not strongly anisotropic. Finally, we present $\kappa_{Bol}$ values for CT AGN identified in X-ray surveys as a function of their observed $L_X$, where an estimate of their intrinsic $L_{X}$ is not available, and redshift, useful for estimating the bolometric output of the most obscured AGN across cosmic time.

Tables

Table 1. Table 1 Basic properties of the galaxies in our sample

Name	Mag	Morphology	Distance	$M_{\mathrm{BH}}$	Ref
(1)	(2)	(3)	(4)	(5)	(6)
Circinus	10.0 ($I$)	SA(s)b?	4.2	1.7$\pm$0.3	a
NGC 424	12.8 ($I$)	(R)SB0/a?(r)	50.6
NGC 1068	9.9 ($I$)	(R)SA(rs)b	14.4	8.0$\pm$0.3	b
NGC 1194	12.5 ($i$)	SA0^+?	58.9	65.0$\pm$3.0	c
NGC 1320	12.5 ($V$)	Sa? edge-on	39.1
NGC 1386	10.76 ($R$)	SB0^+(s)	15.9	1.2$\pm$1.1	d
NGC 2273	10.15 ($I$)	SB(r)a?	28.9	7.5$\pm$0.4	c
NGC 3079	9.5 ($I$)	SB(s)c edge-on	19.2	2.4${}_{-1.2}{}^{}{}_{}{}^{+2.4}$	e
NGC 5643	10.6 ($I$)	SAB(rs)c	13.9
NGC 7582	9.2 ($I$)	(R′)SB(s)ab	22

Equations

$$anisotropy\equiv \frac{{\kappa }_{\mathrm{Bol}}(lo{g}_{10}(N_{\mathrm{H}}/{\mathrm{cm}}^{-2})=24-26)}{{\kappa }_{\mathrm{Bol}}(lo{g}_{10}(N_{\mathrm{H}}/{\mathrm{cm}}^{-2})=20-22)}$$

$${\kappa }_{\mathrm{Bol}}^{\prime }\equiv \frac{L_{\mathrm{Bol}}}{L(0.5-8\mathrm{keV},observed)}$$

Full text

1. Introduction

The bolometric luminosity, $L_{\rm Bol}$, of an accreting supermassive black hole (SMBH), otherwise known as an active galactic nucleus (AGN), describes the integrated emission from the accretion process, which traces the mass accretion rate onto the SMBH ($L_{\rm Bol}=\eta\dot{m}c^{2}$, where $\dot{m}$ is the mass accretion rate and $\eta$ the accretion efficiency). Thus $L_{\rm Bol}$ is an important parameter for understanding the growth of SMBHs. The emission from the accretion disk, which is the primary power generation mechanism, is reprocessed by a number of components in the vicinity of the disk, one of which is a hot corona of electrons that Compton scatters the optical and UV disk emission into the X-ray regime (e.g. Haardt & Maraschi, 1991, 1993).

The fraction of the disk emission that is up-scattered in to the X-ray regime is parameterized by the X-ray bolometric correction factor (from here on $\kappa_{\rm Bol}$), which is defined as $L_{\rm Bol}$/$L_{\rm{X}}$, where $L_{\rm{X}}$ is the X-ray luminosity in the 2–10 keV band. Many works have investigated $\kappa_{\rm Bol}$, finding that it is dependent on $L_{\rm Bol}$ (e.g. Marconi et al., 2004; Steffen et al., 2006; Hopkins et al., 2007; Lusso et al., 2012; Liu et al., 2016) and Eddington ratio (\lambda_{\rm Edd}$$\equiv$$L_{\rm Bol}/$L_{\rm Edd}$, where $L_{\rm Edd}=4\pi$G$M_{\rm BH}$ $m_{\rm p}c/\sigma_{\rm T}\simeq 1.26\times 10^{38}$($M_{\rm BH}$/$M_{\odot}$)  erg s*-1* and $M_{\rm BH}$ is the mass of the black hole, e.g. Wang et al., 2004; Vasudevan & Fabian, 2007, 2009; Lusso et al., 2010, 2012; Jin et al., 2012; Fanali et al., 2013; Liu et al., 2016).

Characterizing $\kappa_{\rm Bol}$ and its dependencies is important for understanding accretion physics and for estimating $L_{\rm Bol}$ when it is not possible to observe the intrinsic disk emission, but where $L_{\rm{X}}$ is known. This can be the case for obscured AGN, where gas and dust in the line of sight extinguishes the optical and UV emission from the accretion disk but X-rays from the corona penetrate through (for all but the most extreme absorbing columns N_{\rm{H}}$$<10^{24} cm*-2*). While the dependencies of $\kappa_{\rm Bol}$ have been well established for unobscured, type 1 AGN, only a few studies have focussed on obscured, type 2 AGN (e.g. Pozzi et al., 2007; Vasudevan et al., 2010; Lusso et al., 2011, 2012).

Investigating $\kappa_{\rm Bol}$ for obscured AGN is important since the majority of AGN in the Universe are obscured (e.g. Martínez-Sansigre et al., 2005; Ueda et al., 2014; Buchner et al., 2015; Aird et al., 2015). It also has potential for testing the AGN unification scheme (e.g. Antonucci, 1993; Urry & Padovani, 1995), the simplest form of which describes the differences between type 1 and type 2 AGN as solely due to orientation, where type 2 AGN are more inclined systems and our view of the central engine is through a toroidal structure of gas and dust. The most extremely obscured sources, so-called Compton-thick (CT) AGN (N_{\rm{H}}$$>1.5\times 10^{24} cm*-2*) constitute some $\sim 20-40$% of the AGN population (e.g. Burlon et al., 2011; Brightman & Nandra, 2011; Brightman & Ueda, 2012; Buchner et al., 2015) and host some of the most highly inclined systems, revealed through the detection of disk megamasers (Zhang et al., 2006; Masini et al., 2016). However, for CT AGN, flux suppression is high even in the X-ray band and the effect of Compton scattering on the X-ray spectrum is dependent on the geometry of the obscuring material (e.g. Brightman et al., 2015) making the intrinsic $L_{\rm{X}}$ difficult to estimate. For this reason $\kappa_{\rm Bol}$ has not previously been investigated for CT AGN.

At energies $>$10 keV, while the effect of Compton scattering remains, the flux suppression is lower due to the declining photoelectric absorption cross section with increasing energy. Therefore, NuSTAR (Harrison et al., 2013), with its sensitivity at these energies, is ideal for estimating the intrinsic $L_{\rm{X}}$ for CT AGN. For this, X-ray spectral models that take into account the absorption and Compton scattering are needed (e.g. Ikeda et al., 2009; Murphy & Yaqoob, 2009; Brightman & Nandra, 2011; Liu & Li, 2014). Figure 1 illustrates this point, showing the NuSTAR data of the well-known CT AGN in the Circinus galaxy (Arévalo et al., 2014), fitted with the Brightman & Nandra (2011) torus model, also showing the intrinsic X-ray spectrum inferred using the model parameters. The figure shows that a greater fraction of X-ray flux emerges above 10 keV in the source, than below 10 keV. Since its launch in 2012, NuSTAR has observed a large number of CT AGN, with $L_{\rm{X}}$ estimated from both the mytorus model of Murphy & Yaqoob (2009) and the torus model of Brightman & Nandra (2011) by various authors (e.g. Puccetti et al., 2014; Arévalo et al., 2014; Baloković et al., 2014; Gandhi et al., 2014; Bauer et al., 2015; Brightman et al., 2015; Koss et al., 2015; Annuar et al., 2015; Rivers et al., 2015; Marinucci et al., 2016; Ricci et al., 2016; Masini et al., 2016; Farrah et al., 2016; Boorman et al., 2016).

As well as being reprocessed by the hot corona into the X-rays, the AGN disk emission is also reprocessed by the dust in the torus into the infrared (e.g. Pier & Krolik, 1992). The structure of the dust torus does not necessarily have the same geometry as the X-ray absorbing material, which is gas that can exist within the dust sublimation radius. As in the X-ray band, torus models have been calculated to model the infrared emission (e.g. Nenkova et al., 2008; Hönig & Kishimoto, 2010; Stalevski et al., 2012; Efstathiou et al., 2013). A natural parameter derived from these models is $L_{\rm Bol}$. Since significant infrared emission is also emitted by dusty star formation in the host galaxy, high-spatial resolution IR data or broadband spectral energy distribution (SED) modeling are required to isolate the AGN and model the torus emission (e.g. Farrah et al., 2003; Stierwalt et al., 2014). Alonso-Herrero et al. (2011) presented the results from fitting of Nenkova et al. (2008) clumpy torus model to high-spatial resolution IR spectroscopy and photometry of 13 nearby Seyfert galaxies, finding that their $L_{\rm Bol}$ estimates agreed well with other estimates from the literature. A further expanded study in the IR was conducted by Ichikawa et al. (2015), which presented an analysis of 21 nearby AGN, with significant overlap with the sample of AGN with X-ray torus modeling.

One such source in common is the CT AGN in the Circinus galaxy. Along with the NuSTAR data in Figure 1, we plot the high spatial resolution IR data along with the fit using the IR torus model. The inferred accretion disk spectrum is also shown.

The aim of this paper is to take advantage of the recent advances in both IR and X-ray torus modeling that produce estimates of $L_{\rm Bol}$ and intrinsic $L_{\rm{X}}$ respectively and derive $\kappa_{\rm Bol}$ values for CT AGN. We start in Section 2 where we describe our sample selection. In Section 3 we collect and compare results from the literature on the two widely used X-ray torus models, mytorus (Murphy & Yaqoob, 2009) and torus (Brightman & Nandra, 2011) and two widely used IR torus models from Fritz et al. (2006) and Nenkova et al. (2008). We assess the systematic differences, if any. Following this we test if the $\kappa_{\rm Bol}$ values we estimate for CT AGN are consistent with established relationships between $\kappa_{\rm Bol}$ and $L_{\rm Bol}$ and $\kappa_{\rm Bol}$ and $\lambda_{\rm Edd}$ as determined from unobscured AGN. Next we compare our new $\kappa_{\rm Bol}$ results for CT AGN to results from previous studies for less obscured systems in order to explore any dependence of $\kappa_{\rm Bol}$ on $N_{\rm{H}}$ and probe orientation effects. We then present $\kappa_{\rm Bol}$ for CT AGN as a function of observed $L_{\rm{X}}$ and redshift, useful for studies of CT AGN in surveys where there is not a good estimate of the intrinsic $L_{\rm{X}}$. We discuss our results in Section 4 and present our conclusions in Section 5. We define $L_{\rm Bol}$ as the total of the inferred disk emission (from IR torus modeling) together with the intrinsic $L_{\rm{X}}$ (from X-ray torus modeling) in order to be consistent with previous works (e.g. Marconi et al., 2004; Vasudevan et al., 2010). We assume a flat cosmological model with $H_{\rm 0}$=70 km s*-1* Mpc*-1* and $\Omega_{\Lambda}$=0.73.

2. Sample Selection and Luminosity Estimates

We compile $L_{\rm{X}}$ measurements from X-ray torus modeling of NuSTAR data and $L_{\rm Bol}$ results from IR spectral/SED modeling from the literature, finding 10 local CT AGN where both of these exist. We find five sources from the sample of Ichikawa et al. (2015), who used the clumpy torus models of Nenkova et al. (2008) to calculate $L_{\rm Bol}$, fitting over the range 1.25–30 $\mu$m. We find a further four sources from the sample of Gruppioni et al. (2016), who rather than using high-spatial resolution IR data to isolate the AGN emission, carry out SED decomposition to isolate the AGN emission from the host galaxy, using the approach described by Berta et al. (2013). They use the torus model of Fritz et al. (2006), which models a smooth distribution of dust and calculate $L_{\rm Bol}$ over the 1–1000 $\mu$m range. Finally, Woo & Urry (2002) calculated $L_{\rm Bol}$ for a large number of AGN by simply integrating over the observed multiwavelength SED. This was a far less sophisticated approach to $L_{\rm Bol}$ estimation than IR torus modeling since it presumably does not account for host-galaxy emission. We compare these $L_{\rm Bol}$ estimates for four sources where overlap with the IR torus modeling exists. We find one CT AGN where X-ray torus modeling has been conducted and an $L_{\rm Bol}$ estimate exists from Woo & Urry (2002), NGC 2273, which we include in our sample.

We list some basic observational properties of our sample in Table 2. Due to the detailed torus modeling involved, these sources are necessarily nearby ($D<60$ Mpc). Our sample also contains six megamaser AGN indicating that they have high inclinations, since these are required to produce this emission. Furthermore, the Keplerian motion of the masing material provides an accurate measurement of $M_{\rm BH}$ (e.g. Kuo et al., 2011) and allows us to test the relationship between $\kappa_{\rm Bol}$ and $\lambda_{\rm Edd}$. We also list the $M_{\rm BH}$ estimates in Table 2.

The different torus models used to calculate $L_{\rm{X}}$ and $L_{\rm Bol}$ have properties that are inherent to each, which we describe here. The Nenkova et al. (2008) models assume a dust torus consisting of clouds that are distributed with axial symmetry and whose number per unit length depends on the distance from the illuminating source and the angle from the equatorial plane. This torus is illuminated by an intrinsic disk spectrum which takes the form of a piecewise power-law distribution described in Rowan-Robinson (1995), where $\lambda F_{\lambda}=\lambda^{1.2}$ for $\lambda\leq 0.01\mu$m, $\lambda F_{\lambda}=$ constant for $0.01\leq\lambda<0.1\mu$m, $\lambda F_{\lambda}=\lambda^{-0.5}$ for $0.1\leq\lambda<1\mu$m and $\lambda F_{\lambda}=\lambda^{-3}$ for $1\mu$m$\leq\lambda$. Integrating over this assumed disk spectrum yields $L_{\rm Bol}$. The anisotropy of this clumpy torus is discussed at length in Nenkova et al. (2008) and depends on the various parameters of the torus. For example, the torus becomes less anisotropic when the power-law index of the radial distribution of clouds increases, i.e. steeper. This is a free parameter in the model and hence fitted for in SED modeling. The anisotropy is also strongly wavelength dependent, with the torus being being particularly isotropic at 12$\mu$m.

While the Nenkova et al. (2008) model assumes a clumpy distribution of dust, the Fritz et al. (2006) model also assumes smooth distribution, but that also depends on the radial distance from the source and the equatorial angle. An intrinsic disk spectrum that illuminates the torus isotropically in the form of a piecewise power-law distribution that is similar but not identical to that assumed by the Nenkova et al. (2008) models. Here $\lambda F_{\lambda}\propto\lambda^{1.2}$ for $0.001\leq\lambda\leq 0.03\mu$m, $\lambda F_{\lambda}\propto$ constant for $0.03\leq\lambda<0.125\mu$m and $\lambda F_{\lambda}\propto\lambda^{-0.5}$ for $0.125\leq\lambda<20\mu$m. The degree of anisotropy from this torus is rather higher than for the clumpy torus, and depends on the viewing angle and the equatorial optical depth. Again these are free parameters of the model and are fitted for in SED modeling. $L_{\rm Bol}$ is calculated from a bolometric correction factor given the best-fit template Gruppioni et al. (2016).

The X-ray torus models of Murphy & Yaqoob (2009) and Brightman & Nandra (2011) both model smooth distributions of gas. mytorus assumes a ‘doughnut’-like geometry with a circular cross-section, whereas the torus model assumes a ‘spherical’ torus with a biconical cut out. Both models assume a intrinsic source spectrum that takes power-law form with $F_{\gamma}\propto E^{-\Gamma}$ ($\lambda F_{\lambda}=E^{-\Gamma+2}$). For sight lines through the torus, the anisotropy in the NuSTAR band is negligible.

The luminosities that we have compiled here are a collection of literature values that also depend on the distance assumed by each author, which can often have large discrepancies due to the nearby nature of these galaxies. For example, Brightman et al. (2015) assume a distance of 6.2 Mpc to the Circinus galaxy based on the Hubble flow distance for the intrinsic $L_{\rm{X}}$ estimate from the torus model, whereas Arévalo et al. (2014) assume a distance of 4.2 Mpc based on the Tully estimate for the intrinsic $L_{\rm{X}}$ estimate from the mytorus model. Furthermore, Ichikawa et al. (2015) assume a distance of 4 Mpc for the $L_{\rm Bol}$ estimate. Since luminosity scales with distance squared, this difference leads to a factor of $\sim 2$ discrepancy which we must account for when calculating and comparing $\kappa_{\rm Bol}$ values. We do this by taking the luminosity and the distance assumed by each author and correcting the luminosity assuming the distance that we list in Table 2.

We list the intrinsic $L_{\rm{X}}$ and $L_{\rm Bol}$ estimates in Table 2 along with the corresponding $\kappa_{\rm Bol}$ values which have been corrected for distance. Our sample spans a range of L_{\rm{X}}$$\approx 10^{41.5}-10^{44}  erg s*-1*, L_{\rm Bol}$$\approx 10^{42}-10^{45}  erg s*-1*, M_{\rm BH}$$\approx 10^{6}-7\times 10^{7} $M_{\odot}$ and \lambda_{\rm Edd}$$\approx 0.01-0.3. All our sources are Compton thick by selection with N_{\rm{H}}$$=10^{24}-10^{25} cm*-2*, with the exception of NGC 1320 that has N_{\rm{H}}$$>10^{25} cm*-2* (Brightman et al., 2015).